A set of BC₂F₅ BILs, derived from a cross between an elite rainfed lowland cultivar Swarna (O. sativa) also known as MTU 7029 as a recurrent parent and a wild accession O. nivara IRGC81848 as a donor parent was developed by Swamy (2009) and these BILs were self-pollinated in consecutive generations to obtain BC₂F₈ families by single panicle selection (Supplementary Table 1).

Ninety four BC₂F₈ lines along with recurrent parent Swarna were grown in wet seasons (Kharif) of 2014 and 2015 at Indian Institute of Rice Research (IIRR), Hyderabad. Experiments were conducted using Randomized Complete Block Design (RCBD) with five replications each. The following traits were evaluated. DFF – number of days from sowing to the time that 50% of the plants showed flowering; DM – duration in days from sowing to the time when more than 80% of the grains on the panicles were fully ripened; PH - length in centimeters from the soil surface to the tip of the highest panicle at the time of harvest; Number of tillers per plant (NT) – number of tillers at the time of harvest; Number of productive tillers per plant (NPT) – number of panicle-bearing tillers at the time of harvest; PW – weight of five panicles per plant; YLDP-weight of the harvested seeds per plant; BY weight of dried and cleaned seeds from 30 plants and Biomass (BM) per plant- weight of well-dried mature harvested plants without panicles. LS means were calculated on data pooled by years using PB tools (Version 1.4¹) and were used in further statistical analysis and QTL mapping.

Correlation was calculated based on two sample t-test with equal variances. CORREL function of the analysis tool pack was used and then imported in excel to find the correlation coefficient between two variables. Correlation among traits was computed at P < 0.05 and P < 0.01, respectively.

Total genomic DNA was isolated from fresh leaf samples of 94 BC₂F₈ plants and the parents following CTAB method (Doyle and Doyle, 1987). In all, 324 markers were tested for parental polymorphism. One hundred and twenty four SSRs were polymorphic of which 111 gave clear bands and also segregated in BC₂F₈ (Supplementary Table 2). PCR was carried out in thermal cycler (G-STORM, United States) with a final reaction volume of 10 μl containing 15 ng of genomic DNA, 1X assay buffer, 200 μM of dNTPs, 1.5 mM MgCl₂, 10 pmol of forward and reverse primer and 1 unit of Taq DNA polymerase (Thermo Scientific). PCR cycles were programmed as follows: initial denaturation at 94°C for 5 min followed by 35 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 1 min and a final extension of 10 min at 72°C. Amplified products were resolved in 3% agarose gel prepared in 0.5 × TBE buffer and electrophoresed at 120 V for 2 h. Gels were stained with ethidium bromide and documented using gel documentation system (Alpha Imager, United States).

Linkage map was constructed based on genotypic data of 94 BILs using 111 polymorphic SSR markers on all chromosomes using MAP function (BC₂RIL) of QTL IciMapping v4.1² using the Kosambi mapping function (Kosambi, 1944). QTL detection was carried out by Composite Interval Mapping, Inclusive Composite Interval Mapping (ICIM) and Single trait Multiple Interval Mapping (SMIM) method in QGene 4.4.0 software (Joehanes and Nelson, 2008). Analysis was undertaken using automatic parameter setting and controlling marker forward stepwise. The threshold of LOD for declaring the presence of significant QTL for each trait was determined using 1000 permutations and α = 0.05 in QGene 4.4.0 software. At α = 0.05, the LOD threshold values ranged from 2.8 to 8.8.

Chromosomal segment substitution lines were identified using genotypic data of 111 polymorphic loci in 94 BILs in the background of the recurrent parent Swarna using the software CSSL Finder³. Statistical analysis was carried out using PB tools (Version 1.4⁴) for the test of significance.

The mean values of parents and BILs in each year for nine traits and number of BILs showing significant trait increase over recurrent parent Swarna is given in Table 1. The highest range of variation was observed for PH, NT, YLDP, and BY in 2014 and for PH, PW, BY, and BM in 2015. Positive transgressive segregation was observed for seven traits in both years compared with Swarna. In 2 years an average of 34% of BILs exhibited 15% to 70% increase over Swarna for PH, PW, YLDP, BY, and BM. Significant pair-wise comparison (at P = 0.05) of replicated phenotypic data of the BILs with Swarna as control revealed that BILs 166-9S, 235S, 10-2S, 10-3-4S and 84S had significantly higher yield with mean of 14.43 to 26.40 g per plant but only in 1 year. Likewise, three BILs for DFF and PH, 9 BILs for BM and 30 BILs for PW were significantly different from Swarna but only in 1 year. The highest number of lines (11 lines) with significant improvement over Swarna was observed for the trait PW. One BIL, 220S showed significantly higher PW over Swarna in both years. Two BILs 14-3S and 166-32S had significantly lower PW compared to Swarna in 2014 (data not shown). All traits followed a normal distribution except NT in 2014 and PW in 2015 as shown in Supplementary Figures 1, 2.

Significant correlations were observed among the traits in both the years at P < 0.05 and P < 0.01 (Supplementary Table 3). In both years, significant positive correlation was observed between DFF and DM; PH with PW, YLDP and BM; NT and NPT; PW and YLDP; YLDP with BY and BM; and BM and BY. Significant negative correlation was observed between PH and NPT in both years. In 2014, DFF was positively correlated with NT and YLDP, but negatively with BY in 2015. NT and NPT showed significant positive correlation with YLDP, BY, and BM only in 2014. In 2015, positive significant correlation was observed between PH and BY; PW and BY, but were negatively correlated in 2014. Negative correlation was observed for DM with PW and BY; PH with NT in both years, but significant only in 2015.

Correlation analysis of both years mean data showed that YLDP has significant positive correlation with PH, NT, NPT, PW, BM, and BY. Significantly correlated traits DFF and DM were negatively correlation with both PW and BY. PH had significant positive correlation with PW, YLDP, and BM but negative correlation with NT and NPT. BM also showed significant positive correlation with PH, NT, NPT, and PW.

Considering all the 111 loci and 94 BILs, a total of 7488 alleles were detected covering all chromosomes. Swarna homozygous alleles accounted for 84%, O. nivara homozygous alleles 11% and heterozygous alleles 4%. The highest number of O. nivara alleles were detected on chromosome 1 (128 alleles), chromosome 2 (120 alleles) and chromosome 8 (108 alleles) and lowest number of O. nivara alleles were detected on chromosome 10 (10 alleles). The O. nivara introgressions among BILs ranged from 1.15 to 27.0%, with a mean of 10.1%. Number of heterozygotes at any locus ranged from 0 to 40 (42%). RM125 on chromosome 7 followed by RM519 on chromosome 12 showed the highest number of heterozygotes. BIL 138S had was highest number of 24 heterozygotic loci (21% of all loci).

Linkage map was constructed based on genotypic data of 94 BILs using 111 polymorphic SSR markers on all chromosomes. In all, 15 QTLs were identified for all traits except for BM, considering data of both years using CIM, ICIM, and MIM (Figure 1). Five QTLs were identified in 2014 and 10 QTLs in 2015 (Tables 2, 3). Two QTLs were common in both years for DM and PH considering CIM, ICIM and MIM. PV explained by these 15 QTLs ranged from 12 to 56%. O. nivara alleles were trait enhancing in 26% of QTLs for 3 traits.

In all, 15 QTLs were identified, of which only 2 QTLs were common in both years and their PV ranged from 13 to 27%. These were qDM2.1 and qPH1.1. The common QTL qDM2.1 was identified at chromosomal region RM8080-RM6318 and explained PV of 13% with LOD score 2.9. A major QTL qPH1.1 at marker interval RM226-RM431 was identified in both years with average LOD of 4.5 explaining PV of 20%. O. nivara allele was trait-enhancing in qPH1.1 and Swarna allele at qDM2.1 in both years.

A total of 5 QTLs were identified, of which 2 were also identified in 2015. One QTL qPH1.2 was identified at RM128-RM226 region with PV of 13%. One QTL each for NPT (qNPT1.1) and PW (qPW9.1) were identified with PV 13% each and LOD score of 3 and 3.5, respectively. The increasing effect of these three QTLs qPH1.2, qNPT1.1 and qPW9.1 was from O. nivara.

In all, 10 QTLs were identified, of which 2 were also identified in 2014. The PV explained by these 10 QTLs ranged from 12 to 56%. Three QTLs each were identified for YLDP and BY. Three QTLs qYLDP1.1, qYLDP5.1, and qYLDP12.1 showed PV ranging from 36 to 56% and trait enhancing alleles from Swarna. Of these three QTLs, one major effect QTL qYLDP12.1 was identified at chromosomal region RM519-RM235 with highest LOD of 16.7 and 56% PV. Three major QTLs qBY2.1, qBY6.1, and qBY8.1 were identified with PV ranging from 27 to 33%. One QTL each for DFF (qDFF2.1) and NT (qNT1.1) were identified with PV 12 and 22.5%. Two common QTLs qDM2.1 and qPH1.1 were identified in both years.

Chromosomal segment substitution line Finder output showed that 74 CSSLs had homozygous chromosome segments from O. nivara substituting Swarna segments (Figure 2). These 74 CSSLs together showed 89% coverage of the wild genome based on the 111 SSRs marker data. Small regions on chromosomes 4, 6, 7, 9, 11, and 12 were not represented in the 74 CSSLs. The average number of substituted segments per chromosome in these was 6.5. The number of substituted segments in each CSSL ranged from 2 to 18, with an average of 9.5% (Supplementary Figure 3). Chromosome 1 was represented by 8 CSSLs, chromosome 2 by 13 CSSLs, chromosome 3 by 9 CSSLs and chromosome 10 by 4 CSSLs.

Chromosomal segment substitution line 142S had only three substituted segments and they were on chromosome 10 and showed significantly higher BM than Swarna. Two lines 220S and 166-23-1S showed significantly higher PW than Swarna in both the years.

Yield enhancing QTLs have been reported previously from O. nivara in BC₂F₂ families (Kaladhar et al., 2008; Swamy et al., 2011, 2014). A total of 28 QTLs for yield traits were reported in BC₂F₂ population derived from the same cross of Swarna/O. nivara IRGC81848 and 78% of the loci from O. nivara were trait enhancing (Swamy et al., 2014). Three common QTLs for traits PH (qPH1.1- RM431) on chromosome 1, DM (qDM2.1-RM3874) on chromosome 2, and YLDP (qYLDP12.1- RM519) on chromosome 12 were identified in both BC₂F₂ (previous study) and BC₂F₈ (present study) and these three are thus significant major effect QTLs. The first common QTL qPH1.1 is close to the well known semi- dwarf locus sd-1, the green revolution gene and explained PV of 22 and 17.5% in 2014 and 2015, respectively. Thus, it shows the robustness of our study as these major QTLs were identified in BC₂F₂ and BC₂F₈ and in both years. The major effect QTL qPH1.1 was identified in both years and in both generations (BC₂F₈ and BC₂F₂) indicating that it is also a stable QTL. RM431 the peak marker is within the region of QTL qDTY1.1 reported previously for yield under drought and being transferred into submergence tolerant versions of three high yielding mega rice varieties Swarna-Sub1, Samba Mahsuri-Sub1, and IR 64-Sub1 using MAS (Vikram et al., 2011; Singh R. et al., 2016). In our study, O. nivara allele of qPH1.1 increases PH. The semi dwarfing gene sd-1, has been widely utilized in rice production, however, attention has to be paid to breed taller plants with strong culm to enhance plant yield by increasing biomass (Han et al., 2017). They identified QTLs for PH in a RIL population derived from Zhenshan 97 and Xizang 2. Two QTLs qph1 and qph7.1 detected in 3 years, explained 13% PV with large additive effect of 12 cm from Xizang 2 allele. In our study, O. nivara alleles increased PH by 11 cm in both the years. RM431 flanks PH QTLs reported previously from elite/ wild crosses using accessions of O. nivara, O. rufipogon and from landraces also (Wickneswari et al., 2012; Eizenga et al., 2013; Mohammadi et al., 2013). In addition to height, RM431 was linked with QTLs for grain number/panicle, harvest index (Li et al., 2012). This indicates that alleles at the vicinity of RM431 are associated with different yield related traits in different genetic backgrounds.

The second common QTL qDM2.1 at chromosomal region RM3874 falls within a metaQTL MQTL2.3 reported for PW (Swamy and Sarla, 2011; Swamy et al., 2011). The third common QTL qYLDP12.1 at RM519 region also harbors QTLs qrd12.1 for rachis diameter, qnpt12.1 for number of productive tillers, qnsp12.1 for number of spikelets and qnfg12.1 for number of filled grains (Swamy, 2009; Swamy et al., 2014). There are two previous reports on identification of common QTLs in BC₂F₂ and later generations. Rangel et al. (2008) identified one common QTL for grain yield on chromosome 1 at RM1 locus from BG90-2/RS-16 (O. glumaepatula) in both BC₂F₂ and BC₂F₈. Similarly, Wickneswari and Bhuiyan (2014) identified two common QTLs qSPL-1-1 and qSPL-8 for spikelets per plant in BC₂F₂ and BC₂F₅ from MR219/Oryza rufipogon Griff. IRGC105491.

Comparing QTLs in BC₂F₈ with BC₂F₂ we found there were few other chromosomal regions common in BC₂F₈ and BC₂F₂ but linked with different traits. In BC₂F_8, RM434 was linked with qPW9.1 where as in BC₂F₂ RM434 was linked with qyldp9.1. Likewise, RM128 was linked with qPH1.2 in BC₂F₈ and qnpt1.1 in BC₂F₂; RM223 for qBY8.1 in BC₂F₈ and qnfg8.1and qyldp8.1 in BC₂F₂; RM519 for qYLDP12.1 in BC₂F₈ and qnpt12.1, qnsp12.1 and qnfg12.1 in BC₂F₂. There could be several reasons for this observation. Earlier, 100 SSRs were used to genotype 227 BC₂F₂ mapping population whereas in this study 111 different SSRs were used to genotype 94 BC₂F₈ BILs and only 37 markers were common with the previous study. The detection of novel QTLs in BC₂F₈ might be due to changes in genetic background from BC₂F₂ to BC₂F₈ or differences in the size of mapping population and the distribution of SSR markers used in the two studies. Also environment conditions were different in the two generations and this might have contributed to differences in phenotype and QTLs.

Considering that the available population size was small, the phenotypic variation explained by QTLs is quite likely to be overestimated. However, since many of the major effect QTLs detected in our study were also identified for the same or different trait previously, it indicates the robustness and importance of these QTLs in regulating the phenotype of yield associated traits. The significant QTLs identified in this study can be further evaluated for use in marker assisted transfer to adapted varieties to improve not only yield but several related traits as QTLs for drought and salinity are being transferred into high yielding mega varieties (Singh R. et al., 2016).

Chromosomal segment substitution line are an important genetic resource in rice to discover novel genes by focusing on small chromosomal regions (Ali et al., 2010; Subudhi et al., 2015; Ramos et al., 2016). A set of 74 CSSLs with substituted chromosomal segments of O. nivara in the genetic background of Swarna were identified. Graphical genotypes of CSSL library revealed 89% coverage of O. nivara genome. A few chromosomal regions were not represented by substituted segments on chromosomes 4, 7, 9, and 11. This could be due to less number of BILs used or the presence of lethal alleles, hybrid sterility and gametophyte lethal genes or even low recombination in these chromosomal locations. Introgression of O. nivara chromosomal segments in the genetic background of 93-11 was reported by Ma et al. (2016). The coverage of the O. nivara genome by the ILs was 94.96%. Several sets of CSSLs have been developed using other wild species such as O. rufipogon (Furuta et al., 2014; Qiao et al., 2016; Ogawa et al., 2016), O. minuta (Guo et al., 2013), O. meridionalis (Arbelaez et al., 2015), O. longistaminata (Ramos et al., 2016). A set of 198 CSSLs was developed from a cross between 93-11 and O. rufipogon, and introgressed segments covered 84.9% of the wild rice genome (Qiao et al., 2016). Subudhi et al. (2015) developed 74 CSSLs covering 99% of the weedy rice donor PSRR-1 genome. In their study, donor segments per line ranged from 1 to 3 with 64% of CSSLs with single homozygous donor segments. In our study number of donor segments per line ranged from 3 to 18 as CSSLs were identified from a set of available BILs at BC₂F₈ generation without MAS. A few backcrosses can be made along with MAS of target regions to recover more background genome of this set or selected CSSLs.

In our previous studies, two stable lines 166S and 14S were reported to be efficient in compartmentalization of Na+ in leaf tissue and grain yield of 166S was least affected by salt stress (Divya et al., 2016; Pushpalatha et al., 2016). Also, three lines 24S, 70S, 14-3S were identified as heat tolerant lines for spikelet fertility and YLDP in both wet and dry seasons (Prasanth et al., 2016). In present study, these five lines 14S, 24S, 70S, 166S and 14-3S were part of the set of CSSLs and they were found to have 11.5–18.6% chromosomal segments from O. nivara. Thus, several elite BILs such as these can be used in prebreeding programs for identification of candidate genes for different yield related traits. Among the CSSLs, 220S showed significantly higher PW over the parent Swarna in both years. 10-2S showed highest YLDP in both years over Swarna. These two lines, 220S and 10-2S showed significantly higher yield than Swarna, so they can be used to dissect target QTL regions. In addition, another 20 CSSLs showed different yield traits as significantly higher than Swarna and these marker defined CSSLs can be further utilized for fine mapping.