The X chromosome plays a central role in the evolution of reproductive isolation, but few studies have examined the genetic basis of X-linked incompatibilities during the early stages of speciation. We report the results of a large experiment focused on the reciprocal introgression of the X chromosome between two species of house mice, Mus musculus and M. domesticus. Introgression of the M. musculus X chromosome into a wild-derived M. domesticus genetic background produced male-limited sterility, qualitatively consistent with previous experiments using classic inbred strains to represent M. domesticus. The genetic basis of sterility involved a minimum of four X-linked factors. The phenotypic effects of major sterility QTL were largely additive and resulted in complete sterility when combined. No sterility factors were uncovered on the M. domesticus X chromosome. Overall, these results revealed a complex and asymmetric genetic basis to X-linked hybrid male sterility during the early stages of speciation in mice. Combined with data from previous studies, we identify one relatively narrow interval on the M. musculus X chromosome involved in hybrid male sterility. Only a handful of spermatogenic genes are within this region, including one of the most rapidly evolving genes on the mouse X chromosome.
THE X chromosome often plays a central role in the evolution of reproductive isolation between animal species with sex chromosomes. For example, studies in Drosophila have shown that substitution of the X chromosome between species has a larger impact on hybrid male fertility than introgression of autosomes (Dobzhansky 1936; Orr 1987; Coyne and Orr 1989; Masly and Presgraves 2007). Likewise, levels of gene flow between naturally hybridizing species tend to be lower on the X chromosome relative to the autosomes (Tucker et al. 1992; Dod et al. 1993; Besansky et al. 2003), suggesting selection against X-linked incompatibilities. The contribution of the X chromosome seems particularly strong when considering reproductive isolation between closely related species. Hybrid incompatibilities between incipient species are often sex limited and manifest in the heterogametic sex first (i.e., Haldane's rule; Haldane 1922; Laurie 1997; Orr 1997) due, in part, to the exposure of hemizygous X-linked incompatibilities (Muller 1942; Turelli and Orr 1995; Turelli and Orr 2000). Nevertheless, we know relatively little about the specific details of the genetic basis and architecture of X-linked reproductive isolation during the early stages of speciation.
Hybrid sterility and other forms of intrinsic postzygotic isolation are thought to be a consequence of deleterious epistasis between interacting genes that have diverged between species [i.e., Dobzhansky–Muller incompatibilities (D–M incompatibilities); Bateson 1909; Dobzhansky 1937; Muller 1942]. Hybrid male sterility is especially relevant to the onset of intrinsic isolation between species with heterogametic males because it has been shown to evolve much more quickly than female sterility or inviability in either sex (Wu 1992; True et al. 1996; Coyne and Orr 1997; Tao et al. 2003; Wu and Ting 2004; Masly and Presgraves 2007). Multiple factors may contribute to the rapid evolution of hybrid male sterility, including the exposure of recessive X-linked D–M incompatibilities (Muller 1942; Turelli and Orr 1995; Turelli and Orr 2000), the rapid evolution of male-limited traits (Wu and Davis 1993; Wu et al. 1996; Presgraves and Orr 1998), and a higher rate of positive selection on the X chromosome (Charlesworth et al. 1987). These forces might quickly lead to a complex genetic architecture for hybrid male sterility, as seen in Drosophila by both the rapid accumulation of X-linked sterility factors (True et al. 1996; Tao et al. 2003; Masly and Presgraves 2007) and the rapid evolution of complex epistasis within sets of interacting male sterility loci (Davis and Wu 1996; Sawamura et al. 2004).
House mice are well suited to address the genetic architecture of hybrid male sterility during the early stages of speciation. House mice are a well-studied model organism (Dietrich et al. 1996; Mouse Genome Sequencing Consortium 2002; Su et al. 2004; Shifman et al. 2006) consisting of at least three closely related species: Mus domesticus, M. musculus, and M. castaneus (also referred to as subspecies of M. musculus: M. m. domesticus, M. m. musculus, and M. m. castaneus). Divergence of these lineages initiated ∼500,000 years ago (She et al. 1990; Boursot et al. 1993). M. castaneus is distributed throughout southeastern Asia, M. musculus occurs in eastern Europe and Asia, and M. domesticus occurs in western Europe, the Middle East, Africa, Australia, and the Americas (Boursot et al. 1993). House mice live in close association with humans, and their ranges have been extended through human-mediated dispersal. All three pairwise combinations of these taxa hybridize to some extent in nature along established hybrid zones (M. domesticus–M. musculus; Boursot et al. 1993) or in mosaic or admixed populations (M. castaneus–M. musculus, Boursot et al. 1993; M. domesticus–M. castaneus; Orth et al. 1998).
The X chromosome is thought to play a primary role in isolation between M. domesticus and M. musculus. In the European hybrid zone, the X chromosome shows lower levels of gene flow relative to the autosomes (Tucker et al. 1992; Dod et al. 1993, 2005; Munclinger et al. 2002; Macholán et al. 2007; Teeter et al. 2008). Payseur et al. (2004) surveyed 13 X-linked loci across a transect in southern Germany and found significantly reduced introgression at several adjacent markers located in the middle of the chromosome, suggesting that this genomic region harbors at least one hybrid incompatibility. Moreover, some laboratory crosses between wild-derived inbred strains of M. domesticus and M. musculus result in asymmetric F1 hybrid male sterility (Britton-Davidian et al. 2005; Good et al. 2008). In these asymmetric crosses, sterility only occurs when the maternal strain is of M. musculus origin. This pattern is consistent with hybrid sterility caused by epistatic interactions between the hemizygous M. musculus X-linked or mitochondrial mutations and dominant autosomal and/or hemizygous Y-linked mutations in M. domesticus (Britton-Davidian et al. 2005; Good et al. 2008).
Two recent studies offer further insights into the genetic basis of X-linked hybrid male sterility between M. musculus and M. domesticus. First, Storchová et al. (2004) substituted the X chromosome of M. musculus (wild-derived strain PWD/Ph) onto the genomic background of M. domesticus as represented by the classic laboratory strain C57BL/6J. The resulting X (heterosomic) consomic females were normal while consomic males were sterile with significantly smaller testes, lower sperm count, and abnormal sperm head morphology (Storchová et al. 2004). Second, Oka et al. (2004) observed male-limited sterility when they introgressed the X chromosome of M. m. molossinus (laboratory strain MSM/Ms) into C57BL/6J. M. m. molossinus is derived from an ancient hybridization event between M. castaneus and M. musculus. Likewise, C57BL/6J and other classic laboratory strains of house mice include substantial genetic contributions from both M. musculus and M. castaneus (∼70–90% M. domesticus; Wade et al. 2002; Frazer et al. 2007; Yang et al. 2007). Both Storchová et al. (2004) and Oka et al. (2004) mapped multiple X-linked quantitative trait loci (QTL) contributing to hybrid male sterility in later backcross generations. QTL confidence intervals overlapped between the studies in some instances, including the identification of a major sterility QTL in the same general central region of the X chromosome that shows low introgression across the southern Germany transect (Payseur et al. 2004). However, since the genetic composition of classic inbred strains appears to have been strongly influenced by selection against hybrid incompatibilities (Payseur and Hoekstra 2005), the relevance of incompatibilities mapped using C57BL/6J to reproductive isolation in the wild is uncertain.
Here, we report the results of a large experiment focused on the introgression of the X chromosome between M. musculus and M. domesticus. Our ultimate goal is to elucidate the genetic basis and architecture of X-linked hybrid male sterility between these species. Our results extend previous studies on X-linked reproductive isolation in mice in three important ways. First, our introgression design is reciprocal. These data provide the first direct test for hybrid incompatibilities on the M. domesticus X chromosome and allow us to fully evaluate basic predictions of the D–M model of epistatic incompatibilities (Muller 1942; Coyne and Orr 2004). Second, this is the first study to examine the genetic basis of X-linked hybrid sterility using inbred strains exclusively derived from wild mice. Third, whereas previous studies introgressed the entire X chromosome, we primarily focused on smaller chromosomal segments. By selecting recombinants, we evaluated the independent contributions of specific introgressed regions spanning the X chromosome.
MATERIALS AND METHODS
Two breeding colonies were established using wild-derived inbred strains purchased from the Jackson Laboratory (http://www.jax.org). We used the LEWES/EiJ strain to represent M. domesticus (hereafter domesticusLEWES). This strain was originally isolated from natural populations in Lewes, Delaware, and is thought to descend from recent human-mediated dispersal of M. domesticus from western Europe. M. domesticus from eastern North America are genetically indistinguishable from European M. domesticus in allozyme studies (Selander et al. 1969a,b). For M. musculus, we used the PWK/PhJ strain isolated from the central region of the Czech Republic (hereafter musculusPWK, Gregorová and Forejt 2000). In genetic analyses both strains cluster most closely with other wild-derived strains of M. musculus and M. domesticus, respectively (Petkov et al. 2004) and have a standard house mouse karyotype (2n = 40). Recently, some putative cases of introgression between M. domesticus and M. musculus have been reported for a few standard wild-derived strains of mice, including musculusPWK (Yang et al. 2007). The contribution of M. domesticus to the genome of musculusPWK has not been quantified, but levels of introgression are likely to be fairly low on the basis of available data (Gregorová and Forejt 2000). All mice were maintained in accordance with the Institutional Animal Care and Use Committee regulations at the University of Arizona Central Animal Facility.
We performed two parallel introgression experiments using a standard backcross design (Figure 1). In experiment one, three genomic regions spanning the M. musculusPWK X chromosome were introgressed into the M. domesticusLEWES autosomal background. Following an initial intercross between ♀ musculusPWK × ♂ domesticusLEWES, F1 female offspring were backcrossed to ♂ domesticusLEWES for a total of nine generations (i.e., to generation N10). Each generation female offspring were screened at 18 diagnostic X-linked microsatellite markers (see below) and females heterozygous for at least one of the three targeted regions were used for the next backcross generation. In experiment two, we repeated this general design to introgress the same three genomic regions of the M. domesticusLEWES X chromosome into the M. musculusPWK genomic background. A total of seven backcross generations (i.e., to generation N8) were performed in experiment two.
Data from previous studies were used to guide our decision on how to partition the X chromosome for our introgression experiments. Both genetic mapping (Oka et al. 2004; Storchová et al. 2004) and hybrid zone data (Payseur et al. 2004) suggest that the middle of the X chromosome is involved in reproductive isolation, including at least one QTL of major effect on hybrid male sterility, Hstx1, that maps to near marker DXMit119 (68.7 Mb, 29.5 cM; Storchová et al. 2004). We targeted three overlapping regions corresponding to the proximal (marker interval 6.8–62.3 Mb), central (62.3–100.3 Mb), and distal (100.3–162.6 Mb) regions of the X chromosome.
Treatment of experimental males:
Males were weaned into same-sex sibling groups of up to four and split into individual cages at ∼60 days old. All males were sacrificed as adults (all males ≥70 days old and ≥ 11.0 g body weight).
Quantification of male reproductive phenotypes:
While the ability to sire offspring may be taken as the most direct assessment of male fertility, this criterion provides little insight into the underlying phenotypic basis of reproductive failure. To provide a broad assessment of male reproductive condition we considered several general male reproductive phenotypes, including testis weight, sperm count, sperm motility, sperm head morphology, and seminal vesicle weight. In addition, we examined the reproductive success of some males using two different mating assays (see below).
Our methods for quantifying sperm count and motility have been described in detail previously (Good et al. 2008). Briefly, we evaluated sperm motility and count using a Makler counting chamber (Sefi-Medical Instruments, Haifa, Israel) and a light microscope at 200× magnification. Sperm from the caudal epididymides were incubated at 37° in modified phosphate-buffered saline for 10 min. Following incubation, 5 μl of mixed sperm suspension were transferred to the Makler chamber and the number of motile and immotile sperm was determined over a fixed area and observation time. Sperm counts were estimated with a Makler chamber using heat-shocked aliquots of the same sperm suspension.
To examine sperm head morphology, 25 μl of sperm suspension were spread onto a microscope slide and allowed to air dry. The slides were then fixed using a 1% acetic acid solution in 95% ethanol for two min and allowed to dry. Fixed preparations were then stained in 1% Eosin Yellow (Sigma) for five min, rinsed three times in 70% ethanol, dried, and mounted with 30 μl of Permount (Fisher). Mounted preparations were evaluated at 400× magnification using a phase contrast microscope. Sperm head morphology was classified into four different categories on the basis of the degree of abnormalities observed: (1) normal head morphology with a long apical hook (Russell et al. 1990), (2) heads with a shortened hook and/or flattened shape, (3) shortened heads with a very short or absent apical hook, and (4) a general class of severely abnormal shape. We considered a minimum of 100 individual sperm heads per male.
For a subset of males we examined fertility on the basis of two different mating assays. First, we considered fecundity on the basis of reproductive output following pairing with musculusPWK females. Males (>90 days old) were caged with a single female for 17 days and, when possible, each male was paired successively with at least two different females. Females were sacrificed and the number of embryos was determined. Second, we performed an in vivo assay of male fertilization potential by pairing males with females that were treated with hormones to induce ovulation (following Nagy et al. 2003). Because musculusPWK females do not respond to the hormone regiment, we used F1 females from a cross between musculusPWK females and M. musculus males from the CZECHII/EiJ wild-derived inbred strain (Jackson Laboratory, http://www.jax.org). Female progeny (5 weeks old) were intraperitoneally injected with 5 units of pregnant mare serum gonadotropin (PMSG, Calbiochem), followed 48 hr later with 5 units of human chorionic gonadotropin (hCG, Calbiochem). Ovulation occurs ∼12 hr after administration of hCG. Immediately following injection with hCG, we paired females individually with males. Females were sacrificed after 20 hr and cumulus masses were dissected from the oviducts and treated with hyaluronidase to isolate oocytes. We counted the number of oocytes with two pronuclei, two polar bodies, or neither. The two pronuclei stage occurs ∼6 hr after fusion of sperm and egg, while the two polar body stage occurs ∼8 hr after fusion (Nagy et al. 2003). Either stage was taken as evidence of fertilization. If we did not observe a vaginal plug, semen contents in the uterus, sperm around the oocytes, or any fertilized embryos, we assumed that mating did not occur.
Each generation we screened females at 18 simple sequence-length polymorphism (SSLP) markers across the X chromosome (Dietrich et al. 1994, 1996). The screened loci, physical locations in megabases (Mb) based on NCBI mouse build 36 (Ensembl release 46, August, 2007; www.ensembl.org), and the consensus genetic location in centimorgans (cM) based on the Mouse Genome Database (MGI 3.54; Eppig et al. 2005) were as follows: DXMit26 (6.8 Mb, 1.5 cM); DXMit137 (14.9 Mb, 5.7 cM); DXMit81 (32.6 Mb, 9.25 cM); DXMit50 (37.1 Mb, 14.1 cM); DXMit166 (47.9 Mb, 15.5 cM); DXMit141 (55.8 Mb, 19.0 cM); DXMit46 (62.3 Mb, 24.5 cM); DXMit119 (68.7 Mb, 29.5 cM); DXMit195 (81.8 Mb, 32.0 cM); DXMit128 (89.5 Mb, 34.7 cM); DXMit158 (100.3 Mb, 43.6 cM); DXMit65 (104.9 Mb, 48.5 cM); DXMit97 (115.5 Mb, 49.0 cM); DXMit173 (125.7 Mb, 50.5 cM); DXMit67 (136.7 Mb, 60.0 cM); DXMit153 (146.4 Mb, 62.2 cM); DXMit121 (156.8 Mb, 67.0 cM); and DXMit160 (162.6 Mb, 73.3 cM). Experimental males from the N2, N4, N6, N7, N8, N9, and N10 generations were also screened with the same 18 markers.
DNA was extracted from fresh tissue (tail tip, ear notch, or liver) using a PureGene DNA isolation kit (Gentra Systems, Minneapolis). We used 10-μl reactions and, when possible, loci were multiplexed into groups of two or three. The forward primer of each locus was labeled with 6-FAM or HEX. We followed standard polymerase chain reaction (PCR) protocols for these loci (Dietrich et al. 1994, 1996) and we optimized annealing temperature, primer concentration, and the number of amplification cycles as necessary. Primer sequences were retrieved directly from Ensembl (http://www.ensembl.org), and specific PCR conditions and primer combinations are provided in the supplemental material. Genotypes were resolved on an ABI 3100 (Applied Biosystems) with a 500 ROX ladder and scored using Genotyper v2.0-3.7 (Applied Biosystems).
A total of 346 males from late backcross generations were used to map quantitative trait loci for several male reproductive phenotypes. For experiment one, we considered 204 males from the N6–N8 generations. For experiment two, 142 males from the N4, N6, and N7 generations were used. Most males included in QTL analyses were resolved for their entire 18-locus genotype and all males were successfully genotyped at a minimum of 14 markers.
Two complementary approaches on the basis of interval mapping were implemented using the QTL Cartographer software package. First, we used the program Zmapqtl to identify X-linked QTL using composite interval mapping (CIM; Zeng 1994). This general method combines background markers as cofactors with standard interval mapping (Lander and Botstein 1989) to increase the power and resolution of QTL identification. For each phenotype, we calculated the likelihood L1 that a given position contributes an additive effect (a ≠ 0) and compared this model to the likelihood L0 of the null hypothesis of no additive effect (a = 0) at the test site. Significance of additive effects was determined using a standard likelihood ratio test statistic [LRS = −2ln(L0/L1)] and critical values for the experiment were calculated with Zmapqtl using 1000 permutations (Churchill and Doerge 1994). The most informative markers for controlling genetic background were selected on the basis of the results of a forward stepwise regression conducted with SRmapqtl. All analyses were performed using a window size of 10 cM and a walking distance of 1 cM. Dominance effects could not be evaluated because males were hemizygous for all X-linked markers.
Following CIM analyses, we estimated a multiple QTL model using multiple interval mapping (MIM; Kao et al. 1999). Our MIM model was constructed using a stepwise approach (Kao et al. 1999; Zeng et al. 2000): (1) an initial model of multiple putative QTL was defined on the basis of the results of CIM; (2) the significance of individual QTL in the model was evaluated, and QTL with LRS scores below the CIM experimentwise critical values of α = 0.05 were removed; (3) positions of significant QTL were refined; (4) the X chromosome was scanned for additional additive QTL; and (5) pairs of all identified QTL were tested for significant epistatic interactions using a forward stepwise search. In addition to CIM and MIM analyses, we also applied simple linear regression and standard interval mapping (IM; Lander and Botstein 1989) to our data. Because our inferences were generally corroborated by these simpler models, albeit with less resolution, we focus on the results of CIM and MIM. A graphical comparison of results for CIM and IM are provided in supplemental Figure 1.
Analysis of candidate genes on the X chromosome:
We evaluated potential candidate genes for hybrid male sterility using a variety of sources. X-linked protein-coding genes were identified from Ensembl (release 46, August 2007; www.ensembl.org) on the basis of NCBI mouse build 36. Gene expression during spermatogenesis was based on data from a recent Affymetrix microarray analysis (Affymetrix 430 2.0) on four distinct classes of germ cells: premeiotic type A and B spermatogonia, pachytene spermatocytes, and postmeiotic round spermatids (Namekawa et al. 2006). Affymetrix probes were matched to protein-coding genes using the Ensembl Biomart tool. Testis specificity was calculated as the amount of testis expression relative to the median expression in a panel of 61 mouse tissues (GNF Symatlas, symatlas.gnf.org; Su et al. 2004).
Male reproductive phenotypes in the parental inbred strains:
Data from musculusPWK and domesticusLEWES for a suite of standard male reproductive phenotypes are presented in Table 1. In general, musculusPWK and domesticusLEWES were similar in overall body weight, sperm count, sperm motility, and paired seminal vesicle weight. A significant reduction in testis weight was observed in domesticusLEWES (Wilcoxon P < 0.001). This difference was strain specific and is not related to divergence between these species (Good et al. 2008). In general, testis weight and sperm count were both significantly correlated with body weight in both strains (Good et al. 2008). We used two different approaches to account for this correlation. First, for all comparisons we report testis weight relative to body weight (relative testis weight, RTW, milligrams of testis per gram of body weight). Second, in all quantitative genetic analyses we used residual trait scores on the basis of a standard least squares regression on body weight.
Asymmetric sterility in F1 hybrid males:
A previous study showed that crosses between a female musculusPWK and a male domesticusLEWES yield hybrid male offspring with abnormal spermatogenesis while the reciprocal cross produces reproductively normal F1 males (Good et al. 2008). These abnormal hybrid males were characterized by significant reductions in testis weight (Table 2) and sperm count (see also Table 2 in Good et al. 2008). Histological analyses revealed primarily postmeiotic abnormalities in spermatogenesis with a general reduction in the number of maturing spermatids (Good et al. 2008). Because the reciprocal F1 hybrids shared the same autosomal genotype, these data suggested that one or more genes on the M. musculus X chromosome, the M. musculus mitochondrial DNA, and/or the M. domesticus Y chromosome influenced spermatogenesis in F1 hybrid males. A reduction in testis weight was observed in both young males (60-day old) and older males (∼120-day old), demonstrating that this difference does not represent a developmental delay (Table 2).
Experiment 1: introgression of the musculusPWK X chromosome into domesticusLEWES
Introgression of the M. musculusPWK X chromosome into the M. domesticusLEWES autosomal background caused severe reductions in male fertility. We considered several male reproductive phenotypes in a total of 306 males sampled across nine backcross generations (N2–N10). We begin by reporting the general qualitative patterns observed during the early backcross generations and then provide a detailed account of the genetic basis of sterility in later backcross generations.
X-linked male sterility during the early backcross generations:
We examined X chromosome genotypes and several male reproductive phenotypes for 35 males from the first backcross generation (N2). These N2 males demonstrated some recovery of testis weight (RTW) relative to F1 males from the ♀ musculusPWK × ♂ domesticusLEWES (RTW = 4.2 mg/g, SD = 0.7; Wilcoxon P < 0.0001). Only 3 N2 males showed dramatic reductions in testis size (average RTW = 2.5 mg/g, SD = 0.6) and all 3 were hemizygous for the entire or most of the musculusPWK X chromosome. However, we did observe some variation in X-linked sterility in this first backcross generation, including multiple males with large introgressions of the musculusPWK X chromosome that had relatively normal fertility parameters. By the N4 generation (n = 55 total males, RTW = 4.2 mg/g, SD = 0.9), males with large introgressions of the musculusPWK X chromosome (>100 Mb of musculusPWK X chromosome, n = 8) were characterized by reduced testis weight compared to control backcross males (n = 12) with a domesticusLEWES X chromosome (Wilcoxon P = 0.012). There were two striking exceptions to this trend: the two most severely abnormal males (testis weight <25 mg and no sperm in their caudal epididymides) in this generation both had a domesticusLEWES X chromosome. The cause of this is unclear, but may be due to particular autosomal interactions that were uncovered in this but not subsequent generations. Dramatic reductions in male fertility were not observed in males with the domesticusLEWES X chromosome genotype in any of the later backcross generations.
Multiple factors on the musculusPWK X chromosome contributed to reduced testis weight, sperm count, and abnormal sperm head morphology:
By the N6 generation none of the introgressed males in our experiment possessed a complete musculusPWK X chromosome. Nevertheless, it became apparent that multiple loci on the musculusPWK X chromosome were involved in male sterility and that the severity of reproductive disruption depended on the number of incompatibilities present in a given male. One simple pattern consistent with a multigenic basis to sterility was a significant negative linear correlation between the residual testis weight and the approximate length of the introgressed musculusPWK X chromosome, regardless of the location of the introgressed region (Figure 2; n = 202 N6–8 males, r2 = 0.45, P < 0.0001). Residual sperm count also showed a significant, albeit weaker, negative correlation with musculusPWK X chromosome introgression length (n = 198 males, r2 = 0.12, P < 0.0001). Body weight, seminal vesicle weight, and sperm motility were all uncorrelated with introgression length (all P > 0.40).
Of the three targeted chromosomal regions (Figure 1), introgression of the proximal portion of the musculusPWK X chromosome had the largest individual impact on male reproduction and resulted in a significant reduction in both testis weight and sperm count relative to control backcross males (Wilcoxon P < 0.01; Figure 3). Only slight reductions in testes weight and no reductions in sperm count were observed for males with either the central or distal musculusPWK X chromosome introgressions. Nevertheless, larger introgressions spanning the proximal, central, and distal portions of the musculusPWK X chromosome resulted in an even greater reduction in both testis weight and sperm count than introgression of the proximal segment alone (Figures 3).
Sperm head morphology was even more strongly influenced by the genotype of the X chromosome in the later backcross generations. In mice, a rounded head shape with a long curved apical hook defines normal sperm morphology (Russell et al. 1990). We observed normal head morphology for the majority of sperm from either parental strain (domesticusLEWES, musculusPWK) and for control backcross males with a domesticusLEWES X chromosome (Figure 4). In contrast, most males with large portions of the musculusPWK X chromosome had sperm with severely abnormal head morphology; smaller introgressions produced less severe abnormalities, including a noticeable shortening of the apical hook and/or a flattened, less rounded head shape (Figure 4). Overall, we observed a very significant correlation between the proportion of sperm with abnormal head morphology and musculusPWK X chromosome introgression length (arcsine square-root transformed, n = 174, r2 = 0.75, P < 0.0001). As with testis weight, the proximal half of the musculusPWK X chromosome had the strongest influence on sperm head morphology. Figure 4B shows the negative influence of progressively larger introgressions of the musculusPWK X chromosome (proximal to distal) on sperm head morphology. Males with ≥81.9 Mb of the proximal musculusPWK X chromosome produced essentially no sperm with normal head morphology (Figure 4B).
We localized four nonoverlapping intervals in recombinant males (0–37.1 Mb; 47.9–81.8 Mb; 89.5–125.7 Mb; and 125.7 Mb–telomere) that resulted in abnormal head morphology. Each interval represents the minimal genotype in each region that resulted in at least one male with proportionally fewer sperm with normal head morphology relative to the entire range of values observed in all control N6–8 males (range = 0.67–0.96 normal sperm, n = 61). Thus, at least four independent factors on the musculusPWK X chromosome contributed to abnormal sperm head morphology.
QTL mapping of sterility factors:
We used a combination of CIM and MIM to further resolve the number, location, and phenotypic effects of X-linked QTL contributing to reduced testis weight (n = 204), sperm count (n = 200), and abnormal sperm head morphology (n = 174) in males from the N6, N7, and N8 generations. Residual sperm count followed a normal distribution (Shapiro-Wilk W-test, P = 0.880). However, both testis weight and sperm head morphology were not normally distributed due to strong left skews. To reduce the influence of outliers, we performed a Box-Cox transformation on testis weight and an arcsine square-root transformation on sperm head morphology. In both cases the distributions were dramatically improved but still deviated from normality (both characters Shapiro-Wilk W-test, P < 0.03).
Residual testis weight, residual sperm count, and the proportion of abnormal sperm head morphology all showed evidence of multiple significant QTL on the basis of CIM. Figure 5 shows the likelihood of odds scores (LOD ≈ 0.217 likelihood ratio static) for all three traits as a function of genetic position on the X chromosome. For testis weight, LOD scores exceeded critical values (LOD = 2.82, α = 0.05) for much of the first third of the X chromosome (DXMit26–DXMit119) and between DXMit173 and DXMit67 on the distal end (Figure 5). The QTL with the largest effect on testis weight was estimated to occur near (position 18.5 cM) marker DXMit141 and explained 21.7% of the total phenotypic variance (LOD = 16.52). A similar pattern was found for sperm count. The QTL with the largest LOD score for sperm count (LOD = 10.69) was estimated to occur at position 15.1 cM and explained 21.9% of the total phenotypic variance.
The likelihood surface was more peaked in our analysis of sperm head morphology (proportion of abnormal sperm heads, arcsine square root). As with our analysis of testis weight and sperm count, the strongest influence was found for loci from the first third of the chromosome. Within this region, there were two distinct peaks and multiple peaked shoulders. The QTL with the largest influence on sperm head morphology accounted for 55.5% of the total phenotypic variation and was located at position 20.0 cM (LOD = 43.4). The central and distal end of the X chromosome also showed evidence of QTL for abnormal sperm head morphology, including one or more QTL near the center of the X chromosome where QTL were not observed for testis weight and sperm count. We also considered abnormal sperm head morphology using an index for each male on the basis of the proportion of different classes of abnormalities within a given male (index modified from Oka et al. 2004) and this alternative quantification produced very similar results in CIM analysis (analysis not shown).
Inspection of Figure 5 suggested that several of the QTL influencing testis weight, sperm count, and sperm morphology were nonindependent. We detected significant pairwise correlations between all three characters (all P < 0.0001; testis weight × sperm count, r2 = 0.26; testis weight × sperm head morphology, r2 = 0.50; sperm count × sperm head morphology, r2 = 0.18). The first principle component (PC1) of the covariance of these three traits accounted for 85.1% of the total variation and CIM analysis of PC1 produced qualitatively similar results (analysis not shown).
Resolution of the location and phenotypic effects of individual QTL identified by CIM was clearly influenced by the occurrence of multiple closely linked QTL. To help address this issue we constructed a multiple QTL model using MIM (Table 3 and Figure 5; Kao et al. 1999). In the MIM framework, R2 is the sum of the contributions of all singly significant QTL to the total genotypic variance divided by the total phenotypic variance and provides an estimate of broad sense heritability. Three QTL remained significant for testis weight, and heritability was moderately high (R2 = 0.631) while only two QTL remained significant for sperm count and heritability was fairly low (R2 = 0.241). For both traits, all QTL identified represented musculusPWK alleles that resulted in hybrid abnormalities and all loci of large effect (R2 > 0.10) were localized proximal to DXMit46 (24.5 cM, 62.3 Mb). By contrast, seven positions remained significant for abnormal sperm head morphology and these loci explained most of the phenotypic variation observed in the trait (R2 = 0.944). Three QTL of major effect were located proximal to DXMit119 (29.5 cM, 68.7 Mb) and one was found distal to DXMit97 (49.0 cM, 115.5 Mb). In addition, one locus contributed an additive effect in the opposite direction (Table 3), consistent with a positive influence of a musculusPWK allele within this region. We found no evidence for significant epistatic interactions between the identified QTL for any of the three characters.
Male fecundity and fertility of select genotypes:
We used two different mating assays to evaluate the relative reproductive success of males from three genotypic classes from the N9 or N10 generations: control (domesticusLEWES), proximal (<DXMit119, 29.5 cM, 68.7 Mb), and all major QTL (<DXMit153, 62.2 cM, 146.4 Mb; see Table 3). The average fecundity (measured as the average number of embryos produced per male) of proximal males was only marginally reduced relative to control backcross males (Wilcoxon P < 0.04; Table 4). However, we observed a dramatic and consistent reduction between these two genotypic classes for in vivo fertilization rate (Fisher's exact test, P < 0.0001; Table 4). Both males containing all major QTL from the musculusPWK X chromosome were completely sterile in our fecundity and fertilization assays and failed to sire offspring or fertilize eggs in repeated trials.
Experiment 2: introgression of the M. domesticusLEWES X chromosome into M. musculusPWK
Asymmetric sterility in the reciprocal F1 crosses between M. domesticusLEWES and M. musculusPWK suggests that the domesticusLEWES X chromosome was not involved in F1 hybrid male sterility (Good et al. 2008). Nevertheless, this asymmetric pattern does not preclude the existence of factors on the M. domesticusLEWES X chromosome contributing to hybrid male sterility because epistatic interactions between hemizygous X-linked loci and recessive autosomal incompatibilities are masked in the F1 generation. Therefore, we introgressed three genomic regions spanning the domesticusLEWES X chromosome onto the musculusPWK genomic background. We considered a total of 227 males in four backcross generations (N2, N4, N6, and N7). In contrast to experiment 1, we observed no introgressed hybrid males with strongly reduced reproductive parameters relative to the parental strains. However, M. musculusPWK had significantly larger testis than M. domesticusLEWES (Table 1) and our experiment identified two X-linked QTL contributing to differences in testis size between these strains.
X-linked loci contributed to differences in testis size between M.musculusPWK and M. domesticusLEWES:
The average RTW of male progeny from the first backcross generation (N2) was intermediate between M. domesticusLEWES and M. musculusPWK, increased across later backcross generations, and approached the values characteristic of M. musculusPWK (Table 5). To determine if X-linked loci contributed to differences in testis weight between M. domesticusLEWES and M. musculusPWK we genotyped 142 males from the N4, N6, and N7 generations. In these males, relative testis weight showed a significant negative correlation with the approximate length of the introgressed domesticusLEWES X chromosome (n = 142, r2 = 0.21, P < 0.0001), consistent with multiple X-linked loci contributing to differences in testis weight between the strains.
QTL mapping of X-linked factors influencing testis weight:
CIM and MIM analysis of testis weight (residuals of Box-Cox transformed values) of 142 males identified two major QTL in the central (34.7 cM, 34.0–44.6 2-LOD confidence interval) and distal (61.0 cM, 50.5–68.0) regions of the X chromosome, respectively (Figure 6). A third putative QTL was borderline significant (LOD = 2.55) on the basis of a critical value of LOD = 2.26 (α = 0.05, 1000 permutations) in CIM but was not significant in our MIM model. Overall, the additive effects of these two major QTL explained 46.8% of the variance in testis weight in our MIM model (R2 = 0.468), with the largest contribution coming from distal most QTL (61.0 cM, R2 = 0.278). We did not detect any significant epistatic interactions between these two QTL.
Introgression of the M. musculus X chromosome into a wild-derived M. domesticus genetic background produced male-limited sterility, consistent with previous experiments using classic inbred strains to represent M. domesticus. QTL analysis of late backcross generation males implicated multiple hybrid incompatibilities and thus a fairly complex genetic basis to X-linked sterility. In some instances our patterns were in agreement with previous mapping experiments, while in other cases we found evidence for apparently novel hybrid incompatibilities. No sterility factors were uncovered during the reciprocal introgression of a wild-derived M. domesticus X chromosome onto a M. musculus genetic background. However, we mapped two loci involved in differences in testis weight between the two wild-derived strains we considered, including one QTL coincident with a sterility factor on the M. musculus X chromosome. Below we discuss the broader significance of our findings in relation to the evolution of D–M incompatibilities during the early stages of speciation. We then contrast our results with previous studies looking at X-linked reproductive isolation between these species, including patterns of gene flow in the hybrid zone. Finally, we discuss specific X-linked genes that are potential candidates for hybrid male sterility in mice.
The evolution of X-linked hybrid male sterility:
The simplest form of the D–M model involves epistatic interactions between two loci with alternative alleles fixed between species (e.g., AAbb and aaBB). Under this scenario, the genetic basis of hybrid dysfunction will be asymmetric and arise from interactions between the A and B alleles (i.e., incompatibilities should not involve interactions between a and b; Muller 1942; Orr 1995). This prediction was uniformly supported in our reciprocal introgression experiments with asymmetric genetic interactions evident for all of the X-linked loci underlying hybrid male sterility between M. domesticus and M. musculus. These data provide strong support for the D–M model and add to data from other reciprocal introgression experiments in animals where individual sets of epistatic interactions have been shown to be asymmetric (e.g., Wu and Beckenbach 1983; Vigneault and Zouros 1986; Orr 1989; Wittbrodt et al. 1989; Johnson et al. 1992).
All of the D–M incompatibilities we identified in later backcross generations involved epistatic interactions between incompatibilities on the M. musculus X chromosome and the M. domesticus genomic background. F1 hybrid male sterility was also asymmetric in crosses between M. domesticusLEWES and M. musculusPWK (Table 2; Good et al. 2008). While asymmetric sterility at the F1 and later backcross generations could reflect the same incompatibilities, the genetic composition of hybrid males was fundamentally different at these two stages of the experiment. In particular, the effects of recessive autosomal hybrid incompatibilities will be masked in F1 males but exposed in later backcross generations. Thus, both the genetic basis of X-linked sterility and the overall sterility phenotype could change over the course of our experiment. We observed no evidence for a shift in the developmental timing of sterility. Hybrid male sterility in F1 males (♀ musculusPWK × ♂ domesticusLEWES) involves primarily postmeiotic disruptions in spermatogenesis, resulting in morphologically abnormal sperm, reduced testis size, and lower sperm count (Good et al. 2008) and a similar phenotype was manifest in N4 and later-generation males with large introgressions of the musculusPWK X chromosome (Figure 4B). However, we did observe some variation in patterns of sterility in the N2 generation that may reflect important differences in the genetic basis of sterility in F1 and later backcross generation males.
The occurrence of multiple linked sterility factors makes it difficult to resolve the total number of incompatibilities on the musculusPWK X chromosome. There were at least four nonoverlapping intervals on the X chromosome that contributed to abnormal sperm head morphology, and MIM analysis implicated a total of seven QTL for this trait (Figure 5, Table 3). Overall, QTL within these four regions made largely additive contributions to hybrid male sterility (Table 3), resulting in complete sterility when all major QTL were combined on the same genotype (see also Storchová et al. 2004). By contrast, males with the proximal portion of the musculusPWK X chromosome were still capable of siring relatively large litters (Table 4). While the majority of sperm produced by these proximal males presented some morphological aberrations, some normal sperm were also observed and abnormalities were generally much less severe than sterile males with larger introgressions (Figure 4B). Interestingly, we did detect a dramatic reduction in the fertilization potential of proximal males relative to control backcross males (Table 4). This discrepancy between a striking reduction in fertilization potential when paired with “super-ovulated” females and a weak reduction in male fecundity (i.e., the number of late-stage embryos produced) when paired with noninduced females suggests that overall litter size was not sperm limited in these benign laboratory conditions and that these subfertile males nevertheless ejaculate enough viable sperm to assure a normal litter size. Nevertheless, it seems likely that these subfertile mutations could strongly reduce male fitness in nature where females often mate with multiple males (Dean et al. 2006), generating intense postcopulatory sperm competition among males.
We did not detect significant epistatic interactions between QTL identified in our MIM models; however, we cannot rule out interactions among tightly linked loci within these regions. Indeed, we observed abnormal sperm head morphology in multiple males containing a relatively small interval in the center of the musculusPWK X chromosome, but we were unable to fine-scale map the phenotype with smaller introgressions (data not shown). Similar patterns have been observed when mapping X-linked sterility in Drosophila (Perez and Wu 1995; Davis and Wu 1996), and complex epistasis within linked regions may be a common pattern of hybrid male sterility (Sawamura et al. 2004). While our data do not speak to the number of interacting partners in specific D–M incompatibilities (but see Oka et al. 2007), X-linked hybrid male sterility in mice does appear to have a relatively complex, multigenic basis overall.
At least two other studies have mapped X-linked QTL underlying differences in testis weight between inbred strains of mice (Le Roy et al. 2001; Bolor et al. 2006), suggesting the X chromosome plays an important role in regulation of mouse spermatogenesis. The QTL that we have identified on the X chromosome further support this view. This is noteworthy for two reasons. First, the X chromosome in mammals comprises a fairly small portion of the overall genome (i.e., ∼4% of known protein-coding genes are X linked in mice) compared to some other species (e.g., 20–40% of the Drosophila genome is X linked). Second, like Drosophila, genes involved in spermatogenesis are proportionally less frequent on the mouse X chromosome than the autosomes due to transcriptional inactivation of most X-linked genes during and after meiosis (Khil et al. 2004). Therefore, X-linked spermatogenesis genes comprise a very small portion of the overall testis transcriptome (∼2.4%; Khil et al. 2004), especially when considering genes involved in postmeiotic spermiogenesis. Thus it seems likely that there are many additional sterility factors that are on autosomes. In Drosophila, X-linked male sterility factors are approximately two and a half times more frequent than autosomal incompatibilities (Tao et al. 2003), a pattern referred to as the “large X effect.” If a similar ratio occurs in mice, assuming a minimum of 4 X-linked sterility factors, we might expect on the order of 65 autosomal recessive hybrid male sterility factors. It is important to bear in mind, however, that a large X effect has not yet been demonstrated in mice, and this calculation is thus very rough. Two recent studies constructed large sets of consomic lines substituting individual chromosomes from M. m. molossinus (MSM/Ms; Takada et al. 2008) or M. musculus (PWK/Ph; Gregorová et al. 2008) on the background of the classic laboratory strain C57BL/6J. Both experiments uncovered putative incompatibilities involving most autosomes, with deleterious effects on strain reproduction and viability. The extent to which these results involve male sterility phenotypes has not been determined, and some evidence exists for postzygotic isolation between M. musculus and M. domesticus due to hybrid inviability (Sage et al. 1986) and female sterility (Britton-Davidian et al. 2005).
Comparison with previous studies on reproductive isolation in mice:
Two previous studies have focused on the genetic basis of X-linked sterility between M. musculus and M. domesticus (Oka et al. 2004; Storchová et al. 2004). Both studies reported male-limited sterility related to the introgression of the M. m. musculus (Storchová et al. 2004) or M. m. molossinus (a hybrid between M. musculus and M. castaneus; Oka et al. 2004) X chromosome onto the genetic background of the classic laboratory strain C57BL/6J. The degree of concordance between our data and these studies depends on which sterility-related phenotypes were considered. Neither of the previous studies detected QTL contributing to testis weight on the proximal portion of the X chromosome (all QTL > 60.2 Mb, Oka et al. 2004; Storchová et al. 2004), the same region where we detected the largest phenotypic effects (Figure 5, Table 3). In general, the predominant influence of the proximal portion of the X chromosome in our study stands as the strongest discrepancy with previous mapping efforts. This difference may reflect genetic differences between the different introgressed M. musculus X chromosomes and/or the genomic background of C57BL/6J vs. domesticusLEWES.
All three studies consistently identified multiple QTL influencing abnormal sperm head morphology spanning the X chromosome. Our minimum estimate of four sterility factors related to abnormal sperm head morphology exceeds the estimates of both previous studies, where the independent influences of specific intervals were not considered in detail (Oka et al. 2004; Storchová et al. 2004). Oka et al. (2004) described three major QTL (sperm head abnormality 1, Sha 2, Sha 3) that overlap with our proximal, central, and distal QTL intervals (Figure 5). Sha 2 also overlapped with the primary hybrid sterility-related QTL (hybrid sterility-X1 or Hstx1) identified by Storchová et al. (2004) and maps to the same region where we identified two QTL for abnormal sperm head morphology and one QTL for reduced testis weight (Table 3). When considered jointly, these data strongly suggest that one or more genes involved in reproductive isolation between M. musculus and M. domesticus reside within this region (see below).
Several parallels between our data on the genetics of hybrid male sterility and patterns of gene flow across the European hybrid zone merit further consideration. First, the two X-linked markers with lowest levels of gene flow across the hybrid zone are located in the central (Pola1, 89.9Mb) and distal (Btk2, 132.1Mb) regions of the chromosome (Payseur et al. 2004). Pola1 is distal to our major central QTL, as well as the Sha 2/Hstx1 interval. However, mapping resolution on the X chromosome has proven to be fairly limited in the hybrid zone and several adjacent markers also show reduced gene flow (Payseur et al. 2004), resulting in a large region of restricted gene flow (64.9–99.8 Mb) that partially overlaps with QTL identified in our study as well as Sha 2/Hstx1. Btk2 is clearly contained within the confidence interval of our distalmost QTL for reduced testis weight (Table 3).
Second, asymmetric introgression has been noted across several hybrid zone transects with a tendency for more introgression of M. domesticus alleles into M. musculus (Vanlerberghe et al. 1986; Tucker et al. 1992; Dod et al. 1993, 2005; Teeter et al. 2008). In light of our data on asymmetric X-linked sterility, it is tempting to speculate that this pattern reflects a general accumulation of more incompatibilities on the M. musculus background. However, the fact that asymmetric introgression in the hybrid zone is seen for loci throughout the genome suggests that it may be driven primarily by demographic processes rather than selection on individual loci.
Finally, levels of gene flow across the entire X chromosome are lower than levels generally found on the autosomes (Tucker et al. 1992; Dod et al. 1993; Payseur et al. 2004; Macholán et al. 2007; Teeter et al. 2008). This pattern would not be expected if the X chromosome harbored only one or a few incompatibilities of major effect. Our mapping study was designed to detect fairly specific abnormalities in male reproduction (of moderate to large effect) while hybrid zone data should be sensitive to a broad range of isolation phenotypes (Payseur et al. 2004). Thus, concordance between our data and hybrid zone studies is consistent with the notion that hybrid male sterility plays an important role in reproductive isolation in nature, but does not address the contribution of other X-linked phenotypes to speciation in house mice.
Candidate genes for hybrid male sterility:
Our QTL analysis indicated that one or more loci involved in reduced testis weight and abnormal sperm head morphology occur within the 12.88 Mb interval between DXMit141 and DXMit119. This general QTL interval represents the clearest region of overlap between the current work and previous studies (Oka et al. 2004; Storchová et al. 2004). In the work of Storchová et al. (2004), the proximal boundary of major QTL for reduced testis weight and sperm count is clearly demarcated at DXMit144 (60.24 Mb). If we further restrict our interval to this proximal boundary, then the region of interest is reduced to 8.44 Mb (DXMit144 at 60.24 Mb to DXMit119 at 68.66). Of the 1026 protein-coding genes currently annotated on the mouse X chromosome, 41 are located within this region. However, only 9 of these loci (Table 6) were expressed during spermatogenesis in a recent survey of X-linked expression in male germ cells (Namekawa et al. 2006). Of these 9 genes, 3 (Ctag2, 4933436I01Rik, and 1700020N15Rik) show high expression in testis relative to other tissues (>3× median across 61 tissues; Su et al. 2004). All 3 genes are expressed primarily in postmeiotic round spermatids (Namekawa et al. 2006) and have no known function. 1700020N15Rik is a paralog to ENSMUSG00000056815 and both genes match the same probe on the Affymetrix 430 2.0 chip used by Namekawa et al. (2006). Thus, putative spermatogenesis expression of these 2 genes was ambiguous.
Multiple genes involved in reproductive isolation in Drosophila have been shown to be rapidly evolving (Ting et al. 1998; Barbash et al. 2003, 2004; Presgraves et al. 2003; Brideau et al. 2006) and positive selection is thought to be an important process in the evolution of D–M incompatibilities (Coyne and Orr 2004). The two testis-specific genes with unambiguous orthologs in rat (4933436I01Rik and Ctag2) show elevated rates of protein evolution, as is common for testis- and other tissue-specific genes in mice (Winter et al. 2004; Good and Nachman 2005; Dean et al. 2008). In particular, 4933436I01Rik is among the most rapidly evolving genes on the X chromosome (highest 3% dN:dS values based on pairwise comparison with one-to-one rat orthologs) and was originally identified in a genomewide cDNA screen of adult testis (Bono et al. 2003). Inspection of known SNPs (dbSNP version 26) in wild-derived inbred strains of Mus revealed three nonsynonymous substitutions between M. musculus (sequenced strain PWD/PhJ) and M. domesticus (sequenced strain WSB/EiJ). Further population genetic analyses will be needed to evaluate whether 4933436I01Rik and other testis-expressed genes within this interval have evolved due to positive directional selection.
Finally, hybrid male sterility in mice may be influenced by locus-specific or X chromosomewide transcription misregulation in hybrid testis. Disrupted genomic imprinting has been put forth as a general mechanism for asymmetric isolation in mammals (Vrana 2007) and stands as a formal possibility for X-linked sterility in house mice (Storchová et al. 2004; Homolka et al. 2007). Likewise, expression divergence between M. musculus and M. domesticus of specific X-linked loci could also be involved in the disruption of spermatogenesis in hybrids. Abnormal patterns of male-biased gene expression are common in Drosophila F1 hybrids and may contribute to male-limited sterility (Michalak and Noor 2003; Ranz et al. 2004). Interestingly, a recent study found that the testis shows high levels of expression divergence between M. musculus and M. domesticus relative to liver and brain (Rottscheidt and Harr 2007). Rottscheidt and Harr (2007) did not find strong evidence for misexpression in testis of F1 hybrid males from reciprocal crosses between M. musculus and M. domesticus. However, they did not evaluate the fertility of F1 males in their experiment and F1 hybrid male sterility is polymorphic between these species (Forejt and Iványi 1975; Britton-Davidian et al. 2005; Vyskocilová et al. 2005; Good et al. 2008). Thus, the potential relevance of these data to reproductive isolation remains speculative. We are in the process of examining patterns of testis expression in M. musculus, M. domesticus, reciprocal F1 hybrids, and select congenic models for which detailed male reproductive data are also available.
We thank T. Shiroishi for sharing published data on the genetics of hybrid male sterility in mice. M. A. Handel, C. W. Birky, B. Payseur, A. Doyle, and T. Quill provided useful advice on experimental design and methodology. C. W. Birky and M. Worobey provided equipment used during the course of this experiment. We are grateful to T. Shiroishi, A. Oka, R. Storchová, B. Payseur, M. Macholán, two anonymous reviewers, and members of the Nachman lab for critical comments on this manuscript. We thank B. Gibson, M. Longhi, K. Smith, M. Sans-Fuentes, M. Rand, and the staff of the University of Arizona Central Animal Facility for assistance with mouse husbandry. J.M.G. received support from the National Science Foundation (NSF) Integrative Graduate Education and Research Traineeship (IGERT) Program in Genomics at the University of Arizona, an NSF Doctoral Dissertation Improvement grant (DEB0608452), and small research grants from the Department of Ecology and Evolutionary Biology, University of Arizona. This work was also supported by NSF grants (DEB0213013, DEB0642778) and a National Institutes of Health grant (RO1 GM-074245) to M.W.N.
↵1 Present address: Max Planck Institute for Evolutionary Anthropology, D-04103 Leipzig, Germany.
Communicating editor: D. M. Rand
- Received December 7, 2007.
- Accepted June 13, 2008.
- Copyright © 2008 by the Genetics Society of America