Impaired neuronal KCC2 function by biallelic SLC12A5 mutations in migrating focal seizures and severe developmental delay

Epilepsy of infancy with migrating focal seizures (EIMFS) is one of the early-onset epileptic syndromes characterized by migrating polymorphous focal seizures. Whole exome sequencing (WES) in ten sporadic and one familial case of EIMFS revealed compound heterozygous SLC12A5 (encoding the neuronal K+-Cl− co-transporter KCC2) mutations in two families: c.279 + 1G > C causing skipping of exon 3 in the transcript (p.E50_Q93del) and c.572 C >T (p.A191V) in individuals 1 and 2, and c.967T > C (p.S323P) and c.1243 A > G (p.M415V) in individual 3. Another patient (individual 4) with migrating multifocal seizures and compound heterozygous mutations [c.953G > C (p.W318S) and c.2242_2244del (p.S748del)] was identified by searching WES data from 526 patients and SLC12A5-targeted resequencing data from 141 patients with infantile epilepsy. Gramicidin-perforated patch-clamp analysis demonstrated strongly suppressed Cl− extrusion function of E50_Q93del and M415V mutants, with mildly impaired function of A191V and S323P mutants. Cell surface expression levels of these KCC2 mutants were similar to wildtype KCC2. Heterologous expression of two KCC2 mutants, mimicking the patient status, produced a significantly greater intracellular Cl− level than with wildtype KCC2, but less than without KCC2. These data clearly demonstrated that partially disrupted neuronal Cl− extrusion, mediated by two types of differentially impaired KCC2 mutant in an individual, causes EIMFS.

GABA A and glycine receptors that are responsible for the Cl − influx 8 . The presence of alternative first exons with different promoters provides two isoforms of KCC2a and KCC2b (see Fig. 1B). Mice deficient for both KCC2 isoforms die at birth due to severe motor defects, and KCC2b-specific knockout mice survive for up to 2 weeks, but die due to spontaneous seizures [9][10][11] , suggesting indispensable roles for KCC2 in proper mammalian brain function.
Recently, heterozygous missense mutations in SLC12A5 were shown to be associated with febrile seizures and idiopathic generalized epilepsy in humans 12,13 , and very recently, autosomal recessive SLC12A5 mutations were reported to cause EIMFS 14 . However, in the former two reports, the mutations were identified based only on the targeted DNA sequencing of SLC12A5, and possible causative mutations in other genes were not clearly excluded. In the more recent study 14 , all nonsynonymous mutations in the patients were systematically listed by whole exome sequencing (WES) analysis, and the SLC12A5 mutations were selected as the most plausible causes based on several criteria. Nevertheless, the Cl − extrusion function of KCC2 was not properly assessed in that study, as discussed in detail below. Therefore, the data did not allow for an estimation of intraneuronal Cl − levels in the patients.
In this study, we identified novel compound heterozygous SLC12A5 mutations in three families, including four affected individuals. Functional analysis using the gramicidin-perforated patch-clamp technique confirmed significant, but not complete, loss of KCC2 function in the patients. Individual mutations in each patient were found to impair KCC2 function to different degrees. Thus, our data demonstrated that partial loss of neuronal KCC2 function by biallelic mutations might cause migrating focal seizures, which are characteristic of EIMFS.

Identification of biallelic SLC12A5 mutations in individuals with EIMFS.
To identify the genetic basis of autosomal recessive EIMFS, WES was performed in two Japanese siblings with EIMFS (individuals 1 and 2, Fig. 1A). A total of 309 and 272 rare protein-altering and splicing-affecting variants were identified per individual, in which 122 variants were common in two (Supplementary Table S1). We focused on genes with two heterozygous variants (possible compound heterozygous variants) or homozygous variants that were consistent with an autosomal-recessive trait, and found that SLC12A5 was a solo candidate. Sanger sequencing validated the c.279 + 1G > C and c.572C > T (p.A191V) variants in two siblings, which were transmitted from their mother and father, respectively (Fig. 1A). The unaffected older brother had only the c.279 + 1G > C variant. We then searched the WES data of 10 sporadic cases with EIMFS for SLC12A5 mutations, and found another Malaysian patient (individual 3) with compound heterozygous SLC12A5 mutations: c.967T > C (p.S323P) and c.1243A > G (p.M415V) (Fig. 1A).
To investigate the possible involvement of SLC12A5 mutations in other types of infantile epilepsy, we also searched the WES data of 526 patients for biallelic SLC12A5 mutations, and examined an additional 141 patients by SLC12A5-targeted resequencing as a second cohort. Following SLC12A5 resequencing, in which the mean depth of SLC12A5 coding sequences was 244 (range 41 to 465), we identified a Japanese patient with compound heterozygous SLC12A5 mutations [c.953G > C (p.W318S) and c.2242_2244del (p.S748del)], who was diagnosed as unclassified intractable epilepsy (individual 4, Fig. 1A). Other biallelic mutations were unidentified in the WES data of 526 epileptic patients. These six mutations were absent in dbSNP 138, our in-house 575 control exomes, the Exome Variant Server, and EXaC database (Supplementary Table S2). Four missense mutations and an in-frame amino acid deletion occurred at evolutionarily conserved amino acids (Fig. 1B). At least two of three web-based prediction tools (SIFT, Polyphen-2, and MutationTaster) predicted that the four missense mutations could affect protein function (Supplementary Table S2).
To examine the mutational effect of c.279 + 1G > C, reverse transcriptase-PCR was performed using total RNA from lymphoblastoid cell lines (LCLs) derived from individuals 1 and 2. Results demonstrated that the c.279 + 1G > C mutation caused a deletion of exon 3 from the SLC12A5 mRNA, resulting in an in-frame 44-amino acid deletion (p.E50_Q93del) (Fig. 1C,D). All six mutations were located on both KCC2a and KCC2b (Fig. 1B,E), and affected the N− and C− terminal regulatory domains (p.E50_Q93del and p.S748del, respectively) 15 , transmembrane domains (p.A191V and p.M415V), and the large extracellular loop (p.W318S and p.S323P) adjacent to four conserved cysteines (C287, C302, C322, C331), which is required for KCC2 activity 16 . Differentially impaired Cl − extrusion function of two KCC2 mutants in individual epileptic patients. To assess the mutational effects of KCC2 on Cl − extrusion function, the HEK293 cells stably expressing the α1 type glycine receptor (GlyR) 17 were transfected with the mutants or wildtype (WT) KCC2. Then we compared reversal potentials of GlyR-mediated Cl − currents, which reflect the equilibrium potentials for Cl − (E Cl ), i.e. [Cl − ] i controlled by KCC2, in the transfected cells using the gramicidin-perforated patch-clamp technique. We used a voltage ramp from −80 to −10 mV and determined E Cl as the voltage level at which the GlyR current became zero, corresponding to the level at the intersection of superimposed current traces obtained before and during application of 100-300 μM glycine ( Fig. 2A and Supplementary Fig. S1; inward and outward currents beyond the intersection indicate efflux and influx of Cl − through GlyRs, respectively. See Materials and Methods and the legend of Supplementary Fig. S1 for details). Thus, a greater negative E Cl indicated greater extrusion of Cl − by KCC2.
First, we co-transfected the cells with a pair of two different KCC2 mutants, i.e. a pair of E50_Q93del and A191V mutants or a pair of S323P and M415V mutants, mimicking the condition in individuals 1 and 2 or individual 3, respectively. We confirmed that the E Cl s in cells expressing the two mutants in individuals 1 and 2 (−47.9 ± 3.1 mV, n = 12) and in individual 3 (−42.3 ± 3.9 mV, n = 11) were significantly more positive than the E Cl in WT-expressing cells (−59.9 ± 2.9 mV, n = 12; Fig. 2B). However, the E Cl s in mutant-expressing cells were  The humps of GlyR currents were generated during glycine application at the holding voltage of −40 mV, and the current responses to voltage ramps were generated before and during the humps. Note that the current levels immediately before and after a ramp response during a GlyR current hump were almost unchanged, and therefore the time course of the humps was not affected by ramp responses. This confirmed that the net Cl − flux across the cell membrane during a ramp response did not significantly alter E Cl . See also Supplementary Fig. S1. Bottom traces are the expanded traces of single voltage ramps (upper traces) and superimposed current responses to voltage ramps before and during glycine application (lower traces). Dotted lines indicate the voltage levels at which the superimposed current traces intersected, corresponding to E Cl . (B) Plot of E Cl in cell groups of Mock (n = 10), WT (n = 12), E50_Q93del & A191V (n = 12), and S323P & M415V (n = 11). *P < 0.03, **P < 0.01 by REGW F-test. (C) Plot of E Cl in cells transfected with single vectors encoding WT (n = 11), E50_Q93del (n = 12), A191V (n = 10), S323P (n = 10), and M415V (n = 10). **P < 0.01 by Dunnett's two-sided t-test. significantly more negative than in cells transfected with the vector containing no KCC2 (Mock, −30.2 ± 2.9 mV, n = 10; Fig. 2B). Thus, the Cl − extrusion function of KCC2 was partially reduced by mutations in cells mimicking patient status.
To determine whether individual KCC2 mutants contributed equally to the reduction, we also measured and compared E Cl s in cells only transfected with one type of KCC2 mutant. We found that the E Cl s in cells expressing the E50_Q93del mutant in individuals 1 and 2 (−36.5 ± 3.1 mV, n = 12) and in the cells expressing the M415V mutant in individual 3 (−26.5 ± 3.2 mV, n = 10) were much more positive than that in WT-expressing cells (−53.6 ± 3.8 mV, n = 11). The E Cl s in cells expressing the A191V mutant in individuals 1 and 2 (−45.2 ± 3.7 mV, n = 10) and in the cells expressing the S323P mutant in individual 3 (−47.8 ± 3.5 mV, n = 10) also seemed to be more positive than that in WT-expresing cells, but did not reach statistical significance (Fig. 2C). Thus, individual patients were found to have one KCC2 mutant with a severely impaired Cl − extrusion function and a second mutant with mildly impaired Cl − extrusion function.
Unaltered cellular distribution and cell surface expression of KCC2 by the mutations. To address the cause of reduced Cl − extrusion function of KCC2 mutants, we next compared the cellular distribution of KCC2 mutants with WT KCC2 using immunofluorescence staining. In all WT-expressing and mutant-expressing HEK293 cells mimicking conditions in normal control, individuals 1 and 2 (E50_Q93del and A191V), and individual 3 (S323P and M415V), and also in the cells expressing individual mutants, KCC2 immunoreactivity was similarly detected in the plasma membrane and the perinuclear region ( Fig. 3 and Supplementary Fig. S2). This suggests that the mutations did not greatly alter KCC2 subcellular localization.
We also compared total and cell surface expression levels of individual KCC2 mutants with those of WT KCC2 using the surface protein biotinylation and immunoblotting assay. Results showed no clear differences in total KCC2 expression between WT-and mutant-expressing cells (Fig. 4A). Moreover, the differences in the ratio of surface expression to total KCC2 expression between WT-and mutant-expressing cells did not reach statistical significance (Fig. 4B). Therefore, mutations in the patients did not greatly alter total or cell surface expression levels of KCC2. Table 1, and case reports are available in the supplementary material. The onset of seizures was within 2 months of age, and various types of focal seizures were observed. Ictal electroencephalography (EEG) showed alternating seizures, accompanied by migrating foci from one hemisphere to the other, in three individuals with EIFMS ( Fig. 5A,B). Interictal EEG was initially normal in 2 of 3 patients, and subsequently showed slow-wave activity or focal epileptic discharges compatible with EIFMS. Individual 4 was not definitively diagnosed with EIMFS, due to the lack of former medical records including ictal electroencephalograms. However, multifocal seizures, which started at variable parts of the body and migrated to other regions, were observed during infancy by her parents, which strongly suggested EIFMS. All four individuals exhibited severe developmental delay, hypotonia, and postnatal microcephaly. Brain MRI showed cerebral atrophy in four individuals, and delayed myelination and thin corpus callosum in three and two individuals, respectively. Individual 4 showed progressive cerebellar atrophy and hippocampal atrophy ( Fig. 5C-J). The combination of potassium bromide and high-dose phenobarbital was effective for individuals 1 and 2, and a ketogenic diet controlled seizures in individual 3. Seizures of individual 4 were intractable.

Discussion
In this study, we identified four patients exhibiting severe infantile epileptic seizures with compound heterozygous mutations in SLC12A5, which encodes the neuronal K + -Cl − co-transporter KCC2. Three of the patients were diagnosed with EIMFS, and multifocal migrating seizures were also observed in an additional patient (individual 4). Together with the previous report 14 , our data strongly indicate that biallelic SLC12A5 mutations cause migrating focal seizures, which is characteristic of EIMFS.
Functional analysis of the four KCC2 mutants revealed that each of the three patients (individuals 1-3) has two KCC2 mutants exhibiting differentially impaired Cl − extrusion function. The combination of mutants caused an average positive E Cl shift by 12-18 mV from the E Cl in WT-expressing cells (Fig. 2B), corresponding to an increase in [Cl − ] i by 10-15 mM. This positive E Cl shift in neuronal populations could result in an increased fraction of neurons that exhibit an excitatory response to GABA [18][19][20][21] . In the brains of patients with temporal lobe epilepsy, interictal discharge events detected during local field recording or intracranial EEG recording in epileptic foci are preceded by synchronous burst firing of GABAergic interneurons, and the firing triggers not only inhibitory postsynaptic potentials in most pyramidal neurons but also out-of-phase firing of a proportion (~20%) of pyramidal neurons exhibiting depolarizing GABA responses [19][20][21][22] . This out-of-phase firing contributes to the pathological high-frequency fast ripple component of the discharges 23,24 . Thus, similar out-of-phase firing activities might also take place in the brains of our patients, which would lead to a migrating focal seizure event when additional vulnerability factors are imposed on a part of the brain [20][21][22][23][24] . A significant suppression of seizures with a high dose of phenobarbital, a GABA A receptor enhancer, in our patients (individuals 1 and 2; but its effect was temporary; see Table 1 and Supplementary Information for details) implies that strong enhancement of inhibitory GABA action on surrounding neurons is necessary for diminishing the effect of out-of-phase firing, even of a minor proportion of neurons.
Individual mutations found in this study did not greatly alter cellular distribution ( Fig. 3 and Supplementary  Fig. S2), surface or total expression levels of KCC2 (Fig. 4). Nevertheless, our immunostaining and blotting would have not resolved subtle differences in the surface expression level, which might explain the subtle differences in Cl − extrusion function especially between p.A191V and p.S323P mutants and WT KCC2. The surface expression of KCC2 may be determined by the phosphorylation/dephosphorylation balance of the residues, especially in the  4 has not yet been examined. It is also possible that the impaired Cl − extrusion function of our mutants would be due to impairment in an intrinsic transport property of KCC2. Although the tertiary structure and ion-transporting structural element of KCC2 remain unknown, the mutations may affect Cl − binding of KCC2 or cause some derangement in the KCC2 structure, thereby impairing Cl − extrusion. Further studies are needed to elucidate whether individual mutations affect either or both the trafficking and the intrinsic transport property of KCC2.
The KCC family members are thought to function as oligomers 15 . A previous report about a KCC3 mutant defective in its surface expression suggested its dominant-negative effect on the expression of not only WT-KCC3 but also WT-KCC2 through forming heteromers 25 . Thus it might be possible that individual KCC2 mutants found in this study exert some effect on other mutants or WT-KCC2 through oligomerization. We have not determined to what extent such interactions actually work in patients and also in heterozygous healthy parents. However, given that the mean E Cl s in the cells expressing KCC2 mutants as in the patients (Fig. 2B) were similar to or more negative than the average of the mean E Cl s in cells expressing individual mutants (Fig. 2C), the severely impaired mutants of E50_Q93del and M415V would not have a strong dominant-negative effect. This further implies that the mean neuronal [Cl − ] i level in heterozygous healthy parents would be lower than that in patients, because of the presence of normal KCC2 in the parents. However, the differences in neuronal [Cl − ] i between the parents and the patients or normal controls, if any, would be too small to detect by our E Cl measurement. In any case, our genetic evidence certainly indicated that compound heterozygous mutations in SLC12A5 only match the presence of symptoms in the patients. Thus, even though the mean neuronal [Cl − ] i level in the heterozygous parents might actually be slightly higher than that in normal controls, it should not be the level causing diseases. Therefore, the number of neurons firing out-of-phase, which predispose an individual to an ictal event, would not be significantly increased in the parents.
A recent study also reported four infantile patients exhibiting migrating focal seizures with mutations in SLC12A5 14 . Two patients had compound heterozygous missense mutations c.1277T > C (p.L426P) and c.1652G > A (p.G551D), and the other two (one deceased) possessed a homozygous missense mutation c.932T > A (p.L311H). Functional analysis concluded that the Clextrusion function of L426P and G551D mutants was completely lost, whereas the L311H mutant was still partly functional, and that these functional losses were due to reduced surface expression and glycosylation of these mutants 14 . However, the study measured E Cl under whole-cell patch-clamp conditions, in which the basal [Cl − ] i during recordings was determined by the Cl − concentration in the pipette solution. Therefore, the data did not provide information about the impact of the mutations on neuronal [Cl − ] i levels in the patients. Moreover, the authors used whole-cell pipette solution containing 110 mM Cs + , instead of K + . KCC2 excludes Cl − with K + out of the cells using the K + transmembrane gradient, but the replacement of intracellular K + with Cs + is known to block KCC2-mediated Cl − extrusion [26][27][28] . Therefore, in the preceding study 14 , KCC2 activity must have been inhibited and E Cl would have not been correctly recorded. Conversely, using Cl − -impermeable gramicidin channels as the current mediator incorporated into the patch membrane, our study clearly demonstrated, for the first time, that E Cl in cells expressing the KCC2 mutants in patients shows a positive shift, but remains more negative than expected E Cl level in the absence of KCC2 (Fig. 2B). Therefore, we confirmed that KCC2 mutant function in our patients was reduced, but still functional, although the collectively reduced function of two mutant alleles is sufficient to cause severe epileptic seizures. The onset of seizures in patients within a few days after birth (Table 1) was compatible with the period of increasing functional KCC2 expression [29][30][31][32][33][34] , and this may also support our conclusion.
In conclusion, our data demonstrated that individual mutations in EIMFS patients causes variable loss of KCC2 function, and that the combinatory effect of partial loss of KCC2 function in each patient results in focal seizures, severe developmental delays, and postnatal microcephaly.

Materials and Methods
Patients. A total of 10 sporadic cases and one family with two affected siblings, who had an initial diagnosis of EIMFS, were analyzed by WES as an initial cohort. Patients with mutations in known epilepsy genes related to EIMFS (KCNT1, SCN1A, SCN2A, SCN8A, PLCB1, SLC25A22, and TBC1D24) 2-7 were excluded from the study. Additionally, we searched WES data from 526 patients with infantile epilepsy, and examined 141 patients with infantile epilepsy by SLC12A5-targeted resequencing as a second cohort. DNA was extracted from peripheral blood leukocytes using standard methods. DNA was extracted from saliva samples from the father and elder brother of individuals 1 and 2, as well as from the elder brother of individual 4, using Oragene (DNA Genotek). Detailed clinical information was obtained from corresponding clinicians. Written informed consent was obtained for all individuals. Experimental protocols were approved by the Institutional Review Board of Yokohama City University School of Medicine, and were carried out in accordance with the approved guidelines.

year
Rolling over 2 years 6 months, but incomplete at 4 years 6 months -4 months to 5.5 months, regressed after 5.5 months, able to roll over again at 9-10 months   reverse transcription, and 2 μl cDNA was used for PCR. PCR conditions and primer sequences are shown in Supplementary Table S3. PCR products were electrophoresed on a 1.5% agarose gel. PCR bands were cut from the gel, purified using the QIAEXII Gel Extraction Kit (Qiagen), and sequenced.

Expression vectors.
A full-length human cDNA of SLC12A5 transcript variant 2 (clone ID: RC223680) was obtained from Origen (Rockville, MD). Site-directed mutagenesis using a KOD-Plus-Mutagenesis kit (Toyobo) was used to generate SLC12A5 mutants, including c.148_279del (p.E50_Q93del), c.572C > T (p.A191V), c.967T > C (p.S323P) and c.1243A > G (p.M415V). All variant cDNAs were verified by sequencing. WT and mutant SLC12A5 cDNAs were cloned into either the pCIG-HA or pCIR-HA vector, in which a N-terminal HA-tag sequence was introduced by PCR to parental pCIG or pCIR vectors 36,37 to express N-terminal HA-tagged KCC2b as well as nuclear-localized EGFP or DsRed. Co-expression of different mutants was confirmed by the presence of both EGFP and DsRed in the nucleus.
Cell culture and transfection. A stable HEK293 cell line expressing GlyRα1 (HEK293-GlyRα1) was generated as previously described 17 , except for the use of the pCMV-GlyRα1WT vector 38  , and 10 μM bumetanide to block endogenous Na + -K + -2Cl − cotransporters in HEK293 cells. The liquid junction potential between these solutions was 2.8 mV and was corrected online. Cells were placed on a small glass-bottom recording chamber filled with 0.5 ml of external solution, and the cells expressing nuclear EGFP and/or DsRed were selected under epifluorescent illumination. A > 5 GΩ (usu. ∼10 GΩ) gigaseal was first formed, and then we typically waited for 1-1.5 hours until the series resistance (Rs) was reduced to <100 MΩ by insertion of gramicidin into the patch membrane before recordings. The Rs during recordings was compensated for by 70%. Our high-quality gigaseal recording resulted in a very low basal current level (6.8 ± 0.8 pA in absolute value, n = 98) at E Cl . Thus IR-drop errors through Rs (81.5 ± 4.3 MΩ) with 70% Rs compensation were 0.16 ± 0.02 mV in absolute value, indicating negligible IR-drop errors in our E Cl values. The holding voltage was set at −40 mV and 1 s voltage ramps from −80 to −10 mV were applied before and during bath application of 100-300 μM glycine. The current levels immediately before and after a voltage ramp during a glycine-induced current hump were almost unchanged ( Fig. 2A and Supplementary Fig. S1), confirming that the net Cl − flux across the cell membrane during a voltage ramp did not significantly alter E Cl . E Cl measurements in a single cell were repeated more than 3 times at >3 min intervals ( Supplementary Fig. S1), and the average over three successive measurements was adopted as the final E Cl value. This value was plotted in the graphs in Fig. 2B,C. When the variation of three successive E Cl values did not converge within ±1 mV, the cell was discarded from the data. All experiments were performed at 26-28 °C.

Immunoblotting. Surface biotinylation experiments were performed using a Pierce Cell Surface Protein
Isolation kit (Thermo Fisher Scientific) according to the manufacturer's protocol. Briefly, HEK293 cells expressing WT KCC2 or KCC2 mutants were washed with ice-cold PBS and then labeled with 0.25 mg/ml sulfo-NHS-SS-biotin for 30 min at 4 °C. Excess biotin was quenched with glycine buffer. The cell lysates were centrifuged (10,000 g for 10 min), the supernatant was isolated with NeutrAvidin gel, and the bound proteins were then eluted with SDS-PAGE sample buffer. Total cell lysate and biotinylated proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane. The blots were blocked in 1% bovine serum albumin and incubated overnight at 4 °C with following primary antibodies: rabbit anti-KCC2 (1:1000, Millipore, #07-432), and mouse anti-transferrin receptor (TfR) (1:500, clone H68.4, Zymed Laboratories). The cells were then incubated with horseradish peroxidase-conjugated secondary antibody (GE Healthcare) for 1 h at room temperature. Immunoblots were visualized with enhanced chemiluminescence (ECL) exposed onto Polaroid instant films through the ECL Mini-camera (GE Healthcare). Band intensities were measured using ImageJ software. Surface and total KCC2 band densities were normalized to the TfR band density. TfR is a membrane protein unrelated to KCC2 and was used as a loading control.