Mice with a Targeted Disruption of the Cl⁻/HCO₃⁻ Exchanger AE3 Display a Reduced Seizure Threshold

Abstract

Neuronal activity results in significant pH shifts in neurons, glia, and interstitial space. Several transport mechanisms are involved in the fine-tuning and regulation of extra- and intracellular pH. The sodium-independent electroneutral anion exchangers (AEs) exchange intracellular bicarbonate for extracellular chloride and thereby lower the intracellular pH. Recently, a significant association was found with the variant Ala867Asp of the anion exchanger AE3, which is predominantly expressed in brain and heart, in a large cohort of patients with idiopathic generalized epilepsy. To analyze a possible involvement of AE3 dysfunction in the pathogenesis of seizures, we generated an AE3-knockout mouse model by targeted disruption of Slc4a3. AE3-knockout mice were apparently healthy, and neither displayed gross histological and behavioral abnormalities nor spontaneous seizures or spike wave complexes in electrocorticograms. However, the seizure threshold of AE3-knockout mice exposed to bicuculline, pentylenetetrazole, or pilocarpine was reduced, and seizure-induced mortality was significantly increased compared to wild-type littermates. In the pyramidal cell layer of the hippocampal CA3 region, where AE3 is strongly expressed, disruption of AE3 abolished sodium-independent chloride-bicarbonate exchange. These findings strongly support the hypothesis that AE3 modulates seizure susceptibility and, therefore, are of significance for understanding the role of intracellular pH in epilepsy.

Deviations from the normal intracellular pH (pH_i) influence diverse neuronal functions (7). Many of these effects reflect an interaction of pH_i with cellular proteins, such as ion channels, ion transporters, and signal transduction pathways. A fine-tuning of pH_i is thus crucial for normal cell function. On the other hand, neuronal activity itself causes significant pH shifts in neurons (8). Several transport mechanisms contribute to keep the pH_i at a constant level. Generally, pH_i regulation is grouped into acid extrusion and acid loading. Acid extrusion is mainly accomplished by sodium/proton exchange, sodium-dependent chloride/bicarbonate exchange, and sodium/bicarbonate cotransport (7). Sodium-independent chloride/bicarbonate exchange mediates acid loading and is conducted by the four members of the anion exchanger family (26), namely, AE1 (encoded by SLC4A1), AE2 (SLC4A2), AE3 (SLC4A3), and AE4 (SLC4A9). SLC4A1 encodes the major intrinsic membrane protein of the erythrocyte and the basolateral Cl⁻/HCO₃⁻ exchanger of the acid-secreting type A intercalated cells of the kidney. Mutations of SLC4A1 can either cause the human red blood cell disorder hereditary spherocytosis or distal renal tubular acidosis (1). AE2 is broadly expressed in various tissues, including many epithelia (26). An AE2-knockout mouse displayed growth abnormalities, early lethality, and defective gastrointestinal pH regulation (12), whereas testicular dysfunction was found in another AE2-knockout mouse (21). The most recently identified family member, AE4 (34), is expressed mainly in intercalated cells of the kidney and some other tissues in a species-specific manner (16, 37). The physiological role of AE4 is unknown, and it is yet unclear whether AE4 functions in a sodium-independent manner (26). The expression of AE3 is unique among the various AEs, as it is predominantly expressed in the brain and heart (17). The SLC4A3 gene employs two different promoters to generate the brain variant bAE3 and the cardiac variant cAE3, which has a shorter amino-terminal amino acid sequence (20). In neurons bAE3 is thought to be important for the removal of excessive intracellular HCO₃⁻ (17) and hence regulation of intracellular pH_i. Recently, mapping of a susceptibility locus for common idiopathic generalized epilepsy identified the chromosomal region 2q36 (27) that includes SLC4A3, the gene encoding the sodium-independent Cl⁻/HCO₃⁻ exchanger AE3. Subsequently, a common polymorphism within the coding sequence of SLC4A3, which entails the amino acid exchange Ala867Asp, has been shown to be associated with idiopathic generalized epilepsy (28). It is still unclear whether the above-mentioned polymorphism 867Asp itself confers the increased risk for epileptic seizures or another gene in close proximity of SLC4A3 is involved.

To understand the role of AE3 in vivo and to address the putative relevance of AE3 for seizure susceptibility, we generated a knockout mouse model for Slc4a3. AE3-deficient mice appeared healthy, were fertile, and did not show obvious behavioral abnormalities. Though disruption of AE3 did not result in overt epilepsy or neurodegeneration, AE3-knockout mice were more sensitive to seizure-inducing agents. While withdrawal of extracellular chloride resulted in a strong alkalotic shift in the CA3 region in hippocampal slices from wild-type mice, this shift was absent in AE3-knockout hippocampi. This impairment of pH regulation may relate to neuronal hyperexcitability in AE3-knockout mice. These findings support the hypothesis that AE3 is a susceptibility gene for epilepsy.

MATERIALS AND METHODS

Experiments were approved by the Ministry of Science and Public Health of the City State of Hamburg, Germany. All experimental procedures complied with the regulations of the U.S. National Institutes of Health and with those of the Society of Neuroscience (USA).

Generation of AE3-knockout mice.

A clone isolated from a 129/SvJ mouse genomic λ library (Stratagene) was used to construct the targeting vector. A 12.2-kb XmaIII/XmnI fragment including all exons of the Slc4a3 gene was cloned into the pKO-V901 plasmid (Lexicon Genetics) with a phosphoglycerate kinase (pgk) promoter-driven diphtheria toxin A cassette. A pgk promoter-driven neomycin resistance cassette flanked by loxP sites was inserted into the Tth111I-site in intron 5. A third loxP site and an additional EcoRV site were inserted into the EcoRI site in intron 19. The construct was electroporated into R1 mouse embryonic stem cells. Neomycin-resistant clones were analyzed by Southern blotting using EcoRV and an external ∼500-bp probe. Correctly targeted embryonic stem cells were transiently transfected by electroporation with a plasmid expressing cre recombinase (pPGKcrebpA; a generous gift from K. Rajewsky) to remove the neomycin cassette and exons 6 to 19 of the Slc4a3 gene. Two independent embryonic stem cell clones were injected into C57BL6 blastocysts to generate chimeras that were backcrossed with C57BL6. Studies were performed in a mixed 129SV/C57BL6 background in the F2 and F3 generations. We used littermates as controls. For genotyping, genomic DNA from tail biopsies was prepared by proteinase K (Roche) digestion and subsequent purification with phenol-chloroform. Mice were genotyped either by Southern blotting as above or by PCR. For PCR genotyping, the sense primer F1 (GCCACCAGGGGAATGACAAGCCCG) and the antisense primers R1 (CTGGAGACCTGGGGGTTGGGCTAA) and R2 (TCTCTAGACACCTAGCTCCCAACA) were used in a single PCR mix. F1 started at position g.2735, R1 at position g.10211, and R2 at position g.3311, the numbers referring to the position relative to the initiation codon ATG of the Slc4a3 gene. The primer pair F1/R1 amplified an 810-bp knockout allele, and the primer pair F1/R2 amplified a 384-bp wild-type allele.

Northern analysis and in situ hybridization.

Total RNA was isolated from various murine tissues using the RNeasy kit (QIAGEN) according to the manufacturer's instructions. Ten micrograms of total RNA was separated by electrophoresis and blotted following standard protocols. A multiple tissue Northern blot was hybridized with a random-primed [α-³²P]dATP-labeled full-length cDNA probe of AE3 (accession number gi:6678022). For expression analysis during embryonic development, a commercial poly(A)⁺ blot from whole mouse embryos was used (Clontech). The following mouse probes were used for differential expression analysis: gi:1330598 (Slc4a1, nucleotides 3810 to 4294), gi:6678020 (Slc4a2, nucleotides 183 to 2191), gi:200722358 (Slc4a4, full length); gi:63660353 (Slc4a7, full length), gi:10946959 (Slc4a8, full length), gi:27370243 (Slc4a9, full length), gi:24660267 (Slc4a10, full length), gi:63492386 (Slc4a11, full length), gi:1357164 (Slc9a1, nucleotides 3790 to 4126), gi:26097164 (Slc9a2, nucleotides 1740 to 2252), gi:51767973 (Slc9a3, nucleotides 1620 to 1900), gi:26351734 (Slc9a4, nucleotides 1225 to 1794), gi:63578911 (Slc9a5, nucleotides 44 to 978), gi:21410135 (Slc9a8, full length), gi:6677992 (Slc12a4, full length), gi:33859679 (Slc12a5, full length), gi:19526768 (Slc12a6, full length), gi:6755533 (Slc12a7, full length), gi:6677990 (Slc12a2, full length), gi:28628356 (Slc26a11, nucleotides 123 to 705). The numbers refer to the position relative to the initiation codon ATG of the respective cDNA. In situ hybridization was essentially performed as described elsewhere (14) using a probe spanning nucleotides 2090 to 2515 of the AE3 cDNA (accession number gi:6678022) labeled with [α-³⁵S]UTP. The day of plug was not counted for specification of embryonic stages.

Western analysis.

The AE3 antiserum was raised in rabbits against the amino-terminal epitope WPHDPDAKEKPLHM (amino acids 478 to 491 of the bAE3 isoform) coupled via N-terminal cysteine to a keyhole limpet hemocyanin carrier. The peptide CWPHDPDAKEKPLHM was coupled to Sulfolink-Sepharose (Pierce) following the instructions of the manufacturer. Subsequently, the Sepharose was incubated with an aliquot of the rabbit serum and then washed with TBS (20 mM Tris, pH 7.5, 500 mM NaCl) and phosphate-buffered saline (PBS; 136.5 mM NaCl, 2.65 mM KCl, 8.3 mM Na₂HPO₄). Antibodies were eluted with 0.1 M glycin (pH 2.5), and the pH was adjusted to pH 7.0 by adding Tris-HCl, pH 8.5. The AE3 antibody did not work for immunohistological detection of AE3 in tissue sections.

For the preparation of brain and heart membrane fractions, the tissues were homogenized mechanically (Ultra Turrax; IKA-Werke, Germany) in PBS with Complete protease inhibitor cocktail (Roche). Residual debris was pelleted by centrifugation (two times at 1,000 × g), and the supernatant was recovered and subsequently centrifuged at 132,000 × g for 30 min. The resulting membrane-enriched fractions were resuspended, and the protein contents were determined using the bicinchoninic acid method (Pierce).

Protein lysates from astrocyte or mixed hippocampal neuronal cultures were prepared by mechanical homogenization as described above. Triton X was added to a final concentration of 1%, and the lysates were incubated under constant rotation at 4°C for 1 h. Five micrograms of total protein per lane was separated on reducing 7.5% sodium dodecyl sulfate-polyacrylamide gels and then transferred to polyvinylidene difluoride membranes (Amersham Biosciences) using a semidry blotting system. Western blots were either probed with the rabbit AE3 antiserum (1:50) or with mouse anti-GFAP (1:1,000; Chemicon). Detection was done with a chemiluminescence kit (Renaissance; DuPont).

Cell culture.

Mixed hippocampal neuron/glia cultures of P₁ mice were prepared exactly as described previously (5). For astrocyte cultures brains of P₁ mice were dissected into small pieces after removal of the meninges. Subsequently, the tissue was incubated in 1% trypsin (Gibco) and mechanically dissociated. The cells were seeded on tissue-grade culture dishes in Dulbecco's modified Eagle's medium with 10% fetal calf serum (Gibco) and cultured for 14 days.

Morphological analysis.

Adult mice were deeply anesthetized using Ketasyl and perfused transcardially with PBS followed by 4% paraformaldehyde in PBS for 5 min. Organs were removed and postfixed overnight. The organs were subsequently dehydrated in an ascending isopropanol series and paraffin embedded. Ten-micrometer sections were cut, deparaffinized, and hematoxylin and eosin stained using standard procedures.

Behavioral tests.

For the activity assay, the home cage activity of single mice (n = 10 for each genotype) was monitored for 7 days with an infrared motion detector developed in-house with a sampling frequency of 1 Hz and a bin size of 4 min. For the light-dark assay, mice were placed into a brightly lit chamber with a dark area (n = 10 for each genotype). The time spent inside or outside the dark area and the transitions between the bright and dark areas were measured for 10 min using an infrared movement sensor as described above. Experiments were performed 1 h after onset of the dark phase. For the rotarod assay, mice were individually placed on a rotarod (TSE Systems, Germany), and the time spent on the rod was measured for a period of up to 600 s. Mice clinging to the rotating rod without running were classified as failing to cope with the test. Three trials per day were performed on three consecutive days (knockout, n = 7; wild type, n = 8).

Seizure susceptibility.

Pentylenetetrazole (PTZ; Sigma) dissolved in PBS was administered intraperitoneally (i.p.) at two different doses (40 and 60 mg/kg of body weight) in a total volume of 200 and 300 μl, respectively. After injection, the animals were watched closely for 10 min. The latency of the first myoclonic jerk (focal seizure), clonic seizure (clamping of the forefeet), or generalized (tonic/clonic) seizure was measured. Latencies of animals not experiencing seizures were set at 600 s. Bicuculline (Sigma) was dissolved in 0.1 N HCl and then diluted with PBS. Bicuculline was administered intraperitoneally at a dose of 4 mg/kg body weight. Scoring of seizures was done as described for PTZ. Pilocarpine (Sigma) was dissolved in PBS and injected i.p. at a dose of 350 mg/kg body weight. At 30 min before pilocarpine treatment, mice were injected with 1 mg/kg body weight methyl-scopolamine (Sigma) to block peripheral cholinergic actions of pilocarpine. The time until the first generalized seizure was scored.

ECoG recordings.

Experiments were performed on seven wild-type and seven knockout male mice aged between 8 weeks and 6 months. Following anesthesia with pentobarbital (50 mg/kg i.p.; Sigma Chemical Co., St. Louis, MO), the mouse was positioned to a stereotaxic instrument (Kopf Instruments) with bregma and lambda in a horizontal plane. To obtain bilateral epidural electrocorticogram (ECoG) recordings, silver electrodes were positioned over the central region (AP, −1.0 mm; L, 2 mm from bregma) of both hemispheres and fixed on the skull with dental acrylic cement (Paladur; Heraeus Kulzer GmbH, Wehrheim/Ts, Germany). Additionally, reference and ground electrodes were implanted over the nasal and cerebellar regions, respectively. The electrode ensemble was fed through a rubber-like socket and fixed on the skull with dental cement. Xylocaine cream (2%; ASTRA, Germany) was applied to all pressure points and wound edges. Three days after surgery the ECoG was recorded for more than 4 h to get a spectrum of different behavioral states (i.e., wake, sleep) using a swivel connector. EcoG recordings were performed using the Spike2 software package (Cambridge Electronic Design, Cambridge, United Kingdom). Recorded electrical activities were fed through a differential amplifier (EXT-20F or DPA 2F; npi Electronic GmbH, Tamm, Germany) filtered by band-pass filters at 0.3 and 30 Hz, transformed by an A/D interface (CED 1401plus; Cambridge Electronic Design, Cambridge, United Kingdom), and stored online on a personal computer. Additionally, data were stored on a magnetic tape recorder via a neuro-corder (DR-890; NeuroData, Instrument Corp., Delaware Water Gap, PA) for offline analysis. Offline analysis of fast Fourier transformation (FFT) was conducted using the Spike2 software package (Cambridge Electronic Design Limited, Cambridge, United Kingdom). The FFT spectrum between 1 and 30 Hz was averaged to 1-Hz bins for a 20-s period of each behavioral stage (wake, sleep).

pH_i measurements.

To test whether lack of AE3 affects pH_i regulation of adult neurons, optical recordings of the CA3 region of hippocampal slices were performed. Synaptic connections between hippocampal neurons are preserved in this type of slice preparation. In brief, the spontaneous activity of CA3 pyramidal neurons can be modulated by the excitatory inputs from neurons within the dentate gyrus and inhibitory GABA-ergic inputs from interneurons. We chose the stratum pyramidale, since in this area neuronal somata are tightly packed (Fig. 4C) and the volume fraction of glial cells amounts to only about 6% (10). Hippocampi were excised from isofluran-anesthetized 3-month-old AE3-knockout or wild-type mice. Slices (400 to 500 μm thick) were cut with a guided razor blade and preincubated in a CO₂/HCO₃⁻-buffered modified artificial cerebrospinal fluid (mACSF) containing the following (in mM): NaCl (124), KCl (3), CaCl₂ (0.75), MgSO₄ (1.3), KH₂PO₄ (1.25), NaHCO₃ (26), and glucose (10) at 28°C. pH was adjusted to 7.35 to 7.40 by gassing with 5% CO₂, 95% O₂. After a preincubation period of at least 1 to 2 h in the presence of 1 μM tetrodotoxin (Sigma, Germany), slices were loaded with 5 μM 2′,7-bis(2-carboxyethyl)-5(6)-carboxyfluorescein-acetoxymethyl ester (BCECF-AM; Molecular Probes, Leiden, The Netherlands) for 10 min in ACSF (as with mACSF, but with 1.75 mM CaCl₂) and transferred to an optical recording chamber (volume, 2 ml) which was mounted on the stage of an upright Olympus Bx50Wi microscope. Slices were superfused with ACSF (4.5 ml/min) for at least 30 min before optical measurements were started. Temperature was kept at 32 ± 1°C. To obtain a nominally Cl⁻-free superfusate, NaCl, KCl, and CaCl₂ of the ACSF were isoosmotically replaced by the respective gluconate salts as described elsewhere (6). Substitution of chloride by gluconate results in a lowering of free Ca²⁺ from 1.75 mM to about 0.33 mM. We conducted control experiments shown in Fig. 3F to exclude an interference of this lowering of Ca²⁺ with the alkalinization upon withdrawal of external Cl⁻. A lowering of the Ca²⁺ concentration did not result in a change of the baseline pH_i or affect the alkalinization upon Cl⁻ withdrawal (Fig. 3F). Removal of extracellular chloride results in the reversal of chloride currents through GABA_A receptor channels, depolarization, and enhanced bioelectric activity (38), which causes intracellular neuronal acidification (8). Hence, tetrodotoxin was added during preincubation, which irreversibly suppressed activity-dependent shifts of pH_i upon removal of extracellular chloride. 4,4′-diisothiocyanato-stilbene-2,2′-disulfonic acid (DIDS; Sigma) was added freshly to ACSF to a final concentration of 200 μM.

FIG. 4.

Histological analysis of AE3-knockout mouse brains. Ten-micrometer-thick horizontal (A and B) or coronal (C to N) sections from different brain regions of 10-month-old mice stained with hematoxylin-eosin are shown. (A and B) Overview. Bar, 1.5 mm. (C and D) Hippocampus. Bar, 200 μm. (E and F) Temporal cortex. Bar, 200 μm. (G and H) Cerebellum: Purkinje cells and granular cell layer. Bar, 100 μm. (I and J) Pontine nuclei. Bar, 50 μm. (K and L) Higher magnification of CA1 pyramidal cells. Bar, 40 μm. (M and N) Deep cerebellar nuclei. Bar, 50 μm. (O and P) Retinae. Bar, 50 μm. No obvious abnormalities were observed.

FIG. 3.

Effect of the disruption of AE3 on Cl⁻/HCO₃⁻ exchange in adult hippocampal CA3 pyramidal cells. (A to C) Mean changes in steady-state pH_i upon chloride withdrawal (bars) are shown for a population of ∼20 neuronal somata within hippocampal slices from wild-type (A and B [+/+]) and AE3-knockout (C [−/−]) mice. Upward deflection represents an increase of pH_i. The alkalotic pH_i shift in wild-type animals (A) was lost upon pretreatment with 0.2 mM DIDS (B). Cl⁻ withdrawal was without effect on pH_i in knockout preparations (C). (D) Mean (±SEM) steady-state pH_i before (black columns) and 10 min after (open columns) Cl⁻ withdrawal measured in wild-type (+/+) and AE3-knockout (−/−) neurons. Cl⁻ concentration was either 130 mM (+Cl⁻) or 0 mM (-Cl⁻) due to equimolar replacement by gluconate. (E) Addition of 0.2 mM DIDS did not change pH_i in wild-type or knockout animals. (F) Ca²⁺ lowering due to gluconate substitution for Cl⁻ did not affect steady-state pH_i or alkalinization upon Cl⁻ withdrawal. The level of significance was tested by Student's t test for paired samples (the asterisk in panel D marks statistical significance; n.s., nonsignificant).

Slices were viewed with a 20× water immersion objective (Olympus; numerical aperture, 0.5) and illuminated with alternating light (440 nm and 490 nm) provided by a 100-W halogen lamp and a computer-operated filter wheel (Sutter Instruments) connected to the microscope by an optical fiber as described previously (18). The 440-nm light path was equipped with a neutral density filter to obtain a BCECF fluorescence ratio for excitation at 440 nm/490 nm of 1.0 at pH 7.0. Loss of fluorescence intensity at excitation with 440 nm was <0.5%/min, indicating that structures under investigation were in good condition. Fluorescence image pairs were captured every 20 s by a charge-coupled device camera (PTI; Surbiton, Surrey, England). A region of interest (area of ∼3,000 to 5,000 μm²) was defined, which was adapted to the curvature formed by the layer of tightly packed neuronal somata. This area comprised fluorescence signals from up to 40 neurons. Background fluorescence was determined from hippocampal slices not loaded with BCECF-AM using identical camera and illumination settings. Before calculating the ratio between emissions after excitation at 440 nm and 490 nm, background fluorescence was subtracted. Acquisition and processing of images were performed with a CARAT system (O. Ahrens, Bargteheide, Germany). Ratios were converted into pH_i values by means of a standard curve obtained by the in vitro calibration method (4) adapted to water immersion optics (3). Changes of pH_i values (mean ± standard error of the mean [SEM]) upon Cl⁻ withdrawal were evaluated by measuring the difference of the mean pH_i immediately before and at the end of Cl⁻-free conditions. Student's t test for paired samples was used with a P value of ≤0.05 considered significant. Calculations were carried out with the Excel 5.0 software.

RESULTS

AE3 expression is established early.

To analyze the expression pattern of AE3 during development and in the adult mouse, a Northern blot analysis was performed. Already at embryonic day 11 (E11), we detected AE3 transcripts of two different sizes, the upper presumably representing the neuronal and the lower the cardiac-specific isoform transcripts. For both transcript sizes, signal intensity increased towards E17 (Fig. 1A). In the adult mouse, Northern blot analysis of different tissues confirmed a predominant expression of Slc4a3 in brain, heart, and retina (Fig. 1C). The band corresponding to AE3 transcripts in the heart was smaller compared to brain and retina, as expected. In addition, AE3 transcripts were identified in adrenal glands. Larger bands were observed for the retina and the adrenal gland, which may be yet-uncharacterized splice variants. In situ hybridization on sagittal sections of embryos at different stages of development demonstrated a widespread expression of AE3 in the central nervous system already at E12 which paralleled neuronal differentiation and increased in intensity towards E18 (Fig. 1B). In the adult mouse brain, transcripts of AE3 were detected in nearly every neuronal population, with particularly high expression levels in the hippocampus and the cortex, while fiber tracts were devoid of signals (Fig. 1D). In Western blot assays, AE3 was detected in primary mixed neuronal cultures from P₁ mice but not in extracts from cultured astrocytes (Fig. 1E). These data confirm a predominant neuronal and cardiac expression of AE3 already during embryonic development.

FIG. 1.

Transcript distribution of Slc4a3. (A) Hybridization of a commercial poly(A)⁺ RNA filter of whole embryos at different stages with an Slc4a3-specific probe revealed expression as early as E11. (B) Autoradiograms of sagittal sections of E12, E14, E16, and E18 mouse embryos hybridized with a [³²P]UTP-labeled RNA probe corresponding to that used for panel A demonstrated that a neuronal expression pattern is established early. The ventral surface is to the left. The section of the E18 embryo is lateral to the spinal cord. (C) In a multiple-tissue Northern blot assay from adult animal organs hybridized with the same probe as in panel A, a prominent band was observed in brain, heart, adrenal gland, and retina. Hybridization with an actin probe served as a loading control (not shown). The cardiac transcript is smaller than the brain transcript. In adrenal gland and retina an additional transcript of ∼6 kb was observed. (D) Dark-field photomicrographs of emulsion-dipped coronal sections of an adult mouse brain hybridized with an AE3-specific probe showed a broad neuronal expression pattern without labeling of fiber tracts and a particularly strong labeling in the stratum pyramidale of the hippocampal formation (signals appear white). (E) In Western blot assays, a strong signal for AE3 was detected in protein lysates of mixed glia/neuronal cultures but not of cultured astrocytes. cb, cerebellum; h, heart; m, mesencephalon; mo, medulla oblongata; sc, spinal cord; t, telencephalon.

Disruption of the Slc4a3 gene.

Our targeting strategy resulted in the deletion of exons 6 to 19, which code for transmembrane domains 1 to 7 of the predicted 12 transmembrane domains of AE3 (Fig. 2A). This leads to a frameshift and a premature stop codon. Matings of heterozygous animals resulted in homozygous offspring in the expected Mendelian ratio (∼25%). Genotypes were either determined by PCR (Fig. 2B) or by Southern blot analysis using an additional EcoRV site in the targeted allele (Fig. 2C). In adult AE3-knockout mice, no full-length or truncated AE3 transcripts were detected by Northern analysis using a full-length cDNA probe (Fig. 2D). AE3 protein was absent in lysates from brain and heart of adult AE3-knockout mice as shown by Western blot analysis with an antiserum against AE3 (Fig. 2E).

FIG. 2.

Targeted disruption of the murine Slc4a3 gene. (A) Schematic diagram of the murine Slc4a3 locus. By homologous recombination of the targeting construct, a neomycin resistance cassette flanked by two loxP sites and a third loxP site with an additional EcoRV restriction site was introduced into the Slc4a3 locus. After transient expression of cre recombinase, clones with a deletion including exons 6 to 19 were used for the generation of AE3-deficient mice. (B) Genotyping of tail biopsy DNA from wild-type (+/+), heterozygous (+/−), and knockout (−/−) mice by PCR. (C) Southern blot analysis with a probe located at the 3′ end of Slc4a3 (A) resulted in a ∼4.0-kb band, compared to ∼18.0 kb for the wild-type allele. (D) Northern blot analysis of total RNA from 3-month-old animals with a full-length AE3-cDNA as a probe demonstrated complete absence of AE3 transcripts in knockout brain and heart. (E) Western blot analysis of brain and heart membrane-enriched fractions from 3-month-old animals confirmed this result at the protein level. Note the differing molecular masses of the brain and the heart variant as indicated by the arrows. The bands of lower molecular size represent unspecific cross-reactivity of the antibody.

AE3 is the dominant sodium-independent anion exchanger in the hippocampal CA3 region.

The term hippocampus generally applies to the dentate gyrus, regions CA1 to CA3, and the subiculum. Information flow through the hippocampus proceeds from dentate gyrus to CA3 to CA1 to the subiculum, with additional input information and outputs at each stage. The perforant path, which brings information primarily from the entorhinal cortex, is generally considered the main source of input to the hippocampus. The main output pathways of the hippocampus arise from region CA1 and the subiculum. In hippocampal neurons of wild-type animals, the reversal of the transmembrane chloride gradient causes reverse Cl⁻/HCO₃⁻ exchange and hence an intracellular alkalosis secondary to the influx of bicarbonate (25). To compare Cl⁻/HCO₃⁻ exchange in AE3-knockout and wild-type mice, we measured pH_i changes evoked by withdrawal of extracellular chloride in the stratum pyramidale of hippocampal slices. The mean steady-state pH_i before Cl⁻ withdrawal was 6.95 ± 0.05 (standard deviation [SD]) in wild-type controls and slightly, though not significantly higher, in AE3-knockout preparations (pH_i, 7.00 ± 0.09) (Fig. 3D). As expected, chloride removal resulted in a significant alkalotic shift of 0.18 ± 0.04 (SD) pH units in preparations from wild-type animals (n = 18 from four animals; P < 0.001) (Fig. 3A and D). In AE3-knockout preparations (n = 20 from five animals), alkalosis upon Cl⁻ withdrawal was absent (Fig. 3C and D). Preincubation with 0.2 mM DIDS for 20 min abolished the alkalotic shift upon Cl⁻ removal in wild-type slices (Fig. 3B), indicating that the alkalotic shift is mediated by a DIDS-sensitive anion exchange. The addition of DIDS itself did not result in a significant change of pH_i in wild-type and knockout preparations (Fig. 3E). These data show that AE3 is the dominating Na⁺-independent Cl⁻/HCO₃⁻ exchanger within the pyramidal cell layer of the hippocampal CA3 region.

AE3-knockout mice are viable and appear normal.

AE3-knockout mice were viable, appeared to develop and move normally, and could not be distinguished from heterozygous or wild-type littermates by observation. Histological analysis of brain, retina, heart, adrenal gland, and kidney did not reveal gross structural abnormalities in knockout mice at the age of 3 months, 7 months, or 10 months (Fig. 4A to P and data not shown).

No major behavioral abnormalities were detectable. Activity measurements for a 7-day period did not show substantial differences between the genotypes (Fig. 5A). Activity peaked in the early evening hours in both wild-type and knockout mice (data not shown). Exploration behavior as determined in the light/dark test was not changed in AE3-knockout mice compared to wild-type littermates (Fig. 5B), and the motor coordination performance did not differ between the genotypes in the rotarod test (Fig. 5C).

FIG. 5.

Behavioral analysis of AE3-knockout mice. (A) Overall activity was determined by recording mouse movements for 7 days. Though AE3-knockout mice moved less, this difference did not reach significance (n = 10 per genotype; mean ± SEM). (B) Light/dark test. AE3-knockout and wild-type mice were placed into a brightly lit cage with a dark house. Mice were observed for 10 min, and the positions of the mice were determined every second using an infrared sensor. No obvious difference was observed for the overall time spent in the dark or bright part of the cage between genotypes (n = 10 per genotype; mean ± SEM). (C) Rotarod assay. AE3-knockout (n = 7) and wild-type (n = 8) mice were placed individually on the rotarod three times daily on three consecutive days, and performance was evaluated for 10 min. The numbers at the bottom indicate the individual training sessions. No major difference could be observed.

AE3-knockout mice have a strongly reduced seizure threshold.

Though no spontaneous seizures were observed in AE3-knockout mice, we tested whether AE3-knockout mice are more prone to seizures by challenging them with the proconvulsant substances bicuculline, PTZ, and pilocarpine. In comparison to their wild-type littermates, latencies until onset of seizures were strongly reduced in AE3-knockout mice for each of the three seizure phases scored (myoclonic jerk and clonic and generalized seizure) after intraperitoneal application of 4 mg/kg body weight of bicuculline in 10 animals (Fig. 6A). Likewise, the time until the first generalized seizure after i.p. application of 350 mg/kg body weight pilocarpine was nearly halved in knockout mice from a mean latency of 2,082 s (±758 s [SD]) in wild-type animals to 1,119 s (±436 s [SD]) (Fig. 6B; n = 8 for each genotype). PTZ was used as a third pharmacological paradigm to test for seizure susceptibility. With a dose of 40 mg/kg body weight injected i.p., knockout mice displayed reduced latencies until the onset of all three seizure phases (Fig. 6C; n = 15 for each genotype). While 100% of the knockout mice (n = 15) exhibited initial focal seizures (phase 1) after application of 40 mg/kg PTZ and as much as 86% of them subsequently progressed to generalized seizures (phase 3), only 6 of 15 (40%) wild-type mice experienced focal seizures and 27% had generalized seizures (Fig. 6D). With a PTZ dose of 60 mg/kg body weight most animals of both genotypes exhibited all three phases of seizures, but latency until onset of each seizure phase was again shorter in knockout animals compared to wild-type mice (Fig. 6E; n = 16 for each genotype) and the outcome was fatal for 69% (11 out of 16) of the knockout mice compared to 13% (2 out of 16) of the controls (Fig. 6F).

FIG. 6.

Reduced seizure threshold of AE3-knockout mice. (A) Adult AE3-knockout mice have shortened seizure latencies after i.p. injection of bicuculline (4 mg/kg body weight). The observation period lasted 600 s (n = 10 per genotype). (B) Latencies until onset of generalized seizures after i.p. injection of pilocarpine (350 mg/kg body weight) in eight animals for each phenotype. The observation period was 1 h (3,600 s). (C) Seizure latencies after i.p. injection of PTZ (40 mg/kg body weight; n = 15 per genotype). (D) Percentage of animals exhibiting the different seizure stages after injection of PTZ (40 mg/kg body weight). (E) Seizure latencies after i.p. injection of PTZ (60 mg/kg body weight; n = 16 per genotype). (F) The mortality after i.p. injection of PTZ (60 mg/kg body weight) was drastically increased in AE3-knockout mice. Shown are mean values ± SEM. MJ, myoclonic jerk; CL, clonic seizure; GEN, generalized seizure.

The ECoG of freely moving AE3-knockout mice was compared to wild-type littermates. AE3-knockout mice as well as wild-type littermates expressed regular field potential patterns on the ECoG without any epilepsy-like abnormalities like spike wave complexes or high-voltage spike activities (Fig. 7A and B). The ECoG of both genotypes showed similar patterns during different behavioral states (e.g., wake, sleep). During wake state, both groups showed predominant frequencies at a theta frequency range of about 4 to 10 Hz; during sleep state the ECoG of animals of both genotypes had a low-frequency activity at 2 to 3 Hz. The FFT power spectra did not show significant differences between the genotypes in the wake or the sleep state (data not shown).

FIG. 7.

Electrocorticogram analysis. Neuronal activity in the left hemisphere of a freely moving wild-type (A) and AE3-knockout mouse (B) is shown. Representative recordings from one of seven adult animals per genotype are shown. Original traces of field potential recordings during wake and sleep states did not reveal sharp waves or spike wave-like activity in AE3-knockout mice.

Thus, though AE3-knockout mice are more sensitive to seizure-inducing agents, they do not display an obvious epileptic phenotype or seizure-like events in EcoG recordings.

Comparative expression profiling in AE3-knockout brains.

AE3-knockout mice exhibit a surprisingly mild phenotype. As this might be a consequence of a compensatory up-regulation of other transporters involved in the regulation of pH_i, the expression of members of the Slc4a, Slc26a, and Nhe (Slc9a) gene family in the brain was analyzed by Northern blotting. As AE3 also modulates the intracellular chloride concentration, our analysis included members of the cation chloride cotransporter (Slc12a) gene family. We did not detect any significant differences in expression levels of any of the genes analyzed in AE3-knockout brains (Fig. 8).

FIG. 8.

Expression of genes involved in Cl⁻ transport or pH regulation. Members of the Slc4a, the Kcc, the Slc9a, and the Slc26a family, which are known to transport Cl⁻, bicarbonate, or protons, did not show altered expression levels by Northern analysis of whole-brain RNA preparations in AE3-knockout mice. Hybridization with an actin probe served as a loading control (not shown). For the Slc26a family only Slc26a11 is shown, as we did not detect expression of other family members in the brain.

DISCUSSION

Epilepsy is a very common neurological disorder, affecting about 1% of the population, and is part of the clinical picture in more than 200 genetic disorders (32). A strong genetic influence has been suggested for idiopathic epilepsy, and both monogenic inheritance and familial clustering without clear inheritance patterns have been described, suggesting that several genetic factors contribute to seizure susceptibility. Many genes mutated in monogenic epilepsies have been identified during the past years, most encoding ion channels (22, 24, 29), but only little is known about minor alleles conferring seizure susceptibility (33). While AE3-deficient mice appeared healthy and did not show alterations of brain morphology, overall activity, exploration behavior, or electrocorticograms, they exhibited a reduced seizure threshold after administration of epileptogenic drugs. Remarkably, a genome-wide linkage analysis mapped one susceptibility locus for idiopathic generalized epilepsy (IGE) in humans to the chromosomal region 2q36 (27) that harbors SLC4A3, the gene encoding AE3. Since SLC4A3 is broadly expressed in the brain, it is a promising candidate gene for IGE (28). Mutation analysis of SLC4A3 identified a high frequency of the A allele of the c.2600C>A variant (Ala867Asp) (28) in 16 patients from multiplex families with IGE and close linkage of the disease locus to D2S371 assigned to the chromosomal region 2q36. Comparison of 366 unrelated IGE patients with a control group of 183 normal individuals revealed a moderate but significantly increased risk for carriers of the c.2600A allele to develop IGE (28). It is not clear from these findings whether it is the Ala867Asp variant that confers susceptibility to epileptic seizures or another variant of SLC4A3, or a different gene in close proximity. Ala867 is not well conserved throughout evolution, and the functional consequence of the amino acid change, if any, is unknown. In view of our findings that AE3 deficiency in mice contributes to seizure susceptibility, our data support the assumption that a reduction in function of SLC4A3 contributes to increased susceptibility to epileptic seizures in humans.

Our present study identifies AE3 as the main sodium-independent Cl⁻/HCO₃⁻ exchanger in the CA3 region of the hippocampus. Under physiological conditions, this type of anion exchange activity will usually transport HCO₃⁻ out of the cell, which results in intracellular acidification. However, Cl⁻/HCO₃⁻ exchange can be reversed upon extracellular chloride removal and then leads to intracellular alkalinization, which is used to measure Cl⁻/HCO₃⁻ exchange activity. Though AE3 is expressed early, cultivated fetal hippocampal neurons from rats did not display active Cl⁻/HCO₃⁻ exchange in an earlier study (25). This prompted us to use hippocampal slices of adult mice for the analysis of pH regulation in AE3-knockout mice. As alkalinization upon chloride removal was absent in AE3-knockout compared to wild-type hippocampal slices, it can be concluded that AE3 is the dominant hippocampal anion exchanger.

The pathophysiology of the increased seizure susceptibility of AE3-knockout mice is not yet clear. Because of the physiological gradients for HCO₃⁻, Cl⁻/HCO₃⁻ exchange activity will result in the transport of Cl⁻ into the cell and HCO₃⁻ out of the cell. As a consequence, intracellular bicarbonate concentration may be slightly increased in AE3-knockout mice because of compromised outward transport. As GABA_A receptors are also permeable for HCO₃⁻, this may favor depolarizing GABA responses, due to HCO₃⁻ efflux (15, 31). On the other hand, neurons of AE3-deficient mice may have a slightly reduced intracellular Cl⁻ concentration, which should augment rather than weaken GABA-ergic inhibition. Intracellular alkalosis has been shown to be associated with increased neuronal excitability, while acidosis seems to dampen epileptic activity (36). Despite a small tendency toward increased steady-state pH_i in AE3-knockout neurons, we were unable to detect significant differences in baseline pH_i under our experimental conditions. Likewise, a contribution of DIDS-sensitive Cl⁻/HCO₃⁻ exchange to steady-state pH_i was not detectable in our slice preparations and suggests that AE3 is not very active at resting pH. However, it cannot be excluded that the pH_i of AE3-knockout neurons is raised due to impaired bicarbonate extrusion in certain microdomains, such as dendritic compartments, and that these changes may underlie the lowered epileptic threshold. Our pH measurements were carried out in the pyramidal cell layer of the hippocampal CA3 region, where neuronal somata are densely packed and astrocytes contribute to only approximately 6% of the total cellular volume (10); hence, we conclude that sodium-independent Cl⁻/HCO₃⁻ exchange is lost in CA3 pyramidal neurons. Yet, Cl⁻/HCO₃⁻ exchange has been reported in astrocytes (9, 30), but the molecular identity of the astrocyte anion exchanger has not yet been resolved. According to our data, Cl⁻/HCO₃⁻ exchange in astrocytes is not mediated by AE3, as AE3 was not detected in astrocyte lysates. Furthermore, the expression analysis for AE3 in the brain by in situ hybridization suggested a neuronal expression pattern with strong labeling of regions with densely packed neuronal somata and absence of AE3 transcripts in fiber tracts.

The disruption of AE3 did not result in morphological changes, as has been observed, e.g., in mice deficient for NHE1. Both spontaneous NHE1-mutant mice and NHE1-knockout mice displayed high mortality, neurodegeneration, and slow-wave epilepsy (2, 11). NHE1 is an acid extruder and, indeed, a significant lowering of the mean baseline pH_i in neurons from knockout mice was observed. Though we observed a small alkalinization of 0.05 pH units in the AE3-knockout compared to wild type, consistent with a role for AE3 in base extrusion, these changes were smaller than in NHE1-deficient mice and did not reach statistical significance. Careful examination of the brain of AE3-knockout mice did not reveal obvious morphological changes or degeneration, as shown for the brain of NHE1-knockout mice, thus making it difficult to delineate the exact nature of the epilepsy of these mice. In addition, an up-regulation of voltage-gated sodium channels in NHE1-knockout mice may be important for the epileptic phenotype (13, 39).

In view of the complete absence of Cl⁻/HCO₃⁻ exchange in hippocampal CA3 neurons of AE3-knockout mice, the mild phenotype of the mutant mice is surprising. It cannot be explained by a compensatory transcriptional up-regulation of other known transporters involved in pH_i homeostasis, as members of the Slc4a, Slc9a, and Slc26a families were expressed at equal levels in AE3-knockout brains. In our Northern analysis we found Ae2 (Slc4a2) and Slc26a11 transcripts in whole-brain RNA, but in the brain Ae2 is mainly expressed in the choroid plexus and obviously does not compensate for Ae3 (19). Slc26a11 belongs to a gene family (35) which includes members that mediate sodium-independent Cl⁻/HCO₃⁻ exchange under certain conditions (23). However, according to our data Slc26a11 probably does not function as a Cl⁻/HCO₃⁻ exchanger in hippocampal neurons.

In summary, we report the generation of an AE3-knockout mouse model. These animals show a reduced seizure threshold in response to proconvulsant substances. The exact cellular events by which AE3 deficiency confers seizure susceptibility are not yet resolved and are likely to be complex. It is tempting to speculate that the clinical course of various types of epilepsy might correlate with anion exchange activity and/or pH_i regulation in the central nervous system. Thus, our model provides a tool to investigate the role of sodium-independent Cl⁻/HCO₃⁻ exchange for neuronal excitation and pH_i regulation in more detail and strengthens the idea that AE3 has a role in regulating seizure susceptibility in humans.