FIGURE 5 Alternative gene replacement designs. The use of site-specific recombination (SSR) systems (gray arrows) permits useful modifications to the basic knockout construct design. An excisable selection cassette flanked by 34-bp loxP sites allows added Cre recombinase to delete the marker after positive selection in ES cells but before introduction into mice, thus preventing read-through transcription by the selectable marker promoter from transcribing downstream gene elements or interfering with the expression of neighboring genes. A knock-in is similar, but it replaces an endogenous exon with a mutant exon (asterisk), leaving only a single 34-bp loxP recognition site within an intron. Rather than obliterating gene function completely, a gene knock-in allows the gene to continue expressing in a normal pattern and thus allows for analysis of more subtle or gain-of-function mutations. An inducible, or conditional, knockout utilizes two different SSR systems (e.g., Cre-loxP and flipase-/ri) to allow the targeted allele to function normally (b) until the investigator decides to induce deletion of the "floxed" exon (c). This permits mutations that would normally result in embryonic lethality to be studied in adult mice or limits the mutation to a specific subset of cells. The diagonal lines represent recombination events; a, b, and c represent sequential steps in the modification of targeting constructs following homologous recombination in embryonic stem cells.

audiogenically induced seizures in 6 cases, and increased resistance to induced seizures in 2 cases. A mouse model of Smith-Magenis syndrome (SMS), which relied on gene replacement to alter an entire chromosome region rather than just a single gene, also exhibited spontaneous seizures (Walz et al., 2003). In some mutated strains, the seizure or excitability phenotype was not investigated (or at least not reported), presumably because the model was generated for purposes other than epilepsy research. In other cases, independent models were created by targeting the same gene; in these cases differences in targeting strategies, mouse genetic strains, or subsequent methods of analyses led to discordant reported phenotypes. Finally, in some cases the mouse phe-notype was similar (i.e., obviously relevant) to the clinical phenotype of human epilepsy patients carrying mutations in the same gene; in other cases, there was no apparent correspondence between the mouse and human phenotypes. A few of these cases are considered in greater detail to illustrate particular points regarding the development and analysis of mouse gene replacement models of epilepsy:


Spontaneous mutations in the voltage-gated Ca2+ gene, Cacna1a, are responsible for a phenotype that includes ataxia and generalized absence-like seizures (Fletcher et al., 1996) as well as an altered threshold for cortical spreading depression (Ayata et al., 2000) in the tottering mouse. A functionally null knockout of mouse Cacna1a also produces ataxia and absence-like seizures but results in lethality by postnatal day 30. Mutations in human Cacna1a are associated with several distinct disorders, including familial hemi-plegic migraine (FHM), episodic ataxia type 2 (EA-2), and spinocerebellar ataxia type 6 (SCA-6) (Ophoff et al., 1996; Zhuchenko et al., 1997). Given these phenotypes, the spontaneous and knockout mutations in mouse Cacna1a originally appeared to be more accurate models of human ataxia and FHM caused by mutations in this gene. Recently, however, Jouvenceau et al. (2001) and Imbrici et al. (2004) identified human mutations that were associated with inherited ataxia plus primary generalized epilepsy. Thus Cacna1a provides an example of a mouse model of epilepsy where the complex phenotype (ataxia plus seizures) corresponds to only a small subset of human patients with mutations in this gene.


Synchronized thalamocortical spike-wave discharges are a basic pathophysiologic finding in absence epilepsy and are believed to depend strongly on the activity of the low-voltage-activated T-type Ca2+ channel. Three genes can encode a pore-forming alpha subunit capable of mediating T-type Ca2+ currents: Cacna1g, Cacna1h, and Cacna1i. To examine the role of one of these genes in absence seizures, Kim et al. (2001) inactivated mouse Cacna1g by gene replacement. These authors reported a lack of burst firing by thalamocortical relay neurons and resistance to the generation of spike-and-wave discharges in response to GABAb receptor activation. Subsequent analysis of double-mutant mice (Cacna1g -/- and Cacna1a -/-) extended these findings to show a complete loss of T-type Ca2+ currents in thal-amocortical neurons and no evidence of the spike-wave discharges seen in Cacna1a single-mutant mice. Mutations in human Cacna1g have not yet been identified. Whereas the Cacna1g knockout mouse cannot be considered a model of epilepsy in the strictest sense (because it does not exhibit spontaneous seizures), it nonetheless demonstrates the importance of the ajG T-type Ca2+ channel for mediating absence seizures in the thalamocortical pathway and the value of the gene replacement approach for elucidating seizure mechanisms.


Haug et al. (2003) reported mutations in the human voltage-dependent Cl- channel gene CLCN2 in patients with a variety of seizure types, including childhood absence epilepsy (CAE), juvenile absence epilepsy (JAE), juvenile myoclonic epilepsy (JME), and epilepsy with grand mal seizures on awakening (EGMA). Bosl et al. (2001) previously described the phenotype of mice with targeted inacti-vation of the Clcn2 gene but reported the conspicuous absence of expected seizures. Instead, Clcn2-null mice exhibited a severe degeneration of cells in the retina and testes, suggesting a possible role for Clcn2 in controlling the ionic environment of these cells. The basis for this dramatic interspecies difference in mutant phenotypes is unclear.

Gabbrl, Gabral, Gabrb3, Gabrg2

Genes encoding subunits of the ionotropic GABAa receptors and metabotropic GABAB receptors are excellent candidates for involvement in epilepsy, given the central role of GABA in mediating inhibition in brain. Mutations in two of these genes have been associated with epilepsy in humans: GABRA1 with juvenile myoclonic epilepsy (Cos-sette et al., 2002), and GABRG2 with generalized epilepsy with febrile seizures plus (GEFS+), childhood absence epilepsy, and febrile seizures (Baulac et al., 2001; Wallace et al., 2001). Mice carrying knockouts in these genes and two other GABA receptor genes (Gabbr1 and Gabrb3) have been produced. Null mutations in Gabbr1 resulted in spontaneous clonic, tonic-clonic, and absence seizures as well as hyperalgesia, memory impairment and premature death (Prosser et al., 2001; Schuler et al., 2001). Seizures were not reported to result from inactivation of Gabra1 (Sur et al., 2001; Vicini et al., 2001). A knockout of Gabrb3 caused fre quent myoclonus and occasional epileptic seizures in addition to cleft palate (Homanics et al., 1997). It is possible that the Gabrb3-null mouse seizures are a model of the seizures observed in some Angelman and Prader-Willi syndrome patients because this gene is contained within the region (15q11-q13) affected by the large deletions and translocations that are associated with these two syndromes (Dooley et al., 1981; Veenema et al., 1984; Wagstaff et al., 1991). However, mutations in human GABRB3 were also recently found in patients with chronic insomnia (Buhr et al., 2002). Finally, a knockout mutation of mouse Gabrg2 was shown to result in perinatal lethality in one study (Gunther et al., 1995), but to cause epilepsy with lethality around 4 weeks in another (Schweizer et al., 2003). The reason for this age disparity in mortality between the two knockouts is unclear, but it might relate to differences in the genetic background of the mice.


No mutations in the human GAD2 (glutamate decar-boxylase 2) gene have yet been associated with seizures, although this gene is believed to be critical for mediating inhibition in brain through production of the inhibitory neu-rotransmitter, GABA. Three knockout studies of mouse Gad2 have been reported. Interestingly, the phenotypes varied. In one case, the Gad2-null mice exhibited no spontaneous seizures, but seizures were more easily induced by PTZ or picrotoxin (Asada et al., 1996). In the other two studies, spontaneous seizures and increased mortality were observed (Kash et al., 1997; Yamamoto et al., 2004). Interestingly, Kash et al. (1997) reported that the seizures were genetic strain-dependent—which might explain the pheno-typic variability observed in the different studies.


AMPA-type ionotropic glutamate receptors, composed of four related subunits (GRIA1-4) in various combinations, mediate the fast component of excitatory postsynaptic currents in CNS neurons. Inclusion of the GRIA2 subunit in AMPA receptors decreases Ca2+ permeability significantly (Hollmann et al., 1991). The human and mouse GRIA2 genes both undergo functionally important mRNA editing at a site corresponding to the second transmembrane domain of the receptor. No human mutations in GRIA2 have yet been associated with epilepsy and, consistent with this, both knock-out and knock-in mutations of mouse Gria2 protein coding sequences do not result in seizure phenotypes (Jia et al., 1996; Kask et al., 1998; Sans et al., 2003). Interestingly, a knock-in mutation that alters the intronic mRNA editing site of Gria2 resulted in seizures and early lethality (Brusa et al., 1995; Feldmeyer et al., 1999). This example demonstrates a potential role for mRNA editing in epilepsy, and shows that gene replacement approaches can be useful for studying non-coding gene regions as well as coding sequences.


Mutations in human KCNA1 usually result in episodic ataxia-myokymia syndrome without any seizures or EEG abnormalities (Browne et al., 1994), but rare mutant alleles have been identified that result in tonic-clonic and simple partial seizures with myokymia and no ataxia (Eunson et al., 2000). A spontaneous mutation in mouse Kcna1 is associated with megencephaly and recurrent behavioral seizures (Donahue et al., 1996; Petersson et al., 2003). Complete inactivation of mouse Kcna1 via gene replacement resulted in frequent spontaneous seizures through adult life as well as abnormal axonal action potential conduction in peripheral nerve which may be related to the pathophysiology of episodic ataxia-myokymia (Smart et al., 1998).


The protein products of the KCNQ2 and KCNQ3 genes co-assemble into a channel to produce the M-type K+ current in neurons, a slowly activating and deactivating current important for regulating intrinsic membrane excitability and response to synaptic input. Mutation of either of these genes in humans results in benign familial neonatal convulsions (BFNC) (Biervert et al., 1998; Charlier et al., 1998; Singh et al., 1998). Complete inactivation of Kcnq2 by gene replacement in mice resulted in homozygous perinatal lethality. Mice heterozygous for the mutation did not exhibit spontaneous seizures but showed a decreased threshold for seizures induced by the convulsant PTZ (Watanabe et al., 2000). Peters et al. (2005) used a novel transgenic approach to conditionally suppress M-channels in mouse brain and demonstrated their function in neuronal excitability and behavior. In this way these investigators obtained information similar to what might have been obtained by a knock-in mutation with incomplete loss of function but in only a fraction of the time.

Scnlb, Scn2a, Scn2b

Mutations in human SCN1B are associated with generalized epilepsy with febrile seizures plus (GEFS+) (Wallace et al., 1998). A complete knock-out of the orthologous mouse gene, Scn1b, also results in spontaneous generalized seizures and should be a good model for studying the human disorder (Chen et al., 2004). Targeted inactivation of the closely related mouse Scn2b gene did not cause spontaneous seizures but did lead to increased susceptibility to pilo-carpine-induced seizures (Chen et al., 2002). Mutations in human SCN2A are also associated with generalized epilepsy with febrile seizures plus (GEFS+) (Sugawara et al., 2001) but cause benign familial neonatal-infantile seizures

(BFNIS) in other cases (Heron et al., 2002). While a transgenic mouse overexpresing a mutant form of Scn2a does exhibit spontaneous seizures (Kearney et al., 2001), the functionally null knockout of Scn2a causes recessive perinatal lethality with no seizures reported (Planells-Cases et al., 2000). A wide variety of different human Na+ channel mutations have been reported, associated with a range of seizure phenotypes; it seems likely that these genes will continue to provide a rich source of mutations to be modeled by knock-in approaches in the future.

Limitations and Pitfalls Expense and Time

Gene knockout models are at least an order of magnitude more costly to produce, on average, than standard transgenic models. Much of this extra cost is due to the extensive embryonic stem cell culturing and analysis that must be carried out to produce a knockout model. Knockout models also require significantly more time to generate than trans-genics. Hemizygous transgenic mice (the founder generation) can be born as few as about 20 days after the transgene construct is microinjected into donor zygotes. Knockout mice, on the other hand, typically require several months after the targeting construct is built just to obtain a clonal embryonic stem cell line with an appropriately targeted gene. The embryonic stem cells must then be microinjected into blastocysts and implanted into surrogate mothers to wait another 20 days before heterozygous chimeric mice can be born. These heterozygous chimeric mice must then be raised to breeding age and interbred at least once to produce homozygous germline mutants (total time from construct to mutant is 1 to 2 years).

Genomic Redundancy and Cooperative Pathways

One of the problems that complicates evaluation of mouse gene knockout phenotypes is the relatively common phenomenon of redundant or interacting genetic pathways (for an example of how this can occur in a mouse epilepsy model, see Burgess et al., 1999). Thus the absence of an abnormal phenotype in a knock-out mouse does not necessarily indicate that the targeted gene is not important for regulating neuronal excitability or synchronization. Elucidating its role might only be revealed by inactivating a compensating gene or gene pathway.

Incomplete Inactivation

When the goal of a gene knockout project is to obliterate gene function to create an epilepsy phenotype, but a targeting strategy cannot be designed that will completely delete all coding sequences of the target gene, the targeted allele might express a truncated form of the protein and thus retain partial function or gain a novel function. A targeted allele might skip a deleted region by alternative pre-mRNA splicing, use alternative or cryptic promoters to produce a novel protein fragment, or transcribe straight through an inserted selection cassette (allowing initiation of translation from downstream start codons). Even a small protein fragment could retain the ability to interact with other proteins in the original pathway or multi-subunit complex and produce a dominant-negative effect. This pitfall can be minimized by careful initial design of the targeting construct and by post knockout screening for residual mRNA and protein products expressed from the targeted locus.

Selection Cassette Interference

The selection cassettes used in standard replacement-type constructs usually contain very strong promoters. Transcription from these promoters may lead to transcription of any remaining parts of a target gene or interference with the normal expression pattern of a neighboring gene and thus confuse interpretation of the experiment. The solution to this problem is to remove the selection cassette following confirmation of homologous recombination in embryonic stem cells using one of the available recombinase systems, such as CRE or FLP (see preceding discussion).

Disruption of Overlapping or Adjacent Genes

The creation of a significant alteration in a target gene could also lead to the unintentional mutation of as yet unidentified genes, for example, genes residing in introns of the target gene or overlapping the target gene but encoded by the opposite strand. It is also conceivable that the target gene contains within its structure one or more regulatory elements important for controlling the expression of unrelated nearby genes. A related possibility is that any alteration in chromatin structure in a knocked-out gene, regardless of how minor, could be propagated to affect one or more adjacent genes. If any of these situations occurred, the resulting phenotype would not accurately reflect alterations in the target gene alone. A brief analysis of the expression function of adjacent genes is recommended to detect this potential problem.

Was this article helpful?

0 0

Post a comment