Dissection of Seven Saline Tolerance Genes in a Geng/Japonica Population Reveals the Subspecie-Specific Selection of RST^1530A in Rice

1. Introduction

The Huang-Huai-Hai (3H) rice region, a vital area for geng/japonica rice cultivation in China, typically refers to the agricultural zone formed by the alluvial plains where the Yellow, Huai, and Hai Rivers converge [1,2]. Rice yield in this region plays a crucial role in ensuring food security. The 3H rice region is bordered by the Bohai and Yellow Seas of China. Due to seawater intrusion, rising groundwater levels, and strong evaporation, the area of saline-alkali soil continues to expand, increasingly subjecting geng/japonica rice to salt stress during its growth [3,4,5]. This type of abiotic stress not only inhibits seed germination and seedling growth and development but also significantly reduces yield [6,7]. Therefore, breeding salt-tolerant geng/japonica rice varieties has become an urgent priority to ensure food security in this region [8]. Although many QTLs and genes have beenidentified to confer rice salt tolerance (ST), most of them remain at the theoretical level [9]. Only a few loci, like “Saltol”, have been reported as being applied for breeding salt-tolerant varieties [10]. Other loci that have been found to be related to ST need further functional tests in natural varieties, especially in those cultivated in saline-alkali areas， and varieties of the 3H rice region are appropriate for such assessment.

In the field of rice ST research, several key genes involved in ion balance, signal transduction, and transcriptional regulation have been identified from various materials [11,12]. However, most studies focus on xian/indica rice or model varieties, and the natural variation and breeding potential of these genes in geng/japonica rice—particularly local germplasm from the 3H region—remain unclear. For instance, SKC1, also referred to as OsHKT1;5, is a Na⁺ selective transporter of the HKT family in rice [13]. Expressed mainly in root stele and xylem parenchyma cells, it unloads Na⁺ from the xylem to reduce sodium buildup in aerial tissues and improve plant salt resistance [13,14,15]. It is the causal gene underlying the major ST QTL “Saltol” on rice chromosome 1 [10,16]. OsWRKY53, mapped by genome-wide association studies (GWAS), is a key negative regulator gene of ST, directly repressing the expression of the major salt-tolerant gene SKC1 (OsHKT1;5) and OsMKK10.2 [17]. Natural variation of OsWRKY53 enhances ST predominantly in xian/indica rice varieties [17,18]. STG5 was identified via a combined mutli-omics and genetic mapping strategy and was found to regulate HKT family genes to contribute to ST [19]. The ST haplotype of STG5 is also predominantly enriched in xian/indica rice populations [18,19]. RST1, also referred to as OsARF18, negatively regulates ST in rice by directly repressing the transcription of OsAS1, a key gene in nitrogen metabolism [20,21]. Loss of RST1 function improves nitrogen utilization, reducing NH₄⁺ accumulation, maintaining a lower Na⁺/K⁺ ratio, and ultimately enhancing ST [20]. Two key single nucleotide polymorphism (SNP) sites with transversions from G to C and C to G, respectively, cause the G530A and E611G amino acid substitutions and have been reported to improve ST [18]. OsHKT1;1 and OsHKT2;3 were two HKT family genes discovered by GWAS, regulated by HST1, OsWRKY53, and STG5 [17,20,22,23,24]. OsSTL1 encodes a C2C2-type zinc finger protein on chromosome 4 and positively regulates ST in rice seedlings [25]. All these genes have not been investigated in specific cultivars in the 3H rice region in China, hindering the potential application of these ST genes for improving the ST of these varieties.

Natural variations cause different functional levels of gene alleles [26,27]. Some ST genes have been reported to be classified into haplotypes based on SNP differences, which provides us with a tool for functional evaluation and utilization of these genes [28,29]. Here, we collected nine SNPs distributed in seven STt associated genes reported to be associated with ST, including OsSTG5^12S, SKC1^140A, SKC1^184H, OsHKT1;1^94K, OsHKT2;3^I77T, OsWRKY53^173G, OsSTL1^289S, RST1^530A and RST1^E611G [13,17,18,19,21,23,24,25]. We investigated the distribution of these genes in a population of 215 geng/japonica rice varieties in the 3H rice region in China. These varieties could be divided into 14 haplotypes. Then based on haplotype comparison, we evaluated the functional levels of these SNPs in the geng/japonica rice population. The results showed that OsSTL1^289S, OsHKT1;1^94K, RST1^530A and SKC1^140A significantly contributed to improving ST, whereas OsWRKY53^173G and STG5^12S did not confer detectable benefits in this population. Among these SNPs, RST1^530A contributes the most to ST. Then we performed functional validation of RST1^530A compared with RST1^530G based on chromosome segment substitution lines (CSSLs), combined with transcription and biochemical analyses. We found that RST1^530A represses OsAS1 expression weaker than RST1^530G and genetically associates with ST. Subspecies cluster analysis shows that RST1^530A is mainly present in xian/indica rice varieties. A few geng/japonica rice varieties, such as Yanfeng47 and Lindao20, have acquired chromosome fragments carrying RST1^530A from the ancestors of xian/indica rice during domestication, thereby enhancing their ST. These varieties could be used as appropriate donors to contribute RST1^530A for geng/japonica rice ST improvement.

2. Materials and Methods

2.1. Rice Resources and Growth Conditions

The 215 Huang-Huai-Hai geng/japonica rice varieties were used in this study (Supplementary Table S1); among them, varieties 1-194 were from the collections as previously described [2], while varieties 195-215 were collected by Mr. Ruihua Zhang (Linyi Academy of Agricultural Science). Prior to the experiment, seeds were dried at 37 °C for 3 d and then soaked in double-distilled water (ddH₂O) for 3–4 d until germination. Germinated seeds were then sown in 96-well black plastic plant boxes containing ddH₂O for 2 d. Subsequently, the water was replaced with Yoshida rice nutrient salts (Coolaber, Beijing, China). The nutrient solution was renewed every 3 d to ensure a stable nutrient supply. Rice seedlings were grown in a phytotron within the Plant Growth Breeding System (PGBS, Wuhan Greenfafa Institute of Novel Genechip R&D Co., Ltd., China) greenhouse with 12 h of light at 30^◦C and 12 h of darkness at 26^◦C, and the humidity maintained at 70% [18].

2.2. SNP Genotyping Based on KASP and Cluster Analysis

Nine ST associated nucleotide polymorphisms (SNPs) including OsSTG5^12S, SKC1^140A, SKC1^184H, OsHKT1;1^94K, OsHKT2;3^I77T, OsWRKY53^173G, OsSTL1^289S, RST1^530A and RST1^E611G were used for genotyping analysis. Kompetitive Allele Specific PCR (KASP) serves as a robust tool for genotyping SNPs within gene sequences [2,30]. In this assay, each SNP was designed with a group of KASP primers listed in Supplementary Table S2. One genotyping sample was a 5 μL mixture consisting of 50–100 ng genomic DNA, 0.05 μM each forward primer, and 0.15 μM reverse primer, added with 2.5 μL 2×Master Mix for ASPCR V1 (HC Scientific, Chengdu, China). The PCR program was as follows: First, initial hot-start activation at 95 °C for 10 min; Then 10 touchdown cycles including 95 °C denaturation for 20 s and starting at 61 °C annealing/extension for 40 s, decreasing by 0.5 °C per cycle; Followed by 26 standard cycles, including 95 °C for 20 s and 56 °C for 40 s; Finally, fluorescence endpoint signals of FAM and HEX were scanned after amplification, and genotype clustering analysis was conducted using LGC SNP viewer software to distinguish homozygous reference alleles, homozygous variant alleles and heterozygous genotypes. Based on the genotyping of the nine SNPs, varieties with identical types at all SNPs were clustered into one haplotype. Then the genetic distance was calculated between different haplotypes based on the neighbor-joining method to create the phylogenetic tree [31].

2.3. Phenotype Assessment of ST

Phenotypic assessment of ST was performed as previously described [18]. In brief, two–weeks-old seedlings had the nutrient solution replaced with 150 mM NaCl to simulate a salt stress environment. After being treated and re-watered for certain days, the relative shoot fresh weight (rSFW) and dead leaf rate (DLR) were measured to characterize ST. The rSFW was the ratio of shoot fresh weight (SFW) measured under treatment to SFW in control. The DLR was the ratio of the number of dead leave to the total number of leaves.

Fv/Fm (the maximum quantum efficiency of PSII) was reported to be significantly positively correlated with ST in rice and is a key physiological indicator for evaluating ST [32,33]. Here we used the ratio of Fv/Fm under salt stress to Fv/Fm before treatment as relative Fv/Fm (rFv/Fm) to characterize the effect of salt stress on photosynthetic rate. The Fv/Fm was detected by the PathoScreen plant pathological phenotype measurement system (PhenoVation, Netherlands).

2.4. RNA Extraction and Quantitative Real-Time PCR

Total RNA was extracted from plant tissues using the RNApure Fast Plant Kit (CWBIO, Beijing, China) according to the manufacturer’s instructions. Briefly, tissue samples were ground in liquid nitrogen and lysed with RSL buffer. After gDNA removal, RNA was bound to the spin column, washed with RW1 and RW2 buffers, and eluted with RNase-free water. Reverse transcription was done for the synthesis of the first-strand cDNA using the HiFiScript gDNA Removal cDNA Synthesis Kit (CWBIO). Then, qRT-PCR was performed on the Archimed X4 Detection System (ROCGENE, Beijing, China) using the UltraSYBR Mixture kit (CWBIO). The amplification program was set as follows: initial denaturation at 95 °C for 3 min, followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s. Melting curve analysis was conducted from 65 °C to 95 °C to verify amplification specificity. Relative gene expression levels were calculated using the 2−∆∆Ct relative quantitative method with a reference gene for normalization [34]. The primers used for RST1, OsAS1, and OsACTIN are listed in Supplementary Table S3.

2.5. FBP-LUZ Reporter Assay for Transcription Repression Analysis

The fungal bioluminescence pathway (FBP) is an autonomous, caffeic acid-dependent luminescence system originating from fungi. The expression level of LUZ in the system determines the intensity of bioluminescence when caffeic acid is sufficient. FBP-LUZ was validated to function well in Nicotiana benthamiana leaves and has been used to evaluate the effect of transcription factors on target promoters [35]. Unlike conventional luciferase reporters, it enables continuous, substrate-free bioluminescence, making it a powerful tool for non-invasive monitoring of gene expression in planta [36]. We constructed the FBP-LUZ system in three parts. One is the reporter system, consisting of the target promoter linked with LUZ in the plasmid, pGreenⅡ 0800. The second is the transcription factor expression system driven by the Cauliflower Mosaic Virus 35S promoter (p35S) in the plasmid pCAMBIA1300, into which thecoding sequence of RST1^530G and RST1^530A were inserted, respectively. The final one is the caffeic acid synthesis system, called “CHN,” constructed with four genes (HispsS, H3H, CPH and NPGA) driven by p35S in the plasmid pCAMBIA1300. These plasmids were separately transformed into Agrobacterium tumefaciens strain GV3101 for transient expression. They were then co-infiltrated into 4-week-old N. benthamiana leaves for 48 h [37,38]. The salt induction method involved spraying a 100 mM NaCl solution evenly onto leaves 45 h after inoculation and treating them for 3h . Then, luminescence signals were detected by the NightShade LB985 plant imaging system (Berthold Technologies, Germany), which operates at a spectral range of 350–1050 nm. The luciferase deviation was quantified to evaluate transcriptional repression efficiency.

2.6. Statistical Method for Data Analysis

For comparisons between two independent groups, the independent-samples t-test was adopted, and two-tailed P values were calculated. For comparisons involving three or more groups, one-way analysis of variance (one-way ANOVA) was performed, and two-tailed P values were calculated. A statistically significant difference was defined as P < 0.05.

3. Results

3.1. Distribution of ST Genes in a Huang-Huai-Hai Geng/Japonica Rice Population.

To investigate the distribution of ST genes in the geng/japonica rice population, we selected seven cloned ST genes including STG5, SKC1, OsHKT1;1, OsHKT2;3, OsWRKY53, OsSTL1 and RST1, and investigated their functional single nucleotide polymorphism (SNP) loci in 215 geng/japonica rice varieties belonging to the 3H geng/japonica rice region in China. The functional SNP loci included OsSTG5^I12S, SKC1^P140A, SKC1^R184H, OsHKT1;1^L94K, OsHKT2;3^I77T, OsWRKY53^A173G, OsSTL1^P289S, RST1^G530A and RST1^E611G according to previous reports [18]. The number in the superscript at each site represents the position of the amino acid, and the amino acid typing before and after the number corresponds to the salt-sensitive and salt-tolerant types respectively. We used KASP to realize the typing of the above nine SNP loci. The genotyping maps of the nine SNP loci were shown in Figure 1A-I, which showed the availability of these primers, and the certain genotypes of each SNP in each variety were listed in Supplementary Table S1.

Then, we compared the typing of various SNP loci in different varieties and found that 71.2% (153/215) of geng/japonica rice varieties had the same typing at each locus (Figure 2). The 153 varieties were classified as the major haplotype group of Hap1, and only two out of nine SNP loci, including OsHKT2;3^77T and RST1^611G were ST type in the Hap1 varieties. The other haplotype groups (Haps) were listed from Hap2 to Hap14 in the order of the number of varieties included. The genetic distances between different haps were calculated and shown in Figure 2. Correspondingly, OsHKT2;3^77T and RST1^611G, were shared with up to 99.53% and 100%, respectively, in the investigated varieties. ST types at other SNP loci were much less than OsHKT2;3^77T and RST1^611G; listed from high to low order, they were OsSTL1^289S (14.42%, 31/215), STG5^12S (7.44%, 16/215), OsWRKY53^173G (6.98%, 15/215), SKC1^140A (3.72%, 8/215), RST1^530A (3.72%, 8/215), OsHKT1;1^94K (2.32%, 5/215), and no variety had the ST type, SKC1^184H.

3.2. Functional Assessment of Six ST Genes in the Geng/Japonica Rice Population

To evaluate the contribution of each SNP locus to ST, we compared ST levels between specific haplotype groups according to the method previously described [18]. Because most varieties belonged to Hap1, which contains three conserved SNPs, we selected several varieties of Hap1 as a control and compared ST between the control and varieties representing each of the other haplotypes (Hap2–4, 6, 7, and 10). As mentioned above, Hap1 had just two ST-type SNP loci, while Hap2 carried three ST type SNP loci, with an additional locus of OsSTL1^289S. Therefore, we compared ST between Hap2 and Hap1 to evaluate the function of OsSTL1^289S. Similarly, we compared Hap1 with Hap3, Hap4, Hap6, Hap7, and Hap10, respectively, to assess the ST function of OsWRKY53^173G, OsSTG5^12S, OsHKT1;1^94K, RST1^530A, and SKC1^140A. Selected varieties were treated with 150 mM NaCl for seven days, then re-watered for an additional five days. Then, relative shoot fresh weight (rSFW) was measured as an ST index. The results showed that the genotypes of OsSTL1^289S, OsHKT1;1^94K, RST1^530A, and SKC1^140A significantly contributed to ST improvement, with contribution rates of 33.3%, 12.5%, 57.1%, and 16.50%, respectively (Figure 3A,D–G), while OsWRKY53^173G and STG5^12S did not show significant contributions (Figure 3B,C,G).

3.3. RST1^530A Impaired the Repression of OsAS1 and Genetically Associated with ST

In the ST comparison experiment, we noticed that Hap7 showed the highest ST level compared with Hap1, indicating that RST1^530A plays a pivotal role in improving ST. The result was consistent with our previous assessment of a xian/indica rice population [18]. Previous studies have shown that RST1 is a negative regulator of ST, reducing rice ST by repressing the target gene, OsAS1 [20]. OsAS1 participates in the primary assimilation of NH⁴⁺ and catalyzes the synthesis of asparagine in roots. The loss of function of RST1 results in the upregulation of OsAS1, which enhances nitrogen utilization efficiency by promoting asparagine synthesis and mitigates excessive NH⁴⁺ accumulation, thereby improving plant ST [20,21]. We hypothesized that the variation of RST1^G530A caused a nonsynonymous mutation might affect the RST1-OsAS1 regulation axis and impact rice ST. To verify this hypothesis, three varieties of Hap7 (Yanfeng47, Lindao20, and Xuxiangjing6) and three varieties of Hap1 (Nipponbare, Jingjing818, and Jinyuan45) were used for salt-tolerant comparison. Simultaneously, we detected the expression of RST1 and OsAS1 of these varieties before and after salt stress in root tissue. The results showed Hap7 had higher rSFW and lower DLR than Hap1 (Figure 4A–C). qRT-PCR showed that both RST1 and OsAS1 were upregulated after 3 hours of treatment with 150 mM NaCl. The average fold changes of RST1 were 1.94 in Hap7 and 2.60 in Hap1, and the average fold changes of OsAS1 were 6.97 in Hap7 and 4.55 in Hap1 (Figure 4D,E). These results indicated that RST1^530A leads to slightly more impaired repression of OsAS1 expression than RST1^530G under salt stress, which confers stronger ST in Hap7 than Hap1.

3.4. An Introgression Line of RST1^530G Impaired the ST and Increased the Repression of Salt-induced OsAS1 Expression

To further validate the effect of RST1^530G/A on ST, forward genetic assessment was performed. Huanghuazhan (HHZ) was a conventional planting xian/indica rice variety carrying RST1^530A, while Jizi 2 (JZ2) was a conventional geng/japonica rice variety carrying RST1^530G. Using HHZ as the recurrent parent and JZ2 as the donor introgression parent, lines of HHZ genetic background with a chromosome segment substitution line (CSSL) of JZ2 ranging from Chr6_27105342 to Chr6_29615495 were generated and named HHZ^CSSL6-JZ2. We planted HHZ and HHZ^CSSL6-JZ2 lines in Yoshida solution and compared ST between them (Figure 5A,B). The results showed that HHZ^CSSL6-JZ2 had significantly lower rSFW and rFv/Fm, and a higher dead leaf rate (DLR) than HHZ (Figure 5C–E). Then we compared the transcription level of RST1 and its target gene OsAS1 in the roots of HHZ and HHZ^CSSL6-JZ2 respectively at 0 h, 3 h, and 24 h post-treatment with 150 mM NaCl. The results indicated that the expressions of RST1 and OsAS1 were both induced by salt stress as reported before (Figure 5F,G). The expression level of RST1 was equal in HHZ and HHZ^CSSL6-JZ2 before or after salt stress, while the expression of OsAS1 was significantly higher in HHZ than in HHZ^CSSL6-JZ2 at 3 h and 24 h post-salt treatment (Figure 5F,G). The phenotype and transcription level comparison verified that RST1^530A confers to ST and is impaired in repressing OsAS1 expression.

3.5. FBP-LUZ Based Assay Revealed that RST1^530A Impaired the Repression of Salt-Induced OsAS1 Expression

To directly reflect the influence of different Haps of RST1 on the promoter of OsAS1, we performed an FBP-LUZ (Fungal Bioluminescence Pathway-based LUZ reporter system) assay. The FBP-LUZ assay was recently developed for detecting transcription factor (TF) activity, which requires no exogenous substrates and possesses the advantages of high sensitivity, low background, and stable signals [36]. The genomic region spanning the 2 Kb sequence upstream of the start codon of OsAS1 was used as the promoter and inserted upstream of the reporter gene LUZ, designated as pAS1:LUZ. It was co-transiently expressed with RST1^530G or RST1^530A protein, respectively. The leaves of N. benthamiana were divided into four parts; each part was co-expressed using the CHN expression system. As demonstrated in Figure 6A,B, compared to pAS1:LUZ and a control of JAMYB+pAGO18:LUZ [36], there was no change for the co-expressed groups of RST1^530G+pAS1:LUZ and RST1^530A+pAS1:LUZ without salt treatment. Under stress, the luciferase activity of both transient groups, RST1^530G+pAS1:LUZ and RST1^530A+pAS1:LUZ, was significantly lower than that ofpAS1:LUZ, and that of RST1^530A+pAS1:LUZ was significantly higher than RST1^530G+pAS1:LUZ (Figure 6C,D). These results demonstrated that both Haps of RST1 repressed the promoter activity of OsAS1, and RST1^530A exhibited weaker inhibition than RST1^530G.

3.6. Evolutionary Analysis Reveals RST1^530A of Geng/Japonica Rice Varieties Originated from Specific Xian/Indica ones

Only three varieties carrying RST1^530A were identified in Hap7 (Figure 2), including Lindao20 and its parent line Yanfeng 47 (Figure 7A). We previously revealed that the proportion of RST1^530A in a xian/indica rice population reached 93.64% as reported [18], while the proportion of RST1^530A was only 3.7% in the 3H geng/japonica rice population (Figure 7B). Thus, we speculated that RST1^530A in a few geng/japonica rice varieties may originated from xian/indica rice varieties during the process of subspecies hybridization breeding. To verify this assumption, we investigated the sequence of the RST1 gene in Lindao20 and Yanfeng47. Lindao20 was a progeny selected from a cross of Lindao10 (♂) with Yanfeng47 (♀), which originated from the donor parent xian/indica rice variety Aijiaonante by the Yinan County Rice Research Institute in China (Figure 7A). We found that both Lindao20 and Yanfeng47 had six variations compared to Nipponbare in the RST1 coding region, including five SNPs at 565 bp, 1612 bp, 1743 bp (RST1^G530A), 1986 bp, 2102 bp, and one “CCG” INDEL at 744 bp (Figure 7C). While Lindao10 had an identical sequence of RST1 to Nipponbare (Figure 7C), this indicated that the RST1 gene in Lindao20 originated from Yanfeng47. Aijiaonante was an early xian/indica rice variety used as one of the donor lines on the inhierarchical diagram of Lindao20 (Figure 7B). We found that Aijiaonante had the same sequence of RST1 as Lindao20 (Figure 7C), suggesting that RST1^530A in Lindao20 originated from xian/indica Aijiaonante. Additionally, LOC_06g47100 and LOC_06g47220 are two flanking genes, which are located 30 Kb upstream and 38 Kb downstream of RST1, respectively. We cloned each gene from all four rice varieties (Lindao20, Yanfeng47, Aijiaonante, and Lindao10) and aligned them with the reference sequence of Nipponbare. Five SNPs for each gene were mined, and they are classified into two Haps, with Lindao20, Yanfeng47 and Aijiaonante grouped together while Lindao10 and Nipponbare grouped as the other Haps (Figure 7C,D). These results suggest that a chromosomal segment of over 68 Kb carrying the RST1^530A allele transferred from the xian/indica rice variety Aijiaonante into the geng/japonica rice varieties Lindao20 and Yanfeng47.

4. Discussion

4.1. Contrasting Frequencies of ST SNPs Between Xian/Indica and Geng/Japonica Populations

Previously, alleles associated with ST had been found largely differentiated between xian/indica and geng/japonica rice populations [39]. However, the specific ST genes were not well-documented. A striking finding emerging from the comparison between our previous xian/indica study [18] and the present geng/japonica survey is the dramatic difference in the frequencies of ST alleles. Among the nine investigated SNPs, seven showed a markedly lower frequency in the geng/japonica population (215 Huang-Huai-Hai varieties) than in the xian/indica population (550 accessions). Specifically, OsSTL1^289S is fixed in xian/indica (100%) but accounts for only 14.4% in geng/japonica; OsWRKY53^173G drops from 92.6% to 7.0%; RST1^530A declines from 93.6% to 3.7%; SKC1^140A from 79.3% to 3.7%; and SKC1^184H is completely absent in geng/japonica. Only two ST-alleles, OsHKT2;3^77T and RST1^611G, shared similarly high frequencies in both subspecies (88.6–99.5% and 93.8–100%, respectively), suggesting they may be ancient [40], functionally indispensable components of the ST machinery that were already fixed before subspecific divergence.

What might explain this pronounced subspecific divergence? Several hypotheses can be proposed. First, differential selection pressures during domestication and breeding likely played a major role. Xian/indica rice is predominantly cultivated in southern China and Southeast Asia, where hot, humid climates often coincide with naturally or anthropogenically salinized soils (e.g., coastal areas, irrigation-induced salinity) [41,42,43]. In such environments, alleles that confer even modest ST would be strongly favored. In contrast, the 3H region, the key geng/japonica production area in northern China, experiences lower temperatures and historically less acute soil salinity, and breeding efforts have focused primarily on yield and cold tolerance, with only indirect or weak selection for ST. Second, the reproductive barrier between xian/indica and geng/japonica (hybrid sterility) has restricted gene flow, preventing the spread of xian/indica-derived tolerant alleles into geng/japonica gene pools [43,44]. Third, some ST-alleles, such as SKC1^184H, may have originated relatively recently within the xian/indica lineage and have not yet had sufficient time or opportunity to introgress into geng/japonica [18,39]. Finally, the possibility of epistatic constraints cannot be ruled out: a ST-allele that is beneficial in a xian/indica background might be neutral or even detrimental in a geng/japonica background, thus being purged by selection [45]. Regardless of the exact causes, the observed rarity of most ST-alleles in geng/japonica strongly indicates that genetic improvement of ST in this subspecies will require targeted introgression of favorable alleles from xian/indica donors.

4.2. Subspecies-Specific Functional Comparisons of Three Representative SNPs Reveal the Critical Role of Genetic Background

In both xian/indica and geng/japonica populations, RST1^530A exhibits the strongest contribution to ST (confirmed by haplotype comparisons in xian/indica; 57.1% improvement in rSFW in geng/japonica) [18]. Notably, this conclusion directly corrects the earlier report that considered RST1^530G as tolerant and RST1^530A as sensitive [20]. Our xian/indica study already demonstrated through rigorous haplotype contrasts (Hap12 vs. Hap11) that RST1^530A is really the tolerant allele [18]. The present study further validates this finding in a geng/japonica background: Hap7 varieties carrying RST1^530A (e.g., Yanfeng47, Lindao20) display significantly higher rSFW and lower DLR than that of Hap1 varieties under 150 mM NaCl treatment. More importantly, using chromosome segment substitution lines and the FBP-LUZ transcriptional reporter system, we provide the first mechanistic evidence in geng/japonica: the RST1^530A protein exhibits weaker repression of its target gene OsAS1 expression compared to RST1^530G, resulting in higher up-regulation of OsAS1 under salt stress (Figure 5 and Figure 6). This likely promotes asparagine synthesis in nitrogen metabolism, alleviates excessive ammonium accumulation, and ultimately enhances ST [21]. Thus, RST1^530A is a functionally conserved, mechanistically well-defined core ST allele across subspecies, making it a priority target for cross-subspecies breeding.

SKC1^140A contributes a 23.8%–27.5% increase in rSFW in xian/indica [18] but only 16.5% in geng/japonica (Figure 3). Although significantly positive in both backgrounds, its effect is attenuated in geng/japonica. This may be attributed to the absence of the synergistic SKC1^184H allele in the geng/japonica gene pool. It is proposed that the two amino acid substitutions (P140A and R184H) lie on different sides of a cytoplasm-exposed loop of the SKC1 protein and may exert additive effects on transport activity or phosphorylation status [40]. In xian/indica, ~31.6% of accessions carry both alleles and exhibit higher ST; in geng/japonica, SKC1^184H is entirely missing, so SKC1^140A acts alone and consequently less effectively. This again illustrates that ST is a polygenic trait, and the contribution of a single locus is conditioned by the presence of interacting partners in the genetic background.

The most striking discrepancy between our xian/indica and geng/japonica studies concerns OsWRKY53^173G. In xian/indica, this allele confers a 38.6% improvement in rSFW [18] and the effect was shown to depend on the presence of SKC1^140A and/or SKC1^184H (Figure 5). In the present geng/japonica population, however, we did not detect a significant positive contribution of OsWRKY53^173G to ST; instead, a slight negative trend was observed. Why does this happen? It is reported that OsWRKY53 acts as a transcriptional repressor that directly binds the SKC1 promoter and suppresses its expression [17]. Yet our haplotype comparisons and expression analyses in natural populations do not support a simple repression model; rather, a positive correlation was observed. A plausible explanation is that in genetic backgrounds already carrying SKC1^140A/184H, OsWRKY53^173G may, via feedback loops or interactions with other transcription factors, ultimately enhance SKC1 expression or function. In geng/japonica backgrounds that lack these SKC1 tolerant alleles (frequency of SKC1^140A is only 3.7%, and SKC1^184H is zero), this regulatory circuit fails to operate, and OsWRKY53^173G might even impose a metabolic burden. Thus, the two studies together uncover an important epistatic network: the ST function of OsWRKY53^173G depends on the prior existence of SKC1 tolerant alleles. This finding provides a clear example of why a gene effective in xian/indica may not be directly transferable to geng/japonica and emphasizes that genetic background must be carefully considered in breeding.

4.3. Origin of RST1^530A: Introgression from Xian/Indica into Geng/Japonica

Pedigree analysis and SNP linkage markers demonstrate that the RST1^530A allele in the geng/japonica varieties Yanfeng47 and Lindao20 is not endogenous to geng/japonica but was introgressed from the xian/indica variety Aijiaonante via a chromosomal segment. Several lines of evidence support this conclusion: (1) Aijiaonante, Yanfeng47 and Lindao20 share an identical combination of SNPs in the RST1 coding region, including the key G→A transition at position +1743 bp (RST1^530A). (2) SNP markers located 30 Kb upstream (LOC_06g47100) and 38 Kb downstream (LOC_06g47220) of RST1 form the same haplotype in these three varieties, whereas the geng/japonica varieties Lindao10 and Nipponbare belong to a different haplotype, indicating that at least a 68 Kb chromosomal segment was introgressed. This finding has important breeding implications [46]. First, it demonstrates that xian/indica rice is a valuable donor reservoir for ST genes in geng/japonica. Second, Yanfeng47 and Lindao20, which have already acquired RST1^530A through conventional crossing, exhibit good ST and can serve as direct donor parents for transferring this allele into other elite geng/japonica varieties. Third, the KASP markers we developed can efficiently track the introgressed segment, facilitating marker-assisted selection.

5. Conclusions

In this study, we investigated the distribution of nine ST associated SNPs from seven cloned genes in 215 geng/japonica varieties from the Huang-Huai-Hai region. Compared with our previous work in xian/indica rice, the geng/japonica population exhibits markedly lower frequencies of most tolerant alleles, with the exceptions of OsHKT2;3^77T and RST1^611G. Functional assessment revealed that OsSTL1^289S, OsHKT1;1^94K, RST1^530A and SKC1^140A significantly improve ST in geng/japonica, among which RST1^530A contributes the most (57.1% increase in rSFW). Notably, OsWRKY53^173G and OsSTG5^12S do not confer detectable benefits in the geng/japonica background, likely due to the absence of interacting SKC1 tolerant alleles. Mechanistically, we demonstrated that RST1^530A enhances ST by more weakly repressing its target gene OsAS1 compared with RST1^530G. Furthermore, we traced the origin of RST1^530A in geng/japonica to a xian/indica donor (Aijiaonante) through a chromosomal segment of at least 68 Kb, as evidenced by pedigree and SNP linkage analyses. Collectively, this study provides a subspecies-specific evaluation of seven major ST genes, corrects the functional assignment of RST1^530A, and identifies both a key tolerant allele and its carrier varieties (Yanfeng47, Lindao20) as practical resources for marker-assisted breeding of ST geng/japonica rice.