DISCUSSION

The effect of X-inactivation on recombination and TRD: We observed variability in the level of TRD at Om among the offspring of (PGK × DDK)F₁ females. We tested the hypothesis that this variability was due to epigenetic differences between the females in X-inactivation phenotype. Among 535 offspring of 38 females, we found that the level of TRD at Om was correlated with the X-inactivation phenotype of the female (Figure 3). Because TRD at Om is the result of MII meiotic drive and MII meiotic drive is affected by recombination (Pardo-Manuel de Villenaet al. 2000a), we examined whether a female's X-inactivation phenotype might affect TRD by altering the overall level of meiotic recombination in her oocytes. Analysis of available genotype information on all 535 offspring at X chromosome loci (de la Casa-Esperónet al. 2000) indicated that the level of recombination on chromosome X was correlated with the X-inactivation phenotype of the F₁ females. This observation argued that any effect on recombination was not specific to chromosome 11. Therefore, we extended the haplotype analysis to all 20 chromosomes, using only the offspring of females from the two extremes of the X-inactivation distribution. Comparison of estimates of tetrad distributions (E₁, E₂, E₃) over all 20 chromosomes indicates that there are significantly more crossovers in oocytes from more DDK X-active females than in oocytes of less DDK X-active females.

We interpret these data to mean that an allelic difference at an X-linked locus (or loci), subject to X-inactivation, is responsible for the observed difference in recombination. The expression of a DDK allele at this locus increases levels of meiotic recombination relative to the PGK allele and, as a consequence, also increases levels of MII meiotic drive favoring the inheritance of DDK alleles at Om (Pardo-Manuel de Villenaet al. 2000a).

Our analysis assumes that the X-inactivation phenotype measured on tail biopsies from each female is representative of the population of primordial germ cells that gave rise to the offspring in which recombination was analyzed. Although there is no way to conduct a post hoc analysis of this assumption, previous studies have shown similar X-inactivation phenotypes among different tissues of a given individual, while larger differences may be observed between females with the same genotype (Plengeet al. 2000). In any case, if the X-inactivation phenotype of a female does have an effect on recombination, misclassification of a female is expected to obscure such an effect rather than enhance it.

One descriptive measure of the level of difference between the two groups of females is that there are ∼20% more multiple crossover tetrads (E₂ + E₃) in more DDK X-active females than in less DDK X-active females. Although a 20% increase in the fraction of tetrads having more than one crossover may not seem large, one must keep in mind that the number and distribution of recombination events during meiosis is very well controlled. More than three crossovers are rarely observed (reviewed in Kleckner 1996; Roeder 1997) because the presence of one crossover interferes with the occurrence of additional events on the same chromatid in a distance-dependent manner [crossover or chiasma interference (Sturtevant 1915; Muller 1916)]. While chiasma interference suppresses the number of crossovers on a chromosome, multiple studies also indicate that a minimum of one crossover per chromosome arm is required to ensure proper chromosome segregation (Hulten 1974; Dutrillaux 1986; Pardo-Manuel de Villena and Sapienza 2001). In other words, one crossover is required but more than one is suppressed by chiasma interference. In fact, >80% of the total variance in recombination rates among mammals can be explained by a requirement for one recombination event per chromosome arm (Dutrillaux 1986; Pardo-Manuel de Villena and Sapienza 2001). In this light, a 20% difference between the two groups of females does not seem insignificant.

Implications of the tetrad model for testing differences in recombination: A major premise of our analysis is that an increase in the level of recombination is accomplished by increasing the proportion of tetrads that have multiple crossovers. In practice this is accomplished by increasing the proportion of E₂ tetrads at the expense of E₁ tetrads. This results in the counterintuitive prediction that the largest class of chromosome haplotype recovered (single recombinants) is uninformative for discerning differences in recombination because one-half of the products of both E₁ and E₂ tetrads are single recombinants (Figure 1). If the great majority of tetrads are E₁ and E₂, no difference in the number of single recombinants recovered is predicted (or observed; Table 1) by shifting E₁ tetrads to E₂ tetrads. An increase in the overall level of recombination as a result of shifting tetrads from the E₁ to the E₂ class will appear as the loss of chromosomes from the nonrecombinant class and the addition of an equal number of chromosomes to the double recombinant class (Table 1).

In a related vein, we note that our estimates of E₃ tetrad frequency in the two groups of females (Table 2) are counter to what might be expected; i.e., there is a larger estimated fraction of E₃ tetrads in the less DDK X-active females. However, the significance of this observation, in both the statistical and biological senses, is unclear. We recovered only 12 triple recombinant chromosomes (and 3 of these occurred within a single individual) among >3200 examined. [We note, however, that recovery of this number does not appear to be low. In comparison, Broman et al. (2002) identified only 3 triple recombinant chromosomes in an experiment of similar size (3760 chromosomes examined) and with much higher density map coverage.] Both the relative rarity of triple recombinant chromosomes and the large uncertainty associated with the estimation of E₃ tetrads (because it is based on recovery of triple recombinant chromosomes in one-eighth of the products of all E₃ tetrads) make it difficult to draw any robust conclusions about the effect of X-inactivation on the proportion of E₃ tetrads. In spite of the errors resulting from including triple recombinants in our tetrad estimation, recombination differences are still statistically significant between less DDK X-active and more DDK X-active females.

A role for the X chromosome in sex-specific recombination rates? While some studies have focused their interest on variability in recombination over particular intervals in different organisms (Reeveset al. 1991; Yuet al. 1996; Simianeret al. 1997), our study addresses the control of overall recombination. Studies on genome-wide recombination in humans have shown interindividual variation in female recombination (Bromanet al. 1998), as we observe in the mouse. Additional differences in the number and distribution of recombination events have been described between different chromosomes (Kaback et al. 1992, 1999; Lawrieet al. 1995; Andersonet al. 1999), between different inbred mouse strains (Speed 1977; Andersonet al. 1999), and between male and female meiosis (Polani 1972; Speed 1977; Jagiello and Fang 1979; Lawrieet al. 1995). The latter observation has led some authors to propose that levels of recombination are affected by one or more genes on the X chromosome (Lynnet al. 2000; Bromanet al. 2002). Because the only X chromosome in mammalian spermatocytes is thought to be inactive, in contrast to the two active X chromosomes in oocytes, the product(s) of X-linked gene(s) could be responsible for changes in the baseline rate of recombination.

How does X-inactivation affect a process that takes place in a cell with two active X chromosomes? An important question that remains unanswered by our study concerns the mechanism by which the X-inactivation phenotype of a female may affect the expression of an X-linked gene involved in meiotic recombination. X chromosome reactivation occurs in the germ cells of the fetal ovary at or shortly before the onset of meiosis (Kratzer and Chapman 1981; McLaren 1983). Although the reported timing of these two processes varies among different studies (Kratzer and Chapman 1981; Monk and McLaren 1981; Tamet al. 1994), it seems plausible that the inactive X chromosome must be reactivated before pairing and recombination (Kratzer and Chapman 1981). If the gene product affecting recombination levels is synthesized by the oocyte itself, it is possible that a delay in the reactivation of the allele that was inactive in the primordial germ cell, or a comparatively “long-lived” product of the single active allele prior to X chromosome reactivation, would be compatible with our observations. Another possibility is that the product of the X-linked gene is not expressed by the oocyte itself but is provided by adjacent somatic cells. The somatic pattern of X-inactivation of an individual female would, in such a case, be reflected in her oocytes by the relative amount of product from each allele that reached the oocytes at the time of meiotic recombination.

Some potential consequences of an unusual maternal effect: The phenomenon we have described indicates that a mother may have unexpected effects on the transmission of genes to her offspring. In this regard, it is worth noting that the definition of a maternal effect is that the genotype of the mother determines the phenotype of the offspring. In the situation we have described in this report, it is the phenotype of the mother (the fraction of her cells with a particular X chromosome active) that influences the genotype of her offspring. Although maternal effects have been described in many organisms, we are unaware of any other examples of this type.

The epigenetic effect that we observe on the level of female meiotic recombination has the potential to alter the behavior of genes in natural populations in several ways: (1) alteration of allele frequencies at loci subject to meiotic drive (Buckleret al. 1999; Pardo-Manuel de Villenaet al. 2000a); (2) alteration of the linkage relationship between alleles at loci that have been coselected or co-adapted during evolution (Wu and Palopoli 1994); and (3) effects on the level of chromosome nondisjunction during meiosis, with possibly widespread consequences (Jacobs and Hassold 1995; Warburton 1997; Brown et al. 2000).