Proteins of the bone morphogenetic protein (BMP) family are known to have a role in ocular and skeletal development; however, because of their widespread expression and functional redundancy, less progress has been made identifying the roles of individual BMPs in human disease. We identified seven heterozygous mutations in growth differentiation factor 6 (GDF6), a member of the BMP family, in patients with both ocular and vertebral anomalies, characterized their effects with a SOX9-reporter assay and western analysis, and demonstrated comparable phenotypes in model organisms with reduced Gdf6 function. We observed a spectrum of ocular and skeletal anomalies in morphant zebrafish, the latter encompassing defective tail formation and altered expression of somite markers noggin1 and noggin2. Gdf6+/− mice exhibited variable ocular phenotypes compatible with phenotypes observed in patients and zebrafish. Key differences evident between patients and animal models included pleiotropic effects, variable expressivity and incomplete penetrance. These data establish the important role of this determinant in ocular and vertebral development, demonstrate the complex genetic inheritance of these phenotypes, and further understanding of BMP function and its contributions to human disease.
The Growth Differentiation Factors (GDFs) are members of the bone morphogenetic proteins (BMP) sub-family of transforming growth factor-beta (TGF-β) signaling ligands, known to regulate patterning during development (1). We previously demonstrated GDF6's role in ocular development by characterizing a GDF6-encompassing copy number variant in a patient with ocular colobomata, and recapitulating the human phenotype in gdf6a morphant zebrafish (2). Such results, combined with comparable findings in Xenopus (3), suggested that GDF6 mutations may underlie a range of ocular phenotypes, including one or more components of the microphthalmia, anophthalmia or coloboma (MAC) developmental spectrum. In addition, in keeping with the broad and incompletely defined roles of BMPs, it was probable that GDF6 mutations contributed to other human disorders.
The BMPs were originally identified through their capacity to induce bone formation (4,5) and are now recognized to have critical roles patterning a diverse range of tissues including bone, heart, lungs and kidney (6–10). Over the last 10 years significant progress characterizing the phenotypes resulting from mutation of BMPs or their receptors has ascribed human disease phenotypes to one-quarter of the BMP family (Online Mendelian Inheritance of Man, http://www.ncbi.nlm.nih.gov/sites/entrez?db=omim), predominantly type I or type II BMP receptors [ACVR1 (11), ACVRL1 (12), BMPR1A (13), TGFBR1 (14), BMPR1B (15), BMPRII (16)] rather than signaling ligands [BMP4 (17), BMP15 (18,19), GDF5 (20–24)]. This distribution likely reflects the challenges discerning human disease phenotypes for ligands that exhibit functional redundancy (25). The contrast between the limited human phenotypes of individual ligands compared with the broad expression pattern and expansive range of phenotypes associated with BMP loss in model organisms implies that only a fraction of the human phenotypes attributable to each have been identified. This is clearly illustrated by the data from BMP4, where human mutations are primarily associated with microphthalmia (17), while murine loss of function causes defects in multiple systems: cardiovascular (26,27), craniofacial (28), ocular (25,29), reproductive, limbs, digits (30) and auditory (31). Increasing the proportion of BMPs with a defined function is thus of dual clinical and scientific importance—advancing understanding of the molecular basis of human disease, and in view of the disparate early and late-onset functions of individual BMPs (20,23,24,32–34), offering potential to inform understanding of multiple aspects of human genetics.
Accordingly, we set out to define GDF6's function in greater detail and present data demonstrating a broad role in human disease. At the outset of our studies, two lines of evidence indicated that, like other BMPs, GDF6 at a minimum possessed an appreciable skeletal developmental role. The first was data from a murine model demonstrating Gdf6's regulation of murine skull and carpal/tarsal joint formation (35), while the second, linkage mapping of two skeletal disorders [Klippel-Feil (KF) syndrome (36) and Split hand and foot malformation (SHFM) (37)] to the vicinity of GDF6 (Fig. 1A), was also compatible with a skeletal function. KF syndrome, characterized by defective cervical, thoracic or lumbar vertebral segmentation (38), is frequently associated with other skeletal (scoliosis, rib abnormalities, Sprengel's deformity) and non-skeletal (deafness, cardiovascular, ocular and renal) anomalies (39–41). Some of these (e.g. deafness and rib anomalies) correlate with the extensive expression pattern of Gdf6 (35). Furthermore, existence of a KF subtype with an ocular phenotype [Wildervanck syndrome (42)] is compatible with the possibility that GDF6 causes both ocular and skeletal disease. The mapping of the sixth SHFM (43) locus to a large GDF6-encompassing interval (37) raised the possibility that this gene might underlie this skeletal phenotype.
In light of these factors supporting GDF6 involvement in human ocular and skeletal disorders, we screened patients with ocular and skeletal developmental anomalies for GDF6 mutations. We investigated the biological significance of a representative subset of the mutations identified with a reporter gene assay and western blot analysis, and characterized the effects of perturbed Gdf6 function in two model organisms (murine and zebrafish). Our results indicate that GDF6 mutations result in variable ocular and skeletal phenotypes with evidence of non-Mendelian inheritance, according closely with comparable features in animal models of decreased Gdf6 function. Such findings have implications for deciphering interactions between members of the BMP family and provide a model for studying the contribution of complex inheritance patterns to human disease (44,45).
Sequencing and mutation analysis
DNA samples from 489 patients with ocular anomalies (micropththalmia, anophthalmia and coloboma), and 81 patients with vertebral segmentation anomalies were screened for GDF6 mutations by sequencing amplicons encompassing the two exons and splice sites. A subset of 32 samples were screened for copy number alteration by real-time quantitative PCR (qPCR) (TaqMan®), yielding normal results. Hybridization of a DNA sample from the proband of the 8q21.11-q22.3 linked SHFM pedigree to a CGH array with a mean 6 kb oligonucleotide probe spacing (46), identified no copy number variations in the 3 Mb region encompassing GDF6.
Sequencing identified two heterozygous sequence changes in exon 1 (125 g→t; 356a→g) and five heterozygous changes (647 g→a, 746c→a, 758a→t, 980c→a and 1271a→g) in exon 2 of GDF6. These sequence alterations, which were absent from dbSNP (http://www.ncbi.nlm.nih.gov/SNP/index.html) and a minimum of 366 control chromosomes (Table 1), result in the following amino acid alterations: Gly42Val (G42V), Gln119Arg (Q119R), Asp216Gly (D216G), Ala249Glu (A249E), Gln253Leu (Q253L), Pro327His (P327H) and Lys424Arg (K424R) (Fig. 1B). Four amino acid alterations (Q119R, D216G, Q253L and P32H) were associated with ocular phenotypes, two with skeletal phenotypes (G42V and K424R), while one alteration (A249E) identified in three probands, was associated with either ocular or skeletal phenotypes (Table 1).
|Proband||Mutation||Phenotype||Penetrance||Enzyme (no. of controls)|
|1||A249E||Coloboma||Post-axial polydactyly||Incomplete||BsrBI (0/646)|
|3||A249E||None||Klippel-Feil||Full (50)||BsrBI (0/646)|
|8||G42V||None||Spondylothoracic dysostosis||–||SphI (0/366)|
|Proband||Mutation||Phenotype||Penetrance||Enzyme (no. of controls)|
|1||A249E||Coloboma||Post-axial polydactyly||Incomplete||BsrBI (0/646)|
|3||A249E||None||Klippel-Feil||Full (50)||BsrBI (0/646)|
|8||G42V||None||Spondylothoracic dysostosis||–||SphI (0/366)|
‘–’ Denotes unknown because of the nature of DNA collection from which samples were derived.
The seven residues affected vary in their degree of conservation, with two conserved in all vertebrates (K424R and D216G), four conserved in mammals (G42V, Q119R, Q253L, and P327H), while A249E is the least conserved (Fig. 1B). Derivation of the DNA samples from historical collections, some established more than a decade ago, limited the ability to recall family members. However, for three of the seven amino acid alterations additional DNA samples could be obtained, allowing segregation to be assessed. In each of these cases, A249E (proband 1), Q253L (proband 5) and P327H (proband 6), the presence of the alteration in an unaffected parent indicates incomplete penetrance (Fig. 1G). Sequence and structure analyses demonstrate that K424R affects the mature TGF-β domain that is invariant in vertebrates (Fig. 1B) and maps to the convex surface of the cystine knot, which is the region of interaction with ligands and soluble inhibitors. Activin A, another member of the BMP superfamily, interacts with the minimal inhibitory module of follistatin through multiple contacts mapped to this surface, with charged residues in both molecules forming an extensive network of hydrogen bonds (47). Replacement of this lysine with a larger arginine residue increases the similarity of GDF6 to GDF5 and GDF7 (members of the same clade as GDF6) that have an arginine in this position, and may potentially decrease the specificity of the interaction with GDF6's receptor(s). Similarly, the residue altered by D216G is invariant in 15 of the 19 other BMPs [including GDF5 and GDF7 (Fig. 1B), as well as GDF 8, 9, 11; BMP 2, 4–8, 10, 15 (ClustalW alignment, data not shown)]. Of the mutations identified in this study, the A249E alteration affects the least conserved residue and is located in the TGF-β prodomain, a region thought to facilitate correct folding of the mature secreted peptide (48). A249E was identified in three probands from geographically disparate countries [Table 1; (49,50)] and genotyping with eight microsatellites and six single nucleotide polymorphism (SNPs) of which four were informative (Fig. 1H), demonstrated that at least two of these individuals carry different haplotypes (Fig. 1I).
Reporter gene and protein expression assays
To characterize the functional consequences of the amino acid changes, two alterations (A249E and K424R) were selected that lie in different GDF6 domains (Fig. 1B). A functional assay employing a SOX9-responsive reporter was used to evaluate BMP signaling. SOX9 plays a central role in chondrogenesis, and as its expression/activity is exquisitely sensitive to changes in BMP/GDF activity, this reporter gene provides a reliable read-out on the status of BMP/GDF signaling (51,52). Accordingly, to assess chondrogenic potential, expression constructs for GDF6 and mutants A249E and K424R were co-transfected with a SOX9-responsive reporter gene into primary limb mesenchymal (PLM) cells and the activities of each determined in replicate experiments. Consistent with the role of GDF6 in skeletal development (35), expression of wild-type GDF6 led to 3.4-fold increased SOX9 reporter activity. In contrast, expression of GDF6–A249E and GDF6–K424R constructs resulted in 2.9-fold and 2.4-fold increases in reporter gene activity (P < 0.034 and P < 0.002; t-test) (Fig. 2A). Such diminished activation of the reporter by the mutant GDF6 constructs provides evidence that these mutations alter GDF6 function.
Western blot analysis of the whole cell lysates and media from wild-type and mutant GDF6 was next undertaken to determine the effects of the two representative mutations on protein expression and secretion. GDF6 cDNA was mutagenized for the two mutations (A249E and K424R), tagged with a V5 epitope, and transiently transfected into COS7 cells. Proteins collected from the media and whole cell lysates were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, transferred to membranes and probed with V5 antibody. The western blot revealed the presence of pro-GDF6 (55 kDa) in both media and cell lysates, with the mature GDF6 ligand (approximately 16 kDa) present only in the media as a doublet (Fig. 2B). Although neither mutant constructs exhibited abnormal formation of pro-GDF6 nor mature GDF6 protein, the level of secreted mature GDF6 was reduced with A249E (23%) and K424R (83%) mutants compared with wild-type [ImageJ (53)] (Fig. 2B). Inclusion of α-tubulin (whole cell lysate) and secreted alkaline phosphatase (supernatants) provided controls for consistent loading and secretion (Fig. 2B). Co-transfection of the GDF6–V5 constructs with β-galactosidase enzyme construct and subsequent assay for β-galactosidase activity (Promega) revealed equal transcription efficiency between wild-type and mutant V5-tagged GDF6 (data not shown).
In light of the variable human phenotypes observed with GDF6 mutations (Fig. 1C–F, Table 1) and the incomplete penetrance evident with three of seven mutations (Fig. 1G), we studied a gdf6a zebrafish model to determine if comparable features were present. gdf6a function was inhibited using morpholino antisense oligonucleotides (MO). The first (gdf6aMO1) targets the 5′-splice site, while the second (gdf6aMO2) targets the intron 1–exon 2 boundary (2). One cell zebrafish embryos were injected with gdf6aMO1 or gdf6aMO2 at a concentration of 5–10 mg/ml (2,54) together with 2.5 mg/ml targeting the translation start site of p53 (p53MO) (55) to reduce apoptotic cell death (56). The prevalence of skeletal anomalies (curled or kinked tails) was much lower than that of ocular anomalies (coloboma and microphthalmia) at each morpholino dose (5 or 10 ng) and time-point studied (48, 72, 96 hpf). For instance, at a 5 ng gdf6a MO dose, 17% of 96 hpf morphants exhibited skeletal anomalies compared to 82% with ocular anomalies (Fig. 3C) (n = 96, χ2, P < 0.001). Notably, a dose-dependent effect was seen with both phenotypes and morpholinos. To provide an additional control, a mismatch (mm) morpholino variant of gdf6aMO1 containing five nucleotide substitutions was injected at the same concentrations as gdf6aMO1 and gdf6aMO2, without yielding ocular or skeletal phenotypes (Fig. 3A and B).
In view of the markedly different prevalences of gdf6a MO-induced skeletal and ocular anomalies, and the incomplete penetrance observed with A249E, Q253L and P327H, we next investigated the level of gdf6a mRNA in phenotypically normal zebrafish morphants. Wild-type and morphant gdf6a zebrafish of 48 hpf were collected and divided into three groups (wild-type, morphants exhibiting a phenotype, and phenotypically normal morphants) based on objective microscopic appearance. Fifty embryos from each group were pooled, RNA extracted, cDNA prepared and gdf6a mRNA variants amplified as described elsewhere (2). Observation of appreciable reductions in the level of correctly spliced gdf6a mRNA in phenotypically unaffected morphant embryo pools (Fig. 3D) demonstrate that zebrafish with reduced levels of gdf6a mRNA may appear phenotypically normal.
Further characterization of morpholino-induced zebrafish skeletal phenotypes was undertaken to document the axial skeletal changes in gdf6a morphants and permit study of genes whose expression may be regulated by GDF6 mutation. At 48 hpf, morphants exhibited kinked (mild), bent (moderate) and curly (severe) tail phenotypes, compared with the straight tails of wild-type zebrafish (Fig. 3E–H). Whole-mount in situ hybridization were undertaken on wild-type and gdf6a morphant zebrafish (at 10 somites, 18 hpf and 2 dpf) using digoxigenin-labeled antisense RNA probes to somite markers myod, her7 and unc45; paralog gdf5; and BMP antagonists, gremlin1, noggin1 and noggin2. No significant differences in the expression of myod, her7 and unc45 were observed (data not shown), however, expression of gdf5 was reduced in the developing axial jaw and gill cartilage elements in gdf6a morphants (Fig. 3I and J) with gdf6a not required for gdf5 expression in lateral jaw cartilage elements. Expression of noggin1 and noggin2 (57), were reduced in the newly formed caudal somites of gdf6a morphants (92% [23/25]) and the ventral aspects of the somites of gdf6a morphants (94% [17/18]), respectively (Fig. 3K–N). In situ observations were validated by qPCR of noggin1 and noggin2; additional comparisons performed with noggin3 revealed significantly decreased expression in gdf6a morphants compared with wild-type (Fig. 3O). Since expression of gremlin1 in the nasal retina is strongly reduced in gdf6a morphants (Fig. 3Q), both human orthologs (GREMLIN1, GREMLIN2) and the related BMP antagonist, NOGGIN, were sequenced in 96 MAC patients without any mutations being identified (data not shown).
In order to study a mammalian model organism, Gdf6+/− mice (Mouse Genome Informatics [MGI: 3604391]), were next examined. Analysis of the 43 offspring generated by seven Gdf6+/− × Gdf6+/− crosses revealed non-Mendelian ratios [Gdf6−/− (n = 1), Gdf6+/− (n = 25), Gdf6+/+ (n = 17)] and reduced litter size (mean, n = 6). Variable and asymmetric ocular Gdf6+/− phenotypes were observed, including: optic disc excavation (eight of 12), microphthalmia (one of 12) and marked asymmetry (six of 12) (Fig. 4A–F). Ocular histology revealed scleral canal enlargement in Gdf6+/− mice (n = 6) that corresponded with the clinically apparent in vivo optic nerve head cupping (Fig. 4G and H). Photopic and scotopic electroretinograms did not demonstrate any significant difference in the amplitude (a- and b-wave) or implicit times (latency) between wild-type and Gdf6+/− mice (n = 14) (data not shown). Two Gdf6+/− mice were screened for skeletal phenotypes using high-resolution micro-CT, however, no appreciable differences were evident at 2 months of age compared with wild-type littermates and in particular, no vertebral fusions were present.
Although our data provide clear evidence for GDF6's role in ocular and skeletal development, certain features are incompatible with simple Mendelian inheritance. Incomplete penetrance and species-specific discrepancies in GDF6-attributable phenotypes were revealed by integrating analyses of a large patient cohort with two animal models. Such features, as well as phenotypic differences at the level of individual mutations, and in one case with the same mutation on different genetic backgrounds, provide evidence of more complex genetic mechanisms.
Incompletely penetrant phenotypes associated with three separate GDF6 mutations demonstrate that a single mutant allele can be insufficient to cause disease, a view supported by analogous findings from zebrafish gdf6a morphants with a common genetic background (Fig. 3D). Considerable reliance can be placed on these findings because of the profound human ocular phenotypes that are incompatible with incomplete ascertainment, corroborative findings from luciferase and Western analyses of GDF6 mutations, and dominantly inherited Gdf6 murine phenotypes (35). This intra-familial variability is highly relevant to human health since elucidating the underlying epistatic interactions would enhance understanding of complex genetic mechanisms, improve genetic counseling for affected pedigrees and, in the longer term, offer potential for modulating these effects therapeutically. However, the complexity of the BMP signaling pathway represents one challenge to deciphering the mechanism(s) involved.
BMPs are synthesized as pro-proteins that are sequentially cleaved and processed to yield disulfide-linked dimers. After binding to type II and type I BMP receptors, the heterotetrameric receptor complex results in phosphorylation and activation of consecutive tiers of downstream SMADs (58,59). This multi-step, indirect and non-linear pathway, in which individual receptors subserve multiple ligands, provides scope for factors modulating ligand activity. The extensive functional redundancy of BMP signaling is well recognized (60) and includes large numbers of paralogs, antagonists and receptors (25,61) as well as intracellular inhibitors (62,63). In this context, observation of altered gdf5 expression in skeletal elements of gdf6a zebrafish morphants (Fig. 3I and J) accords with close paralogs possessing potentially related skeletal functions. Evidence that BMP antagonists are similarly affected by decreased gdf6 function is provided by altered gremlin (ocular) (Fig. 3P and Q) and noggin 1 and noggin 2 (skeletal) expression (Fig. 3K–N). The contribution of such genes to ocular phenotypes remains to be defined although screening of a cohort of 96 MAC DNA samples for mutations in these (three of 13) BMP antagonists did not identify any significant sequence changes. Other genes recently shown to modulate the BMP signaling pathway include: co-receptors DRAGON and RGMa that facilitate ligand-binding (64,65); murine convertase Pcsk5, which cleaves the pro-domain from mature Gdf11 ligand (66); and histone deacetylase 3 (67). Taken together with the key evolutionarily conserved role that gradients of BMP activity have in patterning the vertebrate's dorso-ventral axis, it is thus plausible that buffering by parologs/BMP antagonists and other genes may compensate for mildly perturbed GDF6 function, and combined with stochastic effects ensure that disease phenotypes do not manifest.
In addition to incomplete penetrance, variable phenotypes were observed in nine probands and in both zebrafish and murine models. Patients exhibited either ocular, skeletal or oculo-skeletal anomalies (Table 1). The former represent part of a heterogeneous developmental spectrum, in which unilateral or bilateral disease, variation in ocular size (microphthalmia) and embryonic fissure closure defects (coloboma) may be present (68,69), whereas the skeletal anomalies encompass spondylothoracic dysostosis, cervical and rib fusions, hemi-vertebrae and post-axial polydactyly. Although comparable ocular and skeletal defects are present in gdf6a zebrafish morphants, the prevalence of each differed markedly (Fig. 3A–C). Even though no patient had both vertebral fusions and MAC (Table 1) (2), a subset of morphants exhibited both phenotypes. The Gdf6+/− mice studied exhibited variable and asymmetric ocular phenotypes (Fig. 4) with interesting parallels to the optic nerve cupping seen in glaucoma patients. However, preliminary analysis indicated that no significant skeletal changes were present, and in particular no vertebral fusions were apparent on either micro-CT or MRI. Since these results contrast with those from a second Gdf6+/− strain where skeletal but no ocular anomalies were reported (35), the implication is that genetic background may influence which phenotypes predominate in each species (70–72). This is supported by the differing phenotypes caused by A249E mutations, a likely mutational hotspot owing to the high (85%) adjacent GC-content (73,74). These phenotypes comprise either microphthalmia, coloboma and post-axial polydactyly, or KF; and as at least two of the three probands are unrelated (Fig. 1H and I), different phenotypes can be generated by the same mutation. Although we first reported mutations in GDF6 in two patients with vertebral fusions [ARVO meeting, Fort Lauderdale, USA, 2007] (49), whilst this manuscript was under revision, mutations were also reported in KF (50). This publication is helpful in demonstrating that the disease phenotype in the large pedigree from which proband 3 is derived only exhibits axial skeletal disease, and that this segregates in an autosomal dominant manner.
The seven mutations identified by screening a large patient cohort, demonstrate that approximately 1% and 4% of the ocular and skeletal phenotypes studied are attributable to GDF6, in keeping with the genetic heterogeneity of these disorders (36,75–82). The expression pattern apparent on in situ hybridization (data not shown) correlates with the phenotypic spectrum observed. Several strands of evidence demonstrate that the amino acid alterations observed represent pathogenic mutations, including the high degree of evolutionary conservation evident in six of the seven mutations, their absence from a large number of control chromosomes (Table 1) and the SNP database, together with the complementary assays used to characterize a representative subset of mutations. The combination of significantly reduced reporter activity compared with wild-type GDF6 (Fig. 2A) accords with the reduced levels of mature ligand detected in the mutants by western blot analysis (Fig. 2B), indicating that A249E and K424R represent hypomorphic mutations. Although biochemical characterization of every mutation is impracticable, the findings from A249E, affecting the least conserved residue (Fig. 1B) indicate that alterations affecting more invariant residues are likely to have comparable effects. The distribution of mutations throughout the pro- and mature domains correlates with findings in GDF5 (21,23,24,83–85) and contrasts with localization of mutations in just the prodomain of BMP4 (17) and BMP15 (18,19,86,87). The absence of frameshifts or truncations is compatible with such mutations either resulting in phenotypes not represented in the DNA collections screened, or alternatively, by dimerizing with the normal allele and inducing nonsense-mediated decay, resulting in dominant negative effects that are incompatible with viability [as seen with BMP15 (ovarian dysgenesis 2) (88) and BMPR1B (A2 brachydactyly) (15)]. Although our data do not allow us to differentiate between these possibilities, the non-Mendelian ratios of Gdf6−/− mice (1 of 43) imply the absence of GDF6 function results in reduced viability.
Other interesting features of our data include evidence that GDF6 may underlie a broader range of phenotypes. Two patients were found to have additional systemic anomalies: proband 4 (K424R) had a fused (horseshoe) kidney in addition to rib fusions and hemi-vertebrae, whereas proband 5 (Q253L) had a single testis, in addition to microphthalmia. Such findings accord with BMPs' roles in renal (89,90) and gonadal (91) development and suggest that in addition to highly variable ocular or skeletal disorders, GDF6 may also underlie a spectrum of seemingly sporadic anomalies in other organs. Combined with altered gdf5 expression observed in gdf6a morphants in skeletal elements (Fig. 3G and H), this raises the possibility that GDF6 mutations may cause a broader range of disorders. Indeed, a locus for non-syndromic cleft lip and palate has recently been mapped to a region encompassing GDF6 (90). In view of the increase in GDF5-attributable phenotypes that have been identified over the last few years [brachydactyly type A1 (24), brachydactyly type C (22), acromesomelic dysplasia (20), chrondrodysplasia (21), symphalangism (24) and multiple synostosis syndrome 2 (33)], the spectrum of phenotypes ascribed to GDF6 is expected to expand. Finally, where family data were available, segregation of ocular and skeletal phenotypes appear dominant and fully or incompletely penetrant (Table 1). This accords with the presence of ocular defects in haploinsufficient Gdf6 mice (Fig. 4) and in a patient with a GDF6-encompassing deletion (2), as well as reported dominant inheritance of A249E and an 8q22.2-22.3 inversion in KF pedigrees (50). Interestingly, screening revealed no truncation mutations (missense, nonsense) or homozygous mutations.
In summary, these experiments identify seven GDF6 mutations; four in patients with ocular anomalies and three associated with skeletal phenotypes. The pleiotropic effects (ocular and/or skeletal) of the specific mutations in combination with variable expressivity (coloboma/microphthalmia or vertebral fusion/hemi-vertebrae) and variable penetrance (full/incomplete) describe a novel and complex non-Mendelian inheritance pattern for the corresponding diseases. Our data demonstrate that GDF6 mutations account for 1% of MAC and 4% of vertebral fusion cases implicating perturbed TGF-β signaling in a proportion of ocular and skeletal disorders, which helps identify candidates from the large TGF-β family that merit further investigation. Through the altered noggin1, noggin2, noggin 3, gremlin1 and gdf5 expression observed in gdf6a morphants, we begin to define genes downstream of GDF6/gdf6a. By extending GDF6-attributable phenotypes to other organ systems, these findings provide a potential explanation for some of the reported pedigrees with inherited oculo-skeletal disease. In light of gdf6a's role as the key determinant of zebrafish dorso-ventral patterning (92), we hypothesize that this gene possesses a related function in higher vertebrates and thus may be responsible for a wider range of human diseases beyond skeletal and ocular disorders. An approach integrating analyses of patients and animal models of impaired GDF6 function may aid in elucidating this gene's broader functions as well as the factors mediating penetrance and phenotypic variance that are significant to the pathogenesis of human disease.
MATERIALS AND METHODS
DNA was obtained from blood samples of patients with ocular anomalies (n = 489) and vertebral segmentation anomalies (n = 81) from six centers in four countries. Four patients from the screening panel exhibited both ocular (Duane's syndrome, strabismus, coloboma and visual impairment) and skeletal (polydactyly, vertebral fusions, hemi-vertebrae) phenotypes. Vertebral defects were determined by physical examination, supplemented by radiography in affected individuals. Ethical approval for this study was obtained from the University of Alberta Hospital Health Research Ethics Board, and informed consent was obtained from all participants.
Three pairs of primers amplifying the two exons of GDF6 were designed using Primer3 (http://www.broad.mit.edu/cgi-bin/primer/primer3_www.cgi) and sequences available from Ensembl (http://www.ensembl.org/index.html) and NCBI (http://www.ncbi.nih.gov) (primer sequences and conditions available on request). Briefly, genomic DNA (50 ng) was amplified with Taq polymerase with 10% glycerol and 5% formamide at an annealing temperature of 57°C using standard methods. Amplicons spanning both exons were sequenced on an ABI Prism 3100 capillary sequencer (Applied Biosystems, Foster City, CA, USA) and the data analyzed using Sequencher 4.5 software (GeneCodes, Madison, WI, USA). BsrBI and Tsp509I restriction enzymes were used to screen 646 control chromosomes for A249E and K424R mutations, respectively, whereas 366 control chromosomes were screened with the following enzymes: HpyCH4V (Q253L and P327H), SphI, (G42V), BsrI (Q119R) and TaqαI (D216G). Digestion products were scored following electrophoresis on a 1% agarose gel with ethidium bromide.
Quantitative polymerase chain reaction and array CGH
qPCR was used to screen a subset of 32 patients for copy number alterations of GDF6 (Taqman, Applied Biosystems). Primers were designed using Primer Express software (ABI) (available on request) and qPCR was performed together with connexin 40 (TaqMan Gene Expression Assay ID: Hs99999170_s1; Applied Biosystems) as an internal control. Samples were cycled 40 times at 95°C for 15 s and 60°C for 1 min (ABI Prism, 7000, Sequence Detection System). A DNA sample from the proband of the SHFM pedigree that maps to 8q21.11-q22.3 was hybridized to an oligonucleotide array comprising 385,000 probes at a mean spacing of 6-kb as described elsewhere (46).
Reporter gene assay
After adding a Kozak consensus sequence to permit ribosomal binding on wild-type, A249E and K424R human GDF6 cDNA, sequences were initially cloned into TOPO4 (Invitrogen), then CS2 vector and 1–200 pg of capped, poly-A tailed mRNA (mMessage mMachine®, Ambion Inc, Austin, TX, USA) were injected into 1-cell embryos. For the biochemical assay, PLM cells were harvested from embryonic age (E) 11.5 CD-1 mouse embryos as previously described (51,52). Transfections were carried out with Effectene (Qiagen, Mississauga, ON, CAN) according to the manufacturer's instructions in 384-well plates. Briefly, DNA-transfection mixtures were aliquoted into wells followed by the addition of approximately 100,000 PLM cells, and wells were topped-up to a total volume of 100 µl. Media was replenished 24 h post-transfection, and lysates were collected 48 h post-transfection. Luciferase activity was measured using the Dual Luciferase Kit (Promega, Madison, WI, USA) and firefly luciferase was normalized to an internal Renilla luciferase control. Luciferase assays were performed in quadruplicate and repeated three times.
Western blot analysis
The coding sequence of GDF6 and mutations A249E and K424R (minus the stop codon) were incorporated into Gateway ENTR-D-TOPO, integrated into mammalian V5-tagged Destination vector (Invitrogen, Carlsbad, CA, USA), and confirmed by sequencing. Transient transfections in COS7 cells were performed with FuGENE (Roche, Diagnostics, Indianapolis, IN, USA) according to manufacturer's instructions on 100 mm plates with culture media and lysates collected 48 h post-transfection as previously described (93). Proteins were extracted, separated on a 15% SDS–PAGE gel and transferred to nitrocellulose membranes (BioRad, Hercules, CA, USA), which were incubated with anti-V5 (1:10,000), secreted alkaline phosphatase (1:5000) or α-tubulin (1:10,000) primary antibody (AbCam, Cambridge, MA, USA), and subsequently with anti-mouse or anti-rabbit IgG-HRP (1:5000, Jackson Laboratories, West Grove, PA, USA). The antibodies were detected by chemiluminescence (Pierce, ThermoScientific, Rockford, IL, USA). β-Galactosidase was co-transfected with GDF6+V5 vectors and examined with a β-galactosidase enzyme assay (Promega) for transfection control.
Zebrafish phenotyping and in situ hybridization
Zebrafish (AB strain) knockdown experiments were performed as described earlier (2), with an additional gdf6a morpholino containing five nucleotide substitutions as a control (sequences available on request). RNA whole-mount in situ hybridizations for myod, her7, unc45, nog1, nog2, gremlin1 and gdf5 were performed as previously described (94). Stained embryos were mounted in 70% glycerol and photographed with a Qimaging micropublisher digital charge-coupled device camera. For qPCR validation of in situ hybridizations, zebrafish embryos were collected at 24 h post-fertilization, total RNA extracted (RNaqueous, Ambion, Foster City, CA, USA) and cDNA synthesized (Stratagene, La Jolla, CA, USA). Primers were designed for noggin1, noggin2 and noggin3 using the ‘Roche Applied Science Universal Probe Library’ https://www.roche-applied-science.com/sis/rtper/upl/index.jsp (available on request). Expression levels were quantified by qPCR using SYBR Green chemistry (Stratagene), where ef1α was used as a control. All PCRs were performed twice in triplicate.
Murine genotyping and phenotyping
Gdf6+/− mice [strain Gdf6tm1Lex (MGI:3604391) http://www.informatics.jax.org] were housed and handled in accordance with the University of Alberta Animal Policy and Welfare Committee protocols. Mice were genotyped using allele-specific primers (sequences and conditions available on request) on DNA derived from ear-notched tissue. Mice were anaesthetized using isoflurane and pupils were dilated with Tropicamide (Alcon, USA). Images were initially captured with a digital camera through an OPMI® VISU 160 microscope (Zeiss, Germany). Subsequent examinations were undertaken using an endoscope with an attached otoscope (1218AA; Karl Storz, Tuttlingen, Germany) (95) and still images and videos were captured on a Telecam SLII camera (reference 202130-20, Karl Storz) with a Xenon lamp light source (481C, Karl Storz) and processed with Pinnacle Studio™ software (Avid Technology, Inc., MA, USA).
Micro-CT imaging was performed on a Microtomograph 1076 (Skyscan NV, Aartselaar, Belgium). Serial cross-sectional images were produced of isotropic 18 µm3 voxels, from the 180° angular rotational scans (0.5° incremental steps) and reconstructed using a modified Feldkamp algorithm (40). All image data were Gaussian-filtered and globally thresholded using standardized minimum and maximum cross-section to image conversion values of 0.0–0.0600, respectively, to extract the mineralized phase representing the three-dimensional (3D) bone architecture. Qualitative assessment of skeletal formation was performed from rendered visualizations of the 3D architecture, while quantitative analysis of the vertebral bodies was undertaken with morphometrical analysis software (CTan, Skyscan NV, Aartselaar, Belgium) for the bone volume ratio [volume of bone/total volume of bone and soft tissue], as described elsewhere (41).
The Canadian Institutes of Health Research (to O.J.L. and T.M.U.), Alberta Heritage Foundation for Medical Research and Canadian Foundation for Innovation (to O.J.L.), National Scientific Engineering Council (to C.R.F. and A.J.W.) and Alberta Ingenuity Fund (to A.J.W.). A.J.W. and O.J.L. are recipients of Canada Research Chairs.
The authors thank the patients for participating in this study; Dr Fred Berry for very helpful discussions; Drs Karen Temple, David Fitzpatrick and Raymond Clarke for provision of DNA samples; and Anthony Lott, Helen Chung, Mathew Laroque, May Yu, Tim Footz, Yoko Ito, Hermina Strungau, Timothy Erickson, B. Hemadevi and B. Suganthalakshmi, Drs Stacey Bleoo, Martin Somerville and Gino Fallone for their assistance.
Conflicts of Interest statement. None declared.