Mutations in zinc finger 407 [ZNF407] cause a unique autosomal recessive cognitive impairment syndrome

A pregnant woman and her first cousin spouse, both of Qatari ethnic origin, presented for prenatal diagnosis because of family history of two previous male children with marked cognitive impairment and dysmorphic features, as well as, two early first trimester abortions. The cause of the miscarriages is unknown. To identify the gene and mutation responsible for the phenotype and to propose a possible underlying molecular mechanism, Homozygosity mapping, candidate gene screening, Next Generation Exome Sequencing and bioinformatics analyses of the resulting data were utilized.

Homozygosity mapping was performed for all family members utilizing the Human Mapping 370 K-Cyto12 SNP genotyping array [Illumina, USA]. For the determination of the homozygosity intervals, data were analyzed by the HomozygosityMapper software [1].

Sanger sequencing using Big-dye terminator v.3.1 cycle sequencing [Applied Biosystems, USA] was performed on an ABI 3730 automatic sequencer [Applied Biosystems, USA] to screen for mutations in candidate genes, perform population frequency studies for variants and determine co-segregation of variants with the disease phenotype within the family.

Whole Exome target enrichment Next Generation Sequencing was performed on ABI SOLiD4 platform [Applied Biosystems] according to manufacturer’s specifications. DNA library preparation was with TargetSeq™ Exome Enrichment system [Applied Biosystems] as multiplex fragments libraries utilizing both the SOLiD^® Fragment Library Construction Kits and SOLiD^® Fragment Library Barcoding Kit Module 1–16 for the SOLiD^® 4 System. Bead preparation and enriching was done on an EZ Bead Emulsifier, Amplifier and Enricher utilizing E80 scale. Sequencing modality was with multiplex fragment paired-end.

Bioinformatic analyses of whole Exome NGS data was as follows: Raw data files (in a proprietary XSQ file format) were analyzed with the LifeTechnologies LifeScope v2.5.4 software running on a dedicated cluster to align the reads produced by the SoLID to a hg19 whole genome reference sequence, sourced from the University of California, Santa Cruz Genome Informatics Group (UCSC). The aligned BAM files were validated, duplicate sequences were identified and removed and incorrectly identified Mate-Pairs were corrected using the Picard v1.87 software. The Genome Analysis Tool Kit (GATK) v3.0.0 was applied to the ‘corrected’ output files to recalibrate the base quality scores, using machine learning to model any systematic errors in the data; carry out localized realignments around possible insertion/deletion sequences to ensure mapping accuracy; identify viable variants from the sequence reads; and recalibrate the variants to ensure accuracy of the variant calling, in a variant-type specific manner. Once a suitable list of variants was produced in this manner, the list was filtered using in-house scripts to confirm variant zygosity and identify those variants that conform to the inheritance model. These variants were annotated using an in-house script in conjunction with Annovar to produce an annotated list of variants with the most recent information available on a number of reference websites.

As the three-dimensional [3-D] structure of the ZNF407 protein is not available in the Protein Data Bank (PDB), sequence homology methods were utilized to construct 3-D models of the wild type and mutant zinc finger domains 18 and 19 of ZNF407. BLAST-searches (http://​blast.​ncbi.​nlm.​nih.​gov/​Blast.​cgi) were performed within the PDB for crystal structures of homologous proteins. Once these were identified, the CLUSTALW software (http://​www.​ebi.​ac.​uk) was utilized to insure correct sequence alignments between domains 18 and 19 of ZNF407 and the retrieved zing finger structures found in the PDB. Focus was on complexes with double-stranded DNA only so as to have the proper bioactive conformation of the tandem finger. A high sequence homology with the Aart protein in complex with dsDNA (PDB code 2I13) was identified. The 3-D model of ZNF407 was then generated with the Modeller software (version 9v8) [2] based on the crystal structure of Aart in complex with DNA as a template (PDB code 2I13). A cycle of energy minimizations was then performed to obtain a stable conformation of the 3-D model. The quality of the model generated was finally assessed with DOPE score [3] and the Ramachandran plot of the best model was calculated by ProCheck [4] to discriminate unfavorable amino-acids backbone conformations. Three-Dimensional ribbon representations of the models were obtained with the molecular graphics program PyMOL (http://​www.​pymol.​org). To generate the 3-D model of the mutant, Serine 1685 was replaced for a Tryptophan residue in the second finger of the 18–19 tandem finger domain and submitted to the same protocol as that used for the wild type complex. Thereafter both models were compared for an insight into the structural and thermodynamic stability effects of the mutation. To generate the 3-D model of the mutant, Serine 1685 was replaced for a Tryptophan residue in the second finger of the 18–19 tandem finger domain and submitted to the same protocol as that of the wild type complex. Thereafter, both models were compared for an insight into the structural and thermodynamic stability effects of the mutation.

The study was conducted in accordance with the provisions of the Declaration of Helsinki and an informed consent was obtained by the guardian of the affected participating family member.

Homozygosity mapping, screening of positional candidate genes, whole Exome Next Generation Sequencing and data mining were utilized in the effort to identify the disease causing gene and mutation for this apparently novel rare autosomal recessive disease. Out of the five potentially damaging variants identified within the homozygosity intervals, the IGDCC4: c.G727A/p.D243N [rs35223184]; and C15orf39: g.G46293956A/p.R1029K [rs149175372] were immediately excluded since they represent known common polymorphisms with >1% population frequency [1000 Genomes and NCBI dbSNP Build 135].

The third identified variant, TLN2 [c.G6716A/p.R2239H], affects a non-conserved amino acid. The “alternate” amino-acid Histidine is found as the “normal” amino-acid at that position in numerous non-human species. A gene trap mouse model with disrupted Taln2, as well as Taln2 knockout mice are viable and fertile but may display a mild dystrophic phenotype in skeletal and cardiac muscles [5‐7] without any neurological findings as in the family under investigation. These render this variant as extremely unlikely pathogenic in relation to the clinical phenotype of the patients.

The fourth identified variant c.G1487T/p.G496V affects a non-conserved amino acid in a known Promyelocytic Leukemia tumor suppressor [PML] gene in Acute Promyelocytic Leukemia (APL). PML mutations have been detected in APL as well as in other hematopoietic malignancies. Functional loss of PML does not relate at all to the phenotype of the family [8], and again this variant is extremely unlikely to be pathogenic in relation to the phenotype under investigation.

The fifth variant, a c.C5054G/p.S1685W at the zinc finger transcription factor gene (ZNF407), was found to affect an evolutionarily highly conserved amino acid. It has damaging effects according to PolyPhen and SIFT protein-modeling software [9, 10], and it co-segregates with the disease phenotype within the family. It is absent in any of the publically available variant databases [1000 Genomes and NCBI dbSNP Build 135]. Furthermore, it is absent in 400 ethnically matched control chromosomes, making it the only and most likely disease-causing gene and mutation.

Zinc finger genes are found ubiquitously, and categorized into several types based on the structure of their binding residues, and their biological roles within the cell [11, 12]. Through their ability to bind to DNA, RNA & proteins, zinc finger proteins facilitate numerous cellular functions and processes including, but not limited to, the control of hematopoiesis, regulation of gene expression, cell differentiation, and development [13]. Their DNA-binding domain is comprised of tandem repeats of two, three or more fingers that can have different binding specificities. These tandem arrays, separated by amino acid linkers, bind through sequence-specific contact to 2–4 bases in the major groove of the DNA, spaced at 3-bp intervals [14, 15]. Optimal binding is achieved through the sequential binding and wrapping of the zinc finger domains around the DNA [15].

ZNF407 constitutes one of the 481 ultra-conserved elements in the human genome, rendering it a functionally essential gene [16]. It presents three isoforms produced by alternative splicing. Isoform 1 (Q9C0G0-1) is the “canonical” sequence. Isoform 2 (Q9C0G0-2) differs from the canonical sequence in amino acids 1811–1815 and amino acids 1816–2248 are missing. In isoform 3 (Q9C0G0-3), amino acids 1625–1660 are modified and residues 1661–2248 are missing. This isoform thus does not possess zinc fingers 18 to 22.

The homology modeling approach is justified since the identity rate between the sequence of an unknown structure and a sequence whose experimentally available structure is high. Figure 3 shows the sequence alignment between fingers 18 and 19 of the target sequence (ZNF407), as annotated in UniProt, and the sequence corresponding to fingers 2 and 3 of the high resolution crystal structure (Protein Data Bank code 2I13, subunit A) of a tandem 6-finger zinc finger domain (Aart) bound to a double-stranded DNA and designed to recognize ANN triplets [17]. This alignment shows a high degree of identity between the two sequences (Figure 3). In spite of the 2-residue insertion in the N-terminus of zinc finger 18 of ZNF407 (Threonine-Tryptophan, Figure 3), the high sequence identity (~40%) between the two sequences allows us to model-build a 3D structure of the ZNF407-DNA complex in order to visualize the effects of the Serine to Tryptophan mutation. Since no specific DNA sequence for ZNF407 binding are available, the interaction of tandem fingers 18 and 19 with the DNA sequence was modeled after the Aart-DNA complex. Figure 4 shows diagrams of the 3D model of wild type ZNF407 tandem finger domains 18 and 19 wrapped around the major groove of the double-stranded DNA sequence. The zinc finger linkers are five residues long in general (residues i to i + 4) and usually contain a Proline residue at position i + 4. The Proline side chain is known for its rigidity because of its cyclic character. Its presence at position i + 4 thus restrains the conformation of the linker, reducing its flexibility, and properly positioning the α-helix of the C-terminal finger for interaction with the target DNA. This is illustrated in the case of fingers 19 and 20 (Figure 4a). Accordingly, by reducing the overall conformational entropy, complex formation is favored energetically and structurally.

The presence of Serine in position i + 4 of the ZNF407 18-19 linker is unusual and carries with it added backbone flexibility. Nevertheless, this potential flexibility is reduced by an H-bond between the hydroxyl group of S1685 at i + 4 and the carboxyl side chain of E1683 at i + 2, locking the relative spatial position of the 18 and 19 fingers. This stabilizes the linker in a β-strand-like conformation and favors the interaction with DNA by reducing the number of bioactive conformations available for binding (Figure 4b). A Serine to Tryptophan mutation in the 18–19 zinc finger linker abolishes this H-bond, leading to a more flexible linker, and thus to a decrease in the probability of zinc finger-DNA complex formation of downstream partners (Figure 4c). The expected physiological effects of this physico-chemical perturbation include interfering with the expression of genes under ZNF407 control.

DNA- and RNA-binding zinc finger proteins are transcriptional regulators with a major function controlling developmental cascades of gene expression especially during fetal brain development. Mutations in such proteins have been found to interfere with normal brain development manifested during the developmental period, thus causing severe mental retardation. They have been associated with non-syndromic X-linked mental retardation [ZNF81: OMIM 314998 [18]; ZNF674: OMIM 300573 [19], ZNF711: OMIM 300803 [20], autosomal recessive non-syndromic intellectual disability [ZC3H14, OMIM 613279] [21] and impairment in adaptive behavior that manifest during development.

ZNF407 mRNA, as well as the corresponding protein are expressed in most normal tissues with moderate nuclear and cytoplasmic positivity. These include adult, embryonic and fetal central and peripheral nervous systems (http://​www.​nextprot.​org/​db/​entry/​NX_​Q9C0G0/​expression). It is not known if there is tissue specificity for each of the three isoforms or what tissues express the two isoforms affected by the c.C5054G/p.S1685W mutation. To date, twenty polymorphisms for ZNF407 have been reported in the Domain Mapping of Disease Mutations database (DMDM) [22]. The natural variants reported in the UniProt database entry Q9C0G0 (N69S, G512R, N972T, S1259L, A1913T) are included in DMDM. None of these, though, have been associated with known diseases. However, deletion of a critical region on 18q23 that includes the ZNF407 gene and two other zinc finger genes (ZNF516 and ZNF236), among other genes, is correlated with congenital aural atresia (CAA), which manifests a subset of phenotypes recognized by the 18q deletion (18q-) syndrome [23]. However, the clinical picture described in this syndrome is probably due to haploinsufficiency of several genes, as we could not find any symptomatic presentation in obligate carriers. Analysis of several CAA patients led to defining the breakpoint of the deletion within ZNF407[24]. Another identified translocation breakpoint in the third intron of the ZNF407 gene causes a reduction in the transcript of its isoform 1, resulting in non-syndromic intellectual impairment and autism [25]. Furthermore, two de novo damaging missense mutations [c.A1436G/p.Y460C; c.C3640G/p.P1195A] in the very long linker region between zinc fingers 3 and 4, and between 11 and 12, respectively, of the ZNF407 gene were found in one intellectual impairment patient each. Both patients were heterozygous, each for their relevant mutation.

Next Generation Exome and Genome sequencing have clearly demonstrated that humans harbor numerous deleterious mutations with no apparent ill effect [26, 27]. The built-in redundancy in the human genome provides alternate pathways that make possible compensation for deleterious mutations in important genes without which there would be incompatibility with life. Thus, deleterious mutations resulting from NGS data need to be supported by additional genomic or functional data whenever possible. The ZNF407 c.C5054G/p.S1685W mutation is the only mutation that lies within one of the homozygosity intervals that is certain to harbor the offending gene that could not be excluded as the disease causing mutation. It co-segregates with the disease phenotype within the family, it is not found as a polymorphic variant in publically available databases or in the ethnic population and it affects an evolutionarily highly conserved amino acid. Normal sequence variants [polymorphisms] have not been reported in ZNF407 linker regions (between zinc fingers) providing further evidence of the sequence specificity required by these regions for normal function and subsequently for the damaging effects of the mutation. Extensive protein modeling clearly indicates that the mutation interferes with the normal function of the ZNF407 protein.