Development of Chromosome Segment Substitution Lines (CSSLs) Derived from Guangxi Wild Rice (Oryza rufipogon Griff.) under Rice (Oryza sativa L.) Background and the Identification of QTLs for Plant Architecture, Agronomic Traits and Cold Tolerance

Common wild rice contains valuable resources of novel alleles for rice improvement. It is well known that genetic populations provide the basis for a wide range of genetic and genomic studies. In particular, chromosome segment substitution lines (CSSLs) ais a powerful tool for fine mapping of quantitative traits, new gene discovery and marker-assisted breeding. In this study, 132 CSSLs were developed from a cultivated rice (Oryza sativa) cultivar (93-11) and common wild rice (Oryza rufipogon Griff. DP30) by selfing-crossing, backcrossing and marker-assisted selection (MAS). Based on the high-throughput sequencing of the 93-11 and DP30, 285 pairs of Insertion-deletions (InDel) markers were selected with an average distance of 1.23 Mb. The length of this DP30-CSSLs library was 536.4 cM. The coverage rate of substitution lines cumulatively overlapping the whole genome of DP30 was about 91.55%. DP30-CSSLs were used to analyze the variation for 17 traits leading to the detection of 36 quantitative trait loci (QTLs) with significant phenotypic effects. A cold-tolerant line (RZ) was selected to construct a secondary mapping F2 population, which revealed that qCT2.1 is in the 1.7 Mb region of chromosome 2. These CSSLs may, therefore, provide powerful tools for genome wide large-scale gene discovery in wild rice. This research will also facilitate fine mapping and cloning of QTLs and genome-wide study of wild rice. Moreover, these CSSLs will provide a foundation for rice variety improvement.


Introduction
Wild rice (Oryza rufipogon Griff.) contains many novel and useful alleles that control tiller number, shattering, dormancy, pericarp color, mating type, panicle architecture and grain size and number [1]. Therefore, the potentially beneficial genes in wild rice are an important goal to improve cultivated rice (Oryza sativa L.) [2]. Although many quantitative trait loci (QTLs), for plant architecture, agronomic traits and cold tolerance (CT) have been identified in rice [3][4][5], however, there are few reports on those traits that were discovered in same chromosome segment substitution lines (CSSLs). The hybridization between Oryza sativa and wild rice and use of marker-assisted selection (MAS) of leaf margin color (LMC), tiller angle (TA), heading date (HD), plant height (PH), grain shattering (SH), apiculus color (AC), stigma color (SC), glume color (GC), number of grains per panicle (NGPP), 1000-grain weight (GWT), grain length (GL), grain width (GW), grain length to width ratio (GLWR), awn length (AL), seed coat color (SCC) and cold tolerance (CT). The data for TA was recorded on a scale from 1 to 9 representing the angles of 0-10 • , 11-20 • , 21-30 • , 31-40 • , 41-50 • , 51-60 • , 61-70 • , 71-80 • and 81-90 • , respectively [17][18][19][20]. The HD was recorded when the first panicle to emerge reached about 2-cm-long and the number of days from sowing to heading were scored for each plant. PH was measured for each plant at the mature stage from the base of the stem to the tip of the higher panicle. NGPP, GL and GW were recorded according to the previously established methods [19,20]. The GLWR and 100-GWT of the filled grains were investigated after the rice was harvested at the mature and naturally dried stage [10].

Evaluation of Cold Tolerance at Seedling Stage
To preliminarily evaluate the cold tolerance variation of CSSLs, the experiment was conducted in controlled conditions at seedling stage. Thirty seedlings of each CSSL were planted in soil. Plants were grown in a controlled environment at a day temperature of 25 • C and night temperature of 19 • C till three leaf stage. At three leaf stage, seedlings were exposed to cold stress according to a previously established method [21]. The depth of the water was about 5 cm measured from the surface of the soil in the trey. The cold stress treatment lasted 5 days and the conditions alternated between 10 • C for 10 h during the day and 8 • C for 14 h at night. After the cold treatment, the seedlings were subjected to natural standard growth conditions at 26 • C and the survival rate was investigated. After 5 days of treatment, cold tolerance was evaluated on the basis survival rate and injury level. The experiment was repeated three times under the same cold stress treatment. The average data of three replications were used. The data for CT were recorded on a scale from 0 to 9, representing the survival rates of 0-9%, 10-19%, 20-29%, 30-39%, 40-49%, 50-59%, 60-69%, 70-79%, 80-89% and 90-100%, respectively [21]. The secondary mapping population used to fine-map the major QTL for cold tolerance was developed by backcrossing a CSSL(RZ34) with the recipient parent (93-11). Three hundred and eleven F 2 plants of that cross were genotyped using five InDel markers and fifty-seven cold tolerant and twenty-one cold sensitive plants selected for phenotypic evaluation under cold stress treatment.

Construction of CSSLs and Genome Sequencing and Development of InDel Markers
CSSLs of common wild rice were constructed by hybridization, backcrossing and marker-assisted selection (MAS) according to the previously described method [22,23]. The genomic DNA of DP30 and 93-11 was prepared and whole-genome re-sequencing (WGRS) was performed on an Illumina HiSeq2500™ by Novo Generation Company, Beijing, China. The standard Illumina protocol was followed for sample preparation and sequencing. The quality trimming (phred quality score, <Q 20 ) was carried out by using FastQC [24] and the Cutadapt software was used for adapter trimming with the parameters of −O 5 and −m 32 [25]. The Burrows-Wheeler Aligner (BWA) software was used to map clean reads to the 93-11-reference genome [26,27]. InDel polymorphisms were detected by the GATK tool software with the defined length of insertions and deletions between 1 bp and 10 bp [28]. The larger sized (≥2 bp) InDel regions and high sequencing depths (DP, ≥5-fold) were extracted to design the InDel markers. The primer pairs were designed based on parental sequence differences by the DNAMAN v6.0 software and screened using the NCBI (https://www.ncbi.nlm.nih.gov/) database.

DNA Isolation and PCR Amplification
Genomic DNA was extracted from fresh leaf tissues using the CTAB method described by the previously established protocol [29]. The PCR amplification, separation of PCR products and confirmation and genotyping of the electrophoretic bands of the PCR products were performed according to previously established methods [30][31][32].

QTL Mapping and Data Analysis
The lengths of the substituted segments in CSSLs were assayed according to the previously established method [10]. A chromosomal segment flanked by two donor markers was considered to be 100% donor type, and the length of the segment was considered to be the minimum length of a substituted segment (L-min). A chromosome segment flanked by two recipient markers was considered as 0% donor type, and the length of the segment was considered to be the maximum length of a substituted segment (L-max). A chromosome segment flanked by one marker of donor type and one marker of recipient type was recognized as 50% donor type. The length of each substituted segment in the CSSLs was calculated as the estimated length (L), which is the average of L-min and L-max.
Genotypic graphics and chromosome genetic maps of the CSSLs were generated using the Graphic Geno-Types 2 software (GGT2.0) [10]. A putative QTL was declared at the significance level of p ≤ 0.001 in a CSSL. If several CSSLs with overlapping substituted segments shared the same QTL, a substitution mapping approach was employed to localize the QTL to a smaller genomic interval [33]. QTL nomenclature was performed according to McCouch and CGSNL: each QTL name is italicized and starts with a lowercase letter "q" to indicate that it is a QTL, followed by a two to five letter standardized "trait name", a number designating the rice chromosome on which it occurs (1-12), a period (".") and a unique identifier to differentiate individual QTLs for the same trait that resides on the same chromosome [34]. The linkage map of QTLs was constructed using MapChart 2.2 [35]. The statistical analysis of the phenotypic and genotypic data of the DP30-CSSL populations was performed by QTL IciMapping 4.1.0 software [36]. Base on the permutation test to set LOD value ≥2.5 as the thresholds for QTL analysis. The chromosomal genetic map was constructed using MapChart 2.3 software.

Phenotypic Variation of Plant Architecture, Agronomic Traits and Cold Tolerance in DP30-CSSLs
We evaluated the variation in plant architecture, agronomic and cold tolerance traits (DP30, DP30-CSSLs and 93-11) and calculated the phenotypic values of these traits during two seasons (fall and spring). The DP30-CSSLs showed significant variation in all of these traits. Furthermore, this variation in HD, GWT, GL, GW, GLWR and AL were normally distributed, while TA, PH, NGPP, SH and CT followed biased distribution ( Figure S1). First, we speculated that the biased distribution was due to continuous backcrossing, that the CSSLs without related trait substitution segments were similar to 93-11. Second, we counted 132 lines, so an insufficient population may have led to biased distribution. Third, these distributions were frequently skewed because of interactions between the alleles, nonallelic genes or other environmental factors.

Genome Re-Sequencing and Selection of InDel Markers
The DP30 genome was sequenced by Illumina high-throughput sequencing technology, and the resulting sequence was assembled using IRGSP-1.0, with the 93-11 genome as a reference ( Figure 1A). The DP30 genome is displayed according to the input reads (Table S1). The sequencing results reveal that 1,894,103 bp across all 12 chromosomes were different in DP30 and 93-11, representing 6.1% of the genome ( Figure 1B). InDel molecular markers were designed such that the base mutation exceeded 20 bp. Based on the lengths of the PCR products (150-200 bp) and the results of gel electrophoresis, we selected 285 InDel markers with an average distance of 1.2 Mb ( Table 1). The sequences of the 285 InDel markers (Table S2).

Development of the DP30-CSSLs
The procedure we developed the DP30-CSSLs is shown in Figure 2. The F 1 plants were obtained by crossing DP30 and 93-11. Molecular MAS began using the BC 4 F 1 , resulting in 176 BC 5 F 3 lines, 32 BC 6 F 3 lines and 22 BC 7 F 3 lines containing the target segments. Finally, we selected 132 CSSLs from these 230 lines (candidate plants) for genetic and phenotypic analysis ( Figure 2, Table S3).

Development of the DP30-CSSLs
The procedure we developed the DP30-CSSLs is shown in Figure 2. The F1 plants were obtained by crossing DP30 and 93-11. Molecular MAS began using the BC4F1, resulting in 176 BC5 F3 lines, 32 BC6 F3 lines and 22 BC7 F3 lines containing the target segments. Finally, we selected 132 CSSLs from these 230 lines (candidate plants) for genetic and phenotypic analysis ( Figure 2, Table S3).

Substitution Segments of DP30-CSSLs
In total, we identified 80 BC4 F2 plants from 171 CSSLs and named as DP30-CSSLs. Then, 24 BC4 F2 plants were selected and 44 BC5 F2 plants were obtained after continuous backcrossing. Finally, 18 BC6 F2 were obtained with a relatively complex genetic background to continue backcrossing. The 132 CSSLs formed a DP30-CSSLs population, and the target replacement segments of these CSSLs accumulated a total length of 536.4 Mb, covering 91.42% of the DP30 genome. The CSSLs were arranged according to the position of their target substitution segments. The 99 substitution lines

Substitution Segments of DP30-CSSLs
In total, we identified 80 BC 4 F 2 plants from 171 CSSLs and named as DP30-CSSLs. Then, 24 BC 4 F 2 plants were selected and 44 BC 5 F 2 plants were obtained after continuous backcrossing. Finally, 18 BC 6 F 2 were obtained with a relatively complex genetic background to continue backcrossing. The 132 CSSLs formed a DP30-CSSLs population, and the target replacement segments of these CSSLs accumulated a total length of 536.4 Mb, covering 91.42% of the DP30 genome. The CSSLs were arranged according to the position of their target substitution segments. The 99 substitution lines contained only one substitution segment from DP30. The total length of these CSSLs substitution segments was 359.66 Mb and the coverage rate was 61.36% ( Figure 3).   Chromosomes with more than 90% coverage of the DP30 genome by chromosome substitution segments include chromosomes 1, 2, 3, 5, 7, 11 and 12. Among them, chromosome 5 had the highest coverage rate (97%), while the lowest coverage rate belonged to chromosome 8 ( Table 2).
The average length of DP30-CSSLs population substitution target segment was 4.06 Mb, which ranges from 0.5-22 Mb. The length of the substitution segment of 25 CSSLs was less than 2 Mb; The length range of substitution segments of 55 substitution lines was 2-4 Mb and 30 CSSLs were 4-6 Mb and the length of the substitution segment of 6 CSSLs was larger than 8 Mb.

Wild rice QTLs in the DP30-CSSLs
We detected the genotypes of all CSSLs in DP30-CSSLs (Table S3). In total, 36 QTLs were identified in the CSSLs.
Three CSSLs (RZ142, RZ8 and RZ13) showed significant differences in HD when compared with 93-11 ( Figure S2C). These three lines had an overlapping segment near C11-4 on chromosome 11, so we identified a new QTL (qHD11.1) in this region that has negative additive effects on HD (Table 3; Figure S6).  List of wild rice QTLs identified in this study. Position-physical position of molecular markers on chromosomes (Mb); length-length of the overlapping part of the lines; a-additive effect; Gene-cloned genes in the overlapping segments. TA-tiller angle; HD-heading date; PH-plant height; NGPP-number of grains per panicle; GWT-1000 grain weight; GL-grain length; GW-grain width; GLWR-grain length to width ratio; AL-awn length; SH-grain shattering; and CT-cold tolerance.

Construction of a Secondary Population and Mapping of qCT2.1
Our results showed that there was no significant difference in the appearance of RZ34 and 93-11 seedlings before cold treatment ( Figure 6A). RZ34 had a lesser degree of wilting after cold treatment compared with 93-11 ( Figure 6B). Furthermore, 93-11 died three days after the cold treatment, while RZ34 survived and continued to grow ( Figure 6C). The results show that RZ34 seedlings were more tolerant to cold stress than 93-11. All the F1 individuals in the RZ34 line exhibited a cold tolerant phenotype. We selected some of the F1 plants and the related molecular markers showed that these plants were heterozygous for the tested genotypes. We obtained a total of 311 plants in the F2 population, among them 223 plants were cold tolerant, and 88 plants were susceptible to cold stress. This represented a 3:1 segregation ratio for cold tolerance which consistent with Mendelian's rules (χ 2 = 1.630 < χ 2 0.05,1 = 3.84), demonstrating that this cold tolerance trait was controlled by a single dominant QTL.

Construction of a Secondary Population and Mapping of qCT2.1
Our results showed that there was no significant difference in the appearance of RZ34 and 93-11 seedlings before cold treatment ( Figure 6A). RZ34 had a lesser degree of wilting after cold treatment compared with 93-11 ( Figure 6B). Furthermore, 93-11 died three days after the cold treatment, while RZ34 survived and continued to grow ( Figure 6C). The results show that RZ34 seedlings were more tolerant to cold stress than 93-11. All the F 1 individuals in the RZ34 line exhibited a cold tolerant phenotype. We selected some of the F 1 plants and the related molecular markers showed that these plants were heterozygous for the tested genotypes. We obtained a total of 311 plants in the F 2 population, among them 223 plants were cold tolerant, and 88 plants were susceptible to cold stress. This represented a 3:1 segregation ratio for cold tolerance which consistent with Mendelian's rules (χ 2 = 1.630 < χ 2 0.05,1 = 3.84), demonstrating that this cold tolerance trait was controlled by a single dominant QTL.

Construction of a Secondary Population and Mapping of qCT2.1
Our results showed that there was no significant difference in the appearance of RZ34 and 93-11 seedlings before cold treatment ( Figure 6A). RZ34 had a lesser degree of wilting after cold treatment compared with 93-11 ( Figure 6B). Furthermore, 93-11 died three days after the cold treatment, while RZ34 survived and continued to grow ( Figure 6C). The results show that RZ34 seedlings were more tolerant to cold stress than 93-11. All the F1 individuals in the RZ34 line exhibited a cold tolerant phenotype. We selected some of the F1 plants and the related molecular markers showed that these plants were heterozygous for the tested genotypes. We obtained a total of 311 plants in the F2 population, among them 223 plants were cold tolerant, and 88 plants were susceptible to cold stress. This represented a 3:1 segregation ratio for cold tolerance which consistent with Mendelian's rules (χ 2 = 1.630 < χ 2 0.05,1 = 3.84), demonstrating that this cold tolerance trait was controlled by a single dominant QTL.  We identified the QTL qCT2.1 in the segment of RZ34 ( Figure 7A) between markers JM2-3 (21.7 Mb) and C2-22 (29.18 Mb). Based on the tracking and overlap of three CSSLs (RZ37, RZ38 and RZ39) segments, the range of qCT2.1 was reduced to 4.47 Mb ( Figure 7A). Five InDel markers (dxw-3, dxw-9, dxw-8, dxw-5 and dxw-4) were developed between C2-18 and C2-20 (Table S4). In order to reduce phenotypic identification error and improve the accuracy of QTL mapping, we selected individual plants with extreme phenotypic values to detect genotypes. We identified the phenotypes of 311 plants in the F2 population. Among them, the genotypes of 56 cold-tolerant plants (scale 9-7) and 22 cold-sensitive plants (scale 0) were detected for linkage analysis. Five recombinant plants (R1, R2, R3, R4 and R5) were selected from the constructed F2 population (Table S5). These plants confirmed that qCT2.1 is located in a 1.7 Mb region between molecular markers dxw-4 and dxw-9 ( Figure 7B) and that the LOD value of qCT2.1 was 8 ( Figure 7C).

Constructing CSSLs in Guangxi Wild Rice (O. rufipogon Griff.) DP30
Since QTL detection is based on the natural allelic differences between parental cultivars, it is important to select parental cultivars that show large phenotypic variation in the target traits [49]. There is rich polymorphism between Guangxi wild rice DP30 (O. rufipogon Griff.) and cultivated rice

Constructing CSSLs in Guangxi Wild Rice (O. rufipogon Griff.) DP30
Since QTL detection is based on the natural allelic differences between parental cultivars, it is important to select parental cultivars that show large phenotypic variation in the target traits [49]. There is rich polymorphism between Guangxi wild rice DP30 (O. rufipogon Griff.) and cultivated rice 93-11 (O. sativa L.) due to their distant genetic bases [10]. In this study, we performed MAS using Guangxi wild rice DP30 (O. rufipogon Griff) as a donor parent to develop 132 DP30-CSSLs. The coverage rate of the DP30 wild rice was 91.55%. The average length of each replacement segment was 0.5-22 Mb. Compared with the previously reported CSSLs in wild rice, the DP30-CSSLs had a higher substitution segment coverage rate and more polymorphic primers for MAS [50,51]. Clearly, the genomic constitution quality of the CSSLs population is important. The chromosome position and genetic effect of QTLs locating on dp30-CSSLs will be more accurately assessed.

Identification of Quantitative Trait Loci and Measurement of Various Traits in Fall and Spring
Previous reports have shown that the CSSLs of wild rice are effective in the mining and transferring of wild alleles into cultivated rice [10]. Furthermore, several studies have reported the development of CSSLs and the identification of several agronomic and plant architecture traits [17,[35][36][37][38][39][40][41]. Ma et al. (2019) detected eighteen QTLs were two known grain length-and width-related genes and four novel QTLs. In addition, two QTLs were verified, and two novel QTLs were identified, for panicle neck length, a domestication-related trait [49]. Tan et al. (2004) identified quantitative trait loci (QTLs) associated with plant height and the days to heading in the BC 3 F 2 population. Putative QTLs derived from O. rufipogon were detected for plant height on chromosome 1 and identified 6 QTLs for days to heading on chromosomes 1, 3, 7, 8 and 11 [50].
In the present work, we compared the phenotypes of DP30-CSSLs to the phenotypes of known genes and used related molecular markers to confirm whether any of these genes/allele genes were present in the segments. We identified five QTLs related to TA, qTA7.1 and PROG1 were in the same position on the chromosome; qTA9.1 may be the allelic form of TAC1 (Table 3). PROG1 (LOC_Os07g05900) controls the creeping growth habits of common wild rice [17]. TAC1 (LOC_Os09g35980) is a recently discovered gene that controls the TA of rice corresponding to a major QTL [20]. We found that these five QTLs had positive additive effects on TA (Table 3). A QTL named qHD11.1 had negative additive effects on HD. Three of the QTLs identified in this study (Table 3) had positive additive effects for phenotypic variations in PH in both fall and spring. qPH1.1 and Sd1 were in the same position on the chromosome ( Table 3). The Sd1 (LOC_Os01g66100) gene, which controls gibberellin biosynthesis, was among these QTLs [36]. Three QTLs showed negative additive effects on NGPP and qGWT1.1 may be the allele of OsAGPL2 (Table 3). OsAGPL2 (LOC_Os01g44220) is a member of the OsPYL gene family that regulates the filling rate of grains, leading to lower final grain weight and yields [37]. Except for the QTLs locus near the OsAGPL2 gene, we also identified eight other QTLs. All of them showed negative additive effects on 100GWT in both seasons ( Table 3). The SG1 (LOC_Os09g28520) gene has shorter grains than the wild type and a dwarf phenotype [39]. Similar to the SG1 gene, one new QTL we identified, which showed negative additive effects on GW (Table 3). We identified five AL-related QTL, qAL4.1 and qAL4.2 as alleles of An-1 and An-2, respectively (Table 3). An-1 (LOC_Os04g28280) regulates the formation of awn primordia, promoting awn elongation and increasing the length of grains [40]. An-2 (LOC_Os04g43840) also increases the length of awns and makes them spinier ( Figure S4) [41]. The QTL qSH4.2, which is related to grain shattering, the SH4 (LOC_Os04g57530) was in the same chromosome position [42] (Table 3). Cold tolerance in seedlings is one of the important traits for the stable production of rice [43][44][45]. Here, we identified seven QTLs related to cold tolerance including these loci in a similar region with these three previously cloned genes. qCT1.1, qCT5.1 and qCT6.1 may be the allelic form of OsRAN1, OsiSAP8 and OsPYL9, respectively (Table 3). OsRAN1 (LOC_Os010611100) participates in cell division and the cell cycle and promotes the formation of intact nuclear membranes, thus improving the cold tolerance of rice [43]. OsiSAP8 (LOC_Os06g41010) is a zinc finger protein gene that enhances salt, drought and cold stress tolerance in rice [44]. OsPYL9 (LOC_Os06g33690) is a member of the OsPYL gene family and is a possible abscisic acid (ABA) receptor [45]. In addition, there are six quality trait loci, which contain three alleles [46][47][48]. The alleles of eleven cloned genes showed the reliability of DP30-CSSLs. No cloned genes were found in other QTL, which may contain new genes.

Construction of Secondary Population and Mapping of qCT2.1
Map-based cloning and mapping of cold tolerance genes in rice have always been a classical method for cold tolerance research in rice.
Previous studies used different populations to obtain some cold tolerance genes of rice [51,52]. According to the published data, more than 250 QTL of low-temperature tolerance has been found on 12 chromosomes of rice. In DP30-CSSLs, qSCT-3-1 was identified in the RM15031-RM3400 region of the long arm of chromosome three near to the centromere, and the genetic distance between the linkage markers was found to be 1.8 cM [53]. In this study, F 2 populations were constructed by Guangxi common wild rice seedling cold-tolerant segment replacement line RZ34 and cold-sensitive recurrent parent 93-11. Through map-based cloning, it was found that the main cold tolerance QTL qCT2.1 of rice at the seedling stage was located on chromosome 2 and located in the range of 1.7 Mb between molecular marker dxw-4 and dxw-9. To date, there is no cloned cold tolerance gene at the seedling stage in this interval. qCT2.1 could enhance cold tolerance at the seedling stage, which has a strong dominant effect, so it is expected to be used in rice breeding.
Rice breeding entered to a new era with the utilization of MAS and whole-genome sequencing to link genotypes with phenotypes. The introduction of wild rice CSSLs promoted gene QTLs mapping and genomic research. This study also suggests using Guangxi common wild rice accessions will provide a broad platform for genomic research and may lead to the discovery of new QTLs that will benefit rice breeding.

Conclusions
This study aims to use wild rice to develop CSSLs for cultivated rice and use these CSSLs for comparative mapping of traits related to plant architecture and yield. We focused on germplasm innovation in rice through the identification and transfer of beneficial genes/QTLs from the wild species. We introduced the DP30-CSSL library platform to facilitate pre-designed breeding of cultivated rice to utilize favorable alleles dispersed in Guangxi wild rice resources. The QTLs presented here are expected to provide further clues to identifying underlying mechanisms involved in plant architecture and improved grain. Our ongoing experiments are aimed at confirming the genomic regions and narrowing down of number of genes reported within the QTLs in the present study through comprehensive studies involving high-resolution linkage mapping via high-throughput genotyping by sequencing of advanced generation progenies.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/9/980/s1, Figure S1: Frequency distributions of quantitative traits in DP30-CSSLs during fall and spring, Figure S2: Phenotypic variation in plant-architecture-related traits and heading date in CSSLs and 93-11, Figure S3: Cold tolerance phenotype of some CSSLs and 93-11 at seedling stages (bar = 10 cm), Figure S4: Phenotypic variation of leaf traits in CSSLs and 93-11, Figure S5: Phenotypic variation of in apiculi, glumes and seed coats in CSSLs and 93-11, Figure S6: The distribution of the 36 QTLs in DP30-CSSLs. The molecular markers are shown on the left and the QTL sites are shown on the right. A QTL overlap indicates that there are two QTL in the same region. Table S1: Whole-genome re-sequencing analysis of wild rice DP30. Table S2: The sequence of DP30-CSSLs molecular markers, Table S3: Substitution segments of DP30-CSSLs population, Table S4: The markers sequences of InDel molecular markers for mapping QTL qCT2.1, Table S5: Genotypes and phenotypes of secondary F 2 populations.