A decade of research on the 17q12-21 asthma locus: Piecing together the puzzle

Abstract

Chromosome 17q12-21 remains the most highly replicated and significant asthma locus. Genotypes in the core region defined by the first genome-wide association study correlate with expression of 2 genes, ORM1-like 3 (ORMDL3) and gasdermin B (GSDMB), making these prime candidate asthma genes, although recent studies have implicated gasdermin A (GSDMA) distal to and post-GPI attachment to proteins 3 (PGAP3) proximal to the core region as independent loci. We review 10 years of studies on the 17q12-21 locus and suggest that genotype-specific risks for asthma at the proximal and distal loci are not specific to early-onset asthma and mediated by PGAP3, ORMDL3, and/or GSDMA expression. We propose that the weak and inconsistent associations of 17q single nucleotide polymorphisms with asthma in African Americans is due to the high frequency of some 17q alleles, the breakdown of linkage disequilibrium on African-derived chromosomes, and possibly different early-life asthma endotypes in these children. Finally, the inconsistent association between asthma and gene expression levels in blood or lung cells from older children and adults suggests that genotype effects may mediate asthma risk or protection during critical developmental windows and/or in response to relevant exposures in early life. Thus studies of young children and ethnically diverse populations are required to fully understand the relationship between genotype and asthma phenotype and the gene regulatory architecture at this locus.

Keywords: GSDMA; GSDMB; ORMDL3; PGAP3; Wheezing; gene expression; genome-wide association study; immune cells; lung cells.

FIG 1.

The 17q12–21 asthma locus. A, Regional association plot of 17q SNPs with asthma in TAGC (children plus adults). The lead SNP, rs2952156 (purple diamond), is in ERBB2 in the proximal 17q region. All other SNPs are colored based on their LD with the lead SNP (see inset). Modified with permission from Demenais et al. B, Location of genes at the 17q locus. The 6 genes highlighted in this review are shown in blue, orange, and green to correspond to the proximal, core, and distal regions, respectively. C, Frequencies of asthma-associated alleles at 17 SNPs in 1000 Genomes reference panels: African Americans (American of African Ancestry in SW USA [ASW] and African Caribbeans in Barbados [ACB]), Europeans (CEU, British in England and Scotland [GBR], and Toscani in Italy [TSI]), Latinos (Mexican Ancestry from Los Angeles, USA [MXL] and Puerto Ricans from Puerto Rico [PUR]), East Asians (Han Chinese in Beijing [CHB], Japanese in Tokyo [JPT], Southern Han Chinese [CHS], Chinese Dai in Xishuangbanna [CDX], and Kinh in HoChi Minh City [KHV]). D, LD (r²) among the 17 SNPs described in Fig 2 in African Americans (upper panel), Europeans (lower left panel), Latinos (lower middle panel), and East Asians (lower right panel) determined by using the same 1000 Genomes reference panels as in Fig 1, C. Asthma-associated alleles are shown for each SNP. *Associated allele is ancestral. Data for Fig 1, C and D, are from ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/.

FIG 2.

Seventeen SNPs in the extended 17q12–21 region that are reported to be associated with asthma, gene expression, or epigenetic modification. Base pair position from build hg19 is shown. SNP type is shown as ancestral > derived alleles. TAGC, Transnational Genetics of Asthma Consortium.

FIG 3.

SNPs in the 17q core region are associated with early-life onset and risk of or protection from wheezing illness. A, Ordered-subset regression analysis of asthma was performed for 11 SNPs associated with asthma in the Epidemiological Study on Genetics and Environment of Asthma sample (EGEA; n = 1511). The maximal difference in likelihood ratio test statistics for association between the ordered age-specific subsets and the total sample was for an onset of asthma of 4 years or younger. SNP rs9303277 showed the largest difference in the likelihood ratio test statistics. The EGEA included children from 388 nuclear families with at least 1 asthmatic member living in one of 5 French cities. Modified with permission from Bouzigon et al. B, Risk for asthma based on the rs7216389 genotype is confined to children who wheezed in the first 3 years of life (Childhood Origins of Asthma [COAST], n = 200; COPSAC, n 5 297) or in the first year of life (Protection against Allergy Study in Rural Environments [PASTURE] birth cohort, n = 696). Dashed horizontal lines show the overall prevalence of asthma at age 6 years in each sample. COAST, COPSAC, and PASTURE are birth cohort studies in Madison, Wisconsin; Copenhagen, Denmark; and rural areas of Europe, respectively. The children in COAST had at least 1 parent with asthma or allergies, the children in COPSAC had a mother with asthma, and the children in PASTURE were unselected with respect to asthma or allergy. Modified from Bønnelykke et al. C, In the PASTURE study children with the asthma-associated allele rs8076131-A and exposed to animal barns have significantly fewer wheezing illnesses in the first year of life. Used with permission from Loss et al.

FIG 4.

Relative expression of 17q genes in different tissues and cells. A, PBLs (n = 112). B, Lung CD4⁺ TRMs (n = 18). C, BECs (n = 85). D, ASMCs (n = 67). Expression levels of ORMDL3 and PGAP3 were most correlated in PBLs (ORMDL3 vs PGAP3 r = 0.410, P = 7.3 × 10⁻⁶ cf. ORMDL3 vs GSDMB r = 0.369, P = 6.2 3 10⁻⁵), BECs (ORMDL3 vs PGAP3 r = 0.693, P = 1.9 × 10⁻¹³ cf. ORMDL3 vs GSDMB r = 0.320, P = 2.8 × 10⁻³), and ASMCs (ORMDL3 vs PGAP3 r = 0.450, P = 1.3 × 10⁻⁴ cf. ORMDL3 vs GSDMB r = 0.221, P = .07). ORMDL3 expression was not correlated with expression of any 17q genes in lung CD4⁺ TRMs (ORMDL3 vs PGAP3 r = 0.077, P = .76 cf. ORMDL3 vs GSDMB r = 0.400, P = .10). Gene expression levels in Fig 4, A-C, are from RNA-seq; counts are shown as relative expression normalized to all genes detected as expressed within each cell type. Gene expression in Fig 4, D, is based on microarrays; expression levels were normalized to have a mean of zero and scaled between −5 and 10. Methods for PBLs, CD4⁺ TRMs, and ASMCs are described in the Methods section in this article’s Online Repository; methods for BECs are reported in Nicodemus-Johnson et al.

FIG 5.

Expression of ORMDL3 and PGAP3 transcripts in BECs from 85 adults. Correlation between ORMDL3 (x-axis) and PGAP3 (y-axis) in 85 subjects (r = 0.693, P = 1.9 × 10⁻¹³, as Fig 4). Expression levels are shown as read counts adjusted for unwanted variation, as described in Nicodemus-Johnson et al.

FIG 6.

GTEx project eQTLs for 17q genes in whole blood and lung tissue at a false discovery rate of 5%. 17q SNPs were eQTLs for only 2 of the 6 genes in whole blood and 4 of the 6 genes in lung tissue; the remaining genes (without significant eQTLs) are not shown. SNPs in the proximal region are shown on a blue background, SNPs in the core region are shown on an orange background, and SNPs in the distal region are shown on a green background (see Fig 1). The GTEx project sample composition is approximately 84% white, 14% African American, 1% other, and 1% unknown (https://gtexportal.org/home/ tissueSummaryPage). n.s., Not significant.

FIG 7.

eQTLs for ORMDL3 and GSDMB in Nigerian (YRI) LCLs and associations with asthma in African Americans. LD between each SNP and the lead eQTL (rs12936231) are shown in the third column. eQTLs and GWAS P values of .01 or less are shown in boldface. SNPs in the proximal region are shown on a blue background, SNPs in the core region are shown on an orange background, and SNPs in the distal region are shown on a green background (see Fig 1). n.i., No information available (SNPs not imputed).

FIG 8.

Allele-specific chromatin modification and gene expression at the 17q locus. A, Schematic representation of the CTCF-binding motifs that overlap with the asthma-associated SNPs rs12936231 in ZPBP2 and rs4065275 in ORMDL3 (modified from Schmiedel et al). Relative expression of ORMDL3 from the C-A haplotype was inferred from 1 YRI subject who was heterozygous C-A/G-A (ORMDL3 fragments per kb million [FPKM] = 26; see Fig 8, B). Predicted haplotype frequencies are based on allele frequencies and LD estimates (see Fig 1). B, ORMDL3 and GSDMB gene expression in LCLs and phased genotype data for 29 YRI subjects who are homozygous for 3 of the rs12936231-rs4065274 haplotypes; no subjects were homozygous for the C-A haplotype. The y-axis shows FPKM, a measure of gene expression after normalizing for sequence depth and gene length.

FIG E1.

LD (r²) among the 17 SNPs described in Fig 2 in Nigerians (YRI) by using the 1000 Genomes reference panel.