UBC Faculty Research and Publications

Altered DNA methylation is associated with aberrant gene expression in parenchymal but not airway fibroblasts… Clifford, Rachel L; Fishbane, Nick; Patel, Jamie; MacIsaac, Julia L; McEwen, Lisa M; Fisher, Andrew J; Brandsma, Corry-Anke; Nair, Parameswaran; Kobor, Michael S; Hackett, Tillie-Louise; Knox, Alan J Mar 5, 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-13148_2018_Article_464.pdf [ 1.28MB ]
JSON: 52383-1.0364161.json
JSON-LD: 52383-1.0364161-ld.json
RDF/XML (Pretty): 52383-1.0364161-rdf.xml
RDF/JSON: 52383-1.0364161-rdf.json
Turtle: 52383-1.0364161-turtle.txt
N-Triples: 52383-1.0364161-rdf-ntriples.txt
Original Record: 52383-1.0364161-source.json
Full Text

Full Text

RESEARCH Open AccessAltered DNA methylation is associated withaberrant gene expression in parenchymalbut not airway fibroblasts isolated fromindividuals with COPDRachel L. Clifford1*, Nick Fishbane2, Jamie Patel1, Julia L. MacIsaac3, Lisa M. McEwen3, Andrew J. Fisher4,Corry-Anke Brandsma5,6, Parameswaran Nair7, Michael S. Kobor3, Tillie-Louise Hackett2,8† and Alan J. Knox1†AbstractBackground: Chronic obstructive pulmonary disease (COPD) is a heterogeneous disease of the lungs that iscurrently the fourth leading cause of death worldwide. Genetic factors account for only a small amount ofCOPD risk, but epigenetic mechanisms, including DNA methylation, have the potential to mediate theinteractions between an individual’s genetics and environmental exposure. DNA methylation is highly celltype-specific, and individual cell type studies of DNA methylation in COPD are sparse. Fibroblasts are presentwithin the airway and parenchyma of the lung and contribute to the aberrant deposition of extracellularmatrix in COPD. No assessment or comparison of genome-wide DNA methylation profiles in the airway andparenchymal fibroblasts from individuals with and without COPD has been undertaken. These data providevaluable insight into the molecular mechanisms contributing to COPD and the differing pathologies of smallairways disease and emphysema in COPD.Methods: Genome-wide DNA methylation was evaluated at over 485,000 CpG sites using the Illumina InfiniumHumanMethylation450 BeadChip array in the airway (non-COPD n = 8, COPD n = 7) and parenchymal fibroblasts(non-COPD n = 17, COPD n = 29) isolated from individuals with and without COPD. Targeted gene expression wasassessed by qPCR in matched RNA samples.Results: Differentially methylated DNA regions were identified between cells isolated from individuals with andwithout COPD in both airway and parenchymal fibroblasts. Only in parenchymal fibroblasts was differential DNAmethylation associated with differential gene expression. A second analysis of differential DNA methylationvariability identified 359 individual differentially variable CpG sites in parenchymal fibroblasts. No differentiallyvariable CpG sites were identified in the airway fibroblasts. Five differentially variable-methylated CpG sites,associated with three genes, were subsequently assessed for gene expression differences. Two genes (OAT andGRIK2) displayed significantly increased gene expression in cells isolated from individuals with COPD.Conclusions: Differential and variable DNA methylation was associated with COPD status in the parenchymalfibroblasts but not airway fibroblasts. Aberrant DNA methylation was associated with altered gene expressionimparting biological function to DNA methylation changes. Changes in DNA methylation are therefore implicatedin the molecular mechanisms underlying COPD pathogenesis and may represent novel therapeutic targets.Keywords: DNA methylation, Fibroblasts, COPD, Airway, Parenchyma* Correspondence: R.clifford@nottingham.ac.uk†Equal contributors1Nottingham NIHR Biomedical Research Centre, Nottingham MRC MolecularPathology Node, Division of Respiratory Medicine, University of Nottingham,Nottingham University Hospitals NHS Trust, City Hospital, Nottingham, UKFull list of author information is available at the end of the article© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Clifford et al. Clinical Epigenetics  (2018) 10:32 https://doi.org/10.1186/s13148-018-0464-5BackgroundChronic obstructive pulmonary disease (COPD) affects300 million people worldwide, is currently the fourthleading cause of death in the world, and is expected tobecome the third leading cause of mortality worldwideby 2020 [1]. Clinically defined by airflow obstruction thatis not reversible, COPD is a heterogeneous diseasewithin the lung involving parenchymal lung destructionresulting in loss of elastic recoil (emphysema) and smallairway disease [2, 3]. COPD is caused by exposure tonoxious particles or gases that are predominantly intro-duced into the lung by cigarette smoking, or other expo-sures such as biomass fuels. The irreversible airflowobstruction is not reversed by cigarette smoking cessa-tion, and to date, there are no therapies that can modifydisease activity or progression. A better understandingof the molecular processes underlying the pathology ofCOPD is crucial to the design of novel therapeutics [4].Genome-wide genetic association studies (GWAS)have identified four susceptibility loci for COPD that areassociated with smoking behavior and have been wellreplicated [5–12], but each variant explains only a smallamount of the risk for COPD [1]. It is therefore sug-gested that epigenetic processes may contribute toCOPD risk by mediating the link between genetic vari-ation and environmental exposure [1]. DNA methylationis a heritable, tissue-specific epigenetic modification toDNA that regulates gene expression [4]. It is fundamen-tal to normal development and is known to play a cru-cial role in a number of chronic inflammatory diseasesincluding cancer and aging [13]. The reversibility ofDNA methylation makes it an attractive target for drugdesign. There is strong evidence of dysregulated DNAmethylation in association with COPD [1, 13]. For ex-ample, 349 CpG sites are differentially methylated inwhite blood cells of individuals with COPD compared toindividuals without COPD [1]. Furthermore, severalstudies have assessed aberrant DNA methylation in thewhole lung tissue with varying levels of significance de-pending on sample number [13], integration with geneexpression [2], and integration with GWAS results [14].However, in terms of understanding molecular processesin COPD, these studies are limited by two factors: eitherthey were performed in blood, having relevance as blood-based biomarkers of disease but with limited translationto lung pathology, or they were performed in the wholelung tissue where cell type-specific methylation profileswill likely have masked disease relevant alterations. Fur-ther, mixed cell populations in whole tissue complicatesubsequent cell type-specific mechanistic studies tounderstand disease biology. An exception is the identifica-tion of 1260 CpGs differentially methylated in small air-way epithelial cells between cells isolated from formersmokers with and without COPD, which was associatedwith altered expression of 471 genes [4]. This studystrongly suggests that cell type-specific alterations toDNA methylation exist in association with COPD status.In the present study, we focused on DNA methylationin the lung fibroblasts. Fibroblasts are mesenchymal cellsfound in the stroma of many tissues and in the adventitiaof the vasculature, airways, and parenchyma of adult lungs[15]. They are crucial for stem cell maintenance, lung re-pair, and the homeostasis of the extracellular matrix. Theairway and parenchymal fibroblasts from COPD patientsdiffer in physiological extracellular matrix (ECM) produc-tion [16–18], response to TGFβ [16], response to steroids[16], and proliferation rate [17]. Fibroblasts isolated fromthe lung parenchyma of individuals with COPD are lesscontractile [19, 20], less active to chemoattractant migra-tion [19], and express and secrete increased levels of in-flammatory cytokines CXCL8 and IL-6 [21] compared toparenchymal fibroblasts from individuals without COPD,suggesting they have diminished capacity to mediate re-pair responses potentially contributing to emphysema de-velopment and an enhanced pro-inflammatory profile.Fibroblasts isolated from the intrapulmonary airwayswithin the lungs of COPD patients have been shown tohave increased extracellular matrix deposition [16, 17].The different pathophysiology of emphysema and smallairway disease in COPD highlights the necessity to studyfibroblasts from these two locations distinctly and in com-parison to fully understand the molecular mechanismsunderlying COPD pathogenesis.In this study, we aimed to identify differences ingenome-wide DNA methylation profiles and their asso-ciations with differences in gene expression, between theairway and parenchymal fibroblasts isolated from indi-viduals with and without COPD, to further understandin a cell-specific manner the molecular mechanismsunderlying the pathogenesis of COPD. We found thatparenchymal fibroblasts, but not airway fibroblasts, hadDNA methylation associated alterations in gene expres-sion, implicating DNA methylation as a molecularmechanism underlying parenchymal dysfunction inCOPD. These data provide novel evidence that the air-way and parenchymal fibroblasts are epigenetically dif-ferent in COPD and suggest that alterations to DNAmethylation contribute to COPD pathogenesis.MethodsIsolation and culture of airway and parenchymalfibroblastsPrimary cultures of airway and parenchymal fibroblastsfrom patients with and without COPD (airway: non-COPDn = 8, COPD n = 7; parenchymal: non-COPD n = 17,COPD n = 29, detailed demographics provided in the “Re-sults” section) (defined by the Global Initiative for ChronicObstructive Pulmonary Disease (GOLD) guidelines usingClifford et al. Clinical Epigenetics  (2018) 10:32 Page 2 of 14spirometry) were established from lung biopsies or intra-pulmonary airways and parenchymal lung tissue obtainedfrom lung cancer resections (from disease-free areas) sur-geries, donor lungs, and explant lungs from individualswith COPD undergoing lung transplantation. Airway andparenchymal fibroblasts were derived using the outgrowthtechniques as previously described [22, 23]. Briefly, 2-mm2tissue explants were placed in 6-well tissue culture plateswith DMEM (Sigma) containing 10% fetal bovine serum(GIBCO, Life Technologies), penicillin (100 U/ml), strepto-mycin (100 μg/ml), and l-glutamine (4 mM) in a 5% CO2-humidified incubator. Media were replaced regularly untilcellular outgrowth reached confluence. Tissue pieces wereremoved and destroyed, and cells were harvested usingtrypsin/EDTA solution (Sigma). All samples were gener-ated from cells at passage 4 except a single airway fibro-blast non-COPD donor that was collected at passage 3.Cells at the required passage were grown to confluence in6-well plates and serum starved for 24 h prior to lysis forDNA and RNA isolation. The tissue was obtained, andcells were extracted with the approval of each of the re-search ethics boards for each of the academic institutionsinvolved: Newcastle University (NRES Committee: New-castle and North Tyneside 1 ref:11/NE/0291), McMasterUniversity (Hamilton Integrated Research Ethics BoardRef:00–1839), University of British Columbia (ProvidenceHealth Care Research Ethics Board Ref:H13-02173), Uni-versity of Nottingham (East Midlands Research EthicsCommittee ref: 08/H0407/1), and University Medical Cen-ter Groningen. This study was conducted according to thenational ethical and professional guidelines on the use hu-man body material (“Code of conduct; Dutch federation ofbiomedical scientific societies”; https://www.federa.org/codes-conduct) and the Research Code of the UniversityMedical Center Groningen (https://www.umcg.nl/EN/Research/Researchers/General/ResearchCode/Paginas/default.aspx). The demographics of the subject lung fibro-blasts assessed are given in Table 1.DNA and RNA isolationDNA and RNA were simultaneously isolated from eachsample using the AllPrep DNA/RNA Mini Kit (Qiagen)as per manufacturer’s instructions and assessed for qual-ity and quantity using a NanoDrop™ 8000 Spectropho-tometer (Thermo Fisher Scientific).Bisulfite conversion and DNA methylation arraysSeven hundred fifty nanograms of purified genomicDNA was bisulfite converted using the EZ DNAMethylation Kit (Zymo Research) as per the manufac-turer’s instructions. Specific incubation conditions forthe Illumina Infinium Methylation Assay were used asper the manufacturer’s protocol Appendix. Sampleswere eluted in 12 μl. Bisulfite-converted DNA wasassessed for concentration and quality using aNanoDrop™ 8000 Spectrophotometer (Thermo FisherScientific), and 160 ng of the conversion product wasused for genome-wide DNA methylation quantifica-tion at over 485,000 CpG sites using the IlluminaInfinium HumanMethylation450 BeadChip array, ac-cording to the manufacturer’s protocols.Data quality control and normalizationIDAT files produced by GenomeStudio were importedinto the R statistical software (version 3.2.1) using theminfi package (v. 1.14.0) [24]. The 65 known qualitycontrol SNP probes were used to cluster all samples todetect anomalies within the samples from the samedonor. Probes were excluded from further analysis ac-cording to several criteria: first, 1402 probes were foundTable 1 Donor demographicsAirway fibroblasts Parenchymal fibroblastsNon-COPD COPD Non-COPD vsCOPD p valueNon-COPD COPD Non-COPD vsCOPD p valueAirway vs parenchymalCOPD p valueN 8 7 17 29Gender(M/F)2/6 6/1 0.0187 9/8 18/11 0.544 0.2336AgeMean (SD)63.36 (10.06) 68 (5.63) 0.3097 65.06 (11.31) 65.1 (9.60) 0.9994 0.4489Pack-yearsMean (SD)21.6 (18.89) 53.4 (29.84) 0.0221 31.8 (25.52) 39.8 (19.94) 0.3345 0.1798FEV1%Mean (SD)100.8 (11) 69.7 (18.5) 0.0046 95.6 (17.2) 47.8 (26.3) < 0.0001 0.0457FEV1/FVC %Mean (SD)77.42 (7.2) 59.8 (12) 0.0035 76.6 (6.4) 44.8 (15.5) < 0.0001 0.0182GOLD stage NA 2× GS1, 3×GS2, 2× GS3NA NA 3× GS1, 10× GS2,3× GS3, 13× GS4NA NAM male, F female, COPD chronic obstructive pulmonary disease, SD standard deviation, FEV forced expiratory volume, FVC forced vital capacity, NA not applicableClifford et al. Clinical Epigenetics  (2018) 10:32 Page 3 of 14to have either a detection p value < 0.05 in at least 1% ofsamples or had less than three bead count in at least 5%of samples; second, the 65 SNP probes; third, 59,593probes were found to be cross-hybridized to other partsof the genome, known to be polymorphic at the CpG orexamine single nucleotide polymorphisms [25]; finally,9925 sites on the X chromosome or the Y chromosome.Four hundred fourteen thousand five hundred ninety-two probes remained for analysis.Filtered probes were normalized using the funtoo-Norm algorithm [26], which extends the funNormprocedure [27] and is purported to correct for un-wanted variation while preserving important differ-ences in methylation patterns between different celltypes. We employed the normalization option of prin-cipal components regression with five principal com-ponents. Two values of DNA methylation were used,beta values (β values) and M values. β values are theratio of all methylated probe intensities over total sig-nal intensities (methylated and unmethylated) andhave a range from 0 to 1. They approximately repre-sent percent methylation. M values are the log trans-formation of β values and are more statisticallyrobust [28]. All statistical analyses were performedusing M values, while β values were used forvisualization and interpretability purposes. Principalcomponent analysis was performed for quality controlof the M values.Differential DNA methylation analysisAll statistical analyses were performed using R statisticalsoftware (version 3.2.1). Probes with DNA methylationlevels significantly different between non-COPD donorsand COPD donors in the airway and parenchymal fibro-blasts separately were identified using the limma pack-age using multivariable linear regression on M valuesadjusting for covariates [29] followed by control of the pvalues of the main effect coefficient for false discoveryrate via the Benjamini-Hochberg procedure [30] acrossall CpG sites on the array. For airway fibroblasts, bio-logical sex and smoking pack-years were included as co-variates. As no independently significant sites wereidentified using limma, we looked at aggregated sites toidentify differentially methylated regions (DMRs) usingthe DMRcate package in R [31], which uses Gaussiankernel smoothing to find patterns of differential methy-lation, agnostic to genomic annotation. We used the au-thors’ recommended bandwidth (λ) of 1000 base pairsand scaling factor (C) of 2. For airway fibroblasts, weused a nominal p value cutoff of p < 0.005, which yieldedan analysis of 3432 CpG sites. For the parenchymalfibroblast analysis, all donors were sex- and pack-yearsmatched so a more lenient nominal p value cutoff of p <0.01 was used, which yielded an analysis of 2837 CpGsites.Variable DNA methylation analysisVariable DNA methylation analysis was performed usingthe iEVORA algorithm in R [32] and the recommendedq value cutoff for variably methylated points (DVPs) ofp < 0.001.Reverse transcription and qPCR0.5 μg of RNA was reverse transcribed using SuperScriptIV (Invitrogen), as per the manufacturer’s instructions.The resulting 20 μl cDNA samples were diluted to atotal volume of 200 μl using nuclease-free water. cDNAwas amplified using PerfeCTa SYBR Green FastMix(Quanta bio), with 2 μl template and 200 nM primers ina 10-μl reaction using a Stratagene Mx3000P/3005P sys-tem. Thermal cycler conditions included incubation at95 °C for 10 min, followed by 40 cycles of 95 °C for 10 s,60 °C for 30 s, and 72 °C for 20 s. Data were collected inMxPro; a single product was confirmed by melt curveanalysis, and Ct values were exported to Excel for ana-lysis. Expression was expressed by the ΔΔCt methodrelative to β2-microglobulin (β2M) Ct and mean non-COPD target/β2M ΔCt. Each cDNA was run in triplicatefor both the target gene and housekeeping gene. Themean of the triplicate Cts were taken and the ΔCt be-tween target and housekeeping gene calculated. Themean ΔCt for non-COPD samples was calculated andΔΔCts were calculated relative to the mean non-COPDΔCt. Primer sequences are as follows: β2-microglobulin,forward 5′-AATCCAAATGCGGCATCT-3′, reverse5′-GAGTATGCCTGCCGTGTG-3′; TMEM44, for-ward 5′-GGCACTGGACCTCGCTATTA-3′, reverse5′-CAGGCTCGATGGTCAGCTC-3′; NXN, forward5′-AGACTCTGTTTGGGAGCACG-3′, reverse 5′-TGACTTTGCGAAAGCCATGC-3′; HLX, forward 5′-CGTTTCCAGGTCCCTATGCT-3′, reverse 5′-CGGTTCTGGAACCACACCTT-3′; SPON2, forward 5′-TCCCACGTGGTTGCAGATAC-3′, reverse 5′-TTCCGAAACCGCCCCATTTA-3′; TRPV3, forward 5′-GTGGCCTGCCTGGCG-3′, reverse 5′-GCTTTCATGGCTGGTGAGGT-3′; OAT,forward 5′-CGCTGTCAGATCTGTGGTTT-3′, reverse5′-ACTCCGCGACTAAGTACA-3′; and GRIK2, forward5′-CATGCAGCAAGGTTCTGAGC-3′, reverse 5′-CACTGTCAGAAAGGCGGCTA.Bisulfite PCR-pyrosequencingBisulfite PCR-pyrosequencing was used to validate differ-ences in DNA methylation cg16009558 (GRIK2). BisulfitePCR-pyrosequencing assays were designed with PyroMarkAssay Design 2.0 (Qiagen). The regions of interest wereamplified by PCR using the HotstarTaq DNA polymerasekit (Qiagen) as follows: 15 min at 95 °C (to activate the TaqClifford et al. Clinical Epigenetics  (2018) 10:32 Page 4 of 14polymerase), 45 cycles of 95 °C for 30s, 58 °C for 30s, and72 °C for 30s, and a 5 min 72 °C extension step. For pyrose-quencing, single-stranded DNA was prepared from thePCR product with the Pyromark™ Vacuum Prep Worksta-tion (Qiagen), and sequencing was performed using se-quencing primers on a Pyromark™ Q24 pyrosequencer(Qiagen). The quantitative levels of methylation for eachCpG dinucleotide were calculated with Pyromark Q24 soft-ware (Qiagen). Primer sequences were forwardbiotinylated-5′-ATTTTAGTTTTTTTTATTTAATTTTGGTTT-3′, reverse 5′-CAAAAATTTTACCAAACCCTATTCTACT-3′, sequencing 5′-ACACTACTACACAACTTCTAA-3′. Samples were the same bisulfite converted samplesused on the array. COPD versus non-COPD donors weredemographically matched.ResultsDemographic dataGenome-wide DNA methylation data analysis was per-formed using airway fibroblasts from 7 patients withCOPD and 8 non-COPD controls and parenchymal fi-broblasts from 29 patients with COPD and 17 non-COPD controls (Table 1). Due to the nature of the surgi-cal procedures from which the lung tissue was taken,there were no paired airway and parenchymal cells fromthe same donor in the current dataset. There were nosignificant differences between COPD and non-COPDparenchymal fibroblast samples based on their age, sex,or pack-years smoking history. There were significantdifferences between COPD and non-COPD airway fibro-blast samples based on their sex and pack-years smokinghistory, and these were included as covariates in the stat-istical models. There was also a significant difference inCOPD severity between the airway and parenchymal fi-broblasts as defined by FEV1% and FEV1/FVC%.Differentially methylated DNA regions were associatedwith COPD status in the airway and parenchymalfibroblastsTo understand whether differential DNA methylation wasassociated with COPD in either airway or parenchymal fi-broblasts, we assessed genomic DNA methylation using theIllumina Infinium HumanMethylation450 BeadChip. Afterquality control, 414,592 probes remained for analysis.We looked to identify individually differentially methyl-ated CpG sites with linear regression, with COPD statusas the variable of interest. Linear modeling did not identifyany independently significant sites, after adjusting for mul-tiple tests, in either airway or parenchymal fibroblasts.Subsequently, we assessed differential methylation on ag-gregated sites using the regional DNA methylation Rpackage DMRcate [31], including all CpGs specified by anominal p value limit of 0.005 for airway fibroblasts (3432CpG sites) and a more lenient 0.01 for parenchymalfibroblasts (2837 CpG sites). In airway fibroblasts, 887 dif-ferentially methylated regions were identified in COPDthat contained at least three CpG sites. Six hundred fifty-two of these regions were annotated to a known gene, and35 had a maximum difference in DNA methylation (i.e., atleast one probe displayed a mean difference in methyla-tion) of 20% (Δβ = 0.2) (Fig. 1a and Additional file 1: TableS1). Twelve of these regions had increased DNA methyla-tion associated with COPD, while in 23 regions, DNAmethylation was decreased with COPD status. The five re-gions with the largest maximum difference in DNAmethylation were associated with the genes: TMEM44(Fig. 1b, max Δβ = 0.26), RPH3AL (Fig. 1c, max Δβ = 0.39),HLA-DP1 (Fig. 1d, max Δβ = 0.31), WNT3A (Fig. 1e, maxΔβ = 0.27), and HLA-DRB5 (Fig. 1f, max Δβ = 0.28). Theregions associated with TMEM44, RPH3AL, and HLA-DRB5 did not display a consistent hypomethylation orhypermethylation association with COPD across all of theregion CpGs. However, the regions associated with HLA-DPB1 andWNT3 displayed consistent hypomethylation inassociation with COPD status.In parenchymal fibroblasts, we identified 44 DNA differ-entially methylated regions containing at least three CpGsites, 39 of which were annotated to a gene but only threewith a maximum Δβ of 0.2 (Fig. 2a and Additional file 2:Table S2). Two regions associated with genes HLX (Fig. 2b)and LOC100130872-SPON2 (Fig. 2c) were hypermethy-lated with COPD status while a single region associatedwith NXN was hypomethylated (Fig. 2d).Differentially methylated DNA regions were associatedwith changes in steady-state gene expression with COPDstatus in parenchymal but not airway fibroblastsTo infer a biological significance of the differentiallymethylated DNA regions on cell function in COPD, weperformed targeted gene expression analysis for the fiveairway fibroblast regions with the largest maximum differ-ence in DNA methylation and the three parenchymal fibro-blasts regions. In airway fibroblasts, gene expression ofRPH3AL, HLA-DP1, WNT3A, and HLADRB5 was not de-tectable. Primer functionality was confirmed by positiveamplification in cDNA from the whole lung samples (datanot shown). TMEM44 expression was detected but showedno difference in expression between cells isolated from in-dividuals with COPD versus those without COPD (Fig. 3a).In parenchymal fibroblasts, all transcripts exceptLOC100130872 were detectable. NXN showed no differ-ence in expression between cells isolated from individualswith COPD versus those without COPD (Fig. 3b). However,both HLX (Fig. 3c, p value = 0.0011 unpaired t test) andSPON2 (Fig. 3d, p value = 0.0016 unpaired t test) showed asignificant decrease in gene expression in cells isolated fromindividuals with COPD versus those without COPD.Clifford et al. Clinical Epigenetics  (2018) 10:32 Page 5 of 14Variable DNA methylation is more strongly associatedwith differential steady-state gene expression with COPDstatus in parenchymal fibroblastsIdentifying differentially methylated CpG sites refers tocomparing mean DNA methylation between cases andcontrols and is a standard analytical approach for identi-fying disease-associated CpG sites. Recently, the import-ance of increased DNA methylation variability has beenidentified as a predictor of progression to neoplasia inprecursor cervical cancer legions [32] and has been asso-ciated with type 1 diabetes in three immune effector celltypes [33]. Differentially variable CpG positions (DVPs)can identify larger differences in CpG methylation al-though in a smaller number of the samples [33].As such, we assessed DNA methylation variabilityusing the iEVORA algorithm in R and the associationwith COPD status in the airway and parenchymal fibro-blasts. Using the recommended p value cutoff of 0.001,no DVPs were identified in airway fibroblasts; however,359 CpG sites were identified in parenchymal fibro-blasts. Three hundred twenty-seven of these CpG sitesdisplayed greater variability (as assessed by a largerab cdfeFig. 1 Regional DNA methylation differs in airway fibroblasts isolated from donors with COPD versus non-COPD. a Summary of regional DNAmethylation differences between DNA isolated from donors with and without COPD. Circle size represents the number of probes per regions.Orange = hypomethylated in COPD samples, blue = hypermethylated in COPD. b–f Detailed plots of the five regions with the greatest maximumdifference in DNA methylation between DNA isolated from donors with and without COPDClifford et al. Clinical Epigenetics  (2018) 10:32 Page 6 of 14standard deviation) in cell samples isolated from individ-uals with COPD, 32 CpGs displayed greater variability innon-COPD cells (Additional file 3: Figure S1a). The ma-jority of DVPs were located within CpG islands, but thedistribution of location did not differ from the distribu-tion of the full analysis probes set (Additional file 3: Fig-ure S1b). Two hundred eighty-seven DVPs were geneannotated, with 45 within 200 bp of a transcription startsite. We hypothesized that if the variation in CpGmethylation is contributing to dysfunctional gene ex-pression in parenchymal fibroblasts in COPD, the CpGsites would have relatively stable methylation (low vari-ation) in cells isolated from individuals without COPDand become more variable in cells isolated from individ-uals with COPD. We therefore followed these criteria toidentify CpG sites most likely to be associated with dys-functional gene expression: (1) CpGs with greater vari-ation in cells isolated from individuals with COPD thanthose without, n = 327; (2) CpGs with stable methylationin cells from individuals without COPD as repre-sented by a standard deviation across samples of <0.03, n = 261; (3) CpGs with the greatest variability in cellsisolated from individuals with COPD, as represented by astandard deviation > 0.1, n = 40; and (4) CpGs are within200 bp of a transcription start site and potentially mostlikely to modulate gene expression, n = 5. These remainingab cdFig. 2 Regional DNA methylation differs in parenchymal fibroblasts isolated from donors with COPD versus non-COPD. a Summary of regionalDNA methylation differences between DNA isolated from donors with and without COPD. Circle size represents the number of probes perregions. Orange = hypomethylated in COPD samples, blue = hypermethylated in COPD. b–d Detailed plots of the three regions with a maximumdifference in DNA methylation between DNA isolated from donors with and without COPD of greater than 20% (beta value difference of 0.2)Clifford et al. Clinical Epigenetics  (2018) 10:32 Page 7 of 14five CpG sites were annotated to the promoters of threeseparate genes, TRPV3 (cg11475555, Fig. 4a), OAT(cg02065151, Fig. 4b), GRIK2 (cg24753760, Fig. 4c;cg16009558, Fig. 4d; and cg06247406, Fig. 4e). GRIK2CpG cg16009558 was validated by pyrosequencing andconfirmed a significant correlation between pyrosequenc-ing and array methylation (r = 0.6329, p value = < 0.001,Fig. 4f).We assessed the potential biological relevance of variableDNA methylation at these sites by measuring steady-stategene expression. While TRPV3 showed a non-significanttrend toward increased gene expression in cells isolatedfrom individuals with COPD (p = 0.0618 Welch’s t test,Fig. 5a), both OAT (p = 0.0109 Welch’s t test, Fig. 5b) andGRIK2 (p = 0.0358 Welch’s t test, Fig. 5c) showed a signifi-cant increase in gene expression in cells isolated from indi-viduals with COPD versus those without COPD. Inparticular, in cells from four COPD donors, GRIK2 showeda greater than 20-fold increase in gene expression. Subse-quently, we asked, for these specific genes, whetherexpression correlated with DNA methylation in parenchy-mal fibroblasts isolated from individuals with COPD. OATexpression did not correlate with methylation ofcg02065151 (Fig. 6a). Although TRPV3 showed a non-significant trend toward increased gene expression betweenparenchymal fibroblasts isolated from individuals with andwithout COPD, TRPV3 expression did correlate withcg11475555 methylation, suggesting DNA methylationmay play a role in regulating TRPV3 expression (Fig. 6b).The expression of GRIK2 strongly correlated with methyla-tion of all three GRIK2 CpGs determined as DVPs inCOPD (cg24753760, Fig. 6c; cg06247406, Fig. 6d;cg16009558, Fig. 6e), highly suggestive of aberrant GRIK2CpG methylation leading to increased GRIK2 expression inCOPD parenchymal fibroblasts.DiscussionThe main novel findings of this study are that alterationsto individual CpG site methylation occur with COPDstatus in lung cells, that dysfunction is more stronglya bc dFig. 3 Expression of genes associated with differentially methylated DNA regions. a TMEM44 expression in airway fibroblasts isolated fromindividuals with and without COPD. b NXN expression in parenchymal fibroblasts isolated from individuals with and without COPD. c HLXexpression in parenchymal fibroblasts isolated from individuals with and without COPD. d SPON2 expression in parenchymal fibroblasts isolatedfrom individuals with and without COPD. **p < 0.01 compared with non-COPD control by unpaired t test, n = 15 non-COPD, n = 29 COPDClifford et al. Clinical Epigenetics  (2018) 10:32 Page 8 of 14associated with cell function (gene expression) in paren-chymal fibroblasts than airway fibroblasts and that vari-ability in DNA methylation may represent a strongeranalytical method to identify aberrant DNA methylationassociated gene expression than differential methylationin a heterogeneous disease like COPD.DNA methylation is considered to be an importantbiological mechanism, which integrates genetic and en-vironmental risk factors contributing to COPD patho-genesis. DNA methylation is aberrant in the blood andwhole lung tissue DNA isolated from individuals withCOPD; however, these mixed cell samples, and with re-gard to blood separation from the site of disease, com-plicate defining molecular mechanisms underlyingCOPD pathogenesis. Fibroblasts are critical to the de-position of the physiological extracellular matrix (ECM)and are considered to contribute to pathological ECMscar tissue deposition in many obstructive airway dis-eases including COPD. Fibroblasts reside within the lungparenchyma and airways, and it is becoming increasinglyclear that these two locations define a distinct phenotypeof fibroblast that likely contributes differentially to lungdisease pathogenesis. In this paper, we use purified pop-ulations of the airway and parenchymal fibroblasts fromindividuals with and without COPD to identify disease-associated, cell type-specific alterations to DNAmethylation.We have considered whether any technical issues mayhave affected our results. Our study was performed onhuman airway and parenchymal fibroblasts, expanded invitro to passage 4 (except a single airway non-COPDdonor that was collected at passage 3). This provides thebenefit of a single cell type population for DNA methy-lation analysis, which we know to be highly sensitive toa bc de fFig. 4 CpG methylation is differentially variable in parenchymal fibroblasts isolated from donors with COPD versus non-COPD. DNA methylationarray data for five CpGs identified as differentially variable in association with COPD status. a cg11475555 within 200 bp of the TRPV3 genetranscription start site. b cg02065151 within 200 bp of the OAT gene transcription start site. c cg24753760. d cg16009558. e cg06247406 all within200 bp of the GRIK2 gene transcription start site. f Pyrosequencing validation of cg16009559Clifford et al. Clinical Epigenetics  (2018) 10:32 Page 9 of 14mixed cell populations. By using cells at the same pas-sage, grown under the same conditions, it is unlikelythat the differences in DNA methylation and gene ex-pression are due to cell culture effects. Furthermore, thecells used in the study were collected from four differentsites to minimize isolation and culture technique effects.Secondly performing the Illumina Infinium Human-Methylation450 BeadChip array on bisulfite DNA doesnot allow 5-methylcytosine to be distinguished from 5-hydroxymethylcytosine. 5-Hydroxymethylcytosine is an oxi-dized version of 5-methylcytosine produced by Ten-eleventranslocation (TET) enzymes as an intermediate base dur-ing DNA demethylation. 5-Hydroxymethylcytosine is asso-ciated with gene transcription and gene translation but viapotentially different mechanisms to 5-methylcytosine.Further investigation of the mechanisms by which DNAmethylation regulates gene expression in the targets iden-tified here is warranted and should include Tet-assistedbisulfite pyrosequencing to distinguish between 5-methylcytosine and 5-hydroxymethylcytosine. Finally, theIllumina Infinium HumanMethylation450 BeadChip array,while covering a valid representation of the genome, is nottruly genome-wide, and an analysis using bisulfite sequen-cing or the more recently released Illumina EPIC arraymay highlight further modifications to fibroblast DNAmethylation in association with COPD status.Our initial linear modeling approaches did not identifyany individual differentially methylated CpG sites associ-ated with COPD, potentially due to the large number offeatures and low sample number, despite the current studybeing the largest to date of purified cell types in COPD.There are data reduction techniques that we could haveused, for example, selecting CpG sites within close (200/1500 bp) proximity to transcription start sites, as these areconsidered more likely to drive transcriptional changes.However, a secondary part of this study was to also under-stand how different types of DNA methylation, for ex-ample, promoter versus gene body, affected steady-stategene expression. For this reason, we chose to undertakeregional analysis using the R package DMRcate, consider-ing that while individual site differences may be small, ifthey are persistent across a region, the power to detectthem will be greater. The regional analysis did identifyregions of DNA differentially methylated between cellsisolated from individuals with and without COPD. Fur-thermore, the location of these regions differed betweenairway and parenchymal fibroblasts, confirming that dif-ferent molecular mechanisms are induced in these twocell types in association with COPD.To assess the likelihood of statistically significant dif-ferentially methylated DNA regions having downstreambiological effects, the absolute difference between themeans of the β values of the COPD and non-COPDwere calculated for each CpG within a region andabcFig. 5 Variable CpG methylation in parenchymal fibroblasts fromCOPD donors is associated with differential gene expression. Geneexpression data generated by qPCR for the three genes associatedwith variable CpG site methylation in parenchymal fibroblasts fromCOPD donors. a TRPV3. b OAT. c GRIK2. *p < 0.05, **p < 0.01compared with non-COPD control by Welch’s t test, n = 15 non-COPD,n = 29 COPDClifford et al. Clinical Epigenetics  (2018) 10:32 Page 10 of 14referred to as the delta beta (Δβ). We considered a max-imum difference in methylation of greater than 20%, thatis, at least one CpG site within the designated regionmust have had a mean difference in methylation be-tween samples from COPD and non-COPD donors ofgreater than 20% (Δβ = 0.2), to be of biological interest.After this filtering step, 45 regions remained of interestin the airway fibroblasts, while only 3 remained for par-enchymal fibroblasts. This suggested that airway fibro-blasts are likely predisposed to larger alterations in DNAmethylation than parenchymal fibroblasts and thereforethat they may contribute to a greater extent to COPDpathogenesis than previously considered. However, tounderstand the biological significance of these DNAmethylation differences in greater detail, it was import-ant to establish whether DNA methylation changes werealtering gene expression. We assessed gene expressionfor all three parenchymal fibroblast regions and the fiveregions from airway fibroblasts with the largest max-imum difference in DNA methylation between non-COPD and COPD samples. None of the airway fibro-blast regions was associated with a difference in steady-state gene expression, while two of the three parenchy-mal regions did coincide with a significant difference ingene expression. This highlights the necessity for inte-grating DNA methylation and gene expression data toconfer biological importance. However, it does notexclude that the differences in DNA methylation in theairway fibroblasts may affect gene expression at regionswe did not assess by qPCR or in response to, for ex-ample, lung/cell exposure to environmental COPD riskfactors (cigarette smoke/air pollution), and this warrantsceda bFig. 6 GRIK2 CpG site methylation correlates with gene expression. Correlations between gene expression levels and DNA methylation (hererepresented as M values, the log transformation of β values which are more statistically robust) for the three differentially variable CpG sites/genes. a OAT. b TRPV3. c–e GRIK2. R2 values and p values are shown within the plotsClifford et al. Clinical Epigenetics  (2018) 10:32 Page 11 of 14further investigation. Of interest, although one of thesignificant parenchymal fibroblast regions was within1500 bp of the transcription start site for SPON2, the re-gion for HLX, which also showed a significant differencein gene expression, is located within the gene body, em-phasizing that location within the gene also does not ne-cessarily confer effect at the gene expression level. Weisolated DNA and RNA simultaneously from the samesamples, a valuable resource, and a limitation of thisstudy is that we analyzed our RNA by targeted qPCR ra-ther than genome-wide microarray or sequencing, whichwould have been interesting to gain more informationregarding the association between gene expression andDNA methylation.Evidence for a greater dysfunction in DNA methyla-tion regulated gene expression in parenchymal ratherthan airway fibroblasts was further strengthened by asecondary analysis of DNA methylation variability thathas previously been validated in other diseases [32, 33].Identifying differentially methylated CpG sites comparesthe mean level of DNA methylation between cases andcontrol while assessment of differential variability(DVPs) essentially identifies individual sites displaying“epigenetic outliers” in heterogeneous populations. Thecurrent analysis did not identify any DVPs in airway fi-broblasts but did identify 359 DVPs in parenchymal fi-broblasts. As genome-wide gene expression was notavailable for these samples to allow full integration toDNA methylation data, we used a filtering process toidentify those CpG sites we hypothesized to be the mostlikely to identify differential disease-associated gene ex-pression (gene annotated, greater variability in COPDthan non-COPD, stable variation in non-COPD and lar-gest variation in COPD, within 200 bp of a transcriptionstart site) and assessed the expression of the identifiedgenes by qPCR. The five sites that remained associatedwith three different genes, two of which showed a sig-nificant difference in expression between samples fromCOPD and non-COPD donors (OAT and GRIK2) and athird that displayed a strong trend (p = 0.0618) towarddifferential expression (TRPV3). The higher success rateof identifying differences in DNA methylation associatedwith differential gene expression by variation analysisthan absolute differential analysis suggests that differen-tial variation analysis may be a preferable method foridentifying DNA methylation regulated alterations ingene expression associated with heterogeneous diseaseslike COPD. However, we acknowledge that a full integra-tion of genome-wide DNA methylation and gene expres-sion data and permutation testing is required todefinitively prove this finding.The largest difference in gene expression was seen forGRIK2 with some donors displaying a greater than 80-foldincrease in expression compared to non-COPD donors.GRIK2 expression also directly correlated with DNAmethylation levels in matched donor samples, indicatingthat DNA methylation is tightly associated with GRIK2gene expression. The direct relationship between bothGRIK2 and OAT expression and DNA methylation wassurprising, that is, for both OAT and GRIK2, DNA methy-lation was higher in samples from COPD donors thannon-COPD donors, a phenomenon at gene promotersthat is normally associated with gene repression. Despitethis, qPCR data showed increased expression of both OATand GRIK2 in samples from COPD donors. Again, thishighlights that although general trends have been identi-fied for the regulation of gene expression by DNA methy-lation, the more studies that are performed, the moreexceptions to those trends arise and the more important itis to look at DNA methylation in context with gene ex-pression data rather than as an indicator of gene expres-sion. Furthermore, future studies are warranted tounderstand the mechanisms that mediate the relationshipbetween DNA methylation and gene expression. Inaddition to differences in 5-methylcytosine and 5-hydroxymethylcytosine ratios mentioned above, thesecould include associations with methyl-DNA binding pro-teins, modifications to histone tails, and the ability to re-cruit specific transcription factors. This is likely to notonly be DNA region specific but also cell type and diseasecontext specific, and regimens to include these mechanis-tic studies should be included in future studies of DNAmethylation in COPD.GRIK2 belongs to the kainate family of glutamate re-ceptors, and a small nucleotide polymorphism(rs6570989) in GRIK2 has been associated with a statusof “current smoker” [34]. Our parenchymal fibroblastcohort was matched for smoking pack-years but did in-clude a mixture of current, ex-, and non-smokers, andwe were therefore interested in whether the variability inDNA methylation was driven by smoking history. How-ever, neither GRIK2 DNA methylation levels nor geneexpression levels correlated with pack-years (Add-itional file 4: Figure S2a and b) or smoking status (Add-itional file 4: Figure S2c and d). We also assessed GRIK2single nucleotide polymorphisms (SNPs) in the donorsand the T allele SNP associated with current smokingwas only identified in a single non-COPD donor, to-gether suggesting that COPD status rather smoking sta-tus and genotype are driving the observed associationbetween GRIK2 DNA methylation and gene expression.ConclusionsTo conclude, we show that differences in DNA methyla-tion are associated with COPD status and that these al-terations can be associated with differences in functionalgene expression for specific genes. Differences in DNAmethylation profiles differ between airway and parenchymalClifford et al. Clinical Epigenetics  (2018) 10:32 Page 12 of 14fibroblasts suggesting that DNA methylation contributesdifferently to disease pathology in these two distinct tissuetypes with different functions within the lung. The associ-ation between gene expression and DNA methylation wascomplex and not predictable by the traditional dogma ofDNA methylation regulated gene expression. Finally, asses-sing differentially variable DNA methylation may be a bet-ter approach to identifying aberrant DNA methylationregulated gene expression in heterogeneous diseases suchas COPD.Additional filesAdditional file 1: Table S1. (.MWD): DNA regions differentiallymethylated with COPD status in airway fibroblasts. Summary of the 35gene annotated regions containing a minimum of three CpG probes anda maximum difference in DNA methylation of at least 20% (difference inbeta value of 0.2). A positive delta beta identifies decreased DNAmethylation in cells isolated from individuals with COPD, while anegative value identifies increased DNA methylation in cells isolated fromindividuals with COPD. (DOCX 18 kb)Additional file 2: Table S2. (.MWD): DNA regions differentiallymethylated with COPD status in parenchymal fibroblasts. Summary of thethree gene annotated regions containing a minimum of three CpGprobes and a maximum difference in DNA methylation of at least 20%(difference in beta value of 0.2). A positive delta beta identifies decreasedDNA methylation in cells isolated from individuals with COPD, while anegative value identifies increased DNA methylation in cells isolated fromindividuals with COPD. (DOCX 14 kb)Additional file 3: Figure S1. (.EPS): CpG methylation variation inparenchymal fibroblasts in COPD samples. A) Plot of the difference inmethylation standard deviation of samples from donors with COPD anddonors without COPD at the 359 CpGs differentially variably methylatedbetween samples from donors with COPD and donors without COPD. B)Representation of the CpG type with which DVPs were associated versusCpG types of all CpGs included in the analysis. (EPS 1922 kb)Additional file 4: Figure S2. (.EPS): GRIK2 methylation and expressiondo not correlate with smoking pack-years or smoking status. A) Correlationplot of GRIK2 cg16006558 methylation with smoking pack years. B) Cor-relation plot of GRIK2 expression with smoking pack-years. Neither CpGmethylation nor gene expression correlates with smoking pack-years. C)GRIK2 expression separated by smoking status. D) GRIK2 CpG methylationseparated by smoking status. Smoking status does not drive alterations toGRIK2 DNA methylation or gene expression. (EPS 2269 kb)AbbreviationsCOPD: Chronic obstructive pulmonary disease; DVP: Variably methylatedpoints; ECM: Extracellular matrix; FEV: Forced expiratory volume; FVC: forcedvital capacity; GOLD: Global Initiative for Chronic Obstructive PulmonaryDisease; GWAS: Genome-wide genetic association studies; SD: Standarddeviation; TGFβ: Transforming growth factor beta; Δβ: Delta betaAcknowledgementsNot applicableFundingFunding bodies had no role in the design of the study or collection,analysis, or interpretation of data and in writing the manuscript. RLCwas supported by the European Respiratory Society (ERS) and theCanadian Thoracic Society (CTS)/Canadian Lung Association, joint ERS/CTS Long-Term Research Fellowship LTRF2013, funding from RosetreesTrust, The Henry Smith Charity, and a current Fellowship from theMedical Research Foundation (MRF)/Asthma UK. Support for this studywas provided by Canadian Institutes of Health Research (CIHR) andBritish Columbia Lung Association operating grants. T-LH was supportedby CIHR, Michael Smith Health Research Foundation (MSHRF), and ParkerB. Francis New Investigator awards. PN is supported by the Frederick E.Hargreave Teva Innovation Chair in Airway Diseases. AK and AF receivedfunding for fibroblast harvesting from MRC COPDMAP Workpackage 3.MSK is the Canada Research Chair in Social Epigenetics, Senior Fellow ofthe Canadian Institute for Advanced Research and Sunny Hill BCLeadership Chair in Child Development. LM McEwen is supported by aCIHR Doctoral Research Award (F15-04283).Availability of data and materialsThe datasets generated and/or analyzed during the current study areavailable in the Gene Expression Omnibus (GEO) repository, https://www.ncbi.nlm.nih.gov/geo/ REF (to be updated at publication).Authors’ contributionsRLC devised the study; cultured the cells; isolated the DNA; analyzed thearray, qPCR, and pyrosequencing data; and wrote the manuscript. NFanalyzed the array data. JLM and LMM designed the microarray experimentsand performed the sample array processing. JP generated the qPCR andpyrosequencing data. AJF provided the cells and contributed to themanuscript preparation. MSK facilitated RLC’s analysis of array data andcontributed to the manuscript preparation. CAB and PN provided the lungtissue and isolated cells. TLH provided the lung tissue and isolated cells,provided a statistician (NF) for the analysis, and contributed to themanuscript preparation. AJK oversaw the project and contributed to themanuscript preparation. All authors read and approved the final manuscript.Ethics approval and consent to participateThe tissue was obtained and cells extracted with the approval of each of theresearch ethics boards for each of the academic institutions involved:Newcastle University (NRES Committee: Newcastle and North Tyneside 1Ref:11/NE/0291); McMaster University (Hamilton Integrated Research EthicsBoard Ref:00-1839); University of British Columbia (Providence Health CareResearch Ethics Board Ref:H13-02173); University of Nottingham (EastMidlands Research Ethics Committee Ref: 08/H0407/1); and UniversityMedical Center Groningen. This study was conducted according to thenational ethical and professional guidelines on the use human body material(“Code of conduct; Dutch federation of biomedical scientific societies”;https://www.federa.org/codes-conduct) and the Research Code of theUniversity Medical Center Groningen (https://www.umcg.nl/EN/Research/Researchers/General/ResearchCode/Paginas/default.aspx).Consent for publicationNot applicableCompeting interestsThe authors declare that they have no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Nottingham NIHR Biomedical Research Centre, Nottingham MRC MolecularPathology Node, Division of Respiratory Medicine, University of Nottingham,Nottingham University Hospitals NHS Trust, City Hospital, Nottingham, UK.2Centre for Heart Lung Innovation, University of British Columbia, Vancouver,Canada. 3Department of Medical Genetics, Centre for Molecular Medicineand Therapeutics, BC Children’s Hospital Research Institute, University ofBritish Columbia, Vancouver, Canada. 4Institute of Cellular Medicine,Newcastle University, Newcastle upon Tyne, UK. 5Department of Pathologyand Medical Biology, University of Groningen, University Medical Center,Groningen, Groningen, The Netherlands. 6GRIAC (Groningen ResearchInstitute of Asthma and COPD), University of Groningen, University MedicalCenter, Groningen, The Netherlands. 7Firestone Institute for RespiratoryHealth, St Joseph’s Healthcare and Department of Medicine, McMasterUniversity, Hamilton, Ontario, Canada. 8Department of Anaesthesiology,Pharmacology, & Therapeutics, University of British Columbia, Vancouver,Canada.Clifford et al. Clinical Epigenetics  (2018) 10:32 Page 13 of 14Received: 2 January 2018 Accepted: 25 February 2018References1. Qiu W, Baccarelli A, Carey VJ, Boutaoui N, Bacherman H, Klanderman B,Rennard S, Agusti A, Anderson W, Lomas DA, DeMeo DL. Variable DNAmethylation is associated with chronic obstructive pulmonary disease andlung function. Am J Respir Crit Care Med. 2012;185:373–81.2. Yoo S, Takikawa S, Geraghty P, Argmann C, Campbell J, Lin L, Huang T, TuZ, Foronjy RF, Spira A, et al. Integrative analysis of DNA methylation andgene expression data identifies EPAS1 as a key regulator of COPD. PLoSGenet. 2015;11:e1004898.3. Sood A, Petersen H, Blanchette CM, Meek P, Picchi MA, Belinsky SA,Tesfaigzi Y. Wood smoke exposure and gene promoter methylation areassociated with increased risk for COPD in smokers. Am J Respir CritCare Med. 2010;182:1098–104.4. Vucic EA, Chari R, Thu KL, Wilson IM, Cotton AM, Kennett JY, Zhang M,Lonergan KM, Steiling K, Brown CJ, et al. DNA methylation is globallydisrupted and associated with expression changes in chronic obstructivepulmonary disease small airways. Am J Respir Cell Mol Biol. 2014;50:912–22.5. Repapi E, Sayers I, Wain LV, Burton PR, Johnson T, Obeidat M, Zhao JH,Ramasamy A, Zhai G, Vitart V, et al. Genome-wide association studyidentifies five loci associated with lung function. Nat Genet. 2010;42:36–44.6. Hancock DB, Eijgelsheim M, Wilk JB, Gharib SA, Loehr LR, Marciante KD,Franceschini N, van Durme YM, Chen TH, Barr RG, et al. Meta-analyses ofgenome-wide association studies identify multiple loci associated withpulmonary function. Nat Genet. 2010;42:45–52.7. Lambrechts D, Buysschaert I, Zanen P, Coolen J, Lays N, Cuppens H, Groen HJ,Dewever W, van Klaveren RJ, Verschakelen J, et al. The 15q24/25 susceptibilityvariant for lung cancer and chronic obstructive pulmonary disease isassociated with emphysema. Am J Respir Crit Care Med. 2010;181:486–93.8. Young RP, Hopkins RJ, Hay BA, Epton MJ, Black PN, Gamble GD. Lungcancer gene associated with COPD: triple whammy or possibleconfounding effect? Eur Respir J. 2008;32:1158–64.9. Young RP, Hopkins RJ, Whittington CF, Hay BA, Epton MJ, Gamble GD.Individual and cumulative effects of GWAS susceptibility loci in lung cancer:associations after sub-phenotyping for COPD. PLoS One. 2011;6:e16476.10. Young RP, Whittington CF, Hopkins RJ, Hay BA, Epton MJ, Black PN, GambleGD. Chromosome 4q31 locus in COPD is also associated with lung cancer.Eur Respir J. 2010;36:1375–82.11. Van Durme YM, Eijgelsheim M, Joos GF, Hofman A, Uitterlinden AG,Brusselle GG, Stricker BH. Hedgehog-interacting protein is a COPDsusceptibility gene: the Rotterdam Study. Eur Respir J. 2010;36:89–95.12. Chappell SL, Daly L, Lotya J, Alsaegh A, Guetta-Baranes T, Roca J, RabinovichR, Morgan K, Millar AB, Donnelly SC, et al. The role of IREB2 andtransforming growth factor beta-1 genetic variants in COPD: a replicationcase-control study. BMC Med Genet. 2011;12:24.13. Sundar IK, Yin Q, Baier BS, Yan L, Mazur W, Li D, Susiarjo M, Rahman I. DNAmethylation profiling in peripheral lung tissues of smokers and patientswith COPD. Clin Epigenetics. 2017;9:38.14. Morrow JD, Cho MH, Hersh CP, Pinto-Plata V, Celli B, Marchetti N, Criner G,Bueno R, Washko G, Glass K, et al. DNA methylation profiling in human lungtissue identifies genes associated with COPD. Epigenetics. 2016;11(10):730–39.15. Phan SH. Biology of fibroblasts and myofibroblasts. Proc Am ThoracSoc. 2008;5:334–7.16. Brandsma CA, Timens W, Jonker MR, Rutgers B, Noordhoek JA, Postma DS.Differential effects of fluticasone on extracellular matrix production byairway and parenchymal fibroblasts in severe COPD. Am J Physiol Lung CellMol Physiol. 2013;305:L582–9.17. Hallgren O, Nihlberg K, Dahlback M, Bjermer L, Eriksson LT, Erjefalt JS,Lofdahl CG, Westergren-Thorsson G. Altered fibroblast proteoglycanproduction in COPD. Respir Res. 2010;11:55.18. Dessalle K, Narayanan V, Kyoh S, Mogas A, Halayko AJ, Nair P, Baglole CJ,Eidelman DH, Ludwig MS, Hamid Q. Human bronchial and parenchymalfibroblasts display differences in basal inflammatory phenotype andresponse to IL-17A. Clin Exp Allergy. 2016;46:945–56.19. Togo S, Holz O, Liu X, Sugiura H, Kamio K, Wang X, Kawasaki S, Ahn Y,Fredriksson K, Skold CM, et al. Lung fibroblast repair functions in patientswith chronic obstructive pulmonary disease are altered by multiplemechanisms. Am J Respir Crit Care Med. 2008;178:248–60.20. Campbell JD, McDonough JE, Zeskind JE, Hackett TL, Pechkovsky DV,Brandsma CA, Suzuki M, Gosselink JV, Liu G, Alekseyev YO, et al. A geneexpression signature of emphysema-related lung destruction and itsreversal by the tripeptide GHK. Genome Med. 2012;4:67.21. Zhang J, Wu L, Qu JM, Bai CX, Merrilees MJ, Black PN. Pro-inflammatoryphenotype of COPD fibroblasts not compatible with repair in COPD lung.J Cell Mol Med. 2012;16:1522–32.22. Pechkovsky DV, Hackett TL, An SS, Shaheen F, Murray LA, Knight DA.Human lung parenchyma but not proximal bronchi produces fibroblastswith enhanced TGF-beta signaling and alpha-SMA expression. Am J RespirCell Mol Biol. 2010;43:641–51.23. Noordhoek JA, Postma DS, Chong LL, Menkema L, Kauffman HF, Timens W,van Straaten JF, van der Geld YM. Different modulation of decorinproduction by lung fibroblasts from patients with mild and severeemphysema. COPD. 2005;2:17–25.24. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP,Hansen KD, Irizarry RA. Minfi: a flexible and comprehensiveBioconductor package for the analysis of Infinium DNA methylationmicroarrays. Bioinformatics. 2014;30:1363–9.25. Price ME, Cotton AM, Lam LL, Farre P, Emberly E, Brown CJ, Robinson WP,Kobor MS. Additional annotation enhances potential for biologically-relevant analysis of the Illumina Infinium HumanMethylation450 BeadChiparray. Epigenetics Chromatin. 2013;6:4.26. Oros Klein K, Grinek S, Bernatsky S, Bouchard L, Ciampi A, Colmegna I,Fortin JP, Gao L, Hivert MF, Hudson M, et al. funtooNorm: an Rpackage for normalization of DNA methylation data when there aremultiple cell or tissue types. Bioinformatics. 2016;32:593–5.27. Fortin JP, Labbe A, Lemire M, Zanke BW, Hudson TJ, Fertig EJ, GreenwoodCM, Hansen KD. Functional normalization of 450k methylation array dataimproves replication in large cancer studies. Genome Biol. 2014;15:503.28. Du P, Zhang X, Huang CC, Jafari N, Kibbe WA, Hou L, Lin SM. Comparison ofBeta-value and M-value methods for quantifying methylation levels bymicroarray analysis. BMC Bioinformatics. 2010;11:587.29. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. Limmapowers differential expression analyses for RNA-sequencing and microarraystudies. Nucleic Acids Res. 2015;43:e47.30. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practicaland powerful approach to multiple testing. Journal of the Royal StatisticalSociety Series B (Methodological). 1995:289–300.31. Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, VL R, Clark SJ,Molloy PL. De novo identification of differentially methylated regions in thehuman genome. Epigenetics Chromatin. 2015;8:6.32. Teschendorff AE, Gao Y, Jones A, Ruebner M, Beckmann MW, Wachter DL,Fasching PA, Widschwendter M. DNA methylation outliers in normal breast tissueidentify field defects that are enriched in cancer. Nat Commun. 2016;7:10478.33. Paul DS, Teschendorff AE, Dang MA, Lowe R, Hawa MI, Ecker S, Beyan H,Cunningham S, Fouts AR, Ramelius A, et al. Increased DNA methylationvariability in type 1 diabetes across three immune effector cell types. NatCommun. 2016;7:13555.34. Vink JM, Smit AB, de Geus EJ, Sullivan P, Willemsen G, Hottenga JJ, Smit JH,Hoogendijk WJ, Zitman FG, Peltonen L, et al. Genome-wide association studyof smoking initiation and current smoking. Am J Hum Genet. 2009;84:367–79.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Clifford et al. Clinical Epigenetics  (2018) 10:32 Page 14 of 14


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items