Comparative genome mapping of the deer mouse (Peromyscus maniculatus) reveals greater similarity to rat (Rattus norvegicus) than to the lab mouse (Mus musculus)

Background Deer mice (Peromyscus maniculatus) and congeneric species are the most common North American mammals. They represent an emerging system for the genetic analyses of the physiological and behavioral bases of habitat adaptation. Phylogenetic evidence suggests a much more ancient divergence of Peromyscus from laboratory mice (Mus) and rats (Rattus) than that separating latter two. Nevertheless, early karyotypic analyses of the three groups suggest Peromyscus to be exhibit greater similarities with Rattus than with Mus. Results Comparative linkage mapping of an estimated 35% of the deer mouse genome was done with respect to the Rattus and Mus genomes. We particularly focused on regions that span synteny breakpoint regions between the rat and mouse genomes. The linkage analysis revealed the Peromyscus genome to have a higher degree of synteny and gene order conservation with the Rattus genome. Conclusion These data suggest that: 1. the Rattus and Peromyscus genomes more closely represent ancestral Muroid and rodent genomes than that of Mus. 2. the high level of genome rearrangement observed in Muroid rodents is especially pronounced in Mus. 3. evolution of genome organization can operate independently of more commonly assayed measures of genetic change (e.g. SNP frequency).


Background
The cricetid genus Peromyscus constitutes the most abundant and speciose group of North American mammals. Though superficially similar in appearance to rats and mice, deer mice represent a more distantly related lineage. Mouse and rat are thought to have diverged from each other ~10-12 million years ago (mya) while they last shared a common ancestor with the deer mouse (P. man-iculatus) lineage ~25 mya [1]. The P. maniculatus species complex is a series of semi-interfertile populations spanning nearly every habitat on the continent and is consequently an emerging tool for the study of natural mammalian genetic variation. Facilitating such research is the existence of captive stocks derived from individual populations. Utilizing two of these stocks, we have developed a comparative genomic map for the deer mouse to further research of this genus and to provide insight into the genome rearrangements seen in rats, mice, and other mammals.
Comparative genomic analyses can reveal substantial amounts of information about the biology and evolution of species and are one of the keys to deciphering the roles that genomic structure and organization play in areas such as development, gene expression, and speciation. These analyses, however, are limited to portions of the genomes that have been mapped in all the species being compared and may be compromised by uncertainty of gene orthology and order between any two species.
Although whole-genome sequences are available for many species, proper genome annotation is difficult and typically requires additional resources (e.g. meiotic linkage and radiation hybrid cell maps) [2]. As a result, both cytogenetic methods and genetic linkage mapping are still essential tools for the analysis of genomic organization. While cytogenetic methods are effective for discerning large regions of chromosomal homology and conserved synteny, linkage maps are able to detect rearrangements of gene order within these fragments and pinpoint the locations of synteny breakpoints. Such detailed genomic comparisons require ordered linkage maps that include orthologous Type I (gene coding) loci to provide landmarks that can be identified in the genomes of multiple species. Comparative analyses using such a combined approach may reveal many more chromosomal rearrangements and novel synteny groups.
Two of the most complete mammalian genomic maps are associated with the most used biomedical models, the rat (Rattus norvegicus) and mouse (Mus musculus), which both belong to the rodent family Muridae. Rodentia is the largest mammalian order, containing > 2000 of the ~4600 recognized species and the murids constitute the majority of these [3]. Murid genomes analyzed to date not only show more rapid nucleotide mutation rates [1,4] but also higher rates of chromosomal rearrangement than other mammals. Murid chromosomal divergence rates are estimated to be one rearrangement per million years; ten times the rate for most mammalian genomes [5,6]. Furthermore, such events are punctuated over time rather than having a steady-state mutation pattern [7]. This elevated rate of rearrangement has resulted in greater karyo-typic divergence between rat and mouse than between much more distantly related species (e.g., humans vs. domestic cats) [8] and has hampered reconstructions of ancestral rodent and mammalian genomes [9][10][11][12]. As a result, interpreting the evolutionary trajectory of chromosome segments between these model organisms and humans has proven difficult [13]. An outgroup to the two mapped murid genomes that is less divergent than human could alleviate these problems by aiding in the construction of a more accurate ancestral rodent genome.
Here we describe the initial results of the first intermediate-density comparative genomic map for the deer mouse covering an estimated 35% of the deer mouse genome. These data suggest the deer mouse genomic organization more closely resembles that of rat than mouse, despite the much more recent common ancestor shared by the latter two species. Considered with cytogenetic data [14] and ancestral karyotype reconstructions [10,12], our analysis further suggests that the deer mouse and Rattus genomes have undergone fewer large scale rearrangements than Mus.

Design
We employed a standard backcross design for these studies utilizing P. maniculatus bairdii stock derived from Washtenaw Co. MI (BW) and P. polionotus stock derived from Ocala Nat'l forest in Florida (PO). While neither population is completely inbred, both originated from a limited number of founders and have been maintained as closed colonies. Thus, identifying fixed differences between the two (e.g., SNPs) was typically not difficult.
Map construction for P. maniculatus was conducted using both the Rattus and Mus maps as references [15,16] and using assays designed to span rat-mouse synteny breakpoints. We use the term breakpoint when referring to a break between linkage groups as defined by Pevzner and Tesler [13]. In all, we have genotyped 103 Type I gene markers from 18 different Rattus chromosomes on our backcross panels. Table 1 presents the complete list of markers used in our mapping study, their respective locations in the mouse and rat genomes, and the source of the primer sequences. The figures presented here though, focus on the markers around the breakpoint regions between Mus and Rattus genomes that are informative for this comparative analysis.
Our backcross panels facilitated linkage of markers at distances ranging from 1.2 cM to 35.8 cM. There are a few instances, however, where marker co-segregation occurs. These may be due to recombination "cold-spots", segmental inversions between the BW and PO strains, or sim- ply the interval distance may be below our mapping panel resolution.

Mapping of loci from Rattus Chrs 10 and 14
Our previous analysis of the Peromyscus genome using loci from Mus Chr 11 [17] indicated that there are two separate deer mouse linkage groups. ZOO-FISH data by Mlynarski et al. (submitted, BMC Evolutionary Biology) also supported this conclusion. These linkage groups correspond to Rattus Chrs 10 and 14 and the resulting chromosomal breakpoint is shared with the Rattus genome relative to the Mus genome ( Figure 1). This breakpoint is also shared in other species, including human, chimpanzee, dog, and pig [11,18]. This conserved similarity led us to consider whether the deer mouse genome might share a higher degree of chromosomal similarity with Rattus than to Mus.
To explore this possibility, we expanded the existing linkage groups using loci whose orthologs lie on Rattus Chrs 10 and 14 but are not located on Mus Chr 11 ( Figure 1). For the Rattus Chr 10 homology group, we generated markers for Btbd12 and Gbl, which correspond to regions of Mus Chr 16 and 17, respectively. Single loci for each segment were sufficient because the Mus Chrs 16 and 17 segments are small and most of Rattus Chr 10 is homologous with Mus Chr 11. We found both markers to be clearly linked to the terminal locus from Rattus Chr 10, Xbp1, with very high LOD scores (>30) indicating rat genome homology. However, all three markers co-segregated. At a distance of only 1.9 Mb in the Rattus genome, these markers are likely to be closer in the deer mouse than our panel is able to resolve.   (Figure 1). The marker intervals for this group were larger than those of the Rattus Chr 10 markers and allowed us to map 83.0 Mb (~82%) of Rattus Chr 14 in the deer mouse with only a few markers. Our results again show a greater similarity between Peromyscus genome organization and the Rattus genome than either to the Mus genome.
The high degree of Rattus genome similarity that we found for these two deer mouse linkage groups warranted exam-ination of additional informative regions where synteny breakpoints occur between the Mus and Rattus genomes. While not comprehensive for every chromosome, this strategy accelerated our examination of chromosome evolution for the deer mouse genome and helped determine the best reference genome for future deer mouse genome mapping.

Mapping the Informative Regions of Rattus Chr 1
We next chose markers that spanned Rattus Chr 1, focusing our efforts within the regions surrounding the breakpoints between Mus Chrs 7, 17, 10, and 19 ( Figure 2).   Figure  2) with LOD scores all well above the 3.0 threshold. Our results for the Mus Chr 7 region were also concordant with a previous study [19] from which we utilized two of the same markers, Usp29 and H19.

Comparison of the organization of genes on
We also found genome homology between Rattus and the deer mouse genome by markers spanning the breakpoint between the Mus Chr 7 and Chr 19 regions of Rattus Chr 1. The Mus Chr 7 marker H19 is linked to the Mus Chr 19 marker Prdx5 at a distance of 6.5 cM (LOD = 12.2). Prdx5 is also linked to a second Mus Chr 19 marker Prpf19 at a distance 4.9 cM (LOD = 13.4 cM).
Overall, our data for Rattus Chr 1 loci show that the two deer mouse linkage groups span two Mus genome breakpoints but only one Rattus genome breakpoint, which shows a continued bias towards similarity with the Rattus genome. Our results also imply that the Mus genome has been more rearranged in this region.

Breakpoint Mapping of Rattus Chr 4 and Chr 6 Loci
To broaden the scope of the deer mouse map and help reduce bias resulting from any localized phenomenon such as segmental inversions, we acquired data for markers from multiple Rattus chromosomes. Rattus Chrs 4 and 6 were priority candidates because of the simple breakpoint arrangements and well-conserved gene orders between Rattus and Mus. Rattus

Mapping Loci from Mus Chr 1/Rattus Chrs 5, 9, and 13
To avoid possible bias towards finding only Rattus genome similarity, we also selected markers to span Rattus synteny breakpoints in relation to the Mus genome. This involved markers that span approximately 89% of Mus Chr 1 but are located separately in Rattus on Chrs 5, 9, and 13 ( Figure 5). With exception of the Rattus Chr 9 marker Col9a1, we found linkage between all of the markers within their Rattus chromosome homology groups but not between them, indicating a bias toward Rattus genome similarity. At the telomeric end of Mus Chr 1, we detected linkage between the Rattus Chr 13 markers Acbd3 and Glul at a distance of 21.2 cM (LOD = 9.6) and between Glul and Bcl2 at 35.8 cM (LOD = 3.8) ( Figure 5). However, we did not detect linkage between Bcl2 and the Mus Chr 9 marker Fn1, as would be expected by Mus genome homology.
Amongst the Rattus Chr 9 markers, we found Col3a1 and Fn1 were linked at a distance of 18.8 cM (LOD = 10.3).
However, Col3a1 is surprisingly not linked to Col9a1, which represents a deviation from the Rattus genome. To confirm these results, we also tested Mut, a marker that is closely linked to Col9a1 on Rattus Chr 9 ( Figure 6). Mut is located 7.13 Mb from Col9a1 on Rattus Chr 9 but in Mus is located separately on Chr 17. Mut did not link to Col3a1 but did co-segregate with Col9a1 (LOD = 12.0), thus identifying a linkage similarity between the Rattus and deer mouse genomes. The breakpoints present in the deer mouse and Mus maps for this region are offset and may represent breakpoint area re-usage and a rearrangement hotspot [13]. However, more detailed mapping using markers located between Col9a1 and Col3a1 on Rattus Chr 9 are needed to refine the breakpoint location. We discovered additional Rattus Chr 9 similarity using the marker Lama, which represents a second and separate region of Mus Chr 17 than that of Mut ( Figure 6). We found Lama1 and Fn1 are linked in the deer mouse at a distance of 32.7 cM (LOD = 4.7).
Comparison of the organization of genes on Rattus Chr 9, and Mus Chrs 1 and 17, with the linkage of their orthologous genes in Peromyscus

Mapping Loci from Mus Chr 8/Rattus Chrs 16, 17, and 19
As another test of Peromyscus genome homology to Mus, we performed an analysis using six loci from Mus Chr 8 that form two linkage groups in the Rattus genome ( Figure  7). We selected markers in each group that are less than 20.Mb apart in the Rattus genome to facilitate deer mouse linkage detection. Additionally, the two markers that flank the breakpoint, 9130404H06Rik and Elmod2, are about 16.0 Mb apart in Mus, which is well within the range of our mapping panel.
As with the Mus Chr 1 analysis, we found linkage with highly significant LOD scores but only within the individ-  Figure 8) and Mus Chr 5 has four major regions representing four Rattus chromosomes and three very small segments representing three additional Rattus chromosomes (Figure 9). Positive linkage results for these highly rearranged chromosomes in the deer For Mus Chr 17, six markers (Tcp10b, Mas1, and Pacrg from Rattus Chr 1; Gbl from Rattus Chr 10; Lama1 from Rattus Chr 9; and Sos1 from Rattus Chr 6) were tested for linkage in all possible arrangements despite having already been assigned to other deer mouse linkage groups. No linkage was found other than that which already existed amongst the Rattus Chr 1 markers Tcp10b, Pacrg, and Mas1 (Figure 8).
We conducted a similar test for several markers from Mus Chr 5 (Figure 9). Eleven markers representing five Rattus chromosomes were tested for linkage. As in the Mus Chr 17 analysis, linkage was found only between markers located on the same Rattus chromosomes.
Two Mus Chr 13 markers, Clptml1 and Spz, also failed to show linkage despite being only about 19.6 Mb apart in the Mus genome, thus further reinforcing the linkage group disparities between the deer mouse and Mus genomes. Clptml1 is located in Rattus Chr 1 and Spz1 is on Rattus Chr 2 ( Figure 10).

Genome Mapping and Genomic Evolution
Development and availability of multiple mapping tools is essential for accurate and timely exploration of any species' genome. Three methods have already been employed in mapping small portions of the deer mouse genome in the form of cytogenetics [14,20], meiotic segregation analysis [17,[21][22][23][24][25], and a whole-genome radiation hybrid cell panel [17]. These tools are most powerful when used in combination, as exemplified by Rowe et al. [2] for Mus and by Menotti et al. [26] for the cat.
Here we significantly expand the Peromyscus meiotic segregation mapping data using two P. maniculatus × P. polionotus interspecific backcross panels and present the most comprehensive comparative linkage mapping data for the deer mouse to date using Type I gene markers. In addition to providing an important genetic tool for Peromyscus research, we tested whether the deer mouse genome displayed organizational homology to that of Mus musculus, Rattus norvegicus, or a combination of both. Our results indicate a large degree of gene order and synteny conservation by the deer mouse genome with that of Rattus.
Our analysis was done by establishing linkage over approximately 35% of the deer mouse genome using gene markers that predominantly spanned junctions of largescale genome rearrangements between the Rattus and Mus genomes. While using the Rattus genome as the reference, we tested 13 Mus genome breakpoints. Ten of the 13 breakpoints spanned by the Rattus genome were similarly linked in the deer mouse genome. In contrast, only one of 12 Rattus genome breakpoints that we examined while using the Mus genome as a reference closely coincided with any linkage breakpoints that we found in deer mouse genome. These data demonstrate that the organization of the deer mouse and Rattus genomes are more similar to each other than either is to Mus.
There are three instances, however, where the deer mouse map differs from both the Rattus and Mus maps. Two examples are located between markers Allc and 493504H06Rik ( Figure 4) and between markers Col9a1 and Col3a1 ( Figure 6). Approximately 21 Mb separates the latter pair in both Rattus and Mus, so additional markers will need to be applied to the deer mouse panel to better pinpoint the location of this breakpoint. The third instance is the unique deer mouse linkage of Sec61a2 to Pmfbp1. Collaborative efforts have also helped to inform, as well as confirm, some of these data using additional tools such as ZOO-FISH analyses using flow-sorted whole chromosome probes (Mlynarski et al., submitted BMC Evolutionary Biology). The strong organizational similarity of the deer mouse genome with the Rattus genome rather than the more morphologically similar Mus musculus suggests that a significant amount of rearrangement occurred in the Mus genome after the divergence of the cricetid and murid lineages. Concomitantly, our results suggest that the genomic organizations of Rattus and Peromyscus are more representative of the ancestral muroid genome than the Mus genome, which is in agreement with previous literature that indicated a higher rate of genome rearrangement for Mus [5,6]. Most eutherian genomes have 30 to 40 blocks of homology with the human genome while the Mus genome is extraordinary with approximately 200 homology blocks. However, the Mus genome is not unique in having higher relative rearrangement rates, as the canine and gibbon genomes have approximately twice the average number of homology blocks [27].
Our results also show that genome rearrangement can act independently from other forms of genome evolution, such as sequence mutation. Although rodent sequence mutation rates are higher compared to other mammals, such measurements of genome evolution show Mus and Rattus shared a common ancestor significantly more recently than either have with Peromyscus. Our data does not propose to change this phylogeny but rather merely highlights that DNA sequence variation and chromosome rearrangement are independent processes and greater understanding of both processes can provide different insights into the evolution of the structure and function of the eukaryotic genome.

Development of a P. maniculatus Backcross Panel
We chose interspecific backcross analysis in order to maximize genetic polymorphism and for the ease of linkage analysis [28]. The two species used in the cross were the deer mouse (P. maniculatus bairdii; BW) and the old field mouse (P. polionotus subgriseus; PO) and were obtained from the Peromyscus Genetic Stock Center at the University of South Carolina [29]. We set up the initial crosses in only one direction, BW females × PO males, to generate interspecific hybrid F1's. The direction of this cross is essential, as the reciprocal cross results in lethal overgrowth of the offspring [30].
We created two separate backcross panels, BX116 and BX2, for the linkage analysis. For the BX116 panel, we bred twelve hybrid (plt BW × PO) F1 animals with 12 unrelated plt BW animals to generate 116 backcross progeny. Backcrosses to BW can be performed using both female and male F1 hybrid animals, as both matings will give viable offspring. However, all but one of the matings used for this panel were F1 × BW (Ǩ × ǩ).
We used a similar strategy for the BX2 panel but the plt BW stock was substituted with wild-type BW stock. The plt allele originated in a different subspecies of P. maniculatus than the BW stock to which it was crossed. This additional backcross panel was designed to circumvent intraspecific SNP variation and possible recombination problems due to chromosomal inversions that are known to exist within some P. maniculatus sub-species.
To create the BX2 panel, we crossed four F1 males from separate unrelated Ǩ BW × ǩ PO matings to non-sibling BW females, which were generated from separate matings. This resulted in four unrelated backcross families. We obtained twenty-two Ǩ BW × ǩ F1 offspring from each of three of these backcross matings and 20 offspring were obtained from the fourth for a total of 86 backcross animals. These were grouped in a 96-well tray along with the eight parentals and a positive and a negative control. We employed this strategy to minimize variation within families while maximizing information between families.
Similarly, the strategy of crossing F1 hybrid males with BW females minimized variation due to gender-based differences in recombination frequency.
We extracted genomic DNA from all backcross parents and progeny for the BX116 panel from 1.0 cm tail snips with the Qiagen DNEasy Tissue Kit using the manufacturers protocol (Qiagen, Inc.). Genomic DNA for the BX2 panel animals was extracted using a phenol/chloroform/ isoamyl alcohol extraction method to increase yields.

Marker Development
We obtained primer sequences for some Type I markers from published sets of orthologous gene markers. These are termed UMPS, CATS, and TOASTs [31][32][33] We designed all the markers to be ~400 bp-1500 bp and to span an intron to increase polymorphism detection. This was done by aligning deer mouse DNA sequences to Mus musculus genomic sequences using cross-species megaBLAST (NCBI). Some critical markers however, spanned larger introns and resulted in amplicons slightly larger than the ideal size parameters.  [37], or with the SNP-RFLPing program [38]. Candidate enzymes were chosen from those predicted by the software. RFLP tests for each marker and enzyme were conducted according to manufacturer's protocols on 10 μL PCR products from a template test panel consisting of DNA from BW, PO, a BW/PO mix (equal ratio), and a negative using only TE. All RFLP products were analyzed by gel electrophoresis. Any essential markers that could not be genotyped by either size polymorphism or RFLP were sequenced on all the backcross animals and their parents.

PCR Typing of the Backcross Panel
We tested the backcross panel parental mice with each marker prior to use on the backcross panel. We typed then typed the markers on all possible backcross animals whose parents exhibited the expected genotype. On the BX116 panel, no fewer than 73 animals were used for data to establish linkage between any two markers with exception of Gbl to Hba and Hba to Canx, of which both only used 39 animals due to the small usable data set for Hba. For the BX2 panel, no fewer than 60 animals were typed between any two markers with the exception of Mut and Col9a1, for which only 40 animals could be genotyped in common.

Data Analysis
We performed backcross linkage analysis using Map Manager QTX software [39]. A minimum LOD score of 3 was used to establish linkage. Ordering of markers typed on the backcross data was determined by subjecting the data to the "ripple test", which evaluated local permutations and selected the optimal order based on minimum breakage. Once linkage and gene order was established with high degree of confidence, omitted or unavailable genotypes could sometimes be inferred by Map Manager, as double crossovers between closely linked markers are rare. The procedure of inferring genotypes did not change any gene orders but typically tightened linkages slightly and raised LOD scores.

Authors' contributions
CR was the lead researcher, conceived the study, participated in its design and coordination, developed markers and assays, participated in the design of the backcross panel, performed linkage analysis, performed molecular genetic experiments, performed sequence alignments, and drafted the manuscript. AL developed markers and assays, performed molecular genetic experiments, and performed sequence alignments. JG developed markers and assays, participated in the design of the backcross panels, performed molecular genetic experiments, performed sequence alignments, and edited the manuscript. PV developed markers and assays, performed linkage analyses, participated in the design of the study, and contributed to and edited the manuscript. RO participated in the design and coordination of the study and helped to draft the manuscript. MD was the principle investigator on the project, participated in its design and coordination, performed sequence alignments, participated in the design of the backcross panels, and contributed in drafting the manuscript and figures. All authors read and approved the final manuscript.