CCDC88A mutations cause PEHO-like syndrome in humans and mouse

P rogressive E ncephalopathy with oedema, H ypsarrhythmia and O ptic atrophy (PEHO) is a rare, neurodegenerative disorder of unknown aetiology. Nahorski et al . identify the first causative recessive mutation in CCDC88A , which encodes the actin-binding protein Girdin. The phenotype and brain anatomy of the Ccdc88a knockout mouse resemble those of human PEHO syndrome.


Introduction
Progressive encephalopathy with oedema, hypsarrhythmia and optic atrophy (PEHO) syndrome is rare, of unknown origin, and causes profound intellectual disability with little if any developmental progress. It was first described in Finnish children in 1991, from where the most complete phenotypic data originates (Salonen et al., 1991). Onset occurs usually within the first few weeks of life with affected children displaying seizures, hypotonia and lack of visual fixation (Somer, 1993;D'Arrigo et al., 2005). Those affected thence exhibit early arrest of motor and mental development, optic atrophy, with radiological examination demonstrating severe cerebellar and brainstem atrophy from birth, with cerebral atrophy developing at a later stage (D'Arrigo et al., 2005). Usually microcephaly is present at birth, but inevitably there is a progressive reduction in brain growth (secondary microcephaly). Prognosis is poor with most patients dying before 10 years of age. Inheritance is predicted to be autosomal recessive (Somer, 1993). Clinically there is phenotypic overlap of PEHO syndrome with the early-infantile epileptic encephalopathies (Cross and Guerrini, 2013). Cases lacking either optic atrophy or cerebellar hypoplasia are often termed PEHO-like (Chitty et al., 1996).
The genetic cause(s) and the underlying pathophysiology of PEHO syndrome are unknown; and there is no known curative therapy. We report a large consanguineous family with three affected members in two nuclear families. In affected individuals we identified a homozygous frameshift mutation in CCDC88A, the gene that encodes the actin binding protein girdin (GIRDers of actIN filament, also known as APE GIV and HkPP) (Anai et al., 2005;Enomoto et al., 2005;Le-Niculescu et al., 2005;Simpson et al., 2005). We show that the mutation results in the absence of full length, functional CCDC88A protein. To further investigate the function of CCDC88A in mammalian brain function we examined CCDC88A-deficient mice and compare their phenotype and CNS architecture to our patients. Our results reveal novel insights into the underlying developmental, neuroanatomical and molecular defects of CCDC88A deficiency in mammals.

Materials and methods
Materials and methods are described in detail in the Supplementary material.

Subjects
The UK National Research Ethics Service, Cambridge Central Research Ethics committee gave permission for our study. The authors cared for the family for over a decade; all medical care was within the National Health Service. Results and notes were examined and recorded.

Exome sequencing
Genomic DNA from affected Patients 1 and 2 ( Fig. 2A) was extracted and analysed by exome sequencing (SureSelect Human All Exon 50Mb Kit, Agilent Technologies). Potentially pathogenic mutations were identified by comparison with the GRCh37 reference human genome, filtering for changes with allele frequencies of 5 1 in 500 in the 1000 Genomes project, and focused on homozygous changes affecting both siblings. Any changes were confirmed and tested for expected segregation, by polymerase chain reaction (PCR) and Sanger sequencing.

Reverse transcription polymerase chain reaction analysis
To analyse possible nonsense mediated decay of mutant CCDC88A, blood samples were taken from Patient 1 and their parents, mRNA extracted and reverse transcribed to cDNA. Full details of reverse transcription PCRs (RT-PCRs) undertaken and primers used are detailed in the Supplementary material.

CCDC88A cloning and expression
CCDC88A cloned into the pCAGGS vector has been described previously (Enomoto et al., 2005). The c.2313delT mutation was introduced into the wild-type CCDC88A sequence by sitedirected mutagenesis using the QuikChange II Site-Directed Mutagenesis kit (Agilent), and a haemagglutinin (HA) tag added to the N-terminus. Expression of wild-type and mutant proteins were analysed by western blot.

CCDC88A knockout mouse
The CCDC88A knockout mice and conditional Nestin-cre knockout mice strains have been reported previously (Kitamura et al., 2008;Asai et al., 2012). Full details of brain dissection, staining and analysis are described in the Supplementary material.

Clinical features of the family
The clinical features of the affected children are summarized in Table 1, and individual histories are given in the Supplementary material.
All affected children presented at birth with microcephaly [Occipito-frontal head circumference (OFHC) of À 3 standard deviation (SD) to À 4 SD, typically 529 cm] and moderately severe hypotonia. Weight and length at birth were normal. Excepting the microcephaly, they were not dysmorphic, nor did they have any extracranial congenital anomalies. All spent 2-4 weeks in hospital until satisfactory breathing and feeding were established. All developed seizures associated with anoxia presenting at birth or by 1 month. In all three subjects this progressed to flexion infantile spasms with hypsarrhythmia on EEG within the first year of life. Seizures became tonic/clonic by the third year, fits occurred frequently (at least weekly throughout life), status epilepticus necessitating hospital admission at least once in all three, and their seizures were persistently difficult to control necessitating frequent changes of medication and dosage. Over the first 6 months poor visual attention, fixing and following were noted, and all were diagnosed as having severe cortical visual inattention. All had moderate optic atrophy but otherwise a normal retina on fundoscopy; one had minor bilateral megalocornea (10 mm at 6 months, but there were no associated problems from this, such as glaucoma). Puffiness of the maxillary region of the face, and the dorsum of the hands and feet was present from birth, and persisted, but did not spread in extent (Fig. 1). Over the first 3 years of life it became clear that all had profound cognitive delay. Severe motor delay was also present with no development, with a combination of central hypotonia and peripheral hypertonia. Thoracolumbar kyphoscoliosis (with no vertebral anomalies) and hip flexion deformity developed by 3 years in two (the third not being old enough). The degree of microcephaly gradually increased over the years of follow up to À 5 SD to À 6 SD by 3 years. MRI was performed on all three affected children and showed the consistent and symmetrical findings shown in Fig. 1: widespread coarse pachygyria with the posterior regions of the brain more affected than anterior, polymicrogyria prominent in the sylvian fissures, dilated ventricles (reflecting brain substance loss and not raised intraventricular pressure), hypoplastic corpus callosum, subependymal cysts, and hypoplastic pons. One of the three affected children had a small cerebellum shown on an MRI scan at 3 months.
The two families were related, and each set of parents were first cousins (Fig. 2). The parents were healthy, and of normal cognitive and physical abilities, and had no abnormal neurological features. The family were ethnic white Caucasians. Each of the two families also had normal children. In one family a deceased daughter had complete sex reversal and absent adrenal glands, but no PEHO features. No unusual events occurred in the pregnancy or delivery of the affected children, nor was there any exposure to potential teratogens during the pregnancies. We concluded that the condition was inherited as an autosomal recessive trait. The affected children in this family fulfilled the diagnostic criteria for PEHO syndrome; however, the presence of structural brain anomalies (polymicrogyria nor pachygyria) were not described in the original Finnish cases, and so we considered that the better phenotypic diagnosis for the family was PEHO-like syndrome.

Identification of CCDC88A mutation
To identify the causative gene mutation in this family, exome analysis was performed on genomic DNA from two affected children (Patients 1 and 2, Fig. 2A), the raw sequences mapped to the GRCh37 reference human genome, and any possible mutations shared in both children analysed for their potential to cause PEHO syndrome. We prioritized the search for mutations that would potentially alter the encoded protein (nonsense/frameshift/splice site/missense mutations), and to those that were both homozygous and within concordant homozygous segments shared between the two cousins. Only one deleterious mutation was identified, which was present in both. This was a homozygous frameshift mutation in the CCDC88A gene (c.2313delT) creating a TAG stop codon at the site of the frameshift (p.Leu772X) (Fig. 2B). Segregation of the mutation was confirmed by Sanger sequencing as homozygous in all three affected children and heterozygous in all of their parents (Fig. 2C). The mutation was not identified in the 1000 Genomes Project or ExAC database, and we did not find it in 4 200 ethnically matched control chromosomes tested by Sanger sequencing.
CCDC88A encodes the actin binding protein girdin, which has previously been demonstrated as a binding partner of Akt and to play essential roles in the migration of fibroblasts, endothelial cells, cancer cells and neuronal cells (Enomoto et al., 2005(Enomoto et al., , 2009Jiang et al., 2008;Kitamura et al., 2008). Mice with germline deletion of girdin exhibit neuronal migration defects which result in hypoplasia of the olfactory bulb, granule cell dispersion in the dentate gyrus and postnatal lethality, demonstrating a crucial role for girdin in the early development of the mouse brain .
The girdin protein consists of a conserved N terminal domain showing significant homology to the microtubule binding region of the Hook protein family and proposed dimerization domain (Enomoto et al., 2006), a coiled coil central domain, known to bind members of the G family of heterotrimeric G proteins (Le-Niculescu et al., 2005) and a C-terminal domain that specifies binding partners including actin and Akt (Fig. 2B) (Enomoto et al., 2006). We hypothesized that the introduction of a premature stop codon early within the coiled coil domain would likely result in nonsense mediated decay and hence complete loss of the translated product. However, we were unexpectedly able to identify a reduced amount of full length mRNA, carrying the p.Leu772X in the lymphocytes of Patient 1 (Fig. 2D and Supplementary Fig. 1), and were able to exclude the possibility of abnormal splicing of the exon containing the c.2313delT mutation ( Supplementary  Fig. 1). While it is highly likely that this mutation would nonetheless produce a complete knockout of the girdin protein, mRNA extracted from the brains of a Ccdc88a knockout mouse produced by excising exon 3 to produce a premature stop codon also displayed presence of the mutated Ccdc88a transcript , we proceeded to confirm the pathogenicity of the truncated girdin by cloning it into HEK293 cells as a haemagglutinin-tagged fusion protein. Wild-type girdin produced a single band at $250 kDa by western blot, whereas introduction of the c.2313delT mutation by site-directed mutagenesis resulted in a truncated product of around 105 kDa in size (Fig. 2E). Many of girdin's reported functions are reliant on its interaction with C-terminal binding partners (including actin, Par-3, Akt, multiple receptor tyrosine kinases and Gi3; Enomoto et al., 2005Enomoto et al., , 2006Ohara et al., 2012;Lin et al., 2014;Bhandari et al., 2015) and its guanine nucleotide exchange factor domain, membrane binding region and a number of functional phosphorylatable residues lie downstream of the premature stop codon introduced (Bhandari et al., 2015). It seems probable that, in the unlikely event that the truncated protein is produced in vivo, the function of the mutant protein would be highly disrupted due to its absent C terminal functions.

Features of PEHO syndrome in Ccdc88a knockout mice
Having proven the potential pathogenicity of the p.Leu772X mutation in cellular models we asked if the knockout of CCDC88A in the mouse might mimic features of PEHO syndrome and could be used to identify molecular defects in the mouse brain responsible for the PEHO phenotype, which are unable to be investigated in affected humans. Mice with a knockout of CCDC88A have previously been reported to show a mesial-temporal lobe epilepsy phenotype, and this was again found . The mice also show postnatal lethality and postnatal growth retardation being reported underweight at death . This phenotype is generally consistent with that of PEHO syndrome.
We proceeded to assess the brains of Ccdc88a deficient mice at an anatomical level. Completeness of corpus callosum development in Ccdc88a deficient mice was assessed with Nissl stained coronal sections of neonatal (postnatal Day 0) brain tissue. There was no global difference in appearance of corpus callosum between wild-type and Ccdc88a deficient mouse, while there was a significant developmental defect in the caudal end of corpus callosum in knockout mice (Fig. 3A-C). This may match the finding of a hypoplastic corpus callosum observed in CCDC88A mutant human patients. Layer structures of cerebrum at postnatal Day 12 were compared between wild-type and knockout using Kluver-Barrera staining and immunohistochemistry with layer specific markers (Cux1 for layer II-III, and Ctip2 for layer V). With Kluver-Barrera staining, no 'polymicrocyria' like lesions were observed. Immunohistochemistry with layer markers didn't detect any abnormal layer structure, except that staining intensity of Ctip2 was slightly less in knockout mice. To assess optic nerve atrophy, the cross-section area of optic nerves were measured with ImageJ (National Institute of Mental Health, Bethesda, MD), and averaged. There was no significant difference between wild-type and knockout mice (wild-type versus knockout, n = 6 versus 5, mean 0.0241017 versus 0.025981 mm 2 , SD 0.003509 versus 0.003313 mm 2 , t-test P-value 0.384). To assess microcephaly, analysis of overhead view pictures of brains from nestin-Cre driven Ccdc88a flox/flox mice (cKO) at postnatal Day21 were carried out using ImageJ. Projected area of cerebrum of cKO was significantly smaller than that of Ccdc88a wild-type/flox nestin-Cre (-) (control), while areas of cerebellum of cKO and control were similar ( Fig.  3D and E).

Discussion
We investigated a multi-affected consanguineous family of British origin with three children born with features of PEHO syndrome. By comparing the exomes of consanguineous first cousins we detected a single nucleotide deletion in exon 15 of CCDC88A (c.2313delT). This results in a frameshift and incorporation of a stop codon (p.Leu772X). Our results confirm an autosomal recessive pattern of inheritance for PEHO syndrome, and identify CCDC88A as a further gene essential for normal human brain development.
CCDC88A encodes the actin binding protein girdin, a multi-domain, non-receptor guanine exchange factor (GEF) for G proteins Gi1, G2 and G3, which has been attributed a wide range of functions including roles in cell migration and Akt/EGFR signalling (Enomoto et al., 2006;Beas et al., 2012;Bhandari et al., 2015), actin organization and cell motility (Enomoto et al., 2005), neurite outgrowth and migration, axon/dendrite development (Enomoto et al., 2009;Wang et al., 2011) and to regulate early endosome maturation (Beas et al., 2012). Girdin can link actin filaments through its C terminal domain and simultaneously bind the plasma membrane through the adjacent membrane binding domain at the protein C terminus. This allows girdin to function in actin cytoskeletal remodelling during cell mobility (Hu et al., 2015). Other genes reported to cause cortical lamination defects (lissencephaly, pachygyria and polymicrogryia), severe developmental delay and seizures, are also generally cytoskeletal proteins (e.g. the tubulins) or act upon cytoskeletal proteins [LIS1 (PAFAH1B1), DCX, reelin (RELN), NUDE (NDE1) and NDEL1], and similar to CCDC88A, have complex multiple interactions, and functions (Guerrini and Dobyns, 2014). Postnatally, girdin is expressed in the majority of mammalian tissues; however, during embryonic development, the expression pattern of CCDC88A is specifically restricted, both spatially and temporally to regions of the fore-, mid-and hindbrain as well as the dorsal root ganglia, somites, limbs, nasal processes and branchial arches (Simpson et al., 2005). This specific expression profile implies a potentially important role during embryonic development.
The identification of a truncating mutation in CCDC88A in individuals with PEHO syndrome confirms a crucial function for girdin during embryonic development. Patients with a mutation in CCDC88A resulting in either a truncated girdin or absence of the protein due to  Paraffin embedded brain tissues from wild-type mice (WT) or straight Ccdc88a deficient mice (KO) were coronally sectioned at 7-mm thickness. All sections were serially numbered using the rostral end of hippocampus as a reference section. Rostral sections are positively numbered, and caudal sections were negatively numbered. (B) Nissl stained coronal sections of wild-type and knockout brain. Positions of corpus callosum (CC) relative to the rostral end of hippocampus were similar between wild-type and knockout mice (at around + 120 section), while the caudal end of knockout mice corpus callosum was relatively displaced forward compared with the caudal end of wild-type corpus callosum (wild-type at around À20 section level, knockout at around the reference section). hipp = rostral end of hippocampus. Black square indicates the area of each inset at bottom right corner. (C) Statistical analysis of positions of corpus callosum relative to the rostral end of hippocampus. Three replicates from each genotype were analysed. Section numbers of the rostral end of corpus callosum from the reference section were averaged. In accordance with the observations in B, the caudal ends of corpus callosum of knockout were significantly displaced to rostral side. Values are shown as mean plus standard error. T-test P values less than 0.05 were considered significant. (D) Overhead view of Ccdc88a wild-type/flox:nestin-Cre(À) control mouse (Con) and Ccdc88a flox/flox:nes-Cre( + ) conditional knockout mouse (cKO). (E) Horizontal projected area was measured and analysed. Cerebrum of cKO was significantly smaller than control mice, while areas of cerebellum were not different between controls and cKO mice. nonsense mediated decay, displayed all of the necessary clinical criteria to be diagnosed with PEHO syndrome. All presented at birth with microcephaly and moderately severe hypotonia and developed seizures that progressed into infantile spasms with hypsarrhythmia on EEG. All also displayed progressive brain atrophy, optic atrophy, severe cognitive delay and puffiness of the maxillary region of the face, and the dorsum of the hands and feet. Reassessment of knockout mice for Ccdc88a demonstrated further confirmation of CCDC88A as a PEHO-causing gene, with mice displaying a similar phenotype with mesial-temporal lobe epilepsy and early demise, and structural brain developmental defects affecting the corpus callosum and cerebrum. A lack of macroscopic pachygyria was expected as the murine cerebral cortex is 'smooth' and without a gyral pattern. However, the finding of only a minor (layer 5) cortical lamination defects in Ccdc88a mice may have been unexpected; the human/ mouse correlations for other similar brain-cytoskeletal genes are variable and unexplained: e.g. DCX mutations cause pachygyria in humans and the knockout mouse shows only hippocampal lamination defects (des Portes et al., 1998;Corbo et al., 2002); NDE1 mutations cause lissencephaly in humans and the knockout mouse has microcephaly, but cortical lamination is mostly preserved (Feng and Walsh, 2004;Wynshaw-Boris et al., 2010).
Recently, an individual with PEHO syndrome was reported who harboured a de novo missense mutations in KIF1A, and the authors noted up to six similar previously reported cases (Langlois et al., 2015). Of other individuals with significant mental handicap and KIF1A missense mutations, most had prominent and progressive spasticity and neuropathy-findings atypical of PEHO syndrome (Okamoto et al., 2014;Esmaeeli Nieh et al., 2015;Lee et al., 2015;Ohba et al., 2015). Furthermore, KIF1A mutations have also been reported to cause dominant and recessive pure spastic paraplegia, recessive sensory neuropathy and dominant mental retardation (OMIM: 601255). Thus, the PEHO syndrome phenotype can also have a dominant, and not just a recessive, inheritance pattern caused by a subset of KIF1A missense mutations.
To conclude, we report the first identification of a gene, CCDC88A, which when mutated causes an autosomal recessive PEHO-like neurodevelopmental phenotype. We detail the human phenotype caused by the lack of functioning GIRDIN, and found similar neurological features and anatomy in CCDC88A knockout mouse.