The early-matured aromatic long-grain japonica rice variety QG10, which was developed by our own group, was licensed for release in 2019 and is now widely planted in Heilongjiang province in Northeast China. It was a semi-dwarf rice variety with a lot of superior traits such as long panicle, long grain, aromatic, and good quality ( Figures 1A-D ). It was selected from the cross between two aromatic japonica/Geng rice Wuyoudao4 (WYD4) and Suigeng4 (SG4) ( Figure 1E ). The seedlings of QG10 were grown on the agricultural farm of the Qiqihar Branch of Heilongjiang Academy of Agricultural Sciences. Field management practices were performed according to the most commonly followed agricultural practices of local farmers. The leaves, stems, roots, and panicles at heading stages from plants grown in the experimental station were collected in liquid nitrogen for isolating RNA. The young leaves of a single young plant were used to isolate genomic DNA.

The high molecular weight genomic DNA of QG10 was extracted from the 15-day-old leaf tissues following a modified CTAB method. Whole genome sequencing was done following the standard instructions of the Ligation Sequencing Kit (Nanopore, Oxford shire, UK). The quantified DNA was randomly sheared, and fragments of ∼20 kb were enriched and purified. Then, a 20-kb library was constructed and sequenced on the Nanopore PromethION platform according to the manufacturer’s protocols (Jiang et al., 2020).

De novo genome assembly of Nanopore sequence was performed as follow: The raw Nanopore reads were error-corrected and assembled using CANU (v1.7.1) (Koren et al., 2017), followed by Smartdenovo (https://github.com/ruanjue/smartdenovo) assembly, followed by three rounds of polishing with Racon (Vaser et al., 2017), followed by three rounds with Pilon v0.3.0 using the Illumina PCR-free paired-end reads (Walker et al., 2014). Genome completeness was also assessed using the algae dataset of BUSCO v2.0 (Simão et al., 2015).

A Hi-C fragment library with a 300-700 bp insert size was constructed from the genomic DNA of the QG10. Briefly, adapter sequences of raw reads were trimmed and low-quality PE reads were removed for clean data. The library was sequenced on the Illumina HiSeq4000™ platform by Biomarker Technologies (Beijing, China). The clean Hi-C reads were first truncated at the putative Hi-C junctions and then the resulting trimmed reads were aligned to the assembly results with the software package bwa aligner (Li and Durbin, 2009). Only uniquely alignable pair reads whose mapping quality of more than 20 retained for further analysis. Invalid read pairs, including dangling-end and self-cycle, re-ligation, and dumped products, were filtered by the software package HiC-Pro v2.8.1 (Servant et al., 2015). The 96.25% of unique mapped read pairs were valid interaction pairs and were used for the correction of scaffolds and clustered, ordered, and orientated scaffolds onto chromosomes by the software package LACHESIS (Burton et al., 2013).

Before chromosome assembly, we first performed a preassembly for error correction of scaffolds which required the splitting of scaffolds into segments of 50 kb on average. The Hi-C data were mapped to these segments using the software package BWA (version 0.7.10-r789) (Li and Durbin, 2009). The uniquely mapped data were retained to perform assembly by using the software package LACHESIS (Burton et al., 2013). Any two segments which showed an inconsistent connection with information from the raw scaffold were checked manually. These corrected scaffolds were then assembled with the software package LACHESIS. After this step, placement and orientation errors exhibiting obvious discrete chromatin interaction patterns were manually adjusted.

The RNA of QG10 were isolated from the mixed tissues (leaves, culms, roots, and panicles) following the manufacturer’s protocol (Wang et al., 2023). We then performed the sequencing on the Illumina HiSeq 2500 platform according to the manufacturer’s instructions. The repetitive sequence of the genome based on the principle of structure prediction and de-novo prediction was constructed with the software package LTR_FINDER v1.05 (Xu and Wang, 2007) and the software package RepeatScout v1.0.5 (Price et al., 2005). The PASTE Classifier was used to classify the database (Hoede et al., 2014). Then it was merged with the database of Repbase as the final repeat sequence database (Jurka et al., 2005). And then the software package RepeatMasker v4.0.6 was used to predict the repeat sequence of the QG10 genome based on the constructed repetitive sequence database (Tarailo‐Graovac and Chen, 2009). the software packages Genscan (Burge and Karlin, 1997), Augustus v2.4 (Stanke and Waack, 2003), GlimmerHMM v3.0.4 (Majoros et al., 2004), GeneID v1.4 (Alioto et al., 2018), and SNAP (Korf, 2004) were used for de-novo prediction. The software package GeMoMa v1.3.1 was used for prediction based on homologous species (Keilwagen et al., 2016; Keilwagen et al., 2018). The software packages Hisat v2.0.4 (Kim et al., 2015) and Stringtie v1.2.3 (Pertea et al., 2015) were used for assembly based on reference transcripts, and the software packages TransDecoder v2.0 and GeneMarkS-T v5.1 (Tang et al., 2015) were used for gene prediction. The software package PASA v2.0.2 was used to predict Unigene sequences without reference based on transcriptome data (Campbell et al., 2006). Finally, the software package EVM v1.1.1 (Haas et al., 2008) was used to integrate the prediction results obtained by the above three methods. The predicted gene sequences were compared with NR, KOG, GO, KEGG, TrEMBL, and other functional databases by the software package BLAST v2.2.31 (-evalue 1e-5) (Altschul et al., 1990) to perform KEGG pathway, KOG function, GO function and other genes functional annotation analysis.

The whole-genome assemblies sequences of QG10 were compared with the rice reference genome sequence (Oryza_sativa_MSU7 version) using the software package MUMmer v3 (Kurtz et al., 2004). According to the results from the software package MUMmer, the sequence variations and SVs were further re-called using the software package BLAST. The synteny/inversion comparison were analysis by using GenomeSyn_Win.v1 (Zhou et al., 2022). At the site of each sequence variant, the genotypic information for QG10, Nipponbare, and the elite variety having important genes was called according to the results of the one-to-one alignments. The allelic information of sequence variants was detected based on gff files from the Oryza_sativa_MSU7 version. The software packages ClustalW v1.8.3(Thompson et al., 1994) and BLAST v2.2.31 were used for re-detected the sequence variations and detailed haplotype analyses for the well-characterized genes in rice (Zhao et al., 2018b).

We sequenced QG10 genomic DNA to generate about 73.52Gb of Nanopore sequencing raw data. After data quality control, the clean data volume was 63.23Gb containing 3,279,893 reads with a total of 166.34-fold sequencing depth. The reads with 10 -20 kb and 20 - 30 kb sequencing length were account for 51.56% ( Table 1 ). The mean reads length of clean sequencing data was 19.28 kb with an N50 length of 25.64 kb. The clean Nanopore sequencing data was re-corrected with the software package Canu (Koren et al., 2017). Then the third-generation sequencing data was re-corrected with the software package Racon (Vaser et al., 2017) for three rounds. Then, the second-generation data were used for three rounds of correction by Pilon (Walker et al., 2014) software, and the stain was removed according to NT alignment. Finally, we obtained a 380.15 Mb genome sequence with the contig N50 was 12.24 Mb. The completeness estimated by Benchmarking Universal Single-Copy Orthologs (BUSCO) was 98.12%.

We conducted the assembly in a stepwise fashion following a previously reported approach (Jiang et al., 2020). The Hi-C sequencing raw data was filtered, and the splice sequences and low-quality reads were removed to obtain high-quality clean data. The mapped data was obtained by sequence alignment of clean data with the preliminarily assembled genome. Finally, effective Hi-C data were used for further assembly of the draft genome sequence. LACHESIS software was used for clustering, sorting, and orientating the preliminary assembled genome sequence, and finally, the genomic sequence at the chromosome level was obtained. Finally, we assembled the genome (QG10) into 77 contigs with an N50 length of 11.80 Mb in 27 scaffolds with an N50 length of 30.55 Mb. The assembled genome size was 378.31Mb and the 65 contigs constituted approximately 99.59% of the whole genome. Visualization of the Hi-C signals indicated that 12 square matrix areas in the Hi-C heat map displayed significant differences from the background. These scaffolds were anchored into chromosomes 1–12, respectively ( Figure 2A ; Table 2 ).

According to the whole genome comparison, the genome of QG10 showed four sequence inversions with a length of about 0.5-2 Mb compared with the Nipponbare genome at the position of about 14-16 Mb on chromosome 4, 30-31 Mb on chromosome 5, 5.5-6 Mb on chromosome 8, and 5.5-6 Mb on chromosome 10 ( Figure 2B ). We identified 1,080,819 SNPs and 682,392 InDels between QG10 and Nipponbare on 12 chromosomes ( Figures 2C, D ).

We annotated the repeat regions in our QG10 assembly by Repeat Masker and detected 492,503 repetitive regions with 177.52 Mb repeat length that contained 242,211 Class I retrotransposons, 223,439 Class II DNA transposons, 833 Potential Host Gene, and 2,222 simple sequence repeats ( Table 3 ). The repeat regions make up 46.7% of the QG10 assembly genome.

We predicted 39,465 genes by the Ab initio method, 56,999 genes by the Homology-based method, and 24,998 by RNA-seq. Finally, a total of 57,599 genes were integrated with the prediction results obtained by the above three methods by using the software package EVM v1.1.1 ( Table 4 ). A total of 723 tRNA, 306 rRNA, 194 miRNA, and 5,392 pseudogenes were also predicted. 94.11% of the genes could be annotated into NR, GO, KOG, KEGG, and other databases ( Table 5 ).

We investigated the sequence variations in 19 cloned genes controlling grain size, including GW2 (Song et al., 2007), GS2 (Hu et al., 2015), qTGW2 (Ruan et al., 2020), BG1 (Liu et al., 2015), OsLG3 (Yu et al., 2017), OsLG3b (Yu et al., 2018), GS3 (Fan et al., 2006), qTGW3 (Hu et al., 2018), GS5 (Li et al., 2011), GW5 (Weng et al., 2008), GS6 (Sun et al., 2013), GW6 (Shi et al., 2020), TGW6 (Ishimaru et al., 2013), GW6a (Song et al., 2015), GL6 (Wang et al., 2019), GLW7 (Si et al., 2016), GW7 (Wang et al., 2015), qGW8 (Wang et al., 2012), and GS9 (Zhao et al., 2018a), to explain the long grain of QG10. A total of five grain size elite alleles (qTGW2^Nipponbare , qTGW3^Nanyangzhan , GW5^IR24 , GW6^Suyunuo , and qGW8^Basmati385 ) were identified controlling grain size in QG10 ( Figure 3 ). The qTGW2 allele in QG10 was identical to Nipponbare having the key variants (G/A) at -1818 bp in the promoter region ( Figure 3A ). The qTGW3 allele in QG10 had the key splicing-site mutation as Nanyangzhan, which was a long-grain indica rice ( Figure 3B ). The GW5 allele in QG10 was found without the critical loss of the 1,212 bp deletion mutation as IR24 a long narrow grain indica rice ( Figure 3C ). The GW6 allele in QG10 had the key mutation (6 bp) as Suyunuo, which was a wider grain indica rice ( Figure 3D ). The qGW8 allele in QG10 had the five variants as Basmati385, a long narrow grain indica rice ( Figure 3E ). Tracing the origin of these genes, it was found that qTGW3 (Nanyangzhan type), GW5 (IR24 type), GW6 (Suyunuo type), and qGW8 (Basmati385 type) belonged to indica subspecies, while qTGW2 (Nipponbare type) were mainly derived from japonica subspecies. In addition, these genes controlling grain type are all rare genotypes in japonica rice and have important application value in long grain type japonica rice breeding.

Cold tolerance is the key agricultural trait controlling rice production and geographic distribution. We investigated the sequence variations in 10 cloned genes controlling cold tolerance, including bZIP73 (Liu et al., 2018), COLD1 (Ma et al., 2015), Ctb1 (Saito et al., 2010), CTB2 (Li et al., 2021b), CTB4a (Zhang et al., 2017), HAN1 (Mao et al., 2019), ltt1 (Xu et al., 2020), OsLTPL159 (Zhao et al., 2020), qLTG3-1 (Fujino et al., 2008), and qPSR10 (Xiao et al., 2018), to explain the high cold tolerance of QG10. A total of four elite alleles (COLD1^Nipponbare , bZIP73^Nipponbare , CTB4a^{Kunmingxiaobaigu} , and CTB2^{Kunmingxiaobaigu} ) were identified as controlling cold tolerance in QG10 ( Figure 4 ). The COLD1 allele in QG10 was identical to Nipponbare to have the key SNP in the fourth exon region ( Figure 4A ). The bZIP73 allele in QG10 was found having the key SNP mutation (G/A) as Nipponbare ( Figure 4B ). The CTB4a allele in QG10 was found to have the ten mutations as Kunmingxiaobaigu, which was a cold tolerance variety from Yunnan Province ( Figure 4C ). The CTB2 allele in QG10 was found also have the ten key SNPs as Kunmingxiaobaigu ( Figure 4D ). The COLD1 (Nipponbare type) and bZIP73 (Nipponbare type) all belong to japonica subspecies. The CTB4a (Kunmingxiaobaigu type) and CTB2 (Kunmingxiaobaigu type) were all rare alleles in Northeast japonica rice and have important application value in Northeast japonica rice breeding.

The heading date is one of the most important factors determining rice distribution and the final yield. We investigated the sequence variations in 11 cloned genes related to early heading under long-day conditions, including DTH7 (Gao et al., 2014), Ghd7 (Xue et al., 2008), Ghd8 (Yan et al., 2011), Ehd1 (Doi et al., 2004), Ehd3 (Matsubara et al., 2011), Ehd4 (Gao et al., 2013), Hd1 (Yano et al., 2000), Hd3a (Kojima et al., 2002), Hd6 (Takahashi et al., 2001), Hd16 (Hori et al., 2013), and Hd17 (Matsubara et al., 2012). Among these heading date genes, only the DTH7, Ghd7, and Hd1 haplotypes were found to have non-functional alleles. The DTH7 allele in QG10 was found to have the three mutations as Kitaake, which originated at the northern limit of rice cultivation in Hokkaido, Japan ( Figure 5A ). Kitaake is reported insensitive to day length, short in stature, and completes its life cycle in about 9 weeks (Jain et al., 2019). The Ghd7 allele in QG10 was found having the critical mutations as Hejiang19, which is an early-maturity rice variety in Heilongjiang Province ( Figure 5B ). The Hd1 allele in QG10 was found to have the non-functional allele as Longgeng31, which is the major plant rice variety in Heilongjiang Province ( Figure 5C ). These results indicated that the heading gene combinations of Hd1, DTH7, and Ghd7 determined the early heading in QG10 in the northernmost province of China.

Blast is one of the most devastating rice diseases in Heilongjiang Province. We investigated the sequence variations in 14 cloned rice blast-resistant genes, including Pi5 (Lee et al., 2009), Pi21 (Fukuoka et al., 2009), Pi36 (Liu et al., 2007), Pi37 (Lin et al., 2007), Pi54 (Sharma et al., 2010), Pi56 (Liu et al., 2013), Pia (Okuyama et al., 2011), Pish (Takahashi et al., 2010), Pit (Hayashi and Yoshida, 2009), Pita (Bryan et al., 2000), Pigm (Deng et al., 2017), Pid2 (Chen et al., 2006), Pid3 (Shang et al., 2009), and Pid4 (Chen et al., 2018), to explain the blast resistance. Finally, only two blast resistance genes, Pia and Pid4, were found in QG10. The Pia allele in QG10 was found to have the resistant genotype as Akihikari, which encodes a nucleotide-binding site (NBS) and a C-terminal leucine-rich repeat (LRR) domain protein ( Figure 6A ). The Pid4 allele in QG10 was found to have the resistant genotype as Digu, which encodes a coiled-coil nucleotide-binding site leucine-rich repeat (CC-NBS-LRR) protein ( Figure 6B ). The Pia (Akihikari type) was the major blast-resistant gene in Northeast China. The Pid4 (Digu type) was a rare allele in japonica rice and has important application value in Northeast japonica rice breeding.

Rice stripe virus (RSV), an RNA virus belonging to the genus Tenuivirus and transmitted by small brown planthoppers, causes one of the most destructive rice diseases (Wang et al., 2014). RSV has become more and more serious in Heilongjiang province in recent years. But, almost the majority of japonica varieties cultivated in Heilongjiang are highly susceptible to RSV (Wang et al., 2014). The STV11 was the first cloned resistant gene of RSV, which encodes a sulfotransferase (OsSOT1) catalyzing the conversion of salicylic acid (SA) into sulphonated SA (SSA) (Wang et al., 2014). The STV11 allele in QG10 was found to have the resistant genotype as Kasalath, which is a high-resistance indica landrace ( Figure 6C ). The STV11^QG10 was a useful resistant gene in japonica rice breeding.

Fragrant rice is popular among consumers worldwide because its market price is much higher than that of nonfragrant rice. Fragrant was found to be controlled by BADH2 in rice (Chen et al., 2008). The BADH2 in QG10 was the typical badh2-E2 type of a 7-bp deletion as DHX2 that caused fragrance ( Figure 7A ). We also investigated the sequence variations in three cloned grain number genes (Gn1a (Ashikari et al., 2005), GNP1 (Wu et al., 2016), and SPIKE (Fujita et al., 2013)), and three lodging-resistance genes (Sd1 (Spielmeyer et al., 2002), SCM2 (Ookawa et al., 2010), and SCM3 (Yano et al., 2014)). Only the SCM3 allele in QG10 was found to have the key allele as Chugoku117, which is a stronger culms rice variety in Japan ( Figure 7B ). There was no elite grain number gene found in QG10. NRT1.1B is a key gene controlling the nitrogen-use efficiency (NUE) in rice (Hu et al., 2015). The NRT1.1B allele in QG10 was found to have the indica variation, which has higher nitrate absorption activity ( Figure 7C ).

Japonica/Geng and Indica/Xian are the two major subspecies of Asian cultivated rice (Zhang et al., 2016a). Owing to long-term differentiation and adaptation, both Indica and Japonica rice contain many favorable genes. Therefore, combining the favorable genes of the two subspecies has great value for creating genotypes with greater yield potential, stronger stress resistance, and better quality (Gu, 2010). Over the past 50 years, the combination of plant ideotypes and favorable vigor through hybridization between indica and japonica rice has greatly contributed to yield improvements in modern japonica rice in Northeast China (Tang and Chen, 2021). In recent years, a series of high-yielding and good-quality japonica cultivars have been obtained from hybridization of Indica/Japonica and the cultivation area of them was more than 4 million ha in Northeast China (Cui et al., 2022). The new fragrant early japonica rice cultivar QG10 was derived from a cross between ‘Wuyoudao4 and Suigeng4’, which were all derived from the hybridization of Indica/Japonica. In recently, Wang et al. (2023) chose six interrelated modern Chinese temperate japonica varieties and six related Japanese japonica varieties to investigate genome enhancement in temperate japonica varieties during modern breeding. They found many large SVs in Zhonghua11 (ZH11), Liaogeng5 (LG5), and Daohuaxiang2 (DHX2/WYD4). These large-fragment in the same location introgression from indica were also found in QG10 on chromosomes 4, 8, and 10 ( Figure 2B ). Several indica superior alleles including qTGW3^Nanyangzhan , GW5^IR24 , GW6^Suyunuo , qGW8^Basmati385 , Pid4^Digu , STV11^Kasalath , NRT1.1B^IR24 , and badh2-E2 were also be found in QG10. This information indicated that the indica genome introgression was common in the modern temperate japonica rice breeding in Northeast China. Superior alleles were found in Qigeng10 and had important value for breeding in Northeast China

Greater yield potential, stronger stress resistance, and better quality (longer grain and fragrant) are key agronomy traits that directly influence the market price of rice. Consumers in East Asia, including North China, Japan, and Korea tend to prefer longer fragrant japonica rice (Lu et al., 2022). So, the longer fragrant japonica rice from Northeast China, represented by Daohuaxiang2 (DHX2/WYD4), is the most famous rice in the Chinese market. QG10 was derived from DHX2 and solved some defects of DHX2 including poor lodging resistance, lack of cold tolerance, weak blast resistance, and late maturity. Therefore, the construction of a high-quality genome of ‘QG10’ is essential for further improvement of this cultivar or its progenies, as well as accelerating the process of fragrant japonica rice breeding, by providing genomic resources that could be directly applied to fragrant japonica rice cultivars. In this study, we found five superior alleles (qTGW2^Nipponbare , qTGW3^Nanyangzhan , GW5^IR24 , GW6^Suyunuo , and qGW8^Basmati385 ) controlling long grain size in QG10. To compare the phenotype of different gene haplotypes in rice germplasm, we investigated the grain shape traits and days to heading of 3k cultivars in the website (https://www.rmbreeding.cn/)(Wang et al., 2020). The results showed that the functional haplotype of QG10 controlling longer and slider grain ( Figures 8A-E ) and early heading (Figure 8f-h). Most of them was belong to the rare alleles for controlling longer grain and were less application in rice breeding in Northeast China. The blast resistant alleles (Pid4^Digu ), RSV allele (STV11^Kasalath ), and NUE allele (NRT1.1B^IR24 ) were also belong to the rare alleles for rice breeding in Northeast China. In the future, we will develop molecular assisted markers for the improvement of japonica rice varieties in Northeast China, and expanded the gene pool of japonica rice in Northeast China.