*4.2. LD Pattern*

LD decay a ffects the precision of GWAS and was influenced by population structure, allele frequency, recombination rate and selection. Previous studies reported that LD decay in rice ranged between 120 Kb and 270 Kb [18,23,31,32]. In this panel, the LD decay was about 150 Kb for the whole genome (Figure 3e), consisting with previous reports [18,23,32,33].

### *4.3. GWAS Was E*ff*ective to Identify QTLs Controlled Complex Quantitative Traits*

So far, several previous studies have reported the associated QTLs and candidate genes for rice mesocotyl elongation using linkage mapping or GWAS [20,25,34–36]. In this study, the loci identified on chromosome 7 (10.03–13.61 Mb) at the 6 cm sowing depth of soil was overlapped with QTLs *qFML7-1*, *qFML7-2*, *qFML7-3*, *qFML7-4* and *qFML7-5* [26]. The QTL on chromosome 1 (14.1–17.3 Mb) were overlapped with *qml1-1* [37,38] and *qML1* [23]. Furthermore, two QTLs on chromosome 2 (5.6–8.7 Mb) and 6 (7.3–9.7 Mb) identified under water and sand were overlapped with the QTL *qML2* and *qML6* detected from Shennong 265/Lijiangxintuanheigu RIL population [20]. Ouyang et al. (2015) have reported *qml3-1*, *qml4-1* and *qml5-1* by a RIL population [39], which nearly coincide with the QTL identified on chromosome 3 (15.2–15.3 Mb), 4 (8.3 Mb) and 5 (5.8–6.3 Mb) in our study. Zhang et al. (2006) detected four QTLs for submergence tolerance in seedling stage from a DH population, which explained phenotypic variations from 10.7% to 29.3% [35]. Among these, *qLOE-8* were overlapped with the QTL regions located on chromosome 8 (208–5.2 Mb) identified in sand. Lee et al. (2012) reported two QTLs associated with mesocotyl elongation, *qMel-1* (R<sup>2</sup> = 37.3%) and *qMel-3* (R<sup>2</sup> = 6.5%) in weedy rice on chromosomes 1 and 3, respectively [40]. Moreover, Lu et al. (2016) reported 17 loci and 289 candidate genes for rice mesocotyl by GWAS on chromosomes 1 (4), 4 (4), 5 (1), 6 (3), 7 (1), 9 (3) and

11 (1), respectively [18]. QTLs identified on chromosomes 9 (1.3–2.8 Mb) and 12 (13.3–15.0 Mb) were overlapped with those QTLs reported by previous studies [18,40].

Wu et al. (2015) found that when under water culture and at a significant level of –log10(P) ≥8.0, a total of 13 loci were detected to be significantly associated with mesocotyl length, while only three significant SNPs were declared when the medium was merely changed to sand, and two of the three SNPs were co-localized under both conditions [17]. Lu et al. (2016) reported that six and seven loci were detected under two environments, respectively [18]. Only two of them on chromosome 6 were identified under both environments. In di fferent studies, nearly 40 loci significantly associated with mesocotyl length have been detected on all twelve chromosomes by GWAS, and most of these loci in each study cannot be validated by others [17,18,25,26]. The interaction of G (genotype) × E (environment) can influence QTL and association mapping results, indicating that the e ffect of identified genomic regions was required to be estimated for each environment [41,42]. Only the major genomic regions could be co-localized under multi-environments.

### *4.4. Potential Candidate Genes for Mesocotyl Elongation were Identified*

According to the result of GWAS, 107 candidate genes were found to be related to mesocotyl elongation. Among them, five involved in the biological metabolism of phytohormones, cell elongation and division were selected as the high-confidence candidate genes. Bioinformatics analysis indicated that loci on chromosome 5 (5.6–6.1 Mb) corresponding to zinc finger CCCH type family protein (*LOC\_Os05g10670*) and transcription factor jumonji (*LOC\_Os05g10770*). CCCH-type zinc finger proteins comprise a large family that is induced by drought, high temperature stress and hydrogen peroxide, and is also induced by abscisic acid, methyl jasmonate and salicylic acid [43,44]. Noh et al., (2004) reported that jumonji class transcription factor controls stem elongation, transposon activity and panicle development [45]. Besides, Feng et al. (2017) have identified several candidate genes for mesocotyl in rice including zinc-finger protein genes, which involved in the JA biosynthesis and signaling pathways [29]. Besides, a candidate gene on chromosome 9 (1.2–1.5 Mb) encoding cytokinin-O-glucosyltransferase 2 were identified. CK is a class of plant hormones that were first identified as cell division-promoting factors and were subsequently identified as factors that control various processes in plant growth and development, including mesocotyl elongation [46]. Besides, CK played an important role in the biosynthesis of BRs, a group of steroid plant hormones essential in plant growth and development [47]. Two candidate genes in loci 4.3–4.5 Mb and 14.1–14.5 Mb on chromosomes 12 were identified, which encodes flavin monooxygenase and 9-cis-epoxycarotenoid dioxygenase 1, respectively. Flavin monooxygenase catalyzes hydroxylation of the amino group of tryptamines, a rate-limiting step in tryptophan-dependent auxin biosynthesis, which regulate many processes in plant development [48]. The 9-cis-epoxycarotenoid dioxygenase is essential for the biosynthesis of ABA [49]. The 9-cis-epoxycarotenoid dioxygenase gene could lead to the over-production of abscisic acid, which plays an important role in mesocotyl elongation [50]. Expression of the five candidate genes in di fferent accessions indicated that five genes were all functional in regulating mesocotyl elongation. The most significant di fference in expression level was observed in gene, *LOC\_Os12g24800*, which indicated that this gene could be the most likely functional gene for mesocotyl length. These results also proved that the loci identified by GWAS were reliable. Mesocotyl elongation is a consequence of complicated biological processes and its mechanism remains unclear; more detailed experimental analyses are needed to confirm the function of candidate genes in mesocotyl elongation.
