Using genomic resources for linkage analysis in Peromyscus with an application for characterizing Dominant Spot

Background Peromyscus are the most common mammalian species in North America and are widely used in both laboratory and field studies. The deer mouse, P. maniculatus and the old-field mouse, P. polionotus, are closely related and can generate viable and fertile hybrid offspring. The ability to generate hybrid offspring, coupled with developing genomic resources, enables researchers to conduct linkage analysis studies to identify genomic loci associated with specific traits. Results We used available genomic data to identify DNA polymorphisms between P. maniculatus and P. polionotus and used the polymorphic data to identify the range of genetic complexity that underlies physiological and behavioral differences between the species, including cholesterol metabolism and genes associated with autism. In addition, we used the polymorphic data to conduct a candidate gene linkage analysis for the Dominant spot trait and determined that Dominant spot is linked to a region of chromosome 20 that contains a strong candidate gene, Sox10. During the linkage analysis, we found that the spot size varied quantitively in affected Peromyscus based on genetic background. Conclusions The expanding genomic resources for Peromyscus facilitate their use in linkage analysis studies, enabling the identification of loci associated with specific traits. More specifically, we have linked a coat color spotting phenotype, Dominant spot, with Sox10, a member the neural crest gene regulatory network, and that there are likely two genetic modifiers that interact with Dominant spot. These results establish Peromyscus as a model system for identifying new alleles of the neural crest gene regulatory network.


Background
Peromyscus are the most common and widely distributed mammalian species in North America and includes the deer mouse (P. maniculatus), the old-field mouse (P. polionotus), and the white-footed mouse (P. leucopos) [1]. Peromyscus have similar characteristics to old world mice, including the house mouse, Mus musculus, which is the most widely used model system for biomedical research. However, Peromyscus and Mus are more distantly related than their appearance suggests, sharing a last common ancestor 25 million years ago [2]. Peromyscus are more closely related to hamsters than to Mus.
Mus is an exceptional model system for biomedical research, having a wide range of genetic resources, including a fully sequenced and annotated genome, various inbred strains, and the ability to make specific genetic modifications. However, in developing Mus laboratory strains, human selection resulted in a genome with contributions from three different Mus species, and the process of inbreeding resulted in a lack of genetic diversity [3]. As a result, genome wide association studies (GWAS) using Mus typically produce linkage to genomic intervals that are too large to facilitate candidate gene identification [4]. Furthermore, the lack of genetic diversity has reduced the pool of traits that could be identified and eventually characterized in laboratory mice. Recognizing the limitations of inbred Mus lines for GWAS and quantitative trait locus (QTL) mapping studies, the Collaborative Cross was initiated to generate genomic variation by combining eight inbred lines and three wild subspecies of Mus [5,6]. By intercrossing the Collaborative Cross strains, a Diversity Outbred stock was generated to reduce the linkage disequilibrium blocks characteristic of the founder inbred lines [6,7]. The Diversity Outbred stock has been used to map numerous quantitative traits observed in the founding inbred lines, including serum cholesterol levels and heart size [6,8,9].
Another option is to use the variation observed in existing outbred stocks to generate smaller linked intervals that facilitate gene discovery. For instance, a GWAS study conducted on outbred Mus stocks for variability in high density lipoprotein and hypercholesterolemia produced a linked interval of less than 100 kb, which lead to the identification of allelic differences in Apoa2 associated with hypercholesterolemia [10]. The key to generating small linked intervals is the frequency of genetic variation. In inbred strains, the lack of genetic variation generates long syntenic regions that result in large blocks of linkage disequilibrium. Peromyscus bred in captivity are susceptible to inbreeding depression; therefore, the Peromyscus Genetic Stock Center (PGSC) at the University of South Carolina (UofSC) maintains outbred stocks of P. maniculatus (BW), P. polionotus (PO), and P. leucopus (LL). These stocks are maintained by breeding individuals that do not have a common grandparent [11]. Because these outbred stocks have not undergone intentional selection, they maintain both genetic and phenotypic diversity, similar to natural populations [1,12]. These genetically diverse laboratory stocks provide another animal model with physiological and behavior variation that differs from the Diversity Outbred stocks and other outbred Mus stocks, providing an opportunity to study how genetic variation may predispose individuals to disease [13,14]. Additional genetic variation can be generated in Peromyscus because P. maniculatus and P. polionotus can produce viable and fertile hybrid offspring [1].
P. maniculatus and P. polionotus are species with distinct behavioral and metabolic differences [1]. The ability to generate hybrids between two Peromsycus outbred stocks enables researchers to use QTL analysis to identify genetic loci that underlie the quantitative traits, behavior, or metabolic differences between the two species. For instance, P. polionotus build better nests than P. maniculatus [15]. A QTL associated with nest building is located on chromosome 4, and the vasopressin gene is located within the QTL. There are no non-synonymous changes in the coding region of vasopressin between the two species; however, P. maniculatus express 2.8-fold more vasopressin mRNA in the hypothalamus than P. polionotus. This species-specific expression difference results in the species-specific nest building behaviors, as increasing vasopressin in P. polionotus inhibits nest building behavior [15]. Other QTL analyses utilizing P. maniculatus and P. polionotus have identified genetic loci associated with coat color differences and burrowing behavior [16,17].
There are many behavioral differences between P. maniculatus and P. polionotus that are associated with human health. For instance, The BW stock of P. maniculatus has a high incidence of repetitive behavior, while the PO stock of P. polionotus has no repetitive behavior [18,19]. In humans, repetitive behavior is observed in people with autism and obsessive-compulsive disorder [20]. A QTL analysis for repetitive behavior using P. maniculatus and P. polionotus hybrids could identify the genetic loci associated with this behavior and provide an animal model for understanding complex behaviors associated with human disease.
The genomic tools that facilitate QTL analysis in Peromyscus are actively being developed. The first genomic assembly containing almost 31,000 scaffolds of P. maniculatus was released in 2013 (GenBank assembly GCA_ 000500345.1). Raw whole genome sequencing reads for P. polionotus were deposited in the Sequencing Read Archive (SRA) in July 2015 (SRX179420, SRX179421, and SRX179422). An improved, chromosome level assembly for P. maniculatus followed in November of 2018 (GenBank assembly GCA_003704035.1), along with chromosome level assemblies of P. polionotus (Gen-Bank assembly GCA_003704135.2) and P. leucopus (GenBank assembly GCA_004664715.1), in April of 2019 [21]. These resources facilitate the design of linkage mapping studies by providing species specific allelic information across the genome. Restriction site-associated DNA sequencing, RADseq, and its variant double digest RADseq, ddRADseq, facilitate linkage mapping studies by providing high throughput genome wide allelic information, and have been instrumental in several Peromyscus QTL studies [15,16,22,23]. While a reference genome is not necessary for RADseq experiments, it facilitates these experiments by providing a scaffold on which to map allelic information.
The continued development of Peromyscus genomic resources is required to facilitate linkage analysis studies in Peromyscus. We performed a comparative analysis of the P. maniculatus and P. polionotus genomes, using available genomic datasets, to identify polymorphisms between the two species that can be used for linkage analysis. We used the polymorphic data to design a candidate gene linkage mapping study to identify a genetic locus associated with the Dominant spot trait in P. maniculatus. This information was also used to identify genetic differences associated with known physiological and behavioral differences between the two species.

Results
Analysis of genomic variations between P. maniculatus and P. polionotus The ability to generate fertile and viable hybrid offspring from P. maniculatus and P. polionotus provides an opportunity to identify the genetic loci associated with phenotypic traits that differ between the two species. Capitalizing on this rich genetic resource requires identifying the allelic differences between the two species. We sought to identify allelic differences between P. maniculatus and P. polionotus, using publicly available genomic resources. The first assembly of the P. maniculatus genome was used as the reference genome, and the P. polionotus sequencing reads deposited in the SRA database were mapped against the P. maniculatus reference genome. For polymorphism determination we required 10 independent P. polionotus sequencing reads to be mapped to a specific P. maniculatus genomic location (data file available at https://osf.io/4eypx/). With these criteria we determined that 81.9% of the P. maniculatus genome was covered by 10 P. polionotus sequencing reads, resulting in 38,166,334 polymorphisms between P. maniculatus and P. polionotus. Among these variations are 34,084,607 single nucleotide polymorphisms (SNPs) and 4,081,727 insertions or deletions (INDELs) between BW and PO (Table 1). Lowering the read mapping requirement to five sequencing reads increased the number of identified polymorphisms to 40,815,360. However, we chose to keep the more stringent criteria for subsequent analysis, recognizing that we are likely underestimating the number of polymorphisms between the species. Using the more stringent 38,166,334 polymorphisms results in a variant rate of one variation every 68 base pairs. We used SnpEff to annotate the potential functional effects of the polymorphisms (data file available at https://osf.io/4eypx/), recognizing that individual variant locations can have multiple annotations, which increases the number of annotations above the total number of variants. More than 23 million variants occur in intergenic regions, with more than 7 million occurring within 5 kb upstream or downstream of an annotated gene, potential impacting transcriptional regulatory sites (Table 1). More than 45 million variants occur in an annotated transcript region, including > 769,000 in annotated exons and > 44 million in annotated introns (Table 1). Within exons, the majority occur in 3′ untranslated regions (UTR).
We characterized the SNPs that occur between P. maniculatus and P. polionotus for changes in predicted protein coding sequences and found that there are 11, 588 genes that contain a coding sequence SNP. Of these variants, 70.2% are silent, 29.6% are missense, and 0.2% are nonsense variants. We identified a set of 10,405 genes that contain a nonsynonymous change, which might result in a functional difference in the protein between the two species. Using the annotated Mus genome as a reference, we conducted a gene ontology (GO) term analysis on the list of 10,405 genes with missense or nonsense variants to determine if any biological processes or molecular functions are over or underrepresented within this list of genes. There are 2060 biological processes that are over-represented, and 63 that are under-represented (Table 2 and Additional Files 1 and 2). The processes with the most statistically significant enrichment are typically broad in GO terminology, including cellular process, cellular metabolic process, and metabolic process. P. maniculatus and P. polionotus are known to have distinct physiological differences, including an almost four-fold increase in blood cholesterol levels in P. polionotus compared to P. maniculatus and a two-fold increase in blood triglyceride levels in P. polionotus over P. maniculatus [24]. We searched the list of overrepresented GO terms for terms with a possible relationship with cholesterol and triglycerides and found that the GO terms cholesterol metabolic process,  cholesterol homeostasis, cholesterol transport, triglyceride metabolic process, triglyceride biosynthetic process, and triglyceride catabolic process are enriched in this list of genes with potential functional protein changes (p = 3.90 × 10 − 6 , 3.05 × 10 − 5 , 3.86 X10 − 4 , 9.86 × 10 − 6 , and 1.81 × 10 − 5 , respectively) (Additional File 1). Each of these GO terms contains from 20 to 77 different genes with nonsynonymous changes, suggesting that there is substantial genetic complexity that may underlie the metabolic differences between the two species. In contrast, there are 63 GO terms that are under-represented in the list, including sensory perception of smell, sensory perception of chemical stimulus, sensory perception, G protein-coupled receptor signaling pathway, and nervous system process, suggesting that these biological processes are more conserved between the two species. The P. maniculatus laboratory stock, BW, is known to have a significant incidence of repetitive, or stereotactic, behavior, including repetitive jumping [18,25,26]. The P. polionotus laboratory stock, PO, does not display stereotactic behaviors. In humans, repetitive movements are associated with autism, obsessive-compulsive disorder, and other neurologic disorders, including tics and schizophrenia [20,27]. BW animals have been used to study how neuroactive drugs and environmental enrichment can reduce repetitive behavior [28,29]. BW animals are also less social than PO animals, another hallmark of autism in humans [18,30]. Our expectation is that there are genetic loci that underlie the stereotactic and social behaviors and we sought to determine if there is genetic variation between BW and PO that may be associated with these behavioral differences. We examined the list of overrepresented GO terms for processes that may be related with autism associated behaviors and found that locomotory behavior and social behavior are both enriched GO terms (p = 1.94 × 10 − 5 and 6.79 × 10 − 4 , respectively) (Additional File 1). In addition, we selected a list of candidate genes that all have a high confidence of being associated with autism in humans from the Simons Foundation Autism Research Initiative (SFARI) database (Table 3). We then identified sequence variations that occur between P. maniculatus and P. polionotus in this list of autism candidate genes. Each gene analyzed has multiple sequence variations between the two species that could result in a functional change to the protein, including missense changes, nonsense changes, in-frame deletions, and nucleotide variations in splicing regions. In addition, there are numerous differences in untranslated regions, introns, and upstream and downstream sequences that could result in differences in transcript or protein levels, if they occur in regulatory regions.
We wanted to further explore how polymorphisms between BW and PO might result in functional changes by examining polymorphisms in one autism candidate gene, recognizing that there are no identified connections between there autism candidate genes and the behavioral differences between BW and PO. ASH1L is a chromatin modifying protein that is associated with transcriptional activation [31]. De novo mutations in ASH1L have been identified in multiple people with autism, while rarely occurring in controls [32][33][34]. We aligned select mammalian ASH1L protein sequences to identify highly conserved amino acids and determine if Peromyscus amino acid substitutions occur in these conserved residues (Fig. 1). None of the nonsynonymous changes between P. maniculatus and P. polionotus ASH1L are in the ASH1L conserved protein domains, SET, BROMO, PHD, and BAH, and generally occur in regions with less conservation between mammalian ASH1L proteins. However, at positions 61, 484, 770, 1632, and 2814 there are amino acid substitutions in one Peromyscus species where the amino acid is conserved in the other mammals. The S484A, S770P, and T1632P substitutions are intriguing as they remove potential phosphorylation sites in one of the Peromyscus species. The potential functional impact of these amino acid substitutions on ASH1L function will require further characterization.
We also examined possible transcriptional regulatory changes between P. maniculatus and P. polionotus for ASH1L by generating a VISTA plot to identify conserved non-coding sequences (CNS) in the ASH1L locus [35]. A conserved non-coding sequence occurs in intron 3 of ASH1L in 100 vertebrate species (UCSC Genome Brower: Human GRCh38/hg38 chromosome 1: 155,459,751-155,478,012) and is also found in BW and PO (Fig. 2) [36]. Within this CNS there are three SNPs between P. maniculatus and P. polionotus. Two of the three Peromyscus SNPs are in positions that are not conserved between a group of seven mammalian species. However, one SNP occurs in a region of 16 nucleotides that are completely conserved within the selection of mammalian species (Fig. 2c). We used PROMO to identify potential transcription factor binding sites within this region and found that in six mammalian species, including P. maniculatus, the conserved sequences contain a potential NKX2-1 binding site [37,38]. However, in P. polionotus the SNP removes the NKX2-1 binding site and generates a potential EBF1 binding site.

Restriction enzyme recognition sites in P. maniculatus
A QTL analysis or GWAS using Peromyscus is likely to utilize RADseq. RADseq is a flexible approach to genomic analysis, as the choice of restriction enzyme used to digest the genomic DNA can be varied to customize the number of sequenced sites, known as RAD markers, across the genome [39]. The number of RAD markers generated is twice the number of restriction enzyme recognition sites. An enzyme that cuts more frequently will generate more RAD markers and, therefore, provide more allelic information than an enzyme that cuts less frequently. We used the P. maniculatus reference genome to determine the number of cuts sites and the average fragment size for the enzymes listed in Table 4. This data provides a range of restriction enzymes with recognition sites from approximately 1000 bp apart (DraI) to approximately 1 million bp apart (AscI), enabling an informed choice for restriction enzyme selection in P. maniculatus RADseq projects. RADseq generates about 400 bp of sequence information flanking a restriction enzyme recognition site. Because a sequence variant between P. maniculatus and P. polionotus occurs approximately every 68 base pairs, it is likely that RADseq analysis on F 1 hybrids will generate informative allelic information at most RAD markers.

Linkage analysis of dominant spot
Dominant spot is a spontaneous mutation that arose within a wild population of P. maniculatus near Morrison, Illinois [1,40]. The Dominant spot trait (S) is maintained as heterozygotes on the BW laboratory stock of P. maniculatus at the PGSC. PGSC breeding records suggest that S/S homozygotes are likely embryonic lethal. We crossed BW S/+ adults and generated timed pregnancies and observed resorbing embryos at embryonic day of development 14.5 (e14.5), and approximately one quarter of e13.5 embryos have a variable phenotype that includes morphological defects consistent with embryonic lethality (Fig. 3).
We sought to perform linkage analysis to identify genetic loci associated with Dominant spot by crossing BW S/+ (Fig. 4a) with the PO laboratory stock of P. polionotus (+/+). F 1 hybrids of P. maniculatus and P. polionotus exhibit developmental dysgenesis [1,41]. When female P. maniculatus are crossed with male P. polionotus, the hybrid offspring are smaller than either parent, but are viable and fertile. In contrast, female P. polionotus crossed with male P. maniculatus result in overgrown fetuses with developmental defects and are not viable. Therefore, a male +/+ PO was crossed with S/+ BW to generate six F 1 hybrids with forehead spots (Fig. 4b).
Disrupted pigmentation patterns in laboratory mice, Mus musculus, are readily identifiable, and characterization of the causative mutations for these spotting defects has identified key members of a neural crest gene regulatory network necessary for normal neural crest development [42]. We pursued a candidate gene approach as a first step towards linking Dominant spot with a specific genomic region. Edn3, Ednrb, Kit, Kitl, Mitf, Pax3, Ret, Snail, and Sox10 are all known to cause spotting phenotypes in M. musculus; therefore, we sought to identify allelic differences in P. maniculatus and P. polionotus for each gene to determine if any of these candidate genes are linked with Dominant spot. For our list of candidate genes, we identified a sequence variant that removes a restriction enzyme recognition site in one Peromyscus species. We will call these sites restriction fragment length polymorphisms (RFLPs) because of their similarity to the technique used for genomic variation analysis (Table 5). We then designed polymerase chain reaction (PCR) primers flanking the site to generate an RFLP site specific amplicon. BW and PO genomic DNA, along with genomic DNA from S/+ N 2 animals was PCR amplified and then digested with the appropriate restriction enzyme (Fig. 5 and Additional File 3). The S mutation arose in P. maniculatus and has been maintained on the BW stock; therefore, the S mutation occurs in a BW allele. If a candidate gene is linked with the Dominant spot trait, then all S/+ N 2 animals will have both a BW allele and a PO allele for that candidate gene. If an S/+ N 2 animal has only PO alleles for a candidate gene, then that candidate gene is not linked with the Dominant spot trait. From our list of candidate genes, eight of the candidate genes are not linked with Dominant spot as there are multiple S/+ N 2 individuals Dominant spot (Additional File 4). Among the 53 genes contained in the linkage interval, only Sox10 has a defined role in neural crest development. Therefore, we favor the possibility that the S mutation disrupts Sox10 function. We have sequenced the Sox10 exons, exon/intron junctions, promoter, and several conserved enhancer regions but have not identified a sequence variation that disrupts Sox10 function. We are expanding the sequencing analysis to include the entire linked region.
In generating the N 2 generation, we noticed that the spot size was smaller on the F 1 and N 2 animals compared to the originating BW background. The average spot size for 25 S/+ BW animals is 77.6 ± 36.6 mm 2 . We crossed one BW S/+ with PO +/+ and generated six F 1 S/+ animals, which had very small spots in comparison, 3.75 ± 1.56 mm 2 , suggesting that PO alleles have a dominant effect on the S/+ spot size phenotype. The six F 1 S/+ were backcrossed with PO +/+ and in the PO N 2 generation the spot size for 46 affected animals averaged 14.5 ± 13 mm 2 , which is significantly smaller than the spot size of S/+ BW animals (Welch's t [27] = 8.34, p = 5.38 × 10 − 9 ), suggesting that genetic background has a significant impact on the S/+ phenotype. A histogram for spot size for S/+ on BW and PO N 2 illustrates the quantitative nature of the phenotype and the shift in spot size in the PO N 2 animals (Fig. 6). We used the backcross data to estimate the number of loci that affect spot size. The six F 1 animals with small spots produced 46 offspring with spots, of which 17 (37%) resembled the F 1 parent. In this backcross experiment there are only two possible genotypes for any gene. If one unlinked locus determines the spot size phenotype, then there are two possible genotypes that interact with Dominant spot and 50% of the offspring are expected to resemble the F 1 parent. If two unlinked loci determine the spot size phenotype, then 25% should resemble the parent, and if three unlinked loci are involved then 12.5% should resemble the parent. Using a X 2 analysis, we can reject a model for three interacting loci (X 2 (1, N = 46) = 23.19, p = 1.47 X 10 − 6 ), but not models for one interacting locus (X 2 = (1, N = 46) = 3.13, p = 0.077) or two interacting loci (X 2 (1, N = 46) = 2.82, p = 0.093). These data suggest that there are one to two modifiers that cause the observed variability in the Dominant spot trait.
Discriminating between these two possibilities will require a larger sample size. Further analysis indicates that there is a significant loss of affected animals in the N 2 backcross. A total of 125 offspring were produced in the N 2 generation. Of these animals, 46 pups had forehead spots, representing the S/+ genotype, while 63 are expected (X 2 (1, N = 125) = 9.25, p = 0.0024). Analysis of the PGSC breeding records for BW S/+ X BW +/+ indicate that the S/+ genotype is produced at the expected frequency (310 total offspring of which 165 have spots, (χ2 (1, N = 310) = 1.29, p = 0.256)). These results suggest that there is a significant loss of the S/+ phenotype in the PO N 2 offspring, resulting from either a lethality of S/+ in PO N 2 animals or  Fig. 4 Variation in spot size for the Dominant spot trait. a Large spot size on an S/+ P. maniculatus. b Small spot size on an S/+ F 1 hybrid. c Intermediate spot size on an S/+ PO N 2 . Images were acquired by the authors because some PO N 2 S/+ animals have a phenotypic rescue and do not have a forehead spot. In the backcross, PO N 2 animals without spots should have two PO alleles for Sox10, barring a rare recombination event that separates the Sox10 RFLP site from the unknown causative mutation. We genotyped 50 PO N 2 individuals without spots at the Sox10 RFLP site and found that 21 are BW/ PO, supporting the hypothesis that a modifier can phenotypically rescue the S/+ genotype. We are currently conducting a QTL analysis to identify loci associated with the spot size phenotype.

Discussion
In 2001 Dewey and Dawson described Peromyscus as the "Drosophila of North American Mammalogy" [43]. At the time, the nearly completed Mus and human genome projects suggested that less traditional model systems would benefit from the application of the rich genetic resources provided by the completed Mus and human genomes. They believed that the increase in genomic resources would enable Peromyscus geneticists to identify the loci associated with specific behaviors and physiological traits associated with specific species or subspecies of Peromyscus. They were correct. The ability to generate fertile hybrids of P. maniculatus and P. polionotus coupled with QTL analysis has led to the identification of many loci associated with specific behaviors, such as burrow building and paternal parenting [15,16]. The recent generation of genomic assemblies for P. maniculatus, P. polionotus, and P. leucopus should only quicken the pace of discovery [21].
To facilitate the use of Peromyscus for gene discovery, we have used available genomic resources to identify polymorphisms between P. maniculatus and P. polionotus. Our conservative estimate demonstrates that these closely related species have a polymorphism approximately every 68 base pairs. This genetic diversity supports the use these species for QTL analysis, as the sequence variation prevents large blocks of linkage disequilibrium, a characteristic that has been problematic for QTL analysis in Mus [4]. Most of the sequence variation occurs in intergenic or intronic regions; however, more than 10,000 genes have a polymorphism that results in a missense or nonsense variation between the two species, which could result in a functional difference between the P. maniculatus and P. polionotus alleles. GO term analysis on this list of proteins results in over 2000 GO terms that are overrepresented. This list is not necessarily informative based on the descriptive nature of GO terms at the top of the list, such as cellular or metabolic process, but it is more meaningful when used in combination with known physiological or behavioral differences between the species. P. polionotus have a 2fold increase in blood triglycerides compared to P. maniculatus [24], and there are 62 genes associated with GO terms related to triglyceride metabolism. This selection of triglyceride associated genes provides a potential candidate gene list for identifying genetic variations associated with varying triglyceride levels. All these genes have human homologues, suggesting that variations identified in Peromyscus may inform human genetic variability associated with triglyceride levels.

* Polymorphism and primers identified by and verified in our analysis
The list of triglycerides associated candidate genes is a conservative estimate based on the possibility of protein functional changes. The number of genes is likely much higher, as changes in regulatory regions that affect transcriptional levels are also possible. We demonstrated the potential for polymorphisms affecting transcription by analyzing the CNS for ASH1L, an autism candidate gene [34]. We show that non-coding polymorphisms change predicted transcription factor binding sites for NKX2-1 and EBF1 in the ASH1L CNS. Mutations in NKX2-1 cause benign hereditary chorea and have recently been associated with autism [44]. Nkx2-1 is expressed in the medial ganglionic eminence (MGE) and is critical for the production of inhibitory gamma amino butyric acid (GABA)eric cortical interneurons [45]. Reductions in these inhibitory neurons are associated with autism [46]. Common genetic variations in EBF1 have also been associated with premature birth and autism [47]. Ebf1 is expressed in the lateral ganglionic eminence (LGE), which generates striatal projection neurons, including medium spiny neurons that are preferential lost in Huntington's disease [48,49]. Ebf1 is essential for striatum formation and the generation of direct striatal projection neurons that project to the substantia nigra [50,51]. Functional experiments are necessary to determine the potential impact of this and other non-coding polymorphisms on ASH1L transcription. Our comparative approach for both protein and DNA conservation demonstrates that the variations between BW and PO provides a rich source for functional analysis. Future QTL analysis between P. maniculatus and P. polionotus may help to narrow these candidate gene lists to identify genes that underlie the phenotypic difference between these species. Doing so will provide an animal model that will correlate genotypic variability with autism like behaviors. In addition, Peromyscus is an excellent model system for testing of future therapeutics, as the outbred nature of the stock will better model the variability of the human population than inbred Mus lines.
We combined linkage analysis, using P. maniculatus and P. polionotus, the polymorphisms between the two species, and a candidate gene approach to link the Dominant spot trait with Sox10. Mutations in Sox10 are known to cause belly spotting and megacolon in Mus and Waardenburg syndrome, types 2E and 4C in humans [52][53][54]. Waardenburg syndrome is genetically heterogeneous and results from mutations in Sox10, Pax3, Mitf, Snai2, Ednrb, and Edn3, as well as many cases with unidentified mutations [55]. The phenotypic presentation of Waardenburg syndrome is variable, even in families with the same mutation, suggesting that there are genetic modifiers that affect the severity of the phenotype [56][57][58]. We see similar phenotypic variability in Dominant spot. P. maniculatus with the Dominant Fig. 5 Linkage analysis for spotting mutation candidate genes. Genomic DNA from BW and PO control animals, along with a selection of S/+ F 1 animals, was PCR amplified for a set of candidate genes (Edn3, Ednrb, Kit, Kitl, Mitf, Pax3, Ret, Snai2, and Sox10), known to cause pigmentation defects in Mus and humans. Each amplicon contained a restriction enzyme recognition site polymorphism between BW and PO (see Table 5). PCR amplicons were digested with the appropriate restriction enzyme and analyzed by agarose gel electrophoresis. Sox10 is the only candidate gene were all S/+ N 2 animals have both a PO and BW allele, demonstrating linkage with the Dominant spot phenotype. Each image was cropped to remove excessive space spot trait have a variable forehead spot size, tending towards larger spots. Dominant spot on the PO N 2 background have significantly smaller spots. We believe that there are likely one to two genetic modifiers that interact with Dominant spot to cause the variability in spot size. Intriguingly, the Dominant spot trait is underrepresented on the PO N 2 background, and likely results from a gene interaction that rescues the Dominant spot phenotype. A QTL analysis to identify loci associated with spot size variability is underway and may identify novel members of the neural crest developmental gene regulatory network (GRN) or novel alleles of known members of the GRN, which could expand our understanding of neural crest and melanocyte development. In addition to Dominant spot, the PGSC maintains the Variable white and Tan streak stocks, which also have pigmentation defects that are likely caused by mutations in the neural crest GRN [59,60]. The improved genomic resources for Peromyscus combined with the three stocks of P. maniculatus with pigmentation defects and the inherent variability found in these outbred stocks makes Peromyscus an excellent model system for studying neural crest biology.

Conclusions
The availability of genomic resources for P. maniculatus and P. polionotus facilitates the use of these rodent species to identify genomic loci associated with quantitative traits. We have conducted a linkage analysis for the Dominant spot trait, which arose in P. maniculatus by crossing it with P. polionotus. Using available genomic information, we identified RFLP polymorphisms between the two species in a set of candidate genes associated with pigmentation disruptions in Mus and humans and determined that a region of chromosome 20 containing Sox10 is linked with the Dominant spot trait. Further analysis of the available genomic data demonstrates that there is significant genetic complexity between P. maniculatus and P. polionotus that underlies both physiological and behavior differences, including blood chemistry and stereotypic behaviors. QTL analysis for these traits will provide correlated loci that will aid in the identification of functional polymorphisms that generate phenotypic differences.

Polymorphism analysis
The first genomic assembly of the P. maniculatus genome, National Center for Biotechnology Information (NCBI) accession number GCF_000500345.1, was locally downloaded and used as the reference genome. Whole genome sequence reads, NCBI accession numbers SRX179420, SRX179421, and SRX179422, for P. polionotus were downloaded from the SRA database. After these sequences were converted into FASTQ format using the SRA Tool Kit (v. 2.1.16 centOS Linux 64-bit), the sequence reads, SRX179420 and SRX179422, which contained paired-end sequences reads, were split into two files, using Python script. The forward and reverse paired-end reads were linked to each other and the indexed adapter sequences were trimmed, using the Solexa QA package v. 1.13 [61]. End sequences were trimmed when the Phred quality score dropped below Q = 20. Trimmed PO sequence reads were aligned to the reference genome sequence, using the Burrows-Wheeler Aligner (BWA) program (v. 0.6.1-r104) [62]. The BWA default values for mapping were used, except that seed length (−l) was set to 28 and maximum differences in the seed (−k) equaled 1. Following alignment to the reference genome, data from each P. polionotus dataset was merged into one file. Sequence polymorphisms between the aligned P. polionotus sequence reads and the P. maniculatus reference genome were identified using the pileup function in SAMtools utilities (v. 0.1.16) [63]. The minimum read depth was set to 10 and the consensus sequence was generated with filter command "-uf". SnpEff was utilized, using default parameters, to identify possible functional consequences of the SNP/Indel variant list produced from BCFtools. Data files for the sequence polymorphisms and annotations are available as .vcf files at https://osf.io/4eypx/. GO analysis was performed on the list of genes containing nonsynonymous changes, as determined above. This list was compared to the Mus GO term database using the R packages Gostats [64] to identify highly statistically significant over and underrepresented GO terms with a p < 0.001. ASH1L protein alignments were performed using CLC Main Workbench 8 (Qiagen). Protein sequences for human (NCBI accession number NP_060959.2), dog (NCBI accession number XP_537251.2), cow (NCBI accession number NP_001179672.1), Mus (NCBI accession number NP_619620.3), and rat (NCBI accession number NP_001101159.1) were imported into CLC Main Workbench from NCBI. Predicted BW ASH1L protein sequence was generated from the P. maniculatus reference genome (NCBI accession number GCF_000500345.1 and GeneID 102,924,929) and the P. polionotus protein sequence was generated from the P. maniculatus sequence with predicted nonsynonymous substitutions based on our polymorphism analysis.
For restriction enzyme recognition site frequency analysis, the number of restriction enzyme recognition sites and average fragment length for each enzyme was calculated using the R package DECIPHER v2.0 using the P. maniculatus genome assembly (NCBI accession number GCF_000500345.1) [66].

Peromyscus
All animal research was approved by the UofSC Institutional Animal Care and Use Committee. All animals used in this study were euthanized by CO 2 asphyxiation, which is an American Veterinary Medical Association approved methodology for euthanasia for small rodents. Dominant spot P. maniculatus (S/+) on the BW laboratory stock and PO laboratory stock P. polionotus were obtained from the PGSC, https://www.pgsc.cas.sc.edu, and housed in solid-bottom opaque plastic cages with a wire bar lid, which serves as a food hopper and water bottle holder, and a filter top. Animals were group housed by sex, with up to six animals per cage, and provided paper-product bedding and nesting material. Peromyscus were housed with a 16 to 8-h light to dark cycle and feed food and water ad libitum. Matings were between one male and one female and pups were weaned at postnatal day 24. To generate timed pregnancies, Dominant spot (S/+) BW females and males were paired together. The following morning phosphate buffered saline, pH 7.3 (PBS) was used for a vaginal lavage on the females, and the wash checked for the presence of sperm. The presence of sperm was used to indicate pregnancy and noon of the day the sperm was detected was designated as e0.5. Embryos were collected on e13.5 and e14.5, photographed, and fixed in 4% paraformaldehyde in PBS overnight at 4 C. After fixation, embryos were washed in PBS and dehydrated into 70% ethanol for storage at − 20 C.
Using Mus linkage analysis as a guide, genotyping 50 mice can demonstrate linkage within 2 centiMorgans with 95% confidence [67]. Therefore, we set a goal of generating approximately 50 Dominant spot animals in a backcross experiment. One S/+ BW female was mated with a +/+ PO male to generate an F 1 generation. Six F 1 S/+ offspring with forehead spots (the Dominant spot trait) were then backcrossed to +/+ PO animals to generate a PO N 2 generation with 125 individuals. Both male and female N 2 offspring were euthanized at weaning, photographed to document the spotting phenotype, and tail snips taken for genotyping. Within the N 2 generation 46 animals had the Dominant spot trait and 79 were wild type. Photographs of forehead spots of 25 Dominant spot BW and 46 Dominant spot PO N 2 were analyzed to determine the area of white fur, using the Fiji distribution of ImageJ [68]. The total number of animals used for all experiments, including the three generations of the backcross experiment and the timed pregnancies was 143.
PCR genotyping was performed for a set of neural crest candidate genes, using primers listed in Table 5, in a 25 μl reaction with a 63°C annealing temperature for 30 cycles, using Dream Taq and Dream Taq Green Buffer (Thermo Fisher Scientific). The PCR amplicon was designed to contain a restriction enzyme recognition site in either the BW or PO allele of the candidate gene (see below for restriction enzyme selection). 0.5 μl of the appropriate restriction enzyme was added directly to the PCR mix and incubated at 37°C for at least 2 h, before agarose gel electrophoresis on a 1.5% agarose gel. Images of Peromyscus, embryos, and electrophoretic gels were adjusted for brightness across the entire image, cropped to remove excessive space, and labels added using Adobe Photoshop.
Additional File 1 Overrepresented GO terms for the list of genes containing a nonsynomomous substitution between P. maniculatus and P. polionotus.
Additional File 2 Underrepresented GO terms for the list of genes containing a nonsynomomous substitution between P. maniculatus and P. polionotus.
Additional File 3. Electrophoresis gel images before cropping used for