Article Text
Abstract
Background Previous genetic association studies have reported evidence for association of single-nucleotide polymorphisms (SNPs) in the NOS2 gene, encoding inducible nitric oxide synthase (iNOS), to variation in levels of fractional exhaled nitric oxide (FENO) in children and adults. In this study, we evaluated 10 SNPs in the region of chromosome 17 from 26.07 Mb to 26.13 Mb to further understand the contribution of NOS2 to variation in levels of FENO.
Methods In a cohort of 5912 adults 25–75 years of age, we investigated the relationship between NOS2 haplotypes and FENO, and effect modification by asthma.
Results Seven common (frequency ≥5%) haplotypes (H1–H7) were inferred from all possible haplotype combinations. One haplotype (H3) was significantly associated with lower levels of FENO: −5.8% (95% CI −9.8 to −1.7; p=0.006) compared with the most common baseline haplotype H1. Two haplotypes (H5 and H6) were significantly associated with higher levels of FENO: +10.7% (95% CI 5.0 to 16.7; p=0.0002) and +14.9% (95% CI 10.6 to 19.3; p=7.8×10−13), respectively. The effect of haplotype H3 was mainly seen in subjects with asthma (−21.6% (95% CI −33.5 to −5.9)) and was not significant in subjects without asthma (−4.2% (95% CI −8.4 to 0.2)). The p value for interaction between H3 and asthma status was 0.004.
Conclusions Our findings suggest that several common haplotypes in the NOS2 gene contribute to variation in FENO in adults. We also saw some evidence of effect modification by asthma status on haplotype H3.
- Asthma
- Complex traits
- Genetic epidemiology
Statistics from Altmetric.com
Introduction
The fractional concentration of nitric oxide in exhaled air (FENO; fraction of exhaled NO) is suggested as a useful biomarker of airway inflammation.1 Studies have shown increased levels of FENO in individuals with asthma2 ,3 and FENO correlates with eosinophilic airway inflammation while inhaled corticosteroids reduce FENO.1 Increased levels of FENO can also be observed in adults and children without respiratory symptoms or asthma, and there are indications that increased FENO may be a marker of future risk of new onset of respiratory symptoms or asthma.4 ,5
Nitric oxide (NO) is formed by conversion of L-arginine to L-citrulline in the presence of one of three distinct isoforms of NO synthase (NOS) known as neuronal NOS (nNOS;NOS1), inducible NOS (iNOS; NOS2) and endothelial NOS (eNOS; NOS3), coded by the three genes NOS1, NOS2 and NOS3, respectively.6 nNOS and eNOS each generate small amounts NO in the lung. iNOS, on the other hand, generates substantial amounts of NO primarily in response to inflammatory stimuli such as cytokines, oxidants and infections.7 The expression of NOS2 in human airway epithelium cells is regulated by various inflammatory (eg, nuclear factor kappa B (NFκB)) and non-inflammatory (eg, Kruppel-like factor) transcriptional factors, suggesting that several diverse pathways are involved.7 ,8 A gene expression study has demonstrated that enhanced expression of the NOS2 gene in the epithelial cells of the airways is related to increased FENO levels in healthy individuals.9 Several candidate gene association studies in both children and adults focusing on NOS genes have reported that polymorphisms in the NOS2 gene influence levels of FENO,10–12 and the importance of NOS2 for FENO was very recently confirmed by a GWAS of FENO in children that identified three variants significantly associated with FENO at a genome-wide level; one of these was a single-nucleotide polymorphism (SNP) (rs944722) in NOS2.13 In terms of the more focused candidate gene studies, we reported in a previous study two tag SNPs in NOS2, rs9901734 and rs3729508, that were associated with FENO levels in adults without asthma or respiratory symptoms or atopy.12 A recent study by Bouzigon et al also provided evidence of a relationship between SNPs in the NOS2 gene and FENO in non-asthmatic adults.11 The earliest of these three studies, by Salam et al,10 studied children and identified comprehensive sets of haplotypes in the promoter, coding and downstream regions in NOS2 and found associations of these haplotypes with FENO in children with and without asthma. However, important details about the contribution of the NOS2 gene to variation in FENO levels are still lacking. In this study, we fine-mapped the previous findings by others and us to identify haplotypes that influence FENO levels. We also investigated possible effect modification by asthma status at the haplotype level.
Materials and methods
Study population
We used the population-based Adult Onset Asthma and Nitric Oxide (ADONIX) cohort of men and women, aged 25–75 years at the time of sampling and recruited between 2001 and 2008 in Gothenburg, Sweden. The ADONIX cohort includes in total 6679 participants. Part of the study population (1737 subjects) recruited during 2001–2003 were also included in our previous work on the NOS genes and FENO.12
The original study design and protocol of the ADONIX study have been described in detail elsewhere.14 ,15 In brief, all participants received a questionnaire related to current respiratory health status, medical history and smoking habits. Participants who responded to the questionnaire were also invited for clinical examination, which included blood sampling and measurements of FENO. In the present study, asthma was defined based on a positive answer to at least one of the questionnaire items: ‘Have you ever had asthma?’; ‘Have you ever had asthma diagnosed by a doctor?’; ‘Have you had an attack of asthma during the last 12 months?’; ‘Have you had asthma during the past month?’ Atopy was defined as the presence of specific serum IgE antibodies (≥0.35 kU/L) to any of eight common inhaled allergens (dog, cat, horse, timothy grass, birch, mugwort, house dust mite and cladiosporum) as determined by the Phadiatop test (PharmaciaDiagnostics; Uppsala, Sweden).16 Participants were classified as smokers or non-smokers based on their reported smoking habits.
Measurements of FENO
FENO was measured with an online NO monitoring system (NIOX; Aerocrine AB, Stockholm, Sweden) at an exhalation flow of 50 mL/s, according to the 2005 ATS/ERS recommendations,17 after at least 4 h of fasting. Exhalations were registered for each subject within 10% deviation, and in triplicate between June 2001 and January 2003, and in duplicate from February 2003 through 2008 according to the revised ATS/ERS recommendations.17 The mean concentration was used for analyses.
SNP selection and genotyping
Ten SNPs in the region spanning from 26.07 Mb to 26.13 Mb of the NOS2 gene (the NCBI build 36) were selected for analysis based on previously published data regarding their significant association with FENO (table 2).10 ,12 Six SNPs (rs9901734, rs2297514, rs2248814, rs12944039, rs3729508 and rs2779248) were main findings from our previous analysis.12 The remaining four SNPs (rs4796017, rs2297520, rs9895453 and rs10459953) were SNPs from the Salam et al Southern California Children's Health Study10 that either were not in close linkage disequilibrium (LD) with any of our earlier findings (their SNP rs4796017) or (their SNPs rs2297520, rs9895453 and rs10459953) were alternative markers in moderate-to-high LD with one of our main findings (rs3729508), as potential alternative better markers for this finding and to ensure successful genotyping of at least one marker for this finding across the entire population sample. SNPs were genotyped using a Sequenom MassARRAY platform (Sequenom San Diego, California, USA) or a competitive allele-specific PCR system, KASPar (KBioscience, Hoddesdon Herts, Great Britain). Genotyping call rate for all SNPs was ≥98%, and all the SNPs were in Hardy–Weinberg Equilibrium (HWE) (p≥0.001). Samples with genotyping success rate ≤80% across the 10 SNPs were excluded from the study. Only subjects reporting European country of birth were included in the present study; 96% were of Swedish origin.
Statistical analysis
Stepwise regression in a forward approach was performed in SAS (V.9.2, SAS Institute) assuming five different genetic models as previously described12 to find a subset of SNPs in the NOS2 gene that were strongly associated with FENO. The SNPs all had prior significant evidence of association with FENO, and nominal p values were used for the purpose of ranking SNPs in the selection process. Essentially, we performed an analysis in two stages using forward stepwise multiple linear regression analysis to identify a reduced set of the strongest independently associated SNPs among those studied, and the most appropriate genetic model for each final SNP. In the first stepwise analysis stage, SNPs with p≤0.2 from the additive genetic model in the single-SNP association analysis were entered into the stepwise analysis, which aimed to determine the most strongly associated SNP or SNPs in a model allowing for independent effects of several SNPs. In the second stage, significant SNPs (p≤0.05) from the first stage stepwise regression model, as well as any SNPs from the initial single-SNP association analyses with p≤0.05 from any non-additive (dominant, recessive, over-dominant and co-dominant) genetic models, were included. This selection aimed to ensure that genetic effects not following an additive model would be adequately captured in this final stage. All of the SNPs thus selected and coded into all five genetic models were used as input for a forward stepwise regression analysis in order to identify the most predictive SNPs, each with the best-fitting genetic model, allowing for independent effects of several SNPs. Each SNP was only allowed to enter in one genetic model coding at any one time. In both stepwise stages, to ensure modelling flexibility, p values were set to 0.10 and 0.20 for a SNP to enter and remain in the model, respectively. The stepwise procedure produced a model with seven SNPs. After discarding the two least significant SNPs with p>0.01 (which turned out to be the two SNPs selected at steps 6 and 7 of the stepwise analysis), five SNPs (corresponding to the five SNPs selected at step 5) were retained for haplotype analysis.
Using the five selected SNPs, we obtained all possible haplotype pairs along with their respective likelihoods for each individual, and resulting population frequencies using an E-M (expectation-maximisation) algorithm as implemented in the ‘haplo.stats’ package in the R statistical program.18 The E-M algorithm estimates haplotype frequencies and posterior probabilities of each pair of haplotypes in each individual given observed genotypic data using a likelihood approach. Since our focus was on identifying more common haplotypes of importance, for optimal power, the haplotype analysis was primarily performed with the limit for ‘rare’ haplotypes set at 5%. As a supporting analysis, the analysis was repeated with the cutpoint set at 1%. In addition, haplotypes were also estimated separately for each of the two haplotype blocks with low LD between them that were identified (two and three of the final five SNPs, respectively). The analysis within each haplotype block sought to identify the best marker of an effect within that block, and the 5-SNP analysis across blocks sought to identify the combined effect pattern of genetic variation in the two blocks on the same chromosome.
A generalised linear model using the haplo.glm function implemented in the ‘haplo.stats’ package was used for estimating the effect of all common haplotypes on levels of FENO using the most common haplotype as the reference group and assuming an additive genetic model. We also evaluated the haplotype effects of the 5-SNP haplotypes on FENO by asthma status by including product term between the haplotypes and asthma in the main effect model.
Since the distribution of FENO was skewed, values were log-transformed prior to analyses. We adjusted all analyses for age, sex, height, atopy and smoking habits. Results are presented as a percentage change in the geometric mean of FENO across a group of subjects, comparing each SNP or haplotype to the respective reference.
Results
In the total cohort (n=6679), DNA was available for 6340 participants. Of these, 5963 (94%) were of European origin and 377 (6%) of non-European origin. We excluded 51 individuals due to poor genotyping quality, leaving 5912 European participants. Among these, 5633 participants had FENO values and constituted the final analysis set. Basic characteristics of the study population and FENO levels are presented in table 1. Geometric mean (±SD) of FENO level was 16.4 (±1.8) parts per billion (ppb) for all subjects. Descriptive statistics and publication source of the 10 SNPs are shown in table 2. The LD pattern of the 10 SNPs in the NOS2 gene, with corresponding r2 values, and identified haploblocks, are shown in figure 1.
The stepwise analysis identified a subset of five SNPs with independent associations with FENO (table 3). The minor allele of one SNP, rs3729508(C/T), showed a negative association, and the other four SNPs (rs4796017(A/G), rs9901734(C/G), rs9895453 (T/C) and rs2779248 (T/C)) showed positive association for the minor allele, with different genetic models. For reference, the single-SNP results for all initial 10 SNPs in the current study population are shown in online supplementary table S1.
The common haplotypes defined by the subset of five SNPs spanning the gene across both identified haploblocks, and their frequencies, are presented in table 4. Seven common haplotypes, each with a frequency of ≥5% in the population, accounted for 84% of all possible haplotype combinations, and rare (<5%) haplotypes for approximately 16%. The most common haplotype (ACTCT) was chosen as the baseline haplotype (H1). We found clear and significant associations in the overall population between three haplotypes (H3, H5 and H6) and FENO (table 4). Haplotypes H5 (GGCTC) and H6 (GGCTT) were significantly associated with higher levels of FENO, +10.7% (95% CI 5.0 to 16.7; p=0.0002) and +14.9% (95% CI 10.6 to 19.3; p=7.8×10−13), respectively, compared with the baseline haplotype. Haplotype H3 (ACCTT) was significantly associated with lower levels of FENO, −5.8% (95% CI −9.8 to −1.7, p=0.006). The analyses separately by haploblock indicated for haploblock A that the main difference was between the patterns AC and G*, with haplotypes including the G allele for rs4796017 showing higher FENO levels (see online supplementary table S3A). For haploblock B, the pattern *CC showed the highest FENO levels, with CT* patterns intermediate (see online supplementary table 3B). There were clear effects in both haploblocks, suggesting that both are important for the overall effect of NOS2 genetic variation, supporting the 5-SNP analysis. When the 5-SNP analysis was repeated with a lower cutpoint (≤1%) for defining rare haplotypes, the results indicated several additional haplotypes with frequencies 1–5% characterised either by the pattern G*CTT (analogous to H6) or a new rarer pattern ***CC, with strong effects (≥+15.0%) and statistical significance in relation to the most common (reference) haplotype (see online supplementary table S2).
Among subjects with asthma, H2, H4, H5 and H7 were associated with lower FENO values, while an opposite effect was indicated among subjects without asthma. For H6, the effect was positive in both groups, but stronger in subjects with asthma. However, these differences were statistically not significant. There was, however, a statistically significant difference in the association between H3 and FENO in subjects with asthma as compared to in subjects without asthma (p value for interaction=0.004), with a markedly stronger negative effect in subjects with asthma than in subjects without asthma (−21.6, 95% CI −33.5 to −5.9 vs −4.2, 95% CI −8.2 to 0.2) (table 5).
Discussion
In this study of Swedish adults, we identified three relatively common haplotypes in the NOS2 gene that were strongly and significantly associated with levels of FENO in an adult general population. One haplotype (H3, ACCTT) was associated with lower FENO and the two other (H5, GGCTC; and especially H6, GGCTT) with higher FENO compared with the most common haplotype (ACTCT). Analyses of haploblocks separately suggested that haplotypes in haploblock A (the first two of the five SNPs) that included the G allele for rs4796017 showed higher FENO levels, and haplotypes in haploblock B (last three SNPs of the five) with the pattern *CC (especially haplotype CCC) showed higher FENO levels. These findings were confirmed and expanded in the 5-SNP analysis including rarer haplotypes (frequency from 1%), where particularly high FENO levels were seen with haplotype combinations G*CTT (H6 and a rarer haplotype), as well as with several rarer ***CC haplotypes. In total, these results suggest that there may be at least two, possibly more variants, in the gene that in particular combinations contribute to variability of effect on FENO, with most of the common and rare variants associated with higher levels of FENO compared with the most common (reference) haplotype, some substantially and significantly higher. In addition to the main effects, we also observed suggested difference between subjects with asthma and free from asthma, with a significant effect modification by asthma status for association with haplotype H3, the haplotype with a negative effect compared with the most common haplotype.
The human NOS2 gene is one of several genes encoding a NOS enzyme isoform, and it has been demonstrated that expression of this isoform was associated with exhaled NO levels in children.9 So far, only our previous study and two others have attempted to study the association between polymorphisms in the NOS2 gene and FENO levels in either adults or children using a candidate gene approach.10–12 In children, various SNPs including rs2297512, rs2774894, rs8081248 and rs4796017 (Salam et al10), and in adults rs9901734, rs2297514 (Dahgam et al12) and, most recently, rs12601458, rs6505510 (Bouzigon et al11) have been reported to be associated with FENO among subjects without asthma. In our analysis, of the 10 investigated SNPs selected from our study and Salam et al, 5 showed significant, and independent, evidence of association with FENO levels. This result is in line with the most recent report from the French Epidemiological study on the Genetics and Environment of Asthma (EGEA) assessing effects of genetic variants of the NOS genes on exhaled NO among non-asthmatic adults.11 However, direct experimental evidence to support a functional role of the studied NOS2 polymorphisms is currently lacking. A recent genome-wide association study clearly supported a role for NOS2 variants in affecting FENO levels in children by identifying a very common NOS2 SNP (rs944722, MAF 0.38) with genome-wide significance, but given its design and size, the study had greatest power for more common SNPs and thus provides little additional evidence regarding involvement of rarer and/or multiple NOS2 SNPs, or the potential NOS2 haplotypes relevant for protein variation and the downstream effect of the gene.13
Haplotype analysis, which may provide additional information beyond individual SNP analysis about the genetic basis of complex traits and can be helpful in understanding the unit of biological function, has become of more widespread interest in finding casual connections in candidate genes studies.19–21 Furthermore, constructing haplotypes from a subset of informative SNPs reduces the haplotype dimensionality and increases power for detecting associations as compared with separate analysis of individual SNPs.21–23 Using this approach in our study, five-SNP haplotype analysis identified seven common haplotypes, constituting 84% of all haplotypes based on the defining SNP set, and revealed that four of these showed strong and significant association with FENO. Although significant associations were found also for individual SNPs, the haplotype analysis revealed very strong associations, and the haplotype model provided a much stronger global p value (3.8×10−28) for association than individual SNPs. The total analysis also clearly suggested that the gene is likely to harbour several variants contributing to the variability in FENO levels. The association of FENO with one haplotype (H3) clearly differed between subjects with and without asthma. Our results generally support the findings in children with and without asthma by Salam et al who also report several haplotypes of NOS2 that were associated with increased FENO levels. However, Salam et al10 used a population-based sample of children that were partly of European and non-European ancestry (non-Hispanic white and Hispanic white), and another set of SNPs to infer all possible haplotypes that occurred to describe variation in levels of FENO. So neither the two study populations nor the haplotype analysis methods are entirely comparable.
Potential clinical implications are interesting to consider, given that the variability in FENO related to the NOS genes is approximately similar to that of non-genetic factors. Individuals with the haplotype that was associated with higher FENO (H6; frequency 10%) had on average 21% higher FENO than individuals with the haplotype that was associated with the lowest FENO levels (H3; frequency 9%). Our previous single-SNP analyses showed that individuals with all of the studied FENO-associated NOS2 and NOS3 SNPs (21% of population) had on average 35% higher FENO than individuals with none of the FENO-associated NOS2 and NOS3 SNPs (4% of the population).12 In the same population as in this study, current smokers had 30% higher FENO than non-smokers, and subjects with asthma had 16% higher FENO than healthy subjects.24 Consequently, the current genetic findings are of sufficient magnitude so that they may be useful to provide improvements in predicting normal FENO levels in the general population. However, at present genotyping is not a routine test, and it is not possible to evaluate measured FENO levels in the scope of presence/absence of NOS polymorphisms in a clinical setting. With further development of genotyping techniques, this may, however, become feasible and would enhance the interpretation of an individual FENO value.
This study extends our previous work on FENO association. The strengths of this study include a large study sample from a homogenous adult population and a strong biological a priori hypothesis. In addition, haplotypes were inferred from a set of SNPs with previously reported association with FENO rather than by just selecting random tag SNPs, to some extent avoiding issues of multiple comparisons that can be of importance when testing large numbers of independent and a priori mainly false hypotheses. The SNPs we used to infer the haplotypes are not located in coding or promoter regions. Potential biological reasons for our results could be that the haplotypes are in strong LD with unmeasured causal genetic alterations or that they are involved in more complex regulation of gene expression.21 Further studies can build on these findings to search for causal genetic variants with respect to the studied gene region.
Population stratification is a common confounding issue in genetic association studies.25 In this study, we attempted to limit potential confounding from population stratification by including only subjects that were homogeneous with respect to ethnic background (ie, subjects reporting European country of birth). To increase the validity of the genetic data used, we employed several quality control methods, for example, adequate call rate and assessment of HWE. SNPs with call rate ≤95% were excluded from the study, and all SNPs that were included in the analyses were in HWE.
In summary, the analysis of haplotypes in the NOS2 gene described the NOS2-related contribution to variation in FENO more in depth than the analysis of individual SNPs. Haplotype-defining SNPs were located relatively far apart in the gene, suggesting that there may be several genetic regions within NOS2 that are responsible for variation in FENO. The observation that both common and rare haplotypes confer higher levels of FENO, some substantially and significantly higher, compared with the most common haplotype, could also indicate a role for this common haplotype in lowering baseline FENO levels. Furthermore, one haplotype in the NOS2 gene contributed to the variation in FENO much more strongly in subjects with asthma than in healthy subjects, whereas the other haplotypes showed less contribution.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
Contributors SD contributed to the development of the research question, the data analysis and interpretation of the results, as well as the manuscript development. A-CO and FN contributed to the design of the study, interpretation and discussion of results, manuscript development and approval of the final version. ATN contributed to the genotyping, interpretation and discussion of results, manuscript development and approval of the final version. LM contributed to interpretation and discussion of results, manuscript development and approval of the final version of the manuscript.
-
Competing interests A-CO has received honoraria from Aerocrine for giving lectures on exhaled NO. FN, ATN and LM and SD have reported that no potential conflicts of interest exist with any companies/organisations whose products or services may be discussed in this article. FN is an employee of AstraZeneca in addition to his adjunct academic affiliation.
-
Patient consent Obtained.
-
Ethics approval The local Ethics Committee at Gothenburg University, Sweden.
-
Provenance and peer review Not commissioned; externally peer reviewed.