Genetic analysis of trade-offs among disease resistance, yield, and quality traits employing genome-wide association mapping in indica rice (Oryza sativa L.)

Abstract

The genetic trade-offs among complex traits are often witnessed in rice, however, very little is known about the contributing genes and mechanisms to exploit in breeding programmes. Here, we aimed to understand the genetic trade-offs among disease resistance, quality, and yield traits employing genome-wide association mapping. In all, 78 common marker-trait associations (MTAs) were identified for the targeted traits. In addition, five pleiotropic MTAs, 17 tightly linked MTAs, and two pleiotropic and tightly linked MTAs were detected for various trait combinations. The majority of MTA clusters were observed for quality traits (15 clusters) followed by the combined yield and quality traits (5 clusters) while only one cluster was found for combined yield and disease resistance traits. Further, the prediction of candidate genes controlling MTA clusters by exploiting the publicly available rice genome databases, revealed D-type cyclin 3;1 and Xyloglucan endotransglucosylase were found to be responsible for controlling grain size traits. We found no significant large linkage drag blocks with major MTAs for the targeted traits indicating that the indica rice genotypes have fewer trade-offs compared to japonica. The current study provides deeper insights into the genetic trade-offs among complex traits in rice, aiding in the meticulous planning of future breeding strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11032-025-01578-w.

Introduction

Rice (Oryza sativa L.) is an important staple cereal crop that provides food and calories to more than 3 billion people on the earth. To meet global demand and diverse consumer preferences, rice breeding efforts must focus on developing high-yielding, disease-resistant varieties with superior grain quality. However, achieving this balance is challenging, as these traits are governed by multiple genes with complex molecular mechanisms. Earlier research findings suggest that these three traits are interfused and found to have genetic trade-offs or antagonistic associations among themselves (Deng et al. 2017).

Genetic trade-offs occur when the improvement of one trait leads to a reduction in the performance of another trait, and vice versa (Dwivedi et al. 2021). The main reason for these trade-offs can be attributed to pleiotropy, linkage disequilibrium, phytohormones and secondary metabolites (Dwivedi et al. 2021). The trade-offs help plants to adapt to their external environment. The elucidation of the mechanism controlling the trade-offs provides opportunities to strategize the crop breeding programmes.

Paying close attention to the available rice cultivars reveals three widely observed antagonistic relationships: “higher yield vs. disease resistance,” “good quality vs. higher yield,” and “good quality vs. disease resistance.” These trade-offs present a significant challenge for breeders striving to develop varieties that balance high yield, good quality, and disease resistance. This antagonistic phenomenon may result from linkage drag, where desirable and undesirable genes are closely linked, or from pleiotropic effects, where a single gene influences multiple trait. Previously several reports have uncovered the trade-offs of complex traits such as source-sink, growth-defence, yield-nutrition etc. For instance, the introgression of Pi2, a blast resistance gene, has been shown to negatively affect the number of spikelets per plant and overall yield (Wu et al. 2019). This trade-off is often linked to energy allocation, as resistance-related genes can trigger defense mechanisms that divert resources from growth and reproduction. Specifically, the activation of defense pathways, such as salicylic acid (SA)-mediated signaling, may suppress auxin-related pathways, leading to reduced yield components (Yang et al. 2012). Additionally, disease resistance genes can influence plant growth through structural modifications, such as cell wall reinforcement, which may further compromise yield potential (Hu et al. 2017). The transfer of another blast resistance gene, Piz-t resulted in a higher chalkiness rate (Xiao et al. 2017), whereas Fukuoka et al. (2009) found that the introgression of the pi21 resistance gene had negative effects on grain quality. Therefore, achieving a balance between yield, quality, and disease resistance warrants a profound understanding of the inherent linkage drag associated with these traits. Recently, it was observed that some blast resistance genes such as pigm encode nucleotide-binding, leucine-rich repeat receptor (Xie et al. 2019), and Bsr-dl encodes C2H2 type transcription factor (Li et al. 2017), which enhance blast resistance without any yield penalty.

Advances in genome sequencing and marker-based breeding methods have facilitated the precise cloning of genes and their effective utilization in rice breeding. Genome-Wide Association Studies (GWAS) play a crucial role in identifying genetic variants linked to specific traits by analyzing genetic markers across diverse populations. They have been successfully employed to uncover genomic regions governing complex traits such as yield, quality, and disease resistance in diverse germplasms (Zhu et al. 2016; McCouch et al. 2016). Further, Xiao et al. (2021) employed GWAS in 200 japonica rice varieties to understand the balance of yield, quality and resistance to blast disease in rice.

In this study, we hypothesize that genetic trade-offs exist between yield, quality, and disease resistance due to the influence of closely linked or pleiotropic genes. To investigate this, we utilized GWAS to identify key genomic regions associated with these antagonistic relationships. Specifically, we aimed to examine the balance between yield, quality, and resistance to major diseases such as blast and bacterial leaf blight (BLB) within a diverse panel of indica germplasm. By uncovering the genetic basis of these trade-offs, our findings contribute to the optimization of rice breeding programs, offering valuable insights for developing high-yielding, disease-resistant, and high-quality rice varieties to enhance overall productivity.

Material and methods

Plant material

In the present study, 230 rice genotypes comprising landraces, indica, japonica, aromatic, and modern cultivars originating from different parts of India were used for phenotyping of yield, quality and disease resistance traits in two consecutive years of 2019 and 2020 at Agricultural Research Station (ARS), Acharya NG Ranga Agricultural University (ANGRAU), Nellore, Andhra Pradesh, India. Among them, 142 representative genotypes were selected for GWAS (Table S1) to identify genomic regions governing yield traits, disease-resistance for blast and bacterial leaf blight (BLB) along with quality traits.

Screening of rice genotypes for blast resistance

For screening of rice genotypes for blast resistance, each genotype was sown in a single row of 50 cm with a row-to-row spacing of 10 cm. After every 10 genotypes, local blast susceptible checks (BPT5204 and NLR34242) were planted. In addition, the entire nursery was surrounded on four sides by two rows of susceptible check varieties (Fig. S1). A high level of farmyard manure was incorporated into the UBN nursery before sowing to increase the disease. Nitrogen fertilizer was applied at a rate of 120 kg/ha of which, half of the nitrogen was applied as a basal dose at the time of sowing and the remaining half was applied 15 days after sowing (DAS). Other fertilizers were applied as per recommended package of practice i.e., 50 kg/ha of both Phosphorous and Potassium. The Magnaporthe oryzae isolate (race I A- 1) was previously collected from rice fields at Nellore and sprayed the inoculum on 21 days old seedlings and the disease was allowed to develop. All the measures were taken to ensure the occurrence of maximum disease pressure including maintaining optimum humidity. The disease reaction observations were scored 40 days after sowing or after the 7th-day post-inoculation, when the susceptible spreader BPT5204 and NLR34242 completely died. The plants were scored with a scale ranging from 0 to 9 based on leaf blast severity by following Standard Evaluation System (SES) scale, International Rice Research Institute (IRRI), Philippines (2002).

Screening of rice genotypes for BLB resistance

The bacterial leaf blight (BLB) infected leaves were collected from the BLB infested field near Nellore, Andhra Pradesh, India, and preserved in the laboratory. Infected rice leaves were cut into small pieces (5 mm of infected tissue and 5 mm of adjacent healthy tissue) and were grinded in mortar and pestle. For inoculation of the pathogen in plots, the clip method (Kauffman et al. 1973) was used. In the field seedlings of the test, all 230 rice genotypes were grown the puddled field with an alpha lattice design. Each rice cultivar was transplanted in two rows of 3 m long and spaced 20 cm apart. BLB susceptible genotype BPT5204 and NLR 34242 were transplanted around the whole trial plot to develop the inoculum pressure. The fertilizer rate used was 120:60:30 kg NPK/ha. Other agronomic practices were followed as per recommendation. Disease scoring was done with a 1–9 scale after three weeks of inoculation. On the basis of disease scoring value, genotypes were classified into different categories according to their resistance level using the standard IRRI procedure.

Phenotyping of rice genotypes for grain yield and quality traits

These rice genotypes were evaluated under field conditions in alpha lattice design with three replications with a spacing of 20 X 15 cm. Important agronomic traits such as plant height (PH), panicle length (PL), number of panicles per plant (NP), number of tillers per plant (NT), number of filled grains per plant (NFG), number of chaffy grains per plant (NCG), total number of grains per plant (NG), spikelet fertility (SF) and single plant yield (SPY) were recorded. After three months of harvesting, threshing, and cleaning, the seeds from individual genotypes were dried under shade until the moisture content reached 14. The seed was dehusked in a Satake laboratory huller (Type THU 35 A) and polished in a Satake Rice Polisher (Type TM 05). The polished seed obtained was then utilized for the analysis of 14 seed-quality traits at the Department of Genetics and Plant Breeding. S.V. Agricultural College, ANGRAU, Tirupati. The hulling, milling and head rice recovery are measured using the following formulas.

Hulling percentage

Hulling percentage is estimated as

Milling percentage

Milling percentage is estimated as

Head rice recovery (HRR)

Head rice recovery or milling recovery is an estimate of head rice (milled rice with more than two-thirds size) expressed in percentage

The standard procedures were followed for recording data of quality traits: Grain length (GL), grain breadth (GB), grain length-breadth ratio (LBR), chalkiness, grain length after cooking (GLAC), grain breadth after cooking (GBAC), linear elongation ratio (LER), breadth elongation ratio (BER), alkali spreading value (ASV)/gelatinization temperature (GT) and gel consistency (GC).

Statistical analysis

Analysis of variance (ANOVA) was performed using the lme4 and car packages in R (version 3.1.1), employing a linear mixed-effects model to account for potential data imbalance. The model included genotype as a fixed effect and replication and year as random effects, treating the year as a pseudo-replication rather than a true multi-year model. Spearman correlation analysis was conducted using the corrplot package to assess relationships between traits. Descriptive statistics, including mean, variance, standard error, and standard deviation, were calculated using the stats package following established statistical methods.

Criteria for classification of rice genotypes

In order to understand the trade-offs among the yield, quality, and disease resistance traits, we have classified all the rice genotypes based on their optimum trait combinations for YLD (grain yield), ASV (alkali spreading value), GC (gel consistency) and LBR (length-breadth ratio or grain size) for grain quality and disease resistance i.e., blast (BR) and BLB using phenotypic data of 2019 and 2020 years. The genotypes having more than 30 g per plant are considered high yielding, genotypes with ASV score of 3–5, GC 40–60 mm, and LBR > 2.1 were considered best quality while genotypes with Blast and BLB resistance score of 1–3 were considered resistant.

Genotyping

DNA extraction was carried out from the leaf samples of rice genotypes using the modified Cetyl Trimethyl Ammonium Bromide (CTAB) method described by Murray and Thompson (1980). The purity and concentration of the isolated genomic DNA samples were estimated by a Nanodrop spectrophotometer (Thermo Scientific, USA). The DNA amplification was carried out using a thermal cycler (Eppendorf 5804R). 2 μl of 50 ng/μl of template DNA was used for each reaction. The master mix was prepared by taking 0.5 μl of 5 pmol primers (both forward and reverse primer), 0.5 μl of 1 mM deoxy trisphosphate nucleotides (dNTPs), 1 μl of 10X PCR buffer with Mg²⁺ (ABM, Canada) and 0.1 μl of 5U/μl Taq DNA polymerase (ABM, Canada) and 5.4 μl of sterile distilled water was added to make up the volume to 8 μl. Subsequently, 8 μl of the master mix was added to each of the wells in the PCR plate well having 2 μl of template DNA to make the final volume 10 μl. After completion, the PCR plate was stored at 4 °C and the amplified products were later resolved on 3% agarose gel and the results were visualized using gel documentation (Alpha Innotech, USA).

Population structure

A total of 60 SSR (Simple Sequence Repeat) markers distributed on chromosomes were used for genotyping of 142 rice genotypes (Table S2). The genotypic data has been used for genetic structure (Q) and clustering of the population using STRUCTURE version 2.2 (Pritchard et al. 2000). The ΔK method (Evano et al. 2005) was used to estimate the most probable K value for the analyzed data, using Structure Harvester ver. 0.6. (Earl and VonHoldt 2012). The K value, which is the optimum number of populations, was selected after ten independent runs with a burn-in period of 100,000 steps with 100,000 Monte Carlo Markov chain (MCMC) replicates. The range of genetic clusters was set from K = 2 to 10. The number of subpopulations (K) was determined based on the posterior probability values (LnP(D) and ΔK). The genotypes with more than ≥ 0.80 Q values were considered “pure” while less than 0.80 were assumed as “admixture”.

Genome-wide association mapping

The phenotypic data of blast and BLB disease score, quality traits, yield, and its associated traits, and genome-wide SNP (Single nucleotide polymorphism) marker data were used to identify the marker-trait association (MTA). The double digest restriction site-associated DNA sequencing (ddRAD) method was used to develop 37,808 SNP markers for genotyping of rice genotypes. The SNP data was filtered for minor allele frequency (MAF) ≥ 0.05 and the maximum missing sites per SNP was fixed to < 20%. The marker-trait associations (MTAs) were identified using the Mixed Linear Model (MLM) method including population structure and family kinship (family relatedness) using GAPIT software (Lipka et al 2012). Manhattan plots were produced and a threshold value for declaring marker-trait association generated through a mixed linear model (MLM) approach was set at –Log p = 4 (i.e., p-value ≤ 0.0001). The modified Bonferroni correction was used to control the genome-wide type I error rate. The MTAs were named according to the trait name, and chromosome number. For example, qASV_2 represents QTL for the trait alkali spreading value identified on chromosome 2. The candidate genes were predicted using the workflow based on integrated multi-omics in rice (Keerthi et al. 2024).

The categorization of MTAs into"pleiotropic"and"tightly linked"was based on their observed genetic and phenotypic relationships.

• Pleiotropic MTAs are those that influence multiple traits, indicating that a single genetic variant or region affects more than one phenotypic characteristic. These MTAs often exhibit a broad effect across various traits, which suggests their involvement in key regulatory pathways or complex biological processes.

• Tightly linked MTAs, on the other hand, are those that appear to be associated with a specific trait or a limited set of traits. These MTAs may be part of a genetic cluster, where closely positioned loci or variants are inherited together, showing strong co-occurrence without influencing other traits significantly.

Results

Phenotyping of yield, disease resistance and quality-related traits

All 230 rice genotypes were screened for blast and BLB resistance for two years 2019 and 2020 (Table S3). In blast screening, in the year 2019, among all the genotypes, none of them showed highly resistant (Score 0), however, 43 genotypes showed resistance (18.7%), 87 were moderately resistant (37.82%), 15 were moderately susceptible (6.5%), 57 were susceptible (24.78%), and 28 were highly susceptible (12.17%) (Table S3). Whereas in the year 2020 also, none of them were found to be highly resistant, however, 41 genotypes have shown resistant (17.82%), 79 were moderately resistant (34.34%), 32 were moderately susceptible (13.91%), 57 were susceptible (24.78%), and 21 were highly susceptible (9.13%). Among all, 112 (48.7%) genotypes have shown resistance to blast consistently in both the 2019 and 2020 years.

In case of BLB screening, among 230 rice genotypes, none of them have shown highly resistant (Score 0), while 56 were resistant, 57 were moderately resistant, 40 were moderately susceptible, 45 were susceptible, and 32 were highly susceptible in the year 2019 (Table S3). Similar to the previous year, in the year 2020 also, none of them were classified as highly resistant, while 38 genotypes showed resistance, 59 were moderately resistant, 55 were moderately susceptible, 50 were susceptible, and 28 were highly susceptible. Among all, 80 (34.8%) genotypes showed resistance to BLB consistently in both the 2019 and 2020 years.

Interestingly, 58 (25.2%) genotypes have shown resistance to both blast and BLB in two years. Among them, the rice genotypes, Tetep, NLR3302, NLR3247, Azucena, Koshihikari, MTU1121, Wild rice 1, and WGL915 are notable and have shown high resistance to both the diseases in two years with consistent disease score of 1(Resistant) (Fig. 1).

Fig. 1.

Phenotyping of rice genotypes for blast and BLB resistance in 2019 and 2020. A Classification of rice genotypes based on blast resistance score. B Classification of rice genotypes based on BLB resistance score. C The leaf symptoms of Blast and BLB resistant genotypes. The detailed methodology of screening of Blast and BLB resistance is mentioned in the Materials and Methods

The rice genotypes were evaluated for nine yield and twelve quality traits in both 2019 and 2020 years (Fig. S2). The mean values of these traits were subjected to ANOVA analysis and found significant genotypic differences for all the traits in both years (Table S4).

The correlation analysis among yield, quality traits, disease resistance of bacterial leaf blight, and leaf blast traits revealed almost similar results in 2019 and 2020 (Fig. 2). In yield traits, PH, PL, NP, and FG are positively correlated with SPY. In the case of quality attributes, grain length is positively correlated with LBR, GLAC, and LER while grain breadth is positively correlated with GBAC and negatively associated with GL, LBR, and BER. ASV is also positively correlated with GL, LBR, and GLAC while GC is significantly associated with GBAC and BER. The blast disease (BR) score has shown a positive correlation with BLB resistance in both years while both diseases are negatively correlated with plant height in 2019.

Fig. 2.

Correlation of grain yield, quality and disease resistance traits among rice genotypes in 2019 (A) and 2020 (B). Slimmer and darker color of ellipse represents stronger correlation. Red colour circles indicate positive correlation while blue colour indicates negative correlation. *Significance at 5% level; ** Significance at 1% level; *** Significance at 0.01% level. The trait codes are mentioned in the Materials and Methods

The correlation analysis conducted across two consecutive years (2019 and 20120) in different rice populations—Aromatic, Indica, Japonica, Landraces, Wild Relatives, and NERICA & Aus—revealed both stable and variable trait associations. In the Aromatic and Indica groups, correlations among yield and yield-contributing traits, such as panicle length, grain number, and 1000-grain weight, remained consistently positive across both years, indicating stable genetic relationships suitable for breeding. Disease resistance traits like BLB and blast showed variable or slightly negative correlations with yield, suggesting potential trade-offs. Japonica genotypes exhibited moderate to strong positive correlations for yield and quality traits, though some year-specific variations pointed to environmental sensitivity. Landraces and wild relatives displayed highly inconsistent correlation patterns across years, reflecting their genetic diversity and strong genotype-by-environment interactions. In contrast, the NERICA & Aus group demonstrated relatively stable correlations, especially among agronomic traits, with weaker associations observed for disease resistance. Overall, while Indica and Aromatic groups showed reliable trait relationships across years, the variability in Landraces and wild types highlights the complexity and potential of these genetic resources for trait exploration.

Classification of rice genotypes to understand the trade-offs among disease resistance, yield and quality traits

Among all genotypes, 82 exhibited consistently high yield (above 30 g per plant). In case of quality traits, the most preferred genotypes with ASV (3–5 score) are 57 and genotypes with GC (30–40 mm) are 84 while genotypes with LBR (above 3) are 175 in both the years (Fig. 3). In total, rice genotypes were classified into 47 optimum trait combinations, of them 15 are 2-trait, 10 are 3-trait, 15 are 4-trait, six are 5-trait and one 6-trait combinations were observed. Among all, a maximum of 79 rice genotypes are under the BR + LBR trait combination category while only two genotypes viz., Kanchan and Pusa 44 are with all the six desirable trait combination such as BLB + BR + GC + ASV + LBR + YLD (Fig. 3).

Fig. 3.

The frequency of rice genotypes with optimum traits (A) and trait combinations (B). YLD (grain yield), ASV (alkali spreading value), GC (gel consistency) and LBR (length-breadth ratio or grain size), BR (Blast resistance) and BLB (Bacterial leaf blight). The criteria used to classify the best genotypes are the genotypes having more than 30 g per plant are considered high yielding, genotypes with ASV score of 3–5, GC 40–60 mm, and LBR > 2.1 were considered best quality while genotypes with Blast and BLB resistance score of 1–3 were considered resistant

Genome-wide association study for blast, bacterial leaf blight, yield and quality traits

The population structure analysis divided all rice genotypes into four populations (Fig. 4). Population1 (pop1), population2 (pop2), population3 (pop3) and population4 (pop4) consists of 55, 14, 44, and 29 genotypes, respectively. Further, based on the Q values, the rice genotypes were categorized as pure or admixture. Pop1 consists of 46 pure (83.6%) and 9 admixed (15.4%) genotypes, pop2 has 12 pure (85.7%) and 2 (14.3%) admixed genotypes, pop3 has 25 pure (56.8%) and 19 (43.2%) admixture genotypes, while pop4 has 15 pure (51.7%) and 14 (48.3%) admixture genotypes (Fig. 4).

Fig. 4.

Population structure analysis of rice genotypes. A Delta k graph of rice genotypes for determining the optimum number of subpopulations. B Stacked bar plot for population structure of rice genotypes at K = 5 generated using STRUCTURE. C Genotype-wise bar plot of rice genotypes using STRUCTURE software

In the present investigation, a total of 96 marker-trait associations (MTAs) were identified in 2019 while 103 were identified in 2020 for blast disease, bacterial leaf blight, quality traits, and yield and its associated traits. The phenotypic variance explained (PVE) for all these MTAs are ranged from 7.68% to 28.61% (Table S5 and Table S6). All MTAs are distributed on 12 chromosomes (Fig. S3 and Fig. S4). A total of 78 common MTAs was found to be consistent over both years and among them, one MTA for blast resistance, none for BLB resistance, 46 for quality traits, and 31 for yield and its associated traits (Table 1). All the common MTAs were distributed on 12 chromosomes (Fig. 5).

Table 1.

The list of common MTAs and their favourable alleles

S.No	Trait	Chromosome	SNP	QTL	P.value	Rsquare. of. Model	Fav. Allele	No of Genotypes
						with.SNP
1	BS	6	UChr6_23358790	qBS_6.2 & qBS_6	0.000769	9.02	GG	103
2	ASV	2	UChr2_24143008	qASV_2	0.000105–0.000139	24.63–26.84	CC	113
3	ASV	5	UChr5_20388634	qASV_5	0.000136–0.000267	24.66–25.65	TT	128
4	ASV	6	UChr6_6717918	qASV_6.1	0.00003–0.000025	26.92–28.5	CC	74
5	ASV	6	UChr6_6283316	qASV_6.2	0.000064–0.000071	25.65–27.34	GG	102
6	ASV	6	UChr6_6327999	qASV_6.3	0.000096–0.000104	25.11–26.85	GG	104
7	ASV	6	UChr6_7702617	qASV_6.4	0.000109–0.000116	24.87–26.79	CC	83
8	BER	1	UChr1_24330887	qBER_1	0.000077–0.000087	13.52–14.42	CC	123
9	BER	4	UChr4_31410420	qBER_4.1	0.00011–0.000088	13.12–14.21	AA	129
10	BER	4	UChr4_30174439	qBER_4.2	0.00013–0.000091	12.86–14.17	CC	127
11	GB	6	UChr6_25072196	qGB_6	0.000022–0.000064	24.39–25.54	CC	121
12	GB	9	UChr9_8923188	qGB_9.1	0.000052–0.00089	19.32–25.8	AA	75
13	GB	9	UChr9_9937779	qGB_9.2	0.000024–0.000886	24.25–19.32	CC	114
14	GB	11	UChr11_19883308	qGB_11	0.00021–0.000137	21.27–24.52	CC	129
15	GB	12	UChr12_21051713	qGB_12	0.000036–0.000053	23.68–25.77	CC	77
16	GC	4	UChr4_22077270	qGC_4	0.000232	11.27	AA	104
17	GC	6	UChr6_3175510	qGC_6.1	0.000039–0.000046	13.78–14.23	TT	60
18	GC	6	UChr6_12283093	qGC_6.2	0.000183–0.000192	11.56–11.81	GG	104
19	GC	9	UChr9_10335879	qGC_9	0.000171–0.000213	11.4–11.92	TT	102
20	GL	4	UChr4_9015490	qGL_4	0.000074–0.000078	24.18–24.97	GG	7
21	GL	5	UChr5_20185460	qGL_5	0.000003–0.000005	28.61–28.96	TT	127
22	GL	6	UChr6_13469745	qGL_6	0.000058	24.58	TT	129
23	GL	7	UChr7_29372206	qGL_7	0.000045–0.000049	24.95–25.54	GG	128
24	GL	9	UChr9_924504	qGL_9	0.00011–0.00015	23.69–24	CC	106
25	GL	10	UChr10_23038717	qGL_10	0.00014–0.00018	23.05–24.08	TT	57
26	HP	4	UChr4_21038673	qHP_4	0.000125–0.000168	11.5–11.88	AA	24
27	HP	7	UChr7_29063682	qHP_7	0.000047–0.000111	12.06–13.48	GG	82
28	HP	12	UChr12_17597775	qHP_12	0.000156–0.000163	11.61–11.47	CC	121
29	GBAC	7	UChr7_1702286	qGBAC_7	0.000253	16.41	CC	130
30	GBAC	12	UChr12_13669926	qGBAC_12	0.000396	15.77	CC	19
31	GLAC	5	UChr5_20185460	qGLAC_5	0.000029–0.000076	17.35–18.05	TT	127
32	GLAC	9	UChr9_924504	qGLAC_9	0.00012–0.000119	15.93	CC	106
33	GLAC	11	UChr11_19337793	qGLAC_11	0.00019–0.000145	15.64–15.98	GG	106
34	GLAC	12	UChr12_9009809	qGLAC_12	0.000078–0.000087	16.56–17.16	GG	84
35	LBR	5	UChr5_20185460	qLBR_5	0.0000010	23.49–23.75	TT	127
36	LBR	6	UChr6_13469745	qLBR_6.1	0.00002–0.000008	19.93–17.1	TT	129
37	LBR	6	UChr6_14083152	qLBR_6.2	0.00018	13.73	TT	119
38	LBR	9	UChr9_10308881	qLBR_9	0.000095–0.00013	14.17–16.21	CC	124
39	LBR	10	UChr10_23038717	qLBR_10	0.000051–0.000089	14.79–17.12	GG	130
40	LBR	12	UChr12_19541378	qLBR_12.1	0.000003–0.000004	19.48–21.69	AA	128
41	LBR	12	UChr12_19479274	qLBR_12.2	0.000041–0.000052	15.61–17.45	TT	89
42	LBR	12	UChr12_19207332	qLBR_12.3	0.000088–0.00016	13.9–16.32	A/C	22
43	LER	1	UChr1_34313806	qLER_1	0.000033	18.52–18.77	AA	58
44	LER	3	UChr3_33208291	qLER_3	0.000203–0.0002	16.11	GG	78
45	LER	10	UChr10_13980725	qLER_10	0.000187–0.00019	15.93	GT	84
46	MP	1	UChr1_17378860	qMP_1	0.000206–0.000249	11.06–11.33	CC	112
47	MP	2	UChr2_4998397	qMP_2	0.000061–0.000084	12.45–13.49	AA	78
48	NCG	5	UChr5_16336557	qNCG_5.1	0.000217–0.000285	12.81–13.21	CC	67
49	NCG	5	UChr5_6781090	qNCG_5.2	0.000209–0.000285	12.81–13.02	CC	135
50	NCG	10	UChr10_23018420	qNCG_10	0.000389–0.000167	12.35–13.36	TT	67
51	NFG	4	UChr4_17481870	qNFG_4	0.000196–0.000274	11.41	GG	77
52	NFG	6	UChr6_23212242	qNFG_6	0.00067–0.000824	9.25–9.81	GG	98
53	NFG	8	UChr8_7882895	qNFG_8.2 & qNFG_8	0.00041–0.000247	10.54–11.06	AA	84
54	GP	1	UChr1_35240947	qGP_1	0.000282–0.000627	9.71–10.95	GG	127
55	GP	4	UChr4_18635459	qGP_4	0.000463	10.16	CC	50
56	GP	10	UChr10_4469220	qGP_10	0.000527–0.001066	8.93–10.01	TT	103
57	GP	11	UChr11_28924638	qGP_11	0.000433–0.000643	10.3–9.67	GG	109
58	GP	12	UChr12_17897107	qGP_12	0.001031	8.97	TT	44
59	NP	2	UChr2_29709549	qNP_2	0.000244–0.000227	10.45–10.56	AC	17
60	NP	9	UChr9_17267693	qNP_9.2 & qNP_9	0.000099–0.000164	11.06–11.83	AA	131
61	NT	2	UChr2_29709549	qNT_2	0.000296–0.000296	10.15	AC	17
62	PH	1	UChr1_39810747	qPH_1	0.000188–0.000326	20.27–21.42	GG	131
63	PH	3	UChr3_7858396	qPH_3	0.000039–0.000068	23.58–22.41	AA	128
64	PH	8	UChr8_18879706	qPH_8	0.000153–0.000191	21.29–21.39	CC	108
65	PL	1	UChr1_29404597	qPL_1	0.000118–0.000145	17.72–19.21	AA	126
66	PL	3	UChr3_10554639	qPL_3	0.000165–0.000263	16.87–18.73	TT	29
67	PL	4	UChr4_595135	qPL_4	0.00018–0.000177	17.41–18.64	TT	120
68	PL	7	UChr7_24091803	qPL_7	0.000096–0.000188	17.35–19.5	AA	118
69	PL	11	UChr11_16012709	qPL_11	0.000062–0.000175	18.65–18.95	TT	100
70	SF	10	UChr10_1107410	qSF_10	0.000267–0.000297	11.52–12.78	CC	89
71	SPY	1	UChr1_42051248	qSPY_1	0.000117–0.00025	13.44–14.19	AA	115
72	SPY	3	UChr3_21727824	qSPY_3 & qSPY_3.1	0.000004–0.000009	18.58–19.28	CC	37
73	SPY	4	UChr4_19708000	qSPY_4	0.00023–0.000061	13.57–15.18	TT	58
74	SPY	7	UChr7_28875695	qSPY_7 & qSPY_7.2	0.000008–0.000034	16.45–18.31	TT	70
75	SPY	8	UChr8_26569293	qSPY_8.1	0.000013–0.000014	17.42–17.93	GG	63
76	SPY	8	UChr8_22115372	qSPY_8.2	0.000069–0.000076	15.01–15.22	TT	77
77	SPY	9	UChr9_6250941	qSPY_9	0.000001	22.13–22.35	TT	59
78	SPY	10	UChr10_13728468	qSPY_10	0.000008–0.000044	16.06–18.42	TT	62

Fig. 5.

Physical map of MTAs commonly detected for Blast, Bacterial Leaf Blight Disease Resistance, Quality, Yield and yield associated traits in 2019 and 2020. The brown colour – Blast resistant MTAs; Green colour-quality trait MTAs; Red colour -yield traits; Chr-Chromosome; Chromosome left side Physical distance (Kb) and the right-side MTAs

The MTA alleles with positive phenotypic effect that resulted in increased trait values are considered as favourable alleles whereas the alleles with negative effect on the trait are considered as unfavourable alleles. A maximum of 135 genotypes are found to have favourable allele (CC) at qNCG_5.2 for no. of chaffy grains followed by qNP_9.2 and qNP_9 MTAs for no. of panicles in 131 genotypes with favourable allele of AA while minimum of seven genotypes are having favourable allele (GG) for grain length MTA, qGL_4 (Table S4). Of 78 common MTAs, the genotype, BPT4358 is found to have a maximum of 65 favourable alleles followed by BPT2613 and BPT2270, and JGL11470 with 64 favourable alleles (Fig. S5 and Fig. S6). Whereas, the aromatic varieties such as Basmati370, Basmati386 and Pusa Sugandh 3 are found to have very a smaller number (30) of favourable alleles while improved Pusa basmati, Pusa basmati 1 and Pusa1121 with only 34 favourable alleles.

We uncovered some of the MTAs related to targeted traits have formed few clusters on different chromosomes and presumed the reason for this phenomenon is either due to pleiotropic genes or tightly linked genes on chromosomes. In the present study also, five pleiotropic MTAs, 17 tightly linked MTAs, and two pleiotropic and tightly linked MTAs were detected for various trait combinations on a majority of the chromosomes except chromosomes 3 and 8 (Table 2). Of 25 MTA clusters identified in the current study, the majority of MTA clusters were observed for quality traits (QLT) (15) followed by yield and quality (YLD + QLT) traits (5) while only one cluster was found for disease resistance and yield (DR + YLD) traits (Fig. 6). The one cluster on chromosome 6 between UChr6_13469745 and UChr6_12283093 that controls quality trait MTAs viz., qGL_6/qLBR_6.1 and qGC_6.2 encompasses a maximum physical distance (1186.6 kb). The superior allele analysis at this MTA cluster shows significant differences at GG (45.3%) and GA (28.7%) alleles and of them the allele GG is the most beneficial.

Table 2.

List of pleiotropic and tightly linked MTA clusters for targeted traits

S. No	SNP	Chr	Common MTAs	Type of MTA cluster	Physical distance (kb)	Trait combinations
1	UChr1_34313806 UChr1_35240947	1	qLER_1, qGP_1	Tightly linked MTA cluster	927.14	YLD + QLT
2	UChr2_29709549	2	qNP_2, qNT_2	Pleiotropic MTA	0.0	YLD
3	UChr4_18635459 UChr4_19708000	4	qGP_4, qSPY_4	Tightly linked MTA cluster	1072.5	YLD
4	UChr4_18635459 UChr4_17481870	4	qGP_4, qNFG_4	Tightly linked MTA cluster	1153.5	YLD
5	UChr4_22077270 UChr4_21038673	4	qGC_4, qHP_4	Tightly linked MTA cluster	1038.7	QLT
6	UChr5_20185460	5	qLBR_5, qGL_5, qGLAC_5	Pleiotropic MTA	0.0	QLT
7	UChr5_20185460 UChr5_20388634	5	qLBR_5, qGL_5, qGLAC_5, qASV_5	Pleiotropic MTA and Tightly linked MTA cluster	203.1	QLT
8	UChr6_6283316 UChr6_6327999	6	qASV_6.2, qASV_6.3	Tightly linked MTA cluster	44.7	QLT
9	UChr6_6717918 UChr6_6327999	6	qASV_6.1, qASV_6.3	Tightly linked MTA cluster	389.9	QLT
10	UChr6_7702617 UChr6_6717918	6	qASV_6.4, qASV_6.1	Tightly linked MTA cluster	984.7	QLT
11	UChr6_13469745 UChr6_12283093	6	qGL_6/qLBR_6.1& qGC_6.2	Tightly linked MTA cluster	1186.6	QLT
12	UChr6_13469745	6	qGL_6, qLBR_6.1	Pleiotropic MTA	0.0	QLT
13	UChr6_14083152 UChr6_13469745	6	qLBR_6.2, qGL_6/qLBR_6.1	Tightly linked MTA cluster	613.4	QLT
14	UChr6_23358790 UChr6_23212242	6	qBS_6.2 & qBS_6, qNFG_6	Tightly linked MTA cluster	146.5	DR + YLD
15	UChr7_29372206 UChr7_29063682 UChr7_28875695	7	qGL_7, qHP_7, qSPY_7 & qSPY_7.2	Tightly linked MTA cluster	497	YLD + QLT
16	UChr9_924504	9	qGL_9, qGLAC_9	Pleiotropic MTA	0.0	QLT
17	UChr9_9937779 UChr9_8923188	9	qGB_9.2, qGB_9.1	Tightly linked MTA cluster	1014.6	QLT
18	UChr9_9937779 UChr9_10308881 UChr9_10335879	9	qGB_9.2, qLBR_9, qGC_9	Tightly linked MTA cluster	398	QLT
19	UChr10_23038717	10	qGL_10, qLBR_10	Pleiotropic MTA	0.0	QLT
20	UChr10_23038717 UChr10_23018420	10	qLBR_10&qGL_10, qNCG_10	Pleiotropic MTA and Tightly linked MTA cluster	20.3	YLD + QLT
21	UChr10_13980725 UChr10_13728468	10	qLER_10, qSPY_10	Tightly linked MTA cluster	252.3	YLD + QLT
22	UChr11_19883308 UChr11_19337793	11	qGB_11, qGLAC_11	Tightly linked MTA cluster	545.5	QLT
23	UChr12_19479274 UChr12_19545378	12	qLBR_12.1, qLBR_12.2, qLBR_12.3	Tightly linked MTA cluster	334	QLT
24	UChr12_17897107 UChr12_17597775	12	qGP_12, qHP_12	Tightly linked MTA cluster	299.3	YLD + QLT

Fig. 6.

Manhattan plot and quantile–quantile (Q-Q) plots for MTA clusters for yield (YLD), quality (QLT) and disease resistance (DR) traits. A Manhattan plot of MTA clusters for qGL_6/qLBR_6.1 and qGC_6.2. B Manhattan plot of MTA clusters located on chromosome 6. C Superior allele analysis of MTA for gel consistency, qGC_6.2. D Manhattan plots showing MTA clusters of different trait combinations. The significance threshold at –Log p = 4 (i.e., p-value ≤ 0.0001) was followed to identify MTAs. The numbers on the X-axis represent rice chromosomes

Prediction of underlying candidate genes for MTAs

The candidate genes for each MTA were predicted based on publicly available rice genome databases and were described trait-wise hereunder (Table S7).

Blast resistance

A total of 11 MTAs was identified for blast disease in both 2019 and 2020 years with a PVE range of 7.68% to 13.02%. All these MTAs were distributed on chromosomes 1,3,4,6,8, 10, 11, and 12. Among them, 7 candidate genes viz., OsAT10, OsRpt5b, OsGH5BG, OsNIN-like4, OsSUT1, OsWSL9, and OsCCR were considered as the possible candidate genes for the trait leaf blast disease resistance MTAs.

Bacterial leaf blight resistance

Four MTAs for trait BLB were identified in 2019 and 2020 years. None of them was found stable in both years. Two MTAs were identified in 2019 with a PVE of 19.83%—19.84%. while the remaining two were observed in 2020 with a PVE of 13.69%—14.43%. All the MTAs were distributed on chromosomes 6, 7, and 11. From these MTA regions, the annotated genes, Yr 10, OsHIR3, and OsRab16 were predicted as the plausible candidate genes for BLB resistance.

Hulling Percentage (HP)

Three MTAs for trait HP were identified on chromosomes 4, 7, and 12 with a PVE range of 11.47–13.47% in 2019 and 2020 with three candidate genes OsPHD20, OsWOX11, and OsGRF7.

Milling Percentage (MP)

A total of three MTAs for MP was identified in two years. Among them, one MTA, qMP_12 was detected in the year 2020 which explains 12.33% of PVE, and the other two MTAs qMP_1, and qMP_2 were identified stably in both years with PVE range from 11.05–13.48%. The annotated genes, OsPSBO, OsTATC, and OsCDPK27 could be the candidate genes governing the trait milling percentage.

Head Rice Recovery (HRR)

A total of seven MTAs was identified for trait HRR with a PVE range from 10.98%—14.35% in both years. Among them, four MTAs viz., qHRR_3, qHRR_7, qHRR_8, and qHRR_11 were identified in 2019 while the remaining viz., qHRR_3, qHRR_7, and qHRR_12 were identified in 2020. All the MTAs were distributed on chromosomes 3, 7, 8, 11, and 12. The annotated genes OS-QR, OsGELP92, OsESP1, Hd5, OsSERK1/OsBAK1, OsPEZ2, OsJAZ6, and RSus3(t) are the most probable candidate genes for the trait head rice recovery.

Grain Length (GL)

For trait GL, six MTAs were identified on chromosomes 4, 5, 6,7, 9, and 10 with a PVE range from 23.69–28.95% in both years. The genes like ASYMMETRIC LEAVES2, OsDegp7. OsDAO1, OsCuDAO1, OsNAS3, CycD3;1 and OsXTH25 were predicted as the probable candidate genes for trait grain length.

Grain Breadth (GB)

Five MTAs were detected for trait GB in both 2019 and 2020 years, with a phenotypic variation range from 19.32%−25.79% and are distributed on chromosomes 6,9,11 and 12. The annotated genes STR-like-7 and OsABCG46 within the MTAs were predicted to be the candidate genes for trait grain breadth.

Length/breadth ratio (LBR)

For trait, LBR eight MTAs were identified in both years, with a PVE range from 13.73%−23.74%, and all were identified in both the years. These MTAs were distributed on chromosomes 5, 6, 9, 10 and 12. The genes LOB domain protein 6, OsDegp7, OsDAO1, OsCuDAO1, OsGLK1, OsSLD, OsWAK86, OsISC1, and OsXTH25 could be the candidate genes for trait LBR.

Grain Length After Cooking (GLAC)

For GLAC four MTAs were identified with a PVE range of 15.64%−18.04% in both 2019 and 2020 and are distributed on chromosomes 5, 9, 11, and 12. The genes, LOB domain protein 6 and OsDegp7, CycD3;1 and nudix hydrolases 13 could be the probable candidate genes governing these MTAs.

Grain Breadth After Cooking (GBAC)

A total of two MTAs viz., qGBAC_7 and qGBAC_12 were detected for trait GBAC on chromosomes 7 and 11with a PVE of 16.40% and 15.77% in 2019 and 2020, respectively. The genes OsGH9B17and OsUPL2 might be the candidate genes for the trait GBAC.

Linear Elongation Ratio (LER)

For trait LER three MTAs were identified on chromosomes 1, 3, and 10 with PVE range from 15.93–18.76%, in both years. The candidate genes OsVQ4, OsZIP33, and OsPME29 were predicted under these MTAs.

Breadth Elongation Ratio (BER)

For trait breadth elongation ratio (BER), three MTAs were identified on chromosomes 1 and 4, with a PVE range of 12.81- 14.42% and commonly identified in both 2019 and 2020 years. The genes OsPOP2, POP2, WRKY gene37, OsWAK53, OsWAK52, Os-FLS2, OsFbox232, Os_F0576, OsCtb, and NAD-dependent epimerase genes were predicted to be the most probable candidate genes for the trait BER.

Alkali Spreading Value (ASV)

For ASV trait, a total of 6 MTAs was identified with a PVE range from 24.62% to 27.34% during 2019 and 2020. All MTAs were distributed on chromosomes 2, 5, and 6. The annotated genes du3, OsCBP20, OsATL12, and OsGH9B9 were the probable candidate genes in the MTA regions governing alkali spreading value.

Gel Consistency (GC)

A total of five MTAs was identified for trait GC with a PVE range from 11.27 to 14.22 in 2019 and 2020. Among them, one MTA, qGC_1 was identified in 2019 while the remaining four MTAs were identified in both years. All MTAs were distributed on chromosomes 1, 4, 6, and 9. In all six candidate genes viz., OsGH9B4, Os2ODD10, OsPUB70, OsSLD, OsWAK86, and OsISC1 were prioritized be the candidate genes for trait gel consistency.

Plant Height (PH)

For trait PH trait, three MTAs were identified in both years, with a PVE range of 20.26–23.57% and all are found to be distributed on chromosomes 1, 3, and 8. The candidate genes OsTRXh10, OsNRT1.7, OsEno2, and cytochrome P450 were predicted in the MTAs controlling PH.

Number of Panicles Per Plant (NP)

For trait NP, four MTAs were identified with a PVE range from 10.02%−11.83%. Among them, two MTAs viz., qNP_2, and qNP_9.2/qNP_9 were identified commonly in both years, one QTL qNP_9.1 was identified in 2019, and the remaining one qNP_8 was detected in only 2020. All MTAs were distributed on chromosomes 2, 8, and 9. The candidate genes HSfB1, OsACO2, AP2/EREBP 109, and OsTET12 were predicted for the MTAs controlling the NP.

Number of Tillers Per Plant (NT)

A total of six MTAs was identified for trait NT in both years, with a PVE range from 8.11%−10.33%. Among them, one MTA, qNT_2 was identified in both the years, two MTAs viz., qNT_9 and qNT_11 were identified in 2019, and the remaining three MTAs viz., qNT_8, qNT_9, and qNT_12 identified in 2020. All the MTAs were distributed on chromosomes 2, 8, 9, 11, and 12. The annotated genes NAD-dependent epimerase, Hwi1, AP2/EREBP109, OsTET12, HSfB1, Amy3A/B/C, and OsPPS are predicted as the probable candidate genes for MTAs of this trait.

Panicle Length (PL)

Seven MTAs were identified for trait PL. Among them, five MTAs were stably identified in both years with a PVE values range of 16.86%−19.50%, while two MTAs, qPL_9 and qPL_9 were identified with PVE values of 18.66%, and 18.98% in 2019 and 2020, respectively. All MTAs were distributed on chromosomes 1, 3, 7, and 9. The genes LOGL1 and OsGASR9 were predicted to be the candidate genes for these MTAs of the trait PL.

Number of Filled Grains Per Panicle (NFG)

For this trait, five MTAs were identified on chromosomes 1, 4, 6, and 8. Among them, three MTAs were identified in both years with PVE ranging from 9.25%−11.40% while the remaining two MTAs viz., qNFG_8.1 and qNFG_1 with PVE values of 10.27%, and 11.17% identified in 2019 and 2020, respectively. The prioritized genes within these QTLs are OsAT10, OsMB15, OsSIG1, and OsIDD11 for the trait NFG.

Number of grains per panicle (GP)

For the trait GP, five MTAs were identified in both years with a PVE range of 8.90%−10.94% and are distributed on chromosomes 1, 4, 10, 11, and 12. The annotated genes, OsLBD1-8, OsCIPK11, OsSnRK3.7, OsFbox203, OsDRP3B OsGL1-11, OsLAC26, and OsGRF7 could be the most probable candidate gene for the MTA regions governing the number of grains per panicle.

Number of chaffy grains per panicle (NCG)

A total of five MTAs was detected for trait NCG with a PVE range of 12.38%−13.36%. Among them, three MTAs viz., qNCG_5.1, qNCG_5.2, and qNCG_10 were stably identified in both 2019 and 2020. The remaining two QTLs viz., qNCG_8 and qNCG_12 were detected in only 2019. All these QTLs were distributed on chromosomes 5, 8, 10, and 12. The annotated genes, OsGELP63, OsGELP64, OsERF042, AP2/EREBP048, OsDREB2B, OsXTH25, OsPLGG2, OsLTPG22 and OsNRTP1 could be the probable candidate genes for the trait NCG.

Spikelet Fertility (SF)

Five MTAs for trait SF were identified with a PVE range of 11.19%−13.36%. The MTA, qSF_10 was stably identified in both years, and qSF_2 was identified in 2019, whereas the remaining three MTAs were identified in 2020 only. All the MTAs were distributed on chromosomes 2, 8, 11, and 12. The annotated genes, OsPXLG4, OsXTH27, OsCDPK5, OsRLCK83, OsMADS6, and OsPIN1B/ OsPIN1b/OsPIN1c were predicted as candidate genes for trait SF.

Single Plant Yield (SPY)

A total of eleven MTAs for trait SPY were identified. Among them, eight MTAs were found stable in both years with a PVE range from 13.44%−22.35%. Whereas the remaining three MTAs were identified in 2020 with PVE ranging from 14.79%−15.90%. All the MTAs were distributed on chromosomes 1, 3, 4, 7, 8, 9, 10, and 11. The annotated genes, OsCML31, OsMSR2, OsIMP-1, OsNAS3, OsSAUR33, OsGSA, OsGSAT, OsGSAM, OsILI7, OsBC1, and OsTSCD11 were predicted as the candidate genes for trait SPY.

Discussion

Rice is continuously threatened by several pests and diseases, which reduce productivity levels substantially. Among major diseases, leaf blast and bacterial leaf blight are most devastating and reported to cause as much as 10–30% yield loss in rice growing areas. Additionally, consumer preferences for eating and cooking quality (ECQ) traits also been changing continuously worldwide. Rice has tremendous variability in the germplasm and so far, more than 1200 varieties have been released from India for various ecosystems. Interestingly, close observation of the rice varieties, it is evident that the three types of antagonistic relationships are obvious such as “higher yield with disease resistance”, “good quality with higher yield”, and “good quality with disease resistance”. However, it is largely unclear the rationale behind this antagonistic relationship of the traits in rice germplasm. Several QTLs and candidate genes are known to control the traits related to yield, quality and resistance to blast and blight diseases in rice. In the present study, we aimed to understand the intrinsic relationship among yield, quality, and disease resistance traits employing Genome-Wide Association Mapping (GWAS) in rice.

Evaluation of rice genotypes for yield, quality and disease resistance

The phenotypic screening of 230 rice genotypes for yield, quality traits along with blast and BLB disease resistance revealed wide variation in the germplasm, which could be a potential source of donors in rice breeding. Screening of rice genotypes for blast and blight resistance revealed different levels of resistance ranging from resistant to highly susceptible in both 2019 and 2020 years. This result indicates that the experimental setup was robust and the location chosen was suitable for the evaluation of genotypes for disease incidence. Among the genotypes screened for blast and BLB disease reaction, 138 and 163 genotypes, respectively showed consistent results in both years. Among them, 112 genotypes have shown resistant to blast while 80 for BLB resisatnce. These consistently shown resistant genotypes with high yielding and excellent cooking quality traits can be used as potential donors for breeding disease-resistant varieties. The remaining 92 and 67 genotypes showed variable results for blast and BLB disease reactions, respectively in both years which might be due to environmental conditions (Suh et al. 2009).

The observed correlations provide important insights for rice breeding programs targeting both agronomic and resistance traits across genetically diverse populations. The consistent positive correlations among yield-contributing traits in Indica and Aromatic groups across two years suggest strong genetic linkages or pleiotropy that can be effectively utilized in selection programs (Falconer and Mackay 1996). These populations, already known for their agronomic superiority and market value, offer reliable material for breeding programs focused on yield improvement.

In contrast, the Japonica group, although showing generally positive correlations, exhibited sensitivity to environmental variation, as evidenced by changes in correlation patterns across years. This aligns with previous findings that Japonica genotypes tend to exhibit genotype × environment (G × E) interactions more strongly than Indica varieties (Zeng et al. 2017).

The highly variable and sometimes inverse correlations seen in Landraces and Wild Relatives highlight the genetic complexity and adaptive divergence of these groups. Their trait associations may be influenced by unknown or interacting loci, environmental cues, or pleiotropic effects, which has also been reported in earlier studies exploring traditional germplasm (Courtois et al. 2012). While these populations may pose challenges in breeding due to their instability, they represent a rich reservoir of untapped alleles for stress tolerance, resilience, and nutritional traits.

Disease resistance traits (BLB and blast) were generally weakly correlated with yield traits, and often negatively associated, especially in high-yielding backgrounds. This trade-off has been observed in several studies, where resistance gene introgression can sometimes result in yield drag if not carefully managed through marker-assisted backcrossing or genomic selection (Collard and Mackill 2008). The NERICA & Aus group, with moderate but stable correlations, represents a promising genetic pool for combining adaptive traits with acceptable agronomic performance, particularly in stress-prone and rainfed regions (Jones et al. 1997).

In summary, the results underscore the importance of tailoring breeding strategies to the specific genetic background and trait architecture of target populations. Stable correlations in elite groups like Indica and Aromatic are suitable for straightforward selection, while genetically diverse groups like Landraces and Wild Relatives demand more sophisticated approaches to capture their beneficial alleles without compromising agronomic performance.

GWAS for yield, quality, and disease-resistance traits

Even though the association mapping panel used in the present study consisting of 142 genotypes was small compared to sample sizes used in other GWAS studies (Xiao et al. 2021; Qiu et al 2015), we could successfully detect previously reported as well as novel QTLs for several important traits, suggesting the association mapping panel used is ideal for GWAS studies. In the present study, as many as 78 common MTAs for the targeted traits were identified in both years and majority of them are novel. Usually, major MTAs can be identified across two years and, therefore, are considered stable. These stable MTAs of promise can be straight away transferred to elite varieties through MAS to enhance blast disease, quality, and yield traits. Further, the stable MTAs can also be targeted for uncovering candidate genes employing various fine mapping and map-based cloning strategies. The novel MTAs are future targets for functional genomics studies.

Prediction of candidate genes underlying MTAs

Attempt to predict the candidate genes underlying 78 common MTAs linked to various traits such as blast and BLB disease resistance, quality, and yield attributing traits employing publicly available rice genome databases revealed several important candidate genes (Table S6). For instance, one candidate gene i.e., D-type cyclin 3;1 (CycD3;1) (LOC_Os09g02360) that accelerates cell division has been predicted for a pleiotropic MTA, UChr9_924504 that controls both grain length (qGL_9) and grain length after cooking (qGLAC_9). This finding was confirmed by Qi et al (2012) where they demonstrated that GL3.1, encodes a protein phosphatase kelch (PPKL) family-Ser/Thr phosphatase found to be interact with Cyclin-T1;3 by dephosphorylation. Further, they demonstrated that the downregulation of Cyclin-T1;3 in rice resulted in a shorter grain. Recently, it was demonstrated that D-type cyclin OsCYCD3;1 is involved in the maintenance of meristem activity to regulate branch formation in rice (Ohyama et al. 2022). Another pleiotropic MTA, UChr10_23038717 that controls two grain size traits i.e., qGL_10 and qLBR_10 found to have a predicted gene Xyloglucan endotransglucosylase 25 (OsXTH25) (LOC_Os10g42670), whose role in controlling the grain size is yet to investigate. Interestingly, one cluster on chromosome 6 that harbours four MTAs viz., qASV_6.1, qASV_6.2, qASV_6.3, and qASV_6.4 is found to be very closely associated with already a cloned gene for ASV i.e., soluble starch synthase (SSIIa) (LOC_Os06g12450). These predicted candidate genes have immense value for further use in rice improvement after functional validation.

In this study, we identified two pleiotropic candidate genes—D-type cyclin 3;1 and Xyloglucan endotransglucosylase/hydrolase (XTH)—associated with grain size traits. While these genes are functionally important for cell division and cell wall remodeling, respectively, their influence appears restricted to yield-related traits (Hu et al. 2017). We did not observe strong pleiotropic MTAs that simultaneously regulated disease resistance, yield, and quality traits. This suggests that the trade-offs among these complex traits in our panel may not be due to shared pleiotropic loci but could instead result from tightly linked genomic regions or downstream regulatory networks. Although a few MTAs for different traits were located in relatively close proximity, distinguishing between pleiotropy and tight linkage remains challenging without high-resolution mapping or functional validation.The current findings highlight the need for integrated multi-trait GWAS approaches and systems biology analyses, such as gene co-expression networks or eQTL mapping, to better capture the complex genetic architecture underlying potential trade-offs (Yan et al. 2021; Korte and Farlow 2013). Future research incorporating transcriptomic, epigenetic, and phenomic data will be critical to unravel the interactions among resistance, yield, and quality traits and to guide more precise trait pyramiding strategies in rice breeding programs.

Genetic analysis of trade-offs between targeted traits

The rice varieties with high-yielding and disease-resistant varieties coupled with good quality are of great demand from farmers, traders and consumers point of view. However, these traits are found to be controlled by multiple genes distributed throughout the genome. Conventional breeding approaches have become successful in developing high yielding, disease resistance with appreciable quality, albeit, in isolation. Due to the difficulty in phenotyping of disease resistance and quality compared to yield traits, a harmonious combination of these traits has become challenging through conventional breeding as they have significant trade-offs among these traits. If we try to enhance one trait, the other traits have to be compromised. To this end, it is very essential to understand the intricacies of the trade-offs among these traits. In the present investigation, to gain more insights into the trade-offs, all the rice genotypes have been classified based on the number of desirable traits possessing.

The classification of rice genotypes based on their optimum trait combinations revealed that very small number of genotypes i.e., nine genotypes (KANCHAN, PUSA44, GANDHI BIRION, SURAKSHA, WGL163, NLR145, WGL915, SWARNASUB1A and SIRI1253) with 5-trait combinations while only two genotypes (Kanchan and Pusa 44) with all 6-trait combinations consistently. This clearly suggest that the presence of obvious trade-offs among these complex traits, which throw challenge to breeders to obtain optimum trait combinations for the targeted traits using conventional breeding methods. However, the genotypes with 5-trait and 6-trait combinations can be used as donor sources in rice breeding for improvement of multiple desirable traits simultaneously. Including these rice genotypes with multiple beneficial traits in rice breeding, the linkage drag that is usually associated with while transferring multiple traits can be effectively reduced.

Previously also numerous trade-offs were uncovered between several pairs of key traits, especially between growth-defence, source-sink, yield-quality etc., in diverse crops (Dwivedi et al. 2021). For instance, Descalsota-Empleo et al. (2019) uncovered QTLs for grain nutritional traits and BLB employing GWAS in rice. Further, Xiao et al. (2021) also investigated the association of yield, quality and blast in japonica cultivars. In the present study, we attempted to understand the trade-offs among different trait combinations employing GWAS. Reportedly, the trade-offs among different traits can be attributed to either pleiotropic or tightly linked genes on chromosomes or due to genotype and environment interactions (GXE) (Dwivedi et al. 2021). In the present study, we identified 24 MTA clusters for various yield, quality, and disease-resistance traits. Strikingly, most of the traits in each MTA cluster have positive correlations. Therefore, all these MTA clusters can be used for the simultaneous improvement of traits in rice breeding. For instance, one MTA cluster on chromosome 6 that governs both blast resistance and the number of filled grain grains has expressed positively to the trait expression. In contrast, Li et al. (2023) through their GWAS study in Soybean, they have reported a negative association of traits, especially the increased frequency of superior alleles controlling yield-related traits in four loci while diminished alleles controlling abiotic stress tolerance in the same loci. However, in agreement with the present investigation, Shamsunnaher et al. (2020) demonstrated that the expression of rice immune gene Xa21 confers drought tolerance. Further, knockout mutants of OsPQT3 displayed enhanced resistance to salt stress and yield (Alfatih et al. 2020). These results reinforce the fact that GWAS is one of the excellent approaches besides genetic engineering approaches to understand the trade-offs among complex traits.

Although, theoretically, there is a possibility of seven trait combinations such as YLD, QLT, DR, YLD + QLT, YLD + DR, DR + QLT and YLD + QLT + DR, strikingly, we observed only four trait combinations i.e., YLD, QLT, YLD + QLT and DR + YLD in our GWAS analysis. These findings offer a great possibility to combine diverse complex traits in a customized manner by the breeders. Only antagonistic pleiotropic genes are difficult to handle in the breeding. However, we observed 4 pleiotropic MTAs for positively correlated traits such as YLD and QLT separately. However, previously different antagonistic and positive trade-offs have been reported for disease traits. For instance, a mutation in the SPL28 gene showed enhanced resistance to blast with a reduction in yield (Qiao et al. 2010). In contrast, it was demonstrated that paired NBS-LRR receptor, pigmR and pigms confer blast resistance without yield penalty via an epigenetic regulatory mechanism (Deng et al. 2017). Likewise, IPA1 was found to improve both grain yield and blast resistance (Wang et al. 2018).

We acknowledge that relying solely on single SNP alleles for genotype–phenotype associations can be limiting, particularly in the context of population structure. To address this, our GWAS and correlation analyses focused primarily on a subset of 142 genotypes predominantly of indica background. While haplotype-based analyses were not performed in this study, we agree that such approaches combining multiple SNPs can offer more robust insights, particularly in resolving the complex genetic architecture underlying agronomic traits (Zhao et al. 2011; Huang et al. 2012; Yan et al. 2021). We intend to incorporate haplotype- and subpopulation-specific analyses in future investigations to refine our understanding of trait correlations and genetic trade-offs.In addition, we recognize that population structure can significantly influence trait correlations due to underlying genetic stratification (Zhao et al. 2011; Huang et al. 2012). Although our study began with 230 genotypes, the GWAS and correlation analyses were conducted on a filtered subset of 142 indica-dominant genotypes to reduce such confounding. The STRUCTURE analysis clearly divided the panel into four subpopulations, but due to the limited sample size within each group, subpopulation-specific correlation analysis was not performed. Future studies involving larger and more balanced subpopulation sizes will aim to investigate trait correlations within subgroups, which could uncover finer-scale trade-offs that might otherwise be masked in whole-panel analysis (Yano et al. 2016).

Additionally, we acknowledge that the present study focused on individual-locus associations, while interactions across different loci (epistasis) can also play a vital role in shaping phenotypic trade-offs, including outcomes like sterility, reduced vigor, and hybrid breakdown. These multilocus interactions may modulate or buffer the effects of individual SNPs, potentially leading to complex trait expression patterns (Carlborg and Haley 2004; Huang et al. 2010; Bian et al. 2020). Due to the scope and design of our current GWAS, epistatic interactions were not explicitly modeled. However, we recognize the value of epistasis analysis, particularly in the context of developing multi-trait resilient rice varieties, and aim to incorporate such analyses in our future research to provide a more comprehensive genetic framework for breeding.

Sometimes, screening of diverse rice varieties by whole-genome level reveals large genetic linkage drags on various chromosomes that might cause the inheritance of desirable and undesirable traits simultaneously without recombination. For instance, the large genetic linkage drags on chromosomes 1, 3, 6, and 9 were observed in japonica varieties (Xiao et al. 2021). These large genomic drags have been found to harbor the genes governing multiple traits such as yield (TGW6, SPX1, and GW6a), blast-resistant gene (Piz-t/Pigm), alkali spreading value gene, Alk and heading date gene, Hd1 (Xiao et al. 2021). In contrast, in the present study, using indica cultivars, we have identified six linkage drag blocks on chromosomes 2, 7, 8 and 11 (Fig. S7). However, there are no significant MTAs were identified in these LD blocks which opens the tremendous possibility of combining the favourable alleles into a single elite variety background in indica varieties.

Our analysis revealed no significant large linkage drag blocks associated with major MTAs for the targeted traits, suggesting that selection in these regions is unlikely to bring along undesirable alleles. This observation implies a reduced potential for genetic trade-offs during trait improvement. Notably, this pattern was more evident in indica rice genotypes, where MTAs were generally discrete and trait-specific. In contrast, japonica genotypes, as reported in previous studies, often exhibit broader linkage blocks that contribute to co-selection of undesirable traits, resulting in phenotypic trade-offs during breeding (Zhao et al. 2011; Huang et al. 2010). The relatively narrow linkage patterns and absence of extensive drag regions in indica lines may, therefore, offer a more favorable genetic architecture for simultaneous improvement of multiple traits without compromising performance. This reinforces the breeding value of indica germplasm for developing climate-resilient and high-yielding cultivars.

Strategies to minimize the genetic trade-offs

With the advances in genomic tools, now it is possible to minimize the unbreakable trade-offs between the key traits, which is otherwise not possible through conventional breeding methods. For instance, exploiting genomic approaches, Xiao et al. (2021) introduced the superior alleles of quality and blast resistance into a high-yielding variety background and developed two elite lines i.e.., XY99 and JXY1, which show high yield and blast resistance with ECQ. In addition, we can also create the favourable allele combinations employing a multi-gene pyramiding strategy as demonstrated by Wang et al (2022) for improvement of quality by developing near-isogenic lines with three important quality genes i.e., Chalk7, GS3, and Chalk5 without the yield penalty in rice. In the present study, by employing genotypes with the most favourable alleles viz., BPT4358, BPT2613, BPT2270, and JGL11470, in rice breeding, it is possible to develop cultivars with harmonious allele combinations for disease resistance, high yield, and ECQ traits. The majority of the MTA clusters observed in the current study belong to the YLD + QLT category and they are majorly due to the tightly linked nature rather than the pleiotropy which leaves enough room to exploit these MTAs in future rice breeding. Even the MTA clusters due to the pleiotropy also can be uncoupled by employing gene editing through CRISPR. Editing the cis-regulatory region in IPA1 (Ideal Plant Architecture 1) through CRISPR–Cas9 was demonstrated to resolve the tradeoff between grains per panicle and tiller number in rice (Song et al 2022). These reports, reiterate the fact that the adequate understanding of trade-offs offer great insights to develop rice varieties with multiple desirable traits.

Conclusions

Genome-wide association study (GWAS) is a powerful approach to precisely map genomic regions governing complex traits. In the present investigation, we uncovered several genomic regions associated with key quality and yield traits, as well as resistance to blast and bacterial leaf blight (BLB). Importantly, we identified a number of novel and stable MTAs that can be directly utilized in breeding programs aimed at developing high-yielding rice varieties with consumer-preferred cooking quality and enhanced disease resistance. The absence of significant large linkage drag blocks associated with major MTAs suggests a lower likelihood of unfavorable allele co-selection, particularly in indica genotypes. This finding indicates fewer genetic trade-offs and offers a more favorable breeding landscape compared to japonica backgrounds. These results provide a clearer understanding of the genetic architecture and compatibility of trait stacking in rice improvement.

The high-density SNP dataset generated in this study also offers valuable resources for germplasm characterization and genomic selection. Furthermore, the predicted candidate genes linked to novel MTAs warrant in-depth functional validation, potentially accelerating gene discovery for key agronomic traits. Additionally, the rice genotypes identified with blast and BLB resistance can serve as important donors for resistance breeding and as genetic materials for identifying novel QTLs.

Taken together, this study provides important insights into the genetic basis of complex trait regulation and trade-offs, offering practical tools and strategies for the development of climate-resilient, high-yielding, and disease-resistant rice cultivars.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

LRV conceived the experiment and drafted the manuscript. VNM and MP conducted the experiments. MK, IK, SC, and NK assisted in field work. SG, LK, SL, NC, VBRL, SNU, and SA have carried out data analysis.

Data availability

All additional data is attached to the supplementary file.

Declarations

Ethics statement

This is an observational study. The Research Ethics Committee of the University has confirmed that no ethical approval is required.

Consent to participate

Informed consent was obtained from all individual participants included in the study.

Consent to publish

As there is no personal data of the authors in the article, no ethical approval is required.

Conflict of interest

Authors declare that they do not have any conflict of interest.