Phenotypic analysis
TGW of the two parents G1025 and K1561 showed highly significant differences under five environments with an average of 16.01 g and 32.07 g (Table 1), respectively. TGW of the 201 individuals was mostly distributed between 20-30 g with an average of 24.60 g, 25.69 g, 22.80 g, 24.97 g, and 25.69 g in 2013NN, 2014NN, 2015NN, 2016NN, 2016WH (Table 1, Figure 1, Table S1), respectively. TGW of 104, 106, 104, 107, and 107 out of the 201 individuals was smaller than the average in 2013NN, 2014NN, 2015NN, 2016NN, 2016WH, and TGW of the remaining individuals was either equal to or larger than the average (Table S1).
QTL mapping of TGW by Simple Sequence Repeats (SSR)
TGW QTLs were preliminarily detected by 300 SSR markers with evenly distribution on the 12 chromosomes. The population were F6, F7 RILs derived from the cross of G1025 and K1561 planted in NN in 2013, 2014. Four QTLs TGW3, TGW7, TGW9.2, and TGW12 were stably detected on the chromosomes 3, 7, 9 and 12 in the two environments (Table 2). TGW12 had the largest effect which located on the region of RM247 and RM7003 (Table 2), so it was selected for further analysis. There were other 166 SSR markers (Table S2) in the region based on the genome sequencing data of Nipponbare [28]. The polymorphism of the 166 SSRs were firstly detected between the parental lines G1025 and K1561. As a result, nine SSRs showed polymorphism but only five displayed clear bands. The five SSRs were further used to detect F6, F7 RILs population in 2013, 2014. Finally, TGW12 was mapped to the 5.1 cM region between RM27638 and RM27748 (Figure 2).
QTL mapping of TGW by SLAF markers
We have developed 5, 521 SLAF markers by SLAF sequencing [29]. To further map TGW QTLs, Those SLAF markers were used to screen F8 RILs in NN in 2015 and F9 RILs in NN and WH in 2016, respectively. A total of eight QTLs were detected in three environments, namely, TGW7, TGW9.1, TGW12 in 2015NN, TGW7, TGW9.2, TGW12 in 2016NN, and TGW7, TGW12 in 2016WH (Table 2, Figure 3). TGW7 and TGW12 were both detected in three environments, and TGW9.1, TGW9.2 was each detected once. TGW7 explained the phenotypes for 8.01%, 10.76%, and 10.43% inheritance with an LOD of 6.94, 7.69, and 7.48 in 2015NN, 2016NN, and 2016WH, respectively, whereas TGW12 showed 22.36%, 17.48%, and 17.95% inheritance explaining for the phenotypes with an LOD of 15.42, 11.99, and 11.96 in the three environments, respectively (Table 2). TGW12 had larger effect than TGW7, which was consistent with the results analyzed by SSR mapping (mentioned above). Further analysis for TGW12 was conducted by comparing the linkage map constructed by SSR and SLAF markers. Consequentially, thirteen SLAF markers fell into the region of RM27638-RM27748, and TGW12 was further narrowed to 241.47 kb region between SLAFs Marker2768345 and Marker2853491 (Figure 2).
Evaluation of TGW12 phenotype and identification of TGW12 segment genetic origin
In order to evaluate whether the phenotypes were determined by TGW12, 16 out of the 201 RILs containing the TGW12 region were identified by means of the markers nearby the region (Figure 4). Then, the phenotypes and genotypes of the 16 RILs were compared. All the 16 RILs with one or two segments of K1561 showed TGW increase than the recurrent parent G1025, suggesting TGW12 control TGW (Figure 4). To clarify whether the increasing effect of TGW12 was originated from O. minuta, the genotypes of G1025, K1561, IR24, and O. minuta were detected using markers nearby or within TGW12. The genotype of K1561 was the same as that of IR24 but different from that of G1025 and O. minuta on the sites of RM27638 and RM27748, which are nearby TGW12 (Figure 2, Figure 5). However, the genotype of K1561 was the same as those of IR24 and O. minuta, but it was different from that of G1025 at the site of Marker2758157, which is within TGW12. We cannot presently draw a conclusion whether TGW12 originated from IR24 or O. minuta based on the above results. It has been suggested that translocation through centric break-fusion occurred more frequently than recombination in the introgression lines with interspecific cross, which didn’t always resulted in an O. minuta chromosome arm onto a complete or incomplete O. sativa chromosome [30]. Thus, we cannot simply identify the TGW12 origination by means of single marker. It is feasible to compare sequence of TGW12 candidate among O. minuta, IR24, and K1561 once it was fine mapped.
Preliminary prediction of candidate genes for TGW12
Analysis of annotated genes indicated that 32 ORFs located in the 241.47 kb region based on Nipponbare genome annotation (http://rice.plantbiology.msu.edu) (Table 3). Among them, 13 ORFs encoded functional proteins and 19 ORFs were annotated as transposon/retrotransposon proteins, hypothetical proteins, or expressed protein. It is worth noting that there were four transcription factors (TFs) among the functional proteins, that is, MADS-box (ORF17, ORF19), ZF-HD protein (ORF24), B-box zinc finger protein (ORF27). Because TFs play crucial roles in regulation of plant growth and development [31], the four TFs were considered preferentially as putative candidate genes of TGW12. Reverse transcript (RT)-PCR were conducted to amplify the CDS (coding domain sequence) of the four transcription factors from the parents G1025 and K1561. Sequence comparison indicated that the amplified sequence of ORF17 in K1561 was 56 bp shorter than that of G1025, which resulted in premature transcription termination that only encoding 45 amino acids, whereas ORF17 encoded 202 amino acids in G1025. Further analysis displayed that the decreased 56 bp in K1561 was due to alternative splicing (AS) in the first intron (Figure 6). There were no difference in the CDS of ORF19, ORF24, and ORF27 between K1561and G1025 (data not shown). Thus, the MADS-box (ORF17) is one putative candidate of TGW12. Of course, other nine functional proteins and the hypothetical proteins, or expressed protein cannot be excluded, which required more experiments to verify.