Parental Smoking Modifies the Relation between Genetic Variation in Tumor Necrosis Factor-α (TNF) and Childhood Asthma

Background Polymorphisms in the proinflammatory cytokine genes tumor necrosis factor-α (TNF) and lymphotoxin-α (LTA, also called TNF-β) have been associated with asthma and atopy in some studies. Parental smoking is a consistent risk factor for childhood asthma. Secondhand smoke and ozone both stimulate TNF production. Objectives Our goal was to investigate whether genetic variation in TNF and LTA is associated with asthma and atopy and whether the association is modified by parental smoking in a Mexican population with high ozone exposure. Methods We genotyped six tagging single nucleotide polymorphisms (SNPs) in TNF and LTA, including functional variants, in 596 nuclear families consisting of asthmatics 4–17 years of age and their parents in Mexico City. Atopy was determined by skin prick tests. Results The A allele of the TNF-308 SNP was associated with increased risk of asthma [relative risk (RR) = 1.54; 95% confidence interval (CI), 1.04–2.28], especially among children of non-smoking parents (RR = 2.06; 95% CI, 1.19–3.55; p for interaction = 0.09). Similarly, the A allele of the TNF-238 SNP was associated with increased asthma risk among children of nonsmoking parents (RR = 2.21; 95% CI, 1.14–4.30; p for interaction = 0.01). LTA SNPs were not associated with asthma. Haplotype analyses reflected the single SNP findings in magnitude and direction. TNF and LTA SNPs were not associated with the degree of atopy. Conclusions Our results suggest that genetic variation in TNF may contribute to childhood asthma and that associations may be modified by parental smoking.

The etiology of asthma and atopy involves interactions between genetic susceptibility and exposure to environmental triggers, such as secondhand smoke, ozone, particulate matter, allergens, and endotoxin (Tatum and Shapiro 2005). TNF has been identified as a candidate gene for ozoneinduced airway inflammation and hyperresponsiveness (Kleeberger et al. 1997), and genetic variation in TNF and LTA has been associated with respiratory effects of ozone in humans (Yang et al. 2005). Parental smoking has been consistently related to childhood asthma [U.S. Department of Health and Human Services (DHHS) 2006] and TNF may influence the lung inflammatory response to tobacco smoke (Churg et al. 2003;Park et al. 2003). Few studies have examined whether exposure to parental smoking modifies the relationship between TNF and LTA polymorphisms and asthma risk. We used the case-parent triad design to investigate the association of TNF and LTA polymorphisms and haplotypes with childhood asthma and atopy in asthmatic children from Mexico City, an area with the highest ozone exposure in North America. We also examined possible effect modification by exposure to parental smoking.

Study design and subject enrollment.
We used the case-parent triad design Wilcox et al. 1998). The study population included 536 case-parent triads and 60 case-parent pairs with adequate DNA samples for genotyping of at least one SNP for the TNF and LTA genes. The cases were children 4-17 years of age with asthma diagnosed by a pediatric allergist at the allergy referral clinic of a large public pediatric hospital in central Mexico City (Hospital Infantil de Mexico Federico Gomez). Children and parents provided blood samples as sources of DNA. A parent, nearly always the mother, completed a questionnaire on the child's symptoms and risk factors for asthma including current parental smoking, parental smoking during the first 2 years of the child's life, maternal smoking during pregnancy, and residential history.
We obtained measurements of ambient ozone from the Mexican government's air monitoring stations (http://www.ine.gob.mx/ dgicurg/calaire/tend/concentra.php). Ozone levels were measured via ultraviolet photometry (analyzer model 400, Advanced Pollution Instrumentation, San Diego, CA, USA). The residence of each child who participated in this study was located using a map, and the closest monitoring station was assigned to that residence (Romieu et al. 2002). The ozone exposure data were collected for the year before the time of entry into the study. The parameter we used was the annual average of the daily maximum 8hr averages. We dichotomized this variable at the median of 67 ppb for stratified analyses.
The protocol was reviewed and approved by the institutional review boards of the Mexican National Institute of Public Health, the Hospital Infantil de Mexico Federico Gomez, and the U.S. National Institute of Environmental Health Sciences (NIEHS). Parents provided the written informed consent for the child's participation. Children also gave their informed assent.
Clinical evaluation. The diagnosis of asthma was based on clinical symptoms and response to treatment by a pediatric allergist [British Thoracic Society/Scottish Intercollegiate Guidelines Network (BTS/SIGN) 2003]. The severity of asthma was rated by a pediatric allergist for 571 cases according to symptoms in the Global Initiative on Asthma schema as mild (intermittent or persistent), moderate, or severe [National Heart, Lung, and Blood Institute (NHLBI) 1998]. At a different point of time, for research purposes, pulmonary function was measured using the EasyOne spirometer (ndd Medical Technologies, Andover, MA, USA) for 446 cases according to American Thoracic Society (ATS) specifications (ATS 1995). The best test of three technically acceptable tests was selected. Spirometric prediction equations from a Mexico City childhood population were used to calculate the percent predicted forced expiratory volume in 1 sec (FEV 1 ) (Perez-Padilla et al. 2003). Children were asked to hold asthma medications on the morning of the test.
Atopy was determined using skin prick tests. The following battery of 24 aeroallergens common in Mexico City was used: Aspergillus fumigatus, Alternaria, Mucor, Blattella germanica, Periplaneta americana, Penicillium, cat, dog, horse, Dermatophagoides (pteronyssinus and farina), Ambrosia, Artemisa ludoviciana, Cynodon dactylon, Chenopodium album, Quercus, Fraxinus, Helianthus, Ligustrum vulgare, Lolium perenne, Plantago lanceolata, Rumex crispus, Schinus molle, Salsola, and Phleum pratense. Histamine was used as a positive control and glycerin as a negative control. Children were considered atopic if the diameter of the skin reaction to at least one allergen exceeded 4 mm. The test was considered valid if the reaction to histamine was ≥ 6 mm according to the grading of skin prick test recommended by Aas and Belin (1973). Skin test data on all 24 aeroallergens were available on 545 cases. SNP selection. We had various data sources available for selection of tagging SNPs. These included resequencing data in individuals of African and European descent by Seattle SNPs (http://pga.mbt.washington.edu) and genotyping data on 10 Mexicans and 38 Mexican-Americans with four grandparents born in Mexico for cosmopolitan haplotype tagging SNPs identified by the NIEHS Environmental Genome Project based on resequencing a representative sample of the U.S. population (http://egp.gs.washington.edu). We identified the common haplotypes using PHASE (Stephens et al. 2001) and then used ldSelect (Carlson et al. 2004) to identify tagging SNPs. We also analyzed the genotyping data from seven Mexicans in SNP500Cancer (Packer et al. 2004) (http://snp500cancer. nci.nih.gov). TNF and LTA are small genes located consecutively in a 6 kb region on chromosome 6, and linkage disequilibrium in this region is high. We selected six tagging SNPs-TNF-1031 (rs1799964), TNF-857 (rs1799724), TNF-308 (rs1800629), TNF-238 (rs361525), LTA-379 (rs2239704), and LTA NcoI (rs909253)-to cover the whole region, including all the common SNPs in the regulatory and coding regions of the genes with known functional importance (Fong and Mark 1995;Knight et al. 2004;Kroeger et al. 1997;Messer et al. 1991;Ozaki et al. 2002;Wilson et al. 1997) or that had been associated with asthma in the literature. As expected, given the increasing documentation of the portability of tagging SNPs across populations (Gonzalez-Neira et al. 2006), especially in the less diverse non-African groups, the tagging SNPs selected provide excellent coverage of common haplotypes in other populations. For example, the six tagging SNPs we selected based on the Mexican data would cover all six common haplotypes (> 5%) in European and five of six common haplotypes in African populations.
Genotyping. We extracted DNA from peripheral blood lymphocyte using Gentra Puregene kits (Gentra System, Minneapolis, MN, USA). We obtained genotypes for the TNF-1031 and TNF-857 SNPs using TaqMan SNP Genotyping Assay (Li H et al. 2006). Primers and probes were purchased from Assay-on-Demand (Applied Biosystems, Foster City, CA, USA). All PCR amplifications were performed using 5′ exonuclease assay on GeneAmp PCR Systems 9700 (Applied Biosystems). The fluorescence of PCR products was detected using ABI Prism 7900HT sequence detection system. The TNF-308, TNF-238, and LTA NcoI SNPs were genotyped using a multiplex PCR and immobilized probe linear array system (Barcellos et al. 2004), provided by Roche Molecular Systems (Alameda, CA, USA). The LTA-379 SNP was genotyped using MGB Eclipse Genotyping Assay (Belousov et al. 2004). Primers and Probes were purchased from MGB Eclipse by Design (Epoch Biosciences, Bothell, WA, USA). All genotyping assays were done by a researcher who was blinded to parent or child status of samples. Sixteen quality control samples were plated per 384-well plate along with 24 control samples with known genotype. An additional six blind replicate samples were included in the analyses. The quality controls and the blind replicates were 100% concordant for all genotyping methods.
Nonparentage was ascertained with a set of short-tandem repeats (AmpFLSTR Profiler Plus; Applied Biosystems) analyzed using Pedcheck software (University of Pittsburgh, Pittsburgh, PA, USA) (O'Connell and Weeks 1998). A total of 596 families had genotyping data for at least one SNP, and 566 families had genotyping data for all six SNPs.
Parental smoking, TNF, and asthma Environmental Health Perspectives • VOLUME 115 | NUMBER 4 | April 2007 Statistical analysis. We used a log-linear likelihood approach to analyze associations between asthma and individual SNPs ). The log-linear likelihood-ratio test is a powerful and more flexible alternative to the transmission disequilibrium test (TDT) and tests the same null hypothesis of no within-family relationship between the variant and the disease (Lake and Laird 2004). Similar to TDT-based methods for the analysis of case-parent data, such as the family-based association test (FBAT) (Horvath et al. 2001), the log-linear model achieves robustness against genetic population structure through stratification on the possible parental mating types (Lake and Laird 2004). The log-linear method has the advantage of providing estimates of the magnitude of associations rather than simply tests of significance ). We calculated relative risks for individual SNPs without restricting to a specific genetic model. The log-linear models of case-parent data are inherently immune to confounding by demographic or lifestyle factors such as parental smoking. However, we examined effect modification by sex, asthma severity, parental smoking, and level of ozone exposure. We calculated tests of interactions for the joint effects of genotype and current parental smoking and parental smoking before the child turned two using the method of Umbach and Weinberg (2000). All analyses were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) and STATA version 8.0 (StataCorp., College Station, TX, USA).
To evaluate whether TNF and LTA polymorphisms influenced the degree of atopy, as assessed by the number of positive skin tests out of 24 performed, we used the polytomous logistic method of Kistner and Weinberg to estimate the linkage and association between TNF and LTA polymorphisms and atopy (Kistner and Weinberg 2004). p-Values were calculated from likelihood ratio tests. We also used this method to analyze the relationship between TNF and LTA SNPs and lung function, as assessed by percent predicted FEV 1 .
We used HAPLIN version 2.0 (http:// w w w . u i b . n o / s m i s / g j e s s i n g / g e n e t i c s / software/haplin) to analyze associations between asthma and TNF and LTA haplotypes. HAPLIN is an extension of the loglinear model from a single locus to loci with multiple haplotypes with unknown phase (Gjessing and Lie 2005). The haplotypes of individuals with unknown phase are constructed from the family information whenever possible, and the remaining haplotypes are estimated by using the expectationmaximization algorithm (Gjessing and Lie 2005). HAPLIN estimates single-and double-dose effects of haplotypes rather than simply tests of significance using maximum likelihood (Gjessing and Lie 2005). We set a threshold of 1% for haplotype frequency, leaving 582 families in the HAPLIN analyses.
We present the p-values computed using the above methods. To address the potential issue of multiple comparisons, we calculated the false discovery rate for each p-value < 0.05 using the method of Storey (Storey and Tibshirani 2003). The false discovery rate is the expected proportion of false positives incurred when a particular test is called significant. However, these corrections will be overly conservative when applied equally to all SNPs. Our prior prediction of positive findings would be greatest for SNPs with known functional importance, such as TNF-308, compared with SNPs chosen only as haplotype tagging SNPs where functional significance is not well characterized. In addition, the false discovery rate does not take into account the correlation between SNPs in a gene.

Results
Clinical characteristics of the asthmatic children with genotyping data are presented in Table 1. The mean (± SD) age of cases was 9.0 ± 2.4 years (range 4-17 years). Most had mild (71.5%) as opposed to moderate or severe asthma (28.5%). Nearly all cases (98.3%) had used medication for asthma in the preceding 12 months. Wheezing in the preceding 12 months was reported by 89.8% and chronic dry cough was reported by 65.4%. For 73.9% of cases, asthma symptoms had interfered with daily activities or school attendance in the preceding 12 months. Among cases with spirometry data, the mean FEV 1 percent predicted was 96.7 ± 20.6. Atopy was present in 91.9% of cases. The highest rates of skin test positivity were seen for dust mite (70.3%) and cockroach (43.1%). Although only 5.8% of mothers reported smoking during pregnancy, 32.6% of cases lived with a smoking parent in early childhood (before 2 years of age) and 50.4% were currently exposed to parental smoking.
The minor allele frequency and genotype frequency distributions of the six tagging SNPs are shown in Table 2. The TNF-308 and TNF-238 polymorphisms were relatively rare in our Mexican population (minor allele frequency = 5% for TNF-308 and 4% for TNF-238). The frequency distributions for all mating types for TNF and LTA polymorphisms are presented in Supplemental  Table 1 for all families, Supplemental Table 2 for families with smoking parents, and Supplemental Table 3 for families with nonsmoking parents (Supplemental Material online at http://www.ehponline.org/docs/ 2007/9740/suppl.pdf). Hardy-Weinberg equilibrium (p > 0.1) was confirmed for all six tagging SNPs in the parents. Pairwise linkage disequilibrium coefficients, D, and r 2 ,  between LTA and TNF SNPs, calculated using FBAT (Horvath et al. 2001) are shown in Supplemental Table 4 in the online Supplemental Material. There was moderate linkage disequilibrium between the LTA-379 and LTA NcoI polymorphisms (r 2 = 0.50), and weak linkage disequilibrium between the LTA-379 and TNF-857 polymorphisms (r 2 = 0.39), the TNF-1031 and TNF-238 polymorphisms (r 2 = 0.25), the LTA NcoI and TNF-857 polymorphisms (r 2 = 0.20), and the LTA-379 and TNF-1031 polymorphisms (r 2 = 0.16). The r 2 values for all other SNP pairs were < 0.1. Neither of the two SNPs in LTA was associated with asthma (Table 3). For TNF, carrying at least one copy of the TNF-308A allele was associated with increased asthma risk [relative risk (RR) = 1.54; 95% confidence interval (CI), 1.04-2.28; p = 0.031; false discovery rate = 0.12] relative to homozygotes for the major G allele (Table 3). Only three cases were homozygous for the minor allele for TNF-238 and none for TNF-308, so we were not able to evaluate the relative risks for homozygotes. The TNF-238, TNF-857, and TNF-1031 polymorphisms gave weaker, and more unstable, evidence of association with asthma (Table 3). Results for the six tagging SNPs did not differ appreciably by sex, asthma severity (mild versus moderate to severe), or ozone level (data not shown). We did not observe a clear pattern of associations between the TNF and LTA SNPs and percent predicted FEV 1 (data not shown).
Because of the consistent association between parental smoking and childhood asthma (DHHS 2006) and the involvement of TNF in cigarette smoke-induced inflammation responses (Churg et al. 2003;Park et al. 2003), we examined the association of each LTA and TNF SNP with asthma stratified by exposure to a smoking parent in the home. Among cases without smoking parents in the home, the TNF-308A allele and TNF-238A alleles showed increased risk of asthma (RR = 2.06; 95% CI, 1.19-3.55; p = 0.0097; false discovery rate = 0.04 for TNF-308A; RR = 2.21; 95% CI, 1.14-4.30; p = 0.019; false discovery rate = 0.04 for TNF-238A) ( Table 3). The p-values for interaction with living with smoking parents were 0.09 for TNF-308 and 0.01 for TNF-238. For the other two TNF SNPs, although the results were not statistically significant, the increased risk of asthma for carrying two copies of the minor allele also appeared to be limited to cases with nonsmoking parents (RR = 2.36; 95% CI, 0.82-6.78 for TNF-1031C; RR = 1.32; 95% CI, 0.76-2.29 for TNF-857T). Among cases with smoking parents in the home, none of the six LTA and TNF polymorphisms were associated with asthma (Table 3).
Haplotype analyses results reflected the single SNP findings in magnitude and direction (Table 4). Among all cases, individuals carrying one copy of the ht5 (CGTCAG) haplotype, containing the TNF-308A allele, exhibited an increased risk of asthma of borderline statistical significance (RR = 1.45; 95% CI, 0.97-2.19). Among cases with nonsmoking parents, carrying one copy of the ht6 (CACCGA) haplotype containing the TNF-238A allele or one copy of the ht5 haplotype exhibited an increased risk of asthma (RR = 2.15; 95% CI, 1.21-3.82; p = 0.0082 for ht5; RR = 2.40; 95% CI, 1.18-4.81; p = 0.014 for ht6). The false discovery rate was 0.03 for the ht5 and ht6 findings. Among cases with smoking parents, LTA and TNF haplotypes were not associated with asthma (Table 4).
We also examined the association of LTA and TNF individual SNPs and haplotypes with asthma stratified by exposure to a smoking parent before child turned two because exposure in early childhood has also been consistently associated with childhood asthma risk (DHHS 2006 --a Only three cases were homozygous for the minor A allele for TNF-238 and none for TNF-308, so we did not calculate those relative risks. *p = 0.031; false discovery rate = 0.12. **p = 0.0097; false discovery rate = 0.04; p-value for interaction = 0.09. # p = 0.019; false discovery rate = 0.04; p-value for interaction = 0.01. consistent with findings for stratifying by current exposure to a smoking parent in the home (data not shown). We examined the association between individual LTA and TNF SNPs and the degree of atopy to aeroallergens, assessed by the number of positive skin tests out of a battery of 24 tests. No significant associations were detected with the number of positive skin tests for the six LTA and TNF SNPs (data not shown).

Discussion
In this case-parent triad study in a Mexico City population with high lifetime exposure to ozone, we found that the A allele of the functionally relevant TNF-308 polymorphism was significantly associated with an increased risk of childhood asthma, especially among children with nonsmoking parents. The TNF-238A allele and the haplotypes containing the TNF-308A allele or the TNF-238A allele were associated with an increased childhood asthma risk predominantly in children with nonsmoking parents.
TNF is a potent proinflammation cytokine and has been consistently implicated in asthmatic inflammation and bronchial hyperresponsiveness in a variety of subcellular, in vitro, ex vivo, in vivo, and genetic studies (Thomas 2001). For example, TNF expression is markedly increased in asthmatic airways compared with normal airways (Bradding et al. 1994). TNF appears to have an important amplifying effect on asthmatic inflammation (Babu et al. 2004) and has also been shown to induce airway hyperresponsiveness in rats and humans (Kips et al. 1992;Thomas et al. 1995). Therefore, genetic polymorphisms that affect gene expression or TNF activity in the airways might be expected to influence asthma risk.
The TNF-308 polymorphism has been frequently studied in asthma and atopy association studies because it has direct functional effects on TNF gene regulation. The TNF-308A allele is a much stronger transcriptional activator than the more common G allele (Wilson et al. 1997) and is associated with higher TNF production (Louis et al. 1998). The TNF-308A allele leads to high binding affinity of nuclear factors to the TNF promoter and gives a high level of gene transcription (Kroeger et al. 1997). Thus, observations from functional studies suggest that the TNF-308A allele is of biological significance. Despite the known effects of tobacco smoke on TNF expression and the welldocumented association between parental smoking and childhood asthma, few studies have evaluated whether exposure to a smoking parent modifies effects of genetic variation in TNF on childhood asthma. In our study, the association was greater among children without smoking parents in the home.
We also found an association between the TNF-238A allele and asthma risk among children with nonsmoking parents in the home. The TNF-238 polymorphism was not in linkage disequilibrium with the TNF-308 polymorphism (r 2 < 0.01) in our Mexican population, although they are only 70 bp apart from each other within the class III region of the MHC on chromosome 6p. There is no strong evidence showing that the TNF-238 polymorphism has a direct effect on gene expression, although studies suggest that this region contains a strong repressor site (-280 to -172) (Fong and Mark 1995). However, the TNF-238 polymorphism may be in linkage disequilibrium with a functional polymorphism that impacts TNF production, either within the TNF gene or another gene within the MHC.
Because the etiology of asthma involves numerous environmental triggers, heterogeneous exposure to environmental stimuli among different populations may cause conflicting results across studies. The influence of genotypes on phenotypes may be different and even opposite at different levels of exposure (Martinez 2005). Therefore, incorporating environmental exposures, such as parental smoking, into association studies is important. Parental smoking, a reasonably valid approach to estimate long-term secondhand smoke exposure in infants and children, is one of the most consistent risk factors for childhood asthma (DHHS 2006).
It is not unexpected to find differing associations between genetic polymorphisms and asthma susceptibility based on parental smoking exposure. We previously reported that the protective effect for the NQO1 Ser allele in GSTM1-null children was limited to those with nonsmoking parents (David et al. 2003). A recent genome-wide linkage study found different regions of linkage to childhood asthma by parental smoking exposure (Colilla et al. 2003). Of note, the chromosome 6p region, containing the TNF gene, was more strongly linked to asthma among subjects who did not live with a smoker during infancy compared with those who lived with smokers (Colilla et al. 2003).
A recent study showed that the TNF-308 polymorphism modified the effect of home exposure to smokers on respiratory illnessrelated school absence among children mostly without asthma (Wenten et al. 2005). Another study in the same population showed that the TNF-308A allele was associated with an increased risk of wheezing, especially among children living in low ozone communities (Li YF et al. 2006). In the present study, we found that TNF polymorphisms and haplotypes were associated with childhood asthma susceptibility predominantly among children who did not live with smoking parents.
Because TNF is a candidate gene for ozone-induced airway inflammation and hyperresponsiveness (Kleeberger et al. 1997), we stratified on the ozone level. It is not surprising that we did not observe effect modification by ozone because our population in central Mexico City was exposed to high lifetime levels of ozone compared with U.S. populations. The median level of the annual average of the daily maximum 8-hr averages for our population was 67 ppb, with an interquartile range of only 12 ppb.
A possible explanation for our finding of an association of TNF polymorphisms with asthma predominantly for children without smoking parents is that the combination of exposure to secondhand smoke and ozone, which both increase TNF production, overwhelms the smaller impact of TNF polymorphisms on TNF expression. Indeed, expression levels of TNF are significantly increased in mice and humans exposed to tobacco smoke (Churg et al. 2003;Park et al. 2003). TNF has been identified as a potential candidate gene for ozone susceptibility (Kleeberger et al. 1997;Yang et al. 2005) and ozone exposure can stimulate TNF secretion from lung cells (Arsalane et al. 1995). Tobacco smoke and ozone are both strong oxidants (Tatum and Shapiro 2005) and potent controllers of TNF production (Arsalane et al. 1995;Churg et al. 2003;Park et al. 2003). In combination, these environmental triggers may have synergistic effects and overcome the smaller effects of TNF polymorphisms. Conversely, in the absence of secondhand smoke exposure, the influence of TNF functional polymorphisms on TNF expression may be greater, which could explain our finding of stronger associations at lower levels of exposure. Of note, for CD14, a pattern recognition receptor in the endotoxin-induced immune response, associations between the CD14-159 polymorphism and asthma and atopy differ at high versus low levels of endotoxin exposure (Martinez 2005).
In summary, we found that TNF polymorphisms and haplotypes were associated with childhood asthma susceptibility, especially among children without smoking parents. These results suggest that the effects of genetic variation in TNF may be more apparent at lower levels of exposure to substances such as secondhand smoke, which strongly influence TNF expression.