**About the Editors**

**Katsuyuki Ichitani**, Associate Professor in Kagoshima University. Born in 1970 Graduated from Kyoto University, Japan, in 1993. Lecturer of Kagoshima University, Japan, from 1998. The degree of doctor was awarded in 1998. Associate Professor of Faculty of Agriculture, Kagoshima University, Japan, from 2004. Major academic field is the genetics of agronomic traits and reproductive barrier in rice

**Ryuji Ishikawa**, Professor in Hirosaki University

Born in 1962

Graduated from Hokkaido University, Japan in 1985.

1988 Assistant Professor in Faculty of Agriculture, Hirosaki University.

The degree of doctor was awarded in 1993.

1993 Associate Professor in Faculty of Agriculture, Hirosaki University.

1997 Associate Professor in Faculty of Agriculture and Life Science, Hirosaki University.

2008 Professor in Faculty of Agriculture and Life Science, Hirosaki University.

Major academic field is rice genetics, evolutionary genetics.

## **Preface to "Genetics in Rice"**

Rice feeds more than half of the world population. Its small genome size and ease in transformation have made rice the model crop in plant physiology and genetics. Molecular as well as Mendelian, forward as well as reverse genetics, collaborate with each other to expand rice genetics. The syntety of rice with other grasses, such as wheat, barley and maize, has helped accelerate their genomic studies.

The wild relatives of rice belonging to the genus Oryza are distributed in Asia, Africa, Latin America and Oceania. Phenotypic and genetic diversity among them contributes to their adaptation to a wide range of environments. They are good sources for the study of domestication and adaptation.

Rice was the first crop to have its entire genome sequenced. With the help of the reference genome of Nipponbare and the advent of the next generation sequencer, the study of the rice genome has been accelerated. Now, 3000 (3K) cultivar genome information, the pangenome information comprising the whole genes among rice as a species, and the genomes of wild relatives of rice are available.

The mining of DNA polymorphism has permitted map-based cloning, QTL (quantitative trait loci) analysis, GWAS (genome-wide association study), and the production of many kinds of experimental lines, such as recombinant inbred lines, backcross inbred lines, and chromosomal segment substitution lines. The genetics of agronomic traits and pest resistance has led to the breeding of elite rice cultivars.

Inter- and intraspecific hybridization among Oryza species has opened the door to various levels of reproductive barriers ranging from prezygotic—e.g., hybrid sterility, male sterility—to postzygotic —e.g., hybrid weakness, hybrid breakdown.

This Special Issue of Plants, Genetics in Rice https://www.mdpi.com/journal/plants/special issues/genetics rice, contains eleven papers on genetic studies of rice and its relatives utilizing the rich genetic resources and/or rich genome information described above.

> **Katsuyuki Ichitani, Ryuji Ishikawa** *Editors*

## *Article* **Rice Novel Semidwarfing Gene** *d60* **Can Be as E**ff**ective as Green Revolution Gene** *sd1*

**Motonori Tomita 1,\* and Keiichiro Ishimoto <sup>2</sup>**


Received: 3 September 2019; Accepted: 7 October 2019; Published: 30 October 2019

**Abstract:** Gene effects on the yield performance were compared among promising semidwarf genes, namely, novel gene *d60*, representative gene *sd1* with different two source IR8 and Jukkoku, and double dwarf combinations of *d60* with each *sd1* allele, in a Koshihikari background. Compared with the culm length of variety Koshihikari (mean, 88.8 cm), that of the semidwarf or double dwarf lines carrying Jukkoku\_*sd1*, IR8\_*sd1*, *d60*, Jukkoku\_*sd1* plus *d60*, or IR8\_*sd1* plus *d60* was shortened to 71.8 cm, 68.5 cm, 65.7 cm, 48.6 cm, and 50.3 cm, respectively. Compared with the yield of Koshihikari (mean, 665.3 g/m2), that of the line carrying Jukkoku\_*sd1* allele showed the highest value (772.6 g/m2, 16.1% higher than Koshihikari), while that of IR8\_*sd1*, *d60* and IR8\_*sd1* plus *d60*, was slightly decreased by 7.1%, 5.5%, and 9.7% respectively. The line carrying Jukkoku\_*sd1* also showed the highest value in number of panicles and florets/panicle, 16.2% and 11.1% higher than in Koshihikari, respectively, and these effects were responsible for the increases in yield. The 1000-grain weight was equivalent among all genetic lines. Except for the semidwarf line carrying Jukkoku\_*sd1*, semidwarf line carrying *d60* was equivalent to line carrying IR8\_*sd1*in the yield of unpolished rice, and yield components such as panicle length, panicle number, floret number /panicle. Therefore, the semidwarfing gene *d60* is one of the best possible choices in practical breeding.

**Keywords:** rice; semidawarf gene; *d60*; *sd1*; yield component; phenotyping; growth

#### **1. Introduction**

Semidwarfing prevents plants from lodging at their full-ripe stage, making them lodging-resistant to wind and rain, enhances their adaptability for heavy manuring and markedly improved the global productivity of rice and wheat between 1960–1990 (up to double yields of rice and quadruple yields of wheat) [1,2]. Semidwarf rice contributes stable production in the monsoonal regions of Asia, where typhoons frequently occur during the yielding season and also brings benefits such as erect leaf angle, reduced photoinhibition, and possibility to plant at higher densities to japonica varieties grown in California and also in South America [3]. However, gene source of semidwarfness is limited. The International Rice Research Institute (IRRI) developed a semidwarf rice variety IR8 in 1966 by using Taiwanese native semidwarf variety Dee-geo-woo-gen (DGWG). IR8 called as Miracle Rice, has been improved with lodging resistance and light-reception attitude, and it brought the Green Revolution in tropical Asia [2]. In Japan, semidwarf cultivars in the Kyushu region were developed in the 1960s using the native semidwarf variety Jukkoku [4]. In the Tohoku region, semidwarf cultivars were developed in the 1970s using the semidwarf mutant Reimei induced by Fujiminori-gamma-ray irradiated [5]. In the United States, Calrose 76 was developed in 1976 by Calrose-gamma-ray irradiated [6,7].

Genetic study has also been devoted on the genes responsible for semidwarfism in rice. First, a recessive semidwarf gene *d47* was identified in DGWG, the parental line of IR8 [8,9]. Next, the

semidwarf gene *sd1* in Calrose 76 was shown to be allelic to *d47* [10,11]. Finally, semidwarf genes in Taichung Native 1 descend from DGWG, Shiranui from Jukkoku, and *d49* in the mutant cultivar Reimei were attributed to the same allele by allelism examination [12–14]. Therefore, only a single semidwarf gene, *sd1*, has been commonly used across the world. A little genetic source of current semidwarf rice cultivars have a risk for environmental change. Thus, it is an emerging subject to find a novel semidawrf gene to replace *sd1* and to utilize it to diversify genetic variations of semidwarf rice worldwide.

A novel semidwarf gene, *d60*, which was found in the mutant Hokuriku 100 induced by irradiation of 20 kR of gamma-ray to Koshihikari, is thus of particular importance [15]. While *sd1* is on rice chromosome 1 [16,17], *d60* is located on chromosome 2 (Tomita et al., submitted to Genes). *sd1* is a defective allele encoding GA20-oxidase gene in a late step in the GA biosynthesis pathway [18–20]. Moreover, unlike *sd1*, *d60* complements the gamete lethal gene, *gal*. Therefore, in the cross between Hokuriku 100 (*d60d60GalGal*) and Koshihikari (*D60D60galgal*), male and female gametes, in which *gal* and *d60* coexistent, become lethal and the pollen and seed fertility in the F1 (genotype *D60d60Galgal*) breakdown to 75%. As a result, the F2 progeny exhibits a unique genotype ratio of 6 fertile long-culm (4*D60D60*:2*D60d60GalGal*: 2 partially fertile long-culm (*D60d60Galgal* = F1 type):1 dwarf(*d60d60GalGal*) [15].

Although there are multiple alleles in *sd1* locus of DGWG, Jukkoku, Reimei, and Calrose 76, the differences in their influences on the yield performance have not been reported. Therefore, investigating the differences in phenotypic traits among the different *sd1* allele-carrying plants, *d60*-carrying plant and their double dwarf plants, will be beneficial for practical selection of *d60* and *sd1*alleles. In this study, semidwarf or double dwarf lines, which were integrated with *sd1* of Jukkoku, *sd1* of IR8, *d60,* or both gene combinations in the genetic background of Koshihikari, were used for investigating the influence of these semidwarf genes on phenotypic traits, especially related to yield performance.

#### **2. Results**

#### *2.1. E*ff*ects of Semidwarf and Double Dwarf Genes on Growth*

The trends in full-length growth, depicted by growth curves, were comparable among all lines. (Figure 1). The full length in lines carrying one or two semidwarf genes was already shorter than that of Koshirikari lines at the time of transplanting (June 7, 28 days after sowing), and the differences became prominent around 64–70 days after sowing (July 13 and 19) (Figure 1, Table 1). The full length of *d60*-carrying line was longer than that of *sd1*-carrying lines at the time of transplanting. However, the full length of line carrying Jukkoku\_*sd1* and that of line carrying IR8\_*sd1* exceeded that of line carrying *d60* on June 23 (43 days after sowing) and on July 13 (64 days after sowing), respectively: full length in lines carrying either Jukkoku\_*sd1* or IR8\_*sd1* was longer than that in line carrying *d60* at the time of final measurement (August 23, 103 days after sowing). Days to heading ranged from 86.5 days of line carrying IR8\_*sd1* to 90.5 days of those carrying *d60*. Such a four-day difference was thought to be little. Therefore, the differences appeared in morphological traits, such as culm length and panicle length, were attributed to genetic reason.

**Figure 1.** Effect of growth of semidwarf and double dwarf gene lines. Ten plants were randomly selected, and the distance between the ground and the highest standing point (i.e., the full length) was measured every week for approximately three months until the panicle emerged. The full length of *d60*-carrying line was longer than that of *sd1*-carrying lines at the time of transplanting. However, the full length of line carrying Jukkoku\_*sd1* and that of line carrying IR8\_*sd1* exceeded that of line carrying *d60* on June 23 (43 days after sowing) and on July 13 (64 days after sowing), respectively: full length in lines carrying either Jukkoku\_*sd1* or IR8\_*sd1* was longer than that in line carrying *d60* at the time of final measurement (August 23, 103 days after sowing).


**Table 1.** Plant length of semidwarf and double dwarf gene lines.

The full length in lines carrying one or two semidwarf genes was already shorter than that of Koshirikari lines at the time of transplanting 28 days after sowing, and the differences became prominent around 64–70 days after sowing. Finally, the full length of semidwarf and double dwarf lines were significantly shorter than Koshihaikri. Color in the boxes of genetic lines coincide the color of growth curve in Figure 1. \*: statistically significant at the 5% level.

Integration of a semidwarf gene (or genes) resulted in a reduction in culm length: the mean culm length of Koshihikari was 88.8 cm, while that of lines carrying Jukkoku\_*sd1*, IR8\_sd1, *d60*, Jukkoku\_*sd1* plus *d60*, or IR8\_*sd1* plus *d60* was 71.8 cm, 68.5 cm, 65.7 cm, 48.6 cm, or 50.2 cm, respectively. Leaf length was shorter in line carrying Kinuhikari\_*sd1* (9–16% reduction compared with Koshihikari) or *d60* (9–18% reduction compared with Koshihikari) than in those carrying Jukkoku\_*sd1* (1–9% reduction compared with Koshihikari (Figure 2). Furthermore, leaves of the semidwarf and double dwarf lines were slightly shorter and straighter (pointing upwards) than in Koshihikari (Figure 3), indicating improved light-reception attitude by the integration of semidwarf gene (or genes). Panicle length was slightly longer (by 2.5%) in line carrying Jukkoku\_*sd1* and slightly shorter in lines carrying Kinuhikari\_*sd1* (by 2.4%) or *d60* (by 3.0%), compared with Koshihikari (Table 2). However, the reduction in panicle length was quite less than that in culm length (22.8% decrease in lines carrying Kinuhikari\_*sd1* vs a 26.1% decrease in lines carrying *d60*). Therefore, the negative effects of semidwarf genes *sd1* and *d60* on panicle length were negligible.

**Figure 2.** Effect of semidwarf and double dwarf genes to leaf length. Upper five leaves, arising from the main culm, were measured. Except for Jukkoku\_*sd1* line. leaves of the semidwarf and double dwarf lines were slightly shorter than that of Koshihikari.

**Figure 3.** Plant phenotype of semidwarf and double dwarf gene lines. Leaves of the semidwarf and double dwarf lines were straighter (pointing upwards) than in Koshihikari, indicating improved light-reception attitude by the integration of semidwarf gene (or genes).

#### *2.2. E*ff*ects of Semidwarf and Double Dwarf Genes on Yield*

The yield components of each genotype are summarized in Table 2. The weight of unpolished rice/1000 grains and the proportions of fertile florets differed only slightly between lines. The effect of these genes on the proportion of fertile florets and the weight of unpolished rice/1000 grains were negligible. The number of panicles/plants was 17.9 in Koshihikari: 20.8 in line carrying Jukkoku\_*sd1* (+16.2% vs Koshihikari), and 15.4 in Jukkoku DW line (−14.0% vs Koshihikari) (Figure 4, Table 3). In addition, the floret number/panicle was 87.3 in line carrying Jukkoku\_*sd1* (+11.1% vs Koshihikari) and 72.1 in Jukkoku DW line (−8.3% vs Koshihikari) (Figure 5, Tables 2 and 4). The number of panicles was larger in line carrying Jukkoku\_*sd1*, while floret density was larger in all semidwarf varieties than in Koshihikari (Figure 6, Table 5). Thus, an increase in both the number of panicles/plant and the floret number/panicle resulted in an increase in the number of panicles/m<sup>2</sup> and a consequent increase in yield (Figure 7, Table 6).

**Table 2.** Effect of semidwarf and double dwarf genes to yield components.


Compared with the yield of Koshihikari (mean, 665.3 g/m2), that of the line carrying Jukkoku\_*sd1* was highest value 772.6 g/m<sup>2</sup> increased by 16.1%, while that of IR8\_*sd1*, *d60* and IR8\_*sd1* plus *d60*, was slightly decreased by 7.1%, 5.5%, and 9.7%, respectively. The line carrying Jukkoku\_*sd1* also showed highest value in number of panicles and florets/panicle, each 16.2% and 11.1% higher than in Koshihikari, which were responsible for the increases in yield. The weight of rice/1000 grains was equivalent among all genetic lines. Except for the semidwarf line carrying Jukkoku\_*sd1*, semidwarf line carrying *d60* was equivalent to line carrying IR8\_*sd1*in the yield of unpolished rice, and yield components such as panicles/m2, floret number /panicle. \*: statistically significant at the 5% level.

**Figure 4.** No. of panicles/plant in semidwarf and double dwarf gene lines. The number of panicles/plants was highest at 20.8 in line carrying Jukkoku\_*sd1* (+10.2% vs Koshihikari).


**Table 3.** Effect of semidwarf and double dwarf genes to No. of panicles/plant.

The number of panicles/plants in line carrying *d60* (17.9) was comparable to that in line carrying IR8\_*sd1* (17.7). \*: statistically significant at the 5% level.

㻕 㻜 **Figure 5.** Floret number/panicle of semidwarf and double dwarf gene lines. The number of panicles/ plants in line carrying *d60* was comparable to that in line carrying IR8\_*sd1*.


**Table 4.** Effect of semidwarf and double dwarf genes to floret number/panicle.

The floret number/panicle was highest at 87.3 in line carrying Jukkoku\_*sd1* (+11.1% vs Koshihikari). \*: statistically significant at the 5% level.

**Figure 6.** Effect of semidwarf and double dwarf genes to floret density. The floret density was larger in all lines carrying one semidwarf gene than that of Koshihikari.


**Table 5.** Effect of semidwarf and double dwarf genes to panicle.

Panicle length was slightly longer (by 2.5%) in line carrying Jukkoku\_*sd1* and slightly shorter in lines carrying IR8\_*sd1* (by 2.4%) or *d60* (by 3.0%), compared with Koshihikari. The reduction in panicle length was quite less than that in culm length (22.8% decrease in line carrying IR8\_*sd1* vs a 26.1% decrease in line carrying *d60*).

**Figure 7.** Yield of semidwarf and double dwarf gene lines. The yield of unpolished rice was 665.3 g/m<sup>2</sup> in Koshihikari, 772.6 g/m2 in line carrying Jukkoku\_*sd1* (+15.9% vs Koshihikari), 617.9 g/m<sup>2</sup> in line carrying Kinuhikari\_*sd1* (−7.1% vs Koshihikari), and 628.5 g/m<sup>2</sup> in line carrying *d60* (−5.5% vs Koshihikari).


**Table 6.** Effect of semidwarf and double dwarf genes to yield.

The yield of line carrying *d60* was comparable to that in line carrying IR8\_*sd1*. \*: statistically significant at the 5% level.

The yield of unpolished rice was 665.3 g/m<sup>2</sup> in Koshihikari, 772.6 g/m2 in line carrying Jukkoku\_*sd1* (+15.9% vs Koshihikari), 617.9 g/m2 in line carrying Kinuhikari\_*sd1* (−7.1% vs Koshihikari), and 628.5 g/m<sup>2</sup> in line carrying *d60* (−5.5% vs Koshihikari) (Figure 7, Table 6). The introduction of Kinuhikari\_*sd1* or *d60* into Koshihikari appears to cause a slight reduction in yield. On the other hand, the yield of DW lines was markedly lower than that of Koshihikari: for Jukkoku DW line (−24.5% vs Koshihikari) and 600.5 g/m2 for IR8 DW line (−9.7% vs Koshihikari) (Figure 7, Table 6). When using the alternative equation, the yield index was higher in all semidwarf-gene-carrying lines than in Koshihikari (Figure 8). The high yield index and lodging resistance of semidwarf varieties suggest that introduction of *sd1* and *d60* into non-dwarf genomes will be beneficial for increasing crop yield. Moreover, only minor differences in the grain appearance were observed among lines including Koshihikari, indicating that the grain quality of semidwarf lines is equivalent to that of Koshihikari. Taken together, semidwarf genes *sd1* and *d60* are useful in the agricultural industry.

The yield index was higher in line carrying IR8\_*sd1* than in those carrying Jukkoku\_*sd1* (Figure 8, Table 7). Although line carrying Jukkoku\_*sd1* gave a higher yield than those carrying IR8\_*sd1*, the higher yield index associated with IR8\_*sd1* than Jukkoku\_*sd1* suggests that the efficiency of the distribution to sink organs (e.g., seeds) is higher. Thus, the yield favors a gain in dry matter, which may be also higher in plants carrying IR8\_*sd1* than in those carrying Jukkoku\_*sd1*. Furthermore, the yield index for Jukkoku DW and IR8 DW are high—-45.6% and 46.5% higher, respectively than that of Koshihikari (Figure 8, Table 7). In order to increase a markedly low yield in DW lines, the use of conditions that favor a gain in dry matter, such as intensive cultivation with heavy fertilization to increase the number of tillers, may be effective.

**Figure 8.** Yield index of semidwarf and double dwarf gene lines. The yield index was higher in line carrying IR8\_*sd1* than in those carrying Jukkoku\_*sd1.* The yield index for Jukkoku DW and IR8 DW are 45.6% and 46.5%, respectively higher than that of Koshihikari.


**Table 7.** Effect of semidwarf and double dwarf genes to yield index.

The high yield index and lodging resistance of semidwarf varieties suggest that introduction of *sd1* and *d60* into non-dwarf genomes will be beneficial for increasing crop yield. \*: statistically significant at the 5% level.

#### **3. Discussion**

As exemplified by IR8, which was the variety behind the Green Revolution, many of the rice varieties cultivated worldwide commonly carry the semidwarf gene *sd1*. Another semidwarf gene *d60* is non-allelic to *sd1* and is of particular interest as a different source of semidwarfism to give genetic diversity among the semidwarf varieties. In this study, semidwarf lines, namely Jukkoku (Jukkoku \_*sd1*), *sd1* of Kinuhikari (IR8\_*sd1*: Kinuhikari maintains *sd1* of IR8 origin), *d60* or *sd1* plus *d60* into the Koshihikari background, were used to investigate influence of these semidwarf genes on phenotypic traits, in relation to yield.

This study showed that all tested semidwarf lines had shorter culm lengths than Koshihikari, indicating improved lodging resistance. The effect on culm length carrying *d60*(65.7 cm) is slight shorter than in those carrying *sd1* (Jukkoku\_*sd1*, 71.8 cm; IR8\_*sd1*, 68.5 cm), Among the genetic lines, line carrying Jukkoku\_*sd1*showed the highest yield of unpolished rice 772.6 g/m2, which is 16.1% higher than in Koshikikari. The Jukkoku\_*sd1* line also showed highest value in the number of panicles, the number of florets per panicle than in Koshihikari. Therefore, it was highly possible that the increasing yield of Jukkoku\_*sd1* line was ascribed to the increasing numbers of panicles and florets. Although the yield of unpolished rice of *d60* line, 628.5 g/m<sup>2</sup> was 5.5% lower than that of Koshihikari (665.3 g/m2), but this is almost equivalent yield performance of IR8\_*sd1* (617.9 g/m2). Ogi et al. (1993) [21] and Murai et al. (2004) [22] reported characteristics of isogenic line carrying *sd1* derived from DGWG, the source of IR8 *sd1*. These isogenic lines showed almost same number of panicles as that of original varieties, 'Norin 29 and 'Shiokari'. Therefore, it was concluded that Jukkoku\_*sd1* especially has potential increasing panicle numbers compared to IR8\_*sd1*. Hence, Jukkoku\_*sd1* appears to confer a pleiotropic effect of increasing panicle number very well in the Koshihikari genetic background. The difference of such as effect between *sd1* alleles may be ascribed to that IR8\_*sd1* suffered 383 bp deficit in the region exon 1-2 of *GA20-ox* [18], whereas, Jukkoku\_*sd1* has only a SNP against the wild type *GA20-ox* [18] and the transcripts existed [23].

This study demonstrated that *d60* confers slightly shorter culm length than IR8\_*sd1*, but almost equivalent yield performance with IR8\_*sd1* together with effects on yield-related phenotypic traits comparable to IR8\_*sd1*, which actually contributed to green revolution [2]. Although many dwarf genes are accompanied with a reduction in panicle length, yield of unpolished rice, and grain thresh ability (which is likely attributed to excessive dwarfing effects), *d60* does not exert such negative effects on yield-related phenotypic traits of rice plant. In conclusion, *d60* is applicable to practical breeding and one of choice for expanding genetic diversity of rice varieties.

#### **4. Materials and Methods**

#### *4.1. Genetic Lines*

The following rice semidwarf or double dwarf lines, Koshihikari, Koshihikari carrying Jukkoku\_*sd1* [Koshihikari\*6//(Kanto 79/Jukkoku F4) B6F4], Koshihikari carrying IR8\_*sd1* [Koshihikari/Kinuhikari F5)], Koshihikari carrying *d60* [Koshihikari Koshihikari\*7//(Koshihikari/Hokuriku 100) B7F3], Koshihikari carrying *d60* and Jukkoku\_*sd1* [Jukkoku\_DW, Koshihikari carrying Jukkoku\_*sd1*/Koshihikari carrying\_*d60* F7), and Koshihikari carrying *d60* and IR8\_*sd1* (IR8 DW: Koshihikari carrying IR8\_*sd1*/ Koshihikari carrying\_*d60* F7) were used in this study. Koshihikari carrying Jukkoku\_*sd1* was developed by six times of backcrosses with Koshihiakri as a recurrent parent using the short stemmed *sd1* homozygous fixed F4 strain in Kanto No. 79 × Jukkoku (*sd1*) as a non-recurrent parent [24]. Koshihikari carrying *d60* was developed by seven times of backcrosses with Koshihikari as the recurrent parent using the short-stemmed F2 plant as a non-recurrent parent [15]. Koshihikari carrying IR8\_\_*sd1* was the short stemmed *sd1* homozygous fixed F7 strain derived from Koshihikari × Kinuhikari. Koshihikari carrying *d60* and Jukkoku\_*sd1*(Jukkoku DW) was double dwarf *d60sd1* homozygous fixed F7 strain derived from Koshihikari carrying Jukkoku\_*sd1* × Koshihikari carrying\_*d60* [15]. Koshihikari carrying *d60* and IR8\_*sd1*(IR8 DW) was double dwarf *d60sd1* homozygous fixed F7 strain derived from Koshihikari carrying IR8\_*sd1* × Koshihikari carrying\_*d60* [15]. Kinuhikari has *sd1* derived from IR8 [25], which suffer 383 bp deficit in the region exon 1-2 from wild type *GA20-ox* [18]. Koshihikari carrying Jukkoku\_*sd1*was plant-variety registered via further 7–8th backcrosses and it was designated as Hikarishinseiki [24], whose *sd1* has only a SNP against wild type *GA20-ox* [18,26]. Genomic *sd1*allele and the RNA transcript in Hikarishinseiki are detectable by diagnosis targeting the SNP [23,27].

#### *4.2. Cultivation*

Rice seeds were taken from stocks kept in a refrigerator. Seeds of each line were immersed in enough water just to cover the seeds. Water was exchanged every day for seven days (May 2 to May 8) during seed soaking and stimulation of germination. Seedlings were grown in nursery boxes (30 × 15 × 3 cm) for approximately 20 days. Seedlings were then individually transplanted into a paddy field (120 m2: 6.0 × 20.0 m) of the University Farm on June 8. Two 4-m2 plots (2 × 2 m) with transplanting densities 22.2 seedlings/m2 (one seedling per 30 × 15 cm, 78 seedlings per field) were prepared for each genetic line (two instances). The paddy field was fertilized by 4.0 kg of basal fertilizer containing nitrogen, phosphorus, and potassium (weight ratio, nitrogen:phosphorus:potassium = 2.6:3.2:2.6) at the rate with 4.3 g/m<sup>2</sup> nitrogen, 5.3 g/m<sup>2</sup> phosphorus, and 4.3 g/m2 potassium evenly across the field. A herbicide (Joystar L Floable, Kumiai Chemical Industry, Tokyo, Japan) was applied on June 20 to kill weeds grown uncontrollably, and the water level was then kept at a high enough level to cover the weeds for a week.

#### *4.3. Growth Analysis*

Ten seedlings were randomly selected for each line at the time of transplantation, and the full length was measured individually. After transplantation, ten plants were randomly selected, and the distance between the ground and the highest standing point (i.e., the full length) was measured every week for approximately three months until the panicle emerged. The time when the tip of the panicle first emerged from the flag leaf sheath was recorded as the heading time for all plants.

#### *4.4. Plant Phenotyping*

After ripening, ten plants typical of each genotype were sampled twice. Sampled plants were air-dried, and were assessed or measured the following traits. Culm length: the length between the ground surface and the panicle base of the main culm was measured at the time of sampling. Leaf length: the lengths of the upper five leaves, arising from the main culm, were individually measured. Length and weight of panicle: the length between the panicle base and the tip of the panicle, and the weight of the panicle, were measured. Total panicle number: the number of panicles were counted by sampled individuals and panicle numbers per 1 m2 area (panicles/m2) were counted twice in each plot of the paddy field. Total floret number: florets were counted to obtain total floret number. Floret number/panicle: the total floret number (including both sterile and fertile florets) was divided by the total panicle number. Proportion of fertile florets: each floret was assessed to determine its fertility. Floret density (floret number/cm): the number of florets per panicle was divided by the length of the panicle. Presence of awns: florets with an awn were counted when counting the florets. Grain threshability: was manually tested during examination of phenotypic traits. Appearance of grains: the size, color, and presence of an awn were observed for assessment of grain quality. Total weight of winnowed paddy: the total weight of winnowed paddy was weighed after grain selection using the salt solution (salt content of 1.06 g/m3). Weight of sieved unpolished rice/1000 grains: obtained by multiplying the total weight of winnowed paddy by 0.84. Weight of plant parts above the ground: the weight of the plant parts above the ground was measured. Yield index: the winnowed paddy weight was divided by the weight of the plant part above the ground to obtain the yield index. The means of traits were statistically compared using the *t*-test.

#### *4.5. Yield*

Yield of unpolished rice was calculated using the following equation.

• Yield of unpolished rice (g/m2) = (number of panicles/m2) × (floret number/panicle) × (proportion of fertile florets) × (weight of unpolished rice/grain)

The following alternative equation (see below) was also used to calculate the comparison of yields:

• Yield = (yield index) × (weight of the plant parts above the ground)

**Author Contributions:** Conceptualization—M.T.; methodology—M.T.; investigation—K.I., M.T.; resources—M.T.; writing—original draft preparation—M.T.; writing—review and editing—M.T.; project administration, M.T.; funding acquisition—M.T.

**Funding:** This work is founded by Adaptable and Seamless Technology Transfer Program (A-STEP) through Target-driven R&D (high-risk challenge type) by Japan Science and Technology Agency (JST) to Motonori Tomita, whose project ID14529973 was entitled "NGS genome-wide analysis-based development of rice cultivars with super high-yield, large-grains, and early/late flowering suitable for the globalized world and global warming", since 2014 to 2018.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **The Development and Characterization of Near-Isogenic and Pyramided Lines Carrying Resistance Genes to Brown Planthopper with the Genetic Background of** *japonica* **Rice (***Oryza sativa* **L.)**

**Cuong D. Nguyen 1,2 , Holden Verdeprado <sup>3</sup> , Demeter Zita <sup>4</sup> , Sachiyo Sanada-Morimura <sup>5</sup> , Masaya Matsumura <sup>5</sup> , Parminder S. Virk 3,6 , Darshan S. Brar 3,7 , Finbarr G. Horgan <sup>8</sup> , Hideshi Yasui <sup>9</sup> and Daisuke Fujita 1,3,4,9,\***


Received: 18 September 2019; Accepted: 7 November 2019; Published: 12 November 2019

**Abstract:** The brown planthopper (BPH: *Nilaparvata lugens* Stål.) is a major pest of rice, *Oryza sativa*, in Asia. Host plant resistance has tremendous potential to reduce the damage caused to rice by the planthopper. However, the effectiveness of resistance genes varies spatially and temporally according to BPH virulence. Understanding patterns in BPH virulence against resistance genes is necessary to efficiently and sustainably deploy resistant rice varieties. To survey BPH virulence patterns, seven near-isogenic lines (NILs), each with a single BPH resistance gene (*BPH2*-NIL, *BPH3*-NIL, *BPH17*-NIL, *BPH20*-NIL, *BPH21*-NIL, *BPH32*-NIL and *BPH17-ptb*-NIL) and fifteen pyramided lines (PYLs) carrying multiple resistance genes were developed with the genetic background of the *japonica* rice variety, Taichung 65 (T65), and assessed for resistance levels against two BPH populations (Hadano-66 and Koshi-2013 collected in Japan in 1966 and 2013, respectively). Many of the NILs and PYLs were resistant against the Hadano-66 population but were less effective against the Koshi-2013 population. Among PYLs, *BPH20*+*BPH32*-PYL and *BPH2*+*BPH3*+*BPH17*-PYL granted relatively high BPH resistance against Koshi-2013. The NILs and PYLs developed in this research will be useful to monitor BPH virulence prior to deploying resistant rice varieties and improve rice's resistance to BPH in the context of regionally increasing levels of virulence.

**Keywords:** rice (*Oryza sativa* L.); brown planthopper; near-isogenic lines; pyramided lines; resistance; virulence

#### **1. Introduction**

The brown planthopper (BPH: *Nilaparvata lugens* Stål.) is a major pest of rice (*Oryza sativa* L.) in tropical and subtropical Asia [1]. BPH damages rice by sucking phloem from the plants (mechanical damage) or by transmitting viruses such as rice grassy stunt virus (RGSV), rice ragged stunt phytoreovirus (RRSV) and rice wilted stunt virus (RWSV) [2,3]. At high planthopper densities, rice crops display patches of desiccated rice known as 'hopperburn.' Insecticides have been widely used to reduce BPH populations [4]. However, insecticides are damaging to human health and the environment, and are increasingly recognized as contributing to BPH outbreaks through physiological and ecological pest resurgence mechanisms [1]. Host plant resistance is considered a potentially effective alternative to harmful insecticides, that reduces BPH damage without detrimental effects on the natural enemies of BPH [5].

To date, more than 34 BPH resistance genes have been identified from rice cultivars and wild rice species. Seven genes: *BPH9*, *BPH14*, *BPH17*, *BPH18*, *BPH26*, *BPH29* and *BPH32* have been cloned and characterized for different BPH resistance levels [6–12]. Four gene clusters (chromosomal regions) strongly associated with BPH resistance have been identified. These occur on chromosomes 4S (short arm), 4L (long arm), 6S and 12 [3,13]. Four genes (*BPH12* from B14, *BPH15*, *BPH17* and *BPH20*) and six quantitative trait loci (QTLs) (*QBph4, QBPH4.1, QBPH4.2, QBph4.2 qBph4.3* and *qBph4.4*) for BPH resistance have been identified on chromosome 4S [13–20]. Among those QTLs, *QBph4* (6.7–6.9 Mbp) from IR02W101 (*Oryza o*ffi*cinalis*) and *QBph4*.2 (6.6–6.9 Mbp) from IR65482-17 (*Oryza australiensis*) were identified at similar locations based on physical distance; *QBPH4.1* (5.8–7.8 Mbp) and *QBPH4.2* (15.2–17.2 Mbp) were identified from Rathu Heenati; *qBph4.3* (0.2–0.7 Mbp) and *qBph4.4* (0.7–13.1 Mbp) were detected in Salkathi. On chromosome 4L, five genes for BPH resistance—*BPH6*, *BPH12(t)* from GSK185-2, *BPH18(t)* from BPH2183, *BPH27* from GX2183 and *BPH27(t)* from Balamawee have been identified [21–25]. Six genes/QTLs have been detected on the short arm of chromosome 6: *BPH3*, *BPH4*, *BPH25*, *BPH32, Qbph6* and *qBPH6(t)* [12,26–30]. Eight genes have been identified on the long arm of chromosome 12—*BPH1*, *BPH2*, *BPH7*, *BPH9*, *BPH10* and *BPH18* from IR65482-7-216-1-2, and *BPH21* and *BPH26* [17,29,31–38]. The BPH resistance genes on chromosome 12 were classified into four allelic types based on their amino acid sequences and different resistance levels: type 1—*BPH1, BPH10, BPH18* and *BPH21*; type 2—*BPH2* and *BPH26*; type 3—*BPH7*; and type 4—*BPH9* [6]. A number of highly resistant rice breeding lines and varieties contain several genes with resistance to BPH and other phloem feeding Hemiptera. These include IR71033-121-15 introgressed from *Oryza minuta* carrying *BPH20*, *BPH21* and *qBPH6(t)* [17,39]; Rathu Heenati carrying *BPH3*, *BPH17*, *QBPH4.1* and *QBPH4.2* [8,13,27]; and PTB33 carrying *BPH2* and *BPH32* [12,40,41].

Monogenic resistance is vulnerable to rapid adaptation by BPH populations. Research indicates that BPH populations have sufficient genetic variability to enable them to overcome specific resistance genes when selected on a resistant host over multiple generations [4,42,43]. In the late 1970s, BPH populations adapted to varieties carrying the *BPH1* and/or *BPH2* genes after these were widely deployed in rice varieties across Asia [4,42,44]. A recent multi-national study has indicated that BPH populations across much of Asia have adapted to feed on rice carrying the *BPH1, BPH2, BPH5, BPH7, BPH8, BPH9, BPH10* and *BPH18* genes [45]. Under laboratory conditions, BPH populations continually reared for between seven to 15 generations on resistant rice varieties were capable of adapting to resistance from a range of genes, including *BPH1, BPH2, BPH3, BPH8, BPH9, BPH10* and *BPH32* [43,46–49]. Through adaptation to resistance genes, BPH acquires stronger virulence against resistance genes and BPH virulence remains stable for several decades [4,50]. Therefore, it is important to preserve the effects of resistance genes by preventing BPH adaptation.

To prevent further adaptation by BPH populations to available resistance genes, a strategy for deploying resistance based on insect virulence is necessary [4]. However, BPH virulence varies under different environments depending on the predominant rice cultivars, BPH migration routes, and the length of population exposure to different resistance genes [4]. Without exposing resistance genes to BPH populations under controlled conditions prior to deployment, the potential effectiveness of the resistance genes for target regions is difficult to predict. In previous studies, the virulence of BPH has been characterized using resistant varieties [45,48,51–53]. However, many BPH-resistant varieties have multiple resistance genes, such that the effects of any single resistance gene cannot be assessed using these varieties. In contrast, the effects of any single resistance gene may be revealed in detail by using near-isogenic lines (NILs) that carry the gene on the genetic background of a susceptible variety. Recently, more than 16 NILs with BPH resistance genes (*BPH3*, *BPH4*, *BPH6*, *BPH9*, *BPH10*, *BPH12*, *BPH14*, *BPH15*, *BPH17*, *BPH18*, *BPH20*, *BPH21*, *BPH25*, *BPH26, BPH30* and *BPH32*) have been developed on the genetic backgrounds of several *indica* and *japonica*-susceptible varieties. These NILs have been evaluated against different BPH populations from China, the Philippines and Japan [14,54–58].

Because the resistance of rice varieties carrying single genes is weaker and less durable (i.e., allowing rapid BPH adaptation) to BPH than varieties with multiple resistance genes, several researchers have proposed the pyramiding of two or more genes to enhance resistance levels and thereby avoid pest adaptation [59]. Combinations of multiple BPH resistance genes have been reported to increase levels of plant resistance to BPH. For example, a pyramided line (PYL) with *BPH14* and *BPH15* enhanced resistance against BPH from China compared to monogenic NILs with either *BPH14* or *BPH15* alone [60]. Similarly, the pyramided lines *BPH6* + *BPH12* PYL and *BPH3* + *BPH27* PYL exhibited greater resistance levels in bulk seedling tests than monogenic lines with each of the genes present alone [14,55], and *BPH17* + *BPH21* PYL had greater resistance against BPH in the Philippines than lines with either gene alone [57]. Pyramiding the *BPH25* and *BPH26* genes into a single rice line was reported to have positive epistatic effects against BPH populations collected in Vietnam, the Philippines and Japan [51,61]. Therefore, the development of rice varieties carrying multiple BPH resistance genes might be an effective way to enhance BPH resistance.

In this study, seven NILs with BPH resistance genes (*BPH2*, *BPH3*, *BPH17, BPH20*, *BPH21, BPH32* and *BPH17-ptb*) and a *japonica* rice genetic background were developed to evaluate the effects of different resistance genes on BPH populations. Based on the NILs developed, 15 pyramided lines (PYLs) carrying two or three resistance genes were developed to enhance levels of resistance against BPH. Additionally, using the NILs and PYLs we developed, the study compared resistance against two BPH populations collected in Japan: the first was collected in 1966 (before resistant varieties were widely released) and the second was collected in 2013 (recently migrated from China to Japan). Comparisons of the reactions by BPH from each population to the NILs and PYLs indicates the utility of resistance genes and their different combinations (some with epistatic effects) against modern BPH populations.

#### **2. Results**

#### *2.1. Development of Seven NILs for BPH Resistance*

Seven NILs with BPH resistance genes from three donor parents on the genetic background of Taichung 65 (T65) were developed through marker-assisted selection (MAS) and backcrossing (Tables 1 and 2). For three donor parents, IR71033-121-15 has *BPH20* and *BPH21*; Rathu Heenati contains *BPH3* and *BPH17*; and PTB33 carries *BPH2, BPH17-ptb* and *BPH32* based on previous studies. For PTB33, there has been no previous report of a BPH resistance gene on chromosome 4S. However, amino acid sequences for the *BPH17* locus in PTB33 were identical to those of Rathu Heenati [8]. Thus, we assume that PTB33 contains a gene for BPH resistance on chromosome 4S and tentatively named this as *BPH17-ptb*. The substituted chromosomal segments of the NILs were detected by polymorphic simple sequence repeat (SSR) markers that were equally distributed across the whole genome (Table 3; Figure 1). The genetic background of *BPH2*-NIL was analyzed using 203 polymorphic SSR markers. The ratio of substituted segments from PTB33 on *BPH2*-NIL was 9.1–14.8% (total 33.9–55.0 Mbp). One substituted segment with a size of 21.3–25.4 Mbp encompassing *BPH2* was detected between RM247 and RM5479 on chromosome 12. The other three segments were detected between RM5426 and

RM248 on chromosome 7 with a size of 3.4–4.2 Mbp, between RM5688 and RM444 on chromosome 9 with a size of 4.2–9.2 Mbp and between RM7492 and RM216 on chromosome 10 with a size of 5.0–16.2 Mbp.

The genetic background of *BPH3*-NIL was confirmed using 195 polymorphic SSR markers, and the ratio of substituted segment from Rathu Heenati was 1.0–3.0% (total 3.8–11.3 Mbp). One segment with a size of 1.6–1.8 Mbp including *BPH3* was detected between MSSR1 and RM1369 on the short arm of chromosome 6. The other substituted segments were detected between RM1359 and RM1155 on chromosome 4 with a size of 0.5–4.2 Mbp and between RM1345 and RM3155 on chromosome 8 with a size of 1.8–4.9 Mbp.

The genetic background of *BPH17*-NIL was surveyed using 173 polymorphic SSR markers. The ratio of substituted segments was 1.0–4.8% (total 3.8–17.6 Mbp) containing one segment located between RM8213 and B40 on chromosome 4, including the *BPH17* region.

The genetic background of *BPH17-ptb*-NIL was analyzed using 229 polymorphic SSR markers, and the ratio of substituted segments from PTB33 on *BPH17-ptb*-NIL was 2.8–7.6% (total 10.5–28.1 Mbp). One substituted segment with 5.8–13.1 Mbp encompassing *BPH17-ptb* was detected between C61009 and B40 on chromosome 4. Two other substituted segments were detected at RM3126 on chromosome 3 and between RM7048 and RM6971 on chromosome 9 (4.7–11.5 Mbp).

The genetic background of *BPH20*-NIL was confirmed using 237 polymorphic SSR markers and the ratio of substituted segments of IR71033-121-15 was 5.6–9.6% (total 20.6–35.5 Mbp). One segment with a size of 13.8–19.9 Mb containing *BPH20* was detected between RM335 and RM5900 on chromosome 4. Two other substituted segments were detected between RM224 and RM5926 on chromosome 11 (1.5–7.7 Mbp) and between RM7315 and RM3103 on chromosome 12 (5.3–7.9 Mbp).

The genetic background of *BPH21*-NIL was surveyed using 229 polymorphic SSR markers, and the ratio of substituted segments from IR71033-121-15 was 7.1–11.6% (total 26.4–43.1 Mbp). One segment with a size of 22.6–23.7 Mbp, including *BPH21*, was detected between RM1880 and RM28493 on chromosome 12. Three other segments were detected between RM6841 and RM3348 on chromosome 5 (2.3–7.2 Mbp), around RM1328 on chromosome 9 and between RM224 and RM5926 on chromosome 11 (1.5–7.7 Mbp).

The genetic background of *BPH32*-NIL was confirmed using 233 polymorphic SSR markers. The ratio of substituted segments of PTB33 on *BPH32*-NIL was 1.9–4.1% (total 7.1–15.1 Mbp). One segment with a size of 1.6–3.2 Mbp containing *BPH32* was detected between RM6775 and RM190 on chromosome 6. Three other segments from the donor parent were detected between RM5755 and RM3280 on chromosome 3 with a size of 4.9–8.1 Mbp, between RM1306 and RM248 on chromosome 7 with a size of 0.4–3.2 Mbp and between RM5349 and RM5961 on chromosome 11 with a size of 0.2–0.6 Mbp.

18




**Table 2.** Details of the seven near-isogenic lines and 15 pyramided lines carrying brown planthopper resistance genes.

**Table 3.** Background survey analysis of seven near-isogenic lines using SSR polymorphic markers.


\* The minimum physical distance of donor segment was calculated by the distance between two markers delimiting the substituted segment and the maximum amount was calculated by two flanking markers of substituted segments.

**Figure 1.** Graphical genotypes of *BPH2*-NIL (**A**), *BPH3*-NIL (**B**), *BPH17*-NIL (**C**), *BPH20*-NIL (**D**), *BPH21*-NIL (**E**), *BPH32*-NIL (**F**) and *BPH17-ptb*-NIL (**G**). The 12 bars indicate 12 chromosomes of rice. Horizontal lines across the chromosomes show the positions of polymorphic SSR markers. Circles indicate the approximate positions of brown planthopper resistant genes. The asterisks (\*) indicate SSR markers that were used for marker-assisted selection.

#### *2.2. Development of 15 PYLs Carrying Two or Three BPH Resistance Genes*

Twelve PYLs carrying two BPH resistance genes (BPH2 + BPH17-PYL, BPH2 + BPH25-PYL, BPH2 + BPH32-PYL, BPH2 + BPH17-ptb-PYL, BPH3 + BPH17-PYL, BPH17 + BPH21-PYL, BPH20 + BPH21-PYL, BPH20 + BPH32-PYL, BPH21 + BPH25-PYL, BPH21 + BPH17-ptb-PYL, BPH25 + BPH17-ptb-PYL and BPH32 + BPH17-ptb-PYL) and three PYLs containing three BPH resistance genes (BPH2 + BPH3 + BPH17-PYL, BPH2 + BPH32 + BPH17-ptb-PYL and BPH20 + BPH21 + BPH32-PYL) were developed using NILs and PYLs with BPH resistance gene(s) (Table 2). The PYLs were confirmed for resistance genes through foreground selection using flanking SSR markers tightly linked to each resistance gene. Most of PYLs were selected from the BC4F3 equivalent generation, except BPH3 + BPH17-PYL from the BC4F4 equivalent generation, BPH20+BPH21-PYL from the BC3F8 generation and BPH32 + BPH17-ptb-PYL from the BC3F8 generation.

#### *2.3. Comparison of Resistance Levels against Hadano-66 by Modified Seedbox Screening Test ( MSST)*

T65 was highly damaged (damage score (DS) = 8.2) by the Hadano-66 population (Figure 2A). The DSs of the donor parents were significantly lower (0.7 for IR71033-121-15, 0.7 for PTB33 and 0.2 for Rathu Heenati) than that of T65. The donor parents also had higher levels of resistance compared with their respective NILs and PYLs. Among the NILs, *BPH2*-NIL (DS: 3.0) and *BPH17*-NIL (3.2) showed the highest resistance levels. The other NILs *BPH3*-NIL (6.0), *BPH20*-NIL (6.0), *BPH21*-NIL (6.5), *BPH25*-NIL (6.7), *BPH26*-NIL (4.8), *BPH32*-NIL (6.7) and *BPH17-ptb*-NIL (5.7), had lower DSs than T65's but were not significantly different from the T65. Damage scores across the 15 PYLs ranged from 2.3 to 6.0. Among PYLs, the DSs of 10 PYLs—*BPH2* + *BPH17*-PYL (2.7), *BPH2* + *BPH25*-PYL (2.5), *BPH2* + *BPH32*-PYL (3.0), *BPH2* + *BPH17-ptb*-PYL (3.0), *BPH17* + *BPH21*-PYL (2.3), *BPH20* + *BPH21*-PYL (2.3), *BPH21* + *BPH25*-PYL (3.3), *BPH21* + *BPH17-ptb*-PYL (2.7), *BPH2* + *BPH3* + *BPH17*-PYL (3.0) and *BPH20* + *BPH21* + *BPH32*-PYL (2.3), were equal to or less than 3.3, while the DSs of five PYLs—*BPH3* + *BPH17*-PYL (5.0), *BPH20* + *BPH32*-PYL (5.3), *BPH25* + *BPH17-ptb*-PYL (6.0), *BPH32* + *BPH17-ptb* (5.0) and *BPH2* + *BPH32* + *BPH17-ptb*-PYL (4.3), were more than 4.3. Although the DSs between NILs and PYLs were not significantly different, the resistance levels of the PYLs tended to be higher than those of the NILs.

Additionally, fresh biomass reduction rates (FBRRs) of the NILs and PYLs were calculated as an indicator of resistance (Figure 2B). T65 had the highest FBRR (89.0%) and was significantly different from the donor parents: IR71033-121-15 (35.7%), PTB33 (39.2%) and Rathu Heenati (20.4%). Among the NILs, *BPH17*-NIL (58.7%) had the lowest FBRR and was significantly different from T65. The other NILs, *BPH2*-NIL (68.6%), *BPH3*-NIL (82.4%), *BPH20*-NIL (77.3%), *BPH21*-NIL (84.3%), *BPH25*-NIL (85.3%), *BPH26*-NIL (73.8%), *BPH32*-NIL (86.7%) and *BPH17-ptb*-NIL (77.6%), had lower FBRRs than T65; however, the differences were not significant. The FBRRs of four PYLs—*BPH2* + *BPH32*-PYL (59.1%), *BPH2* + *BPH17-ptb*-PYL (56.7%), *BPH21* + *BPH17-ptb*-PYL (50.1%) and *BPH2* + *BPH3* + *BPH17*-PYL (57.6%), were less than 60%. The FBRRs of five PYLs—*BPH2* + *BPH17*-PYL (64.5%), *BPH2* + *BPH25*-PYL (68.4%), *BPH20* + *BPH21*-PYL (62.3%), *BPH2* + *BPH32* + *BPH17-ptb*-PYL (64.0%) and *BPH20* + *BPH21* + *BPH32*-PYL (63.2%), ranged from 60% to 70%; and the FBRRs of six PYLs—*BPH3* + *BPH17*-PYL (79.3%), *BPH17* + *BPH21*-PYL (70.9%), *BPH20* + *BPH32*-PYL (71.6%), *BPH21* + *BPH25*-PYL (70.3%), *BPH25* + *BPH17-ptb*-PYL (78.9%) and *BPH32* + *BPH17-ptb*-PYL (74.4%), ranged from 70% to 80%. Additionally, DSs and FBRRs were positively correlated (Pearson's C = 0.89; *p* < 0.001).

**Figure 2.** Damage scores (**A**) and fresh biomass reduction rates (**B**) of near-isogenic lines and pyramided lines infested with the Hadano-1966 *Nilaparvata lugens* population using the modified seedbox screening test at the seedling stage. The lower damage scores and fresh biomass reduction rates indicate higher resistance levels.

#### *2.4. Comparison of Adult Mortality (ADM) of the Hadano-66 Population to the NILs and PYLs*

Levels of adult mortality (ADM) of the donor parents IR71033-121-15, PTB33 and Rathu Heenati (100%) were significantly higher than that of T65 (17.6%) (Table 4). Among the NILs, *BPH2*-NIL and *BPH17*-NIL had the highest ADM rates, 68.9% and 59.0%, respectively. The ADM rates of other NILs were not significantly different from that of T65. The PYLs carrying the *BPH2—BPH2* + *BPH17*-PYL (75.0%), *BPH2* + *BPH25*-PYL (87.5%), *BPH2* + *BPH32*-PYL (84.0%) and *BPH2* + *BPH17-ptb*-PYL (84.0%), showed the highest ADM rates among PYLs with two genes and were higher than any of the corresponding NILs. The ADM rates of seven PYLs—*BPH3* + *BPH17*-PYL, *BPH17* + *BPH21*-PYL, *BPH20* + *BPH32*-PYL, *BPH21* + *BPH25*-PYL, *BPH21* + *BPH17-ptb*-PYL, *BPH25* + *BPH17-ptb*-PYL and *BPH32* + *BPH17-ptb*-PYL, ranged from 50.0% to 68.0%, while the ADM of *BPH20* + *BPH21*-PYL was 33.3%. The ADM rates of PYLs for three genes—*BPH2* + *BPH3* + *BPH17*-PYL (96.0%), *BPH2* + *BPH32* + *BPH17-ptb*-PYL (95.8%) and *BPH20* + *BPH21* + *BPH32*-PYL (92.0%), were higher than those of the corresponding NILs and PYLs for two genes, and were similar to the ADM rates of the donor parents (100%). Furthermore, the ADM rates were negatively correlated with the DSs (Pearson's C = −0.79; *p* < 0.001) and FBRRs (Pearson's C = −0.76; *p* < 0.001).

**Table 4.** The adult mortality of *Nilaparvata lugens* on near-isogenic lines and pyramided lines carrying brown planthopper resistance genes.


Parameter values (means ± standard deviations) followed by the same letter are not significantly different between each *Nilaparvata lugens* population (*p* < 0.05, Tukey–Kramer multiple comparison tests).

#### *2.5. Comparison of ADM Rates for the Koshi-2013 Population on the NILs and PYLs*

T65 was susceptible to the Koshi-2013 population with an ADM of 5.0%. Rathu Heenati had the highest ADM among entries (84.0%), which was significantly higher than T65. The ADM rates of the other donor parents, IR71033-121-15 (44.0%) and PTB33 (36.0%), were lower than that of Rathu Heenati. The ADM rates of the NILs were less than or equal to 20.0%. Among the PYLs, the ADM rates of *BPH2* + *BPH17*-PYL (32.0%), *BPH20* + *BPH32*-PYL (36.0%) and *BPH2* + *BPH3* + *BPH17*-PYL (36.0%) were highest. The ADM rates of the six PYLs, *BPH3* + *BPH17*-PYL, *BPH20* + *BPH21*-PYL, *BPH21* + *BPH25*-PYL, *BPH32* + *BPH17-ptb*-PYL, *BPH2* + *BPH32* + *BPH17-ptb*-PYL and *BPH20* + *BPH21* + *BPH32*-PYL ranged from 20% to 28%. The ADM rates of the other PYLs, *BPH2* + *BPH32*-PYL, *BPH2* + *BPH17-ptb*-PYL, *BPH17* + *BPH21*-PYL, *BPH25* + *BPH17-ptb*-PYL, *BPH2* + *BPH25*-PYL and *BPH21* + *BPH17-ptb*-PYL ranged from 8.0% to 16.0%.

#### *2.6. Agronomic Characteristics of the NILs and PYLs*

Six agronomic traits—days to heading (DTH), panicle length (PL), culm length (CL), flag leaf length (LL), flag leaf width (LW) and panicle number per plant (PN) of the NILs and PYLs are presented in Table 5. The DTHs and PNs of the NILs and PYLs were not significantly different from those of T65. The PLs, CLs, LLs and LWs were similar for NILs and T65, except that the *BPH2*-NIL had longer culms, *BPH25*-NIL had shorter panicles and *BPH3*-NIL had wider flag leaves. The PLs, CLs, LLs and LWs were not significantly different between the PYLs and T65, except that *BPH2* + *BPH17*-PYL, *BPH2* + *BPH25*-PYL, *BPH2* + *BPH32*-PYL, *BPH2* + *BPH17-ptb*-PYL, *BPH21* + *BPH25*-PYL and *BPH2* + *BPH3* + *BPH17*-PYL had longer culms; *BPH21* + *BPH25*-PYL had longer flag leaves; *BPH2* + *BPH32*-PYL, *BPH17* + *BPH21*-PYL and *BPH20* + *BPH32*-PYL had narrower flag leaves; and *BPH3* + *BPH17*-PYL had wider flag leaves.

**Table 5.** Agronomic traits of near-isogenic lines and pyramided lines for brown planthopper resistance genes.


DTH: days to heading, CL: culm length, PL: panicle length, LL: flag leaf length, LW: flag leaf width, PN: panicle number per plant. \* *p* < 0.01; \*\* *p* < 0.05 (Dunnett's multiple comparison tests against Taichung 65).

#### **3. Discussion**

The seven NILs we developed carried BPH resistance genes on the short arm of chromosome 4 (*BPH17*-NIL, *BPH20*-NIL and *BPH17-ptb*-NIL), on the short arm of chromosome 6 (*BPH3*-NIL and *BPH32*-NIL) and on the long arm of chromosome 12 (*BPH2*-NIL and *BPH21*-NIL). One of the resistance genes on chromosome 12, *BPH2*, was originally identified from ASD7 which was used as a donor parent for many modern resistant varieties (e.g., IR36, IR42 and so on) [31,67]. *BPH2* from ASD7 is identical to *BPH26* in DNA sequence and resistance level [10]. *BPH2* from ASD7 was resistant against the Hatano-66 population (synonym of Hadano-66) but susceptible to Nishigoshi-05, a BPH population collected in Koshi, Kumamoto Prefecture in 2005 [51]. PTB33 was reported to carry one dominant and one recessive gene [40] that were confirmed to be *BPH2* and *BPH3* using conventional genetic analysis [41]. However, there was no report of the exact location of *BPH2* from PTB33. In our study, *BPH2*-NIL had similar resistance patterns to *BPH2* in ASD7: *BPH2*-NIL was highly resistant (ADM of 68.9%) against the Hadano-66 population but less effective (ADM of 4.0%) against the recently collected population, Koshi-2013. Moreover, *BPH2*-NIL (PTB33) and *BPH26*-NIL had similar resistance levels against both Hadano-66 and Koshi-2013, suggesting that PTB33, ADR52 and ASD7 might harbor the same resistance gene. Further sequence analysis for *BPH2* from PTB33 is necessary to understand its genetic basis. Another gene on chromosome 12, *BPH21,* was originally identified from IR71033-121-15, an introgression line derived from *O. minuta* and estimated to be located between two markers, S12094A and B122, on the long arm of chromosome 12 [17]. Recently, *BPH21* has been reported to be allelic to *BPH26* [6] and *BPH18* [68] based on amino acid sequences. Both *BPH18* and *BPH26* were isolated and located at 22.9 Mbp on chromosome 12 [9,10]. Therefore, we estimated that the location of *BPH21* was around 22.9 Mbp on chromosome 12, and the region carrying *BPH21* from IR71033-121-15 was selected using RM1246 (19.2 Mbp) and RM28493 (23.3 Mbp) in this study.

The *BPH17* locus on chromosome 4S from Rathu Heenati has been reported by Sun et al. (2005) [16]. *BPH17* was mapped between two markers, RHD9 (6.2 Mbp) and RHC10 (7.0 Mbp), on chromosome 4S and isolated by Liu et al. (2014) [8]. The amino acid sequence and chromosomal location of *BPH17* from Rathu Heenati were the same as those of *BPH17-ptb* from PTB33 [8]. In this study, resistance of *BPH17*-NIL and *BPH17-ptb*-NIL against the Hadano-66 population differed; however, both NILs had similar effects on the Koshi-2013 population. The different resistant levels might be because the loci were derived from different accessions or varieties of rice. Therefore, the amino acid sequences of PTB33 and Rathu Heenati used in this study on the *BPH17* locus should be determined for future research. Additionally, *BPH20* was detected between two markers, B42 (8.7 Mbp) and B44 (8.9 Mbp) on chromosome 4 [17]. Two NILs for *BPH17* and *BPH20* on the genetic background of 9311 varieties developed by Xiao et al (2016) [51] showed different resistance levels against a BPH population from China [68]. In our study, the resistance levels of *BPH17* and *BPH20* were different in both MSST and antibiosis tests against the Hadano-66 population and against the Koshi-2013 population, which corresponds well with previous research by Xiao et al. (2016) [51]. Therefore, the genes on chromosome 4S of IR71033-121-15, PTB33 and Rathu Heenati might be different. To confirm this, further sequence analyses are needed for the three loci *BPH17, BPH17-ptb* and *BPH20*.

Among six genes/QTLs that have been identified on the short arm of chromosome 6 of *O. sativa* and its wild relatives [3,12], *BPH3* and *BPH32* have been widely introduced to elite rice cultivars to improve BPH resistance and were related to durable and broad-spectrum resistance in PTB33 and Rathu Heenati [67,69]. In previous research, *BPH3* was mapped onto chromosome 6 between two markers, RM19291 (1.2 Mbp) and RM8072 (1.4 Mbp) [69]. *BPH32* from PTB33 was identified at the same location as *BPH3* from Rathu Heenati, but the amino acid sequence of *BPH3* was not identical to that of *BPH32* [12]. In our study, the resistance levels of the *BPH3*-NIL were slightly different from those of the *BPH32*-NIL, suggesting that *BPH3* might be different from *BPH32*. A comparison of amino acid sequences between *BPH3* and *BPH32* would be necessary to confirm whether these resistance genes are different.

Among the developed NILs, the *BPH3*-NIL, *BPH17*-NIL, *BPH17-ptb*-NIL and *BPH32*-NIL had around 97.0% of their chromosomal segments from the recurrent parent. This proportion coincides with the theoretical ratio for substituting chromosomal segments from recurrent parents by backcrossing four times. The other NILs had fewer chromosomal segments from T65 than the theoretical rate. The substituted chromosomal segments from the donor parents might be related to undesirable traits such as the suppression of the associated BPH resistance gene. Additionally, due to the low density of available polymorphic SSR markers between T65 and donor DNA around the target genes, the BPH resistance genes on the NILs were selected using two flanking markers that were relatively far apart. Furthermore, *BPH17*, *BPH20* and *BPH17-ptb* on the NILs were selected by flanking markers with longer intervals because of the low density of polymorphic markers between donor parents

and T65 around the chromosome 4S region. The intervals between each of the flanking marker pairs for *BPH17* and *BPH17-ptb* were 3.7 Mbp, and that for *BPH20* was 5.7 Mbp. Similarly, the interval for each of the two flanking markers for *BPH2* and *BPH21* on chromosome 12 was 4.1 Mbp because the exact locations of genes had not been identified before we started to develop the NILs by MAS. Therefore, many of the NILs had relatively long chromosome segments derived from the donor parents and there is a possibility that the remaining chromosomal segments from donor DNA around the target genes included linkage drag associated with susceptibility. In further research, ensuring that flanking markers are tightly linked to target genes will avoid linkage drag from donors through MAS and backcrossing.

An improvement of rice resistance levels against BPH is necessary since many genes have become less effective against BPH across Asia [45]. In this study, we developed 15 PYLs carrying two or three genes for BPH resistance. The PYLs tended to increase resistance against the two BPH populations, Hadano-66 and Koshi-2013. Among the 15 PYLs, 12 and nine PYLs had higher ADM rates than corresponding NILs against Hadano-66 and Koshi-2013, respectively; ten PYLs had lower FBRRs compared to corresponding NILs in the MSST against the Hadano-66 population. For example, *BPH2* + *BPH32*-PYL (84.0%) and *BPH2* + *BPH32* + *BPH17-ptb*-PYL (95.8%) had higher resistance levels than those of the *BPH2*-NIL (68.9%), *BPH32*-PYL (14.0%) and *BPH17-ptb*-PYL (22.0%) in antibiosis tests against the Hadano-66 population. The ADM rates of *BPH2* + *BPH17*-PYL (32.0%) and *BPH2* + *BPH3* + *BPH17*-PYL (36.0%) were higher than those of *BPH2*-NIL (4.0%), *BPH3*-NIL (0%) and *BPH17*-NIL (20.0%) against the Koshi-2013 population. Additionally, the FBRR of *BPH2* + *BPH17* (42.3%) was lower than for *BPH2*-NIL (67.6%) and *BPH17*-NIL (58.8%). However, the effectiveness of the PYLs was not consistently higher than that of the corresponding NILs. The effect of PYLs was influenced by specific interactions between gene loci, the specific BPH populations and the screening methods. For example, the resistance levels of *BPH3* + *BPH17*-PYL (50.0%) and *BPH17* + *BPH21*-PYL (58.3%) were not higher than that of *BPH17*-NIL (59.0%) in antibiosis tests (or ADM rates) against the Hadano-66 population. *BPH2* + *BPH25*-PYL (87.5%) showed higher ADM against the Hadano-66 population in comparison to *BPH2*-NIL (68.9%) and *BPH25*-NIL (16.0%), while the ADM rate of *BPH2* + *BPH25*-PYL (12.0%) was lower than that of *BPH25*-NIL (16.0%) against the Koshi-2013 population. A similar tendency has been reported for gene combinations between *BPH1* and *BPH2* [70]; *BPH18* and *BPH32*; *BPH20* and *BPH32*; and *BPH2, BPH18* and *BPH32* [57]. That the resistance levels of most of the PYLs were not significantly higher than those of the corresponding NILs might be related to the relatively small number of replications used in the bioassays (five replications for the antibiosis and three replications for the MSST).

In a previous study, virulence of a BPH population collected during 2005 in Japan had increased compared with the virulence of a population collected in 1966 [51]. Through antibiosis tests, we evaluated BPH resistance against the populations collected in 1966 (Hadano-66) and in 2013 (Koshi-2013). Both represented BPH arriving as migrants to Japan. The Hadano-66 population was virulent to T65 (with no resistance gene) but avirulent to all plants with resistance genes, including Mudgo (*BPH1*), ASD7 (*BPH2*), Rathu Heenai (*BPH3* and *BPH17*), Babawee (*BPH4*), Chin Saba (*BPH8*), Balamawee (*BPH9*) and two NILs, *BPH25*-NIL and *BPH26*-NIL [51,71]. In the present study, most of the NILs, all of the PYLs and the donor parents were still effective against the Hadano-66 population. In contrast, all of the NILs and most of the PYLs were susceptible to the Koshi-2013 population, suggesting that BPH recently arriving to Japan from China has greater virulence than was evident about 50 years ago (i.e., 1966). Among the PYLs, two PYLs, *BPH20* + *BPH32*-PYL and *BPH2* + *BPH3* + *BPH17*-PYL, had relatively high resistance, suggesting that PYLs with combinations of these genes are likely to provide good resistance against the current BPH populations that arrive to Japan (Koshi-2013). Finding new sources of resistance genes will be necessary to further improve resistance against contemporary BPH populations as they gain virulence.

In comparison to the corresponding NILs and PYLs, the resistance levels of PTB33, Rathu Heenati and IR71033-121-15 were higher. This suggests that PTB33, Rathu Heenati and IR71033-121-15 might also contain other BPH resistance gene(s). The other genetic factor(s) for BPH resistance can be revealed by analyzing the segregating populations derived from crosses between the developed PYLs and their donor parents in future studies. Additionally, Rathu Heenati had *QBPH4.1* (5.8-7.8 Mbp) and *QBPH4.2* (15.2-17.2 Mbp) on chromosome 4S rather than *BPH3* and *BPH17* [13]. Therefore, the NILs and PYLs carrying *QBPH4.1* and *QBPH4*.2 should be developed and evaluated in further analyses. On the other hand, the lower resistance levels of the NILs and PYLs might be related to the relatively high ratio of substituted chromosomal segments from donors in the NILs (from 3.8 to 55.0 Mbp) and PYLs. There is a possibility that the retained donor chromosomal segments in the genetic background of the NILs and PYLs might be linked to the suppression of BPH resistance. To gain further knowledge of BPH resistance controlled by multiple genes, it will be essential to reduce the donor parent chromosomal segments on the NILs and PYLs by further backcrossing and MAS.

#### **4. Materials and Methods**

#### *4.1. Plant Materials*

To develop NILs with BPH resistance genes, a *japonica* rice variety, T65, that is susceptible to BPH, was used as a recurrent parent, and three rice varieties resistant to BPH were donor parents. The donor lines were IR71033-121-15, PTB33 and Rathu Heenati. IR71033-121-15 contains two BPH resistance genes, *BPH20* and *BPH21*, from the wild rice species *O. minuta* (Accession number: IRGC101141) [17]. PTB33 (Accession number: IRGC19325) that originated from India contains *BPH2, BPH17-ptb* and *BPH32.* Rathu Heenati (Acc. no. IRGC 11730), that originated from Sri Lanka, carries *BPH3* and *BPH17* [16,39]. T65 was crossed with these donor parents and F1 plants were backcrossed four times with T65 to generate BC4F1 plants (Figure 3). At each generation of backcrossing, plants carrying BPH resistance genes from the donor parents were selected by MAS using flanking SSR markers of the target BPH resistance genes (Table 1). The selected BC4F1 plants were self-pollinated to produce BC4F3, BC4F4 and BC4F5 plants with BPH resistance genes. Finally, seven NILs with either *BPH2, BPH3, BPH17, BPH20, BPH21, BPH32* or *BPH17-ptb* were developed. The NILs were used to survey the genetic background and evaluate BPH resistance levels as well as agronomic traits. Two additional NILs, *BPH25*-NIL and *BPH26*-NIL were used in the development of the PYLs [54].

**Figure 3.** Breeding scheme for the development of near-isogenic lines and pyramided lines containing brown planthopper resistance genes from donor parents, IR71033-121-15, PTB33 and Rathu Heenati.

#### *4.2. The Development of PYLs with BPH Resistance Genes*

All the PYLs for two or three BPH resistance genes were developed using the NILs descended from the BC4F1 generation, except *BPH20* + *BPH21*-PYL and *BPH32* + *BPH17-ptb*-PYL that were descended from the BC3F1 generation. The F1 plants derived from crosses between NILs were self-pollinated to produce F2 plants. From 96 F2 plants, plants that were homozygous for two or three BPH resistance genes were selected by MAS. Several plants from 96 F3 plants with similar agronomic traits to T65 were selected as final PYLs. The following PYLs carrying two or three BPH resistance genes were evaluated for BPH resistance and agronomic traits: *BPH2* + *BPH17*-PYL, *BPH2* + *BPH25*-PYL, *BPH2* + *BPH32*-PYL, *BPH2* + *BPH17-ptb*-PYL, *BPH3* + *BPH17*-PYL, *BPH17* + *BPH21*-PYL, *BPH20* + *BPH21*-PYL, *BPH20* + *BPH32*-PYL, *BPH21* + *BPH25*-PYL, *BPH21* + *BPH17-ptb*-PYL, *BPH25* + *BPH17-ptb*-PYL, *BPH32* + *BPH17-ptb*-PYL, *BPH2* + *BPH3* + *BPH17*-PYL, *BPH2* + *BPH32* + *BPH17-ptb*-PYL and *BPH20* + *BPH21* + *BPH32*-PYL.

#### *4.3. The MAS for BPH Resistance Genes*

To conduct MAS, approximately 2 cm of leaves from two–week old seedlings were collected and dried in a freeze drier for 48 h, and total DNA was extracted using the potassium acetate method [72]. The genotypes of SSR markers on plants in each generation were determined by polymerase chain reaction (PCR) and electrophoresis. The PCR amplification mix (8 μL) contained 3 μL of 1X GoTaq® Green Master Mix (pH 8.5), 0.25 μM of primer and 4 μL of 20 times-diluted DNA. Each PCR amplification included one cycle at 96 ◦C for 5 min, 35 cycles at 96 ◦C for 30 s, 55 ◦C for 30 s and 72 ◦C for 30 s, followed by one extension cycle at 25 ◦C for 1 min. PCR products were analyzed by electrophoresis at 200 V using 4% agarose gel with 0.5 μg/mL ethidium bromide in 0.5X TBE buffer for 1 h and photographed under ultraviolet light. During MAS for resistance genes on chromosome 4S, the plants with *BPH17* and *BPH17-ptb*, were selected using two markers, RM8213 and MS10, and the plants with *BPH20* were selected using MS10 and RM5900 (Table 1). The plants with *BPH3* and *BPH32* on the short arm of chromosome 6 were selected using two flanking markers, RM508 and RM588. The plants carrying *BPH2* and *BPH21* located on the long arm of chromosome 12 were screened using RM1246 and RM28493. The plants with *BPH25* were selected using S00310 and MSSR1, and the plants with *BPH26* were selected using RM309, RM28438, InD14, RM28466, RM28481 and MSSR2.

#### *4.4. The Genetic Background Survey of the NILs*

In the genetic background survey of the NILs, the bulk DNA from five plants was used. A total of 384 SSR markers distributed on 12 rice chromosomes were used during polymorphism tests with T65 and the donor parents [62]. Among the 384 SSR markers, 254 SSR markers with polymorphisms between IR71033-121-15 and T65 were utilized to identify substituted chromosomal segments from IR71033-121-15 on *BPH20*-NIL and *BPH21*-NIL. Additionally, 244 of 384 SSR markers with polymorphisms between PTB33 and T65 were used to detect substituted chromosomal segments from PTB33 on *BPH2*-NIL, *BPH32*-NIL and *BPH17-ptb*-NIL. To identify substituted chromosomal segments from Rathu Heenati on *BPH3*-NIL and *BPH17*-NIL, 204 of 384 SSR markers with polymorphisms between Rathu Heenati and T65 were used. The whole genome compositions of the developed NILs were graphically displayed following the concept of the graphical genotype proposed by Young and Tanksley (1989) using GGT software version 2.0 [73].

#### *4.5. The BPH Populations and the Characterization of BPH Resistance*

Two BPH populations from Japan (Hadano-66 and Koshi-2013) were used to evaluate the NILs and PYLs for their resistance. Hadano-66 was collected in Hadano City, Kanagawa Prefecture, Japan in 1966 [51], and Koshi-2013 was collected in Koshi City, Kumamoto Prefecture, Japan in 2013. Both BPH strains were maintained on the susceptible *japonica* rice variety, Reiho, at 25 ◦C with 16 h/8 h of light/dark at Kyushu Okinawa Agricultural Research Center of the National Agriculture and Food Research Organization in Japan.

To evaluate resistance, an adaptation of the modified seedbox screening test (MSST) [45,74] was applied at 25 ◦C using the Hadano-66 strain. To conduct the test, 30 seeds of each of the NILs, PYLs and parent lines were sown to single rows in a plastic tray (23.0 × 30.0 × 2.5 cm) with 2.5 cm between successive rows of seedlings. Two sets of trays—one tray infested by BPH and the other without infestation (the control tray), were used to measure the effects of BPH on plant biomass. One row of Rathu Heenati was added as a resistant control, while three rows of T65 were sown at the center and the two edges as a susceptible control. At seven days after sowing (DAS), the plants in the trays were thinned to 20 plants per row. One tray was infested by the second and third instar nymphs at a density of around 20 BPHs per plant. The experiment was replicated three times. When all the plants of T65 were completely desiccated due to BPH feeding, the DSs of all lines were graded following the standard evaluation system for rice of the International Rice Research Institute [75]. The plants from each row in the two trays were cut above the soil surface and weighed. The fresh biomass reduction rate (FBRR) was calculated using the following formula:

$$\begin{array}{c} \text{Fresh biomass reduction rate} \\ \text{(FBRR)} \left( \% \right) \end{array} = \left[ 1 - \frac{\text{Infected plant weight} \left( \text{g} \right)}{\text{Non-infected plant weight} \left( \text{g} \right)} \right] \times 100. \tag{1}$$

#### *4.6. Antibiosis Tests*

Antibiosis tests were conducted at 25 ◦C following the method described by Myint et al. (2009) [51]. Five plants of each NIL, PYL and parent line were individually sown in 200 mL plastic cups. At four weeks after sowing, the plants were trimmed to 15 cm height and covered with a plastic cage with insect screen windows for ventilation. Each cage was infested with five thin-abdomen brachypterous female BPHs. At five days after infestation, the ADM was recorded (i.e., the number of dead females).

#### *4.7. Characterization of NILs and PYLs for Agronomic Traits*

The NILs and PYLs were grown in a paddy field at Saga University (Saga, Japan) in 2018 and characterized for their agronomic traits compared to those of T65. Seedlings were transplanted at 28 DAS as one plant per hill, with 20 cm between hills and 25 cm between rows. Each entry was planted as at least three rows (12 plants per a row). Six agronomic traits: DTH, CL, PL, LL, LW and PN were measured for five plants in the same row. DTH was the days from sowing until 50% of panicles flowered. CL was measured from the soil surface to the panicle neck. PL is the length from tip to panicle neck of the longest panicle. The flag leaf width and length were measured from the largest and longest flag leaf of each sampled plant. Panicle number is the number of reproductive panicles of each plant at maturity.

#### *4.8. Statistical Analysis*

Mean values of BPH resistance (DS, FBRR and ADM) for the NILs and PYLs and agronomic traits were compared using one-way ANOVA. Dunnett's test and Tukey Kramer's test were conducted for multiple comparisons of BPH resistance and agronomic traits, respectively, using R software version 3.5.2.

**Author Contributions:** P.S.V., D.S.B., F.G.H., H.Y. and D.F. designed the research. S.S.-M. and M.M. provided insects for conducting the research. C.D.N., H.V., D.Z. and D.F. performed the research. C.D.N., H.V. and D.F. developed the plant materials. C.D.N. and D.F. wrote the paper.

**Funding:** This work was supported by JSPS KAKENHI, grant numbers JP15H04438 and 17K07606.

**Acknowledgments:** We thank staff of the Insect Pest Management Research Group, Kyushu Okinawa Agricultural Research Center, for rearing and preparing insect populations, and Atsushi Yoshimura, Yoshiyuki Yamagata and the staff belonging to the Plant Breeding Laboratory at Kyushu University for growing and maintaining the plant materials. Furthermore, we thank Kshirod K. Jena for aiding in the development of materials and Elmer Sanchez for developing materials and taking care of the plants. We also wish to thank the Government of Vietnam for the doctoral fellowship granted to C.D.N.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Two SNP Mutations Turned o**ff **Seed Shattering in Rice**

## **Yu Zhang** †**, Jiawu Zhou** †**, Ying Yang** †**, Walid Hassan Elgamal** ‡**, Peng Xu, Jing Li, Yasser Z. El-Refaee** ‡**, Suding Hao and Dayun Tao \***

Yunnan Key Laboratory for Rice Genetic Improvement, Food Crops Research Institute, Yunnan Academy of Agricultural Sciences (YAAS), Kunming 650200, China; zhangyu\_rice@163.com (Y.Z.); zhjiawu@aliyun.com (J.Z.); yaasyang@126.com (Y.Y.); elgamal.rrtc@gmail.com (W.H.E.);

xupeng@xtbg.ac.cn (P.X.); lijinglab@163.com (J.L.); elrefaeey@yahoo.co.in (Y.Z.E.-R.); haosuding@126.com (S.H.)


Received: 29 August 2019; Accepted: 5 November 2019; Published: 6 November 2019

**Abstract:** Seed shattering is an important agronomic trait in rice domestication. In this study, using a near-isogenic line (NIL-*hs1*) from *Oryza barthii*, we found a hybrid seed shattering phenomenon between the NIL-*hs1* and its recurrent parent, a *japonica* variety Yundao 1. The heterozygotes at *hybrid shattering 1* (*HS1*) exhibited the shattering phenotype, whereas the homozygotes from both parents conferred the non-shattering. The causal *HS1* gene for hybrid shattering was located in the region between SSR marker RM17604 and RM8220 on chromosome 4. Sequence verification indicated that *HS1* was identical to *SH4*, and *HS1* controlled the hybrid shattering due to harboring the ancestral haplotype, the G allele at G237T site and C allele at C760T site from each parent. Comparative analysis at *SH4* showed that all the accessions containing ancestral haplotype, including 78 wild relatives of rice and 8 African cultivated rice, had the shattering phenotype, whereas all the accessions with either of the homozygous domestic haplotypes at one of the two sites, including 17 wild relatives of rice, 111 African cultivated rice and 65 Asian cultivated rice, showed the non-shattering phenotype. Dominant complementation of the G allele at G237T site and the C allele at C760T site in *HS1* led to a hybrid shattering phenotype. These results help to shed light on the nature of seed shattering in rice during domestication and improve the moderate shattering varieties adapted to mechanized harvest.

**Keywords:** Seed shattering; *O. barthii*; *O. sativa*; *HS1*; haplotype

#### **1. Introduction**

During rice domestication, seed shattering is one of the most greatly changed traits for seed dispersal. Easy shattering leads to the loss of production [1], and more attention is paid on selection for non-shattering but threshable rice in modern rice breeding [2]. Seed shattering is caused by the formation and degradation of the abscission zone (AZ), which, constituted by a band of small cells, is responsive to signals promoting abscission [3].

Recently, several genes responsible for seed shattering were identified in rice. *qSH1* encodes a BELL homeobox protein. An SNP mutation in the regulatory region of *qSH1* could inhibit its expression, which resulted in defective abscission layer development [4]. The allelic genes of *SH4* [5], *SHA1* [6] and *GL4* [7] encoding a trihelix transcriptional factor, all controlled the seed shattering, however, haplotypes were divergent because of two different SNP mutations. A "G237T" mutation in *SH4* and *SHA1* was responsible for the loss of seed shattering in Asian cultivated rice [5,6], whereas "C760T" transition in *GL4* conferred non-shattering seeds in African cultivated rice [7]. The *SHAT1* gene, which encoded an

APETALA2 transcription factor, was responsible for seed shattering through specifying abscission zone development in rice. The expression of *SHAT1* was positively regulated by the transcription factor *SH4,* which was required for the AZ identification during the early spikelet developmental stage, and *qSH1* functions downstream of *SHAT1* and *SH4*, promoting the AZ differentiation by maintaining the expression of *SHAT1* and *SH4* [8]. *SH5*, which is highly homologous to *qSH1*, also controlled seed shattering by regulating lignin deposition in the pedicel region [9]. *OsGRF4* could increase the expression of two cytokinin dehydrogenase precursor genes resulting in the high cytokinin level, which led to reduced seed shattering [10]. *OSH15* together with *SH5* induced seed shattering by repressing lignin biosynthesis genes [11]; *ObSH3* in *Oryza barthii* encoded a YABBY transcription factor, which was also required for the development of the seed abscission layer [12]. In addition, some other minor genes and allelic interaction at major locus might be involved in the seed shattering domestication as rice underwent a prolonged domestication process, with continuing selection for reduced shattering [13,14].

*O. barthii* is one of the relatives distributed in West Africa, sharing the same AA genome as Asian cultivated rice. Most of *O. barthii* accessions exhibited the seed shattering. The previous report indicated that *GL4* in *O. barthii* was involved in the non-shattering selection during the African cultivated rice domestication [7], but the nature of seed shattering was still not clear. Here, we report that a novel locus, named *hybrid shattering 1* (*HS1*)*,* controlled the seed shattering in the hybrid between *O. barthii* and *O. sativa*. A near-isogenic line (NIL-*hs1*) from *O. barthii,* and its recurrent parent, a *japonica* variety Yundao 1, showed the non-shattering phenotype. Interestingly, the hybrid between two parents showed the seed shattering, similar to the ancestral wild rice. Whether it was shattering or not was dependent on the different haplotypes of two SNPs at *HS1*. This result could help us understand the complex molecular mechanism of seed shattering.

#### **2. Results**

#### *2.1. HS1 Controlled the Hybrid Seed Shattering in Rice*

We developed a NIL-*hs1* carrying genome fragment from *O. barthii* on chromosome 4 in the Yundao 1 genetic background. Surprisingly, Yundao 1 and the NIL-*hs1* showed the non-shattering seed, while F1 hybrid exhibited seed shattering (Figure 1A, Table S1). In order to distinguish the differences in abscission layer structure between F1 hybrid and its parents, longitudinal sections at the seeds base were observed using fluorescent microscopy. The results showed that Yundao 1 displayed the deficiency in abscission zone development near the vascular bundle (Figure 1B), the NIL-*hs1* showed no abscission layer on the palea side and the partial abscission layer on the lemma side between the seed pedicel and the spikelet, respectively (Figure 1D). Conversely, the F1 hybrid had a continuous abscission zone between the vascular bundle and the epidermis (Figure 1C). These results indicated that seed shattering in the F1 hybrid resulted from the complete and continuous abscission layer, whereas the loss of seed shattering in Yundao 1 and NIL-*hs1* was caused by the irregular development of the abscission zone.

#### *2.2. Dominant Complementation of G Allele at G237T Site and C Allele at C760T Site in HS1 Led to Hybrid Seed Shattering Phenotype*

In order to understand whether *HS1* acted as a single Mendelian factor or not, the seed shattering rate was investigated in BC4F1 and BC4F2 populations derived from the cross between Yundao 1 and NIL-*hs1*. All the BC4F1 individuals showed seed shattering. The seed shattering rate in the BC4F2 population was segregated into non-shattering and shattering classes in a 246:205 ratio, which fitted the 1:1 ratio (χ<sup>2</sup> = 1.870, *P* = 0.172) (Figure S1). These results indicated that the seed shattering in F1 hybrid was controlled by a single gene. We designated it as *HS1*.

A population of 790 BC4F2 plants was generated for mapping the *HS1.* Eight polymorphic SSR markers in the introgressed region on chromosome 4 were used for genotyping the 790 individuals in the BC4F2 population. *HS1* was mapped into a 0.4 cM region flanked by RM17604 and RM8220, at genetic distances of 0.3 and 0.1 cM, respectively, and co-segregated with RM17616 (Figure 2). The homozygotes from both parents at RM17616 showed a non-shattering phenotype, whereas the heterozygotes at RM17616 showed a shattering phenotype. Based on the GRAMENE public database (http://www. gramene.org), the physical distance between RM17604 and RM8220 was about 434.6 kb. The mapping region of *HS1* was similar to the location of *SH4*/*GL4* (LOC\_Os04g57530/ORGLA04G0254300) identified from the Asian rice and the African rice [4–6]. In order to confirm whether the *HS1* was allelic to *SH4*/*GL4* or not, the sequence analysis of *SH4* in Yundao 1 and NIL-*hs1* was performed. A total of 13 SNPs, 5 indels in the 2.3 kb of aligned sequenced DNA were identified (Figure 3), which resulted in 1 amino acid insertion, 4 amino acid substitutions, 6 amino acid deletions and pre-stop codon in NIL-*hs1*, respectively (Figure 3). Of these, two base substitutions of G237T and C760T (C760T referred that the C to T SNP mutation in *HS1* was at nucleotide position 760 in *O. barthii*, which was the same as C769T mutation in Yundao 1) resulted in the mutation of Asn79 to Lys79 and Gln258 to a stop codon, respectively. It was reported that the G allele at G237T site and the C allele at C760T site were responsible for the seed shattering during the Asian cultivated rice and the African cultivated rice domestication, respectively [4–6]. Thus, we postulated that *HS1* was identical to *SH4* and *GL4*. The G allele at G237T site and the T allele at C760T in the Yundao 1 background, and the T allele at G237T site and the C allele at C760T in the NIL-*hs1* background all conferred the non-shattering phenotype, whereas the combination of the G allele at G237T site and the C allele at C760T site exhibited the shattering phenotype. Dominant complementation of the G allele at G237T site and the C allele at C760T site in *SH4* led to the hybrid shattering phenotype. Moreover, *SH4* in NIL-*hs1* had a unique deletion of the 227th amino acid residue isoleucine (Ile), compared with that in other AA genome species in the genus *Oryza*.

**Figure 1.** (**A**) The seed shattering rate of Yundao 1 (left), F1 hybrid (middle) and near-isogenic line (NIL-*hs1*) (right). Scale bars = 0.5 cm. (**B–D**) Fluorescence images of a longitudinal section of the spikelet

and pedicel junction in Yundao 1, F1 hybrid and NIL-*hs1*, respectively. (**B**) Yundao 1 showed an incomplete in abscission zone. (**C**) F1 hybrid with a complete abscission layer. (**D**) NIL-*hs1* exhibited a deficiency in abscission layer on the palea side and partial abscission layer on the lemma side. AL: Abscission layer, V: Vascular bundle. White arrow indicates a deficiency in abscission zone. Scale bars = 10 μm.

**Figure 2.** (**A**) Graphical genotypes show that an *O. barthii* chromosomal segment was introgressed into the NIL-*hs1* genome on chromosome 4. (**B**) Genetic mapping of *hybrid shattering 1* (*HS1*) on Chromosome 4, white bar: homozygous Yundao 1; grey bar: heterozygous; black bar: homozygous NIL-*hs1*. "R" means the number of recombinants.

**Figure 3.** The difference in the coding sequence and amino acid of *HS1* between Yundao 1 and NIL-*hs1*. Synonymous mutations and functional mutations were shown in green and red, respectively. The asterisk indicates the stop codon.

#### *2.3. Two SNP Mutations Turned o*ff *Seed Shattering in Rice*

In order to confirm the function of the two SNPs, we reanalyzed the gene sequence of *SH4* in 95 wild accessions of rice, 119 *O. glaberrima* and 65 *O. sativa* using previously published data. All the accessions harboring both the G allele at G237T site and the C allele at C760T site exhibited the shattering phenotype, including 2 *O. longistaminata* accessions, 22 of 28 *O. barthii* accessions, 8 of 119 *O. glaberrima* accessions, 2 *O. glumaepatula* accessions, 20 of 25 *O. nivara* accessions and 30 of 36 *O. rufipogon* accessions. All the accessions (varieties) with either the G allele at G237T site or the C allele at C760T site showed the non-shattering, including 6 of 28 *O. barthii* accessions, 111 of 119 *O. glaberrima* accessions, 5 of 25 *O. nivara* accessions, 6 of 36 *O. rufipogon* accessions and 65 Asian cultivated varieties. These results indicated that the G allele at G237T site and the C allele at C760T site were ancestral alleles in African rice domestication and Asian rice domestication, respectively, whereas the T allele at both sites that resulted from selection pressure were mutation alleles. The ancestral haplotype (the G

allele at G237T site and the C allele at C760T site) induced a shattering phenotype in rice, whereas domestic haplotypes (the G allele at G237T site and the T allele at C760T site, the T allele at G237T site and the C allele at C760T site) all exhibited the loss or reduction of seed shattering (Table 1), which was consistent with our experimental results that the hybrid harboring ancestral haplotype showed a shattering phenotype; however, Yundao 1 and NIL-*hs1* carrying homozygous domestic haplotypes exhibited a non-shattering phenotype.


**Table 1.** The haplotypes of the *SH4* at G237T and C760T sites in the AA genome species of genus *Oryza.*

#### **3. Discussion**

What causes the differences in seed shattering in different species is totally an open question. Loss or reduction of seed shattering represents a key transition to domestication in rice [15,16]. In this study, it was a serendipitous finding that the hybrid F1 derived from the cross between the non-shattering Yundao 1 and NIL-*hs1* that displayed the seed shattering phenotype. And we reported that a novel locus *HS1* controlled the hybrid shattering between *O. sativa* and *O. barthii*. Yundao 1 and NIL-*hs1* showed the irregular and partially developed abscission layer, whereas the F1 hybrid exhibited a continuous abscission layer between seed pedicel and spikelet. *HS1* and *SH4* were mapped into a similar region on chromosome 4 [5], interestingly, *SH4* functioned in the seed shattering on the homozygous background, whereas *HS1* conferred the seed shattering on the heterozygous background. What resulted in this difference? It was suggested that an allelic interaction at a single locus or an epistatic interaction at two independent loci from the non-shattering species *O. sativa* and *O. barthii* determined the phenotype. Interestingly, *HS1* and *SH4* were mapped into a similar region on Chromosome 4 and sequencing analysis confirmed that *HS1*, allelic to *SH4*, carrying ancestral haplotype (the G allele at G237T site and the C allele at C760T site) contributed to shattering in the F1 hybrid. Thus, an allelic interaction at *SH4* was responsible for the hybrid seed shattering. NIL-*hs1* and most of *O. glaberrima* accessions shared the same *SH4* haplotype, and we also found that the hybrid between the *O. glaberrima* and *O. sativa* also displayed the seed shattering phenotype. As we know, it is difficult to obtain the hybrid progeny between *O. glaberrima* and *O. sativa*, one of the reasons is that interspecific hybrid sterility between *O. glaberrima* and *O. sativa* prevents the formation of hybrid offspring, and another reason is that strong seed shattering in the hybrid increases the difficulty of the crossing. Previous studies reported that one single-nucleotide polymorphism, G237T or C760T, controlled seed shattering in rice independently, because one SNP was fixed and not polymorphic in the wild and cultivated accessions [5–7]. In this study, the relationship of the different haplotypes of *SH4* and the shattering phenotype was analyzed from the whole gene sequence viewpoint. Moreover,

two base mutations of G237T and C760T at *SH4* occurred in Trihelix DNA binding domain, indicating that this domain played an important role in seed shattering and either of the nucleotide acid mutations had no effect on the function of the DNA binding domain. The transcription factor *SH4* controlled the AZ identification by positively regulating the expression of *SHAT1,* and *qSH1* could promote the AZ differentiation by maintaining the expression of *SHAT1* and *SH4* [8], suggesting that two amino acid substitutions (Asn79 to Lys79 and Gln258 to a stop codon) in *SH4* might affect the interaction with *SHAT1* and *qSH1*, resulting in the loss of seed shattering. In addition, with the similar genetic background, the shattering degree of Yundao 1 was easier than that of NIL-*hs1*, there were two possibilities: (1) The haplotype of T237 and C760 combination conferred the easier shattering than that of G237 and T760 combination; (2) other genes in the introgressed region could decrease the seed shattering rate by regulating the expression of *SH4.* These results would provide new clues into the molecular basis of seed shattering in rice, and breeders can take the advantage of different haplotype combinations adapted to the moderate shattering degree so as to meet the need for mechanized harvest.

Asian cultivated rice (*O. sativa* L.) was domesticated from wild species *O. rufipogon* thousands of years ago [17,18], whereas *O. glaberrima* Steud. was an African species of rice that was domesticated from the wild progenitor *O. barthii* about 3000 years ago [15,16]. In this study, 79% of *O. barthii* accessions harbored the G allele at G237T site and the C allele at C760T site in *SH4*, but 93% of *O. glaberrima* accessions carried the G allele at G237T site and the T allele at C760T site. It is suggested that the wild-type G at the G237T site was fixed, while the mutated T allele at the C760T site was selected during the African cultivated domestication. The function of the G allele at G237T site (Lysine residue) may be critical for the growth and development of the African wild relative of rice and African cultivated rice so that it was fixed during the gradual domestication process, while the T allele at G237T site contributing to the small grain size was selected, which may be an adaptation to the extreme environment in West Africa, such as drought, soil acidity, iron and aluminum toxicity [7,19,20]. Two haplotypes of *SH4* in *O. rufipogon* existed, whereas the cultivated rice only had one domestic haplotype, these results were also observed in *O. barthii* and *O. glaberrima*. Compared with African cultivated rice, domestic allele in Asian cultivated rice was the G237T mutation, while the C allele at C760T mutation was fixed, indicating that the C alleles at the C760T site might be involved in the selection of a large grain size [7]. This might be one of the reasons that the yield of Asian cultivated rice was higher than that of African cultivated rice. Taken together, the seed shattering characteristics were selected in African cultivated rice and Asian cultivated rice, respectively, which were consistent with the theory that *O. glaberrima* and *O. sativa* were domesticated in parallel [12,18].

#### **4. Materials and methods**

#### *4.1. Plant Materials*

An *O. barthii* accession, Acc.104284, introduced from the International Rice Research Institute (IRRI), as a donor and male parent, was crossed with a temperate *japonica* variety of *O. sativa*, Yundao 1, from Yunnan province, P. R. China. The male gametes in hybrid F1 between *O. barthii* and *O. sativa* were fully sterile, and the female gametes in hybrid F1 were partially fertile, thus, hybrid F1 as the female parent was consecutively backcrossed with Yundao 1 as the male parent. BC3 plants were self-pollinated for 13 generations to produce the BC3F14 introgression lines. Four hundred and twenty-six polymorphic SSR markers evenly distributed on 12 chromosomes of rice were used to evaluate the substituted fragments from *O. barthii.* The results indicated that only 14.8 cM segments on chromosome 4 were substituted by the *O. barthii* genome. The individuals harboring the homozygous genome fragment from *O. barthii* were selected as NIL-*hs1*. NIL-*hs1* was crossed with the recurrent parent Yundao 1, and then was self-fertilized to produce the BC4F2 population. We found the BC4F1 plants all showed seed shattering, while seed shattering and non-shattering were both observed in the BC4F2 population. Seven hundred and ninety individuals were used to mapping *HS1* for seed shattering between *O. sativa* and *O. barthii.*

All plant materials were grown in the paddy field at the Experiment Station, YAAS, located in Jinghong, Yunnan Province, P. R. China.

#### *4.2. Evaluation of Seed Shattering Rate*

The spikelets of each plant were bagged at the stage of heading. Then, the panicles were collected at the stage of maturity, and panicles freely fell 2 m to the plastic box (60 cm × 90 cm). All the grains shredded prior to the test were counted as shattering grains. The shattering rate was calculated by a percentage of shattered seeds to the total seeds. The shattering rate below 5% or above 50%, was defined as the non-shattering type and the shattering type, respectively.

#### *4.3. Microscopy*

Seeds including pedicels were collected at the stage of maturity, and slices made by mature and dry seeds were stained with 1% acridine orange. Abscission layer at seed base was observed by Fluorescent microscopy (OLYMPUS BX53).

#### *4.4. DNA Extraction and SSR Analysis*

The experimental procedure for DNA extraction was performed as previously described [21]; rice SSR markers were selected from the Gramene database (http://www.gramene.org) or previously published SSR markers in rice [22]. PCR was performed as follows: a total volume of 10 μL containing 10 ng template DNA, 1 × buffer, 0.2 μM of each primer, 50 μM of dNTPs and 0.5 units of Taq polymerase (Tiangen Company, Beijing, China). The reaction mixture was incubated at 94 ◦C for an initial 4 min, followed by 30 cycles of 94 ◦C 30 s, 55 ◦C 30 s and 72 ◦C 30 s, and a final extension step of 5 min at 72 ◦C. PCR products were separated on 8% non-denaturing polyacrylamide gel and detected using the silver staining method.

#### *4.5. Linkage Analysis*

A linkage map was constructed on the basis of genetic linkage between the genotype of SSR markers and seed shattering phenotype in the BC4F2 population.

#### *4.6. Sequencing*

In order to compare the sequence difference of *SH4* between Yundao 1 and NIL-*hs1*, the primer (SH4-F: CCGAACACCAAACGCCTCAG, SH4-R: CCGTACTCCCAATACTCGCAGA) was designed on the 5- UTR and 3- UTR region of *SH4* gene for amplifying the target sequence. PCR mixture (25 uL) contained 0.4 mM of each dNTP, 0.3 uM of each primer, 0.5 units of Taq polymerase (KOD FX DNA polymerase, Toyobo, Japan) and template DNA 100 ng in the GeneAMP PCR system 9700 (Applied Biosystems, Foster City, CA, USA). The PCR program was 94 ◦C for 2 min, followed by 30 cycles at 98 ◦C for 10 s, 55 ◦C for 30 s and 68 ◦C for 2 min. PCR products were separated in 1% agarose gels.

#### *4.7. Haplotype Analysis of the SH4 Gene*

To analyze the haplotype of the *SH4* gene, the nucleotide sequence of *SH4* in the AA genome was downloaded from the GenBank database, wild rice genome project and public data [5,7,15,23–25], 2 SNPs of G237T and C760T were analyzed in 279 rice accessions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2223-7747/8/11/475/s1, Figure S1. Frequency distributions of seed shattering rate and the sequence alignment of *HS1*. Figure S2. The information on haplotypes in *SH4* of 279 accessions in AA genome *Oryza*. Table S1. The seed shattering rate of Yundao 1, NIL-*hs1* and F1 hybrid.

**Author Contributions:** Y.Z., J.Z. and D.T. planned and designed the research. Y.Y., W.H.E., P.X., J.L., Y.Z.E.-R. and S.H. performed the laboratory experiments. Y.Z., J.Z. and Y.Y. analyzed the data together. Y.Z., J.Z. and D.T. finished the manuscript. All authors reviewed and approved the final manuscript.

**Funding:** This research was supported by the National Natural Science Foundation of China (Grant Nos. U1502265, 31660380, 31201196), Yunnan Provincial Science and Technology Department, China (Grant Nos. 2015HB079, 2018FG001-086) and the Yunnan Provincial Government (YNWR-QNBJ-2018-359).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**

NIL: Near isogenic line; AL: Abscission layer, V: Vascular bundle.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Allelic Di**ff**erentiation at the** *E1*/*Ghd7* **Locus Has Allowed Expansion of Rice Cultivation Area**

**Hiroki Saito 1,2,\*, Yutaka Okumoto 1, Takuji Tsukiyama 1,3, Chong Xu 1,4, Masayoshi Teraishi <sup>1</sup> and Takatoshi Tanisaka 1,4**


Received: 9 September 2019; Accepted: 25 November 2019; Published: 28 November 2019

**Abstract:** The photoperiod-insensitivity allele *e1* is known to be essential for the extremely low photoperiod sensitivity of rice, and thereby enabled rice cultivation in high latitudes (42–53◦ north (N)). The *E1* locus regulating photoperiod-sensitivity was identified on chromosome 7 using a cross between T65 and its near-isogenic line T65w. Sequence analyses confirmed that the *E1* and the *Ghd7* are the same locus, and haplotype analysis showed that the *e1*/*ghd7-0a* is a pioneer allele that enabled rice production in Hokkaido (42–45◦ N). Further, we detected two novel alleles, *e1-ret*/*ghd7-0ret* and *E1-r*/*Ghd7-r*, each harboring mutations in the promoter region. These mutant alleles alter the respective expression profiles, leading to marked alteration of flowering time. Moreover, *e1-ret*/*ghd7-0ret*, as well as *e1*/*ghd7-0a*, was found to have contributed to the establishment of Hokkaido varieties through the marked reduction effect on photoperiod sensitivity, whereas *E1-r*/*Ghd7-r* showed a higher expression than the *E1*/*Ghd7* due to the nucleotide substitutions in the *cis* elements. The haplotype analysis showed that two photoperiod-insensitivity alleles *e1*/*ghd7-0a* and *e1-ret*/*ghd7-0ret*, originated independently from two sources. These results indicate that naturally occurring allelic variation at the *E1*/*Ghd7* locus allowed expansion of the rice cultivation area through diversification and fine-tuning of flowering time.

**Keywords:** rice; flowering time; photoperiod sensitivity; allelic variation; fine-tuning

#### **1. Introduction**

Rice is a major cereal extensively cultivated in a wide range of latitudes from 55◦ N to 35◦ S. Because rice is formerly a facilitative short-day (SD) plant well adapted to warm climate, photoperiodic control of flowering time is a key factor in the regional and seasonal adaptability of rice varieties [1]. In high latitudes (>ca. 40◦ N), rice cultivation had been impracticable due to the short summer and long-day (LD) more than 15 h during the summer, until early flowering varieties with extremely weak photoperiod sensitivity were raised [2–4]. It was during 1900 to 1930 that such varieties were first released and planted in the northernmost rice cultivation area, Hokkaido, in Japan (42–45◦ N) [5]. The varieties raised for Hokkaido also enabled rice cultivation even in Hei Long Jiang province (43–53◦ N) of China [6].

Also in low latitudes (ca. 20◦ S–20◦ N), a recent rice breeding program aims to produce varieties with weak photoperiod sensitivity (PS), though a long basic vegetative growth period is necessary at the same time, because such a combination of the two traits for heading will permit almost constant and adequate vegetative growth periods under SD (less than 13.5 h) [7–9]. In addition, in middle latitudes (30–40◦ N), there is a close relation between the photoperiod sensitivity of varieties and the latitude of their cultivation area [3,4]. Thus, understanding of the genetic factors responsible for photoperiod sensitivity, as well as basic vegetative growth, will be essential for not only guaranteeing stable rice production but also allowing further expansion of rice cultivation area.

Genetic studies on rice flowering (heading) time started in 1915 [10]. Since then, many flowering time loci were reported: among them, *E1* [11–14], *Photosensitivity 1* (*Se1*) [15], and *Earliness 1* (*Ef1*) [16], have been intensively studied about their genetic characteristics, such as allelic variation, response to photoperiod, geographical distribution, and interaction with other loci. The geographical studies showed that these three loci play especially important roles in regional adaptabilities of Japanese and Taiwanese *japonica* rice varieties and *japonica*/*indica* cross varieties in Korea [4–6,15,17–23].

The Committee on Gene Symbolization, Nomenclature and Linkage Groups of the Rice Genetics Cooperative made a rule that the gene symbols which have been commonly used by many workers in the past should be retained [24], and recommended to categorize flowering time genes into three types, earliness and lateness (gene symbol: *E*), photoperiod sensitivity (gene symbol: *Se*), and basic vegetative growth (gene symbol: *Ef*). With the advance of quantitative trait locus (QTL) analysis, however, it has become difficult to categorize newly found QTLs into three types because they are detected only from flowering time data. Since then, the gene symbols, *E*, *Se*, and *Ef*, did not come to be retained. Recent molecular genetic analyses identified three key flowering time loci, *Heading date 1* (*Hd1*) [25], *Early heading date 1* (*Ehd1*) [26], and *Grain number, plant height, and heading date 7* (*Ghd7*) [27], all of which were named regardless of the rule, and subsequent studies on these three loci provided new numerous molecular-based knowledges of rice flowering. Similarly, about the *E1*, *Se1* and *Ef1* loci, the information useful for rice breeding has been accumulated with enormous numbers until now. Therefore, it is significant to clarify the relationships of the three loci, *E1*, *Se1* and *Ef1*, to the loci named regardless of the rule. To date, the *Se1* and *Ef1* loci proved to be identical with the *Hd1* [21,28] and the *Ehd1* loci [7], respectively.

The *E1* was first identified as a late flowering time locus: the functional allele *E1* is completely dominant over the nonfunctional allele *e1* [11,12]. This locus was also involved in plant height. Later, this locus proved to control PS [13], and its functional allele *E1* was shown essentially important in rice varieties for temperate areas in Japan (30–40◦ N) [17,18] because of firmly inhibiting the panicle primordial differentiation under LD until it becomes SD conditions, and thereby ensuring normal vegetative growth and stable yields. In contrast, a photoperiod-insensitivity allele *e1* was found to be essential for the varieties commercially cultivated in Hokkaido (42–45◦ N), because of its marked reducing effect on photoperiod sensitivity: use of *e1* enabled rice cultivation in high latitudes where LD conditions continue during the summer [5,21]. The *E1* locus was found to be located on chromosome 7, linked to the *rfs* (rolled fine strip) and *slg* (slender glume) loci with recombination values of 16.3% and 9.1%, respectively [29]. This locus has been well investigated for its effects on photoperiod sensitivity and regional adaptabilities of rice plants [5,11–19], but little is known about the relationship with the loci which were identified by molecular genetic analysis and named regardless of the rule. Recently, the *Ghd7* locus was precisely mapped on chromosome 7, and this locus exert major effects on not only heading date but also number of grains per panicle and plant height [27]. In addition, subsequent molecular analyses of the *Ghd7* locus demonstrated that a loss-of-function allele of *Ghd7* is essential for the extremely early flowering of Hokkaido varieties [30–32]. These reports make a conjecture that the *E1* and *Ghd7* are the same locus.

In the present study, we first analyzed the effects of three alleles at the *E1* locus on photoperiod sensitivity using the Taiwanese *japonica* rice variety "Taichung 65 (T65)" harboring *E1* [33], its isogenic line T65m harboring *e1* [16,33–35], and T65w that harbors a chromosome segment of *O. rufipogon* Griff. including the *E1* locus in the genetic background of T65 [36]. Subsequently, we attempted to determine the precise chromosomal location of the *E1* locus using the progenies from T65 × T65w, and then conducted sequence analysis to learn the sequences of the three alleles, also to investigate

the relationship between the two loci, *E1* and *Ghd7*. We finally applied a haplotype analysis of the chromosomal region surrounding the *E1*/*Ghd7* locus to 44 Hokkaido and 50 Japanese-core-collection varieties in order to prove correctness of the findings by Okumoto et al. (1996) [5] and Ichitani et al. (1998) [21] that *e1* is the key allele for establishing the varieties for the northernmost rice cultivation area, and its history and origin.

#### **2. Results**

#### *2.1. Photoperiod Sensitivities of T65, T65w, and T65m*

Days to heading (DH) of T65, T65m, and T65w under a SD were 84.7, 81.4, and 82.0 respectively, while those under a LD were 95.6, 90.4, and 118.0, respectively (Figure 1). Thus, the photoperiod sensitivities of T65, T65m, and T65w were estimated at 11.1, 9.0, and 36.0, respectively. Since T65m is an isogenic line of T65 for the *E1* locus, the weaker PS of T65m was attributable to the photoperiod-insensitivity allele e1 at the *E1* locus. T65w showed far stronger photoperiod sensitivity than T65 and T65m. The genotypic difference between T65w and T65 is only in the chromosome region including the *E1* locus, where only T65w harbors a chromosome segment induced from *O. rufipogon* Griff. Since any other photoperiod sensitivity genes have not yet been reported in this region, we conclude that the chromosome segment introduced from *O. rufipogon* Griff. in T65w certainly harbors a strong photoperiod-sensitivity allele, probably at the *E1* locus.

**Figure 1.** Days to heading of T65, T65m, and T65w under short-day (SD, white bar) and long-day (LD, black bar) conditions.

#### *2.2. Chromosomal Location of the E1 Locus*

The F2 population from the cross between T65 and T65w, comprising 205 plants, showed a continuous distribution of DH within the parental ranges (Figure 2a). We conducted a progeny test using 38 F3 lines, which were derived from randomly selected F2 plants. In the test, all the F3 lines were clearly classified into three groups. The ratio of [T65-type]:[segregating-type]:[T65w-type] lines was 13:14:11, which fitted the 1:2:1 ratio expected for one-locus segregation (χ<sup>2</sup> = 8.904, P > 0.05) (Table S1). In contrast, the F2 population from the cross between T65w and T65m showed a bimodal distribution of DH within the parental ranges, with a clear breakpoint dividing the population into early (T65m-type) and late (T65w-type) groups (Figure 2b). The ratio of early type (34 plants): late type (91 plants) fitted the 1:3 ratio expected for one-locus segregation (χ<sup>2</sup> = 0.570, P > 0.05). In the progeny test, all the 40 F3 lines were clearly classified into three groups. The ratio of [T65m type]:[segregating type]:[T65w type] lines fitted the 1:2:1 ratio expected for one-locus segregation (χ<sup>2</sup> = 0.150, P > 0.05) (Table S2). T65m is an isogenic line of T65 harboring a recessive allele *e1* at the *E1* locus. We accordingly inferred that T65w harbors a novel allele at the *E1* locus, whose heading-date delaying effect was stronger than *E1* in T65. We designated this allele *E1-r* (a novel photoperiod-sensitivity allele at the *E1* locus).

**Figure 2.** Distributions of days to heading in two F2 populations from crosses between (**a**) T65 × T65w and (**b**) T65w × T65m. The black bar indicates the range of days to heading of T65w. The white bar indicates the ranges of days to heading of (**a**) T65 and (**b**) T65m. The arrow indicates the breakpoint between early and late heading groups.

Using 546 F3 plants from the cross between T65 and T65w, we tried to identify the chromosomal location of the *E1* locus. The result showed that the *E1* locus was present in the region with a physical distance of 4.11 Mb between two simple sequence repeat (SSR) markers, RM1253 and RM3635, on chromosome 7 (Figure 3). Subsequently, we attempted to narrow down the candidate region of the *E1* locus, using 1263 F4 progenies derived from several F3 recombinants between RM1253 and RM3635; consequently, the chromosomal location of the *E1* locus was narrowed down to the region with a physical length of approximately 228.1-kb between RM5436 and RM21341 (Figure 3). In this region, 11 genes are reported in Rice Annotation Program Database [37]. Among them, we proposed that Os07g0261200, which was reported as *Ghd7*, a repressor of flowering time under LD conditions [27], was likely to be a candidate of *E1*.

**Figure 3.** Map-based cloning and graphical genotypes of the candidate region of the E1 locus. "T" and "H" at each marker indicate T65 homozygous and heterozygous, respectively. "T" and "W" at plant type indicate T65-type (early heading) and T65w-type (late heading), respectively.

Sequence analyses showed that the sequences of the alleles at the *Ghd7* locus in T65 and T65m were completely consistent with a functional allele *Ghd7-2* [38] and a nonfunctional allele *ghd7-0a* [27], respectively (Figure 4a,b). Since the genotypic difference in flowering time between T65 and T65m is only at the *E1* locus, this suggests that *E1* and *Ghd7* are the same locus (hereafter we tentatively designate *E1* (=*Ghd7*) as *E1*/*Ghd7*), and that T65m flowered earlier than T65 because the former harbors a loss-of-function allele *e1*/*ghd7-0a*. In contrast, the allele of T65w at the *E1* locus harbored four nonsynonymous substitutions and two nucleotide substitutions in the promoter region (Figure 4a,b). Among the substitutions, two in the promoter region were in the transcriptional signal motifs (cis

elements): low temperature response element (LTRE) core actor (located at −284) and the TATA box (located at −564). Thus, the two nucleotide substitutions were considered to modify the expression of *E1*/*Ghd7*. Subsequent expression analysis of *E1*/*Ghd7* showed that the expression of T65w was higher than that of T65 (Figure 4c). This suggests that the late flowering of T65w is caused by high expression of *E1*/*Ghd7* due to the nucleotide substitutions in the cis elements.

**Figure 4.** (**a**) Schematic diagrams of the alleles at the *E1*/*Ghd7* focused on nucleotide substitutions among T65 T65m, and T65w. (**b**) Alignments of amino acid sequences of the alleles at the *E1*/*Ghd7* locus. The white and black characters with black and gray cells indicate amino acid substitutions and CONSTANS, CO-like, and TOC1 (CCT)-motif, respectively. The box indicates the CCT-motif region. *Ghd7-1*, *Ghd7-2* and *Ghd7-3* were functional alleles [29]. (**c**) Comparison of the expression level of the allele at the *E1*/*Ghd7* locus between T65 and T65w.

#### *2.3. A Novel Nonfunctional Allele at the E1*/*Ghd7 Locus*

Okumoto et al. (1996) [5] showed that nine Hokkaido varieties tested all harbored a nonfunctional (photoperiod-insensitivity) allele *e1* at the *E1* locus thorough a conventional genetic analysis, and assumed that this allele has played an essential role in the establishment of rice varieties for the Hokkaido district. To confirm this assertion, we analyzed the presence of the nucleotide substitution from GAG (Glu) to TAG (stop codon) in exon 1 at the *E1*/*Ghd7* locus (Figure 4a) of 44 Hokkaido varieties using a cleaved amplified polymorphic sequence (CAPS) marker. The result showed that 37 varieties harbored the *e1*/*ghd7-0a* allele, and 7 varieties did not (Table S3). This single nucleotide substitution

was not observed in EG5 (Aikoku), which is one of the tester lines for the *E1*, *E2* and *E3* loci involved in the flowering time, and which harbors *e1* allele at the *E1* locus [11–13]. Sequence analysis for the EG5 revealed that a Ty1-copia like retrotransposon (TE) was inserted in the promoter region of the *E1*/*Ghd7* allele (Figure 5a). The seven varieties, which did not harbor the *e1*/*ghd7-0a* allele, also harbored the same TE insertion. We named this novel nonfunctional allele *e1-ret*/*ghd7-0ret*. The expression of the *e1-ret*/*ghd7-0ret* allele was far lower than the *E1*/*Ghd7-2* allele in the Japanese variety "Nipponbare" with the reference genome (Figure 5b), indicating that *e1-ret*/*ghd7-0ret* confers extremely weak photoperiod sensitivity by losing the normal function of the promoter.

**Figure 5.** (**a**) Schematic diagram of the allele at the *E1*/*Ghd7* locus in "EG5 (Aikoku)". (**b**) Comparison of the expression level of the allele at the *E1*/*Ghd7* locus among three Japanes varieties "Nipponbare" (NH), "EG5 (Aikoku)" and "Kirara397". "EG5 (Aikoku)" and "Kirara397" are a tester line for the *E1*, *E2* and *E3* locus and an elite Hokkaido variety, respectively.

#### *2.4. Haplotype Patterns of the Chromosomal Region Surrounding E1*/*Ghd7 Locus*

We surveyed DNA polymorphisms between EG5 (Aikoku) (*e1-ret*/*ghd7-0ret*) and Kirara397 (*e1*/*ghd7-0a*) around *E1*/*Ghd7* locus. Subsequently, we found three polymorphisms (two SNPs and a 20-bp deletion) other than the nucleotide substitution from GAG (Glu) to TAG (stop codon) and the TE insertion. To know the origins of two nonfunctional alleles *e1*/*ghd7-0a* and *e1-ret*/*ghd7-0ret*, we investigated the haplotypes of Hokkaido and Japanese-core collection varieties using five markers surrounding the *E1*/*Ghd7* locus (three single nucleotide polymorphisms (SNPs), a 20-bp deletion, and a TE insertion). The Japanese-core-collection varieties were classified into at least four haplotypes, Hap2, Hap3, Hap4, and Hap5 (Figure 6 and Table S4). In contrast, Hokkaido varieties were classified into two distinct haplotypes, Hap1 and Hap3. This suggests that Hap1 was derived from Hap2 via nucleotide substitution from GAG (Glu) to TAG (stop codon) in exon 1, whereas Hap3 was derived from Hap4 via the TE insertion in the promoter region. These results indicate that two independent mutational events contributed to the occurrence of the two nonfunctional alleles *e1*/*ghd7-0a* and *e1-ret*/*ghd7-0ret*. Interestingly, although Hap1 was found only in Hokkaido varieties, Hap3 was found not only in Hokkaido varieties but also in some Japanese varieties, particularly in the Aikoku-related varieties (Figure 6 and Table S4). Further, the varieties of the Hap3 group, except for Hokkaido varieties, flowered about 20 days later than the Hokkaido varieties, implying that such varieties do not adapt to Hokkaido where autumn comes early (Figure 6 and Table S4). These findings indicate that other genetic factor (s) were involved in the early flowering of Hokkaido varieties belonging to the Hap3 group. We accordingly investigated allelic variations in the *Se1*/*Hd1* and another major photoperiod-sensitivity gene, *Hd5*, which is known to be involved in the PS in the Hokkaido varieties [30–32]. The results showed that varieties with *e1*/*ghd7-0a* flowered early regardless of harboring a functional allele(s) (photoperiod-sensitivity allele) at the *Se1*/*Hd1* and/or *Hd5* locus, whereas varieties with *e1-ret*/*ghd7-0ret* flowered early only when harboring a nonfunctional allele at either of the *Se1*/*Hd1* or *Hd5* locus (Figure 7). These results indicate that coexistence of *e1-ret*/*ghd7-0ret* with a photoperiod-insensitivity allele either at the *Se1*/*Hd1* or at the *Hd5* locus is necessary to promote flowering under LD conditions.

**Figure 6.** Haplotypes around the *E1*/*Ghd7* locus and days to heading of Hokkaido and Japanese core collection varieties.

**Figure 7.** Gene combinations for the *E1*, *Se1*, and *Hd5* loci and days to heading. 1) This "Aikoku" variety belongs to the Japanese core collections. 2) This "Akage" variety belongs to the Hokkaido varieties.

#### **3. Discussion**

Since Hoshino (1915) [10], many genes (loci) controlling flowering time have been reported (reviewed in [39–41]). Among them, the *E1* is an important locus closely associated with the regional adaptability of rice varieties: its photoperiod insensitivity allele *e1* enabled rice cultivation even in Hokkaido, one of the northernmost rice cultivation area (42–45◦ N) [5,13,21]. A dominant photoperiod-sensitivity allele *E1* at the *E1* locus widely deployed among Japanese varieties for all regions other than Hokkaido in Japan [17,18,21]. Ichitani et al. (1998) [21] reported that the *E1* locus was identical to the *Heading date 4* (*Hd4*) locus, which was identified by Quantitative Trait Locus (QTL) analysis of flowering time using progenies from the cross of the *indica* variety Kasalath and the *japonica* variety Nipponbare [25]. Fujino and Sekiguchi (2005) [30] identified two QTLs, *qDTH-7-1* and *qDTH-7-2*, for flowering time using progenies from the cross between two Hokkaido varieties, Hoshinoyume and Nipponbare. They concluded that *qDTH-7-1* is the same locus as the *E1* (*Hd4*). Later, Xue et al. (2008) [27] isolated *the grain number, plant height, and heading date 7* (*Ghd7*) locus on chromosome 7, whose functional allele *Ghd7* encodes a *CONSTANS*, *CO-like*, and *TOC1* (CCT) domain-containing protein that delays flowering under LD conditions. They also reported that a nonfunctional allele *ghd7-0a*, harboring a premature termination in the predicted coding region, deployed among varieties commercially cultivated in Hei Long Jiang province, China (43–53◦ N). In the present study, we successfully determined the chromosomal location of the *E1* locus within a 228.1-kb physical region on chromosome 7 (Figure 3). According to the rice public database RAP-DB [38], 11 loci (genes) exist in this region. Among the genes, only Os07g0261200 (=*Ghd7*) showed a SNP in exon 1 of a photoperiod insensitive allele *e1* in T65m (Figure 4). This substitution was the same as that of the nonfunctional allele *ghd7-0a* [27]. From these results, we concluded that the *E1*, *Hd4*, *qDTH-7-1* and *Ghd7* are the same locus. Then we finally designate this locus as *E1*/*Ghd7*. It is noteworthy that we identified a novel photoperiod-insensitivity allele *e1-ret*/*ghd7-0ret* that harbored a TE insertion in the promoter region of *E1*/*Ghd7*. We conclude that this TE-insertional mutation causes the loss of function of the *E1*/*Ghd7* allele.

In addition, we identified a novel strong photoperiod-sensitivity allele *E1-r*/*Ghd7-r*, which harbors three nonsynonymous substitutions in the coding sequence (CDS) and two SNPs in the promoter region: one is in the LTRE core actor (CCGAC), and the other is in the TATA-box (Figure 5a). The LTRE core actor was identified in the regulatory regions of all cold-induced genes in Arabidopsis [42]. The CCGAC core

motif, also known as C-repeat / drought response element (CRT/DRE), is essential for transcriptional activation in response to cold, drought, and/or high-salt treatments [43]. Kim et al. (2002) [44] showed that light signaling mediated by phytochrome activates cold-induced gene expression through CRT/DRE. The expression of *E1-r*/*Ghd7-r* was higher than that of a PS allele of *E1*/*Ghd7-2* in Nipponbare (Figure 5b). The expression level of *E1*/*Ghd7* is regulated by red light signal and correlated well with its LD specific activity [45,46]. These suggest that the mutations in the promoter region modify the *E1-r*/*Ghd7-r* expression, which delays flowering under LD conditions. Elucidation of the influences of the three amino acid substitutions in the CDS of *E1-r*/*Ghd7-r* awaits further study.

Haplotype analysis showed that two photoperiod-insensitive alleles, *e1*/*ghd7-0a* and *e1-ret*/*ghd7-0ret*, originated independently from two sources (Figure 6). *e1*/*ghd7-0a* widely deployed among improved and landrace varieties in Hokkaido, including "Akage", which was one of the pioneers of the Hokkaido rice varieties in the late 1800's [47]. This indicates that *e1*/*ghd7-0a* is a pioneer allele, leading to raising extremely early heading varieties with extremely weak photoperiod sensitivity during 1900–1930 (see Introduction). In contrast, *e1-ret*/*ghd7-0ret* was detected in "Fukoku", one of the past leading varieties in Hokkaido. "Fukoku" was bred from the cross between the Japanese warm region variety "Nakate-Aikoku", and the Hokkaido variety "Bozu 6". "Nakate-Aikoku" harbors *e1-ret*/*ghd7-0ret*, whereas "Bozu 6" does not. This indicates that *e1-ret*/*ghd7-0ret* of "Fukoku" was derived from "Nakate-Aikoku" (Figure S1). Interestingly, although most of the Aikoku-related varieties harbor *e1-ret*/*ghd7-0ret* (Table S4), none of them flowered as early as "Fukoku". This indicates that some genetic factor(s) other than *e1-ret*/*ghd7-0ret* is also responsible for the early flowering of "Fukoku". The varieties with *e1-ret*/*ghd7-0ret* flowered extremely early when harboring a nonfunctional allele either at the *Se1*/*Hd1* locus or at the *Hd5* locus (Figure 7). Therefore, we conclude that a nonfunctional allele either at the *Se1*/*Hd1* locus or at the *Hd5* locus is necessary for early flowering of the Hokkaido varieties with *e1-ret*/*ghd7-0ret*. Fujino et al. (2013) [32] reported that the nonfunctional allele at the *Hd5* locus was a spontaneous mutant gene that occurred in the Hokkaido local landrace "Bozu," and that this allele contributed to the expansion of rice cultivation to the northern area of Hokkaido. In contrast, the nonfunctional allele at the *Se1*/*Hd1* locus widely deployed among the varieties for the Tohoku-Hokuriku region (37–40◦ N) [15,18]. In the early rice breeding in Hokkaido, many Aikoku-related varieties, which were chiefly cultivated in the Tohoku-Hokuriku region, Japan, were frequently used as cross-parents to increase the genetic diversity and improve grain quality [47]. This suggests that*e1-ret*/*ghd7-0ret* was introduced from the Aikoku-related varieties, and the combination with the nonfunctional allele at the *Hd1* or *Hd5* locus made *e1-ret*/*ghd7-0ret* available for rice breeding programs in Hokkaido. It is also suggested that mutations in the promoter region often make functional differentiations of alleles at the *E1*/*Ghd7* locus, bringing about flowering time diversification in rice.

Recent molecular genetic studies revealed that there are large natural allelic variations in key loci controlling flowering time, such as *Ghd7*, *Hd1*, *DTH2*, and *DTH7*, which contribute to the diversity of flowering time, and to the regional adaptability by adjusting flowering time to distinct environmental conditions [27,32,48–51]. Zhang et al. (2015) [52] and Zheng et al. (2016) [53] classified *Ghd7* alleles into three (strong function, weak function and non-function) and two (function and non-function) groups based on the sequence polymorphisms in the coding region, respectively. On the other hand, Lu et al. (2012) [38] analyzed 104 varieties (*O. sativa*) and three wild rice accessions (*O. rufipogon*) and found that 76 SNPs and six insertions and deletions within a 3932-bp DNA fragment of *Ghd7*. Among them, the functional C/T mutation in the promoter region was related to plant height probably by altering gene expression. In this study, we identified two novel alleles with functional differentiations in the promoter region of the *E1*/*Ghd7* allele. The interaction of these alleles with other flowering time genes except *Hd1* and *Hd5* have not yet been elucidated, additional detailed analysis of their effects on flowering time will contribute to fine-tuning of the flowering time well adapting to the climatic conditions in each area.

Photoperiod sensitivity is an important trait responsible for regional and seasonal adaptability of rice varieties. In this study, we detected two novel alleles at the *E1*/*Ghd7* locus. It is known that many other loci are involved in the photoperiod sensitivity pathway of flowering in rice. Recent studies showed that photoperiod sensitivity loci, *Se1*/*Hd1*, *OsPRR37* and *Ghd8*, each have a large allelic variation [49,54–57]. Therefore, analyzing the functional and inter-locus interactions should be advanced, which will lead to practice the fine-tuning of flowering time in rice breeding programs. In addition, we elucidated that the *E1* and *Ghd7* are the same locus. Until now, numerous genetic information has been accumulated for each locus (*E1*, *Ghd7*). For further genetic analyses of the *E1*/*Ghd7* locus, however, all information about *E1* and *Ghd7* should be available.

#### **4. Materials and Methods**

#### *4.1. Photoperiod Sensitivities of T65, T65m, and T65*

The Taiwanese *japonica* rice variety "Taichung 65 (T65)", and its isogenic line T65m and its near-isogenic line T65w were used. T65 harbors a photoperiod sensitivity allele *E1* at the *E1* locus, while T65m harbors a photoperiod-insensitivity allele *e1*, which is derived from the cross with Bozu5 [16,19,34,35]. T65w is a chromosome substitution line which was developed by introducing the *E1* region of *O. rufipogon* (W107) into the genetic background of T65 [36]. Five seeds were sown on field soil in a 3.6 L pot and covered with granulated soils. Seedlings were thinned to one plant per pot 14 days after sowing, and were grown under two photoperiod conditions, SD (12-h light/12-h dark) and LD (14-h light/10-h dark). Photoperiod treatments were conducted using two growth cabinets without temperature control. In addition to natural daylight (8:00–18:00), supplementary artificial light was used for the 12-h and 14-h light conditions. The degree of photoperiod sensitivity of each line was expressed as a difference of DH between SD and LD. The experiment was conducted by using five plants per line with three replications from 1 May to October 16. Heading date was recorded for each plant when the first panicle emerged from the sheath of the flag leaf.

#### *4.2. Chromosomal Location of the E1 Locus*

Two F2 populations from crosses of T65 × T65w, comprising 205 plants, and T65w × T65m, comprising 125 plants, were subjected to genetic analysis of heading date. They were grown in a paddy field of Kyoto University, Kyoto, Japan (35◦01- N). Seeds were sown on April 25 in 2007, and seedlings were transplanted on May 16 in 2007. The progeny test was conducted with 40 F3 lines (sowing on April 30 in 2008 and transplanting on May 21 in 2008). Each F3 line with approximately 25 plants was the progeny of an F2 plant randomly selected from the F2 population. To narrow down the candidate region of the *E1* locus, the F4 progenies derived from several F3 recombinants between RM1253 and RM3635 were cultivated (sowing on May 1 in 2009 and transplanting on May 22 in 2009).

#### *4.3. Expression Analysis of E1*/*Ghd7*

Plants of T65, T65m, and T65w were grown in a growth cabinet with a temperature controller under a LD condition (14.5-h light, 30 ◦C/9.5-h dark, 25 ◦C) at 70% relative humidity. Seedlings were grown on sand in 3.6 L pots (two plants/pot) with additional liquid fertilizer (Kimura's B Culture Solution, Nippon Medical and Chemical Instruments Co., Ltd., Osaka, Japan). Thirty days before flowering, leaves were collected at 4-h intervals during that day. Total RNAs were extracted with Trizol reagent (Life Technologies Inc., Gaithersburg, Maryland, USA) according to the manufacturer's protocols. Total RNA was subjected to DNA digestion by treatment with RNase-free DNase I (Takara Bio Inc.). The transcriptor first-strand cDNA synthesis kit (Roche Applied Science, Indianapolis, Indiana, USA) was used to reverse-transcribe cDNA from 1 μg of RNA using anchored-oligo (dT)18 primers. Real-time PCR analysis was performed by TaqMan PCR using a LightCycler 1.5 (Roche Applied Science) according to the manufacturer's instructions. The primer sets of *Ghd7* and *UBQ* genes and Universal Probe Library probes of each gene were designed with ProbeFinder version 2.45 (Roche; https://www.roche-applied-science.com/). Primer and probe sequences are shown in Table S5. Expression analysis using the standard curve method was performed to determine the expression level

of each gene. The relative expression level of each gene was calculated using *UBQ* gene. The RNA gene standards for the seven genes were applied to their plasmids prepared by the pGEM-T Easy Vector System (Promega Corp., Madison, Wisconsin, USA) using PCR amplicons from the total RNA of T65.

#### *4.4. Identification of the Genotype at the E1*/*Ghd7 Locus*

To identify the *E1*/*Ghd7* genotype, we developed two DNA markers based on an SNP and a TE insertion. CAPS marker analysis was based on a nucleotide substitution from GAG (Glu) to TAG (stop codon) in exon 1, producing an additional restriction enzyme (*Spe* I) site. A pair of PCR primers (Ghd7\_CAPS\_F: 5- -CCAACTTGCCCTGTCTTCTT-3- , Ghd7\_CAPS\_R: 5- -AGCTGCTGCAAGCCAGTAAT-3- ) was designed to amplify the 950 bp. PCR was performed with a 20 μL reaction mixture containing 2 μL template DNA, 10× PCR buffer, 25 mM MgCl2, 2 mM of each deoxyriboside-triphosphate (dNTP), 0.2 μL *Taq* DNA polymerase (5 U/μL), 4 μL of a 2.5 mM solution of each primer, and 3.2 μL H2O. PCR conditions were as follows: 94 ◦C for 5 min, followed by 35 cycles (1 min at 94 ◦C, 1 min at 60 ◦C, and 2 min at 72 ◦C) with a final extension of 7 min at 72 ◦C. The amplified products were digested with *Spe* I at 37 ◦C for 6 h. After digestion, the nonfunctional allele produced three fragments (81, 287, and 582 bp), whereas the functional allele produced two fragments (81 and 869 bp). Amplicons and digested amplicons were separated on 1% agarose gel. After electrophoresis, the gel was stained with ethidium bromide, and the DNA fragments were visualized under UV light. Insertion and deletion (INDEL) marker analysis was based on an 1897-bp copia-like TE insertion. A pair of PCR primers (Ghd7\_INDEL\_F: 5- -CGTTTCAGCAATAGCATTATGG-3- , Ghd7\_INDEL\_R: 5- -GCGGGTAGTCATCGAACAG-3- ) was designed to amplify the 824 bp in the wild type and the 2721 bp in the insertion type. PCR was performed with a 20 μL reaction mixture containing 2 μL template DNA, 10× PCR buffer, 25 mM MgCl2, 2 mM of each dNTP, 0.2 μL *Taq* DNA polymerase (5 U/μL), 4 μL of a 2.5 mM solution of each primer, and 3.2 μL H2O. PCR conditions were as follows: 94 ◦C for 5 min, followed by 35 cycles (1 min at 94 ◦C, 1 min at 60 ◦C, and 2 min at 72 ◦C) with a final extension of 7 min at 72 ◦C. Amplicons were separated on 1% agarose gel. After electrophoresis, the gel was stained with ethidium bromide, and the DNA fragments were visualized under UV light.

#### *4.5. Haplotype Patterns of the Chromosomal Region around the E1*/*Ghd7 Locus*

Three DNA polymorphisms (two single-nucleotide substitutions and a 20 bp deletion) within the 100-kb chromosomal region surrounding *Ghd7* existed among 44 Hokkaido, 50 Japanese-core-collection [58] and 71 Aikoku-related varieties, which are considered to the derivatives of "Aikoku" variety based on their names. The Japanese rice core collection is a limited set of accessions representing, with a minimum set repetitiveness, the genetic diversity among Japanese rice varieties [58]. The two substitutions (Hap\_SNP1 and Hap\_SNP2) were detected by CAPS marker analyses. Hap\_SNP1 harbors a G-to-T substitution, resulting in changing the recognition site of the restriction enzyme *Hpy188* I. Hap\_SNP2 harbors a C-to-T substitution, resulting in changing the recognition site of *Hpy188* I. Primer pairs for Hap\_SNP1 and Hap\_SNP2 were designed to amplify 623 and 295 bp, respectively (Table S6). In addition, two polymorphism surveys of the *copia*-like TE insertion and the SNP in the first exon of *Ghd7* were performed. The CAPS and INDEL marker analyses to detect the insertion and SNP were described above.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2223-7747/8/12/550/s1, Figure S1: Pedigree of the varieties, which harbor *e1-ret*/*ghd7-0ret* allele. Bold with underline indicates that the varieties harbor *e1-ret*/*ghd7-0ret* allele. Gray indicates varieties whose allele is unknown. Other indicates that the varieties harbor *e1*/*ghd7-0a* allele., Table S1: Frequency distributions of days to heading in F3 lines (T65 × T65w), Table S2: Frequency distributions of days to heading in F3 lines (T65m × T65w), Table S3: Lists of Hokkaido and Japanese core collection varieties used in this study, Table S4: Lists of Aikoku-related varieties used in this study, Table S5: Primer and probe sequences for expression analysis, Table S6 Primer sequences for haplotype analysis. **Author Contributions:** Conceptualization, H.S., Y.O. and T.T. (Takatoshi Tanisaka); Methodology, H.S., Y.O. and T.T. (Takatoshi Tanisaka); Formal analysis, H.S., T.T. (Takuji Tsukiyama) and M.T.; Investigation, H.S. and C.X.; Writing—original draft preparation, H.S.; Writing—review and editing, H.S., Y.O. and T.T. (Takatoshi Tanisaka); Project administration, H.S. and T.T. (Takatoshi Tanisaka).

**Acknowledgments:** We thank the genetic resources center, NARO (National Agriculture and Food Research Organization) for providing the seeds of Hokkaido, Japanese-core-collection and Aikoku-related varieties. T65m was kindly provided from Shigetoshi Sato in Ryukyu University and Kuo Hai Tsai in National Chung-Hsing University. T65w was kindly provided from Yoshio Sano in Hokkaido University. We also would like to thank Syo Asano and Shinsuke Hamada for their supports of our experiments.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
