*2.3. Development of BC3F4 Pyramided Lines Using Marker-Assisted Breeding*

Nine five-gene-pyramided lines at the BC3F4 generation, along with the recurrent and donor parents, were evaluated in the first crop season of 2018 at Taiwan Agricultural Research Institute (TARI), Taiwan. Significant variances were observed between the pyramided lines and parental rice genotypes for plant height, days to 50% flowering, panicle length, panicles/plant, panicle weight, number of grains/panicle, 1000-seed weight, and single plant yield (Table 2). The recurrent parent, TNG82, recorded mean grain yield of 36.8 g/plant, while the donor parent, IRBB66, was 30.1 g/plant. Six of the nine five-gene-pyramided lines, CNYBB5R4-275, -276, -278, -279, -285 and -287, produced significantly higher grain yields per plant than the recurrent parent, which ranged from 37.1 to 44.5 g/plant, and displayed a similar phenotype to the donor parent TNG82 (Figure 6).

**Table 2.** Agro-morphologic traits of parental and five-gene-pyramided BC3F4 genotypes.


LSD, least significant difference at 5% probability level.

**Figure 6.** Phenotype of the five-gene-pyramided BC3F4 genotypes compared with recurrent parental variety TNG82.

A significant difference was noted between the parental rice varieties and pyramided genotypes in grain quality traits (Table 3). The palatability among pyramided BC3F4 genotypes varied between 69.8 (CNYBB5R4-275) and 74.5 (CNYBB5R4-276). The protein content among pyramided BC3F4 genotypes varied between 6.4 (CNYBB5R4-276 and CNYBB5R4-286) and 7.4 (CNYBB5R4-272). The brown rice ratio for the five-gene-pyramided genotypes varied from 72.8% to 79.3%. The four genotypes, CNYBB5R4-272, -275, -276 and -278, were found to have higher head rice ratios, however, the amount of total milled rice was not significantly different from the recurrent parent, TNG82. The evaluation of agronomic traits in BC3F3 and BC3F4 provided us with an important selection criteria, which can select candidate lines with stable agronomic performances and resistance to disease.


**Table 3.** Grain quality of parental and five-gene-pyramided BC3F4 genotypes.

LSD, least significant difference at 5% probability level.

#### **3. Discussion**

Conventional backcross breeding is the primary method used to develop highly BB-resistant rice cultivars, but it cannot accurately transfer multiple genes into the cultivar by phenotypic screening and the process requires a significant amount of time [21,22]. Modified backcross pyramid breeding, combined with molecular marker-assisted selection, has already been demonstrated to increase the precision and efficiency of breeding [23–25]. Due to the relatively large amount of work involved with the MAS process, the conventional backcross breeding approach has been widely adopted in breeding programs designed to breed for BB resistance [10,26–28].

To date, many rice cultivars with broad-spectrum resistance against *Xoo* isolates have been developed; Singh et al. (2001) pyramided three *R* genes, *xa5*, *Xa13* and *Xa21*, in the *indica* rice cultivars PR106 and Jalmagna using MAS to enhance the bacterial blight resistance [6,15]; the four genes *Xa4*, *xa5*, *Xa13* and *Xa21* were introgressed into the recurrent parent lines Jyothi, IR50, Mahsuri, PRR78, KMR3 and Pusa 6B [26,29,30]; different BB-resistant genes, *Xa7*, *Xa21*, *Xa22* and *Xa23*, were also transferred to an elite hybrid rice restorer line, Huahui 1035, in order to improve BB resistance and enhance rice yield [31].

In Taiwan, many *japonica* rice cultivars lack BB resistance, resulting in significant yield loss in severely infected fields. One such variety is Tainung82, which was released in Taiwan for commercial cultivation in 2006. TNG82 is described as a popular *japonica* rice variety, with high-yield potential (6–7 t/ha), excellent grain quality, various culinary applications, and relatively low grain protein content (4.5%–5.5%). As an extremely valuable yet BB-susceptible variety, TNG82 was selected as the focus of this study to increase BB resistance through the introgression of five BB-resistant genes, *Xa4*, *xa5*, *Xa7*, *xa13* and *Xa21*.

The primary purpose of backcross breeding is to transfer one or multiple genes of interest, linked to desirable traits, from donor parents into a base variety for improvement, a process which typically requires six to eight backcrosses to recover the recurrent parent's phenotype [32]. However, in the MAS scheme, three to four generations of backcrossing is generally enough to achieve more than 99% of the recurrent parent genome [33]. The theoretical %RPG of each generation, BC1, BC2, BC3 and BC4, were 75%, 87.5%, 93.8% and 96.9%, respectively. Furthermore, the %RPG can be improved by using MAS for background selection [16,34]. The 80% and 89% recovery rates following two and three backcrosses were obtained from three-BB-gene-pyramided BC2 and BC3 genotypes, via MAS [35]. Balachiranjeevi et al. (2015) transferred the BB gene, *Xa21* and rice blast-resistant gene, *Pi54,* to DRR17A and were able to recover 73.4%, 84.8% and 93.4% RPG in the BC1, BC2 and BC3 generations, respectively.

In this study, the recurrent parent genome recovery rates in BC1F1, BC2F1 and BC2F2 were 73.8%, 83% and 84.7% (Figure 2), respectively. Compared with the theoretical %RPG, a relatively low background recovery rate was obtained, however, the results were consistent with those found in previous studies [36,37]. Marker-assisted backcrossing can accelerate the breeding process and facilitate a speedy recovery for most of the recurrent genome within a few generations [38], however, the population size of each backcross generation, linkage drag, number of background markers used and genetic background between two parents are considered to be factors that reduce the efficiency of MAB and %RPG [32].

Bacterial blight is one of the most destructive diseases affecting rice productivity in Asia. In Taiwan, rice production is frequently affected by BB in the second crop season, resulting in substantial yield loss. In recent years, BB has become a more prevalent threat, due to climate change [39]. XF89-b, a strong and stable Taiwanese epidemic pathogen, has also been used for genetic analysis and the mapping of BB-related resistance genes [40]. In our bioassays, artificial screening of BC2F3 progenies revealed that all genotypes containing at least one BB-resistant gene displayed a degree of increased resistance (Table 1, Figure 4). The BC2F3 progenies that pyramided more than three BB-resistant genes exhibited a very high level of BB resistance against the XF89-b strain, compared to parental lines (Figure 5). The lesion lengths were measured between 0.37 and 1.25 cm (Table 1). The data indicated that multiple BB-resistant genes pyramided in rice can improve resistance to *Xoo*. The BB pyramiding lines are expected to enhance the adaptability and durability necessary to provide resistance against the dynamic nature of the pathogen. In addition, the results suggest that the gene combinations containing the *Xa21* gene were most resistant, as evidenced by shorter lesions lengths, followed by *Xa4* + *Xa21*, *Xa7* + *Xa21* and *xa13* + *Xa21*, while lines with *Xa4* + *xa5*, *xa5* + *xa13* and *Xa7* + *xa13* were

less effective. These results are consistent with previous studies, which have shown the presence of *Xa21* to be correlated with high levels of persistent resistance against BB disease in rice [6,14,15,17,25]. *Xa21* is the cell surface receptor, kinase, which is able to provide resistance to *Xoo* infections; *Xa21* not only suppresses *Xoo* growth, but also triggers broad perturbation in rice transcriptomes and mediated signaling pathways, preventing *Xoo* infections [14].

The agronomic performance evaluation of BC3F4 derived in the genetic background of TNG82 revealed that all pyramided lines for most of the agro-morphological traits were, in general, similar to the recipient parent, TNG82. However, six candidate lines, CNYBB5R4-275, -276, -278, -279, -285 and -287, produced significantly higher grain yields per plant than the recurrent parent, which was further confirmed by the multilocation evaluation. In addition, three candidate lines, CNYBB5R4-276, -278 and -286, were not significantly different in palatability, protein, amylose, brown rice ratio, head rice ratio or total milled rice ratio, indicating that the BC3F4 pyramiding lines had grain quality consistent with TNG82. The data showed that there were no yield or grain quality reductions, but rather improvements, due to the pyramiding of the five BB-resistant genes.
