Cost of genetic incompatibility
The cost of genetic incompatibility is the reduction in litter size that a +/t incurs when mating with a +/t rather than a +/+. In experiment 1, we estimated the cost of genetic incompatibility to females by performing all possible crosses of +/t and +/+ genotypes and counting litter size at birth. Litter size differed between crosses (Table S1; ANOVA, F_{3,49} = 4.04, P = 0.012); +/t females mated to +/t males produced litters significantly smaller than those of all other crosses (for all contrasts, t_{1,49} > 2.34, P < 0.023). Litter sizes at birth may have been underestimated, as we found remains of dead pups in five instances, implicating infanticide, with a further suspicion of such causes in two more cases. To rule out maternal cannibalism, we conducted a second experiment, examining the uteri of females shortly after giving birth.
In experiment 2, we conducted additional crosses to compare fertility, measured as the total number of uterine scars, which indicates the number of embryos implanted in the uterus. Fertility did not differ significantly among crosses (Fig. 3, ANOVA, F_{3,70} = 1.40, P = 0.249). The overall mean number of uterine scars per female was 7.58 (±0.15 SE). This indicates the average litter size at birth that would have resulted from the survival of all implanted embryos, regardless of genotype. Counts of red scars differed among crosses (Fig. 3, ANOVA, F_{3,70} = 16.12, P = 0.001). Red scars indicated that +/t × +/t matings resulted in a mean of 4.18 (±0.42 SE) offspring, significantly fewer than that detected in all other groups (t_{1,70} > 4.67 for all contrasts, P < 0.001), on average a loss of 3.40 pups per litter. Litter size at birth was highly correlated with the number of red scars (Pearson correlation, N = 74, r = 0.92, P < 0.001). Mean litter size at birth (Table S1) also differed significantly among the four types of crosses (ANOVA, F_{3,70} = 14.67, P < 0.001). While yellow scars were found in females of all mating crosses (Fig. 3), indicating prenatal embryonic mortality, there were significant differences between mating crosses (F_{3,70} = 16.73, P < 0.001). More yellow scars were found in +/t × +/t crosses, averaging 3.09 (±0.37 SE) (Fig. 3; t > 5.31 for all comparisons, P < 0.001). Compared with all other crosses combined, which averaged 0.67 (±0.16 SE), an excess of 2.42 yellow scars was found in +/t × +/t matings.
Overall, 79.3% of mating crosses were fecund (yielded offspring). Fecundity, however, did not differ according to type of mating cross (χ^{2}_{3,127} = 1.85, P = 0.603). We also tested for a possible advantage to +/t females mated to +/t males – with a smaller litter size, they might give birth sooner. However, neither litter size nor mating cross predicted time to birth (F_{4,119} = 0.35, P = 0.842). Furthermore, using data from experiment 1, we tested for a difference in pup survival until weaning from different crosses and found no difference among crosses (binomial GLM, χ^{2}_{3,43} = 0.16, P = 0.999).
The cost of genetic incompatibility to +/t females can be calculated by comparison of litter sizes at birth, of the number of red scars, or of yellow scars. Counting litter sizes after birth gave an estimation of reduction in litter size of 40.8% (combining experiments 1 and 2; 95% CI 30.6–50.8). Comparing red scars, which is a better indicator of the number of pups to which a female gave birth, the reduction is similar at 40.3%. Finally, from the excess of yellow scars in +/t × +/t matings compared to the average for all other matings, the reduction in litter size is estimated to be 32.0%. From the male +/t point of view, the cost of mating with a +/t female compared with a +/+ female was a litter size loss of 37.5% (combining experiments 1 and 2; 95% CI 26.8–48.2). From comparison of red scars, the reduction was 36.9%, while the estimate based on yellow scars is the same as for females.
Drive
Pups from the laboratory crosses were genotyped to estimate the degree of drive associated with the t haplotype. As expected, no pups were homozygous for t. Transmission ratios varied with the type of mating cross (GLM, χ^{2}_{2,91} = 72.15, P < 0.001; Fig. 4). When +/t males and +/+ females were crossed, 89.7% of 175 offspring from 30 litters inherited the t. In crosses between +/t males and +/t females, 78.6% of 109 offspring that were born in 32 litters inherited the t haplotype, a significantly lower proportion than in the previous cross (GLM, t = 2.18, P = 0.032). In the reciprocal cross of +/+ male and +/t female, 51.8% of 203 offspring from 32 litters inherited the t haplotype, which was not different from 0.5 (exact binomial test, P = 0.674).
We further investigated the difference in transmission rate of the t haplotype inherited through the male, depending on the female genetic background. Based on 89.7% transmission to offspring in +/t male and +/+ female crosses, and 50% transmission in crosses of +/+ male and +/t females, the proportion of +/t offspring in crosses of +/t males and +/t females was expected to be 90.7%, taking into account t/t lethality. The 95% confidence interval around the estimate of the observed value (78.6%) does not overlap this expected value (Fig. 4).
Drive in the wild population
The degree of transmission bias of the t in litters from the wild population sired by a single genotype was estimated, and analyzed in a binary GLM, using litters as the main unit of analysis. The overall model was significant (F_{2,42} = 17.45, P = 0.003), which was due to a difference between crosses of +/t males mated with +/+ females, in which 24/28 (85.7%) offspring from seven litters inherited the t, compared to crosses of +/+ males mated with +/t females, in which 33/76 (43.4%) offspring from 25 litters inherited the t (t_{1,42} = 3.03, P = 0.004). The proportion of +/t offspring detected in these crosses did not differ from that of crosses of +/t males mated with +/t females, in which 26/40 (65.0%) of offspring in 13 litters inherited the t (GLM, t_{1,42} = 1.60, P = 0.117 for the former and t_{1,42} = −1.89, P = 0.066 for the latter). Transmission rate of the t was similar to the laboratory, for +/t × +/t crosses (Pearson's χ^{2} = 2.28, df = 1, P = 0.131), +/t males with +/+ females (χ^{2} = 0.09, df = 1, P = 0.761) and for +/+ males with +/t females (χ^{2} = 0.64, df = 1, P = 0.425).
MHC genotyping
MHC class II loci Aα and Eβ and the MHC class Ilinked microsatellite D17Mit28 showed similar levels of variation with 4–5 alleles per locus (summarized in Table 2). At each locus, +/t mice had the same allele which was expressed only as a heterozygote variant and was not found in any +/+ mice. Because the t haplotype is lethal in its homozygote form, heterozygote excess is anticipated in +/t mice. This was confirmed by significant heterozygote excess at the three MHC loci (Fisher's exact tests, P ≤ 0.004). +/+ mice exhibited a significant heterozygote deficiency at D17Mit 28 (Fisher's exact test, P = 0.001) and Eβ (Fisher's exact test, P = 0.005), but not at Aα (Fisher's exact test, P = 0.123).
Table 2. Allele frequency and heterozygosity of MHC loci among 14 +/t and 15 +/+ miceAllele  Locus 

D17Mit28  Aα  Eβ 

+/+  +/t  +/+  +/t  +/+  +/t 


1 ( t )  0.000  0.500  0.000  0.500  0.000  0.500 
2  0.233  0.036  0.167  0.036  0.067  0.107 
3  0.067  0.107  0.067  0.000  0.233  0.036 
4  0.367  0.250  0.768  0.464  0.700  0.357 
5  0.333  0.107     
H _{Obs}  0.400  1.000  0.267  1.000  0.200  1.000 
H _{Exp}  0.720  0.688  0.393  0.553  0.467  0.632 
Exact tests 
P  <0.001   <0.001   <0.001  
Sequences of Aα and Eβ indicate that all alleles were unique not only with respect to nucleotide sequences but also at the amino acid level (Fig. 7). Sites involved in antigen binding showed the highest variation with an average of 5.7–7.2 differences in amino acid sequence between alleles compared with 1.8–2.0 differences at neutral sites, indicating that all alleles found might differ in their antigen binding capacities.
These mice were also genotyped for 21 neutral microsatellites. Similar to the pattern observed at the MHC loci, a significant deficit of heterozygotes was found but only for +/+ mice (Table 3, Fisher's exact test, P < 0.001) but the difference between observed and expected heterozygosity dropped to a fourth of the value found at MHC loci. To clarify the selective processes acting on the different loci we excluded the 12 founder individuals, and analyzed the remaining random sample of their descendants (N = 17). For MHC markers, the excess of heterozygotes in +/t remained significant, but a deviation from HW equilibrium was no longer observed in the +/+ population. Neutral markers showed the lowest differences between observed and expected heterozygosity in this sample with no deviation from HW equilibrium, however, a significant heterozygote excess was found in the +/t population (Table 3). Differentiation between the +/t and +/+ population was significant (Fisher's exact tests, P < 0.001) in all analyzed comparisons.
Table 3. Variability in three MHC loci and 21 neutral microsatellites for +/t and +/+ individuals, and divided into the subset of 12 founders and the subset of 17 randomly chosen individualsType of loci  N individuals  N loci  Mean N alleles/locus  H _{Obs}  H _{Exp}  H_{Obs}–H_{Exp}  P 

MHC 
+/+  15  3  3.33  0.289  0.526  −0.238  0.001 
+/t  14  3  4.0  1.000  0.624  0.376  0.000 
MHC – excluding founders 
+/+  7  3  3.00  0.286  0.414  −0.128  0.080 
+/t  10  3  3.33  1.000  0.605  0.394  0.000 
Microsatellites 
+/+  15  21  5.29  0.643  0.728  −0.085  0.000 
+/t  14  21  4.67  0.694  0.615  0.079  0.995 
Microsatellites – excluding founders 
+/+  7  21  4.14  0.703  0.720  −0.017  0.955 
+/t  10  21  4.29  0.705  0.631  0.073  0.999 