Seasonal performance and stability of tropical-adapted maize haploid inducers

The development of maternal haploid inducers adapted to tropical savanna environments is important for efficient implementation of doubled haploid (DH) technology in maize breeding. This study evaluated 14 advanced F₅ tropical-derived haploid inducer lines together with the temperate inducer BHI306 across three growing seasons to characterize induction performance, diagnostic-marker reliability, seasonal consistency, and operational breeding value. Field evaluations were conducted during the 2023/24 and 2024/25 dry seasons and the 2025 rainy season using Pacific789 as the maternal tester. Traits included putative haploid induction rate (HIR), marker misclassification with and without the flat-embryo criterion (MCRF and MCRNF), phenotypically confirmed haploid recovery, red root uniformity (RRU), pollen production capacity (PPD), flowering, and plant architecture. Fixed- and random-genotype REML models were used for entry comparison and estimation of entry-mean broad-sense heritability, respectively. Trait associations were evaluated using genotype-level correlations, and hierarchical clustering together with a predefined weighted selection index combining HIR, MCRNF, and PPD was used to characterize operational performance. Significant overall genotype effects were detected for HIR, MCRF, RRU, and tassel length, whereas MCRNF, phenotypically confirmed haploid induction rates, and most agronomic traits showed no significant overall genotypic differentiation. Significant genotype × season variation occurred for most induction-related and agronomic traits. Entry-mean broad-sense heritability was trait-dependent, ranging from 0 to 0.74. MCRF showed the highest estimate (0.74; 95% CI 0.30–0.89), whereas HIR (0.59) and RRU (0.60) had moderate point estimates but wide confidence intervals, and MCRNF had a boundary estimate of zero. Genotype-level correlations showed that relationships among HIR, marker misclassification, phenotypically confirmed recovery, and plant architecture varied in magnitude and direction among seasons, with several associations not consistently reproduced across seasonal conditions. The weighted selection index identified different seasonal leaders: K8/BHI306-68 in 2023/24, K8/BHI306-53 in 2024/25, and K8/BHI306-61 in the 2025 rainy season. K8/BHI306-61 was the only entry ranked among the top three in all three seasons and was also assigned to Cluster B. Tropical-derived inducer lines retained substantial phenotypic induction capacity under tropical savanna conditions, but their relative performance was strongly trait- and season-dependent. Reliable selection therefore requires integration of induction capacity, diagnostic reliability, reproductive performance, and consistency across seasons rather than reliance on a single trait or analytical approach. K8/BHI306-61 showed the most consistent operational ranking among the evaluated entries and warrants further multi-location, multi-year, and maternal-background validation. Entry-mean repeatability was strongly trait-dependent ($$\:{H}_{\text{e}\text{n}\text{t}\text{r}\text{y}}^{2}=0\text{\--}0.74$$), while widespread G×S effects emphasized the need for multi-season evaluation. Tropical-derived lines showed greater pollen-shedding capacity than BHI306, whereas most plant architectural traits showed no consistent overall genotypic differentiation. Associations among induction capacity, marker misclassification, phenotypically confirmed recovery, and plant architecture varied substantially among seasons. Pl-1 root expression showed weak and inconsistent relationships with R1-nj kernel misclassification, supporting their complementary use in haploid screening. The weighted selection index identified different seasonal leaders; K8/BHI306-61 was the only entry ranked among the top three in all three seasons.