Physiological adaptive traits are a potential allele reservoir for maize genetic progress under challenging conditions

Sustaining genetic progress in maize yield under increasingly challenging environmental conditions is critical given climate change and constraints on irrigation and fertilization. Historically, maize yield gains have come from optimizing crop cycle duration to local environments and, within maturity classes, from large genetic improvements in yield through selection and genomic selection. A central question is whether current rates of genetic gain can be maintained and which traits and alleles underlie this progress. The study aims to link changes in yield over 65 years to underlying physiological, phenological, and architectural traits, and to associated genomic regions, across a wide range of European environmental scenarios. It tests whether selection on yield indirectly targeted constitutive traits with stable effects versus adaptive physiological traits whose yield effects are highly environment-dependent.

Previous work has documented: (i) widespread yield increases due to breeding and agronomy across crops; (ii) major roles of flowering time and maturity adaptation in maize; (iii) substantial within-maturity genetic yield progress (50–75% of total gain) via selection and genomic selection; (iv) identification of genomic regions under selection and temporal yield progress using multi-site trials; and (v) context-dependent QTL effects on yield, with many QTLs showing positive, negative, or neutral effects depending on environment. The authors build on phenomics advances enabling high-throughput measurement of traits like stomatal conductance and canopy architecture, and on models that describe adaptive strategies from growth-oriented to conservative. The study fills a gap by analyzing, within the same genetic materials, the causal pathway from allele changes to trait changes to yield across environmental scenarios.

Genetic materials: A panel of 66 successful European maize hybrids (1950–2015; FAO 280–490) was assembled and genotyped (60 with Axiom 600k, yielding 465,621 SNPs post-QC). Representativeness was checked against a 96 ex-PVP panel and a 250-hybrid diversity panel (dent × common flint tester) via PCA and STRUCTURE analyses. Phenotyping and experiments: - Field trials: 30 experiments across Europe (2010–2017) spanning well-watered and water-deficit conditions and a broad temperature range. A multi-site set of 26 experiments was clustered into six environmental scenarios (cool/warm/hot × well-watered/water deficit) using temperature, soil water potential, VPD, and intercepted radiation indices. An intensive field experiment tested two plant densities (7 vs 9 plants m−2) under irrigated and water-deficit conditions. Grain yield, components (grain number, individual grain weight), anthesis, silking, ear traits, and environmental variables were collected; BLUEs and heritabilities were estimated with spatial mixed models (SpATS). - Controlled-environment phenotyping: Seven platform experiments (PhenoArch, Phenodyn, RootPhAir) measured: 3D architecture and derived canopy traits (rhPAD: height at which 50% of leaf area accumulates; light intercepted at ear layer; radiation use efficiency); reproductive traits (silk length, number of ovary and grain cohorts, number of extruded silks); whole-plant stomatal conductance (model inversion of Penman–Monteith every 15 min); leaf elongation rate and its sensitivity to soil water potential and VPD; root and shoot biomass (aeroponics) and root:shoot ratios. Treatments included well-watered, controlled water deficit (soil water potential targets), and high temperature or high evaporative demand conditions. Statistical analyses: - Genetic gain: Linear regressions of traits on year of release; ANOVA-based model selection to partition variance in grain number into genetic main effects (including year of release), environmental main effects (scenarios), and G × E interaction. Additional partitioning attributed components of genetic gain to vegetative duration, number of extruded silks, and architecture (rhPAD). Factorial regression tested sensitivities to environmental indices (mean soil water potential at flowering; intercepted radiation during vegetative phase). - Bayesian networks: Scenario-specific networks (TABU + BGE score; bootstrap model averaging) assessed conditional dependencies among year of release, phenology, architecture, reproductive traits, intercepted radiation at ear layer, and grain number. Genomics: - Regions under selection (RUS) were detected by (i) Bayenv2 Bayesian XtX scan contrasting the 22 oldest vs 22 most recent hybrids (top 0.05% XtX), and (ii) regression of allele counts on year of release using a mixed model controlling for relatedness (GCTA; threshold −log10 p ≥ 3.5). - Meta-analysis tested enrichment of overlaps between RUS and published QTLs/eQTLs (flowering time, florigens; architecture including rhPAD and intercepted light; stomatal conductance; leaf growth sensitivity; grain number sensitivity to soil water or heat; yield QTLs detected only in stress scenarios) using permutation-based expected overlaps (100,000 iterations) on matched genomic intervals (V2 coordinates). QTL effect stability: Six exemplar QTLs were selected: three adaptive (ABA synthesis; max stomatal conductance; temperature-responsive yield QTL) and three constitutive (rhPAD; flowering time/florigen; additional yield QTL with stable effects). Previously estimated scenario-specific allelic effects were projected onto the 24 fields by assigning each field to an environmental scenario to predict QTL-by-scenario effects and relate them to temporal changes in allele frequencies.

- Genetic gain in grain yield averaged 101 kg ha−1 year−1 across 65 years (n=60; p < 1e−5), accounting for ~75% of on-farm yield increase in Europe over the same period. Gains were similar across environmental scenarios (cool/warm/hot; well-watered/water deficit) and at two plant densities (7 vs 9 plants m−2). The year-of-release × scenario interaction explained only ~0.9% of yield variance. - Yield gains were primarily via increased grain number per unit area across all scenarios; individual grain weight increased by ~14% but was weakly related to yield. Phenology and reproduction: - Whole crop cycle duration was essentially stable, but vegetative phase duration increased (+~10 d at 20 °C; highly significant), grain-filling duration decreased, and the anthesis–silking interval (ASI) shortened markedly. - Reproductive development improved: more ovary cohorts were initiated; silk growth and number of extruded silks increased across environments; the number of fertile grain cohorts increased; grains per cohort remained unchanged. Bayesian network analyses implicated longer vegetative duration (supporting more ovaries) and enhanced silk growth (reducing abortion) as key mediators increasing grain number. Phenology-driven reproductive changes captured ~40% of genetic-progress variance in grain number. Architecture and light interception: - Canopy architecture shifted such that a greater proportion of leaf area was located lower in the canopy (lower rhPAD), driven by larger lower leaves and more erect leaves. - Modeled canopies showed increased light interception in the ear-bearing layer, improving carbon supply to developing ears, reducing abortion, and increasing grains per unit intercepted radiation. Root:shoot ratio and radiation use efficiency increased in modern hybrids. Architectural changes accounted for ~48% of the variance explained by genetic progress in grain number. Adaptive physiological traits: - Despite large heritability and genetic variation, adaptive traits showed little or no directional change across decades: whole-plant stomatal conductance showed no substantial trend; leaf elongation rate sensitivity to drought remained essentially stable (minor decrease under WD only); water-use efficiency was stable. Sensitivity of grain number to soil water status remained essentially unchanged. Genomic signatures: - Regions under selection were significantly enriched for QTLs of flowering time and florigen expression and for architectural traits (rhPAD; intercepted light), with strong allele frequency shifts between old and recent hybrids (enrichment p < 0.001). - There was little or no enrichment/colocalization of RUS with QTLs for stomatal conductance, leaf growth sensitivity, grain number sensitivity to stress, or yield QTLs detected only under stress scenarios. QTL effect stability and selection: - Adaptive QTLs exhibited scenario-dependent and oscillating allelic effects on yield across fields (approximately −0.16 to +0.17 t ha−1), and the frequencies of alleles favorable in favorable conditions did not increase over time (some decreased). - Constitutive QTLs displayed more stable positive effects (about +0.10 to +0.24 t ha−1 on average), and favorable allele frequencies increased; in at least one case the unfavorable allele was eliminated.

The study demonstrates that decades of selection for yield have indirectly targeted constitutive, physiologically simpler traits—phenology and canopy architecture—that have stable, positive effects on grain number across diverse environments. These changes improved reproductive development (more ovary cohorts, shorter ASI, more silks) and enhanced light interception at ear level, contributing to consistent gains even under heat and water deficit. In contrast, adaptive physiological traits (stomatal regulation, growth sensitivity, water-use efficiency) remained largely unchanged, showing no genomic signatures of selection, despite ample heritable variation. A plausible explanation is that allelic effects at adaptive-trait QTLs fluctuate with environmental scenarios and from year to year, producing no consistent selection pressure on a given allele in breeding programs that select primarily on mean yield across heterogeneous conditions. Consequently, while yield has improved in parallel across favorable and unfavorable scenarios, a reservoir of unexploited alleles governing adaptive traits remains. With climate change increasing the frequency of extreme stress events, leveraging this reservoir could enhance yield stability and resilience. However, because constitutive-trait improvements may be approaching biological or genetic limits (e.g., finite scope to extend vegetative duration or further optimize leaf distribution), new strategies that explicitly consider environment-dependent trait effects are warranted.

By integrating multi-environment phenomics, multi-site field trials, and genomic analyses across 66 years of European maize hybrids, the study links sustained yield gains (~101 kg ha−1 year−1) to shifts in phenology and canopy architecture that enhance reproductive success and light interception. It finds negligible directional change and no selection signatures for adaptive physiological traits, explaining their non-selection by the instability of their yield effects across environments. The work identifies adaptive physiological traits as a potential allele reservoir for future breeding, particularly for resilience under climate change. Future research and breeding directions include: (i) incorporating selection for yield stability and resilience to extreme drought and heat; (ii) explicitly exploring the genotype × environment × management space to tailor hybrids to scenarios; (iii) integrating phenomics, modeling, and genomic prediction to select on response curves to environmental drivers; and (iv) targeting adaptive-trait QTLs with environment-specific or index-based selection to exploit untapped variation.

The panel focused on temperate European elite germplasm and may not fully represent subtropical or tropical maize; findings need validation in other regions and heterotic groups. Only a subset of adaptive traits was measured; other physiological processes might also contribute to resilience. Some platform-derived traits rely on model inversion or simulation (e.g., stomatal conductance, intercepted light), which, despite validation, may introduce modeling assumptions. The inference of selection uses statistical scans and colocalization with published QTLs; causal loci and pleiotropy are not definitively proven. Finally, the scenario definitions and projections of QTL effects assume consistent scenario assignment and effect transferability from the diversity panel to the genetic progress panel.