Introduction
Rice is a very important grain crop to feed the world; however, its productivity is significantly affected by the adverse effects of excess water at different stages of growth and development. Rice cultivation in different parts of the world faces this challenge due to excess rainfall or flush flooding, particularly in areas close to rivers where rising water levels contribute to flooding (Mwakyusa et al., Reference Mwakyusa, Dixit, Herzog, Heredia, Madege and Kilasi2023). In many parts of the African continent, excess rainfall due to the impending effects of climate change has been observed, especially during the wet season cultivation August to October over the past decade (Niang et al., Reference Niang, Ruppel, Abdrabo, Essel, Lennard, Padgham, Urquhart, Barros, Field, Dokken, Mastrandrea, Mach, Bilir and White2014). Flooding in rice ecosystems can occur in multiple forms, including flash flooding during germination, complete submergence at the seedling stage, deepwater flooding and stagnant flooding (SF) (Kato et al., Reference Kato, Collard, Septiningsih and Ismail2014). Among these, SF presents a distinct challenge characterized by sustained water levels of 25–50 cm over an extended period, often throughout the crop cycle. This stress is prevalent in rainfed and irrigated (IR) lowland ecologies, especially in flood-prone regions of sub-Saharan Africa and South and Southeast Asia (Agbeleye et al., Reference Agbeleye, Olubiyi, Ehirim, Shittu, Jolayemi, Adetimirin, Ariyo, Sanni and Venuprasad2019).
While stem elongation is required for plant survival under SF, rice varieties adapted to deep water systems performed poorly under SF due to high levels of lodging (Kato et al., Reference Kato, Collard, Septiningsih and Ismail2014; Vergara et al., Reference Vergara, Nugraha, Esguerra, Mackill and Ismail2014). Submergence-tolerant SUB1A-1 allele-harbouring rice varieties also do not perform well under SF that lasts until the physiological maturity. This is primarily because these varieties lack the ability to elongate their stems and are forced to die under SF. In addition, the high-yielding rice varieties suitable for rainfed and IR lowland ecology performed poorly under SF, as water above 20–25 cm depths results in a growth penalty in rice. SF mainly leads to a reduction in the number of tillers, lodging, reduced fertility, small size of panicles and ultimately to yield loss due to the combined effects of all these factors (Singh et al., Reference Singh, Mackill and Ismail2011, Reference Singh, Carandang, Gonzaga, Collard, Ismail and Septiningsih2017; Kato et al., Reference Kato, Collard, Septiningsih and Ismail2014).
In contrast, Oryza glaberrima, the African cultivated rice species, presents a promising alternative due to its long evolutionary history of adaptation to harsh and variable environments with minimal human intervention (Agnoun et al., Reference Agnoun, Biaou, Sié, Vodouhè and Ahanchede2012; Agbeleye et al., Reference Agbeleye, Olubiyi, Ehirim, Shittu, Jolayemi, Adetimirin, Ariyo, Sanni and Venuprasad2019). This species has developed adaptive and protective mechanisms to withstand multiple biotic and abiotic stresses and displays notable plasticity for both deepwater and submergence conditions (Luo et al., Reference Luo, Nakazawa, Sasayama, Fukayama, Hatanaka and Azuma2020). In particular, O. glaberrima is highly adapted to deepwater flooding in the wetlands of West Africa, where it survives through rapid shoot elongation and extension of submerged leaves (Agnoun et al., Reference Agnoun, Biaou, Sié, Vodouhè and Ahanchede2012; Sakagami et al., Reference Sakagami, Joho and Ito2009; Sakagami, Reference Sakagami2012). These traits position O. glaberrima as a valuable genetic resource for enhancing stress resilience in O. sativa, especially in flood-prone environments across sub-Saharan Africa. Additionally, given its robust stress resilience, O. glaberrima warrants greater attention as a potential donor, which holds substantial promise for future rice improvement programmes targeting SF stress (Sarla and Swamy, Reference Sarla and Swamy2005; Ndjiondjop et al., Reference Ndjiondjop, Wambugu, Sangare, Dro, Kpeki and Gnikoua2018).
In our past work, we systematically characterized the entire set of known O. glaberrima accessions and landraces tolerant to SF and identified GERVEX 2674 as one of the highly tolerant lines (Agbeleye et al., Reference Agbeleye, Olubiyi, Ehirim, Shittu, Jolayemi, Adetimirin, Ariyo, Sanni and Venuprasad2019). GERVEX 2674 exhibits superior tolerance to SF, primarily attributed to its enhanced shoot elongation capacity, which enables it to maintain contact with the air above the water surface under prolonged inundation, contributing to better survival and yield stability. These adaptive traits make it a valuable donor parent to develop pre-breeding populations by crossing with O. sativa elite breeding lines. Using O. glaberrima donors directly in mainstream breeding poses several challenges such as shattering and sterility (Cubry et al., Reference Cubry, Tranchant-Dubreuil, Thuillet, Monat, Ndjiondjop, Labadie, Cruaud, Engelen, Scarcelli, Rhoné and Burgarella2018; Wu et al., Reference Wu, He and Wang2023). Therefore, it is important to do pre-breeding to introgress beneficial alleles while simultaneously fixing the undesirable traits before introducing interspecific lines as parents to elite breeding pools. To this end, we developed three interspecific pre-breeding populations by crossing GERVEX 2674 with O. sativa elite lines. The resulting BC2F6 populations were advanced using rapid generation advancement (RGA) through the single seed descent (SSD) technique. In the present study, we evaluated three early-generation interspecific (O. glaberrima/O. sativa) pre-breeding populations: GERVEX 2674/ART27-79-1-3-B-B-B-2, GERVEX 2674/ART28-126-3-2-2-7 and GERVEX 2674/FAROX508-3-10-F44-2-1, consisting of 87, 138 and 167 BC2F6 progenies to study their response to SF stress and a comparative IR control treatment over three cropping seasons. Our objectives were to assess the phenotypic variation introduced by landrace alleles under SF and IR conditions, identify genomic regions associated with adaptive traits and explore the potential utilization of novel genetic resources from these pre-breeding populations for developing resilient rice varieties for flood-prone ecosystems.
Materials and methods
Development of the populations
The entire set of O. glaberrima accessions available from the AfricaRice gene bank, comprising approximately 2200, was systematically characterized for SF tolerance and the tolerant donor accessions were identified (Agbeleye et al., Reference Agbeleye, Olubiyi, Ehirim, Shittu, Jolayemi, Adetimirin, Ariyo, Sanni and Venuprasad2019). One of the tolerant accessions, GERVEX 2674, was used to develop three interspecific pre-breeding populations with O. sativa. The three populations were GERVEX 2674/ART27-79-1-3-B-B-B-2, GERVEX 2674/ART28-126-3-2-2-7 and GERVEX 2674/FAROX508-3-10-F44-2-1, consisting of 87, 138 and 167 BC2F1 progenies, respectively. The three populations (392 progenies) were advanced up to the BC2F6 generation using the SSD approach under RGA across the 2014–2016 cropping seasons (Supplementary Fig. S1). The advanced BC2F6 and BC2F7 populations were utilized in the present study.
Description of experimental site
The experiments were carried out at the Africa Rice Center (AfricaRice) research station at the International Institute of Tropical Agriculture (IITA), Ibadan, Nigeria (latitude 3°54′32″E and longitude 7°29′15″N). SF trials and control trials (IR) were conducted in specially designed deep (∼2 m) tanks and lowland fields, respectively.
Phenotyping of the populations
Seeds of the parents, 392 BC2F6-7 progenies and standard local as well as interracially known tolerant and susceptible checks, were used in this study. To break seed dormancy, all seeds were subjected to heat treatment at 50°C for 2 days prior sowing (Fig. 1). The progenies along with parents and known local and international tolerant and susceptible checks were evaluated using an alpha lattice design with two replicates. The populations were phenotyped for three seasons in 2017 wet season (2017 WS, July–November 2017), 2018 dry season (2018 DS, January–May 2018) and 2018 wet season (2018 WS, July–November 2018). In all the trials, seeds were initially raised in a nursery, and 21-day-old seedlings were transplanted to puddled, well-levelled plots. Each plot consisted of two rows, 2 m in length, with a spacing of 20 cm between plants and rows. Nutrient management included a basal application of NPK fertilizer (15-15-15) at 200 kg/ha, applied a day after transplanting. Urea was used as topdressing at two stages: 30 kg/ha at tillering and another 30 kg/ha at panicle initiation. Weed control involved the application of herbicide in the initial stages, followed by manual weeding as necessary.
Figure 1. Schematic representation of the experimental protocol for phenotyping rice genotypes under stagnant flooding conditions: (A) control condition and (B) stagnant flooding condition.
For the IR (control) condition, a water depth of about 5 cm was maintained throughout the crop cycle. In the SF condition, water was initially maintained at 5 cm above the soil surface for 14 days post-transplanting. Flood levels were then raised to 15 cm and increased incrementally by 5 cm every alternate day until reaching 50 cm. This water level was maintained until physiological maturity, after which the plots were drained to a 2–5 cm water level just before harvest. The SF screening protocol closely followed the method described by Singh et al. (Reference Singh, Mackill and Ismail2011, Reference Singh, Carandang, Gonzaga, Collard, Ismail and Septiningsih2017) at IRRI. The protocol used for SF stress treatment is also summarized in Fig. 1.
Data were collected for several phenological and agronomic traits under both conditions, including days to 50% flowering, plant height, number of tillers per plant, number of panicles per plant and grain yield (GY). Days to 50% flowering were recorded when approximately 50% of the plants in a plot had exerted panicles. Plant height was measured at maturity from the ground to the tip of the panicle on five randomly selected plants per plot. The numbers of tillers and panicles were also counted from five representative plants and averaged. GY was determined by harvesting, threshing and cleaning panicles from each plot, followed by drying to a moisture content of about 14% before weighing.
Statistical analysis of agronomic data
All collected data were analysed using Breeding View within Breeding Management System (BMS) v 3.0.8 (Integrated Breeding Platform (IBP), 2015). A mixed linear model was employed, treating replications and blocks nested within replications as random effects, while genotypes were considered fixed. Significant differences among genotypic means were determined using the least significant difference (LSD) test. Broad-sense heritability estimates were also generated using breeding view. The trait values presented in the tables represent the combined best linear unbiased predictors (BLUPs) across three seasons under both IR and SF conditions.
Genotyping of the populations and quantitative trait locus analysis
Leaf samples from 392 BC2F6 progenies of three populations and their parents were collected, lyophilized and shipped to the Diversity Arrays Technology (DArT), Australia, for genotyping. Phenotypic data collected over three seasons (WS 2017, DS 2018 and WS 2018) were used for quantitative trait locus (QTL) analysis. A linkage map was constructed in QTL ICIMapping v4.2 (Meng et al., Reference Meng, Li, Zhang and Wang2015), retaining DArTseq markers mapped to the 12 rice chromosomes for the 392 recombinant inbred populations derived from O. glaberrima (GERVEX 2674) and O. sativa (ART27-79-1-3-B-B-B-2, ART28-126-3-2-2-7 and FAROX508-3-10-F44-2-1). The genomic data were filtered in Microsoft Excel and converted into ICIMapping-compatible format SNP and BIN functionality (Wang et al., Reference Wang, Li, Zhang and Meng2016). Furthermore, markers >10% missing data, segregation distortion (p < 0.05), or redundancy were excluded (Wang et al., Reference Wang, Li, Zhang and Meng2016). QTL mapping was conducted using the ICIM-ADD method (step = 1.0 cM; PIN = 0.01) with BLUPs of phenotypic traits in the BIP functionality. Logarithm of odds (LOD) threshold for QTL detection was determined by 1000 permutation tests at a significance level of p < 0.05.
Results
Performance of mapping populations and parents under IR vs. SF stress
Under SF conditions, GERVEX 2674 and the O. sativa parents flowered 5–10 days earlier than under IR conditions, with the exception of FAROX508-3-10-F44-2-1, which flowered at the same time under both conditions (Table 1). Plant height increased moderately (20–25%) in all parents under SF compared to IR, suggesting a stress-induced elongation response. However, the number of productive tillers declined significantly under SF conditions in all parents compared to IR, with GERVEX 2674, the SF-tolerant donor parent maintaining a relatively higher number of tillers under SF compared to the O. sativa parents (Table 1).
Table 1. Best linear unbiased predictors (BLUPs) of days to flowering, plant height and number of tillers per plant of GERVEX 2674/O. sativa interspecific pre-breeding populations, parents and checks under SF and control conditions across three seasons in the WS of 2017, DS of 2018 and WS of 2018
WS – wet season; DS – dry season.
The number of panicles and GY declined significantly under SF compared to IR (p < 0.05, Table 2). Among all parents, GERVEX 2674 recorded the highest GY under SF, exhibiting a 29% reduction in GY under SF compared to IR (Table 2). The Sub-1-introgressed checks used in this study, IRRI 119 (internationally known tolerant check) and FARO 66 (local check), both exhibited a decline in days to flowering, plant height, number of tillers, number of panicles and GY under SF compared to IR. Specifically, both IRRI 119 and FARO 66 exhibited 33% and 43% decline in GY under SF compared to IR, respectively (Tables 1 and 2).
Table 2. Best linear unbiased predictors (BLUPs) of grain yield and number of panicles per plant of GERVEX 2674/O. sativa interspecific pre-breeding populations, parents and checks under SF and control conditions across three seasons in the WS of 2017, DS of 2018 and WS of 2018
WS – wet season; DS – dry season.
The progenies of three interspecific pre-breeding populations exhibited wide variation and transgressive segregation for the phenological, morphological and yield-related traits under both IR and SF conditions (Tables 1 and 2). The range of decline in GY in the progenies varied (17–95%) under SF compared to IR. The relative decline in GY under SF was 17–23% in the best performing progenies, 29–35% in parents and about 33–43% in the checks. Some of the progenies that performed consistently better than the parents and checks in terms of GY are given in Tables 1 and 2. The better performing progenies exhibited no significant change in days to flowering, along with moderate elongation, high tillering ability, higher numbers of panicles and less decline in GY under SF compared to the low performing progenies, parents and checks (Tables 1 and 2).
Correlations among the traits under SF stress
A highly significant (p < 0.01) positive correlation was found between GY and two key morphological traits: number of tillers per plant and number of panicles per plant under SF (Table 3). The correlation coefficient between GY and number of tillers was r = 0.60, indicating a strong association, while the correlation between GY and number of panicles was r = 0.51, suggesting a moderate but meaningful relationship. Additionally, the correlation between the number of tillers and the number of panicles was also highly significant (p < 0.01), reflecting the interdependence of the two traits. In contrast, no significant correlation was found between GY and the other traits under SF (Table 3).
Table 3. Correlation coefficients of different traits under stagnant flooding stress in GERVEX 2674/O. sativa interspecific pre-breeding populations. FLW, days to 50% flowering; HT, plant height; TILL, number of tillers per hill; PAN. number of panicles per hill; GY, grain yield.
QTL identification
A genetic linkage map was constructed using 2110 SNPs mapped on the 12 rice chromosomes (Supplementary Fig. S2; Supplementary Table S1). QTL analysis using the Inclusive Composite Interval Mapping (ICIM) method detected 20 QTLs across nine chromosomes (1, 2, 3, 4, 5, 6, 10, 11 and 12), with LOD scores ranging from 3.4 to 12.3 (Supplementary Tables S2 and S3; Supplementary Figs S3–S11). Both parents (tolerant GERVEX 2674 O. glabberima and susceptible O. sativa) contributed positively and negatively to the observed effects. Out of the 20 QTLs, 16 showed positive additive effects for traits including days to flowering, plant height, number of tillers, number of panicles, GY and percentage survival. Four QTL exhibited negative effects on plant height and GY across control and SF environments. Under IR conditions, QTLs for days to 50% flowering (qdtf-1, qdtf-2 and qdtf-11) were identified on chromosomes 1, 2 and 11, collectively explaining 16% of observed phenotypic variation. QTLs for plant height (qpht-12), number of tillers (qtill-1.1, qtill-1.2 and qtill-5) and GY (qgy-3 and qgy-4) were detected. Notably, co-localized QTLs on chromosomes 1 and 5 influenced both number of tillers and number of panicles, accounting for 12% of the variation (Supplementary Table S2). Under SF, QTL for days to 50% flowering (qdtf-3 and qdtf-12), plant height (qpht-1, qpht-2 and qpht-10) and GY (qgy-1, qgy-6.1 and qgy-6.2) were identified. The strongest SF-related QTL, qdtf-12, explained 10.5% of observed phenotypic variance. Additionally, a single QTL (qsurv-12) was identified for survival, contributing 3.1% of the variation (Supplementary Table S3).
Discussion
This study presents the first detailed characterization of interspecific pre-breeding rice populations developed by introgressing genetic diversity from the African landrace O. glaberrima (GERVEX 2674) into O. sativa backgrounds to enhance tolerance to SF. Across three environments, we demonstrated that specific progenies derived from GERVEX 2674 exhibited superior adaptation to SF stress, combining moderate stem elongation, higher tiller retention and reduced GY loss compared to SUB1A-1 introgressed checks and susceptible parents. Unlike prior studies that focused on deepwater or submergence stress, this work addresses the relatively underexplored but agriculturally significant SF stress. The identification of 20 minor-effect QTLs associated with key adaptive traits, many of which are novel and condition-specific, adds to the limited existing knowledge on the genetic basis of SF tolerance. In the present study, the SUB1A-1-harbouring checks IRRI 119 and FARO 66 showed moderate performance under SF, exhibiting about 33% and 43% decline in GY under SF, respectively, compared to control. Many SUB1A-1 varieties have been used previously as checks in SF stress studies and their response to SF stress varied. For instance, Chirang-Sub1 showed tolerance, while Swarna-Sub1 exhibited susceptibility to SF stress (Septinsingh et al., Reference Septinsingh, Pamplona, Sanchez, Neeraja, Vergara, Heuer, Ismail and Mackill2009; Vergara et al., Reference Vergara, Nugraha, Esguerra, Mackill and Ismail2014; Singh et al., Reference Singh, Carandang, Gonzaga, Collard, Ismail and Septiningsih2017). Similarly, a moderate decline in yield under SF compared to the IR condition in SUB1A-1 genotypes has been reported recently (Kato et al., Reference Kato, Collard, Septiningsih and Ismail2019). These findings underscore the limitation in the use of SUB1A-1 introgressed lines under SF, reinforcing the need to explore alternative donor sources such as O. glaberrima.
The tolerant O. glaberrima donor identified for SF tolerance, GERVEX 2674 (Agbeleye et al., Reference Agbeleye, Olubiyi, Ehirim, Shittu, Jolayemi, Adetimirin, Ariyo, Sanni and Venuprasad2019), and the best performing of the three GERVEX 2674/O. sativa interspecific pre-breeding populations exhibited moderate elongation under SF and maintained high tillering compared to susceptible parents and progenies. The progenies with high elongation (160 cm and above) or short height (100 cm and below) exhibited maximum decline in GY under SF (data not shown). Also, the progenies with a higher number of tillers (30% or less reduction in the number of tillers compared to IR control) performed better in terms of GY under SF conditions. Elongation of stems under stagnant floodwater is an adaptive response that enables them to maintain contact with the air above the water surface (Voesenek and Bailey-Serres, Reference Voesenek and Bailey-Serres2013, Reference Voesenek and Bailey‐Serres2015). However, the degree of elongation plays a critical role in determining yield performance. In fact, moderately elongating rice varieties like AC85 maintained high GY under SF stress compared to those with high elongation or no elongation (Vergara et al., Reference Vergara, Nugraha, Esguerra, Mackill and Ismail2014; Kuanar et al., Reference Kuanar, Ray, Sethi, Chattopadhyay and Sarkar2017). High or excessive elongation, on the other hand, likely diverts assimilates towards vegetative growth rather than the developing panicles, leading to reduced GY. Additionally, the rice genotypes with a high number of tillers have a better chance of withstanding lodging under SF conditions and thus produce a high number of panicles and maintain GY. This is further supported by the positive correlation between the number of tillers, number of panicles and grain yield (Table 3). Similar results have been reported in previous studies where it was shown that high tillering leads to high grain yield under SF conditions (Kuanar et al., Reference Kuanar, Ray, Sethi, Chattopadhyay and Sarkar2017). This highlights the importance of selecting moderate elongation and high tillering capacity as breeding targets to enhance yield stability under SF.
The QTL mapping of the GERVEX 2674/O. sativa interspecific pre-breeding populations under SF identified 10 minor-effect QTLs associated with key yield-related traits, explaining just 2.3–10.2% of the observed phenotypic variation. This confirms the quantitative and complex nature of SF tolerance, which is likely controlled by a large number of genes (Kuanar et al., Reference Kuanar, Ray, Sethi, Chattopadhyay and Sarkar2017; Sarkar et al., Reference Sarkar, Chakraborty, Chattopadhyay, Ray, Panda and Ismail2019). The QTL identified on chromosomes 1, 2, 3, 6, 10 and 12 was linked to days to 50% flowering, plant height, GY and survival percentage. These differ from those reported by the previous studies, thereby providing new insights on the genetic architecture of SF tolerance in rice. However, it is important to note that some QTLs were environment-specific, detected only in one or two seasons, indicating the influence of environmental variability on the traits (Ghazy et al., Reference Ghazy, Abdelrahman, El-Agoury, El-Hefnawy, El-Naem, Daher and Rehan2023; Zhu et al., Reference Zhu, Weng, Bush, Zhou, Zhao, Wang and Li2023; Snehi et al., Reference Snehi, Choudhary, Kumar, Jayaswal, Kumar and Prakash2024). This suggests that future QTL validation efforts should consider multi-environment testing to capture stable and consistent loci that can be reliably used in breeding. The tolerant parent GERVEX 2674 contributed favourable alleles for most of the QTLs, particularly those associated with number of tillers, number of panicles, plant height and GY, which are key indicators of SF tolerance. This reinforces its potential value as a genetic donor for the improvement of SF tolerance in rice (Van Ooijen, Reference Van Ooijen1999; Mwakyusa et al., Reference Mwakyusa, Dixit, Herzog, Heredia, Madege and Kilasi2023).
Among the key findings in this study, QTL qdtf-12 on chromosome 12 was a major QTL for days to 50% flowering under SF environment. Co-localized QTLs for number of tillers and number of panicles (qtill-1.1/qpan-1.1 and qtill-5/qpan-5) suggest shared genetic control, as also supported by strong phenotypic correlation. Similarly, qdtf-3 and qgy-3, linked to days to 50% flowering and GY, respectively, exhibited pleiotropic effects across environments. These QTLs with pleiotropic effects may hold greater potential for marker-assisted breeding due to their influence on multiple traits. Furthermore, the QTLs identified in this study were different between the SF and control environments and thus suggest that gene expression is condition-specific and may be influenced by environmental factors and stress-responsive regulatory mechanisms (Kuanar et al., Reference Kuanar, Ray, Sethi, Chattopadhyay and Sarkar2017). This environment-specificity emphasizes the need for validation across seasons and locations to identify stable, broadly effective loci.
In conclusion, the interspecific pre-breeding populations developed and characterized in this study provide a valuable resource for fine mapping, functional validation and introgression of SF tolerance alleles into elite rice backgrounds. Future research should prioritize the validation of identified QTLs across diverse environments and genetic backgrounds to confirm their stability and utility in breeding programmes. Integrating these findings with transcriptomic, physiological and genome-wide association approaches can help uncover the underlying molecular mechanisms and identify robust markers for selection. Moreover, pyramiding favourable alleles for moderate elongation, high tillering and yield stability under SF could accelerate the development of flood-tolerant rice varieties tailored to the needs of smallholder farmers in flood-prone lowland ecosystems in vulnerable regions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1479262125100269.
Funding statement
This research was financially supported by the Ministry of Agriculture, Forestry and Fisheries (MAFF) of Japan through the Global Crop Biodiversity Trust (GCDT) grant number GS16010.
Open access
