Analyses of associations between three positionally cloned asthma candidate genes and asthma or asthma-related phenotypes in a Chinese population

Six genes, plant homeodomain finger protein 11 (PHF11), dipeptidyl-peptidase 10 (DPP10), histocompatibility antigen, class I, G (HLA-G), ADAM metallopeptidase domain 33 (ADAM33), neuropeptide S receptor 1 (NPSR1, previously GPR154 or GPRA), and cytoplasmic FMR1 interacting protein 2 (CYFIP2), have been identified to be associated with asthma and asthma-related phenotypes following linkage studies [1–6]. These genes span a wide range of functions, which were either unknown or would not have been considered to be implicated in the etiology of asthma before their discovery.

PHF11 was positionally cloned by Zhang et al. [1] and has been found to be associated with IgE levels, severe clinical asthma and childhood atopic dermatitis in Caucasian populations [1, 7]. A more recent study demonstrated that knockdown of PHF11 expression reduced expression of the T(H)1 - type cytokines IFN-gamma and IL-2 [8]. Allen et al. identified DPP10 and showed its associations with asthma in Caucasian populations [2]. Other studies have demonstrated that DPP10 modulated functional properties of Kv4-mediated A-type K+ channels in the central neurons via its intracellular and transmembrane domains [9–11]. Nicolae et al. positionally cloned HLA-G and showed that HLA-G was associated with asthma, bronchial hyperresponsiveness (BHR) and atopy in Caucasian populations [3]. HLA-G was expressed in bronchial epithelial cells [3] and may be involved in airway remodeling by inhibiting angiogenesis [12]. ADAM33 was identified by Van Eerdewegh et al. [4], and following functional studies have suggested that it might play roles in airway remodeling by promoting angiogenesis [13]. Laitinen et al. identified NPSR1 as an asthma candidate gene and reported that levels of the NPSR1-B isoform were increased in airway smooth muscle cells and epithelial cells in asthma patients compared to healthy controls, and NPSR1 was up-regulated in a mouse model of ovalbumin-induced inflammation [5]. CYFIP2 was positionally cloned by Noguchi et al. [6] and has been implicated in Rac-1-mediated T cell adhesion [14].

Recently, in studies of childhood asthma, Hersh et al. attempted to replicate associations with five positionally cloned asthma susceptibility genes, including ADAM33, DPP10, NPSR1, HLA-G, and PHF11, using both Caucasian and Hispanic families [15]. However, the replication with asthma was only found in NPSR1 and PHF11 in both cohorts, yet for NPSR1, the opposite alleles of SNPs were associated in either cohort, and for PHF11, there was no overlap in the associated SNPs across the two cohorts; no association was found in ADAM33, DPP10 or HLA-G in either cohort. A more recent study was conducted to replicate associations of 93 previously reported asthma candidate genes, including the six positionally cloned genes, with asthma and asthma-related phenotypes using 5,565 individuals (predominantly Caucasian); however, no significant associations were found for these six genes [16]. Actually, the associations of ADAM33, NPSR1, and CYFIP2 with asthma or asthma-related phenotypes have been studied in Chinese or Japanese populations [6, 17, 18]. However, to our knowledge, there is no further confirmation study conducted for DPP10, HLA-G, and PHF11, and the reported associations in non-Caucasian populations have yet to be examined. In this article, we examined the associations of polymorphisms in the identified top regions of PHF11, DPP10, and HLA-G with asthma and asthma-related phenotypes in a Chinese population, using a population-based study design.

DPP10 Among the five studied SNPs, rs10192393 produced the strongest association signal (Table 3 and additional file 1 - Supplemental Table E2). Compared with the rs10192393 T/T homozygotes, individuals with C/C or C/T genotype had a 1.90 times higher risk of BHR (nominal P = 0.0059) and a 2.03 times higher risk of BHR asthma (nominal P = 0.0067) (Table 3). Among the five SNPs, rs1430090, rs6737251 and rs7580359 were in a haplotype block defined by the solid LD spine (Figure 1) [23]. We further explored the associations with haplotypes constructed by the two SNPs outside the block and by the three SNPs inside the block, respectively. As shown in Table 4, compared to subjects with the TC haplotype constructed by rs10192393 and rs1430092, ones with the CC haplotype had a 1.99 times higher risk of BHR (nominal P = 0.0040) and a 2.15 times higher risk of BHR asthma (nominal P = 0.0041). No significant association was found for haplotypes inside the block. Compared to the study-wide significance level of 0.0041, only the haplotype associations remained significant.

HLA-G In single-SNP analyses, compared with the rs1736935 C/C homozygotes, individuals with T/T genotype had a lower risk of BHR asthma (OR = 0.64, nominal P = 0.044) (Table 3). Under the additive genetic model, we estimated that each additional copy of the T allele reduced the risk of BHR asthma by 19% (OR = 0.81; nominal P = 0.043). In haplotype analyses, we found that compared to the AT haplotype constructed by rs1632949 and rs1736935, each additional copy of the AC haplotype had a 1.40 times higher risk of BHR asthma (nominal P = 0.0086) (Table 4). However, compared to the study-wide significance level of 0.0041, all the associations became not significant.

Because asthma is a clinical syndrome and there are no clinical benchmarks, several asthma-related phenotypes are used to help the diagnosis. Since these phenotypes are objective, quantitative, and of less genetic heterogeneity, they are often used in genetic studies to search for the causal genetic variants for asthma [26]. Evidence has shown that combining asthma-related phenotypes with physician's diagnosis of asthma led to a stronger association to the asthma candidate genes, which may shed light on the different molecular pathways underlying asthma [4, 22]. In our current study, we selected study samples by combining asthma affectation status and three asthma-related phenotypes, including total serum IgE levels, bronchial responsiveness test, and skin prick test. This study design provided more power to examine various mechanisms of asthma. We observed that combining BHR with asthma produced stronger genetic effects (odds ratios) in both DPP10 single SNP and haplotype analyses as compared to BHR alone. This indicates that the gene may participate in asthma susceptibility through the mechanisms of BHR.

Our results confirm previous associations of DPP10 and asthma or asthma-related phenotype, but differ in the details. In the positional cloning study of DPP10, Allen et al. demonstrated that the strongest association signal to asthma came from the allele 3 of a microsatellite marker D2S308 in a population containing three Caucasian cohorts [2]. Studies have shown that DPP10 is strongly expressed in central neural system and is a modulator of voltage-gated K+ channel inactivation [2, 11]. This indicates that DPP10 may be important in neural regulation of airway smooth muscle and control of airway reactivity, which may explain our finding that the haplotype containing the SNP rs10192393 in DPP10, 4 kb upstream of D2S308, was associated with BHR and BHR asthma. For PHF11 and HLA-G, the associations with asthma or asthma-related phenotypes were only significant at the nominal level, and became not significant after the adjustment for multiple testing. However, these results might still be of interest. Zhang et al. identified the locus at the PHF11 gene as a determinant of serum IgE levels [1]. However, we did not find any association between PHF11 and IgE in our Chinese cohort even at the nominal significance level. Recently, Jang et al. provided evidence for an association between rs2247119 and rs1046295 in PHF11 and atopic dermatitis in a young Australian cohort [7]. In their study, the T allele from rs2247119 and G allele from rs1046295 were over-transmitted to children with atopic dermatitis. The correlation between these two SNPs in Chinese population (r² = 0.53) is moderate compared with that in Australian population (r² = 0.92) [7]. In our current study, however, we found that the haplotype containing the C allele of rs2247119 (rather than the T allele) and the G allele of rs1046295 was nominally associated with an increased risk of positive skin reactions to antigens. Nicolae et al. identified HLA-G as asthma and BHR susceptibility gene in two Chicago cohorts by examining pairwise combination of SNPs [3]. The association of SNP rs1632947 in the promoter region of HLA-G with BHR was confirmed in the Hutterite families, where the C allele was over-transmitted to children with BHR. In our current study, we found that the genotype containing the C allele at rs1736935, one of the two selected tag SNPs 215 bp upstream of rs1632947, and the haplotype containing the rs1736935 C allele were nominally associated with an increased risk of BHR asthma. It is probably due to the small effect sizes of the genetic variants and lack of power in the current study that the associations of PHF11 and HLA-G with asthma or asthma-related phenotypes failed to reach the study-wide significance level. Therefore, studies with larger sample sizes can be conducted to further evaluate the nominal associations of PHF11 and HLA-G identified in the current study.

The population-based study design is a common strategy used in genetic studies of complex diseases. Population homogeneity is desirable to avoid population stratification. Our study samples are relatively homogenous. All of them are from a rural area of Anqing, China, a region known for a historically stable population. Not only is there a predominant singular ethnic group, but there exists a uniformity of occupation (farming) and lifestyle throughout Anqing. This population homogeneity minimizes the potential of population stratification that may lead to false-positive associations. Also, in a previous asthma candidate gene study, we tested 119 SNPs from 105 genes for associations with asthma using 170 asthmatic cases and 347 controls selected from the same 3022 families where the current study subjects were sampled [27]. We applied STRUCTURE2.1 to assess the presence of population substructures within our previous study samples. A set of 111 SNPs after excluding potentially associated ones (p < 0.05 in association tests) was used in the analyses. STRUCURE2.1 showed that a single substructure had the highest likelihoods among all possible number of substructures in the samples, which indicated that the samples in our previous study were homogeneous. Even though the study samples were different between the current and previous studies, they were actually selected from the same population pool in the same geographic region, so the conclusion of no population stratification based on one study can be generalized to the other study.

In the current study, several demographic factors, including age, height, weight, and BMI, were slightly imbalanced between asthmatic and non-asthmatic samples. To minimize their potential confounding effects, we included these factors as covariates in all the analyses. Furthermore, we examined the distributions of these factors among samples with different genotypes of studied SNPs and found no significant differences. This indicated that there were no correlations between studied SNPs and four above-mentioned factors. Therefore, the results were not confounded by age, height, weight, and BMI. Also, gender did not bias the results since it was balanced between asthmatic and non-asthmatic samples. Moreover, the genetic effects were similar in gender-specific (males and females) analyses and there was no significant interaction between gender and the DPP10 haplotype in the log-likelihood-ratio test (data not shown). Note that our study subjects include both children and adults. Although evidence showed that childhood and adult asthma differed in several parameters [28, 29], the associations between PHF11, DPP10, HLA-G and asthma or asthma-related phenotypes were found in both groups [1–3, 7]. Therefore, these genes may play roles in the common pathway of childhood and adult asthma. Further subgroup (subjects aged <21 yrs and ≥21 yrs) analyses were performed and similar associations were detected in both younger and older groups; also, there was no significant interaction between age and the DPP10 haplotype by the log-likelihood-ratio test (data not shown). This suggests that age may have a minimal effect on the associations.

In conclusion, our study was the first one to demonstrate that DPP10 is associated with BHR and BHR asthma in a Chinese population. Our study provides supportive evidence to the previously reported associations between the DPP10 gene and asthma or asthma-related phenotypes in Caucasian populations.