Clinical spectrum of STX1B-related epileptic disorders

Objective The aim of this study was to expand the spectrum of epilepsy syndromes related to STX1B, encoding the presynaptic protein syntaxin-1B, and establish genotype-phenotype correlations by identifying further disease-related variants. Methods We used next-generation sequencing in the framework of research projects and diagnostic testing. Clinical data and EEGs were reviewed, including already published cases. To estimate the pathogenicity of the variants, we used established and newly developed in silico prediction tools. Results We describe 17 new variants in STX1B, which are distributed across the whole gene. We discerned 4 different phenotypic groups across the newly identified and previously published patients (49 patients in 23 families): (1) 6 sporadic patients or families (31 affected individuals) with febrile and afebrile seizures with a benign course, generally good drug response, normal development, and without permanent neurologic deficits; (2) 2 patients with genetic generalized epilepsy without febrile seizures and cognitive deficits; (3) 13 patients or families with intractable seizures, developmental regression after seizure onset and additional neuropsychiatric symptoms; (4) 2 patients with focal epilepsy. More often, we found loss-of-function mutations in benign syndromes, whereas missense variants in the SNARE motif of syntaxin-1B were associated with more severe phenotypes. Conclusion These data expand the genetic and phenotypic spectrum of STX1B-related epilepsies to a diverse range of epilepsies that span the International League Against Epilepsy classification. Variants in STX1B are protean and contribute to many different epilepsy phenotypes, similar to SCN1A, the most important gene associated with fever-associated epilepsies.

Genetic generalized epilepsies (GGEs) and genetic epilepsies with febrile seizures (FS) plus (GEFS+) are genetically and phenotypically heterogeneous epileptic disorders. GGEs share common clinical hallmarks such as seizure types, generalized epileptic discharges on EEG, and typical onset between childhood and adolescence. Most GGE cases have a polygenic inheritance, but a few monogenic causes have been identified, such as GABRA1, 1 SLC2A1, 2 CACNA1H, 3 as well as microdeletions, such as 15q13.3. 4 GEFS+ is a familial epilepsy syndrome characterized by focal or generalized febrile and afebrile seizures, and focal or generalized epileptiform discharges on EEG. The clinical presentation may differ considerably among affected individuals within the same family. 5,6 Private mutations have been identified in genes encoding subunits of voltagegated Na + channels (SCN1A, SCN1B) 7,8 and the γ-aminobutyric acid type A receptor (GABRG2, GABRD). 9,10 Most of the genes that are associated with GGE and GEFS+ are also implicated in developmental and epileptic encephalopathies (DEEs), which are characterized by seizure onset in the first years of life, often pharmacoresistant seizures, cognitive regression, and neurologic deficits. For certain genes, such as SCN1A, a clear genotype-phenotype correlation is described: in GEFS+, a greater proportion of missense variants is found, whereas in Dravet syndrome, nonsense mutations or large deletions are more common. 11 Recently, we reported variants in the STX1B gene as a cause of fever-associated epilepsies of variable severity. 12 Two large families showed a rather benign course of GEFS+ syndrome, whereas 4 individuals had a more severe DEE phenotype. An additional 18-year-old patient exhibiting epilepsy with myoclonic-atonic seizures and moderate to severe intellectual disability was reported recently. 13 STX1B encodes syntaxin-1B, a presynaptic protein that is part of the SNARE complex mediating the process of calciumdependent synaptic vesicle release. 14 We aimed to describe additional variants in STX1B, characterize the related clinical and EEG phenotypic spectrum, and establish genotype-phenotype correlations.

Clinical data
We identified 49 individuals harboring heterozygous variants in STX1B (NM_052874.3). The cases were pooled from 14 different centers identified through research studies and clinical diagnostic testing. The cohort consists of 7 families with 2 to 17 affected individuals and 16 sporadic patients. Of these, we previously reported 2 large families and 4 sporadic patients. 12,15,16 We systematically collected clinical information such as age at onset, seizure type, neurologic and cognitive deficits, neuroimaging outcome, and antiepileptic drug (AED) treatment through a standardized questionnaire. We reviewed original EEG recordings (S. Wolking., H.L., T.D., C.M., P.W.-W., D.M.) for all but 3 patients (F3, F15, F18). EEGs for F1 and F2 were described in previous publications. 15,16 Standard protocol approvals, registrations, and patient consents Written informed consent to participate in this study was obtained from all patients or caregivers.

Genetics
We used different next-generation sequencing datasets from multicenter research projects or diagnostic testing services to identify the reported STX1B variants. The methods and the centers are available from Dryad (doi:10.5061/dryad.cf0hj73) and in our previous report. 12 All patients are of European descent.
To evaluate the effect of the missense variants, we used wellestablished in silico prediction tools: SIFT (sorting tolerant from intolerant), 17 PolyPhen-2 (polymorphism phenotyping v2), 18 and MutationTaster. 19 We compared patient STX1B missense variants to variants identified in almost 150,000 control individuals stored in the Genome Aggregation Database (gnomAD, gnomad.broadinstitute.org/). 20 Mutation density was calculated by counting the number of variant positions in gnomAD divided by the window length within a window of 10 nucleotides using a sliding window approach over the coding sequence of the STX1B transcript NM_ 052874.3. We determined paralog conservation as described by Lal et al. 21 In short, we aligned the canonical protein sequences of the 7 paralog genes of the Ensembl SYNTAXIN protein family, STX2, STX1A, STX11, STX1B, STX19, STX3, and STX4 and scored the conservation at each alignment position using JalView. 22 Then we determined the mean and the standard deviation paralog conservation for each single protein of the protein family and a z score (para_zscore) calculated for each residue position. We defined paralog conservation as para_zscore >0.

Data availability
Data not published in this article will be shared as anonymized data by request from any qualified investigator. Glossary DEE = developmental and epileptic encephalopathy; FS = febrile seizures; GEFS+ = genetic epilepsies with febrile seizures plus; GGE = genetic generalized epilepsy; gnomAD = Genome Aggregation Database; GTCS = generalized tonic-clonic seizure; JME = juvenile myoclonic epilepsy; PolyPhen-2 = polymorphism phenotyping v2; SIFT = sorting tolerant from intolerant; SNARE = SNAP (soluble NSF attachment protein) receptor. FS were present in the majority of patients. The first seizure started between 10 months and 5 years (median 20 months). There was a large variety of afebrile seizure types, including generalized tonic-clonic seizures (GTCS) (n = 13), focal impaired awareness seizures (n = 3), atonic seizures (n = 8), tonic seizures (n = 4), and absence seizures (n = 5). Nine patients had infrequent seizures until adulthood that did not require treatment. For most cases, cognition was intact and neurologic deficits were not frequent in this cohort (table 1). Asymptomatic variant carriers were identified in 2 families (4  in F1, 3 in F2), which is of importance for genetic testing and counseling in mild cases. Moreover, in F1, 4 phenocopies were identified (2 patients with a GGE phenotype not matching GEFS+, 1 individual with a single FS, and 1 patient with falls and head nodding in early childhood that were later interpreted as nonepileptic 15 ).The treatment outcome in this group was overall positive. More than 50% of patients were seizure-free without medication (17); the remainder were seizure-free with AED treatment (11). Only patient F7 had ongoing seizures with valproic acid. EEGs showed focal and generalized epileptiform discharges. Available ictal recordings in F8.1 showed focal onset seizure. We retrieved detailed phenotypes for 12 of 15 patients. Epilepsy onset was between 0 months and 3.5 years (median 15 months). Six patients (F5, F14, F16, F17, F20) had FS with onset between 13 months and 2 years. F14 had only a single FS after vaccination. Afebrile seizures comprised GTCS (n = 9), myoclonic (n = 10), atonic (n = 8), tonic seizures (n = 3), atypical absences (n = 3), infantile spasms (n = 1), and hyperkinetic focal seizures (n = 1).
Except for F6, initial development was normal in all patients, with severe global developmental delay or even regression coinciding with seizure onset. In F19 and F21, global developmental delay and seizures were present since birth. Neurologic examination showed anomalies in most patients with ataxia being the most frequent finding.
Seizures were pharmacoresistant in all patients except F18, undergoing on average 9 AED trials. Only F5 and F21 achieved seizure freedom with ongoing treatment. Of note, F4 responded with a significant seizure reduction to the combined therapy of lamotrigine and valproic acid after several unsuccessful AED trials. After the administration of bromide, F14 showed a significant reduction in seizure frequency.
EEGs were available for review in 11 of 15 patients. Interictally, the EEGs in all cases showed multifocal epileptic discharges that were predominantly located in the frontal or temporal region. In addition, we detected generalized discharges in 10 of 11 patients presenting as generalized spikewave or polyspike-wave discharges.  Figure 1A shows the new variants combined with those previously reported in a schematic view of the syntaxin-1B protein. In comparison to the location of loss-of-function variants, missense variants were more frequent in the second part of the gene containing (1) the SNARE motif conveying interaction with the other 2 components of the SNARE complex and (2) the transmembrane domain, which is responsible for anchoring syntaxin-1B into the cell membrane. None of these variants was reported so far in the gnomAD database, currently the largest collection of exomes and genomes publicly available. 20 In general, STX1B is intolerant to missense variants (missense z score >3.63) and haploinsufficiency (pLI score of 0.94) in the Exome Aggregation Consortium browser. 20 Evaluation by different in silico prediction tools (SIFT, 17 PolyPhen-2, 18 MutationTaster 19 ) showed a damaging effect for 6 of 8 variants in 3 of 3 prediction tools (table 2). We calculated the mutational density of missense variants in the general population from the Exome Aggregation Consortium using a sliding window approach and could show that the variants are located in regions of different mutational density ( figure 2B). We also calculated a paralog conservation score distribution for STX1B (para_zscore) 21 and highlighted the patient variants ( figure  2C). This score gives the degree of conservation among paralog genes in the human Syntaxin gene family. Disease variants are enriched in paralog-conserved residues. 21 In our paralog analysis, the SNARE motif turned out to be the most paralogconserved and gnomAD variant-depleted region of the protein with the highest para_zscores ( figure 3, A and B). Furthermore, in a direct comparison of patient and gnomAD missense variant distribution across the linear protein sequence, we observed a significant enrichment of patient variants in the SNARE motif (p = 0.036). We ascertained missense variants in an unbiased manner so that SNARE motif enrichment by chance is improbable. The amino acid positions of most of the detected missense variants showed high or intermediate para_zscores.
There was no significant difference between the para_zscores from missense variant positions in patients compared to those in gnomAD ( figure 3C). We also detected a few diseaseassociated variants in regions with a high mutational density in gnomAD and low para_zscores (p.Arg261Gln, p.Ile282Thr), which were located in the transmembrane domain or the part linking the SNARE motif with the transmembrane domain of the protein.

Segregation of variants in families and family histories
The segregation status for all variants in the respective pedigrees, together with a classification of the phenotypes, is shown in figure 1B. Previously published pedigrees can be found in Schubert et al. 12 We validated inheritance in 6 of the total 23 variants in families with more than one affected (4 new and 2 published). They were mostly cosegregating with their phenotypes but also included both asymptomatic variant carriers and phenocopies (F1, 2,8,13,16,20). In 3 patients, a positive family history of epilepsy was available, but additional family members were not available for genetic testing (F4, 11, 23). The available phenotypes of the other affected individuals are depicted in figure 1B.

Discussion
The current study expands the number of reported patients with STX1B-related epileptic syndromes and describes 17 new variants. Four different phenotypes can be discerned: (1) a benign epilepsy syndrome with febrile and afebrile seizures corresponding to GEFS+, (2) a GGE phenotype, (3) a DEE syndrome with refractory seizures and moderate to severe developmental deficits, and (4) a focal epilepsy phenotype. STX1B-related epilepsies thus show a remarkable phenotypic heterogeneity that is in its extent reminiscent of other epilepsyrelated genes, such as SCN1A, 11 SCN2A, 25 KCNQ2, 26 and STXBP1. 27 In comparison to STXBP1, which is closely related to STX1B (since the respective proteins interact with each other in the synaptic transmitter release machinery) and exhibits in most cases an early onset of epilepsy presenting as Ohtahara syndrome or West syndrome, 28 STX1B-related DEE shows a later onset of epilepsy after the first birthday. Exceptions to this rule are F19 (p.A246P) and F21 (p.Q128Gfs*2) who presented with a seizure onset within the first days of life and are more reminiscent of STXBP1-related syndromes. It is of interest that STX1B-and STXBP1-related DEEs both feature ataxia and other movement disorders. 28 However, in the vast majority of STXBP1-related epilepsy syndromes, moderate to Abbreviations: PolyPhen-2 = polymorphism phenotyping v2; SIFT = sorting tolerant from intolerant. Reported missense variants with in silico prediction scores. The table shows the genomic location of the variants in ascending order, the amino acid change, and the predicted effect of these variants to the protein function by using SIFT, PolyPhen-2, and MutationTaster as well as the paralog conservation score. The Genome Aggregation Database was used as a source for minor allele frequencies.
severe developmental delay is present 27-29 -milder phenotypes, as in our cohort, are not observed. In comparison to other common DEE disorders, the median time of seizure onset differentiates: neonatal or early infantile for SCN2A and KCNQ2 (a later onset with these genes is rarer and associated with different functional consequences-gain-vs loss-offunction mutations 25,26,30 ) and later infantile for SCN1A and STX1B. Also in other aspects, parallels can be drawn with SCN1A-related disorders. Thus, in 5 STX1B-related DEE cases, FS and the presence of myoclonic seizures were reminiscent of Dravet syndrome. 31 However, the presence of tonic seizures in these patients would be unusual for Dravet syndrome. A patient with generalized tonic-clonic, atonic, and myoclonic seizures with onset at 1 year and severe developmental delay was reported before this study with a 1.2 Mb de novo 16p11.2 deletion, 13 including the STX1B gene. The phenotype was classified as epilepsy with myoclonic-atonic seizures, a syndrome that has been associated with SCN1A in several patients. 32 The previously reported patient showed a remarkable resemblance to F6 and F20, who also harbor a comparable deletion. Another similarity with SCN1A is that our cohort features several patients and families with GEFS+ syndrome 7 and one patient with mesial temporal lobe epilepsy and FS (for which common polymorphisms in SCN1A have been described as a genetic risk factor 33 ), and at least one GEFS+ multiplex family with mesial temporal lobe epilepsy has been reported. 34 Of interest is our patient with episodes reminiscent of hemiplegic migraine; SCN1A mutations are a recognized cause of familial hemiplegic migraine. 35 We acknowledge that the association with a migraine-like syndrome in a single patient could also be coincidental. In contrast, classic GGE syndromes, such as represented by the 2 patients with JME in our cohort, are not typically described in patients with SCN1A mutations. Of note, both patients with JME had an atypical response of pharmacoresistant seizures. In the DEE group, 7 of 13 (F3, F5, F6, F14, F17, F18, F19) variants were de novo. This supports the likelihood of a causative role in the context of the observed phenotypes. Three missense variants showed a high paralog conservation (p.C144F, p.G226R, p.A246P), increasing the probability of a functional role. An exception to this rule was p.I282T (F14), a variant that was discovered using a gene panel approach. Possibly, despite the de novo status of this variant, another as yet unknown genetic alteration in a gene not captured by the gene panel might be more relevant in this case. Although variant p.V216E (F4) showed a slightly negative paralog score, this variant lies in a highly conserved region. The patient had a positive family history and previous functional data, i.e., rescue experiments in STX1B knockdown zebra fish larvae 12 support the pathogenicity of this variant. In F13 (p.N189Afs*5), both affected siblings had a similar phenotype. This suggests an inheritance from the mother, who has not been tested, while the father was tested negative. Since the mother and other maternal family members were reportedly not affected, germline mosaicism could be possible. However, other genetic causes cannot be entirely excluded because the variant was found in a gene panel of 85 epilepsy genes. In F15 (p.S258N), the prediction tools suggest a benign variant. Since no family members were tested and the paralog score was rather low, the role of this variant has to be considered carefully. However, the variant is located at the end of the SNARE motif, which supports a causative role. In F16 (p.L221P), the family history suggests an inherited defect. The variant was identified by exome sequencing and no other convincing variants were found. The unaffected mother also carries the variant but other affected family members were not available for testing. Prediction tools indicate a deleterious variant and the variant is located in the middle of the SNARE motif, although the paralog score was rather low; a case of incomplete penetrance is therefore possible or the variant could represent a predisposing factor for epilepsy. In F20, the gene deletion was inherited from the father, who had a milder phenotype, suggesting the presence of other genetic modifiers. In F21 (p.Q128Gfs*2), the parents were not tested and the variant was identified in a gene panel of 115 candidate genes, so that other genetic causes cannot be excluded. In the GEFS+ group, in 3 of 6 cases (F1, F2, F8), a positive family history with cosegregating genetic findings support the causative role of the reported variants. F7 (p.S10Afs7) was confirmed to be de novo. For F9 (p.R245*) and F10 (p.Y140*), the parents were not tested. Since all variants in this group predict a complete loss of function, we consider them as disease-causing and the mechanism for the GEFS+ group seems to be clearly a haploinsufficiency. In the GGE group, neither case had parents who were tested. F11 (p.E210K) has a positive family history that is suggestive of an inherited disease. The causative role of this variant is underscored by the in silico findings that the variant is predicted deleterious, lies in a region of very low mutational density, and shows high paralog conservation. p.K93* (F12) leads to a haploinsufficiency, in line with the variants causing GEFS+. Furthermore, this variant was identified by whole-genome sequencing and no other convincing variants were found. Of interest, in a recent genome-wide association study of common epilepsies, a significant signal in the region of STX1B was detected in the JME cohort. 36 In the FE group, the role of STX1B should be considered more carefully. In F22 (p.R261Q) and F23 (p.V88F), parents were not tested. For p.R261Q, the established prediction tools point toward a deleterious variant, while the position shows medium mutational density and low paralog conservation. However, whole-genome sequencing showed no other likely pathogenic variants. For p.V88F, the used prediction tools show a more coherent picture of a deleterious variant, the paralog score was high, and whole-exome sequencing found no other convincing variants.
Regarding genotype-phenotype correlation, it is notable that missense variants in the functionally relevant SNARE motif were identified in 6 of 7 patients or families (86%) associated with DEE. The seventh patient (p.E210K) had refractory JME. In contrast, all but one of the GEFS+ patients or families had truncating variants (5 of 6, 83%) predicting nonsensemediated decay and complete loss of function of one allele indicating haploinsufficiency as the disease mechanism.
A possible explanation could be that missense variants in crucial protein regions, such as the SNARE motif, may lead to a deficient gene product that interferes with protein-protein interactions and normal presynaptic vesicle fusion (dominant-negative effect), whereas missense variants in the first part of the gene may affect assembly of syntaxin-1B proteins and therefore have similar consequences as a truncation with nonsense-mediated decay. The loss-of-function mechanisms could also be compensatedat least in part-by syntaxin-1A and, thus, be less damaging. 37 We demonstrate that STX1B variants are linked to 4 different epilepsy phenotypes. De novo dominant mutagenesis is the etiology of most DEEs, whereas the more common epilepsies are likely complex genetic disorders. This work suggests a role for ultrarare STX1B variants in the pathogenesis of a broad range of both common and rare epilepsies. There are plausible explanations for genotype-phenotype correlations, such that haploinsufficiency causes rather benign phenotypes, whereas missense variants in functionally critical regions cause more severe phenotypes. Other, unknown factors, which may include genetic modifier, individual genetic background, and epigenetic effects, could also have an important role and explain phenotypic heterogeneity, reduced penetrance, and variable phenotypes among carriers of truncating mutations. Functional studies on a protein and cellular neuronal level will shed more light on the relevance and degree of severity of the described variants and contribute to a better understanding of the dysfunction of syntaxin-1B and synaptic transmission in epilepsy and strengthen our current interpretation of genotype-phenotype correlations.
manuscript for important intellectual content, analysis and interpretation. G. Lesca: phenotyping and acquisition of data, critical revision of the manuscript for important intellectual content. S.M. Sisodiya: phenotyping and acquisition of data, critical revision of the manuscript for important intellectual content, analysis and interpretation. Y.G. Weber: phenotyping and acquisition of data, critical revision of the manuscript for important intellectual content. D. Lal: analysis and interpretation, writing the manuscript. C. Marini: phenotyping and acquisition of data, critical revision of the manuscript for important intellectual content, analysis and interpretation. H. Lerche: study concept and design, analysis and interpretation, writing of the manuscript, study supervision. J. Schubert: study concept and design, analysis and interpretation, writing of the manuscript, study supervision