The CSSL Finder program selected a subset of 125 SSR markers properly distributed across the twelve rice chromosomes. On this basis, searching for CSSL candidates led to a set of sixty-four lines (Figure 1). Average size of the substituted chromosomal segments in the CSSLs was of 10 cM and covered the whole O. glaberrima genome, except for small regions landmarked by markers RM71-RM300 (43.8-65.9 cM) on chromosome 2 and RM185-RM241 (93.8-135.0 cM) on chromosome 4. The proportions of Caiapó and MG12 in the CSSL lines were 87.59% and 7.64%, respectively. The remaining 4.78% corresponded to heterozygotes and missing data. The number of introgressed segments varied between 2 to 8 per line. We observed several lines with a few heterozygous chromosomal regions, for which pollen contamination that occurred in the field between lines of the population is the most probable explanation. Additional backcrossing (2-3) with marker-assisted monitoring is currently carried out to purify the genetic background of the 64 lines.

Correlation coefficients among yield and yield component traits were tested for significance at P < 0.05 and P < 0.01, and are presented in Table 1. Coefficients of phenotypic correlation were low, indicating the complexity of relationships between these traits. Positively correlated traits (P < 0.01) were plant height with yield (R² = 0.376) and panicle length (R² = 0.548), and sterility percentage with tiller number per plant (R² = 0.295). The observed correlation between plant height and yield corroborates previous yield-associated QTL studies in rice 3233. Panicle length is largely proportional to plant height, explaining the relatively high R² value. Negatively correlated traits (P < 0.01) were plant height with 1000-grain weight (R² = - 0.172), and, as expected, sterility percentage with yield (R² = - 0.244).

Fourteen QTLs were found for plant height (PTHT), yield (YLD), tiller number per plant (TINB), 1000-grain weight (TGRWT) and sterility percentage (ST) located on chromosomes 1, 3, 4, 6 and 9. A major QTL for RSNV was detected on chromosome 11 (Table 2; Figures 2, 3). All QTLs were detected by both the single-marker ANOVA1 and interval mapping-based methods (IM and CIM), indicating their robustness for QTL detection for this type of populations.

Using both CSSL Finder and WinQTLCart software, one highly significant QTL with an F = 64.40 could be located on chromosome 11. The QTL region was saturated with downstream and upstream SSR markers delimiting this QTL (Figures 2 and 3). Analysing the recombination events in the region allowed us to semi-fine map the RSNV major QTL, between SSR markers RM202-RM26406 (44.5-44.8 cM).

The phenomenon of segregation distortion (SD), defined as a deviation from the expected Mendelian segregation ratios in a segregating population, has been reported in several crops. In rice, this effect is often due to sterility genes located on several chromosomal regions. Genetic interactions, genes with variable effects in regeneration by anther culture and physiological and/or environmental factors can also lead to SD 40. 37% (74) of the markers showed distortion in favour of MG12 alleles on chromosomes 1, 2, 3 and 6. As expected, the strongest segregation distortion was found at the short arm of chromosome 6, at markers RM6273 and RM204 (0.0-15.8 cM) 414243. This region corresponds to the genomic location of the S₁ locus, a sporo-gametophytic sterility factor identified in previous studies. The other distorted regions matched with the chromosomal locations of O. sativa × O. glaberrima sterility loci described so far: S_33(t) on chromosome 1 44, S_29(t) on chromosome 2 45, S₁₉ and S_34(t) on chromosome 3 4647.

Yield is a complex trait controlled by many genes of major or minor effect 32. QTLs for yield found in the present study were associated with small effects that are co-localized with QTLs of the group of M-QTLs (main-effect QTLs) identified in other studies. M-QTLs represent more than 90% of the QTLs reported to date 48. Also, transgressive segregation was observed for all traits except tillering (Figure 4), demonstrating that interspecific crossing enhanced the possibility of introgressing genetic variability in cultivated rice 4950. Although several QTLs were detected on the short arm of chromosome 6, they should be carefully considered, because their effects could have been overestimated due to the strong segregation distortion affecting this region.

To our knowledge, this is the first identification of a genetic factor underlying resistance to the RSNV disease. In order to better elucidate the bases of genetic control of RSNV resistance, fine mapping of this region is being envisaged using recombinant event analysis in the BC₄F₂/F₃ lines that we produced in 2007.

Breeding strategies such as marker-assisted selection (MAS) or marker-assisted backcrossing (MAB) require comprehensive dissection and understanding of the complex traits measured. Development of genetics resources such as CSSL lines will greatly facilitate the detection of naturally occurring allelic variation in rice and will help to acquire a better knowledge of target traits 9121351. Phenotyping strategies based on CSSL populations present the advantage of a relatively small number of lines to evaluate, with the possibility of replicating evaluations over space and time. This should lead to better quality data in the case of complex, time-consuming or expensive phenotypic evaluations. Genetic dissection of complex traits by associating genetic variation with introgressed fragments allows us to reduce interference effects between QTLs. This helps to understand the genetic bases of reproductive barriers between species, and provides a powerful approach for QTL identification, fine mapping of QTLs, laying the bases for both marker-assisted selection and map-based cloning strategies based on exploitation of wild alleles. Comparison of phenotypic values between any line of the population and the recurrent parent generates high statistical power. CSSL lines can be crossed in different ways in order to study epistatic interactions between QTLs, develop Near-Isogenic Lines (NIL) and do QTL pyramiding 16263152.

Wild and cultivated African rice species have been shown to be valuable sources of alleles associated with traits of agronomic importance 1243. However, they carry many undesirable alleles that may show strong linkage to favorable alleles, linkages that usually are very difficult to break up by conventional crossing. CSSL lines give access to the original exotic allelic source, providing an elegant way of circumventing this issue, thus representing a useful and powerful tool for genetics and breeding approaches. They constitute a very useful genetic resource for studying both inheritance of agronomically important traits and directing their incorporation as progenitors in breeding programs for the development of elite germplasm with exotic characteristics of interest. The set of CSSL lines presented in this study is available to the rice community through both the CIAT Rice Outcome Product Line and the Generation Challenge Programme. Several research teams around the world are already using this population in their effort to locate, map and utilize new alleles associated with traits of economic importance.

The genetic diversity of crop plants has been narrowed down due to the domestication process and decades of selection. Exotic genetic resources such as wild rice species can be successfully exploited to increase allelic variability into elite lines 5354. Within the framework of a Generation Challenge Programme project, we are now developing a series of new CSSL populations, using wild AA-genome rice species (O. rufipogon, O. glumaepatula, O. meridionalis and O. barthii) as donors. Associated partners to this effort are EMBRAPA-CNPAF (Brazil), WARDA (Benin) and Cornell University (USA). These wild species as well as African cultivated rice show adaptation to biotic and abiotic constraints associated with specific geographic regions. Transgressive segregation has been demonstrated in several studies 4955. The development of libraries of introgression lines makes immediate use possible for plant breeders and will simultaneously serve to enhance our understanding of the wild/cultivated allelic genetic interactions. We hope that the results of this work will contribute to a better understanding of plant performance key components and to the development of new improved rice cultivars.

The recurrent parent Caiapó (O. sativa ssp. tropical japonica) is a commercial rice variety developed by EMBRAPA-CNPAF (Goiania, Brazil) and has been cultivated since 1992 in Brazil and other places in Latin America and the Caribbean. This variety is characterized by presenting yields of 2.5 tons/ha under upland conditions, long grain type, medium growth cycle, tolerance to leaf blast (Magnaporthe grisea), moderate resistance to neck blast and tolerance to aluminium toxicity, acid soil conditions and drought 56. The donor parent MG12 (acc. IRGC103544) is an accession of the African cultivated rice species, O. glaberrima. This species is grown in West Africa and shows several negative characteristics with respect to the Asian O. sativa, like shattering, brittle grain and poor milling quality. More importantly, it consistently shows lower yields than O. sativa. However, African rice often shows more tolerance to fluctuations in water depth, iron toxicity, infertile soils, severe climatic conditions and human neglect, and exhibits better resistance to various pests and diseases like nematodes (Heterodera sacchari and Meloidogyne sp.), African gall midge, RSNV and Rice yellow mottle virus (RYMV) 5758596061.

The population was developed at the International Center for Tropical Agriculture (CIAT) headquarters, in Cali, Colombia, starting in 1997. The scheme applied for population development is shown in Figure 5. Accession MG12 was used as the male parent of the F₁ hybrid. F₁ plants were completely androsterile and 20 individuals were randomly selected as females for backcrossing with the recurrent parent Caiapó. A total of 154 BC₁F₁ plants were produced and then successively backcrossed to Caiapó until the BC₃F₁ generation. Anthers were collected from the BC₃F₁ plants and processed through in vitro culture to generate double haploids (DH) as described by 62. As a result, 695 BC₃F₁DH lines were obtained and multiplied for seed under irrigated field conditions in 2000. Subsequently, a subset of 312 BC₃F₁DH lines offering a good representation of the observed phenotypic variability was selected as a mapping population for agronomic evaluation and molecular characterization [Additional file 1: Figure S1].

The mapping population and the parent accessions (as controls) were first evaluated in replicated field plots in Colombia at CIAT headquarters in 2001. Materials were planted under irrigated conditions in a randomized complete block design arranged in two rows, where each row was 5 m long with a spacing of 30 × 30 cm (20 plants/row), with three replications. Transplanting was done at twenty-five days after sowing. Five plants per BC₃F₁DH line were randomly selected and then evaluated for six agronomic traits: plant height (PTHT), tiller number (TINB), panicle length (PNLG), percentage of sterility (ST), 1000-grain weight (TGRWT) and grain yield (YLD). A second field experiment with the BC₃F₁DH lines and the two parents was planted in a randomized complete block design with two replications at the Rice Research Station, Crowley, Louisiana 34 in 2002.

Rice stripe necrosis virus is a furovirus associated with the disease known as crinkling, hence its common name, "crinkle virus". It was first reported in West Africa in the late 1970s 63. Later on, in 1991, the virus was found in South America, in the Colombian Department of Meta and was locally called "entorchamiento" 64. Symptoms include seedling death, foliar striping and severe plant malformation. This disease can provoke yield losses of up to 40% in highly infected fields. Since O. glaberrima was shown to be highly resistant to RSNV 60, we took advantage of the usefulness and potential of the CSSL lines to search for QTLs for RSNV resistance. In order to screen the lines for their resistance to RSNV, infested soil from farmer's field was used as inoculum. The level of soil infestation was tested by planting the highly susceptible rice cultivar Oryzica 3 in several pots containing the infested soil. The infested soil was used if the incidence of RSNV infected plants on the susceptible check was above 80%. The virus incidence on the mapping population was evaluated in 178 lines by counting the number of plants showing the characteristic symptoms of the disease, including: 1) crinkling or deformation, 2) yellow stripes on leaves or foliar striping, 3) stunting of plants (Figure 6) and 4) dead plants. Number of healthy plants was also recorded. The highly susceptible cultivar Oryzica 3 was used in each experiment as a control and indication of the disease pressure. Ten plants per line were evaluated. Lines with a percentage of healthy plants superior to 85% were considered as resistant, while the other ones were considered as susceptible. These evaluations were carried out in the greenhouse of the CIAT's Rice Pathology Laboratory, where the average of both relative humidity was 80 percent and the temperature 25°C. A randomized complete block design with four replications with ten plants per pot was used. The experiment was replicated two times over a period of six months with a total of 80 plants evaluated for each genotype. Two evaluations were made, the first one 30 days after planting and the second one 60 days after planting. Final line reaction was based on the second evaluation. In each experiment the plants were fertilized with a commercial dose of Nitrogen equivalent to 200 KgN/ha in order to favour the development and high incidence of the disease.

Total DNA was extracted from frozen leaf tissue based on a slightly modified version of the Dellaporta protocol (our unpublished data). Subsequently, quality and quantity of DNA was evaluated on 0.8% agarose gel stained with ethidium bromide. A total of 200 polymorphic simple sequence repeats (SSR) loci distributed across the twelve rice chromosomes with an average spacing of 8.0 cM was used. Most of these SSR markers were selected from the Universal Core Genetic Map (UCGM) of rice developed at CIAT Rice Genetics and Genomics group 65. The UCGM was developed from the list of 18,000 SSRs published in IRGSP (2005). Polymerase chain reactions (PCR) were performed in a total volume of 15 μL containing 20 ng/μL of DNA template, 1X PCR buffer, 2.5 mM of MgCl₂ (or 1.5 to 2.0 mM for some specific pairs of primers), 0.2 mM of d-NTP, 0.13 μM of each primer and 1 U/μL Taq DNA polymerase. Amplification was run on MJ Research PTC-225 (384 well) thermocycler with the following program: 94°C for 3 min; 29 cycles at 94°C for 30 s, 55°C for 45 s (modified for some specific pairs of primer), 72°C for 1 min; 72°C for 5 min. PCR products were separated on 4% high-throughput agarose gel for markers that showed a polymorphism size higher than 10 bp, and stained with ethidium bromide. For polymorphism lower than 10 bp, PCR products were separated using 6% denaturing polyacrylamide gel followed by silver staining, as described in the Promega Technical Manual 66.

Selection of a subset of introgression lines that cover the entire donor genome was carried out with the help of the CSSL Finder v. 0.84 computer program 67. CSSL Finder was designed to search for a subset of CSSL that optimizes specific parameters: target size of introgression segments, percentage of donor genome and number of introgressed fragments. It also makes it possible to define the minimum set of lines that cover the entire donor genome, according to the same parameters. Subsequently, graphical genotypes of the candidate lines can be displayed. CSSL Finder is available at no cost at http://mapdisto.free.fr.

As the coordinates of SSR markers of the UCGM are physical positions on the rice pseudomolecules, it was necessary to convert them to centimorgans (cM) in order to obtain QTL confidence intervals comparable to those obtained in other studies. For this purpose, we used a genetic linkage map obtained from a BC₁F₁ population derived from the cross IR64 (O. sativa ssp. indica) × TOG5681 (O. glaberrima) (our unpublished data). The map was constructed using the computer program MapDisto v. 1.7 68 http://mapdisto.free.fr. For each marker, a chi-squared test (P < 0.01) was performed to identify markers with segregation distortion. Correlation between the traits evaluated was calculated using the QGene v. 3.07 program 69, and tested using significance levels of 0.05 and 0.01. As several introgression events are present at each marker position in the complete set of 312 lines, we used standard methods to identify QTLs linked to the segregating traits. A QTL analysis for the evaluated traits was done using both the CSSL Finder v. 0.84 and the MapDisto v. 1.7 programs, which basically perform a single-marker ANOVA1 F-test. We considered the F-test as significant when its value was higher than 15. CSSL Finder was used to display graphical genotyping of subsets of fifteen lines that presented the most extreme phenotypic value for each trait, in order to confirm each detected QTL. Interval mapping (IM) and composite interval mapping (CIM) analyses using WinQTLCart v. 2.5 70 were also performed. Significant QTLs found using F-test, IM and CIM methods were compared with previous studies.