Genetic gains in short-season corn hybrids: Grain yield, yield components, and grain quality traits

Abstract

Era studies are important to understand historical changes in maize (Zea mays) germplasm and estimate genetic gains, yet information for short-season maize hybrids is limited. Here, we determine grain yield genetic gain in Bayer short-season hybrids (100–105 days) and investigate indirect changes made on 17 secondary traits, including yield components (kernel number, weight, and shelling efficiency), and grain quality traits (oil, protein, starch, ethanol, moisture, and test weight). We evaluated 40 maize hybrids released from 1980 to 2020 across 18 environments in the US Corn Belt. Plant density and N-fertilizer were held constant within each environment. Results indicated a linear increase in grain yield (from 11.1 to 15.3 Mg ha⁻¹, 105 kg ha⁻¹ year⁻¹, or 0.8% year⁻¹) with no sign of a plateau. The increase in grain yield was attributed more to increased kernels per m² (0.57% year⁻¹) than kernel weight (0.23% year⁻¹). Grain protein concentration decreased until the late 2000s and plateaued thereafter, while starch and ethanol concentration increased until the early 2000s and plateaued thereafter. However, the total amount of protein, starch, and ethanol increased linearly from 1980 to 2020. We concluded that maize breeding for increased grain yield has indirectly affected many traits at different rates and directions. Our results are encouraging for future progress in grain yield increase, update genetic gain information to 2020 for short-season hybrids, and can inform plant breeders, crop physiologists, agronomists, and crop modelers.

Abbreviation

BLUP

best linear unbiased predictor

1 INTRODUCTION

Hybrid maize (Zea mays) was the initial major driver of early 20th-century grain yield increases before the introduction of synthetic nitrogen in the mid-1950s (Cardwell, 1982). Improved crop harvesting and planting equipment, effective herbicides, and adoption of soybeans into the rotation with maize also improved grain yields throughout the 20th century (Cardwell, 1982). The contribution of breeding to grain yield increase is typically measured in field experiments evaluating historical cultivars (the so-called era studies). Such studies are routinely performed to ensure progress in crop production and estimate genetic gain (Duvick, 2005; Russell, 1974). Over time, numerous era studies have been implemented around the globe for maize (Castleberry et al., 1984; Curin et al., 2020; DeBruin et al., 2017; Derieux et al., 1987; Duvick, 1984; Fernández et al., 2022; Liu et al., 2021; Mueller et al., 2019; Russell, 1991; Smith et al., 2014; Tollenaar, 1989; Wang et al., 2011). Grain yield genetic gain has been routinely reported in all published era studies; however, secondary traits such as grain components and grain quality traits critical to explain grain yield genetic gain have been reported in fewer studies (Chen et al., 2017; DeBruin et al., 2017; Haegele et al., 2013).

Information on secondary traits provides an opportunity to identify potential traits that may require more attention through breeding to maintain grain yield increases in the future. For example, determining genetic gain for kernel number and kernel weight provides physiological insight into grain yield genetic gain. Prior research has shown that the kernel number per m² increases with the year of hybrid release (DeBruin et al., 2017; Duvick, 1984; Haegele et al., 2013). Others have shown kernel weight to rise steadily over time, contributing further to increasing grain yields (Chen et al., 2017; DeBruin et al., 2017; Haegele et al., 2013). Changes in grain quality (starch, protein, and oil) impact grain yield and nitrogen use efficiency due to the metabolic energy required to produce starch compared to protein (Connor et al., 2011; McDermitt & Loomis, 1981). Decreases in grain protein concentration with years of hybrid release have previously been attributed to an indirect selection for improved grain yield (Duvick & Cassman, 1999). Grain protein is the main nitrogen sink in the kernel; thus, a reduction in protein concentration will impact nitrogen use efficiency (Sinclair & Rufty, 2012) and fertilizer recommendations (Woli et al., 2016). Test weight is another trait important for milling and processing of maize. This applies to the United States due to a volumetric standard (bushel) of selling grain. The USDA has set the minimum allowable test weight for number 1 yellow maize to be 56 lbs (720 kg hL⁻¹), and for number 2 at 54 lbs, for an equivalent “bushel” volume of 1.244 cubic feet at 15.5% grain moisture. Maize delivered with test weight below 54 lbs is subject to price discounts. Factors such as kernel size, grain moisture, and density affect the final test weight values (Rankin, 2009). Due to economic penalties, these traits are important to maintain minimum requirements as grain yield increases.

Our literature review revealed that published era studies have primarily focused on relatively long-season maize hybrids averaging around 111-day relative maturity (Chen et al., 2016; DeBruin et al., 2017; Duvick, 1984; Haegele et al., 2013; Mueller et al., 2019; Russell, 1991). Our knowledge of short-season maize hybrids, ∼103 days or less, is limited to only three studies (Badu-Apraku et al., 2016; Derieux et al., 1987; Tollenaar, 1989). The first published information comes from Derieux et al. (1987), who studied 10 short-season hybrids released from 1950 to 1985 across 10 environments in France. Two years later, Tollenaar (1989) published data (nine era hybrids × four plant density) from Ontario, Canada, pertinent to the 1959–1988 year of release. The last era study in west-central Africa investigated 56 hybrids released from 1995 to 2012 across five environments (Badu-Apraku et al., 2016). These three studies reported a grain yield genetic gain from 54 to 100 kg ha⁻¹ year⁻¹ for maize with very limited information on secondary traits.

Core Ideas

• We measured 18 plant traits of 40 maize hybrids across 18 experiments in the US Midwest.
• Grain yield linearly increased by 105 kg ha⁻¹ year⁻¹ from 1980 to 2020.
• Kernels per m² increased at twice the relative rate compared to kernel weight.
• Protein concentration decreased up to 2008 and remained unchanged until 2020.
• Grain moisture at harvest, test weight, and oil concentration remained unchanged.

Interestingly, no era study has been conducted in the US Corn Belt for short-season hybrids. The only study that showed grain yield progress in short-season hybrids in the US Corn Belt was a hybrid × seeding rate trial that did not grow the same hybrids each year in each environment (Assefa et al., 2017). Short-season hybrids are typically recommended for northern US Corn Belt latitudes (Figure 1), and while this region represents a small fraction of maize cultivation in the US Corn Belt (roughly 8.5 million ha, Figure 1), short-season hybrids are also used in lower latitudes, especially in situations of delayed planting or when farmers aim to harvest their crops early (Baum et al., 2020). Therefore, the importance of studying short-season hybrids and accumulating new knowledge is elevated. The aim of this study is to enrich our knowledge base on short-season maize hybrid (∼103-day relative maturity) genetic gains. Our first objective is to estimate grain yield genetic gain for Bayer short-season legacy hybrids. Our second objective is to identify indirect changes made in 16 secondary traits, including grain yield components and grain quality traits and investigate relationships among traits to increase our understanding.

FIGURE 1

Distribution of the 18 experiments (site-year combinations) across the US Corn Belt. The background color indicates corn relative maturity zones (source: Bayer Crop Science). Colored pixels indicate maize cropland. Location names: 1: Dyersville, IA, 2: Britt, IA, 3: Burnside, IA, 4: Marcus, IA, 5: Spencer, IA, 6: Compton, IL, 7: Lanark, IL, 8: Blooming Prairie, MN, 9: Grand Meadow, MN, 10: Morgan, MN, 11: Tracy, MN, 12: Tyler, MN, 13: Wilmont, MN, 14: Atkinson, NE, 15: Dakota City, NE, 16: Carey, OH, 17: Cuba City, WI (2020), and 18: Cuba City, WI (2021). Additional information per experiment is provided in Table 1.

2 MATERIALS AND METHODS

2.1 Environments and genotypes

Eighteen on-farm experiments were conducted in the US Corn Belt in 2020 and 2021 (Figure 1; Table 1). The environments captured a diverse set of soil (Table 1) and weather conditions (Figure S1), along with a range in farmer standard practices for nitrogen fertilizer (146–274 kg N ha⁻¹), plant density (6.6–10.1 plants m⁻²), and planting dates (April 23rd–May 12th; Table 1). The farmers made field preparation and management decisions.

TABLE 1. Soil series, texture, soil organic matter (SOM, 0–30 cm), water holding capacity (WHC, 0–150 cm), planting date, plant counts, nitrogen fertilizer application (N-rate) and grain yield at 15% moisture for each environment (Env). All experiments were in a maize-soybean rotation except #1 and #7, which were in maize-maize. Environments 1–5, 13, 17, and 18 were under minimum tillage, 14 was under no-tillage, while 6–12 and 16 were under conventional tillage. Environments 1–6, 8, and 16 were poorly drained, 10, 11, and 13 moderately drained, while 9, 12, 14, 15, 17 and 18 well drained.

Env.	Soil series	Soil texture	SOM (%)	WHC (mm)	Planting date	Plants per m^2	N-rate (Kg N ha^−1)	Yield (Mg ha^−1)
1	Clyde-Floyd	Complex	7.5	285	April 28	7.5	146	11.7
2	Webster	Clay loam	6.7	270	May 1	7.9	151	13.2
3	Marna	Silty clay	6.2	255	April 27	8.7	208	12.6
4	Primghar	Silty clay loam	5.5	315	April 28	9.4	217	14.5
5	Gillett Grove	Silty clay loam	6	300	May 2	8.0	253	13.0
6	Fella	Silty clay loam	5.5	315	May 8	8.9	218	14.9
7	Psco	Silt loam	3.5	300	May 12	8.3	246	10.5
8	Tripoli	Clay loam	6	150	May 5	6.6	250	14.2
9	Ostrander	Loam	3.4	240	May 1	7.4	182	13.3
10	Normania	Loam	2.7	270	April 25	8.5	211	15.0
11	Complex	Clay loam	4.3	240	May 4	8.5	158	13.7
12	Hokans-Svea	Complex	4.5	270	April 30	7.7	168	12.3
13	Crooksford	Silty clay loam	5.5	300	May 1	10.1	226	13.1
14	Oneil	Loamy sand	1.8	105	April 30	8.8	218	15.6
15	Blyburg	Silt loam	1.5	315	April 29	8.3	274	13.5
16	Lenawee	Silty clay	6.1	285	May 4	7.6	227	12.3
17	Tama	Silt loam	6	300	April 23	9.9	179	15.2
18	Fayette	Silt loam	1.5	285	April 29	8.5	179	13.2

Tillage practices varied across experiments from no-till to conventional tillage (Table 1). The previous crop was soybean for 16 environments and maize for two environments. Experiments were maintained free of weeds, pests, and diseases.

In each experiment, we evaluated 40 maize hybrids, averaging 103-day relative maturity, that were commercially released from 1980 to 2020 by Bayer Crop Science (Table 2). Hybrids were selected based on commercial sale volume across the Corn Belt (Table 2). No GMO (genetically modified organism) trait or trait packages were included. The experimental design was a randomized complete block with two replications totaling 80 plots per environment. The plot size was 5-m long with four rows spaced 76 cm.

TABLE 2. Bayer Crop Science legacy hybrids used in all locations by era of release (ERA), product or parent names (Name), relative maturity (RM in days), and year of release (YOR).

ERA	Name	RM	YOR	ERA	Name	RM	YOR
1985	T1000	100	1983	2005	DKC 52-43	102	2003
	DK524	100	1985		DKC 53-11	103	2004
	LH82 + LH74	104	1985		DKC 54-50	104	2004
	LH74 + LH59	105	1986		DKC 52-21	102	2004
	DK547	104	1986		DKC 52-59	102	2006
1990	DK535	103	1988	2010	DKC 55-24	105	2008
	LH74 + LH61	102	1989		H5222VT3	102	2008
	LH202 + LH163	102	1990		DKC 55-09	105	2009
	LH222 + LH172	101	1991		DKC 53-78	103	2009
	DK512	101	1991		DKC 53-56	103	2012
1995	LH227 + LH172	104	1994	2015	DKC 51-19	101	2013
	DK527	102	1994		DKC 52-04	102	2013
	LH198 + LH176	102	1995		DKC 54-38	104	2014
	DK546	104	1995		DKC 55-20	105	2015
	DK521	102	1995		DKC 52-84	102	2015
2000	LH227 + LH295	100	1998	2020	DKC 55-85	105	2018
	DK537	103	1998		204-25STX	104	2020
	DK520	102	1998		DKC 53-25	103	2020
	DK539	103	1999		DKC 52-34	102	2020
	DKC 51-43	101	2002		DKC 52-16	102	2020

2.2 Soil and weather characterization

Soil information for each environment was obtained from SSURGO (Soil Survey Geographic Database; Soil Survey Staff, 2023). The dominant soil type per experiment was used to determine soil water holding capacity (difference between field capacity and wilting point in millimeters) and organic matter (Table 1). Across experiments, the 0–30 cm soil organic matter ranged from 1.5% to 7.5%, the 0–150 cm water holding capacity ranged from 105 to 315 mm, while soil texture included clay loams, loams, and silty clay loams (Table 1). Our study captured a large range of soil conditions within the US Corn Belt and is a representative sample of the target environments for these hybrids (Figure 1).

For each location, gridded hourly weather data (30 years) was provided by Bayer Crop Science (Muñoz-Sabater et al., 2021). We calculated monthly mean temperatures and total rainfall per environment (Figure S1, Tables S1 and 2). We also calculated weather anomalies (Li et al., 2019) to benchmark our study years with historical years. Anomaly values greater than one in either direction (positive or negative) refer to abnormal conditions (Figure S1). At the start of the growing season (May), rain and air temperatures were near normal (Figure S1). In June, some environments were abnormally warm (anomaly greater than 1) or very warm (anomaly greater than 2) with some being drier than normal. During the critical period (July), most environments had normal conditions with a few exceptions, such as being drier and/or warmer than normal. The primary grain filling period (August) had mostly normal conditions, with three environments being drier and warmer, while three environments experienced the opposite (Figure S1).

2.3 Measurements and calculations

Grain yield was determined from the middle two rows of each plot using precision small plot combine harvesters by Bayer Crop Science. Grain test weight and moisture were also recorded during mechanical harvest. Prior to mechanical harvest, at about physiological maturity, five consecutive ears from the center row were hand harvested for grain yield components and grain quality analysis. The grain from the hand-harvested ears was included in the final plot grain yield estimations. The collected maize ears were air-dried, and the total ear mass was measured. Then the ears were shelled using an AEC (Applied Electronics Corporation) small batch grain sheller, and the kernels were cleaned using aspiration. Kernels were weighed and counted using a Seed Counter 900. The average kernel number per ear was determined from the total kernel number divided by the number of ears collected. The number of kernels per m² was estimated as the total plot yield divided by the 1000 kernel weight at 0% grain moisture. Shelling efficiency was calculated as the total ear mass divided by the total kernel weight in each sample.

The kernels from the hand-harvested ears were analyzed using near-infrared radiation (NIR) to determine oil, protein, starch, ethanol production (function of protein, oil, and starch), and grain moisture using an Infratec whole grain NIR analyzer. NIR machine calibrations were developed by the Iowa State University Grain Quality Lab, which is the adopted standard method of the American Association of Cereal Chemistry (AACC, 1999; Rippke et al., 1995). To assess changes in protein mass per kernel, the 1000 kernel weight was multiplied by the measured grain percentage protein divided by 100. The total protein, oil, starch, and ethanol amount (kg ha⁻¹) was determined as total grain yield multiplied by the corresponding concentration at 0% moisture.

The nitrogen fertilizer use efficiency was calculated as the total plot grain yield (0% moisture) divided by the total amount of nitrogen fertilizer applied. Grain yield per plant was estimated by dividing grain yield with the actual plant counts per plot. Plant counts were recorded early in the season, and the obtained plant counts were very close to the intended plant density.

2.4 Statistical analysis

Genetic gain for grain yield, yield per plant, N-fertilizer use efficiency, kernels per m², kernels per ear, kernel weight, shelling efficiency, test weight, protein mass per ha, protein mass per kernel, starch mass per ha, liters of ethanol per ha, oil concentration, and oil mass per ha per environment followed linear patterns and therefore were described using a linear mixed-effect model using the lme function in the lme4 package in R (Bates et al., 2015; R Core Team, 2022). Model factors included year of release (centered by the mean of 2001.5), environment, and their interaction as fixed effects. Hybrid, hybrid by environment, and replicates within environment were considered random effects. The significance of fixed effects was tested using a type III F-test with Kenward–Roger degrees of freedom approximation using an α of 0.05 (Kenward & Roger, 1997). To quantify the variance explained by genetics, environment, and their interaction, a linear mixed effect model was fitted with the hybrid, environment, and their interaction as random effects.

Initial data exploration on kernel protein concentration, starch concentration, grain moisture, and liters of ethanol per kilogram of grain indicated non-linear patterns. This was confirmed as the linear plateau model had a lower Akaike's information criteria value compared to the linear model (Akaike, 1998). The genetic gain per non-linear trait (i.e., percentage starch and percentage protein) was fit using a linear plateau function with environment as a random effect using the nlme function from the nlme package (Pinheiro et al., 2022). Per environment, genetic gain was fit using nlsLMList and the self-start function SSlinp from the nlraa package (Miguez, 2022), with significance between environments being determined using an F-test from the anova_nlslist function from the nlshelper package (Duursma, 2017).

For all traits, the hybrid best linear unbiased predictor (BLUP) values were estimated and graphed against the year of release (Figures 2-5). The BLUPs represent the true genotypic effect (de la Vega et al., 2007; Felipe et al., 2016). The absolute gain (the slope) was divided by the mean estimate from 1980 to 2020 per trait to determine the relative gain over this specific time period. Furthermore, grain moisture across relative maturities and eras were tested for its influence on grain moisture genetic gain. Grain moisture means for each relative maturity group were tested using a Fisher least significant difference test with the LSD test function from the agricolae package (de Mendiburu, 2021). Pearson correlation coefficients and their significance between traits were analyzed to determine the strength of relationships utilizing a hybrid's BLUP value per trait. Correlations were significant at a p-value ≤ 0.05 with Pearson correlation coefficients and p-values estimated using the cor function from the stats package (R Core Team, 2022) along with cor_pmat function from the rstaix package (Kassambara, 2021).

FIGURE 2

Overall across 18 experiments grain yield genetic gain (a), grain yield per plant genetic gain (grain yield / plant counts; b), and nitrogen fertilizer use efficiency (grain yield / N-fertilizer; c). Points represent the hybrid best linear unbiased predictor (BLUP) values. The regression coefficients are provided in Table S3 while the relative genetic gain of each trait is provided in Figure 4.

FIGURE 3

Environmental and management explanatory variables for grain yield genetic gain. The explanatory variables include average environment grain yield at 15% moisture (panel a; y = 19.4 + 6.4x, R² = 11%), total rainfall from May to September (panel b; y = 49.5 + 0.13x, R² = 16%), average vapor pressure deficit from May to September (panel c; y = 228–130x, R² = 18%), nitrogen fertilizer rate (panel d; y = 164–0.28x, R² = 18%), observed plant density (panel e; y = 81.1 + 2.48x, R² = .01%), and planting date (panel f; y = 373–2.2x, R² = 15%). Each point represents a study environment. Additional correlations can be found in Figure S3.

FIGURE 4

Ranking of the relative genetic gain (percentage per year) for the 18 studied traits. The box plot represents the variability across 17 environments. The dot within each box plot represents the mean value. Plant traits were color-coded per category (e.g., grain quality). Relative genetic gain was estimated by dividing the actual genetic gain with the mean observed value. The * next to kernel weight, shell efficiency, and protein indicates non-significant genetic gain across environments at an α of 0.05.

FIGURE 5

Overall genetic gain across 17 experiments for kernels per m² (a), kernels per year (b), kernel weight (c), shelling efficiency (d), test weight (e), grain moisture (f), protein (%) (g), starch (%) (h), liters of ethanol per kg of grain (i), protein mass per ha (j), starch mass per ha (k), liters of ethanol per ha (l), protein mass per kernel (m), % oil (n), and oil mass per ha (o). Points represent the hybrid best linear unbiased predictor (BLUP) values. The regression coefficients are provided in Table S3 while the relative genetic gains in Figure 4.

3 RESULTS

3.1 Grain yield

Overall average grain yield significantly increased from 11.1 Mg ha⁻¹ (176 bu ac⁻¹) to 15.3 Mg ha⁻¹ (241 bu ac⁻¹) from 1980 to 2020 (p < 0.002, Figure 2; Table S3). The genetic gain averaged 105 kg ha⁻¹ year⁻¹ (1.7 bu ac⁻¹, 0.8% year⁻¹) and ranged from 62 to 165 kg ha⁻¹ year⁻¹ across environments (Figures 2-4). Newer hybrids exhibited smaller variability in grain yield (coefficient of variation for 2015–2020 hybrids of 13%) compared to older hybrids (coefficient of variation for 1985–1990 hybrids of 16%, Table S4). Grain yield per plant increased by 1.26 g plant⁻¹ year⁻¹ (0.79% year⁻¹), with a relative gain identical to grain yield (Figures 2 and 4). The N-fertilizer use efficiency (grain yield/N-fertilizer rate) significantly increased from 47.5 to 65.5 kg grain kg N⁻¹ applied from 1980 to 2020, having a genetic gain of 0.45 kg kg N⁻¹ (0.81% year⁻¹, Figures 2 and 4).

Grain yield genetic gain was positively correlated with the average grain yield per environment and total precipitation from May through September (Figure 3) and negatively correlated with average vapor pressure deficit from May through September, N fertilizer, and planting date (Figure 3). Among these, the vapor pressure deficit had the most significant impact (p = 0.08; R² = 18%) on grain yield genetic gain, which was found to decrease by 13 kg ha⁻¹ year⁻¹ for every 0.1 increase in vapor pressure deficit (Figure 3c). High N fertilizer rates decreased grain yield genetic gain by 0.28 kg ha⁻¹ year⁻¹ for each additional kg N applied (Figure 3d). Delayed planting dates lowered genetic gain by 2.2 kg ha⁻¹ year⁻¹ for each additional day delay (Figure 3f). Planting density did not correlate with the grain yield genetic gain (Figure 3e, Table 1).

3.2 Grain yield components

Kernels per m² significantly increased from 3933 (in 1980) to 4937 (in 2020), having an overall genetic gain of 25.1 kernels m⁻² year⁻¹ with a relative increase of 0.57% year⁻¹ (Figures 4 and 5a). Across environments, the genetic gain ranged from 11 to 43 kernels m⁻² year⁻¹. Increases in kernels m⁻² were accompanied by significant increases in kernel weight, from 242 (in 1980) to 266 (in 2020) mg kernel⁻¹ (Figure 5b), with an overall genetic gain of 0.59 mg kernel⁻¹ year⁻¹ and a relative increase of 0.23% year⁻¹ (Figures 4 and 5; Table S3). Kernel weight genetic gain did not significantly vary across environments (p = 0.42).

3.3 Grain moisture, test weight, and shelling efficiency

Shelling efficiency (kernel weight to ear weight) increased from 85% in 1980 to 87% in 2020 (Figure 5d), having an overall genetic gain of 0.06% year⁻¹ (Figures 4 and 5d). Shelling efficiency genetic gain did not vary across experiments (p = 0.49). Grain test weight remained unchanged from 1980 to 2020 (Figures 4 and 5e). Average grain moisture at harvest decreased from 21.1% to 18.6% until 1998, when further decreases in grain moisture ceased (Figure 5f; Table S3). Relative maturity could have a major impact on harvest moisture. Indeed, 104- and 105-day hybrids had the highest grain moisture at harvest (20%), while the 101- and 102-day hybrids had the lowest (17%; Figure S2). However, relative maturities were well distributed within each era (Table 2), and relative maturity did not affect our genetic gain estimate for grain moisture.

3.4 Grain quality

Protein concentration at 0% moisture significantly decreased from 9.2% to 7.7% until 2008, reaching a plateau thereafter (Figure 5g). The average genetic gain for grain protein from 1980 to 2008 was −0.05% year⁻¹ and ranged from −0.03% to −0.12% year⁻¹ across environments (Figure 4). Starch concentration increased from 71% to 73% until the year 2000, when further changes in starch concentration ceased (Figure 5h). From 1980 to 2000, the starch concentration relative genetic gain averaged 0.06% year⁻¹ (range: 0.01% to 0.16% year⁻¹ across environments). Overall, the protein concentration was more variable than starch across environments (Figure 4). The oil concentration remained unchanged over the years (Figure 5n).

The cumulative protein production per hectare significantly increased from 850 to 977 kg ha⁻¹ over the 40 years, having a significant overall genetic gain of 3.2 kg protein ha⁻¹ year⁻¹ or 0.35% year⁻¹ (Figures 4 and 5j). Different environments significantly affected protein kg ha⁻¹ genetic gain, which ranged from 0.68 to 6.5 kg protein ha⁻¹ year⁻¹. Expressing kernel protein concentration as a mass showed a significant decrease of −0.05 mg kernel⁻¹ year⁻¹ (Figure 5m) with no significant genetic gain differences across environments (p = 0.31). It should be noted that three hybrids from the early eras substantially deviated from the trendline (and this was consistent across environments), which could be influencing the significance of the slope (Figure 5m). Without including those early hybrids, the protein mass per kernel remained unchanged. Total ethanol produced per kg of grain increased slightly up to the year 2000 (Figure 5i). Greater impacts on ethanol production are derived from the increase in grain yield over time with liters of ethanol ha⁻¹ increasing from 4600 to 6548 (Figure 5l). The overall genetic gain significantly increased by 49 L of ethanol ha⁻¹ year⁻¹ (range: 35–74 L ethanol ha⁻¹ year⁻¹ across environments).

3.5 Genetics and environment effects on traits genetic gain

The genetic gain and magnitude of values for some traits were affected more than others by environmental conditions and management practices explored in this study (Figure 6). Grain moisture, grain yield per fertilizer applied, test weight, oil concentration, protein content, grain yield per plant, and oil content were affected the most by environmental variability, which encompasses different soil–weather conditions and management practices (Figure S1, Table 1). Shelling efficiency and starch concentration were explained the most by genetics (Figure 6). All other traits, such as grain yield, were affected nearly equally by genetics and environment (Figure 6). All traits had a ratio of H × E/H <1.0 (see Figure 6) indicating that hybrids performed similarly across environments. Top-performing hybrids were the top performers across environments, even if the absolute values varied by the environment.

FIGURE 6

Variance components (panel a) with proportions of traits explained by hybrid (genetics), environment, H by E (hybrid × environment), and unexplained variance (residual). Traits are ordered by the H by E/H ratio indicating the amount of trait by environment interaction (panel b). All traits had a ratio of H × E/H <1.0 indicating that hybrids performed similarly across environments.

3.6 Correlation among traits

Kernel number per m² had a 43% stronger positive correlation with grain yield (r = 0.79) compared to kernel weight (r = 0.45; Figure S3). Kernel weight and kernel number had a non-significant weak negative correlation with each other (r = −0.19; Figures 7b and S2). Grain yield was also moderately correlated with shelling efficiency (r = 0.56) and starch concentration (r = 0.62; Figure S3). Decreasing grain protein concentration had a moderate negative correlation (r = −0.68) with grain yield, while increasing starch concentration had a moderate positive correlation with grain yield (r = 0.62; Figures 7 and S3). Further investigation of protein concentration and grain yield showed a linear plateau relationship with protein concentrations decreasing up to 14.5 Mg ha⁻¹ (Figure 7a). Decreases in grain protein concentration had a positive impact on starch concentration and subsequently on ethanol-related traits as they are dependent on starch (Figure S3).

FIGURE 7

Relationship between protein concentration and grain yield (panel a; y = 12.9–0.36x if x < 14.5, y = 7.72 if x ≥ 14.5; R² = 0.49; p-value < 0.001) and kernels per m² and kernel weight (panel b; p-value = 0.2; R² = 0.04). Points are the hybrid best linear unbiased predictors (BLUPs). Dashed line in panel b indicates non-significant regression fit.

4 DISCUSSION

4.1 Grain yield and components

Grain yield genetic gain was found to be continually increasing without a sign of a plateau in any of the 18 environments (Figures 2 and 3a). These consistent results confirm the continual improvement of maize hybrids found in past experiments and update genetic gain information in commercial hybrids to the year 2020 (Figure 8). In this study, we observed two- to threefold higher environmental average grain yields compared to previous short-season era studies (Badu-Apraku et al., 2016; Derieux et al., 1987; Tollenaar, 1989) and at the same time the highest genetic gain (Figure 8). This finding is very encouraging for future progress in grain yield increase and highlights the important role of maize breeding programs. The reported genetic gain in this study is similar to those found in longer-season hybrid era studies (66–143 kg ha⁻¹ year⁻¹; Castleberry et al., 1984; Chen et al., 2016; DeBruin et al., 2017; Duvick, 1984; Fernández et al., 2022; Haegele et al., 2013; Smith et al., 2014). Assefa et al. (2017) analyzed plant density by commercial hybrid experiments (1987–2015) across latitudes in North America and reported grain yield genetic gains comparable to those found in our study (Figure 8).

FIGURE 8

Grain yield genetic gains versus mean environmental yield (large symbols) in our study and prior studies examining short-season hybrids. Genetic gain for Tollenaar (1989) is the reported total grain yield slope, Derieux et al. (1987) is referencing the very early and early hybrids, and Badu-Apraku et al. (2016) is referencing the striga-free conditions. Small green symbols refer to data from Figure 3a. Table S5 provides additional information on the literature studies. Grain yields refer to 15% moisture.

A novel result from our study is that the kernel number increased at twice the relative rate compared to kernel weight while also having the strongest positive correlation with grain yield (Figures 4 and S2). This indicates that kernel number per unit area was the dominant mechanism for increasing the grain yield of the 103-day relative maturity hybrids. Our results are consistent with some previous studies for 111-day hybrids (DeBruin et al., 2017; Fernández et al., 2022; Haegele et al., 2013), but in contrast to some others who found kernel weight to be the dominant grain component factor for increased grain yield (Chen et al., 2016; Mueller et al., 2019). Our results show no negative correlation between grain yield components (Figure 7b). While there was no correlation, it is interesting to note the kernel weight hybrid BLUP values of the 2015 and 2020 eras (Figure 5c). There appears to be a potential plateau forming, but future studies will be needed to determine if this is a trend or an artifact of the hybrids selected. Future breeding to increase kernel number will likely require hybrids with improved radiation use efficiency to increase photo assimilate supply and/or further adaptations to produce more kernels with less available resources (Echarte et al., 2004; Messina et al., 2022). As plant density increases in the US Corn Belt (Assefa et al., 2017), it will be critical to maintain biomass production per plant and ensure adequate nitrogen supply so reductions in kernel weight do not occur (Borrás & Gambín, 2010; Olmedo et al., 2023).

We attribute the increased kernel number found in this study to greater kernel set during the critical period due to greater biomass partitioning to the ear (Andrade et al., 1999; Ruiz et al., 2023; Vega et al., 2001). Gambin et al. (2023) evaluated a subset of our 103-day studied hybrids and reported that the increase in kernel weight was due to the greater kernel growth rate during the grain-filling period. The estimated kernel number genetic gain in this study (25 kernels m⁻² year⁻¹ or 1.85 kernels ear⁻¹ year⁻¹, Figure 5a) agrees well with previous studies (Campos et al., 2006; Fernández et al., 2022; Haegele et al., 2013; Wang et al., 2011). However, the obtained kernel weight genetic gain (0.59 mg year⁻¹, Figure 5c), while similar to findings from one study (Haegele et al., 2013), was lower compared to others who reported genetic gains of 1–1.3 mg year⁻¹ (Chen et al., 2016; DeBruin et al., 2017; Fernández et al., 2022; Russell, 1991). Differences in the magnitude of genetic gain among studies are attributed to different relative maturities, hybrids’ year of release examined, and the number of environments.

The obtained increases in shelling efficiency, kernel-to-ear weight ratio, and positive correlations with both kernel number ear⁻¹ and weight show hybrid improvement of biomass partitioning to the grain (Figures 5d and S2). This partially explains the reported increase in the maize harvest index (Ruiz et al., 2023). Russell (1991) reported increases in shelling efficiency from 84.7% in 1930 to 85.6% in 1980. Our study shows a similar continuing increase from 1980 to 2020 (Figure 5d), which brings information up to 2020. Hybrid BLUP values beyond the year 2010 (shelling efficiency of 88%) indicate a possible physiological limitation to increase this trait further (Figure 5d). A minimum amount of cob is needed to support kernel development, where further reductions in cob mass may begin to limit further increases in kernel number and weight.

4.2 Grain quality traits

We observed a plateau in the kernel protein concentration decline after 2008 (Figures 5g and 7) as opposed to a continuous decline reported in previous studies for 111-day hybrids (Duvick & Cassman, 1999; Haegele et al., 2013; Mueller et al., 2019). The data presented by DeBruin et al. (2017) also suggest a potential plateau in kernel nitrogen concentration in 111-day hybrids released after 2005. It should be noted that no information on protein concentration is available in the literature for short-season hybrids. Prior to 2008, we found protein concentration to decrease linearly from 9.2% (in 1980) to 7.7% (in 2008), with an inversely proportional increase in starch concentration (Figure 5h) in agreement with previous findings (Duvick & Cassman, 1999). This suggests that while protein concentration did negatively correlate with grain yield (Figures 7 and S2), a further decrease in protein concentration may not be needed to improve grain yield and that maize breeding has found other avenues to achieve genetic gains.

Grain moisture at harvest and test weight are critical economic traits that have received little attention in previous studies. Post-harvesting grain drying is commonly needed in the northern part of the US Corn Belt (Figure 1) for grain to be suitable for storage. We found no significant changes in grain moisture with the year of hybrid release in accordance with previous studies (Cavalieri & Smith, 1985; Tollenaar, 1989), but we observed a slight decrease in grain moisture until 1998 (Figure 5f). This may be due to the lower number of husk leaves in newer hybrids (Cavalieri & Smith, 1985). The test weight did not change with year of hybrid release in this study but was affected by environmental conditions (Figures 5e and 6). Our finding does not align with findings from Argentina that found test weight to decrease by 0.059 kg hL⁻¹ year⁻¹ from 1965 to 2016 (Abdala et al., 2018). However, they did find genotypes to explain only 23% of the variation similar to our study. Typically, abiotic and biotic stressors during the grain-filling period influence test weight (Rankin, 2009).

Maize grain is a primary feedstock for ethanol production and livestock feed. Combined, these end-use products account for 73% of total US maize domestic demand in the 2021–2022 marketing year (USDA-NASS, 2023). Reducing the energy demand of maize production is the first step to increasing the efficiency of livestock and ethanol production. The observed increase in starch concentration (Figure 5) enhanced kernel composition for the primary needs of these industries. The significant correlation between grain yield loss with greater protein concentration suggests a role of starch contribution toward improving grain yield. Lower grain yields have been previously observed at higher grain protein concentrations focusing on germplasm bred for high and low protein (Uribelarrea et al., 2004). However, the plateau in decreased protein concentration indicated that while decreases in grain protein may appear to contribute to grain yield gains, it may not be necessary (Figures 5 and 7).

4.3 Environmental and management effects on genetic gains

The grain yield genetic gain was consistently positive across environments, but the magnitude of it varied almost threefold across environments (from 62 to 165 kg ha⁻¹ year⁻¹, Figure 3a). Growing season rainfall (R² = 16%) and vapor pressure deficit (R² = 18%) explained part of the variation (Figure 3), reaffirming the critical role of water availability and crop water stress on grain yield genetic gain (Hsiao et al., 2019). Previous studies have also shown greater grain yield genetic gain under well-watered conditions but also improvements under stress conditions (Campos et al., 2006) and overall higher water use efficiencies due to breeding (Curin et al., 2020; Reyes et al., 2015).

We found excessive N fertilization rates (>220 kg N ha⁻¹) to reduce grain yield genetic gain (Figure 3). This is probably attributed to the fact that excessive N fertilizer rates override the breeding-related improvement in nitrogen use efficiency of newer hybrids (Carlone & Russell, 1987; DeBruin et al., 2017; Haegele et al., 2013; Smith et al., 2014). In our study, the increase in nitrogen fertilizer use efficiency of 0.44 kg kg N⁻¹ year⁻¹ (Figures 2 and 4) is within the range previously reported for 111-day hybrids (0.38 at high nitrogen rates and 0.95 at low nitrogen rates; DeBruin et al., 2017). However, previous studies have shown higher grain yield genetic gain with higher N fertilizer rate, which is the opposite of our results (Castleberry et al., 1984; Chen et al., 2016; Haegele et al., 2013; Mueller et al., 2019). The disagreement between the present findings and the literature is because (1) these prior studies explicitly included multiple N-fertilizer treatments within each experiment; (2) our study integrates experiments with many different soil–weather conditions (Table 1; Figure S1).

Planting date had a negative correlation with grain yield genetic gain (R² = 0.15; Figure 3). Planting delays reduce the time available for plant growth, especially in northern latitudes, leading to grain yield reductions when maize is planted after the optimum window (Baum et al., 2019, 2020; Coulter, 2022). In our study, two of the 18 experiments (#6 and #7, Table 1) were planted after optimal planting dates for Illinois (Nafziger, 1995; Figure 3; Table 1). In contrast to planting date, planting density (range explored: 6.6–10 pL m⁻², Table 1) had no apparent influence on grain yield genetic gain. Previous era studies have shown larger grain yield genetic gain differences with larger plant density ranges within experiments (1–7.9 pL m⁻²; Duvick, 2005).

A unique aspect of our study is the information on 16 secondary traits. Maize breeding affected secondary traits at different rates and/or directions (Figure 4). Furthermore, the environmental conditions affected some traits more than others (Figure 6). Shelling efficiency and starch concentration had the least environmental influence, while grain moisture, test weight, protein amount, and oil concentration had the most environmental influence (Figure 6). All other traits showed nearly equal variability between genetics and environment. In general, our findings agree well with those of Tucker et al. (2020), who studied many traits in 57 commercial Bayer hybrids of 95–115-day relative maturity across three high-yielding environments in Illinois. Interestingly, the G × E/G ratio in our study was less than 1.0 for all traits indicating minimal interaction (Figure 6). This suggests hybrid stability across environments where best-performing hybrids are generally always the best.

4.4 Implications, limitations, and future directions

To our knowledge, this is the first multi-faceted (18 traits) and multi-environment (18 experiments) era study examining 40 hybrids of 103-day relative maturity hybrids across the US Corn Belt region where these hybrids are primarily grown (Figure 1). This region accounts for approximately 25% of the US corn production area, and the importance of 103-day hybrids is further increasing as these hybrids are used in other production regions when planting is delayed (Baum et al., 2020) or when farmers want to harvest early to facilitate other farm activities (i.e., install subsurface tile drainage systems; Castellano et al., 2019, or plant cover crops; Martinez-Feria et al., 2016). In addition to grain yield genetic gain, our study provided data on many secondary traits not previously evaluated in the literature for this maturity group (i.e., starch concentration, kernel weight), which are consequential for maize breeders, crop physiologists, and crop modelers. In companion studies regarding our 103-day hybrids, Ruiz et al. (2023) reported increased harvest index and biomass production, Gambin et al. (2023) reported increased kernel growth rate, dos Santos et al. (2023) reported faster leaf appearance rates with year of release, while Elli et al. (2023) reported that the leaf angle genetic gain is slowing down. Future studies could increase our current knowledge base for short-season hybrids by studying other important traits such as plant nitrogen, root mass, and water use to further our understanding of physiological mechanisms.

Over or underestimation of the genetic gain may have occurred in our study (as well as previous era studies) for multiple reasons. First, historical hybrids were grown at current densities and not at the historical densities for which the older hybrids were originally intended to be planted at. A few prior studies have addressed this issue by investigating grain yield genetic gain for different plant densities and reported that the grain yield genetic gain is overestimated when 1930s hybrids are grown at 2010s plant densities compared to using historical plant densities (DeBruin et al., 2017; Duvick, 2005). However, this concern is small in our study because we studied hybrids from 1980 onwards. Many other published era studies have utilized a single density commonly used at the time of the experiment (Campos et al., 2006; Haegele et al., 2013; Reyes et al., 2015).

Old hybrids are also subjected to environments for which they were not originally selected for. However, the weather conditions the crops experienced during experimentation (years 2020 and 2021) were not that different from the 1980s in terms of temperature, precipitation, radiation, and vapor pressure deficit, according to our analysis (Table S2). The only noticeable difference was the CO₂ concentration. Even in this case, the 50 ppm increase in CO₂ from 1980 to 2020 has less than 1% impact on leaf photosynthesis because of the shape of the relationship between CO₂ and leaf photosynthesis in C4 species (Acock & Allen, 1985). Beyond the environment, improvements in planting efficiency (better planters), seed quality, and harvest efficiency are likely to play a potential, though unquantified, role in relative era hybrid performance.

5 CONCLUSION

Grain yield genetic gain for short-season maize hybrids continually increased across 18 US environments without a sign of a plateau, indicating that maize breeding has successfully improved genetic adaptation over the years. A major contribution of this study is the results for 18 secondary traits, which can inform plant breeders of internal maize plant changes and crop modelers to refine prediction models to capture historical trends better. Two key findings from this work are that the kernel number explained about two-thirds of the grain yield increase, while the decline in protein concentration reached a plateau after 2008 in short-season hybrids. We conclude that there are concurrent changes in many secondary traits with different genetic gains and that there are numerous trait-altering routes that maize breeding can continue exploring to increase grain yields in the future.

AUTHOR CONTRIBUTIONS

Kyle King: Conceptualization; data curation; formal analysis; investigation; methodology; visualization; writing—original draft. Antonella Ferela: Data curation; formal analysis; investigation; methodology; software; visualization; writing—review and editing. Tony J. Vyn: Conceptualization; funding acquisition; investigation; methodology; resources; writing—review and editing. Slobodan Trifunovic: Conceptualization; funding acquisition; project administration; resources; writing—review and editing. Doug Eudy: Conceptualization; investigation; methodology; resources; writing—review and editing. Charles Hurburgh: Investigation; resources; writing—review and editing. Kendall R. Lamkey: Funding acquisition; investigation; resources; writing—review and editing. Sotirios V. Archontoulis: Conceptualization; funding acquisition; investigation; methodology; project administration; resources; supervision; writing—original draft; writing—review and editing.

ACKNOWLEDGMENTS

This study was supported by FFAR (project title: Evaluating the relative influence of maize breeding, field management, and environmental setting on crop production and sustainability targets), Bayer Crop Science, Plant Science Institute of Iowa State University, NSF (#1830478), and the USDA Hatch project (IOW10480). The authors thank the Bayer maize breeding testing teams for helping with setting up experiments, managing, and helping at harvest, and Clarice Mensah for assisting with field coordination. The authors also thank Alejo Ruiz, Caio dos Santos, Cintia Sciarresi, Elvis Felipe, Emily Wiley, Mitch Baum, Makis Danalatos, Mickala Stallman, from Iowa State University, Garrett Verhagen and Lia Olmedo Pico from Purdue University for help with data collection. The authors also thank Allan Gaul from Iowa State University for providing shelling equipment for grain processing.

Open access funding provided by the Iowa State University Library.

CONFLICT OF INTEREST STATEMENT

Slobodan Trifunovic and Doug Eudy are employed by Bayer Crop Science. The other authors declare no conflicts of interest.