A novel missense LGI1 mutation produces a secretion-defective LGI1 protein

We treated a 16-year-old Chinese man who developed recurrent epileptic seizures (see Materials and Methods and Additional file). His episodes were characterized by generalized tonic–clonic seizures preceded by auditory auras. His father and grandfather had similar manifestations from the puberty, strongly indicating ADLTE [5, 48]. We applied whole-exome sequencing using a blood sample of the proband, and identified a novel missense LGI1 variant, c.547 T (p.Trp183Arg) (Fig. 1A, B), which is absent from the ExAC, dbSNP, 1000G, and gnomAD databases. Moreover, both his father and grandfather carried the same heterozygotic variant (Fig. 1A), consistent with a pattern of autosomal dominant inheritance. The Trp183 residue is highly conserved throughout vertebrate species (Fig. 1C), and is located in the C-cap domain of LGI1 (Fig. 1C, D). The brain magnetic resonance imaging (MRI) of the proband appeared normal (Fig. 1E). The drug treatment we provided effectively controlled his seizures (see Additional file), and thereby only brief and occasional interictal epileptiform discharges were detected after the treatment (Fig. 1F).

Fig.1

Genetic and expression analysis of the LGI1W183R mutation. A Pedigree of the LGI1W183R variant. (square, male, circles, females; see Additional file for details). B Chromatograms of c.547 T > C within LGI1. (upper, normal sequence; lower, mutant sequence). C Upper, domains within the LGI1 protein: N-cap, LRR, C-cap, and EPTP. The LGI1W183R mutation maps onto the C-cap domain. Lower, alignment analysis of LGI1 orthologues in different vertebrate species shows the conservation of the W183 residue. D Mapping of the LGI1W183R mutation on the 3D structure of LGI1. E Brain MRIs of the proband showing normal brain structure (horizontal, coronal, and sagittal views from left to right). F Representative EEG recording from the proband (arrowhead, an epileptiform discharge during the interictal period; left electrodes in the Bipolar Montage. G mRNA levels of LGI1 in HEK cells transfected with LGI1WT or LGI1W183R. H Total LGI1 protein in cultures transfected with LGI1WT or LGI1W183R. I LGI1 expression in the medium (secreted) and cell lysates of cultures transfected with LGI1WT or LGI1W183R. J LGI1 protein levels at annotated time points in HEK cells transfected by LGI1WT or LGI1W183R and continuously treated with CHX. K Immunoprecipitation with rabbit anti-Flag antibody followed by western blots using mouse anti-Flag antibody and mouse anti-Flag-ubiquitin show no difference in LGI1 ubiquitination level between the two groups (Flag-LGI1WT or Flag-LGI1W183R transfected into HEK cells). See Additional file 4: Table S1 for statistics. **P < 0.01. ***P < 0.001

To characterize this missense mutation, we generated plasmids encoding wild-type LGI1 (LGI1WT) or mutant LGI1W183R, and transfected them into HEK cells. mRNA analysis revealed no difference in the level of transcripts between LGI1W183R and LGI1WT (Fig. 1G). Western blots with anti-LGI1 antibody also showed no difference in the protein level between LGI1W183R and LGI1WT (Fig. 1H). It has been shown that some of LGI1 mutations yield secretion-defective LGI1 proteins [26, 28]. To investigate whether LGI1W183R mutation leads to defective secretion, we measured LGI1 protein in conditioned medium and cell lysates, and found that LGI1WT protein was secreted by HEK cells, whereas LGI1W183R protein was not (Fig. 1I). We compared the stability of LGI1WT and LGI1W183R proteins in the cell lysates using cycloheximide (CHX) treatment. The expressions of both LGI1WT and LGI1W183R proteins were reduced over time during CHX treatment, but the rate of decrease of LGI1WT was much faster than that of LGI1W183R (Fig. 1J), which might be due to the secretion of LGI1WT. To investigate whether the LGI1W183R mutation affects LGI1 ubiquitination, we added ubiquitin to HEK cells and found that both LGI1WT and LGI1W183R proteins were ubiquitylated to the same extent, implying that LGI1W183R mutation does not affect the degradation of LGI1 (Fig. 1K).

Expressing LGI1W183R in excitatory neurons results in epileptic seizures

Previous work has shown that the depletion of LGI1 in excitatory neurons is critical to the onset of seizures [19], highlighting the importance of excitatory neurons in ADLTE. Hence, we hypothesized that the pathogenic mechanism of LGI1W183R mutation is dependent of excitatory neurons. To test this idea, we expressed LGI1W183R solely in excitatory neurons in the brain. First, we applied Cre-Loxp technique to generate LGI1flox/+ mice (Fig. 2A), which were further intercrossed with CaMKII-Cre mice to obtain either CaMKII-Cre;LGI1flox/+ (Het) or CaMKII-Cre;LGI1flox/flox (cKO), where LGI1 was either haploinsufficient or deficient in excitatory neurons (Fig. 2B). Next, we injected AAV9-DIO-LGI1WT-GFP or AAV9-DIO-LGI1W183R-GFP bilaterally into the ventricles of either Het or cKO mice at P0 (Fig. 2B). With this approach, LGI1WT-GFP or LGI1W183R-GFP was specifically expressed in excitatory neurons, in which endogenous LGI1 was deleted or haploinsufficient. Finally, cKO or Het mice with exogenous LGI1WT or LGI1W183R were subjected to video monitoring of autonomous or PTZ-induced epileptic seizures and/or electroencephalogram (EEG) recordings (Fig. 2B). With this strategy, the expression of LGI1WT or LGI1W183R was confirmed by confocal imaging of GFP fluorescence and CaMKIIα expression. We found that the GFP signal was broadly expressed in the cerebral cortex and the hippocampus (Fig. 2C). Moreover, GFP signal was well co-localized with CaMKIIα signal (Fig. 2C), indicating the specific expression of LGI1WT and LGI1W183R proteins in excitatory neurons. Counting the numbers of neurons showed no difference in the ratio of GFP-positive neurons among CaMKIIα-positive neurons between the LGI1WT and LGI1W183R groups at two developmental stages, P17–20 and P35 (Fig. 2D). These results indicate that, with our strategy, exogenous LGI1WT and LGI1W183R proteins are robustly expressed in excitatory neurons upon the viral injection of LGI1WT or LGI1W183R into the ventricles. This conclusion was further strengthened by confocal imaging of GFP and the specific marker proteins for other major types of brain cells, including parvalbumin (PV)-positive interneurons, astroglia, and microglia (Additional file 1: Fig. S1).

Fig.2

Expressing LGI1W183R in excitatory neurons increases seizure susceptibility. A Gene targeting strategy for the generation of LGI1flox/flox mice. B AAV9-DIO-LGI1WT-GFP or AAV9-DIO-LGI1W183R-GFP is expressed in cKO or Het excitatory neurons (CaMKII-Cre). Viruses are bilaterally injected into the ventricles (P0). cKO mice subjected to video observation and electrophysiological recording at P17–20 and Het mice are subjected to PTZ treatment and EEG recording at P35. C Representative images for triple fluorescence, GFP, CaMKII(CKII) and DAPI, in the hippocampus (hip) and temporal cortex (temp lb) of cKO mice (P17) (scale bars: 1 mm (whole brain); 50 μm (magnified). D Number ratios of GFP + vs CaMKII-Cre + cells in cKO and Het mice expressing LGI1WT or LGI1W183R (n = 5 per group). At P17–20, the ratio of GFP + vs CaMKII + is 43 ± 4 (CA1; LGI1WT) and 42 ± 4 (CA1; LGI1W183R), P = 0.86; 41 ± 4 (temp lb; LGI1WT) and 42 ± 5 (temp lb; LGI1W183R), P = 0.86. At P35, the ratio of GFP + vs CaMKII + is 44 ± 5 (CA1; LGI1WT) and 44 ± 5 (CA1; LGI1W183R), P = 0.98; 42 ± 3 (temp lb; LGI1WT) and 38 ± 5 (temp lb; LGI1W183R), P = 0.47. E Kaplan–Meier survival curves. F Quantification of reactions to PTZ injection. Latency to generalized seizure (GS): 383 ± 50 s (Het::LGI1WT; n = 5) and 212 ± 33 s (Het::LGI1W183R; n = 10), P = 0.023. G Representative EEGs and power spectral analysis in Het::LGI1WT and Het::LGI1W183R mice (P35) during PTZ-induced seizures. G Enlarged view of EEGs in G. H Spectral analysis of the EEGs. Grey dots indicate individual data points. *P < 0.05

The loss of LGI1 in excitatory neurons causes epileptic seizures and premature death [19]. Accordingly, we considered whether LGI1W183R protein in excitatory neurons may lead to these phenotypes as well. The majority of cKO::LGI1W183R mice (17/19) died before P21 and their median lifetime was 22 days, as shown in the Kaplan–Meier survival curves (Fig. 2E). In contrast, cKO::LGI1WT, Het::LGI1WT and Het::LGI1W183R littermates survived for > 40 days (Fig. 2E).The cKO::LGI1W183R mice often had spontaneous seizures (generalized tonic or clonic seizures) (Additional file 9: Video S1) at an onset age of P16 with a frequency between 0.25 and 1 per hour and a mean duration of 45.5 ± 18.7 s (Additional file 6: Table S2). In contrast, cKO::LGI1WT, Het::LGI1WT and Het::LGI1W183R mice displayed no autonomous seizures. Because mutant mice with LGI1 haploinsufficiency display increased seizure susceptibility to PTZ [28], we examined PTZ-induced seizures in Het::LGI1WT and Het::LGI1W183R mice at P35. Our results showed that, compared to Het::LGI1WT mice, seizure severity was significantly greater in Het::LGI1W183R mice, as shown by more frequent generalized seizures: the majority of Het::LGI1W183R mice (10/15) were at stages 3/4 and displayed a shortened latency to generalized seizures, whereas the majority of Het::LGI1WT (10/15) mice were at stages 1/2 (Fig. 2F). The behavioral difference was confirmed by EEG recordings. Energy spectra of representative epileptic EEGs recorded from Het::LGI1WT and Het::LGI1W183R mice are shown in Fig. 2G and the absolute power at each firing frequency is shown in Fig. 2H. These results indicated that expressing LGI1W183R in excitatory neurons augmented PTZ-induced seizure severity compared to expressing LGI1WT (Fig. 2G, H). Taken together, we conclude that LGI1W183R in excitatory neurons is sufficient to cause epileptic seizures in mice.

LGI1W183R causes hyperexcitability and firing irregularity in hippocampal pyramidal neurons

To investigate the mechanism by which LGI1W183R regulate neuronal activity, we performed whole-cell recordings in hippocampal CA1 pyramidal neurons of cKO::LGI1WT and cKO::LGI1W183R mice aged P17–20 (Fig. 3A). Single action potentials (APs) were induced by rheobase current injection and their major kinetic parameters were analyzed (Fig. 3B). We found that cKO::LGI1W183R neurons required a smaller rheobase than cKO::LGI1WT neurons, but showed no difference in membrane capacitance between two groups (Fig. 3C). The plots of rate of change membrane potential (dV/dt) vs membrane potential revealed that AP waveform differed between cKO::LGI1WT and cKO::LGI1W183R neurons (Fig. 3D). In cKO::LGI1W183R neurons, AP threshold was more hyperpolarized, the half-width was increased, and the values of dV/dt at + 20 mV and − 40 mV were increased (Fig. 3E). Meanwhile, AP amplitude and resting membrane potential (RMP) were unaltered (Fig. 3E). The altered AP parameters indicated that exogenous LGI1W183R may influence the generation of AP. In fact, rheobase current that induced a single AP in cKO::LGI1WT neurons could evoke doublet APs in cKO::LGI1W183R neurons (Fig. 3F), suggesting that cKO::LGI1W183R neurons are more excitable than cKO::LGI1WT neurons.

Fig.3

Hyperexcitability and spiking irregularity in hippocampal neurons expressing LGI1W183R. A Schematic of whole-cell recording in cKO CA1 pyramidal neurons expressing LGI1WT-GFP or LGI1W183R-GFP (pip: patch pipette). B AP evoked by a rheobase current (black), but not subthreshold currents (grey) (arrows, threshold, RMP, amplitude, and half-width. C Averages of membrane capacitance (Cm) and rheobase. D Left: example APs in cKO::LGI1WT and cKO::LGI1W183R neurons. Right: phase-plane plots for APs. The arrowheads show the measurement of threshold and dV/dt at 0, + 20 (repolarization) and − 40 mV. E Averages of RMP, threshold, half-width, amplitude, dV/dt at 0 mV, dV/dt at + 20 mV, and dV/dt at − 40 mV. F Left: example APs induced by rheobase current in cKO::LGI1WT and cKO::LGI1W183R neurons. Right: probabilities of doublet APs. G Left: example spikes recorded in cKO::LGI1WT and cKO::LGI1W183R neurons responding to 80-pA and 200-pA currents. Right: numbers of spikes as a function of injected currents. H Left: representative 1st and last spikes induced by 200-pA current. Middle: half-widths of 1st spikes induced by different currents plotted against to currents. Right: ratios of last vs 1st spikes were plotted against corresponding currents. I Left: example firing showing spike-timing reliability. Right: plots of intraburst jitters as a function of recording time. J Averages of 1st ISI frequency, CV, and CV2. K Left: example AHPs. Right: plots of fast AHP as a function of spikes. Insets amplification of 2nd and 4th AHPs. See Additional file 8: Table S3 for statistics. Grey dots indicate individual data points. *P < 0.05. ** P < 0.01. *** P < 0.001

Next, the population firings in pyramidal neurons were induced by a depolarizing step current. We found that an 80-pA current produced more spikes in cKO::LGI1W183R neurons than in cKO::LGI1WT neurons (Fig. 3G), suggesting that LGI1W183R increases the firing potential. Unexpectedly, further study demonstrated that a 200-pA current produced fewer spikes in cKO::LGI1W183R neurons, which was manifested by the input–output curves (injected current-number of spikes) obtained from cKO::LGI1WT and cKO::LGI1W183R neurons (Fig. 3G). Previous work has suggested that altered AP half-width is the reason for such bi-directional changes accompanying increasing stimulation intensities [36]. Indeed, we showed that AP half-width was increased by LGI1W183R (Fig. 3E). To test if this is the case for population firing, we measured the half-width of the 1st spike evoked by increasing stimuli (80–200 pA), and found that the values were always larger in cKO::LGI1W183R neurons than in cKO::LGI1WT neurons (Fig. 3H). Furthermore, we calculated the ratio of half-width of the last vs the 1st spikes. Our results demonstrated that: the spike became wider in both cKO::LGI1WT and cKO::LGI1W183R neurons as time passed; but this ratio was always greater in cKO::LGI1W183R neurons than in cKO::LGI1WT neurons (Fig. 3H), which explained the reduced number of spikes upon stimulation with large currents.

Also, we noted that the 1st interspike interval (ISI) in a firing was shorter in cKO::LGI1W183R neurons (Fig. 3G). To clarify this point, we adjusted the intensity of injection currents (60–80 pA) to induce exactly 7 spikes in recorded neurons (Fig. 3I) [49], and measured two parameters: 1st ISI frequency, which was augmented significantly in cKO::LGI1W183R neurons (Fig. 3J), and the regularity of firing, which was characterized by the coefficient of variation of all ISIs (CV) and the coefficient of two consecutive ISIs (CV2) [50]. As shown by event rasters (Fig. 3I) and statistics of CV and CV2 (Fig. 3J), the firing became more irregular in cKO::LGI1W183R neurons than in cKO::LGI1WT neurons. It has been suggested that the ISI depends on the after-hyperpolarization potential (AHP) [34, 35], which can be separated into two parts, fast and slow AHPs [51, 52]. In our hands, we found that the fast AHP was decreased for the second and third spikes in cKO::LGI1W183R neurons, but the difference gradually declined over time (Fig. 3K). Thus, these altered fast AHP may explain the irregularity of spontaneous spikes in cKO::LGI1W183R neurons.

Kv1.1 activity is down-regulated in cKO::LGI1W183R neurons

A putative function of LGI1 is its modulation of glutamatergic transmission through binding ADAMs [24, 25]. However, the LGI1W183R mutation unlikely plays this role, since it yields a secretion-defective LGI1 protein. Indeed, we found that neither the amplitude nor the frequency of mEPSCs was altered by LGI1W183R expression in cKO neurons (Additional file 2: Fig. S2). Alternatively, LGI1W183R may act on ion channels, as we had previously demonstrated that Kv1 is down-regulated in cortical neurons upon LGI1 ablation [18]. To test this possibility, we made whole-cell recordings in cKO hippocampal neurons expressing LGI1WT-GFP or LGI1W183R-GFP with perfusion of DTx-K (a specific Kv1.1 antagonist [53, 54]) (Fig. 4A). By applying a series of stepped voltage pulses to neurons, we obtained Kv1.1 current by subtracting DTx-K-sensitive current from overall K+ current (Fig. 4B). Our results showed an overall decrease in Kv1.1 current in cKO::LGI1W183R neurons (Fig. 4B), indicating that LGI1W183R inhibits Kv1.1 current. This conclusion was strengthened by analyzing the activation and inactivation of Kv1.1 current, which were defined as the currents evoked by depolarizing voltages and the currents evoked by 3-s inactivating pre-pulses, respectively [18]. As shown by normalized conductance recorded at stepped voltages (from − 70 to + 40 mV), both the activation and the inactivation of Kv1.1 current were reduced by LGI1W183R expression (Fig. 4C, D). Further kinetics analysis showed no effect of LGI1W183R on the slope of activation curve and half-activation voltage, and that there was an increase in half-inactivation voltage, but not the slope of the inactivation curve, following LGI1W183R expression (Fig. 4E). These analyses reveal that LGI1W183R exerts strong regulatory effects on Kv1.1 activity, which can alter the waveform of the AP and spiking pattern [36, 55,56,57,58]. To determine the cause of Kv1.1 current reduction, i.e. whether it is due to a reduction of single-channel conductance or the number of active channels, we performed non-stationary noise analysis on the activation of Kv1.1 in cKO neurons expressing LGI1WT or LGI1W183R at a command voltage of + 40 mV [59, 60]. Plotting current variance as a function of current amplitude yielded a parabola, whose parameters are suitable to determine single-channel conductance and the number of active channels [59, 60]. Our analysis indicated that LGI1W183R expression reduced the number of active Kv1.1 channel (8417 for LGI1WT and 5936 for LGI1W183R), while single-channel conductance was not affected (0.31 for LGI1WT and 0.33 for LGI1W183R) (Fig. 4F). These data show that the reduction in Kv1.1 current is due to a reduced number of active channels.

Fig.4

Downregulation of Kv1.1 activity in cKO::LGI1W183R neurons. A Schematic of whole-cell recording in cKO::LGI1WT and cKO::LGI1W183R neurons perfused with DTx-K. B Activated K+ current by stepped voltage pulses (− 70 to + 40 mV) in neurons before and after application of DTx-K (100 nM). Kv1.1 current (DTx-K-sensitive) are from current subtraction. C Kv1.1 current during the activation phase normalized to cell capacitance (current density) and plotted against command voltage. D The inactivation of Kv1.1 current by stepped voltage pulses (− 70 to + 10 mV). E Left, steady-state activation and inactivation curves of Kv1.1 current normalized to maximal conductance. Right, averages of half-voltages for the activation and inactivation curves. F Non-stationary noise analysis of Kv1.1 activation at a command voltage of + 40 mV. Current variance is plotted against the amplitude at a given time point. Single-channel conductance and number of active channels are determined by fitting a parabola to the data points. G Example APs induced by rheobase in LGI1WT and LGI1W183R neurons treated with DTx-K. H Averages of AP threshold, rheobase, half-width, and amplitude with the addition of DTx-K. I Curves of spikes vs injected current in neurons perfused with DTx-K. J Left: averages of half-width of 1st spike induced by different current injections in neurons perfused with DTx-K. Right: ratios of last vs 1st spikes plotted against injected current. K Plots of intraburst jitters as a function of recording time in neurons perfused with DTx-K. L Average values of 1st ISI frequency, CV, and CV2 in neurons perfused with DTx-K. See Additional file 5: Tables S4 for statistics. Grey dots indicate individual data points. *P < 0.05. **P < 0.01. ***P < 0.001

If LGI1W183R reduces Kv1.1 activity, it is reasonable to assume that inhibiting Kv1.1 is able to annihilate the difference in intrinsic excitability between cKO::LGI1WT and cKO::LGI1W183R neurons. To test this idea, we perfused DTx-K onto cKO neurons expressing either LGI1WT or LGI1W183R, and examined the AP and spikes. Under this condition, AP waveform showed similar kinetics in cKO::LGI1WT and cKO::LGI1W183R neurons (Fig. 4G). The statistics showed no difference between cKO::LGI1WT and cKO::LGI1W183R neurons in a number of AP parameters, including threshold, rheobase, half-width, and peak amplitude (Fig. 4H). Again with stepped current injection, we compared the number of spikes, half-width of 1st spike, half-width ratio of the last vs 1st spike, and spiking regularity, which were shown to differ between cKO::LGI1WT and cKO::LGI1W183R neurons (Fig. 3). With the perfusion of DTx-K, no difference was found in the numbers of spikes for all intensities of injected currents, as shown by the input–output curves (Fig. 4I). We induced 7 spikes in cKO::LGI1WT and cKO::LGI1W183R neurons with the perfusion of DTx-K. Likewise, DTx-K eliminated the differences in the half-width of 1st spike and half-width ratio of the last vs 1st spike, when cKO::LGI1WT and cKO::LGI1W183R neurons were injected with the same current (Fig. 4J). In addition, DTx-K application changed the firing of cKO::LGI1WT neurons, making the pattern equal to that of cKO::LGI1W183R neurons (Fig. 4K). The statistics showed that the values of 1st ISI frequency, CV, and CV2 were all increased by DTx-K application in cKO::LGI1WT neurons, while these parameters were not altered in cKO::LGI1W183R neurons (Fig. 4L), thereby eliminating the differences between two groups. Taken together, we conclude that a reduction in Kv1.1 activity is the cause of the abnormal excitability in cKO neurons expressing LGI1W183R.

Kv1.1 control spiking pattern of pyramidal neurons: evidence from computer simulation

To better elucidate the contribution of Kv1.1 to neuronal firing, we constructed a neuronal model containing a repertoire of voltage-dependent ion channels (Fig. 5A) [43, 61,62,63]. The conductance densities were adjusted to generate an AP at a threshold of − 20 mV above RMP, and firing was elicited at suprathreshold currents.

Fig.5

Computer model of CA1 neuronal firing pattern. A A schematic model of a CA1 pyramidal neuron. B APs evoked by rheobase in the cell model containing normal (0.02) or the half (0.01) of Kv1.1. C The same strength of stimulation induces a single AP with Kv1.1 (0.02), but doublet APs with insufficient Kv1.1 (0.01). D A reduction in Kv1.1 results in more firing when the cell receives a 200-pA current injection (400 ms) (gray bars, intervals between 1st and 2nd spikes). E Reduced Kv1.1 results in less firing when the cell receives a 700-pA current injection (400 ms). F The amplifications of 1st and 4th spikes induced by 200-pA current. Note the difference in the half-width between 1st and 4th spikes, or between normal and half Kv1.1 conditions. G The amplifications of 1st and 5th AHPs and fitting analysis (gray lines) of AHP currents with normal or half Kv1.1

Kv1.1 density was set at half of normal to mimic the reduction of Kv1.1 activity caused by LGI1W183R. Our simulation showed that Kv1.1 reduction significantly widened the AP (Fig. 5B). We then applied a rheobase current, which was sufficient to evoke a single AP in a cell containing normal Kv1.1, to a cell containing half of the Kv1.1. We found that the rheobase current induced a single AP in the cell with normal Kv1.1, but induced doublet APs in the cell with insufficient Kv1.1 (Fig. 5C). These results confirm that the pattern of neuronal APs is dependent on Kv1.1.

Next, we investigated the dependence of firing on the density of Kv1.1. Two levels of current injections (200 pA and 700 pA) were applied to the cell model as the low and high intensities of current injection, respectively. With 400-ms duration, 200-pA current induced 4 APs with normal Kv1.1, but 6 APs with half Kv1.1 (Fig. 5D). Moreover, the ISI between first two spikes was reduced in the cell with half Kv1.1 (Fig. 5D). However, 700-pA stimulation led to 8 spikes with normal Kv1.1, but 6 spikes with reduced Kv1.1 (Fig. 5E). The bidirectional change in the number of spikes with low and high intensities of stimulation is consistent with our whole-cell recording results. We amplified the spikes induced by 200-pA stimulation, and found that Kv1.1 reduction increased the half-width of 1st and 4th spikes (Fig. 5F). Moreover, the half-width ratio of 4th/1st spikes appeared more significant with reduced Kv1.1 (Fig. 5F), also consistent with our whole-cell recordings.

The resurgence of AHP current appeared slower with half Kv1.1 (Fig. 5E), which may explain the altered firing pattern. To test this point, we analyzed AHPs in early and late spikes with normal or fewer Kv1.1 channels. The polynomial fitting showed that Kv1.1 reduction caused a different pattern of fast AHP between the 1st to and the 2nd spikes, that is, the AHP tended to depolarize with normal Kv1.1 (slope coefficient: 68), but tended to hyperpolarize with reduced Kv1.1 (slope coefficient: − 214) (Fig. 5G). Interestingly, in the follow-up AHPs, the difference began to decrease, showing that the slope coefficient was 30 for normal Kv1.1 and − 3 for fewer Kv1.1 channels (Fig. 5G). Therefore, these data suggest that Kv1.1 reduction permits different firing pattern by delaying the onset of AHP currents.

Restoring Kv1.1 alleviates seizure susceptibility in cKO::LGI1W183R mice

Having demonstrated that reduced activity of Kv1.1 by LGI1W183R expression is responsible for epileptogenesis, an interesting question was whether the seizures can be ameliorated and whether the lifespan can be prolonged by restoring Kv1.1. To do so, we injected AAV9-DIO-LGI1W183R-GFP with AAV9-DIO-Kv1.1-mCherry or AAV9-DIO-mCherry bilaterally into the ventricles of Het or cKO mice at P0 (Fig. 6A). Using this approach, LGI1W183R and Kv1.1, as exogenous proteins, were simultaneously expressed in excitatory neurons from cKO or Het mice. Later, the mice expressing LGI1W183R and Kv1.1-mCherry or mCherry were subjected to video monitoring of autonomous or PTZ-induced seizures and/or electrophysiological recordings (Fig. 6A). The expression of LGI1W183R and Kv1.1 was confirmed by