Balance between a Higher Degree of Heterosis and Increased Reproductive Isolation: A Strategic Design for Breeding Inter-Subspecific Hybrid Rice

The application of heterosis (hybrid vigor) has brought great success to plant breeding, particularly of hybrid rice, achieving significant yield increases. Attempts to explore the heterosis of inter-subspecific hybrids between indica and japonica rice, which result in even greater yield increases, have greatly increased in the past decades. However, because of the reduced seed setting rate in F1 hybrids as a result of increased reproductive isolation, the application of inter-subspecific hybrids in rice has slowed. Understanding the balance between heterosis and the reproductive isolation of inter-subspecific hybrids will facilitate the strategic design of inter-subspecific hybrid breeding. In this study, five indica and seven japonica rice varieties were chosen as the parental lines of a complete diallel mating design. Data from six group traits from all of the hybrids and inbred lines were collected. We found that the grain weight per plant, grain number per panicle, tiller per plant, thousand grain weight and plant height, which reflected increased heterosis, were associated with the genetic divergence index (GDI) of the parents. Meanwhile, owing to the reduced seed setting rate, which was also associated with the parents' GDI, the grain production of the hybrids was negatively affected. After analyzing the relationships between the GDI of indica-japonica parents and the grain weight per plant of the F1 hybrids, an ideal GDI value (0.37) for the two indica-japonica parents that could provide an optimal balance between the inter-subspecific heterosis and reproductive isolation was proposed. Our findings will help in the strategic design of an inter-subspecific hybrid rice breeding program by identifying the ideal indica and japonica parents for a hybrid combination to achieve hybrid rice with an optimal yield. This strategic design of an inter-subspecific hybrid rice breeding program will be time saving and cost effective.


Introduction
Rice is the main staple food for almost half of the world's population. In China, rice constitutes 40% of the total calorie intake [1]. Asian-cultivated rice (Oryza sativa L.) consists of two subspecies, indica and japonica. These subspecies have significantly diverged at the molecular [2][3][4][5], physiological and biochemical levels [6][7][8]. Regarding morphology and living conditions, indica varieties exhibit greater plant heights, longer leaves, and heat and moisture tolerance but are sensitive to low temperatures. These varieties are cultivated at low latitudes and in humid regions. japonica varieties have lower plant heights and shorter leaves than do those of indica but are tolerant to low temperatures. In addition, these varieties are more suitable for high latitudes and for the lower latitudes of high altitude cultivation. Because of these differences, reproductive isolation emerges in the hybridization between these two subspecies.
Hybrid rice was first widely commercialized in the 1970s in China [9]. Some elite hybrid rice varieties with excellent performances were bred in China, such as Shanyou63 and Luoyou8, which are widely cultivated in Asian countries. However, these varieties are mostly intra-subspecific hybrids, and the heterosis of the hybrids is limited. With the growing global population and decreasing proportion of arable land, increasing rice yield remains an urgent task. The magnitude of heterosis depends on the genetic diversity between the two parents of the hybrids. The greater the genetic difference between the parents, the higher the heterosis [10]. Fortunately, the inter-subspecific hybridization between indica and japonica can gain powerful heterosis compared with intra-subspecies hybridization [11][12][13][14][15]. The utilization of inter-subspecific hybrids is the most feasible approach for realizing super-high yields [1,13]. Hybrids of indica and japonica varieties have a yield advantage of approximately 25% [14]. In addition, Dr. G. Khush of the IRRI has also predicted that the combination of IRRI's new plant types with the heterosis of indica-japonica may increase the yield potential of tropical rice by 50%.
However, breeders are hampered by the reproductive isolation of inter-subspecies. As a type of evolutionary phenomenon, the F 1 hybrids often have a low seed setting rate [16,17]. Even after doubling the number of chromosomes to obtain autotetraploid rice, this problem could still not be resolved [18]. Over recent decades, many studies have investigated the low seed setting rate defect [16,[19][20][21][22][23][24][25]. Explanations for this phenomenon include pollen sterility, embryonic abortion or incompatibility. Because the interaction of indica-japonica hybridization is complex, reproductive isolation is still a major obstacle to breeding inter-subspecific hybrid rice.
Considering that heterosis is determined by the divergence of the parents and that the seed setting rate is controlled by the degree of reproductive isolation in the parental lines, we hypothesized the existence of a critical point at which the conflict between the greater degree of heterosis and the increased reproductive isolation is balanced. At this point, the heterosis of inter-subspecific hybridization can be acquired, and a high seed setting rate can also be realized. In search of this point, twelve rice varieties, five indica and seven japonica, with different proportions of indica-japonica content were selected as the mating parents of complete diallel hybridization. An experimental population with 132 hybrids and twelve parental lines was constructed. The genetic divergence index (GDI) of the indica-japonica parents and the plant height, grain weight per plant, and yield related trait data were collected. Regression analyses were performed to depict the relationship between the yield and GDI of the parents. After integrating the regression results of the raw phenotypic data, relative low-parent heterosis, relative mid-parent heterosis, and relative better-parent heterosis, an ideal GDI was found and was consistent in reciprocal combinations. This result will help to facilitate the inter-subspecific hybrid rice breeding program and contribute to yield improvement.

Materials and Methods
A total of 260 rice varieties (CV, the field studies did not include endangered or protected species), which were collected worldwide (China, India, Italy, Japan, Philippines, respectively, and we received the rice samples from fellow researchers, Guangzhu Zhang from Heilongjiang Academy of Agricultural Sciences, Liangming Chen from Nanjing Agricultural University, Liyon Cao from China National Rice Research Institute, Wuhan Zhang from Hunan Hybrid Rice Research Center, Yixuan Lu from Yunnan Academy of Agricultural Sciences, respectively), were planted in the experimental field of the Engineering Research Center for Plant Biotechnology and Germplasm Utilization, Ministry of Education, Wuhan University, in Wuhan (N30u 329 22.440, E114u 229 18.210; no specific permissions were required for the locations in this study) in May 2011. The DNA from every inbred line was extracted using the CTAB method [26]. The proportions of indica-japonica content named F i of these inbred lines were determined using the InDel marker estimating method. The selected co-dominant InDel markers have been validated to truly reflect the genetic variation and differentiation of indica and japonica subspecies [27]. After PCR reactions, electrophoresis, the banding patterns of these rice varieties could be divided into indica type which was identical to 93-11 and japonica type which was identical to Nipponbare. By calculating the indica or japonica banding type frequency of all the InDel loci, a total of 12 rice varieties whose F i ranged from 0 to 1 were selected as the parental lines (Table 1). R465 and Balilla, two rice accession resources, were introduced from India and Italy, respectively, by the Chinese government. The remaining species are native to low and high latitudes in China and some are outstanding restorer lines, such as Mianhui725, 610234, and 9311K. All of the selected lines have good agronomic traits and are suitable for breeding programs. The phenotypes of the twelve rice varieties are consistent with the F i value from our previous study. Each of the twelve selected inbred lines were planted in the experimental field of the Hybrid Rice Hainan Experimental Base of Wuhan University in Lingshui (N18u 309 22.120, E110u 29 10.720), Hainan Province, in December 2011. A complete diallel mating design was used to obtain the hybrids. Total of 132 bag hybrids were obtained from March to May 2012.
The seeds of twelve rice varieties and 132 hybrids were bagged with transparent plasmic bags and buried in water at 28uC for 48 h. These seeds were then transferred to an incubator at a constant temperature of 28uC for 24 h. All of the seeds were planted in soil in the experimental field of the Hybrid Rice Ezhou Experimental Base of Wuhan University in Ezhou (N30u 229 19.820, E114u 449 59.170), Hubei Province, on May 13, 2012 and were transplanted on June 10, 2012. A randomized block design with three replications was applied. A total of 10 individual plants of each replicate were planted at a spacing of 16.5626.4 cm. Four plants from a cytoplasmic male-sterile line named YTA were located around these individuals to minimize the marginal effect. The field managements were applied as recommended. Due to low germination, the seedling numbers of some combinations were not sufficient for statistical analyses, and therefore, 122 hybrid combinations remained. Five plants in the middle of each repeat were chosen to collect the field data. All of the parental lines and the hybrid plant height (from the surface of the soil to the highest part of the plant) and tiller number were measured during the maturation stage. The grain weight per plant (GWP) was weighed. The grain number per plant and the empty grain number were measured using a Seed Counting Machine (PME-1 Seed Autocounting Machine, Shanke Equipment, Shanghai, China). The grain number per plant divided by the tiller (tillers with grain) number per plant (TPP) determined the grain number per panicle (GNP). The seed setting rate (SSR) was calculated by dividing the full grain number per plant by the total grain number per plant. Means over replications were calculated for each trait and used in the data analysis.
The GDI of the parental lines per hybrid was calculated as GDI = |F ia 2 F ib |, where F ia is parent a9s F i , and F ib is parent b9s F i . The relative low-parent heterosis (LPH) was determined using the equation LPH = (F 1 2 P low )/P low . The relative mid-parent heterosis (MPH) was determined using the equation MPH = (F 1 2 MP)/MP. The relative better-parent heterosis (BPH) was determined using the equation BPH = (F 1 2 P high )/P high . F 1 represents the hybrid data; P low represents the lower data of the parent. MP represents the means of the parent. P high represents the higher data of the parent. All of the trait data analyses were performed using the software SPSS 19.0 (SPSS, Chicago, Illinois, USA).

Indica-japonica inter-subspecific hybridization acquired heterosis in yield, yield-related agronomic traits and plant height
Each of the five traits exhibited heterosis ( Table 2). The grain weight per plant has the strongest relative mid-parent heterosis, which is approximately 24%, consistent with previous reports [14]. Regarding the remaining three yield-related traits, the highest average value of the relative mid-parent heterosis was the tiller number per plant, which was as high as 15.38%. The lowest was the thousand grain weight of 5.10%, and the grain number per panicle was 10.29%. The magnitudes of heterosis in these three traits were all smaller than the grain weight per plant. Considering that the yield is a complex trait, these results may indicate that the heterosis of yield may be a cumulative outcome of yield-related traits. Furthermore, as a typical nutritional trait, the relative midparent heterosis of plant height was as high as 13.45%. The highest percentage of the hybrids displaying positive relative mid-  parent heterosis (MPH$0) was 84.68% in plant height. The remaining three yield-related traits were greater than 70%, and the grain weight per plant was 68.03%. These results demonstrate that heterosis is acquired through inter-subspecific hybridization both in reproductive traits and nutritional trait. We then analyzed the relationships between the GDI and the relative mid-parent heterosis of grain number per plant, tiller per plant, thousand grain weight and plant height (Figure 1). Except for the thousand grain weight, the remaining three traits showed a significant increasing trend of heterosis, demonstrating that the larger the genetic divergence, the higher the degree of heterosis. Therefore, Figure 2. Linear regression between the seed setting rate and the GDI. SSR and GDI represent the seed setting rate and genetic divergence index, respectively. doi:10.1371/journal.pone.0093122.g002 the application of indica-japonica inter-subspecific hybridization seems to be a practical method to improve the rice yield.
The seed setting rate was closely associated with the GDI of indica-japonica parental lines The indica and japonica varieties are two subspecies of Asiancultivated rice. Through evolution, reproductive isolation came to exist in their hybrid offspring. In this study, when the GDI increased, indicating that the degree of reproductive isolation between the parents became more serious, their hybrids' seed setting rate decreased dramatically ( Figure 2). Figure 3 shows the changes in the seed setting rate as the GDI increased. When the parents were closely related, the seed setting rates were greater than 80% (Figure 3a). If the GDI was close to 0.5, the seed setting rate decreased to nearly 60%, which could be regarded as semisterile. In addition, the seed setting rate could reach approximately 40% when the two parents were typical indica and japonica varieties (Figure 3c). The lowest seed setting rate in this study was 10% in the Mianhui725/Balilla combination. Additionally, in the linear regression model, the seed setting rate can be predicted using the equation SSR = 0.8520.37*GDI.

The grain weight per plant decreased significantly when the seed setting rate decreased
The seed setting rate is one of the most important factors in determining the grain weight per plant. A low seed setting rate of inter-subspecific hybridization has puzzled rice breeders for many years. Figure 4 depicts the relationships between the grain weight per plant and the seed setting rate. Hybrids with a high seed setting rate could achieve a high grain yield value. At a seed setting rate value of approximately 85%, the grain weight per plant peaked at approximately 80 g/plant. However, this example may be extreme in the population, and there were still some hybrids with low yield performance at a seed setting rate of 85%. Subsequently, when the seed setting rate reached 90%, the grain weight per plant was quite excellent. No combinations remained under the regression curve. Once the seed setting rate is high enough, the yield will be improved. Therefore, although the yield is a complex trait that is influenced by many agronomic traits, the seed setting rate is a key element in determining the final output. The relationship between these two traits was GWP = 0.735*ln(SSR)+48.751.  Table 3. Correlations between the GDI and the GWP, the relative heterosis of the GWP. Linear regression between the grain weight per plant and the GDI Because the seed setting rate was negatively correlated with the GDI, it was positively correlated with the grain weight per plant; therefore, we analyzed the relationship between the GDI and the grain weight per plant. The correlation results showed that the grain weight per plant and the GDI were significant at the 0.05 level (Table 3). A regression model was subsequently constructed to further investigate the relationship between the grain weight per plant and the GDI (Figure 5a). The regression equation was GWP = 36.08+34.294*GDI-45.342*GDI 2 , which is a quadratic function. The highest point was at GDI = 0.3783, indicating that the ideal GDI of the parents may be 0.3783. Then, the relationships between the GDI and the LPH, MPH, and BPH were analyzed (Figure 5b, c, d). It was surprising that all of these quadratic regression outcomes had approximately the same peak, at 0.37. We thus concluded that when the GDI of the parental lines is 0.37, the hybrids will have the highest probability of obtaining the best yield performance. Considering that cytoplasmic differences may influence the hybrids, this value may not be suitable for traditional breeding applications. We divided the total experimental population into two groups that were reciprocal combinations. Among these reciprocal combinations, those with higher values were classified into group 1, and those with lower values were in group 2. The same regression analyses were performed ( Figure 6). The ideal GDI was close to 0.37, indicating that this value is consistent in different cytoplasm with the same nuclear background.

Discussion
Experiments investigating the relationship between the parental lines and the hybrid performance were conducted over recent decades in order to make use of the heterosis of inter-subspecific hybrids [15,[28][29][30][31][32]. However, no practical theoretical method has been proposed to guide rice breeders for correctly selecting restorer lines for a sterile line. Currently, large amounts of traditional testing hybridizations are still being used, which are time-and energy-consuming.

Higher degree of heterosis from indica-japonica intersubspecific hybridization
Our results showed that the relative mid-parent heterosis of the grain weight per plant can be as high as approximately 24%, which is close to 25% [14]. This result may reflect the cumulative effects of heterosis on yield-related traits. The grain number per panicle, tiller per plant and thousand grain weight displayed increasing trends to some extent. The larger genetic distance, the higher the degree of heterosis. Inter-subspecific hybridization of rice will make a powerful contribution to solving the problems of a growing global population and decreasing arable land and meeting the demands of an environmentally friendly, low carbon output.
The importance of choosing an appropriate estimating method to determine the GDI of indica-japonica parental varieties Knowing that reproductive isolation is caused by the genetic divergence between parents, many studies have estimated the degree of differentiation between the indica-japonica rice varieties [4,[33][34][35][36], and efforts have also been made to determine the relationship between genetic divergence and heterosis [37][38][39]. Most reports correctly concluded that the larger the genetic distance between parents, the stronger the heterosis of the hybrids. However, in regard to the relationships between the yield and heterosis, people found it incomprehensible and obtained different and even contradictory outcomes. The reason for these discrep-ancies may be that the DNA markers chosen to classify the genetic distance of the parents were not suitable. Because there are so many different loci between the sequences of indica and japonica, these differences may not show the true genetic distance. Therefore, selecting an appropriate estimating tool is quite important.
Speciation is connected to reproductive isolation. In intersubspecific hybridizations, the seed setting rate of the hybrids should be a favorable indicator for the real genetic divergence of the parents. Therefore, the correlations between the genetic divergence and the seed setting rate must be verified to be true. In this study, when the GDI increased, the seed setting rate decreased synchronously (Figure 2, Figure 3), indicating that those InDel markers are feasible for and effective at estimating the genetic divergence.
Balance between a higher degree of heterosis and increased reproductive isolation in the inter-subspecific hybridization to achieve a maximum yield The seed setting rate is related to the reproductive isolation of inter-subspecies caused by speciation, and the speciation procedure also introduces heterosis. Therefore, we cannot ignore Figure 6. Linear regression between the GWP and the MPH-GWP of reciprocal combinations and the GDI. a. Quadratic regression between the GDI and the GWP of group 1; the peak is at GDI = 0.3645. b. Quadratic regression between the GDI and the MPH-GWP of group 1; the peak is at GDI = 0.378. c. Quadratic regression between the GDI and the GWP of group 2; the peak is at GDI = 0.3681. d. Quadratic regression between the GDI and the MPH-GWP of group 2; the peak is at GDI = 0.363. doi:10.1371/journal.pone.0093122.g006 reproductive isolation in exploiting heterosis. However, we can find a balance between these two aspects. All of the linear regression results were optimal at GDI = 0.37, and this value remained constant in the reciprocal combinations. At this point, the heterosis of inter-subspecies has been acquired, and reproductive isolation can be avoided. This result provides a theoretical guide to indica-japonica hybrid rice breeding. Rice breeders can evaluate the degree of indica-japonica differentiation (F i ) of the restorer lines and identify suitable indica and/or japonica sterile lines based on the GDI value of 0.37 to achieve hybrid rice combinations with an optimal yield. The strategic design for the breeding program of inter-subspecific hybrid rice will be timesaving and cost-effective.