A diversity set comprising 293 samples of pigmented rice was examined, including 18 purple-colored, 256 variable-purple-colored, 16 red-colored, and 3 light brown varieties ( Supplementary Table S1 ). These samples were selected from the International Rice Research Institute (IRRI) genebank, purified them through single seed dissent and were planted during the 2019 dry season at the experimental station of IRRI in Los Baños, Laguna, the Philippines by following the standard agronomic practices and irrigated conditions. From the harvested paddy samples of pigmented rice core collection, the state of dormancy was terminated by subjecting rice seeds to a temperature of 50°C for duration of five days. The germination process was initiated in accordance with the described method (Cáceres et al., 2017; Tiozon et al., 2023).

The extraction protocol and data process followed the previous method (Tiozon et al., 2024). Briefly, 50 mg of PRS samples were extracted with 1.2 mL methyl tert-butyl ether:methanol, 500 µL of the upper lipid-containing phase was dried in a speedVac concentrator and re-suspended in 250 µL acetonitrile: 2-propanol (7:3, v/v) solution. The workflow included peak detection, retention time alignment, and removal of chemical noise. Identified lipids were confirmed by manual verification of the chromatograms using Xcalibur (version 3.0, Thermo-Fisher, Bremen, Germany). The mass spectra were acquired using an Orbitrap high-resolution mass spectrometer: Fourier-transform mass spectrometer (FT-MS) coupled with a linear ion trap (LTQ) Orbitrap XL (ThermoFisher Scientific, https://www.thermofisher.com) ( Supplementary Tables S2 , S3 ) (Alseekh et al., 2021). Multiple multivariate statistical methods were employed to investigate the variability of the lipidome pattern in germinated rice seeds. To this end, principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), variable importance in projection scores, heatmap, and pathway enrichment analysis for data visualization were performed using MetaboAnalyst 4.0 software (Chong et al., 2019). The fold change (log2FC) comparison of lipids was assessed between different color classifications among germinated sprouts ( Supplementary Table S4 ). The lipophilicity of the lipid groups was estimated based on the LIPID MAPS comprehensive classification system for lipids (Fahy et al., 2009).

The PRS samples subjected to DPPH (2,2-Diphenyl-1-picrylhydrazyl) radical scavenging activity and ABTS (2,2’-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)) assays according to Chu et al. (2020), and the Ferric Reducing Antioxidant Power (FRAP) assay was performed following the previous methods (Tomasina et al., 2012; Chu et al., 2020). The absorbance was measured using a microplate reader (BMG SPECTROstar Nano) at 515 nm for DPPH, 734 nm for ABTS, and 620 nm for the FRAP assay. The results were expressed as mg Trolox equivalent per g of sample. All calibration curves used have R² = 0.999. Correlation analysis was performed to assess the contribution of various lipid compounds to the antioxidant properties of PRS.

The bioactivity of samples was tested against colon carcinoma (HCT116) and lung adenocarcinoma (A549) cell lines (American Type Culture Collection, Manassas, VA, USA) following the previous method with modifications (Brotman et al., 2021). The cell line was maintained and cultured using Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin-streptomycin solution (Gibco, Life Technologies Corporation, New York, United States). The cells were grown in an incubator set at a temperature of 37°C in a 5% CO₂ humidified atmosphere (rh=95%).

Flasks containing cells of >70% confluency were trypsinized and the cells were harvested to obtain a cell suspension containing 50,500 cells/mL. About 198 µL of the cell suspension was transferred into a sterile 96-well microplate and incubated overnight at similar culture conditions. After overnight incubation, 2 µL of serially diluted samples (lipophilic extracts of PRS from superior and inferior MTAs) in DMSO were added into six designated wells, resulting in the extracts’ final concentrations of 240, 120, 60, 30, 15, and 7.5 µg/mL. Two-fold serial dilutions of doxorubicin (2 µg/mL) served as positive control and wells treated with DMSO served as solvent vehicle control. All treatments were performed in triplicate and independently repeated thrice. After 72 h of incubation, the spent medium in the wells was removed and 5 mg/mL MTT (Life Technologies Corporation, Eugene, Oregon) in phosphate buffer solution was added to each well. Plates were re-incubated for four hours and DMSO was added thereafter. The absorbance of each well was read at 570 nm using a microplate reader. The percent growth inhibition was calculated using the equation below. The half maximal inhibitory concentration (IC₅₀) was computed based on the trendline of the sample concentrations between which 50% inhibitory activity falls.

The DNA was extracted from the developing seed of pigmented rice collection and the sequencing libraries were prepared using the previous method (Elshire et al., 2011). The SNP marker data were generated by genotyping by sequencing (GBS) and the variant calling of the pigmented rice lines was performed against the Nipponbare reference genome (MSUv7). The GBS data were screened based on ≥90% call rate, locus homozygosity, and minor allele frequency (MAF) ≥0.05 resulting to 459,826 high-quality biallelic SNPs. Genome-wide association studies (GWAS) based on the mixed linear model for single-locus analysis namely Efficient Mixed-Model Association eXpedited (EMMAX) were conducted using the rMVP (A Memory-efficient, Visualization-enhanced, and Parallel-accelerated tool) R package (Yin et al., 2021). Association of SNP to the lipid expression was considered significant at a Bonferroni correction threshold of P < 1.08736783x10^-07 (or -log₁₀(P) ≥ 6.96362352) using the formula 0.05/m, where m is the number of SNP markers utilized. Furthermore, the population structures were estimated based on the calculation of principal components and kinship matrices ( Supplementary Figure S1 ). MAGMA software was utilized for gene-level analysis to identify significant genes identified based on Bonferroni-significant SNPs to further refine the list of candidates. The software includes internal correction for multiple testing, enabling the use of a significance threshold of P < 0.05 (De Leeuw et al., 2015). The heritability values of the traits were calculated using software tool called genome-wide complex trait analysis (GCTA) (Yang et al., 2011).

Haplotype blocks were examined using the blocks function implemented in PLINK 1.9, and the Haploview program was used to identify tag SNPs based on the threshold of the linkage disequilibrium (LD) coefficient (D’) > 0.8. SNPs associated with a P-value of <0.05 were considered significant and were used to generate the haplotypes (Barrett et al., 2005). Pairwise comparisons between alleles were based on the Mann-Whitney test and further confirmed by a t-test using the ggstatplot package in R (Patil, 2021). The genomic region, including 2 kb upstream of the start codon and 1 kb downstream of 3’ UTR was extracted using samtools (Danecek et al., 2021). Considering this region for each candidate gene identified using single-locus GWAS and known genes based on literature, targeted association analysis was performed through PLINK (Purcell et al., 2007) and EMMAX (Kang et al., 2010) for all traits of interest. Candidate genes with significant SNPs filtered at a 95% confidence level were visualized using Cytoscape (Shannon et al., 2003) and KnetMiner (Hassani-Pak et al., 2021). For the creation of marker-trait association (MTA), significant SNPs within each candidate gene were subjected to pruning (r² < 0.2, window of 75 kb), and their contribution to phenotypic variance explained (PVE) was assessed using the reml function in LDAK. In cases where candidate genes were located within broad association peaks, SNP thinning was performed through LDAK’s heritability model, applying a threshold of r² < 0.98 within a 75 kb window.

The filtered GBS data with 459,826 high-quality biallelic SNPs was utilized to calculate the Tajima’s D index and the F_ST using TASSEL 5.2.87 and VCF tools using a 20-kb window with a step size of 5-kb for each subpopulation following the procedure of Danecek et al. (2011). The Tajima’s D value was calculated in a 20-kb non-overlapping window. The population structure was taken from previous analysis (Mbanjo et al., 2023). Both Tajima’s D index and the F_ST were analyzed genome-wide with the comparison between Indica and Japonica. Genetic regions containing the genes of interests were visualized in the context of selection sweeps. PLAZA 5.0 was employed to investigate gene duplications among the selected genes across different species (Van Bel et al., 2022). Utilizing protein sequences available in the Ensemble database (plants.ensembl.org/index.html), orthologs/homologs were identified through a BLAST search. Subsequently, the construction of a phylogenetic tree was performed using the MEGA X program, with the incorporation of 1,000 bootstrap replicates to gauge confidence levels at each node. Evolutionary distances were then computed using the Poisson correction method (Kumar et al., 2018).

Lipidomic analysis of germinated rice sprouts revealed the major lipid classes – glycerolipids represented by diacylglycerols (DAG) and triacylglycerols (TAG) (42.9%), galactolipids comprised of lyso- counterparts, monogalactosyl diacylglycerols (MGDG), and digalactosyl diacylglycerols (DGDG) (24%), phospholipids consisting of lysophospolipids, phosphatidylcholines (PC), phosphatidylethanolamines (PE) (21.1%), sphingolipids (SL) (8.6%), and carotenoids (3.4%) ( Figure 1A ). Supplementary Figure S2 shows the general structure of these lipid compounds. Changes in the accumulation of lipids with a higher degree of unsaturation appeared to be highly significant between PRS and non-pigmented sprouts ( Supplementary Figure S3A ). Moreover, significantly higher amounts of glycerolipids were observed in pigmented (i.e., purple, red, and variable purple) rice sprouts compared to non-pigmented rice (light-brown) based on average peak intensities of DAGs ( Supplementary Figure S3B ) and TAGs ( Supplementary Figure S3C ). Among glycerolipids, unsaturated diacylglycerols (DAGs; 36:2, 36:3, 34:2) and triacylglycerols (TAGs; 46:3, 50:4, 50:5, 54:4, 56:2, 56:3, 56:6, 58:3, and 58:4) were significantly more abundant in purple rice sprouts compared to non-pigmented samples. Notably, minimal lipid class differences were observed between light brown and variable purple rice, with the exception of slightly elevated levels of lutein, zeaxanthin, TAG 54:3, and TAG 56:3. Within PRS ( Supplementary Figures S3G-I ), carotenoid concentrations were higher in purple rice than in red rice. Furthermore, a consistent downregulation of certain sphingolipids and phosphatidylcholines was observed in PRS.

In general, the germinated sprouts lipophilic extracts exhibit higher antioxidant activity in comparison to mature grains ( Figures 1B-D , Supplementary Table S5 ). The superior lines classified in germinated variable purple rice and germinated purple demonstrated remarkably higher antioxidant activities across all three antioxidant assays for lipophilic extractions. Results also showed a moderately strong positive correlation between the ABTS antioxidant and carotenoids such as alpha- and beta-carotene ( Figure 1E ). Among lipid fractions, MGDG 36:4 showed significant positive correlations with scavenging activity against the ABTS radical ( Figure 1E ). Specific triglycerides with higher degrees of unsaturation, such as TAG 56:6 and 56:5, and DGDG 38:4 showed significant positive correlations with the DPPH radical scavenging activity. On the other hand, DGDG 38:4 and Ceramide (Cer) (t18:0/22:0) were positively correlated with ferric reducing activity. Cluster analysis based on antioxidant values revealed distinct groups among the samples. Specifically, the sPLSDA demonstrated that three lines were black-colored rice samples—GBCR88, GBCR94, and GBCR142—exhibited high antioxidant capacity ( Figure 1F , Supplementary Table S5 ). However, the samples with moderate and low antioxidant capacities lacked clear differentiation. Notably, features such as total antioxidants, FRAP, DPPH, ABTS, alpha-carotene, and beta-carotene displayed relatively high loading scores, contributing significantly to the separation of the samples distinguishing high, medium and low antioxidant activity ( Figure 1G ).

Single-locus GWAS revealed 186 SNPs mapped to 174 candidate genes from different chromosomes associated with 20 lipid compounds of PRS based on Bonferroni correction threshold (P < 1.00x10^-7) ( Figure 2 ; Supplementary Figures S4 , S5 ). The gene-level and targeted association analyses narrowed down the list to 72 candidate genes for different lipid compounds (q-value < 0.01) ( Table 1 ). The combination of single-locus GWAS and targeted association analyses identified 11 candidate genes associated with lipase-related annotations, including six genes encoding GDSL-type esterase/lipase (GELP) proteins ( Supplementary Tables S6 , S7 ).

Out of 73 unique candidate genes identified from single-locus GWAS and literature mining, the gene-targeted association revealed a total of 11 candidate genes from Chromosomes 1, 2, 4, 7, 8, 9, 10, and 12 possessing significantly associated SNPs to different lipids of PRS ( Table 1 , Figure 2B ). The candidate genes include: LOC_Os02g40440 (OsGELP40), LOC_Os10g05088 (OsGELP102), LOC_Os10g30290 (OsGELP107), LOC_Os01g52230 (OsACP1), LOC_Os04g04254 (OsCTB2), LOC_Os04g27720 (OsTPS23), LOC_Os07g23410 (OsFAD2-3), LOC_Os07g45370 (expressed protein), LOC_Os08g08110 (OsDGK2), LOC_Os12g38780 (OsDGK4) and LOC_Os09g27210. Two candidate genes of GELP lipase gene family members from Chromosome 10 were associated with different lipids and carotenoids. LOC_Os10g05088 (OsGELP102), which possessed multiple intronic and exonic SNPs (synonymous) was significantly associated with DAG 36:1, DGDG 34:0, Cert(18:0/c26:1), and violaxanthin. The LOC_Os10g30290 (OsGELP107) gene which possessed different UTR3 and exonic significant SNPs were linked with TAG 46:2, PC 36:5, and TAG 56:5. LOC_Os01g52230 (OsACP1) possessed two significant SNPs with non-synonymous mutation (S01_30025718 and S01_30025731) was associated with TAG 46:2. LOC_Os02g40440 (OsGELP40), also possessed intronic and exonic SNPs significantly linked with PC 36:5, Lyso_MGDG.18.1, and violaxanthin.

The high PVE contributing SNPs from top five genes [(LOC_Os02g40440, OsGELP40), (LOC_Os10g05088, OsGELP102), (LOC_Os10g30290, OsGELP107), (LOC_Os01g52230, OsACP1), and (LOC_Os09g27210 - lecithin-cholesterol acyltransferase] identified superior MTA combinations GGTAAC/ACAAGCTGGGCCC exhibiting higher antioxidant activity measured across three independent antioxidant methods, namely 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2’-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid (ABTS), and ferric ion reducing antioxidant power (FRAP) ( Figure 3 , Supplementary Figures S9 , S10 ). Among the antioxidant traits, FRAP demonstrated the highest heritability, ranging from 0.80 to 0.96 ( Supplementary Table S8 ). The superior MTA combination (GGTAAC/ACAAGCTGGGCCC) exhibited higher levels of alpha-carotene and beta-carotene as well as the highest DPPH and FRAP antioxidant capacities ( Figure 3 , Supplementary Figures S9 , S10 ). Interestingly, this finding is consistent with the observation that DPPH is positively correlated with alpha-carotene and beta-carotene levels in the diverse PRS samples ( Figure 1 ). To further verify the potential anti-cancer properties, the samples of superior and inferior MTA combinations were taken and tested for their inhibitory activity against colon (HCT116) and lung (A549) cancer cell lines. Interestingly, the superior MTA combinations (GGTAAC/ACAAGCTGGGCCC) possess effective inhibitory activity against HCT116 and A549 with average 1/IC50 of 0.03 and 0.02 (mL/µg) compared to the inferior MTA combination (AATGACACAGCCGGGCCC), respectively ( Figures 3G, H ; Supplementary Table S9 ). It can be surmised that the increase in the carotenoids and unsaturated lipids translate to the antioxidant and anticancer properties in the lipophilic extracts of germinated sprouts.

Genetic analysis provided evidence that the GELP gene family significantly influences lipid composition and increased carotenoid changes in pigmented rice sprouts. To delve deeper, OsGELP40 (LOC_Os02g40440), OsGELP102 (LOC_Os10g05088), and OsGELP107 (LOC_Os10g30290) lipase genes were scrutinized to determine if they experienced selection pressures during pigmented rice domestication. Selection scans on three lipase genes located on chromosomes 2 and 10 were conducted through analysis of Tajima’s D index ( Figure 4A ) and F_ST values ( Figure 4B ) using pigmented rice japonica and indica subspecies. OsGELP40 exhibited positive Tajima’s D indices for both indica (0.79) and japonica (2.26), coupled with a relatively low F_ST value of 0.07, indicative of balancing selection that contributes to maintaining genetic diversity within both populations. Conversely for OsGELP102 gene, Japonica displayed a positive Tajima’s D (3.14), while Indica exhibited a negative Tajima’s D (-2.48), suggesting potential demographic events or selection processes leading to distinct patterns of genetic diversity. Additionally, a F_ST value of 0.397 indicated substantial genetic differentiation between Japonica and Indica. Similar patterns were observed in OsGELP107 concerning Tajima’s D and F_ST. The expression of OsGELP40, OsGELP102 and OsGELP107 were found in germinating sprouts ( Supplementary Figure S12 ).

The 3000 rice genome database was queried for superior alleles associated with these genes. Superior alleles linked to OsGELP40 exhibited distribution across rice subspecies, with Indica having the highest percentage. Conversely, OsGELP102-related superior alleles were exclusively found in Indica, not in Japonica ( Figure 4C ). Similar observations were made for OsGELP107 haplotypes, with a small percentage present in temperate Japonica. Figure 4D revealed conserved gene sequences of these GELP genes in O.sativa, O.glumipatula, and O.glabberima. Gene duplications were analyzed across species revealing non-duplicated and block duplications for OsGELP40 in both Indica and Japonica. Interestingly, different patterns for OsGELP102 gene were observed, where in tandem and block duplications were found in Indica and no duplications noted in Japonica ( Figure 4E ). For OsGELP107 no duplications were observed in either subspecies. In the phylogenetic analysis involving wild rice species, it was observed that there is no distinct divergence between domesticated rice and wild rice species ( Figure 4F ).

Although secondary metabolites are predominantly responsible for the coloration of PRS (Tiozon et al., 2021), our study showed that significantly higher proportions of unsaturated DAGs and TAGs were present in PRS compared to non-pigmented samples. Moreover, substantial accumulations of carotenoids and unsaturated lipids under the classification of TAGs, DGDGs, and PCs were observed in the red and purple variants compared to light-brown rice sprouts, while higher carotenoids were observed in purple rice than in red rice. Notably, the fatty acids present within both DAGs and TAGs have been found to exhibit a heightened antioxidant capacity (Khan et al., 2019). Primarily, unsaturated fatty acids can readily donate electrons to unstable free radicals (Khan et al., 2019). Notably, GELP genes such as LOC_Os10g30290 (OsGELP107), LOC_Os10g05088 (OsGELP102), and LOC_Os01g61200 (OsGELP26) have been associated with triglycerides such as TAG 46:1, TAG 46:2, and TAG 56:5 ( Figure 2 ; Supplementary Figure S11 ). Interestingly, these segmentally duplicated genes belong to the same phylogenetic subclade (Chepyshko et al., 2012), parallel to the current findings, which identified these genes to be linked with TAGs of similar chemical structures. Furthermore, few of the galactolipids (DGDG 38:4, DGDG 36:4, and MGDG 36:4) have correlation with antioxidants ( Figure 1E ). GWAS revealed associations of DGDG 38:4 with four candidate genes: LOC_Os04g39610, LOC_Os05g22940 (OsACC1), LOC_Os06g21950 (OsPT10), and LOC_Os06g22080 (OsDGAT2) ( Figure 2 ). One of the top candidate genes, OsDGAT2, confirmed through both GWAS and gene-level analysis, is known to encode for a diacylglycerol O-acyltransferase is involved in the key rate-limiting step in the conversion of DAGs into TAGs through the esterification of the third acyl chain of DAG (Bhunia et al., 2021). OsACC1 gene linked with DGDG 38:4 functioning upstream of DGDG synthesis plays a crucial role in fatty acid synthesis and metabolism (Tiwari et al., 2016; Bhunia et al., 2021).

In general, carotenoids are known for their photoprotective functions by dissipating excess light contributing to scavenge free radicals, and inhibit lipid peroxidation (Rodriguez-Concepcion et al., 2018). Concurrently, Pereira-Caro et al. (2013) revealed higher total carotenoid content was observed in rice samples with black pigmentation compared to red-colored rice. In plants, the relationship between lipids and the carotenoid pathway is complex yet interconnected. Lipids play a crucial role in providing the structural components and energy required for carotenoid biosynthesis, storage, and function. Additionally, lipids can act as carriers and protectors of carotenoids within the plant cells (Barja et al., 2021). Both lipids and carotenoids share common precursor molecules, namely isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), which are derived from the mevalonate (MVA) or the methylerythritol phosphate (MEP) pathways. Furthermore, geranylgeranyl pyrophosphate (GGPP), a precursor for carotenoid biosynthesis, is synthesized from IPP and DMAPP, which are derived from the lipid pathway (Barja et al., 2021). GGPP is a crucial precursor for carotenoids, and its availability is closely linked to lipid metabolism. Compared to other carotenoids like β-carotene and lutein, the levels of neoxanthin in seeds are relatively lower, yet its presence could play a role in enhancing the nutritional value of rice as a dietary source of carotenoids (Saini et al., 2022). GWAS analysis revealed that OsTPS23 (LOC_Os04g27720), which encodes for terpene synthase is significantly linked with neoxanthin. Protein-protein network analysis shows that terpene synthase gene is interacting with geranylgeranyl diphosphate synthase (GGDPS) ( Supplementary Figure S8 ), an enzyme that plays a key role in the biosynthesis of carotenoids in plants. Mechanistically, GGDPS catalyzes the formation of GGPP from farnesyl diphosphate in the plastidic isoprenoid biosynthetic pathway. GGPP is used as a precursor for the synthesis of various carotenoids, which are important for light harvesting and photoprotection in plants (Barja et al., 2021). It can be inferred that PRS potentially possesses higher levels of lipid intermediates, enabling the accumulation of increased carotenoids and lipids.

The lipidome-GWAS identified the importance of GELP genes relating to the elevation in alpha and beta carotenoids in the present study. The superior alleles combined from MTAs (GGTAAC/ACAAGCTGGGCCC) of top five genes including three GELP gene family members exhibits the highest levels of alpha- and beta-carotene, which corresponds to its high DPPH and overall antioxidant activity ( Figure 3 ). Furthermore, the pigmented donor lines exhibiting the GGTAAC/ACAAGCTGGGCCC haplotype represent significant donor candidates due to their consistently elevated antioxidant activity and anticancer property.

Lipases are attributed to the degradation of lipids resulting in complex fatty acids (Howitt and Pogson, 2006). In particular, TAG lipases play a crucial role in lipid homeostasis by breaking down TAGs into free fatty acids and glycerol. These enzymes are involved in various physiological processes, including energy mobilization, signal transduction, and stress responses (Howitt and Pogson, 2006). In addition, the generation and availability of free fatty acids are likely influential factors in the esterification and accumulation of carotenoids. Mechanistically, during germination, it is possible that carotenoids that are stored in lipid droplets during seed maturation are mobilized by lipases upon seed germination that undergo esterification with fatty acids, initiating the potential mobilization of the triglycerides and carotenoids (Abreu et al., 2020). Likewise, observations in other species have highlighted that owing to the lipophilic nature of carotenoids, they can be stored in lipid droplets. Subsequently, these carotenoids may be cleaved by lipases, facilitating their further mobilization and consequently enhancing detectable carotenoid levels (Abreu et al., 2020; Bu et al., 2022). Furthermore, lipids and phenolic compounds are also known to interact in plants to influence color and bioactivity (Karonen, 2022).

The accumulation of lipids can be associated with lipophilic antioxidant capacities, thus potentially representing a novel dietary source for human health. In fact, Figure 1 shows that there is an increased antioxidant capacity of lipophilic fractions after germination. Particularly, PRS with purple and variable purple color showed important donor lines with enhanced antioxidant activities of lipophilic fractions after the germination process. Similarly, Mohd Esa et al. (2013) demonstrated that germinated brown rice has higher lipid antioxidant enzyme activities such as superoxide dismutase and glutathione peroxidase compared with brown and milled rice (Mohd Esa et al., 2013). Lipophilic antioxidants, including lipophilic vitamins and unsaturated fatty acids, exhibit high effectiveness in protecting cell membranes from lipid oxidation. These antioxidants play a crucial role in preventing the oxidative damage of cell membranes by scavenging reactive oxygen species and inhibiting lipid peroxidation processes (Lagouri, 2014). Concurrently, alpha- and beta-carotene, as well as triglycerides with higher degrees of unsaturation, such as TAG 56:6 and 56:5 and DGDG 38:4 demonstrated moderate correlations with certain antioxidant tests ( Figure 1 ). Mechanistically, carotenoids operate through the presence of a conjugated system, wherein the π-electrons are efficiently delocalized across the entire length of the polyene chain. This delocalization enables carotenoids to exhibit unique antioxidant properties, making them important contributors to rice’s dietary properties (Lagouri, 2014). In fact, clinical studies have demonstrated that carotenoids engage in cross-talk with other lipids in the human body, thereby promoting enhanced immune system functions and lowering the risk of developing degenerative chronic diseases (Rodriguez-Concepcion et al., 2018; Marhuenda-Muñoz et al., 2022). These diseases include but are not limited to age-related macular degeneration, type 2 diabetes, obesity, certain types of cancer (breast, cervical, ovarian, colorectal), and cardiovascular diseases (Rodriguez-Concepcion et al., 2018; Marhuenda-Muñoz et al., 2022). Consistent with this feature, carotenoids are known to regulate the hallmarks of cancers through immune modulation, hormone and growth factor signaling, regulatory mechanisms of cell cycle progression, cell differentiation and apoptosis (Niranjana et al., 2015). The anti-cancer effect of carotenoids may be associated with the generation of reactive oxygen species, triggering cytotoxicity and apoptosis which involves the cleavage of caspases-3 and -9, as well as poly-ADP-ribose polymerase, accompanied by reductions in the levels of Bcl-xL (Sathasivam and Ki, 2018). Furthermore, polyunsaturated fatty acids may work through the antioxidant signaling pathway regulation and inflammatory cycle modulation (Calder, 2020). The enhanced antioxidant properties of the lipophilic fraction enriched with carotenoids in PRS in the diet may be beneficial to potentially promote human health by reducing the risks of inflammation, cardiovascular disorder, and cancer (Lagouri, 2014; Calder, 2020).

The phylogenetic tree of the OsGELP gene family suggests that, OsGELP40, OsGELP102, and OsGELP107 were recognized to occupy the same clade (Chepyshko et al., 2012). Notably, OsGELP102 and OsGELP107 genes, exhibiting distinct Tajima’s D indices and relatively high F_ST values, were further identified to be situated on a common subclade (Chepyshko et al., 2012). Besides on their role of lipid mobilization, these GELP genes were also found to be responsive to both biotic and abiotic stress (Yao-Guang et al., 2022). In addition, experimental evidences revealed that these genes may also influence the secondary metabolism pathways, plant development, and morphogenesis (Chepyshko et al., 2012). The accumulation in the lipid metabolites in pigmented rice sprouts that translated to its antioxidant capacity can potentially be attributed to these GELP genes. Interestingly, the GELP genes catalyze the esterification of an antioxidant carotenoid namely lutein in wheat (Watkins et al., 2019). These findings underscore the role of these genes in enhancing the antioxidant and anticancer activity of cereals.

From an evolutionary perspective, GELPs play an important role in cuticle formation, particularly cutin and suberin, which is crucial for the terrestrial colonization of plants (Shen et al., 2022). Recently, Jing et al. (2023) revealed 30 genes associated with domestication in rice where many of these genes are associated with plant growth and response to stress (Jing et al., 2023). Moreover, these GELPs play roles in these functions, prompting an analysis of population genomic parameters such as F_ST and Tajima’s D index in pigmented rice samples of Indica and Japonica which unveils potential differences in their selection process. The negative Tajima’s D index for the OsGELP102 and OsGELP107 genes could be indicative of population expansion, positive selection, or purifying selection in Indica. Furthermore, the relatively high F_ST value suggests strong genetic differentiation between Japonica and Indica populations, indicating limited gene flow between them linked to these genes. In fact, recent genomic and archaeological discoveries indicate that the initiation of indica domestication may have occurred approximately 8,000 years before the present, significantly earlier than the previously speculated timeframe thus suggesting that domestication was likely in progress in northern India before the arrival of japonica (Bates et al., 2017; Jing et al., 2023). It is noteworthy to further scrutinize the selection signals associated with the GELP genes.