Assessing the evolution of wheat grain traits during the last 166 years using archived samples

The study investigates how long-term environmental changes since the Industrial Revolution—particularly rising atmospheric CO₂ concentrations and temperature—have affected wheat grain yield and nutritional quality. Atmospheric CO₂ increased from ~280 ppm (pre-industrial) to ~406 ppm in 2017, with global mean surface temperature increasing by ~0.85 °C between 1880 and 2012. Prior research indicates elevated CO₂ can boost C₃ crop photosynthesis and yields but often reduces grain protein and mineral concentrations, potentially via carbohydrate dilution and changes in plant physiology. Experimental approaches such as growth chambers, OTCs, and FACE systems have limitations in simulating historical conditions, and few studies have examined lower-than-present CO₂ on grain quality. Archived materials (herbaria, long-term experiments) allow direct assessment of grains grown under past environments. This study aims to evaluate the impact of long-term changes in atmospheric CO₂, temperature, and rainfall on wheat grain quality traits (carbohydrates, protein, mineral concentrations) across the last 166 years, and how these changes relate to yield.

• Elevated CO₂ typically increases C₃ photosynthesis and yields but reduces protein and alters lipid and mineral content in grains (meta-analyses and FACE studies).
• Carbohydrate dilution has been proposed to explain lower protein and mineral concentrations under elevated CO₂, alongside reduced transpiration-driven mass flow of nutrients and altered N assimilation.
• FACE approximates field conditions better than chambers but still presents challenges (abrupt CO₂ changes, microclimate effects, replication).
• Prior historical analyses using archived materials indicate declining mineral density in wheat grains over time.
• Use of carbon isotope discrimination (Δ¹³C) as an integrative indicator of stomatal conductance and water-use efficiency is established, with genetic variation influencing Δ¹³C responses.

Grain material collection:

• Wheat (bread and durum) grains were collected from archives across 16 countries (global herbaria samples) spanning 1850–2016 and from the Broadbalk Continuous Wheat Experiment (Rothamsted, UK). Global sampling details (country, location, year, replicates) are listed in Table 1. Samples were intact without visible degradation.
• Broadbalk plots: annual farmyard manure (35 t ha⁻¹ fresh) applied since 1843. Yield data available for 1850–2016; thousand kernel weight (TKW) available 1974–2016. Harvest methods evolved from hand cutting to combine harvesting. Environmental data (1850–2016):
• Global atmospheric CO₂ from European Environment Agency; global temperature trends from IPCC reports. Rothamsted temperature and precipitation provided by the Department of Computational and Analytical Sciences (Rothamsted). Analytical measurements:
• Carbon isotope discrimination (Δ¹³C): 15 mg finely milled grain analyzed via elemental analyzer (EA1108, Carlo Erba) coupled to IRMS (Delta C, Finnigan) in continuous flow. δ¹³C calculated relative to Pee Dee Belemnite standard, then converted to Δ¹³C using Farquhar et al. equation with time-varying δ¹³C_air from published sources.
• Non-structural carbohydrates: Soluble sugars (glucose, fructose, sucrose) extracted with ethanol from ~25 mg milled grain; quantified by ion chromatography (ICS-3000, Thermo Dionex) against 50 mM standards. Starch in pellets solubilized (KOH 0.2 N; pH adjusted to 4.8), hydrolyzed with amyloglucosidase (R-Biopharm kit), and quantified spectrophotometrically at 340 nm.
• Protein content: Determined following Suchy et al. method.
• Elemental composition: C and N (%) via elemental analyzer (EA1108). Macro- and micronutrients (Cu, Zn, Fe, Mn, K, P, Mg, Ca, Na) by ICP-OES (iCAP 6500 Duo, Thermo Fisher Scientific). Statistical analyses:
• Due to heterogeneous locations/years/genotypes, yearly averages were computed; genotype effects were not modeled explicitly. One-way ANOVA assessed year effects; Fisher’s LSD for multiple comparisons. Multifactor ANOVA tested effects of CO₂, temperature, and their interaction. Pearson correlations assessed associations between traits and environmental variables; significance at p<0.05. Software: STATGRAPHICS Centurion 17.1.02 and R (RStudio v3.4.2).

Global archived samples (multi-country):

• Atmospheric CO₂ rose slowly 1850–1965 (+31 ppm) and rapidly 1965–2016 (+82 ppm); global mean temperature increased by ~1.2 °C from 1850 to 2016.
• Δ¹³C: Significant differences among years, but no significant difference between 1850–1955 and 1965–2016 (16.53% vs 16.50%; p=0.940); no significant correlations with CO₂ or temperature.
• Starch: Non-significant 7% increase between periods (2712.72 vs 2900 μmol g⁻¹ DW; p=0.344); positive association with CO₂ (r=0.247; ANOVA CO₂ effect p=0.009) and negative, non-significant correlation with temperature (r=−0.175; ANOVA temperature effect p=0.008).
• Soluble sugars: Glucose increased significantly between periods (1.01 to 1.89 μmol g⁻¹ DW; p=0.027); sucrose and fructose changes not significant. Correlations: CO₂ positively correlated with glucose (r=0.379, p<0.01) and sucrose (r=0.676, p<0.001); temperature positively correlated with glucose (r=0.238) and fructose (r=0.257).
• Protein: Significant 23% decrease from 1850–1955 to 1965–2016 (16.96% to 13.07%; p=0.049). Correlations: protein vs CO₂ r=−0.265 (p<0.01); protein vs temperature r=0.269 (p<0.01). ANOVA indicated significant effects of CO₂ and temperature on protein.
• C and C/N: C% unchanged; C/N ratio increased by 20% between periods (p=0.039). C/N correlated positively with CO₂ (r=0.229, p<0.01) and negatively with temperature (r=−0.396, p<0.001).
• Minerals: General decreasing trend with rising CO₂. Significant negative correlations with CO₂ for Mg (r=−0.560, p<0.001), Zn (r=−0.285, p<0.05), Fe (r=−0.188, p<0.05); temperature significantly affected Ca (positive; r=0.202, p<0.05) and Mn (positive; r=0.364, p<0.001). Multifactor ANOVA showed CO₂ and/or temperature significantly affected several minerals (see Table 6 in paper). Broadbalk (Rothamsted) long-term experiment:
• Yield and TKW: Yield was stable until ~1960, then increased; positively correlated with precipitation (r=0.662, p<0.001), CO₂ (r=0.631, p<0.01) and temperature (r=0.688, p<0.001). TKW decreased since 1974; no significant correlation with yield; CO₂ and temperature significantly affected TKW (ANOVA both p<0.001).
• Δ¹³C: Clear decline over recent decades; strongly negatively correlated with CO₂ (r=−0.688, p<0.001); temperature effect significant in ANOVA; no significant CO₂×temperature interaction.
• Starch: Increased over time, highest in 2016; positively correlated with CO₂ (r=0.458) and temperature (r=0.435); ANOVA showed significant CO₂ effect (p=0.026).
• Soluble sugars: Year-to-year variation; glucose positively correlated with CO₂ (r=0.591, p<0.01) and temperature (r=0.755, p<0.001); ANOVA confirmed significant effects of CO₂, temperature, and their interaction for glucose; sucrose significantly affected by temperature (p<0.001) in ANOVA.
• Protein: Decreased by ~26% from 1850 to 2016; negatively correlated with CO₂ (r=−0.771, p<0.001) and temperature (r=−0.474, p<0.05). ANOVA indicated significant CO₂ effect (p<0.001), temperature not significant.
• C and C/N: C% increased ~3%; C/N ratio increased ~40%. Both positively associated with CO₂ and significantly affected by CO₂ in ANOVA.
• Minerals: Significant negative associations of multiple minerals with CO₂ and temperature; strongest negative correlations with CO₂ for Mn (r=−0.942, p<0.001), Na (r=−0.852, p<0.001), K (r=−0.860, p<0.001), Fe (r=−0.717, p<0.001).

The findings demonstrate that rising atmospheric CO₂ and temperature over 166 years are associated with increased wheat yields but deteriorated grain nutritional quality. At Rothamsted, yield gains since the 1960s align with elevated CO₂, warming, and the introduction of semi-dwarf cultivars that increased harvest index and grain number (while TKW declined). The decline in Δ¹³C at Broadbalk suggests reduced stomatal conductance and increased water-use efficiency under higher CO₂, consistent with lower transpiration. Increased non-structural carbohydrates (starch, sugars) in grains are consistent with enhanced photosynthesis and carbon allocation under elevated CO₂ and modest warming (temperatures below heat-stress thresholds), which likely extended grain filling and improved source-to-sink carbon flow. Conversely, protein and mineral concentrations declined. This is consistent with carbohydrate dilution and physiological mechanisms such as reduced transpiration-driven mass flow of nutrients and possible limitations in nitrogen uptake/assimilation or reduced responsiveness of modern cultivars to N fertilization. The strong rise in C/N ratio supports a shift toward greater carbon accumulation relative to nitrogen. Global Δ¹³C showed no temporal trend, likely due to genotypic and environmental heterogeneity in the worldwide herbarium samples, highlighting the confounding effect of cultivar diversity and site-specific conditions. Overall, the results emphasize trade-offs between yield and grain nutritional quality under historical and ongoing climate change and breeding advances.

Using archived wheat grains (1850–2016) from global herbaria and the Broadbalk experiment, the study shows a long-term trend of increased grain carbohydrates (starch and soluble sugars), decreased protein content, and widespread declines in mineral concentrations, particularly since the 1960s. Yield increased while TKW decreased, indicating more grains rather than heavier grains contributed to yield gains. Mechanistically, elevated CO₂ and modest warming enhanced photosynthesis and carbon allocation, while reduced transpiration and potential N assimilation constraints contributed to lower protein and mineral contents. Breeding and management changes (e.g., dwarfing genes) also played a role in yield increases. Future research should focus on breeding strategies that combine high yield with superior nutritional quality and greater resource-use efficiency under changing climates, leveraging genetic variation in protein and mineral traits, and elucidating physiological and agronomic interventions (e.g., fertilization strategies) to mitigate nutrient declines.

• Herbarium samples often lack complete metadata (exact locations, environmental conditions, cultivars), increasing heterogeneity and limiting environmental attribution.
• Different genotypes and uneven replicate numbers per year; genotype effects were not modeled explicitly—yearly means were used instead.
• Yield and TKW data were only available for the Broadbalk experiment, not for global samples, limiting yield–quality comparisons globally.
• Environmental datasets were aggregated (e.g., global temperature), and global precipitation data were unavailable in Table 2, restricting certain analyses.
• Confounding from historical changes in agronomy and breeding (e.g., introduction of semi-dwarf cultivars, fertilization practices) alongside climate variables cannot be fully disentangled.
• Experimental comparisons with FACE/chamber studies are limited by differences between controlled exposure and natural historical growth conditions.