Abstract

RESULTS

Nested Recombinant Strains to Dissect the El2Region

From initial QTL mapping experiments, the chromosomal localization of El2 was very imprecise (2-lod confidence interval of >20 cM) and a large portion of Chromosome 2 was transferred in the construction of congenic strains. This imprecision resulted partly from the inherent difficulty of QTL mapping in modest-size single generation mapping crosses where trait characteristics of a given genotype can be determined only once, and partly from the segregation of unlinked loci. Therefore, as a prelude to high resolution crosses for fine-structure mapping of El2, we decided to break-up the El2 congenic interval with a series of strains containing nested recombinations, anchored on either the centromeric or telomeric end by the recombination from the EL2E congenic strain. As each recombinant strain is homozygous, multiple individuals could be tested by independent measures of seizure susceptibility to determine which region harbors the main effect. In addition, recombinant strains which retained high seizure susceptibility would serve as renewable resources for further investigation. Individual mice containing recombinations were identified from crosses between EL2E and ABP, outcrossed to ABP, and then inbred lines were established to retain a particular recombination. Simple sequence length polymorphisms (SSLP) found between EL and ABP strains were used to define the recombination points of each strain. The result was eight derivative strains, four containing recombinations on the telomeric half of the El2 congenic interval and four on the centromeric half (Table ​(Table1).1).

Table 1

Genetic Markers that Define the Recombination Boundaries of CongenicStrains

Strain	Proximal boundary	Distal boundary	Interval size (cM)
EL2E	D2Mit5–D2Mit7	D2Mit71–D2Mit51	52.5
1	D2Mit422–D2Mit21	fromEL2E	14.7
2	D2Mit133–D2Mit30	fromEL2E	23.0
3	D2Mit132–D2Mit103	fromEL2E	26.2
4	D2Mit479–D2Mit58	fromEL2E	28.4
5	fromEL2E	D2Mit258–D2Mit136	34.9
6	fromEL2E	D2Mit224–D2Mit193	33.9
7	fromEL2E	D2Mit41–D2Mit43	21.9
8	fromEL2E	D2Mit435–D2Mit299	17.5

Open in a separate window

See Fig. ​Fig.1B1B for a graphical representation. Strains 1–8 are derived from the EL2E strain, as described in the text. Interval size was approximated from the MGD consensus map position midway between the two markers (see http://www.informatics.jax.org). 

Multiple Determinants for Rhythmic TossingSeizures

Seizure susceptibility of recombinant and parental strains was first assessed by analyzing the development of seizures over the duration of the rhythmic tossing assay (Fig. ​(Fig.1A).1A). This plot is useful for qualitative assessment of seizure incidence differences between strains and genders, as well as a starting point for further analyses. In both males and females, the seizure onset of the parental strain EL2E was earlier than in all other strains, and individuals in recombinants 5 and 6 began to seize near the midway point of the testing paradigm, although recombinant 5, which contains more of the donor strain interval, had a steeper onset curve. Recombinant 4 had an earlier onset in females than in males and further delays in seizure onset were seen in other recombinant strains, as well as the ABP parent. In general, the females were more susceptible than males, although this difference was stronger in some regions than in others.

Open in a separate window

Figure 1

Cumulative onset for rhythmic tossing-induced seizures. (A) The frequency of the number of seizures in females or males per strain is plotted for each rhythmic tossing test and points connected by colored lines (key is shown in B). Thin broken horizontal lines within the plot highlight the 50% mark and the fifteenth seizure test, respectively, chosen for subsequent analysis as described in the text. (B) Graphic representation of the EL-like interval (in black; actual values are detailed in Table ​Table1)1) retained in each of the congenic and recombinant strains, on its ABP background (white), with the Chromosome 2 centromere depicted as a ball and the colored lines used for each in A. The text table gives the effective number of tests for 50% (ET₅₀) of each strain to have experienced at least one seizure, as determined by Probit analysis described in the text.

El2 was mapped initially in conventional backcrosses following a single, arbitrary, quantitative measurement, mean seizure frequency per 20 rhythmic tossing tests. For a quantitative assessment in the present study, where different congenic strains and derivative recombinants were compared to each other we first determined the effective test number required for 50% of each group to experience a seizure (ET₅₀). This is a more standardized, albeit conservative approach. Comparisons of ET₅₀s show that no single interval accounts for the high seizure susceptibility of the EL2E strain. Although recombinant 5 did have an ET₅₀ similar to that of EL2E, neither recombinants 4 or 6, which overlap to together span the interval, was as susceptible (Fig. ​(Fig.1B).1B). These data provided the first direct evidence that at least two genes on Chromosome 2 contribute to El2.

To assess the significance of quantitative differences between recombinant strains and at the same time to minimize nongenetic effects, for example, test procedure on outcome (Poderycki et al. 1998), we limited the window of analysis to the earliest test number (15 tests) that showed a maximal difference between parental strains. These data again show that only nested recombinant 5 was as seizure susceptible as EL2E, whereas other strains show only partial susceptibility at best, depending upon gender and the chromosomal interval retained. (Results were not substantially different from those determined after 20 tests; data not shown.) These results are consistent with and extend the ET₅₀ results, showing that at least one gene from the proximal half of the interval and another from the distal half together comprise the seizure susceptibility of the parental congenic strain.

Common Determinants for PTZ and Rhythmic Tossing ThresholdSeizures

To determine whether this apparent genetic complexity was unique to the rhythmic tossing assay, susceptibility to PTZ induced seizures was determined in the same strain set (Fig. ​(Fig.2B).2B). Interestingly for PTZ, the gender bias was reversed when compared to rhythmic tossing, with males more sensitive than females. Nevertheless, the relative patterns of seizure incidence were similar in recombinants which retain the telomeric half of the interval, the main difference being the much higher susceptibility (represented by recombinant strains 7 and 8 in Fig. ​Fig.2B)2B) of the centromeric half to PTZ threshold seizures. In particular, recombinant strains 4 and 6, which contain each respective half-interval, have an increased seizure frequency which is statistically different from ABP in both testing paradigms. Nevertheless, because the recombinant 7 is already sufficient for the PTZ threshold phenotype of the EL2E parent, it is unclear whether the same gene(s) in the centromeric half of the interval are responsible for both rhythmic tossing and PTZ responses.

Open in a separate window

Figure 2

Comparison of seizure frequency utilizing rhythmic tossing. (a) vs. PTZ seizure testing (b), in recombinant strains. The Chromosome 2 area spanned by recombinant strains 1–8, EL2E, and ABP is depicted graphically (actual values given in Table ​Table1),1), between a and b by a white bar for ABP-like and a black bar for EL-like. Females are shown on the left of each histogram pair (white bar) and males on the right (gray bar). Numbers of animals in each rhythmic tossing group were as follows (F,M): ABP (28,22); 1 (44,47); 2 (22,20); 3 (14,13); 4 (20,31); 5 (22,22); 6 (32,38); 7 (32,22); 8 (30,23); EL2E (59,42); (EL2E×ABP)F₁ (16,18). Numbers of animals in each PTZ-tested group were as follows: ABP (35,29); 1 (32,28); 2 (37,29); 3 (29,42); 4 (35,26); 5 (17,12); 6 (21,34); 7 (47,31); 8 (36,25); EL2E (47,40); (EL2E×ABP)F₁ (37,29); (4×7)F₁ (28,29). Pairwise differences in seizure frequency between strains with adjacent recombinations were determined as described in Methods and are represented by quadrants, comparing genders (female, left quadrants; male, right) and seizure tests (rhythmic tossing, top quadrants; PTZ, bottom); significance is depicted by a different color or number of asterisks (dark blue, no asterisk, not significant; light blue, *, P<0.1; yellow, **, P<0.01; pink, ***, P<0.0001). Asterisks above histograms denote within-strain gender differences (same code). The quadrants to the right of strain 8 are from tests in comparison with the ABP/Le strain. Only the bottom quadrants are shown in the pairwise comparison between F1 hybrids because only PTZ was tested in both. Regions defined as a, b, c, and d are described in the text.

DISCUSSION

Seizure susceptibility genes on Chromosome 9 (El1, El4), Chromosome 2 (El2), and Chromosome 10 (El3) were initially mapped as QTL in crosses between the EL and ABP mouse strains (Rise et al. 1991; Frankel et al. 1995a). Congenic strains were then produced containing the EL allele of each of these chromosomes transferred onto an ABP strain background by genotypic selection, and these strains were tested for sensitivity to seizures induced by rhythmic tossing (Frankel et al. 1995b). The relative seizure frequencies of these strains were El2>El1,El4>El3, parallel to respective effects in initial backcrosses. In particular, the EL2E congenic strain (Frankel et al. 1995b) showed that EL strain-derived alleles correlated well with seizure susceptibility in the context of a nonepileptic strain background. This is consistent with the notion that El2 is a robust, major seizure locus suitable for higher resolution mapping.

To reduce the size of the El2 critical interval, we constructed and tested nested recombinant strains derived from the EL2E congenic. Surprisingly, these studies revealed a more complex pattern of inheritance. The most obvious result was that none of the recombinant strains recapitulated the full rhythmic tossing seizure phenotype of EL2E, as determined first by a conservative analysis of phenotype (Fig. ​(Fig.1B)1B) and then by seizure frequency itself (Fig. ​(Fig.2A).2A). It was unclear whether this apparent gene interaction between linked determinants is epistatic or additive. For example, in females, incremental differences between recombinant strains suggested a continuum of loci with additive effects, that is, increasing seizure susceptibility with each interval inherited (Fig. ​(Fig.2A).2A). However, these increments were not usually statistically significant and the effect was less apparent in males (Fig. ​(Fig.2A).2A). Therefore, we conclude that at least one gene in the telomeric half (between D2Mit58 and D2Mit30) and one in the centromeric half (between D2Mit7 and D2Mit41) interact to give the full parental congenic strain phenotype. However, given the apparent trend for seizure frequencies to respond incrementally with added genetic material from the EL strain, it is possible that multiple genes contribute to each.

Parallel experiments were performed utilizing an independent seizure stimulus, PTZ. In the recombinants, PTZ mirrored the rhythmic tossing results in the telomeric half of the interval and had a similar (albeit opposite) effect on gender overall (Fig. ​(Fig.2).2). The fact that similar patterns were obtained for both seizure paradigms lessens the likelihood that at least some of the gene(s) underlying El2 are merely artifacts of the rhythmic tossing procedure, as suggested by others (Poderycki et al. 1998). However, with respect to localization of essential seizure determinants, PTZ proved similarly confusing as rhythmic tossing. For example, PTZ genes in either half of the interval were sufficient to recapitulate the EL2E phenotype, at least in males (Fig. ​(Fig.2B,2B, strain 4 vs. strain 6 or 7). In addition, mice containing only the centromeric half of the EL2E congenic interval appeared exquisitely sensitive to PTZ, that is, they did not respond to rhythmic tossing per se (Fig. ​(Fig.2,2, strains 7 and 8). Nevertheless, as we speculated earlier, by inference there must be a gene(s) in the centromeric half of the interval which contributes to the overall EL2E for rhythmic tossing phenotype by interacting with more telomeric loci. It is therefore possible that the underlying genes are the same, but that PTZ as stimulus is sufficient to push their effects above threshold, whereas rhythmic tossing is not. The alternative hypothesis is that unique PTZ-responsive determinants reside in the centromeric half.

An interesting area in the middle of the EL2E interval negates the gender bias in PTZ seizures, conferring equal sensitivity to both males and females (Fig. ​(Fig.2,2, EL-derived region of strain 7 vs. strain 8). Additionally, an adjacent area (between strains 4 and 7) seems to exacerbate the gender bias. One or both effects could be explained by the presence of a gene which modulates the effect of PTZ on the GABAa receptor. Gender differences in sensitivity to PTZ, but not in GABAa receptor binding, have been previously noted (Kokka et al. 1992). Female rats were shown to have a higher threshold than males to seizures induced by PTZ as a result of neurosteroid modulation by metabolites of progesterone. Further studies would be necessary to distinguish this possibility from the alternative explanation, namely gender limited expression of seizure susceptibility alleles. However, if there are gender bias-enhancement loci, they specifically interact with seizure susceptibility alleles located distally, and are not necessarily a property of the ABP strain background. For example, comparable recombinant strains made by introgressing Chromosome 7 alleles from the epileptic SWXL-4 strain onto an ABP background do not show the same pattern of gender bias (M.E. Legare and W.N. Frankel, unpubl.).

Despite these daunting complexities, several seizure determinants were narrowed to small areas of Chromosome 2. At least two strain pairs with adjacent recombinations show significant differences in seizure susceptibility to either PTZ or rhythmic tossing (Fig. ​(Fig.2).2). For example, an ~1 cM region (Fig. ​(Fig.2,2, region “a”) between recombinants 5 and 6 differed for rhythmic tossing seizures, and an ~5 cM region between recombinants 6 and 7 had a more modest effect (Fig. ​(Fig.2,2, region “b”). In addition, a similar size interval between strains 7 and 8 females showed a significant difference in PTZ seizures (Fig. ​(Fig.2,2, region “c”). Finally, a small interval between strains 4 and 7 might contain an interesting gender-bias enhancement locus (Fig. ​(Fig.2,2, region “d”). However, although effects delimited to 1 cM are encouraging, in light of the overall nature of our results one cannot assume that the region is defined by a single genetic variant. Moreover, the effect of region “a,” while statistically significant, is small and further systematic fine-mapping will require a great deal of progeny-testing of individual recombinants. Further electrophysiological and pharmacological characterization of strain pairs which differ in these regions may help determine whether a region is worth pursuing by a positional candidate approach, and several candidate genes within these intervals could be tested directly.

It is clear that in some polygenic models, such as the epileptic EL mouse, conventional QTL mapping approaches suffice to confirm trait loci with large effects. However, further dissection may reveal them to be multiple loci each with smaller effects. Although among the six seizure frequency QTLs mapped in EL we have evaluated El2 most thoroughly, we are now completing studies on the Chromosome 7 locus Szf1 from the epileptic SWXL-4 mouse and have observed similar results (M.E. Legare and W.N. Frankel, unpubl.). Many published QTL show multiple likelihood curve peaks, suggestive of several underlying loci that are either more loosely linked to each other than in El2, that differ in additive or dominance components, or that exhibit different types of interaction with unlinked loci, any of which may be at least partly resolved by conventional QTL mapping programs, e.g., as for El5. (Frankel et al. 1995a). Nevertheless, in retrospect, our results are not particularly surprising. Based on the number of gene-targeted mouse mutations reported recently to have seizures, one can extrapolate that hundreds of genes would potentially give rise to seizure phenotypes when appropriately mutated. Consequently, between two common mouse strains, a functional polymorphism in a seizure gene might occur every 5–10 cM. Given the low resolution of conventional QTL mapping crosses and the tendency to select for further study regions with large effects, such approaches to detecting and confirming QTL might systematically capture several trait loci.

This phenomenon can be generalized for any trait where many segregating genes could potentially influence phenotype. There are two types of solution for such problems. First, although it is often desirable to study an intact phenotype, for example, tonic-clonic seizures, a study of component phenotypes should reduce the genetic complexity. This approach seems to work for certain systemic complex disorders, such as diabetes (Leiter et al. 1998), although it may be difficult or impractical for composite neurological behaviors. Second, alternatives to conventional QTL detection in mouse strains of choice have been suggested in recent years, including approaches such as advanced intercross (AI) lines (Darvasi and Soller 1995) and heterogeneous stocks (HS) (Talbot et al. 1999). In these cases, the retained nonrecombinant segments are much smaller than in conventional crosses. Thus, local linkages are more readily disrupted and mapped QTL are less likely the result of multiple linked genes. Nevertheless, some of these high-resolution approaches are not compatible with the type of control that is often necessary to understand the mechanism of a genetic effect (i.e., provide a strain pair where the only genetic variable is the locus in question), nor do they address the additional dilemma that many of these natural polygenic variants may be inherently difficult to study because effect sizes are usually quite small. The likelihood that common human genetic disorders are of a similar nature while perhaps justifying the approach, does not solve the problem. Moreover, polygenic models are many generations away from being suitable for genetic isolation and bear the extra burden of proof necessary for distinguishing between strain polymorphism and causative mutation. For seizure disorders, given the large number of target genes and the uncertainty of using multiple crosses to sort through the many possible natural variants to find those that are the most appropriate models for study, the systematic screening for de novo mutations induced by random chemical mutagenesis on a known strain background (King et al. 1997; Brown 1998; Schimenti and Bucan 1998) is an attractive alternative. Mutagenesis screens allow one to rapidly select from a set of single gene mutations of interest those that have the strongest effects and move directly from high-resolution mapping to gene identification.

METHODS

Acknowledgments

Footnotes

REFERENCES