Strong premating divergence in a unimodal hybrid zone between two subspecies of the house mouse


Carole Smadja, Laboratoire Génétique et Environnement, Institut des Sciences de l'Evolution, UMR 5554, Université Montpellier II, C.C. 065, 34095 Montpellier cedex 5, France.
Tel.: (00 33) 4 67 14 46 31; fax: (00 33) 4 67 14 36 22;
e-mail: or


Although selection against hybridization is expected to generate prezygotic divergence in unimodal hybrid zones, such a pattern has been seldom described. This study aims to better understand how prezygotic mechanisms may evolve in such zones. We investigated prezygotic divergence between populations of two subspecies of mice (Mus musculus musculus and M. m. domesticus) located at the edges of their unimodal hybrid zone in Denmark, and we developed an original multiple-population choice-test design, which allows assessment of within and between subspecies variation. Our study demonstrates that a strong assortative preference characterises one of the two subspecies (musculus) and that urinary signals are involved in this subspecies recognition. Taking into account the specific genetic and geographical characteristics of the Danish hybrid zone, we discuss the influence of the above pattern on its fate and the mechanisms that could have favoured this prezygotic divergence, among which the role of recombined populations constituting the core of the zone.


Hybrid zones between divergent geographic taxa are natural settings where the nature of barriers to gene exchange can be explored. Indeed, analyses of post-zygotic isolation mechanisms allow characterization of genetic incompatibilities involved in isolation (toads: Szymura & Barton, 1991; grasshoppers: Butlin, 1998; several eutherian mammals: Searle, 1993), as well as showing that species could be formed through hybridization (e.g. polyploid speciation in plants: Arnold, 1992). More recently, the application of genetic mapping tools to track the genetic architecture of naturally recombined individuals allowed further progress in understanding the mechanisms of genetic isolation (Rieseberg et al., 1999). Nevertheless, several hybrid zones have led to a more complete and integrative understanding once prezygotic isolation mechanisms relying on ecological differentiation (Irises: Cruzan & Arnold, 1994; butterflies: Jiggins et al., 1996), on premating differences (butterflies: Jiggins et al., 1996; grasshoppers: Ritchie et al., 1989) or on post-mating divergence, either through gamete incompatibilities (ground crickets: Howard & Waring, 1991) or pollen precedence/conspecific sperm (Rieseberg et al., 1995; Howard et al., 1998), were taken into account.

Unimodal zones characterized by the predominance of recombined genotypes are the outcome of secondary contacts between taxa that have not diverged to a sufficient extent for prezygotic or post-zygotic characteristics to hinder production of intermediate forms through inter-crosses and multiple back-crosses (Jiggins & Mallet, 2000). Nevertheless, examples exist where, despite possible prezygotic divergence in allopatry, the advent of a contact constitutes a strong disturbance leading to a mismatching in the species recognition system and to hybridization (Rhymer & Simberloff, 1996). In their review, Jiggins & Mallet (2000) pointed out that prezygotic isolation mechanisms are more frequently observed in bimodal hybrid zones where parental forms predominate with rare F1 hybrids than in unimodal hybrid zones. Indeed, unimodal hybrid zones could be evolutionary dead-ends and lead to the extinction of one or both taxa or to their fusion (Paterson, 1978; Liou & Price, 1994). Nevertheless, in general, selection against hybridization occurs in unimodal zones (e.g. Hewitt et al., 1987; Szymura & Barton, 1991), which should favour evolution of prezygotic divergence, and result in the shift of the zone towards bimodality (i.e. reinforcement: Dobzhansky, 1940; Howard, 1993). Under the latter scenario, selection against hybridization could either directly involve the two taxa, or influence each of the two through interactions with neighbouring recombined populations across the zone (e.g. relative preferences in sticklebacks: Nagel & Schluter, 1998).

Prezygotic divergence may be rarely observed in unimodal hybrid zones because such zones are not easy to investigate. Indeed, the question of where, in the zone, divergence is most likely to evolve and be detected may not always be simple to answer. It is often assumed that prezygotic isolation mechanisms should preferentially evolve at the centre of the zone where selection against hybridization is the strongest; however, this means that, in unimodal hybrid zones, prezygotic divergence is to be investigated in recombined populations whose relatedness to one or the other parental type may not be obvious to assess. Populations from the edges of a contact zone might be more relevant to investigation, being part of the hybrid zone, but still clearly belonging to one or the other parental type and as such perhaps easier to investigate in terms of their prezygotic characteristics. Besides, in general, investigations of prezygotic isolation mechanisms along transects crossing the hybrid zones (e.g. Ritchie et al., 1989; Butlin & Ritchie, 1991; Sanderson et al., 1992) do not consider replicate populations for each part of the zone. Consequently, there is a risk that localized processes may mislead the interpretation of a pattern more relevant to the dynamic of the hybrid zone (Butlin, 1994). Besides, only an approach taking into account between population variation could assert between taxa differentiation if a divergence exists. Therefore, we have adopted a multiple-population approach to investigate prezygotic divergence in populations from the edges of a unimodal contact zone. This extensive characterization of premating divergence is considered a necessary step to understand which mechanisms could drive such divergence in hybrid zones.

This study concerns a unimodal hybrid zone between two subspecies of the house mouse, Mus musculus musculus and M. m. domesticus. After a differentiation in allopatry from a common ancestor in the Indian subcontinent (Boursot et al., 1996) M. m. domesticus spread into western Europe and the Mediterranean basin and M. m. musculus into central Europe and northern China following human travel, agricultural developments, and landscapes opening (Auffray et al., 1990). The secondary contact between these two subspecies occurred in the southern part of Europe, the Balkans, 6000 years ago and has progressed through to the North reaching Denmark some 2800 years ago (Auffray et al., 1990; Auffray, 1993). Unimodality of the zone is supported by the existence of recombined hybrids in its core (Dod et al., 1993; Dallas et al., 1995). Moreover, available genetic data indicates that strong selective pressures against hybridization exist, particularly at the centre of the zone (Vanlerberghe et al., 1986; Nancéet al., 1990; Dod et al., 1993; Moulia et al., 1993; Fel-Clair et al., 1996).

A preliminary study testing a pair of populations from the edges of the zone suggested that subspecies recognition occurred between these two populations, and that this mate recognition was mediated by urinary signals (Smadja & Ganem, 2002). Here, we test whether subspecies recognition and assortative preference characterize each subspecies so as to assert premating isolation between the two taxa. For divergence to be relevant to isolation, both subspecies recognition signals and assortative mate preference should characterize all or most of the populations within each subspecies at the edges of the hybrid zone. Using a multiple-population design, we analysed male and female mate preference within four naturally occurring populations of each subspecies, sampled at the edges of the hybrid zone in Denmark (Jutland). We used choice tests to assess preference in laboratory conditions, a good estimator of the propensity to mate in a mouse (Yanai & McClearn, 1973; Egid & Brown, 1989; Coopersmith & Lenington, 1992; Laukaitis et al., 1997; Smadja & Ganem, 2002).

All the populations within each subspecies displayed a consistent pattern of preference, irrespective of which pair of populations was presented as the stimuli during the test, and assortative preference characterized one of the two subspecies. In addition, we performed choice tests during which pools of urine representing each population were presented to the mice, and demonstrated that the expression of assortative mate preference relies on the recognition of signals present in urine. Another short experiment aimed to test whether assortative preference could depend on perception of particular local properties of the signals emitted by the Danish populations of M. m. domesticus (i.e. chromosomal rearrangement, levels of introgression), rather than on subspecies differences. For that, we repeated a choice test replacing the Danish signal by an allopatric M. m. domesticus signal, and revealed a trend for the same assortative preference. We discuss conditions under which patterns of preference evidenced in laboratory conditions could be effective in the field, and consider several scenarios involving different mechanisms to explain potential premating isolation between the two subspecies of mice whose direct contact is prevented by the presence of central recombined populations.

Material and methods

Population samples

Mice were trapped in May 2000 in farmhouses located on the tails of the introgression cline that define the Danish hybrid zone in Jutland (Dod et al., 1993). Trapping took place on several farms on each side of the hybrid zone, and mice from sites located in close proximity to each other were assigned to a single ‘population’ named after the main town in the area. Every population was composed of mice from two or more farms and each subspecies was represented by four different populations: for M. m. musculus: (A) Laasby, (B) Ejstrupholm, (C) Framlev and (D) Odder; for M. m. domesticus: (A) Lunderskov, (B) Sommersted, (C) Fjelstrup and (D) Sonder Bjert (Table 1 and Fig. 1).

Table 1.  Geographical characteristics of trapping sites, and number of mice trapped per farm (only adults were retained for the behavioural tests).
SubspeciesPopulation (named after the closest town)FarmLatitude/longitudeNumber of mice trappedNumber of mice retained for testing
  1. Grouping of farms to constitute populations was based on geographical proximity (‘G1’ = first generation litters of crosses between wild mice).

Mus musculus musculus(A) LaasbyVoel OestermarkN 6228.33; E 545.543211
ElleskovhuseN 6223.20; E 552.29501798
ProvstgaardN 6229.69; E 549.011
TovstrupN 6226.03; E 550.546101
(B) EjstrupholmFrisbaekN 6203.48; E 523.001
Vester GludstedN 6209.32; E 518.022110
LangbankeN 6211.84; E 518.563101
Gammel HampenN 6207.23; E 523.26191165
Store NoerlundN 6210.51; E 514.929734
(C) FramlevBorum OestergaardN 6227.26; E 563.9017532
AboN 6220.46; E 564.51141055
OrmslevN 6220.50; E 566.7114523
(D) OdderHoejbyN 6197.40; E 575.4021927
GosmerN 6199.75; E 574.7011 (G1)83
Total  139804040
Mus musculus domesticus(A) LunderskovBaeklundN 6149.29; E 516.18481459
GelballeN 6149.29; E 522.6021651
(B) SommerstedKastvra AN 6129.22; E 522.15211266
RingtvedN 6125.52; E 522.139321
SillerupgaardN 6131.84; E 517.431
KastvraN 6129.05; E 521.365523
(C) FjelstrupSillerupN 6128.38; E 535.7438743
10 (G1)64
SlusenN 6133.66; E 538.164303
(D) Sonder BjertSkartvedN 6145.88; E 534.229110
BinderupN 6143.27; E 535.102321
Soender StenderupN 6146.51; E 538.651101
LauritzmindeN 6145.05; E 537.06201578
Total  179804040
Total   2961608080
Figure 1.

Geographical location of eight populations of house mice sampled for this study, with reference to the centre of the hybrid zone (dotted line) and to the introgression pattern estimated from genetic studies (gradient of shading from Mus musculus musculus in the north to M. m. domesticus in the south). Black dots, position of farms with reference to their latitude/longitude coordinates; circles, farms grouped to represent a given population. All populations are located more or less at the edges of the hybrid zone.

Mice were housed with members of the same population both during transport from the field and under laboratory conditions (food and water available ad libitum, 12 : 12 photoperiod; lights on between 7 am and 7 pm). The wild mice involved in the behavioural tests were all trapped as adults. For two populations, one of each subspecies, mortality during transport necessitated breeding the survivors in captivity to increase numbers [M. m. musculus D = 9 wild mice + 11 ‘first generation’ (G1) mice and M. m. domesticus C = 10 wild mice and 10 G1 mice; Table 1]. G1 mice were tested when they were sexually mature (>1 month old). Three to four weeks before being tested, the mice were kept singly in small cages (26 × 16 × 14 cm).

Genetic characterization

Levels of introgression of our populations were estimated through an electrophoresis analysis of five allozymic loci which present alternate alleles diagnostic of each subspecies (blood proteins: Es-1 on chromosome 8, Amy on chromosome 3, Es-10 and Np on chromosome 14, Pgm-1 on chromosome 5) (Hunt & Selander, 1973; Boursot et al., 1984). Tissue, buffer systems, starch gels and stains were prepared according to Pasteur et al. (1987). For each population (Table 2), we computed a hybridization index corresponding to the average frequency of M. m. domesticus alleles on the five loci (Vanlerberghe et al., 1986). The index ranges from 0 for a ‘pure’M. m. musculus sample to 1 for a ‘pure’M. m. domesticus.

Table 2.  Hybridization index (HI) based on five diagnostic loci (calculated with reference to the frequency of Mus musculus domesticus types alleles), and karyotypes of the eight populations involved in our study.
Mus musculus musculusMus musculus domesticus
mus Amus Bmus Cmus Ddom Adom Bdom Cdom D
  1. Size of samples involved in these analyses are given between brackets. 2n, diploid number of chromosomes.

 0.061 (24)0.114 (12)0.029 (14)0.027 (11)0.952 (25)0.994 (17)0.971 (8)0.991 (23)
 40 (2)40 (2)40 (2)40 (2)34–37 (5)35–36 (5)34 (5)34–36 (5)

Whereas M. m. musculus populations present the standard karyotype of the genus (2n = 40), Robertsonian translocations are known to occur in several M. m. domesticus populations in Denmark (Nancéet al., 1990; Fel-Clair et al., 1996). We checked the chromosomal status of a sample of mice from each of the populations. Mitotic metaphases were obtained from bone marrow cells after yeast stimulation, following the air-drying technique (Lee & Elder, 1980). The slides were stained with a Giemsa solution, and chromosomal counting was performed using a Zeiss Axiophot fluorescent microscope (Carl Zeiss S.A., Le Pecq, France).

General experimental procedure

All the behavioural experiments took place between 8 and 12 am. When mice were involved in two different series of experiments, the two were performed at least a month apart (see below).

We measured a relative directional preference and presented an individual with a two-way choice (Smadja & Ganem, 2002). Opposite sex stimuli were used to assess sexual preference, and females were tested when sexually receptive (oestrus or proestrus/oestrus) to optimize sexual preference. State of sexual receptivity was controlled by performing vaginal smears on the evening prior to the day of the experiment. A pair of stimuli was composed of a homosubspecific (same subspecies but different population to the test animal) and a heterosubspecific stimulus to assess subspecies preference. The stimuli were mice (potential sexual partners), or their urine (part of their olfactory signature). Urine samples were collected from several individuals of several populations, pooled, and kept at −20 °C prior to testing.

The choice apparatus comprised a Y maze connected to a start box and to two additional peripheral boxes in which live stimulus animals were placed (for a detailed description see Smadja & Ganem, 2002). When the tests involved urinary stimuli, 10 μL urine aliquots were placed on a piece of blotting paper taped on the extremity of the branches. A test lasted 5 min during which the mouse was free to investigate the apparatus and the two stimuli. We controlled for laterality by alternating the position (left and right) of each type of stimulus between individuals within a given population sample. Contact with a stimulus was recorded when the mouse either sniffed or licked the stimulus. When the stimuli were live animals, contact was recorded when the actor sniffed or licked the transparent perforated door that isolated the peripheral boxes, whether the stimulus was standing behind the door or not. Time spent on each branch of the Y maze (‘side’) and time spent investigating each stimulus (‘stimulus’) were recorded using a Psion Organiser (Wageningen, The Netherlands) and the Observer software (The Observer Mobile, version 3.0; Noldus Information Technology, Wageningen, The Netherlands). The patterns of choice based on ‘side’ and ‘stimulus’ were consistent, and hence, only the results involving time spent investigating the stimuli are reported here. Preference is deduced from a relative comparison of the time spent investigating one or the other stimulus.

Specific experimental procedures

Experiment 1: multiple-population analysis of subspecies mate preference

Preference of 160 mice of both sexes belonging to eight populations and two subspecies was tested. This experiment involved live mice as stimuli (Tables 1 and 3).

Table 3.  Experimental setting designed to assess patterns of subspecific mate preference within and among populations of two subspecies of the house mouse in their hybrid zone in Denmark. Within each subspecies, each population has to choose between a different pair of populations (combination of four pairs of populations, a homosubspecific and a heterosubspecific population). Within each population, each mouse is presented with a different pair of mice of the opposite sex (10 pairs of individuals per sex). The design is detailed for one population.
inline image

Experiment 2: test the presence of subspecific signals in the urine

A M. m. musculus population (musculus A: 10 males; 10 females) and a M. m. domesticus population (domesticus A: 10 males; 10 females) were presented with a choice between two pools of urine: a ‘Danish M. m. musculus’ pool of males or females (musculus C and musculus D) vs. a ‘Danish M. m. domesticus’ pool (of males or females domesticus C and domesticus D). We expected that if the urinary stimuli contained the informative signals, mice would show the same pattern of preference in this test that they did when presented with a pair of live animals. The word ‘signal’ refers here to a mixture of molecules.

Experiment 3: influence of a peculiarity in the Danish M. m. domesticus signals on the pattern of choice

For this test, 10 males from M. m. musculus A population were presented with a choice between two pools of urine ‘Danish M. m. musculus’ (musculus C and musculus D) vs. ‘allopatric M. m. domesticus’ (a population from Italy and a population from Israel obtained from the genetic repository of the house mouse, UMR 5000, Montpellier, France). If homosubspecific preference still occurred, we would conclude that preference for M. m. musculus is not due to a lack of interest toward peculiar signals carried by Danish M. m. domesticus stimuli.

Data analysis

Estimation of mate preference

Sexual preference is assessed through a relative comparison of the difference in the duration of investigation of one or the other stimulus. However, an intrinsic property of this type of measure is that the larger the total duration of investigation of both stimuli (Total investigation time), the higher the probability of obtaining a large difference between the duration of each stimulus investigated. In order to control for this bias, we have assessed preference using the following estimator:


A given test was repeated when the mouse did not investigate any stimulus at all during the 5 min. Thus, the Total investigation time never equalled zero and the ratio ranged from −1 to +1. A nil ratio corresponds to an absence of preference, and the sign of the ratio indicates the direction of a choice; when the heterosubspecific stimulus is preferred the ratio is negative, whereas when the homosubspecific stimulus is preferred the ratio is positive.

Statistical analyses

Normality of the distributions of the two variables (Rstimulus and Total investigation time) (JMP software, Sall et al., 1996, Belmont, CA, USA) and equality of variances among samples (Statistica software, Statsoft Inc., Tulsa, OK, USA) were verified, and subsequently only parametric tests (anova and Student's t-tests) were performed (JMP). Backward elimination of statistically not significant factors (all with P > 0.2) was performed in order to obtain the most parsimonious anova model (Sokal & Rohlf, 1981). Multiple comparisons of mean values were performed using post hoc contrast tests (JMP).

First, we asked whether Total investigation time varied between treatments (nested anova: subspecies, populations within subspecies, and sexes within populations; equality of variances verified: Bartlett test inline image = 20.08; ns). Secondly, we asked whether Rstimulus differed between treatments (nested anova: subspecies, populations within subspecies, and sexes within populations; equality of variances verified: Bartlett test inline image = 13.88; ns). Duration of investigation could influence the pattern of preference for other reasons than mentioned above (e.g. intrinsic to the sexes, populations or subspecies). In order to control for this, when the time spent investigating the stimuli varied, we performed a covariance analysis using Total investigation time as a covariate and Rstimulus as the main variable (ancova, equality of variances verified: Box M3 = 2.09; ns). If a significant influence of the covariate was detected, we would conclude that patterns of choice are different between the two subspecies and that they could bias our estimation of their preference.

When urinary stimuli were used, we tested whether the ratio of preference varied between the two subspecies and the sexes within each subspecies (nested anova, equality of variances verified: Bartlett inline image = 4.42; ns).

Finally, the occurrence of a directional preference within each sample was tested by comparing the value of R to a theoretical value of 0 (Ho: no preference; Student's t-test). All tests were two-tailed and the Bonferroni adjustment was applied whenever a procedure involving multiple testing of the same hypothesis was used (Rice, 1989).


Genetic characteristics of the eight populations

Hybridization index and diploid number of each population are given in Table 2. This analysis confirms the position of the eight populations on the tails of the allozymic introgression cline.

Patterns of mate preference between the two subspecies

Variation of mate preference among populations (experiment 1)

We first analysed variation of the Total investigation time that may reflect a behavioural divergence between the two subspecies (differential choice behaviour or sensitivity to the stimuli). We assessed variation of this trait within and between the sexes, populations and subspecies (Table 4). Results revealed that although variation between populations occurred, time of investigation was not significantly different between the two subspecies (anova, Table 5).

Table 4.  Total investigation time (in seconds) during choice tests involving live animals (experiment 1).
SubspeciesPopulationsTotal investigation time   
Females (N = 10)Males (N = 10)Populations (N = 20)Subspecies (N = 80)
  1. Values are mean ± SE. N is sample size.

Mus m. musculusPopulation A (Laasby)62.30 ± 11.1275.80 ± 11.0069.05 ± 7.7754.20 ± 3.03
Population B (Ejstrupholm)66.20 ± 10.0055.90 ± 5.4061.05 ± 5.65 
Population C (Framlev)37.60 ± 3.3944.70 ± 5.7341.15 ± 3.91 
Population D (Odder)48.50 ± 7.3542.60 ± 4.7345.55 ± 4.31 
Mus m. domesticusPopulation A (Lunderskov)56.90 ± 7.5055.70 ± 10.2956.30 ± 6.2043.22 ± 2.68
Population B (Sommersted)38.80 ± 6.5645.60 ± 6.6042.20 ± 4.60 
Population C (Fjelstrup)31.20 ± 5.2234.40 ± 6.9232.80 ± 4.23 
Population D (Sonder Bjert)45.50 ± 7.0037.70 ± 7.8441.60 ± 5.19 
Table 5.  Nested anova testing the influence of subspecies, populations and sexes on Total investigation time and ratio of preference (Rstimulus). For the latter, an ancova testing interaction between the two above variables is also presented. The general and the most parsimonious models are given.
  1. SS, sum of squares.

Experiment 1 (stimuli = live animals; N = 160)
Total investigation time (in seconds)
 General modelSubspecies2020.0511.840.202
Pop[subspecies] (random)9626.662.710.016
Sex[pop; subspecies]2461.680.520.84
 Reduced modelPop[subspecies]19110.465.46<0.0001
 General modelSubspecies0.1711.720.200
Pop[subspecies] (random)0.5360.790.576
Sex[pop; subspecies]0.8981.000.434
 Reduced modelSubspecies3.21129.72<0.0001
Rstimulus (ancova)
Total investigation time0.000410.0030.953
Experiment 2 (stimuli = urines; N = 40)
 General modelSubspecies0.6816.170.018
 Reduced modelSubspecies2.92125.52<0.0001

Patterns of preference assessed through Rstimulus did not differ within a given subspecies nor between the sexes and the populations, suggesting a homogeneity of the patterns of preference in the two subspecies at the border of the hybrid zone (anova, Tables 5 and 6). The only significant variation was between the two subspecies, with the ratio of preference being higher in M. m. musculus than in M. m. domesticus (post hoc contrast test P < 0.0001). Moreover, the Total investigation time did not seem to affect the pattern of variation described above (ancova, Table 5), suggesting that the difference between duration of investigation of homosubspecific vs. heterosubspecific stimulus may as well suffice to estimate preference in the house mouse.

Table 6.  Ratio of preference (Rstimulus: mean ± SE) assessed per population and per sex during choice tests involving live animals as stimuli (experiment 1). Student's t-tests comparing Rstimulus to a theoretical ratio (0) were performed and t- and P-values are given for each population.
SubspeciesPopulationsFemales (N = 10)Males (N = 10)Populations (N = 20)
Mus m. musculusPopulation A (Laasby)0.313 ± 0.0990.323 ± 0.1270.318 ± 0.0784.0500.0007
Population B (Ejstrupholm)0.446 ± 0.0840.144 ± 0.0800.295 ± 0.0664.4400.0003
Population C (Framlev)0.225 ± 0.1010.360 ± 0.0970.293 ± 0.0704.1820.0005
Population D (Odder)0.270 ± 0.1040.299 ± 0.0940.284 ± 0.0684.1590.0005
Mus m. domesticusPopulation A (Lunderskov)−0.009 ± 0.0900.138 ± 0.1570.065 ± 0.0900.7200.480
Population B (Sommersted)0.130 ± 0.086−0.013 ± 0.1250.058 ± 0.0760.7670.492
Population C (Fjelstrup)0.086 ± 0.093−0.061 ± 0.1220.012 ± 0.0770.1630.872
Population D (Sonder Bjert)−0.105 ± 0.122−0.050 ± 0.056−0.077 ± 0.066−1.1810.252

Directional subspecific mate preference within populations and subspecies (experiment 1)

Our study did not demonstrate any directional preference within M. m. domesticusRstimulus = 0.014 ± 0.038, N = 80; t = 0.37, ns), whereas M. m. musculus displayed a highly significant preference for mice of the opposite sex belonging to the same subspecies (Rstimulus =0.296 ± 0.035; N = 80; t = 8.54; P < 0.0001; Table 6). Moreover, the direction of the preference was consistent within each subspecies: the four M. m. musculus populations showed a significant homosubspecific mate preference, whereas none of the four M. m. domesticus populations displayed such a pattern (Student's t-test, Table 6). Similarly, preference within the sexes was consistent, except for the population with the highest hybrid index (‘musculus B’ population), where the males showed a nonsignificant assortative preference (Rstimulus = 0.144 ± 0.08; N = 10; t = 1.788; ns).

Subspecific recognition signals in the urine (experiment 2)

The difference in the patterns of preference of M. m. musculus and M. m. domesticus was confirmed when the mice were replaced by urinary stimuli (anova, Table 5). Moreover, the average ratio of preference of the M. m. musculus population indicates a significant homosubspecific preference (Rstimulus = 0.438 ± 0.069; t = 6.35; P <0.0001) whereas a preference was not detected in M. m. domesticus (Rstimulus = −0.109 ± 0.083; t = −1.32; ns).

Subspecies recognition or lack of interest toward Danish M. m. domesticus peculiar signals (experiment 3)

Although the pattern of preference is not significantly different from zero, this experiment suggests that a trend for assortative mate preference exists (Rstimulus = 0.326 ± 0.152; t = 2.14; P = 0.06).


Premating divergence between the two subspecies

Our multiple population approach takes into account variation in both preferences and signals. Consistency of the responses across M. m. musculus populations indicates both that the assortative pattern of sexual preference is likely to be characteristic of M. m. musculus populations (Fig. 2), and that informative subspecific signals are present in all M. m. musculus and M. m. domesticus populations from the edges of the Danish hybrid zone. As a consequence of M. m. musculus assortative preference, the strength of the premating barriers to reproduction with M. m. domesticus could be particularly high. Moreover, although the house mouse is known as a polygynous species (Anderson, 1978; Berry, 1981), several studies have demonstrated a mutual mate choice of male and female mice in an individual or kin recognition context (D'udine & Partridge, 1981; Christophe & Baudoin, 1998; Isles et al., 2001). Here we show that in a subspecies recognition context, both males and females M. m. musculus have evolved a homosubspecific preference.

Figure 2.

Homosubspecific preference within Mus musculus musculus and absence of preference within M. m. domesticus (mean ± SE) with reference to their hybridization index value. The ‘central’ part of the hybrid zone (between ‘mus B’ and ‘dom A’) is an extrapolation from previous genetic analyses of the centre of the zone. Note (1) the homogeneity of the responses within each subspecies, (2) the ‘asymmetric’ pattern of preference between the two subspecies despite their symmetric position in the hybrid zone.

A directional preference in a two-way choice context could be a positive response to the specific signals carried by the preferred mate, or either a negative or a none response to those carried by the nonpreferred mouse. Our study sets the choice at a geographical level. Preference for M. m. musculus type signals does not seem to rely significantly on particular characteristics of the Danish pool given that a similar response is obtained when a M. m. musculus population from another geographical area is involved in the choice (Smadja & Ganem, 2002). Moreover, when the Danish M. m. domesticus population is replaced by a population from another origin, our study shows a nonsignificant assortative preference displayed by M. m. musculus. The fact that this preference is not as marked as in the other choice contexts (experiments 1 and 2) could be related to the specific chromosomal characteristics of the Danish populations (Robertsonian fusions: Table 3) which may enhance the cost of hybridization (Fel-Clair et al., 1996) and lead to a stronger selective pressure for assortative signals in Denmark. Another possibility, not excluding the previous, could be that the choice decision may also rely on an evaluation of a difference between the two stimuli; hence a lesser divergence between allopatric signals than between contact signals may occur and could explain a less marked preference.

In contrast with the pattern observed in M. m. musculus, preference in M. m. domesticus is neither directional nor assortative among the sexes and across populations. This pattern is in agreement with two studies (Christophe & Baudoin, 1998; Smadja & Ganem, 2002) but not with a third study which evidenced patterns of assortative mate preference in other M. m. domesticus populations from Denmark (G. Ganem, C. Litel-Mass, C. Smadja, B. Dod & R.C. Kann, unpublished data), therefore suggesting that patterns may not be consistent along the edges of the contact zone. Nevertheless, our study reveals a difference in the mate recognition system of the two subspecies, and further analyses of the pattern of mate preference in M. m. domesticus could be important for the understanding of the extent of premating divergence between the two subspecies.

Constraints on effectiveness of premating divergence

Patterns of mate preference suggest that premating isolation between the two subspecies could occur. However, a crucial point is whether the potentialities assessed under laboratory conditions are effective in the field. Occurrence of a geographical or an ecological barrier as well as within-sex competition can overcome mate preference of one of the sexes (Andersson & Iwasa, 1996). In the house mouse, within-sex competition is known to occur and males generally dominate females (Hurst, 1990a,b,c). Moreover, inter-subspecific male–male competition and dominance of M. m. domesticus over M. m. musculus males (Thuesen, 1977; Van Zegeren & Van Oortmerssen, 1981; Munclinger & Frynta, 2000) could influence the pattern of reproduction between the two subspecies, in zones of overlap or contact, by interfering with the access of males M. m. musculus to females of the same subspecies, hence preventing the expression of females, and possibly males, M. m. musculus mate preference. Nevertheless, a M. m. domesticus gene flow is expected to be disadvantageous to M. m. musculus individuals as a result of the cost of hybridization (Moulia et al., 1991; Dod et al., 1993; Forejt, 1996) and could generate selective pressures that would favour evolution of assortative preference.

Effectiveness of mate preference also depends on rates of encounter between individuals of the two subspecies. The two edges of the hybrid zone are approximately 40 km apart, and recombined populations of mice compose the body of the hybrid zone. Direct estimates of dispersal based on capture-mark-release methods seem to agree on a mean active dispersal distance of around 100 m for the house mouse (Myers, 1974; Carlsen, 1993), with some reports of long-distance dispersal of around 500 m (Cassaing & Croset, 1985; Walkowa et al., 1989). However, passive transport of mice over longer distances could also occur. In the Danish hybrid zone, a study involving microsatellite markers suggests that gene flow between the two subspecies in Denmark would mainly follow a stepping stone scheme, i.e. through intermediate populations across the zone although some long distance migration through passive transport can occasionally occur (Dallas et al., 1995). Therefore, direct encounters between populations from the two edges of the hybrid zone are expected to be infrequent in the field, and mice from the edges are more likely to interact with individuals from adjacent populations and patterns of gene flow between the two subspecies probably rely on mate choice characteristics between adjacent populations along the zone. Consequently, if isolation does not occur between neighbouring populations of a slightly different genotype, gradual gene flow between potentially isolated extremities of the hybrid zone would occur (buffer effect) (Howard et al., 1993; Britton-Davidian et al., 2002). Nevertheless, a strong assortative preference seems to have evolved in M. m. musculus, suggesting that the premating barrier could be effective. A process involving preference for the less introgressed population, or own population preference (Nagel & Schluter, 1998) can lead to sexual selection of partners of own type and may explain the pattern disclosed in Denmark. Additionally, a behavioural hybrid disadvantage in the zone would tend to promote isolation (Hatfield & Schulter, 1996; Vamosi & Schluter, 1999; Naisbit et al., 2001).

Asymmetric premating isolation

The different patterns of mate preference displayed by M. m. musculus and M. m. domesticus could lead to an asymmetric sexual isolation between the two subspecies. In other studies where asymmetric patterns of mate preference were found, it was suggested that it could contribute to a shift of the hybrid zone leading to the replacement of one of the subspecies by the other (Bella et al., 1992; Patton & Smith, 1993; Shapiro, 2001). Therefore, we could predict that the asymmetrical mating preference observed in our study could lead to the replacement of the most permissive subspecies (M. m. domesticus) by the more selective subspecies (M. m. musculus). However, in Denmark, dominance of M. m. domesticus males over M. m. musculus could suffice to protect the subspecies from heterosubspecific gene flow, preventing M. m. musculus dispersers access to reproduction with females M. m. domesticus. Under this hypothesis we might expect a symmetric premating isolation between the two subspecies, through two types of mechanisms: dominance and assortative mate preference.

Effective premating isolation and the fate of the Danish hybrid zone

A crucial question is whether the patterns of mate preference assessed in our study would affect the dynamic and the fate of the hybrid zone. At present, the occurrence of many distinct populations recombined to a different extent and not showing a heterozygote deficit confers a unimodal configuration to the Danish house mouse hybrid zone (Dallas et al., 1995). If isolating mechanisms are too weak or not maintained through time, unimodal hybrid zones could be an intermediate stage preceding extinction of one or both taxa or their fusion (Paterson, 1978; Liou & Price, 1994), and this could be the case in Denmark. Nevertheless, our study demonstrates a strong premating divergence at the edges of the house mouse hybrid zone. If this divergence evolved before the contact, in allopatry, the patterns observed at the edges of the zone could be related to gene flow from allopatric populations. Under the latter hypothesis, hybridization would have been the result of an accident (Rhymer & Simberloff, 1996). Moreover, the fact that this divergence is still maintained nowadays could suggest that gene flow from the centre of the zone either does not occur or is too weak in the face of gene flow from outside the zone. Then, we may predict that the evolution of the hybrid zone towards bimodality would depend on the stability of the central recombined populations. Alternatively, premating divergence could have evolved, or further evolved, in the contact zone in response to selection against hybridization (Liou & Price, 1994). Indeed, the hybrid disadvantage demonstrated between the two subspecies (Dod et al., 1993; Moulia et al., 1993; Fel-Clair et al., 1996; Forejt, 1996) could have generated sufficient selection to promote divergence in the contact zone. However, the latter hypothesis raises the question of how selection would be effective given that the two subspecies are physically separated by central recombined populations. Hence, either the divergence may be a residual effect of selection on initial contact, or the recombined central populations may generate or maintain the necessary selective pressures through a process similar to that described earlier in this discussion. If premating divergence has evolved in response to selection against hybridization (reinforcement), a counter effect of gene flow from outside the zone (Sanderson, 1989) could be limited by the particular geographical characteristics of Jutland peninsula in which M. m. musculus is relatively isolated (Fig. 1), and that may be a favourable geographical context for divergence to be maintained. In contrary, M. m. domesticus, which is outside the peninsula, may not benefit from a geographically favourable context, a pattern consistent with the asymmetric divergence suggested by our results. Under the above scenario, again, the dynamic of the recombined populations would determine whether bimodality would evolve or not.

Finally, a shift between unimodality and bimodality may not occur in Denmark because the house mouse intermediate recombined populations may have a dynamic of their own: they occur in similar densities as the other populations (G. Ganem, personal observation, 2002) and seem to have evolved specific adaptations (Alibert et al., 1994; N. Raufaste, A. Orth, C. Smadja, K. Belkhir, F. Bonhomme & P. Boursot, unpublished data): thus, they could be maintained and constitute a stable geographical barrier that could evolve independently of the premating barriers at the edges of the zone.


We wish to warmly thank the Danish farmers of Jutland who kindly accepted our intrusion, and particularly J. Jakobsen and H. Juhl. Special thanks to J. T. Nielsen and R.C. Karn for assistance in the field. We are grateful to R. Osterballe and Givskud zoo for their warm support, and to T. Secher Jensen (Natural History Museum of Aarhus) for his help and interest. We thank P. Boursot, F. Bonhomme, B. Dod and J. Britton-Davidian for discussions on the Danish hybrid zone, and information on molecular characteristics of our mice. We thank N. Pillay, P. Boursot and F. Bonhomme for their comments on the manuscript, and Gilbert Pistre for the design and building of the behavioural apparatus, M. Perriat-Sanguinet for looking so well after our mice, and J. Lopez for her help with the allozymes. This is a contribution of UMR 5554, No ISEM 2003/070.