Introduction

The global need for food and nutritional security to end hunger and hidden hunger has been widely recognized1,2,3,4. Rice is the main source of calories in many South and Southeast Asian developing countries; however, milled rice is less nutritious5,6. Developing high-yielding and nutrient-dense rice varieties is crucial in achieving food and nutritional security7,8,9. Biofortified food crops can help alleviate malnutrition10,11,12,13. Recently, a concerted effort has been done to biofortify food crops with various vitamins and minerals14. Zinc (Zn), iron (Fe), and vitamin A deficiencies affect more than three billion people and can cause anemia, stunting, weakened immunity, blindness, diarrhea, and other health problems. Therefore, these nutrients have been prioritized for biofortification of food crops10,15,16.

The availability of genetic variation for grain Zn content in rice germplasm, particularly in landraces, pigmented rice, wild rice, and aus subgroup, has led to a significant focus on developing Zn-biofortified rice varieties7,17,18. The HarvestPlus and IRRI biofortification teams, along with national partners, are leading genetic biofortification efforts for rice. The first wave of high-Zn rice varieties has been successfully developed and introduced in South and Southeast Asia, and several promising high-Zn rice lines have been evaluated in different Asian and African countries2.

Developing Zn-biofortified rice varieties with higher yield and other desirable traits is challenging for rice breeders because of the genetic complexity of grain yield and grain Zn content, which are greatly influenced by the environmental factors19,20. Although there is often a negative correlation between grain Zn content and yield, a few specific germplasm accessions and populations have shown either no association or a negligible correlation between these two traits paving the way for successful Zn biofortification21,22,23.

Numerous Quantitative Trait Loci (QTL) mapping studies have been conducted in rice using different mapping populations8,17,18,20,24,25,26,27,28,29. Recent developments in high-throughput sequencing and robust statistical analysis have enabled the breeders to uncover more detailed information about the genetic basis of complex traits. Molecular markers have been successfully used by the breeders to develop high-yielding and nutritious rice varieties30,31,32,33. Breeding programs are transitioning from trait-based breeding approaches to modernized product development pipelines to accelerate varietal replacement23. It effectively integrates all the traits necessary for market acceptance into breeding products. As a result, the selection for high grain Zn content is being integrated into elite breeding pool to ensure the constant delivery of sufficient Zn to the diets of nutrition-vulnerable populations.

Multi-parent populations (MPPs) are important resources for exploring plant genetic architecture and identifying QTL because of their superior control over population structure, ability to capture greater genetic diversity, and decreased likelihood of false-positive QTL discovery34,35. The QTL effects in MPPs may exhibit varying degrees of allelic diversity, which affects QTL identification, but larger crossing/breeding combinations can boost QTL detection power despite having fewer parents36. Overall, MPPs have become increasingly popular in QTL mapping.

In this study, we analyzed phenotypic variation, correlations, and principal components of yield, agronomic traits, and grain Zn and Fe contents in two recombinant inbred line (RILs) populations of rice. We identified QTLs for grain yield, agronomic, and nutritional traits using multi-parent population analysis, interval mapping (IM) and inclusive composite interval mapping (ICIM). We also identified RILs with high grain Zn content and acceptable yield potential that can serve as parental lines in rice biofortification programs.

Results

Phenotypic performance

Most traits showed differences among parents (Table 1). Kaliboro had higher values for PH, NT, NP, TGW, GL, GW, and Zn than the other parents across the seasons. Similarly, wider variations and normal distributions were observed for all traits in both populations across the seasons (Fig. 1). DF, PH, TGW, GL, and Zn showed higher heritability, whereas Fe consistently showed lower heritability in both populations. The RILs exhibit transgressive segregation on both sides of the bell curve. The analysis of variance revealed significant genotype by seasonal effects for all traits in both populations, except for DF, PH, NT, Fe and Zn in P1 (Table 1).

Table 1 Phenotypic variation for agronomic, yield, and grain Fe and Zn concentration in two RIL populations.
Figure 1
figure 1

Frequency distributions of agronomic, yield, Fe, and Zn in the P1 and P2 RIL populations of rice evaluated during the 2017 DS, 2017 WS, and 2018 DS. DF days to flowering (days), PH plant height (cm), NT number of tillers, NP number of panicles, PL panicle length (cm), YLD yield (kgha-1), TGW thousand grain weight (g), GL grain length (mm), GW grain width (mm), Fe iron (ppm), Zn zinc (ppm).

Correlation and principal component analysis

Out of 55 possible correlations in each population per season, the number of significant positive correlations was higher than the number of significant negative correlations (Fig. 2). The grain Zn content was consistently and positively correlated with PH, GW, and Fe, while it was negatively correlated with DF, GL, and YLD. Similarly, NT and NP, PL and PH, GW and TGW, GL and PL, and PL and YLD were positively correlated, whereas GL and GW were negatively correlated across the seasons and populations.

Figure 2
figure 2

Correlations among agronomic, yield, Fe, and Zn in the P1 and P2 RIL populations of rice during the 2017 DS, 2017 WS, and 2018 DS. Positive correlations are indicated by dark purple color, while negative correlations are indicated by light purple color. Bigger circles correspond to stronger correlation coefficients. DF days to flowering (days), PH plant height (cm), NT number of tillers, NP number of panicles, PL panicle length (cm), YLD yield (kgha-1), TGW thousand grain weight (g), GL grain length (mm), GW grain width (mm), Fe iron (ppm), Zn zinc (ppm), *p < 0.05, **p < 0.01, ***p < 0.001.

The first two principal components (PC1 and PC2) explained the maximum total variance of 48.31% and 46.96% across seasons in P1 and P2, respectively (Table S1; Fig. 3). DF, NT, NP, GL, GW YLD, and Zn were the major contributors to the total PC1 variance (21–26%) in P1. Similarly, NT, NP, PL, GL, GW, YLD, Fe, and Zn contributed to the PC1 variance (25–29%) in P2. NT, NP, and YLD consistently showed high factor loading (0.4–0.6) across seasons in both populations.

Figure 3
figure 3

Principal component analysis using the key variance components (PC1 and PC2) for yield, agronomic traits, Fe, and Zn of P1 and P2 RIL populations grown in 2017 DS, 2017 WS and 2018 DS. Purple dots inside the box correspond to the RILs of P1 and P2. A small angle between traits (colored in red) implies positive correlation, a large one suggests negative correlation, and a 90-degree angle indicates no correlation between traits. Traits with a high contribution to the population variance are in longer vector lines whereas traits with low contribution are in shorter lines. DF days to flowering (days), PH plant height (cm), NT number of tillers, NP number of panicles, PL panicle length (cm), YLD yield (kgha-1), TGW thousand grain weight (g), GL grain length (mm), GW grain width (mm), Fe iron (ppm), Zn zinc (ppm).

Identification of RILs with higher YLD and Zn

We identified the top five RILs in each population with higher YLD and Zn content (Table 2). In P1, among the top five RILs, Zn ranged from 12.9 to 27.3 ppm. IR120447-B-2-4-2 had the highest YLD of 8511 kgha−1 in the 2017 DS, with Zn of 21.8 ppm in 17DS, while IR120447-B-3-6-4 had the highest Zn of 27.3 and 26.8 ppm in the 2017 and 2018 DS, respectively, with a YLD of 5747.0 kgha−1 recorded in 17DS comparable with check variety IRRI 156. IR120447-B-2-5-2 had the DF of 80 days with a YLD of 5625.0 kgha−1 obtained in 18DS and Zn content of 20.3 ppm in 17DS. Meanwhile, IR120447-B-6-6-1 had the PH of 83.6 cm average across season and a Zn content of 21.6 ppm observed in 17DS. All of the top RILs in P1 possessed alleles that increased YLD, Zn, and Fe, with IR 120447-B-3–6-4 having the highest number of YLD (qYLD4.2, qYLD4.3, qYLD4.4, qYLD6.1, qYLD6.2, qYLD7.1) and Zn (qZn2.1, qZn5.1, qZn5.2, qZn10.1) QTLs but only one for Fe (qFe6.2), whereas the others carried two Fe QTLs. In P2, among selected lines, Zn ranged from 16.5–28.6 ppm and YLD from 3398.3–6541.4 kgha−1. IR120455-B-19-1-1 had the highest YLD average of 5361 kgha–1 across season with Zn from 18.4 ppm in 17WS to 22.6 ppm in 17DS, while IR120455-B-1-6-1 had the highest Zn content of 28.6 ppm in 17DS and YLD of 6388.1 kgha-1 obtained in 17DS. Moreover, IR 120455-B-17-5-1 had the DF of 81.4 average across season with high Zn of 22.3 ppm in 17DS. The YLD of all RILs in P1 like IR 120447-B-2-2-1, IR 120447-B-2-4-2, and IR 120447-B-2-5-2 exhibited higher YLD values compared to IRRI 156 the high yielding check in the 17DS and 18DS. However, in P2, the YLD of the 5 RILs was more consistent across seasons, with IRRI 156 performing relatively well. Three of the RILs in P2 carried combined alleles for YLD, Zn, and Fe with IR 120455-B-1-6-1 containing the greatest number of QTLs (qYLD4.4, qYLD6.2, qZn1.1, qZn2.1, qZn4.1, qZn4.2, qZn5.1, qZn6.3, qFe1.1). These RILs can be a good genetic resource for biofortification breeding programs.

Table 2 Yield, agronomic traits, and Zn concentrations of the parents and top ten high-Zn rice RILs.

Construction of linkage map

We processed the 11,609 SNPs data by removing redundant SNPs and those with a high missing rate (> 30%). In all 2018 and 2213 SNPs were obtained for P1 and P2, respectively (Table S2). The highest and lowest numbers of SNPs were on chromosomes 2 (274) and 9 (89) for P1 and on chromosomes 4 (328) and 10 (70) for P2, respectively. The total lengths of the linkage maps were 1232.2 cM for P1 and 1071.5 cM for P2. The average marker intervals were 0.61 cM and 0.48 cM in P1 and P2, respectively.

QTLs identified using ICIM and IM analysis

Seventy-five QTLs were identified for yield and agronomic traits using either IM or ICIM, and 22 of these QTLs were detected in both the analyses (Table 3; Fig. 4). The PVE of the QTL varied from 5% to 16.6% across the populations. There were nine QTLs for DF on chromosomes 2, 4, 6, 10, and 12, which explained a PV of 5.2–13.6%. The qDF2.2, contributed by IR95040, had the highest PVE (13.6%). Among the 15 QTLs detected for PH on chromosomes 1, 2, 4, 6, 7, 8, and 10, qPH1.2, was consistently identified on chromosome 1 (39.6–40.1 Mb), and the Kaliboro allele at this locus increased PH by 13.2–15.9 cm. Seven QTLs were mapped for NT on chromosomes 3, 4, 5, and 7. The qNT5.1 detected at 5.5–5.9 Mb on chromosome 5 explained the highest PV (16.6%) in P1. Likewise, eight QTLs were uncovered for NP on chromosomes 2, 3, 4, and 7, which explained a PV of 5–9.3%. The qNP7.3, explained 9.3% of the PV in P1. Eleven QTLs were identified for PL on chromosomes 2, 3, 4, 6, 7, and 8, explaining 5.8–13.5% PV. The qPL3.1, had the highest PVE and was contributed by IR95097. qPL4.2, was consistently mapped on chromosome 4 (20.6–20.8 Mb) during both the DS and by both analyses (ICIM and IM). Five QTLs were identified for YLD on chromosomes 4, 6, and 7, with a PVE ranging from 5% to 14.8%. It is interesting to note that a large effect QTL for yield, qYLD4.1 (PVE > 10%) was detected on chromosome 4 (12.6–13.3 Mb) in P2. Four QTLs were detected for TGW on chromosomes 1, 4, 6, and 12, explaining a PV of 5.1–10.7%. The qTGW1.1 had the highest effect (PVE, 10.7%). Likewise, five QTLs responsible for GL located on chromosomes 2, 4, and 10 explained a PVE of 7.9 to 10.6%. Among them, qGL2.2 and qGL10.1 explained PV > 10%. A total of 11 QTLs were found for GW on chromosomes 2, 3, 4, 6, 7, and 10, with PVE that varied from 5.2 to 13.8%. Two large effect QTLs (qGW4.2 and qGW4.5) with PVE > 10% were detected on chromosome 4.

Table 3 QTLs identified for agronomic, yield, and grain Fe and Zn concentration in two RILs populations using inclusive composite interval mapping and interval mapping.
Figure 4
figure 4

Genomic visualization of QTLs identified for yield, agronomic traits, Fe, and Zn in RIL populations using ICIM and IM analyses. The outer circle represents the rice linkage map. The middle circle represents DF, PH, PL, NT, NP, and YLD QTLs. The inner circle represents the QTLs for GL, GW, TGW, Fe, and Zn.

Five QTLs (qFe4.1, qFe6.1, qFe6.2, qFe9.1 and qFe11.1) were detected for grain Fe content on chromosomes 4, 6, 9, and 11 in P1. The Kaliboro allele at qFe4.1 and qFe6.1 increased Fe content. Thirteen QTLs were associated with Zn content on chromosomes 3, 4, 5, 6, 7, 8, 10, and 12, accounted for 5.2–18.3% of PVE. Interestingly, four QTLs were consistently detected on chromosomes 3, 7, and 10 in P1 and on chromosomes 4 and 8 in P2. The qZn3.1 located at 11.5–11.6 Mb on chromosome 3 explained 7.5% and 6% PVE during 17DS and 17WS respectively. While qZn7.3 was located between 28.9–29.1 Mb on chromosome 7 had an additive effect of 2.6 ppm in 17DS and 2.8 ppm in 18DS. The qZn10.1, detected between 19.13–19.14 Mb on chromosome 10 had the highest PVE (> 10%) and additive effect (3.1 ppm), and the trait-enhancing allele was contributed by Kaliboro. Similarly, qZn4.1 had an additive effect of > 2.4 ppm across the three seasons.

QTLs identified using multi-parent population analysis

The MPP QTL analyses detected twenty-two QTLs for yield and agronomic traits, and five QTLs for grain Zn and Fe contents (Table 4; Fig. 5). The qPH1.2 was consistently detected on chromosome 1 with a PVE > 53.5% and a P-value of 37.7. There were five QTLs for DF on chromosomes 2, 5, 7, and 9 (PVE—4.4 to 8.9%; P value -3.7 to 7). NT had one QTL on chromosome 11, whereas NP had two QTLs on chromosomes 2 and 11. Three QTLs were identified for PL on chromosomes 1, 2, and 4, their P-values ranged from 3.7 to 8.4 and PVE from 5.7 to 13%. There were two QTLs for TGW on chromosomes 3 and 5; qTGW3.1, had the highest PVE (6.6%) and P value (6.0). Among the three QTLs detected for GL, qGL3.1 explained 16.5% of the PV. Similarly, three QTL were detected for GW on chromosomes 3 and 4, and their PVE varied from 4.6 to 9.2%. Two QTLs detected for YLD on chromosomes 4 and 6 had PVE of < 10%. For grain Zn content, four QTLs were identified on chromosomes 1, 2, and 5, and the large effect QTL qZn5.2 explained 10.8% PV. There was only one QTL for Fe content on chromosome 1 with a PVE of 5.1%.

Table 4 QTLs identified for agronomic, yield, and grain Fe and Zn concentration by multiparent analysis.
Figure 5
figure 5

Genomic visualization of QTLs identified for yield, agronomic traits, Fe, and Zn in RIL populations using multiparent population analysis. The outer circle represents the rice linkage map. The middle circle represents yield and agronomic, Fe, and Zn QTLs.

Candidate genes underlying consistent QTLs

We shortlisted 53 candidate genes underlying the QTLs detected for agronomic, yield, and nutritional traits (Table 5). Eight candidate genes were associated with two YLD QTLs: OsD11, OsFCP1, DAO, DWA1, Rf17, and OsCRC1 co-located with qYLD4.4, while dwarf35 and OsMIOX were co-located with qYLD6.2. Candidate gene dh1 was associated with qDF2.3, whereas seven genes that known to be associated with plant height (SDG714, OsSSI2, OsIRX10, HDT702, lhd2, OsCKX4, sd1) were identified with qPH1.2. Six candidate genes were co-located with four QTLs such as qNT11.1 (OsARF1), qNP11.1 (OsD27), qPL1.1 (OsFBP1, OsAAP6), and qPL8.2 (WFP, OsIPI1). Similarly, OsRSR1, OsMTP1, OsVPS22, OsLOX2 (L-2), and OsERO1 were co-located with the QTLs for TWG, GL, and GW. We analyzed the expression profiles of the 12 candidate genes using RiceXPro. The 12 agronomic and yield candidate genes included OsIRX10, HDT702, lhd2, and sd1 for PH; OsARF1 for NT; OsFBP1, OsAAP6, WFP, and OsIPI1 for PL; OsRSR1, and OsMTP1 for TGW; and OsD11 for yield (Fig. 7).

Table 5 Summary of candidate genes for yield, agronomic traits, and Fe and Zn concentrations in QTLs identified.
Figure 6
figure 6

Gene networks showing the relationship between the (A) sd1 gene and other putative QTLs on chromosome 1, (B) OsNAS1 gene and other putative QTLs on chromosome 3, (C) OsSAP13 gene and other putative QTLs on chromosome 1; (D) OsRZFP34 gene and other putative QTLs on chromosome 1, (E) OsC3H33 gene and other putative QTLs on chromosome 5, (F) OsC3H37 gene and other putative QTLs on chromosome 5, (G) OsHMA1 gene and other putative QTLs on chromosome 6.

Twenty-six candidate genes were associated with QTLs detected for Zn and Fe contents. Three candidate genes (OsSAP13, OsRZFP34, and OsHMP5) were co-located with qZn1.1. The OsNAS1 and OsSET16 co-localized with qZn3.1, whereas qZn4.1 harbored OsDLN109 and OsMTP7. On chromosome 5, OsC3H33 was co-located with qZn5.1, OsGATA1, and OsC3H37, OsSET22, OsSIRP1, and OsAIR1 were co-located with qZn5.2. Likewise, Osferroportin, OsHMP32, OsHMA1, OsHMA7, OsRFPHC-6, and OsC3H53 were linked to QTLs detected for Zn on chromosomes 6 and 7. The OsHMP7 and OsRFP were co-located with qFe1.1. To comprehend the expression patterns of these genes during various developmental stages, the RiceXPro database was used to examine the expression profiles of several genes. The OsRZFP34, OsHMA1, OsMTP7, OsRFPHC-6 and OsC3H37 were over expressed in the embryo (Fig. 8), while OsNAS1, OsC3H37 and OsRFP over expressed in the endosperm. Furthermore, the ovary exhibited over-expression of several genes that included OsHMP5, OsFerroportin, OsHMA1, OsNas1, OsMTP7, OsRFPC-6, and OsC3H53.

Figure 7
figure 7

Graphical representation of the expression profiles of 12 candidate genes for agronomic and yield traits throughout rice development, based on spatiotemporal profiling across various tissues and organs, with data extracted from RiceXPro (http://ricexpro.dna.affrc.go.jp/).

Discussion

The biofortification of rice with improved grain Zn content is a priority nutritional improvement program that helps tackling Zn malnutrition37,38. The investigation of molecular mechanisms involved in Zn metal homeostasis in rice contributes to the enhancement of Zn biofortification of rice23,27,31. Recent advancements in genotyping techniques, as well as statistical and bioinformatics tools have facilitated accurate and precise QTL mapping in diverse set of populations34,39. In this study, we employed ICIM mapping and MPP analysis to conduct QTL analysis on two biparental populations with shared parentage, and identified robust QTLs associated with Zn content that can be effectively utilized for genomic-assisted breeding34,40.

We observed significant variations in the majority of traits across the three seasons in both RILs populations, with most of the traits showing a normal distribution. The use of high Zn elite breeding lines as recipient parents generates effective variations that can be directly used in pre-breeding, as opposed to utilizing highly contrasting parents, which may result in reduced agronomic performance and lead to the inclusion of undesirable alleles. These variations also imply the quantitative or polygenic inheritance of the traits8,17,31. Our study revealed moderate to high heritability values (H2) for traits such as DF, PH, PL, YLD, TGW, GL, and Zn, which is consistent with results of previous studies conducted using different populations17,18,20,31. These traits can easily be transferred to the next generation through conventional method and permits effective phenotypic selection for their improvement16,41,42. Meanwhile, traits with lower to intermediate heritability can integrate conventional and molecular techniques in improving these traits.

In both RILs populations across seasons, DF, PL, and GL showed a positive correlation with YLD which is expected since these traits are considered as yield components that directly influence the trait. In contrast, Zn exhibited a negative correlation with YLD consistently across populations and seasons that conforms with previous studies8,17,20,31,43. The unfavorable correlation between YLD and Zn presents a challenge for breeders in developing high-yielding Zn-biofortified rice varieties, as it may indicate a different mechanism at the genetic level that controls both traits, having opposite effects on their improvement. To overcome this challenge, appropriate breeding techniques and selection procedures must be applied to identify stable high-Zn donor lines with a greater or acceptable YLD7,23,38. It is essential to properly combine YLD and Zn to develop high-yielding Zn-biofortified rice varieties that address food and nutritional security in developing countries. On the other hand, the present study confirms previous findings that grain Zn exhibited both positive and negative relationships with TGW and GL, respectively, whereas Fe exhibited a negative relationship with both TGW and GL24,31, these can be attributed to the dilution effect where Fe and Zn concentration tend to be diluted as grain related traits increase. However, we also found substantial positive correlations between Zn and Fe in both populations across seasons, consistent with previous findings44,45,46,47,48. These results suggest that common molecular mechanisms, pathways, and gene networks likely underlie the uptake, translocation, and loading of these two elements in rice grains7. However, there were some differences in the deposition of Fe and Zn within the rice grains. While most of the Fe is deposited in the aleurone layer, a significant portion of Zn is loaded into the endosperm49. This variation may be useful in breeding high-Zn rice through biofortification breeding, whereas high-Fe rice is being developed using genetic engineering or genome editing approaches2,23.

Principal component analysis (PCA) is a powerful tool for summarizing multivariate data and achieving the best representation in a lower-dimensional space50. Generally, PCs with vector co-efficients greater than 0.3 are considered important regardless of their direction of influence51. Component loadings for each principal component revealed that YLD, PL, and GL were among the phenotypic traits contributing positive projections, whereas Fe and Zn contributed negatively to PCs in P1. The results revealed that these traits significantly contribute to the overall variability in the population52. The PL and GL can be considered in the selection criteria in selecting high-yielding lines. Moreover, the positive significant contributions of YLD further indicates the higher likelihood of identifying high-yielding lines in the populations. In P2, YLD showed positive projections while Fe and Zn showed negative projections in 2017DS, whereas YLD showed negative projections while Fe and Zn showed positive projections in the two PCs during 2017WS and 2018DS. These opposite projections of YLD and Zn on PCA corroborated the results of the correlation analysis. The results of PCA can be utilized to identify the traits with the greatest impact on the phenotype of RILs for line selection. Interestingly, we observed that some correlated variables, such as NT and NP, Fe and Zn, and GW, had similar PC effects (positive or negative) for each season and population.

In our study, we found more QTLs in P1 compared to its counterpart during WS. The donor parent Kaliboro contributed 71% of the positive alleles. The biparental and MPP analyses identified 93 and 27 QTLs, respectively (Tables 3, 4). The MPP analysis identified only 27 QTLs for 11 traits, three of these QTLs (qPH1.2, qPL4.2, and qZn5.1) were consistently detected across populations and seasons. It also detected two YLD QTLs (qYLD4.4 and qYLD6.2) on chromosomes 4 and 6. Where as, IM identified two QTLs on chromosomes 4 and 7 while ICIM identified five QTLs on chromosomes 4, 6, and 7 that were associated with YLD. Previous studies reported yield QTLs at the same chromosomal regions on chromosomes 4 and 7. Khahani et al. (2021) identified MQTL-YLD4 at 29.15–30.85 Mb on chromosome 4 that harbored multiple genes associated with grain yield67. In another study, Zhang et al. (2021) using a high-density genetic map of hybrid rice, revealed the presence of a consistent QTL for YLD (qGYP-6b) on chromosome 6 across different environments68. These findings highlight the need for further exploration of these specific genomic regions on chromosomes 4 and 6 to identify key regulatory genes influencing YLD in rice.

Our analysis revealed QTLs for Fe on chromosomes 1, 4, 6, 9, and 11 using various methods, which is consistent with previous studies that identified common regions on chromosomes 1, 3, and 7 in rice16,17,18,21,31,69,70,71. Specifically, we identified QTLs for Fe content on chromosome 1 using MPP, on chromosomes 4 and 6 using ICIM, and on chromosomes 9 and 11 using IM. Moreover, a recent review by Swamy et al. (2021) reported several QTL for Fe on chromosomes 1, 3, and 7, and also shortlisted candidate genes for Fe homeostasis38. These results suggest that multiple genomic regions and genes are associated with Fe accumulation in rice, which may provide valuable targets for future breeding efforts to develop Fe-biofortified rice varieties.

In MPP analysis, a consistent QTL for Zn16 was found on chromosome 5 similar to the report of Palanog et al. (2023). Moreover, numerous studies have also reported similar genomic regions making it potential loci for genomics assisted breeding for Zn biofortification24,55,59,64. Zn QTLs were also detected on chromosomes 3, 4, 5, 6, 7, 8, and 10 using ICIM and IM. Previous QTL mapping studies have identified Zn QTLs on all rice chromosomes8,17,18,19,20,27,31,43,44,55,60,61,65,72,73. However, it is important to note that these QTL mapping experiments used different parental genotypes, mapping populations, field sites, and environmental conditions. Hence, further validation is required prior to integrating the Zn QTLs into the biofortification program.

Our candidate gene analysis that were co-localized with QTLs for different agronomic traits such as DH1 for DF74, sd1, OsIRX10, and HDT702 for PH75,76,77, OsD27 for NP78, Wealthy Farmer's Panicle (WFP) and OsIPI1 for PL79,80, OsRSR1 and OsMTP1 for TGW81, and OsVPS22 for GL with proven biological and physiological functions and directly involved with the traits. With the help of KnetMiner, we were able to identify the relationship of the candidate genes with other genes proximately connected with them and regions where they are largely expressed. It appears that there are associations between the sd1 gene and 39 QTLs in the KnetMiner, as well as another gene GA20OX5, which is responsible for shortening plant height (Fig. 6A). We also identified CL4 and d35 gene co-located with YLD QTLs known for their importance in panicle development and regulation of plant height that have contributed to crop productivity82,83. These genes are potential candidates for functional validation and for introgression through MAB to develop high yielding nutritious rice varieties84,85. For instance, OsRSR1 and OsMTP1 were co-located with TGW QTLs that play an essential role in producing higher grain YLD, these genes were also involved in metal transportation within the rice plant86,87.

Figure 8
figure 8

Graphical representation of the expression profiles of 12 candidate genes for Zn and Fe throughout rice development, based on spatiotemporal profiling across various tissues and organs, with data extracted from RiceXPro (http://ricexpro.dna.affrc.go.jp/).

One of the highlights of our study is the identification of putative genes co-located with major Zn QTLs such as OsNAS1, OsSAP13, OsMTP7 OsC3H33, OsC3H37, OsSDG728/SET22, OsSIRP1, OsAIR1, OsHMA1, OsFerroportin, and OsRFP (Fig. 8). These genes are deeply involved in Zn binding and heavy metals homeostasis and have been connected to delivering Zn to developing rice seeds88 (Fig. 6B–G). They are also expressed and upregulated in various parts of rice such as roots, embryo, ovary, stem, anther, pistil, lemma, and palea at various growth stages of rice development (Fig. 8, RiceXPro). The identification of these potential genes offers valuable insights into the molecular processes involved in Zn uptake and transport within rice plants. This emphasizes the significance of understanding the genetic basis of Zn absorption and movement in rice plants, which carries implications for improving both the nutritional quality and productivity of this important staple crop.

Our study has identified promising lines with superior YLD and Zn in both populations. We selected IR 120447-B-2-4-2 with an average YLD of 5222.5 kgha-1 and Zn content of 18.9 ppm in P1 and IR 120455-B-19-1-1 with an average YLD of 5361.4 kgha-1 and Zn content of 20.3 ppm in P2. These promising lines highlight the increased chances of achieving the primary goal of biofortification breeding–to develop high-yielding Zn-biofortified varieties. These lines will be further evaluated and can be nominated for varietal release trials, these lines can also serve as parents in rice biofortification.

Conclusion

Phenotypic evaluation of RILs populations across seasons revealed wider variations for agronomic, yield and nutritional traits. QTL analyses (MPP, ICIM and IM) of multiple populations evaluated for three seasons identified consistent QTLs for agronomic, yield and biofortification traits. They were co-located with candidate genes important in the expression and regulation of these traits. We identified the top five RILs in each population with higher YLD (≥ 5 tons ha-1) and Zn content (12.9–28.6 ppm). Our findings provide breeders with valuable information on the genetic basis of agronomic, yield and Zn traits as well as SNP markers linked to candidate genes of important QTLs. Moreover, the identification of high-YLD and high-Zn rice lines will aid the rapid development of Zn biofortified rice varieties.

Materials and methods

Plant materials and phenotyping

These experiments were conducted at the Zeigler Experiment Station, International Rice Research Institute (IRRI), Los Baños, Laguna, the Philippines. The two RILs populations were derived from crosses of IR95040 x Kaliboro (IRGC 77201-1) (P1) and IR95097 × Kaliboro (IRGC 77201-1) (P2). The parental lines IR95040 and IR95097 are high-Zn breeding lines developed by the Biofortification group of IRRI, while Kaliboro is an Aus-type rice accession from IRRI Genebank with high grain Zn and Fe contents. The two populations, P1 and P2, were composed of 178 and 195 RILs, respectively. The seeds of the two populations and parental lines used in this study were obtained from the Biofortification group of IRRI. Lines were grown in an alpha-lattice design with two replications each during the 2017 dry season (DS) and wet season (WS) and the 2018 DS. Seeds were sown in a seedbed, and 21-days-old seedlings were transplanted. Standard agronomic practices and plant protection measures were applied to ensure good crop growth and complete grain development.

Data on days to 50% flowering (DF) were recorded when 50% of the plants in the plot started flowering. At maturity, three randomly selected plants from the middle of the plot were sampled to measure plant height (PH, cm) and panicle length (PL, cm). Grains were harvested from each plot, dried, weighed, and corrected to 14% moisture content for grain yield (YLD, kg ha-1). Post-harvest data, thousand-grain weight (TGW, g), grain length (GL, mm), and grain width (GW, mm), were also measured following the standard evaluation system of rice95.

Grain Zn and Fe contents were measured using an XRF-Bruker S2 Ranger. Rice samples were obtained from the bulked seeds of each plot. Samples were dehulled, and approximately 5 g of seeds were prepared for grain Zn and Fe content measurements. The samples were analyzed twice, with measurements expressed in parts per million (ppm). We identified ten promising RILs based on agronomic traits, grain yield, and Zn content for Zn biofortification breeding.

Data analysis

The means of the traits were used for statistical analysis. All basic statistical parameters were generated using STAR v. 2.0.1, whereas ANOVA and broad-sense heritability (H2) were analyzed using PBTools v. 1.4 (http://bbi.irri.org/). Histograms, principal component analysis, and Pearson’s correlation were estimated using R-Studio v. 1.2.5 (https://www.rstudio.com/).

DNA isolation and sequencing

Leaf samples were collected from 21-days-old seedlings from each of the 373 RILs and their parents (IR95040, IR95097, and Kaliboro IRGC 77201-1). DNA was isolated using a modified CTAB method115. DNA quality was assessed by electrophoresis on 1% agarose gel. Using this platform, high-quality DNA was obtained for SNP genotyping (Tea Break Bioinformatics, Limited). The ApeKI restriction enzyme was used for library preparation following a protocol modified by Elshire et al. (2011)116. For GBS, a type II restriction endonuclease (ApeKI) was used for DNA digestion, the digested DNAs was ligated to the adapter, and a 96-plex library was constructed according to the GBS protocol. High-density genotyping of all individuals was performed using genotyping-by-sequencing (GBS) with Tea Break Bioinformatics (http://tenformatics.com/index.html).

Sequence data analysis and SNP identification

The genotype sequence SNP data was analyzed using the Tassel 5 GBS pipeline117. SNP calling and filtering were performed using the Tassel pipeline117. The variant call format (VCF) was filtered using a minor allele frequency (MAF) > 0.01, call rate ≥ 95%, and the Hardy–Weinberg equilibrium (HWE) test with a threshold value of P < 0.001. Finally, 221,419 high-quality SNPs were selected for downstream analyses. The SNPs retained after filtering using the Tassel pipeline were manually filtered. SNPs with monomorphic alleles between the two parental lines, high heterozygosity, and more than 10% of missing SNP calls were excluded from further analysis. A total of 11,609 SNP markers were retained for linkage map construction in both populations.

Linkage mapping and QTL analysis

Linkage map construction and QTL analysis were performed using QTL ICIMapping v. 4.1118. The grouping of SNPs was based on their physical map locations on chromosomes. The ordering of the markers within each chromosome was based on recombination events between SNPs. The recombination distance was calculated using the Kosambi function119.

We performed QTL analyses for agronomic, yield and grain micronutrient (Zn and Fe) traits. The critical threshold value for QTL detection was calculated based on Churchill and Doerge (1994)120. SNPs associated with QTL identified by linkage mapping were anchored to the physical map. Information regarding the physical location of SNPs flanking the QTL was used to determine the position of the QTL. All QTL were visualized using the RCircos R package111. To increase the power of QTL detection and compare the effects of QTLs detected in cross-specific models, multi-parent population QTL mapping was performed using the mppR package40.

Candidate gene analysis

Candidate genes associated with agronomic traits, grain yield, grain Zn and Fe contents were analyzed across all detected QTLs. Identification of candidate genes within the QTL regions was performed using available online database: the “Rice Genome Annotation Project” website (http://rice.plantbiology.msu.edu/cgi-bin/ORFinfopage.), the “Gramene QTL Database” (http://archive.gramene.org/qtl/), and Oryzabase (https://shigen.nig.ac.jp/rice/oryzabase/quicksearch/list). Subsequently, locus IDs for these candidate genes were leveraged for trait association and gene ontology annotation. This facilitated the identification of GO terms relevant to the traits of interest. Leveraging literature-based evidence and GO, we identified top candidate genes. Additionally, we utilized Knetminer (http://knetminer.rothamsted.ac.uk/Oryza_sativa/) to construct a network mapping the relationships between candidate genes and related QTLs. The tissue-specific and developmental stage-specific expression patterns of these candidate genes were further elucidated using data from The Rice Expression Profile Database (http://ricexpro.dna.affrc.go.jp), which provided insights into gene expression throughout the rice plant’s life cycle.

This research did not involve any human participants or animals. The study adhered to applicable institutional, national, and international guidelines and legislation, ensuring that all experiments on rice (Oryza sativa L.) plants were conducted in accordance with relevant regulations. Additionally, this research was not conducted with endangered species and was carried out in accordance with the Declaration of the IUCN Policy on Research Involving Endangered Species.