Mining rare haplotypes to accelerate precision breeding in rice

Abstract

Rice (Oryza sativa L.) genomics has made significant progress, with over 13000 genome sequences available and more than 4500 genes functionally characterized. However, limited haplotype‐level information has constrained the effective deployment of these genes in breeding programs. Here, we investigated haplotype variation of 1644 functionally characterized genes representing six major trait categories using the 3000 Rice Genome panel. High haplotypic diversity was observed, with more than 300 haplotypes, which clustered into 2–15 major haplogroups. Several haplotypes showed subpopulation or region‐specific distributions, whereas others were broadly distributed across the panel. To identify rare haplotypes (frequency < 0.1) with potential breeding value, we prioritized 180 genes relevant to defined product concepts and identified 210 rare haplotypes across 51 genes. Notably, a rare haplotype of DEP1 (DEP1‐H2: frequency 0.07) associated with increased panicle length was introgressed into the mega rice variety Swarna through haplotype‐based backcross breeding. The resulting introgression lines showed a 29%–34% increase in panicle length compared to the recurrent parent, confirming the functional advantage of the rare haplotype.

Plain Language Summary

Rice contains thousands of genes, and variation within these genes shapes key agronomic traits such as yield, grain quality, and stress tolerance. In this study, we analyzed 1644 functionally characterized genes in the 3K Rice Genomes panel to understand how gene variants, known as haplotypes, are distributed and how they affect trait performance. We identified 210 rare haplotypes present in only a few rice accessions, some of which showed potential agronomic advantages. To evaluate their breeding potential, a rare haplotype of the DEP1 gene (DEP1‐H2) was introgressed into a widely grown variety, Swarna. The resulting lines showed a 29%–34% increase in panicle length compared with Swarna. These results demonstrate that rare allelic variants can provide measurable benefits and can be effectively used in breeding programs.

Abbreviations

3K RG

3000 Rice Genomes

Admix

admixture

Aro

aromatic

Ind

indica

InDel

insertion or deletion

IRGC

International Rice Genebank Collection

Japx

japonica

KASP

kompetitive allele‐specific PCR

PL

panicle length

SNP

single nucleotide polymorphism

Subtrop

subtropical japonica

Temp

temperate japonica

Trop

tropical japonica

1. INTRODUCTION

Rice (Oryza sativa L.) cultivation faces numerous challenges, including the need to increase yield to meet the growing demand, improve grain quality to address nutritional deficiencies, enhance growth and development for optimal resource utilization, combat biotic and abiotic stresses to ensure stable production, and manipulate phytohormone signaling for desired agronomic traits (Yuan et al., 2021). These complex traits are controlled by multiple genes that interact with each other and with the environment (C. Wang & Han, 2022). Advances in molecular biology, genomics, and high‐throughput sequencing technologies have facilitated the identification and characterization of candidate genes associated with various traits (Yao et al., 2018). Over 3000 rice accessions have been sequenced, and 10,000 rice accessions are being sequenced, revealing millions of genetic variants, including ∼16 million single nucleotide polymorphisms (SNPs) in the 3K Rice Genomes (3K RG) Project (Itoh et al., 2013; W. Wang et al., 2018).

These vast datasets enable detailed studies of genetic diversity, trait mapping, and the identification of evolutionary patterns. Combined with cloning of over 4500 genes, this genomic information accelerates precision breeding for improved yield, stress resilience, and climate adaptation, advancing global food security efforts (F. Huang et al., 2022; Yao et al., 2018). Recent precision breeding frameworks emphasize the strategic use of naturally occurring allelic variation, rather than transgenic approaches, to fine‐tune complex traits such as nutrient content, yield components, and stress tolerance through informed allele selection and deployment (Kaur et al., 2026). However, the genetic variations within these genes, particularly their haplotype diversity, remain largely unexplored.

Haplotypes are sets of closely linked genetic variants that tend to be inherited together, providing valuable insights into the genetic architecture of complex traits (Min et al., 2021). Haplotypes have several advantages over SNPs by (i) providing critical information on the evolutionary history and genetic variants controlling traits; (ii) generating hundreds of thousands of genetic polymorphisms in population samples of reasonable sizes portraying dependence or independence of linked markers with each other; (iii) studying the existence of linkage disequilibrium among markers makes it possible to infer population histories and localize genetic variants underlying complex traits (N'Diaye et al., 2017). Allele‐ and haplotype‐based approaches are increasingly recognized as central to next‐generation precision breeding, as they enable the identification of functionally superior or context‐specific alleles that can be directly introgressed into elite cultivars (Singh, 2025).

Recent research has highlighted the significance of haplotype diversity in crops, linking specific haplotypes to phenotypic variations in key traits (Abbai et al., 2019; Selvaraj et al., 2021; Sinha et al., 2020). In addition to this, several other studies have revealed the allelic diversity present within genes associated with yield potential, grain quality attributes, developmental stages, responses to pests and diseases, as well as abiotic stress tolerance mechanisms. Systematic exploration of allelic variation within cloned and functionally validated genes has emerged as a powerful strategy for trait improvement, particularly in the context of precision breeding and biofortification programs (G. Singh, 2025). Understanding haplotype diversity in these genes is necessary to elucidate the molecular mechanisms underlying these traits.

Rare haplotypes are crucial in rice breeding because they may possess beneficial, rare allelic variations that are not common in the general population but can significantly enhance key traits (Bhat et al., 2021). For instance, studies have shown that rare haplotypes, such as ASR3‐H88 with a low frequency, are associated with a significant improvement in reproductive stage drought tolerance (P. Singh et al., 2024). While another study focused on grain size and weight‐related genes, analyzing 42 functional genes in Indica rice accessions, identified rare haplotypes with frequencies as low as 5% for BG1‐H2 and 10% for GW5‐H1 (Liu et al., 2023). These rare haplotypes add value to key characteristics, guiding breeders in genomic‐assisted breeding strategies and accelerating the development of superior rice varieties.

In this research article, we aim to conduct an extensive haplotype diversity analysis of 1644 key genes that control economically important traits, aligned with the rice product concept. By leveraging available genomic resources, including high‐quality reference genomes, genotyping data from diverse rice accessions, and advanced bioinformatics tools, we explored allelic variation and haplotype patterns across these genes. Recent studies have focused on harnessing the superior/common/frequent haplotypes in breeding programs. However, our study aims to mine the rare haplotypes of reported cloned and/or functionally characterized genes. By focusing on rare allelic variants within the validated genes, this study aligns with emerging precision breeding paradigms that seek to unlock hidden genetic potential without compromising agronomic stability. Rare or unique haplotypes have a very limited occurrence in a given population and may sometimes significantly impact the trait of interest. In this study, we analyzed haplotype diversity, identified rare haplotypes, and validated their impact on breeding. We developed introgression lines carrying the rare haplotypes and observed a significant enhancement of the trait over the recurrent parent. Overall, the findings of this study will not only provide valuable insights into the genetic diversity and haplotype distribution within genes controlling key traits in rice but also highlight the importance of rare haplotypes, thereby assisting breeding programs in addressing the challenges of global food security.

2. MATERIALS AND METHODS

2.1. Rice genomic dataset and subpopulation structure

Genotypic data from the 3K RG panel genomes are available at the Rice SNP‐Seek Database (https://snp-seek.irri.org/), developed by the International Rice Research Institute in the Philippines (Alexandrov et al., 2015; W. Wang et al., 2018). The panel of sequenced rice accessions in this database had the genomic information of a diverse set of more than 3024 rice genotypes originating from 89 countries and was selected from a combined collection of ∼13,000 O. sativa accessions from the ∼180,000 rice accessions conserved in the International Rice GeneBank Collection (IRGC) at IRRI and the China National Crop Gene Bank. The complete details about the 3K RG rice genotypes, including the origin accession number and passport, are provided in W. Wang et al. (2018). For haplotype analysis, resequencing data from approximately 3000 accessions representing 12 subpopulations: aro (aromatic), aus, admix (admixture), ind1A, ind1B, ind2, ind3, indx (where ind is indica), japx (japonica), subtrop (subtropical japonica), temp (temperate japonica), and trop (tropical japonica) were used. Nipponbare was used as the reference genome.

2.2. Extraction of gene information controlling various traits in rice

The funRiceGenes database was utilized to collect relevant details on rice genes responsible for various characteristics and functions (Yao et al., 2018). The genes were further categorized into different trait categories based on their significant functions (Table S1).

2.3. Haplotype diversity analysis

Haplotype diversity analysis of functionally characterized genes was conducted using the 3K RG resequencing dataset described above. To ensure biological relevance, only functionally important sequence variants were considered. These variants included non‐synonymous SNPs and insertions or deletions (InDels) located in exonic regions that resulted in amino acid changes or were predicted to affect gene function, such as splice‐site variants, frameshifts, start‐loss, and stop‐gain mutations. Synonymous variants were excluded as they are less likely to influence protein function. The haplotype construction study was conducted using the SNP‐seek pipeline. The resulting haplotypes were further clustered to yield biologically meaningful haplogroups, reducing redundancy and capturing major functional patterns. Haplogroup clustering was performed using the K‐means algorithm, with the optimal number of clusters determined by the Calinski–Harabasz index, which evaluates clustering quality based on between‐ and within‐cluster variance (Saitta et al., 2008). The rare haplotypes were selected based on the following cutoff values: (i) should have a frequency ≤0.1 (10%), (ii) have more than five and a maximum of 300 genotypes in one haplotype group, and (iii) have a uniform distribution of either indica or japonica genotypes.

2.4. Development and evaluation of introgression lines

To evaluate the impact of rare haplotypes, the DEP1 gene regulating rice panicle length (PL) was selected as a candidate locus. To introgress DEP1‐H2 associated with increased PL, parental lines were selected based on haplotype evidence reported previously in our study (Abbai et al., 2019). The donor line ADT‐12::IRGC 6254‐1, possessing a rare haplotype (DEP1‐H2), was crossed with the recipient parent Swarna, which lacked haplotype DEP1‐H2.

A crossbreeding experiment was conducted during the 2020 dry season (2020DS) at the IRRI‐South Asia Regional Centre, Varanasi (ISARC). Hybridity of the resulting F₁ generation was confirmed using a set of 10 quality control SNP panels (IRRI GSL). Confirmed true F₁ hybrids were backcrossed with Swarna to produce the BC₁F₁ generation, followed by a subsequent second backcross to produce the BC₂F₁ generation. At BC₂F₁, plants were again screened to ensure hybridity using quality control markers. BC₂F₁ plants were selfed and advanced to the BC₂F₅ generation under standard puddled‐transplanted conditions. Foreground selection for DEP1‐H2 was done using kompetitive allele‐specific PCR (KASP) markers to trace the introgression in the backcross‐derived lines. Based on phenotypic and genotypic selection, 31 plants‐derived from BC₂F₂ individuals were grown in 2023WS. Among these BC₂F₃ plants, 14 plants carrying the DEP1‐H2 haplotype were advanced through selfing to the BC₂F₅. The resulting haplotype‐based isogenic lines (HaploIL's) were evaluated during the 2025 wet season (2025WS) at ISARC following standard IRRI agronomic protocols. Each entry was grown in a ∼2‐m² plot (2 m × 1 m) consisting of five rows planted at 20 cm × 15 cm spacing accommodating ∼65 plants per plot. At the physiological maturity stage, PL data were recorded from selected three replicates from each plot. These individuals were subsequently evaluated for grain yield during 2025WS.

2.5. Designing KASP marker

Sequences of the DEP1 gene was obtained from the Rice SNP‐Seek Database and aligned for parent lines, ADT‐12::IRGC 6254‐1 and Swarna, using ClustalW to detect haplotype variations. Following the identification of haplotype variations, KASP markers (snpOS01094, snpOS01095, snpOS01096, snpOS01097, snpOS01098, snpOS01100, and snpOS01101) were designed using the Polymarker software, as described by Ramirez‐Gonzalez et al. (2015). The designed KASP markers were synthesized by Intertek, Hyderabad, a specialized service provider in molecular biology. Genotyping was also conducted at Intertek using their high‐throughput platforms, ensuring precise and reliable genotypic data. This approach enabled the efficient differentiation of haplotypes in the studied rice populations, thereby contributing to understanding genetic diversity in DEP1.

2.6. Statistical analysis

Statistical significance among haplotype groups was tested using Duncan's multiple range test with the agricolae package in R (Felipe de Mendiburu, 2006). Groups with different letters in the graphs indicate significant differences at p < 0.05. A paired t‐test was employed to analyze the differences between the introgression lines and parents.

3. RESULTS

3.1. Genetic profiling reveals extensive diversity in the 3K RG panel

The 3K RG panel was analyzed for haplotype diversity within 1644 functionally characterized and cloned genes in rice among the 12 subpopulations (aro, aus, admix, ind1A, ind1B, ind2, ind3, indx, japx, subtrop, temp, and trop) (Figure 1). These genes were categorized into six main groups based on their functions, that is, (i) yield and yield‐related traits (316 genes), (ii) growth and development (478 genes), (iii) grain quality (204 genes), (iv) abiotic stress (319 genes), (v) biotic stress (145 genes), and (vi) phytohormones (182 genes) from the entire 3K RG panel (Table S1).

FIGURE 1.

Frequency and number of haplotypes present in each trait category and haplotype diversity and distribution across the 3000 rice genomes (3K RG) panel: (A.) Frequency and number of haplotypes present in each trait category, i.e, abiotic stress, biotic stress, grain quality, growth/development, phytohormones, and yield. Each trait category encompasses a different number of genes, as indicated in the Materials and Methods section, (B.) geographical distribution of different haplotypes of Xa10 gene, (C.) geographical distribution of different haplotypes of OsACS5 gene.

Across the panel, a high level of haplotypic diversity was observed. After grouping related haplotypes into haplogroups (see Material and Methods section for more details), 1451 genes (88.26%) exhibited between 2 and 15 distinct haplogroups, indicating widespread allelic variation within functionally important genes (Figure 1A; Table S1). The number and frequency of haplogroups varied substantially among genes, reflecting contrasting evolutionary and domestication histories. Distinct haplogroup distribution patterns were observed across rice subpopulations and geographic regions. For example, the bacterial leaf blight resistance gene Xa10 formed six haplogroups. The predominant haplogroup (Xa10‐H6; 41.87%) was widely distributed across regions, whereas rare haplogroups showed strong subpopulation and geographic specificity. Xa10‐H4 (0.49%) was restricted to indica‐derived accessions, primarily from South and Southeast Asia (Figure 1B; Tables S2 and S3). Other haplogroups displayed intermediate frequencies and were preferentially enriched in specific subpopulations, indicating structured haplotype differentiation within this locus. Similar patterns were observed for genes controlling growth and development.

OsACS5, a gene involved in ethylene biosynthesis and vegetative growth regulation, was resolved into nine haplogroups derived from five non‐synonymous SNPs. OsACS5 has been shown to influence vegetative growth and internode elongation through modulation of ethylene levels, particularly under conditions that promote cell division and elongation in young stems and leaf sheaths (e.g., roles in shoot and stem growth) (Yao et al., 2018). The most frequent haplogroup (OsACS5‐H4; 39.48%) was predominantly distributed across indica, aus, aro, and admix groups, whereas OsACS5‐H9 was rare (0.5%) and confined to a limited number of accessions from China, Nepal, and Myanmar (Figure 1C; Tables S4 and S5). These examples illustrate how both common and rare haplogroups coexist within key regulatory genes. The details of haplotype diversity and haplotype count for the selected genes are provided in Supporting Information Data S1–S6.

Based on the functional relevance and extent of haplotypic complexity, 180 genes (30 genes per functional category) were shortlisted for rare haplogroup variation. To ensure biological robustness and analytical reliability, genes that did not meet diversity criteria were excluded from further analysis. Consequently, we focused only on those genes that exhibited ≥3 distinct haplogroups (Supporting Information Data S1–S6) and harbored two or more non‐synonymous SNPs (Supporting Information Data S7–S12), thereby providing sufficient allelic resolution to discern genuine haplotypic diversity. This stringent selection ensured that the subsequent rare haplogroup characterization was both statistically sound and biologically meaningful, enabling accurate inference of functional divergence among alleles.

Furthermore, we performed an additional analysis to assess the distribution of the selected haplogroups across the 12 major rice subpopulations (aro, aus, admix, ind1A, ind1B, ind2, ind3, indx, japx, subtrop, temp, and trop) represented in the 3K RG panel to evaluate the population‐specific prevalence and differentiation of haplogroups, providing deeper insights into their evolutionary and domestication trajectories. By mapping haplogroup occurrence within each subpopulation, we could discern whether certain allelic variants were subpopulation‐specific, widely shared, or selectively enriched, thereby adding a valuable population‐genomic dimension to our haplotype diversity framework (Supporting Information Data S13–S18). Out of the 180 genes analyzed, rare haplogroup variations were identified in 51 genes (Table 1). These included seven yield‐related genes, 11 quality‐related genes, 10 plant growth‐related genes, five biotic stress‐related genes, and nine each in abiotic stress‐related and phytohormone‐associated genes (Table 1). Further, population‐wise analysis revealed that rare haplogroups were more frequently observed in indica and aus subpopulations, whereas japonica groups generally harbored fewer rare variants (Supporting Information Data S13–S18).

TABLE 1.

Comprehensive list of genes, haplotypes, and haplotype diversity across different trait categories for the identification of unique haplotypes in the 3K RG panel.

Gene	LOC_ID	Trait	Total haplotype group	SNPs	Indels	π	Tajima's D	Hd
D2	LOC_Os01g10040	Yield, panicle or tiller number	15	5	301	0.004	0.587	0.5
DEP1	LOC_Os09g26999	Yield, panicle or tiller number	12	5	147	0.000 a	0.000 a	0.9
LGD1	LOC_Os09g32540	Panicle length and grain size	15	7	31	0.029	0.860	0.9
OsCDC48	LOC_Os03g05730	Yield, panicle or tiller number	8	2	138	0.001	−0.013	0.5
OsSPL14|IPA1|WFP	LOC_Os08g39890	Yield, panicle or tiller number	15	4	128	0.000 a	0.000 a	0.25
TGW6	LOC_Os06g41850	Yield, grain weight	15	3	37	0.001	0.023	0.8
GW6a	LOC_Os06g44100	Yield, grain weight	15	5	10	0.001	−0.010	0.8
WTG1|OsOTUB1	LOC_Os08g42540	Grain quality, grain size and shape	15	4	73	0.000	0.027	0.3
GS9	LOC_Os09g27590	Grain quality, grain size and shape	15	3	19	0.000	0.000	0.7
GW7|GL7|SLG7	LOC_Os07g41200	Grain quality, grain size and shape	5	6	79	0.000	0.012	0.5
LABA1|OsLOGL6	LOC_Os04g43840	Grain quality, grain size and shape	15	1	57	0.000 a	0.000 a	0.7
DEP2	LOC_Os07g42410	panicle length and grain size	12	3	111	0.000 a	0.000 a	0.8
GluB4	LOC_Os02g16830	Grain quality, grain nutritional quality	14	3	25	0.000 a	0.000 a	0.5
GluA	LOC_Os10g26060	Grain quality, grain nutritional quality	15	5	11	0.000 a	0.000 a	0.2
OsCML16	LOC_Os01g04330	Grain quality, grain nutritional quality	4	4	29	0.000	0.010	0.7
OsAPX1	LOC_Os03g17690	Grain quality, zinc	15	2	83	0.000	0.033	0.6
OsNAS1	LOC_Os03g19427	Grain iron	15	2	29	0.000	−0.006	0.4
OsNAS3	LOC_Os07g48980	Grain iron	15	1	34	0.000 a	0.000 a	0.2
OsNRAMP5	LOC_Os07g15370	Growth and development	4	1	444	0.014	0.643	0.1
OsBEIIb	LOC_Os02g32660	Growth and development	6	3	128	0.000	−0.006	0.6
OsEPFL9	LOC_Os01g60900	Growth and development	15	3	11	0.003	0.216	0.6
sd1|GA20ox2	LOC_Os01g66100	Growth and development	4	3	112	0.000	0.002	0.9
D35|OsKOS3|OsKO2	LOC_Os06g37364	Growth and development	6	11	516	0.020	0.440	0.9
d18|OsGA3ox2	LOC_Os01g08220	Growth and development	14	1	37	0.210	0.960	0.1
SLR1|OsGAI	LOC_Os03g49990	Growth and development	14	6	24	0.470	1.520	0.6
LAX1	LOC_Os01g61480	Growth and development	10	2	0	0.230	0.990	0.5
OsCP1	LOC_Os04g57490	Growth and development	15	1	40	0.030	0.620	0
OsCesA7	LOC_Os10g32980	Growth and development	15	1	118	0.000 a	0.000 a	0
eIF4G	LOC_Os07g36940	Rice tungro disease resistance	13	15	135	0.002	0.199	0.9
OsERF922	LOC_Os01g54890	Blast	15	3	0	0.001	−0.009	0.5
BPH29	LOC_Os06g01860	BPH resistance	15	1	26	0.000 a	0.000 a	0.7
RCH10	LOC_Os03g30470	Sheath blight resistance	15	5	7	0.017	0.528	0.7
OsPGIP1	LOC_Os05g01380	Sheath blight resistance	15	3	30	0.000	0.019	0.5
OsLEA3‐2	LOC_Os03g20680	Abiotic stress tolerance	15	3	38	0.001	−0.009	0.7
OsMYB59	LOC_Os01g74410	Abiotic stress tolerance	15	3	37	0.001	0.069	0.8
OsMYBR1	LOC_Os04g49450	Abiotic stress tolerance	15	6	22	0.001	0.069	0.7
OsDSR2	LOC_Os01g62200	Abiotic stress tolerance	15	2	41	0.000	−0.008	0.6
ONAC095	LOC_Os06g51070	Abiotic stress tolerance	14	2	8	0.001	−0.004	0.3
CAA30371.1	LOC_Os04g59420	Abiotic stress tolerance	15	1	78	0.000	0.043	0.6
OsNAC14	LOC_Os01g48446	Abiotic stress tolerance	15	3	13	0.003	0.165	0.2
OsSKL2	LOC_Os10g42700	Abiotic stress tolerance	15	3	34	0.002	0.134	0.6
ASR1	LOC_Os01g72910	Abiotic stress tolerance	15	2	14	0.000 a	0.000 a	0.2
CRL1|ARL1	LOC_Os03g05510	Phytohormones	15	3	4	0.000 a	0.000 a	0.2
CRL4|OsGNOM1	LOC_Os03g46330	Phytohormones	15	3	27	0.044	1.308	0.4
OsCAND1	LOC_Os02g07120	Phytohormones	4	1	224	0.000 a	0.000 a	0.5
LA1|LAZY1	LOC_Os11g29840	Phytohormones	6	9	57	0.001	0.048	0.3
OsIAA18	LOC_Os05g44810	Phytohormones	15	1	42	0.002	0.182	0.2
OHP1|OsAHP1	LOC_Os08g44350	Phytohormones	14	2	82	0.000	0.005	0.6
OsCKX4	LOC_Os01g71310	Phytohormones	15	2	31	0.000 a	0.000 a	0.7
D10|OsCCD8	LOC_Os01g54270	Phytohormones	14	5	169	0.000 a	0.000 a	2.4
OsIPT7	LOC_Os05g47840	Phytohormones	15	1	74	0.001	0.011	0.6

Abbreviation: Hd, haplotype diversity.

^a Nucleotide diversity (π) = 0.0000 means near‐zero diversity, rounding of values <0.0001.

For instance, genes controlling plant architecture (D2) and grain weight (TGW6) displayed rare haplotypes predominantly in the japonica population of East Asian countries (Supporting Information Data S1). While genes influencing grain size and shape, such as OsOTUB1, GW8, GS9, GS2, GS5, GW7, An‐2, and DEP2, exhibited diverse rare haplogroups distributed predominantly in the indica and aus subpopulations, revealing genetic diversity within the 3K RG panel (Figure 2; Supporting Information Data S2, S8, and S14). In the case of genes associated with grain quality, 39 rare haplotype variations were identified across 11 genes that regulate key quality traits, including grain shape and size, starch composition, and iron and zinc content in the grain. It was observed that OsCML16 and GW7 each possessed a single rare haplogroup, whereas OsNAS3, a gene involved in regulating grain iron content, exhibited the highest number of distinct rare haplogroups (seven) (Table 1). Genes from other functional categories, such as yield (Supporting Information Data S1, S7, and S13), grain quality (Supporting Information Data S2, S8, and S14), plant growth and development (Supporting Information Data S3, S9, and S15), biotic (Supporting Information Data S4, S10, and S16), abiotic stress (Supporting Information Data S5, S11, and S17), and phytohormones (Supporting Information Data S6, S12, and S18), also showed the presence of significant rare haplotypes (Table 1), highlighting their potential for trait optimization.

FIGURE 2.

Frequency of the rare and inferior haplotype of important cloned genes affecting yield and related traits in different rice subpopulations. This figure presents: (A) The frequency of rare haplotypes in relation to major and other haplotypes for each gene, and (B) the distribution of rare haplotypes among the 12 sub‐populations. *indicates that these genes have more than one rare haplotype present among the 3K RG population, while the number in parentheses indicates rare haplotype group of each gene.

For the trait of yield, a total of 26 rare haplotypes were identified across seven genes (D2, DEP1, LGD1, OsCDC48, OsSPL14/IPA1/ WFP, TGW6, and GW6a) associated with yield and yield‐related traits (Table 1). The number of rare haplotypes per gene ranged from one to seven (Table 1). Notably, the gene TGW6, which influences grain weight, exhibited the highest diversity with seven rare haplotypes (Table 1). In contrast, OsCDC48 showed the least variation, with only a single rare haplotype detected (Table 1). These genes play critical roles in regulating panicle architecture, tillering, and overall yield potential. Among these, both D2 and DEP1 harbored three rare haplogroups each, represented by 10–28 genotypes within the 3K RG analyzed (Supporting Information Data S1 and S13). Notably, the DEP1‐H2 haplotype exhibited a diverse distribution across five subpopulations, with distinct accessions found in the indx, aus, aro, trop, and temp groups. The distribution of haplotypes associated with yield and yield‐related genes, highlighting the presence of rare haplotypes across different subpopulations within the 3K RG panel, is illustrated in Figure 3A,B. Similarly, the rare haplogroups and their distribution across the rice subpopulations are depicted in Figures S1–S6. The presence of unique, rare, region‐specific haplogroups provides insights into targeted breeding efforts tailored to regional preferences and environmental conditions.

FIGURE 3.

Haplotype distribution and variation of the DEP1 gene. (A.) Haplotype frequency distribution of the DEP1 gene in the3K rice genome panel. (B.) Distribution of the DEP1 haplotypes across different sub‐populations of rice (admix, aro, aus, indica, and japonica subgroups). (C.) Haplotype variation in the DEP1 gene.

3.2. Validation of rare haplotypes of the DEP1‐H2

Among the rare haplotype groups of DEP1 observed in the 3K RG panel, rare haplogroup H2 (Figure 3C) was also found to be the superior haplotype for PL in our previous study (Abbai et al., 2019). To understand the potential of this rare haplogroup, a haplotype‐based backcross breeding approach was employed to develop isogenic lines. The donor parent, ADT‐12::IRGC 6254‐1, which possesses the haplotype H2 for the DEP1 gene, contributes to a PL of 31.74 cm. While the recipient, Swarna, which lacks this superior haplotype, possesses a PL of 21.44 cm.

Four introgressed lines were developed (IR 176319:1‐5‐9‐11‐1, IR 176319:1‐5‐9‐11‐2, IR 176319:1‐5‐9‐11‐3, and IR 176319:1‐5‐9‐11‐4) employing superior haplotype‐based foreground selection in each generation. The lines, namely, IR 176319:1‐5‐9‐11‐1, showed a PL of 27.70 cm, representing a significant enhancement (29%) over the recipient parent, Swarna. Similarly, the lines IR 176319:1‐5‐9‐11‐2 (28.84 cm), IR 176319:1‐5‐9‐11‐3 (28.57 cm), and IR 176319:1‐5‐9‐11‐4 (28.14 cm) exhibited (34%, 33%, and 31%) increase of PL over Swarna (21.4 cm) (Figure 4, Table S6). Interestingly, the developed introgressed lines exhibited a 29%–34% increase in PL compared to Swarna, indicating a considerable and promising improvement in yield potential.

FIGURE 4.

Performance of developed lines for panicle length. (A.) Comparison of panicle length and percentage superiority of introgressed lines over parents. (B.–D.) Phenotypic expression of panicle length in introgressed lines relative to parents.

To assess the agronomic relevance of the introgressed DEP1 haplotype on yield, the grain yield of the selected lines was evaluated. The introgression lines exhibited yield comparable to or exceeding that of the recurrent parent (Table S6). IR 176319:1‐5‐9‐11‐2 with the largest panicle (28.8 cm) showed the highest grain yield (5124.18 kg ha⁻¹) with 9.87% yield advantage over Swarna (4663.74) at ISARC, Varanasi. While IR 176319:1‐5‐9‐11‐3 produced 4862.61 kg ha⁻¹, corresponding to 4.26% yield advantage over the recurrent parent. Collectively, these results confirm that introgression of rare haplotypes DEP1‐H2 increased grain yield in Swarna‐ILs due to an increase in PL, highlighting its potential utility for using rare haplotypes in crop breeding.

4. DISCUSSION

Genetic diversity in genes associated with economically important traits is the foundation for an effective breeding program or crop improvement strategy. In the face of accelerated climate change and the need for region‐specific adaptation, targeting populations with superior or rare genetic variation presents a powerful avenue to enhance crop resilience. Recently, large‐scale databases such as the 3K RG panel and the Rice SNP‐Seek database have provided extensive data on SNPs distributed across the genome (Alexandrov et al., 2015; W. Wang et al., 2018). These databases can be utilized to gain insight into the genetic distribution of various functional genes in rice. Traditional methods of association have relied on common haplotypes, which are easier to detect due to their frequency in the population. However, this approach often overlooks rare haplotypes (Bhat et al., 2021). It is essential to understand that genetic diversity and distribution are crucial for identifying common and unique patterns of genetic variation within the gene pool and have numerous applications. While the available database provides resources to understand the allelic variation, significant challenges remain in characterising haplotype distribution, structure, and the functional validation, particularly in establishing the phenotypic effects of these rare haplotypes.

In this study, we focused on the haplotype diversity of genes and on mining the rare haplogroups from six main groups: yield and yield‐related traits, growth and development, grain quality, abiotic stress, biotic stress, and phytohormones. Our analysis of 1644 genes related to important agronomic traits in the 3K RG panel revealed substantial haplotype diversity, with most genes (88.2%) ranging from 2 to 15 distinct haplotypes, each with varying frequencies and SNP effects (Supporting Information Data S1–S18). This genetic diversity (all haplotype groups) was particularly evident in indica, japonica, and aus subpopulations, representing a valuable resource for targeted crop improvement (Supporting Information Data S1–S18). The distribution patterns of these haplotypes reflect both historical selection pressures and potential breeding value, with some rare haplotypes showing clear geographical associations that likely indicate local adaptation. For example, DEP1, OsNAS1, TGW6, GW6a, OsNAC14, GW7, RCH10, and many other genes have location or population‐specific rare/unique haplotypes. Some of these rare/unique haplogroups are present in indica, japonica, aus, or aro subpopulations, while others are localized in a specific location. The rare haplogroups were predominantly found in the indica population, while the prevalence of rare haplogroups in the japonica population was generally low or nonexistent (Supporting Information S13–S18). The results are consistent with previous studies highlighting the genetic complexity underlying rice yield and related traits (Selvaraj et al., 2021; Sharma et al., 2018; W. Wang et al., 2018; Wu et al., 2020). The geographically restricted distributions suggest these haplotypes may provide adaptive benefits under specific environmental conditions or pathogen pressures, making them particularly useful for regional breeding efforts.

Rare haplotypes, characterized by their low population frequency, often indicate recent evolutionary events or results from transgressive segregation, potentially exhibiting extreme phenotypic expression. Such haplogroups are particularly valuable for breeders aiming to develop improved varieties with novel trait combinations. The study identified 51 rare haplogroups from 180 selected genes. From a breeding perspective, the observed haplotype variation across trait‐specific genes offers a valuable framework for informed parental selection and strategic allele introgression. Within the yield‐related gene group, LGD1 showed the nucleotide diversity (π = 0.029), suggesting the maintenance of divergent alleles in the 3K RG panel, likely shaped by historical selection favoring panicle architecture. In contrast, among the grain quality genes GS9 showed the lowest diversity (π = 0.0001), indicating a more conserved locus, possibly due to the fixation of favorable haplotypes.

For breeders, this distinction highlights where variability exists for exploitation (e.g., LGD1) versus loci already under strong selection pressure (e.g., GS9). Meanwhile, WTG1, with the lower π (0.0004) in this group, suggests narrower diversity, implying fewer options for recombination to improvegrain texture or chalkiness through this locus alone. Targeting genes like GS9 offers a broader scope for breeding grain quality in accordance with consumer preferences across various market segments. Among genes, SLR1/OsGAI exhibited the highest nucleotide diversity (π = 0.470) across all trait categories, reflecting substantial allelic divergence at this key gibberellin‐signaling locus, potentially maintained by diversifying selection.

From a breeding perspective, high‐diversity stress‐responsive loci are priority candidates for multi‐parent crosses aimed at stacking resilience. This genetic richness facilitates the dissection of complex traits and supports the development of cultivars with improved stress tolerance (Scott et al., 2020). Among biotic stress‐associated genes, eIF4G exhibited elevated π (0.002). In hormone signaling, OsCCD8 stood out, exhibiting a near‐zero π value (0.000001), indicative of low natural variation in strigolactone biosynthesis. This makes it difficult to find the allelic variations for modifying plant architecture or stress responsiveness, depending on breeding goals. The haplotype diversity ranged from 0.1 (OsGA3ox2) to 0.95 (eIF4G and LGD1), reflecting varying degrees of allelic richness. High haplotype diversity in eIF4G and LGD1 suggests broad adaptability and potential for recombination breeding, while lower values in genes like OsGA3ox2 caution against overreliance on such loci unless supported by phenotypic validation. By prioritizing genes with both high diversity and balanced allele frequency patterns, trait‐pyramiding strategies have been shown to enhance long‐term genetic gain significantly (Bhat et al., 2021; P. Singh et al., 2024; Yu et al., 2024).

A significant challenge in crop improvement is translating genetic discoveries into practical applications. Our study focused on dep1, a well‐characterized regulator of panicle architecture and grain yield (X. Huang et al., 2009). To further characterize the DEP1 haplotypes, we analyzed them for sequence similarity with the reference DEP1 allele from X. Huang et al. (2009). The major haplotypes were 100% similar to reference DEP1 with H10 being the most common (frequency = 0.38, n = 1143). This suggested that this major haplotype is the most conserved allele in the population. However, the rare haplotype, DEP1‐H2 (ADT12), showed relatively low similarity (∼77.17%) with the reference DEP1 allele and had a low frequency (0.07). This suggested that this is a highly divergent allelic variant of DEP1.

To evaluate the breeding potential of the rare and superior DEP1‐H2 haplotype, we employed a targeted introgression approach, transferring this allele from donor ADT‐12::IRGC 6254‐1 into the genetic background of Swarna, a high‐yielding mega variety. This strategy allowed us to isolate the effect of DEP1‐H2 while minimizing the confounding impacts of other genetic factors. In addition to haplotype identification, the use of KASP markers proved critical for the effective deployment of rare haplotypes in breeding. The custom‐designed KASP markers enabled precise foreground selection of the DEP1‐H2 haplotype during backcrossing, ensuring accurate tracking of the target allele across generations. Of all seven KASP markers screened in this study, “snpOS01097” and “snpOS01101” showed clear polymorphism between the two parental lines. Consequently, BC₂F₃ lines for the ADT‐12 allele in these loci were selected and advanced by selfing. The high frequency of fixed allele combinations in these derived lines is an indicator of the success of marker‐assisted selection in fixing the target haplotype. The integration of haplotype information with KASP‐based selection demonstrates a practical and scalable route for translating population‐level haplotype discoveries into applied rice breeding programs.

Introducing the rare and superior DEP1‐H2 haplotype into the high‐yielding cultivar Swarna resulted in consistent increases in PL (29%–34%) across multiple isogenic lines, demonstrating how haplotype‐guided breeding can enhance yield components. Significantly, this increase in PL was found to be associated with a measurable yield advantage of up to 9.87% under field conditions, with the line having the longest panicle also having the highest grain yield.

This phenotypic effect may be explained by the underlying protein haplotype variation that was found for the DEP1 protein. The protein haplotype variation for the DEP1‐H2‐associated protein haplotype (DEP1_PH2) was found to differ from the major protein by a number of amino acid substitutions that were found to be located in key functional regions of the protein, such as the cysteine‐rich domain that is involved in panicle development. The structural implications of the protein were analyzed and the amino acid substitutions were found to result in a difference in ligand binding properties and a decrease in the interaction energy with RGB2, a key protein in the G‐protein signaling pathway. This success was achieved through careful selection of both the target haplotype and the genetic background of the recurrent parent's genome while maintaining the beneficial effects of the haplotype (Figure 4; Table S6). This provides clear evidence that rare and valuable haplotypes that have been identified in population studies can be effectively utilized to bring about measurable improvements in agronomic traits.

In order to understand the context of the results that have been presented here, it is essential to understand the context by comparing the effects of the DEP1‐H2 allele with the previously characterized DEP1 alleles. There are differences in the context and nature of phenotypic effects between this and previous reports. For instance, while the classical dep1 allele of X. Huang et al. (2009) is known to carry traits such as dense and erect panicles, higher branching, and up to ∼40% improvement in grain yield per plant, mostly in Japonica. Whereas, in this study, the rare haplotype DEP1‐H2 was found to carry a 29%–34% increase in PL and up to ∼10% yield advantage in the elite indica mega variety Swarna in field conditions. Consistent with our previous findings (Abbai et al., 2019), DEP1‐H2 was associated with longer panicles (∼23–31 cm range) and grouped among the superior haplotypes for PL.

It is worth noting, however, that phenotypic differences in DEP1‐H2 and other dep1 alleles in previous reports might be influenced by genetic background and epistatic interactions with other yield‐related loci such as Gn1a, GS3, and Ghd7, considering that grain yield is a complex trait influenced by various components. In addition, in an elite genetic background where grain yield potential is already optimized, it is inherently more difficult to make further gains, and therefore, any level of improvement is significant and valuable in breeding programs.

These findings have important implications for rice breeding and genetics. The greater diversity of rare haplotypes in indica and aus subpopulations, compared to their scarcity in japonica, likely reflects differences in domestication history and breeding intensity. This pattern suggests that indica and aus varieties may be particularly valuable sources of new genetic variation for breeding programs. The strong geographical patterns observed in many haplotypes, such as Xa10‐H4 in South and Southeast Asia, or DEP1_PH2 in East Asia, reveal the importance of developing breeding approaches tailored to local conditions and genetic resources. Further investigations into the underlying molecular mechanisms and interactions associated with the rare haplotype can deepen our understanding and open up new opportunities for targeted genetic improvement in rice varieties.

Our work highlights the importance of examining not only single genetic variants but also complete haplotypes, protein‐level effects, and phenotypic outcomes. While individual mutations can sometimes have significant effects, exploring combinations of variants within haplotypes provides a better understanding of functionally crucial genetic variation and its phenotypic consequences. The protein haplotype analysis we conducted for DEP1 provides a valuable approach for understanding how genetic changes impact protein structure and function, enabling more informed selection of breeding targets. Expanding such analyses to additional genes, combined with advanced phenotyping and population genetics methods, could significantly improve our ability to identify and use valuable haplotypes for crop improvement.

5. CONCLUSION

This study provides a comprehensive haplotype‐level assessment of 1644 cloned and functionally characterized genes across the 3K Rice Genome panel, revealing extensive allelic and haplotypic diversity within genes controlling key agronomic traits. A large proportion of genes exhibited multiple haplogroups with distinct population and geographic distributions, underscoring the rich but unevenly utilized genetic variation present in global rice germplasm. Importantly, the systematic mining of rare haplotypes uncovered a subset of low‐frequency alleles with clear biological and breeding relevance.

The successful introgression of the rare DEP1‐H2 haplotype in the elite variety Swarna for increased PL and yield advantage provides evidence that rare haplotypes discovered at the population scale can be effectively translated for breeding purposes. Several additional rare haplotypes, especially for grain quality and abiotic stress tolerance, are also promising candidates for validation. Stacking multiple superior haplotypes at different loci may provide an effective breeding approach to exploit additive and epistatic effects to develop optimized combinations for increased stability, adaptability, and performance in target environments. Overall, the present study provides a framework for haplotype breeding at scale. Future work will be directed toward the comparative evaluation and stacking of multiple rare haplotypes to fully exploit their breeding potential for climate‐resilient and high‐performing rice varieties.

AUTHOR CONTRIBUTIONS

Swati Tyagi: Data curation; formal analysis; investigation; methodology; software; validation; visualization; writing—original draft. Anoop Kishor Singh Gurjar: Data curation; formal analysis. Sonali Vijay Habde: Investigation; resources; writing—review and editing. Krishna Tesman Sundaram: Investigation; resources; writing—review and editing. Namrata Devidas Patil: Investigation; resources; writing—review and editing. Diksha Rai: Data curation; visualization. Ajay Bogati: Formal analysis; visualization; writing—review and editing. Pronob J. Paul: Formal analysis; visualization; writing—review and editing. Ankur Kumar Singh: Data curation; visualization. Ajay Tripathi: Writing—review and editing. Sanjay Kalia: Resources; writing—review and editing. Ajay Kohli: Writing—review and editing. Arvind Kumar: Conceptualization; supervision; writing—review and editing. Pallavi Sinha: Conceptualization; funding acquisition; methodology; project administration; resources; supervision; validation; writing—original draft. Uma Maheshwar Singh: Conceptualization; funding acquisition; investigation; project administration; resources; supervision; validation; writing—original draft. Vikas Singh: Conceptualization; funding acquisition; investigation; project administration; resources; supervision; validation; writing—original draft.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supplementary Information

Figure S1. Distribution of haplotype diversity in genes governing major traits in rice.

Figure S2. Frequency of the rare and inferior haplotype of important cloned genes affecting grain quality and related traits in different rice subpopulations.

Figure S3. Frequency of the rare and inferior haplotype of important cloned genes affecting growth and development‐related traits in different rice subpopulations.

Figure S4. Frequency of the rare and inferior haplotype of important cloned genes affecting biotic stress‐related traits in different rice subpopulations‐.

Figure S5. Frequency of the rare and inferior haplotype of important cloned genes affecting abiotic stress‐related traits in different rice subpopulations.

Figure S6. Frequency of the rare and inferior haplotype of important cloned genes affecting phytohormone‐related traits in different rice subpopulations.

Table S1. The aplotype diversity for genes governing specific traits (yield, grain quality, growth and development, biotic, abiotic, and phytohormones) in the 3K RG panel.

Table S2. Distribution of SNPs, Indels in the Xa10 gene, and number of haplotypes present, their sub‐populations, and distribution frequencies in the 3K RG panel.

Table S3. The SNP effect of various SNPs present in the Xa10 gene.

Table S4. The distribution of SNPs, Indels in the OsACS5 gene, and number of haplotypes present, their sub‐populations, and distribution frequencies in the 3K RG panel.

Table S5. The SNP effect of various SNPs present in the OsACS5 gene.

Table S6. Performance of advanced breeding lines for panicle length over parents.

Data S1. List of candidate genes associated with yield traits, showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S2. List of candidate genes associated with grain quality traits, showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S3. List of candidate genes associated with growth and development‐related traits, showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S4. List of candidate genes associated with biotic stress showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S5. List of candidate genes associated with abiotic stress showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S6. List of candidate genes associated with phytohormones showing the number of haplogroups, their frequencies, and distribution across 12 rice subpopulations.

Data S7. The predicted impact of SNPs on the coding sequence of the candidate genes governing yield and related traits.

Data S8. The predicted impact of SNPs on the coding sequence of the rice candidate genes governing grain quality and related traits.

Data S9. The predicted impact of SNPs on the coding sequence of the rice candidate genes governing growth and development‐related traits.

Data S10. The predicted impact of SNPs on the coding sequence of the rice candidate genes governing biotic stress and related traits.

Data S11. The predicted impact of SNPs on the coding sequence of the rice candidate genes governing abiotic stress and related traits.

Data S12. The predicted impact of SNPs on the coding sequence of the rice candidate genes governing phytohormones‐related traits.

Data S13. Distribution of the haplotypes in the 12 major subpopulations for yield‐related genes.

Data S14. Distribution of the haplotypes in the 12 major subpopulations for grain quality‐related genes.

Data S15. Distribution of the haplotypes in the 12 major subpopulations for growth and development‐related genes.

Data S16. Distribution of the haplotypes in the 12 major subpopulations for biotic stress‐related genes.

Data S17. Distribution of the haplotypes in the 12 major subpopulations for abiotic stress‐related genes.

Data S18. Distribution of the haplotypes in the 12 major subpopulations for phytohormones‐related genes.

Tyagi, S. , Gurjar, A. K. S. , Habde, S. V. , Sundaram, K. T. , Patil, N. D. , Rai, D. , Bogati, A. , Paul, P. J. , Singh, A. K. , Tripathi, A. , Kalia, S. , Kohli, A. , Kumar, A. , Sinha, P. , Singh, U. M. , & Singh, V. K. (2026). Mining rare haplotypes to accelerate precision breeding in rice. The Plant Genome, 19, e70292. 10.1002/tpg2.70292

DATA AVAILABILITY STATEMENT

The authors confirm that the data supporting this study's findings are available in the article as supporting information.