The compact genome of torafugu (Takifugu rubripes) (~400 Mb) is a useful model for annotating the vertebrate genomes. However, the currently incomplete assembly of the torafugu genome with 28% of scaffolds, in terms of chromosomal assignment and orientation unknown, has decreased its value for various genetic and genomic analyses. This is partly due to the usage of a natural heterozygous male torafugu as a starting material. As the heterozyogus genome possesses polymorphic sequences or might even contain copy number variations (CNV) between homologous chromosomes, the construction of accurate genome assembling has become extremely difficult.

1. Induction of mitotic gynogenesis in torafugu

For the purpose of reducing the polymorphism levels and improving the quality in genome assembling, we created homozygous torafugus through artificial induction of mitotic gynogenesis. First, diluted milt from 2 males was irradiated with 3 different UV dosages (40, 80 and 160 mJ/cm2) for a complete inactivation of sperm genetic materials. For each UV treatment, 150 g of eggs from 1 female were fertilized with 15 mL of irradiated milt mixed from 2 males, followed by an incubation in fresh seawater at 18.0°C. At 180 min post-fertilization, cold shock was applied to approximately 140 g of eggs in each UV treatment by soaking eggs in the icy cold seawater for 45 min. The remaining 10 g of eggs were served as haploid controls. Totally, we generated 7 groups named DG1 for the diploid gynogenesis and HC1 for the haploid control treated with a 40 mJ/cm2 UV dosage; named DG2 for the diploid gynogenesis and HC2 for the haploid control treated with a 80 mJ/cm2 UV dosage; named DG3 for the diploid gynogenesis and HC3 for the haploid control treated with a 160 mJ/cm2 UV dosage; and named NC for the normal control. Among all 3 DG groups, the highest development rate and hatching rate of 28.2% and 4.0% were achieved in the DG1 group. No fry survived to hatch in all 3 HC groups, meanwhile a microscopic analysis showed that the embryos had typical haploid syndromes, indicating the complete activation of paternal genetic materials by UV treatments.

2. Homozygosity analyses by microsatellite genotyping

Microsatellite genotyping was employed to assess the homozygosity levels of 16 mito-gynogenetic fry randomly sampled from the DG1 and DG2 groups (8 fry from either the group). Among the 156 microsatellite loci genotyped in 2 males and 1 female torafugus, a total of 56 microsatellite loci, sharing no common allele between the male and the female were selected to genotype the gynogenetic offspring. Each genotyping was performed by a microsatellite markers of post-PCR fluorescent labeling method and analyzed by capillary electrophoresis in a genetic analyzer. Through comparing the location of genotype peaks in the mito-gynogenetic or the normal offspring with those in the parents, the inheritance patterns clearly revealed that there was no genetic contribution from the paternal genomes to 16 mito-gynogenetic fry. As a result, the homozygosity levels of 16 mito-gynogenetic fry reached 100% at 56 unambiguous microsatellite loci, with an average density of 2-3 loci for every chromosome. The first instance of mitotic gynogenesis induction in torafugu has been successfully established.

3. Deep sequencing of torafugu genomes

Subsequently, we performed deep sequencing of a wild-type and a 5-month old mito-gynogenetic torafugu genomes using 2 Illumina next-generation sequencing (NGS) platforms (Illumina GA IIx and Hiseq 2000). For the wild-type torafugu, the Illumina GA IIx produced 84,857,156 reads in 101-bp length for 1 paired-end (PE) library of a 400-bp insert size while the Illumina Hiseq 2000 produced 278,642,344 and 244,796,700 reads in 76-bp length for 2 mate pair (MP) libraries of 2-Kb and 5-Kb insert sizes. For the mito-gynogenetic torafugu, the Illumina Hiseq 2000 produced 283,351,680 and 253,179,572 reads for 2 PE libraries of 300-bp and 500-bp insert sizes as well as 248,467,078 and 339,507,094 reads for 2 MP libraries of 2-Kb and 5-Kb insert sizes. The sum of base calls possessed physical coverages of approximately 120× and 280× to the theoritical genome size (~400 Mb) for the wild-type and the mito-gynogenetic torafugu, respecitively.

4. Genome-wide homozygosity analyses

Since the microsatellite genotyping at limited loci failed to reveal the whole-genome homozygosity levels of the mito-gynogenetic torafugu, our research established a genome-wide single nucleotide polymorphism (SNP) genotyping method based on the NGS data. After the pre-processing to raw reads, a total of 74,235,872 high-quality and de-duplicated reads from either the wild-type or the mito-gynogenetic torafugu DNA libraries were mapped against the reference sequences (the fifth fugu genome assemblies) with 1 mismatch allowed by Bowtie. The potential SNPs (pSNPs) between the reference and the mapped NGS reads (known as inter-pSNPs) were called with variant frequency of 100% at 26-34× depth of coverage by the CLC genomics workbench. A total of 39,499 inter-pSNPs were detected in the wild-type torafugu genome while a total of 93,525 inter-pSNPs were detected in the mito-gynogenetic torafugu genome. These 2 individuals shared 8,609 inter-SNPs in common. Among the remaining 30,890 (39,499-8,609) and 84,916 (93,525-8,609) inter-pSNPs in 2 individuals, the SNPs between homologous chromosomes within 1 individual (known as intra-SNPs) were called at the variant frequency ranged from 40%-60% at 26-34× depth of coverage as well. As a result, no intra-SNPs was detected in the mito-gynogenetic torafugu genome while 5,621 intra-SNPs were detected in the wild-type torafugu genome. Thus, it has been confirmed that the mito-gynogenetic torafugu was a doubled-haploid individual.

5. De novo assembly of torafugu genomes

According to the results of genome-wide SNP analyses, we estimated the effect of homozygosity levels on genome assembling by importing pre-processed reads containing nearly the same number of bases from the wild-type (5,880,348,666 bp) and the doubled-haploid (5,880,348,674 bp) torafugu DNA libraries into the SOAPdenovo genome assembler with the same parameter settings. The N50 size and the maximum length of contigs constructed from the reads of the doubled-haploid torafugu libraries were increased by more than 5 times and 2 times compared with the values of the wild-type torafugu. This has strongly suggested that different genome-wide homozygosity levels had great effects on genome assembling and the utilization of a doubled-haploid genome has improved the assembling quality by decreasing the polymorphic pathways of de Bruijn graphs during k-mer overlapping connection.

To achieve a good-quality assembling of the doubled-haploid torafugu genome, we employed 3 genome assemblers (SOAPdenovo, IDBA-UD and an embedded de novo assembler of the CLC genomics workbench) to investigate their different performance. The setting of multiple k-mer values for SOAPdenovo was ranged from 63-77, with a gradient of 2-mers every assembling while for the CLC genomics workbench, k-mer values of 53, 57, 61 and 64 were assigned. For IDBA-UD, genome assembling was performed under a gradient k-mer values ranged from 31-81, with a step addition of 25-mer. Among 3 assemblers, SOAPdenovo generated the largest contig of 75,131 bp and scaffold of 5,484,273 bp under the k-mer value of 71 and 73, respectively. After removal of potential sequencing errors with the k-mers of single-occurrence, the torafugu genome was assembled into 185,868 scaffolds containing estimated 380,443,322 bp of residues with a maximum scaffold size of 4,727,828 bp by SOAPdenovo. A further gap closure to the gap regions of the scaffolds increased the number of residues without Ns to 361,764,858 bp, and closed 47,203 out of 130,146 gaps. Finally, an alignment of the largest scaffold against 22 sequences of the fifth fugu genome assembly, showed a highly-identical match with a sequence on chromosome No. 9. Thus, the accuracy of de novo assembled torafugu sequences have been proved.

6. Conclusions

We confirmed the complete homozygosity of the mito-gynogenetic torafugu by microsatellite genotyping and genome-wide SNP analyses. With the utilization of the doubled-haploid torafugu as a starting material, we obtained an approximately 378 Mb of sequence assembly for the torafugu genome with an estimated coverage depth of 50× from the Illumina paired-read sequences in less than 5 months. The quality of the torafugu genome assembled from short reads was proved to be valid.

高等生物のゲノム解析において、全ゲノムショットガン法は有効な手段であるが、各種多型の存在が高精度ゲノムをアセンブルするための一つの障害要因になっている。本研究では、魚類における代表的なモデル生物であるドラフグにおいて、すでに存在しているゲノムドラフトシーケンスよりはるかに高精度のゲノムシーケンスを解読、構築するために、相同染色体がホモ接合しており多型の寄与を除外できる第一卵割阻止型の雌性発生個体をも作出し、ゲノムシーケンシングに資することを目的に研究をおこなったものである。

論文内容の概要は

(1)第一卵割阻止型雌性発生トラフグ作出条件の検討

UV照射量や卵割阻止操作の有無などの検討を行い、十分で効率的UV照射量を見いだしたのとともに、卵割阻止を行なわない場合は、半数体症候群個体が生じるが、阻止することにより正常な個体が発生することを確認し、多数の第一卵割阻止型雌性発生個体を得ることに成功した。

(2)マイクロサテライトマーカーを用いたホモ接合性の評価

得られた第一卵割阻止型雌性発生個体がホモ接合しているかを検証するため、各染色体から2、3種類のマイクロサテライトマーカー(のべ56種類)を用いて雌性発生個体16匹の多型タイピングを行なった。その結果、全ての個体の全てのマーカーでホモ接合していることが確認できた。

(3)全ゲノムSNPタイピングによるホモ接合性の評価

第一卵割阻止型雌性発生個体のホモ接合性をより包括的に検証するため、通常の二倍体の1個体、第一卵割阻止型雌性発生の1個体を対象に次世代シーケンサーによる解読を行ない、全ゲノム的にSNPのタイピングを行なった。配列が1塩基異なる部位が多型か、リピート間の差異かを判別するのは容易ではないが、SNPである確度の高いサイトに関してSNPの個数を比較した結果、二倍体では5,621個のSNPサイトが見いだされたのに対して、雌性発生個体では0個であった。このことから作出した個体が全ゲノム的にホモ接合していることが確かめられた。

(4)ゲノムアセンブルによる有効性の確認

(3)の二倍体個体、雌性発生個体について、解読したRaw readsの塩基数が同一になるようにデータ量をあわせてアセンブルした結果、雌性発生個体のN50 Contig長は約2kbであるのに対して、二倍体では0.35kbと雌性発生個体の方が5倍以上のN50 Contig長を得ることができた。このことから雌性発生個体を全ゲノムショットガン法で解読することの有効性が示された。

以上の結果により、審査委員一同は、本論文が博士(農学)の学位論文として価値あるものと認めた。