*2.3. Measurement of Mesocotyl Length*

To evaluate the variation of mesocotyl length (the distance from the basal part of seminal root to the coleoptilar node) at varied sowing depths, two replications were set up, and in each replication 15 good-quality seeds from each accession were sown at a depth of 2, 4 and 6 cm in soil contained by plastic trays. The size of plastic tray contained 50 holes with the size of 9.5 cm depth, 4.5 cm top diameter and 2.1 cm bottom diameter. After sowing, the plastic tray was kept in a plastic pallet with 3-cm-deep soil, and the whole system were maintained in a 30 ◦C dark incubator and the soil in each pallet was kept water saturated for seed germination and seedling growth. Emergence rate was recorded every day, until the emergence rate of any accession reached 100%. Three days after that, seedlings from each hole were carefully excavated and washed for measuring mesocotyl length using Image J (https://imagej.en.softonic.com/).

To quantify the mesocotyl length under di fferent culture media, sand and water culture were used, in addition to soil culture. Fifteen good-quality seeds were sown at a depth of 6 cm in sand, the culture method and sampling timing in sand was the same as that in soil. As for water culture, 15 good-quality seeds were put on a gauze in each hole of a plastic tray (50 holes with a depth of 4.8 cm, top diameter at 4.1 cm and bottom diameter at 2.1 cm). The plastic tray was put on a plastic pallet containing water in level of 2 cm from the bottom. Each plastic tray was wrapped with tin foil paper to avoid illumination and kept in phytotron with constant temperature of 30 ◦C for 10 days. After that, ten uniform seedlings out of 15 seeds of each accession were used for mesocotyl length measurement with Image J. Two replications were set up in both culture methods and mean of the two replications was used for GWAS analysis.

### *2.4. Population Structure and LD Decay Analysis*

The population structure of the 208 accessions was analyzed using Admixture 1.3.0 [27] based on 2000 SNP markers with default cross-validation (K value ranged from 2 to 5). An adhoc quantity statistic ΔK was used to predict the real number of subpopulations and k = 2 was chosen. Besides, the principal component analysis (PCA) and neighbor-jointing trees were also used to validate population stratification with the software Tassel v5.0 (https://www.maizegenetics.net/tassel).

LD decay analysis was done for the whole population. LD decay was measured by correlation coe fficients (*r*2) for all pairs of SNPs within 500 Kb that were calculated using PopLDdecay v3.27 (https://github.com/BGI-shenzhen/PopLDdecay) with the following parameters: -MaxDist 500-MAF 0.05-Het 0.88-Miss 0.999.

### *2.5. Genome-Wide Association Analysis and Candidate Genes Identification*

Associations between genotypic and phenotypic data were analyzed using the kinship matrix in an MLM (Mixed linear model) by GAPIT (http://www.zzlab.net/GAPIT/) based on R 3.6.1 to control background variation and eliminate the spurious MTAs. Due to Bonferroni-Holm correction for multiple testing ( α = 0.05) was too conserved and only a few significant MTAs were detected with this criterion, markers with an adjusted –log10 (*p*-value) ≥ 6.0 were regarded as the significant ones.

Candidate genes for the loci consistently identified in two or more environments were identified. The following steps were conducted to identify candidate genes for important QTL. Firstly, found all the genes located in LD block region around the peak SNP (± 150 Kb based on LD decay analysis) of each important QTL from the MSU Rice Genome Annotation Project (http://rice.plantbiology.msu.edu/ cgi-bin/gbrowse/rice/). Then, all available SNPs located inside of these genes were searched. The genes (except for expressed protein, hypothetical protein, transposon protein and retrotransposon protein) with SNP in coding region and which could further lead to sense mutation were considered as candidate genes. Besides, mesocotyl elongation is highly regulated by various phytohormones, including

strigolactones (SLs), cytokinin (CTK), brassinosteroid (BR), abscisic acid (ABA), jasmonates (JAs), gibberellins (GA) and auxin (IAA). Thus, these genes involved in the metabolism of the phytohormones talked above were selected as high-confidence candidate genes for mesocotyl elongation.
