The clinical and genetic heterogeneity of paroxysmal dyskinesias

The contributions of different genes to inherited paroxysmal movement disorders are incompletely understood. Gardiner et al. identify mutations in 47% of 145 individuals with paroxysmal dyskinesias, with PRRT2 mutations in 35%, SLC2A1 in 10% and PNKD in 2%. New mutations expand the associated phenotypes and implicate overlapping mechanisms.


Introduction
Paroxysmal dyskinesia was first reported in 1892 by Shuzo Kure in a 23-year-old Japanese man, who had frequent movement-induced paroxysmal attacks from the age of 10 years. At that time the diagnosis was referred to as atypical Thomsen's disease (Kure, 1892). Later, Gowers (1901) described a similar child, but he considered this movement disorder an epileptic phenomenon, and in 1940, Mount and Reback (1940) described a 23-year-old with involuntary writhing and posturing of the trunk and extremities and labelled this condition paroxysmal dystonic choreoathetosis. Kertesz (1967) and Weber (1967) described families with this condition termed paroxysmal kinesigenic choreo-athetosis and familial paroxysmal dystonia, and Demirkiran and Jankovic (1995) amalgamated the many terms used, suggesting three subtypes, comprising paroxysmal kinesigenic (PKD or PKC), non-kinesigenic (PNKD), and exercise-induced dyskinesia (PED) (Bruno et al., 2004(Bruno et al., , 2007Bhatia, 2011). A fourth type, paroxysmal hypnogenic dyskinesia (PHD), characterized by attacks of dyskinesia during sleep, was previously included, but has since been recognized as autosomal dominant nocturnal frontal lobe epilepsy (Sohn and Lee, 2011).
The most common of the paroxysmal movement disorders is PKD, in which attacks are precipitated by voluntary movements such as standing from a sitting position, or the transition from walking to running. Onset is usually in childhood, and attacks are often controlled by carbamazepine (Bhatia, 2001(Bhatia, , 2011Erro et al., 2014). PKD is frequently preceded by infantile convulsions, often with choreoathetosis. The gene responsible for PKD proved elusive for many years, but was recently identified as PRRT2, which encodes a small proline-rich transmembrane protein (Chen et al., 2011;Wang et al., 2011;Cloarec et al., 2012;de Vries et al., 2012;Gardiner et al., 2012;Guerrini and Mink, 2012;Hedera et al., 2012;Heron et al., 2012;Liu et al., 2012;Scheffer et al., 2012). The function of the protein is unknown, but it has been shown to interact with the synaptic protein SNAP25 (Lee et al., 2012). Mutations in the PRRT2 gene account for a large proportion of PKD and several groups have reported mutations in this gene (Chen et al., 2011;Wang et al., 2011;Cao et al., 2012;de Vries et al., 2012;Friedman et al., 2012;Gardiner et al., 2012;Heron et al., 2012;Lee et al., 2012;Liu et al., 2012;Ono et al., 2012;Ishii et al., 2013;Specchio et al., 2013).
Attacks of PNKD are usually triggered by alcohol, coffee or strong emotion. They last longer than attacks of PKD, often from 10 min to 1 h, but can last as long as 12 h. However, they are much more infrequent and occur only a few times a year (Mount and Reback, 1940;Bhatia, 1999;Lombroso and Fischman, 1999;Vercueil, 2000;Lee et al., 2004;Engelen and Tijssen, 2005;Friedman et al., 2009;Ghezzi et al., 2009;van Rootselaar et al., 2009;Benz et al., 2012;Pons et al., 2012). The gene responsible for PNKD was identified as the MR-1 gene in 2004, but it is now referred to as PNKD (Raskind et al., 1998;Lee et al., 2004;Rainier et al., 2004). To date three mutations in this gene have been reported; p.A7V, p.A9V and p.A33P, the first two of which have been found in multiple unrelated patients (Lee et al., 2004;Friedman et al., 2009;Ghezzi et al., 2009;Shen et al., 2011;Pons et al., 2012;Erro et al., 2014). Recent work from Shen et al., (2015) has shown that PNKD interacts with the synaptic active zone proteins RAB-interacting molecule (RIM)1 and RIM2, and modulates neurotransmitter release. The mutant protein is less effective at inhibiting exocytosis. Lance (1977) described a family with exercise-induced dystonia with attacks lasting between 5 and 30 min, once or twice per month. This disorder is now termed PED (Lance, 1977). PED is thought to be the rarest of the three paroxysmal movement disorders, where attacks are induced by physical exertion after long periods of exercise. The condition can be associated with migraine, hemiplegia, ataxia and epilepsy (Zorzi et al., 2003;Bhatia, 2011). Mutations in the SLC2A1 gene, which encodes the glucose transporter type 1 protein, have recently been found to be responsible for causing PED, often called GLUT1 deficiency syndrome 2 (Wang et al., 2000;Vermeer et al., 2007;Suls et al., 2008). SLC2A1 mutations also cause GLUT1 deficiency syndrome 1, a phenotypically variable syndrome that often includes ataxia, microcephaly, intellectual dysfunction, dystonia, epilepsy and low fasting glucose levels detected on CSF analysis (Wang et al., 2000;Vermeer et al., 2007;Suls et al., 2008;Schneider et al., 2009;Fung et al., 2011;Gokben et al., 2011;Hashimoto et al., 2011;Bawazir et al., 2012;Agostinelli et al., 2013;Muhle et al., 2013;Weller et al., 2015).
The majority of published reports on paroxysmal movement disorders are single families, small series or single gene studies with little known about the gene mechanisms.
Here, we carry out the first large screening study of the three main paroxysmal dyskinesia genes [the total coding regions of SLC2A1 and PRRT2 and exons one and two (the only exons in which mutations have been previously identified) of PNKD] in a large referral series of 145 paroxysmal movement disorders and in a further 53 genetically undefined patients with episodic ataxia or familial hemiplegic migraine. We identify the mutation frequency and spectrum as well as genetic and phenotypic heterogeneity, describe novel mutations, and investigate the mutation mechanisms amongst the paroxysmal dyskinesias.

Materials and methods
Patients and unaffected family members were recruited through the laboratory with consent and ethical approval (NHNN studies 06/N076 and 07/Q0512/26); they were seen either at the National Hospital in Queen Square, or referred from other centres for genetic testing with local approval. Patients were diagnosed with a paroxysmal dyskinesia or movement disorder based on recognized criteria (Bruno et al., 2004(Bruno et al., , 2007Kinali et al., 2004;Bhatia, 2011) by the authors. Acquired causes were excluded using clinical investigation prior to genetic testing. Episodic ataxia and familial hemiplegic migraine cases were negative for mutations in the KCNA1 and CACNA1A genes by direct sequencing of all codons. DNA was extracted from blood of affected patients and unaffected family members using standard diagnostic laboratory methods.

Sequencing
Polymerase chain reaction (PCR) was used to amplify the three coding exons and flanking introns of the PRRT2 gene, the 10 coding exons and flanking introns of the SLC2A1 gene, and the first two coding exons and flanking introns of the PNKD gene (Supplementary Table 1). For each gene the longest transcript was used for primer design and sequencing: PRRT2-001: ENST00000358758; SLC2A1-001: ENST00000426263; PNKD-001: ENST00000273077. PCR amplification was performed using 10 pmol of both forward and reverse genomic primers (synthesized by Sigma-Aldrich) and FastStart TM Taq DNA polymerase (Roche). Each purified product was then sequenced using forward or reverse primers, as well as internal sequencing primers to ensure complete coverage of in the case of exon 2 of PRRT2 with Applied Biosystems BigDye Õ terminator v3.3 sequencing chemistry as per the manufacturer's instructions. The resulting reactions were resolved on an ABI3730XL genetic analyser (Applied Biosystems) and analysed with SeqScape v2.5 software (Gene codes).
In developing our genetic analysis strategy for diagnostics we also developed a custom Illumina sequencing gene panel (Illumina Inc.). This panel included the PRRT2, SLC2A1 and PNKD genes. These genes had a mean coverage of 269 Â, 196Â and 178 Â, respectively and 24 samples were analysed in this way. All regions of the genes were covered and no coverage gaps had to be completed by Sanger sequencing. The analysis of data consisted of mapping the raw data to the hg19 human reference assembly using Novoalign software, and PCR duplicates were removed using the Picard software. Indels were called using the GATK package and variants annotated using SAMtools. Mutations were verified in both directions. Mutation position was labelled from the transcriptional start site of the genes, according to the standard nomenclature.

Expression methods
Regional distribution of PRRT2, SLC2A1, PNKD, KCN1A, SNAP25 and CACNA1A mRNA expression in the normal human brain was determined using microarray analysis of human post-mortem brain tissue from the UK Human Brain Expression Consortium (Trabzuni et al., 2011). Brain tissues originating from 134 control Caucasian individuals were collected by the Medical Research Council (MRC) Sudden Death Brain and Tissue Bank (Edinburgh, UK). The following brain regions were included in the analysis: cerebellum, frontal cortex, hippocampus, medulla, occipital cortex, putamen, substantia nigra, temporal cortex, thalamus and white matter. Total RNA was isolated from these tissues using mRNeasy 96-well kit (Qiagen) before processing with the Ambion Õ WT Expression Kit and Affymetrix GeneChip Whole Transcript Sense Target Labeling Assay, and hybridization to the Affymetrix Exon 1.0 ST Array. The probe set defining each gene mRNA was determined using the Affymetrix Netaffx annotation file (HuEx-1_0-st-v2 Probe set Annotations, Release 31). The combined signal of the gene probe sets were used to determine mRNA expression.
Sequencing of PNKD and PRRT2 cDNA from affected patient fibroblast mRNA was carried out to assess the presence of nonsense-mediated decay and to indicate the presence of a truncated protein in mutations that affect the last exon of the gene. Fibroblasts were first taken with informed consent and mRNA was extracted using a Qiagen miRNA kit. cDNA was synthesized from the mRNA with SuperScript Õ II reverse transcriptase according to the manufacturer's protocol, 1000 ng of mRNA was used as template with random oligonucleotide primers. The PNKD C-terminal and the PRRT2 (across the whole gene) of the resulting cDNA product was then amplified by 35 cycles of PCR and sequenced by the above method, using primers designed to amplify only cDNA and not genomic DNA.

Results
Mutations in the PRRT2 gene were found in 53 families or sporadic cases, with nine different mutation types (Figs 1-5 Figure 2 Family tree and mutation chromatograms. Filled symbols indicate family members that are affected, unfilled symbols are unaffected. The proband is indicated with a black arrow. + /À denotes an individual that is heterozygous for the mutation shown, À/À does not carry the mutation. (C) genes. In each case mutations that have been previously reported to cause a paroxysmal movement disorder are shown above the gene, and mutations found in this paper are shown below (blue have previously been reported, red are novel).
and Table 1). A male to female ratio of 2:1.3 was identified in those expressing a phenotype, and the patient demographic was 56% British and a mixture of other populations accounting for the other 44%. As widely reported, by far the most common mutation (44 families, 82%) was an insertion of a cytosine into a string of nine cytosines, resulting in a frame shift mutation and premature stop codon (p.R217Pfs*8). Each of the other nine mutations accounted for one family and the majority were loss-of-function. These mutations were found in families with a number of different ethnicities and there was no common background haplotype. Four mutations were novel and two of the mutations (p.G305W and p.C332_V333insD) have only been reported by us in the past. We include them here, as well as the cases with p.R217Pfs*8 mutations, for the assessment of the frequency of PRRT2 mutations in our cohort (Gardiner et al., 2012;Silveira-Moriyama et al., 2013). The p.P215R variant is also included in the mutation table; it has a frequency of 57:10 000 in the ExAC database and not seen in 488 UK control subjects. The pathogenicity of this change is still uncertain. The p.P216H variant has been found in our patient series but was also found in the UK control population at a rate of 1%. Mutations in the PRRT2 gene were mainly associated with paroxysmal kinesigenic dyskinesia with a number of associated phenotypes (Table 1) including: (i) episodic ataxia; (ii) benign epilepsy; (iii) PED; and (iv) migraine and familial hemiplegic migraine. Fifty-one patients were part of the paroxysmal dyskinesia series and the remaining two were from the episodic ataxia and familial hemiplegic migraine series.
Migraine and hemiplegic migraine were by far the most common associated phenotypes (Table 1). Interestingly, the majority of patients were given symptomatic treatment, mainly with carbamazepine; it has been widely reported that patients with PRRT2-positive PKD are more likely to respond well to the drug than patients without a mutation (Li et al., 2013;Mao et al., 2014). There did not appear to be a correlation between genotype and efficacy of treatment in our cohort. Initially the extended Indian families were taking phenytoin, which was then usually switched to carbamazepine, and lamotrigine in one patient. Depending on availability some of the extended Indian family patients still take phenytoin. Patient 48, who did appear to benefit from even high doses of the drugs. No treatment was being given in three families, at patients' request. A family with episodic ataxia and one with familial hemiplegic migraine alone were identified with PRRT2 mutations. The familial hemiplegic migraine family proband presented as an infant with infrequent seizures until age 2 years and then developed   typical hemiplegic migraine attacks. The sister, father and two cousins also had classical hemiplegic migraine and the attacks in the proband persisted until now (aged 18 years) but responded to carbamazepine. Fourteen SLC2A1 mutations were identified in the paroxysmal dyskinesia series (10%) and one in the episodic ataxia and familial hemiplegic migraine series (Figs 1, 2 and Table 2). In general these were complex cases that had been heavily investigated prior to obtaining a genetic diagnosis. Eight had PED, often associated with other features such as epilepsy and migraine. Three had PKD (one with epilepsy) and one had PNKD, two with episodic ataxia and one with myotonia and dystonia, as discussed below. Eleven of the mutations had previously been reported as being pathogenic. The p.C201R mutation has not before been identified but presented with a PNKD phenotype and was present in the affected mother. p.C201R is not present in population databases, but is not well conserved and predicted to be benign by PolyPhen-2 but damaging by SIFT. p.T60M is present in 0.00015% of the population, is moderately conserved and is predicted to be damaging by PolyPhen-2 but tolerated by SIFT. This mutation has been reported in association with seizures in the past but like in our family, there was reduced penetrance. Patient 66 had sequence variants in both SLC2A1 and PRRT2 (p.R333Q and p.P216H, respectively) and a PKD phenotype, but the PRRT2 mutation is unlikely to be pathogenic as it is present in 1% of controls we analysed, and the p.R333Q mutation has been reported previously as pathogenic.
The majority of PRRT2 mutations are predicted to be loss-of-function and likely lead to haploinsufficiency. It has been previously demonstrated that mutations p.Q163X, p.G192WfsX8 and p.R217PfsX8 result in nonsense-mediated decay . This is not the case for all mutations as cDNA created from two of our mutations; p.G305W and p.*341Lext27 (a stop codon mutation extending the protein, HGVS standard nomenclature used; den Dunnen and Antonarakis, 2000) do not affect mRNA or lead to a longer transcript (Figs 3 and 4A). The mechanism behind these two mutations is likely to be the same as those causing nonsense-mediated decay with lack of association in the SNAP25/SNARE complex and greater vesicle release (Fig. 5). SLC2A1 mutations were associated with a wide spectrum of clinical features. Family 56 was identified with a novel heterozygous mutation at p.G76V that was not present in 488 controls and 6502 exomes in the exome variant server. This patient was a 26-year-old, diagnosed with attention-deficit hyperactivity disorder as a child and since then has had episodes of 'wobbly' eyes, legs and arms, and abnormal arm posturing that last 5-10 min, several times per day. Triggers for these episodes included tiredness, sudden movement, intercurrent infection or illness and excitement. He experienced episodes of weakness and painful cramps in his hands and his legs. He has tried carbamazepine, which helped a little, and acetazolamide may have helped reduce the frequency of these attacks.
He underwent repeat long exercise testing (McManis) and this showed significant decrement, accompanied by weakness of the exercised hand muscles. This was most unexpected given that SLC2A1 is best known as a brain transporter; however, there is some evidence of the protein having an additional important role in skeletal muscle (Andrisse et al., 2014). This result was repeated and abnormal spanning over several years. The significant decrement on McManis testing ranged from 51-66%. The clinical diagnosis at that time suggested a periodic paralysis phenotype but the movement disorder was not consistent with this.
In the PNKD gene, four mutations were identified (Figs 1-3 and Table 3). Three were in the paroxysmal dyskinesia series and one in a familial hemiplegic migraine family. The mutations associated with paroxysmal dyskinesias were in phenotypically typical PNKD families with non-kinesigenic precipitants such as stress or strong coffee. These mutations have been reported in the past and these were in two unrelated families with p.A7V and one with p.A9V. In the familial hemiplegic migraine family the mutation was novel and the female proband presented at 42 years of age with a typical attack of hemiplegic migraine with headache, abnormal vision and left-sided motor and sensory weakness that lasted for 45 min to an hour in duration. She had a normal MRI shortly after the event and other cardiac investigations were unremarkable, and the hemiplegic migraine resolved. A few months later she had a similar hemiplegic migraine attack. Her paternal great uncle and father had similar attacks. Her father presented at a similar age and to date has had over 50 hemiplegic migraine attacks, often without a headache. He has presented to the emergency department many times concerned that this was a stroke and has been extensively worked up but imaging and other investigations have been normal. A heterozygous mutation of c.1022delC; p.P341fs*2 was identified in the PNKD gene in the proband and father, not in the mother. We analysed cDNA, from mRNA extracted from patient fibroblasts. The deletion was present in the mRNA, indicating that nonsense-mediated decay would not occur, although nonsense-mediated decay is dependent on cell type and therefore it is possible that it could occur in neurons. This mutation therefore caused the formation of a truncated PNKD in the mRNA (Figs 3 and 4B). Although functional work was not carried out the truncating effect of this mutation is likely to have an abnormal effect on exocytosis due to impaired interaction between PNKD and RIM/ RIM1 (Fig. 5).

Discussion
High prevalences of PRRT2, SLC2A1 and PNKD mutations were identified in this large, mainly London based paroxysmal movement disorder referral series. Although we have a multi-ethnic population the results corroborate smaller individual gene series ( Fig. 1 and Tables 1-3). There was a spectrum of clinical features and many patients had additional clinical features such as seizures. The frequency of migraine and hemiplegic migraine was highly associated with these phenotypes although this is also common in the general population. Some individuals in the extended PKD families did not have a movement disorder at all or were affected by seizures or hemiplegic migraine alone. The usual mechanism for PRRT2 mutations is loss of function due to nonsense-mediated decay, leading to haploinsufficiency (Figs 2-5) and likely lead to a lack of SNAP25/SNARE interaction and increased vesicle release. Segregating PRRT2 missense mutations were also identified where there was no change in the PRRT2 mRNA, but we expect a loss of SNAP25/SNARE interaction or prevention of the PRRT2 protein from anchoring to the presynaptic membrane, and thus leading to a similar lack of inhibition of vesicle release due to reduced tethering (Fig. 5).
Fewer mutations were identified in the SLC2A1 and PNKD genes, and primarily in patients with PKD and PNKD ( Fig. 1 and Table 2). The patients with SLC2A1 mutations had the broadest spectrum of clinical phenotypes. There was overlap clinically with PKD (as in the p.R223W family) and PNKD (as with the p.C210R family). This group were the most extensively investigated before a genetic diagnosis was sought, and fasting CSF glucose was frequently low in affected individuals with a more complex phenotype associated with seizures but usually normal with a movement disorder alone. There was also a greater rate of an incorrect clinical diagnosis and overlap with other channelopathies, as with the family with the p.G76V mutation and abnormal McManis testing, and in the family with the p.R333Q mutation and unusual tongue dystonia as part of the phenotype. These families are similar to those first described in 1892 as atypical Thomsen's disease (Kure, 1892). The p.R333Q had an additional variant in the PRRT2 gene (p.P216H), which may be benign or modifying the effect of the p.R333Q mutation. In addition there was evidence of reduced penetrance in SLC2A1, most clearly in the family with the p.T60M mutation that presented with paroxysmal attacks, headaches and nystagmus where the father and brother had the mutation but were unaffected (see family tree, Fig. 2). The p.T60M mutation has previously been identified in idiopathic epilepsy, further extending the heterogeneity.
In the episodic ataxia cohort, one family was identified with a mutation in the PRRT2 gene, one with a defect in the SLC2A1 gene and two familial hemiplegic migraine families were identified, one with a PRRT2 mutation and one with a novel PNKD mutation. The familial hemiplegic migraine families were of most interest as they have a typical phenotype and the mutations segregate in the family. The novel PNKD mutation is a frameshift deletion located in exon 10, which is predicted to cause a truncated protein, this segregated with the disease, predicted pathogenic and was not identified in controls (Figs 2, 3 and 4B). Alternate splicing of the PNKD gene results in three isoforms of the protein of varying length; PNKD-S, PNKD-M (both expressed ubiquitously), and PNKD-L (expressed in the CNS) (Shen et al., 2011). All previously reported mutations are located in the 5' end of the gene, found in both PNKD-L and PNKD-S but not PNKD-M. This mutation, instead affects PNKD-L and PNKD-M and the location and truncating effect of the change in shortening the PNKD protein is likely to lead to reduced RIM/RIM1 binding (Shen et al., 2015) in the SNARE complex and abnormal vesicle release (Fig. 5).
While there is a great deal more to be understood, it seems likely that these three paroxysmal genes are acting on the presynaptic terminal, possibly with overlapping pathways, and thus result in a similar dysregulated and possibly increased vesicular release. Although there is clinical overlap, there are also additional clinical features. This overlap is seen in the brain expression patterns where genes with a similar mechanism have identical regional expression patterns (Supplementary Fig. 1) as for PRRT2, SNAP25, KCNA1 and CACNA1A (all presynaptic) where they share highest expression levels in the cerebellum, and frontal, temporal and occipital cortices as compared with SLC2A1 and PNKD. This could explain the subtle phenotypic differences and the regional effect on vesicle release. It has recently been reported that overexpression of wild-type PNKD in rat hippocampal cultures reduced neurotransmitter release in comparison to an empty vector, whereas overexpression of mutant PNKD did not. This suggested that PNKD also has a role in regulating presynaptic exocytosis . It is also known that PRRT2 interacts with SNAP25, a protein important in facilitating synaptic exocytosis . Therefore, we suggest a possible disease mechanism whereby both PNKD and PRRT2 perform similar roles in restricting synaptic exocytosis. Disease-causing mutations that either reduce levels of PRRT2 or disrupt PNKD function reduce this restriction and result in excessive neurotransmitter release (Fig. 5). It is unclear how SLC2A1 mutations contribute to this theory, but it has been shown that they result in reduced glucose transport into the brain, so perhaps glucose is also involved in the regulation of exocytosis. The functional consequence of the regional expression patterns remains to be seen but may indicate that SLC2A1 and PNKD pathways are more closely related to dystonic genes located in the basal ganglia and brainstem.
Little is known about how disruption of these proteins results in migraine, a clinical manifestation that has been seen frequently here and elsewhere. However, in a recent study, transgenic mice with human monogenic migraine gene mutations (thus mimicking the types of migraine seen in this cohort) were shown to display increased glutamatergic neurotransmission and cerebral hyperexcitability (Ferrari et al., 2015). This finding indicates that the lack of neurotransmitter release regulation postulated here could also result in the migraine exhibited. There is clearly a large pathophysiological overlap between all of these related neurological disorders, which required further investigation to be understood more fully.
Overall this work reveals a wide spectrum of mutations and phenotypes and has expanded the broad phenotypic spectrum of these paroxysmal movement disorders, suggesting where possible, as part of the investigative work-up, all three genes should be analysed in these conditions. We also highlight novel mutations and a likely distinct mechanism for 3' PNKD mutations that lead to PNKD-L dysregulation. There is genetic and phenotypic overlap amongst other episodic movement disorders with episodic ataxia, the neuronal channelopathies and familial hemiplegic migraine all being identified with defects in these three genes.