Development and evaluation of near-isogenic lines for brown planthopper resistance in rice cv. 9311

Brown planthopper (BPH) is the most destructive pest of rice in Asia. To date 29 BPH resistance genes have been identified, but only a few genes are being used in breeding due to inefficient markers for marker-assisted selection (MAS) and little knowledge of the real effects of the genes. In this study we individually transferred 13 genes or QTLs (Bph14, QBph3, QBph4, Bph17, Bph15, Bph20, Bph24, Bph6, Bph3, Bph9, Bph10, Bph18 and Bph21) into cultivar 9311 by marker assisted backcross breeding (MABB). Through positive and negative selection we narrowed the segments from donors containing Bph14, Bph15, Bph6 and Bph9 to 100–400 kb. Whole-genome background selection based on a high resolution SNP array was performed to maximize reconstitution of the recurrent parent genome (RPG 99.2–99.9%). All genes reduced BPH growth and development and showed antibiotic responses in seedlings. Based on genetic effects and amino acid sequences of genes in three clusters we inferred that Bph10 and Bph21 might be identical to Bph26, whereas Bph9 and Bph18 were different. Bph15 might be same with Bph17, but QBph4, Bph20 and Bph24 might be different. We believe that these NILs will be useful in rice BPH resistance research and breeding.

developed 20 . However, only a few genes have been exploited successfully in breeding for BPH resistance due to inefficient markers and inadequate knowledge of the actual effects of the resistance genes.

Results
Development of monogenic NILs. To develop monogenic NILs, 13 BPH resistance genes or QTLs, identified in nine donor accessions were individually incorporated into 9311 by MABB (Fig. 1). The entire scheme took 9 crosses, 4 generations of backcrossing and one generation of selfing. In each backcross generation, individuals heterozygous at the target locus were further backcrossed to the recurrent parent 9311.
Positive and negative selection. Among the 13 genes, four (Bph14, Bph15, Bph6 and Bph9) were transferred by positive and negative selection (Fig. 2). Taking Bph6 as an example, closely linked markers Y37 and RM17008 were used for positive selection during the introgression process. Two other markers, J6-7 and J6-10, approximately 28 kb upstream of Y37 and 32 kb downstream of RM17008, respectively, were used in negative selection. Among 2,000 BC 1 F 2 progenies (segregating at the Bph6 locus), by selection on one side of the Bph6 locus, one plant heterozygous in the region near the Bph6 allele and homozygous for the 9311 region at the J6-7 locus, was selected and then backcrossed to 9311 to produce the BC 2 F 1 . Then five selected plants heterozygous at Bph6 were selfed to produce a BC 2 F 2 population. Among 3,000 BC 2 F 2 progenies, two plants heterozygous for Bph6, and homozygous for the 9311 allele at the J6-10 locus, were selected. Finally, the linked segments around Bph14, Bph15, Bph6 and Bph9 were narrowed to less than 100 kb, 400 kb, 100 kb and 200 kb, respectively (Fig. 2).
Background selection. The selected recombinants B14-2, B15-2, B6-3 and B9-3 heterozygous at the Bph14, Bph15, Bph6 and Bph9 loci, respectively, were each backcrossed to 9311 to produce advanced backcross (BC) generation materials. In the BC 3 to BC 4 generations, the 6 K SNP chip 30 was used to select individuals with the highest RPG. Selected individuals were self-pollinated to produce the BC 4 F 2 generations. As shown in Fig. 3     Agronomic performance of NILs and their hybrids. A number of agronomic traits were measured, including days to heading (DTH), plant height (PH), panicle number (PN), number of grains (NG), number of grains per panicle (NPG), spikelet fertility (SF), 1000-grain weight (GW), and yield per plant (YD). Bph3-NIL and Bph21-NIL showed significant decreases in GW compared to 9311, QBph3-NIL had significantly higher SF and GW, resulting in increased yield ( Table 1). The improved hybrids H2613S/QBph4-NIL and H2613S/Bph15-NIL had significantly higher SF compared to the conventional hybrid H2613S/9311. H2613S/Bph24-NIL showed significantly higher NG and NPG, but lower GW, leading to an equal yield compared to the conventional hybrid (Table 2). However, for most traits comparisons, there were no significant differences between the NILs and 9311, improved hybrids and conventional hybrids (Tables 1 and 2).   Seedling response. After about 14 days of BPH infestation in the greenhouse the 9311 control showed 100% wilting whereas the NILs were surviving. In short, the NILs harboring single genes or QTL on chromosome 4 (QBph4, Bph17, Bph15, Bph20, Bph24 and Bph6) had higher resistance than those on chromosome 12 (Bph10, Bph18 and Bph21), except for Bph9 (Fig. 4A). The Bph24-NIL had the lowest response score (1.3), representing the highest level of seedling resistance among the 13 NILs. To explore the potential usefulness of these genes in hybrids, 13 monogenic hybrids between H2613S and the NILs were also evaluated. The hybrid response closely paralleled those of corresponding NILs (Fig. 4B). Hybrids carrying Bph14, QBph4, Bph17, Bph6, Bph3, Bph9 and Bph10 showed lower levels of resistance than the corresponding NILs, indicating incomplete dominance of these genes. However, there were no significant differences between the response scores of the remaining six monogenic NILs and their hybrids, indicating complete dominance of those genes (Fig. 4C).

Honeydew excretion and survival rate of BPH on NILs.
To determine whether the presence of resistance genes affected BPH growth and development, we compared the areas of honeydew excretion for BPH feeding on each NIL. The results were in accordance with the seedling responses (Fig. 5A,B), suggesting that honeydew deposition is a simple measurable indicator of BPH fitness on the lines. The excretion areas were grouped into four classes: the smallest, QBph3, Bph9, Bph15 and Bph20 (approximate area 10 mm 2 ); small, QBph4, Bph3, Bph6, Bph14, Bph17 and Bph24 (20 mm 2 ); higher, Bph10, Bph18, and Bph21 (50 mm 2 ); the highest, 9311 (130 mm 2 ) (Fig. 5A).
BPH survival rates on the NILs mostly paralleled the data for honeydew accumulation and seedling response. The lower effectiveness of the NILs with QBph4 and Bph24 compared to those with Bph17, Bph15 and Bph20 suggests these two genes might have different mechanisms of resistance. To explore whether the genes in the clusters surrounding Bph14, Bph26 and Bph17 were the same, we sequenced and compared the amino acid sequence of the Bph26 alleles from Bph9-NIL, Bph10-NIL and Bph21-NIL, the Bph14 allele from the QBph3-NIL and Bph14-NIL, and the Bph17 allele from the QBph4-NIL, Bph15-NIL, Bph17-NIL, Bph20-NIL and Bph24-NIL. The proteins encoded by Bph26 alleles from the monogenic NILs carrying Bph10 and Bph21 were identical in amino acid sequence as the cloned Bph26. Bph18 and Bph26 are different alleles with many amino acid substitutions 12 . However, due to inability to obtain the PCR product of the third exon of the Bph26 allele from the Bph9-NIL using six pairs of specific primers, only the first and second exons were compared and a few nucleotide polymorphisms causing amino acid substitutions were detected (Fig. 6A). Compared to Bph14, the Bph14 allele from QBph3-NIL had a number of amino acid substitutions in the LRR domain (Fig. 6B). Compared to the amino acid sequence of the cloned Bph17, that from Bph15-NIL was the same, while that from NILs of QBph4, Bph20 and Bph24 showed several substitutions (Fig. 6C).

Discussion
To achieve improvement in target traits by MABB, breeders aim to minimize the introgressed segments from donors in order to reduce linkage drag, and to maximize reconstitution of recurrent parent genomes. Previously, improved lines contained large fragments (> 1,000 kb) of the target gene regions, due to lack of suitable closely linked molecular markers and limited knowledge of the actual chromosomal locations of the resistance genes 14 .
To reduce linkage drag, we performed positive and negative selection to obtain resistant recombinants between flanking markers in target regions based on high resolution physical maps of BPH resistance genes in two large backcross populations (BC 1 F 2 and BC 2 F 2 ). Introgressed segments containing four genes (Bph14, Bph15, Bph6 and Bph9) were finally narrowed to less than 400 kb (Fig. 2). In contrast, the introgressed segments for the remaining nine genes exceeded 1,000 kb as negative selection was not employed (Supplementary Fig. S1).
In previous MABB programs, background selection was based on RFLP and SSR markers which had low resolution and inadequate whole genome coverage 15,21 . With development of next generation sequencing technology, large numbers of SNPs became available, and two breeding chips RICE6K and RiceSNP50 with high-throughput SNP arrays were developed in China 30,31 , making efficient whole-genome background selection a reality. Background selection with high resolution SNP markers was used in rice breeding to improve blast resistance and wide compatibility 16,32 . In the present study, the breeding chip RICE6K was employed in background selection for improving BPH resistance. Undesirable donor segments in each NIL could be viewed on the haplotype map produced by the chip. For example, after backcrossing and background selection, Bph6-NIL and Bph9-NIL only had four and three short segments from donors, and the RPG recoveries were 99.8 and 99.9%, respectively  A and B) and survival rates of BPH on NILs (C). Uppercase letters above the error bars in A indicate significant differences in ranking by Duncan's multiple range test at P < 0.01. Error bars, SEM. Tests were conducted in eight replications. (Fig. 3). Our work demonstrated that positive and negative selection of target loci and whole genome background selection based on the 6 K SNP chip was a powerful way of developing monogenic NILs.
To date, eight BPH resistance genes, Bph1, bph2, Bph3, Bph6, Bph14, Bph15, Bph18 and Bph27 (t) have been individually incorporated into indica or japonica varieties by MABB [18][19][20][21][33][34][35] . These NILs or ILs carrying single resistance genes have been reported as resistant to one or more BPH biotypes predominating in various countries. However, the real effects of these genes cannot be compared using the original source genotypes due to the diverse genetic backgrounds in which additional resistance QTLs may be present. In the present study, 13 BPH resistance genes were separately introduced into cv. 9311 by MABB. Generally, the NILs carrying QBph4, Bph15, Bph17, Bph20, Bph24 and Bph6 on chromosome 4 showed significantly higher resistance than those carrying Bph10, Bph18 and Bph21 on chromosome 12, indicating that the gene cluster on chromosome 4 was more effective in conferring BPH resistance. The response scores of most NILs consistently matched seedling response, honeydew deposition area, and BPH survival rate scores, indicating a similar resistance mechanism. However, the Bph24-NIL showed the highest level of seedling resistance and the least honeydew excretion, but a higher BPH survival rate, indicated that Bph24 might mediate a resistance mechanism than antibiosis (Figs 4 and 5). However, the RPG coverage differs among the 13 NILs and this may have some effect on the agronomic performance and BPH responses (Table 1 and Supplementary Fig. S1). Previously, Hu et al. 18 proved that improved hybrid rice containing Bph14 and Bph15 showed enhanced resistance compared to a conventional line. In our study, hybrid F 1 descendants of H2613S and NILs showed higher resistance than conventional hybrid rice that lacked resistance genes . Bph14, QBph4, Bph17, Bph6, Bph3, Bph9 and Bph10 showed significantly less resistance in hybrids than corresponding NILs, indicating incomplete dominance (Fig. 4).
It is noteworthy that multiple BPH resistance genes cluster together on rice chromosomes; eight genes (Bph1,  bph2, Bph9, Bph10, Bph18, Bph19, Bph21 and Bph26) are clustered on chromosome 12 L, and six (QBph4, Bph15, Bph12, Bph17, Bph20 and Bph24) on chromosome 4S 36 . These gene clusters might involve different genes, different alleles at a single locus, or even the same gene with different haplotypes 21 . Based on response phenotypes of the NILs and comparisons of amino acid sequence with the cloned Bph26, Bph14 and Bph17, some aspects were resolved. The amino acid sequences of Bph10 and Bph21 were identical to Bph26 and the response phenotypes of the two monogenic NILs were similar to that of Bph26-NIL, whereas Bph9 and Bph18 were different (Figs 4, 5 and 6). We inferred that Bph10, Bph21 and Bph26 might be the same gene, but Bph9 and Bph18 were likely different alleles in this locus. QBph3 has a number of amino acid substitutions compared with Bph14. Hu et al. 22 reported that QBph3 and Bph14 were tightly linked on chromosome 3 L, but the QBph3-NIL showed a higher degree of resistance than Bph14-NIL. Thus, they might be alleles or linked genes that mediate different resistance mechanisms. Bph15 shared the same amino acid sequence as Bph17, and the Bph15-NIL had a similar BPH response phenotype to the Bph17-NIL with resistance scores of 2.4 versus 2.7, honeydew deposition area of 11.8 vs 18.1, and survival rates 33% vs 38%. However, there were differences between Bph17 and the alleles from the QBph4, Bph20 and Bph24 NILs in amino acid sequence and response phenotype. These results indicated that Bph15 was likely to be identical to Bph17, whereas QBph4, Bph20 and Bph24 might be different. Lv et al. 35 and Hu et al. 22 reported that Bph15 and QBph4 were located proximally to Bph17, thus differing from the present findings. One possible reason is that besides Bph17 there is another resistance gene/QTL in the donor B5.
Cv. 9311 is an elite restorer line for two-line hybrid rice and HL CMS three-line hybrid rice because of its good adaptation, ideal plant type, good grain quality and high yield potential. However, due to the absence of disease (bacterial blight, blast) and insect (stem borer, BPH) resistance, the commercial application of 9311 is limited. Our 9311 NILs with high BPH resistance will provide a choice of parent lines for use in producing hybrid rice.

Methods
Collection of agronomic trait data from field experiments. The 26 NILs and hybrids containing the BPH resistance genes were planted in a randomized complete block design at Wuhan in the summer of 2015. Plots of each line consisted of two rows with 10 plants per row at a planting density of 17 cm between plants in a row and 27 cm between rows. The central eight plants from each plot, were used to measure agronomic traits including plant height (PH), days to heading (DTH), panicle number (PN), number of grains (NG), number of grains per panicle (NPG), spikelet fertility (SF), 1,000 grain weight (GW), and yield per plant (YD). There were three replications for each NIL and hybrid combination.
Seedling response assays. The BPH bioassay was performed as a modified bulk seedling test in the greenhouse, following the method of Pathak et al. 39 . Seeds of 9311 and test lines were sown as random groups in 50 cm × 30 cm × 10 cm (height) plastic trays. Seedlings were thinned to 12 plants per line at the three-leaf stage and infested with second and third instar nymphs at a density of 15 insects per seedling. When all 9311 seedlings (control) had died [10-12 days after infestation (DAI)] the plants in other lines were examined, and each seedling was given a score of 1, 3, 5, 7 or 9 according to the criteria described by Huang et al. 40 where higher scores indicate greater susceptibility to BPH. These tests were performed in three replications.
Honeydew area. Determination of areas of honeydew deposition followed the method of Du et al. 11 .
Thirty-day-old NILs and 9311 (control) were covered by inverted transparent plastic cups placed over a filter paper resting on plastic Petri dishes. After starving for 2 h, five fifth instar BPH nymphs were placed in each cup. Two days later, the filter papers were oven dried for 30 min at 60 °C and treated with 0.1% solution of ninhydrin in acetone. Areas of ninhydrin-positive deposits were measured using Image J software. Tests were conducted in eight replications. BPH survival rates. Survival rates were calculated following the method of Du et al. 11 . To determine nymph survival rates on rice lines, each plant was infested with 20 second and third instar nymphs and covered with a cylindrical plastic cup. Survival rates calculated as percentages of surviving nymphs divided by the total number of nymphs released at the beginning were recorded daily for 9 days.