Introduction

Domestication is the process in which a population of wild animals or plants has evolved to match human requirements through a series of artificial selection. Because domestication has had strong influence on human culture and history, it is of particular interest to know when and from where the domesticated species originated, for what progenitors have been utilized, and how cultivars were created. In case of the Asian cultivated rice Oryza sativa, these basic questions remained unsolved, because the relationship of wild and cultivated rice varieties is quite complex. O. sativa consists of diverse varieties, many of which can be classified into two major groups, japonica and indica. There are two alternative hypotheses about the process of domestication from its progenitor O. rufipogon. One suggests that japonica and indica were derived from a single ancestor and then diverged, while the other suggests that these two groups were independently domesticated from different varieties of O. rufipogon that diverged much earlier than domestication. Since either of the hypotheses was apparently supported by several lines of evidence, the origin(s) of the cultivated rice and divergence time of japonica and indica remained a matter of debate. In addition, it was suspected that some genes might have been transferred between japonica and indica through hybridization, but the amount and directions of genes transferred were unknown. Here, using large-scale data generated by the next-generation sequencers, I conduct genome and transcriptome analyses of wild and cultivated rice varieties. I attempt to reveal the origin(s) and divergence time of the Asian cultivated rice, and to elucidate contribution of gene flow between japonica and indica to the process of their domestication.

Results

1. Independent domestication revealed by large-scale data analysis

Gene tree discordance, which is widely observed in closely related species, is a serious problem that gene trees are inconsistent with an expected species relationship because of ancestral polymorphisms. To examine gene tree discordance in rice, the genome sequences of a japonica cultivar Nipponbare (NB) and an indica cultivar 93-11 were compared with 2,026 full-length cDNA sequences obtained from O. rufipogon W1943, which was sampled in Jiangxi, China. As a result, a total of 167 gene trees could be reconstructed with statistical significance, and as theory suggests, all the three possible tree topologies were observed. However, one topology that supported closeness of japonica and W1943 were clearly predominant, whereas the other two were minor (Figure 1A). Therefore, it is suggested that the major topology showing a sister group of japonica and W1943 represents the species relationship of these groups and the others were caused by ancestral polymorphisms. This result shows that, despite the gene tree discordance, a careful examination of large-scale data should be useful to solve phylogeny of closely related species.

To answer the question of the origin(s) of the cultivated rice, RNA-seq data of two diverse varieties (W1943 and W0106) of O. rufipogon were used. Because W0106 was sampled in Orissa, India, it was expected to show a phylogenetic pattern different from that of W1943. In addition, the genomic sequences of another indica cultivar Guangluai-4 (GLA4), for which ~20X coverage Illumina reads were available, was employed. Aligning all the short reads to the japonica genome, I obtained 6,142,852 sites that were shared among all the four groups and 8,868 parsimony-informative sites were found. Under the maximum parsimony principle, a tree topology can be inferred at each of these 8,868 sites. It was expected that one of the three possible unrooted tree topologies, which is consistent with the evolutionary relationship of these groups, should be predominant, whereas the other two minor topologies would be observed because of ancestral polymorphisms. The data clearly indicates that at 7032 sites (79%) japonica and indica are close to W1943 and W0106, respectively, and japonica and indica do not form a sister group (Figure 1B). Hence, I conclude that japonica and indica were derived independently from distinct groups of O. rufipogon.

2. Gene flow from japonica to indica

If ancestral polymorphisms are the only cause of gene tree discordance, similar numbers of minor tree topologies should be observed. However, the minor tree showing that japonica and indica consist a sister group was supported greater than the other one (Figure 1). In a genome alignment of NB and 93-11 (Figure 2A) and that of NB and GLA4 (Figure 2B), several nearly identical regions, which were possibly due to hybridization between japonica and indica, were found. Thus, a larger number of trees showing japonica-indica closeness can be accounted for by hybridization events after their split.

In many cases, the positions of the nearly identical regions were similar between two alignments (Figure 2). This is possibly because hybridization occurred in the common ancestor of the two indica cultivars. To investigate this possibility, a window analysis of 10 kb was performed to compare nucleotide differences between NB and 93-11 (djp-9311) and those between NB and GLA4 (djp-GLA4). Since 93-11 and GLA4 are equally distant to NB, djp-9311 and djp-GLA4 should be scattered around the diagonal line in an x-y plot, although they may slightly fluctuate by chance (Figure 3A, region I). In addition, because recent hybridization before the split of 93-11 and GLA4 was expected, there should be windows with small djp-9311 and djp-GLA4 (Figure 3A, region II). In fact, the x-y plot of djp-9311 and djp-GLA4 showed that a significant number of dots were located in the regions II (Figure 3B), suggesting hybridization before divergence of the indica cultivars.

Since recent gene flow was found between japonica and indica, next question was its amount and direction. If a genomic region was introgressed from japonica to indica, the region should be more closely related to W1943 than to W0106. This can be examined by comparing djp/in-W1943 and djp/in-W0106, where jp/in means the nearly identical regions between NB and GLA4. It was found that at least 13 Mb of the nearly identical regions, which contain ~2,000 genes, were introgressed from japonica to indica, while only 0.77 Mb were from indica to japonica. These observations suggest that there might have been much more genes transferred from japonica to indica. Since relatively strict criteria were used to define nearly identical regions, these should be regarded as minimum estimates of the amount of introgressed genes.

3. Japonica and indica groups diverged earlier than domestication

The estimated divergence time of japonica and indica ranges from 8,200 years to 0.44 million years in previous studies. This considerable range of the time might be due to inappropriate assumption such as single origin hypothesis or disregard of ancestral polymorphisms and hybridization. It is believed that molecular data such as gene sequences obtained from different species or groups are useful to accurately estimate divergence time. However, because genes (alleles) are diverged before the divergence of two groups (Figure 4), observed divergence time of genes (tobs) is always larger than the real divergence time of groups (t). The difference between tobs and t is due to ancestral polymorphisms and depends on the ancestral effective population size. Here I employed a model that includes ancestral polymorphisms and estimated both divergence time and the effective population size using the maximum likelihood method. The number of nucleotide differences was counted in alignments of 2 kb between NB and GLA4 and a distribution of the differences was created. The genome alignment of these two cultivars consists of two different regions that diverged at ancient time and were introgressed recently. Thus, the distribution of the nucleotide differences can be a combination of two distributions (Figure 5A), each of which has its own divergence time and effective population size. Moreover, the proportions of the two distributions, which correspond to the amounts of introgressed and non-introgressed regions, should be estimated. As a result, the distribution based on estimated parameters (Figure 5B, black curve) fit well with the observed distribution (Figure 5B, gray bars). The estimated divergence time of japonica and indica was 120,000 years, which is much earlier than the widely accepted domestication time, 9,000 years. Thus, this result provides a strong support that japonica and indica were independently derived from different varieties of wild rice. The ancient effective population size was 125,000, while the effective population size of the japonica group before hybridization was 6,700. This reduced effective population size of japonica might be related to a severe bottleneck in japonica. It is reasonable to think that only a small number of japonica ancestors were selected during domestication. It is noteworthy that the estimated proportion of the introgressed regions was 21%, which is larger than my previous estimate. This inconsistence may be due to the strict criteria used for selecting nearly identical regions in previous analysis. These results show that japonica genes may have played important roles in shaping current indica cultivars.

Conclusion

Several significant aspects of molecular evolution of rice were revealed. First, this study provides clear evidence that japonica and indica were originated from different varieties of wild rice. Intriguingly, W1943 was sampled near the Middle Yangtze, which is well-known as the center of early rice cultivation, whereas W0106 was sampled in East India, which is close to the Middle Ganges valley where archeological evidence indicates rice cultivation of ~7,000 years ago. This geological evidence and the phylogeny disclosed in this study suggest that rice was domesticated multiple times in distant places. Second, a significant portion (21%) of the indica genome was derived from japonica. This finding suggests that even though japonica and indica diverged much earlier than their domestication and were separately domesticated, they recently crossed, so that some japonica genes contributed to create the modern indica rice. The introgressed genes detected here are possibly related to important traits, and can be useful resources for breeding of rice in the future.

Figure 1. Ratios of gene trees. (A) NB, 93-11, and W1943 were used. (B) NB, GLA4, W1943, and W0106 were used. J: japonica; I: indica.

Figure 2. Alignment in chromosome 3 (A) between NB and 93-11 and (B) between NB and GLA4. Thick bars below x-axis represent "nearly identical regions" that showed <0.2 nt / kb differences and >75 kb.

Figure 3. (A) Expected and (B) observed differences between japonica and indica.

Figure 4. Observed and real divergence time.

Figure 5. (A) Expected and (B) observed distributions.

本論文は、イネ栽培種並びに野生種のゲノム比較解析によって、栽培化過程におけるイネの歴史や進化的変化を明らかにしたものであり、以下の3章からなる。

第1章において、アジアの野生種であるオリザ・ルフィポゴンと栽培種であるオリザ・サティヴァの系統関係を解き明かした。まず、サティヴァに属する主要な2グループのジャポニカ及びインディカのゲノム配列を、ルフィポゴンのW1943(多年生)の完全長cDNA配列と比較解析した。その結果、イネの大規模解析に於いては遺伝子系統樹の不整合性(gene tree discordance)といわれる現象を考慮に入れなければならないことが判明した。栽培種と野生種の系統関係を更に深く解明すべく、2つの栽培種と2つの野生種を含む4つのグループでゲノム全体をカバーするようなデータを使用した。ルフィポゴンのW1943とW0106(一年生)においては、全トランスクリプトームのショットガン法による配列(RNA-seq)を、また、インディカの広陸矮4においてはゲノムの短鎖配列を用いた。この解析でジャポニカはW1943により近く、一方でインディカはW0106により近かったことから、ジャポニカとインディカは異なる野生イネから得られたことが示された。

第2章においては、ジャポニカとインディカの間の遺伝子交流がはっきりと同定された。これは、第1章の樹形のパターンが、ジャポニカとインディカのグループ間での交雑が生じたことを示唆したことによる。この交雑はインディカ栽培種が分岐して多様化する以前に起こったのかもしれない。加えて、遺伝子交流の規模と方向を詳細に調べたところ、およそ5%の遺伝子が導入されており、幾つかの既知の栽培化関連遺伝子を含めて、大半はジャポニカからインディカに入っていた。

第3章において、交雑と祖先多型も考慮に入れた上でジャポニカとインディカの分岐時間を推定したところ、およそ120,000年前であった。さらに、有効な集団の大きさを推定すると、ジャポニカの祖先集団は野生イネの祖先に比べてはるかに小さく、ジャポニカの栽培化過程において強い瓶首効果のあったことが示唆された。

結論として、本研究は、ジャポニカとインディカが異なる栽培化の歴史を経ながらも、それらの間で非常に多くの遺伝子が動いていたことを明らかにした。従って、本論文は、現在見られるアジアの栽培イネがどのようにして形成されたかを詳細に解明したものであると認められる。

なお、本論分第1章と第2章は、農業生物資源研究所の松本隆、伊藤剛、沼寿隆、坂井寛章、川原善浩、呉健忠、水野浩志との共同研究であるが、論文提出者が主体となって大規模配列解析及び検証を行ったもので、論文提出者の寄与が十分であると判断する。

したがって、博士(生命科学)の学位を授与できると認める。