Novel evidence of association with NSCL/P was shown for SNPs in FOXF2 gene in an Asian population

Abstract

Background

The forkhead box F2 gene (FOXF2) located in chromosome 6p25.3 has been shown to play a crucial role in palatal development in mouse and rat models. To date, no evidence of linkage or association has been reported for this gene in humans with oral clefts.

Methods

Allelic transmission disequilibrium tests were used to robustly assess evidence of linkage and association with nonsyndromic cleft lip with or without cleft palate (NSCL/P) for 9 SNPs in and around FOXF2 in both Asian and European trios using PLINK.

Results

Statistically significant evidence of linkage and association was shown for two SNPs (rs1711968, and rs732835) in 216 Asian trios where the empiric P values with permutation tests were 0.0016 and 0.005, respectively. The corresponding estimated odds ratios for carrying the minor allele at these SNPs were 2.05 (95%CI=1.41, 2.98) and 1.77 (95%CI=1.26, 2.49), respectively.

Conclusions

Our results provided statistical evidence of linkage and association between FOXF2 and NSCL/P.

INTRODUCTION

Oral clefts (which include cleft lip, cleft lip and palate, and cleft palate) represent a collection of common birth defects that severely influence quality of life of the affected and their families (Wehby et al., 2010). There is strong evidence for some genetic etiology for oral clefts from many family, twin and population studies (Marazita, 2012), and a large number of candidate genes have been suggested as important in the etiology of these complex and heterogeneous malformations (Jugessur et al., 2009; Dixon et al., 2011). Genome-wide association studies (GWAS) have confirmed at least a dozen genes playing some role in the etiology of oral clefts in recent years (Birnbaum et al., 2009; Beaty et al., 2010; Mangold et al., 2010; Grant et al., 2009), although specific causal variants and biological mechanisms remain unclear.

The human forkhead box F2 gene (FOXF2), located in chromosome 6p25.3, has been shown to play a crucial role in palatal development in mouse models (Wang et al., 2003). However, no evidence of linkage or association has been reported for this gene in humans including linkage studies of multiplex NSCL/P families recruited from various populations (Marazita et al., 2004 &2009), two candidate gene studies of population-based Scandinavia NSCL/P case-parent trios (Jugessur et al., 2009&2010) and several GWAS (Birnbaum et al., 2009; Beaty et al., 2010; Mangold et al., 2010; Grant et al., 2009).

To clarify the potential role that FOXF2 may play in the etiology of this common and complex disorder in humans, we performed allelic transmission disequilibrium tests (TDT) for NSCL/P with markers in and near FOXF2 gene using both Asian and European case-parent trios. Our results provide statistical evidence of association between markers in FOXF2 and NSCL/P in Asian trios, and provide new insight to the etiology of this birth defect.

MATERIALS AND METHODS

Ethics statement

Institutional review boards (IRB) or ethical committees from the following institutions reviewed and approved the study: KK Women’s and Children’s Hospital in Singapore, National University of Singapore, Chang Gung Memorial Hospital in Taiwan, Yonsei University in South Korea, and Johns Hopkins University in the US. Written informed consent was obtained from adult participants (including biological parents of all probands and probands old enough to give their own consent/assent) and parents or guardians of the minor participants.

Sample description

This study includes 216 Asian NSCL/P probands and their biological parents, recruited from cleft lip and palate treatment centers in Taiwan, Singapore, Korea, and 75 European NSCL/P trios, recruited from Maryland (US) and Singapore (n=1) (Table 1). All probands received clinical genetic assessment by a health professional to check for other birth defects or developmental delays and were diagnosed as having a non-syndromic, isolated oral cleft.

Table 1.

Race and gender for 297 NSCL/P trios

Race&Site	NSCL/P
	Male	Female	Total
Asian
Korea	22	18	40
Singapore	21	9	30
Taiwan	95	51	146
Subtotal	138	78	216
European
Maryland	43	30	73
Singapore	1	1	2
Subtotal	44	31	75
AA
Maryland	1	2	3
Other
Singapore	2	1	3
Total	185	112	297

AA: African American

SNP selection, DNA, and genotyping

Nine SNPs spanning 7 691 base pairs (bps) in physical distance were genotyped in and near FOXF2 for the current study. SNPs were chosen with a goal of identifying an average of one SNP per 5 kilo bps of physical distance. The physical distance between adjacent genotyped SNPs ranged from 56 to 2909 bps. The target region, represented by the 9 genotyped SNPs, covered the whole gene from its 5′end (rs1711971) to its 3′ end (rs1737762) involving all the two exons (exon 1: rs11759800; exon 2:rs2293783) and the only intron (rs3799321, rs1711968, rs9405473, rs732835, and rs3823148) basing on variant NM_1400521 (Fig. 1, http://www.ncbi.nlm.nih.gov/gene/2295). SNPs with “SNP scores” >0.6, high validation levels in dbSNP, and high heterozygosity levels were given priority in selection. Three (rs1711968, rs732835, and rs2293783) of the four SNPs released by the HapMap project (HapMap Data Rel 24/phaseII Nov08, on NCBI B36 assembly, dbSNP b126) were included in the marker panel for the current study. Among these 9 SNPs, rs1711971 was the only SNP genotyped for both the current study and the published GWAS panel (Birnbaum et al., 2009; Beaty et al., 2010; Mangold et al., 2010; Grant et al., 2009).

Fig. 1.

Schematic genomic structure of FOXF2 and coordinates of 9 SNPs genotyped in this study and SNPs representing those genotyped using Illumina 610-Chips for NSCL/P GWAS. Variant NM_001452.1 of the human FOXF2 gene is shown in the figure with 2 exons aligned from left to right. Filled black boxes represent exons. Scale on top of the figure shows position of the gene and coordinates for the 9 SNPs (black) genotyped in this study and some of the 32 SNPs (gray) genotyped using Illumina 610-Chips in NSCL/P GWAS basing on GRCh build 38. There are 5 SNPs (rs1737753, rs10498654, rs1711969, rs1878475 and rs1711970) genotyped between rs12524544 and rs1711971 at the 5′ of the gene, and there are another 24 SNPs (rs4959557, rs1711973, rs746095, rs731394, rs9501718, rs9328051, rs11753773, rs9378623, rs1711959, rs1711960, rs9392288, rs1555110, rs1737786, rs9502924, rs11242687, rs9405475, rs2493158, rs9502928, rs1922932, rs10458124, rs7774941, rs17202895, rs9328053, rs932410) genotyped between rs6596817 and rs2816251 at the 3′ of the gene in the NSCL/P GWAS panel.

Genomic DNA was extracted from peripheral blood using protein precipitation method and stored at −20°C. Aliquots (4μg) of each genomic DNA sample were dispensed into a bar-coded 96-well microtiter plate at a concentration of 100ng/μl and genotyped by Illumina’s GoldenGate chemistry (Oliphant et al., 2002) at the Genetic Resources Core Facility (GRCF) of Johns Hopkins University. Two duplicates and four controls from the Centre d’Etude du Polymorphisme Humain (CEPH) collection were included in each plate to evaluate genotyping consistency within and between plates.

Statistical analysis

(1) SNP screening and preliminary analysis

We first screened SNPs for minor allele frequency (MAF) and Hardy-Weinberg equilibrium (HWE) using parents’ genotypes, for Asian and European trios, respectively. SNPs with MAF>1% and showing adherence to HWE (P>0.01) were eligible for association analysis. This screening procedure resulted in 6 and 5 eligible SNPs in Asian and European NSCL/P trios, respectively. Further, one additional SNP was excluded due to very low number of informative families in European trios. As a result, six SNPs in Asian and four SNPs in European trios were finally used in association analysis. SNP screening analysis and pairwise linkage disequilibrium (LD, measured as r²) estimation for eligible SNPs were all carried out using Haploview (v4.2, http://www.broadinstitute.org/haploview/haploview) (Barrett et al., 2005).

In addition, data from the 1000 genome project (Genomes Project Consortium, 2010) for Asian samples (Han Chinese in Beijing, CHB; Han Chinese South, CHS; Japanese individuals, JPT; as well as the combined Asian sample, Asian), and the CEPH collection were used to estimate LD patterns among SNPs genotyped for the current study and those genotyped in the GWAS panel using Haploview.

(2) Association analysis

We performed an allelic TDT analysis, originally proposed by Spielman et al., (1993) on each eligible marker using PLINK (v1.07; http://pngu.mgh.harvard.edu/purcell/plink/) (Purcell et al., 2007) and used Family Based Association Test (FBAT; http://www.biostat.harvard.edu/~fbat/fbat.htm) to calculate the number of informative families (Laird et al., 2006). In addition to the Bonferroni correction to each P-value generated from allelic TDT analysis, an overall empiric P value for each SNP was also generated from 10,000 permutations in Haploview to correct for multiple comparisons.

RESULTS

Preliminary analysis

Three and four monomorphic SNPs among the 9 genotyped SNPs in or near FOXF2 for the current study were dropped from association analysis in the Asian NSCL/P trios (rs1711971, rs1737762 and rs11759800) and in the European NSCL/P trios (rs11759800, rs3799321, rs9405473 and rs3823148), respectively. The number of informative European trios (n=3) for rs1711971 was too low for meaningful association analysis, and was excluded as well. The remaining 6 SNPs in Asian parents and 4 SNPs in European parents were all compatible with HWE (at p>0.01). Genotyping call rates in Asian and European founders were no lower than 98.3% and 97.8%, respectively (Table 2).

Table 2.

MAF for SNPs in and near FOXF2 gene in Asian and European NSCL/P trios

SNPs	Position build 38	Asian				European
		HW pval	% Geno	MAF (%)	Minor allele	HW pval	% Geno	MAF (%)	Minor allele
		rs1711971	1388187	1.00	100.0	0.0	T	1.00	99.3	2.6	T
rs11759800	1391096	1.00	100.0	0.0	G	1.00	100.0	0.0	G
rs3799321	1391817	0.99	98.3	4.1	C	1.00	74.6	0.0	C
rs1711968	1391873	0.12	99.8	22.4	C	0.24	98.6	45.6	C
rs9405473	1392224	1.00	100.0	2.9	T	1.00	100.0	0.0	T
rs732835	1392765	0.40	99.8	26.0	C	0.05	97.8	20.7	C
rs3823148	1392958	1.00	99.8	2.6	G	1.00	99.3	0.0	G
rs2293783	1394808	0.62	99.8	23.8	C	0.36	100.0	22.1	C
rs1737762	1395878	1.00	99.3	0.0	A	1.00	97.8	25.2	A

Among the Asian founders, SNPs rs1711968, rs9405473 and rs732835 formed the only linkage disequilibrium (LD) block where rs1711968 and rs732835 were in strong LD with r²= 0.72. Among European founders, the only LD block was formed by SNPs rs732835, rs2293783 and rs1737762 where rs1711968 was independent and r² between rs1711968 and rs732835 was 0.31 (Fig. 2).

Fig. 2.

The LD patterns in founders for Asian (left) and European (right) NSCL/P probands. Numbers shown on the figure represent r² between SNPs.

No evidence of strong LD was shown for SNPs eligible for the current study with SNPs included in the GWAS panel (r2 ranged from 0 to 34) using data from 1000 genomes for CHB, CHS, JPT, and the combined Asian group (Fig. S1).

Association analysis

Statistically significant linkage and association was shown for three SNPs (rs1711968, rs732835 and rs2293783) in the NSCL/P Asian trios by allelic TDT analysis for each marker individually using PLINK. The corresponding Bonferroni corrected P-values were 0.00072, 0.0043 and 0.032. The overall empiric P values for these three SNPs over 10,000 permutations (using Haploview) were 0.0016, 0.005 and 0.054, respectively. The corresponding estimated odds ratios (OR) for carrying the minor allele for rs1711968 and rs732835 were 2.05 (1.41, 2.98) and 1.77 (1.26, 2.49), respectively (Table 3). Significance held for rs1711968 and rs732835 for Taiwan trios (Table S1).

Table 3.

TDT analysis for SNPs in and near FOXF2 gene in 216 Asian and 75 European NSCL/P trios

SNPs	Position build 38	Minor allele	MAF (%)	FAM	T	NT	OR (95%CI)	P Value	Bonferroni P Value	Empiric P Value
Asian
rs3799321	1391817	C	4.1	32	18	16	1.13 (0.57, 2.21)	0.73	1	1
rs1711968	1391873	C	22.4	101	84	41	2.05 (1.41, 2.98)	0.00012	0.00072	0.0016
rs9405473	1392224	T	2.9	21	14	8	1.75 (0.73, 4.17)	0.2	1	0.85
rs732835	1392765	C	26.0	119	92	52	1.77 (1.26, 2.49)	0.00086	0.0043	0.005
rs3823148	1392958	G	2.6	19	13	7	1.86 (0.74, 4.66)	0.18	1	0.75
rs2293783	1394808	C	23.8	114	85	53	1.60 (1.14, 2.26)	0.00645	0.032	0.054
European
rs1711968	1391873	C	45.6	32	17	22	0.77 (0.41, 1.46)	0.42	1	0.98
rs732835	1392765	C	20.7	25	9	20	0.45 (0.20, 0.99)	0.041	0.16	0.063
rs2293783	1394808	C	22.1	27	10	23	0.43 (0.21, 0.91)	0.024	0.1	0.093
rs1737762	1395878	A	25.2	27	20	14	1.43 (0.72, 2.83)	0.3	1	0.95

^*FAM: number of informative families, T: number of transmitted alleles, NT: number of un-transmitted alleles; Empiric P Value: P values for 10,000 permutations in Haploview

Among the European trios, nominally significant evidence in the allelic TDT analysis was seen for two adjacent SNPs (rs732835 and rs2293783) which were in strong LD with one another (r ² = 0.87). However, after Bonferroni correction they were no longer significant (Table 3).

DISCUSSION

The FOXF2 gene is one of the important candidate genes for oral clefts. It has 1 intron and 2 exons and encodes a transcription factor protein which belongs to a large family of evolutionarily conserved DNA-binding proteins (http://genome.ucsc.edu/cgi-bin) (Clark et al., 1993). It was found to be expressed in developing palate, tongue, oral cavity as well as other organs by in situ hybridization in mouse (Wang et al., 2003; Aitola et al., 2000) and rat (Aitola et al., 2000). The embryonic expression levels in FOXF2-deficient homozygous mutant mouse were the same in the wild-type mouse for several other candidate genes for oral clefts including MSX1, MSX2, TBX1, FOXC2 (MFH1), and FOXF1. All the 34 examined FOXF2-deficient mice were born with cleft palate without any combination of other craniofacial malformations (Wang et al., 2003). In addition, heterozygous DNA sequence variations in FOXF2 were identified in 2 of the 18 patients with both disorder of sex development and cleft palate but not in any of the 20 normal controls (Jochumsen et al., 2008).

In the current analysis, statistically significant evidence of linkage and association was shown for NSCL/P with two adjacent intronic SNPs (rs1711968 and rs732835) in FOXF2 in Asian case-parent trios after correction for multiple comparison. These statistical results are consistent with the findings in animal models and provide further support for FOXF2 as a potential susceptibility gene for NSCL/P in humans. The failure to replicate these significant findings in our European trios may be explained by etiologic heterogeneity, the small sample size available for analysis or missing of important SNPs in the marker panel.

In analyses of Scandinavian nationwide population-based NSCL/P and NSCP case-parent trios, none of the three tested SNPs (rs1711970, rs732835, rs2293783) in or near FOXF2 in that study showed evidence of linkage or association with either NSCL/P or NSCP (Jugessur et al., 2009&2010). However, our significant findings for rs732835 in Asians do not necessarily conflict with this Scandinavian study because different population origins may well explain the inconsistency.

Significant linkage or association identified with rs1711968 for Asian NSCL/P in the current study does not conflict with the published NSCL/P GWAS either (Birnbaum et al., 2009; Beaty et al., 2010; Mangold et al., 2010; Grant et al., 2009). In one of these GWAS for NSCL/P, rs1711971 is the only SNP tested in FOXF2 (Grant et al., 2009). For the other 3 GWAS for NSCL/P, rs1711971 is the only SNP in common (Birnbaum et al., 2009; Beaty et al., 2010; Mangold et al., 2010) in FOXF2 with those in the marker panel for the current study. None of the other 32 SNPs genotyped in the GWAS panel for FOXF2 fell into the 7 691 base pair region where the 9 SNPs tested in this study are located. Also, because neither of the statistically significant SNPs (rs1711968 and rs732835) identified in the current study has been included in the previous NSCL/P GWAS, and they are unlikely in tight LD with SNPs genotyped in the GWAS panel, our study can be seen as an independent analysis although 58.3% (126 trios) of our Asian sample went into one of these GWAS where only one SNP (rs11753773, P=0.01) reached nominal significance of linkage or association with NSCL/P (Beaty et al., 2010). Because rs1711968 is a common SNP and located in the intronic region of the gene, the statistically significant association seen between NSCL/P and rs1711968 in our Asian trios could possibly be explained by LD with an unobserved causal variant.

This study reported the first significant linkage and association between markers in FOXF2 and NSCL/P in humans, building upon previous experimental animal studies. Further exploration is warranted to replicate these findings in other populations.

Supplementary Material

Supp Fig S1.1

Fig. S1 The LD patterns for CHB, CHS, JPT, combined Asian group, and CEPH using data from1000 genome project involving SNPs in FOXF2 genotyped for the current study and for the NSCL/P GWAS using Illumina 610-Chips. Numbers shown on the figure represent r² between SNPs.