A panel of 182 rice genetic resources was used for phenotypic and genotypic evaluation to identify genomic regions associated with PHS resistance. The set of genetic resources comprised 106 Japonica, 35 Indica, 33 Admixed, 6 Aus, and 2 Aromatic types. Of the 182 rice genetic resources, 116 were obtained from the National Institute of Crop Science, and the remaining were sourced from the National Agrobiodiversity Center.

The experiment was conducted in 2024 at the Experimental Farm, College of Agriculture and Life Sciences, Kyungpook National University. Seedlings were transplanted at a spacing of 30 × 15 cm, with one seedling per hill. Fertilizer was applied at rates of 9.0–4.5–5.7 kg/10a (N–P₂O₅–K₂O), following national crop fertilizer guidelines (National Institute of Agricultural Sciences, 2022).

PHS resistance was evaluated by recording the heading date of each rice genetic resource and harvesting the main panicle 40 days after heading, corresponding to an accumulated growing degree day value of 1,000 °C (Kang et al., 2018). To ensure phenotypic reliability, seed viability was assessed after the PHS evaluation by conducting an independent post-harvest germination test using harvested seeds. For each rice genetic resource, 30 seeds were placed in a Petri dish with three biological replicates and incubated at 25 °C for 7 days under standard germination conditions. All rice genetic resources used in this study showed germination rates above 70% in this test. Three biological replicates were included per genetic resource. Panicles were fully wrapped in tissue paper to facilitate moisture absorption and placed in stainless steel trays (325 × 265 × 63 mm). Samples were incubated in a growth chamber at 25 °C and 100% relative humidity for 7 days (Rural Development Administration, 2012). After incubation, the germination rate was calculated as the percentage of germinated seeds among the total number of filled seeds per panicle. The mean value of three replicates was used to determine the final PHS rate.

Seeds were considered germinated when the coleoptile visibly emerged from the hull, while unfilled or defective grains were excluded. Based on germination rates, PHS resistance was classified into five categories: degree 1 (≤ 20%), degree 3 (21–40%), degree 5 (40–60%), degree 7 (60–80%), and degree 9 (81% ≤) (Rural Development Administration, 2012). Table 1 presents the classification criteria. Rice genetic resources with degrees 1 or 3 were considered resistant, while those with degrees 5, 7, or 9 were considered susceptible.

Single-nucleotide polymorphism (SNP) genotyping was performed using the 580K Axiom Rice Genotyping Chip (580K_KNU chip), developed from eight genomic data sources (Kim et al., 2022). Genomic DNA samples were hybridized to the array and scanned on the GeneTitan^® platform, Affymetrix, Santa Clara, CA, USA. SNP calling was conducted with Genotyping Console v4.2, Affymetrix, Santa Clara, CA, USA, and further refined using the SNPolisher R package v3.0. SNPs were aligned to the IRGSP-1.0 (japonica), MH63RS2 (indica), and Oryza rufipogon reference genomes. High-quality SNP markers were selected for GWAS. They were filtered using the following criteria: minor allele frequency (MAF) > 0.05, missing rate < 0.02, heterozygosity rate < 0.05, removal of non-polymorphic SNPs, and sequencing depth > 10×. After filtering, 289,569 SNPs were retained for GWAS.

GWAS was performed using two multi-locus models—Bayesian information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK) and Multi-Locus Mixed Model (MLMM) implemented in the GAPIT package in R. Before association testing, population structure was assessed by principal component analysis (PCA), with the first three principal components included as covariates alongside a kinship matrix. MLMM iteratively incorporates significant markers as covariates, simultaneously detecting multiple loci contributing to phenotypic variation (Segura et al., 2012). BLINK filters redundant markers using linkage disequilibrium (LD) and applies the Bayesian Information Criterion (BIC) for model selection, enhancing statistical power while controlling false positives (Huang et al., 2019). For multiple testing correction, a Bonferroni adjustment at α = 0.05 was applied based on 289,569 SNPs, yielding a genome-wide significance threshold of p < 1.726 × 10^-7 (−log₁₀p = 6.76). GWAS results were visualized using Manhattan and quantile–quantile (QQ) plots generated with the “qqman” R package (Turner, 2018). Significant SNPs were annotated by assigning open reading frames (ORFs) within a ±150-kb window around each SNP as candidate genes.

Variation in PHS among rice genetic resources was evaluated using one-way analysis of variance (ANOVA) in R version 4.3.1 software. When ANOVA results were significant (p < 0.05), group comparisons were performed using Duncan’s multiple range test through the “agricolae” R package (de Mendiburu, 2023) to identify statistically significant differences among genetic resources.

Two-locus haplotype analysis of SDR4 and qPH7 was performed using PCR amplification with Solg™ e-Taq DNA Polymerase (SolGent, Daejeon, Korea) (Table 2). For the SDR4-SacII marker (Sugimoto et al., 2010), thermal cycling conditions included an initial denaturation at 94 °C for 5 min; 35 cycles of 94 °C for 20 s, 55 °C for 25 s, and 72 °C for 1 min; with a final extension at 72 °C for 5 min. For the PH_1_13 marker (Lee et al., 2023), the same conditions were used except for the annealing step at 54 °C for 45 s. PCR products for SDR4 and qPH7 were digested with the restriction enzymes SacII and DdeI, respectively. Digested products were separated on a 1.5% agarose gel and stained with SYBR™ Safe DNA Gel Stain (Thermo Fisher Scientific, Waltham, MA, USA). Bands were visualized using the DAVINCH Gel Imager CG-550 (DAVINCH-K, Seoul, Korea).

PHS was evaluated in 182 rice genetic resources using the predefined criteria (Table 1). PHS rates ranged from 0% to 95.7%, with an average of 20.8 ± 25.2% (Figure 1). The distribution was right-skewed (skewness = 1.36; kurtosis = 0.87). The median value was 8.5%, indicating that most rice genetic resources exhibited relatively low sprouting levels. The resistant control cultivar, Joun, showed an average PHS rate of 13.61 ± 4.23%, whereas the susceptible control, Jopyeong, exhibited a significantly higher rate of 44.23 ± 5.30%.

PHS rates were classified into five degrees according to the established evaluation criteria (Table 1; Figure 2). Among the 182 genetic resources, 20.9% were classified as susceptible (degree 5, 7, or 9), while the remaining 79.1% were classified as resistant (degree 1 or 3). The resistant and susceptible control cultivars corresponded to degree 1 and degree 5, respectively. To evaluate PHS variation among ecotypes within the population, the genetic resources were grouped into five ecotype categories and evaluated for PHS levels based on predefined criteria (Supplementary Figure S1). Japonica (group I) and Indica (group II) showed the widest PHS variation, with Indica showing a lower median value than that of Japonica. In contrast, Admixed (group III) predominantly exhibited low PHS rates.

Overall, 289,569 SNPs were analyzed across 182 resources using an SNP chip and next-generation sequencing (NGS) (Supplementary Table S1). SNP distribution varied across the 12 rice chromosomes, with an average of 24,131 SNPs per chromosome. Chromosome 1 showed the highest number (37,852), whereas chromosome 12 had the lowest (16,551). SNP density also differed among chromosomes, averaging 1.32 SNPs/Mb (Supplementary Figure S2). Chromosome 12 exhibited the highest density (1.66 SNPs/Mb), while chromosomes 2 and 3 showed the lowest (1.09 SNPs/Mb).

A phylogenetic analysis of 182 rice genetic resources was performed using genome-wide SNP data to assess genetic relationships and population structure (Supplementary Figures S3A, B). The neighbor-joining tree analysis revealed five distinct ecotype groups based on genetic similarity: (i) Japonica, (ii) Indica, (iii) Admixed, (iv) Aus, and (v) Aromatic (Supplementary Figure S3A). These groups reflect unique genetic backgrounds and ecological adaptations, with pronounced divergence between Japonica and Indica.

Genetic structure was further validated through PCA (Supplementary Figure S3B), which supported the same five-group clustering pattern. PC1 and PC2 explained 60.38% and 4.63% of the total genetic variation, respectively. The PCA results showed clear genetic differentiation among ecotypes, providing complementary evidence to the phylogenetic analysis. This grouping establishes the basis for interpreting phenotypic variation in traits such as PHS resistance.

GWAS were performed using genotypic and PHS rate data from 182 rice genetic resources to identify SNPs associated with PHS resistance. The GAPIT package in R was used to implement the BLINK and MLMM models. As illustrated in the Manhattan and quantile–quantile (QQ) plots (Figure 3), SNP associations were evaluated across the 12 rice chromosomes using a genome-wide significance threshold of −log₁₀(P) = 6.76, based on a Bonferroni correction using 289,569 SNPs. The QQ plots showed that the observed p-values closely followed the expected distribution, indicating minimal genomic inflation and adequate control of population structure. BLINK identified significant SNPs on chromosomes 7, 8, and 11 (Figure 3A), while MLMM detected lead SNPs on chromosomes 7 and 11 (Figure 3B). The SNPs on chromosomes 7 and 11 were detected at identical loci in both models, whereas the SNP on chromosome 8 was unique to BLINK. These lead SNPs on chromosomes 7, 8, and 11 were considered QTLs and designated as qRPH7, qRPH8, and qRPH11, respectively (Table 3).

qRPH7 showed strong statistical significance in both BLINK and MLMM, with –log₁₀(p) values of 8.01 and 9.04 and phenotypic variance explained (PVE) values of 43.69% and 80.00%, respectively. qRPH11 showed –log₁₀(p) values of 7.52 in the BLINK model and 6.72 in the MLMM model, the latter being slightly below but close to the predefined significance threshold (6.76). Despite this marginal significance, qRPH11 consistently explained a substantial proportion of phenotypic variance, with PVE values ranging from 13.07% to 33.89% across the two models. Based on its consistent detection and large effect size, qRPH11 was retained as a candidate QTL and selected for subsequent analyses. qRPH8 was detected exclusively by BLINK, with a –log₁₀(p) value of 8.07 and a PVE of 32.84%. Despite differences between the models, both BLINK and MLMM consistently identified three QTLs significantly associated with PHS resistance.

To assess the effects of the identified QTLs, the 182 rice genetic resources were classified into five groups (I-V) based on their QTL combinations, and PHS was confirmed (Figure 4). Group I included all three QTLs—qRPH7, qRPH8, and qRPH11. Groups II–IV each contained two QTLs: Group II possessed qRPH7 and qRPH11; Group III had qRPH7 and qRPH8; and Group IV included qRPH8 and qRPH11. Group V included genetic resources carrying only qRPH7. Group I showed the lowest mean PHS rate, and Groups I and II exhibited significantly lower PHS rates than those of other groups.

To identify candidate genes related to PHS resistance, ORFs located within ±150 kb of the three QTLs (qRPH7, qRPH8, and qRPH11) identified through GWAS were examined. The analysis prioritized genes annotated in the Rice Annotation Project database and functionally related to PHS, germination, seed dormancy, and ABA/GA signaling pathways (Supplementary Table S2). Within the qRPH7 interval, two major loci linked to PHS resistance—SDR4 and qPH7—were identified (Figure 5A). Additionally, the qRPH11 region harbored OsPK1, a gene implicated in hormonal regulation related to PHS resistance (Figure 5B). In contrast, no annotated ORFs with known roles in PHS, seed dormancy, or ABA/GA signaling were identified within the ±150 kb region flanking qRPH8. Thus, leaving the functional candidate gene(s) underlying this QTL unresolved.

Among the 182 rice genetic resources, 167 carrying qRPH7 were selected to analyze the genotypes of the candidate loci SDR4 and qPH7 (Supplementary Figure S4). Consequently, 53 genetic resources carried SDR4, 42 carried qPH7, and 39 possessed both genes. In contrast, 14 carried SDR4 alone, three carried only qPH7, and 111 carried neither locus.

Among the 167 rice genetic resources excluding group IV (Figure 4), those carrying qRPH7 were classified into four combination types (A–D) based on the presence or absence of SDR4 and qPH7, and their PHS rates were compared (Table 4). Type D, comprising 111 rice genetic resources lacking both SDR4 and qPH7, was selected for additional analysis to evaluate the effects of other QTLs. Type A (SDR4 + qPH7) exhibited the lowest mean PHS rate (2.0 ± 2.5%), followed by Type C (qPH7, 3.8 ± 5.5%), Type B (SDR4, 6.1 ± 5.6%), and Type D (neither locus), which had the highest rate at 24.8 ± 24.7%. The overall mean PHS rate was 17.5 ± 22.7%. According to PHS classification criteria, Types A, B, C, and the overall mean were categorized as degree 1 (≤ 20%), whereas Type D corresponded to degree 3 (21–40%).

To further dissect the genetic basis of PHS resistance, the 111 rice genetic resources classified as Type D in Table 4 (lacking both SDR4 and qPH7) were analyzed based on the presence or absence of qRPH8 and qRPH11 (Table 5). Based on the presence or absence of qRPH8 and qRPH11, these resources were further classified into four types (a–d), and their PHS rates were compared. Among these groups, type a (qRPH11 + qRPH8) showed a mean PHS rate of 20.6 ± 26.4%, type b (qRPH11) exhibited 18.3 ± 19.5%, type c (qRPH8) recorded a markedly higher rate of 67.8 ± 27.1%, and type d (none) showed 46.0 ± 26.9%. The overall average PHS rate was 24.8 ± 24.7%, corresponding to degree 3 (21–40%). Based on the PHS classification criteria, type b was categorized as degree 1, type a as degree 3, type d as degree 5, and type c as degree 7. Duncan’s multiple range test (p < 0.05) revealed significant differences in PHS rates among types: types c and d were grouped as “a”, while types a and b were grouped as “b”, indicating that groups carrying qRPH11 exhibited significantly lower PHS rates.

The effects of different QTL combinations on PHS resistance were evaluated (Figure 6). Strong resistance was observed in genetic resources carrying qRPH7, qRPH8, and qRPH11 simultaneously (groups 1–3), which showed low PHS rates of 2.1 ± 2.5%, 4.6 ± 8.4%, and 5.7 ± 6.4%, respectively. Similarly, genetic resources harboring qRPH7 and qRPH11 (groups 4–6) demonstrated very strong resistance, with PHS rates of 1.4 ± 2.7%, 6.7 ± 4.5%, and 0.2%, respectively. In contrast, group 7, which was comparable to groups 1–3 but lacked either SDR4 or qPH7, and group 8, which was comparable to groups 4–6 but lacked SDR4 or qPH7, exhibited moderate resistance, with PHS rates of 20.6 ± 26.4% and 18.3 ± 19.5%, respectively. However, both groups exhibited considerable phenotypic variation. By comparison, group 9 (qRPH7 and qRPH8) and group 10 (qRPH7 alone) exhibited high PHS rates of 67.8 ± 27.1% and 46.0 ± 26.9%, respectively. Similarly, group 11 (qRPH8 and qRPH11) showed a high PHS rate of 55.6 ± 31.6%. In group 12, the additional of qRPH7 along with qRPH8 and qRPH11 did not reduce the PHS rate, which remained high at 57.4 ± 23.8%.

Rice (Oryza sativa) is a critical global crop, but its productivity and quality are highly susceptible to environmental threats such as PHS. While previous studies identify loci associated with PHS resistance, most have focused on specific cultivars, leaving the broader molecular mechanisms unresolved.

In the GWAS analysis using the BLINK model, three putative QTLs—qRPH7, qRPH8, and qRPH11—were detected (Figure 3, Table 3). Among these, qRPH7 and qRPH11 were also identified by MLMM, with qRPH7 accounting for a high proportion of phenotypic variance 33.89~80.00% (Table 3). This overlap between models highlights the robustness of these loci. Conversely, qRPH8 was detected only by BLINK, and its inconsistent phenotypic association suggests a limited contribution to PHS resistance.

In rice genetic resources, three major QTLs—qRPH7, qRPH8, and qRPH11—that exhibit additive cumulative effects on PHS resistance (Figure 4). The lowest mean PHS incidence (4.5 ± 10.4%) was observed in Group I, which harbors all three QTLs, suggesting that the combination of these loci is highly effective in enhancing resistance. In contrast, Group II (qRPH7 and qRPH11) demonstrated a lower PHS incidence (16.1 ± 18.6%) and was classified as resistant; however, the wide phenotypic variance observed in this group suggests potential influence from genetic background or environmental factors. The effect of a single QTL was observed only in Group V, which contains qRPH7 alone. Since no genetic resource lines individually carried qRPH8 or qRPH11, their single effects could not be evaluated. While qRPH7 exhibited the highest PVE, its solitary presence in Group V did not confer significant resistance.

Although qRPH7 and qRPH11 were consistently detected across GWAS models, the wide phenotypic variation among qRPH7 carriers indicates strong inter-locus interactions. qRPH11 acts additively to enhance qRPH7-mediated resistance, whereas qRPH8 functions as a context-dependent antagonist that modulates resistance expression.

To identify candidate genes influencing PHS, ORFs within the three significant QTLs were functionally annotated based on a comprehensive review. The analysis focused on ORFs annotated in the Rice Annotation Project database (https://rapdb.dna.affrc.go.jp/) that are associated with PHS, general germination, seed dormancy, and ABA/GA signaling pathways (Supplementary Table S2). Within the genomic region surrounding qRPH7 (± 150 kb), the presence of two major loci associated with PHS resistance—SDR4 and qPH7—were confirmed (Figure 5). Among the 167 rice genetic resources carrying qRPH7, 56 (33.5%) possessed at least one of these two major loci, SDR4 or qPH7 (Supplementary Figure S4). Within this group, 14 carried only SDR4, 3 carried only qPH7, and 39 possessed both loci.

Within the fine-mapped qPH7 region, Os07g0584366 exhibited nearly ninefold higher expression in the PHS-resistant donor Wandoaengmi6 compared to susceptible japonica cultivars, supporting its association with qPH7-linked resistance (Lee et al., 2023). In addition, the seed dormancy regulator Sdr4 has been shown to contribute substantially to natural variation in dormancy through both haplotype variation and differences in expression level. The Kasalath allele (Sdr4-k) confers deeper dormancy and stronger resistance to PHS, whereas the Nipponbare allele (Sdr4-n) is associated with reduced dormancy and increased susceptibility (Sugimoto et al., 2010). Previous studies have further reported that higher expression of Sdr4 is associated with suppressed germination and enhanced seed dormancy, indicating that Sdr4 function depends not only on allelic variation but also on transcriptional regulation (Chen et al., 2021). Moreover, Sdr4 acts downstream of the ABA-responsive transcription factor OsVP1, functioning as part of a regulatory network that integrates ABA signaling during seed maturation. Together, these results show that candidate genes located within the qRPH7 region are functionally associated with seed dormancy and ABA-mediated signaling pathways.

qRPH7 is located within a genomic region where the previously reported PHS-resistance QTL and genes, qPH7 and SDR4, are positioned in proximity within the ±150 kb interval. However, in the present study, PHS resistance was observed in rice genetic resources carrying qRPH7 despite the absence of both SDR4 and qPH7. This finding suggests that qRPH7 may contribute to PHS resistance independently of these previously characterized genes and further raises the strong possibility that additional PHS-associated candidate genes are present within this region. Therefore, the PHS resistance associated with qRPH7 cannot be sufficiently explained solely by the effects of SDR4 or qPH7. Future studies will aim to identify novel PHS-related candidate genes within the qRPH7 interval through gene expression profiling and transcriptome analyses. In addition, the genetic effects and potential interaction mechanisms among SDR4, qPH7, and qRPH7 in regulating PHS resistance will be further investigated.

Within the ±150 kb genomic region surrounding qRPH11, 52 ORFs were identified, among which OsPK1—a gene known to regulate the balance between ABA and GA—was the only locus associated with PHS or seed germination (Figure 5B). In rice, OsPK1 is a metabolism-related gene that contributes to growth regulation and environmental adaptation (Zhang et al., 2012). By modulating the ABA/GA balance, OsPK1 integrates hormonal signaling and functions as a molecular link between stress responses and growth suppression. Consistent with this role, knockout mutants of OsPK1 (ospk1) were reported to accumulate higher levels of abscisic acid (ABA) while suppressing gibberellin (GA) biosynthesis, resulting in an altered ABA/GA balance, enhanced dormancy, and increased oxidative stress. In contrast, no annotated ORFs with known functions associated with PHS, seed dormancy, or ABA/GA signaling were identified within the ±150 kb region flanking qRPH8.

Among the 167 genetic resources carrying qRPH7, four QTL combination types were classified based on the presence of SDR4, qPH7, or both, with some combinations significantly associated with reduced PHS rates (Table 4). Genetic resources in types A, B, and C—each carrying either SDR4, qPH7, or both—consistently exhibited low PHS rates, indicating strong resistance. In contrast, type D, which possesses qRPH7 but lacks SDR4 and qPH7, demonstrated a markedly broader distribution and higher mean PHS rates. This divergence suggests that qRPH7 alone is insufficient to confer stable resistance and highlights the significant individual and combined contributions of SDR4 and qPH7 in enhancing seed dormancy and suppressing PHS.

In the Type D subtype, the roles of qRPH8 and qRPH11 were investigated to further dissect the genetic architecture underlying PHS resistance in the absence of SDR4 and qPH7 (Table 5). Among the four genotypic combinations evaluated (types a–d), types a and b, carrying qRPH11, consistently exhibited lower PHS rates compared to those that lack this locus (types c and d). Type b, carrying qRPH11 alone, exhibited the lowest PHS rate (18.3 ± 19.5%) and was classified as degree 1 resistance, suggesting that qRPH11 positively contributes to resistance, either independently or in combination. Type c, which carries only qRPH8, exhibited the highest PHS rate (67.8 ± 27.1%), suggesting that in certain genetic backgrounds, qRPH8 may function as an epistatic gene, suppressing the effects of other resistance loci or even promoting susceptibility. qRPH8 exhibits epistatic interactions that antagonize, rather than enhance, PHS resistance, complicating its functional interpretation given its variable phenotypic expression. Overall, qRPH8 is inferred to act as a negative regulator of other PHS resistance–associated QTLs, thereby increasing the PHS rate.

Duncan’s multiple range test results further support the significant contribution of qRPH11 to PHS resistance, with types a and b forming distinct statistical groups compared to those of types c and d. These findings suggest an additive effect of qRPH11, while the apparent lack of beneficial effect from qRPH8 raises concerns about its utility in breeding programs and warrants further functional characterization.

The indica ecotype BALA displayed a PHS rate of 18.0% despite the absence of the qRPH7 locus, suggesting that alternative genetic factors contribute to PHS resistance independently of qRPH7. Developing a segregating population from this genetic resource would facilitate further investigation of the underlying mechanisms. Figure 4 illustrates two genetic resources in Group I that appear as outliers, exhibiting high PHS rates of 49.2% and 49.8%, respectively. These genetic resources lacked SDR4 and qPH7, despite carrying qRPH7 and qRPH11, suggesting that the absence of SDR4 and qPH7 may have a greater effect on PHS susceptibility than the presence of qRPH7 and qRPH11. Therefore, functional analysis incorporating SDR4 and qPH7 will be essential to elucidate the genetic interactions among these loci.

Overall, these results highlight how qRPH7, qRPH8, and qRPH11 interact synergistically and antagonistically in modulating PHS resistance. To further contextualize these findings within the broader genetic framework, all 12 possible QTL–loci combinations were analyzed (Figure 6), which revealed the complex genetic architecture underlying PHS resistance. Based on these observations, multi-locus combinations were then examined to determine how additive and epistatic interactions collectively shape PHS resistance.

PHS resistance was evaluated using 12 allelic combinations derived from three QTLs (qRPH7, qRPH8, and qRPH11) and two loci (SDR4, qPH7) (Figure 6). The results indicate that both the additive effects of individual QTLs and their genetic interactions are essential for determining PHS resistance. Strong resistance was observed in groups harboring all three QTLs (Groups 1–3) and in those carrying qRPH7 together with qRPH11 (Groups 4–6). In contrast, Groups 7 and 8, which had the same QTL combinations as Groups 1–3 and 4–6, respectively, but lacked SDR4 and qPH7, exhibited lower average resistance and greater variation, indicating that the effect of qRPH7 depends on the presence of SDR4 and qPH7. A comparison between Groups 8 and 10 further supports this finding: Group 10 (qRPH7 alone) exhibited high susceptibility, while Group 8 (qRPH7 + qRPH11 without SDR4 and qPH7) showed overall resistance, suggesting that qRPH11 acts additively to enhance the effect of qRPH7.

More specifically, Group 7 comprised three highly resistant genetic resources (≤ 20% PHS) and two susceptible (>40%), while Group 8 included 83 genetic resources, of which 55 were highly resistant (≤ 20%), 17 moderately resistant (≤ 40%), and 11 susceptible. These findings suggest that in the absence of SDR4 and qPH7, qRPH11, in combination with qRPH7, contributes to resistance. However, some genetic resources remain susceptible, indicating that minor QTLs or background genetic variation may also influence PHS. The most notable finding was the antagonistic epistasis of qRPH8. Among combinations lacking SDR4 and qPH7, groups carrying qRPH7 with qRPH8 or qRPH8 with qRPH11 (Groups 9, 11, 12) exhibited high PHS rates. In contrast, resistance was observed when all three QTLs (qRPH7, qRPH8, qRPH11) were present (Groups 1–3, 7). These findings suggest that qRPH8 suppresses the effect of qRPH7 or qRPH11 when present individually, leading to susceptibility, but when all three QTLs are combined, this antagonistic effect is neutralized. Thus, qRPH8 may function as an antagonistic regulator, modulating the effects of other major QTLs rather than acting only as a minor contributor.

Collectively, these findings indicate that PHS resistance is regulated by complex interactions among qRPH7–SDR4–qPH7, qRPH11, and qRPH8, rather than by a single major locus. Specifically, qRPH7 functions in an SDR4- and qPH7-dependent manner and is further enhanced by qRPH11, while qRPH8 exerts antagonistic epistasis by suppressing or modifying the effects of the other loci. These findings highlight that PHS is a typical polygenic trait, governed by additive effects and complex interactions among multiple loci.

Our results show a complex genetic interplay among multiple loci contributing to PHS resistance in rice. The consistent effects of SDR4, qPH7, and qRPH11 suggest that combining these loci through marker-assisted selection could substantially enhance resistance. In contrast, the effects of qRPH8 are inconsistent or adverse, highlighting the need for careful interpretation of its role. Future studies should investigate potential epistatic interactions among these loci and account for environmental influences that may affect the phenotypic expression of resistance.

In breeding, these findings provide a practical framework for improving PHS resistance in rice. We propose a targeted pyramiding strategy incorporating SDR4, qPH7, and qRPH11 to develop rice cultivars with enhanced PHS resistance. The inclusion of qRPH8 in breeding programs should be carefully considered, as its phenotypic effects are inconsistent. Using these validated loci in marker-assisted selection may accelerate the development of resilient varieties, particularly under humid and warm conditions that increase PHS risk. Moreover, integrating genotype-by-environment interaction analyses will be essential to ensure stable resistance across diverse cultivation settings.