Journal Publication

Exploring gene-gene interactions in the etiology of neural tube defects

by Lorenzo D. Botto and Pierpaolo Mastroiacovo

	Abstract
	Introduction
	Methods
	Results
	Discussion
	References
	Tables

Abstract

The role of susceptibility genes in the etiology of birth defects is unclear, but may involve in some cases multiple alleles at multiple loci. We suggest a simple epidemiologic approach to explore gene-gene interactions, and use it reevaluate data from a recent case-control study on the possible association of neural tube defects (NTDs) with specific mutations of two genes, 5,10 methylene-tetrahydrofolate reductase (MTHFR) and cystathionine-beta synthase (CBS). We found that, compared to the common genotype, homozygosity for the MTHFR mutation alone was associated with a twofold increased risk for NTDs, while homozygosity for the CBS mutation alone was not a risk factor. However, individuals homozygous for the mutations at both loci had a fivefold greater risk for NTDs than those with the reference genotype. Though the original study was too small to detect statistically significant differences among most of the risk estimates, these results, if confirmed by independent and larger studies, suggest that gene-gene interaction may play a role in modulating the susceptibility to NTDs in a proportion of affected individuals. This approach, moreover, could be a valuable adjunct to the study of gene-gene interactions in the etiology of human disease.

Introduction

The role of genes in the etiology of neural tube defects (NTDs) is unclear. Because of the protective role of folic acid, recent genetic research has focused on genes that encode enzymes involved in folate-related metabolism, including 5,10 methylene-tetrahydrofolate reductase (MTHFR) and cystathionine-beta synthase (CBS). Specific polymorphisms for both genes have been identified, including an MTHFR allele with a C to T transition in position 677 (MTHFR-T)(Kang et al., 1991) and a CBS allele with a 68 base-pair insertion (CBS-I)(Sebastio et al., 1995). Both alleles could contribute to folate-related NTD risk as they both may be associated with abnormal metabolism of folate and homocysteine: the MTHFR-T allele encodes for a heat-labile protein with reduced enzymatic activity, and the CBS-I allele has been described in homocystinuria patients (Sebastio et al., 1995). Recently, homozygosity for MTHFR-T has been shown to be a risk factor for NTDs (Ou et al., 1996; van der Put et al., 1995; Whitehead et al., 1995).

Disentangling the contribution of multiple alleles at multiple loci, however, may not be straightforward, even in the relatively simple case of two alleles at two loci.

The challenges of doing so can be illustrated by a recently published paper (Ramsbottom et al., 1997) in which the authors looked at the relationship between specific MTHFR and CBS polymorphisms and NTD risk in an Irish population.

In that study, the authors conclude that "when this [insertion/I128T] CBS haplotype was analyzed in combination with the MTHFR TT genotype, it did not confer an increased risk of NTD This study therefore suggests that [...] deficient CBS function does not contribute significantly to the aetiology of NTDs. Investigations into genetic factors involved in NTD should therefore focus on other enzymes involved in [homocysteine] metabolism" (p.42). However, the results of a reanalysis of their data suggest that a further look to the combined effect of CBS and MTHFR mutations is probably warranted.

Methods

To explore the possible interaction between the two genes, we applied to the data a simple model that can be used in the context of both case-control and cohort studies (Table 1). For this analysis, we considered only two genes, and each genotype was considered dichotomous. Thus, the layout turns out to be a two-by-four table. Though simplistic, this model illustrates the general approach that has been first suggested for the study of gene-environment interactions (Khoury et al., 1988).

From this layout, estimates of relative risks and odds ratios can be derived for each genotype (Table 1). These estimates may be adjusted for potential confounders using standard analytic techniques such as logistic regression. These estimates can also be used to address the issue of the overall contribution of the genotypes to a disease among the group of cases (attributable fraction among cases, AFC) or in a particular population (population attributable fraction, AF). Measures of attributable fraction can be computed in several ways, depending on the type of epidemiologic study (Khoury et al., 1993). In the context of the case-control study that we reanalyzed, we computed the population attributable fraction as (Miettinen, 1974)

AF= f_c (R-1)/R

where f_c is the fraction of cases with the genotype(s) under study and R is the odds ratio (as the estimate of relative risk of disease). The attributable fraction among cases was computed as AFC= (R-1)/R.

Using the paper by Ramsbottom et al. as the source of the information, we rearranged their data, to the best of our abilities, following the layout in Table 1, so that we could compare NTD risk of people with different combinations of alleles at the two loci with that of people with a reference genotype (here, the common genotype in that population). For table 2 we computed exact (Fisher) confidence intervals of the odds ratios.

The identification and measurement of interaction in the context of epidemiologic studies has been matter of discussion for years (see Yang and Khoury, 1997 for a review of the concept of interaction with a focus on gene-environment interaction). The discussion largely centers on how to assess the relationship between the observed joint effect of multiple risk factors (acting through a common pathway) and the effect of each of the risk factors taken alone. Two major models of interaction, additive and multiplicative, have been suggested. To test for the presence of interaction, we used log-linear models where the relationship of the relative risks among the strata was assessed on both a multiplicative (Breslow and Day, 1980) and an additive scale (Rothman et al., 1980; Hosmer and Lemeshow, 1992; Assmann et al., 1996). Specifically, we were able to formally test, for instance, whether the odds ratio associated with coexisting mutations at both loci was significantly different from that associated with mutations at each locus alone, on an additive or a multiplicative scale.

Results

We found that homozygosity for both MTHFR-T and CBS-I is associated with a fivefold greater risk for NTDs compared to the reference genotype (Table 2). MTHFR-T homozygosity alone is associated with a twofold increased risk and that CBS-I alone apparently is not an independent risk factor (Table 2).

Discussion

The results of our reevaluation of a recent case-control study of NTDs (Ramsbottom et al., 1997) suggest that coexisting mutations at two loci might influence the risk of NTDs, and that the magnitude of the joint effect may be larger than that predicted on the basis of the effect of each locus alone.

Unfortunately, the original study was not large enough to show statistically significant differences among some of these risk estimates. In particular, although we used several approaches to formally test whether the odds ratio associated with both mutations (5.2) was significantly different from that associated with MTHFR-T alone (2.0), including tests for departure from multiplicative or additive models of interaction, we found no significant differences between the two. Nonetheless, because of the substantial (though nonsignificant) differences in these odds ratios, it would be important to explore further, in a larger, independent study, the possibility that gene-gene interactions may modify NTD risks. The public health relevance of these findings in a particular population may depend in part on the genotype frequency. In the study by Ramsbottom et al., for instance, homozygosity at both loci for the mutant alleles was not uncommon in the control population (1 in 86), and, if the observed association with NTD risk is causal, accounted for 4.4% of cases of spina bifida (Table 2). In other populations these figures may well be different. Another aspect that warrants further study is the possible interaction of these genotypes with environmental factors, in particular folate consumption, to understand whether particular genotypes may define high-risk groups of individuals who may benefit particularly from preventive measures such as dietary or supplemental folate use.

From a methodologic perspective, we would like to point out some possible advantages in using the "two-by-four table" approach when studying potential gene-gene interactions. First and foremost, laying out the results in such a fashion may help researchers to explore the data carefully. Second, it allows the contribution of genotypes at each locus to be assessed independently and in combination, both in terms of relative risk and attributable fraction. Third, the data so displayed can be used easily in formal tests for statistical interaction. Finally, the basic approach can be readily expanded to the case of multiple alleles at multiple loci. As such, the approach exemplified by the two-by-four table could be a valuable adjunct to the study of gene-gene interactions in the etiology of birth defects and human disease in general.

Acknowledgments: We would like to thank J. David Erickson, D.D.S., Ph.D., Quanhe Yang, Ph.D., and Muin J. Khoury, M.D., Ph.D., for helpful discussions. P.M. gratefully acknowledges the financial support of TELETHON Italy Grant No E.439.

References

1. Assmann SF, Hosmer DW, Lemeshow S, Mundt KA. Confidence intervalsfor measures of interaction. Epidemiology 1996;7(3)286-90.
2. Breslow, NE, Day NE. Statistical Methods in Cancer Research, Volume 1: the Analysis of Case-Control Studies. Lyon: International Agency for Research on Cancer, 1980.
3. Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology 1992;3(5):452-60.
4. Kang S-S, Wong PWK, Susmano A, Sora J, Norusis M, Ruggie N. Thermolabile methylenetetrahydrofolate reductase in patients with coronary heart disease. Am J Hum Genet 1991;48:536-545.
5. Khoury MJ, Adams MJ, Flanders WD. An epidemiologic approach to ecogenetics. Am J Hum Genet 1988; 42:89-95.
6. Khoury MJ, Beaty TH, Cohen BH. Genetic Epidemiology. Oxford University Press. New York and Oxford, 1993: 77-79.
7. Miettinen OS. Proportion of disease caused or prevented by a given exposure, trait or intervention. Am J Epidemiol 1974;99:325-32.
8. Ou CY, Stevenson RE, Brown VK, Schwartz CE, Allen WP, Khoury MJ, Rozen R, Oakley GP Jr, Adams MJ Jr. 5,10 Methylene-tetrahydrofolate reductase genetic polymorphism as a risk factor for neural tube defects. Am J Med Genet 1996; 63(4):610-614.
9. Ramsbottom D, Scott JM, Molloy A, Weir DG, Kirke PN, Mills JL, Gallagher PM, Whitehead AS. Are common mutations of cystathionine beta-synthase involved in the aetiology of neural tube defects? Clin Genet 1997;51(1):39-42.
10. Rothman KJ, Greenland S, Walker MA. Concepts of interaction. Am J Epidemiol 1980;112:467-70.
11. Sebastio G, Sperandeo MP, Panico M, de Franchis R, Kraus JP, Andria G. The molecular basis of homocystinuria due to cystathionine beta-synthase deficiency in Italian families, and report of four novel mutations. Am J Hum Genet 1995; 56(6):1324-1333.
12. van der Put NM, Steegers-Theunissen RP, Frosst P, Trijbels FJ, Eskes TK, van den Heuvel LP, Mariman EC, den Heyer M, Rozen R, Blom HJ. Mutated methylenetetrahydrofolate reductase as a risk factor for spina bifida. Lancet 1995 Oct 21;346(8982):1070-1071
13. Whitehead AS, Gallagher P, Mills JL, Kirke PN, Burke H, Molloy
14. AM, Weir DG, Shields DC, Scott JM. A genetic defect in 5,10 methylenetetrahydrofolate reductase in neural tube defects. Q J Med 1995;88(11):763-766.
15. Yang Q, Khoury MJ. Evolving methods in genetic epidemiology: III. Gene-environment interaction in epidemiologic research. Epidemiol Rev 1997; 19(1):33-43.

 

Table 1 A simple gene-gene interaction model in the context of case-control and cohort studies.

II. Cohort Study

Genotype: + = presence of a given genotype - = absence of given genotype
F_AB: Frequency of disease among people with genotype A and B
F_A-: Frequency of disease among people with genotype A alone
F_-B: Frequency of disease among people with genotype B alone
F_--: Frequency of disease among people without genotype A or B

Table 2. Estimated relative risk for neural tube defects associated with selected mutations of the MTHFR and CBS genes, alone or in combination (modified from Ramsbottom et al., 1997)

MTHFR: + = homozygous C677 mutation - = heterozygous C677 mutation or no mutation .
CBS: + = Insertion mutation - = no insertion mutation.
AFC(%): Attributable Fraction (percent) among cases exposed to risk factor.
AF (%): Attributable Fraction (percent) among all cases in the population.
Genotype frequency (%) in controls: frequency of genotype among individuals in the control group.