Introduction

Single or multiple Robertsonian (Rb) fusions are involved in evolutionary divergence of many animal and plant species (Capanna, 1982; Baker & Bickham, 1986; Jones, 1990; Bidau, 1991; King, 1993). It is expected within the framework of several chromosomal speciation models, that hybrids between Rb divergent populations should show reduced fertility due to the meiotic consequences of fusion heterozygosity, especially the unbalanced segregation of trivalents or higher order multivalents (Baker & Bickham, 1986; Bidau, 1991; King, 1993; Searle, 1993). Such structural hybrids may be synthesized in the laboratory or found in nature, usually in hybrid zones at the contact of two chromosomal races. Chromosomal hybrid zones involving Rb differences have been reported for mammals (Searle, 1993) and insects (Barton & Hewitt, 1981; Bidau & Tosto, 1991; Bidau, 1991) and there are also Rb polymorphisms unrelated to hybridization (Bidau, 1990; Nachman, 1992).

A further factor needs to be considered in relation to Rb variation. Chromosomal rearrangements can exert effects on recombination, visualized cytologically as changes in chiasma frequency and localization (John, 1983, 1990; Bidau, 1990, 1993). This repatterning of chiasmata may adversely influence orientation and segregation of both normal bivalents (Moens & Spyropoulos, 1995) and the heterozygous chromosome configurations (Mirol & Bidau, 1992; Bidau & Martí, 1995) found in hybrids between Rb chromosomal races. Chiasma repatterning in hybrids could also disrupt coadapted supergenes, thereby potentially reducing fitness in the F2 generation (Bidau, 1996 and references therein).

The West European house mouse, Mus musculus domesticus, is a well-known case of Rb variation in nature (Capanna, 1982; Redi & Capanna, 1988; Nachman & Searle, 1995). A number of Rb chromosomal races have been described and several hybrid zones identified involving multiple differences (Searle, 1993). Although many studies have been devoted to chromosome pairing and segregation in a variety of chromosomal conditions in the house mouse (Gropp & Winking, 1981; Redi & Capanna, 1988; Searle, 1988, 1993), almost nothing is known about the chiasma effects of Rb fusions in natural populations of this species.

In this paper we analyse chiasma frequency and distribution in male house mice from a hybrid zone between two Rb chromosomal races that occur in northern Scotland. The John O’Groats race has 2n=32 chromosomes and is fixed for fusions Rb(4.10), Rb(6.13), Rb(9.12) and Rb(11.14). This race forms a hybrid zone with the standard and widespread 2n=40 (all telocentric) race, in the counties of Caithness and Sutherland (Searle, 1991; Searle et al., 1993). In this paper, we describe a differential effect of heterozygous and homozygous Rb fusions on chiasma frequency and localization in male mice from the John O’Groats–standard hybrid zone.

Materials and methods

The 31 male house mice studied in this paper were live-trapped during two field trips (September and October 1994) from a variety of farms where there was already information on karyotypes (Searle et al., 1993).

Mice were killed by cervical dislocation. In order to identify the karyotype of each individual, metaphase preparations were obtained from a suspension of bone marrow cells exposed to colcemid according to the method of Ford (1966). G-banding followed the method of Seabright (1972). The right testis was used for meiotic air-dried preparations according to the method of Evans et al. (1964); C-banding of meiotic cells followed Sumner’s (1972) protocol. Twenty late diakinesis/metaphase I cells were scored per individual. Chiasmata were counted per telocentric chromosome or metacentric chromosome arm and classified by inspection after dividing them in three equal parts, as proximal (P), interstitial (I) or distal (D). The chromosome constitution of the males analysed is shown in Table 1.

Table 1 Diploid numbers and status of Robertsonian fusions in individuals of Mus musculus domesticus from the John O’Groats–standard hybrid zone. The individuals belong to the following localities: Thrumster 2 (1001, 1003–1004), Mains of Olrig (1005, 1008–1009), Dunnet 2 (1010), Ratter (1012), Bardnaclavan (1013–1020, 1023–1028), West Canisbay (1035–1037), Crakraig (1038, 1040–1043) and Brora (1052–1053). Grid references for these localities are provided in Searle et al. (1993) except Thrumster 2 (3342/9442), Ratter (3267/9724) and West Canisbay (3342/9718)

Results

Chiasma frequency and distribution in the standard and John O’Groats races

In all cases the XY bivalent was excluded from the analysis of chiasmata. Standard 2n=40 males displayed a basically P-D chiasma distribution in autosomal bivalents with a predominance of D chiasmata (Table 2). Numbers of chiasmata per cell ranged from 19 to 28 with a mean chiasma frequency (±SE) of 22.83 ± 0.31. Cell mean frequencies of P, I and D chiasmata were, respectively: 5.88 ± 0.35; 2.78 ± 0.22 and 14.18 ± 0.23. Α majority of bivalents was monochiasmate (mean: 14.95 ± 0.49) and the rest bichiasmate (mean: 4.05 ± 0.49). Bichiasmate configurations tended to occur in the larger bivalents and most of them (93.5%) were of the P-D type.

Table 2 Chiasma distribution in male house mice from the John O’Groats–standard hybrid zone. In the cases of 2n=32 and hybrid individuals, chiasma patterns are shown separately for telocentric bivalents and Robertsonian configurations. The values in parentheses in the first column correspond to the number of mice studied

Populations of the John O’Groats race are monomorphic, all individuals sharing the same 2n=32 karyotype that includes four homozygous Rb fusions: 4.10, 6.13, 9.12 and 11.14. Chiasmata per cell ranged from 19 to 25 with a cell mean chiasma frequency of 22.28 ± 0.29. The difference in cell mean chiasma frequencies between the John O’Groats and standard races is statistically significant (F=2.59; d.f.=1, 66; P < 0.05).

The frequencies of P, I and D chiasmata in John O’Groats race individuals were, respectively: 4.37± 0.29, 2.53 ± 0.29 and 15.38 ± 0.32. This chiasma distribution is significantly different from that of the 2n=40 individuals (contingency χ2=11.86; d.f.=2; P < 0.05) (Table 2). However, it is clear that this difference relates to the differential behaviour of the telocentric and metacentric bivalents. Chiasma distribution in both groups is notably different: the telocentric bivalents show a typical P-D pattern whereas the metacentrics have a basically distal one (Table 2). This difference is statistically highly significant (χ2=70.75; d.f.=2; P < 0.001); furthermore, when chiasma distribution of the telocentric bivalents alone is compared to that of the 2n=40 males, both patterns are not significantly different (χ2=2.13; d.f.=2; 0.25 < P< 0.50).

The difference in chiasma distribution between telocentrics and metacentrics also involves a decrease of chiasma frequency in the latter. Cell mean chiasma frequency of the 2n=32 individuals is lower than that of the 2n=40 ones (22.28 vs. 22.83) as shown before. However, 27.46% of the telocentric bivalents of the John O’Groats race are bichiasmate while only 3.70% of the metacentric arms have two chiasmata in the same race. Thus, mean chiasma frequencies per bivalent arm are 1.28 and 1.04 for telocentrics and metacentrics, respectively.

Chiasma frequency and distribution in males from the John O’Groats–standard hybrid zone

As indicated above, individuals homozygous for the four John O’Groats race metacentrics predominate in the north-eastern corner of Caithness. Over the rest of Caithness and Sutherland, chromosome arms 4, 6, 9, 10, 11, 12, 13 and 14 occur in a polymorphic state which has been interpreted as representing a hybrid zone between the 2n=32 and 2n=40 races (Searle, 1991; Searle et al., 1993). An analysis of chiasma frequency and distribution was performed in males from several localities within the hybrid zone, showing different homozygous and heterozygous combinations of the four Rb metacentrics (Table 1). Although cell mean chiasma frequency showed wide inter-individual variation in males from the hybrid zone (20.29–23.75), there is a general tendency for a decrease in chiasma frequency with increase in the number of Rb configurations (bivalents and trivalents in the case of Rb homozygosity and heterozygosity, respectively). Thus, 2n=40 individuals (no Rb configurations) have a pooled frequency of 22.83, whereas those with Rb configurations (bivalents and/or trivalents) have: 22.43 (two Rb configurations), 21.90 (three Rb configurations) and 21.42 (four Rb configurations including 2n=32 individuals); no individuals with only one configuration were observed; a highly significant negative correlation exists for these mean values (r=−0.9913; t=10.65; d.f.=2; P < 0.01). As shown above, in the John O’Groats race, the Rb bivalents show a differential behaviour with respect to the telocentric bivalents regarding chiasma frequency and distribution. It was expected that the same phenomenon would occur in hybrids with various combinations of Rb bivalents and trivalents, that could account for the decrease in chiasma frequency. The situation is however, more complex.

The telocentric bivalents show the same chiasma distribution as those of the 2n=40 and 2n=32 individuals (contingency χ2=6.56; d.f.=4; 0.10 < P < 0.25) (Table 2). The Rb configurations show, as expected, a differential chiasma pattern but bivalents and trivalents behave in a radically different way. First, all the metacentric bivalents show the same distal chiasma pattern and reduction of chiasma frequency as those of the John O’Groats race regardless of the rest of the karyotype (χ2=3.32; d.f.=2; 0.10 < P < 0.25) (Table 2); this behaviour is present in all metacentric bivalents of all individuals analysed and seems to be independent of the arm combination involved as the comparison in Table 3 shows for 4.10 and 11.14. The trivalents however, not only show usually higher mean chiasma frequencies than the metacentric bivalents as a per arm comparison demonstrates (F=23.33; d.f.= 11 834; P < 0.01), but have a completely different pattern of chiasma distribution in which pronounced increases of P and I chiasmata occur (Table 4). This difference between Rb bivalents and trivalents is statistically highly significant (χ2=188.51; d.f.=2; P < 0.001). Furthermore, the increase is highly significant for both P (F=142.71, d.f.=11, 884; P < 0.01) and I chiasmata (F=501.90, d.f.=11, 884, P < 0.01) when tested independently. In fact, trivalents show an even higher proportion of P + I chiasmata than telocentric bivalents (Table 2), although the latter have a significantly higher chiasma frequency per arm than trivalents (F=38.00; d.f.=1, 1420; P < 0.01). However, when both types of chiasmata are tested independently, telocentric bivalents have more P chiasmata than trivalents (F=18.12; d.f.=1, 992; P < 0.01) and fewer I chiasmata (F=8.375; d.f.=1, 992; P < 0.01). Furthermore, trivalent behaviour is independent of the trivalent involved, as the data included in Table 3 show. Relative to telocentric bivalents, Rb bivalents have a more pronounced reduction of chiasma frequency per arm than trivalents (F=168.32; d.f.=1, 2196, P < 0.01) and also significantly less P (F=314.27; d.f.= 11, 800; P < 0.01) and I chiasmata (F=16.16; d.f.= 11, 800; P < 0.01). It is worth noting that pairing behaviour of both types of Rb configuration is highly regular as indicated by the very low and similar frequency of chiasma failure: 3.2% for bivalents and 3.3% for trivalents.

Table 3 Frequencies of total (T), proximal (P), interstitial (I) and distal (D) chiasmata in three types of Robertsonian trivalent (III) and two types of Robertsonian bivalent (II) of male house mice from the John O’Groats–standard hybrid zone. The chiasma distributions of the three trivalents do not differ from each other (G = 2.242; d.f. = 4; 0.50 < P < 0.75) nor do the chiasma distributions of the two bivalents (G = 0.762; d.f. = 2; 0.50 < P < 0.75). The values in parentheses in the first column correspond to the number of mice studied
Table 4 Frequencies of total (T), proximal (P), interstitial (I), distal (D) and proximal plus interstitial (P+I) chiasmata per meiotic configuration in male house mice carrying Robertsonian fusions from the John O’Groats–standard hybrid zone

Discussion

Robertsonian rearrangements occupy a prominent place in animal cytogenetics because they are commonly observed and easily analysed in cytological preparations (White, 1978; King, 1993). They have been studied as spontaneous mutants, balanced polymorphisms and within hybrid zones or in synthetic hybrids, and have been implicated in chromosomal speciation models (White, 1978; Capanna, 1982; Baker & Bickham, 1986; Sites & Moritz, 1987; Redi & Capanna, 1988; Searle, 1988, 1993; Bidau, 1991; King, 1993; Bidau & Martí, 1995). Although models vary in a number of details, most assume that the fertility of heterozygous hybrids is reduced relative to homozygotes. Underdominance of the rearrangements is attributed to segregation failure of multiple configurations, and most meiotic studies have been centred in these aspects (Gropp & Winking, 1981; Redi & Capanna, 1988; Bidau, 1991, 1996; Mercer et al., 1992; Bidau & Martí, 1995; Hauffe & Searle, 1998). However, except for extensive studies on the grasshopper Dichroplus pratensis, a species with a complex Robertsonian system (Bidau, 1990, 1991, 1993, 1996; Mirol & Bidau, 1992; Bidau & Martí, 1995), data on chiasma variation are scarce for comparable Rb systems such as Mus musculus domesticus and the common shrew (Sorex araneus) (Searle, 1986; Wallace et al., 1992).

Detailed analysis is relevant because number and position of chiasmata affect the symmetry of heterozygous meiotic configurations which in turn can affect metaphase I orientation and anaphase I disjunction (Sybenga, 1975; Mirol & Bidau, 1992; Bidau & Martí, 1995; Bidau, 1996). This is pertinent, since a common effect of chromosomal rearrangement in well studied groups such as grasshoppers is the repatterning of chiasmata in the chromosomes involved (Bidau & Martí, 1995), but almost no information exists on mammals. Repatterning of chiasmata could explain the adaptive value or the underdominance of the rearrangements by creating conditions for supergene formation or disruption, respectively (Bidau & Martí, 1995; Bidau, 1996).

Although many studies have analysed recombination and chiasmata in M. musculus, either because of their intrinsic genetic interest or in relation to other phenomena, most have been performed on laboratory strains (Searle et al., 1970; Kyslikova & Forejt, 1972; Polani, 1972; Berry et al., 1973; Lyon, 1976; Speed, 1977; Maudlin & Evans, 1980; Nijhoff & de Boer, 1981; Gorlov et al., 1992; Lawrie et al., 1995; Nachman & Churchill, 1996). However, in several cases detailed studies of male and female chiasma distribution are available (Polani, 1972; Speed, 1977; Nijhoff & de Boer, 1981; Lawrie et al., 1995).

Quantitative chiasma studies in feral Rb mice are almost nonexistent (Wallace et al., 1992). The John O’Groats–standard hybrid zone was formed through the reproductive interaction of the 2n=32 race from John O’Groats with the widespread 2n=40 standard race and is a clear example of a staggered hybrid zone that results in zonal populations where most individuals are homozygous for some, but not all, of the metacentrics that characterize the metacentric race (Searle, 1991, 1993; Searle et al., 1993). Our results indicate that three different effects of the four Rb fusions on chiasmata are present in these mice. First, increasing numbers of Rb configurations are significantly negatively correlated with cell mean chiasma frequency; this agrees with the case of D. pratensis, in which populations with increasing frequencies and numbers of Rb fusions show decreasing mean populational chiasma frequencies (Bidau, 1990); the same effect is observed within populations. Second, as in D. pratensis, reduction of chiasma frequency is due to an intrachromosomal effect of the fusions accompanied by a redistribution of chiasmata (Bidau, 1990). This effect has not been previously reported in feral mice (Gropp & Winking, 1981). Telocentric bivalents are unaffected and retain the same chiasma characteristics in the presence or absence of the Rb chromosomes; indeed, the 2n=40 males studied by us show mean chiasma frequencies that fall within the range of chiasma frequencies of standard laboratory mice (Searle et al., 1970; Kyslikova & Forejt, 1972; Polani, 1972; Berry et al., 1973; Speed, 1977; Maudlin & Evans, 1980; Nijhoff & de Boer, 1981; Gorlov et al., 1992; Lawrie et al., 1995); the chiasma distribution that we observed in telocentric bivalents is likewise similar to that previously found in laboratory mice (Maudlin & Evans, 1980; Lawrie et al., 1995).

However, in Rb mice from the John O’Groats–standard hybrid zone, there are differences between homozygous and heterozygous configurations. Whereas in D. pratensis, both Rb bivalents and trivalents show a displacement of chiasmata towards distal ends of chromosome arms in both sexes (Bidau & Martí, 1995), in Caithness mice, only Rb bivalents show such repatterning, which is independent of the Rb chromosomes involved (Table 3). All trivalents show the opposite behaviour: a significant increase of P and I chiasmata.

There is evidence that these effects of Rb fusions could be more widespread in the house mouse. (1) Polani (1972) studied laboratory mice homozygous for a 6.15 Rb fusion and found a strong tendency to distal localization of chiasmata in the short arm, although chiasma frequency did not seem to be affected. (2) Maudlin & Evans (1980) studied chiasma formation in oocytes of laboratory mice homozygous for the Rb(5.15)3Bnr translocation and found a lack of interference across the centromere regarding the number of chiasmata, but a strong tendency to distal localization in both arms and an elimination of chiasmata around the centromere. (3) Recently, Dumas & Britton-Davidian (2000) reported what seems to be a similar situation to ours in two races (2n=40 and 2n=22) of M. m. domesticus and their hybrids in Tunisia. (4) These opposite effects are not restricted to the Western European house mouse. The same differential chiasma formation occurs in Rb homozygotes and heterozygotes of the Ctenomys perrensi species complex (Caviomorpha, Ctenomyidae) from Argentina (C. Lanzone, personal communication) and in the South American marsh rat Holochilus brasiliensis (Nachman, 1992).

Why does this different behaviour of Rb bivalents and trivalents occur? The results suggest that different mechanical rules operate for chiasma formation in house mice relative to D. pratensis. In D. pratensis, a simple model explains the parallel chiasma repatterning of both configurations: since in grasshoppers pairing usually starts at chromosome ends attached to the nuclear envelope and assuming that chiasmata are formed more readily in regions that pair first (Jones, 1987), a Rb fusion (homozygous or heterozygous) instantly reduces the early pairing ends from four to two: the distal ones (Martí & Bidau, 2000; Bidau, 1993; Bidau & Martí, 1995). In house mice this is not applicable. First, trivalents and bivalents behave in opposite ways; second, recent observations do not suggest that telomeric sequences initiate synapsis in the house mouse (Moens et al., 1995).

According to these effects on chiasma formation, what are the possible evolutionary influences of Rb fusions in natural populations of house mice? They could exert a form of recombination control. Lyon (1976) proposed that recombination was not random in M. musculus, based on the analysis of the then known genetic map, since genetic markers were not evenly distributed along mouse chromosomes, but clustered. Lyon suggested that this could be a direct reflection of chiasma localization. This hypothesis was indirectly supported by the patterns of chiasma location found in mouse bivalents, and was demonstrated by Nachman & Churchill (1996), who analysed the distribution of markers along the M. musculus microsatellite map. Thus, changes in chiasma frequency in Rb homozygotes and heterozygotes of M. m. domesticus, as we observed, may change the recombination frequencies of different parts of the genome, for example by the creation of relatively recombination-free zones in Rb homozygotes (in the proximal and interstitial regions of the metacentrics).

With respect to the John O’Groats–standard hybrid zone, the distribution of chiasmata in hybrids influences the gene flow across the zone. The presence of chiasmata near the centromere in Rb heterozygotes enhances the possibility of gene flow for loci along the whole of the Rb chromosome. This reduces the opportunity for genetic differentiation of the John O’Groats and standard races while in contact (see Searle, 1993).

In experiments in which wild Rb metacentrics were introduced into standard laboratory mouse genomes, some metacentrics suppressed pericentromeric genetic recombination in heterozygous individuals (Cattanach, 1978; Davisson & Akeson, 1993). These cases were attributed by Davisson & Akeson (1993) to delayed pairing of pericentromeric areas of trivalents, and would seem to contradict our findings. However, there are many Rb metacentrics that do not suppress pericentromeric recombination (Davisson & Akeson, 1993), and even the cases of true suppression could be due to interactions between the genetic background and the metacentric chromosomes peculiar to certain lines of laboratory mice. The situation is clearly different from that of the wild populations studied in this paper in which the Rb polymorphisms arose from a past event of hybridization and underwent many generations of natural selection.

The occurrence of I and P chiasmata in Robertsonian heterozygous mice is interesting in other respects. First, in relation to the low anaphase I nondisjunction suffered by such individuals (only 2.7% per heterozygous configuration; Wallace et al., 1992). Distal location of chiasmata is considered important for correct segregation (Sybenga, 1975). Clearly, our results cast doubt on this. Secondly, such a distribution of chiasmata would fail to protect from recombination any supergene formed by Robertsonian fusion in the metacentric race.