A hospital-based case-control study was conducted to identify interactions between the 538(T→C) polymorphic site of bone morphogenetic protein 4 gene ( BMP4T538C ) and exposures in pregnancy with nonsyndromic cleft lip, with or without cleft palate (nsCL/P). Associations between offspring polymorphism of BMP4T538C , paternal smoking, paternal high-risk drinking, maternal passive smoking, and maternal multivitamin supplement with nsCL/P were analyzed by logistic regression analysis. BMP4T538C polymorphism, maternal passive smoking exposures and maternal multivitamin use were associated with the risk of nsCL/P but paternal smoking and paternal high-risk drinking were not. Gene–environment interactions were analyzed using the multifactor dimensionality reduction (MDR) method. The two-factor model including maternal passive smoking and BMP4T538C, was the best for predicting nsCL/P risk with a maximum cross-validation consistency (10/10) and a maximum average testing accuracy(0.605; P < 0.0001). The findings suggested that: BMP4T538C could be used as a genetic susceptibility marker for nsCL/P; maternal passive smoking exposure is a risk factor for nsCL/P; maternal multivitamin supplements are a protective factor; the synergistic effect of BMP4T538C and maternal passive smoking could provide a new tool for identifying individuals at high risk of nsCL/P, and provides additional evidence that nsCL/P is determined by genetic and environmental factors.
Nonsyndromic cleft lip with or without cleft palate (nsCL/P) is a common malformation that affects approximately one in 600 newborn babies . Its etiology is complex and incompletely understood, but is generally considered to be determined by environmental and genetic factors . Epidemiological studies have suggested that a proportion of nsCL/P may be associated with maternal exposures during the critical early period of pregnancy; these exposures include prescription drugs , cigarette smoking , alcohol , and deficiency of nutritional folic acid . Many linkage and association studies have been carried out for nsCL/P and some potential candidate loci and candidate genes have been identified, including (but not limited to) MSX1 , TGFA , TGFβ1 , TGFβ2 , TGFβ3 , RARA , and MTHFR . The emphasis has been on determining the role of gene–environment interactions in the development of nsCL/P, but most research has involved the independent effects of different factors. A few gene–environment interactions have been reported usually with inconsistent findings. These inconsistencies may be attributable to population differences, confounding, low power, sampling variation, or differences in the approaches used to assess gene–environment interactions.
A genome scan meta-analysis for nsCL/P and multipoint parametric heterogeneity LOD (a log odds ratio) score identified significant linkage to the chromosomal region 14q21-24. Bone morphogenetic protein 4 ( BMP4 ), located to 14q22–23 , is expressed in the maxillary prominence , one of the regions of pre-fusion in development of the face in embryos, which forms the lateral parts of the upper lip and the secondary palate . A CL/P phenotype was presented in BMP4 knockout mice . On the basis of these study results, the only non-synonymous single-nucleotide polymorphism (SNP) of BMP4 , T538C ( rs17563 ), in the SNP database was detected in this research. The interaction effects between BMP4T538C and some common environmental exposures, including exposure to parents’ cigarette smoke and alcohol intake and mother’s multivitamin supplement, on the development of nsCL/P were analyzed in a Han Chinese population using the multifactor dimensionality reduction (MDR) method.
Materials and methods
The study population included 200 nonsyndromic CL/P patients, aged between 6 months and 15 years (mean age: 4.9 ± 6.23 years), and 200 phenotypically normal individuals as the controls. The subjects comprised 172 cases and 184 controls, whose genotypes had been previously reported and 28 cases and 16 controls who were recruited later. All subjects were genetically unrelated ethnic Han Chinese. The selection criterion for patients was that they had to have CL/P with no other malformations. Individuals with any other anomalies, such as neural tube defects, monogenic traits (e.g. Van der Woude’s), syndromes (e.g. trisomy), and sequences (e.g. Pierre Robin) were excluded. All the cases included in the study were diagnosed by clinical examination by well-trained dentists and oral surgeons and were evaluated by at least one dysmorphologist who confirmed their nonsyndromic status. Members of the control group were made up of outpatients with trauma and inpatients with bone fracture treated in the same hospital. Criteria for entry to the control group included no evidence of any other serious illness, especially hereditary diseases, and no family history of CL/P. The control group was matched to the case group by gender and age and socio-economic status.
Mothers of case and control patients were interviewed in the hospital with a questionnaire after informed consent was obtained. The information obtained included demographics (parents’ age, schooling, socio-economic status, children’s age, gender), family history of oral cleft, parents’ tobacco and alcohol consumption, mother’s multivitamin supplement intake and passive smoking.
For smoking, a woman was asked if she and her child’s natural father had smoked during the 3 months before conception, and also if they had smoked in the first trimester of pregnancy. The average number of cigarettes smoked per day was determined if smoking was reported. To assess passive smoke exposure, the mother was asked whether anyone smoked inside her home or near her at work during the same period.
Reports of paternal and maternal alcohol consumption were collected for the 3-month period before conception and the first trimester of pregnancy. When drinking was reported, the number of times they drank per month, the typical number of alcohol units they drank per occasion, and the kinds of drinks (e.g. beer, wine) were ascertained.
Maternal multivitamin supplementation was reported separately for the 3-month period before conception and the first trimester of pregnancy. Mothers who took supplementary vitamins were asked how many vitamins they took a week and for how long.
At the end of the interview, an approximately 5-ml blood sample was collected from each participant for individual genotype analysis.
DNA extraction and genotyping
Genomic DNA was extracted from ethylenediamine tetraacetic acid (EDTA)-anticoagulated peripheral blood leukocytes by the salting-out method .
The BMP4T538C ( rs17563 ) genotype was determined by a polymerase chain reaction-restriction fragment length polymerase (PCR-RFLP) assay. The PCR primers were designed based on the Gen-Bank reference sequence . Primer sequences for PCR reactions were 5’-CCTAACTGTGCCTAG-3’ (forward) and 5’-CATAACCTCATAAATGTTTATACGG-3’ (reverse). The PCR reactions were performed in a total volume of 25 μl containing 100 ng genomic DNA, 20 pM of each primer, 0.2 mmol/l dNTPs, 20 mmol/l Tris-HCl (pH 8.8), 10 mmol/l KCl, 10 mmol/l (NH 4 ) 2 SO 4 , 2 mmol/l MgSO 4 , 0.1% Triton X-100, and 1 unit of Taq polymerase(New England BioLabs Ipswich, MA). The PCR cycle conditions consisted of an initial denaturation step at 94 °C for 12 min followed by 35 cycles of 1 min at 94 °C; 1 min at 58 °C; 1 min at 72 °C; and a final elongation at 72 °C for 7 min. The PCR products were digested for 6 h at 37 °C with HphI (New England BioLabs). The digested PCR products were resolved on 8% polyacrylamide gels and visualized by a silver staining protocol. The accuracy of the screening method was confirmed by direct sequencing of amplified DNA from randomly selected samples (5%), and no difference in results was found between the two methods (data not shown).
Using the SPSS 11.5 statistical software package, the Pearson χ 2 test was used to test for differences between the case and controls subjects for the categorical variables of gender, socio-economic status, paternal schooling and maternal schooling. The Wilcoxon rank sum test was used to test the difference of child’s age, paternal age and maternal age between groups. Unconditional logistic regression analysis was conducted to assess the independent main effects of the SNP of BMP4T538C and environmental exposures, including paternal smoking, paternal alcohol consumption, multivitamin supplement and passive smoking. Maternal smoking and maternal alcohol consumption were excluded from the analysis because none of the mothers reported smoking or drinking.
Potential interactions between BMP4T538C and environmental exposures were assessed using the open-source MDR software package available from. www.epistasis.org/ .
MDR is a nonparametric and genetic model-free method for detecting and characterizing high-order gene–gene interactions or gene–environment interactions in case-control and discordant-sib-pair studies with relatively small samples . MDR defines a single variable that incorporates information from several loci and/or environmental factors that can be divided into high risk and low risk combinations. This new variable can be evaluated for its ability to classify and predict risk using cross-validation and permutation tests. To search for the best n-factor ( N = 1, 2, 3, 4, 5) model, all the data were divided into a training set including 9/10 of the data and a testing set including 1/10 of the data. Training set is used to search for the best model associated with the high-risk pattern depending on the ratio of cases to controls and the testing set is used to control the goodness-of-fit of the model. This complete test was repeated 10 times using different random seeds to reduce the probability of biased results due to the chance division of the data in training and testing sets. The authors selected the best MDR model as the one with the maximum testing accuracy and a high cross-validation consistency. Statistical significance was evaluated by comparing the average testing accuracy from the observed data with the distribution of average testing accuracies under the null hypothesis of no association derived empirically from 1000 permutations. MDR results were considered statistically significant at the 0.05 level.
In all analyses, environmental exposures of interest were considered as dichotomous variables. The strategies used to dichotomize these variables are described below.
Regarding paternal smoking, it is hard to classify fathers who stopped smoking “when their wives became pregnant”, because their own recognition of pregnancy may or may not have preceded the crucial time of lip and/or palate formation. The authors decided to define a smoker as a man who smoked at least one cigarette per day in either the 3-month before conception or the first trimester of pregnancy. An individual who had never smoked or occasionally smoked was considered a nonsmoker.
Similarly, paternal alcohol consumption was categorized as yes/no according to whether there was high-risk drinking in the 3 months before conception or the first trimester of pregnancy. High-risk drinking was defined as drinking on average 40 g/day or more of pure alcohol. The level of drinking was estimated on the basis of the quantity, frequency of consumption, and alcoholic content of different kinds of drinks (wine, beer, liquor).
If the mother reported using multivitamins in the 3 months before conception or the first trimester of pregnancy, she was classified as a multivitamin user. A mother was considered a passive smoker if she reported someone smoking inside her home or near her at work and the exposure lasted for at least 15 min/day for more than 1 day per week.
Characteristics of the study population
The characteristics of the study groups are presented in Table 1 . There were no significant differences between the cases and controls for child’s age, child’s gender, and socio-economic status, which suggested that the matching based on these variables was adequate. Maternal age and paternal age were similar between the two groups. Educational level was lower for the mothers and fathers of cases compared with those of controls. The 200 cases consisted of 141 (71%) with cleft lip only (CLO) and 59 (30%) with cleft lip with cleft palate (CLP). 3 (2%) cases reported a family history of oral clefts, while none of controls reported that.
|Variable||nsCL/P cases (n = 200)||Controls (n = 200)||P|
|Child’s age (mean ± SD, years)||4.9 ± 6.23||5.1 ± 2.92||0.201|
|Maternal age (mean ± SD, years)||24.5 ± 4.76||25.6 ± 2.62||0.335|
|Primary school or less||59(29.5)||46(23.0)|
|College or higher||17(8.5)||49(24.5)|
|Paternal age (mean ± SD, years)||26.7 ± 3.87||28.2 ± 2.09||0.212|
|Primary school or less||44(22.0)||31 (15.5)|
|College or higher||19(9.5)||70(35.0)|
|Socio-economic status a||0.190|
Main effects of gene and environmental exposures
The authors estimated the relative risk of genotypes of BMP4T538C , paternal smoking, paternal high-risk drinking, maternal multivitamin supplement and maternal passive smoking on nsCL/P occurrence using odds ratios ( ORs ) and 95% confidence intervals ( CIs ). Maternal schooling and paternal schooling were considered as potential confounding factors when testing for association between nsCL/P and exposures. Adjusted ORs and 95% CIs were computed using logistic regression. The results are shown in Table 2 . The results of genotypes are in agreement with the preceding report . Homozygosity for the BMP4T538C polymorphism ( CC ) resulted in a significantly raised odds ratio for the CC genotype compared with the TT genotype ( OR = 2.44; 95%CI : 1.23, 4.90). The authors did not find an increased risk of nsCL/P for the heterozygous genotype compared with the TT genotype. The frequency of the 538C allele was 41% (162/400) in nsCL/P and 32% (128/400) in controls, and the 538C allele showed a significantly elevated risk ( χ 2 = 6.25, P < 0.05 ), compared with the 538T allele. The current study shows higher frequencies of paternal smoking and paternal high-risk drinking among nsCL/P cases compared with controls (53% vs. 46% and 29% vs. 19%, respectively). The risk for nsCL/P was not materially increased owing to their 95%CIs including 1.0. Maternal multivitamin supplement was associated with a significantly decreased risk of nsCL/P ( OR = 0.59; 95%CI : 0.38, 0.92), while maternal passive smoking was associated with an increased risk of nsCL/P ( OR = 1.72; 95%CI : 1.08, 2.74).
|No. of cases||No. of controls||Odds ratios a||95% confidence intervals|
|Genotypes of BMP4T538C|
|Paternal high-risk drinking|
|Maternal multivitamin supplement|
|Maternal passive smoking|
Potential interactions between gene and environmental exposures
MDR software was used to identify which combination of factors is the best model to predict the high-risk subgroup of subjects. Table 3 summarizes the results of MDR analysis for all number of factors evaluated ( N = 1, 2, 3, 4, 5). The best model for each order is shown along with its cross-validation consistency, average testing accuracy and significance level as determined by permutation testing. The model including only maternal passive smoking was the best single factor model for predicting nsCL/P risk, with a cross-validation consistency of 10/10 and an average testing accuracy of 0.585. Among two-factor models, the combination of maternal passive smoking and BMP4T538C proved to be the strongest, with a cross-validation consistency of 10/10 and an average testing accuracy of 0.605. When three factors were considered, the combination of maternal multivitamin supplement, maternal passive smoking, and BMP4T538C was the strongest model, with a cross-validation consistency of 8/10 and an average testing accuracy of 0.585. A model containing maternal multivitamin supplement, maternal passive smoking, paternal smoking and BMP4T538C was the strongest of all the four-factor models, with a cross-validation consistency of 7/10 and an average testing accuracy of 0.591. When all five factors were considered, the cross-validation consistency and the average testing accuracy of the model were, respectively, 10/10 and 0.599. Among all models, the two-factor model, including maternal passive smoking and BMP4T538C, had a maximum cross-validation consistency and a maximum average testing accuracy (P < 0.0001). The authors chose the two-factor model as the best model.
|Best model||Cross-validation consistency||Average testing accuracy||Permutation test P value|
|Maternal passive smoking||10/10||0.585||0.0007|
|Maternal passive smoking||10/10||0.605||<0.0001 a|
|Maternal multivitamin supplement||8/10||0.585||<0.0001|
|Maternal passive smoking|
|Maternal multivitamin supplement||7/10||0.591||<0.0001|
|Maternal passive smoking|
|Maternal multivitamin supplement||10/10||0.599||<0.0001|
|Maternal passive smoking|
|Paternal high-risk drinking|