Rice is originated from wild rice (O. rufipogon Girff.) through domestication. This considerably changed the morpho-physiological characteristics of wild rice to in-crease cultivation efficiency, grain quality, and rice yield (Fuller et al. 2010; Huang et al. 2012). Compared to wild rice, cultivated rice typically exhibits favorable charac-teristics including erect growth, no or short awns, increased spikelet number per panicle, closed panicle, and reduced seed shattering and dormancy, all of which changed dramati-cally during domestication (Kovach et al. 2007; Sweeney and McCouch 2007).

Awn, a long extension of the lemma tip, helps seed dis-persal by wind and by sticking to human cloths or animal fur while providing seed protection from predators under natural conditions (Elbaum et al. 2007; Svizzero et al. 2019). In barley and wheat, awns also photosynthesize which contributes to grain filling (Abebe et al. 2009; Maydup et al. 2010; Guo and Schnurbusch 2016). In contrast, rice awns are not photosynthetically active because of the absence of chlorenchyma tissues (Toriba and Hirano 2014; Ntakirutimana and Xie 2019). Moreover, in agriculture, awns are inconvenient for harvesting, threshing, packing and storage (Takahashi et al. 1986). Hence, rice domesti-cation has led to cultivated rice with no or short awns (Luo et al. 2013).

Regent studies of the genetic mechanisms underlying the development of rice awns have suggested that awn devel-opment is a complex trait controlled by multiple genes. A total of 21 QTLs with major and minor effect on rice awn length in rice have been reported in the Gramene database (https://archive.gramene.org/). However, only a few major QTLs have been identified and characterized at the molecular level. An-1/RAE1 encodes a basic helix-loop-helix (bHLH) transcription factor that regulates the formation of awn primordia, cell division and grain length and reduces the grain number in wild rice (Luo et al. 2013). An-2/LABA1 encodes a cytokinin synthesis enzyme that promotes awn elongation by increasing cytokinin concentration in the awn primordia. It reduces the number of grains per panicle and tiller number per plant (Gu et al. 2015; Hua et al. 2015). RAE2/GAD1/GLA encodes a secreted signal peptide that regulates awn development as well as the number of grains per panicle and grain length (Bessho-Uehara et al. 2016; Jin et al. 2016; Zhang et al. 2019). The YABBY trans-cription factor DL, auxin responsive factor OsETTIN2, and RNA-dependent RNA polymerase SHL2 are also involved in awn formation (Toriba and Hirano 2014).

To identify novel genomic regions associated with awn length, we have performed QTL-seq (Takagi et al. 2013), a combination of bulk segregant analysis (BSA) and whole genome re-sequencing of DNA pools. An F₂ population derived from a cross between Tun Sart (an awnless cul-tivar) and Sobaekmangsudo(an awned cultivar) were used. This study will provide a valuable genetic resource for future molecular breeding in rice.

Plant materials

Two japonica rice cultivars [awnless Tun Sart (IT 004483) and awned Sobaekmangsudo (IT 006737)] were used as parental lines to develop the 197 F₂ population. [To simplify the following description, we represent Tun Sart as RWG-45 and Sobaekmangsudo as RWG-111]. RWG-45 and RWG-111 are part of the 137 KRICE_CORE pop-ulation (Kim et al. 2016) and both seeds were received from the Rural Development Administration (RDA) gene bank, Korea. F₂ population was obtained from Pusan national University, Miryung, Korea.

Awn length evaluation

Phenotyping was carried out at Pusan University, Miryang, Korea using 197 individual F₂ plants of RWG-45 × RWG-111. Three main panicles of each plant were used for pheno-typing and the awn length of the whole panicle was re-presented by the average of apical spikelets on each primary branch. Measuring the awn lengths was carried out two weeks after heading to avoid the awn breakage.

Construction of segregating pools

All young leaves of 197 individual F₂ plants were col-lected separately for total genomic DNA extraction using CTAB method (Porebski et al. 1997), with minor modi-fications. The genomic DNA of 23 individuals with ex-tremely short awn (ESA) and 30 extremely long awn (ELA) were selected as two bulked pools of the F₂ popula-tion. Isolated DNA was quantified using a Nanodrop spectrophotometer (Thermo Scientific, Wilmington, USA). Equal amounts of DNA from the ESA and ELA individuals were mixed.

QTL-seq analysis

Total genomic DNA was extracted from two bulked pools, and it was used to construct paired-end libraries with an insert size of 151 bp using TruSeq Nano DNA Kit (Illumina, San Diego, CA, USA). These libraries were sequenced using the Illumina NGS platform at Macrogen (Seoul, Korea). After sequencing, raw reads filtering was performed by fastp program (Chen et al. 2018). This data was aligned to the Nipponbare reference genome (IRGSP) by using the BWA program (Li and Durbin 2009). Samtools and GATK were used to clean the BAM file and for SNP variation calling, respectively (Li et al. 2009; McKenna et al. 2010). VCF file was filtered by vcftools (Danecek et al. 2011) to obtain high quality genotype data. The QTL-seq pipeline (QTL-seq version 2.1.3) (Takagi et al. 2013; Sugihara et al. 2022) was used for mapping the QTLs for Awn genes.

Prediction of candidate genes

To predict possible candidate genes associated with awn length, the following strategies were employed. First, we compared the DNA sequences of genes within the QTL regions between the two parents using the whole-genome DNA re-sequencing results to predict the candidate genes. Second, comparing the results with known genes/QTLs for awn length on 12 chromosomes in rice. Third, candidate genes were reselected according to their functional anno-tation from the rice genome database (http://rice.uga.edu/).

Evaluation of the awn length

The awn length of the two parental cultivars, RWG-45 and RWG-111 (Fig. 1A) along with their 197 F₂ population (Fig. 1B), were evaluated two weeks after heeding. Sig-nificant difference of the awn length was observed between the two parents. The F₂ population showed awn length variation from 0.3 to 112.5 mm (Fig. 1B). Among the 197 F₂ individuals, 23 extremely short awn and 30 extremely long awn plants were selected to prepare the ESA-pool and ELA-pool, respectively, which were then used for DNA re-sequencing.

Figure 1. Comparison of the awn phenotypes between RWG- 45 and RWG-111. (A) Phenotypic comparison of seed arrays from mature spikelets of RWG-45 (left) and RWG-111 (right). (B) Frequency distribution of the awn length in the F₂ population (RWG-45 × RWG-111).

Whole-genome sequencing and SNP identification

We performed high-throughput genome sequencing using four samples including RWG-45, RWG-111, ESA-pool, ELA-pool and obtained a total of 238.9 million reads and 32.6 Gb of raw data (Table 1). After cleaning the data by fastp, the average GC content was 42.310% and the Q30 of all the samples reached more than 91%. The mapped ratios between samples and the Nipponbare genome were 98.34%, 98.93%, 98.89%, and 98.94%, respectively. Most samples except RWG-111 (81.92%) showed properly paired ratio higher than 94% and unmapped ratio of all the sam-ples was lower than 1.7%. The average genome-coverage depth was 22X and the genome coverage was higher than 98%. These results suggest that the resequencing quality is confirmed and they could be used for the following analysis.

Table 1 . Quantity of genome sequence obtained for each sample.

Sample ID	Total reads	Total base	GC (%)	AT (%)	Q20 (%)	Q30 (%)	Mapped (%)	Properly paired (%)	Unmapped (%)	Average depth	Genome coverage (%)
RWG-045	31,616,276	3,190,458,278	39.8	60.2	97.0	92.1	98.34	94.18	1.66	10X	98.33
RWG-111	36,820,718	3,679,271,766	41.9	58.1	97.1	92.4	98.93	81.92	1.06	11X	98.50
ESA-pool	84,862,982	12,814,310,282	43.7	56.3	96.5	91.1	98.89	95.79	1.04	34X	99.98
ELA-pool	85,611,904	12,927,397,504	43.9	56.1	96.8	91.6	98.94	95.92	1.09	33X	99.98

We obtained a total of 735,910 variants those were including 583,569 SNPs and 152,341 indels from the sam-ples. Among the 1,213.9 K annotations, 619.6 K, 51.4 K, and 22.3 K were located at intergenic region, intron, and exon, respectively. 10.2 K and 8.4 K of the annotations lo-cated at the exon region were non-synonymous and syn-onymous, respectively (Table 2).

Table 2 . SNPs identified among two parents and two mixed pools.

Type	Number	Ratio (%)
SNP	583,569	79.30
MNP	0	0
INS	75,284	10.23
DEL	77,057	10.47
3’UTR	11,772	0.97
5’UTR	8,661	0.71
Downstream	205,221	16.93
Exon	22,380	14.52
Intergenic	619,687	51.12
Intron	51,482	4.25
Splice site acceptor	75	0.01
Splice site donor	81	0.01
Splice site region	1,360	0.11
Transcript	79,433	6.55
Upstream	212,115	17.50
Missense	10,208	54.32
Nonsense	160	0.85
Silent	8,425	44.83

QTL-seq analysis and sequence comparison of the candidate genes

Two major peaks on chromosome 4 and 8 were iden-tified for awn length and named as qAwn-4 and qAwn-8, respectively (Fig. 2). The qAwn-4 was spanning 7.5-Mb (12.8-20.3 Mb) intervals on chromosome 4 and the accu-racy of this QTL was ascertained by a valid 99% Δ(SNP- index) significance level (Fig. 2A, Table 3). The other QTL, qAwn-8, was spanning 4.9-Mb (22.3-27.2 Mb) inter-vals on chromosome 8 and the accuracy of this QTL was ascertained by a valid 95% Δ(SNP-index) significance level (Fig. 2B, Table 3). The qAwn-4 coding region har-bored 1,352 SNPs and 302 indels, and the qAwn-8 coding region harbored 880 SNPs and 254 indels (data not shown).

Table 3 . QTLs associated with awn development identi-fied using QTL-seq.

QTL name	Chr.	Start (Mb)	End (Mb)	Peak
qAwn-4	4	12.8	20.3	‒0.6010
qAwn-8	8	22.3	27.2	‒0.4153

Figure 2. Single nucleotide polymorphism (SNP)-index charts. (A and B) SNP-index charts of awnless-pool (green), awned- pool (orange), and corresponding Δ(SNP-index) plots (blue) with 95-99% confidence interval borders of RWG-45 × RWG-111 for chromosome 4 (A) and chromosome 8 (B). Average values of Δ(SNP-index) are plotted with a 2 Mb sliding window and a 50 kb increment.

Genomic sequence comparison was made between the two parents based on the genes that were previously reported on the identified QTL regions. The An-1 (Os04g0350700), a major gene that regulates awn development, was located in the qAwn-4 region. One SNP was identified in the second exon between awnless and awned plants in Os04g0350700 (Fig. 3). Another major gene RAE2 (Os08g0485500) was located in the qAwn-8 region. However, no nucleotide variants were found between awnless and awned plants indicating the existence of a novel gene contributing to awn length (data not shown). Further analysis will be performed to narrow down the candidate region and to discover the novel genes.

Figure 3. Sequence variation in an-1, awn-4, and Awn-4. The common variations among Nipponbare, RWG-45, and RWG-111 are indicated in this figure. Black bars represent introns and grey boxes represent coding regions. Bar = 1 kb.

Morpho-physiological traits of wild species have been modified to meet human needs during crop domestication. Wild rice typically exhibits long awns that help in seed dispersal and provide protection from predators under nat-ural conditions. However, in agriculture, long awns are inconvenient for pre-harvesting and post-harvesting because of its structure. Hence, rice domestication has led to culti-vated rice with no or short awns. Recently, studies have been conducted to elucidate the genetic mechanisms under-lying the rice awn development. Although several genes/ QTLs associated with awn development have been detected, only a few major QTLs have been cloned and characterized at the molecular level.

To identify the novel QTLs for awn length, two DNA pools with extreme phenotypic difference were used to perform QTL-seq analysis. We used the Δ(SNP-index) algorithm approach to map QTL regions at the 95% or 99% significance level. Two highly significant peaks (qAwn-4 and qAwn-8) were detected on chromosome 4 and 8, with the former mapped between 12.8-20.3 Mb and the latter between 22.3-27.2 Mb (Fig. 2).

QTLs and genes for awn length on chromosome 4 and 8 have been previously reported. The An-1 encoding a bHLH transcription factor was identified from wild rice (O. rufipogon) and it regulates long awn formation (Luo et al. 2013). This gene was detected in qAwn-4 region. In chromosome 8, the location of qAwn-8 in the present study was containing the previously reported gene, RAE2 that encodes one member of the epidermal patterning factor-like protein (EFPL) family regulating awn formation (Bessho- Uehara et al. 2016; Jin et al. 2016; Zhang et al. 2019).

Comparison between the genomic sequences of the two parents revealed that one SNP difference in the coding region of An-1 (Fig. 3), in which +244-bp (A > G) changes the amino acid sequence, suggesting that this SNP might be involved in awn formation and elongation. However, no SNPs were found between the sequence of the two parents in the coding region of RAE2 suggesting that a novel gene associated to awn length might exist in the interval of qAwn-8 which warrants further exploration. Therefore, further discovery of the novel genes will provide insights into the genetic architecture of awn length.

This work was supported by the Rural Development Administration, Republic of Korea (RS-2022-RD010201).