Candidate gene analysis in human neural tube defects

Abstract

Biochemical and developmental pathways, mouse models, and positional evidence have provided numerous candidate genes for the study of human neural tube defects. In a survey of 80 studies on 38 candidate genes, few found significant results in human populations through case-control or family-based association studies. While the folate pathway has been explored extensively, only the MTHFR 677C > T polymorphism was significant, and only in an Irish population. Developmental pathways such as the Wnt signaling pathway and Hox genes have also been explored without positive results. More than 90 mouse candidates have been identified through spontaneous and knockout mutations, but only the T locus (mouse Brachyury gene) showed association in an initial study that was not confirmed on follow-up. Positional candidates have been derived from cytogenetic evidence, but preliminary genomic screens have limited power due to small sample sizes. Future studies would increase their power to detect association by using more samples. In addition a clarification of the phenotype would be beneficial as many studies used different inclusion criteria. Incorporating several types of data could highlight better candidates, as would looking beyond the traditional sources for candidate genes. Recent studies of an energy metabolism gene (UCP2) and vitamin B metabolism (Transcoalbumin) have produced promising results. Utilizing other model organisms may also be beneficial, as in a recent study from a chick model of NTDs in NCAM1. New approaches combined with traditional methods and increased sample sizes will help prioritize human NTD candidate genes and clarify the complex etiology of this condition. © 2005 Wiley-Liss, Inc.

INTRODUCTION

Candidate genes for human neural tube defects have previously been supported by one of three types of evidence: biochemical pathways such as folate metabolism, mouse model genes, and positional candidates. The folate pathway provides candidates due to the known decrease in NTD incidence in the children of mothers who take folic acid periconceptionally. Mouse lines with an increased rate of spontaneous NTDs and mouse knockouts of specific genes have implicated over 90 candidate genes [Harris and Juriloff, 1999]. Positional candidates are derived from regions identified through genomic screens, which have been difficult to conduct due to the limited availability of families with multiple affected members. Most screens to date have had little power to detect major genes, but future studies will improve as more samples become available. Cytogenetic rearrangements and association of NTDs with trisomy 13 and 18 have implicated genomic positions as well, albeit large ones.

Few of the candidate genes studied in human neural tube defects have proved to have significant impact on the development of NTDs. Future research is needed to combine the strengths and weaknesses of traditional approaches while incorporating new sources of candidates. Other model organisms such as the chick, zebrafish, or sea urchin might clarify the numerous candidates seen in mouse models and be suitable organisms for testing intervention techniques in the future. Exploring other developmental and metabolic pathways, such as the retinoic acid pathway, could provide new insight as well. The definition of the neural tube defect phenotype has varied widely in these studies, with some utilizing only lumbosacral myelomeningocele and others a broad range of NTDs such as other open defects like anencephaly as well as closed defects like encephalocele or lipomyelomeningocele.

Biochemical Pathways

Folic acid is known to reduce the incidence of neural tube defects but the exact mechanism is unclear [Milunsky et al., 1991; MRC Vitamin Study Research Group, 1991; Centers for Disease Control and Prevention, 1992]. Folate is essential to the carbon transfer necessary for DNA synthesis, cell division, and tissue growth [Botto and Yang, 2000]. It is also necessary for DNA methylation, which plays an important role in gene expression and chromatin structure. Blood folate levels have a strong genetic component with an estimated heritability of 46% [Morrison et al., 1998], yet maternal folate supplementation can only prevent 50%–70% of NTDs [Chatkupt et al., 1994]. Folic acid has also been demonstrated to decrease the incidence of phenytoin induced NTDs in developing chick embryos, with relevance to humans due to the increased risk of NTDs associated with maternal exposure to anti-epileptic drugs [Guney et al., 2003]. Folate deficiency is one contributor to the multifactorial etiology of NTDs, and genes in this metabolic pathway have been the basis for many candidate gene studies.

Folate deficiency is one contributor to the multifactorial etiology of NTDs, and genes in this metabolic pathway have been the basis for many candidate gene studies.

Several potential genes have been derived from the folic acid pathway (outlined in Fig. 1): cystathionine-β-synthase (CBS), methionine synthase (MS), and 5,10-methylenetetrahydrofolate reductase (MTHFR). CBS converts homocysteine to cystathionine, thus inefficiency in this enzyme may lead to elevated homocysteine levels, as is often observed in mothers of children with NTDs [van der Put et al., 1995]. MS and MTHFR are key enzymes in the methylation cycle. To date, no association studies have found evidence to support a role for MS or CBS alone in NTD, although there is weak support that MS might work in conjunction with MTHFR [Morrison et al., 1998; Trembath et al., 1999].

Figure 1. Folic acid metabolism and the role of genes in the pathway [Sharp and Little, 2004].

Large doses of Vitamin A during pregnancy have been associated with congenital malformations including NTDs in animals [Elwood et al., 1992]. Anti-epileptic agents, such as valproic acid, are known to increase the risk of NTDs and may interfere with retinoic acid metabolism [Ross et al., 2000]. While the limited studies conducted to date have failed to find a significant association possibly due to small sample sizes, this pathway may still prove to be a fruitful avenue of research with candidates such as retinaldehyde dehydrogenase or cellular retinoic acid binding proteins. Related metabolic pathways have the potential for involvement in NTDs as well.

Developmental Pathways

Experimental systems have long been used to elucidate early developmental pathways. From early physical ablation experiments to later genetic disruptions, these complex signaling systems have been painstakingly deciphered. Interruption of several of these key pathways can lead to neural tube defects in animal models.

Wnt pathway signaling via β-catenin is essential in the development of nematodes, Drosophila, and vertebrates. Early in vertebrate development the functions of this pathway include: dorsalization of the body, posteriorization of the neural plate, midbrain development, and somite dorsoventral organization [National Research Council, 2000]. The Wnt pathway also acts in conjunction with Jun N-terminal Kinase (JNK) to establish planar cell polarity. Members of the Wnt pathway, such as Disheveled, have been investigated as candidate genes in human NTDs. Figures 2 and 3 depict the Wnt signaling pathways [National Research Council, 2000]. The hedgehog pathway is also used extensively in the early development of Drosophila and vertebrates. Functions such as notochord induction of the neural tube floor plate, notochord and floor plate induction of the somite sclerotome, and dorsoventral organization of the neural tube make this pathway a strong candidate for human NTDs [National Research Council, 2000]. In mice, null mutants of Sonic Hedgehog and Patched receptor cause spinal cord defects and an open neural tube, respectively, and these genes have been investigated as candidates in human NTDs as well [Zhu et al., 2003a]. Figure 4 depicts this pathway [National Research Council, 2000].

Figure 2. Wnt signaling via β-catenin [National Research Council, 2000].

Figure 3. Wnt signaling via JNK [National Research Council, 2000].

Figure 4. Hedgehog signaling through the patched receptor [National Research Council, 2000].

Homeobox (HOX) genes play a vital role in the proper development anterior-posterior (A-P) segments in Drosophila and vertebrates. The banded expression patterns determine segmentation boundaries along the A-P axis and these patterns of expression are strikingly similar between Drosophila and vertebrates, as is the organization of the genes within the genome [National Research Council, 2000]. These genes are necessary for proper development of vertebral segments and have thus been investigated as candidates for human NTDs [Volcik et al., 2002a].

Positional Candidates

Genomic screens can be a powerful tool to locate regions that contain risk-conferring genes for human neural tube defects, but ascertaining samples from families with multiple affected individuals is a difficult process. Recently a sufficient number of families was collected through the NTD Collaborative Group and the resulting genomic screen has implicated several potential regions [Speer et al., 2003]. The association of NTDs with trisomies 13 and 18 has implicated these chromosomes, and several cytogenetic rearrangements have provided positional candidates as well [Melvin et al., 2000]. Expansion on these types of studies can focus the genomic regions that may contain candidate genes.

HOW CANDIDATE GENES ARE INVESTIGATED

Case-Control Studies

Retrospective case-control studies have been classically used in epidemiology and they are readily applied to genetic studies. Typically a 2 × 2 table is constructed assessing the presence or absence of a risk factor (an allele or genotype for genetic studies) in a case population as compared to a control population [Kahn and Sempos, 1989]. Table I outlines the construction of a 2 × 2 table for testing the risk conferred by an allele, but this example can be extended to a comparison of a risk genotype to other genotypes separately or as a group.

Table I. 2 × 2 Table for Construction an Odds Ratio From Case-Control Data*

Risk allele	Cases	Controls	Total
* This score test is constructed from counts of the risk allele in case and control populations.
Present	a	b	a + b = m1
Absent	c	d	c + d = m2
Total	a + c = n1	b + d = n2	m1 + m2 = n1 + n2 = t

An odds ratio (OR) is constructed to compare the frequency of the risk allele in case to controls and confidence intervals can be calculated to test for significance.

• 

If the frequency of the risk allele is the same in cases and controls, the OR will be near 1. If the 95% confidence interval does not contain 1, than the results are significant with a P-value less than 0.05; if this interval does contain 1, then the results are not significant at the 0.05 level. However, using the traditional 0.05 P-value is not entirely appropriate in these situations due to multiple testing issues. In a genome-wide setting, it has been proposed that several levels of significance be used in publications: suggestive, significant, and highly significant with point wise significance levels of 7 × 10⁻⁴, 2 × 10⁻⁵, and 3 × 10⁻⁷, respectively [Lander and Kruglyak, 1995]. There are several variations on these types of P-value adjustments to account for smaller samples sizes or the type of standard error used [Kahn and Sempos, 1989].

For equivalent sample sizes, a case-control method will always have more power than family-based methods, but there are potential pitfalls to using this method. If the case and control samples are not taken from identical populations, the measured differences between the groups may not be due to the risk factor under study. In genetic studies this is particularly problematic because underlying ethnic stratification within the population can lead to inherit differences in allele frequencies.

If the case and control samples are not taken from identical populations, the measured differences between the groups may not be due to the risk factor under study. In genetic studies this is particularly problematic because underlying ethnic stratification within the population can lead to inherit differences in allele frequencies.

If case and control populations are not truly interbreeding, any detected genetic differences may not be due to the analyzed gene although this is a point of contention within the field [Thomas and Witte, 2002; Wacholder et al., 2002]. Case-control methods do not directly study transmission from parent to child—the key determining factor in genetics—so alternative methods are necessary.

Family-Based Studies

Family-based studies of candidate genes offer a solution to population stratification issues by using unaffected members of families or non-transmitted alleles in place of the control population. While these studies require more effort to ascertain, considerable resources to genotype several members of a family, and offer less power than case-control methods, they can improve the certainty of the results.

The most common NTD families available for research are small nuclear families, which are ideally suited to the transmission disequilibrium test (TDT). The TDT has developed into one of the most widely used tests of association for candidate genes, and this method sired several variations, such as the PDT and Sib-TDT that can incorporate several types of family structures including missing parents and unaffected siblings.

The TDT has developed into one of the most widely used tests of association for candidate genes, and this method sired several variations, such as the PDT and Sib-TDT that can incorporate several types of family structures including missing parents and unaffected siblings.

[Spielman et al., 1993; Spielman and Ewens, 1998; Martin et al., 2000].

The TDT test also uses a 2 × 2 matrix, but in a very different way (see Table II). Within parent-child triads, the transmitted and non-transmitted alleles can be counted if the parents are heterozygous at that particular locus. The alleles transmitted to an affected child are compared to the non-transmitted alleles across a large number of samples to look for a deviation from the 50/50 expectation under Mendelian segregation patterns assuming no linkage. If affected children have received the disease-associated allele more often than expected by chance, then there is evidence for association of that allele and susceptibility to disease. Population stratification is not an issue when potential products of the same mating are used for comparison.

Table II. TDT 2 × 2 Table Construction*

Transmitted allele	Non-transmitted allele
	M1	M2	Total
* The transmitted and non-transmitted alleles are counted for each sampled parent. Only the heterozygous parents will contribute to the score test.
M1	n11	n12	n1
M2	n21	n22	n2
Total	n1	n2	2n

Transmissions from homozygous parents are not informative, and therefore n₁₁ and n₂₂ do not contribute to the test statistics because they will always transmit the same allele. The requirement of heterozygous parents makes determining sample size a priori difficult because it will vary according to the heterozygosity of the marker in the parental samples.

Multiallelic markers were one of the first variants of the TDT to be developed [Sham and Curtis, 1995]. Expansions on the TDT have been made to utilize other family structures, the PDT [Martin et al., 2000], extended haplotypes, TDTHAP [Clayton, 1999], and allowing for errors, TDT-AE [Gordon et al., 2001]. TDT-PC can be used to calculate the power of a study given the samples size and other marker parameters [Chen and Deng, 2001].

The TDT and TDT-like tests rely on counts of transmissions and are consequently considered score tests. The Genotype Relative Risk (GRR) model is a likelihood-based test of association that also utilizes parent-child triads and can test for parent-of-origin effects as well as incorporate other parameters such as environmental risk factors [Schaid and Sommer, 1993]. The GRR method has been expanded upon in a TDT-based log-linear model that is more accessible to researchers through a SAS based program [Weinberg et al., 1998].

Candidate Genes Studied in Human Neural Tube Defects

Many candidate genes have been investigated in human NTD populations, yielding few positive results. Table III summarizes the results of these studies, including why the candidate was investigated, what population was studied and what type of test was used. The types of NTDs included in the study sample are also included when it was available. While this table attempts to compare a majority of work in the field, it cannot capture every aspect of these publications, and the original literature should always be consulted for details of these studies.

Table III. Candidate Gene Chapter

Human gene	Human locus	Type of candidate	Study rationale	Reference	Population studied	Sample size	Diagnoses included	Type of study	Summarized results	Test used (p-value)	Conclusion
BHMT (betaine-homocysteine methyltransferase)	5q13.1-q13.2	Folate Metabolism	Enzyme of the homocysteine methylation pathway	Morin et al. [2003]	Canadian	54 patients, 57 mothers of patients, 93 control children and 86 control mothers	Spina bifida	Frequency of the G742A polymorphism	No significant difference was found between cases and controls	Odds Ratio with 95% CI (not sig)	No evidence for association
BMP4 (bone morphogenic protein 4)	14q22-q23	Mouse Model	Knockout mice exhibit little mesoderm differentiation causing disorganized structures such as a small neural plate and open cranial folds	Felder et al. [2002]	German	179 cases, 161 controls	Non-syndromic spina bifida aperta (meningocele and myelomeningocele)	Screen for mutations including T455C polymorphism	Four mutations found in four case and unaffected individuals	Equal frequencies	No evidence for association
BRCA1	17q 21	Mouse Model	Null mutation in mice have an increased frequency of NTD	Morrison et al. [1998]	Dutch and British	79 cases and their families	Spina bifida spina bifida occulta, and encephalocele	Screen for A4956G and A1186G	No transmission disequilibrium	TDT (not sig)	No evidence for association
CBS (cystathionine beta-synthase)	21q22.3	Folate Metabolism	Major enzyme regulating homocysteine levels and elevated maternal plasma levels have been observed in some NTD pregnancies	Ramsbottom et al. [1997]	Irish	83 cases, 79 mothers of cases, 201 control infants, 241 control mothers	NTD (myelomeningocele, anencephaly, encephalocele)	Examined frequency of G307S and I278T alleles as well as a 68bp insertion	No increased frequency	Odds Ratio with 95% CI (not sig)	No evidence for association
				Morrison et al. [1998]	Dutch and British	79 cases and their families	Spina bifida spina bifida occulta, and encephalocele	Screen for a novel polymorphism and T2199C	No transmission disequilibrium	TDT (not sig)	No evidence for association
				Zhao et al. [2000]	Chinese	40 mothers of cases, 36 control mothers	NTD	Frequency of T833C and G919A in mothers of NTD pregnancies	No significant difference was found between NTDs mothers and non-NTDs mothers	Chi-square test (not sig)	No evidence for mothers' genotypes
CRABP1 (cellular retinoic acid binding protein 1)	15	Retinoic Acid Pathway	Exposure to high levels or retinoic acid can cause NTDs in humans and CRABP1 is highly expressed in the developing embryo	Dickerson et al. [2002]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	screen for polymorphisms	2 polymorphisms in both the gene and promoter region did not show association	PDT (not sig) FBAT (not sig)	No evidence for association
CRABP2 (cellular retinoic acid binding protein 2)	1q21.3	Retinoic Acid Pathway	Exposure to high levels or retinoic acid can cause NTDs	Dickerson et al. [2002]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	3 polymorphisms did not show significant association	PDT (not sig) FBAT (not sig)	No evidence for association
CSK	15q23-q25	Mouse Model	Involved in the organization of the cytoskeleton; mice deficient for CSK have an increased risk of NTDs	Klootwijk et al. [2003]	Dutch	38 multiplex families; 79 case samples	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms in the coding region and part of the introns	3 polymorphisms detected in both patients with NTDs and controls with similar frequencies	Frequency too low for TDT based log linear	No evidence for association
CYP26 (cytochrome P450 retinoic acid-metabolizing enzyme)	10q23-q24	Retinoic Acid Pathway	Exposure to high levels or retinoic acid can cause NTDs, CYP26's could provide some protection from high levels of retionoic acid	Dickerson et al. [2002]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	Polymorphisms did not show significant association	PDT (not sig) FBAT (not sig)	No evidence for association
DVL2 (Disheveled)	17	Mouse Model	Essential for neural tube closure (and other functions) in mice	Speer et al. [2003]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	8 novel SNPs identified, but none associated with NTDs	n/a	No evidence for association
FR-alpha (folate receptor alpha)	11q13.3-q13.5	Folate Metabolism	Primary folate receptor responsible for binding and importing folate; mutations may reduce folate levels	Trembath et al. [1999]	Midwestern US	128 families from Iowa 35 from Minnesota, and 9 from Nebraska (96% Caucasian)	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms	No consistent polymorphisms found	n/a	No evidence for association
FR-beta (folate receptor beta)	11q13.3-q13.5	Folate Metabolism	Primary folate receptor responsible for binding and importing folate; mutations may reduce folate levels	Trembath et al. [1999]	Midwestern US	128 families from Iowa 35 from Minnesota, and 9 from Nebraska (96% Caucasian)	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms	Polymorphisms found were not significant	Odds Ratio with 95% CI (not sig)	Lack of association is consistent with biochemical mechanisms of folate deficiency
GCPII (glutamate carboxypeptidase II)	11p11.2	Folate Metabolism	H475Y polymorphism decreases enzyme activity and is associated with decreased plasma folate levels and increased plasma total homocysteine	Vieira et al. [2002]	Caucasian	128 Iowa families, 35 from Minnesota, 9 from Nebraska, 53 families from the NTD Collaborative Group, and 41 CEPH families	Individuals with NTDs and their immediate families	Tested for allelic variation	Polymorphisms tested were not significant	TDT and Chi-square tests (not sig)	No evidence for association
				Afman et al. [2003]	Dutch Caucasians	96 cases, 113 mothers of cases, 97 fathers, and 101 controls	Spina bifida	Test for H475Y polymorphism	Polymorphism tested was not significant	Odds Ratio genotype test (not sig)	No evidence for association
HOX Gene Family	four clusters: A on 7, B on 17, C on 12, and D on 2	Mouse Models	Members of all four clusters of the HOX genes are implicated in neural tube closure in mice	Volcik et al. [2002a]	American Caucasians and Hispanics of Mexican descent	459 patients and their parents	Isolated myelomeningocele	Screen for polymorphisms	Polymorphisms found were not significant allele-wise or genotype-wise	ETDT (not sig)	No evidence for association
MACS (human homologue of MARCKS)	1p34-1pter	Mouse Model	MARCKS is important to the development of the central nervous system; mice lacking the protein have a higher frequency of NTDs	Stumpo et al. [1998]	Caucasian	43 simplex families	L-S Myelomeningocele	No polymorphisms within the gene, tested markers flanking the gene	Not significant, but the test had low power	TDT (not sig)	No evidence for association
MLP (MARCKS-like protein)	1p34-1pter	Mouse Model	MARCKS is important to the development of the central nervous system; mice lacking the protein have a higher frequency of NTDs	Stumpo et al. [1998]	Caucasian	43 simplex families	L-S Myelomeningocele	Tested MLP1 polymorphisms	Not significant, but the test had low power	TDT (not sig)	No evidence for association
				Klootwijk et al. [2003]	Dutch	38 multiplex families; 79 case samples	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms	No polymorphisms found	n/a	No evidence for association
Msx2 (muscle segment homeobox 2)	5q34-q35	Mouse Model	Genetic alterations of MSX2 have been shown to cause failure of cranial neural tube closure in mice.	Stegmann et al. [2001]	German and Italian	204 cases (10 anencephaly, 8 encephalocele, and 183 spina bifida aperta) and 222 German controls	Grouped into 3 groups: 1. Anencephaly (including craniorachischisis) 2. Encephalocele 3. Spina Bifida Aperta	Mutation screen	Found polymorphisms, but not significant	2-sided Fisher exact test (not sig)	No evidence for association
MTHFD1 (methylenetetra-hydrofolate dehydrogenase/methenyltetrahydrofolate-cyclohydrolase/formyltetrahydrofolate synthetase)	14q24	Folate Metabolism	MTHFD encodes a single protein with three catalytic properties important in folate metabolism	Hol et al.[1998]	Dutch	38 familial cases, 79 sporadic cases, and 300 controls	NTD (myelomeningocele, anencephaly, encephalocele)	SSCP screen for mutations	One polymorphism found in 3 members of one family (NTD, SBO, and unaffected) two other polymorphisms for in both cases and controls	n/a	No evidence for a major role however the identification of a mutation in one family suggests that this gene can act as a risk factor for human NTD.
				Brody et al. [2002]	Irish	319 complete triads, 22 cases, 13 mothers of cases, 2 fathers of cases, an additional 83 mothers of cases, and two control populations of 699 individuals and 318 pregnant women	NTD (spina bifida, encephalocele, anencephaly, anencephaly plus spina bifida)	R653Q variant	Excess of the Q allele in mothers of cases and preferential transmission to cases	Odds Ratio (p = 0.003) and traditional TDT (p = 0.015) and TDT based log linear (p=0.007) tests also significant	Mothers with the MTHFD1 “QQ” genotype have an increased risk (∼1.5-to2.0-fold) of having an NTD-affected pregnancy.
MTHFR (5,10-methylenetetrahydrofolate reductase)	1p36.3	Folate Metabolism	MTHFR is an enzyme within the folate pathway, and the C677T allele decreases the activity of folate-dependent re methylation of homocysteine	Van der Put et al. [1995]	Dutch	55 cases and controls	NTD	Screen for variants	C677T found to be significantly different in cases than controls	Odds Ratio (p < 0.05)	The C677T mutation is a genetic risk factor for spina bifida
				Morrison et al. [1998]	Dutch and British	79 cases and their families	Spina bifida spina bifida occulta, and encephalocele	Screen for C677T and C1068T	Equal transmission, but approaches significance in conjunction with MS	TDT (not sig)	Possible association in conjunction with MS
				Shields et al. [1999]	Irish	271 cases and 218 families	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for “thermolabile” variant (C677T)	The T allele frequency is higher in cases than controls and the TT genotype is significant in cases and their mothers	Odds Ratio (p = 0.0005) and TDT based log linear (p < 0.05)	The T allele increases risk to cases a modest additional risk is conferred by a maternal TT genotype
				Stegmann et al. [1999]	German	148 cases and 174 controls (population-based test) 77 parent child triads and 110 mother-child pairs (family-based test)	Non-syndromic spina bifida at any level (plus 8 anencephalics and 2 encephalocele for molecular analysis)	Test for association with C677T and A1298C polymorphisms	No significant differences in cases and controls; equal transmission observed	Likelihood ratio test (not sig) and TDT (not sig)	No evidence for association
				Trembath et al. [1999]	Midwestern US	128 families from Iowa 35 from Minnesota, and 9 from Nebraska (96% Caucasian)	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms (C677T mutation)	Not significant in myelomeningocele cases, however the mutation was significantly different in non myelomeningocele (lipomyelomeningocele, intradural lipoma and sacral hypoplasia/agenesis)	TDT (p = 0.03 in non-myelomeningocele NTDs)	Evidence for in increased risk in non-myelomeningocele cases
				Christensen et al. [1999]	Canadian (75% English or French)	56 cases, 62 mothers of cases, 97 control children, 90 control mothers	NTD (myelomeningocele, anencephaly, encephalocele)	“Thermolabile” (C677T) mutation screen	Significant when mother and child are homozygous for the variant allele	Odds Ratio (p < 0.05)	Increased risk for NTD associated with the C677T polymorphism
				De Marco et al. [2002]	Italian	203 case, 98 of their mothers, 67 fathers, and 210 controls	Myelomeningocele, meningocele, lipoma, lipomyeloschisis, dermal sinus, tight filum terminalis	Screen for A1298C	Cases, mothers, and fathers with the CC allele are significantly different than controls	Odds Ratio (all p < 0.05)	Evidence for in increased risk when cases or their mothers have a C allele or fathers have a CC genotype
				Johanning et al. [2002]	84% Caucasian, 16% African-American	77 case and 77 controls	NTD (myelomeningocele, anencephaly, encephalocele)	Thermolabile mutation: alanine to valine	Heterozygotes and homozygotes had higher risk for NTD's, with a more significant difference from 1998-1994 than those born after 1994	Odds Ratio with 95% CI (p < 0.05)	Increases risk for heterozygotes and valine homozygotes
				Cunha AL et al. [2002]	Brazilian	25 cases with mothers, 75 controls	NTD (myelomeningocele, anencephaly, encephalocele)	C677T and A1298C mutations in MTHFR gene	Genotypes were associated with metabolite blood levels	No difference in case and control frequencies	Thermolabile mutation may affect vitamin B12 and homocysteine metabolism, which possibly could contribute to NTDs
				Gonzalez-Herrera et al. [2002]	Hispanic (Yucatan area)	65 cases, 60 of their mothers and 110 controls	Spina bifida	Screen for C677T	The C677T variant frequency is not different in cases than controls	Odds Ratio (not sig)	No evidence for association
				Rampersaud et al. [2003]	American Caucasian	175 cases and their families and 195 controls	L-S Myelomeningocele without folic acid supplementation	Screen for “thermolabile” variant (C677T)	Evidence of association with cases only, but no evidence of unequal transmission or a previously reported association with CBS	Odds Ratio (p < 0.05), PDT (p > 0.1)	Evidence for an increased risk in addition to other candidate genes
				Pietrzyk et al. [2003]	Polish	104 cases, 106 mothers of cases, and 100 adult controls	Non-syndromic, isolated spina bifida (separated into lumbosacral and thoracolumbar groups)	Screen for C677T	Statistically significant differences in cases and controls seen in homozygous mothers and children	Odds Ratio significant in cases (p = 0.049) and mothers (p = 0.007)	Increased risk associated with the C677T polymorphism in homozygous cases and mothers
				Revilla et al. [2003]	Spanish	27 cases, 28 mothers of cases, 23 siblings of cases, and 159 controls	NTD	Screen for C677T and A1298C	There was no significant difference in these genotypes between cases and controls. Therefore we conclude these polymorphisms have no association with NTDs in the Spanish population.	Chi-square test (not sig)	No evidence for association
MTR or MS (Methionine Synthase)	1q43	Folate Metabolism	converts intracellular folate and homocysteine to tetrahydrofolate and methionine. Tetrahydrofolate is a crucial ingredient in biosynthesis of DNA and RNA, while methionine is important in numerous methylation reactions	Morrison et al. [1998]	Dutch and British	79 cases and their families	Spina bifida spina bifida occulta, and encephalocele	Screen for C5049A and A2756G	Equal transmission, but approaches significance with MTHFR	TDT (not sig)	Possible association in conjunction with MTHFR
				Brody et al. [1999]	Irish	85 case-parent triads	Spina bifida	Screen for polymorphisms	2 polymorphisms 900 and 154 kb away from MS not associated	TDT (not sig)	No evidence for association
				Trembath et al. [1999]	Midwestern US	128 families from Iowa 35 from Minnesota, and 9 from Nebraska (96% Caucasian)	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms	Found polymorphisms, but no significant differences in cases and controls	Chi square with 95% Confidence Interval	No evidence for association
				Christensen et al. [1999]	Canadian	56 cases, 62 mothers of cases, 97 control children, 90 control mothers	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for 2756A > G polymorphism	Polymorphism less common in cases	Odds Ratio (p < 0.05)	The polymorphism was associated with a reduced risk for NTD
				De Marco et al. [2002]	Italian	203 case, 98 of their mothers, 67 fathers, and 210 controls	Myelomeningocele, meningocele, lipoma, lipomyeloschisis, dermal sinus, tight filum terminalis	Screen for A2756G variant	No differences between case and control frequencies	Odds Ratio (not sig)	No evidence for association
				Gueant-Rodriguez et al. [2003]	Southern Italian	40 cases and 58 age and sex matched controls	Myelomeningocele, all with failure of closure at or below position L-2	Evaluated the association of a MTR A2756G polymorphisms alone and with MTHFR C677T	MTR polymorphisms significantly different in cases than controls	Odds Ratio (p = 0.046)	Evidence for increase risk from the polymorphism
MTRR (methionine synthase reductase)	5p15.2-15.3	Folate Metabolism	Activates cobalamin-dependent methionine synthase as part of the homocysteine re methylation pathway; impairment of folate and cobalamin metabolism has been observed in families with NTDs.	Wilson et al. [1999]	Canadian	56 cases, 58 mothers of cases, 97 control children, 89 control mothers	Spina bifida	66A > G polymorphism check	Polymorphism only significant when cobalamin is low	Odds Ratio (p < 0.05 with low cobalamin)	Polymorphism increases risk of NTD when cobalamin status is low
				Gueant-Rodriguez et al. [2003]	Southern Italian	40 cases and 58 age and sex matched controls	Myelomeningocele, all with failure of closure at or below position L-2	Evaluated the association of a MTRR A66G polymorphisms alone and with MTHFR C677T	MTRR polymorphisms were associated with NTD risk in cases having a MTHFR 677 CC wild genotype	Odds Ratio (p = 0.023)	Evidence for risk after accounting for the MTHFR allele
				Pietrzyk et al. [2003]	Polish	104 cases, 106 mothers of cases, and 100 adult controls	Non-syndromic, isolated spina bifida (separated into lumbosacral and thoracolumbar groups)	A66G polymorphism	Statistically significant differences in cases and controls seen in homozygous mothers and children	Odds Ratio significant in cases (0.034) and mothers (0.039)	Increased risk associated with the A66G polymorphism in homozygous cases and mothers
NAP1L2 (nucleosome assembly protein 1-like 2)	Xq12-q24	Mouse Model	Mouse models exhibit NTDs closely resembling spina bifida and anencephaly in humans; plays a role in the cell cycle regulation of developing neurons.	Rogner et al. [2002]	United Kingdom population of mixed ethnic origins	114 cases	Spina bifida occulta to craniorachischisis, but were predominantly spina bifida aperta (57) and anencephaly (36)	Search for polymorphisms	Polymorphisms found were not associated	n/a	In the context of a multifactorial origin, the polymorphisms within the 5′ CpG island of NAP1L2 may contribute to the complex etiology of NTDs
NCAM1 (neural cell adhesion molecule 1)	11q23.1	Chick Model	Controls cell migration in neural tissues; cell adhesion molecules are disturbed in spontaneous NTDs in chicks	Bastress et al. [2005]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	5 polymorphisms screened, one is significant	PDT (p = 0.041), FBAT (p = 0.00045)	Significant evidence of association in this population
Noggin	17q22	Mouse Model	Mouse models show fully penetrant skeletal abnormalities and defects in growth and patterning of the neural tube.	Felder et al. [2002]	German	179 cases	Non-syndromic spina bifida aperta (meningocele and myelomeningocele)	Search for polymorphisms	One polymorphism present in one SB patient, her unaffected father, and in one control individual	n/a	No evidence for association
				Bauer et al. [2002]	American Caucasian	202 cases	143 lumbosacral myelomeningocele (143), thoracic myelomeningocele (12), thoracic myelomeningocele (23), lipomyelomeningocele (12), and miscellaneous (22)	Search for allelic variances	One variant found in a single case an unaffected family members	n/a	No evidence for association
PAX Gene Family	multiple sites	Mouse Models	PAX genes regulate normal development, Pax 1 and Pax3 (splotch) knock out mice exhibit severe NTDs	Chatkupt et al. [1995]	17 American and Dutch	17 informative multiplex families	NTD, including SBO	Linkage analysis in Pax 3	No evidence around PAX3 with a dominant or recessive model	Limited power for a linkage study	No evidence of linkage
				Hol et al. [1996]	Dutch	38 familial cases, 79 sporadic cases, and 300 controls	All spina bifida except 1 encephalocele and 1 craniorachischisis	Allelic association tests	Pax1 polymorphism found in one case and the maternal grandmother, Pax3 polymorphism found in cases and controls at equal frequencies	n/a	No evidence for association
				Volcik et al. [2002b]	59%Hispanic, 35% White, 6% other	459 patients and their parents	Isolated myelomeningocele	Screen for polymorphisms	Positive transmission of alleles for markers within PAX1, PAX7, and PAX8 were detected but not in phase with disease	TDT (Pax1 p = 0.019, Pax7 p = 0.011, Pax8 p = 0.013)	Associations may be with a disease locus within the same region as these genes, therefore future studies should focus on this area
				Trembath et al. [1999]	Midwestern US	128 families from Iowa 35 from Minnesota, and 9 from Nebraska (96% Caucasian)	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms in Pax 3	All polymorphisms found in both cases and controls	n/a	No evidence for association
PDGFRA promotor	4q12	Mouse Model	Pax-1 mouse models have indicated that deregulated expression of the gene encoding the platelet-derived growth factor alpha receptor (PDGFRA) causes congenital NTDs.	Joosten et al. [2001]	Dutch	49 familial, 76 sporadic, and 77 controls	Spina bifida	Tested for presence of specific PDGFRA promotor haplotype combinations	Heterozygotes have an increased risk for malformations	Chi-square test (not sig)	This promoter most likely acts in combination with other adverse factors
				Joosten et al. [2002]	Dutch	49 familial, 76 sporadic, and 77 controls	Spina bifida	Expression studies in human cell line culture	PDGFRA expression enhanced by treatment with retinoic acid or cyclic AMP	n/a	Improper PDGFRA expression may play a role in the etiology
RALDH2 (retinaldehyde dehydrogenase)	10q23-q24	Retinoic Acid Pathway	Exposure to high levels or retinoic acid can cause NTDs,	Dickerson et al. [2002]	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	Out of 9 polymorphisms, one was significant in one of 3 tests	FBAT (p = 0.02) PDT (p > 0.05) TRANSMIT(p > 0.05)	Limited evidence for association, but not likely to be important in the context of the whole genome
RFC-1 (reduced folate carrier protein)	21q22.3	Folate Metabolism	A80G polymorphism causes lower plasma folate levels	De Marco et al. [2001]	Italian	203 unrelated nonsyndromic cases, 98 mothers and 67 fathers	NTD (myelomeningocele, anencephaly, encephalocele)	Check for A80 G polymorphism, along with MTHFR A1298 C allele frequency	The frequency of the polymorphism in cases, mothers, and fathers is higher than in controls	Chi-square test, p < 0.05 for cases, mothers, and fathers	Both RFC-1 and MTHFR polymorphisms may play a role in NTD risk in the Italian population
				Shaw et al. [2002]	California	133 case infants and 188 control infants	NTD (anencephaly, spina bifida cystica, craniorachischisis, or iniencephaly)	Check for A80G polymorphism	However some evidence suggestive of an interaction between infant G80/G80 genotype and maternal supplemental vitamin use on the occurrence of spina bifida	Odds Ratio (not sig)	No evidence for association
				De Marco et al. [2003]	Italian	174 cases, 43 mothers of cases, 53 fathers of cases, and 156 controls	Non-syndromic NTD	Check for A80G polymorphism, along with known MTHFR polymorphisms	GG genotype more common in the NTD cases and mothers; no evidence for an association between NTD phenotype and combined MTHFR C677T/RFC-a A80G genotypes;	Odds Ratio (p < 0.05 for homozygous cases and mothers)	Evidence for increased risk to cases and mothers, but not in combination with MTHFR
SHH (Sonic Hedgehog)	7q36	Chick and Mouse Models	SHH patterns the midline neuro-axis and distal limb structures and abnormal expression is seen in mice with NTDs	Vargas [1998]	Irish	78 families, 49 with NTD (others have sacral agenesis or polydactyly)	NTD	Screen for mutations	One sequence variation was found in two unrelated individuals with NTD and in none of the normal control samples	n/a	No evidence for association
				Kirillova et al. [2000]	French	25 embryos with axial structure abnormalities	Craniorachischisis and spina bifida	In situ hybridization for expression	Abnormal patterns of expression observed	n/a	Abnormal expression associated with neural tube defects in embryos
				Zhu et al. [2003a]	Midwestern US	50 cases and 50 controls (pilot study)	Spina bifida	Screen for mutations or polymorphisms	No known mutations previously observed were observed, one intronic polymorphism identified	n/a	No evidence for association
SHMT (serine hydroxymethyltransferase	cytosolic form: 17p11.2 mitochondrial form: 12q13.2	Folate Metabolism	SHMT catalyzes a reaction involved in metabolism of folate dependent homocysteine and elevated homocysteine levels and decreased plasma folate levels were observed in mothers of children with NTDs	Heil et al. [2001]	Dutch	109 cases, 120 mothers of cases, and 420 controls	Spina bifida	Screen for mutations or polymorphisms	Two polymorphisms found, but neither significant	Pearson Chi Square (not sig)	No evidence for association
SLUG (encodes a zinc finger protein of the Snail family of transcription factors)	8q11	Mouse, Chick and Frog Models	SLUG is selectively expressed in the dorsal part of the developing neural tube; ablation and antisense experiments in chicken suggest that SLUG may be an important factor during neural tube closure	Stegmann et al. [2001]	German	150 cases (11 familial) and 136 controls	Spina bifida aperta	Screen for polymorphisms	One polymorphism found in a case but also the unaffected parent	n/a	No evidence for association
T (human analogue of Brachyury gene in mice)	6q27	Mouse Model	Mouse knock outs of its homologue, T (Brachyury), have notochord abnormalities	Morrison et al. [1998]	Dutch and British	79 cases and their families	Spina bifida spina bifida occulta, and encephalocele	Screen for TIVS7C and A530G	Preferential transmission of the TIVS7C allele	TDT (p = 0.03)	Significant evidence for association
				Trembath et al. [1999]	Midwestern US	13 informative sporadic cases and their families	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for T1VS7-2 polymorphism	Equal transmission seen	TDT (not sig)	No evidence for association
				Shields et al. [2000]	Irish	218 case-parent triads	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for T1VS7-2 polymorphism in intron 7 of T gene	Allele markedly associated with cases born before 1980, but not with more recent cases.	TDT (p = 0.01), MGRR (p = 0.02), and TDT based log-linear (p = 0.006)	This polymorphism is a modest, but important risk factor and works via different mechanism than MTHFR polymorphism
				Richter et al. [2002]	German	183 cases and 266 controls (familial study)	NTD (myelomeningocele, anencephaly, encephalocele)	Check for T1VS7 allele	No differences between case and control frequencies	McNemar's test (not sig)	No evidence for association
				Speer et al. [2002]	American Caucasian	86 simplex families	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for T1VS7-2 polymorphism	Some evidence of association found for one polymorphism in one test but not across the region	PDT (p = 0.04) Haplotype Pattern Mining (p = 0.62)	Regional haplotype did not have significance, so the T locus is not a major risk factor
TC (transcobalamin)	11q11-q12	Vitamin B Pathway	Low plasma TC levels have previously been associated with an increased risk for having a child with an NTD	Afman et al. [2002]	Dutch	42 mothers of cases and 73 control mothers	Non-syndromic NTD	Screen for variants in coding region of the TC gene	No effect on homocysteine and NTD risk could be detected with any variant discovered.	Odds Ratio with logistic regression (not sig)	No evidence for association with risk or homocysteine levels
				Gueant-Rodriguez et al. [2003]	Southern Italian	40 cases and 58 controls	Myelomeningocele, all with failure of closure at or below position L-2	Evaluated the association of a TC C777G polymorphisms alone and with MTHFR C677T	Polymorphism not significant alone, but is in combination with the MTHFR CC genotype	Odds Ratio (p = 0.028)	Evidence for TC playing a role with MTHFR
TERC (telomerase RNA component)	3q21-q28	Mouse Model	Knockout mice have chromosomal instability which has been implicated in failed neural tube closure	Benz et al. (in press)	American Caucasian	477 sporadic and familial cases and their families, 153 controls	L-S Myelomeningocele	Screen for polymorphisms	Two novel SNPs found, but neither was significant	PDT (not sig)	No evidence for association
TFAP2-alpha (transcription factor activating enhancer-binding protein 2 alpha)	6p24	Mouse Model	AP-2 null and chimeric mice exhibit exencephaly.	Stegmann et al. [1999]	German and Italian	204 cases (10 anencephaly, 8 encephalocele, and 183 spina bifida aperta) and 222 German controls	Grouped into 3 groups: 1. Anencephaly (including craniorachischisis) 2. Encephalocele 3. Spina Bifida Aperta	Mutation screen	Found polymorphisms, but not significant	2-sided Fisher exact test (not sig)	No evidence for association
				Klootwijk et al. [2003]	Dutch	38 multiplex families; 111 case samples	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms of coding region and part of the introns	Found three polymorphisms and the C1257T allele approached significance	TDT based log linear (not sig)	1257C > T allele may confer an increased risk
UCP2 (uncoupling protein 2)	11q13	Energy Metabolism	Polymophisms are capable of affecting energy metabolism, body weight regulation, and possibly preventing the buildup of reactive oxygen species, all factors that could contribute to neural tube defect risk through maternal obesity and diabetes.	Volcik et al. [2003]	California	133 case infants and 188 control infants	NTD (anencephaly, spina bifida cystica, craniorachischisis, or iniencephaly)	Searched for insertion/deletion at 3′ UTR	Combined homozygosity for both UCP2 variants resulted in a threefold or more elevated risk of SB (neither was significant individually))	Odds Ratio (p < 0.05)	Compelling evidence for UCP2 as an NTD risk factor
X-chromosome		Positional Candidate	A large Icelandic pedigree displayed what appeared to be X-linked spina bifida and anencephaly.	Newton R. et al. [1994]	Icelandic	One large family with multiple affected members (43 sampled)	Spina bifida and anencephaly	Linkage analysis	Apparent X-linked spina bifida and anencephaly haplotypes were extensively analyzed and found to exclude linkage to the X chromosome	Linkage and haplotype analysis	No evidence for linkage in this family
ZIC 2 (transcription factor)	13q32	Positional Candidate	ZIC2 is in the chromosome 13q32 critical deletion seen in the 13q-syndrome (included encephalocele and anencephaly); targeted mutation of ZIC2 in mouse models result in NTDs	Brown et al. [2002]	Children's Memorial Hospital Spina Bifida Clinic (Chicago)	192 cases	95% L-S myelomeningocele (5% encephalocele and other NTDs)	Screen for mutations	Found polymorphisms but in too few samples to produce significant results	Attempted TDT, but there were only 9 suitable triads	ZIC2 mutations are, at most, a very infrequent cause of NTDs
				Zhu et al. [2003b]	Hispanic	69 controls and cases	NTD (spina bifida or anencephaly)	Screen for 9H or 10H alleles	No differences between case and control frequencies	Odds Ratio (not sig)	No evidence for association
ZIC 3 (zinc finger protein of cerebellum)	Xq26.2	Mouse Model	Execephaly and tail defects in Bent tail mice and sacral anomalies and NTDs in patients with ZIC3 mutations	Carrel et al. [2001]	Australian, British, Icelandic	3 multiplex families and 5 cases	NTD (myelomeningocele, anencephaly, encephalocele)	Screen for polymorphisms	No mutations or polymorphisms were identified	n/a	No evidence for association
				Zhu et al. [2003b]	Hispanic	35 cases and controls	NTD (spina bifida or anencephaly)	Screen for mutations or SNPS	No mutations or polymorphisms were identified in any of the three exons studied	n/a	No evidence for association

CANDIDATE GENES WITH CONFIRMED ASSOCIATIONS

Originally identified as a risk factor for vascular disease, the 677C → T thermolabile isoform of MTHFR has been associated with NTDs in some populations. MTHFR reduces 5,10-methylenetetrahydrofolate to 5-methylenetetrahydrofolate (the predominant circulatory form of folate) for use as a carbon donor for the re-methylation of homocysteine to methionine [Frosst et al., 1995]. Low MTHFR activity is associated with low plasma folate and high homocysteine levels. An early case-control study in 55 Dutch patients found a higher incidence of the 677C > T mutation in patients and parents of NTD cases than in the control samples [van der Put et al., 1995]. A much larger study of 218 Irish families supported this finding and also found a modest additional risk associated with the maternal TT genotype. “These results favor a biological model of MTHFR-related NTD pathogenesis in which suboptimal maternal folate status imposes biochemical stress on the developing embryo, a stress it is ill equipped to tolerate” if it is homozygous for the thermolabile mutation [Shields et al., 1999; Botto and Yang, 2000; Arole et al., 2003].

Other studies failed to find evidence of association with MTHFR in different populations [Rampersaud et al., 2003]. Those studies that did find an association could only attribute 11%–19% of NTDs to this gene [Ou et al., 1996]. The 677C > T variant of the enzyme has 50%–60% lower activity than the wild-type, however homozygotes with a good diet would have normal folate levels [Botto and Yang, 2000]. The thermolabile form of MTHFR could be a risk factor in populations with poor folate nutrition, which would explain the conflicting studies [Shields et al., 1999]. MTHFR contains another silent polymorphism, T1059C, that was associated with NTDs in a small Iowa subset of a larger study, but this association has not been confirmed in other studies and it is unclear what the functional significance of this variant would be. Studies of other folate metabolism genes have produced mixed results as well [Christensen et al., 1999; Wilson et al., 1999; Brody et al., 2002].

Several candidate genes from mouse models of NTDs have been studied in human populations. The T locus is the human homolog of the Brachyury gene in mice, which is vital to axial development and the formation of the posterior mesoderm. No evidence has been found for it being a major locus in human NTDs [Trembath et al., 1999; Speer et al., 2002]. An early study found limited evidence in a small association study, but it could only account for 6%–18% of NTD incidence [Morrison et al., 1998]. Splotch mice have a homozygous mutation in the Pax3 gene and exhibit NTDs, but no evidence has been found for this gene to be a major NTD risk factor in humans either [Trembath et al., 1999; Speer et al., 2002]. As Table III illustrates often compelling animal model candidates fail to be associated in human populations.

A few human NTD candidate gene studies have recently been published that look beyond the standard sources. Uncoupling Protein 2 (UCP2) functions in energy metabolism and was associated in a case-control study of a Californian population [Volcik et al., 2003]. A Vitamin B metabolizing enzyme, transcobalomin (TC) was not associated in a Dutch population, but did have significant results in a Southern Italian study [Afman et al., 2002; Gueant-Rodriguez et al., 2003]. Utilizing evidence from a chick model of NTDs involving cell adhesion molecules, there is recent evidence to support a role for neural cell adhesion molecule 1 (NCAM1) in the etiology of NTDs. While these findings may not hold up to future investigation, they delve into new sources of NTD candidate genes. Mouse models and folate metabolism will always provide new NTD candidates, but new research investigating different pathways and under utilized model organisms may provide keys to NTD research.

CONCLUSIONS

Candidate gene testing in human neural tubes has proved to have a burden of plenty—too many possible candidates exist and few that are studied have yielded positive results. Efficiently prioritizing these possibilities based on one source of data such as an animal model, metabolic pathway, or positional location is difficult, if not impossible. Incorporating several types of data may lead to a convergence of information, highlighting crucial regions or pathways. Comparing several model organisms will clarify which genes play a fundamental role in neural tube closure. Detailed clarification of the phenotype is of critical importance to these studies. The most conservative approach is to include only the narrowest group, but this limits the samples available to most studies. Increased sample numbers in candidate gene studies will directly increase their power to detect a significant association. While human NTD samples are difficult to obtain, redoubling efforts to ascertain more families from a variety of populations will enable genomic screens and candidate gene testing to be more effective in the future.