21
Copyright Ó 2011 by the Genetics Society of America DOI: 10.1534/genetics.110.123851 Evolution of the Genomic Recombination Rate in Murid Rodents Beth L. Dumont 1 and Bret A. Payseur 2 Laboratory of Genetics, University of Wisconsin, Madison, Wisconsin 53706 Manuscript received October 5, 2010 Accepted for publication December 6, 2010 ABSTRACT Although very closely related species can differ in their fine-scale patterns of recombination hotspots, variation in the average genomic recombination rate among recently diverged taxa has rarely been surveyed. We measured recombination rates in eight species that collectively represent several temporal scales of divergence within a single rodent family, Muridae. We used a cytological approach that enables in situ visualization of crossovers at meiosis to quantify recombination rates in multiple males from each rodent group. We uncovered large differences in genomic recombination rate between rodent species, which were independent of karyotypic variation. The divergence in genomic recombination rate that we document is not proportional to DNA sequence divergence, suggesting that recombination has evolved at variable rates along the murid phylogeny. Additionally, we document significant variation in genomic recombination rate both within and between subspecies of house mice. Recombination rates estimated in F 1 hybrids reveal evidence for sex-linked loci contributing to the evolution of recombination in house mice. Our results provide one of the first detailed portraits of genomic-scale recombination rate variation within a single mammalian family and demonstrate that the low recombination rates in laboratory mice and rats reflect a more general reduction in recombination rate across murid rodents. T HE rate of meiotic recombination is an important parameter in genetics and evolution. Recombina- tion rate determines the effects of selection on nearby polymorphisms (Maynard Smith and Haigh 1974; Charlesworth et al. 1993), dictates the magnitude of correlations between genetic variants in populations (Pritchard and Przeworski 2001), shapes features of the genomic landscape (Duret and Arndt 2008), and determines the resolution of genetic mapping experi- ments. In addition, recombination ensures that ho- mologous chromosomes align and disjoin correctly during the first meiotic division. Homologous chro- mosome pairs that fail to undergo a recombination event face a high risk of nondisjunction, an outcome with strongly deleterious consequences for the organ- ism (Hassold and Hunt 2001). Variation in recombination rate has been docu- mented within genomes, between individuals, and among species (Coop and Przeworski 2007). Much of this variation is thought to derive from differences in the density and intensity of recombination hotspots (Coop et al. 2008; Webb et al. 2008; Paigen and Petkov 2010), short genomic regions with very high rates of crossing over. Recombination rates on this fine scale evolve rapidly (Ptak et al. 2005; Winckler et al. 2005; Jeffreys and Neumann 2009), and loci that confer differences in hotspot activity within and between species have been identified ( Jeffreys and Neumann 2005; Baudat and de Massy 2007; Parvanov et al. 2009, 2010; Baudat et al. 2010). Even as information about the mechanisms of fine- scale recombination rate variation accumulates, the genetics and evolution of a more basic parameter—the total number of crossovers in a meiotically dividing cell—remain poorly understood. Mammals exhibit a wide range of genomic recombination rates (Burt and Bell 1987; Coop and Przeworski 2007; Dumont and Payseur 2008), providing an excellent system for evaluating evolutionary trends on this scale. Average genomic recombination rates are phylogenetically distri- buted across mammalian taxa (Dumont and Payseur 2008), indicating that more closely related species have more similar average genomic recombination rates. House mice and rats, two murid rodent species, provide an especially illustrative case: compared with other eutherian mammals, these species have markedly re- duced genomic recombination rates (Burt and Bell 1987; Dumont and Payseur 2008). For example, the genetic linkage maps of house mice and rats are scarcely half as long as the human map (Broman et al. 1998; Steen et al. 1999; Kong et al. 2002; Cox et al. 2009), despite the approximate conservation of physical ge- nome size between these species (Gibbs et al. 2004). Supporting information is available online at http://www.genetics.org/ cgi/content/full/genetics.110.123851/DC1. 1 Present address: Department of Genome Sciences, University of Washington, Seattle, WA 98195. 2 Corresponding author: University of Wisconsin, Laboratory of Genetics, Genetics/Biotechnology 2428, 425-G Henry Mall, Madison, WI 53706. E-mail: [email protected] Genetics 187: 643–657 (March 2011)

Evolution of the Genomic Recombination Rate in Murid Rodents · Evolution of the Genomic Recombination Rate in Murid Rodents ... data from genetic linkage maps ... Evolution of the

Embed Size (px)

Citation preview

Copyright � 2011 by the Genetics Society of AmericaDOI: 10.1534/genetics.110.123851

Evolution of the Genomic Recombination Rate in Murid Rodents

Beth L. Dumont1 and Bret A. Payseur2

Laboratory of Genetics, University of Wisconsin, Madison, Wisconsin 53706

Manuscript received October 5, 2010Accepted for publication December 6, 2010

ABSTRACT

Although very closely related species can differ in their fine-scale patterns of recombination hotspots,variation in the average genomic recombination rate among recently diverged taxa has rarely beensurveyed. We measured recombination rates in eight species that collectively represent several temporalscales of divergence within a single rodent family, Muridae. We used a cytological approach that enables insitu visualization of crossovers at meiosis to quantify recombination rates in multiple males from eachrodent group. We uncovered large differences in genomic recombination rate between rodent species,which were independent of karyotypic variation. The divergence in genomic recombination rate that wedocument is not proportional to DNA sequence divergence, suggesting that recombination has evolved atvariable rates along the murid phylogeny. Additionally, we document significant variation in genomicrecombination rate both within and between subspecies of house mice. Recombination rates estimated inF1 hybrids reveal evidence for sex-linked loci contributing to the evolution of recombination in housemice. Our results provide one of the first detailed portraits of genomic-scale recombination rate variationwithin a single mammalian family and demonstrate that the low recombination rates in laboratory miceand rats reflect a more general reduction in recombination rate across murid rodents.

THE rate of meiotic recombination is an importantparameter in genetics and evolution. Recombina-

tion rate determines the effects of selection on nearbypolymorphisms (Maynard Smith and Haigh 1974;Charlesworth et al. 1993), dictates the magnitude ofcorrelations between genetic variants in populations(Pritchard and Przeworski 2001), shapes features ofthe genomic landscape (Duret and Arndt 2008), anddetermines the resolution of genetic mapping experi-ments. In addition, recombination ensures that ho-mologous chromosomes align and disjoin correctlyduring the first meiotic division. Homologous chro-mosome pairs that fail to undergo a recombinationevent face a high risk of nondisjunction, an outcomewith strongly deleterious consequences for the organ-ism (Hassold and Hunt 2001).

Variation in recombination rate has been docu-mented within genomes, between individuals, andamong species (Coop and Przeworski 2007). Muchof this variation is thought to derive from differences inthe density and intensity of recombination hotspots(Coop et al. 2008; Webb et al. 2008; Paigen and Petkov

2010), short genomic regions with very high rates of

crossing over. Recombination rates on this fine scaleevolve rapidly (Ptak et al. 2005; Winckler et al. 2005;Jeffreys and Neumann 2009), and loci that conferdifferences in hotspot activity within and betweenspecies have been identified ( Jeffreys and Neumann

2005; Baudat and de Massy 2007; Parvanov et al.2009, 2010; Baudat et al. 2010).

Even as information about the mechanisms of fine-scale recombination rate variation accumulates, thegenetics and evolution of a more basic parameter—thetotal number of crossovers in a meiotically dividingcell—remain poorly understood. Mammals exhibit awide range of genomic recombination rates (Burt andBell 1987; Coop and Przeworski 2007; Dumont andPayseur 2008), providing an excellent system forevaluating evolutionary trends on this scale. Averagegenomic recombination rates are phylogenetically distri-buted across mammalian taxa (Dumont and Payseur

2008), indicating that more closely related species havemore similar average genomic recombination rates.House mice and rats, two murid rodent species, providean especially illustrative case: compared with othereutherian mammals, these species have markedly re-duced genomic recombination rates (Burt and Bell

1987; Dumont and Payseur 2008). For example, thegenetic linkage maps of house mice and rats are scarcelyhalf as long as the human map (Broman et al. 1998;Steen et al. 1999; Kong et al. 2002; Cox et al. 2009),despite the approximate conservation of physical ge-nome size between these species (Gibbs et al. 2004).

Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.110.123851/DC1.

1Present address: Department of Genome Sciences, University ofWashington, Seattle, WA 98195.

2Corresponding author: University of Wisconsin, Laboratory of Genetics,Genetics/Biotechnology 2428, 425-G Henry Mall, Madison, WI 53706.E-mail: [email protected]

Genetics 187: 643–657 (March 2011)

Similarly, crossover counts visualized as chiasma inmeiotically dividing cells are considerably lower in miceand rats than in other eutherian taxa (Burt and Bell

1987).The rate of recombination is correlated with chro-

mosome arm number in mammals (Dutrilleux 1986;Pardo-Manuel de Villena and Sapienza 2001; Li andFreudenberg 2009), suggesting that low recombina-tion in laboratory rodents may partially derive fromtheir unique karyotypes. The house mouse genome isorganized into 19 acrocentric autosomes and a pair ofsex chromosomes, while the rat genome is composed of20 acrocentric autosomes and a sex chromosome pair.In contrast, most mammalian genomes are partitionedover a larger number of bi-armed chromosomes. Com-pared with other mammals, house mice and rats areclear outliers with regard to the number of chromosomearms in their haploid genome [mean across mammals is�30 (Pardo-Manuel de Villena and Sapienza 2001)].Nonetheless, even after adjusting recombination ratesby chromosome arm number (or chromosome num-ber), house mice and rats still possess unusually lowrecombination rates for mammalian species (Dumont

and Payseur 2008).The low rate of recombination in house mice and rats

could reflect natural selection to reduce recombinationintensity along the rodent lineage. There is significantgenetic variation for recombination rate among inbredstrains of house mice (Koehler et al. 2002; Dumont

et al. 2009), suggesting that the wild populations fromwhich they are derived harbor ample raw material toenact an evolutionary response. Theoretical models in-dicate that a reduction in the rate of recombination isselectively favored when high-fitness alleles are linked incoupling phase (Felsenstein 1965; Barton 1995), butthe extent to which such conditions are upheld innature remains unknown. On the other hand, the in-breeding and reduction in population size that accom-panied the creation of the laboratory inbred strainsmight be expected to inadvertently select for increasesin recombination rate to facilitate purging combina-tions of deleterious, recessive alleles from the genome(Felsenstein 1974; Otto and Barton 2001). Consis-tent with this latter prediction, domesticated plants andanimals have elevated recombination rates relative totheir wild progenitors and nondomesticated species(Burt and Bell 1987; Ross-Ibarra 2004).

To begin to understand the evolutionary processescontributing to reduced recombination in house miceand rats, recombination rate estimates from additionalrodent taxa are required. Genetic maps have beenconstructed for Peromyscus polionotus (Steiner et al.2007) and the Syrian hamster (Okuizumi et al. 1997),both yielding total map lengths �15–25% shorter thanthose of laboratory mice or rats. Chiasma counts areavailable for several additional rodent species, includingwood mice, grass mice, water rats, Mongolian gerbils,

steppe lemmings, and European hamsters (Dutrilleux

1986; Burt and Bell 1987). However, chiasma countsare difficult to relate directly to meiotic recombinationrate, as they are commonly biased downward (Morton

1991; Nilsson et al. 1993). Nonetheless, data fromgenetic linkage maps and chiasma counts seem tosuggest that all rodents—not only house mice andlaboratory rats—are characterized by low recombina-tion rates. However, no comparative recombination ratedata sets have been collected by a common set ofinvestigators using the same method (thereby minimiz-ing experimental and observer biases), nor have suchdata sets been examined in an evolutionary framework.

In this study, we harness the power of an immunocy-tological method for measuring genomic recombina-tion rates in individual males to quantify recombinationrate variation across rodents. This approach uses fluo-rescently labeled antibodies to visualize discrete foci ofMLH1 along the synaptonemal complex at late pachy-tene of prophase I (Anderson et al. 1999; Koehler et al.2002). MLH1 is a mismatch repair protein that localizesto late recombination nodules corresponding to sitesof reciprocal crossing over. Multiple studies haveconfirmed that MLH1 foci faithfully depict both thenumber and the distribution of recombination eventsin male meiosis (Baker et al. 1996; Anderson et al. 1999;Koehler et al. 2002). We measure the genomic rateof recombination in eight rodent taxa, focusing specifi-cally on the murid rodents (mice and rats). We documentconsiderable variation for recombination rate over arange of evolutionary time scales and find that a low re-combination rate is a characteristic feature of Muridae.

MATERIALS AND METHODS

Animal husbandry and experimental crosses: Wild-derivedinbred strains CAROLI/EiJ (Mus caroli), CAST/EiJ (Musmusculus castaneus), CZECHI/EiJ (M. m. musculus), PANCEVO/EiJ (Mus spicilegus), PERA/EiJ (M. m. domesticus), PWD/PhJ(M. m. musculus), and WSB/EiJ (M. m. domesticus) werepurchased from the Jackson Laboratory and housed in theUniversity of Wisconsin School of Medicine and Public Healthmouse facility according to animal care protocols approved bythe University of Wisconsin Animal Care and Use Committee.Mice were provided with food (2016S Teckland Global Rodentdiet) and chlorinated water ad libitum. Adult males from eachinbred strain and F1 males from intra- and inter-subspecificcrosses were sacrificed at 8–10 weeks of age, with the exceptionof PANCEVO/EiJ (sacrificed at 53.5 weeks).

Wild-derived inbred strains are difficult to breed. In ourcare, CAST/EiJ males performed poorly in all crosses, butmost notably in crosses to conspecific females. Our analysesused CAST males purchased directly from the vendor tocircumvent this challenge. Likewise, our breeding pair ofPANCEVO/EiJ failed to produce a single litter after 9 months.The single PANCEVO male considered in this study was themale originally purchased for breeding purposes.

Males from inbred strain CIM (M. m. castaneus) wereobtained from Francxois Bonhomme’s stock repository at theUniversite Montpellier II. Animals were sacrificed shortly after

644 B. L. Dumont and B. A. Payseur

arrival to the University of Wisconsin-Madison (aged 24.5–35.5weeks). Because these animals were not reared under constantlaboratory conditions, phenotypic variation among inbredCIM animals could have partially derived from differences inenvironmental exposure.

Testis tissue from adult male rats (COP/CrCrlWKY/NCrl con-genic animals) was kindly donated by Michael Gould’slaboratory at the University of Wisconsin. Melissa Gray pro-vided testis tissues from Gough Island mice maintained as anoutbred colony at the University of Wisconsin.

Wild Peromyscus maniculatus and Microtus pennsylvanicuswere live-trapped in the Biocore Prairie on the LakeshoreNature Preserve on the University of Wisconsin-Madisoncampus. Approximately 60 Sherman traps (3 3 3.5 3 9 in.)were baited with a peanut-butter-and-oatmeal mixture, setovernight, and checked at dawn. Handling and euthanasia ofwild rodents was performed as per protocols approved by theUniversity of Wisconsin Animal Care and Use Committee.

Spermatocyte spreads and immunostaining: Spreads ofearly meiotic cells were prepared as described by Peters

et al. (1997). Briefly, the left testis of sexually mature males wasremoved, weighed, and rinsed in sterile 13 PBS. The outertissue coating of the testis was punctured to allow a smallvolume of seminiferous tubules to be extracted. Tubules wereincubated in a hypotonic solution (30 mm Tris, 50 mm sucrose,17 mm citric acid, 5 mm EDTA, 2.5 mm dithiothreitol, 0.5 mm

phenylmethanesulfonyl fluoride) for �45 min at room tem-perature. Tubules were then transferred to a small volume(20 ml) of 100 mm sucrose solution deposited on a clean glassslide and shredded using fine-gauge forceps. Tubular rem-nants were removed, and an additional 20 ml of 100 mm

sucrose was added to the cell slurry. The solution was agi-tated by pipetting, and 20 ml was deposited onto each of two3- 3 1-in. glass slides coated with 100 ml 1% paraformalde-hyde supplemented with Triton X-100 (0.15%; pH 9.2). Theslides were gently rocked to distribute cells across their surfaceand allowed to dry overnight in a room-temperature hu-mid chamber. Dried slides were then washed briefly in0.4% PhotoFlo (Kodak), air dried, and subjected toimmunostaining.

The immunostaining protocol was adapted from Anderson

et al. (1999) and Koehler et al. (2002). A 103 concentration ofantibody dilution buffer (ADB) was prepared [2.5 ml normaldonkey serum ( Jackson ImmunoResearch), 22.5 ml 13 PBS,0.75 g bovine serum albumin (Fraction V; Fisher Scientific),and 12.5 ml Triton X-100] and sterilized by vacuum filtration(0.45 mm; Millipore). Slides were blocked in 13 ADB (dilutedin 13 PBS) for �30 min and then lightly drained by touchingthe edge of the slide to a clean paper towel. All antibodydilutions were made into 13 ADB and all incubations wereperformed in a 37� humid chamber. A 60-ml aliquot of primaryantibody cocktail [1:50 rabbit polyclonal antibody againstMLH1 (Calbiochem) and 1:50 goat polyclonal antibodyagainst SYCP3 (SantaCruz Biotechnology)] was dispensed oneach slide. Slides were cover-slipped, sealed with rubbercement, and incubated overnight. The rubber cement wasthen carefully removed and coverslips were soaked off in 13ADB. Slides were washed twice for 30 min each in 13 ADB. A60ml volume of 1:100 Alexa 488 donkey anti-rabbit secondaryantibody (Molecular Probes) was deposited on each slide.These slides were cover-slipped, sealed with rubber cement,and incubated overnight. After soaking off coverslips in 13ADB, 60 ml of 1:100 Alexa 568 donkey anti-goat secondaryantibody (Molecular Probes) was applied to each slide. Slideswere sealed with a parafilm ‘‘coverslip’’ and incubated for 2 hr.Slides were then washed three times for 1 hr each in 13 PBS,air-dried, and mounted in a drop of ProLong Gold antifademedia (Molecular Probes).

Imaging and pachytene cell scoring: Cells were visualizedusing a Zeiss Axioskop microscope equipped with an AxioCamHRc camera and a 1003 objective lens. Late pachytene cellsthat were damaged during preparation, displayed bulbouschromosome termini (indicative of transition into diplotene),lacked clear cell boundaries, or displayed flagrant defects insynapsis were not imaged. Images were captured in AxioVision(Rel. 4.8) software and stored as high-resolution tiff files.Images were subsequently cropped and the fluorescentintensity was adjusted using ImageJ software. Numbers ofautosomal MLH1 foci in late pachytene cells were scoredblindly with respect to species designation. Exceptions includecells from rat, M. pennsylvanicus, and P. maniculatus, whichcould not be scored naively due to the unique number ofchromosomes in these species genomes. However, to the bestof our knowledge, MLH1 foci analysis has never beenperformed in these taxa, and we had no a priori informationregarding the expected number of MLH1 foci per cell to biasour scoring. Only cells characterized by (i) the completemerger of SYCP3 signals from the two homologs, (ii) a fullcomplement of chromosomes, (iii) clear, brightly stainedMLH1 foci, and (iv) minimal background fluorescence werescored. We retained only cells with at least one MLH1 focus oneach synaptonemal complex, excepting the possibility of oneachiasmate bivalent; cells with two or more synaptonemalcomplexes lacking a MLH1 focus were extremely rare andlikely represented staining artifacts.

The average number of MLH1 foci in late pachytenespermatocytes from Mus spretus (wild-derived inbred strainSPRET/EiJ) was obtained from the literature (Koehler et al.2002).

Total autosomal synaptonemal complex length was mea-sured using Micromeasure (v. 3.2) (Reeves and Tear 2000).

Phylogenetic and statistical analyses: Phylogenetic treeswere constructed from partial coding sequence data from Cytband Irbp; these were the only two loci with publicly availablesequence data for all taxa considered in this analysis. Bothtrees were rooted using the outgroup taxon Spermophilustridecemlineatus. Optimal sequence alignments were con-structed using default parameters in ClustalW (Thompson

et al. 1994) and manually verified in BioEdit (v. 7.0.5; http://www.mbio.ncsu.edu/BioEdit/bioedit.html). GenBank acces-sion numbers are provided in the supporting information,Table S1.

We used PhyML (Guindon and Gascuel 2003) to identifythe model of molecular evolution that yielded the highestlikelihood support for these data as assessed by the Akaikeinformation criterion. A general time-reversible model ofmolecular evolution with the proportion of invariant sitesand optimal nucleotide frequencies estimated from the dataand gamma distributed rate variation across four classes wasselected for the Cytb locus. The Tamura and Nei (1993) modelwith the proportion of invariant sites and optimal transition/transversion ratio estimated from the data was selected for theIrbp locus. Although these models are heavily overparameter-ized for the small data set considered here, simpler models ofDNA sequence evolution provide qualitative results identicalto those presented below (data not shown).

All statistical analyses were performed in the R environmentfor statistical computing (R Development Core Team 2008).The extremely unequal sample sizes in some strain, sub-species, and species comparisons complicate pairwise compar-isons. For each comparison, we pared down the larger dataset using a random subsample equal in size to the smaller dataset and compared this subsample to the entire smaller data set.Results from Mann–Whitney U-tests presented below reflectthe mean P-value of 10,000 tests performed on independentsubsamples of the larger data set. In comparisons that pool

Recombination Rate Evolution in Rodents 645

variation among inbred strains within a subspecies, we usedequally sized random draws from each strain. Similarly, we usea balanced subsample in species analyses where subspecies-level data are available.

RESULTS

We scored the total number of MLH1 foci on au-tosomal synaptonemal complexes in seven or morespermatocytes from each of 64 adult male rodents fromeight taxa (Table 1). Representative images of latepachytene spermatocytes stained with fluorescentlylabeled antibodies against MLH1 and SYCP3 are shownin Figure 1.

The mean number of MLH1 foci varies tremendouslyamong taxa, ranging from a low of 21.78 foci in a wild-derived inbred strain of M. m. castaneus (CAST/EiJ) to ahigh of 34.46 foci in the rat (Table 1; Figure 2). Theestimated mean number of MLH1 foci in strain CAST/EiJ is in good agreement with a previous estimate fromthis strain, and the average MLH1 foci counts in the twoinbred strains of M. m. domesticus are similar to thosefrom common inbred laboratory strains of house mice(Koehler et al. 2002), which are predominantly of M. m.domesticus origin (Yang et al. 2007). Assuming that eachfocus corresponds to a genetic map length of 50 cM,total male map lengths differ by .600 cM among theserodent species. This variation is remarkable in light ofthe rough conservation of overall genome size in muridrodents (Gregory 2005).

A minimum of one crossover per chromosome arm isrequired to ensure the correct alignment of homolo-gous chromosomes at the metaphase plate and theirproper segregation during meiosis I, although recentstudies have suggested that this constraint could berelaxed to one crossover per chromosome (Borodin

et al. 2008; Hassold et al. 2009). The average number ofrecombination events per autosome or per autosomalarm still varies markedly among the eight rodent taxaexamined here (Table 1), suggesting that observed var-iation in raw MLH1 foci counts does not solely derivefrom differences in gross genome organization amongrodent species.

The standard deviation (SD) of MLH1 foci countsalso varies across taxa (Table 1). In general, animals withhigher recombination rates tend to have greater vari-ability in their MLH1 foci counts (Spearman’s r ¼ 0.57,P ¼ 9.51 3 10�7). Interestingly, two taxa for whichsamples came from outbred individuals, P. maniculatusand M. pennsylvanicus, have the lowest SDs in genomicrecombination rate. Both species have average MLH1foci counts that approximate their autosomal chromo-some numbers (Table 1), indicating that very rigidregulatory mechanisms must be in place to ensure thatone (and usually only one) crossover is allocated to agiven chromosome. In addition, both of these specieshave fewer MLH1 foci than chromosome arms (Table

1), complementing previous assertions that only onecrossover per chromosome is necessary for correct ho-molog segregation (Borodin et al. 2008; Hassold et al.2009), at least in some taxa.

Polymorphism and subspecies divergence for maleMLH1 foci counts in house mice: Our data set includesMLH1 foci counts from multiple males from each of twowild-derived inbred strains representing each of thethree principal subspecies of house mice, M. m. muscu-lus, M. m. domesticus, and M. m. castaneus (Table 1). Thishierarchical structure allowed us to test for evolutionarydifferences in genomic recombination rate betweentaxa at various stages of early divergence. We findsignificant differences in MLH1 foci counts fromspermatocytes of independent inbred strains of thesame house mouse subspecies [Mann–Whitney U-testwith subsampling to account for differences in samplesize (Table 2)]. Likewise, we find significant variationamong laboratory-reared outbred males from a GoughIsland population of presumed M. m. domesticus origin(Kruskal–Wallis rank-sum test, x2¼ 9.7544, P¼ 0.0076).Combined, these findings indicate that natural popula-tions of house mice harbor polymorphism for globalrecombination rate, confirming previous predictionsbased on analyses of laboratory mice (Dumont et al.2009). Similarly, differences in MLH1 foci number be-tween the three subspecies of house mice are highlystatistically significant in pairwise comparisons [Mann–Whitney U-test with subsampling (Table 2)].

Although there are significant differences in MLH1foci count both within and between subspecies of housemice, subspecies divergence is considerably greaterthan intra-subspecies polymorphism (Table 1; Figure2). Independent wild-derived inbred strains within M.m. castaneus and M. m. domesticus differ by an average ofless than one MLH1 focus, while inbred strains of M. m.musculus differ by a mean of approximately three foci. Incontrast, the M. m. musculus genome undergoes approx-imately six recombination events more than either theM. m. domesticus or the M. m. castaneus genome at mei-osis. Much of the remaining variation in the MLH1 focicounts among house mice reflects differences amongcells sampled from a single animal, with a lessercontribution of variation among animals of a commonstrain.

Our MLH1 data set features ordinal data, with un-balanced sampling at each hierarchical level and un-equal variances among strains, among subspecies, andamong animals. These features violate key assumptionsof standard parametric (e.g., ANOVA) and nonparamet-ric (e.g., Kruskal–Wallis one-way ANOVA) methods,limiting further statistical analysis of these data. None-theless, we note that a fully nested sum-of-squares-basedapproach for partitioning the variance in MLH1 focicounts recapitulates the visually observable trendsenumerated above. Namely, most of the variation inMLH1 counts among wild-derived inbred strains of

646 B. L. Dumont and B. A. Payseur

TABLE 1

MLH1 foci counts in Muridae

MLH1 count

Species Subspecies Strain AnimalNo. ofcells Mean SD Range

Per chromosome(per chromosome arm)

M. musculus M. m. musculus CZECHI 1 25 26.76 1.83 23–31 1.41 (1.41)2 23 27.17 2.12 24–32 1.43 (1.43)3 24 26.67 2.08 22–30 1.40 (1.40)4 19 26.16 2.87 20–32 1.38 (1.38)5 22 26.68 2.21 23–31 1.40 (1.40)6 22 27.54 2.60 23–32 1.45 (1.45)7 21 26.52 2.11 22–31 1.40 (1.40)

Total 156 26.80 2.25 20–32 1.41 (1.41)PWD 1 15 29.60 2.57 23–32 1.56 (1.56)

2 18 29.39 2.95 23–34 1.55 (1.55)3 21 31.19 2.06 27–35 1.64 (1.64)4 21 30.62 1.86 26–34 1.61 (1.61)5 21 29.19 2.58 24–34 1.54 (1.54)6 22 30.55 2.09 26–35 1.61 (1.61)7 16 28.56 2.83 23–32 1.50 (1.50)8 22 29.77 2.39 25–35 1.57 (1.57)9 18 29.00 2.83 23–34 1.53 (1.53)

10 18 30.83 2.20 27–36 1.62 (1.62)Total 192 29.92 2.51 23–36 1.57 (1.57)

M. m. domesticus WSB 1 20 22.10 1.68 19–25 1.16 (1.16)2 22 21.91 1.72 20–25 1.15 (1.15)3 20 22.10 1.68 20–26 1.16 (1.16)4 19 22.74 1.45 20–26 1.20 (1.20)5 18 22.67 1.81 20–26 1.19 (1.19)

Total 99 22.28 1.67 19–26 1.17 (1.17)PERA 1 32 23.91 2.01 20–27 1.26 (1.26)

2 21 24.00 2.17 21–30 1.26 (1.26)3 21 23.57 1.78 21–27 1.24 (1.24)4 25 22.56 1.61 20–26 1.19 (1.19)5 25 23.12 1.56 20–26 1.22 (1.22)6 18 22.94 2.24 19–28 1.21 (1.21)7 22 22.60 1.71 20–25 1.19 (1.19)8 29 23.00 1.58 20–27 1.21 (1.21)9 28 22.11 1.75 19–27 1.16 (1.16)

10 24 22.21 1.67 19–26 1.17 (1.17)11 24 24.17 1.90 21–28 1.27 (1.27)12 23 22.91 1.65 20–27 1.21 (1.21)

Total 292 23.09 1.89 19–30 1.22 (1.22)M. m. castaneus CAST 1 37 22.00 2.00 19–26 1.16 (1.16)

2 29 21.76 1.70 19–25 1.15 (1.15)3 25 21.48 1.45 19–26 1.13 (1.13)

Total 91 21.78 1.76 19–26 1.15 (1.15)CIM 1 16 24.69 2.33 21–29 1.30 (1.30)

2 50 22.44 1.77 19–27 1.18 (1.18)3 31 22.32 2.27 19–27 1.17 (1.17)4 10 21.70 2.00 20–26 1.14 (1.14)

Total 107 22.67 2.19 19–29 1.19 (1.19)Gough Island Laboratory outbred 1 23 24.09 2.63 19–29 1.27 (1.27)

2 21 25.86 3.00 19–30 1.36 (1.36)3 21 23.09 2.21 18–26 1.22 (1.22)

Total 65 24.34 2.83 18–30 1.28 (1.28)

M. spicilegus PANCEVO 1 44 24.18 1.77 21–28 1.27 (1.27)

M. caroli CAROLI 1 17 29.06 2.28 24–32 1.53 (1.53)2 20 29.55 2.58 23–34 1.56 (1.56)

(continued )

Recombination Rate Evolution in Rodents 647

house mice is accounted for by subspecies differences(�60%), with much of the residual variation explainedby within-animal sampling variation (�30%). Subspe-cies polymorphism and variation among animals from asingle inbred strain each account for a small percentageof the variance in MLH1 foci counts among rodents. Weobtain quantitatively similar results when an identicalanalysis is performed on rank-transformed MLH1counts. In light of the departures from model assump-tions, we caution against over-interpretation of the roughpercentages presented here. However, the general pat-terns of recombination rate variation that we identify areconsistent across statistical approaches.

Species divergence for male MLH1 counts inMuridae: We also tested for differences in MLH1 focicount among murid rodent species. Here, we pooledthe three house mouse subspecies, treating them as asingle species. We again exclude the Gough Island micefrom this analysis because of their uncertain taxonomicstatus within M. musculus. Most pairwise species compar-isons are statistically significant (Mann–Whitney U-test;Table 3). The differences in MLH1 foci number be-tween species become even more dramatic when valuesare transformed to remove the effect of karyotype (i.e.,by subtracting one focus for each additional chro-

mosome arm in the species with the more complexkaryotype).

Most of the observed variation in MLH1 foci countsamong rodent species derives from differences amonganimals within species (Figure 2), a pattern that largelystems from the considerable variation within andbetween animals in the M. musculus species complex(Figure 2). Using nested sum-of-squares methodology,we estimate this percentage to be �45%, although weacknowledge the considerable uncertainty associatedwith this value due to violation of parametric assump-tions. A large fraction of the divergence in MLH1 countsis also due to divergence among murid rodent taxa(�25%). We conclude that subspecies and species-leveltaxonomic status explains the majority of the variationin recombination rate among the murid rodents that wesurveyed.

Relationship between recombination and DNA se-quence divergence: Mean MLH1 foci counts appear tobe randomly distributed with respect to the rodentphylogeny (Figure 2). M. m. castaneus and M. m.musculus, two of the most closely related taxa included inthis analysis, exhibit relatively large differences in recom-bination rate. Two highly divergent taxa, M. pennsylva-nicus and P. maniculatus, have very similar recombination

TABLE 1

(Continued)

MLH1 count

Species Subspecies Strain AnimalNo. ofcells Mean SD Range

Per chromosome(per chromosome arm)

3 21 30.29 2.74 26–36 1.59 (1.59)Total 58 29.67 2.56 23–36 1.56 (1.56)

R. norvegicus COP 1 20 34.55 2.68 30–40 1.73 (1.73)2 21 34.10 2.51 30–39 1.70 (1.70)3 16 33.81 1.83 31–37 1.69 (1.69)4 19 34.79 1.81 30–37 1.74 (1.74)5 20 34.95 2.06 31–38 1.75 (1.75)

Total 96 34.46 2.22 30–40 1.72 (1.72)

P. maniculatusa Wild outbred 1 20 23.50 0.83 22–26 1.02 (0.67)2 16 24.44 1.03 23–26 1.06 (0.70)3 21 23.71 0.96 22–26 1.03 (0.68)4 22 23.41 0.85 22–26 1.02 (0.67)5 20 23.45 0.94 22–25 1.02 (0.67)6 19 23.21 0.42 23–24 1.01 (0.66)7 20 24.60 1.05 23–27 1.07 (0.70)8 10 23.20 0.42 23–24 1.01 (0.66)

Total 148 23.70 0.98 22–27 1.03 (0.68)

M. pennsylvanicus Wild outbred 1 7 23.29 1.70 22–27 1.06 (0.97)2 23 23.87 1.18 22–26 1.08 (0.99)3 20 23.75 1.77 21–29 1.08 (0.99)

Total 50 23.74 1.50 21–29 1.08 (0.99)

a P. maniculatus possesses considerable chromosomal diversity within the limits of a conserved 24-chromosome karyotype. Ani-mals sampled from the American Midwest commonly have between 35 and 38 autosomal chromosome arms (GREENBAUM et al.1978). Calculations presented in the table assume 35 autosomal chromosome arms to minimize interspecific variation arisingfrom this uncertainty.

648 B. L. Dumont and B. A. Payseur

rates. If recombination rate differences evolved accord-ing to a simple Brownian motion model, we wouldexpect pairwise divergence in recombination rate toaccumulate linearly with the square root of divergencetime. Using DNA sequence divergence (measured as

the square root of the summed branch lengths) as aproxy for relative divergence time, this prediction is notsupported (Figure 3; Irbp tree: Spearman’s r ¼ �0.074,P ¼ 0.6689; Cytb tree: Spearman’s r ¼ �0.073, P ¼0.6719), even when metrics insensitive to species varia-tion in chromosome number or chromosome armnumber are used (chromosome number: Irbp tree—Spearman’s r ¼ �0.072, P ¼ 0.6755; Cytb tree—Spearman’s r ¼ �0.089, P ¼ 0.6064; chromosome armnumber: Irbp tree—Spearman’s r ¼ 0.244, P ¼ 0.1515;Cytb tree—Spearman’s r ¼ 0.233, P ¼ 0.1706).

Interspecific covariation between MLH1 foci countand synaptonemal complex length: Previous studieshave established a strong positive relationship betweeninter-individual variation in total autosomal synaptone-mal complex (SC) length (the sum of SC lengths acrossautosomes) and total autosomal MLH1 foci count inboth mice (Lynn et al. 2002) and humans (Lynn et al.2002; Tease and Hulten 2004). Assuming constancy ofgenome size, differences in SC length largely reflectvariation in the density of chromatin packing at meiosis,with longer SCs resulting from more loosely woundDNA (Kleckner et al. 2003). Since recombination mustoccur between synapsed homologs in the context of theSC, it has been suggested that longer SCs have higherrecombination rates because they provide more physi-cal space for crossovers subject to strong positive in-terference to occur (Lynn et al. 2002).

We measured the total autosomal SC length in $30cells from each of the eight rodent taxa, including bothof the wild-derived inbred strains considered for eachM. musculus subspecies. To ask whether species differ-ences in MLH1 foci count have coevolved with changesin SC length, we examined the correlation betweenthese two variables. There is a strong, positive correla-tion between these two measures (Spearman’s r ¼0.976, P ¼ 0.0004; Table S2). However, because rodents

Figure 1.—Representative pachytene spermatoctyes from(A) inbred M. m. domesticus strain PERA/EiJ, (B) wild-caughtP. maniculatus, (C) inbred laboratory rat strain COP, and (D)intra-subspecific PWD/PhJ3CZECHI/EiJ (M. m. musculus) F1

hybrid. SYCP3, a component of the lateral elements of thesynaptonemal complex, is stained in red. Sites of recombina-tion along the synaptonemal complex are denoted by greenMLH1 foci. The white arrows point to the heterogametic sexchromosomes. Only MLH1 foci on autosomal bivalents werescored in this study. Cells displayed in A–D have 22, 23, 39,and 28 MLH1 foci, respectively. Images are not scaled equally.

Figure 2.—Variation in MLH1foci counts across nine murid ro-dents. The mean (61 standarddeviation) for each animal is dis-played. Data from M. spretus(Mspr) individuals were collectedby Koehler et al. (2002). Housemice belonging to the M. muscu-lus species complex are boxed to-gether, with animals fromdifferent wild-derived inbredstrains of a given subspecies de-noted by different shades of thesame color. Note that the tree de-picts only the branching relation-ships among species; branchlengths are not scaled to evolu-tionary divergence. Mmc: M. m.castaneus; Mmm: M. m. musculus;Mmd: M. m. domesticus; Mspi: M.spicilegus; Mcar: M. caroli; Rnov:R. norvegicus; Mpen: M. pennsylva-nicus; Pman: P. maniculatus.

Recombination Rate Evolution in Rodents 649

share a common evolutionary history, estimates ofMLH1 counts and SC lengths are not phylogeneticallyindependent, which could result in an overestimate ofr. After transforming these measures into phylogeneti-cally independent contrasts (Felsenstein 1985), we stillfind a very strong positive correlation between the SClength and MLH1 count (Irbp tree: Spearman’s r ¼0.943, P ¼ 0.017; Cytb tree: Spearman’s r ¼ 0.964, P ¼0.003). As predicted, species with higher mean MLH1counts have longer total autosomal SC lengths.

Genetic inference of recombination rate modifierswithin and between subspecies: The significant differ-ences in genomic recombination rate that we observedamong inbred strains belonging to the M. musculusspecies complex suggest that these closely related taxaharbor genetic differences at loci that affect malerecombination rate. To begin to explore the geneticbasis of this recombination rate divergence, we analyzedthe inheritance of total MLH1 foci count and autosomalSC length in F1 hybrid males derived from reciprocal

intra-subspecific and inter-subspecific crosses of housemice.

Reciprocal F1 males from crosses between wild-derived inbred strains of M. m. musculus (CZECHI/EiJand PWD/PhJ) display statistically significant differ-ences in mean MLH1 foci count (Mann–Whitney U-testperformed on 10,000 subsamples, P ¼ 4.45 3 10�5;Table 3; Figure 4A) and SC length (P ¼ 1.69310�17;Table S3). Similarly, reciprocal F1 males from crossesbetween WSB/EiJ and PERA/EiJ (M. m. domesticus) havedistinct genomic recombination rates (Mann–WhitneyU-tests on 10,000 subsamples, P ¼ 0.0224; Figure 4B),although SC lengths are similar for both sets of F1’s (P¼0.66; Table S3). Because reciprocal F1’s are geneticallyidentical, barring the parental origin of their sex chro-mosomes and mitochondria, these findings providepreliminary evidence for sex-linked and/or mitochon-drial factors contributing to subspecies polymorphismin recombination rate. However, we cannot rule out thepossibility that these patterns are influenced by imprint-ing or maternal effects (Paigen et al. 2008; Ng et al. 2009),especially given that our analysis is limited to males.

Reciprocal inter-subspecific F1 crosses also reveal aputative contribution of sex-linked loci to recombina-tion rate divergence between subspecies. We identifysignificant differences in MLH1 counts between re-ciprocal F1 classes for the three sets of inter-subspecificcrosses examined here (average P-value from Mann–Whitney U-test of 10,000 subsamples; CAST and PWDF1’s: P ¼ 7.593 3 10�11; CAST and WSB F1’s: P ¼ 0.0129;WSB and PWD F1’s: P ¼ 1.50 3 10�10; Table 4; Figure 4,C–E). SC lengths from F1’s between PWD and CASTand between CAST and WSB follow this trend (PWDand CAST F1’s: P ¼ 0.0004; CAST and WSB F1’s: P ¼0.0191; PWD and WSB F1’s: P ¼ 0.2256; Table S3).Interestingly, mean MLH1 foci counts for CAST3

WSB (dam 3 sire) F1 males are significantly higherthan those of either parent (average P-value fromMann–Whitney U-test; CAST and CAST3WSB F1: P ¼

TABLE 2

Differences in MLH1 counts within and betweensubspecies of M. musculus

Comparison P a

Intra-subspecificCIM–CAST 0.0092 (0.0603)b

PERA–WSB 0.0095CZECHI–PWD ,10�20

Inter-subspecificc

Mmc–Mmd 0.0225Mmc–Mmm �10�50

Mmd–Mmm �10�50

a Reported P-value is the mean of 10,000 P-values fromMann–Whitney U-tests performed on independent randomsubsamples of the larger data set.

b P-value after removal of outlier CIM animal 1.c Comparisons use inbred laboratory strains only.

TABLE 3

Differences in MLH1 foci counts between murid rodent species

M. musculus M. spicilegus M. spretus M. caroli R. norvegicus M. pennsylvanicus P. maniculatus

M. musculus — 0.24 0.48 5.25*** 10.04*** 0.68 0.72M. spicilegus 0.24 — 0.72 5.49*** 10.28*** 0.44 0.50M. spretus 0.48 0.72 — 4.77 9.56 1.16 1.20M. caroli 5.25*** 5.49*** 4.77 — 4.79*** 5.93*** 5.97***R. norvegicus 9.04*** 9.28*** 8.56 3.79*** — 10.72*** 10.76***M. pennsylvanicus 5.68*** 5.44*** 6.16 10.93*** 14.72*** — 0.04P. maniculatus 16.72*** 16.48*** 17.20 21.97*** 25.76*** 11.04*** —

Raw differences in mean MLH1 counts between species pairs are given in the upper triangle, with the significance of the as-sociated Mann–Whitney U-test denoted by an asterisk (P-value averaged over 10,000 subsamples of the data; see materials and

methods; *P , 0.05; **P , 0.01; ***P , 0.001). Differences in MLH1 counts adjusted for differences in chromosome arm num-ber between species are presented in the data group in the lower triangle, with significance denoted. Significance was not assessedin comparisons with M. spretus. The M. musculus value represents the mean of the means of six wild-derived inbred strains.

650 B. L. Dumont and B. A. Payseur

8.54 310�5; WSB and CAST3WSB F1: P¼ 0.0170). Thisobservation points to a possible epistatic interactionbetween CASTand WSB alleles, with at least one locus inthe interaction localizing to the CAST X chromosomeor mitochondria or the WSB Y chromosome. However,we note that this result is based on MLH1 countsgathered from a single WSB3CAST F1 male, and thepossibility that this pattern is due to heterosis cannot beformally ruled out.

Male inter-subspecific F1 animals appear to have MLH1counts that are more similar to those of the paternalparent (Figure 4, C–E). To assess this pattern statistically,we evaluated the departure of the mean MLH1 foci countin F1 animals from the midparent value [(mean MLH1foci count in parental stain A 1 mean MLH1 foci count inparental strain B)/2]. We generated bootstrap samples

from the MLH1 foci count data for each of the twoparental strains and the F1 animals. We then calculatedthe mean MLH1 foci count from the bootstrapped F1 dataand subtracted the two-parent average MLH1 foci countdetermined from the parental bootstrapped samples.This procedure was replicated 10,000 times to derive adistribution of differences in mean MLH1 foci countbetween the F1’s and the midparent value. If F1 MLH1 focicounts equally resemble those in the two parents, thisdistribution should be centered on 0. For all inter-subspecific comparisons, except WSB3CAST F1’s, westrongly reject this null hypothesis (P , 10�4; P ¼0.4889 for WSB3CAST F1’s). In these significant cases,the mean of the bootstrap distribution is shifted in thedirection expected if phenotypic resemblance is strongestwith the paternal parent (Figure S1). In contrast, mostintra-subspecific F1’s more closely resemble the maternalparent (P , 10�3 for CZECHI3PWD, PWD3CZECHI,and PERA3WSB F1’s; P ¼ 0.3057 for WSB3PERA F1’s;Figure S2). These results suggest two possibilities. First,the sex chromosome genetic architecture of recombina-tion rate polymorphism and divergence may be funda-mentally distinct. Second, if differences in imprintingstatus at recombination rate modifiers evolve sufficientlyslowly, parent-of-origin gene-silencing effects could varybetween subspecies but not within subspecies.

DISCUSSION

Divergence in genomic recombination rate on vari-able temporal scales: We observed variation in genomicrecombination rate over a range of evolutionary timescales. We documented polymorphism in recombina-tion rate within subspecies of house mice, as well as

Figure 3.—Correlation between divergence in meanMLH1 foci count and pairwise DNA sequence divergence cal-culated as the sum of branch lengths from maximum likeli-hood phylogenies constructed from (A) CYTB and (B)IRBP coding sequences.

Figure 4.—Variation inmean MLH1 foci counts(61 standard deviation)among F1 intra- and inter-subspecific hybrids. Recip-rocal F1 intra-subspecificcrosses were conducted be-tween parental wild-derivedinbred strains (A) CZE-CHI/EiJ and PWD/PhJand (B) WSB/EiJ andPERA/EiJ. Inter-subspecificF1 males were generatedfrom reciprocal crosses be-tween strains (C) WSB/EiJand PWD/PhJ, (D) WSB/EiJ and CAST/EiJ, and (E)CAST/EiJ and PWD/PhJ.For F1 animals, the mater-nal parent is the strain listedfirst (i.e., strain C is themother in cross C3P).

Recombination Rate Evolution in Rodents 651

marked divergence in genomic recombination ratebetween closely related subspecies (Table 2). Ouranalysis also reveals significant divergence betweenMus species and among more distantly related rodenttaxa (Table 3). Most of the variation in MLH1 foci countis explained by species or subspecies identity, even afteradjusting recombination rates to account for karyotypicdifferences among species (Table 1; Table 3).

The amount of divergence in genomic recombina-tion rate does not scale with evolutionary divergencebetween rodent species inferred from DNA sequences.This result suggests that the rate of evolution in re-combination rate has fluctuated during the history ofmurids. Multiple evolutionary processes could be re-sponsible for this rate heterogeneity. Unique selectiveregimes operating along distinct lineages could resultin local accelerations or decelerations in the rate ofrecombination rate evolution, masking a potential cor-relation between the phylogeny and the trait. Recombi-nation rates in mammals are subject to strong selectivepressures to maintain a basal threshold that ensures aminimum of one crossover per chromosome (or chro-mosome arm). On the other hand, very high rates of

recombination can stress the limits of the cell’s DNArepair capacity, predisposing genomes to nonhomolo-gous recombination and deleterious genomic rearrange-ments. Within these bounds, natural selection couldact to optimize the rate of recombination according tospecies-specific criteria. For example, high deleteriousmutation loads could select for higher recombinationrates (Kondrashov 1984). However, this possibility re-mains speculative, as there is no direct empirical ev-idence that genomic recombination rates reflect optimaobtained through natural selection.

Alternatively, rapid, stochastic changes in the geno-mic rate of recombination within the upper and lowerbounds defined by the constraints on the meioticprocess could conceivably give rise to the pattern thatwe document. Such a scenario distills to a random walkwith two reflective boundaries in phenotypic space.When the rate of recombination hits the lower boundof one crossover per chromosome (or chromosomearm), selective forces will function as a reflective bar-rier, pushing the rate back up into the neutral realm.Conversely, when the rate of recombination hits theupper limit defined by preservation of genome integrity,

TABLE 4

MLH1 counts in intra-subspecific F1 males

MLH1 count

Mother 3 father Animal No. of cells Mean SD Range Per chromosome

PWD3CZECHI 1 19 30.32 2.33 26–34 1.602 19 29.79 2.51 25–34 1.573 17 29.65 2.71 24–35 1.564 21 28.62 1.77 26–33 1.51

Total 76 29.57 2.37 24–35 1.56

CZECHI3PWD 1 21 27.71 2.12 24–33 1.462 24 28.25 2.19 23–32 1.493 23 27.70 2.57 24–34 1.464 19 27.79 2.32 22–33 1.465 20 27.85 1.84 24–32 1.466 20 27.40 2.14 23–31 1.447 14 27.57 2.06 24–32 1.45

Total 141 27.77 2.17 22–34 1.46

WSB3PERA 1 18 22.33 2.45 19–26 1.182 18 22.39 1.97 19–26 1.183 20 22.75 1.77 19–25 1.204 17 22.35 2.06 19–27 1.185 23 22.96 2.51 18–30 1.21

Total 96 22.58 2.16 18–30 1.19

PERA3WSB 1 23 23.57 1.85 21–29 1.242 23 23.39 1.85 20–27 1.233 17 23.65 2.23 20–29 1.244 27 23.44 1.91 20–28 1.235 26 24.46 2.18 21–29 1.296 21 23.10 1.95 20–26 1.227 29 22.48 1.88 18–26 1.18

Total 166 23.42 2.02 18–29 1.23

652 B. L. Dumont and B. A. Payseur

it will promptly reflect downward. In this manner, highlydivergent taxa could possess similar genomic recombi-nation rates by the chance convergence of their randomwalks. Similarly, closely related species could havemarkedly different genomic recombination rates if theirassociated random walks are initially characterized bysteps in opposite directions.

We attempted to use phylogenetic comparative meth-ods to assess the fit of evolutionary models of neutrality,stabilizing selection, and lineage-specific evolution toMLH1 foci counts, but simulations revealed that thesmall size of our data set combined with the excess ofvery short branch lengths yield insufficient statisticalpower to discriminate between alternative evolutionaryhypotheses (data not shown). Similar reductions inpower likely accompany our tests for correlations betweenrecombination rate divergence and DNA sequence di-

vergence, suggesting caution in our interpretation. Theconsideration of additional rodent taxa, along withdevelopment of phylogenetic comparative methods formodeling dually bounded quantitative traits, will berequired to evaluate the ideas advanced here.

Reconciling evolutionary patterns within muridrodents and across mammals: We previously showedthat average genomic recombination rates of 13 mam-malian species are phylogenetically distributed, withmore closely related species possessing more similarrates (Dumont and Payseur 2008). This result stands instark contrast to the absence of a phylogenetic signatureamong murids.

There are several potential explanations for this dis-crepancy. First, we may lack power to detect a phyloge-netic signal. Although we document considerabledivergence in recombination rate in Muridae (Table

TABLE 5

MLH1 counts in inter-subspecific F1 males

MLH1 Count

Mother 3 father Animal No. of cells Mean SD Range Per chromosome

WSB3CAST 1 63 22.03 2.16 19–28 1.16

CAST3WSB 1 47 23.13 1.74 20–26 1.222 24 22.67 1.81 19–26 1.193 20 22.75 2.00 20–26 1.204 21 22.95 2.20 20–28 1.21

Total 112 22.93 1.88 19–28 1.21

WSB3PWD 1 18 26.94 2.71 21–31 1.422 20 28.10 2.47 24–34 1.483 26 27.69 2.35 23–32 1.464 12 26.25 2.73 24–32 1.385 24 26.79 2.78 21–32 1.416 18 27.33 3.33 23–33 1.447 19 26.89 2.51 22–31 1.42

Total 137 27.21 2.69 21–34 1.43

PWD3WSB 1 18 24.06 2.71 19–29 1.272 19 24.21 2.84 19–29 1.273 18 25.06 2.01 22–30 1.324 20 24.65 1.57 22–28 1.305 23 24.35 2.39 20–29 1.28

Total 98 24.46 2.32 19–30 1.29

PWD3CAST 1 40 22.27 1.85 19–26 1.17

CAST3PWD 1 22 27.82 2.15 23–31 1.462 19 26.42 1.80 23–30 1.393 22 27.18 1.92 23–30 1.434 23 26.61 2.13 23–31 1.405 23 26.78 2.15 23–30 1.416 20 27.15 2.11 24–30 1.437 23 27.00 2.43 22–31 1.428 20 27.80 2.04 24–31 1.469 16 25.25 2.05 21–28 1.33

10 17 26.12 2.57 22–31 1.3711 23 27.04 2.40 21–31 1.42

Total 228 26.89 2.22 21–1 1.42

Recombination Rate Evolution in Rodents 653

1), the two most phenotypically divergent taxa that weexamine—the house mouse and rat—have highly sim-ilar genomic recombination rates when viewed in thecontext of the far greater variability observed amongmammals (Burt and Bell 1987; Dumont and Payseur

2008). In addition, because of our concentrated sam-pling within the Mus genus, most species pairs areseparated by little divergence time. Although littlevariation in both recombination rates and branchlengths leaves little power to detect an associationbetween the phylogeny and recombination rate, certainpatterns argue against the existence of a strong phylo-genetic signal. For example, some closely related speciespairs exhibit very large differences in mean MLH1 focicounts and some divergent species pairs have nearlyidentical values.

Methodological differences might also generate discrep-ancies between the two studies. Dumont and Payseur

(2008) estimated recombination rates from geneticmaps, whereas a cytological approach was used here.There are small, albeit systematic, differences in malerecombination rates estimated by cytological vs. geneticapproaches (Sun et al. 2004). Slight variations in thetemporal loading of MLH1 onto DNA sites of recombi-national repair could lead to uncounted recombinationevents by the method used here (Cheng et al. 2009). Onthe other hand, genotyping error can induce artificialinflation in genetic linkage map lengths, especially interminal chromosomal regions where flanking geno-type data are absent (Broman et al. 1998). In addition, asmall subset of crossovers appears to be repaired by anon-MLH1 pathway in mammals (Holloway et al.2008) and will obviously go undetected by the MLH1method. However, it seems unlikely that solely method-ological differences underlie the observed difference.Indeed, average male chiasma counts produce a strongerphylogenetic signal in mammals than recombination ratesestimated from linkage maps (Dumont and Payseur

2008).Third, most of the taxa considered in the mammalian

analysis were (partially) outbred or domesticated spe-cies, whereas the analysis of recombination rate varia-tion among murid rodents presented here relies heavilyon estimates obtained from fully inbred animals. Re-combination rates commonly increase in response toinbreeding (Burt and Bell 1987; Ross-Ibarra 2004),but a sample of outbred Gough Island mice haverecombination rates comparable to, if not slightly higherthan, wild-derived inbred strains of M. m. domesticus. Inaddition, we note that most hybrid F1 males from intra-and inter-subspecific crosses of house mice have recom-bination rates intermediate between the values for thetwo parental strains, revealing minimal evidence for in-breeding depression in recombination rate.

Finally, these seemingly contradictory patterns couldindicate that the inferred evolutionary processes shap-ing species differences in genomic recombination de-

pend on the scale of taxonomic divergence. Variation inrecombination rate among rodents may reflect stochas-tic fluctuations or selection toward local optima withinthe confines of the upper and lower bounds describedabove, whereas variation in recombination rates amongdivergent mammals may indicate gradual evolutionaryshifts in the precise placement of these meioticallydefined constraints. For example, genetic changes inthe DNA repair machinery could render a given clademore proficient at the repair of meiotic double-stranded breaks, thereby pushing the upper boundaryfurther upward. Similarly, evolutionary changes in co-hesion proteins, protein components of the synaptone-mal complex, or the efficacy of the meiotic spindlecould exempt a clade from the ‘‘one crossover perchromosome arm’’ rule, shifting the lower bound tothe threshold determined by the number of chromo-somes. Over time, these boundary shifts could lead tosizable changes in the range of genomic recombinationrates that can be realized between divergent taxonomicgroups, giving rise to an apparent phylogenetic signal inrecombination rate. This hypothesis remains untested,but genetic studies of species differences in genomicrecombination rate over different evolutionary timescales could shed light on this possibility.

Covariation of synaptonemal complex length andrecombination rate across Muridae: The synaptonemalcomplex is a tripartite protein structure that physicallylinks homologous chromosomes at meiosis (Moses

1968). A previous report identified a positive correla-tion between autosomal SC length and recombinationrate within and between inbred house mouse strainsand among humans (Lynn et al. 2002). Our analysisextends these results to show that genomic recombina-tion rate and SC length are positively correlated be-tween species. It remains unclear whether changes in SClength drive the evolution of recombination rate, re-combination rate evolution catalyzes adjustments in SClength, or the evolution of both SC length and re-combination rate is a correlated response to changes ina third variable.

Total SC length is likely a function of physical genomesize and the average size of DNA loops emanating fromthe chromatin scaffold at meiosis (Kleckner 2003). Ifgenome sizes are largely conserved among rodents(Gregory 2005), the observed differences in SC lengthacross taxa may predominately reflect changes in thedegree of chromatin condensation at meiosis. Chroma-tin structure can evolve rapidly between species (Henik-

off et al. 2001; Ferree and Barbash 2009), a fact thatnominates heterochromatin-binding proteins, histonemodification proteins, and other loci regulating chro-matin states as candidates underlying the differences ingenomic recombination rate between murid rodentspecies documented here.

Genetic basis of subspecies polymorphism anddivergence in genomic recombination rate: Males from

654 B. L. Dumont and B. A. Payseur

reciprocal F1 directions from all crosses examined dis-played significantly different mean MLH1 foci counts(Tables 4 and 5; Figure 4), highlighting a putative rolefor sex-linked loci contributing to variation in genomicrecombination rate within and between subspecies.Epigenetic phenomena such as imprinting and thematernal environment may also give rise to this pattern.Unfortunately, our study design does not allow us todisentangle these various contributions to differencesbetween reciprocal F1 males.

Our restriction of MLH1 foci counts to the autosomesindicates that if sex-linked modifiers segregate betweenhouse mouse strains, they must exert trans effects onrecombination rate (sex-linked cis effects may also exist,but our design could not measure them). MLH1 focicounts are challenging to obtain in females (since pro-phase I is completed in the fetal ovary). However, whencoupled with known sex differences in genomic recom-bination rate between male and female house mice(Reeves et al. 1990; Cox et al. 2009), these results raisethe question of how these modifiers might affect re-combination rates in females.

Although modifiers of local- and genomic-scale re-combination rate variation have been identified inhouse mice (Shiroishi et al. 1991; Heine et al. 1994;Baudat and de Massy 2007; Grey et al. 2009; Parvanov

et al. 2009; Baudat et al. 2010; Parvanov et al. 2010),only one sex-linked modifier of genome-wide recombi-nation rate has been discovered (de la Casa-Esperon

et al. 2002). The locus identified by de la Casa-Esperon

et al. (2002) is X-linked and segregates between inbredlaboratory strains of mice. Given that the genomes ofcommon lab strains bear small contributions from M. m.musculus and M. m. castaneus (Yang et al. 2007), the sex-linked genetic factors that we find in crosses involvingwild-derived inbred strains could colocalize with thispreviously identified locus. Further efforts to refine thegenomic location of the modifier(s) identified here,evaluate their impact on female recombination rate (ifin fact X-linked or mitochondrial), and deduce theirmolecular functions will provide important insights intothe evolution of recombination rate differences be-tween species.

Low genomic recombination rate in Murid rodents:Our survey of variation in mean MLH1 foci countsuggests that the low genomic recombination rateobserved in laboratory mice and rats relative to othermammals is a defining characteristic of the muridrodent family. Rats, the species with the highest meanMLH1 foci count among the species considered here,still have far fewer MLH1 foci per meiosis than humans[mean¼ 49.8 foci/meiosis (Sun et al. 2004)], macaques[mean ¼ 39.0 (Hassold et al. 2009)], or dogs [mean ¼40.0 (Basheva et al. 2008)]. In contrast, most speciesconsidered in this analysis have higher MLH1 focicounts than the common shrew (Borodin et al. 2008),a pattern presumably reflecting the marked reduction

in chromosome number in shrews (2N¼ 20–33) relativeto the murids examined here (2N ¼ 40–46). Ouranalysis rules out the possibility that the comparativelylow recombination rates in laboratory mice and rats arean incidental by-product of adaptation to the laboratorysetting or inbreeding, as wild-caught and outbred micealso conform to this trend. Moreover, this pattern doesnot appear to stem solely from karyotype distinctionsbetween rodents and other mammals (Dumont andPayseur 2008). Our survey of recombination rate var-iation within a single rodent family provides an im-portant first step toward determining the underlyingmechanism of low recombination rates in this speciosemammalian order.

We thank Terry Hassold and Terrah Hansen for instruction on theMLH1/SYCP3 immunostaining approach, Michael Gould for pro-viding the rat tissue samples used in this study, and Peter Ryan andMelissa Gray for Gough Island mouse tissues. We acknowledge KarlBroman and Cecile Ane for useful discussions on statistical issues andphylogenetic comparative methods. We also thank Lauren Brooks,Robert Fettiplace, Michael Joyce, Mara McDonald, and Yushi Oguchifor field assistance. B.L.D. was supported by a National ScienceFoundation (NSF) Predoctoral Fellowship. This work was funded byNSF Doctoral Dissertation Improvement Grant DEB 0909779 and NSFGrant DEB 0918000.

LITERATURE CITED

Anderson, L. K., A. Reeves, L. M. Webb and T. Ashley, 1999 Dis-tribution of crossing over on mouse synaptonemal complexes us-ing immunofluorescent localization of MLH1 protein. Genetics151: 1569–1579.

Baker, S. M., A. W. Plug, T. A. Prolla, C. E. Bronner, A. C. Harris

et al., 1996 Involvement of mouse Mlh1 in DNA mismatch re-pair and meiotic crossing over. Nat. Genet. 13: 336–342.

Barton, N. H., 1995 A general model for the evolution of recom-bination. Genet. Res. 65: 123–144.

Basheva, E. A., C. J. Bidau and P. M. Borodin, 2008 Generalpattern of meiotic recombination in male dogs estimated byMLH1 and RAD51 immunolocalization. Chromosome Res. 16:709–719.

Baudat, F., and B. de Massy, 2007 Cis - and trans-acting elementsregulate the mouse Psmb9 meiotic recombination hotspot. PLoSGenet. 3: e100.

Baudat, F., J. Buard, C. Grey, A. Fledel-Alon, C. Ober et al.,2010 PRDM9 is a major determinant of meiotic recombinationhotspots in humans and mice. Science 327: 836–840.

Borodin, P. M., T. V. Karamysheva, N. M. Belonogova, A. A.Torgasheva, N. B. Rubtsov et al., 2008 Recombination mapof the common shrew, Sorex araneus (Eulipotyphla, Mammalia).Genetics 178: 621–632.

Broman, K. W., J. C. Murray, V. C. Sheffield, R. L. White and J. L.Weber, 1998 Comprehensive human genetic maps: individualand sex-specific variation in recombination. Am. J. Hum. Genet.63: 861–869.

Burt, A., and G. Bell, 1987 Mammalian chiasma frequencies as atest of two theories of recombination. Nature 326: 803–805.

Charlesworth, B., M. T. Morgan and D. Charlesworth,1993 The effect of deleterious mutations on neutral molecularvariation. Genetics 134: 1289–1303.

Cheng, E. Y., P. A. Hunt, T. A. Naluai-Cecchini, C. L. Fligner, V. Y.Fujimoto et al., 2009 Meiotic recombination in human oocytes.PLoS Genet. 5: e1000661.

Coop, G., and M. Przeworski, 2007 An evolutionary view of humanrecombination. Nat. Rev. Genet. 8: 23–34.

Coop, G., X. Wen, C. Ober, J. K. Pritchard and M. Przeworski,2008 High-resolution mapping of crossovers reveals extensivevariation in fine-scale recombination patterns among humans.Science 319: 1395–1398.

Recombination Rate Evolution in Rodents 655

Cox, A., C. L. Ackert-Bicknell, B. L. Dumont, Y. Ding, J. T. Bell

et al., 2009 A new standard genetic map for the laboratorymouse. Genetics 182: 1335–1344.

de la Casa-Esperon, E. J. C. Loredo-Osti, F. Pardo-Manuel de

Villena, T. L. Briscoe, J. M. Malette et al., 2002 X chromo-some effect on maternal recombination and meiotic drive inthe mouse. Genetics 161: 1651–1659.

Dumont, B. L., and B. A. Payseur, 2008 Evolution of the genomicrate of recombination in mammals. Evolution 62: 276–294.

Dumont, B. L., K. W. Broman and B. A. Payseur, 2009 Variation ingenomic recombination rates among heterogeneous stock mice.Genetics 182: 1345–1349.

Duret, L., and P. F. Arndt, 2008 The impact of recombination onnucleotide substitutions in the human genome. PLoS Genet. 4:e1000071.

Dutrilleux, B., 1986 Role of chromosomes in evolution: a new in-terpretation. Ann. Genet. 29: 69–75 (in French).

Felsenstein, J., 1965 The effect of linkage on directional selection.Genetics 52: 349–363.

Felsenstein, J., 1974 The evolutionary advantage of recombina-tion. Genetics 78: 737–756.

Felsenstein, J., 1985 Phylogenies and the comparative method.Am. Nat. 125: 1–15.

Ferree, P. M., and D. A. Barbash, 2009 Species-specific heterochro-matin prevents meiotic chromosome segregation to cause hybridlethality in Drosophila. PLoS Biol. 7: e1000234.

Gibbs, R. A., G. M. Weinstock, M. L. Metzker, D. M. Muzny, E. J.Sodergren et al., 2004 Genome sequence of the brown Norwayrat yields insights into mammalian evolution. Nature 428: 493–521.

Greenbaum, I. F., R. J. Baker and J. H. Bowers, 1978 Chromosomalhomology and divergence between sibling species of deer mice:Peromyscus maniculatus and P. melanotis (Rodentia, Cricetidae). Evo-lution 32: 334–341.

Gregory, T. R., 2005 Animal Genome Size Database. http://www.genomesize.com.

Grey, C., F. Baudat and B. de Massy, 2009 Genome-wide control ofthe distribution of meiotic recombination. PLoS Biol. 7: e35.

Guindon, S., and O. Gascuel, 2003 A simple, fast, and accurate al-gorithm to estimate large phylogenies by maximum likelihood.Syst. Biol. 52: 696–704.

Hassold, T., and P. Hunt, 2001 To err (meiotically) is human: thegenesis of human aneuploidy. Nat. Rev. Genet. 2: 280–291.

Hassold, T. J., T. Hansen, P. Hunt and C. VandeVoort,2009 Cytological studies of recombination in rhesus males. Cy-togenet. Genome Res. 124: 132–138.

Heine, D., S. Khambata, K. S. Wydner and H. C. Passmore,1994 Analysis of recombination hotspots associated with thep haplotypes of the mouse MHC. Genomics 23: 168–177.

Henikoff, S., K. Ahmad and H. S. Malik, 2001 The centromeresparadox: stable inheritance with rapidly evolving DNA. Science293: 1098–1102.

Holloway, J. K., J. Booth, W. Edelmann, C. H. McGowan andP. E. Cohen, 2008 MUS81 generates a subset of MLH1–MLH3-independent crossovers in mammalian meiosis. PLoSGenet. 4: e1000186.

Jeffreys, A. J., and R. Neumann, 2005 Factors influencing recombi-nation frequency and distribution in a human meiotic crossoverhotspot. Hum. Mol. Genet. 14: 2277–2287.

Jeffreys, A. J., and R. Neumann, 2009 The rise and fall of a humanrecombination hotspot. Nat. Genet. 41: 625–629.

Kleckner, N., A. Storlazzi and D. Zickler, 2003 Coordinate var-iation in meiotic pachytene SC length and total crossover/chi-asma frequence under conditions of constant DNA length.Trends Genet. 19: 623–628.

Koehler, K. E., J. P. Cherry, A. Lynn, P. A. Hunt and T. J. Hassold,2002 Genetic control of mammalian meiotic recombination. I.Variation in exchange frequencies among males from inbredmouse strains. Genetics 162: 297–306.

Kondrashov, A. S., 1984 Deleterious mutations as an evolutionaryfactor. 1. The advantage of recombination. Genet. Res. 44: 199–217.

Kong, A., D. F. Gudbjartsson, J. Sainz, G. M. Jonsdottir, S. A.Gudjonsson et al., 2002 A high-resolution recombinationmap of the human genome. Nat. Genet. 31: 241–247.

Li, W., and J. Freudenberg, 2009 Two-parameter characterizationof chromosome-scale recombination rate. Genome Res. 19:2300–2307.

Lynn, A., K. E. Koehler, L. A. Judis, E. R. Chan, J. P. Cherry et al.,2002 Covariation of synaptonemal complex length and mam-malian meiotic exchange rates. Science 296: 2222–2225.

Maynard Smith, J., and J. Haigh, 1974 The hitch-hiking effect of afavourable gene. Genet. Res. 23: 23–35.

Morton, N. E., 1991 Parameters of the human genome. Proc. Natl.Acad. Sci. USA 88: 7474–7476.

Moses, M. J., 1968 Synaptinemal complex. Annu. Rev. Genet. 2:363–412.

Ng, S. H., R. Madeira, E. D. Parvanov, L. M. Petros, P. M. Petkov

et al., 2009 Parental origin of chromosomes influences cross-over activity within the Kcnq1 transcriptionally imprinted domainof Mus musculus. BMC Mol. Biol. 10: 43.

Nilsson, N. O., T. Sall and B. O. Bengtsson, 1993 Chiasma andrecombination data in plants: Are they compatible? TrendsGenet. 9: 344–348.

Okuizumi, H., T. Ohsumi, N. Sakaki, H. Imoto, Y. Mizuno et al.,1997 Linkage map of Syrian hamster with restriction landmarkgenomic scanning. Mamm. Genome 8: 121–128.

Otto, S. P., and N. H. Barton, 2001 Selection for recombination insmall populations. Evolution 55: 1921–1931.

Paigen, K., and P. Petkov, 2010 Mammalian recombination hotspots: properties, control, and evolution. Nat. Rev. Genet. 11:221–233.

Paigen, K. J., P. Szatkiewicz, K. Sawyer, N. Leahy, E. D. Parvanov

et al., 2008 The recombinational anatomy of a mouse chromo-some. PLoS Genet. 4: e1000119.

Pardo-Manuel de Villena, F., and C. Sapienza, 2001 Recom-bination is proportional to the number of chromosome armsin mammals. Mamm. Genome 12: 318–322.

Parvanov, E. D., S. H. S. Ng, P. M Petkov and K. Paigen, 2009 Trans-regulation of mouse meiotic recombination hotspots by Rcr1. PLoSBiol. 7: e1000036.

Parvanov, E. D., P. M. Petkov and K. Paigen, 2010 Prdm9 controlsactivation of mammalian recombination hotspots. Science 327: 835.

Peters, A. H. F. M., A. W. Plug, M. J. van de Vugt and P. de Voer,1997 A drying-down technique for the spreading of mamma-lian meiocytes from the male and female germline. ChromosomeRes. 5: 66–68.

Pritchard, J. K., and M. Przeworski, 2001 Linkage disequilibriumin humans: models and data. Am. J. Hum. Genet. 69: 1–14.

Ptak, S. E., D. A. Hines, K. Koehler, B. Nickel, N. Patil et al.,2005 Fine-scale recombination patterns differ between humansand chimpanzees. Nat. Genet. 37: 429–434.

R Development Core Team, 2008 R: A Language and Environmentfor Statistical Computing. R Foundation for Statistical Computing,Vienna. http://www.R-project.org.

Reeves, A., and J. Tear, 2000 MicroMeasure for Windows, version3.3. http://www.colostate.edu/Depts/Biology/MicroMeasure.

Reeves, R. H., M. R. Crowley, B. F. O’Hara and J. D. Gearhart,1990 Sex, strain, and species differences affect recombinationacross an evolutionarily conserved segement of mouse chromo-some 16. Genomics 8: 141–148.

Ross-Ibarra, J., 2004 The evolution of recombination under do-mestication: a test of two hypotheses. Am. Nat. 163: 105–112.

Shiroishi, T., T. Sagai, N. Hanzawa, H. Gotoh and K. Moriwaki,1991 Genetic control of sex-dependent meiotic recombinationin the major histocompatibility complex of the mouse. EMBO J.10: 681–686.

Steen, R. G., A. E. Kwitek-Black, C. Glenn, J. Gullings-Handley,V. Van Etten et al., 1999 A high-density integrated genetic link-age and radiation hybrid map of the laboratory rat. Genome Res.9: 1–8.

Steiner, C. C., J. N. Weber and H. E. Hoekstra, 2007 Adaptive var-iation in beach mice produced by two interacting pigmentationgenes. PLoS Biol. 5: e219.

Sun, F., M. Oliver-Bonet, T. Liehr, H. Starke, E. Ko et al.,2004 Human male recombination maps for individual chromo-somes. Am. J. Hum. Genet. 74: 521–531.

Tamura, K., and M. Nei, 1993 Estimation of the number of nucle-otide substitutions in the control region of mitochondrial DNAin humans and chimpanzees. Mol. Biol. Evol. 10: 512–526.

656 B. L. Dumont and B. A. Payseur

Tease, C., and M. A. Hulten, 2004 Inter-sex variation in synaptone-mal complex lengths largely determines the different recombina-tion rates in male and female germ cells. Cytogenet. GenomeRes. 107: 208–215.

Thompson, J. D., D. G. Higgins and T. J. Gibson, 1994 CLUSTALW: improving the sensitivity of progressive multiple sequencealignment through sequence weighting, position-specific gappenalties and weight matrix choice. Nucleic Acids Res. 22:4673–4680.

Webb, A. J., I. L. Berg and A. Jeffreys, 2008 Sperm cross-over ac-tivity in regions of the human genome showing extreme break-

down of marker association. Proc. Natl. Acad. Sci. USA 105:10471–10476.

Winckler, W., S. R. Myers, D. J. Richter, R. C. Onofrio, G. J.McDonald et al., 2005 Comparison of fine-scale recombinationrates in humans and chimpanzees. Science 308: 107–111.

Yang, H, T. Bell, G. Churchill and F. Pardo-Manuel de Villena,2007 On the subspecific origin of the laboratory mouse. Nat.Genet. 39: 1100–1107.

Communicating editor: J. C. Schimenti

Recombination Rate Evolution in Rodents 657

GENETICSSupporting Information

http://www.genetics.org/cgi/content/full/genetics.110.123851/DC1

Evolution of the Genomic Recombination Rate in Murid Rodents

Beth L. Dumont and Bret A. Payseur

Copyright � 2011 by the Genetics Society of AmericaDOI: 10.1534/genetics.110.123851

B. L. Dumont and B. A. Payseur 2 SI

FIGURE S1.—Mean MLH1 foci count in intersubspecific F1 males most closely resembles the paternal parent phenotype. We

used a bootstrapping randomization procedure (see main text) to derive an empirical distribution of differences between the F1

mean MLH1 foci count and the midparent average MLH1 foci count. If F1 animals equally resemble both parents, this

distribution should be centered on 0. In contrast, if F1 animals are more similar to the parent with higher (lower) mean MLH1 foci count, this distribution will be shifted to the right (left). For PWDxWSB (maternal x paternal parent) F1 animals, the

distribution is clearly shifted left, indicating that the F1s more closely resemble the low recombination rate WSB paternal parent

(A). In the reciprocal F1 cross, WSBxPWD F1 animals have mean MLH1 foci counts that more closely resemble those in the high

recombination rate PWD paternal strain (B). The distribution for CASTxWSB F1s is shifted to the right, as predicted if the mean

MLH1 foci count in the F1s is more similar to the WSB paternal parent phenotype (C). The WSBxCAST F1s have a mean

MLH1 foci count that is roughly equidistant between that of the two parents (D). Again, CASTxPWD F1s more closely resemble

the paternal strain (E), whereas the mean MLH1 foci count in PWDxCAST F1s is more similar to that of CAST (F).

PWDxWSB F1

Fre

quency

-6 -4 -2 0 2 4 6

0

1000

2000

3000

AWSBxPWD F1

-6 -4 -2 0 2 4 6

0

1000

2000

B

CASTxWSB F1

Fre

quency

-6 -4 -2 0 2 4 6

0

1000

2000

3000

CWSBxCAST F1

-6 -4 -2 0 2 4 6

0

1000

2000

3000

D

CASTxPWD F1

Fre

quency

-6 -4 -2 0 2 4 6

0

1000

2000

3000

EPWDxCAST F1

-6 -4 -2 0 2 4 6

0

1000

2000

3000

F

F1 Mean MLH1 Foci Count - Midparent Mean MLH1 Foci Count

More WSB-like More PWD-like

More WSB-likeMore CAST-like

More CAST-like More PWD-like

More WSB-like More PWD-like

More CAST-like More WSB-like

More

CAST-like

More PWD-like

B. L. Dumont and B. A. Payseur 3 SI

FIGURE S2.—Mean MLH1 foci count in intrasubspecific F1 males most closely resembles the maternal parent phenotype. We

used a bootstrapping randomization procedure (see main text) to derive an empirical distribution of differences between the F1 mean MLH1 foci count and the midparent average MLH1 foci count. If F1 animals equally resemble both parents, this

distribution should be centered on 0. In contrast, if F1 animals are more similar to the parent with higher (lower) mean MLH1

foci count, this distribution will be shifted to the right (left). For PWDxCZECHI (maternal x paternal parent) F1 animals, the

distribution is shifted to the right, indicating that the F1s more closely resemble the higher recombination rate PWD maternal

parent (A). In the reciprocal F1 cross, CZECHIxPWD F1 animals have mean MLH1 foci counts that are more similar to those in

the lower recombination rate CZECHI maternal strain (B). The distribution for PERAxWSB F1s is shifted to the right, as

predicted if the mean MLH1 foci count in the F1s is more similar to the PERA maternal parent (C). The WSBxPERA F1s have a

mean MLH1 foci count that is roughly equidistant between that of the two parents (D).

PWDxCZECHI F1

Fre

quency

-3 -2 -1 0 1 2 3

0

500

1000

1500

2000

APWDxCZECHI F1

-3 -2 -1 0 1 2 3

0

500

1000

1500

2000

B

PERAxWSB F1

Fre

quency

-3 -2 -1 0 1 2 3

0

500

1000

1500

2000

CWSBxPERA F1

-3 -2 -1 0 1 2 3

0

500

1000

1500

2000

D

F1 Mean MLH1 Foci Count - Midparent Mean MLH1 Foci Count

More

CZECHI-like

More

PWD-likeMore

CZECHI-like

More

PWD-like

More

WSB-like

More

PERA-likeMore

WSB-like

More

PERA-like

B. L. Dumont and B. A. Payseur 4 SI

TABLE S1

NCBI GenBank sequence accession numbers

Species IRBP Locus CYTB Locus

Mus musculus castaneus * AY057805.1

Mus musculus musculus ** AY057804.1

Mus musculus domesticus *** AY057807.1

Mus spicilegus AB125809.1 AY057809.1

Mus caroli AB125797.1 AY057812.1

Mus spretus AJ698883.1 AY057810.1

Rattus norvegicus AJ429134.1 GU592997.1

Peromyscus maniculatus AY163630.1 FJ800584.1

Microtus pennsylvanicus AM919415.1 AF119279.1

Spermophilus tridecemlineatus AF297278.1 AF157877.1

* Sequence pulled from whole-genome sequence of inbred mouse strain CAST/EiJ

** Sequence pulled from whole-genome sequence of inbred mouse strain PWK/PhJ *** Sequence pulled from whole-genome sequence of inbred mouse strain WSB/EiJ

Whole genome sequences are available from the Wellcome Trust Sanger Institute at

http://www.sanger.ac.uk/resources/mouse/genomes/.

B. L. Dumont and B. A. Payseur 5 SI

TABLE S2

Synaptonemal complex lengths in murid rodents

Species Number of Cells

Measured

Mean Synaptonemal

Complex Length (μm) Standard Deviation

Mus musculus castaneus

CAST/EiJ 30 242.79 18.34

CIM 32 234.56 19.16

Mus musculus musculus

CZECHI/EiJ 30 283.08 26.40

PWD/PhJ 30 281.40 32.24

Mus musculus domesticus

PERA/EiJ 255 219.16 23.06

WSB/EiJ 85 224.30 27.12

Mus spicilegus 32 266.23 16.21

Mus caroli 30 299.18 20.94

Rattus norvegicus 29 303.40 18.84

Peromyscus maniculatus 126 244.66 31.57

Microtus pennsylvanicus 44 244.84 24.20

B. L. Dumont and B. A. Payseur 6 SI

TABLE S3

Synaptonemal complex lengths in F1 house mice

Cross

(Mother x Father)

Number of Cells

Measured

Mean Synaptonemal

Complex Length (μm) Standard Deviation

Intersubspecific

WSB x CAST 30 231.67 21.77

CAST x WSB 30 245.14 20.43

WSB x PWD 120 241.36 21.34

PWD x WSB 29 232.96 21.07

CAST x PWD 30 275.55 21.80

PWD x CAST 30 251.16 27.54

Intrasubspecific

WSB x PERA 81 228.09 17.07

PERA x WSB 142 227.84 18.64

PWD x CZECHI 30 293.51 27.46

CZECHI x PWD 30 260.56 24.45