Mutations in the PH Domain of DNM1 are associated with a nonepileptic phenotype characterized by developmental delay and neurobehavioral abnormalities

Abstract Background Dynamin 1 is a protein involved in the synaptic vesicle cycle, which facilitates the exocytosis of neurotransmitters necessary for normal signaling and development in the central nervous system. Pathogenic variants in DNM1 have been implicated in global developmental delay (DD), severe intellectual disability (ID), and notably, epileptic encephalopathy. All previously reported DNM1 pathogenic variants causing this severe phenotype occur in the GTPase and Middle domains of the dynamin 1 protein. Methods We used whole‐exome sequencing to characterize the molecular basis of DD and autistic symptoms in two identical siblings. Results The twin siblings exhibit mild to moderate ID and autistic symptoms but no epileptic encephalopathy. Exome sequencing revealed a genetic variant, c.1603A>G (p.Lys535Glu), in the PH domain of dynamin 1. Previous in vitro studies showed that mutations at Lys535 inhibit endocytosis and impair PH loop binding to PIP2. Conclusions Our data suggest a previously undescribed milder phenotype associated with a missense genetic variant in the PH domain of dynamin 1.


| BACKGROUND
Neurotransmission in the central nervous system depends on synaptic vesicle transport. Neurotransmitters are taken up by exocytotic vesicles which fuse with the plasma membrane, releasing neurotransmitters into the synapse (Sudhof, 2004). Neurotransmitters in the synapse are degraded or taken up into the presynaptic axon. Vesicle endocytosis, when vesicles bud off from the plasma membrane into the presynaptic axon, is mediated by proteins including amphysin, endophilin, clathrin, and dynamin (Rizzoli, 2014). Newly synthesized or retrieved neurotransmitters enter these new vesicles and restart the cycle (Takei, 1996).
The Dynamin family of proteins interact with clathrin in the vesicle budding cycle and participate in the scission between the cell membrane and vesicle wall. After the invagination of the lipid membrane, dynamin binds GTP and its alpha helixes wrap around and pinch the budding vesicle. Next, dynamin uses the energy from GTP hydrolysis to sever the membrane tubules and complete vesicle fission (Antonny et al., 2016). This process is critical for the synaptic vesicle cycle (McNiven, 2000). Dynamin proteins have five domains; the GTPase domain is the largest and best understood, followed by a middle domain, pleckstrin homology (PH) domain, coiled coil (CC) domain, and a proline-rich domain (PRD) (McNiven, 2000). The PH domain is thought to interact directly with the lipid bilayer (Vallis, Wigge, Marks, Evans, & McMahon, 1999).
Here, we report 8-year-old identical twin sisters with a de novo p.Lys535Glu likely pathogenic novel variant in exon 15 of DNM1 (NM_004408.3), the first documented pathogenic variant in the PH domain. Although all previously reported patients with DNM1 pathogenic variants exhibited epileptic encephalopathy and severe neurological phenotype, the two probands in this report had no seizures and exhibited mild to moderate DD/intellectual disability (ID) and autism spectrum disorder. Our data suggest an emerging and previously undescribed phenotype associated with a specific DNM1 missense genetic variant in the PH domain.

| CLINICAL REPORT
Patients 1 and 2 are monozygotic triplet sisters who presented for evaluation of DD, autism spectrum disorder, some dysmorphic features, and hypotonia. They were born at 36 weeks gestation, which was conceived naturally, via an uncomplicated cesarian section secondary to triplet pregnancy. There were no complications after delivery. They have a healthy triplet brother. Patient 1 had a birth weight of 2.8 kg (>97th percentile for triplet gestation), a birth length of 49 cm (>97th percentile), and an occipitofrontal circumference (OFC) of 33.8 cm (80th percentile). Patient 2 had a birth weight of 2.55 kg (50th-90th percentile for triplet gestation), a birth length of 48.2 cm (>97th centile), and OFC of 33.1 cm (65th percentile). Their triplet brother had a birth weight of 2.38 kg (50th-90th percentile for triplet gestation), a birth length of 46.9 cm (97th percentile), and OFC of 32 cm (40th percentile).
Developmental concerns were first noted at 4 months of age when both sisters were noted to lack visual tracking, but this spontaneously resolved at 7 months of age. Patient 1 and 2 started receiving physical, occupational, and speech therapies at 1 year of age due to delayed speech and motor milestones. Both sisters started sitting at 7-8 months, crawling at 11 months, and walking at 18 months. At 18 months, Patient 1 could say 5-10 words, but lost most of this vocabulary by age 2. Patient 2 had delayed language acquisition but no language loss. Their triplet brother started sitting at 6 months, crawling at 9 months and walking at 15 months. He required speech therapy for 9 months during third year of life for mild speech delay. The brother is in mainstream classes, never required any resources at school, and does not exhibit neurobehavioral abnormalities. The sisters were diagnosed with autism at 3 years of age and have been receiving applied behavior analysis (ABA) therapy since. At the age of 5, Patient 2 had one episode of febrile seizure. She had a normal brain MRI after this event. She also had a few staring episodes, but her EEG at 7 years of age was normal.
Developmental evaluation of both sisters at age 8 years revealed that Patient 1 can say 15 words but rarely speaks. She can add and subtract double digits, recognize sight words, and use a few signs. She sometimes becomes agitated and may bite or hit herself or others. Patient 2 can use 50 words, combine 2-3 words, use many signs, write her name, recognize some sight words, and complete some basic addition and subtraction. She occasionally becomes agitated, may hit herself or others, and refuses to cooperate with therapy or schoolwork. Patient 2 receives Ritalin and Clonidine for behavioral problems. A formal neuropsychological assessment was performed at 8 years of age for both sisters and the detailed results are provided in Appendix S1.
The genetic workup of the Patient 1 and 2 started at 3 years of age and included normal karyotype, chromosomal microarray analysis, and fragile X testing. Because of muscle weakness and elevated blood lactate, the two sisters were diagnosed with a nonspecific mitochondrial disease. However, mtDNA sequencing and deletion/duplication studies and a mitochondrial nuclear gene panel were normal except for a variant of unknown significance, p.Asp85Asn, in KCNE1. This finding prompted a cardiology consultation and both sisters had a normal echocardiogram and EKG. The exome results (see below) prompted an electroencephalogram which was normal in both sisters.

| MATERIALS AND METHODS
The Institutional Review Board of Washington University School of Medicine approved our study protocol and informed consent was obtained from the parents for publication of genetic results and photographs. Clinical wholeexome sequencing was performed using a trio design with one of the identical twin sisters submitted as a proband and the second sibling and their parents used for segregation analysis. For more details see Appendix S2.

| MOLECULAR RESULTS
Exome sequencing of the sisters identified a de novo, heterozygous, likely pathogenic variant designated as c.1603A>G (p.Lys535Glu) in DNM1. This variant has not been observed in approximately 6,500 individuals that were sequenced as part of the National Heart, Lung, and Blood Institute (NHLBI) exome sequence project or in ExAC database (http://exac.broadinstitute.org/). Both parents tested negative for this variant.
The twin sisters studied here share the de novo variant, c.1603A>G (p.Lys535Glu) in exon 15 of DNM1, classified as likely pathogenic. This variant has not previously been reported in patients or databases of healthy controls. It is a nonconservative amino acid substitution that occurs at an evolutionary conserved position ( Figure S1). Previous in vitro studies presented below further support the variant's pathogenicity.
Eight previously reported patients with DNM1-related disorder carry pathogenic variants in the GTPase domain and two carry variants in the middle domain (Figure 2a). To our knowledge, the monozygotic twin sisters in this  study have the first reported PH likely pathogenic variant. Our report indicates this variant in the PH domain causes less functional impairments and leads to a less severe phenotype than variants in the GTPase or middle domains. Dynamin is a family of proteins with GTPase activity that facilitate vesicle scission (Achiriloaie, Barylko, & Albanesi, 1999). In vitro studies showed that the expression of certain GTPase mutants decreased endocytosis activity by interfering with vesicle scission during synaptic vesicle endocytosis and a middle domain mutation disrupted higher order DNM1 oligomerization (Dhindsa et al., 2015). Experiments also showed that when residue 44 in the GTPase domain of DNM1 was mutated, the protein had reduced binding affinity for GTP and reduced fission activity (Lee, Frank, Marks, & Lemmon, 1999).
The PH domain is primarily responsible for binding PIP2, yet the PH loop has been suggested to have constrictive functioning in the oligomer helix (Achiriloaie, Barylko, & Albanesi, 1999). This structure bends to insert into the plasma membrane and induce curvature before fission. In vitro dominant negative inhibition variants of the PH domain abolish phosphoinositide binding and knock out all synaptic vesicle reuptake (Lee et al., 1999). Variants in the PH domain of dynamin 1 have been previously studied by Vallis et al. (1999) by creating a Lys535Ala mutant. The position Lys535 was chosen because it comes into proximity to the 5'phosphate of the inositol ring in inositol-1,4, 5-trisphosphate (IP3). The point mutation described in that article is at the same amino acid position as both patients in our study. While Patients 1 and 2 have a Lys535Glu variant, the mutated residue in that study is a Lys535Ala substitution. Each of these amino acid substitutions are significant because they do not share the same structure or chemical properties as lysine. Multiple PH domain mutants were compared, and while others did not show changed levels of transferrin uptake, a well-established endocytosis assay, Lys535Ala showed inhibition of endocytosis and impairment of PIP2 binding (Vallis et al., 1999). These findings support the critical role of Lys535 in the coordination of phosphoinositide binding, an essential step in endocytosis (Vallis et al., 1999).
A similar study by Achiriloaie et al. (1999) compared PH domain variants K535M and K561M to a GTPase domain variant N272 (missing most of the GTPase domain) and wild-type DNM1. The GTPase domain variant and K535M both inhibited endocytosis, while the K561 mutant and wild type DNM1 showed no significant difference in endocytosis (Achiriloaie, Barylko, & Albanesi, 1999). This implies that not all amino acid substitutions in the PH domain are significant and reinforces the claim that Lys535 is essential to the PH loop binding to PIP2 and promoting endocytosis. PH domain mutations at other sites may not have a clinically discernable phenotype. We also examined whether the position of the pathogenic variants affects different isoforms of Dynamin 1, but found all variants involve all isoforms (Figure 2b). F I G U R E 2 (a) Exons 1-23 of DNM1, based on transcript DNM1-009 (ENST00000341179.11). The genetic variant described in this paper and previously reported pathogenic variants are shown in the diagram. The 5 important Dynamin 1 domains are depicted. (b) Diagram of the four common DNM1 isoforms (RefSeq: NM_004408.3, UniProt: Q05193; dark blue rectangles). Numbers refer to amino acids. The main protein isoform is #1 and has 864 amino acids. Light blue rectangles indicate different amino acids compared to isoform 1. All DNM1 variants are shown in relation to the different isoforms In summary, Dynamin 1 is mainly expressed in the central nervous system, explaining the neurological phenotype in DNM1-related disorder. Previously reported patients carry pathogenic variants in the GTPase or middle domains and presented with epileptic encephalopathy and severe neurodevelopmental symptoms. Here, we report the first documented DNM1 variant in the PH domain associated with a milder phenotype without epilepsy. Our data suggest a novel, possibly unique, and less severe phenotype resulting from a variant in the PH domain of dynamin 1. Reports of additional patients are needed to better define the genotype-phenotype correlation.