Specific deletion of Na_V1.1 sodium channels in inhibitory interneurons causes seizures and premature death in a mouse model of Dravet syndrome

Generation of a Floxed Na_V1.1 Mouse Line.

Homologous recombination was used to replace exon 25 of the endogenous Scn1a gene using a targeting vector containing the exon flanked by LoxP sites and a neomycin-selection cassette, which itself was flanked by flippase recognition target (FRT) sites (Fig. 1A). Before breeding animals for experiments, the neomycin-selection cassette was removed to avoid possible impairment of Scn1a function by mating a heterozygous flox mouse with a flippase-expressing mouse. This genetic cross excises the neomycin-selection cassette and leaves only a single FRT site (Fig. 1A). Flox heterozygous (F/+) and heterozygous mice maintained on a C57BL/6 background are indistinguishable from WT littermates and breed, thrive, and survive normally. Breeding with a globally deleting Meox2-Cre strain (Jackson Labs) excised exon 25 and resulted in the expected truncated PCR product (Fig. 1B) in DNA samples from F/+:Meox2-Cre⁺ and F/F:Meox2-Cre⁺ mice at postnatal day (P) 14. Immunoblotting of protein samples from these mice showed a substantial reduction in Na_V1.1 protein levels in F/+:Meox2-Cre⁺ mice and no detectable Na_V1.1 protein in F/F:Meox2-Cre⁺ mice at P14 (Fig. 1C). Immunocytochemical studies with specific antibodies against Na_V1.1 protein revealed a visually evident reduction in staining of the cell bodies of GABAergic inhibitory neurons in layers IV and VI in somatosensory cortex, which contain primarily interneurons, whereas there was no evident decrease in immunostaining of neuronal cell bodies in layer V, where the cell bodies of excitatory pyramidal neurons are located (Fig. 1D). Further quantitative analysis of the selective deletion of Na_V1.1 channels is presented below.

Specific Deletion of Na_V1.1 Protein in GABAergic Neurons in Dlx-I12b-Cre Mice.

The Dlx gene family encodes homeobox transcription factors that are essential for specification, maturation, and survival of interneuron progenitors (20). Mating with the Dlx1/2-I12b-Cre line (Dlx-Cre) (23) mouse targets Cre recombinase expression to forebrain interneurons, and deletion of Dlx1 leads to interneuron loss and seizures (22). Our C57BL/6 F/F mice were crossed with Dlx1/2-I12b-Cre⁺ (Dlx-Cre⁺) mice to produce heterozygous loss of Na_V1.1 channels in interneurons of the cerebral cortex and hippocampus on a C57BL/6:CD1 background. This genetic cross was consistent through all experiments and all comparisons were made between Dlx-Cre⁻ and Dlx-Cre⁺ littermates.

Double labeling with antibodies recognizing Na_V1.1 and GABA revealed a specific decrease in immunostaining for Na_V1.1 in GABA-expressing inhibitory interneurons, as illustrated in the dentate gyrus from F/F:Dlx-Cre⁺ and F/F:Dlx-Cre⁻ mice (Fig. 2 A–F). The arrowheads indicate individual examples of neurons that are double labeled in F/F:Dlx-Cre⁻ mice (Fig. 2 A–C) but only labeled with GABA in F/F:Dlx-Cre⁺ mice (Fig. 2 D–F). Arrows indicate clusters of two to four neurons that are double-labeled in F/F:Dlx-Cre⁻ mice but not in F/F:Dlx-Cre⁺ mice (Fig. 2 A–F). Similar images were obtained for the visual and somatosensory cortex (Fig. S1). Quantification of this double-labeled immunostaining showed a significant reduction in the fraction of Na_V1.1/GABA double-labeled interneurons in F/F:Dlx-Cre⁺ brain sections compared with F/F:Dlx-Cre⁻ controls across several regions of the cerebral cortex and hippocampus (Fig. 2G). As expected, F/F:Dlx-Cre⁺ brain slices retained normal numbers of Na_V1.1-expressing layer V pyramidal cells in visual cortex and somatosensory cortex at P14 compared with F/F:Dlx-Cre⁻ slices (Fig. 2H). GABA⁺ neurons do not all contain Na_V1.1, and Cre expression is not expected to result in complete loss of Na_V1.1 expression in all cells because of differences in timing and extent of activation by the Dlx enhancer. We determined the average pixel intensity of Na_V1.1 staining in GABA⁺ and GABA⁻ staining cells and observed that the intensity of immunostaining of Na_V1.1 in remaining Na_V1.1/GABA double-labeled interneurons in F/F:Dlx-Cre⁺ mice was reduced compared with F/F:Dlx-Cre⁻ littermates, but no change in Na_V1.1 staining intensity was observed in Na_V1.1⁺, GABA⁻, excitatory pyramidal cells in layer V of the cerebral cortex, or in dentate granule cells (Fig. 2I). These results demonstrate selective deletion of Na_V1.1 channels in GABAergic interneurons in the hippocampus and cerebral cortex.

Heterozygous Conditional Deletion of Na_V1.1 Leads to Premature Death Following Spontaneous Seizures.

F/+:Dlx-Cre⁺ mice died prematurely compared with F/+:Dlx-Cre⁻ littermates, with initial deaths beginning late in the third postnatal week and only 30% of animals remaining at 90 d (Fig. 3A). Video recordings in a separate population of animals captured spontaneous seizures in approximately one-third of the F/+:Dlx-Cre⁺ mice, which were scored 1–5 using the Racine scale (n = 10/35) (24). Seizures and death in these animals followed a stereotyped progression. Seizures began with forelimb clonus (Racine 3) and progressed in both frequency and severity through forelimb clonus with rearing (Racine 4), eventually culminating in generalized tonic-clonic seizure (GTC, Racine 5), and finally death 4–48 h after initial seizure onset. This time course of seizures and death is illustrated for a typical animal in Fig. 3B. Age at death ranged from P18 to P22 and all deaths were observed immediately following a Racine 5 GTC seizure (Fig. 3 B and C). Animals died an average of 17.9 h after seizure onset, during which time they experienced an average of nine seizures (Fig. 3 D and E). This timing of death in the population of mice monitored for spontaneous seizures correlated with the time of highest incidence of death in the total population, suggesting seizures precede death in the general population. Throughout this time period, no F/+:Dlx-Cre⁻ animals were observed to have seizures or die (n = 28). These results show that the premature death observed in F/+:Dlx-Cre⁺ mice is comparable to that in Na_V1.1 global heterozygotes (14) and therefore provide support for the conclusion that heterozygous deletion of Na_V1.1 channels in forebrain GABAergic inhibitory neurons is sufficient to cause premature death in the mouse model of DS.

Heterozygous Conditional Deletion of Na_V1.1 Leads to Thermal Sensitivity to Seizure.

Patients with DS (13) and our mouse model of this disease (15) are susceptible to seizures evoked by elevation of body temperature. When the core body temperature of our F/+:Dlx-Cre⁺ animals was increased, seizures were induced in all animals tested with mean temperature for seizure induction of 39.6 °C (Fig. 4A) (F/+:Dlx-Cre⁺, n = 17). Conversely, no F/+:Dlx-Cre⁻ animals exhibited thermally induced seizures (Fig. 4A; F/+:Dlx-Cre⁻, n = 10). Seizures provoked by thermal induction ranged in severity from simple forelimb clonus (Racine 3) to full GTC (Racine 5), but the majority of animals had Racine 4 seizures (Fig. 4B). The two animals that progressed to GTC died soon thereafter. These results show that the sensitivity to thermal induction of seizures in our F/+:Dlx-Cre⁺ is indistinguishable from global Na_V1.1 heterozygotes, supporting the conclusion that heterozygous deletion of Na_V1.1 channels in forebrain GABAergic inhibitory neurons is sufficient to cause the susceptibility to thermally induced seizures observed in our mouse model of DS.

A subset of animals was implanted with four EEG leads at P24, and EEGs were recorded during thermal induction on the following day after the mice recovered from surgery and were free from the effects of anesthesia. In each F/+:Dlx-Cre⁺ animal tested, electrographic seizures and behavioral correlates were closely linked (F/+:Dlx-Cre⁺, n = 4) (Fig. 4C). In contrast, none of the F/+:Dlx-Cre⁻ mice experienced either electrographic seizures or behavioral correlates (F/+:Dlx-Cre⁻, n = 5) (Fig. 4C). The pattern of epileptiform electrical activity preceding and during the seizures induced in F/+:Dlx-Cre⁺ mice was similar to those characterized previously in global Na_V1.1 heterozygotes (14, 15) and in children with DS (25).

Lack of Effect of Heterozygous Conditional Deletion of Na_V1.1 in the Heart.

Na_V1.1 channels are expressed in the mouse heart (26); therefore, it is possible that a direct effect of our gene deletion on cardiac function could contribute to seizures or premature death. In an extensive study using sensitive detection methods, Potter et al. demonstrated that Dlx-I12b-Cre expression is restricted to the forebrain and no expression was observed in the peripheral nervous system or heart (23). Our findings that deletion of Na_V1.1 channels is specific to inhibitory neurons and not excitatory neurons further demonstrates the specificity of this gene deletion and indicates that no leaky Cre expression has occurred in our F/+:Dlx-Cre⁺ mice. Consistent with these studies, ECG recordings at rest revealed no significant differences between F/+:Dlx-Cre⁻ (n = 5) and F/+:Dlx-Cre⁺ (n = 6) animals in heart rate (F/+:Dlx-Cre⁻, 672.2 ± 24; F/+:Dlx-Cre⁺, 658 ± 57), PR interval (F/+:Dlx-Cre⁻, 33 ± 2; F/+:Dlx-Cre⁺, 32 ± 2), QRS interval (F/+:Dlx-Cre⁻, 9.4 ± 1; F/+:Dlx-Cre⁺, 8.3 ± 1), or QT interval (F/+:Dlx-Cre⁻, 23 ± 5; F/+:Dlx-Cre⁺, 20 ± 1.5) (Fig. 4D, P > 0.05). These results demonstrate that selective deletion of Na_V1.1 channels in forebrain neurons has no effect on intrinsic action potential generation and conduction in the heart and therefore support the conclusion that premature deaths in our conditional deletion mouse model of DS are initiated by seizures.

Generation and Maintenance of Floxed Mouse Line.

All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Washington, Seattle. Exon 25, the second-to-last coding exon of Scn1a, was cloned into plasmid 4517D (a gift from G. Stanley McKnight, University of Washington, Seattle) containing a single LoxP site, FRT-flanked neomycin-selection cassette, and multiple cloning sites. The NheI/SalI sites were used to insert a 6-kb long arm containing exon 26 (DIV S3-CT) and the XhoI/NotI sites were used to insert a 1.5-kb short arm containing exon 25 (DIV S1-S3) and second LoxP site. The plasmid was linearized at the AscI site, electroporated into embryonic stem cells, and used to generate chimeric mice. Animals were genotyped using FHY311 (5′-CTTGATGTGTTGAAATTCAC-3′) and FHY314 (5′-TATAGAGTGTTTAATCTCAAC-3′): WT allele, 846 bp; floxed allele, 1019 bp; and excised allele, 258 bp. The neomycin-selection cassette was removed following mating with a flippase-expressing mouse (B6.SJL-Tg(ACTFLPe)9205Dym/J; Jackson Labs), and the resulting neomycin-excised animals were backcrossed with C57BL/6J (Jackson Labs) for 10 generations. Dlx-I12b-Cre mice (obtained from John L. Rubenstein, University of California, San Francisco) were received and maintained on a CD1 (Jackson Labs) background. Mouse lines were maintained independently and mated together to generate F₁, F/+:Dlx-Cre⁺ and F/+:Dlx-Cre⁻ littermates. All experimental comparisons were made between F₁ littermates.

Immunohistochemistry.

Animals were perfused at P14 with 4% (wt/vol) paraformaldehyde in PB, brains were removed and saturated in 30% sucrose, sliced to 50 μm, and double labeled as free-floating slices (30). Primary antibodies were affinity-purified rabbit anti-Na_V1.1 (1:150, SP11A) (3) and guinea pig anti-GABA (1:600; Abcam). Secondary antibodies were goat anti–guinea-pig IgG labeled with Alexa 555 (1:400; Invitrogen), goat anti-rabbit IgG labeled with Alexa 488 (1:400; Invitrogen). Tissue samples from F/F:Dlx-Cre⁻, F/F:Dlx-Cre⁺ and global Scn1a knockout animals were processed simultaneously. Gain- and offset-matched images were collected on a Leica SL confocal microscope at the Keck Imaging Facility of the University of Washington. Sections stained in the absence of primary antibody showed no detectable labeling, similar to the global Na_V1.1-knockout tissue stained with Na_V1.1 antibodies.

Spontaneous Seizure Recording and Analysis.

Animals were continuously video monitored from P19 to P27. The resulting video files were reviewed at ∼8× speed. Suspected seizure events were reviewed at 2× speed and scored from 1 to 5 for seizure severity on the basis of the Racine scoring system: 1, mouth and facial movements; 2, head nodding; 3, forelimb clonus, usually one limb; 4, forelimb clonus with rearing; and 5, generalized tonic-clonic seizure (GTC), rearing, clonus, and falling over (24).

Thermal Induction and EEG.

Each animal’s core body temperature was continuously monitored by a rectal temperature probe and controlled by a feedback circuit in line with a heat lamp. Body temperature was increased in 0.5 °C steps at 2-min intervals until a seizure occurred or a temperature of 42 °C was reached. The animal was then cooled and returned to its home cage. Thermal induction was performed on P22, and the induction process and resulting seizures were recorded by video monitoring. Seizures were scored on the Racine scale using the same criteria as in spontaneous seizure monitoring. EEG surgeries were performed at P23 with the animals under ketamine/xylazine anesthesia and aseptic conditions. Animals were implanted with four to six small platinum wire electrodes and allowed to recover for 24 h before referential EEG recording during thermal induction (15).

Electrocardiogram Recordings.

Surgeries to implant ECG electrodes were performed under ketamine/xylazine (130/8.8 mg/kg) anesthesia and aseptic conditions. A small midline skin incision was made above the skull and a second incision above the thorax. Two silver electrodes were tunneled s.c. from the head incision to the thoracic incision site and anchored in place with suture. One electrode was positioned near the heart apex and the other near the right forelimb. Mice were allowed to recover from surgery overnight. On experiment day, mice were transferred to the recording chamber and allowed to acclimate for 30 min before recording. ECG recordings were collected in conscious mice in daytime on a PowerLab 35 data acquisition unit using LabChart software (ADInstruments), at 1-kHz sampling rate and processed off-line with a 3-Hz highpass filter.