The Nasal Methylome and Childhood Atopic Asthma

Abstract

Background

Given strong environmental influence on both epigenetic marks and allergic asthma in children, the epigenetic alterations in respiratory epithelia may provide insight into allergic asthma.

Objective

To identify DNA methylation and gene expression changes associated with childhood allergic persistent asthma.

Methods

We compared genomic DNA methylation patterns and gene expression in African American children with persistent atopic asthma[N=36] versus healthy controls[N=36]. Results were validated in an independent population of asthmatic children[N=30] using a shared healthy control population[N=36] and in independent population of Caucasian adult atopic asthmatics[N=12] and controls[N=12].

Results

We identified 186 genes with significant methylation changes, differentially methylated regions(DMRs) or differentially methylated probes(DMPs), after adjustment for age, gender, race/ethnicity, batch effects, inflation, and multiple comparisons. Genes differentially methylated included those with established roles in asthma and atopy, genes related to extracellular matrix, immunity, cell adhesion, epigenetic regulation, and airflow obstruction. The methylation changes were substantial (median 9.5%, range:2.6–29.5%). Hypo- and hyper-methylated genes were associated with increased and decreased gene expression respectively (P<2.8x10⁻⁶ for DMRs and P<7.8x10⁻¹⁰ for DMPs). Quantitative analysis in 53 differentially expressed genes demonstrated that 32(60%) have significant methylation-expression relationships within 5kb of the gene. 10 loci selected based on the relevance to asthma, magnitude of methylation change, and methylation-expression relationships were validated in an independent cohort of children with atopic asthma. 67/186 genes also have significant asthma-associated methylation changes in nasal epithelia of adult Caucasian asthmatics.

Conclusions

Epigenetic marks in respiratory epithelia are associated with allergic asthma and gene expression changes in inner-city children.

INTRODUCTION

The increase in the prevalence, incidence, and severity of asthma over the last 20 years(1) provides strong evidence that exposures play an important role in this disease. While common genetic variants explain only a small portion of asthma heritability(2), epigenetic changes could potentially explain both the non-Mendelian(3) and parent-of-origin(4) patterns of inheritance that are characteristic of asthma. Additionally, epigenetic marks can be influenced by the environment(5), these marks have been shown to affect the expression of transcription factors that alter the maturation of T lymphocytes(6–8), and we have demonstrated a causal relationship between DNA methylation and both Th₂ immunity and allergic airway disease in mice(9).

Recently, we have shown that DNA methylation marks in peripheral blood mononuclear cells (PBMCs) are associated with allergic asthma(10) and account for 13.5% of the variation in serum IgE concentrations(11). However, the airway epithelium is the primary interface with the environment, interacts with allergens(12) and other environmental stimuli(13), and represents a potentially important mediator of allergic airway disease. Gene expression profiles of the asthmatic airway epithelium have identified genes associated with exposure to endotoxin and house dust mite allergen(14) as well as cigarette smoke(15), asthma(15), and Th₂-high vs –low subphenotypes of disease(16). More recently, it has been demonstrated that gene expression in nasal epithelia is a proxy measure for gene expression of the lower airway epithelium(17). However, no study to date has comprehensively characterized genome-wide DNA methylation patterns and associated changes in gene expression in atopic asthmatic nasal epithelia.

METHODS

Study Populations

Our primary study population consisted of inner-city children aged 10–12 years with both atopy and persistent asthma (cases, N=36) or without atopy or asthma (healthy controls; N=36) (10). Study subjects were recruited by six sites supported by the Inner City Asthma Consortium (ICAC) from census tracts that contain at least 20% of households below the U.S. government poverty level (Supplemental Table S1). We limited our study population to 36 cases and 36 controls with at least 80% ciliated epithelial cells visualized from slides obtained from nasal brushings and by expression of the FOXJ1 gene (Supplemental Table S2). The validation population consisted of 30 African Americans aged 10–12 years with atopic asthma, using the same definition as the derivation cohort, collected by the ICAC independent of the primary study population. The original 36 control samples were used in the validation analysis and will be referred to as ‘shared controls’. The second validation population consisted of 24 Caucasian adults (age range 24–74), 12 with asthma and 12 without asthma, all with at least 80% ciliated epithelial cells visualized from slides obtained from nasal brushings, recruited at National Jewish Health. In this cohort, all asthmatics were also atopic, as assessed by positive skin prick testing for multiple allergens or RAST/Phadiatop tests. Control were required to have IgE level at visit < 100 but did not have additional allergy testing performed.

DNA Methylation and Gene Expression Data Collection

DNA methylation was measured on Illumina’s Infinium Human Methylation 450k BeadChip and validated internally and externally using pyrosequencing with custom designed primers (Supplemental Table S3). Gene expression was assessed on Agilent Human Gene Expression arrays (G3 SurePrint 8x60k). DNA methylation and gene expression array data have been deposited to the Gene Expression Omnibus (GEO) (GSE65205).

Overview of Statistical Analyses

The goal of our analyses was to determine whether DNA methylation and gene expression changes in nasal epithelia are associated with atopic asthma. Overall study design and workflow are presented schematically in Supplemental Figure S1. We identified DNA methylation changes associated with asthma for both single CpG motifs (differentially methylated probes [DMPs]) and differentially methylated regions (DMRs). While identification of DMPs is the most commonly used method for identification of methylation changes in Illumina arrays(18), our rationale for identification of DMRs is three-fold: first, identification of regions is conceptually consistent with what is known about DNA methylation patterns in the human genome(19); second, it increases power to detect associations(20); and third, it has been used in other diseases(21, 22). We also performed two exploratory analyses: (1) to determine if any of the DMRs or DMPs identified in nasal epithelia were associated with nasal corticosteroid use among asthmatics, and (2) to test whether any of the 81 DMRs we previously identified as associated with asthma in PBMCs(10) are also associated with asthma in nasal epithelia.

Statistical Analyses of DNA Methylation Data

Data from the methylation array were normalized using the SWAN method(23) and the normalized M-values were used in all downstream analyses while beta-values, on the scale of 0–100%, are used for tables and figures. Differences between cases and controls are reported as percent methylation changes, using beta-values. We filtered out probes on the 450k Illumina array with known SNPs in European (CEU) and African (YRI) populations within the CpG motif.

We performed the analysis to identify differential methylation in three steps. In brief, these are: infer PEER factors to account for unobserved batch effects(24), fit linear models with limma(25), and use comb-p(26) with a window size of 300 base pairs (bp) to identify regions of sustained low p values or DMRs from the p-values reported by limma. DMPs were identified by calculating q values, corresponding to false discovery rate (FDR), from linear model p-values using the method of Benjamini and Hochberg(27). Following examination of q-q plots, we performed global adjustment for inflation of p values using standard methodology. Details of the analyses are outlined in the diagram in Supplemental Figure S2 and in the methods in the Online Supplement.

Statistical Analyses of Gene Expression Data

The R package limma(25) was used to background-correct, normalize (quantile) and fit linear models for expression data; p-values were based on the moderated t-statistic and q-values were calculated from p-values using the method of Benjamini and Hochberg(27).

Analysis of DNA Methylation and Gene Expression

To understand the relationship of DMRs with gene expression changes, we considered inversely correlated (canonical) vs. positively correlated pairs limiting the analysis to genes within 5kb of a DMR. We calculated the enrichment of inversely correlated pairs in relation to all pairs using the binomial test. To integrate the expression and methylation data, we used a method derived from the comb-p(26) except instead of testing each methylation site for its relation to asthma status, we tested methylation regions and expression probes within 5kb.

Additional Methods

Additional methods are available in the Online Supplement.

RESULTS

There were no significant demographic differences between allergic asthma cases (N=36) and controls (N=36), although collection site approached statistical significance (Table 1). Per study design, all allergic asthma subjects were atopic based on positive skin prick test to at least one indoor allergen (data not shown) and total IgE concentrations, and had airflow limitation as indicated by spirometric measurements (Table 1).

Table 1.

Demographic and clinical characteristics of asthma subjects and controls

	Allergic Asthma Subjects (N=36)	Control Subjects (N=36)	p- value †	Replication Allergic Asthma Population (n=30)
Site: Boston	8 (22.2%)	15 (41.7%)	0.071	--
Dallas	4 (11.1%)	1 (2.8%)		6 (20.0%)
Denver	7 (19.4%)	3 (8.3%)		--
Detroit	4 (11.1%)	1 (2.8%)		--
New York	6 (16.7%)	3 (8.3%)		3 (10.0%)
Washington D.C.	7 (19.4%)	13 (36.1%)		--
Chicago	--	--		7 (23.3%)
Cincinnati	--	--		14 (46.7%)
Age at recruitment (yr)	11.0 [10.0, 12.0]	11.0 [10.0, 12.0]	0.522	11.0 [10.0, 12.0]
Gender (Male)	19 (52.8%)	17 (47.2%)	0.641	19 (63.3%)
Participant race: African American (Yes)	33 (91.7%)	33 (91.7%)	>0.991	30 (100.0%)
Participant race: Hispanic or Latino (Yes)	7 (19.4%)	3 (8.3%)	0.171	0 (0.0%)
How often exposed to smokers: Daily	10 (27.8%)	8 (22.2%)	0.591	8 (26.7%)
Dog living in the home in the last 6 months (Yes)	11 (30.6%)	13 (36.1%)	0.621	9 (30.0%)
Cat living in the home in the last 6 months (Yes)	9 (25.0%)	12 (33.3%)	0.441	3 (10.0%)
Any water/dampness in the last 12 months (Yes)	9 (25.0%)	4 (11.1%)	0.131	10 (33.3%)
Gas stove, gas range or gas oven (Yes)	22 (61.1%)	24 (66.7%)	0.621	12 (40.0%)
Allergic to any indoor allergen (Yes)	36 (100%)	0 (0%)	**	30 (100%)
Cockroach IgE (IU/mL)	0.34 [0.34, 1.53]	0.34 [0.34, 0.34]	**	0.44 [0.34, 25.2]
Total serum IgE (kU/l )	366.0 [185.0, 785.0]	29.0 [16.5, 49.5]	**	519.5 [181.0, 883.0]
Baseline - FEV1(% predicted) ‡	88.8 ± 17.5	105.4 ± 11.2	**	97.3 ± 19.3
Baseline - FEV1/FVC §	72.6 ± 10.1	84.6 ± 6.4	**	79.6 ± 8.1
Asthma Medications: Albuterol	34 (94.4%)	0 (0%)	**¶	30 (100.0%)
Inhaled Steroids	29 (80.6%)	0 (0%)		22 (73.3%)
Nasal Steroids	9 (25%)	0 (0%)		25 (83.3%)
Montelukast	15 (41.7%)	0 (0%)		11 (36.7%)

^*Data are presented as mean ± standard deviations, medians [interquartile range] or numbers (%).

^†P-values were calculated with the use of the Chi-Square¹ test for categorical variables and the Mann-Whitney² test for continuous variables.

^‡FEV₁ denotes forced expiratory volume in one second

^§FVC denotes forced vital capacity

^¶Medication data missing on 14 controls

^**By protocol design

After adjusting for age, gender, race/ethnicity, unknown batch effects, removing probes with known European and African SNPs at the CpG motif, and performing global adjustment for inflation, we identified 119 genome-wide significant DMRs associated with 118 unique genes (Figure 1A; Table 2 and Supplemental Table S4) and 118 single CpG motifs (DMPs) associated with 107 unique genes (Figure 1B; Table 2 and Supplemental Table S5). The median percent methylation change between allergic asthmatics and controls was 6.8% (range 2.6–28.8%) for DMRs and 13.6% (range 5.2–29.5%) for DMPs. 39 genes were identified in both the DMR and DMP analysis but a number of genes only have DMPs or DMRs as these two analytical approaches are complementary. Some genes/loci that have a highly significant CpG (DMP) will not be represented in the DMR analysis because there are no neighboring CpGs with low p values as Illumina array does not interrogate all CpGs and sequence data would be needed to examine this further. On the other hand, DMR analysis identifies regions of sustained low p values in multiple CpG sites where coverage on the Illumina array is available but not all of these loci have one highly significant CpG (DMP). Representative histograms of percent methylation estimates (beta values) in allergic asthmatic and controls are shown in Supplemental Figure S3. Among the 186 (DMR and/or DMP) allergic asthma-associated differentially methylated genes in the nasal epithelia are genes with established roles in asthma and atopy - arachidonate 15-lipoxygenase (ALOX15), calpain 14 (CAPN14), histamine N-methyltransferase (HNMT), and periostin (POSTN) (Figure 1C). Moreover, components of the extracellular matrix (COL16A1, COL5A2, COL5A3, ELN, HAS3, MMP14), genes related to immunity (IFNGR2, HLKA-DPA1, LAG3, NFIL3, PRF1, TNFSF13), cell adhesion (CTNND1, EPPK1, GJA4), epigenetic regulation (ATXN7L1, H1F0, HIST1H1D, METTL1), airway obstruction (GABRG3), obesity (C1QTNF1, GPC4), and autophagy (AMBRA1) are also differentially methylated in nasal epithelia of atopic asthmatics compared to controls (Figure 1C). None of the allergic asthma-associated DMRs/DMPs were significantly associated with nasal corticosteroid use among asthmatics after adjustment for multiple comparisons. Moreover, the small degree of overlap in allergic asthma-associated epigenetic changes in PBMC and nasal epithelia (3 of the 81 PBMC DMRs are significant in nasal epithelia at q<0.05; Supplemental Table S6) is not surprising given that PBMCs are a surrogate for an asthma-associated immune response, while nasal epithelia are directly exposed to the environment that may influence DNA marks and the course of asthma. Moreover, nasal epithelia are a more pure cell population than PBMCs and some of the methylation changes present in specific mononuclear cell subsets such as Th2 T-cell subsets for example may be diluted by the presence of other cell types. The 3 DMRs that are present in both cell types may be associated with genetic variants (methylation quantitative trait loci or mQTLs)(28). We also performed an enrichment analysis of methylation marks within areas of open chromatin, as assessed by DNAse I hypersensitivity site (DHS) analysis, in cell lines profiled by ENCODE(29). This analysis identified strongest enrichment of allergic asthma-associated DNA methylation changes in DHSs in epithelial cells (q<6.57x10⁻¹¹), suggesting that our analysis identified DNA methylation changes in open areas of chromatin in the cell type of interest (Supplemental Figure S4). To begin to explore the influence of environmental exposures on the nasal epithelial epigenome, we identified 48 DMRs that are significantly associated with environmental tobacco smoke (ETS) after adjusting for age, gender, race/ethnicity, unknown batch effects, and allergic asthma. 48 DMRs are in 46 unique genes (Supplemental Table S7), one of which (SFRP2) overlaps with the analysis of direct cigarette smoke exposure on methylation of small airway epithelia that identified three genes after adjustment for genome-wide comparisons(30).

Figure 1.

Differentially methylated regions (DMRs) (A) and differentially methylated single-CpG probes (DMPs) (B) in nasal epithelia are associated with asthma after controlling for age, gender, race/ethnicity, technical variables, and batch effects. (A) Manhattan plot of the adjusted p-values for disease status (asthma/control) from the linear model. Each dot represents a p value for a probe on the Illumina 450k array that has been adjusted by the significance of neighboring probes within 300 bases according to their correlation. Probes within statistically significant DMRs after adjustment for genome-wide comparisons are identified by darker larger symbols. (B) Manhattan plot of the false discovery rate (FDR) adjusted p-values (q-values) for disease status (asthma/control) from the linear model. Probes with q<0.05 are highlighted by darker larger symbols. (C) Venn diagram depicting gene-based overlap of DMRs and DMPs. Representative genes of clinical/biological relevance to asthma that are discussed in the text are highlighted

Table 2.

10 differentially methylated regions (DMRs) and 10 differentially methylated probes (DMPs) with the most pronounced DNA methylation changes in allergic asthmatics compared to controls.

Type	Chr	Start	End	Adjusted p value	% Methylation Difference	Nearest Gene	Gene Distance	CpG Island Distance
DMR	chr11	36030085	36030086	0.002895	−28.8%	LDLRAD3	0	63631
DMR	chr7	1.06E+08	1.06E+08	1.32E-06	−27.3%	ATXN7L1	−4084	3152
DMR	chr12	58162286	58162287	2.12E-05	−27.0%	METTL1	63	2286
DMR	chr10	4386801	4386802	2.28E-05	−26.1%	LINC00703	−39635	−481323
DMR	chr15	1.02E+08	1.02E+08	9.70E-05	−25.1%	PCSK6	0	51207
DMR	chr22	19471092	19471093	4.13E-04	−24.4%	CDC45	0	3281
DMR	chr3	1.02E+08	1.02E+08	3.78E-04	−23.2%	LOC152225	177435	325159
DMR	chr15	39544143	39544144	0.01364	−22.9%	C15orf54	0	−328383
DMR	chr15	45449436	45449437	6.83E-07	−22.7%	DUOX1	0	4731
DMR	chr8	1.45E+08	1.45E+08	0.003634	−21.5%	EPPK1	−17562	3314
DMP	chr16	88558222	88558223	1.43E-05	−29.5%	ZFPM1	0	0
DMP	chr11	36030085	36030086	6.23E-06	−28.8%	LDLRAD3	0	63631
DMP	chr7	1.06E+08	1.06E+08	4.09E-09	−27.3%	ATXN7L1	−4084	3152
DMP	chr12	58162286	58162287	5.03E-08	−27.0%	METTL1	63	2286
DMP	chr10	4386801	4386802	5.03E-08	−26.1%	LINC00703	−39635	−481323
DMP	chr15	1.02E+08	1.02E+08	2.39E-07	−25.1%	PCSK6	0	51207
DMP	chr22	19471092	19471093	8.17E-07	−24.4%	CDC45	0	3281
DMP	chr6	2977292	2977293	4.99E-04	−24.1%	SERPINB6	−4893	5330
DMP	chr3	1.02E+08	1.02E+08	8.15E-07	−23.2%	LOC152225	177435	325159
DMP	chr15	39544143	39544144	2.62E-05	−22.9%	C15orf54	0	-328383

Gene expression analysis of nasal epithelia identified 53 differentially expressed genes (Figure 2A; Supplemental Table S8A). Among differentially expressed genes are genes associated with asthma and allergy - ALOX15, bradykinin B1 receptor (BDKRB1), cadherin 26 (CDH26), calpain 14 (CAPN14), cathepsins C and G (CTSC, CTSG), cystatin SN (CST1), gelsolin (GSN), mast cell carbohypeptidase CPA3 and tryptases (TPSAB1 and TPSD1), NADPH oxidase 1 (NOX1), neurotrophic tyrosine kinase receptor (NTRK2), plasminogen activator inhibitor 2 (SERPINB2), and von Willebrand factor (VWF). Moreover, a number of immune related genes with high relevance to asthma (CASP4, CCK, CCL5, CCL26, CD3G, CD55, CD6, CD74, CXCR6, HLA-DMB, HLA-DOA, KLRB1, TCRA) are also differentially expressed in atopic asthmatic compared to control samples.

Figure 2.

DNA methylation changes are associated with changes in gene expression in nasal epithelia. (A) Asthma-associated gene expression changes after adjusting for age, gender, and race/ethnicity. Blue dots indicate 27 genes with q<0.05 for the asthma vs control comparison with high functional relevance to asthma and atopy. The size of the dot is proportional to asthma vs control fold change. Expression changes in genes within 5 kilobases of the nearest DMR (B) and DMP (C). In both panels, x-axis methylation difference is represented by the mean % methylation difference in asthma subjects compared to controls; y-axis expression difference is represented by the mean fold change in asthma subjects compared to controls (on the log₂ scale). The blue symbols represent hypomethylated genes that were associated with increased gene expression as well as some hypermethylated genes associated with decreased gene expression. The red symbols represent methylation changes that were not associated with expected gene expression differences. Upward triangles indicate DMR/DMP location upstream of the gene, circles represent DMRs in the gene body, and downward triangles refer to DMR/DMPs downstream of the gene.

Given the substantial number of immune–related differentially expressed genes, we examined the function, general tissue/cell expression patterns(31), and immune cell specific expression patterns(32) of all 53 differentially expressed genes to discern which genes were likely contributed by immune cells in the respiratory epithelium (Supplemental Table S8B). This analysis revealed 11 of the 53 of the genes as highly likely to be contributed by immune cells – B-cells (HLA-DMB and HLA-DOA), mast cells (CPA3, CTSG, TPSAB1, and TPSD1), NK cells (KLRB1), T cells (CD3G, CD6, TCRA) and multiple immune cells (CST1). This is consistent with published findings in the lung epithelium; for example, differential expression of mast cell genes (CPA3, TPSAB1, and TPSD1) in the airway epithelium of asthmatics with different immune subtypes of asthma (Th2 high and Th2 low) is due to accumulation of intraepithelial mast cells with a unique protease phenotype in Th2-high asthma(33). The remaining 42 genes are specifically expressed by epithelial cells, contributed both by epithelial and immune cells, or have unknown expression patterns.

The relationship between DNA methylation and gene expression revealed enrichment of hypomethylated genes associated with increased gene expression, and hypermethylated genes associated with decreased gene expression (binomial P<2.8x10⁻⁶ for enrichment in inverse relationships in asthma-related DMRs [Figure 2B] and P<7.8x10⁻¹⁰ for allergic asthma-associated DMPs [Figure 2C]) within 5 kilobase (kb) distance of the transcription start site (TSS). However, a number of DNA methylation changes were not associated with canonical gene expression differences (Figure 2B and 2C). While these results are similar when we use 2kb or 3kb distance, beyond 5kb of the TSS we begin to lose the enrichment in anti-correlated methylation-expression pairs. We did not observe an enrichment in negatively correlated methylation-expression relationships in gene promoters compared to gene bodies, as has been previously reported(34); however, this analysis was underpowered. We implemented QTL mapping to identify methylation marks that control gene expression (methyl-eQTR or quantitative trait region).

We performed two analyses – one centered on differentially methylated genes (union of DMRs and DMPs from Figure 1; 186 unique genes) and the other focused on 53 differentially expressed genes (Figure 2A). Of the 186 allergic asthma-associated differentially methylated genes, 158 have methylation and expression data within 5kb distance and, of these, 16 (10%) have significant relationships of DNA methylation and gene expression within 5kb (Supplemental Table S9A). Of the 53 allergic asthma-associated differentially expressed genes, 32 (60%) have significant relationships of DNA methylation and gene expression within 5kb (Supplemental Table S9B). We observed a strong enrichment in significant methylation-expression pairs with 5kb of differentially expressed genes compared to all methylation-expression pairs within 5kb distance of each other in the genome (enrichment p=3.2x10⁻⁶). Both analyses identified predominantly canonical inverse correlations of methylation and expression (15/ o f the 16 genes in the methylation-centered analysis and 32 of the 53 in the expression-centered analysis). Importantly, both analyses demonstrated that ALOX15 has significant relationship of DNA methylation and gene expression. Upstream regulator analysis of the 47 allergic asthma-associated genes that demonstrate canonical inverse relationships between methylation and expression, using Ingenuity Pathway Analysis (IPA), revealed a significant enrichment (p<1x10⁻⁴) in cytokines (IL-13, IL-4, IL-6, IFN-γ , and others) as well as transcription factors (CIITA) and growth factors (TGF-β) known to regulate gene expression profiles in asthma (Supplemental Table S10). Protein-protein interaction (PPI) analysis of the 47 allergic asthma-associated genes with inverse relationships of methylation and expression revealed a network of proteins with the largest hub being RIPK2 (Figure 3). RIPK2 or CARD3 is a component of signaling complexes in both the innate and adaptive immune pathways, is critical for NOD-mediated NF-κB activation and cytokine production(35), and silencing of its expression attenuates allergic airway inflammation in mice(36). Smaller hubs in the network include other proteins important in immunity and identified in our analysis. Taken together these analyses support biological and disease relevance of our findings.

Figure 3.

Protein-protein interactome analysis of 47 allergic asthma-associated genes with inverse relationship of methylation and expression. The interactome was created using NetworkAnalyst(70) and the InnateDB PPI dataset. The nodes are colored based on their methylation and expression (green are downregulated and hypermethylated while red are upregulated and hypomethylated). The sizes of nodes are proportional to their betweenness centrality values.

To determine the validity of these findings, we selected 10 loci for pyrosequencing and included an independent population of children with allergic asthma from the inner city (Table 1). 10 loci were chosen to represent genes with known relevance to asthma, loci with largest methylation changes, and methylation marks that affect gene expression. Four allergic asthma-associated DMRs or DMPs were selected based on the relevance to asthma (ALOX15, HLA-DPA1, GJA4, and POSTN), three were selected based on the extent of methylation differences between asthmatics and controls (LDLRAD3, ATXN7L1, and METTL1), and three additional loci were selected based on the allergic asthma specific methylation-expression relationships (CCL5, CTSC, and CXCR6). Nine of these 10 loci validated internally in the original study population (one-tailed t-test P<0.05), and the CTSC locus approached statistical significance (P=0.12) (Table 3 and Supplemental Table S11) if using nominal p values and 8/10 if stringent Bonferroni correction is applied. More importantly, all 10 loci validated in an independent population of children with allergic asthma residing in the inner city compared to controls from the original population (shared control design; one-tailed t-test P<0.05; Table 2) and methylation levels in two sets of allergic asthmatic cases are comparable (Supplemental Table S11). Significance levels were very similar if adjustment for age, gender, and race/ethnicity was included, demonstrating the robustness of these associations with allergic asthma. We also examined expression of these ten genes by qPCR and showed that half of them have significant changes in allergic asthma compared to controls and are also significantly inversely correlated with methylation in the original study population as well as independent allergic asthmatics (Table 3).

Table 3.

Pyrosequencing and qPCR validation of selected DNA methylation marks in the same asthma cases and controls and an independent population of asthmatics. Green denotes hypomethylation and decreased expression while red indicated hypermethylation and increased expression in asthma cases compared to controls.

(A) Selected DMRs/DMPs from the Illumina analysis
DMR/DMP Coordinates	Gene	Illumina Adjusted p Value	Illumina Methylation Change
chr17:4541333-4541334	ALOX15	7.82x10−4	−10.1
chr6:33041220- 33041697	HLA- DPA1	1.04x10−6	9.8
chr1:35258778- 35258933	GJA4	6.00x10−3	−9.8
chr13:38172802-38172803	POSTN	7.25 x10−4	−11.6
chr11:36030085-36030086	LDLRAD3	6.23x10−6	−28.8
chr7:105521115-105521116	ATXN7L1	4.09x10−9	−27.3
chr12:58162286-58162287	METTL1	5.03x10−8	−27
chr3:45984742-45985168	CXCR6	7.30x10−4	5.9
chr17:34202460-34202461	CCL5	0.017	1.9
chr11:88059526-88059527	CTSC	1.51x10−3	−1.1

(B) Internal validation
Coordinates of Measured Cs	Gene	Methylation Change	One-tailed t-test	Expression Fold Change	One-tailed t-test	Methylation-Expression Correlation	Correlation p value
chr17:4541333	ALOX15	−8.7	1.55x10−7	2.0	0.0258	−0.39	5.53x10−4
chr6:33041528	HLA- DPA1	9	2.10x10−4	0.6	0.0361	−0.46	2.90x10−5
chr1:35258850, 35258855	GJA4	−3.4, −7.8	6.0x10−5, 1.74x10−3	1.7	0.0913	−0.0061, −0.028	0.9595; 0.810
chr13:38172802	POSTN	−8.9	1.70x10−4	5.8	1.84x10−4	−0.35	1.71x10−3
chr11:36030086	LDLRAD3	−20.9	2.18x10−7	1.2	0.263	−0.093	0.424
chr7:105521115	ATXN7L1	−24.8	2.60x10−14	0.9	0.375	0.048	0.682
chr12:58162286	METTL1	−22.8	8.25x10−11	0.8	0.214	0.054	0.643
chr3:45984839, 45984841	CXCR6	5.7, 6.0	9.89x10−4, 1.48x10−3	0.5	0.0197	−0.32; −0.27	5.42x10−3; 0.0193
chr17:34202460	CCL5	4.4	0.048	0.8	0.152	−0.29	9.94x10−3
chr11:88059526	CTSC	−4.7	0.12	1.6	9.30x10−5	−0.36	1.40x10−3

(C) External validation
Coordinates of Measured Cs	Gene	Methylation Change	One-tailed t-test	Expression Fold Change	One-tailed t-test	Methylation-Expression Correlation	Correlation p value
chr17:4541333	ALOX15	−10.6	7.50x10−10	1.5	0.0697	−0.092	0.449
chr6:33041528	HLA- DPA1	8.1	1.27x10−3	0.5	0.00109	−0.47	4.20x10−5
chr1:35258850, 35258855	GJA4	−4.7, −13.7	2.03x10−8, 2.26x10−8	7.1	1.27x10−5	−0.16; −0.13	0.172; 0.303
chr13:38172802	POSTN	−13.6	6.5x10−9	7.3	9.73x10−5	−0.57	<1x10−5
chr11:36030086	LDLRAD3	−23.8	8.80x10−9	1.5	0.0282	−0.13	0.298
chr7:105521115	ATXN7L1	−31.1	4.91x10−20	0.9	0.346	0.11	0.347
chr12:58162286	METTL1	−27.1	6.02x10−14	1.3	0.149	−0.076	0.537
chr3:45984839, 45984841	CXCR6	4.2, 4.7	0.0227, 0.0188	0.7	0.0632	−0.30; −0.31	0.0125; 9.49x10−3
chr17:34202460	CCL5	7.8	1.6x10−3	1.0	0.419	−0.25	0.0393
chr11:88059526	CTSC	−14.5	8.2x10−5	1.5	0.00232	−0.46	7.30x10−5

To further examine validity and generalizability of our findings, we profiled nasal epithelia from adult Caucasian subjects, 12 atopic asthmatic and 12 controls (cohort characteristics in Supplemental Table S12), on Illumina methylation arrays. Using the same methods applied to our primary cohort, we identified 1766 DMPs and 370 DMRs, representing 1108 unique genes, significantly associated with asthma (genome-wide adjusted p<0.05) after adjustment for age, gender and batch effects. Of the 186 genes with significant methylation changes in the primary cohort of atopic asthmatic African American children from the inner city, 67 of them (36%) have at least one genome-wide significant DMP and/or DMR in adult Caucasian asthmatics (Supplemental Table S13). Similarly, 56 genes (30%) also overlap with genes that have IL-13 responsive CpGs in cultured nasal epithelial cells from 57 unrelated adult lung donors(37). These results demonstrate external validity and generalizability of some of our findings to asthma regardless of age, race/ethnicity, specific exposures, and conditions (cultured vs fresh cells, nasal vs lung epithelia).

DISCUSSION

Our findings demonstrate that methylation marks in the nasal epithelia of children with allergic asthma are associated with changes in gene expression. Moreover, the magnitude of these allergic asthma-associated epigenetic signals is substantial when compared with other diseases(38, 39). These findings suggest that genes that have been previously reported to be differentially expressed in the airway epithelium of allergic asthmatics(14–16) may be regulated by epigenetic mechanisms.

The effect of aeroallergens(12) and other environmental stimuli(13) on the respiratory epithelium is a key mediator of asthma. Gene expression profiles of the allergic asthmatic airway(14–16) and nasal(17) epithelia demonstrate that these cells are involved in an active biological process. Our findings demonstrate a clear relationship between DNA methylation and gene expression in the nasal epithelia of young allergic asthmatics using a stringent analytical pipeline to ensure that methylation changes we identified are not false positives due to influence of demographic factors or technical variables. More than half of the differentially expressed genes in our analysis have significant associations with methylation marks; this includes asthma and allergy genes ALOX15(40), CAPN14(41), CTSC(42), CST1(43), GSN(44), NTRK2(45), and VWF(46) as well as a number of immune and extracellular matrix genes. Moreover, genes have previously identified as differentially expressed in the respiratory epithelia but did not reach statistical significance in our expression analysis, such as periostin(15, 17), are also differentially methylated. Finally, genes with the largest absolute changes in methylation observed in allergic asthma (LDLRAD3, ATXN7L1, and METTL1) have not previously been implicated in allergic asthma and represent biological candidates for future investigation. There is recent evidence that expression of ALOX15 is regulated by demethylation of H3K27me3 at the ALOX15 promoter following IL-4 treatment in epithelial cells, providing support for the specific genes identified in our analysis(47).

Importantly, the magnitude of DNA methylation changes at some of the loci is large, consistent with that observed in other diseased tissue(38, 39) and these loci may become important therapeutic targets in the future, provided additional evidence from cohorts with larger sample sizes. DNA methylation changes have been shown to drive tumor formation and malignant progression(48), appear to critical to disease pathogenesis, and represent novel therapeutic targets in cancer. DNA methyltransferase (DNMT) inhibitors, such as 5-azacitidine and decitabine, have been approved for the treatment of myelodysplastic syndrome(49, 50), and are being tested in solid tumors(51, 52). Methylation changes in nasal epithelia of allergic asthmatics are on average ~10 times larger than DNA methylation changes in peripheral blood that are associated with childhood allergic asthma(10) or other diseases of the airway such as COPD(18). The strength of the DNA methylation signal in the nasal epithelia of children with asthma was detectable even in the limited sample size of the study population. Moreover, many of the methylation marks validated in an even smaller cohort of adult Caucasian asthmatics, demonstrating generalizability of our findings to asthma. Given similarities in the nasal and airway/bronchial epithelial transcriptomes(17), it is logical to speculate that DNA methylation in nasal epithelia could be used as a biomarker of exposure or disease; however, further work in larger cohorts will be needed to explore this biomarker potential..

The majority of allergic asthma-associated methylation marks we identified in the nasal epithelia are hypomethylated and within gene bodies. Our results are in concordance with previous reports that have shown hypomethylation of specific genes in asthma(10, 53–55) and other airway diseases such as COPD(18). Similarly, cigarette smoke exposure is more often associated with hypomethylated regions in airway epithelia(30) and large blocks of hypomethylation have been identified in cancer(56, 57). While early studies focused specifically on promoter methylation as a mechanism of gene regulation(51), more recent research has demonstrated the significance of gene body methylation(58, 59) in transcriptional regulation in malignant(34) and non-malignant diseases(22). Our study is the first to show an association of gene body methylation and gene expression in allergic asthma.

There are several potential causes for these methylation changes. First, nasal epithelia are subject to a number of environmental exposures that are relevant to allergic asthma and are also known to shape the epigenome; these include cigarette smoke(30, 60), air pollution(61), and farming aerosols(62). While we were unable to evaluate the impact of exposures on the epigenome due to lack of exposure assessment in our study, this is an important future direction for the field. Second, epigenetic marks may influence the severity of allergic asthma. Previous work has shown association of DNA methylation in specific gene loci (IL-6 and iNOS) in the nasal epithelia(53) and buccal cells (IFN-3)(63). Finally, microbial species are also known to influence the epigenome. Differential effect of rhinovirus infection in asthmatics (N=6) compared to non-asthmatic (N=3) nasal epithelia on methylation of a small noncoding RNA locus (SNORA12) has been recently demonstrated(64).

There are several limitations to our study. First, our study design precludes us from differentiating epigenetic marks associated with allergy versus those associated with asthma. However, since most children with asthma also are atopic, our findings among children allergic asthma are generalizable to a larger population of children with asthma. Secondly, we were able to evaluate the impact of exposures on the epigenome in limited fashion using questionnaires and not exposure assessments. We did not collect longitudinal samples and were unable to examine longitudinal variation in DNA methylation due to changes in exposure. The current study is an observational study with small sample size showing an association between methylation biomarkers and childhood persistent asthma, and further work will be required to address the stability of the biomarker over time. The relationship between the environment and epigenome, especially in relation to conditions like allergic asthma, remains a particularly important area of future research. Thirdly, our small sample size precluded us from assessing methylation marks in relation to disease heterogeneity and asthma endotypes and future work in larger cohorts will be required to address this question. Fourthly, while all our subjects are self-reported African American or of Dominican/Haitian descent, there is substantial genetic admixture of African ancestry in these populations(65, 66) and we did not examine the effect of genetics on DNA methylation(67, 68). The fourth limitation is the inability to detect differences in methyl- vs. hydroxyl-methyl cytosine, which is thought to be a mark for de-methylation(69). Finally, our association analysis is unable to distinguish between DNA methylation marks being the cause or consequence of disease.

Despite these limitations, our study identified substantial and consistent DNA methylation changes in the nasal epithelia of children with allergic asthma residing in the inner city compared to non-diseased controls, and these methylation marks were internally and externally validated in both children and adults. These methylation changes are associated with gene expressionand need to be further evaluated in larger cohorts for their potential as biomarkers and therapeutic targets . The peculiar spatial and biological relation of epigenetic changes in the nasal epithelia of asthmatics suggests that these cells and these mechanisms may provide insight into the etiology and pathogenesis of this persistent public health concern(1).

KEY MESSAGES.

• Expression of genes related to extracellular matrix, immunity, cell adhesion, epigenetic regulation, and airflow obstruction are epigenetically regulated in nasal epithelia and associated with asthma.

• The methylation changes in nasal epithelia are substantial (median 9.5%, range: 2.6–29.5%) and similar in magnitude to those observed in other diseases.

• More than one third of methylation changes identified by our analysis appear to be generalizable to atopic asthma regardless of age, regardless of age, race/ethnicity, and specific exposures.

ABBREVIATIONS

DMP

Differentially methylated probe

DMR

Differentially methylated region

FDR

false discovery rate

ICAC

Inner City Asthma Consortium

PBMC

peripheral blood mononuclear cells

PEER

Probabilistic estimation of expression residuals

QTL

Quantitative trait locus

CAPSULE SUMMARY

DNA methylation changes substantial in magnitude are associated with atopic asthma and gene expression in African American inner city children.