|Home | About | Journals | Submit | Contact Us | Français|
Cleft lip with or without cleft palate (CLP) and cleft palate only (CPO) have an inherited component and, many studies suggest, a relationship with folate. Attempts to find folate-related genes associated with clefts have, however, often been inconclusive. This study examined four SNPs related to folate metabolism (MTHFR 677 C→T, MTHFR 1298 A→C, MTHFD1 1958 G→A, and TC II 776 C→G) in a large Irish population to clarify their relationship with clefts.
Cases and their parents were recruited from major surgical centers performing cleft repairs in Ireland and a support organization. Data on risk factors, medical history, and DNA were collected. Controls were pregnant women from the greater Dublin area (n = 1,599).
CLP cases numbered 536 and CPO cases 426 after exclusions. CPO mothers were significantly more likely than controls to be MTHFR 677 TT, OR 1.50 (95% CI: 1.05–2.16; p = .03). Log-linear analysis showed a borderline association (p = .07). Isolated CPO case mothers were significantly more likely than controls to be homozygous for the MTHFD1 1958 G→A variant, OR 1.50 (95%CI: 1.08–2.09; p = .02). When multiple cases were added, both CPO cases and case mothers were significantly more likely to be AA (p = .02 and p = .007, respectively). The CLP case-control and mother-control analyses also showed significant effects, ORs 1.38 (95% CI: 1.05–1.82; p = .03) and 1.39 (95% CI: 1.04–1.85; p = .03), respectively.
Associations were found for both CPO and CLP and MTHFD1 1958 G→A in cases and case mothers. MTHFR 677 C→T could be a maternal risk factor for clefts but the association was not strong. Because multiple comparisons were made, these findings require additional investigation. Given the known association between MTHFD1 1958 G→A and NTDs, these findings should be explored in more detail.
There is considerable evidence suggesting that folate-related genes play a role in the etiology of oral facial clefts. Clefts are known to have a strong genetic component. Nonsyndromic clefts are complex traits and it is likely that genetic factors interact with environmental factors. Folate has long been considered one such factor. A number of observational studies (Czeizel et al., 1996, 1999; Itikala et al., 2001; Loffredo et al., 2001; Shaw et al., 1995; Tolarova and Harris, 1995; van Rooij et al., 2004; Boyles et al., 2008), although not all (Hayes et al., 1996; Czeizel, 1993), have reported lower rates of clefts in the offspring of mothers who took folic acid during the periconceptional period. Some studies have found a decline in cleft rates since food fortification with folic acid began (Canfield et al., 2005; Yazdy et al., 2007), although others have not found a change (Ray et al., 2003; Simmons et al., 2004). These observations have led investigators to look for associations between clefts and folate enzyme genes, including methylenetetrahydrofolate reductase (MTHFR) (Jugessur et al., 2003; Gaspar et al., 1999; Blanton et al., 2002; Martinelli et al., 2001; Prescott et al., 2002; Shotelersuk et al., 2003; Shaw et al., 1998; van Rooij et al., 2003; Vieira et al., 2005; Mills et al., 1999) and methylenetetrahydrofolate dehydrogenase—the trifunctional cytoplasmic enzyme methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase/formyltetrahydrofolate synthetase—MTHFD1 (Boyles et al., 2008; Mostowska et al., 2006). Because of the interrelationship between vitamin B12 and folate, the plasma vitamin B12 transporter, transcobalamin 2 (TCII), has also been investigated for a role in clefts (Boyles et al., 2008). The results of these investigations have been mixed: some demonstrating strong associations and others failing to find any significant associations. This lack of consensus may be the result of genetic differences, nutritional factors, or the lack of sufficient numbers of subjects to find an effect.
We sought to clarify the role of common variants in folate genes that are known to be expressed during development using a large, homogeneous population from an area, Ireland, where food fortification is not mandated, and folic acid supplement use is uncommon.
Subjects with cleft palate only (CPO) or cleft lip with or without cleft palate (CLP) along with their mothers and fathers were identified by several mechanisms. Samples were collected at the Dublin Cleft Centre in Ireland, which consists of two units with a single database and common protocols between two surgeons and the cleft teams. The majority (80%) of Irish patients with cleft lip and palate are treated through the Centre. Subjects were also recruited from the membership of the Cleft Lip and Palate Association of Ireland as previously described (Mills et al., 1999). Cleft subjects or their parents were either interviewed or completed the interview questionnaire themselves and, when necessary, had an interviewer review their answers with them in order to collect data on socio-demographic variables, obstetrical history, family history, and relevant data on the affected pregnancy including medication use, folic acid exposure, alcohol, and smoking. Recruitment was conducted in two phases with a more detailed questionnaire being used in the second phase (985 families) than the first phase (103 families). Medical information was provided by the attending surgeons for 87% of the cleft cases. Written informed consent was obtained from all participants. Ethical approval was granted by the Research Ethics Committees of the Health Research Board of Ireland, the participating hospitals, and the Institutional Review Board at NIH.
Between 1986 and 1990, blood samples were obtained from a population of 56,049 pregnant women attending the three main maternity hospitals in the Dublin area. Controls for this study were a random sample (n = 1,000), selected after all mothers who had given birth to a child with a known malformation, including NTD, orofacial cleft, congenital heart defect, Down syndrome, or limb deformity had been removed. Mothers with a previous history of an NTD-affected pregnancy were also excluded. Additional controls (n = 599) selected from the same pool of 56,049 mothers for previous work (Parle-McDermott et al., 2006) were used to augment the 1,000 random controls for this analysis. Subjects gave consent for the samples.
Whole blood or buccal swab samples were used for the extraction of genomic DNA using a QIAamp DNA Blood Mini Kit (Qiagen, UK). For all polymorphisms, genotyping quality was tested by repeat genotyping approximately 10% of the samples using the initial genotyping method. In addition, the assays had been previously verified by replicate genotyping approximately 10% of samples using an independent assay. All genotypes were independently double scored with repeat genotyping reaching 95% agreement with the original genotyping results in order to achieve a high level of quality control. The MTHFD1 1958G→A, R653Q, (rs2236225) polymorphism assay was performed by restriction fragment length polymorphism (PCR-RFLP) using Msp I as previously described (Brody et al., 2002). The MTHFR 677C→T, A222V (rs1801133) polymorphism was performed by PCR-RFLP using Hinf I as previously described (Frosst et al., 1995). The MTHFR1298A→C A429E (rs1801131) polymorphism was PCR amplified as described in van der Put et al. (1998) and genotyping was carried out via allele specific oligonucleotide analysis as described in Parle-McDermott et al. (2003). The TCII 776C→G, P259R, (rs1801198) polymorphism was genotyped using an allele-specific primer extension assay and scored by matrix-assisted laser desorption/ionization-time of flight MALDI-TOF mass spectrometry (Sequenom, San Diego, CA). The success rates and agreement rates for repeat genotyping were, respectively, for MTHFD1 1958G→A, R653Q 99.2% and 98.4%, for MTHFR 677C→T, A222V 99.9% and 95.1%, for MTHFR1298A→C, A429E 99.3% and 95.6% and for TCII 776C→G, P259R 93.5% and 94.9%. Not all subjects were genotyped for all SNPs. At the start of the study it was decided that when it was clear after the first group of subjects was genotyped that the results would remain negative even if all the remaining samples were genotyped, no further analysis would be performed for cost and logistic reasons. In some instances sample was not available to perform all SNP analyses.
Tests for Hardy-Weinberg equilibrium (HWE) were conducted within each subject class (case, control, mother, and father) for each SNP by chi-squared test for genotype proportions.
Cases were analyzed as CLP or CPO. Initial analysis was restricted to isolated defects only, with additional analysis on the full group of isolated or multiple defects. Each case-control group was tested for association with each SNP using the two genotype groups corresponding to either a dominant or recessive disease model using Fisher’s exact test for 2 × 2 tables. The strength of association was characterized by the odds ratio (OR) estimates from logistic regression. Similarly, mother-control associations were tested for each SNP using the same methodology.
Triads (case, mother, and father) were used to test for genotype relative risk using two log-linear models. The first uses a single parameter for the number of risk alleles in the case genotype, giving a 1-df test of association equivalent to the transmission disequilibrium test (TDT) of Spielman et al. (1993). This model also uses a single parameter for maternal genotype, yielding an analogous test of association. The second log-linear model uses two parameters each for the case and maternal genotypes, allowing a separate effect for one or two copies of the risk allele (Weinberg et al., 1998). This last model was also fit using the EM algorithm to incorporate the incomplete families. Finally, to address the possibility of an interaction effect of gene and exposure, a separate log-linear model was fit for each of the following exposures: alcohol use, tobacco use, family history of clefts, and folic acid supplementation (broken into five time periods; one before and four after conception). Each log-linear model had separate terms for case genotype with and without the exposure. Tests for interaction were based on the 4-df Wald test for case genotype*exposure.
There were 1,088 cases recruited for this study. Five were excluded because the mother had diabetes, 12 because the mother had epilepsy, and 27 because the mother was exposed to potentially teratogenic drugs. In addition, we dropped nine cases with chromosomal abnormalities, 49 with other syndromes, three because the type of cleft could not be classified, and 21 who were not genotyped. Of the remaining cases, 536 were CLP. This group consisted of 494 cases with isolated defects, 23 with one additional defect, 18 with multiple defects, and one with Pierre Robin. There were 426 cases with CPO. This group consisted of 321 cases with isolated defects, 15 with one additional defect, 21 with multiple defects, and 69 with Pierre Robin Sequence. There were 783 complete triads (case, mother, and father), 134 case-mother pairs, eight case-father pairs, 36 cases without parents, and one mother without a case.
Study subjects’ characteristics are shown in Table 1. More than a third (38%) of the mothers smoked during the affected pregnancy and 55% drank some alcohol. Most women had previous pregnancies; 3% reported previous stillbirths and over 30% had at least one prior pregnancy loss. There was at least one affected sibling in 5% of the families. Family history was positive in 16% of the fathers and 20% of the mothers.
In the primary analysis only isolated cases of CLP and CPO and their parents were included. In the secondary analysis, nonsyndromic cases with other defects were included to see if the results were changed. ORs are reported for the recessive model except as noted. Genotype distributions in all groups were in HWE except as noted below. The genotyping results are shown in Table 2.
It has been noted previously by us that the control subjects were not in HWE for MTHFD1 1958G→A, R653Q, and the possible biological reasons have been discussed (Brody et al., 2002).
There was no significant association between case status and the Q allele (OR 1.31; CI: 0.94–1.82; p = .13). The TDT and log linear analyses did not show any association between the Q allele and case status. Mothers were significantly more likely to be QQ compared to controls (OR 1.50; CI: 1.08–2.09; p = .02) but not in the log-linear analysis (p > .05). Adding the multiple case families changed the results for the case analysis. The case-control OR became significantly positive (OR 1.41; CI: 1.05–1.90; p = .02), but the log-linear analysis did not become significantly positive. The statistical findings in the mother-control analysis became more positive (OR 1.51; CI: 1.13–2.03; p = .007). The log-linear analysis remained negative.
Both the case-control and mother-control analyses showed a significant effect, ORs 1.38 (CI: 1.05–1.82; p = .03) and 1.39 (CI: 1.04–1.85; p = .03), respectively. TDT testing did not show increased transmission of the Q allele (p = .16). There was borderline evidence (p = .09) for a maternal association in the log-linear analysis but not a case effect. When the multiple case families were added to the analysis, there were minor changes. The case-control effect became slightly more pronounced (p = .02) while the mother-control finding became slightly weaker (p = .05). The overall maternal effect in the log-linear analysis remained the same, but the log-linear examination of the effect of one maternal copy of the variant allele versus none became significantly positive (p = .03). It is worth noting that the term for two copies of the variant allele versus zero was not significantly positive.
CPO cases did not show a significant association with the MTHFR 677C→T, A222V variant. The OR was 0.96 (CI: 0.64–1.43; p = .92). The TDT and log-linear analyses showed no association. On the other hand, there was evidence that the variant was a maternal risk factor for CPO. Mothers were significantly more likely than controls to be TT (OR 1.50; CI: 1.05–2.16; p = .03). Log-linear analysis showed a borderline association in mothers (p = .07).
When multiple cases were included in the analysis there was no change in the findings regarding cases. The effect in mothers was less strong; the OR was 1.37 (CI: 0.98–1.90; p = .07). No significant effect was found in the log-linear analysis (p = .15).
The 677C→T SNP was not associated with being a CLP case in the isolated case-control analysis (OR 1.08; CI: 0.78–1.50), nor was it preferentially transmitted (p > .05). The 677C→T SNP was also not more common in mothers in the mother-control analysis (OR 1.00; CI: 0.71–1.41) or in the log-linear analysis. When the multiple defect families were added to the analysis, the results were not substantially changed.
MTHFR 1298A→C (A429E) is well known to be in linkage disequilibrium with MTHFR 677 C→T. Mothers in the CPO group were not in HWE.
The OR in the isolated case-control analysis was 0.83 (CI: 0.49–1.40; p = .53). The TDT analysis showed no association between CPO and the C allele (p > .05). Log-linear analysis showed no case effect. Both the mother-control and log-linear analyses showed no significant association. The combined analysis of isolated and multiple cases also showed no significant associations between the C allele and case or mother status.
As was the case with CPO, there was no association between the C allele and CLP. In the isolated case families, the case-control analysis, mother-control analysis, TDT, and log-linear analyses revealed no significant associations (all p > .05). Adding the multiple group did not change the results substantially.
There was no association between having the 776C→G (P259R) variant of the TC II gene and CPO in isolated cases. The ORs for being either a case (1.05; CI: 0.73–1.52) or a case mother (1.00; CI: 0.68–1.46) were not statistically significant. Neither log-linear analyses for maternal or case associations, nor TDT testing, showed any association between CPO and the G allele (all p > .05). The analysis was repeated adding multiple cleft cases; all p values remained nonsignificant.
In the CLP cases (OR 1.05; CI: 0.77–1.43) and mothers (OR 1.10; CI 0.81–1.50), the findings were similar. There were no significant associations between the 776C→G SNP and CLP in the case-control, mother-control, log-linear, or TDT comparisons (all p > .05). Adding multiple CLP cases did not alter the above results. However, the mother-control comparison did show a significant (p = .04) effect (OR 1.35; CI: 1.02–1.77) when a dominant gene model was used. The log-linear analysis also produced a significant (p = .03) maternal effect for one copy of the gene. It should be noted that the dominant model using only isolated cases (p > .05), the log-linear analysis testing the effect of two copies of the gene (p > .05), and the recessive model reported above did not show a significant effect.
Next, all the SNP analyses were tested for interactions with environmental factors. Alcohol and tobacco use, positive family history, and folic acid use preconception and in each of the first 4 months of pregnancy were considered. There were no significant interactions with any of these factors in any of the SNPs tested. Adding incomplete triads to the analyses did not change any of the results substantially. We examined the CLP and CPO cases that had one or more additional defects to determine whether any patterns suggestive of syndromes were present. None were found. The other defects are shown in Table 3.
We examined four SNPs from genes related to folate directly or indirectly via vitamin B12. MTHFR 677 C→T and 1298 A→C have been studied extensively, but with inconsistent results. MTHFD1 1958 G→A and TC II 776C→G have only been examined in a few studies (Boyles et al., 2008), also with conflicting results.
The most interesting results appeared in our investigation of MTHFD1 1958 G→A. As has been reported by us previously (Brody et al., 2002), the control population was not in HWE. This finding complicates the interpretation of the case-control and mother-control analyses. We found that mothers of isolated (p = .02) and multiple defect CPO cases were significantly (p = .007) more likely to be homozygous for the A (Q) variant than controls, although A was not a significant risk factor in the log-linear analysis. In the case-control analysis, adding multiple cases increased the case control OR from 1.31 (p = .13) to 1.50 and produced a statistically significant result (p = .02). As in the mothers, the log-linear analysis was not significant. In the CLP group, both cases and mothers showed a significant association with the variant allele (both p = .03). The log-linear analysis showed at most a borderline association (p = .09) in the mothers and no significance in the cases. In contrast to our findings, the other studies that looked at cases (Boyles et al., 2008; Palmieri et al., 2008) found no effect, as did the two studies that looked for an effect in mothers (Boyles et al., 2008; Mostowska et al., 2006).
Interestingly, MTHFD1 has been shown by our group to be a maternal, but not an embryonic, risk factor for NTDs in two studies of Irish subjects (Brody et al., 2002; Parle-McDermott et al., 2006). MTHFD1 is a logical candidate gene to explore for possible associations with birth defects. It catalyzes the conversion of tetrahydrofolate to the corresponding 10 formyl, 5,10 methenyl, and 5,10 methylene derivatives. 10-Formyltetrahydrofolate and 5,10 methenyltetrahydrofolate act as carbon donors for the de novo synthesis of purines and pyrimidines and hence are required for DNA synthesis. Insufficient DNA synthesis by the mother could damage the developing embryo.
In contrast to our smaller, earlier study, we found that the MTHFR 677C→T variant was not a risk factor in CPO or CLP cases in any of our analyses. The same was true in mothers of CLP cases. In mothers of isolated CPO cases, however, the MTHFR 677 C→T variant was significantly more common in the mother-control analysis and a borderline risk factor (p = .07) in the log-linear analysis. Thus, the MTFHR 677 C→T variant in mothers may play a role in CPO in the Irish population, although these findings could be due to chance.
Results of previous investigations have been mixed. Some studies of MTHFR 677 C→T have shown no association with CLP in either cases or mothers while others have reported associations in mothers, or in mothers who did not use folic acid periconceptionally (Jugessur et al., 2003; Gaspar et al., 1999; Blanton et al., 2002; Martinelli et al., 2001; Prescott et al., 2002; Shotelersuk et al., 2003; Shaw et al., 1998; van Rooij et al., 2003; Mills et al., 1999). One recent study (Chevrier et al., 2007) found that the TT genotype was significantly protective in their case control analysis, although the authors noted that the sample size was small and that the TT rate in controls was higher than expected. CPO has not been studied as extensively. Shaw et al. (1999) found no association between MTHFR 677 C→T and CPO. Jugessur et al. (2003) found no association in mothers but an increased risk when the mother was heterozygous CT.
The MTHFR 1298 A→C variant was not found to be a risk factor for CPO or CLP in any of our analyses. Neither mothers nor cases showed any associations in the case-control, mother-control, log-linear, or TDT analyses. These results are consistent with the findings of a recent meta-analysis (Verkleij-Hagoort et al., 2007) that showed ORs close to 1 for CLP in both mothers and cases. CPO was not included in the meta-analysis. The only previous study (Jugessur et al., 2003) of 1298 A→C in CPO found, as we did, no association in either cases or mothers.
Our investigation of the role of the B12 transporter gene variant 776 C→G showed no association with either CPO or CLP. Only two other studies have examined this variant. Boyles et al. (2008) found no association except in the parent of origin test, and this was based on only 10 cases in which the allele was inherited from the father and six in which it was inherited from the mother. Martinelli et al. (2006) found a positive association with CLP. There are several possible reasons for these discordant results. Their subjects came from genetically different populations. Martinelli et al. (2006) had a higher proportion of familial cases (87/218) and chance may have played a role; the authors noted that their findings required confirmation. Thus, our data and data from the other large study, Boyles et al. (2008), do not suggest that TC II 776C→G is an important risk factor in CLP, but additional investigation would help to clarify the situation.
Our study has numerous strengths. It included a large number of triads and controls. The Irish population is genetically homogeneous, which reduces concerns that stratification could have occurred and increases our power. We were able to document other potential effect modifiers and confounders including smoking, alcohol, and folic acid use. The study has some limitations as well. Folate levels during the pregnancy of interest could not be measured. Our study included many tests of genotype-phenotype association so that our statistically significant findings could have occurred by chance. Syndromic cases were identified by the attending surgeons and excluded, but clinical geneticists did not examine the cases. We did, however, record the other defects seen in cases with multiple defects and there were no patterns of defects suggestive of syndromes present.
In summary, ours is the first study to find that the MTHFD1 1958G→A variant is a maternal risk factor for CLP. MTHFR 677 C→T was not a risk factor in cases; however, there was modest evidence that it was a maternal risk factor in CPO by both mother-control and log-linear analyses. In light of the fact that multiple comparisons were made, the positive findings require additional investigation. Nonetheless, our findings suggest that the MTHFD1 1958 G→A variant may be important in the etiology of CPO and CLP and this variant merits additional attention.
This work was supported by the Intramural Research Programs of NICHD and NHGRI.
The authors gratefully acknowledge the patients and families who participated in the study, the Cleft Lip and Palate Association of Ireland, and the Dublin Cleft Center team.
The data in this manuscript have not been presented at any meetings.
†This article is a US Government work and, as such, is in the public domain in the United States of America.