Physiological determinants of heterosis in hybrid rice (Oryza sativa L.): identifying the optimal phenotyping stage for yield prediction

Abstract

Introduction Hybrid rice productivity depends on heterosis, yet the developmental stage at which physiological traits become sufficiently integrated to reliably predict grain yield across diverse environments remains poorly understood. Although physiological phenotyping is increasingly used in crop improvement, the biological basis for selecting the most informative developmental stage for early yield prediction has received limited attention. Methods A panel of 39 rice genotypes, including hybrids, parentals, and check varieties, was evaluated across four environments. Fourteen physiological and agronomic traits were measured at the early (30 days after sowing, DAS) and mid-vegetative (60 DAS) stages. Best linear unbiased predictors (BLUPs) derived from mixed models were used to estimate genotype performance from an unbalanced multi-environment dataset, followed by multivariate analysis, leave-one-season-out cross-environment validation, and development of a physiology-based selection index. Results BLUP-based multivariate analyses revealed a distinct productivity axis separating hybrids from parental lines, indicating that heterosis was associated with coordinated physiological and structural development rather than enhancement of individual traits alone. High-performing two-line and three-line hybrids exhibited broadly similar physiological patterns, suggesting shared physiological mechanisms underlying superior yield performance. Traits measured at 60 DAS consistently outperformed those measured at 30 DAS for predicting grain yield across environments. Chlorophyll content (SPAD60) and plant height (PH60) emerged as the most robust predictors, while stomatal conductance (FLGS60) provided complementary physiological information. Leave-one-season-out validation demonstrated that mid-vegetative traits achieved the highest predictive performance (r ≈ 0.38), whereas early-stage traits showed weak and inconsistent predictive ability. Combining early and mid-stage measurements did not improve prediction. A selection index based on key 60 DAS traits achieved a 39.08% yield advantage over the population mean. Conclusion This study demonstrates that the mid-vegetative stage represents a critical developmental transition during which physiological stabilization, canopy development, biomass consolidation, and coordinated source-sink function converge to support reliable yield prediction. Rather than identifying individual predictor traits alone, our findings establish a developmentally informed physiological framework explaining why 60 DAS provides the most biologically informative stage for hybrid evaluation across environments. This framework offers a practical basis for physiology-guided phenotyping, cost-efficient early hybrid selection, and improved decision-making in hybrid rice breeding programs.