Through multiple stages involving positional cloning, candidate gene sequencing and developmental gene expression analysis, we have identified FOXE1 as a major disease gene within the previously published 9q22–33 linkage peak. An initial panel of 300 SNPs narrowed the 20 Mb 2-LOD interval to 357 Kb. Fine mapping data that excludes the two genes (XPA and HEMGN) flanking this interval and expression data, negative for C9orf156 and positive for Foxe1, converge on FOXE1 as the etiologic gene.
Within a 200 kb critical region, a 20 kb disease haplotype with greater effect in Colombian and Scandinavian samples was identified that extends 19.2 kb upstream of the FOXE1
start codon (Fig. ). Although all populations analyzed in this study contributed to the FOXE1
association signal, SNP and haplotype frequencies found within each population indicate that this predisposing variant is more common in populations of Caucasian descent. This is in contrast to the IRF6
predisposing SNP rs642961 G/A (9
) that is more common in populations of Asian descent. Our data indicate two additional FOXE1 risk alleles. Independent signals from a haplotype spanning 95.5 kb starting 21.1 kb 3′ of FOXE1
in our Filipino population indicate an Asian-specific allele (Figs C and ). Previous reports of a translocation breakpoint 55 kb 3′ of FOXE1
in two siblings with bilateral cleft lip and palate lend credence to this hypothesis (33
). Filipinos have demonstrated similar evidence for two or more variants contributing to clefting at the IRF6
). Finally, the presence of a third allele in a separate LD block is supported by significant SNPs and haplotypes containing rs1443432 and rs780258 67 kb upstream of FOXE1
(Figs –). Finding multiple risk alleles are to be expected for a complex disease such as orofacial clefting.
The association patterns suggest the phenotypic spectrum of FOXE1 variants involves all combinations of primary and secondary palatal clefts. These findings differ from those of the IRF6 rs642961 variant disrupting an AP-2α-binding site that was not associated with CP in Filipinos. Furthermore, in contrast to IRF6, FOXE1 markers could not distinguish CLPO or CL/P from CLO in our populations. Both isolated cleft lip and isolated cleft palate can be separated from, or lumped together with patients having CLPO based on embryology, recurrence and syndromic gene findings. FOXE1 appears to present one unifying path for these phenotypes.
Of note is the maternal over-transmission of FOXE1
alleles in the Colombian and Filipino samples compared with all other Caucasian samples as well as the paternal over-transmission observed in families from the USA and Scandinavia. This may reflect a parent of origin effect such as imprinting, although there is no existing evidence to suggest imprinting for the FOXE1
locus. Preferential maternal over-transmission could also indicate a maternal genetic effect. There is dissimilarity between the maternal and paternal ancestral lineages of Colombian families with a predominance of Native American mitochondrial haplotypes (87.5%) and Caucasian Y chromosome haplotypes (57.4%), indicating admixture events between Caucasian immigrant men and Native American women (34
); a common finding amongst South American populations (35
). A Brazilian orofacial cleft study not only demonstrated similar differences between ancestral maternal and paternal contributions, but also showed significantly higher frequency of Native American mitochondrial and Y haplotypes in cases compared with controls (37
). Furthermore, evidence of genetic heterogeneity due to Native American ancestry was underscored by the presence of association between CL/P and IRF6
= 0.023) and of CL with RFC1
= 0.017) in non-carriers of the Native American haplogroup D (38
Similar parent of origin effects for orofacial clefting have been described for CBS
). Most interesting were the results reported by Shi et al
), indicating a maternal genetic effect for IRF6
with the fetal risk allele being protective when the mother is a risk allele carrier.
Only two FOXE1
coding mutations, I59S and P208R, were identified in our screen of over 200 patients. Neither mutation is present in 24 control individuals from this study or in 186 controls in a candidate gene mutation screen (15
). P208R is immediately adjacent to the A207V identified in a patient with CL/P (15
). I59S is located in the first alpha helix of the forkhead DNA-binding domain within which two mutations, S57N and A65V, have been described in families with features of the Bamforth–Lazarus syndrome (BLS) (46
). The S57N and A65V mutations have been shown to significantly reduce DNA binding and transcription activation. Furthermore, an I84S mutation at the corresponding FOXL2
amino acid as FOXE1
I59 has been reported in a family with blepharophimosis-ptosis-epicanthus inversus syndrome (48
). Therefore, the I59S mutation in our study may have similar effects on FOXE1
function. The lack of a more severe phenotype in our proband may be due to only one mutated FOXE1
copy, whereas BLS is recessive. The phenotypic spectrum of FOXE1
coding mutations may include features less severe than BLS, such as hypothyroidism. Of note, FOXE1
variants confer a 5.7-fold for thyroid cancer (49
The variants identified by sequencing the noncoding regions of FOXE1, PTCH, TGFBR1 GABABR2, ZNF189, FGD3 could presumably have etiologic effects. Although none were associated with CL/P, they occurred at conserved nucleotides and some were predicted to affect TFBSs, including SP1, AP1, AP3 and POU1F1 that are involved in facial development.
is a member of a transcription factor family that contains a DNA-binding forkhead domain and that are involved in embryonic pattern formation. Newborn mice null for Foxe1
exhibit cleft palate and thyroid anomalies (50
is expressed in the secondary palate epithelium in both mice at E13.5–E15.5 (51
) and humans at week 11 of gestation (52
). The specific expression pattern of Foxe1
at the point of fusion between maxillary and nasal processes, described for the first time in this study, strongly suggests FOXE1
as a key player in primary palatogenesis.
is a downstream target of the Shh/Gli
pathway in hair follicle morphogenesis (53
). Thus, it is plausible that there are two disease genes in the 9q22–q33 region since the SHH
was associated with CL/P in our Filipino data set and there is ample evidence that SHH
signaling is involved during primary palatogenesis (54
). Furthermore, there is overlapping Foxe1
expression in the caudal epithelium of the mesial nasal and maxillary processes.
The paucity of FOXE1
mutations despite exhaustive searches (15
) and in light of the highly significant association described here indicate that causal mutations are in nearby noncoding regions that regulate FOXE1
expression, most likely within the 70 kb high LD block containing the 5′-UTR and promoter. Recently, a variant affecting MYF-5 DNA binding in a cluster of GLI
-binding sites 1.2 kb upstream of the transcription start site has been described in 11of 25 patients with NSCLP (55
). We found this variant in 9 of 184 case and 2 of 186 control chromosomes. While this is coincident with etiologic effects, this variant is too infrequent to explain the strong association results unless it splits a common haplotype into two disease alleles.
Efforts to find other causal variants will require in depth sequencing and characterization of regulatory function in highly conserved regions within our putative disease haplotypes. Utilizing the Vertebrate Multiz Alignment (28 species) and thePhastCons track of the UCSC browser (56
), we performed a preliminary screening of the 200 kb critical region with a set threshold score above 350. Fourteen regions with high PhastCons scores (Table ) were identified. One region that contains the most significantly associated SNP also harbors five GLI2
-binding sites (57
). The first of these GLI2-binding sites is located 20 bp downstream of rs3758249, suggesting further a connection of FOXE1
to the SHH/GLI
pathway for primary palatogenesis.
Regions harboring potentially regulatory elements
In the aggregate, our findings place FOXE1 as the third locus along with IRF6 and 8q24 in which common variants have a substantial impact on the occurrence of cleft lip and palate. It affords a new pathway to investigate biologically and new tools to improve genetic counseling.