Mutations of ephrin-B1 (EFNB1), a marker of tissue boundary formation, cause craniofrontonasal syndrome

Materials and Methods

Mutation Screening of EFNB1 and X-Inactivation Analysis. The study was approved by the Oxfordshire Clinical Research Ethics Committee, and informed consent was obtained before performing sampling procedures. DNA was obtained from whole blood samples, cultured fibroblasts, lymphoblastoid cell lines, or cranial periosteum by phenol-chloroform extraction. The following GenBank accession nos. were used for EFNB1 primer design: cDNA, NM_004429; genomic, AL136092. Standard PCRs (30 μl) contained 15 mM Tris·HCl (pH 8.0), 50 mM KCl, 2.5 mM MgCl₂, 100 μM dNTP (each), 0.5 μM primers, 0.75 units of Amplitaq Gold polymerase (Applied Biosystems), and 0.15 units of Pwo polymerase (Roche). Amplicons were analyzed by denaturing HPLC (dHPLC) on a Wave 3500HT (Transgenomic, Omaha, NE). Primer sequences, PCR amplification, and dHPLC analysis conditions are provided in Table 2, which is published as supporting information on the PNAS web site. The coding regions of all five exons of EFNB1 were analyzed by dHPLC in every affected subject. Amplicons exhibiting aberrant migration patterns were sequenced by using BigDye (version 3) on a 3100 DNA sequencer (Applied Biosystems). Mutations were confirmed either by restriction enzyme digestion or by allele-specific oligonucleotide hybridization (see Table 3, which is published as supporting information on the PNAS web site). Correct biological relationships between children and their parents were inferred from the concordant segregation of at least eight microsatellites of 78% average heterozygosity, each located on different chromosomes. The minimum probability of correct paternity in these trios, based on the conservative assumption that the father's alleles corresponded to those of highest frequency in a search of The Genome Database (April 8, 2004; http://gdb.wehi.edu.au/gdb), was 99.1%.

In addition to the analysis of 20 CFNS families reported here, 155 further unrelated patients with craniosynostosis of unknown cause (mutations in FGFR1, FGFR2, FGFR3, TWIST, MSX2, and ALX4 having been excluded) were screened for mutations in EFNB1. Samples from males were doped with an approximately equal amount of PCR product from a control male before heteroduplexes were made. A C→T variant (dbSNP: rs2230423) was encountered in 16 of 230 chromosomes, but no pathogenic changes were found.

X-inactivation analysis of the androgen receptor (AR) locus was undertaken, as described (17, 18). The positive control sample was from a female with Melnick–Needles syndrome (18). Measurements were made in triplicate for samples from blood and in duplicate for samples from cranial periosteum.

RNA in Situ Hybridization. Whole-embryo in situ hybridization was performed as described (19). The Efnb1 probe (15) was characterized by sequence analysis. Expression was visualized by means of alkaline phosphatase-conjugated antidigoxygenin antibody. Embryos were subsequently embedded in Cryo-M-Bed compound (Bright Instruments, Huntingdon, England) and cut on a cryostat at 15 μm. Specimens were viewed and photographed by using a MP3 dissecting microscope (Wild, Heerbrugg, Switzerland) and a DMRBE light microscope with Photoautomat (Leica, Deerfield, IL).

Results

Heterozygous Mutations of EFNB1 in CFNS. We PCR amplified EFNB1, which comprises five exons and encodes a protein of 346 aa, and analyzed the gene in 24 affected females from 20 unrelated families. We identified pathogenic mutations in every case (Table 1). There were 17 distinct mutations, one of which was a frameshifting single-nucleotide deletion (Fig. 2a), the remainder being single-nucleotide substitutions that either alter a gt/ag splice site (three mutations), predict nonsense codons (two mutations) (Fig. 2b), or encode missense changes in the amino acid sequence (11 mutations). One mutation, 451G→A (encoding Gly151Ser) was recurrent (four families); this transition arises in a CpG doublet (Fig. 2c). Nine of these different mutations were shown to have arisen de novo from the unaffected biological parents (Table 1), demonstrating their pathogenic nature conclusively. The remaining mutations, for which both parents were not available, either grossly disrupt the EFNB1 sequence or, in the case of missense changes, occur at highly conserved positions (see below) and were not observed in 180 normal control chromosomes from the same population. Transmission of the mutation was observed in the four affected mother–daughter pairs available for analysis. We conclude that EFNB1 mutations cause CFNS, and we find no evidence of genetic heterogeneity in affected females.

Mutations Predict Complete or Partial Loss of Ephrin-B1 Function. The location of the mutations in ephrin-B1 in relation to known structural motifs and selected homologous sequences is shown in Fig. 3. The extracellular portion of murine ephrin-B2 consists of an eight-stranded β-barrel crystal structure with Greek key folding topology (20). The frameshift, splicing, and nonsense mutations truncate the ephrin-B1 protein in the extracellular region, suggesting complete loss of function, and the 1A→G (Met1Val) mutation should prevent normal translation initiation. The remaining missense mutations occur (with the exception of Ile62Thr, which arose de novo) at positions that are identical in the three human B-type ephrins, and they all are located in structurally important regions (see Fig. 3 legend) (21). Additional evidence that these are loss-of-function mutations is provided by genetic and biochemical data from the Caenorhabditis elegans ephrin-A homologue VAB-2. The VAB-2 mutations Met1Leu and Pro108Leu, which occur at positions equivalent to Met-1 and Pro-119 in human ephrin-B1 (Fig. 3), were described, respectively, as showing weak and intermediate defects in epithelial and neuronal morphogenesis (22). Loss of binding of VAB-2 to its Eph receptor VAB-1 was observed for the Pro108Leu substitution (22).

Although CFNS subjects differed in the extent of craniosynostosis and occurrence of additional clinical features, we did not observe any distinct genotype–phenotype correlation in the heterozygous females between truncating and missense mutations (Table 1), suggesting that all mutations cause a comparable disturbance of ephrin-B1 function. At this stage, we cannot exclude the possibility that a genotype–phenotype correlation exists for hemizygous males, in whom viability might depend on residual ephrin-B1 activity.

A Boundary of Efnb1 Expression Corresponds to the Position of the Future Coronal Suture. We examined the RNA expression of murine Efnb1 at embryonic day (E)9.5 and E10.5, the time at which the frontal-parietal boundary is established. Confirming and extending previous work (15), in situ hybridization shows that the most rostral Efnb1 expression domain corresponds specifically to the neural crest-derived, future frontal bone territory and underlying telencephalic vesicle neuroepithelium but is absent from the mesenchyme that is destined to form the parietal bone (Fig. 4).

X-Inactivation in CFNS Females. Expression data from somatic cell hybrids indicate that EFNB1 is normally subject to X-inactivation (24). We investigated X-inactivation in DNA extracted from whole blood in 18 affected females informative for the AR microsatellite assay, a widely used method for identifying skewed X-inactivation in X-linked disorders (17, 18, 25). We did not observe extreme skewing (>90:10) in any individual, and all but one sample showed skewing within the normal range (50:50 to <80:20) (Fig. 5). In four mother–daughter pairs, we could deduce (directly in the daughters and by virtue of close genetic linkage between AR and EFNB1, which are separated by only ≈1.3 Mb, in three of the mothers) whether the bias of X-inactivation was toward or against the mutant allele; in six of seven cases, there was a weak bias toward preferential inactivation of the mutant allele (Fig. 5). Because the X-inactivation status in blood is likely to differ from that in the coronal suture, we obtained samples of cranial periosteum overlying the entire length of both coronal sutures from a familial case (subject ID 3167) at the time of reconstructive surgery. The level of inactivation of the mutant X chromosome in eight evenly spaced specimens varied from 48.6% to 84.3% (data not shown). We conclude that normal ephrin-B1 function is not required for cell survival either in blood or in cranial periosteum.

Discussion

The identification of significant mutations in all 20 unrelated CFNS females studied, including nine different de novo mutations, provides convincing evidence that EFNB1 is the major CFNS locus. The diversity of the mutations, the disruptive nature of several of them, the location of missense mutations in regions of secondary structure or receptor binding, and the lack of genotype–phenotype correlation indicate reduction or loss of ephrin-B1 function. Our findings are consistent with recent linkage data for one family (12) but are difficult to reconcile with an earlier report that mapped CFNS to Xp22 (13). None of the patients that we investigated were included in either of these studies, so mutation analysis of EFNB1 in patients from the linkage panels should establish whether the previous findings are explained by genetic heterogeneity, coincidence, or other factors. Our findings are of particular interest in relation both to the pathogenesis of the component malformations of CFNS, especially craniosynostosis, and because of the apparent reversal in phenotypic severity between the sexes for an X-linked disorder.

Craniosynostosis affects ≈1 in 2,500 births and previously identified single-gene mutations in the FGFR1, FGFR2, FGFR3, TWIST, and MSX2 genes (26) account for ≈25% of cases overall (A.O.M.W. and S.A.W., unpublished data). Efforts to elucidate the pathogenesis of craniosynostosis in mice have focused pre-dominantly on the E14–E16 period, when the orthologous genes are expressed in the maturing cranial sutures (27, 28). CFNS characteristically affects the coronal suture, which separates the frontal and parietal bones (Fig. 1b); recent evidence from mice indicates that these bones have distinct embryological origins (frontal bone from neural crest and parietal bone from mesoderm) (23). This finding raises the question of what molecular signals mark the separate identities of these tissues at the time that the position of the future coronal suture is established. Ephrins are good candidates for such signals because the spatial distribution of their biological activity (within the A and B classes) is often complementary to their cognate Eph receptors (29), and this interaction appears to mediate repulsive signaling between distinct cell types (reviewed in refs. 30 and 31). The pattern of Efnb1 expression is exactly as expected for a molecule required for boundary formation at the coronal suture (compare Fig. 4c with Fig. 3D in ref. 23), and the demonstration of EFNB1 mutations in CFNS provides genetic confirmation of the importance of ephrin-B1 in this process. An important focus for future work may be to define further the molecular processes that lead to neural crest/mesoderm boundary formation and, hence, coronal suture positioning. Interestingly, there is evidence for direct interactions between the ephrin-B1 and fibroblast growth factor signaling pathways (32, 33), suggesting a possible mechanism by which boundary formation might lead to the onset of growth-related signaling in the coronal suture.

Why do mutations in EFNB1 apparently lead to craniosynostosis only in females? Although mice harboring the Efnb1 null mutation were reported to have shortened skulls (16), no information is available regarding the presence of craniosynostosis in these animals. More instructively, heterozygous female Efnb1^+/– mice exhibited additional limb phenotypes (preaxial polydactyly and syndactyly) that are not present in either Efnb1^– males or Efnb1^–/– females (16, 34). This appearance was demonstrated to correlate with a patchy distribution of Efnb1 expression in the limbs of Efnb1^+/– females in a pattern reciprocal to the receptors EphA4 (34), EphB2, and Ephb3 (16). Moreover, there was marked up-regulation of Efnb1 expression in regions where the X chromosome bearing the wild-type allele was active (16). Our finding that X-inactivation is not highly skewed in CFNS females, either in blood or cranial periosteum (Fig. 5), is consistent with observations in mice and leads us to propose that the craniosynostosis results from disturbance in the formation of the normally sharp neural crest/mesoderm tissue boundary at the future coronal suture, caused by patchy abnormalities of signaling arising from random X-inactivation and subsequent sorting of ephrin-B1 positive and negative cells (processes that would be specific to females). The observation that males inheriting the CFNS mutation do not have craniosynostosis (2, 3, 5, 11, 12) suggests functional redundancy in the mechanisms of cranial suture determination, perhaps involving other members of the ephrin/EphR signaling system (29).

Supplementary Material