Refining Genotypes and Phenotypes in KCNA2-Related Neurological Disorders

Pathogenic variants in KCNA2, encoding for the voltage-gated potassium channel Kv1.2, have been identified as the cause for an evolving spectrum of neurological disorders. Affected individuals show early-onset developmental and epileptic encephalopathy, intellectual disability, and movement disorders resulting from cerebellar dysfunction. In addition, individuals with a milder course of epilepsy, complicated hereditary spastic paraplegia, and episodic ataxia have been reported. By analyzing phenotypic, functional, and genetic data from published reports and novel cases, we refine and further delineate phenotypic as well as functional subgroups of KCNA2-associated disorders. Carriers of variants, leading to complex and mixed channel dysfunction that are associated with a gain- and loss-of-potassium conductance, more often show early developmental abnormalities and an earlier onset of epilepsy compared to individuals with variants resulting in loss- or gain-of-function. We describe seven additional individuals harboring three known and the novel KCNA2 variants p.(Pro407Ala) and p.(Tyr417Cys). The location of variants reported here highlights the importance of the proline(405)–valine(406)–proline(407) (PVP) motif in transmembrane domain S6 as a mutational hotspot. A novel case of self-limited infantile seizures suggests a continuous clinical spectrum of KCNA2-related disorders. Our study provides further insights into the clinical spectrum, genotype–phenotype correlation, variability, and predicted functional impact of KCNA2 variants.


Introduction
Potassium channels are involved in many vital processes such as neuronal excitability, neurotransmitter release, and cellular osmoregulation [1,2]. KCNA2, encoding for the voltage-gated potassium channel subfamily A member 2 (K v 1.2, OMIM * 176262) is a member of the shaker-like delayed rectifier potassium channel family [3]. It is predominantly expressed in axons and presynaptic terminals of the central nervous system [4,5]. Homo-or heterotetramers formed by α-subunits containing variable proportions of K vsubfamily members contribute to functional channel characteristics [6]. K v 1.2 consists of six transmembrane segments (S1-S6) forming the voltage-sensing domain (S4) and the ion-conducting pore (S5-S6) with its selectivity filter localized in the pore loop [7,8]. The Cterminal and N-terminal cytosolic domains are less conserved and regulate tetramerization and the association to modifying subunits [9,10]. Several antibody-mediated neurological diseases in children and adults have been linked to proteins associated with K v 1 channel components, such as leucine-rich glioma-inactivated 1, contactin-associated protein 2, and particularly K v 1 channel subunits [11,12].
In 2015, pathogenic de novo variants in KCNA2 were first described in six individuals with early-onset severe epilepsy, cognitive impairment, and ataxia [13]. Two phenotypically distinct epilepsy syndromes were delineated correlating with gain-of-function (GOF) and loss-of-function (LOF) changes of the net potassium current of K v 1.2 [13,14]. The recurrent KCNA2 variant c.881G > A, leading to an amino-acid exchange at position 294 within the voltage sensing domain, was identified in individuals with hereditary spastic paraplegia (HSP) as an additional phenotype [15]. In 2017, novel and previously reported individuals were summarized into three distinct functional and clinical subgroups of K v 1.2 dysfunction: GOF, LOF, and a mixed dysfunctional type (GOF/LOF) [16].
In this study, we present additional individuals with known and novel variants in KCNA2. We reevaluate phenotypic, functional, and genetic data from all published cases to refine and further delineate the assumed phenotypic and functional subgroups. We analyze the genetic distribution of all known pathogenic and likely pathogenic variants in KCNA2 and predict the functional impact of all possible missense variations in KCNA2.

Results
We identified 115 individuals with pathogenic and likely pathogenic (P/LP) KCNA2 variants from 28 publications (78 cases) and ClinVar database entries (37 cases). Additionally, we report clinical data from seven novel cases, six individuals with P and LP KCNA2 variants, and one individual with a VUS. After the exclusion of all individuals without sufficient clinical information, we included 76 of 115 cases into further analyses (see Section 4).
In 40 individuals, variants occurred de novo, while 17 segregated with the disease in the family. In 21 cases, information on inheritance was unavailable. Thirty different alterations of KCNA2 are on record to date, with 26 single-nucleotide variants (SNV), one in-frame deletion, and three different deletions (Tables 1 and 2). GOF: gain-of-function effects, LOF: loss-of-function effects compared to wild-type channels; all variants are described according to the transcript NM_004974.3.

Novel Cases with KCNA2 Variants in this Study
Data from seven previously unpublished cases with known pathogenic (n = 5) and novel variants (n = 2) in KCNA2 were ascertained. Genetic and phenotypic data are summarized in Table 3. The localization and structural consequence of the novel missense variant p.(Tyr417Cys) is depicted in Figure 1. (E) and Pro407 (F) disrupts its symmetry and structure, likely impairing interactions with S4-S5 during gating-related conformational changes. All three substitutions are located in the S6 helix and in proximity to the S4-S5 linkers, which is suggestive for the probable affection of associated interactions leading to altered channel gating.

Clinical Characteristics of Individuals with Reported and Novel KCNA2 Variants
Mean age at inclusion and publication was 17 years (median age 10 years, range 0-81 years). No individual had died until the last reported study visit. Nearly all (91%) reported cases developed seizures. Mean age at seizure onset was 13 months (median age 8 months, range 0-156 months). In total, 41 male and 33 female individuals were included (three cases with missing information). No relevant differences in gender distribution on clinical manifestation or genotype were observed. Seven individuals were reported as not having seizures at all, of whom five harbor the variant p.(Arg294His) associated with hereditary spastic paraplegia and ataxia (HSP) [15,20] and two harbor the homozygous variant p.(Arg65*) associated with familial intellectual disability [41]. Associated movement and motor disorders were reported in 61/64 cases (95%), with ataxia in 64% of reported cases, tremor in 23%, and increased muscle tone (spasticity), in 16%. Febrile seizures were described in 28% (21/76) of cases. 30% (21/69) of the patients achieved seizure control ) as well as the amino acid substitutions Leu405 (E) and Ala407 (F) of the corresponding proline residues. The PVP motif interacts with the S4-S5 linker, and both prolines of this symmetric motif (D) are important for channel gating. The substitution of Pro405 (E) and Pro407 (F) disrupts its symmetry and structure, likely impairing interactions with S4-S5 during gating-related conformational changes. All three substitutions are located in the S6 helix and in proximity to the S4-S5 linkers, which is suggestive for the probable affection of associated interactions leading to altered channel gating.

Clinical Characteristics of Individuals with Reported and Novel KCNA2 Variants
Mean age at inclusion and publication was 17 years (median age 10 years, range 0-81 years). No individual had died until the last reported study visit. Nearly all (91%) reported cases developed seizures. Mean age at seizure onset was 13 months (median age 8 months, range 0-156 months). In total, 41 male and 33 female individuals were included (three cases with missing information). No relevant differences in gender distribution on clinical manifestation or genotype were observed. Seven individuals were reported as not having seizures at all, of whom five harbor the variant p.(Arg294His) associated with hereditary spastic paraplegia and ataxia (HSP) [15,20] and two harbor the homozygous variant p.(Arg65*) associated with familial intellectual disability [41]. Associated movement and motor disorders were reported in 61/64 cases (95%), with ataxia in 64% of reported cases, tremor in 23%, and increased muscle tone (spasticity), in 16%. Febrile seizures were described in 28% (21/76) of cases. 30% (21/69) of the patients achieved seizure control (missing data in 11 cases). Of the cases with available data, most individuals (82%, 59/72) show intellectual disability of various severity. Thirteen individuals were reported with a normal cognitive development, including all five cases harboring the p.(Met255_Ile257del) deletion [3]. Cranial magnetic resonance imaging (available in 58 cases) was abnormal in 23 cases (40%), with cerebellar atrophy being most commonly reported (n = 18).
Compared to individuals with a mixed dysfunctional type (LOF/GOF), individuals with LOF variants showed better development before seizure onset (p = 0.012), later onset of epilepsy (p = 0.033), more favorable epilepsy outcome (seizure-free) (p = 0.014), and less severe intellectual disability (p = 0.029). Compared to individuals with LOF/GOF variants, epilepsy onset is significantly later in GOF individuals (p = 0.008), whereas development before seizures, epilepsy outcome, and intellectual disability did not differ significantly. Comparing individuals with LOF and GOF variants, development before seizures, epilepsy onset, seizure outcome, and intellectual disability was not significantly different. Figure 2 illustrates the age at onset of epilepsy in the different subgroups.

Prediction of Functional Relevance of KCNA2 Variants
To evaluate the likelihood of pathogenic variant effects, we combined the distribution of known pathogenic and benign variants with computational pathogenicity scores for all biologically possible missense variants of KCNA2. The REVEL meta-score combines

Prediction of Functional Relevance of KCNA2 Variants
To evaluate the likelihood of pathogenic variant effects, we combined the distribution of known pathogenic and benign variants with computational pathogenicity scores for all biologically possible missense variants of KCNA2. The REVEL meta-score combines variant annotations of 13 individual pathogenicity scores and showed high overall performance in the discrimination of pathogenic and benign variants in large clinical datasets [42,43]. High REVEL scores overlap with areas of pathogenic variant enrichment in KCNA2 (Figure 3). Several peaks, corresponding to the voltage sensing domain (S4), the pore-loop (S5-S6) and the ion conducting pore with the gating mechanism (S5 and S6), were identified and are supported by similar tendencies in the evaluation of further established pathogenicity scores, such as CADD, M-CAP, PolyPhen, and SIFT ( Figure S1). The density of putatively benign variants from the gnomAD database (green) shows enrichment at the less conserved N-and C-termini. In one specific region within S5, putatively benign variants are enriched, mainly being the result of tolerated changes at position Ala329.

Discussion
Recently, different neurological disorders, including epilepsy syndromes, ataxia, and hereditary spastic paraplegia were associated to Kv1.2 channel dysfunction. In this study, we present additional patients with known and novel KCNA2 variants. By analyzing phenotypic, functional, and genetic data from all published reports and novel cases, we refine

Discussion
Recently, different neurological disorders, including epilepsy syndromes, ataxia, and hereditary spastic paraplegia were associated to K v 1.2 channel dysfunction. In this study, we present additional patients with known and novel KCNA2 variants. By analyzing phenotypic, functional, and genetic data from all published reports and novel cases, we refine and further delineate phenotypic and functional subgroups. Overall, the most prevalent manifestation was early-onset developmental and epileptic encephalopathy in combination with inconstant movement and motor disorders, mainly ataxia. Similar to other disorders related to voltage-gated ion channel genes (e.g., KCNA1, KCNQ2), a continuous spectrum from mild to severe neurologic manifestations and paroxysmal movement disorders can be delineated [44,45]. By comparing phenotypic characteristics from patients with functional data on mutational effects, carriers of variants leading to complex and mixed channel dysfunction that are associated with a gain and loss of potassium conductance more often show early developmental abnormalities and an earlier onset of epilepsy compared to patients with pure loss-of-function or pure gain-of-function variants, respectively.

Novel Cases of KCNA2 Variants and Clinical Presentation
We here report seven children with three known pathogenic variants and two novel variants in KCNA2. The variant p.(Tyr417Cys) was identified in a three-year-old boy (#1) with benign epilepsy of infancy (BFIS) and normal development. No other variants in known genes for BFIS were identified. The variant was inherited from an unaffected father. It is classified as a variant of unknown significance (VUS) according to the American College of Medical Genetics and Genomics (ACMG) (PM2, PP2, PP3) [46]. This variant is located at the cytosolic border of S6 in a region enriched for pathogenic variation and is predicted as damaging ( Figure 3D, Table S1). The amino acid Tyr417 is spared from variation in the gnomAD database. Previous functional investigations of the homologous amino acid exchange in drosophila p.(Tyr485Cys) have shown that cysteine replacement of tyrosine at this position leads to a slight reduction of single channel conductance (γ), an increased deactivation rate, and a small decrease in the time constant for "ON gating currents" [47,48]. The clinical phenotype with self-limiting infantile-onset seizures in this boy is partially overlapping with the mild end of KCNA2-related epilepsy. Therefore, a role of the specific KCNA2 variant p.Tyr417Cys in benign infantile epilepsy (BFIS) appears possible, with a reduced penetrance, similar to other known genetic causes of infantile seizures such as PRRT2 [49], SCN2A [50], and SCN8A [51]. The novel variant p.(Pro407Ala) was found in a three-year-old girl (#2) with neonatal-onset developmental and epileptic encephalopathy and profound intellectual disability (ID). The variant occurred de novo and has not been described before. It is absent from controls in the gnomAD database and predicted as damaging. According to ACMG criteria, this variant is likely pathogenic (PM6, PM2, PP2, PP3) [46]. The variant leads to an exchange from proline to alanine at position 407 within a region that is enriched for pathogenic variants. It disrupts the highly conserved proline(405)-valine(406)-proline(407) (PVP) motif, which is essential for channel gating. Variants affecting prolines in the PVP motif and in closely related residues have been identified in KCNA2and recently in KCNA1-associated early-onset epileptic encephalopathy [16,23,36,39,52]. Previous in vitro data of p.(Pro407Ala) in the Shaker K vchannel suggested a non-conducting mutant channel as the underlying dysfunction [53]. This novel variant underlines the importance of the PVP motif regarding channel function.
The variant p.(Pro405Leu) affects the first proline of the PVP motif. We report two additional children with infantile-onset epilepsy and prolonged febrile seizures, harboring this recurrent variant. One of the two children had moderate intellectual disability, while the other one was described as normally developed. The mean age at seizure onset in all reported individuals with p.(Pro405Leu) was 9 months, and the main seizure types included focal and generalized seizures as well as febrile seizures. Continuous spikes and waves during sleep (CSWS) were reported in 8/13 individuals [13,16,23,34,37,38]. While both our cases with p. (Pro405Leu) show sleep-activated spikes, CSWS was incomplete or unilateral, reflecting a milder epilepsy syndrome and suggesting interindividual variability.
In twins, one boy and one girl (#6 and #7), we identified the recurrent variant c.881G>A; p.(Arg294His), which was described previously in families with hereditary spastic paraplegia [15,20]. Similar to family 3 from Helbig et al., the presenting symptom in both children was epilepsy with different focal and generalized seizure patterns, including absences and status epilepticus. While previous reports showed spastic paraplegia (6/7 individuals) and ataxia (4/7 individuals) as prominent manifestations, we here present two additional cases with seizures that included tonic-clonic and absence seizures in both children [15]. Interestingly, ataxia was present in the boy only, and ataxia was markedly worsened when valproate was temporarily applied, resembling absence epilepsy with ataxia (AEA) due to loss-of-function mutations in CACNA1A [54]. Similar to previous AEA families, intrafamilial variability was present, as the girl showed isolated seizures with age-appropriate cognitive development, the boy had a severe complex developmental disorder, and the father did not show neurologic symptoms. The further family history revealed a sister of the unaffected father who had intellectual disability and an unspecified movement disorder. The mother of the father (grandmother of the reported twins) had a progressive gait disorder from advanced age resembling spastic paraplegia. Unfortunately, genetic testing was not available.

Functional Classification and Phenotypic Spectrum of Published KCNA2 Variants
Previously, three distinct subgroups of KCNA2 variants with a genotype-phenotype correlation based on the electrophysiological channel properties were suggested [16]. We here analyze all available clinical reports and accordingly are able to group cases into these functional subgroups (LOF, GOF, and LOF/GOF). We show significant differences between subgroups regarding development before seizure onset, age at seizure onset, seizure outcome (seizure freedom), and intellectual disability. The combined dysfunctional effects of LOF/GOF variants are leading to more severe phenotypes compared to simple LOF or GOF variants. However, isolated LOF or GOF effects on K v 1.2 both lead to epileptic encephalopathy as well. Recently, Shore et al. (2020) provided data, showing how GOF variants can lead to epilepsy [55]. By expressing known epilepsy-causing KCNT1-gainof-function variants in excitatory and inhibitory neurons, they were able to demonstrate specific mutational effects affecting inhibitory neurons only, especially interneurons with fast-spiking activity, promoting network hyperexcitability and hypersynchronicity. Celltype-specific functional changes from distinct variants might also explain the more severe epilepsy course in children carrying variants with mixed LOF and GOF effects. It can be hypothesized that specific effects dominate in different neuronal subpopulations, leading to an impaired GABA-ergic response from interneurons on one side and to impaired glutamatergic transmission in excitatory neurons on the other side, affecting different inhibitory and excitatory networks. Interestingly, the homologous shaker mutation of the LOF variant p.(Arg294His) was shown to lead to proton currents in addition to the pure loss-of-K v 1.2 function when expressed in Xenopus laevis oocytes [15,56]. These effects are difficult to assess by most in vitro models, but in vivo neuron-subtype-specific dysfunctional effects with a loss of net potassium currents and additional a gain-of function from leaky proton currents might explain that the children described here exhibit focal and generalized absence seizures, which were previously attributed to gain-of-function effects in K v 1.2.
Apart from the electrophysiological classification of variants, which allows assigning some phenotypic features, recurrent variants in each functional subgroup show distinct variant-specific characteristics. Individuals with the variant p.(Pro405Leu) are more likely to have febrile seizures and CSWS than other individuals in the LOF subgroup. Individuals with the variant p.(Arg294His) developed spastic paraplegia and ataxia as prominent manifestations. Seizures were described only in two previously reported cases [15] and in the two individuals reported here. Five individuals with the variant p.(Met255_Ile257del) presented with episodic ataxia, normal intellectual abilities and self-limited epilepsy, forming another distinct phenotype. In the GOF subgroup, apart from two individuals, all carried the recurrent variant p. (Arg297Gln). Therefore, characteristics of this subgroup are largely based on this variant. In the LOF/GOF subgroup, all but one individual with the variant p.(Thr374Ala) (n = 7) had neonatal refractory epilepsy. All showed severe intellectual disability, and in four of seven individuals, spastic cerebral palsy was reported (see Table S2: Summary of clinical data).

Prediction of Functional Relevance and Modelling of KCNA2 Variants
By combining the distribution of known pathogenic and benign variants with prediction scores for all possible missense variants, we further define protein regions with enrichment of pathogenic variants and highly predicted mutational effects, such as the voltage-sensing domain, the pore-loop, and the pore-forming domains. Both novel variants described here are affecting residues within these regions. Homology modeling of p.(Pro407Ala) and p.(Pro417Tyr) suggests effects on protein stability and protein structure of K v 1.2 ( Figure 1). Importantly, we here complement the phenotypes associated with variants in both prolines of the PVP motif that similar to KCNA1 are linked to severe early onset developmental and epileptic encephalopathy [52].

Database Research
Reports and entries of KCNA2 variants containing clinical and/or genetic data were identified using PubMed, Human Gene Mutation Database (HGMD), and ClinVar ( Figure 4). PubMed literature research was performed using the keywords "KCNA2" and "voltagegated potassium channel K v 1.2". ClinVar database entries were selected for variants classified as "pathogenic" or "likely pathogenic". Reported variants were checked for duplicate mentions. Data acquisition was finished 30 April 2020. In addition, we collected clinical and genetic data from seven previously unpublished cases from European epilepsy centers with putatively causative variants in KCNA2. After the exclusion of one individual with a paternally inherited variant of uncertain significance (VUS) and all individuals without sufficient clinical information (37 ClinVar, 8 HGMD/Pubmed), 76 individuals were included into further statistical analyses ( Figure 4).
Individuals were classified according to functional consequences of the mutated K v 1.2 channel when functional data from in vitro analysis were available. Quantitative variables are illustrated by sample size, mean, and standard deviation. Statistical comparisons between means of two cohorts were performed with Welch's two-sample t-test. Frequencies of descriptive variables are depicted with number, sample size, and percentages. Statistical comparisons of the frequency of descriptive variables between cohorts were conducted using the two-tailed Fisher's exact test. iants classified as "pathogenic" or "likely pathogenic". Reported variants were checked for duplicate mentions. Data acquisition was finished 30 April 2020. In addition, we collected clinical and genetic data from seven previously unpublished cases from European epilepsy centers with putatively causative variants in KCNA2. After the exclusion of one individual with a paternally inherited variant of uncertain significance (VUS) and all individuals without sufficient clinical information (37 ClinVar, 8 HGMD/Pubmed), 76 individuals were included into further statistical analyses (Figure 4). Individuals were classified according to functional consequences of the mutated Kv1.2 channel when functional data from in vitro analysis were available. Quantitative variables are illustrated by sample size, mean, and standard deviation. Statistical comparisons between means of two cohorts were performed with Welch's two-sample t-test. Frequencies of descriptive variables are depicted with number, sample size, and percentages.

Computational Pathogenicity Analysis and 3D Protein Structure Modeling
To visualize differences in variant distribution between affected individuals and unaffected controls, we compiled all currently known pathogenic and benign KCNA2 variants and their respective allele count using the databases PubMed, HGMD, ClinVar, and gno-mAD. gnomAD variants referring to the canonical Ensembl transcript ENST00000485317.1 were selected by the categories "Missense" and "pLOF" (excluding frameshift variants). Variants were standardized to the NM_004974.3 transcript of the GRCh37/hg19 human reference genome using the Mutalyzer Nomenclature Checker web tool [57]. Additionally, all biologically possible missense variants of the KCNA2 coding sequence were computed and translated to the corresponding genomic position. Subsequently, established variant effect prediction (VEP) scores (REVEL, CADD, MetaLR, MetaSVM, M-CAP, PolyPhen-2, SIFT) were annotated using the VEP web tool of Ensembl  [58][59][60][61][62][63][64][65]. Resulting VEP scores of the canonical Ensembl feature and variant distributions were plotted along the primary structure of the K v 1.2 protein using the geom_smooth and geom_density function, respectively, of the ggplot2 library in RStudio (v. 1.2.5042; RStudio, Inc., Boston, MA, USA) [66]. The REVEL meta-score demonstrated high overall performance regarding discrimination between pathogenic and benign single nucleotide variants in large clinical variant datasets [43]. Therefore, the REVEL score is primarily used in this study to assess the pathogenicity of variants. A complementary synopsis of all annotated VEP score results is depicted in Supplementary Figure S1.

Limitations
Due to limited and heterogeneous quality of data in publications, phenotypic features might not have been completely accessible. Therefore, the number of patients with available information is specified with each clinical feature. Due to differing age at imaging, the detectability of cerebellar changes might vary between reports. Despite the reasonable ability of the software programs to predict pathogenicity, they still cannot discriminate among functional LOF, GOF, and LOF/GOF-type variants. This is a critical limitation of computed predictions of variant effects in clinical practice and for genetic counseling. Furthermore, some recurrent variants represent a significant proportion of the cohort and variant-specific effects may dominate the phenotypic presentation of the whole functional subgroup.

Conclusions
Our study provides further insights into the clinical spectrum, genotype-phenotype correlations, variability, and predicted functional impact of KCNA2 variants. The location of variants reported here highlights the importance of the PVP motif in transmembrane domain S6 as a mutational hotspot. A novel case of self-limited infantile seizures suggests a continuous clinical spectrum of KCNA2-related disorders. The prediction of functional relevance of variants allowed us to identify hotspots of functional impact and recurrent mutations.   Institutional Review Board Statement: The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the ethics committee of the University of Heidelberg (S-318/2018). All individuals or their legal guardians gave their informed consent for data sharing and genetic testing.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.