Homozygous mutation of STXBP5L explains an autosomal recessive infantile-onset neurodegenerative disorder

Abstract

We report siblings of consanguineous parents with an infantile-onset neurodegenerative disorder manifesting a predominant sensorimotor axonal neuropathy, optic atrophy and cognitive deficit. We used homozygosity mapping to identify an ∼12-Mbp interval identical by descent (IBD) between the affected individuals on chromosome 3q13.13-21.1 with an LOD score of 2.31. We combined family-based whole-exome and whole-genome sequencing of parents and affected siblings and, after filtering of likely non-pathogenic variants, identified a unique missense variant in syntaxin-binding protein 5-like (STXBP5L c.3127G>A, p.Val1043Ile [CCDS43137.1]) in the IBD interval. Considering other modes of inheritance, we also found compound heterozygous variants in FMNL3 (c.114G>C, p.Phe38Leu and c.1372T>G, p.Ile458Leu [CCDS44874.1]) located on chromosome 12. STXBP5L (or Tomosyn-2) is expressed in the central and peripheral nervous system and is known to inhibit neurotransmitter release through inhibition of the formation of the SNARE complexes between synaptic vesicles and the plasma membrane. FMNL3 is expressed more widely and is a formin family protein that is involved in the regulation of cell morphology and cytoskeletal organization. The STXBP5L p.Val1043Ile variant enhanced inhibition of exocytosis in comparison with wild-type (WT) STXBP5L. Furthermore, WT STXBP5L, but not variant STXBP5L, promoted axonal outgrowth in manipulated mouse primary hippocampal neurons. However, the FMNL3 p.Phe38Leu and p.Ile458Leu variants showed minimal effects in these cells. Collectively, our clinical, genetic and molecular data suggest that the IBD variant in STXBP5L is the likely cause of the disorder.

Introduction

A considerable proportion of rare disease has a genetic aetiology (1). Despite contemporary progress in the identification of causative gene mutations, finding answers for many families, often with only one or a small number of affected individuals, is challenging. This is mostly because the power of genetic mapping is limited (2) by the genetic model and the number of meiosis separating affected individuals. Historically, families with consanguinity have been instrumental in disease gene identification. The application of massively parallel sequencing in such families has led to disease gene discovery ‘en masse’, underscoring the importance of the knowledge of genetic relationships for disease variant (and gene) identification (3–5).

We report a brother and sister, children of first cousin parents, with a mixed central and peripheral neurodegenerative disorder dominated by infantile-onset axonal neuropathy, optic atrophy and cognitive delay. By homozygosity mapping, we identified a homozygous IBD variant in STXBP5L (c.3127G>A, p.Val1043Ile, syntaxin-binding protein 5-like) as the most likely cause. In addition, we identified variants in FMNL3 (c.114G>C, p.Phe38Leu and c.1372T>G, p.Ile458Leu), predicted to be deleterious by in silico analysis, using a compound heterozygous model of inheritance. While both genes were investigated, our follow-up cell and molecular studies suggest that the variant in the STXBP5L gene, which is involved in a highly conserved pathway of vesicular traffic, is most likely responsible.

Results

Homozygosity mapping and identification of STXBP5L and FMNL3 gene mutations

A family of middle-eastern ancestry was identified with two siblings of consanguineous parents affected by a phenotypically consistent neurodegenerative disorder (Fig. 1A; detailed clinical description in Materials and Methods). Homozygosity mapping using a rare, autosomal recessive model identified a single region of homozygosity by descent (HBD) shared between the two affected individuals on chromosome 3 between single-nucleotide polymorphisms (SNPs) rs13089846 and rs1355563 [chr3: 110 496 376–122 397 033 bp (hg19; UCSC genome browser)], (Fig. 1B, Supplementary Material, Fig. S1). This interval achieved the maximum attainable LOD score of 2.31 (Fig. 1B). The region covers 11 900 657 bp containing ∼98 annotated genes and 8 miRNA. We employed three sequencing strategies to analyse this family for potentially disease-causing variants: (i) targeted enrichment and massively parallel sequencing (Roche 454) of all sequences coding for exons, miRNA and conserved domains from the region of IBD sharing; (ii) family-based (parents and affected sibs) whole-exome sequencing and (iii) family-based whole-genome sequencing (both Illumina). Finally, all sequence gaps in coding regions of the IBD interval that were not adequately covered by next generation sequencing were amplified by polymerase chain reaction (PCR) and Sanger sequenced to ensure 100% coverage. Only one novel variant in the HBD interval resulted in an altered protein sequence; it occurred in syntaxin-binding protein 5-Like (STXBP5L c.3127G>A, p.Val1043Ile [CCDS43137.1]) (Fig. 1C). Family-based whole-exome and whole-genome sequence data confirmed the homozygous variant in STXBP5L but also identified compound heterozygous variants in FMNL3 on chromosome 12q13.12 (c.114G>C, p.Phe38Leu and c.1372T>G, p.Ile458Leu; [CCDS44874.1]) that segregated with disease under a recessive inheritance model (Fig. 1C). No other unique coding sequence variants or CNVs identified segregated with the disease in the family.

Figure 1.

(A) Pedigree with affected male and female siblings of consanguineous parents indicating likely autosomal recessive inheritance. *DNA collected for mapping and subsequent segregation of mutations. (B) Homozygosity mapping for chromosome 3 performed using data obtained from Illumina 660 Quad SNP chips. A single peak was identified using MERLIN between rs2673391 and rs1355563 using both parametric (black, LOD = 2.31) and non-parametric (green, LOD = 1.81) analyses. (C) Segregation of the mutation in STXBP5L and FMNL3 with disease was confirmed by Sanger sequencing in all available family members. Sequence traces for the parents and the affected sibs III:2 and III:6 are shown. (D) STXBP5L has 14 WD repeats (as indicated by the key); the Val1043Ile mutation occurs in the 14th repeat (indicated by the arrowhead and line), which is adjacent to the C-terminal R-SNARE domain (as indicated by the key). (E) FMNL3 has multiple domains (GBD, GTPase-binding domain; DID, diaphanous inhibitory domain; DD, dimerization domain; formin homology 1 (FH1), 2 (FH2) and 3 (FH3) domains; DAD, diaphanous autoregulatory domain). The Phe38Leu and Ile458Leu are located in the N-terminus and between DD and FH1 domains, respectively. Sequences related to (F) STXBP5L and (G) FMNL3 were identified by tblastn search, as well as HomoloGene and OrthoDB (6). Full-length protein sequences were then aligned using the CLUSTALW algorithm and sorted by distance; a section of the full-length alignment of each protein is shown with the mutant residue marked with the arrowhead.

Functional studies on STXBP5L gene mutation

STXBP5L is a large multidomain protein with an N-terminal WD domain composed of 10 WD40 repeats, a variable domain containing a further 4 WD40 repeats and a C-terminal R-SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) domain (7). The STXBP5L p.Val1043Ile mutation is located within the fourth WD40 repeat of the variable domain (Fig. 1D). Comparison of sequences similar to STXBP5L by CLUSTALW alignment showed conservation of p.Val1043 in either STXBP5L or the closely related STXBP5 (more commonly referred to as Tomosyn-1) in all mammals and birds but not fish or amphibian species examined (Fig. 1F). To exclude the possibility that a mutation in STXBP5L might be compensated for by STXBP5, we determined the expression of these genes by semi-quantitative RT-PCR in 18 different human tissues including disease target tissues, such as brain and spinal cord. STXBP5L and STXBP5 expression in brain and spinal cord was confirmed by real-time RT-qPCR. Taken together, the data showed significantly higher STXBP5L expression compared with the STXBP5 expression in brain and spinal cord (Fig. 2A), an expression pattern consistent with the observed pathology in the affected children. The STXBP5L p.Val1043Ile change is highly conserved but otherwise predicted to have minimal impact on protein function (Table 1). We therefore sought to determine the effect of the STXBP5L p.Val1043Ile mutation using a SNARE-dependent human growth hormone (hGH) secretion assay in rat PC12 cells (7). Expression of the STXBP5L p.Val1043Ile mutant enhanced inhibition of K⁺-stimulated exocytosis when compared with wild-type (WT) STXBP5L and vector-only transfected cells (Fig. 2B). The STXBP5L p.Val1043Ile-mediated inhibition of exocytosis was comparable with STXBP5, which was used as a positive control (Fig. 2B). Differences in inhibition were not due to varying levels of the expressed proteins (Fig. 2B, inset). To confirm this unexpected enhanced exocytosis inhibition in cells expressing the STXBP5L p.Val1043Ile protein, we performed patch clamp exocytosis assay on PC12 cells expressing WT or MT (p.Val1043Ile) STXBP5L protein (Fig. 2C). These data showed that calcium-dependent exocytosis was reduced by 42% in the PC12 cells expressing the STXBP5L p.Val1043Ile compared with those expressing the WT protein. However, there was no statistical difference between the PC12 cells transfected with WT or enhanced green fluorescent protein (EGFP) (Fig. 2C).

Figure 2.

(A) Expression of STXBP5L, the closely related STXBP5 and FMNL3 compared with PPIA as measured by semi-quantitative RT-PCR, size-fractionated by agarose gel (3%) electrophoresis, stained with ethidium bromide and visualized under UV light. STXBP5L and STXBP5 expression was also assayed in brain, spinal cord and heart by real-time RT-qPCR and normalized to HPRT1 expression (right panel). Note that STXBP5L expression is significantly (Student's two-tailed t-test) higher than STXBP5 in the brain and spinal cord. (B) Effect of STXBP5L p.Val1043Ile mutation (STXBP5L-MT) compared with STXBP5L-WT and STXBP5 on the secretion of hGH in rat PC12 cells ectopically expressing comparable levels of these proteins (inset western blot; anti-Myc immunoreactive signals from three independent western blots quantified using ImageJ software and normalized to the housekeeping anti-β-tubulin signals were as follows: Lane 2, 1.00 ± 0.09; Lane 3, 1.00 ± 0.15 and Lane 4, 0.94 ± 0.05). STXBP5L-MT exhibits enhanced suppression of hGH secretion compared with STXBP5L-WT. For each sample, data were normalized to unstimulated (light bars) levels of hGH. Columns represent the mean of three independent experiments performed in triplicate; error bars show SDs. Statistical significance was measured between all groups using a Student's two-tailed t-test. All data were corrected for multiple comparisons using the false discovery rate step-up method. (C) Membrane capacitance (calcium-dependent exocytosis) is reduced in the STXBP5L-MT compared with STXBP5L-WT. The figure shows the change in membrane capacitance at 30, 60, 90 and 120 s after calcium infusion in PC12 cells ectopically expressing with WT, MT or GFP alone proteins. The WT versus MT were significantly different at P < 0.05 at the 60 and 90 s time points.

Table 1.

Summary of rare missense variants co-segregating with the disease in the family

Chr	Position (hg19)	Ref.	Obs.	Gene	Variant	Genotype	SIFT	Polyphen2 (prediction)	Log-rank test (prediction)	Mutation taster (prediction)	Mutation assessor (prediction)	CADD (PHRED scaled C-score)	RVIS (percentile)	HI	FATHMM	GERP++	PhyloP	SiPhy
3	121 132 111	G	A	STXBP5L	c.3127G>A, p.Val1043Ile [CCDS43137.1]	Hom	0.98	0.18 (Benign)	0.000111 (Damaging)	0.013596 (Neutral)	0.695 (Neutral)	16.36	24.63	38%	0.93	6.04	2.873	14.1631
12	50 045 947	T	G	FMNL3	c.1372T>G, p.Ile458Leu [CCDS44874.1]	Het	0.54	0.33 (Benign)	0.000355 (Damaging)	0.922444 (Damaging)	0.6 (Neutral)	6.28	8.1	71.80%	1.02	5.36	2.171	14.6471
12	50 100 850	G	C	FMNL3	c.114C>G, p.Phe38Leu [CCDS44874.1]	Het	0	0.117 (Benign)	0.000378 (Neutral)	0.851383 (Damaging)	2.52 (Medium)	15.03	8.1	71.80%	−2.41	2.05	0.656	6.6819

SIFT, Polyphen-2, log-rank test, Mutation taster, Mutation assessor, FATHMM, GERP++, PhyloP and SiPhy annotations applied using ANNOVAR (8) and the data collated in dbNSFP from the original prediction programmes (9); CADD (combined annotation dependent depletion) (10); RVIS (residual variance intolerance score; a gene level annotation) (11); HI (haploinsufficiency index; a gene level annotation) (12).

We also determined the cellular consequences of WT or STXBP5L p.Val1043Ile overexpression in mouse E18.5 primary hippocampal neurons. We observed that the overexpression of STXBP5L-WT, but not STXBP5L p.Val1043Ile, promotes significant axonal outgrowth in these neurons, suggesting a loss of function of STXBP5L p.Val1043Ile protein in inducing axonal growth (Fig. 3A–C). Our data do not, however, establish if the axonal phenotype is the consequence of altered exocytosis.

Figure 3.

(A) STXBP5L, but not variant STXBP5L, promotes axonal outgrowth. Primary hippocampal neurons expressing EGFP only or with either STXBP5L (STXBP5L-WT) or the mutant STXBP5L p.Val1043Ile (STXBP5L-MT) protein were stained for marker proteins of dendritic (MAP2, cyan) and axonal (TAU1, red) structures. Cells were counterstained with DAPI (blue). Representative images of transfected neurons are shown. (B) Mean primary axonal length of transfected neurons. (C) Mean number of neurite (i.e. axonal, dendritic and total) termini. *P < 0.01 Student's two-tailed t-test. Error bars represent ±SD. Experiment conducted in triplicate with at least 30 neurons scored per replicate (in total, Control n = 107, STXBP5L-WT n = 100 and STXBP5L-MT n = 102). (D) Mean primary axonal length and (E) Mean number of neurite (i.e. axonal, dendritic and total) termini of neurons overexpressing FMNL3-WT, p.Phe38Leu, p.Ile458Leu or p.Phe38Leu and p.Ile458Leu. No difference in axonal outgrowth or arborization among different treatments was observed. Error bars represent ±SD. Experiment conducted in triplicate with at least 35 neurons scored per replicate (in total, FMNL3-WT n = 118, p.Phe38Leu n = 116, p.Ile458Leu n = 112, p.Phe38Leu and p.Ile458Leu n = 111).

Functional studies on FMNL3 gene mutations

Human FMNL3 is member of a large family of dimeric multidomain proteins that perform a range of essential functions such as long-range vesicle transport, cell mobility, heart morphogenesis and formation of dendritic spines in neurons through their actin nucleation and elongation activities (13). Although all 15 members of the human formin family contain a formin homology (FH2) domain, there is a considerable difference in domain architecture among different formins that are responsible for a range of cellular functions (14). The p.Phe38Leu and p.Ile458Leu changes are located in functionally undefined regions of the FMNL3 protein (Fig. 1E). Compared with STXBP5L and STXBP5, a modest variation was detected in FMNL3 expression among different human tissues (Fig. 2A). Furthermore, we detected no difference in axonal outgrowth or arborization in mouse primary hippocampal neurons overexpressing FMNL3-WT, FMNL3 p.Phe38Leu, FMNL3 p.Ile458Leu or combined FMNL3 p.Phe38Leu and p.Ile458Leu (Fig. 3D and E).

Discussion

We sought to resolve the genetic aetiology of a unique mixed central and peripheral neurodegenerative disorder, dominated by infantile-onset axonal neuropathy (the characteristics of which suggest a component neuronopathy), progressive optic atrophy and central manifestations with marked cognitive delay and generalized cerebral atrophy evident upon neuroimaging. We chose to limit our search space initially to a region of HBD sharing taking advantage of consanguinity in the family (3,15), which strongly supported the likelihood of a homozygous variant underpinning the disorder (LOD > 2). However, the degree of inbreeding was close (first cousin); thus, there was the possibility that the inbreeding was coincidental and the hypothesized genetic model of homozygosity for the causal variant was incorrect. We identified a likely causative mutation in STXBP5L in the HBD interval, but when we broadened our search space using whole-exome and whole-genome sequence data, we also identified compound heterozygous missense variants in FMNL3 segregating with the disease, which is consistent with a recessive genetic model, but not a homozygous genetic model. This gene is located in a region that is IBD = 2 for both affected siblings, but not HBD. Thus, based on the evidence from the linkage analysis models, STXBP5L is the stronger candidate, but FMNL3 cannot be excluded. Hence, we sought to further functionally characterize each to identify whether one or both mutant genes contributed to the disease.

The mutation in STXBP5L is located in a residue that is conserved in mammals and birds and is located within a known structural domain of the protein (7). Protein modelling predicts that the STXBP5L p.Val1043Ile change does not alter the structure of the protein but most likely impacts local interaction with other cellular factors (Uri Ashery, personal communication). While STXBP5L has been poorly studied until now, more recent functional data on rat STXBP5L and its paralog STXBP5 isoforms show that both proteins cause inhibition of exocytosis in neurosecretory cells (7). Mouse Stxbp5 and Stxbp5l have 82% amino acid sequence similarity in the WD40 domains, and the equivalent residue to STXBP5L, p.Val1043 is identical in both proteins (16). Stxbp5l expression in the brain increases throughout embryonic development and shows highest expression in the CA2 region of the hippocampus and cerebellum (16). STXBP5L and STXBP5 regulate docking and fusion of synaptic vesicles with the presynaptic membrane, a process of crucial importance for normal neurotransmission. In addition, whereas STXBP5 acts as a regulator of neurite growth (17), STXBP5L is now reported to be a negative regulator of insulin secretion (18), suggesting that these proteins may have multiple overlapping as well as independent cellular roles. Consistent with this idea, STXBP5 and STXBP5L are differentially expressed in different tissues and during development (16,18). We observed a significantly higher level of STXBP5L expression than STXBP5 in spinal cord. As these two proteins are predicted to perform similar functions in the cell, it may be that one can functionally compensate for the other. Thus, the presence of higher levels of STXBP5L in spinal cord raises the possibility that this tissue could be more susceptible to the adverse effects of the STXBP5L p.Val1043Ile variant. STXBP5 was first identified as a binding partner of Syntaxin 1, and this interaction was shown to reduce Ca²⁺-dependent exocytosis mediated through Syntaxin 1 (19). Since then, STXBP5 has been shown to be an integral part of the exocytosis machinery in both neuronal and non-neuronal cell types (20–23). Most importantly, down-regulation of exocytosis requires structural integrity of the entire N-terminal portion of the protein excluding the R-SNARE domain (24). The STXBP5L p.Val1043Ile substitution is within the region of the protein required for the inhibition of exocytosis. Indeed, increased suppression of hGH and patch clamp exocytosis by STXBP5L p.Val1043Ile suggests that the mutation contributes to the severe neurodegenerative clinical outcomes in the affected siblings. It must be noted that these two exocytosis assays represent different readouts. The hGH assay measures total release via exocytosis, whereas the membrane capacitance approach measures total net change in plasma membrane size and as such represents the balance between exocytosis and endocytosis occurring simultaneously. Whether STXBP5L or STXBP5L p.Val1043Ile affect endocytosis as well as exocytosis cannot be determined from these results but may be worth additional study. Furthermore, if Stxbp5l has any effects on the vesicle fusion pore, which it may given the interactions of tomosyn proteins with members of the SNARE protein complex (19,25–27), then this may also confound direct comparisons between the two approaches we have used in our study. If WT or STXBP5L p.Val1043Ile also regulates the fusion pore, this will also effect exocytosis output.

The apparent gain-of-function effect of the STXBP5L p.Val1043Ile does not imply dominance because the interaction of WT and STXBP5L p.Val1043Ile proteins was not assessed in these assays. There are precedents for masking of apparently dominant biochemical effects by mutations that are inherited in a recessive manner such as β-amyloid precursor protein p.Ala673Val, which is prevented from enhancing β-amyloid aggregation in the presence of the WT protein (28). In addition, the affected individuals have two copies of STXBP5L p.Val1043Ile instead of a single copy in the unaffected parents and four siblings. Taken together, it is therefore possible that disease symptoms are caused by increased dosage of the STXBP5L p.Val1043Ile protein and/or unmasking of its cellular effects in the absence of WT STXBP5L, although direct evidence supporting these possibilities is lacking. Notably, recent systematic investigations on Stxbp5l null mice showing that Stxbp5l supports normal motor performance by regulating transmitter release suggest potential involvement of Stxbp5l in neuromuscular disorders (29).

Stxbp5 has higher levels of gene expression during embryonic stages of development whereas Stxbp5l is more active during postnatal stages (16). The developmental switching to Stxbp5l as the predominant paralog in the neonatal mouse is consistent with the age of onset of disease in the affected individuals. Furthermore, a deletion containing STXBP5 coding sequence in a patient with autism, intellectual disability and seizures (30) and the recent demonstration of STXBP5 p.Leu412Val in individuals with the same phenotype, and STXBP5 p.Tyr502Cys in individuals with seizures (31), now provide examples indicating that perturbed STXBP5 function can lead to a variety of neurodevelopmental and neuropsychiatric disorders.

The FMNL3 variants change amino acid residues that are highly conserved among diverse organisms. Human formins are an architecturally complex family of proteins (14). Whereas some of the human formins have been functionally well-characterized, studies on the FMNL subgroup have lagged behind. The p.Phe38Leu and p.Ile458Leu changes are located in the functionally uncharacterised regions of the FMNL3 protein. Formins play a role during axonal morphogenesis and have been implicated in numerous diseases (32–35). For example, a number of INF2 mutations are shown to cause axonal loss in focal segmental glomerulosclerosis-associated Charcot–Marie–Tooth neuropathy (32). FMNL2, the formin closest to FMNL3, has been implicated in colorectal cancer and mental retardation (36,37), but FMNL3 mutations have not been associated with any disease yet. FMNL3 overexpression induces formation of filopodia and membrane protrusions (38). However, we observed no significant effect of exogenous p.Phe38Leu and p.Ile458Leu, separately or together, on mouse hippocampal neurons.

Predicated upon the combined genetic and molecular data, we contend that the homozygous STXBP5L variant p.Val1043Ile gives rise to the observed phenotype; however, the possibility that compound heterozygous variation in FMNL3 is complicit cannot be absolutely excluded.

Materials and Methods

Clinical description: Case 1 (III:2) was a female child born at 38 weeks of gestation; her birth was uncomplicated and followed an uneventful antenatal period. In utero movements were appropriate, and there was no polyhydramnios. The perinatal period was dominated by irritability, poor feeding and prominent gastric reflux. Growth was proportionate with expected values with the exception of weight loss at end-stage disease; head circumference was preserved within the 10th–25th centiles (gender matched) throughout. The child was generally hirsute with isolated premature pubarche recorded at 29 months. A progressive axonal sensorimotor neuropathy (confirmed neurophysiologically) manifested at 3 months of age, presenting with global involvement of the appendicular and axial musculature, with marked hypotonia, loss of myotatic reflexes and muscle wasting. Fasiculations were not evident. A thoracolumbar scoliosis and distal joint contractures progressed from this time. Independent sitting and mobility were never achieved. Sensory testing proved grossly intact to limited assessment. Dystonic posturing and upper limb dyskinesia presented at 8 months. At 12 months, mixed semiology seizures manifest, with generalized tonic–clonic activity and independent myoclonic events that were refractory to anticonvulsant therapy; at times evolving to status epilepticus. Restricted lateral gaze reflected supranuclear pathology at 8 months, whereas superimposed ocular paresis eventuated with end-stage disease. Progressive optic atrophy was prominent from 12 months of age. Auditory acuity appeared clinically intact throughout the disease course. Cognitive development was markedly delayed, plateauing at attainment of reactive smiling only. Death occurred at 30 months of age secondary to respiratory compromise.

Case 2 (III:6) was a male child born at 40 weeks of gestation who manifested a comparable, yet attenuated, disease course to his sister. His birth was uncomplicated and followed a normal antenatal period. Early feeding difficulties were transient and his perinatal period otherwise uncomplicated; right-sided cryptorchidism was noted. Growth remained proportionate with the exception of decreased body mass from 12 months of age; head circumference continued between the 50th and 75th (gender matched) centiles. Clinical manifestations of axonal neuropathy were evident from 3 months; however, disease progression proved slower than his sibling. Motor ability peaked with attainment of limited truncal rotation and sufficient upper limb power to enable reaching and object manipulation at 12 months. Lower limb power was disproportionately affected, and independent sitting and mobilization were not achieved. Further loss of axial and appendicular strength ensued with marked thoracolumbar scoliosis, loss of antigravity power and an inability to grasp objects at 72 months. Early loss of myotatic reflexes proved a consistent feature, with preservation of adductor responses only at 72 months. Bulbar dysfunction manifest from 36 months necessitated gastrostomy insertion. Bilateral buphthalmos, secondary to congenital glaucoma, was evident in infancy and while the child remained visually attentive at 72 months, deterioration of acuity occurred in the context of progressive optic atrophy. Saccadic slowing and restriction of gaze in both the horizontal and vertical planes was additionally noted. Auditory acuity remained intact to clinical assessment at 72 months. Cognitive development peaked with an established vocabulary of three words and a limited ability to identify body parts at 24 months. Subsequent loss of these skills eventuated, and at 72 months, the child was predominantly non-verbal, although he demonstrated reactive smiling to vocal and visual stimuli. Secondarily generalized seizures manifested at 72 months with subsequent development of non-lesional epilepsia partialis continua, characterized by left facial rhythmic clonic activity, proving refractory to anticonvulsant therapy. Unlike his sister, extrapyramidal features did not occur. Death occurred at 7 years and 8 months of age secondary to respiratory compromise in association with viral pneumonia.

Extensive investigations in both children failed to identify a biochemical abnormality or inborn metabolic aetiology to explain the observed pathology; screening molecular analysis proved non-contributory, including copy number variant analysis (Illumina HumanCytoSNP-12), CMT mutation panel, PLA2G6 sequencing (infantile neuroaxonal dystrophy), SMN1/2 MLPA (spinal muscular atrophy) and screening of common mitochondrial mutations. Limited electrodiagnostic assessment performed at 13 and 14 months, in Cases 1 and 2, respectively, demonstrated reduced amplitude of motor and sensory action potentials and slowed conduction velocities in all nerves screened. F-waves and H-reflex were not performed. Corresponding electromyography revealed large amplitude motor unit action potentials consistent with denervation; fibrillation potentials and positive sharp waves were not reported. Electroencephalography in Case 2 proved age appropriate at 9 and 42 months, with development of bilaterally asynchronous epileptiform discharges of spike-wave and poly-spike morphology at 72 months of age; focal discharges were not evident with onset of epilepsy partialis continua. Despite prominent seizures, serial interictal studies were normal in Case 1 (epileptiform features were only noted in the context of electrical status). Electroretinography, undertaken in Case 2 only, was normal at 18 and 36 months, whereas concomitant visual-evoked responses demonstrated diffuse, low-amplitude responses to flash stimuli and absent pattern reversal potentials below 210′ consistent with substantially reduced visual function.

Magnetic resonance imaging at 12 months of age (Case 1) revealed generalized cerebral atrophy, most prominently involving the temporal lobes and without associated signal change upon T2-weighted sequences. T1-weighted series of the spinal cord were normal. Serial imaging at 8, 40 and 63 months (Case 2) demonstrated mild cerebral atrophy, again in the absence of radiological leukodystrophy; atrophy of the intraorbital optic nerves was demonstrated at 40 months. Spinal series were normal, aside from the noted scoliosis and a capacious central canal. Single voxel proton magnetic resonance spectroscopy (TE 31 and 144 ms) performed at 63 months with voxels sited over the striatal grey matter and parietal white matter demonstrated a reduction in NAA/creatine within the striatal spectra only, suggestive of non-specific neuronal loss in this region. Additional spectral sampling was not performed.

Histological and ultrastructural examination of skeletal muscle (quadriceps), performed at 10 and 15 months in Cases 1 and 2, respectively, were non-specific, with features of neurogenic atrophy in each. Sural nerve biopsies, at 10 and 28 months, in Cases 1 and 2, respectively, revealed occasional autophagic vacuoles within degenerating axons; spheroids were absent and myelin sheaths structurally intact. Histology of skin (Cases 1 and 2), hepatic tissue, conjunctiva and rectal mucosa (Case 1 only) was normal.

Homozygosity mapping

Genotyping was carried out on all individuals from Generations II and III by the Australian Genome Research Facility using Infinium HumanHap660W-Quad BeadChip arrays (Illumina). A total of 12 163 SNPs with an average heterozygosity of 0.49 based on Caucasian (CEU) Hap Map data were chosen for analysis across the 23 chromosomes using the Perl script linkdatagen_illumina.pl (39). The program FEstim was used to estimate the inbreeding coefficient for genotyped individuals (40,41). The estimated inbreeding coefficients of 0 in the parents of Generation II and median inbreeding coefficient for the six children in Generation III of 0.068 validate the known first cousin relationship between the parents of these siblings and lack of further inbreeding loops above the Generation II. Linkage analysis was performed using the program MERLIN (42), for the autosomal chromosomes 1–22 assuming a fully penetrant recessive disease model with a disease allele frequency of 0.00001.

Sequence enrichment and massively parallel sequencing

A 385K NimbleGen array (Roche/NimbleGen) was designed to capture sequences corresponding to exons of known genes, miRNA and conserved sequences with LOD score of >50 based on the 28-way alignment of vertebrate species (43). Sequences were enriched from genomic DNA extracted from blood of III:6 and sequenced on one-fourth of a Roche 454 pyrosequencer run as described previously (44). Whole-exome sequencing was carried out by a service provider (Otogenetics, Norcross, GA, USA). Genomic DNA from II:1, II:2, III:2 and II:6 was enriched using a SeqCap EZ Human Exome Library v2.0 (Roche/NimbleGen). Enriched DNA was sequenced using a 100-bp, paired-end read strategy on an HiSeq 2000 (Illumina). Segregation was carried out by Sanger sequencing on PCR amplicons of genomic DNA extracted from blood generated using primers STXBP5LEx25F1, 5′ TGCTTCTAACAACCTGCATATGAT and STXBP5LEx25R1, 5′ CATTCTCTCAACCATGTCCAA and cycling conditions 95°C for 5 min; 35 cycles of 95°C for 15 s, 60°C for 15 s and 72°C for 20 s. hFMNL3-F1, 5′ CCGGCAAGATGCCGATGCCTGAG and hFMNL3-R1, 5′ TCAGAGGTTCCTGGGAAGACACTGGA for c.114G/C, and hFMNL3-F2, 5′ TTCTGCCACTCTGAGCCTCTGAGT and hFMNL3-R2, 5′ TGGCATGCCCTCACTCAGCTCTGCA for c.1372T/G. Platinum Taq polymerase High Fidelity (Life Technologies) with cycling conditions 94°C for 1 min; 35 cycles of 94°C for 15 s, 57°C for 15 s and 68°C for 30 s were used.

Sequenom genotyping assay

A specifically designed multi-plexed MALDI-TOF mass spectrometry (Sequenom) assay was used to genotype 84 middle-eastern controls for the STXBP5L homozygous variant (45). Genotypes were called by the MassARRAY System Typer version 4.0 software (Sequenom). Details of primers are available on request.

Bioinformatic analysis

Mapping and variant calling of 454 sequences was performed as previously described (44). Novel variants were defined by the following filters; at least 80% variant reads, absent from dbSNP130 and absent from the Nijmegen in-house variant database, which includes 1000 genomes data not already in dbSNP as well as over 100 in-house exome sequences. In total, this represents >2200 chromosomes screened. Variants were categorized based on genomic context relative to known genes with prioritization given to variants predicted to alter protein sequences (non-synonymous, nonsense and splice variants). Exome and whole-genome sequences were mapped to GRCh37/hg19 human genome reference sequences with the Burrows–Wheeler Aligner (46). Variants were called using the Genome Analysis Toolkit Unified Genotyper (GATK v2.8.1-g932cd3a) (47) and annotated with ANNOVAR (8).

Orthologues of STXBP5L were identified by tblastn query of the NCBI non-redundant (nr) database using a 150 aa window flanking p.Val1043. Full-length sequences were then aligned by CLUSTALW2 (48) using the European Bioinformatics Institute server. Alignments were visualized using jalview (49).

STXBP5L expression in human tissue panel

cDNA was synthesized from 2 μg of total RNA of a human tissue Master Panel (Clontech) with random hexamer primer using SuperScript III reverse transcriptase according to the recommended protocol (Life Technologies). The levels of STXBP5L, STXBP5 and FMNL3 expression were determined by semi-quantitative PCR on these cDNA samples. STXBP5L, STXBP5 and PPIA (housekeeping gene encoding peptidylprolyl isomerase A) were assayed using KAPA HiFi PCR Kit (Kapa Biosystems) under the following conditions: 98°C for 3 min; 24 cycles of 98°C for 10 s, 60°C for 10 s and 72°C for 15 s. FMNL3 expression was determined using Platinum Taq polymerase High Fidelity (Life Technologies) with cycling conditions 94°C for 1 min; 28 cycles of 94°C for 15 s, 59°C for 15 s and 68°C for 20 s. Primers used were as follows: STXBP5L-1811-F, 5′ AGGACAGTATTCCATGCCTCAATG and STXBP5L-1905-R, 5′ ATCTACCCACACCAATTGAATAAC; STXBP5-1744-F, 5′ CATCCATCTACCAGTAGCAGTTCATC and STXBP5-1869-R, 5′ CCAAACCAACTGAATAACTAGTTCTGT; hFMNL3-F 5′ GAACTAGAGAAGCAGCTGCTACAG and hFMNL3-R2. PCR amplification of PPIA was used to normalize cDNA levels among different tissues under the following conditions: 98°C for 3 min; 22 cycles of 98°C for 10 s, 60°C for 10 s and 72°C for 15 s. PPIA primers used were as follows: forward, 5′ GGCAAATGCTGGACCCAACACAAA and reverse, 5′ TGCTGGTCTTGCCATTCCTG. PCR products were resolved on 3% agarose gels and detected by staining with ethidium bromide. STXBP5L and STXBP5 expression was also determined in brain, spinal cord and heart by real-time RT-qPCR using SYBR Green master mix (Bio-Rad) and normalized to HPRT1 expression determined by real-time RT-qPCR using primers 5′ TGACACTGGCAAAACAATGCA and 5′ GGTCCTTTTCACCAGCAAGCT with cycling conditions 95°C for 10 min; 40 cycles of 95°C for 15 s and 60°C for 20 s.

Expression plasmids

Human STXBP5L and STXBP5 open-reading frames (ORFs) were PCR-amplified from an adult brain cDNA using Platinum Taq DNA polymerase High Fidelity (Life Technologies) and following primers:

STXBP5L-EcoRI-F 5′ GCAGGCTTCCGAATTCCCATGAAGAAGTTTAATTTCCGAAAAGTTTTG and STXBP5L-XhoI-R 5′ GCTGGGTCCTACTCGAGTCAGAATTGGTACCATTTCTTATCCTTG; STXBP5-XhoI-BamHI-F 5′ CACACACACTCGAGGGATCCCCATGAGGAAATTCAACATCAGGAAGGTGCT and STXBP5-NotI-KpnI-R 5′ CACACACACAGGTACCGCGGCCGCTCAGAACTGATACCACTTCTTATCTTTG.

STXBP5L and STXBP5 ORFs were cloned at EcoRI-XhoI and BglII-NotI sites, respectively, in-frame with the Myc-tag coding sequence in the pCMV-Myc vector (Clontech). Val1043Ile mutation was introduced in the pCMV-Myc-STXBP5L expression construct by overlap PCR using STXBP5L-V to I-F 5′ CAGGCATTATACCTGATATCTCCTACTGAAATTCAG and STXBP5L-V to I-R 5′ CTGAATTTCAGTAGGAGATATCAGGTATAATGCCTG. The constructs were confirmed by DNA sequencing.

The pcDNA3-FMNL3-Flag construct expressing C-terminally Flag-tagged human FMNL3 protein reported previously (38) was mutated to generate pcDNA3-FMNL3-Flag-Phe38Leu and pcDNA3-FMNL3-Flag-Ile458Leu constructs using Prime STAR Mutagenesis Kit (TAKARA) and the following primers:

FMNL3-Flag-Phe38Leu-F 5′ AAAGGTTGGCCCTGGTGCTGAGCTCC and FMNL3-Flag-Phe38Leu-R 5′ CCAGGGCCAACCTTTCCTCCAGCTCA; FMNL3-Flag-Ile458Leu-F 5′ GAGGCTCCTTAAAGAGAAGGAGGAGG and FMNL3-Flag-Ile458Leu-R 5′ TCTTTAAGGAGCCTCCGCAGGGTGTG. The constructs were confirmed by DNA sequencing.

hGH secretion assay

The hGH secretion assay was performed as previously described (7) with some modifications. Briefly, PC12 cells were plated in triplicate in Nunc Nunclon Surface 24-well plates and co-transfected with hGH (kindly provided by Prof Edward L Stuenkel, University of Michigan, MI, USA) and either pCMV-Myc, pCMV-Myc-STXBP5L-WT, pCMV-Myc-STXBP5L-MT or pCMV-Myc-STXBP5 expression constructs. The amount of transfected DNA for each treatment was equalized by adding varying amounts of the pCMV-Myc plasmid. Secretory assay was performed 48 h after transfection. The cells were washed with physiological saline solution (1× PSS; 15 mm HEPES, 145 mm NaCl, 5.6 mm KCl, 0.5 mm MgCl₂, 2.2 mm CaCl₂, 5.6 mm glucose, 2 mg/ml sodium ascorbate and 2 mg/ml fatty-acid-free bovine serum albumin, pH 7.4) for 5 min. The 1× PSS was then replaced with either 1× PSS or 1× stimulation solution (15 mm HEPES, 95 mm NaCl, 56 mm KCl, 0.5 mm MgCl₂, 2.2 mm CaCl₂, 5.6 mm glucose, 2 mg/ml sodium ascorbate and 2 mg/ml fatty-acid-free bovine serum albumin, pH 7.4) for 20 min at room temperature. The unstimulated and stimulated solutions were harvested and stored at −20°C until analysed. The cells were lysed in 1× lysis buffer (50 mm Tris–HCl, pH 8.0, 250 mm NaCl, 1% Triton X-100, 50 mm NaF, 0.1 mm Na₃VO₄). The levels of secreted and total hGH were measured using an hGH enzyme-linked immunosorbent assay kit (Roche). That PC12 cells used for secretion assays expressed comparable levels of the Myc-STXBP5L and Myc-STXBP5 proteins were validated by western blotting with mouse anti-Myc antibody as reported (50).

Patch clamp exocytosis assay

PC12 cells plated on poly-l-lysine-coated petri-dishes at a single-cell density were transfected with pEGFP-C1 alone or with either pCMV-Myc-STXBP5L-WT or pCMV-Myc-STXBP5L-MT expression constructs. That EGFP-positive PC12 cells expressed the STXBP5L proteins was established in pilot immunofluorescence experiments where Myc-STXBP5L proteins were detected using the mouse monoclonal anti-Myc antibody (Santa Cruz; 9E10) and Alexa Fluor 555 donkey anti-mouse IgG (Life Technologies). Four independent experiments showed that >85–90% EGFP-positive PC12 cells expressed the STXBP5L proteins. Capacitance measurements in the presence of high intracellular Ca²⁺ were utilized as an assay for exocytotic capacity in EGFP-positive PC12 cells. Whole-cell patch clamp was performed using an EPC-10 patch clamp amplifier and PatchMaster software (HEKA Electronik, Lambrecht/Pfalz, Germany). Patch pipettes were pulled from borosilicate glass and fire-polished, with resistance of 3–5 MΩ. Internal solution contained (mm): 70 K₂SO₄; 10 NaCl; 10 KCl; 10 HEPES; 10 EGTA; 10 CaCl₂, adjusted to pH 7.2 (calculated free Ca²⁺ concentration of 55 μm) (51). External solution contained (mm): NaCl, 150; KCl, 2.8; HEPES, 10; MgCl₂, 2; CaCl₂, 10; Glucose, 10; adjusted to pH 7.2 with NaOH. Capacitance measurements utilized the software Lock-in module of the PatchMaster software. Following attainment of the whole-cell configuration and equilibration of the high-Ca²⁺ intracellular solution, cell capacitance change was measured over time using a 30-mV peak-to-peak sine wave at a resting membrane potential of −80 mV. Data were expressed as change in capacitance per unit time. All experiments were carried out at room temperature (22–24°C).

Neuronal cell assay

Morphometric analyses were undertaken as previously described (15,52). Mouse primary hippocampal neurons were isolated from embryos harvested at E18.5 as described (53) and nucleofected immediately using nucleofector kit according to the manufacturer's instructions (Lonza) with plasmid expressing EGFP (pMAX-EGFP) together with either an empty plasmid (Control), expressing STXBP5L-WT, the variant STXBP5L (STXBP5L-MT), FMNL3-WT, FMNL3 p.Phe38Leu, FMNL3 p.Ile458Leu or FMNL3 p.Phe38Leu and FMNL3 p.Ile458Leu protein. Cells were cultured for 5 days in vitro and fixed with 4% PFA, and immunofluorescence was performed as previously described (15). Immunofluorescent staining of axons and dendrites were visualized using TAU1 and MAP2 marker proteins, respectively.

Funding

This work was supported by philanthropic funds, The Channel 7 Children′s Research Foundation of SA (R.K., E.H., M.A.C. and J.G.), the Netherlands Organization for Health Research and Development (A.H. and J.A.V.), an Australian Research Council Future Fellowship (M.B.), National Health and Medical Research Council Research Fellowship (J.G.) and Program Grants (M.B. and J.G.) and MS McLeod Fellowship of the Women′s and Children′s Hospital Foundation (M.A.C.).