Position of the HMHZ in Central Europe

Frequencies of the X- and Y-linked markers display roughly the same geographic distribution as morphological data (Fig. 1A-B and D, respectively). Contrary to these data sets, in northern and southern regions of the study area M. m. domesticus mtDNA is found markedly eastward within an otherwise M. m. musculus genetic background (Fig. 1C). This inconsistency challenges the mtDNA diagnostic marker as the most relevant trait for delimiting distribution ranges of the mouse subspecies. In the following, because of the higher effective population size of the X chromosome relative to the Y chromosome and stronger selection on X-linked loci in comparison to autosomal loci (for references see Introduction), we chose HIX for delineating the course of the HMHZ in the region under study.

Starting from the Baltic Sea, the zone runs between Weitendorf and Reinstorf (localities #14, #36 in Fig. 1A), in the direction from Wismar to Schwerin (Lake Schweriner). Then it turns south-west between Lalendorf and Groß Wokern (#37 and #38, both M. m. musculus) and Lübz and Suckow (#15 and #16, both M. m. domesticus) to Lake Müritz and then it turns south to Berlin. Although our sampling on the M. m. domesticus side does not allow us to precisely describe its course here, morphological data suggest that this subspecies is present in Berlin (#311 in Fig. 1D). Mice sampled in Hohenstein (N = 4, #42 in Fig. 1A), located about 25 km from the eastern suburbs of Berlin, are M. m. musculus at all five X-linked markers. South of Berlin the zone runs between Elsterwerda (#44) and Lauchhammer (#43) in the west (both prevailingly M. m. musculus) and Kreischau and Mehderitzsch (#82 and #81, both predominantly M. m. domesticus, Teeter et al. 2010) with the centre close to Zschirla (#80, HIA ≈ 0.70, HIX ≈ 0.66) between Leipzig and Chemnitz (Teeter et al. 2010). Further to the south, the zone runs through the westernmost tip of the Czech Republic (Munclinger et al. 2002, Božíková et al. 2005, Macholán et al. 2007, 2008, 2011, Dufková et al. 2011, Wang et al. 2011).

The position of the HMHZ in Bavaria is not as easy to detect as in northern Germany. The zone seems to bend south-eastwards where the River Wondreb/Odrava crosses the Czech-German border and then follows the state border where the contact between mouse subspecies is hampered by presence of the Oberpfälzer Wald/Český les Mts. Stronger gene flow can be surmised in the gap between this mountain range and the more south-eastern Bayerischer Wald/Šumava Mts., along the River Kouba/Chambach lowland. This river is the tributary of the River Regen which flows into the River Danube in Regensburg where occurrence of M. m. musculus was assumed by Zimmermann (1949) and Kraft (1985). Both authors reported morphologically musculus-like mice some 25 km north of this city (Parsberg/Oberpfalz, Amberg, Neumarkt, Burglengenfeld, localities # 277–281 in Fig. 1D), whereas our genetic data from mice collected from nearby localities (Gebenbach #26, Pessensricht #27, Frechetsfeld #28, Unterkatzbach, Fig. 1A) suggest presence of M. m. domesticus in this region. Further south, the HMHZ turns to Munich. Based on the genetic data gathered from the Bavarian transect (Sage et al. 1986b, Tucker et al. 1992, Payseur et al. 2004, Teeter et al. 2008, 2010, Wang et al. 2011) it was suggested that the River Isar forms a barrier between the subspecies (Sage et al. 1986b, 1993). Our sampling at four localities along both banks of this river south of Munich suggests a position of the zone east of the Isar; the HMHZ is thus likely to bend to the south-east in Munich with the presumed centre between Palnkam (#33) and the River Inn.

The M. m. musculus populations from the north-eastern Schleswig-Holstein (Prager et al. 1993, 1996, 1997, localities #4–8 in Fig. 1B, C) seem to have disjunct distribution. We are unable to infer whether this isolate was a part of the continual range of consubspecifics detected here further to the east (#36 in Fig. 1A) and separated from it by invasion of M. m. domesticus (represented by localities #14 in Fig. 1A, #13 in Fig. 1D) or is related to M. m. musculus populations in Fehmarn Isl. (Prager et al. 1993, 1996, 1997, localities #1–3 in Fig. 1B, C) and Sweden/Norway (Jones et al. 2010).

The length of the MHMZ covered in this study is about 850 km. Along its course, the zone runs through different environments. From a perspective point of view it would be interesting to find if the zone is solely maintained by migration of parental taxa in the zone and selection against hybrids or whether and to what extent the zone interacts with geographic barriers posed by mountains, riverine systems or artificial structures such as autoroutes. For example, in southern Germany such a barrier forming a step in gene frequency was suggested to occur across the floodplain of the River Isar (Sage et al. 1986b, 1993). While this river can reduce gene flow in the region north of Munich, our data suggest this may not be the case for the region south of Munich, where M. m. domesticus is present on both banks of this river. On the other hand, we did not find any support in our data for association of the HMHZ centre with a previously suggested geographic barrier in northern Germany, where the zone was suggested to be associated with the River Elbe (Zimmermann 1949). Rather it seems that this river transverses the HMHZ in Saxony and is far away from the zone in other places.

Introgression varies among different parts of the mouse genome

Since position of the HMHZ was approximated from X-linked markers, we used HIX as the proxy to which introgression patterns at the autosomal, Y-linked and mtDNA markers are compared. Each of these biparental, paternal and maternal markers displays a unique pattern of introgression in the concordance plots (Fig. 2).

Fig. 1.

Map of the study area with house mouse localities sampled and analysed for X-linked markers (A), Y-linked markers (B), mtDNA (C), and morphological traits (D). The symbols are: — M. m. domesticus (defined as HI ≤ 0.2), — polymorphic/heterozygous populations (0.2 < HI < 0.8) and — M. m. musculus (HI ≥ 0.8), except panel D, in which classification of these categories were taken from original studies (Zimmermann 1949, Kraft 1985) and their designation of mice to either category was accepted (with the same symbols for domesticus, hybrids and musculus as in panels A-C).

Autosomal loci. Changes in frequencies at autosomal loci are strongly correlated with HIX (linear fit, Pearson's r = 0.97, 93.9 % of the variation explained). This concerted transition of gene frequency along the diagonal is termed here as the symmetric bidirectional pattern of introgression (Fig. 2A). On the contrary, introgression of the haploid loci deviates from this pattern and cannot be modelled as a linear fit (lack of fit gives P = 0.02 for the Y-linked loci and P = < 0.0001 for the mtDNA loci).

Y chromosome. Unidirectional asymmetric introgression characterises introgression of the Y-linked loci. Out of 224 populations analysed for both X a Y loci, there are 11 cases, when the domesticus Y was found in the musculus background (Fig. 2B). Does it mean that the unidirectional asymmetric introgression documented for the Y chromosome does not hold?

Due to continual sampling in the Czech transect within 1991–2011, we obtained some evidence that these cases represent rather results of longdistance migration events of M. m. domesticus into M. m. musculus territory. Introgressed alleles and haplotypes are expected to persist there for limited time, but disappear later due to colonization from neighbouring parental demes. For example, two samples were analysed from Rudolec, a village about 12 km east of the zone centre. Both samples contained the domesticus Y in 1999 (frequencies of musculus Y: 0.33 and 0.4). From 2000 up to 2003, when last mice were sampled at the localities, only the musculus Y was detected at both sampling sites, increasing over-year average frequencies within the localities to 0.5 and 0.86, respectively (Macholán et al. 2008, Fig. 2B). Although we cannot make such inference from other populations because they were sampled only once, the rarity of such atypical findings (4.9 %) makes introgression of domesticus Y rather exceptional and does not break the rule of the unidirectional asymmetric introgression for the Y chromosome. On the contrary, data from other segments of the HMHZ are in conformity with this empirical rule. In Transcaucasia and north-eastern Iran the presence of two mtDNA lineages indicates that these are regions of admixture between the mouse subspecies; however, in both regions only the musculus Y chromosome was detected (Boissinot & Boursot 1997). Of the 34 males trapped on the domesticus side in the Danish transect, six (18 %) had musculus Y chromosomes (Dod et al. 1993). Similarly, Jones et al. (2010) found the musculus Y chromosome to be widespread in otherwise M. m. domesticus-type populations in Norway.

Fig. 2.

Frequencies at autosomal (A), Y-linked (B) and mtDNA loci (C) plotted against hybrid index based on five X-linked loci. Parental M. m. domesticus is in the left bottom corner (coordinates: 0,0) and parental M. m. musculus in right upper corner (1,1).

mtDNA. Bidirectional asymmetric introgression characterises mtDNA (Fig. 2C). Alien allelic variants can introgress into either mouse genome, although the findings of domesticus mtDNA in M. m. musculus populations seem to be more prevalent (Fig. 2C). While introgression at autosomal and Y-linked loci is mostly limited to the HMHZ, mtDNA can be found far away from its centre. This phenomenon is not locally restricted. For example, locality Parlino in northwestern Poland (#46 in Fig. 1C) with the domesticus type of mtDNA is 80 km from Lindhorst, Uckerland (#39 in Fig. 1A) which is still within the musculus range. Likewise, Długoszyn (#47 in Fig. 1C) with domesticus mtDNA is about 100 km east of Berlin, and the presence of domesticus mtDNA within M. m. musculus territory was also reported from southern Bavaria by Božíková et al. (2005) (cf. Simbach, #132 in Fig. 1C, ca. 90 km east of the River Isar). Introgression of musculus mtDNA onto domesticus background seems to be more restricted, up to 25 km behind the zone centre in the Czech transect (Macholán et al. 2008).

Why do we see different introgression patterns?

Differences in concordance and symmetry observed in pairwise comparison of introgression between four sets of markers can reflect variation in effective population sizes between the marker types, recombination rates, and effects of stochastic or systematic factors.

The effective population size decreases from autosomal loci (N_e), to X-linked loci (¾ N_e) to mtDNA and Y-linked loci (¼ N_e). Hence the markers will be differentially subject to drift, with corresponding consequences for probability of fixation of alternative variants (see Polechová & Barton 2011). Another important factor potentially responsible for differences in introgression patterns can be recombination. Among the markers scored in this study, only those on the autosome and X can recombine away from their flanking background. When two individuals, each of different mouse subspecies, will meet and mate, recombination at each subsequent generation will break parental combinations. If both marker sets are neutral, they will spread across the hybrid zone, segregate in Mendelian fashion and occupy any place with coordinates given by their frequencies. Obviously, this is not the picture observed in Fig. 2A. Instead we see a nearly linear relationship between the two types of loci. This observation can be partly ascribed to the fact that diagnostic loci were chosen in all the studies covered in this paper (see Material and Methods). By definition, diagnostic loci are fixed for alternative alleles, and thus we cannot expect their segregation in localities geographically distant from the hybrid zone centre. However, such localities will all be located either in the bottom left corner in Fig. 2A (coordinates with HIs (0,0) defining parental M. m. domesticus) or in upper right corner (1,1) defining parental M. m. musculus. Thus we can conclude that both types of loci, summarised in terms of HIs, interact only within the HMHZ and are not able to introgress into musculus or domesticus genomes.

A more appropriate explanation of the lack of introgression of markers depicted in Fig. 2A is based on the tension zone model. According to this model the zone is maintained by a dynamic equilibrium between influx of parental gene combinations into the zone and selection against hybrids and recombinants (Barton & Hewitt 1985). Selection has been evidenced in the HMHZ both on the phenotype level (Britton-Davidian et al. 2005, Turner et al. 2012, Albrechtová et al. 2012) and genetic level (Raufaste et al. 2005, Macholán et al. 2007, 2011, Janoušek et al. 2012, see Baird & Macholán 2012 for recent review). These authors have come to agreement that the HMHZ is maintained through weak selection acting on many loci, rather than strong selection affecting a single or a few loci. In the tension zone model the selected loci will not behave independently resulting in strong effective selection (Barton & Gale 1993). As a result allele frequencies at the selected loci will change in a concerted way as observed in Fig. 2A. Consequently, the presence of recombination and selection can largely explain the concordance in frequency changes between autosomal and X-linked loci.

MtDNA and most of the Y chromosome do not recombine and, in addition, have uniparental inheritance. From the population-genetic point of view, we consider two possible scenarios of mouse dispersal in a hybrid zone to explain their different introgression. Both scenarios start with colonization of new habitats by a female from one source and a male from another source subspecies. While recombination at autosomes and the X chromosome will create a wide range of hybrids with HI between 0 and 1, both haploid chromosomes will be in the absence of gene influx always fixed for their alternatives depending on their maternal or paternal origin. Consequently, their frequencies will tend to be positioned along the horizontal axes with fixed musculus or domesticus alleles in Fig. 2B, C, just depending on the maternal or paternal source of the haploid marker/locus.

Polymorphism at mtDNA and Y chromosomes observed along the vertical axes in Fig. 2B, C can only be detected under two alternative scenarios, in which a deme is invaded by mice carrying advantageous gene variants. First, suppose an advantageous allele enters a deme of the other subspecies. Its spread will initially be delayed due to incompatibilities brought about by loci at the same and/or other chromosomes (Piálek & Barton 1997). However, the positively selected locus will ultimately recombine and/or segregate away from counterselected loci, cross the zone and spread through the other subspecies' range. The spread can be stopped by the presence of either a physical barrier or other incompatible allele(s) at the same locus. Fixation of musculus mtDNA and Y variants on the domesticus genetic background is indicated in Fig. 2 B, C by positions of localities with HI(_{mtDNA, Y}) = 1 and HI_X = 0. Increasing frequencies at both mtDNA and Y loci along vertical axis at HI_X = 0 are suggestive of remnants of the spread of advantageous musculus alleles to domesticus background from low frequencies until fixation. The symmetric gradient of domesticus mtDNA at HI_X = 1 (Fig. 2C) then documents the spread domesticus mtDNA into musculus populations. Although the pattern shown in Fig. 2C is drawn from the Czech transect (Božíková et al. 2005) it is not unique for this transect as it was also observed in Bulgaria (Vanlerberghe et al. 1988a). Interestingly, in Fig. 2B, C most populations are positioned within a space out of the selection domain discussed above for autosomal and X-linked loci. So far this argument was used as if selection operates directly on any mtDNA- or Y-linked locus. The observed introgression can suggest positive selection at other loci and hitch-hiking with sex-subspecific traits (see, e.g. Macholán et al. 2008) or positive epistatic or pleiotropic interactions.

An alternative to the spread-of-advantageous-alleles model is past movement of the HMHZ leaving salient of alien allele(s) that locally persist as footprints of previous distribution (Macholán et al. 2011). While the majority of the alien alleles will be replaced by gene flow from neighbouring populations coupled with genetic drift (Polechová & Barton 2011), slow removal of non-native alleles can indicate some selective advantage, frequency-dependent selection or preferential mate choice for that marker. Both scenarios are grounded on assumption that the loci are fixed for alternative alleles in each subspecies. However, as argued by Vošlajerová Bímová et al. (2011) some loci thought to be diagnostic may represent ancestral polymorphism and caution must be taken before interpretation of results.

No studies focusing on effects of mtDNA on house mouse phenotype are known; thus only genotyping other loci in mice from localities with introgressed mtDNA can help to decipher whether this chromosome is experiencing positive selection either as a whole or due to a specific locus mutation. On the contrary, we found evidence that sperm count and motility were more favourable in domesticus males harbouring introgressed Y musculus than the same males with native Y chromosome (Albrechtová et al. 2012). Improvement of sperm quality due to asymmetric replacement of the Y chromosome is in agreement with the pattern of introgression envisaged in Fig. 2C.