Construction of High-Resolution RAD-Seq Based Linkage Map, Anchoring Reference Genome, and QTL Mapping of the Sex Chromosome in the Marine Medaka Oryzias melastigma

Medaka (Oryzias sp.) is an important fish species in ecotoxicology and considered as a model species due to its biological features including small body size and short generation time. Since Japanese medaka Oryzias latipes is a freshwater species with access to an excellent genome resource, the marine medaka Oryzias melastigma is also applicable for the marine ecotoxicology. In genome era, a high-density genetic linkage map is a very useful resource in genomic research, providing a means for comparative genomic analysis and verification of de novo genome assembly. In this study, we developed a high-density genetic linkage map for O. melastigma using restriction-site associated DNA sequencing (RAD-seq). The genetic map consisted of 24 linkage groups with 2,481 single nucleotide polymorphism (SNP) markers. The total map length was 1,784 cM with an average marker space of 0.72 cM. The genetic map was integrated with the reference-assisted chromosome assembly (RACA) of O. melastigma, which anchored 90.7% of the assembled sequence onto the linkage map. The values of complete Benchmarking Universal Single-Copy Orthologs were similar to RACA assembly but N50 (23.74 Mb; total genome length 779.4 Mb; gap 5.29%) increased to 29.99 Mb (total genome length 778.7 Mb; gap 5.2%). Using MapQTL analysis with SNP markers, we identified a major quantitative trait locus for sex traits on the Om10. The integration of the genetic map with the reference genome of marine medaka will serve as a good resource for studies in molecular toxicology, genomics, CRISPR/Cas9, and epigenetics.

water in Asian regions including Pakistan, India, Burma, and Thailand (Naruse 1996). O. melastigma has been acknowledged as a potential model fish for marine ecotoxicological studies and is useful for the evaluation of acute and/or chronic toxicity, and embryo toxicity testing (Chen et al. 2011;Dong et al. 2014;Kim et al. 2016;Kong et al. 2008;Shen et al. 2010;Wang et al. 2011;Kim et al. 2016). The genus Oryzias has been divided into three major monophyletic species groups; the latipes, the javanicus, and the celebenesis groups, while O. dancena, a closely related species of O. melastigma, has been phylogenetically placed in the javanicus group (Takehana et al. 2005). Phenotypic distinction between male and female in medaka is distinguished by a number of secondary sex characters including shape and size of dorsal and anal fins due to morphologically indistinguishable sex chromosomes (Schartl 2004). Moreover, Oryzias species shows the both XY and ZW sex-determining systems. O. latipes, O. dancena, and O. minutillus have an XX/XY sex-determining system and O. hubbsi has a ZZ/ZW (Matsuda et al. 1998;Takehana et al. 2008). For example, sex-determining gene dmrt1bY has been identified in a few species of the O. latipes group and sox3 Y in O. dancena and some species of the celebenesis group (Kondo et al. 2003;Matsuda et al. 2002;Myosho et al. 2015). There is a lack of studies on sex-determining genes in O. melastigma. Therefore, more genomic resources specifically devoted for the marine medaka O. melastigma are required.
A genetic linkage map is a very useful tool to understand genetic architecture such as chromosome structure, segregation distortion regions, recombination rate, and recombination hotspots (Meyer and Van de Peer 2005;Zhu et al. 2014;Guo et al. 2013;Shifman et al. 2006). Furthermore, it provides a framework for mapping the chromosomal location of single-gene traits and quantitative traits of interest, and helps to facilitate candidate gene cloning, and comparative genomic analysis with some genome information together (Lee et al. 2005;Amores et al. 2011;Feng et al. 2015;Zhu et al. 2013;Li et al. 2015;Shao et al. 2015;Wang et al. 2015;Xiao et al. 2015;Kanamori et al. 2016). A high-density genetic map also plays an important role in assembling whole genome sequences by examining the accuracy of de novo genome assembly (Dukić et al. 2016;Rhee et al. 2017;Wang et al. 2017). Indeed, the importance of a high-density genetic map has been demonstrated during the de novo genome assembly in teleost fish, as it validated the presence of additional genome duplication (Meyer and Schartl 1999;Meyer and Van de Peer 2005;Volff 2005). In addition, restriction-site associated DNA (RAD) sequencing based on next-generation sequencing (NGS) enables the rapid discovery of genome-wide genetic markers and highthroughput single nucleotide polymorphism (SNP) genotyping in mapping families and facilitates the construction of high-density genetic linkage maps in both model and non-model organisms (Baird et al. 2008;Davey and Blaxter 2010;Amores et al. 2011;Davey et al. 2011;Etter et al. 2011).
We have recently published a reference genome assembly (total genome length 779.4 Mb) of O. melastigma as a model species in environmental toxicology (Kim et al. 2018). In this study, we constructed a high-density genetic linkage map of O. melastigma using an F1 full-sib family. Using the genetic linkage map, sex QTL has been mapped in the genome of O. melastigma and the previous genome assembly was anchored onto the linkage map to improve the contiguity of the assembly. The development of a high-density genetic map is imperative to facilitate both genetic and genomic studies in O. melastigma. The present study will assist in a better understanding of genome-based research in molecular toxicology, genomics, CRISPR/Cas9, and epigenetics.

Mapping cross
The marine medaka O. melastigma used in this study were obtained from the City University of Hong Kong (kindly provided by Dr. Doris W.T. Au) and maintained at Nagahama Institute of Bio-Science and Technology, Nagahama, Japan. A male and a female fish were bred to produce F1 progenies. In total, 58 F1 individuals were used to create a linkage group (LG).

RAD sequencing
Genomic DNA was extracted from muscle tissue using DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The size and quality of DNA isolated was checked on 1% agarose gel by electrophoresis and the concentration was measured using Qubit florometer (Thermo Fisher Scientific, Waltham, MA, USA). Genomic DNA (40 ng) of each sample was digested with BglII (5 Unit) and EcoRI (5 Unit), ligated with a Y-shaped adaptor, amplified by polymerase chain reaction (PCR) with KAPA HiFi HS ReadyMix (Kapa Biosystems, Wilmington, MA, USA), and fragments were selected with E-Gel Size Select (Life Technologies, Carlsbad, CA, USA). The mean size of the selected fragments was 333 bp (CV 16.4%). Further details of the library preparation method are described in a previous study by Sakaguchi et al. (2015). RAD sequencing (RAD-seq) was performed using 58 F1 individuals and both parents with HiSeq2500 (Illumina, San Diego, CA, USA) with eight cycles for index read and 51 cycles for the reads of interest. For each parental sample, the same amounts were aliquoted in four different reaction tubes and sequencing of each reaction was carried out to reduce PCR amplification bias. All procedures related to RAD-seq including the library construction were performed by Clockmics Inc. (Osaka, Japan).
Extracting RAD-tags and SNP genotyping by Stacks Quality filtration of sequence reads was performed using Trimmomatic v.0.33 (Bolger et al. 2014) with parameter options of -0.33.jar SE -phred33 TOPHRED33 ILLUMINACLIP TruSeq3-PE-2.fa:2:30:10 LEADING:19 TRAILING:19 SLIDINGWINDOW:30:20 AVGQUAL:20 MINLEN:51. RAD-tag extraction and genotyping were performed with Stacks v.1.47 software (http://creskolab.uoregon.edu/stacks/) . The sequence reads were aligned to the available reference genome (GCF_002922805.1, Kim et al. 2018) using GSnap (https://github.com/ juliangehring/GMAP-GSNAP) with default parameters (-t 30 -n 1 -m 5 -i 2), which converted to BAM files. All RAD-tags catalog from the parental samples were extracted by Stacks using the ref_map.pl pipeline with the parameters -m10 and -P 3 and genotyping was called by the parameters of minimum number of 5 reads to call a homozygous genotype, a minimum minor allele frequency of 0.1 to call a heterozygote, and a maximum minor allele frequency of 0.05 to call a homozygote. Among RAD-tags, single nucleotide polymorphism (SNP) markers with maximum likelihood of 0 were selected for mapping and the SNP markers with genotypes of at least 53 F1 offsprings (. 90%) were collected for map construction using the command genotypes -r 53 of Stacks v.1.47. Data of raw sequences were deposited in the Sequence Read Archive (SRA) (http://www.ncbi.nlm.nih.gov/sra) under the accession number PRJNA514812.

Linkage map construction
Linkage analysis of genetic markers (SNPs) was performed using Join-Map 5.0 (Wageningen, Netherlands, Van Ooijen 2018). The SNP markers with a significant segregation distortion (x 2 test, P , 0.01) were removed from the analysis of linkage map construction. Linkage groups were identified by the grouping parameters of independent Likelihood Odds Ratio (LOD) threshold of 5 provided in JoinMap 5.0. Map distances were calculated by the Kosambi's mapping function and mapping algorithm used for building linkage map was regression mapping based on the default parameters. The regression mapping adds loci one by one starting from the most informative pair of loci. The best position of each added locus is searched by comparing the goodness-of-fit of the calculated map for each tested position. JoinMap performed three rounds of marker positioning with a jump threshold of 5 and we took second round of map as a final map as recommended by the manual. The linkage map was visualized using MapCart 2.32 (Voorrips 2002). The name of the linkage groups was matched with the homologous chromosomes of Japanese medaka.

Comparative analysis of two medaka genomes
The final linkage map based genome assembly of marine medaka was compared with the genome of Japanese medaka (Hd-rR strain) to compare the similarity between two medaka genomes using Mummer v.3.0 (http://mummer.sourceforge.net/manual/#coords).

Sex linkage analysis
The mapping panel consisted of 27 males and 31 females. In order to examine sex determining mechanism in O. melastigma by linkage analysis, two completely linked sex markers (sex_xy and sex_zw) were additionally added in the genotype files. For sex_xy and sex_zw genotypes, male individuals were coded as heterozygotes and homozygotes, respectively, while female individuals were coded as homozygotes and heterozygotes, respectively. The linkage analysis of sex markers was performed with JoinMap 5.0 (Wageningen, Netherlands, Van Ooijen 2018). Since the sex trait is qualitative, the phenotype for sex was converted into binary code; male 1 and female 0. For QTL analysis, standard interval mapping was performed and significance was determined by permutation test (n = 1000) using MapQTL 6.0 (van Ooijen et al. 2002).

RESULTS
Constructing genetic linkage map of the marine medaka Oryzias melastigma Stacks software extracted 113,367 RAD-tags from O. melastigma genome. Among them, the number of putative SNP markers was 34,040. To distinguish polymorphisms from sequencing errors, we collected 24,441 SNP markers with Likelihood Ratio of 0, including 8,518 SNP markers that have been located in the same RAD-tags more than twice. Among the remaining 15,922 loci, 4,497 SNP loci were successfully genotyped in at least 53 F1 individuals (.90%). After removing the markers that showed a segregation distortion (P , 0.01), 3,732 were finally used for building a genetic map. The 3,730 markers were grouped into 24 LGs with a LOD $ 5.0 by independence LOD and two remaining markers were not linked with any of those groups. The regression mapping function by JoinMap positioned 2,481 SNPs in the second rounds of mapping by comparing the goodness-of-fit (Suppl. File S1). The 24 LGs were consistent with the number of chromosomes (n = 24) in O. melastigma (Uwa et al. 1983). The total map length was 1,784 cM and each LG included 57-173 markers with an average marker interval of 0.72 cM ( Figure 1 and Table 1).

Comparative genomic analysis of two medaka genomes
The final genome assembly integrated with a genetic map provides an efficient resource for comparative genomic analysis with other medaka genomes such as a Japanese medaka (O. latipes). The 24 genetic map-based scaffolds showed good homology in gene contents and sequences similarity with chromosomes in O. latipes, Most LG-based scaffolds of marine medaka showed collinear relationships completely or in a majority with the counterpart chromosomes of O. latipes (Figure 2 and Suppl. Fig. S3). Other LG/ chromosomes in O. melastigma showed disrupted collinearity due to intrachromosomal rearrangements or possible errors in linkage map. Reversely matched parts in the collinearly related scaffolds were most likely caused by the undetermined orientation of scaffolds involved in the anchoring process. In addition, some LG-based scaffolds showed in/del regions (Om05, Om08, Om11, Om14, Om15, Om16, and Om17) compared to Japanese medaka chromosomes.

Mapping of sex-determining regions
The linkage analysis of two sex markers (sex_xy and sex_zw) showed that both markers were more or less linked to markers in the Om10. Sex_xy showed the highest LOD scores with 17018 (LOD = 17.06) and strong linkages with other markers in this linkage group (Suppl .  Table S4). Although sex_zw showed the highest LOD value (7.42) with 16660, which also had the same LOD with sex_xy, the LOD scores of sex_zw with other markers indicated weak or suspect linkage (Suppl .  Table S4). This investigation suggested that the O. melastigma has XY sex-determining system. A significant QTL for sex was detected in Om10 with the LOD significance threshold of 5.3 based on the permutation test ( Figure 3A). Among 2,481 markers mapped in the Om linkage groups, 58 markers showed the significant association with sex (Suppl. Table S5), most of which were markers from the RACA27. The most significant two markers associated with sex were 17040 (24.3 cM) and 17018 (24.4 cM) with LOD of 35 (Suppl. Table S5). The genotypes of those markers were completely linked with the sex phenotype of all individuals, except for one animal (d004), which could be caused by wrong phenotyping or sex-reversal. When the animal was removed from the QTL analysis, LOD scores of 17018 and 17040 increased to 881.12 and 299.4, respectively ( Figure 3B). Physical location of sex-determining region was mapped between 10,353,256 bp and 10,538,878 bp in Om10, spanning approximately 186 kb, and 10 genes were located in the regions without any obvious candidate gene ( Figure 3C). However, the sox3 was located on the 10,625,885 bp, which was 87 kb apart from the mapped region (Suppl. Table S3; Figure 3C).

DISCUSSION
A high-resolution genetic map is very useful in diverse genomic research and has been applied to many fish species Li et al. 2015;Shao et al. 2015;Xiao et al. 2015;Wang et al. 2015;Kanamori et al. 2016). In this study, a high-density genetic map was constructed in O. melastigma using RAD sequencing and was used for verifying the previously published marine medaka reference genome and for aligning the scaffolds at the chromosomal level. The genetic map of O. melastigma consists of 24 LGs with 2,481 SNP markers, which cover 24 chromosomes (Uwa et al. 1983). The 810 genetic markers were anchored to the genetic map (Table 2 and Suppl. Figure 2), which mapped 90.7% (713 Mb) of the reference genome assembly sequence onto the genetic map (Table 3). Although all SNP markers were extracted by the alignment of RAD sequences against the reference genome of marine medaka (Kim et al. 2018), the number of anchored markers were lower than anticipated. Many markers in the LGs were excluded from the integrating process due to inconsistency between the genetic map and reference assembly, indicating possible errors in marker order and/or de novo assembly. Indeed, in most cases the markers tended to be highly concentrated in the narrow regions, suggesting that more recombinant individuals will be required to obtain a more precise marker order. Also, it is likely that the size of the mapping family was not big enough for the marker concentrated regions, compared with the number of markers analyzed, resulting in relatively low number of anchored markers. Despite this shortage, the marine medaka genetic map was still integrated with the reference assembly, considering other fish species had a similar or slightly lower level of scaffolds mapping onto the genetic map (Tine et al. 2014;Li et al. 2015;Wang et al. 2017). Previously, we have developed a reference genome for the marine medaka O. melastigma in several steps (Kim et al. 2018). De novo assembly was performed using the combination of Platanus and Haplomerger2 assemblers (Kajitani et al. 2014;Huang et al. 2017). The contiguity of de novo genome assembly was further increased by RACA (Kim et al. 2013), which assisted in the construction of highly ordered and oriented scaffolds of de novo assembly and reassembled scaffolds into longer chromosomal fragments using the comparative n■  (Kim et al. 2018). Taken together, the genetic map-integrated assembly improved the reference sequences by mapping 90.7% of genome sequences onto 24 LGs with 94% determined orientation of mapped scaffolds ( Table 2). The BUSCO and N50 demonstrated a better quality of integrated final genome assembly with values increased to 96.8% and 29,987,720 bp, respectively (Tables 3 and 4). Quantitative trait locus (QTL) analysis identified that Om10 was strongly associated with sex in O. melastigma and two markers, 17040 and 10718, showed the most significant association for sex, without any recombinant between genotype and phenotype. LOD scores of 17040 (LOD = 34.99) is little higher than that of 10718 (LOD = 34.98), which is likely associated with the missing genotype in one individual for 10718 (Suppl . Table S5). Thus, sex determining locus should be located between 17018 and 17040.
Although no obvious candidate gene for sex was found in the mapped sex-determining regions, we noticed sox3 outside the mapped region ( Figure 3C). It was intriguing because a cis-regulatory element in the downstream region of sox3 on Y chromosome plays a key signal to sex determination in O. dancena (Takehana et al. 2014), suggesting that the same cis-regulatory element might be present in the mapped regions of O. melastigma. For further analysis, fine mapping with additional recombinants and the functional analysis of candidate genes or regulatory elements need to be performed in the mapped sex-determining region.
In summary, a high-density genetic map is a very useful resource to determine the accuracy of de novo genome assembly, especially with massively parallel short sequencing reads. In addition, the integration of a genetic map with reference genomes will be a useful resource in various genomic studies including comparative genomic analyses, fine mapping of QTL, positional cloning of candidate genes, and CRISPR/Cas9 studies.

ACKNOWLEDGMENTS
This study was supported by a grant from the National Research Foundation (2017R1D1A1B03036026) to Bo-Young Lee.