**1. Introduction**

Rice (*Oryza sativa*) is one of the most important cereal crops grown worldwide. Dry direct-seeded rice refers to the process of establishing the crop from seeds sown on non-puddled and unsaturated soil; in contrast, seedlings from nursery are transplanted into puddled or submerged soil in transplanted rice [1]. Compared with traditional transplanted rice, dry direct-seeded rice has been proposed as a water-e fficient and labor-saving approach, which can reduce the cost of water and labor at about 50% [2,3]. Also, dry direct-seeded rice could e fficiently utilize early-season rainfall in drought-prone environments and complete its growth cycle within the wet season in rainfed lowlands [4]. To date,

direct seeding has been adopted by more than 25% of the worldwide rice cultivation area [1]. Dry direct-seeded rice is becoming a popular option in Philippines, India, Thailand, Cambodia, Laos and Indonesia in tropical Asia as well as the US, Australia and Latin America [5,6].

However, a few constraints in dry direct-seeded rice, e.g., poor seedling establishment, weed infestation and lodging susceptibility could lead to large reduction in grain yield and quality [7–11]. Seedling establishment is very important for high yielding and controlling weed infestation in direct seeding rice [10]. Meanwhile, deep sowing could be good for root and basal internodes elongation in deeper soil, which could gain the stability of rice plant and increase lodging resistance of direct seeding rice [12]. However, deep sowing is always in contradiction to seedling uniformity and leads to grain yield reduction [13]. Rapid and good seedling establishment is important for weed competitiveness and good harvesting. Mesocotyl, an organ between the coleoptilar node and the basal part of seminal root in young monocot seedlings, plays a key role in pushing buds out of the deep water or soil during germination for successful seedling establishment. Longer mesocotyl can facilitate seedling establishment under deep sown conditions in dry direct-seeded rice [14–18]. Besides, Mgonja et al. (1988) also reported the strong association between mesocotyl length and seedling vigor [19]. However, the molecular mechanisms of rice mesocotyl length variation are poorly understood.

Genome-wide association study (GWAS) based on linkage disequilibrium (LD) has been widely adopted to identify loci significantly associated with important and complex agronomic traits in rice [20–22]. Extremely high resolution can be achieved by dense SNPs identified in diverse germplasm panels based on the next-generation sequencing (NGS) or SNP chip approaches [20,21,23,24]. Thousands of rice landraces or cultivars have been screened for mesocotyl length by GWAS and dozens of QTL have been reported in previous studies [17,18,25,26]. Wu et al. (2015) screened 270 rice accessions and 16 loci were identified associated with mesocotyl elongation [17]. Furthermore, 469 Indica accessions were used to measure mesocotyl length, and 23 loci were significantly associated with mesocotyl length [18]. Zhao et al. (2018) evaluated the mesocotyl length of 621 rice accessions and detected 13 QTLs [26]. Also, Sun et al. (2018) have identified three QTL for mesocotyl length from 510 rice accessions [25].

Identifying the QTLs in multi-environments could provide accurate information for gene cloning and molecular breeding. In this study, GWAS based on resequencing was conducted in a set of 208 rice accessions for mesocotyl elongation under sand, water and soil (2.0 cm, 4.0 cm and 6.0 cm). The objectives of this study were: (1) To dissect the genetic architecture of mesocotyl elongation, (2) to identify SNPs significantly associated with mesocotyl length, (3) to search for candidate mesocotyl elongation genes for further study, and (4) to select accessions with longer mesocotyl for the breeding of mechanized dry direct-seeded rice.
