First-Generation Linkage Map of the Gray, Short-Tailed Opossum, Monodelphis domestica, Reveals Genome-Wide Reduction in Female Recombination Rates

Corresponding author: Paul B. Samollow, Southwest Foundation for Biomedical Research, 7620 NW Loop 410, San Antonio, TX 78245-0549., pbs{at}darwin.sfbr.org (E-mail)

Communicating editor: S. P. OTTO

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The gray, short-tailed opossum, Monodelphis domestica, is the most extensively used, laboratory-bred marsupial resource for basic biologic and biomedical research worldwide. To enhance the research utility of this species, we are building a linkage map, using both anonymous markers and functional gene loci, that will enable the localization of quantitative trait loci (QTL) and provide comparative information regarding the evolution of mammalian and other vertebrate genomes. The current map is composed of 83 loci distributed among eight autosomal linkage groups and the X chromosome. The autosomal linkage groups appear to encompass a very large portion of the genome, yet span a sex-average distance of only 633.0 cM, making this the most compact linkage map known among vertebrates. Most surprising, the male map is much larger than the female map (884.6 cM vs. 443.1 cM), a pattern contrary to that in eutherian mammals and other vertebrates. The finding of genome-wide reduction in female recombination in M. domestica, coupled with recombination data from two other, distantly related marsupial species, suggests that reduced female recombination might be a widespread metatherian attribute. We discuss possible explanations for reduced female recombination in marsupials as a consequence of the metatherian characteristic of determinate paternal X chromosome inactivation.

METATHERIAN ("marsupial") mammals have a long history in research as models for organismal and cellular physiology, endocrinology, and developmental patterns and processes and more recently have taken their place alongside eutherian ("placental") mammals as serious subjects for genetically oriented biomedical and evolutionary research (reviewed by SAMOLLOW and GRAVES 1998 ; GRAVES and WESTERMAN 2002 ). This increased interest in the genetic characteristics of marsupials reflects the growing utility and importance of comparative models in biomedical research (WOMACK 1997 ; GRAVES 1998 ; POLLOCK et al. 2000 ; POSTLETHWAIT et al. 2000 ) and the broadening recognition of the value of the unique phylogenetic position of marsupials in the mammalian lineage. Marsupials are phylogenetically distinct from the more commonly used mammalian biomedical models, all of which are eutherian species. However, marsupials and eutherians are more closely related to one another than to any other vertebrate model species (e.g., birds, amphibians, fishes). Thus, marsupials represent a unique midpoint between eutherian and nonmammalian vertebrate models.

As a legacy of their common ancestry, marsupials and eutherians share genetic mechanisms and molecular processes that represent fundamental (ancestral) mammalian characteristics. Nevertheless, since their divergence from a common ancestor 150–180 million years ago (MYA; HOPE et al. 1990 ; KUMAR and HEDGES 1998 ; WOODBURNE et al. 2003 ) eutherian and marsupial mammals have evolved many distinctive morphologic, physiologic, and genetic variations on these elemental mammalian designs. These phylogenetically restricted differences can be used as comparative tools for examining the underlying molecular and genetic processes that are common to all mammalian species and thereby help to reveal how variations in these mechanisms contribute to differences in gene regulation, expression, and function.

Monodelphis domestica (also known as the laboratory opossum) is a small South American marsupial that has become widely used as a model organism for comparative research on a broad range of topics that are relevant to human development, physiology, and disease susceptibility (reviewed by VANDEBERG 1990 ; VANDEBERG and ROBINSON 1997 ). Because of its small size (100–150 g), rapid growth and maturation (maturity at 5 months), favorable reproductive characteristics (mean litter size of approximately eight; up to three litters per year), and simple husbandry (rodent cages and commercial feed; VANDEBERG 1999 ), M. domestica has become the most extensively used laboratory-bred marsupial for basic biologic and biomedical research worldwide (SAMOLLOW and GRAVES 1998 ; and recent database searches). Examples of recent research involving M. domestica include: gene and genome evolution, genomic imprinting (autosomal and X-linked genes), photobiology and DNA repair, molecular characterization of ultraviolet radiation-induced skin and eye neoplasias, structure and evolution of mammalian antigen receptors, lipoprotein metabolism and response to dietary fat and cholesterol, neurophysiology, normal and regenerative neurodevelopment (central and peripheral nervous systems), craniofacial ontogeny and evolution, skeletal and skeletomuscular development, reproductive endocrinology, and more.

To further expand the potential of this species as a research model, we have undertaken construction of a linkage map of the M. domestica genome, consisting of both anonymous markers and functional gene loci. The primary objective is to establish a resource that will enable the localization of quantitative trait loci (QTL) that contribute to normal and abnormal physiologic and developmental variation and will provide comparative information regarding the evolution of gene synteny and linkage relationships among distantly related mammalian species.

An important outcome of our ongoing linkage analyses has been to extend and refine fundamental ideas and speculations regarding sex-specific recombination rates in marsupials. Early linkage studies involving small numbers of genes in two distantly related marsupial species yielded the surprising result that meiotic recombination in female marsupials was much lower than that in males. This pattern is contrary to the general mammalian one in which recombination rates are similar between the sexes or are reduced in males: e.g., humans (BROMAN et al. 1998 ; KONG et al. 2002 ), dogs (MELLERSH et al. 1997 ; NEFF et al. 1999 ), pigs (ARCHIBALD et al. 1995 ; MIKAWA et al. 1999 ), cattle (BARENDSE et al. 1997 ; KAPPES et al. 1997 ), and mice (DUNN and BENNETT 1964 ; DAVISSON et al. 1989 ; COPELAND et al. 1993 ).

Specifically, in the fat-tailed dunnart, Sminthopsis crassicaudata (Dasyuridae; an Australian marsupial family of small, mouse-like carnivores), four pairwise gene combinations that exhibited no recombination in females exhibited modest to high recombination frequencies (rf, 0.16–0.37) in males, and two pairwise combinations that had minimal recombination in females (rf, 0.06 and 0.12) were unlinked in males (BENNETT et al. 1986 ). A cytological examination of meiotic cells revealed reduced chiasma frequency in females as compared to males and restriction of female chiasmata primarily to terminal (subtelomeric) regions, suggesting a causal relationship between chiasma characteristics and recombination rates (BENNETT et al. 1986 ). Soon thereafter, HAYMAN et al. 1988 reported similar cytological phenomena in M. domestica (Didelphidae; an American family composed of the opossums and woolly opossums), and subsequent inheritance studies revealed drastically reduced recombination among six loci in females of this species as well (VAN OORSCHOT et al. 1992B , VAN OORSCHOT et al. 1993 ; PERELYGIN et al. 1996 ). The combined weight of these limited findings prompted guarded speculation that reduced female recombination might be a general phenomenon in marsupials (VAN OORSCHOT et al. 1992B , VAN OORSCHOT et al. 1993 ; HOPE 1993 ).

Most recently, ZENGER et al. 2002 published a comprehensive linkage map of the tammar wallaby, Macropus eugenii (Macropodidae; an Australian family composed of the kangaroos, wallabies, and related forms), the first of its kind for any marsupial species. Composed of 60 loci and believed to cover 71% of the M. eugenii genome, the map shows marked heterogeneity of recombination rates including regions of severely reduced female recombination interspersed with intervals in which female and male recombination rates are equivalent. Overall, the female map is 78% the size of the male map, and in none of the interlocus intervals does female recombination exceed that in males.

The linkage map of M. domestica presented in this article is composed of 83 loci distributed among eight autosomal linkage groups and the X chromosome and represents the most extensive linkage data for any marsupial species. Although the map is still under construction, the available data indicate that recombination is considerably more male biased in M. domestica than in the tammar wallaby and is by far the most male-skewed pattern known for any vertebrate species.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Animal resources and genetic mapping panel:
Gray, short-tailed opossums (M. domestica) were obtained from the research colony maintained at the Southwest Foundation for Biomedical Research (SFBR), San Antonio, Texas. The "GMBX" mapping panel was established by crossing members of laboratory stocks that were descended from the most geographically separated populations available (1200 km) at the inception of the study: population 1 (from Exu, in the state of Pernambuco, Brazil) and population 3 (Joaima, Minas Gerais, Brazil). Two wild-captured Pop3 males were crossed to six pedigreed Pop1 females to produce the F₁ generation. Nineteen F₁ offspring of both sexes were crossed to 22 pedigreed Pop1 mates to produce the backcross generation composed of 313 progeny in 30 sibships. Insufficient Pop3 animals were available to produce the reciprocal backcross. Tissue and blood samples were collected from all of the animals produced by this crossing scheme with the exception of the six Pop1 grandmothers. Thus, the 356-member GMBX panel is composed of 30 backcross families including 313 backcross progeny, all of their parents (19 F₁ and 22 Pop1 mates), and the two Pop3 grandfathers.

Sample collection and preparation:
Whole blood and various solid tissues (including liver, brain, kidney, heart, skeletal muscle, and ear pinna) were collected from lethally anesthetized animals under a protocol approved by the SFBR Institutional Animal Care and Use Committee. Dissected tissues were immediately frozen on dry ice and stored in small, heavy-duty, zipper-type plastic bags at -80°. Blood was allowed to clot for 1 hr at ambient temperature and then separated into clot and serum fractions. Both fractions were stored at -80° in 50-µl aliquots, which were sealed in plastic tubing "pillows" following the method of Cheng and co-workers (CHENG et al. 1986 ; CHENG and VANDEBERG 1987 ). Crude extracts of solid tissues and blood clots were prepared by standard methods (e.g., SELANDER et al. 1971 ; SAMOLLOW et al. 1987 ) for application to electrophoretic or isoelectric focusing gels. Serum samples for protease inhibitor (PI) and transferrin (TF) analysis were used directly without additional preparation. For AT3 and C6 analyses, 20-µl serum samples were mixed with 14 µl of 30 mM sodium acetate-95 mM (NH₄)₂SO₄ buffer (pH 5.0) containing 0.008 units of neuraminidase and incubated overnight at 4° prior to use. High-molecular-weight genomic DNA was prepared using routine methods adapted from a variety of sources. Briefly, frozen tissue (usually liver) was ground to a powder in liquid nitrogen, mixed with a 100 µg/ml proteinase K solution, and incubated at 55° overnight with gentle rotation. The resulting solution was RNase treated (20 mg/ml) for 1 hr at 37°, and the DNA was extracted by phenol/chloroform phase separation. The DNA was purified by ethanol precipitation and resuspended in 1x Tris-EDTA (TE) and stored at 4° (or -80° for archival storage) until used.

Polymorphisms—inferred locus homologies and inheritance:
Twelve previously published genetic polymorphisms and 72 newly developed ones were used as genetic markers to genotype all 356 members of the M. domestica GMBX mapping panel. The genetic markers included 21 coding (type I) and 63 anonymous (type II) genetic loci (Table 1 and Table 2, respectively). All polymorphisms were tested for Mendelian inheritance in a subset of families before being screened in the full mapping panel.

Homologies (putative orthologies) of the type I loci to other vertebrate genes were inferred by a variety of approaches. For polymorphic proteins, criteria included: (1) uniqueness of loci in mammals [i.e., only a single adenylate kinase (AK1) is known in mammalian erythrocytes and only one gene encoding GAPD and one gene encoding GPT are known to be expressed in mammals]; (2) immunological cross-reactivity with antibodies developed for other mammalian species—AT3 (BELL et al. 1992 ), C6 (VAN OORSCHOT et al. 1993 ), PI (VAN OORSCHOT et al. 1992B ), and TF (P. B. SAMOLLOW and S. M. MAHANEY, unpublished data); (3) characteristic tissue-specific expression and/or coenzyme requirements—ALDOC (brain-specific; SOKOLOVA et al. 1997 ), DIA1 (predominant erythrocytic diaphorase, NADH requirement; P. B. SAMOLLOW and S. M. MAHANEY, unpublished data), and LDHB (predominant heart isoenzyme); and (4) subcellular localizations (P. B. SAMOLLOW and S. M. MAHANEY, unpublished data)—ACP2 (lysosomal), ACONC (cytosolic), and ME1 (cytosolic).

For PCR-based, type I DNA polymorphisms (AMG, GPD, SP17, and TP53), DNA fragments were generated using primers derived from cloned M. domestica sequences. Homology inferences are based on the strong amplification of single fragments and lack of additional amplification products under high stringency conditions. With the exception of NRAS, hybridization probes for Southern blots were derived from marsupial sources: PHL and the anonymous U15557 probes from M. domestica and the HBE probe from the fat-tailed dunnart, S. crassicaudata. These yielded unique, strong hybridization signals at high stringency, indicative of homology between the probe and the target sequence. NRASLA and NRASLB were detected using a human NRAS genomic probe (MURRAY et al. 1983 ). Although originally reported as homologous to human NRAS (VAN OORSCHOT et al. 1991 ; PERELYGIN et al. 1994 ), results from the present study indicate that NRASLA and NRASLB are unlinked to one another, thus obviating orthology inferences. Preliminary DNA hybridization studies using M. domestica-derived HRAS and KRAS genomic probes (data not shown) have failed to clarify the relationships between these polymorphic NRAS fragments and other RAS family genes. We propose that both polymorphic fragments be classified as NRAS-LIKE loci until their homologies are ascertained by sequencing studies.

Microsatellite and randomly amplified polymorphic DNA (RAPD) polymorphisms were developed for this study using PCR primers designed from cloned M. domestica DNA sequences and commercial RAPD primers, respectively (described below).

All type I protein and DNA polymorphisms exhibited strict codominant inheritance, and no unusual segregation patterns (i.e., segregation distortions) were detected for these or any of the loci examined in this study. Microsatellite variation was primarily codominant, although nonamplifying (null) alleles were detected at seven loci (Table 2). In these cases, individuals exhibiting dominant phenotypes were included in the genotype data set only if their genotypes could be unambiguously inferred from pedigree data. Specifically, individuals that could be either homozygous for an amplifiable allele or heterozygous for that allele and a null allele were excluded from the data set. Apparent null homozygotes were reamplified and rescored multiple times to verify the absence of an amplifiable allele.

RAPD phenotypes encompassed both dominant/recessive and codominant patterns of inheritance (Table 2). Dominant/recessive inheritance occurred at 25 of the 28 RAPD loci, with the presence (P) of an amplimer on a gel dominant to its absence (A). For loci with this inheritance pattern, individuals with dominant phenotypes were included in the genotype data set only if their genotypic status (homozygous P/P or heterozygous P/A) could be unambiguously inferred from pedigree data. For some loci, this criterion excluded entire families (one parent known or suspected of being homozygous P/P) from consideration and thereby reduced the overall informativeness of those loci for mapping purposes. One RAPD locus exhibited codominant variation in amplimer size, and the two remaining RAPD polymorphisms were essentially restriction fragment length polymorphisms (RFLPs) detected by restriction digestion of the RAPD-PCR amplimer. In general, the RAPD polymorphisms, by dint of their dominant inheritance patterns, were the least informative of the polymorphisms used in this study.

Protein polymorphisms:
Protein polymorphisms were detected using electrophoretic and isoelectric focusing methods modified from a variety of published sources. Tissues used and electrophoretic method/buffer combinations are listed in Table 3. Horizontal starch gel electrophoresis (SGE) was conducted in 12% starch gels (e.g., SHAW and PRASAD 1970 ; SHAKLEE and KEENAN 1986 ; MORIZOT and SCHMIDT 1990 ). Vertical polyacrylamide gel electrophoresis (PAGE) was conducted using a 10% acrylamide (1:37.5 bis:acrylamide) running gel overlaid with a 4% acrylamide (1:37.5 bis:acrylamide) stacking gel (e.g., GAHNE et al. 1977 ; SAMOLLOW et al. 1987 ). Cellulose acetate gel electrophoresis (CAGE) was conducted in Cellogel (Chemtron, Milan, Italy) cellulose acetate medium (e.g., RICHARDSON et al. 1986 ). Polyacrylamide gel isoelectric focusing (PAGIEF) was performed in 5% polyacrylamide (1:29 bis:acrylamide) gels prepared by two methods. For AT3, 0.45-mm-thick gels were prepared according to the riboflavin photopolymerization protocol of POLLITT and BELL 1983 . For PI, 0.30-mm gels containing 10% sucrose (w:v) were produced by standard TMED/ammonium persulfate polymerization. Agarose gel isoelectric focusing (AGIEF) was performed in 0.5-mm, 1.25% isoagarose (FMC BioProducts, Rockland, ME) gels containing 13.75% sucrose (w:v).

Visualization ("staining") of 13 protein polymorphisms on gels was accomplished by a variety of methods modified from published and unpublished sources (Table 3). Details of these procedures are available online at the GENETICS supplemental information website: http://www.genetics.org/supplemental.

Southern blotting/RFLP analysis:
Restriction enzyme digests of M. domestica genomic DNA were electrophoretically separated on 1% agarose gels and then transferred and fixed to nylon filters (Hybond-N+; Amersham Biotech, Piscataway, NJ) by standard methods (e.g., SAMBROOK et al. 1989 ). Radiolabeled (³²P) probes for HBE, NRAS, PHL, and U15557 (described below) were prepared by random priming (Amersham Biotech oligolabeling kit). Labeled HBE and PHL probes were hybridized to filters for 16 hr at 65°, after which the filters were washed three times to a final stringency of 2x SSC/0.1% SDS at 65°. Hybridization and wash conditions for NRAS and U15557 are given by PERELYGIN et al. 1994 , PERELYGIN et al. 1996 (respectively). The hybridization probes used for Southern blot analyses of RFLPs were:

• HBE: embryonic ß-globin genomic DNA fragment, clone pSG-2H, from the dasyurid marsupial, S. crassicaudata (HOPE et al. 1992 ), gift of Rory Hope.

• NRAS: human NRAS proto-oncogene genomic DNA fragment, clone p52c- (MURRAY et al. 1983 ) from American Type Culture Collection (no. 41030).

• PHL: probe generated by PCR amplification from M. domestica genomic DNA using primers designed from published M. domestica photolyase cDNA sequence data (KATO et al. 1994 ). Forward primer was 5'-AAAAAGGGAGGAGCAGAAGGC-3'; reverse primer was 5'-GTCATAATGGCGAATGGTAGCAC-3'. The amplified fragment was cloned into pT7Blue(R) and transformed into NovaBlue competent cells (Novagen, Madison, WI).

• U15557: anonymous M. domestica genomic DNA fragment clone (PERELYGIN et al. 1996 ), GenBank accession no. U15557.

Polymerase chain reaction amplimer-restriction fragment length polymorphisms:
PCR was used to generate amplimers that were subsequently subjected to restriction endonuclease digestion to reveal restriction site polymorphisms (PCR-RFLPs):

• G6PD: PCR primers were designed from published G6PD genomic DNA sequence of the Virginia opossum, Didelphis virginiana (KASLOW et al. 1987 ) and used to amplify an 2400-bp M. domestica genomic DNA fragment. This fragment was partially sequenced (data available on request), and the resulting data were used to design new, M. domestica-specific PCR primers: forward primer (exon 5), 5'-CATCAAAGAAACTTGCATGAGCCAG-3' and reverse primer (exon 8), 5'-ATCTGGAGGAGGTGGTTCTGCATCA-3'. Amplification was achieved using a "touchdown" PCR procedure: initial 5-min denaturation step at 95°; 10 cycles of 40-sec denaturation at 94°, 30-sec annealing beginning at 69° for the first cycle and decreasing 1° each subsequent cycle, and 60-sec extension at 72°; 30 identical cycles of 40 sec at 94°, 30 sec at 59°, and 60 sec at 72°; and a final 5-min extension at 72° (unless otherwise specified, all PCR procedures for this study were conducted using ABI 9600 or 9700 thermocyclers). The amplified M. domestica fragment contained a polymorphic PstI restriction site. This PCR-RFLP was visualized on 1% agarose gels (1x TE) stained with ethidium bromide.

• SP17: M. domestica SP17 cDNA sequence data (GenBank accession no. AF054290 and M. G. O'RAND, personal communication) were used to design primers for sequencing the intron 1 region of the M. domestica SP17 gene. The intron 1 boundary was inferred from gene structure information of human, mouse, and rabbit homologs. PCR primers were designed to amplify an 2400-bp fragment of intron 1, which was partially sequenced (data available on request) and found to contain a single nucleotide polymorphism (SNP) in an HaeIII restriction site. The PCR primers were: forward, 5'-CACTGTATTCCATTTTACTC-3' and reverse, 5'-TTCCTACTTTACATATGAGG-3'. Amplification was accomplished using touchdown PCR: initial 5-min denaturation step at 95°; 10 cycles of 40-sec denaturation at 94°, 30-sec annealing beginning at 67° for the first cycle and decreasing 1° each subsequent cycle, and 150-sec extension at 72°; 30 identical cycles of 40 sec at 94°, 30 sec at 57°, and 150 sec at 72°; and a final 7-min extension at 72°. The polymorphism was analyzed as an HaeIII PCR-RFLP on 1% agarose gels stained with ethidium bromide.

• TP53: Published M. domestica TP53 exon 4–11 cDNA sequence data (KUSEWITT et al. 1999 ) and unpublished partial intron sequence data (D. F. KUSEWITT, personal communication) were used to generate sequencing primers for intron regions. Intron 7 was partially sequenced (data available on request) and found to contain an SNP in an AluI restriction site, which was analyzed as a PCR-RFLP pattern visualized on 1% agarose gels stained with ethidium bromide. The PCR primer sequences for amplification of the 850-bp, intron 7-containing fragment of the TP53 gene were: forward, 5'-GCGCCCAATCCTGACTATCAT-3' and reverse, 5'-AGGAAGGGACAGGTAGAAGGA-3'. PCR amplification was accomplished using touchdown PCR: initial 5-min denaturation at 95°; 10 cycles of 30-sec denaturation at 94°, 30-sec annealing beginning at 78° for the first cycle and decreasing 2° for each subsequent cycle, and 60-sec extension at 72°; 30 identical cycles of 30 sec at 94°, 30 sec at 58°, and 60 sec at 72°; and a final 5-min extension at 72°.

• AMG: M. domestica AMG cDNA sequence data (HU et al. 1996 ) were used to design primers to amplify a region spanning 22 bp of exon 5, all of intron 5, and 69 bp of exon 6. The amplified fragment varied from 430 to 450 bp, indicating an 20-bp insertion/deletion (indel) polymorphism. The primers were: forward, 5'-TCAAAGCATGATGCGACAGC-3' and reverse, 5'-CTGAGATAGCACTGGGATGA-3'. Touchdown PCR procedures were identical to those for G6PD except that the initial and final annealing temperatures were 68° and 58°, respectively. The indel polymorphism was visualized on 1-mm thick, 6.5% polyacrylamide (1:37.5 bis:acrylamide) gels stained with ethidium bromide.

RAPD analysis:
RAPD polymorphisms were detected using arbitrarily primed PCR as described by WILLIAMS et al. 1990 . Briefly, arbitrary 10-base oligonucleotides (Operon RAPD 10-Mer kits A and B; Operon Technologies, Alameda, CA) were used alone or in pairs to amplify random sequences from genomic DNA. The arbitrary primers used in this study are listed in Table 2. RAPD-PCR reaction mixtures consisted of 2.0 mM MgCl₂, 0.1 mM each dNTP (dATP, dCTP, dGTP, and dTTP), arbitrary primer(s) (0.4 µM if a single primer was used; 0.2 µM each if two primers were used), 5 units AmpliTaq Polymerase (Applied Biosystems, Foster City, CA), 1x PCR AmpliTaq PCR buffer, and 25 ng of genomic DNA. Cycle parameters (ABI 480 thermocycler) were: denaturation at 94° for 2 min, followed by 45 cycles of 1 min at 94°, 1 min at 36°, and 2 min at 72°. Resultant amplification products were run out on 1.4% agarose gels and visualized by ethidium bromide staining. RAPD banding patterns on gels were scrutinized forvariation in three categories: (1) PA, presence/absence variation of a band at a specific position on the gel, presumed to reflect priming sequence variation or large insertion/deletion variants that preclude successful amplification; (2) SV, size variation of a fragment revealed by differences in the position of a band on a gel presumed to reflect small insertion/deletion variation; and (3) RFLP, length variation of restriction endonuclease-digested RAPD-PCR amplification product.

Microsatellite development and polymorphisms:
Detection of short tandem repeat sequences (microsatellites) was conducted following the methods outlined in HILLIS et al. 1996 . Briefly, a small-fragment genomic DNA library was constructed by cloning partially Sau3AI-digested and size-selected (200–800 bp) genomic DNA into the pT7 Blue T-Vector (Novagen) according to the manufacturer's instructions. The library (maintained in Nova Blue cells) was plated, grown, and transferred to nylon membranes, which were probed at high stringency with (CA)₁₅ or (GT)₁₅ probes (Genosys Biotechnologies, The Woodlands, TX) to detect the presence of dinucleotide repeat sequences in the plasmid vectors. DNA from strongly hybridizing clones was isolated and purified (Wizard Plus Mini-preps DNA purification system; Promega, Madison, WI) and then used for sequencing analysis. Sequencing of plasmid DNA was accomplished either by standard dideoxy sequencing methods (e.g., SAMBROOK et al. 1989 ) with ³⁵S-labeled dATP using the Sequenase version 2.0 sequencing kit (Amersham Biotech) or by fluorescence-based cycle sequencing methods (ABI PRISM Big Dye Terminator Cycle Sequencing ready reaction kit; Applied Biosystems). For cycle sequencing, PCR reaction products were "cleaned" with shrimp alkaline phosphatase and exonuclease I (Amersham Biotech) according to the method of NICKERSON et al. 1998 and sequenced on an Applied Biosystems (ABI) 377 automated DNA sequencer following the manufacturer's instructions. The DNA sequence data were analyzed using Sequencing Analysis Version 3.3 and Factura Version 2.2 software. Data from the forward and reverse reactions were used to determine the base sequences of the unique (nonrepetitive) regions flanking the repeat region and the nature and size of the repeat region itself. PCR primer pairs were designed to match unique flanking sequences, and DNA samples from the smallest subset of completely informative individuals in the GMBX panel (those that contained all alleles that contributed to the backcross generation) were amplified and screened for variation in amplimer length at each microsatellite locus.

Genotyping and data quality:
Genotypes were scored independently by two observers. Samples yielding ambiguous scorings were rerun and rescored until agreement was reached or it was decided that the sample could not be scored for the particular marker. In rare cases, wherein an individual's genotype at a particular microsatellite locus was unambiguous but could not be reconciled with parental and sibling genotypes even after repeated typings (apparent mutations), the individual's genotype was excluded from the mapping analysis. The overall quality of the genotyping data was assessed by random retyping of an average of 17.2% (range, 10.6–23.0% per locus) of previously typed samples. Discrepancy rates were lowest for type I DNA (0.19%) and protein (0.39%) polymorphisms and somewhat higher for microsatellite (0.82%) and RAPD polymorphisms (1.0%). The average random repeat discrepancy rate across all loci was 0.77%. All discrepancies were reconciled, and many nonrandom sample repetitions were performed; thus the final genotyping error rate was slightly lower than that suggested by these random repeat error rates. Overwhelming agreement of parent-offspring genotypes suggests that there were no pedigree errors among the 356 animals used in the mapping analysis.

Map construction:
The GMBX panel was constructed by crossing members of two outbred populations that had fixed genetic differences at some loci, but shared alleles at other loci. Therefore, the number of informative meioses varied substantially across loci (Table 1 and Table 2). The average number of informative meioses per locus, 150 in females and 177 in males, was sufficient to assure high levels of support for locus order for most loci.

Construction of the M. domestica linkage group maps proceeded in four phases: identifying linked loci, ordering linked loci, removal of double recombinants, and final map construction. We used the computer program Crimap (GREEN 1990 ) to perform a series of two-point (pairwise) linkage analyses and assigned markers to linkage groups. To be included within a linkage group, loci had to be linked to at least one other locus within the group at a LOD 3.00 (1000:1 odds). We next ordered the loci within each autosomal linkage group using the expert system program MULTIMAP (MATISE et al. 1994 ), which implements routines of the computer program Crimap. In MULTIMAP, the two loci with the highest pairwise heterozygosity are chosen to start the map and additional loci are added in order of informativeness. At each step, MULTIMAP determines a reasonable order for the set of loci, as well as validating that this order has a higher likelihood than other orders, by inverting groups of three loci and keeping the order with the highest likelihood. This process continues until no more loci can uniquely be added to the map, on the basis of a criterion of LOD 2.0 (100:1 odds) as statistical support for locus order. After ordering the loci within each linkage group, we used Simwalk2 to detect possible double recombinants (WEEKS et al. 1995 ; SOBEL et al. 2001 ). Simwalk2 uses Markov chain Monte Carlo and simulated annealing algorithms to perform these multipoint haplotype analyses and reports the overall probability of mistyping, i.e., evidence of a double recombinant, at a locus. We removed individual genotypes that had >50% likelihood of mistyping. Finally, MULTIMAP was rerun to obtain estimates of sex-specific orders and distances. Map distances were calculated as centimorgans (cM) using the Kosambi mapping function (see OTT 1999 ). With the exception of linkage group 7, which was composed of essentially four recombining loci, we again required LOD 2.00 as statistical support for order.

Sex-specific differences:
We tested for sex-specific heterogeneity in linkage group length by using the likelihood ratio test to compare the likelihood of each linkage group with sex-specific recombination vs. the null hypothesis of sex-equal recombination. The likelihood ratio test is asymptotically distributed as a chi square with degrees of freedom equal to the number of intervals minus one. We also tested for sex-specific heterogeneity of recombination within each interval by comparing the maximum LOD for linkage when male and female recombination is equal [Z(_m = _f): the null hypothesis] vs. the maximum LOD obtained with sex-specific recombination [Z(_m, _f)] (OTT 1999 ). The resulting likelihood ratio test, ² = 2 x ln(10) x [Z(_m, _f) - Z(_m = _f)], is asymptotically distributed as a chi-square distribution with 1 d.f.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Linkage map construction:
Two-point linkage analysis using a linkage criterion of LOD 3.0 identified eight autosomal linkage groups (LGs 1–8) and an X chromosomal linkage group (LG X) comprising 83 of the 84 loci examined (LDHB failed to link to another locus with LOD 3.0). Loci were ordered within LGs 1–8 at a minimum criterion of LOD 2.0 (100:1 odds) for inclusion. Following detection and elimination of improbable double recombinants, sex-specific locus orders and recombination distances were recalculated to yield female- and male-specific recombination maps for LGs 1–8. Multipoint analysis was not pursued for the 3 loci in LG X. Data pertaining to the number of loci, statistical support levels, and sizes of the sex-average and sex-specific linkage groups obtained from multipoint linkage analyses are listed in Table 4. Diagrammatic representations of the eight autosomal linkage group maps for both sexes, including locus orders, map positions, and interlocus interval support levels are shown in Fig 1. Locus C7 was not examined in the present study but was placed on the map by inference from the finding of VAN OORSCHOT et al. 1993 that C6 and C7 are extremely tightly linked in M. domestica ( = 0.00, Z_max = 31.18).

View larger version (51K): In this window In a new window Download PPT slide   Figure 1. Female- and male-specific linkage group maps of M. domestica. The female map is to the left in each linkage group. Locus abbreviations lie between the female and male maps, and locus positions are indicated by their distance in centimorgans from the "top" of the linkage group. Boxes enclose nonrecombining loci, which can differ between sexes. Bars to the right of the male map indicate the most likely ranges for loci that could not be ordered with high confidence. Numbers to the extreme right of each linkage group indicate the statistical support (LOD) for each interval, given the locus order shown.

Not all of the 80 loci assigned to autosomal linkage groups could be unambiguously ordered. Five RAPD loci (Mdo1LR04, Mdo2LR06, Mdo3LR08, Mdo4LR12, and Mdo6LR18 in LGs 1, 2, 3, 4, and 6, respectively) could be placed only within multilocus regions rather than at specific points on the corresponding linkage group map. Three additional RAPD loci (Mdo1LR05, Mdo3LR11, and Mdo7LR22 in LGs 1, 3, and 7, respectively) could not be placed on the maps with any confidence despite linking strongly to one or more loci within their corresponding linkage groups. Eight locus pairs exhibited no recombination in either sex. Lack of recombination for two of these pairs could be attributable to a low number of pairwise informative meiosis rather than to tight linkage (AT3-Mdo2LR10 with 11 pairwise informative meioses and AMG-Mdo5LR17 with 20 pairwise informative meioses), but the remaining six locus pairs had between 44 and 426 pairwise informative meioses (total for sexes combined), so tight linkage is the more likely explanation in these cases.

Statistical support for map order is very good; average LOD is 27.9 across all 56 recombining intervals. Only 3 intervals had multipoint support levels less than LOD 3.00. The structure of LG 7 is problematic. This linkage group contains only six loci, two pairs of which exhibit no recombination; thus, the LG 7 map has only 3 interlocus intervals. The overall order for LG 7 is supported only at LOD 1.6, although support for 2 of the LG 7 intervals is strong for the given order (LOD 22.0 and LOD 5.8). We therefore consider the map for LG 7 as provisional with regard to the placement of Mdo7LR27.

Map size and sex-specific recombination rates:
Two striking features of the M. domestica linkage map are its small overall size and the very large differences in recombination rates between sexes. The sex-average map size of only 633.0 cM is, to our knowledge, the smallest of any vertebrate species. This small size is due in part to the substantial reduction in recombination seen in females as compared to males (see below), but even the larger male map, which spans 884.6 cM, is smaller than any other vertebrate linkage map, regardless of sex.

Noninclusion of large portions of the genome does not appear to be an adequate explanation for the compact linkage map of M. domestica for several reasons. The M. domestica karyotype is composed of eight pairs of large autosomes and the X and Y sex chromosomes (PATHAK et al. 1993 ). The coalescence of 83 genetic markers into eight autosomal linkage groups and one X chromosomal linkage group in the M. domestica linkage map is consistent with this species karyotype. Failure of LDHB to link to any of the linkage groups is almost certainly due to its lack of variation (only 19 informative meioses) rather than to its location in an uncharted portion of the genome. Further, the smallest autosomal linkage group in the M. domestica map includes six loci, and LG X includes all of the X-linked loci examined. There are no other unlinked loci, nor are there any locus pairs or triplets that do not link to one of the nine linkage groups. Moreover, LGs 1–8 were identified during an early phase of mapping using only 46 loci. Since that time, no newly examined locus has failed to link to one of the existing linkage groups. Another indication that no large, undetected regions of the genome remain is that the continuing addition of new loci has not dramatically altered the size of the map. For example, a series of locus additions that increased the map content from 59 loci to its current 83 loci (a 46% increase) increased the overall map size by only 13%. For all of these reasons we believe that the eight autosomal linkage groups correspond to the eight autosomes of this species and that the current linkage map encompasses a very large proportion of the genome. If this is true, M. domestica possesses a generally low rate of meiotic recombination and particularly so in females.

Overall, the male genetic map is twice the size of the female map (884.6 vs. 443.1 cM), and male linkage group sizes exceed those for each of the corresponding female linkage groups (Fig 2). These differences are statistically significant for the five larger linkage groups (LGs 1–4 and 6), but not for the three smaller ones, which have fewer loci (Table 4). Of the 56 recombining intervals, 20 (representing 45.9% of the total sex-average map length) show significantly higher male recombination, while only two (representing 6.7% of total sex-average map length) exhibit higher female recombination (Table 5). Male recombination appears to be higher on average over the remaining 34 (nonsignificant) intervals as well, as judged from the summed lengths of these intervals: 345.5 cM in males and 250.3 cM in females. Only LGs 5 and 6 exhibit no significant differences in interval lengths between sexes, but, as mentioned, the overall length of LG 6 is significantly greater in males than in females. LG 5, which is the smallest with regard to length and number of loci, is the only linkage group that exhibits no significant between-sex differences either in total length or at the interlocus interval level.

View larger version (54K): In this window In a new window Download PPT slide   Figure 2. Relative sex-specific linkage group sizes of M. domestica. Linkage groups are arranged in descending order of male linkage group size, with male linkage groups above and corresponding female linkage groups below. Linkage group sizes (in centimorgans) are indicated above and below the male and female bars, respectively. Probabilities are for tests of male vs. female differences in linkage group size (see footnote b of Table 4).

The magnitude of the recombination rate differences varies widely among linkage groups, with male linkage group sizes ranging from 1.2 to 5.3 times the size of the corresponding female linkage groups. Consideration of deviations of sex-specific recombination rates from sex-average rates for each interlocus interval (illustrated in Fig 3) suggest that sex differences, while widespread, are exaggerated in specific chromosomal regions rather than uniform across a chromosome. Interpretation of these regional and linkage-group level differences is hampered by the varying levels of informativeness among the loci studied; some seemingly large differences in interlocus distances between the sexes undoubtedly failed to reach significance because of the low number of informative meioses among the loci involved. However, 43 of the 56 interlocus intervals are longer in males than in females, and the male map significantly exceeds the female map for almost half (46%) of its length. The overall impression is that male recombination is greater than female recombination for most of the genome, but that the intensity of the difference varies among chromosomes and chromosomal regions.

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. Sex-specific deviations from the sex-average recombination rate in M. domestica. Interval numbers correspond to interlocus intervals in Fig 1, beginning at the top of the linkage group diagram; e.g., interval 3 of LG 1 corresponds to AK1–U15557, and interval 4 of LG 2 corresponds to the nonrecombining MdoL005–MdoL006 "interval." The horizontal line represents the sex-average recombination rate. Overlapping male and female symbols on the horizontal line indicate that recombination occurred in the interval, but that the sex-specific rates were identical. Solid squares indicate intervals with no recombination in either sex.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Sex-average linkage map:
The length of the M. domestica sex-average linkage map, excluding X-linked loci, is only 633 cM. As argued above (RESULTS), the map appears to represent nearly full coverage of the M. domestica genome. However, an estimate of the total linkage map size must account for undetected regions beyond the terminal markers of each linkage group and for the length of the X chromosome. To estimate the length of putative undetected telomeric regions, we used the simple expedient of assuming them to be the same size as the average interlocus distance (ILD) of the mapped loci (e.g., FISHMAN et al. 2001 ). Specifically, the average ILD was calculated as the total map length divided by the number of recombining intervals: 633/56 = 11.3 cM. A commonly used alternative method (e.g., MIKAWA et al. 1999 ; ZENGER et al. 2002 ) uses the total number of linked autosomal loci minus the number of linkage groups (80 - 8 = 72) as the denominator. This more liberal approach yields an average ILD estimate of 8.79 cM.

We have no estimate of the length of LG X, but physically the X chromosome is less than half the size of the smallest M. domestica autosome (PATHAK et al. 1993 ). It seems reasonable to assume that its recombinational length would not exceed that of the average of the three smallest autosomes; thus we use the average length of the three smallest linkage groups, 54.3 cM, as our estimate of LG X length. Combining the lengths of LGs 1–8 with the estimate for LG X and 18 linkage group ends yields an estimate of 633 + 54.3 + (18 x 11.3) = 890.7 cM for the M. domestica sex-average recombinational map. To our knowledge, this is the most compact sex-average linkage map of any vertebrate species.

The lengths of mammalian sex-average linkage maps are highly variable, indicating a broad range of recombination rates among species. Examples range from 1398 cM for laboratory mice (RHODES et al. 1998 ) to >3500 cM in humans (KONG et al. 2002 ), cattle (BARENDSE et al. 1997 ; KAPPES et al. 1997 ), and sheep (MIKAWA et al. 1999 ). However, the length of the mouse linkage map could underestimate the true recombination rate of individual mouse species or strains because it was assembled using combined data from both conspecific crosses and interspecific hybrids (COPELAND et al. 1993 ; DIETRICH et al. 1996 ). Among maps constructed using strict conspecific crosses, the 1510-cM dog (Canis familiaris) map (NEFF et al. 1999 ) and 1790-cM rat (Rattus norvegicus) map (BONNE et al. 2002 ) are the smallest, but both are incomplete. The full-coverage dog and rat maps are estimated to be 2700 and 1989 cM, respectively. Among other vertebrates, estimated map lengths vary from 1350 cM in rainbow trout (Oncorhyncus mykiss; SAKAMOTO et al. 2000 ) to 7291 cM in hybrid crosses of the salamanders Ambystoma mexicanum and A. tigrinum (VOSS et al. 2001 ).

The tammar wallaby, M. eugenii, is the only other marsupial for which a comprehensive linkage map has been published (ZENGER et al. 2002 ). The evolutionary distance between M. eugenii and M. domestica is 60–70 million years (KIRSCH et al. 1997 ; SPRINGER 1997 ; GRAVES and WESTERMAN 2002 ), which is roughly comparable to that between humans and mice. The M. eugenii map length of 828.4 cM includes the X chromosome, but is not corrected for putative undetected telomeric regions and is estimated to represent only 71% of a total length of 1172 cM. This total map length is 31.6% larger than our estimate of the M. domestica sex-average map, but is still quite small by vertebrate standards. The M. domestica genome (3.6–3.7 pg of the DNA/haploid genome, estimated by flow cytometry; J. DANKE, unpublished data) is 6% larger than that of humans, while that of M. eugenii may be as much as 18% larger than the human genome [data of HAYMAN and MARTIN 1974 , cited by SHARP and HAYMAN 1988 and GREGORY 2001 ]. Clearly, recombination rates (centimorgans per megabases) in these two distantly related marsupial species are very low relative to those of other mammals and of vertebrates in general.

It is tempting to propose that the short map lengths of M. domestica and M. eugenii are related to the low chromosome numbers of these two species. However, an examination of total map lengths (in centimorgans) and chromosome numbers (n) among the 19 nonmetatherian vertebrates for which sufficient data are available, suggests that chromosome number does not explain the majority of the observed variation in map lengths (Fig 4). Among eutherian mammals, for example, dogs have more than twice as many chromosomes as cats (n = 39 and 19, respectively), but the dog linkage map (2700 cM) is substantially shorter than the cat map (3300 cM). Humans and Syrian hamsters, on the other hand, have nearly identical chromosome numbers (23 and 22, respectively), but their map lengths (3600 and 2000 cM, respectively) are widely divergent. Even comparing the marsupials, M. eugenii has fewer chromosomes than M. domestica, while its map is nearly a third longer. Moreover, regression analyses (excluding the clearly anomalous amphibian relationship) of map length on chromosome number for various vertebrate groupings reinforce the visual impression that chromosome number is a poor predictor of map length. The correlations (r²) between map length and chromosome number for the various groups were: eutherians alone, 0.12 (P = 0.277); eutherians + chicken, 0.22 (P = 0.107); eutherians + fishes, 0.07 (P = 0.322); and eutherians + chicken + fishes, 0.16 (P = 0.102). However, when the two marsupials were added to the analyses, all of the corresponding correlations became stronger (r² ranged from 0.32 to 0.51) and were statistically significant (P = 0.003–0.012). Because the marsupial data were extreme values and provided strong leverage points for the regression analyses, the implications of this latter outcome are not clear. In general, however, there is little support for the concept of a robust relationship between map length and chromosome number among the vertebrate species examined. More data from marsupials and other vertebrates with low chromosome numbers will be needed to adequately test this hypothesis.

View larger version (20K): In this window In a new window Download PPT slide   Figure 4. Total map length vs. chromosome number in vertebrates. See text for discussion. Data sources are as follows: human (BROMAN et al. 1998 ; KONG et al. 2002 ), baboon (ROGERS et al. 2000 ), mouse (RHODES et al. 1998 ), rat (BONNE et al. 2002 ), Syrian hamster (OKUIZUMI et al. 1997 ), cat (MENOTTI-RAYMOND et al. 1999 ), dog (NEFF et al. 1999 ), cattle (average of estimates of BARENDSE et al. 1997 ; FERRETTI et al. 1997 ; KAPPES et al. 1997 ), pig (MIKAWA et al. 1999 ), sheep (MADDOX et al. 2001 ), deer (SLATE et al. 2002 ), horse (GUERIN et al. 2003 ), M. domestica (this study), M. eugenii (ZENGER et al. 2002 ), chicken (GROENEN et al. 2000 ), salamander (VOSS et al. 2001 ), channel catfish (WALDBIESER et al. 2001 ), medaka (NARUSE et al. 2000 ), rainbow trout (mean of estimates derived from YOUNG et al. 1998 ; SAKAMOTO et al. 2000 ), Xiphophorus sp. (S. KAZIANIS, R. S. NAIRN, R. B. WALTER, D. A. JOHNSTON, J. KUMAR et al., unpublished data), and zebrafish (SHIMODA et al. 1999 ). Vertebrate species with substantially incomplete maps were excluded from this analysis. For consistency with the other species examined, the published total map lengths for rat, deer, horse, and medaka were adjusted for missing telomeric regions by adding twice the average interlocus distance for the particular species map to each of the identified linkage groups.

Interspecific synteny comparisons:
Inspection of the linkage group affiliations of the 20 type 1 loci mapped in this study is inconclusive with regard to conservation of gene synteny with eutherian species. SP17 and TF are syntenic in M. domestica (LG 4), mice (Mmu 9), and humans (Hsa 3q), but LG 4 also includes HBE, which is not syntenic with SP17 or TF in mice or humans (a fourth coding locus in LG 4, PHL, has no confirmed homolog in eutherians). C6/C7 and GPT are syntenic in M. domestica (LG 2) and mouse (Mmu 15), but not in humans (Hsa 5p and Hsa 8q, respectively). A series of two-point linkage comparisons reported by SOKOLOVA et al. 1997 suggested that NF1, ERBB2, RARA, HOX2, MPO, PRKCA, and ALDOC are syntenic in M. domestica, and the present study shows that ALDOC and TP53 are syntenic on M. domestica LG 3. These 8 loci are known to be syntenic on Mmu 11 and Hsa 17q, suggesting that a substantial 8-locus synteny may have been conserved among M. domestica, mice, and humans. However, LG 3 is not entirely homologous to Hsa 17q and Mmu 11, as it also contains loci found on Hsa 1p (NRASLB), Hsa 1q (AT3), and Hsa 6q (ME1); these loci are similarly dispersed in the mouse genome. The impression of mixed conservation and rearrangement relative to eutherian genomes agrees with extensive physical mapping results from M. eugenii and a few other marsupial species (reviewed by SAMOLLOW and GRAVES 1998 ). Unfortunately, except for several X- and Y-linked genes, there is almost no overlap among the type I loci mapped in M. domestica and M. eugenii (SAMOLLOW and GRAVES 1998 ; and more recent articles by DELBRIDGE et al. 1999 ; HAWKEN et al. 1999 ; CHARCHAR et al. 2000 ; WATERS et al. 2001 ), so there is currently no opportunity for synteny comparisons between these two marsupial species.

Sex-specific recombination:
The small size of the M. domestica linkage map is remarkable in its own right, but more surprising is the difference in sex-specific recombination rates implied by the female- and male-specific linkage maps. The combined male linkage map (884.6 cM) is twice the size of the female map (443.1 cM; the female to male recombination ratio, F/M, is 0.50), and each of the individual linkage groups is larger in males than in females. This sex difference is extraordinary not only for its magnitude, but also for the direction of the sex-specific recombination bias, which is unlike anything observed in other vertebrates except in two distant marsupial relatives, M. eugenii and S. crassicaudata (fat-tailed dunnart). Unfortunately, linkage data from S. crassicaudata (BENNETT et al. 1986 ) are very limited (six loci) and not directly comparable to those from M. domestica or M. eugenii. However, interesting comparisons are possible with the M. eugenii map. The reduction in female recombination in M. domestica is greater in both extent and magnitude than that seen in M. eugenii. The M. eugenii map exhibits significant sex differences in only 8 interlocus intervals covering 26.2% of the sex-average autosomal map length (calculated from data of ZENGER et al. 2002 ), and the overall F/M ratio is 0.78, compared with 0.50 in M. domestica. After removal of the 8 significant intervals from the M. eugenii map, the F/M ratio increased to a sex-equal 1.01 (ZENGER et al. 2002 ), while in M. domestica, removal of the 20 significant intervals yields an F/M ratio of 0.72, a ratio that is more skewed than the F/M ratio for the full M. eugenii map with all 8 significant intervals included.

Sex-specific differences in meiotic recombination rate are common among a broad range of animal and plant species. HALDANE 1922 and HUXLEY 1928 each noted this general trend and suggested that in species where the sexes had substantially different recombination rates, it was the heterogametic sex that had the reduced rate. While neither author proposed this situation to be inviolable, and exceptions are known [e.g., flour beetles, Tribolium castaneum (SOKOLOFF 1964 )], this prediction has turned out to be accurate in the great majority of species in which strong sex-specific differences in recombination frequency have been documented on a genome-wide basis (regional sex-specific reversals over short map distances are known in many species, but these do not affect the genome-wide generalization).

Most eutherians examined exhibit substantially higher female than male recombination rates, and, among those with extensive maps, the F/M ratio is almost always >1.0; e.g., 1.56–1.65 in humans (BROMAN et al. 1998 ; KONG et al. 2002 ); 1.36–1.41 in dogs (MELLERSH et al. 1997 ; NEFF et al. 1999 ), and 1.30–1.55 in pigs (ARCHIBALD et al. 1995 ; MIKAWA et al. 1999 ). In cattle, the female map is just slightly larger than the male map: F/M = 1.03–1.06 (BARENDSE et al. 1997 ; KAPPES et al. 1997 ). Genome-wide female bias in recombination is also evident in the baboon linkage map (J. ROGERS and M. C. MAHANEY, personal communication) and in linkage data from each of six rhesus macaque chromosomes examined to date (J. ROGERS, personal communication). A possible exception to this pattern has been observed in sheep, in which the F/M ratio = 0.83; but this ratio must be regarded as provisional because peculiarities of the mating scheme used to produce the sheep maps could have contributed to an underestimate of male map length (MADDOX et al. 2001 , p. 1285).

The tendency for female map length to exceed that of males appears to be greatly exaggerated in fishes: zebrafish F/M = 2.74 (SINGER et al. 2002 ); catfish F/M = 3.18 (WALDBIESER et al. 2001 ); rainbow trout F/M = 3.25 (SAKAMOTO et al. 2000 ), but absent in birds. In the only avian map available, that of chicken, F/M = 0.99 (GROENEN et al. 1998 , GROENEN et al. 2000 ), and a cytologic study in pigeons revealed equal numbers of chiasmata in female and male meiotic cells (PIGOZZI and SOLARI 1999 ). Separate female and male maps are not available for amphibians, but recombination data for six autosomal loci in the frog Rana brevipoda and a pair of loci in its congener R. japonica indicate that strongly female-biased recombination occurs in at least some amphibians (SUMIDA and NISHIOKA 2000 ). Average chiasma counts for male crested newts (Triturus cristatus) exceed those of females on all chromosomes (WALLACE et al. 1997 ), suggesting higher male than female recombination rates (chiasma-based F/M = 0.77) in this species. However, the chiasmata in oocytes are strongly clustered around centromeres, while those in spermatocytes are restricted to the more distal regions of the chromosome arms; thus the male and female recombining regions are largely distinct and nonoverlapping. This dimorphism, together with the lack of genetic mapping data for this species, inhibits interpretation of these cytological observations with regard to recombination across the genome as a whole. In summary, we have been unable to identify a single unambiguous case of excess male recombination, as gauged from genome-wide genetic data, in any nonmetatherian vertebrate species.

The occurrence of reduced female recombination in both American and Australian marsupial species, which last shared a common ancestor at least 60 MYA (KIRSCH et al. 1997 ; SPRINGER 1997 ; and discussions by KIRSCH and MAYER 1998 ; SPRINGER et al. 1998 ), tempts speculation that reduced female recombination is a general characteristic that distinguishes metatherian from eutherian species. With so few mammalian species maps completed, such speculation might be premature; however, cytologic data from the brushtail possum, Trichosurus vulpecula, (family Phalangeridae) also suggests reduced female recombination, although less extreme than that in M. domestica and S. crassicaudata (HAYMAN and RODGER 1990 ). The only contrary finding comes from a cytologic study of the brush-tailed bettong, Bettongia penicillata (family Potoroidae), in which no obvious difference in sex-specific chiasma number or distribution was detected (HAYMAN et al. 1990 ). Unfortunately, linkage data are not available for T. vulpecula or B. penicillata. Thus, reduced female recombination surely does occur in three marsupial species and perhaps a fourth. Notwithstanding the data from B. penicillata, it seems probable that reduced female recombination is a very common, if not universal, marsupial characteristic.

What influences recombination rates in marsupials?
Marsupial genomes are about the same size as those of eutherian mammals (GREGORY 2001 ) but are far more conservative with regard to chromosomal compartmentalization (discussed by SAMOLLOW and GRAVES 1998 ; GRAVES and WESTERMAN 2002 ). Whereas eutherian chromosome numbers range from 2n = 6 to 2n = 118, those of marsupials range only from 2n = 10 to 2n = 32, and in >90% of marsupial species examined, 2n = 14–22. On average, then, marsupials package the same amount of DNA into fewer, larger chromosomes than do their eutherian counterparts. Low chromosome number reduces the opportunity for genome reshuffling via interchromosomal assortment, but this tendency can be compensated by an increase in intrachromosomal recombination (crossing over). That such compensation does not appear to occur in M. domestica (2n = 18), M. eugenii (2n = 16), or S. crassicaudata (2n = 14) suggests that the intergenerational reshuffling of genetic information in these marsupials is considerably less intense than that in eutherians and other vert