Mice Expressing Only Monosialoganglioside GM3 Exhibit Lethal Audiogenic Seizures*

Gangliosides are a family of glycosphingolipids that contain sialic acid. Although they are abundant on neuronal cell membranes, their precise functions and importance in the central nervous system (CNS) remain largely undefined. We have disrupted the gene encoding GD3 synthase (GD3S), a sialyltransferase expressed in the CNS that is responsible for the synthesis of b-series gangliosides. GD3S−/− mice, even with an absence of b-series gangliosides, appear to undergo normal development and have a normal life span. To further restrict the expression of gangliosides, the GD3S mutant mice were crossbred with mice carrying a disrupted GalNAcT gene encoding β1,4-N-acetylgalactosaminyltransferase. These double mutant mice expressed GM3 as their major ganglioside. In contrast to the single mutant mice, the double mutants displayed a sudden death phenotype and were extremely susceptible to induction of lethal seizures by sound stimulus. These results demonstrate unequivocally that gangliosides play an essential role in the proper functioning of the CNS.

Gangliosides are glycosphingolipids that contain sialic acid (reviewed in Ref. 1). They are found on the external leaflet of the plasma membrane on eucaryotic cells and are most abundant in the central nervous system (CNS) 1 where they represent the major sialoglycoconjugate. Because of their dramatic changes in expression during neuronal development and differentiation (3)(4)(5)(6), as well as their prominence in the mature CNS, gangliosides have long been assumed to have fundamental roles in CNS development and function.
In the ganglioside biosynthetic pathway (1) (see Fig. 2A), lactosylceramide serves as the core structure. The first ganglioside synthesized, GM3, 2 is produced by the transfer of an ␣2,3-linked sialic acid residue to lactosylceramide. Subsequently, GM3 can be modified by the action of ␤1,4-N-acetylgalactosaminyltransferase (GalNAcT, EC 2.4.1.92) to produce GM2 and other complex gangliosides. Alternatively, GM3 can be modified by the action of GD3 synthase (CMP-sialic acid: GM3 ␣-2,8-sialyltransferase, EC 2.4.99.8) to produce the disialoganglioside GD3, which diverts the pathway to the synthesis of b-and c-series gangliosides. Gene targeting in mice has been a particularly fertile approach for uncovering the functions of gangliosides in the CNS. Disruption of the GalNAcT gene (7) blocks the synthesis of complex gangliosides and results in the expression of only the simple gangliosides GM3 and GD3. Surprisingly, these mutant mice are viable, with a normal life span and a CNS that is largely intact both morphologically and functionally (8,9). These mice do, however, exhibit an agerelated dysmyelination process that is associated with axonal degeneration (10). The mechanism for dysmyelination may be the absence of neuronal ganglioside ligands for myelin-associated glycoprotein (MAG) resulting in myelin instability. Ultimately, motor defects are observed in aged, 12-month-old Gal-NAcTϪ/Ϫ mice, suggesting a role for complex gangliosides in long-term CNS maintenance (11).
We have now produced mice with a disrupted GD3 synthase (GD3S) gene (12), of which expression is relatively high in the CNS. Even though these mice lack the disialylated b-series gangliosides in their CNS, they show no overt phenotype. By combining the GalNAcT and GD3S mutant alleles, we have established double knockout mice (DKO) that express only monosialylated GM3 as their major CNS ganglioside. In contrast to each of the single knockout mice, the DKO mice show a sudden death phenotype and a severe CNS disturbance that is manifested by an exquisite susceptibility to lethal, soundinduced seizures. These results provide compelling evidence for the critical and essential role of gangliosides in CNS function.

EXPERIMENTAL PROCEDURES
Targeted Disruption of the Mouse GD3S Gene and Generation of Mutant Mice-To generate the GD3S knockout mice, a genomic fragment was isolated from a 129/Sv strain library (Stratagene, catalog no. 946306) containing the fifth exon of the 5-exon GD3S gene (13). For knocking-in the LacZ reporter gene and targeted inactivation of the GD3S gene, a LacZ-neo (neomycin-resistant gene) cassette (14) was * This work was supported in part by funding from the Deutsche Forschungsgemeinschaft (SFB 284 to K. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  ʈʈ To whom correspondence should be addressed: Bldg. 10, Rm. 9N-314, National Institutes of Health, Bethesda, MD 20892-1821. 1 The abbreviations used are: CNS, central nervous system; DKO, double knockout mice; MAG, myelin-associated glycoprotein; GD3S, the gene encoding GD3 synthase; GalNAcT, the gene encoding ␤1,4-Nacetylgalactosaminyltransferase; TLC, thin layer chromatography; Xgal, 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside; kb, kilobase pair; PBS, phosphate-buffered saline; E, embryonic day (e.g. E11.5). 2 The nomenclature for gangliosides follows the system of Svennerholm (2).
inserted into the BamHII site of the fifth exon. The resulting targeting vector contained an ϳ11-kb GD3S 5Ј fragment and a 2.2-kb GD3S 3Ј fragment separated by the LacZ-neo cassette (Fig. 1A). The LacZ coding region is preceded by an internal ribosomal entry sequence. Therefore, targeted insertion generates a bicistronic transcription unit in which the expression of the ␤-galactosidase reporter protein is under the control of GD3S transcriptional regulatory elements. The herpes simplex virus thymidine kinase (TK) gene was positioned outside the homologous sequence to deter random integration events. Targeting of TC1 embryonic stem cells and establishment of chimeric mice were performed as in past experiments (8,15). Male GD3SϪ/Ϫ mice were crossed to female GalNAcTϪ/Ϫ mice (8) to generate GD3Sϩ/Ϫ GalN-AcT ϩ/Ϫ double heterozygotes. These double heterozygote mice were intercrossed to generate GalNAcTϪ/Ϫ GD3SϪ/Ϫ double mutant mice. Genotyping was accomplished by polymerase chain reaction and Southern analysis of genomic DNA isolated from tail biopsies.
Glycosyltransferase Assays, Lipid Analysis, and MAG Determina-tion-GD3 and GM3 synthase activities were determined as previously described (16,17). The lipid analysis of mouse brain was done as described (18). The TLC solvent system was CHCl 3 /MeOH/0.22% CaCl 2 (60:35: Each lane was loaded with the equivalent of 10 mg wet weight of brain. The ganglioside bands were detected with a phosphoric acid/copper sulfate reagent (15.6 g of CuSO 4 -(H 2 O) 5 and 9.4 ml of H 3 PO 4 (85%, w/v) in 100 ml of water) (19). Lipid-linked sialic acid determinations were done according to Svennerholm (20) modified by the method of Miettinen and Takki-Luukkainen (21). Determination of MAG levels was accomplished by Western blotting.
For histological analysis, the mice were perfused with PBS and 10% phosphate-buffered formalin. The brain, spinal cord, and visceral organs were dissected out, fixed in 10% buffered formalin, and processed for embedding in paraffin. The pathology of the organs were surveyed on the sections stained with hematoxylin and eosin, and brain and spinal cord were further examined with Luxol-Fast Blue and Nissl staining.
Audiogenic Seizure Induction-The mice were 2-4 months old at the time of testing. An individual mouse was put in a conventional cage with bedding and allowed to explore for at least 1 min. After the acclimation period, a key bundle was shaken steadily 15 cm above the cage by the experimenter for 5-10 s. The key bundle consisted of five keys, and stimulus noise level was within the range of 83 to 86 decibels. Background noise level in the room was 52 decibels. Observations were made for at least 60 s after the start of key shaking.

RESULTS AND DISCUSSION
Generation of GD3 Synthase Knockout Mice-To disrupt the GD3S gene in mice, we created a targeting vector that would both render GD3 synthase inactive and allow monitoring of the GD3S locus expression (Fig. 1A). The targeting vector was designed to introduce a LacZ reporter gene into the last exon of the GD3S gene. This targeting strategy would cause an early termination of the GD3 synthase protein, resulting in the elimination of the highly conserved sialyl S motif, a critical portion of the catalytic domain (22). Removal of the sialyl S motif from the GD3 synthase in this manner renders the enzyme inactive (16,23). After electroporation of the targeting vector into embryonic stem cells, we obtained clones with a homologously recombined GD3S allele. Targeted embryonic stem cells were used to create chimeric mice and, subsequently, heterozygous mice. Crossbreeding of the heterozygous mice resulted in the birth of homozygous mutant mice (Fig. 1B) at the expected Mendelian frequency, indicating an absence of embryonic lethality associated with this mutation.
GD3 synthase activity was not detectable in extracts of GD3S Ϫ/Ϫ brain (Fig. 1C). However, GM3 synthase, the sialyltransferase that catalyzes the preceding step of ganglioside biosynthesis, was readily detectable in the same extracts (Fig. 1C).
We next determined the embryonic expression pattern of the GD3S gene by whole mount staining of heterozygous and homozygous mutant embryos with X-gal. At E11.5, very strong GD3S expression was found in developing spinal cord. The eyes and the top of the metencephalon were also positive (not shown). Throughout later developmental stages, E13.5-E17.5, the targeted GD3S gene expression continued to be relatively restricted to the developing CNS (Fig. 1D).
As would be predicted from the location of GD3 synthase in the biosynthetic pathway of gangliosides ( Fig. 2A), the brain ganglioside pattern of GD3SϪ/Ϫ mice was devoid of the bseries structures, GD1b and GT1b, and was dominated by the A, schematic representation of the GD3S targeting strategy. The structure of the 3Ј end of the mouse GD3S locus is shown at the top, the structure of the targeting vector in the middle, and the predicted structure of the homologous recombined locus is on the bottom. B, BamHI; H, HindIII; probe, probe for Southern blot; WT, wild type; RV, EcoRV. B, identification of mice homozygous for the GD3S locus. Southern analysis yielded a 4.5-kb HindIII band corresponding to the wildtype GD3S allele in wild-type DNA using the external probe shown in A. DNA from mice homozygous for the disrupted GD3S allele yielded a 7-kb band. C, GD3SϪ/Ϫ mice are devoid of GD3 synthase activity. As demonstrated by TLC, GD3 synthase activity was not detectable in the extract from the GD3SϪ/Ϫ mouse brain but was readily apparent in the extract from the wild-type mouse brain. Both extracts contained GM3 synthase activity as indicated by the synthesis of radiolabeled GM3. The TLC migration positions of labeled GD3 and GM3 are indicated by arrows. D, GD3SϪ/Ϫ, embryos were obtained and were incubated with X-gal to visualize expression from the GD3S locus. Embryos were sectioned along the saggital plane prior to staining to expose internal structures. Note the high level of expression in the CNS. a-series gangliosides, GM1 and GD1a (Fig. 2B). The absence of b-series gangliosides further documents the disruption of the GD3S gene and validates the ganglioside synthesis scheme proposed by Sandhoff and colleagues (1,24). The GD3SϪ/Ϫ mice were viable and fertile, had normal growth, and were without gross behavioral abnormalities. Histologic examination of brains from GD3SϪ/Ϫ mice did not reveal any obvious abnormalities. The GD3S Ϫ/Ϫ mice did not show observable demyelination as determined by Luxol Fast Blue staining of brain sections.
Establishment of Mice That Express Only GM3 GalNAcTknockout mice express predominantly GM3 and GD3 gangliosides in their CNS due to the absence of ␤1,4-N-acetylgalactosaminyltransferase, yet they have fundamentally normal nervous system development and a normal life span (8,9). We further narrowed the expression of ganglioside structures by crossing the two types of ganglioside mutant mice. Mice doubly mutant for the GalNAcT and GD3S genes were born at the expected Mendelian frequency. As predicted from the biosynthetic pathway of gangliosides ( Fig. 2A), elimination of both transferases in the DKO mice resulted in the expression of GM3 as their major CNS ganglioside (Fig. 2B). Although roughly similar amounts of total brain ganglioside were present in wild-type and mutant mice, DKO mice contained markedly reduced levels of lipid-linked sialic acid in brain compared with wild-type and single mutant mice (Fig. 2C).
After their weaning we noted that the DKO mice had an extremely high mortality rate. We followed the life span of a group of DKO mice and found that by 13 weeks of age about 50% had died of unknown causes (Fig. 3A). The overall mortality was 92% during the observation period (36 weeks). We searched for pathological changes in the DKO mice. We examined brain sections by Nissl, and Luxol Fast Blue staining but could not discern any gross pathological changes, including neuronal cell loss or significant demyelination. Histological examination was done with mice after timed sacrifice and after sudden death. It had been noted previously that expression of MAG, which binds certain gangliosides, was reduced in the GalNAcT mutant mice, presumably because of a deficiency of ganglioside ligands for MAG (10). We found that the single mutant GD3S mice had a level of MAG similar to wild-type mice (not shown). This finding can be explained by the presence of GD1a (Fig. 2, A and B), a relatively high affinity MAG ligand (25). In the DKO mice, MAG expression was similar to that of the single mutant GalNAcT mice (not shown), presumably because of a comparable deficiency of MAG ligands in the two types of mutant mice.
Again, as had been noted in the GalNAcT Ϫ/Ϫ mice, the male DKO mice displayed a defect in sperm maturation resulting in sterility (not shown) (8,26). No pathologic abnormalities were found in visceral organs in the DKO mice.
Double Mutant Mice Are Sensitive to Sound-induced Seizures-We noticed that some of the DKO mice showed apparently spontaneous and/or handling-induced seizures. In one case, a mouse died after such a handling-induced seizure attack. Therefore, we hypothesized that the reason for sudden death in the DKO mice may be enhanced susceptibility to lethal seizure. Sound is a classic means of inducing seizures in certain susceptible strains of mice (27). We used a relatively mild sound stimulus, key jangling, to test the DKO and single mutant mice for sensitivity to audiogenic seizure. We found that the double mutant mice were exquisitely sensitive to seizure induction by this method (Fig. 3B). The typical seizure began with a brief wild running phase, followed by a clonic-tonic seizure. In most cases the seizure resulted in death. All DKO mice tested were induced to have a clonic-tonic seizure. Ultimately, 9 of 10 DKO mice died during tonic phase, apparently because of respiratory arrest. None of the wild-type or GD3S Ϫ/Ϫ mice showed any seizure response to the sound stimulus. One of the nine GalNAcT Ϫ/Ϫ mice responded with wild running activity and a clonic seizure, but the episode did not progress further to lethality.
Why Are Mice That Express Only Monosialoganglioside GM3 Exquisitely Susceptible to Lethal Seizures?-Seizure susceptibility in mice has been shown to result from mutations in a large number of genes with diverse functions, including a variety of proteins involved in neuronal cell signaling (28). Gangliosides have been proposed to act as modulators of transmem-

FIG. 2. Ganglioside and MAG expression in ganglioside mutant mice.
A, the partial biosynthetic pathway of gangliosides with blocks in the pathway predicted in the single and double mutant mice. B, brain ganglioside patterns from wild-type (WT) and ganglioside mutant mice. The positions of major brain gangliosides are shown on the sides of the panel. C, lipid-linked sialic acid levels from brains (n ϭ 3) of wild-type and ganglioside mutant mice. An estimation of the total brain ganglioside concentration (nmol/10 mg wet weight) based on scanning the TLC plates and the sialic acid determination is as follows: wild-type, 7-10; GD3SϪ/Ϫ, 11-16; GalNAcTϪ/Ϫ, 12-14; GD3S/GalNAcTϪ/Ϫ, 13-14. brane signaling (29,30). Recently, Kasahara et al. (31,32) have shown that GD3 ganglioside is associated with the Src family kinase, Lyn, and with the glycosylphosphatidylinositol-anchored signaling protein, TAG-1, in glycosphingolipid-enriched microdomains. It is also of interest that mice deficient in Fyn kinase, another Src family kinase, are susceptible to audiogenic seizures (33). In the DKO mice, the undersialylated ganglioside complement could alter the properties of these glycosphingolipid-enriched membrane microdomains, resulting in impaired neuronal cell signaling and seizure susceptibility.
A role for gangliosides in the development of the CNS has been suggested based on the striking changes in ganglioside composition during the differentiation of neurons (3)(4)(5)(6). Thus, it is also possible that the seizure susceptibility in DKO mice may be a result of abnormal CNS development. Although his-tologic examination of DKO mutant brains revealed no obvious abnormalities, subtle defects cannot be ruled out.
Finally, our findings may have relevance for human epilepsy, a disorder that burdens about 1% of the population (28). Our results raise the possibility that mutations altering the ganglioside synthesis pathways may be factors in contributing to some forms of this disorder. FIG. 3. Life span and audiogenic seizures of ganglioside mutant mice. A, groups of mice (n ϭ 9) of each of the genotypes were monitored for survival from an age of 4 weeks until 35 weeks. B, mice from each genotype (wild-type (WT), n ϭ 5; GalNAcTϪ/Ϫ, n ϭ 9; GD3SϪ/Ϫ, n ϭ 5; GD3S/GalNAcT Ϫ/Ϫ, n ϭ 10) were exposed individually to a sound stimulus created by jangling a set of keys for ϳ5 s. The behavior of the mice was observed for the following: wild running, clonic seizure, tonic seizure, and death. *, one GalNAcTϪ/Ϫ mouse responded to the sound stimulus with wild running and a clonic seizure but not a tonic seizure or death. **, p Ͻ 0.001; 2 test.