Main Text
Dual sensory impairment due to cataracts and sensorineural hearing loss is a well-recognized consequence of infectious teratogenic exposure (i.e., fetal rubella syndrome), but only rarely observed as a developmental defect in genetic disease phenotypes. In 1996, Gripp and co-workers described two unrelated subjects with congenital cataracts and sensorineural deafness associated with intellectual disability, short stature, brachycephaly, and a distinctive flat facial appearance and considered this trait to represent a previously unrecognized syndrome (MIM 601088).1 In their clinical report,2 Aymé and Philip discussed on the similarities between the clinical features exhibited by their case and others previously reported by other authors4 and those of the patients reported by Gripp and co-authors.1 Aymè and Philip concluded that all these cases were clinically related to the patient originally described by Fine and Lubinsky.3 For this reason, the authors proposed the term “Fine-Lubinsky syndrome” to define this developmental disorder. Since then, a few additional cases exhibiting features fitting or partially overlapping this condition(s) have been reported,5–9 and whether these phenotypes represent variable manifestations of a single nosologic entity remained unresolved. Autosomal recessive inheritance was suggested, based on affected siblings.7 Here, whole-exome sequencing (WES) on a single affected individual and Sanger sequencing on a selected cohort of subjects with phenotype suggestive of FLS were used to identify a narrow spectrum of missense mutations in v-maf avian musculoaponeurotic fibrosarcoma oncogene homolog (MAF [MIM 177075]) as the molecular cause underlying this previously poorly understood multisystem disorder, and delineate its clinical phenotype. The provided biochemical and functional data demonstrate that the mutations identified in this study specifically affect the phosphorylation of MAF promoted by the protein GSK3, which is a serine/threonine kinase that requires a specific recognition motif for its action—i.e., the presence of a proline residue adjacent to the serine/threonine residue that is substrate of its action. The impaired phosphorylation at those sites affects MAF ubiquitination, which, in turn, impairs degradation of the mutated (unphosphorylated and unubiquitinilated) protein, generally mediated by the proteasome complex. Finally, these mutations are able to induce neurodevelopmental defects in vivo (zebrafish), thus representing dominant-acting mutations.
Thirteen subjects were included in this study. All individuals were clinically assessed by experienced clinical geneticists. Among them, nine subjects had previously been reported.1,2,5–8 Clinical features are described in detail in Table S1. Clinical data and biological material collection and storage were attained from the participating families after written informed consent was secured, following procedures in accordance with the ethical standards of the responsible committees on human experimentation (institutional and national). Genomic DNA was isolated from peripheral blood leukocytes, skin fibroblasts, hair bulb cells, and/or buccal mucosal epithelial cells, using standard protocols. We performed WES on genomic DNA extracted from circulating leukocytes of a single affected subject (case 8) (Figure 1A, CaGi_UCSC). Exome capture was performed using NimbleGen SeqCap EZ Exome V. 3.0 (Roche) and sequencing by a HiSeq2000 instrument (Illumina). WES data analysis was performed using an in-house implemented pipeline.10 For sequencing statistics, see Table S2. Data annotation predicted 11,168 high-quality variants having functional impact (i.e., non-synonymous and splice site changes). Among them, 259 private, rare (minor allele frequency < 0.001), or clinically associated changes were retained for further analyses. After excluding the presence of variants compatible with autosomal recessive transmission (Table S3), we reasoned that the clinical symptomatology might be caused by a de novo event. Candidates were stratified through a mixed filtering/prioritization strategy taking into account the predicted impact of each variant and the functional relevance of individual genes on the developmental processes altered in the disorder. Only changes (private, clinically associated, or having unknown frequency or minor allele frequency < 0.001) predicted to be deleterious by the Combined Annotation Dependent Depletion (CADD)11 (score > 15.0) or Database for Nonsynonymous SNPs’ Functional Predictions (dbNSFP) Support Vector Machine (SVM)12 (radial score > 0.0) algorithm were retained and prioritized on the basis of the functional relevance of genes using GeneDistiller. Genes were ranked based on combinations of terms from the OMIM clinical synopsis for MIM 601088 and 601353 (i.e., cataract, deafness, mental retardation, facial dysmorphism, short stature, and seizure) as keywords, using similarity of expression patterns and protein-protein interactions as major weights. We obtained the highest score for MAF, a gene whose mutations had previously been reported to cause autosomal dominant congenital cataracts and lens abnormalities (MIM 610202).13,14 Sanger sequencing confirmed heterozygosity for the c.161C>T (p.Ser54Leu) change in the proband, and sequencing of parental DNAs revealed only the reference allele, evidence for its de novo origin (Figure S1). STR genotyping (AmpFlSTR Identifiler Plus [Life Technologies]) confirmed paternity. The variant was documented in the proband’s skin fibroblasts as well as hair bulb and buccal epithelial cell specimens, strongly arguing against the possibility of a somatic event (Figure S1). All the other candidate variants turned out to be inherited from one of the unaffected parents (Table S4).
To confirm the causal involvement of MAF, we scanned the entire coding sequence of the gene (NM_005360.4 and NM_001031804.2) for mutations in DNA samples from 12 additional subjects (Table S1), including a sib pair, with features overlapping the conditions delineated by Gripp et al.1 and Aymé and Philip,2 by direct sequencing. Primer pairs designed to amplify the MAF coding exons and their intron boundaries (NC_000016.10, 79593848..79600725) are listed in Table S5. We identified heterozygous missense mutations in seven unrelated individuals (Table 1 and Figure S1). Two different changes affected Thr58 and Pro59, while others involved adjacent residues, including the previously identified c.161C>T substitution. In all family trios for which parental DNA samples were available, genotyping documented the de novo origin of each mutation, and STR analysis confirmed paternity (Table 1). Sanger sequencing of DNA from available oral mucosal epithelial cells (case 11-1), or epithelial cells and fibroblasts (case 4-1) supported the germline origin of mutations. All changes were predicted to impact protein function by dbNSFP and/or CADD (Table 1), and affected residues conserved among orthologs and paralogs (Figure S2).
MAF is a basic leucine zipper (bZIP)-containing transcription factor of the AP1 superfamily.15,16 It is important for lens and eye development17,18 and controls multiple physiological processes, including mechanosensory function, and chondrocyte and T cell differentiation.19–21 Immunohistochemical analyses in mouse embryos documented wide Maf expression (Figure S3). Consistent with previous reports, we observed Maf staining in the lens, dorsal spinal cord, dorsal root ganglia, skin, kidney, hypertrophic chondrocytes of vertebrae, rib and limb cartilage, and the cartilage primordium of the basioccipital bone.22,23 In line with the sensorineural hearing loss occurring in all mutation-positive subjects, we detected a specific and strong signal in cochlear cells of E14.5 embryos.
Similar to other “large” MAF subfamily members (i.e., MAFA, MAFB, and NRL), MAF’s structure is characterized by a C-terminal extended homology region and bZIP domain mediating DNA binding, and a N-terminal transactivation domain required for transcriptional activity and regulatory function (Figure 1B). The latter contains four GSK3 phosphorylation motifs, highly conserved among large MAF proteins (Figure 1C). In MAFA, the sequential phosphorylation of these serine/threonine residues promotes ubiquitination and rapid degradation, but also increases transactivation potential.24,25 Remarkably, all identified MAF mutations clustered within these motifs. Three affected residues, Ser54, Thr58, and Ser62, are known GSK3 phosphorylation target sites.26 The remainder did not involve phosphorylatable residues, but were predicted to affect GSK3-mediated phosphorylation by altering proline residues adjacent to either a phosphorylation site (Thr58) or the C-terminal priming site (Ser70), whose phosphorylation is absolutely required for GSK3 function. To explore the impact of the p.Pro59His, p.Pro59Leu, and p.Pro69Arg changes, we performed molecular dynamics (MD) simulations on complexes formed by full-length GSK3 and ten residue-long peptides of MAF corresponding to the segment that interacts directly with the GSK3 binding cleft, encompassing both the GSK3 target and pSer/pThr primed residues (Table S6). The starting coordinates for the ATP-bound GSK3 were taken from the crystallographic structure of GSK3B complexed with AMP-PNP (PDB entry 1pyx).27 Each decapeptide was set in an extended conformation along the catalytic cleft of GSK3 as specified in Table S6. The MD simulations were carried out according to the protocol previously described.10 The Gromos 53a6 force field was used, with the exception of the partial charges of pSer/pThr,28,29 and the parameters for ATP, obtained from quantum mechanical calculation of the molecular system reported in Figure S4. For p.Pro59His and p.Pro59Leu decamers, the conformation of the trimer comprised between the substrate and primed residues was rearranged considerably during the simulations (Figure 2A and Figure S5), with the correct orientation of the substrate residue in the GSK3 active site being destabilized (Figure 2A). p.Pro69Arg, introducing a cationic residue in the proximity of the positively charged GSK3 priming pocket formed by residues Arg96, Arg180, and Lys205, caused a general rearrangement of the adjacent pSer70, pulling it away from the binding pocket (Figure 2B). Overall, our simulations indicated consistently that all disease-causing MAF mutations inhibit GSK3-mediated phosphorylation through impaired association and/or catalysis, by perturbing the interaction with the priming site (p.Pro69Arg) or the active site (substitutions affecting Pro59).
To explore the mutations’ functional impact directly, we evaluated MAF phosphorylation status. The disease-causing p.Ser54Leu, p.Thr58Ala, p.Thr58Ile, p.Pro59Leu, p.Pro59His, and p.Pro69Arg (FLS-like disorder), and p.Arg288Pro (c.863G>C) changes, the latter considered as representative of lesions associated with isolated cataract,13 were introduced into the MAF cDNA cloned in pCS2+ vector using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies). Consistent with previous reports,26 Western blot analysis of transiently transfected COS1 cell lysates documented two MAF states: a slower-migrating, fully phosphorylated form, and a faster-migrating, unphosphorylated form. In cells expressing wild-type MAF, the phosphorylated protein (upper band) predominated, while unphosphorylated MAF (lower band) was barely detectable (Figure 3A, upper panel). Similarly, the mutant carrying the p.Arg288Pro substitution in the DNA binding domain, previously associated with isolated lens and eye defects, was efficiently phosphorylated. This was in sharp contrast to all MAF mutants identified in the present study, which accumulated in cells as unphosphorylated proteins. GSK3-mediated phosphorylation represents a regulatory mechanism promoting MAFA ubiquitination and degradation.24,25 Based on the high conservation of the GSK3 recognition motif and MAF being a GSK3 substrate, we hypothesized that the amino acid changes in our affected subjects might mediate inefficient protein clearance. On Western blot analyses, we noted increased protein levels (Figure 3A, upper panel) and decreased ubiquitination (Figure 3A, middle panel) for the disease-causing MAF mutants when compared to the wild-type protein (Figures 3A and 3B). Treatment with cycloheximide (CHX), a protein synthesis inhibitor, showed that the half-life of wild-type MAF was much shorter than that of the mutants (Figure 3B). Indeed, a complete disappearance of the protein was observed upon 4 hr CHX treatment, while the steady-state level of the mutants was largely unchanged. Consistently, treatment with MG132, which specifically inhibit proteasomal function, stabilized the protein level of wild-type MAF, while it did not have any significant effects on mutants (Figure 3B). Taken together, these results showed that mutations prevented MAF degradation and enhanced their stability. Of note, a partial phosphorylation was apparent for the p.Pro69Arg MAF mutant, which was associated with increased degradation via proteasome, even though less efficiently compared to wild-type MAF. This finding suggests a milder perturbing role of the proline-to-arginine substitution on GSK3-mediated phosphorylation at Ser66 compared to the other disease-causing amino acid changes, possibly due to the peculiar effect of the introduced arginine residue, which was documented to primarily affect MAF interaction with the GSK priming site. Confocal microscopy of transfected COS1 cells confirmed the nuclear localization of all tested mutants and their higher abundance within cells (Figure 3C). Moreover, treatment with CSK buffer prior fixation indicated that the syndrome-causing mutants retained efficient interaction with chromatin suggesting that they bind to DNA, in contrast to the DNA binding-impaired cataract-associated p.Arg288Pro mutant (Figure 3C and Table S7). Transactivation assays using luciferase as reporter under control of the IL4 promoter documented that COS1 cells transiently expressing the cataract-causing mutant allele had barely detectable reporter induction (Figure 3D). In contrast, cells expressing the p.Ser54Leu, p.Thr58Ala, p.Thr58Ile, p.Pro59Leu, p.Pro59His, or p.Pro69Arg MAF coding alleles showed efficient induction of luciferase levels, though not reaching the levels of the wild-type protein, suggesting that, despite their stabilization and much higher levels, these mutants are less active, at least under these specific conditions.
Next, we conducted gene-expression profiling analyses on primary skin fibroblasts from two unrelated subjects (4-1 and CaGi_UCSC) to explore more globally the impact of dysregulated MAF function on gene expression (Table S8 for details). Approximately 6% of genes in mutation-positive subject-derived cells were expressed differentially compared to control fibroblasts (ATCC code PCS-201-012). Gene ontology enrichment analysis of these differentially expressed genes revealed an overrepresentation of genes coding for proteins associated with developmental programs (tissue morphogenesis, branching morphogenesis, urogenital system, and bone development) and cellular processes (cytoskeletal rearrangement, cell migration and adhesion, and response to extracellular stimuli) (Table S8). Publicly available microarray data (ArrayExpress accession code E-GEOD-51231) allowed selection of genes whose expression is positively controlled by MAF. Remarkably, these putative targets were observed to be enriched significantly in up- or downregulated transcripts in MAF mutation-positive fibroblasts (Figure S6), suggesting a complex, promoter-specific dominant dysregulatory function of these mutations.
Because cognitive deficits, with or without brain defects, are invariably present among subjects with MAF mutations affecting the GSK3 recognition motif, but absent in individuals with congenital cataracts caused by MAF mutations involving the DNA binding domain,13,30–33 we posited that expression of the mutant alleles here identified should induce defects in neurogenesis, whereas alleles driving dominant cataracts through a haploinsufficiency model should not. To test this hypothesis in vivo, we analyzed the impact of these mutation classes on the integrity of the central nervous system (CNS) using a zebrafish model. The optic tectum is a major component of the vertebrate midbrain, comprising a structure equivalent to the mammalian superior colliculus, with external layers collecting sensory information and internal layers having a motor-function. We previously showed that measurement of the area of this structure represents a robust surrogate for brain volume and that reduction of the absolute area of the tectum correlates with neurodevelopmental defects in humans.34,35 We injected clutches of 50–100 embryos with human mRNA encoding wild-type MAF, each of four alleles carrying mutations discovered in the present study, or the cataract-associated allele. Experimental work was carried out under protocols approved by the Institutional Animal Care and Use Committee of the Duke University, as previously described.34–36 Following quantitative measurement of the surface area of the tecta (blind to injection cocktail; replicated), we observed that expression of either wild-type and cataract-associated (c.863G>C; p.Arg288Pro) MAF mRNAs did not induce appreciable brain volume differences. In contrast, injection of human MAF mRNA encoding each of the p.Ser54Leu, p.Thr58Ala, p.Pro59Leu, or p.Pro69Arg changes caused a statistically significant reduction of the size of the optic tecta (p < 0.0001; Figure 4). These data were concordant with our in vitro model and provided further evidence for the differential impact of these two mutation classes on CNS development.
Individuals with MAF mutations shared congenital or early onset cataracts, sensorineural hearing loss, developmental delay/intellectual disability, seizures, brachycephaly, midfacial hypoplasia, and reduced growth (final height −2.25 to −4 SD) as major characteristics (Figure 1A and Table S1). Bilateral hearing loss was diagnosed in early childhood and had at least a moderate to severe sensorineural component in all, in some cases requiring hearing aids. Limited information on inner ear imaging is available. Facial features were distinctive and recognizable, and included a short nasal tip and long philtrum, a small mouth, and small/low-set posteriorly angulated ears. Late closing anterior fontanel, radio-ulnar synostosis or limited elbow movement occurred in several subjects, as did pericardial effusion and idiopathic chondrolysis of the hip. The three adult females had mammary gland hypoplasia, and two lacked axillary hair. Two individuals showed hypopigmented retinal lesions. Mesangiocapillary glomerulopathy occurred in the oldest affected individual. While intellectual function was not assessed systematically, impairment was universally present. Adaptive function in adult individuals ranged from severely autistic, nonverbal and requiring medication for behavior problems to working in a supervised community environment and living with guardians. Seizures and abnormal EEG findings consistent with focal and diffuse abnormal activity were present in all individuals, most often diagnosed in early childhood. Brain imaging studies commonly showed prominence of the axial and extra-axial fluid spaces; other anomalies like a Chiari 1 malformation or J-shaped sella occurred in single cases. The absence of mutations among subjects with only partly overlapping features (e.g., absence of brachycephaly and cataracts in sibs 10 and 11, and facies not suggestive in case 13) implies that mutations affecting residues within the GSK3 recognition motif of MAF cause a clinically homogeneous phenotype, and that cataracts, sensorineural hearing loss, intellectual disability, seizures, brachycephaly, Down syndrome-like facial appearance, and reduced growth are likely to represent cardinal features of the disorder.
Transcription factors operate in developmental processes to mediate inductive events and cell competence, and perturbation of their function or regulation can dramatically affect morphogenesis, organogenesis, and growth.37 Here we establish that missense mutations impairing GSK3-mediated MAF phosphorylation severely perturb multiple developmental processes. Inactivating missense mutations affecting the MAF DNA binding domain had been reported to cause congenital cataracts, microcornea and iris coloboma.13,30–33 Contrasting those mutations, we describe a class of missense changes not impairing DNA binding, but instead precluding post-translational modifications required for proteasomal degradation, resulting in mutants being more stable than wild-type MAF. In addition to defective degradation, this MAF mutation class showed perturbed in vitro transactivation activity, and endogenous expression in primary fibroblasts was associated with both up- and downregulation of genes identified as positively controlled MAF targets. These findings support the idea that, besides promoting degradation, GSK3-mediated MAF phosphorylation impacts protein activity through other mechanisms, as previously demonstrated for MAFA,24 and suggest a complex pathogenetic mechanism involving protein stability and functional dysregulation. Of note, dominant missense mutations affecting residues in the same regulatory motif of MAFB and NRL, including those homologous to Ser54, Pro59, Thr62, and Pro69, have been reported to cause multicentric carpotarsal osteolysis (MCTO [MIM 166300]) and autosomal dominant retinitis pigmentosa (RP27 [MIM 613750]) (Figure S2), respectively,38,39 further highlighting the critical role of this domain.
Early onset cataracts and hearing loss have rarely been reported in genetic conditions, with the collagenopathies, Alport (MIM 104200), Stickler (MIM 108300) and Marshall (MIM 154780) syndromes being the most common. This work identified MAF mutations as principal cause for a disorder combining cataracts and hearing loss in a multisystem developmental syndrome independently recognized by Gripp et al.1 and Aymé and Philip.2 Though the clinical presentation of subjects within this study resembles that described in the context of FLS, we note that this overlap is only partial. Moreover, FLS has been applied to clinically variable phenotypes of likely heterogeneous etiology (here exemplified by cases 10, 11, and 12). We therefore propose the eponym Aymé-Gripp syndrome for the disorder caused by mutations affecting residues of the GSK3 phosphorylation motifs in order to distinguish the phenotype described here from that reported by Fine and Lubinsky.3
By regulating the spatio-temporal expression of tissue-specific genes, MAF proteins act as key regulators of terminal differentiation in many tissues and organs, including bone, brain, kidney, lens, pancreas, and retina.40,41 While apparently no gross anomalies are associated with Maf haploinsufficiency in mice, Maf−/− pups die soon after birth and exhibit defective lens formation and eye development,17,18 chondrocyte terminal differentiation,21 as well as differentiation and function of mechanoreceptors and neurons with mechanosensory function.20,42 In contrast, a semi-dominant, missense mutation (c.881G>A, p.Arg291Gln) affecting the DNA-binding domain of Maf and causing a reduced transactivation activity of the transcription factor has been associated with congenital cataract in heterozygote mice,14 recapitulating the hypomorphic cataracts-associated MAF mutations in humans.13,30–33 Finally, a functionally distinct missense change (c.269A>T, p.Asp90Val) affecting the N-terminal transactivation domain and promoting enhanced transactivation function in Maf has been shown to cause a dominant isolated cataract phenotype.43 Contrary to these loss-of-function, gain-of-function, and haploinsufficiency models, we here showed a distinct, dominantly acting effect of MAF mutations underlying a complex developmental disorder affecting multiple organs and tissues. As such, the pleiotropic effect of impaired MAF phosphorylation in Aymé-Gripp syndrome expands the perturbing consequences of dysregulated MAF function for multiple developmental programs, establishing its role in morphogenesis, CNS development, hearing and growth, and delineates a novel instance of protein dosage effect in human disease.