Yield Stability Differs in Commercial Maize Hybrids in Response to Changes in Plant Density , Nitrogen Fertility , and Environment

Continued yield increases in modern commercial maize (Zea mays L.) hybrids will require increased plant density, improved nitrogenuse efficiency, and breeding for a hybrid’s potential yield response to this management. The objective of this study was to determine the genetic variation of commercial hybrids in response to plant density and nitrogen (N) fertilizer levels to assist breeding programs to select hybrids with high yield stability or adaptability to crop management. From 2011 to 2014, 101 hybrids were grown in eight different environments at two planting densities (79,000 and 110,000 plants ha-1) and three N rates (0, 67, and 252 kg N ha-1). Broad-sense heritability increased with increased N rate and plant density. Increased plant density altered yield from -0.60 Mg ha-1 to +0.58 Mg ha-1 under high N conditions, whereas the yield response to increased N ranged from +4.47 to +5.64 Mg ha-1. Hybrids that combined above-average yield under unfertilized and low-N conditions exhibited greater-than-average yield stability across environments under high-N conditions. Hybrid yield stability variance was larger under high-N than under low-N conditions because of greater genotype × environment interaction. Hybrids that were adaptable to high plant density and N conditions exhibited greater-than-average yield potential and yield variation across environments. Selecting hybrids with both high yield and yield stability may be difficult, as yield under lower N levels and yield increases with high N fertilization were negatively correlated. Affiliations: A.T. Mastrodomenico, J.W. Haegele, J.R. Seebauer, and F.E. Below, Dep. of Crop Sciences, Univ. of Illinois, Urbana, IL 618014730; A.T. Mastrodomenico, current address: PR-445 Road, km 56.5, Limagrain, Londrina, Brazil PR 86115-000; J.W. Haegele, current address: WinField United, Mahomet, IL 61853. Received 6 June 2017. Accepted 14 Oct. 2017. *Corresponding author (fbelow@illinois.edu). Assigned to Associate Editor Manjit Kang. Abbreviations: BLUP, best linear unbiased predictor; CH, Champaign, IL; CRM, corn relative maturity; DK, DeKalb, IL; HB, Harrisburg, IL; KN, kernel number; KW, kernel weight; NUE, nitrogen-use efficiency; V2-V4, between the twoand four-leaf growth stages. Published in Crop Sci. 58:230–241 (2018). doi: 10.2135/cropsci2017.06.0340 © Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA This is an open access article distributed under the CC BY license (https:// creativecommons.org/licenses/by/4.0/). Published January 15, 2018

increased maize yield during the past 20 yr (Egli, 2008).In comparisons of the genetic gain of maize hybrids under different levels of N fertility, newer hybrids (developed in 1990s) exhibited greater yield than older hybrids (developed in 1970s) under low-and high-N conditions (Tollenaar et al., 1997;O'Neill et al., 2004).Current maize hybrids have greater total N uptake and N utilization (the ratio of yield increase to the difference in plant N content of a fertilized compared to an unfertilized crop); and approximately 70% of the genetic gain in maize yield under high-N conditions is attributable to yield improvement under low-N conditions (Haegele et al., 2013).
Current commercial breeding programs select and develop elite hybrids under optimal agronomic inputs (high-N fertilizer level and the contemporary average plant density of producers in the area), only evaluating a hybrid's yield responses to different crop-management practices at the precommercial stage.Therefore, there are limited reports on genetic variability of elite hybrids for different N use efficiency (NUE) traits (Bertin and Gallais, 2001).The genetic improvement in the yield response to N fertilizer in maize hybrids is well documented (Ding et al., 2005;Coque and Gallais, 2007;Haegele et al., 2013).However, these previous studies used a small representation of elite maize hybrids across a limited number of environments, which may have underestimated the genetic variation of current maize hybrid yields in response to N fertilizer supplementation.
In addition to the genetic improvements in NUE, tolerance to increased plant density is one of the most valuable agronomic advances since the development of the maize hybrid (Duvick, 1977).Maize yield increases across different N fertilizer and plant density conditions were associated with greater N uptake and greater ear sink strength (kernel weight and number) during reproductive development (Ciampitti et al., 2013).The kernel weight component has been found to be more related to the yield increases than to kernel number, and was associated with a longer grain-filling period, improved biomass remobilization during reproductive development, enhanced stress tolerance to N loss, and higher plant densities (Chen et al., 2016).
The most recent maize yield record in the United States (31.5 Mg ha -1 ) was achieved with a plant density (128,000 plants ha -1 ) much greater than that commonly used (National Corn Growers Association, 2015), indicating that the average yield of maize is still far from reaching a plateau and that current maize hybrids exhibit greater crowding-stress tolerance when compared to  The experiment was planted using a precision plot planter with variable seeding rate capability (SeedPro 360, ALMACO, Nevada, IA).Plots were 5.6 m in length, with 0.76-m row spacing and two rows in width.At planting, Force 3G insecticide [(tefluthrin 2,3,5,6- Greensboro, NC] was applied in-furrow at a rate of 0.15 kg a.i.ha -1 to control soil pests.Preemergence herbicide Lumax EZ (mixture of S-metolachlor, atrazine, and mesotrione; Syngenta Crop Protection, Greensboro, NC) was applied at a rate of 7 L ha -1 to control early-season weeds.Postemergence herbicide Roundup [N-(phosphonomethyl)glycine; Monsanto, St. Louis, MO] was applied at a rate of 1.75 L ha -1 when necessary.

MATERIALS AND METHODS
Field evaluations used 101 representative elite singlecross maize hybrids; all were commercially available at the time and adapted to the state of Illinois.These 101 commercial maize hybrids had a variety of biotechnological traits and seed treatments (i.e., fungicides, insecticides, and/or nematicides), were from 11 different seed brands, and had relative maturities ranging from 101 to 117 d (Supplementary Table S1).Due to the latitudinal diversity of the experimental sites, the nature of corn hybrid relative maturities, and the availability of seed, not all hybrids were planted in each site or year.However, on average, 42 hybrids were planted in each environment.

Treatments
To assess the ability of the hybrids to tolerate high plant density conditions (Ruffo et al., 2015), two plant densities (79,000 and 110,000 plants ha -1 , denoted as standard and high plant density, respectively) were used.The standard plant density is based on contemporary farmer practices in the state of Illinois.Final plant stands were recorded prior to harvest.Unfertilized (0 kg N ha -1 ) check plot yields were used to estimate hybrid N stress tolerance, whereas 67 and 252 kg N ha -1 were used to estimate the yield response to low and high N fertilizer, respectively.Nitrogen treatments were broadcast applied as urea their predecessors (Tollenaar and Wu, 1999).One of the possible reasons for the success of increasing plant density to improve maize yield is attributable to the fact that the yield potential of individual plants has not increased in the past 80 yr, rather maize hybrids have better stress tolerance, including the ability to tolerate increased plant densities (Tollenaar and Lee, 2002;Duvick, 2005).Increased plant density typically reduces the yield of individual plants, but increases light interception, and as a result, kernels produced per unit area, thereby increasing the area-wide source-sink ratio (Borrás et al., 2003).In turn, the greater number of potential kernels produced per unit area in density-tolerant hybrids may minimize the yield decreases caused by environmental stresses.
Genotypes that are tolerant to abiotic and biotic stresses are expected to have improved yield stability, that is, improved ability to maintain consistent yield across different environmental conditions (Tollenaar and Lee, 2002).On the other hand, genotypes that are responsive to high-yield environments are considered adaptable genotypes and are expected to show improved yield under favorable agronomic conditions.Stability and adaptability classifications were first proposed by Finlay and Wilkinson (1963) and are based on a hybrid's performances in relation to the corresponding environmental indices (average performance of multiple hybrids in a particular environment).Hybrids with high yield stability have been further characterized as "work-horse" hybrids, whereas hybrids with high adaptability are dubbed as "race-horse" hybrids (Tollenaar and Lee, 2002).
One of the future challenges of maize breeding will be selecting genotypes with higher yields in response to higher plant density, with concurrent improvement in yield stability across environments (Tokatlidis and Koutroubas, 2004).Genetic gain for yield was reduced when maize hybrids were grown under high plant density conditions (de Leon and Coors, 2002;Fasoula and Tollenaar, 2005).These authors attributed the higher genetic gain observed at low plant densities to improved prolificacy and the yield potential of individual plants.However, most current elite hybrids, grown at 79,000 plants ha -1 or greater, are single-eared plants and are better adapted to increased plant densities than older hybrids (Tollenaar et al., 1992).Breeding programs targeting for maize hybrids with improved tolerance to high plant density and utilization of N fertilizer could benefit from a comprehensive evaluation of a representative number of commercial maize hybrids, N rates, plant densities, and environments to develop future selection strategies.Therefore, the objective of this research was to measure the genotype × environment × management interaction by evaluating an extensive assortment of current maize hybrids for yields and classify them for yield stability and crop-management adaptability to improve future breeding programs.
(46-0-0) between the V2 and V4 developmental stages in each environment (Ritchie et al., 1997).Nitrogen application dates were 18 June 2013 and 17 June 2014 at DeKalb, IL; 3 June 2011, 13 June 2013, and 29 May 2014 at Champaign, IL; and 01 July 2011, 25 June 2013, and 13 June 2014 at Harrisburg, IL.The experimental design was a randomized complete block with four replications within each environment in a split-split-plot arrangement.The main plot was hybrid, the split plot was N fertilizer rate, and the split-split plot was plant density level.

Yield and Yield Component Measurements
At maturity, plots were harvested with a two-row plot combine (SPC40, ALMACO, Nevada, IA).Grain yield is reported as Mg ha -1 at 15.5% grain moisture.A representative grain sample from each plot was collected during harvest, from which 300 random kernels were selected and weighed to estimate mean kernel weight (KW).Kernel number (KN) per unit area was estimated from the total plot grain weight, mean individual kernel weight, and final plant density.Seed protein and oil concentrations were estimated using Near Infrared Transmittance (NIT) spectroscopy (Infratec 1241, FOSS, Eden Prairie, MN) from the sample that was used for measuring yield components.

Statistical Analysis and Derived Measurements
Statistical analysis was performed using PROC MIXED statement in SAS version 9.4 (SAS Institute, 2013).Nitrogen fertilizer rates and plant density levels were included in the model as fixed effects, whereas environment, block, and hybrid were considered random effects.The designation of hybrid as a random effect was to allow an inference about the distribution of current maize hybrid yield performances due to the other sources of variation and not examine specific hybrids.The interactions between fixed effects and random effects were included in the model as random effects.The normality of residuals, outlier observations, and assumptions of homoscedasticity were assessed using PROC UNIVARIATE in SAS.
As not all hybrids were planted in every environment, and the objective of this study was to make an inference about the distribution of current maize hybrid performances, best unbiased linear predictors (BLUPs) were calculated within each N fertilizer by plant density treatment using restricted maximum likelihood estimation.Therefore, the phenotypic observations (y ijk ) were modeled according to Eq. [1]: where Y ijk is the phenotypic observation of ith environment within jth block, for kth hybrid, μ is the overall mean, E i is the random effect of ith environment (i = 1, 2,…, and 8), B (i)j is the random effect of jth block ( j = 1, 2, 3, and 4) nested within ith environment, G k is the genetic random effect of kth hybrid (k = 1, 2, …, and 101), (G × E) ik is random effect of the interaction between kth hybrid and ith environment, and e ijk is the random error term.Variance component estimates from this model were used to calculate the broad-sense heritability (H 2 ; the ratio between the genotypic and phenotypic variances) per hybrid mean basis (Holland et al., 2003).
Nitrogen-use efficiency was calculated as the ratio of grain yield increase from the amount of N fertilizer supplied to grain yield from the unfertilized control treatment (Moll et al., 1982).In addition, yield stability and adaptability were calculated using PROC REG, regressing the BLUP estimates from each hybrid (sum of the E i , G k, and G × E ik effects) against the environmental indices (E i ).Pearson's pairwise correlation coefficients (r) between the hybrid b-values (slopes derived from the yields of an individual hybrid across environments when the latter are arranged in increasing average yield order) for different N rates and population densities were calculated using PROC CORR.The mean of all hybrid b-values evaluated was set to 1.0 and hybrids with a b-value equal to 1 were considered average hybrids.High stability hybrids were defined as those with regression slopes of b < 1 ("work-horse"), whereas hybrids with b > 1 were classified as high adaptability, or "racehorse" hybrids (Tollenaar and Lee, 2002).
Hybrids were separated into four groups based on comparison of their yield performance to the mean distribution across environments for four phenotypic traits.The phenotypic traits considered in this study were: (i) unfertilized check plot yield (yield at 0 kg N ha -1 and 79,000 plants ha -1 ), (ii) low N yield response (yield change between 0 and 67 kg N ha -1 at 79,000 plants ha -1 ), (iii) yield response to high N (yield change between 0 and 252 kg N ha -1 at 79,000 plants ha -1 ), and (iv), yield response to plant density (yield change between 79,000 and 110,000 plants ha -1 with 252 kg N ha -1 ).Mean separations for different hybrid groups were analyzed in PROC MIXED using hybrid group as a fixed effect at the 5% significance level.The test for equal stability variances across hybrid groups was performed using the Brown-Forsythe method via PROC GLM (Brown and Forsythe, 1974).

Weather Conditions
Air temperature and rainfall amounts from all environments were obtained from the National Oceanic and Atmospheric Administration (NOAA) and are presented in Supplementary Table S2.July 2011 was hot and dry; at the Champaign, IL site, precipitation was 79 mm below the 10-yr average, whereas both minimum and maximum average daily temperatures were greater than the 10-yr average at Champaign (by +3°C) and Harrisburg (by +2°C) environments, N fertilizer increased maize yield, but plant density did not alter the overall average hybrid yield within an N rate (Table 2).The environments used in this study provided different crop-growing conditions (Supplementary Table S2).Consequently, across the eight location-year environments, environmental indices for the average yield with high N conditions (252 kg N ha -1 ) deviated from the overall mean by -1.6 to +2.4 Mg ha -1 at the standard plant density (79,000 plants ha -1 ) and by -1.9 to +2.3 Mg ha -1 at the higher plant density (110,000 plants ha -1 ) (Fig. 1).Although hybrids exhibited similar yields at both plant densities, the highestyielding hybrids within each environment usually resulted from the higher plant density and the highest rate of N supply (data not shown).
Maize genetic improvement for NUE has been more attributed to improved yield under low N than to yield increases with N fertilizer (Haegele et al., 2013).In this study, maize yield under low N conditions (0 kg N ha -1 ) accounted for, on average, 55 and 52% of the yield under high N at the standard and high plant densities, respectively (Table 2).Low N (67 kg N ha -1 ) increased yield by +2.8 and +2.9 Mg ha -1 over the unfertilized control, and the yield response to high fertility (i.e., the yield increase between 0 and 252 kg N ha -1 ) accounted for +5.1 and +5.6 Mg ha -1 at the standard and high plant densities, respectively.As a result, the NUE at the low (Supplementary Table S2).The late vegetative to early reproductive developmental stages of maize usually occur during July to August in the U.S. Corn Belt; these are important growth stages in determining maize yield.Data from 2012 of the trial were excluded from the analysis because of severe drought stress.In contrast, the weather of 2013 provided ideal environmental conditions for maize development and yield at all three sites, with daily temperatures similar to or cooler than the 10-yr average, and aboveaverage precipitation during June (+83, +49, and +22 mm at DeKalb, Champaign, and Harrisburg, respectively).In 2014, above-average precipitation also occurred during June (+117, +98, and +20 mm at DeKalb, Champaign, and Harrisburg, respectively, compared with the 10-yr average), and below-average temperature occurred in July, with both minimum and maximum average daily temperatures deviating from the 10-yr average by -2 to -3°C for all sites.The environmental conditions in 2014 led to a statewide record corn yield of 12.5 Mg ha -1 (USDA National Agricultural Statistics Service, 2017).

Hybrid Yields and Variance Components
Overall, maize grain yield was affected by the environment, hybrid, nitrogen rate, and their interactions (Table 1).In contrast, the population change was only significant in combination with environment (P = 0.012).Averaged across Table 1.Analysis of variance for yield of 101 maize hybrids (Hyb.)grown at eight environments (Env.; covering three locations and three years), two plant density populations (Pop.), three N fertilizer rates (Nrate), and four replications (Rep.).Table 2. Plant density and N fertilizer effects on yield from BLUP analyses, variance components from Eq. [1], broad-sense heritability (H 2 , estimated on hybrid-mean basis), and nitrogen-use efficiency (NUE).Values are averaged across 101 maize hybrids grown at three locations (DeKalb, Champaign, and Harrisburg, IL) for three years (2011, 2013, and 2014).N rate was greater than the NUE at high N. Increased plant density may increase NUE under greater N fertilizer rates (165 and 330 kg N ha -1 ) (Boomsma et al., 2009).However, the data in Table 2 show no change in NUE with increasing plant population.While other studies found maize hybrids showing improved tolerance to crowding stress (Tollenaar and Lee, 2002) in combination with NUE improvement (Tollenaar and Wu, 1999), possibly the diverse range of hybrids in this study might have concealed the improvements in these parameters.Environmental (s 2 E ), genetic (s 2 G ) and the genotype × environment interaction (s 2 G × E ) variances for yield differed across N rates and plant densities (Table 2).While the environmental variance for yield decreased under higher N fertility conditions, the genetic variance tended to increase.Other investigators have also reported a reduction in environmental variance for yield because of better agronomic conditions (Bänziger and Cooper, 2001).Also, high N conditions may have reduced the soil heterogeneity and allowed for an increased genetic effect (Bertin and Gallais, 2001).At the standard plant density (79,000 plant ha -1 ), N fertilizer increased yield s 2 G × E over the unfertilized control, but at the high plant density, the s 2 G × E remained similar regardless of the N fertilizer rate.
High N fertility conditions can increase heritability (Brun and Dudley, 1989;Bänziger et al., 1997;Bertin and Gallais, 2001) or decrease heritability (Agrama et al., 1999) depending on the germplasm and the agronomic conditions used for evaluation.Higher heritability under high N conditions may increase the effectiveness of selection and be used as indirect selection for maize genetic improvement under low N (Gallais et al., 2008).Conversely, a reduced response to selection has been found for maize hybrids grown under high plant densities because of reduced stand uniformity and individual plant yield (Fasoula and Tollenaar, 2005).In this study, genetic variance for yield increased by 41% and broadsense heritability increased by 15% with increased plant density (Table 2).Therefore, current maize hybrids are more tolerant of crowding stress as a group, and under these conditions, the greater variability in yield potential allows for improved NUE and yield.

Yield Components and Grain Quality
Although plant density treatments did not affect overall yields, they influenced yield components (Fig. 2).Increased plant density has been found to reduce total leaf area per plant, individual kernel weight (KW), and kernel number (KN) per plant (Borrás et al., 2003).Under low N conditions, KW and KN were not affected by plant density.However, the highest N rate increased both KW and KN by 16 and 34% at the standard plant density, and by 13 and 40% at high plant density, respectively, over the unfertilized controls.Kernel number often exhibits greater plasticity than KW when maize hybrids are exposed to different agronomic conditions (Sadras and Slafer, 2012;Boomsma et al., 2009).Similarly, KN was more correlated to yield than KW across all N and plant density treatments (Table 3).Maize seed set and the resulting KN is sensitive to N-stress (Below et al., 1981).Accordingly, the highest correlation coefficients were found between yield when unfertilized and KN at both plant densities.Correlations between KW and KN with yield tended to decrease with increased N rate, regardless of the plant density; but KW was more correlated to yield at the standard plant density than at the high plant density.Reduced correlations between yield components and yield with increased N fertilizer may be associated with increased genetic variance (Table 2) and with specific hybrid grain characteristics.Individual mean KW was under greater genetic control than KN at 0 and 67 kg N ha -1 , but under high N, both yield components were highly heritable.
In addition to yield components, both N fertilizer and plant density had an effect on grain quality (Table 4).Grain oil concentration was stable across plant density treatments.Therefore, because of increased yield, the highest N fertilizer rate increased oil content by approximately 35% over the unfertilized control, regardless of the plant density.In addition, the highest N fertilizer rate increased grain protein content by 60% at both plant densities over the unfertilized control because of increases in both yield and protein concentration.Although grain protein and oil proportions have a negative relationship (Simmonds, 1995), increases in N availability increased the content per unit area of both traits, whereas no changes in grain quality were observed with increased plant density.Nitrogen fertility affected protein more than oil content, most likely because of the inherent composition of proteins.These results using 101 commercial hybrids revealed less genetic variation in grain protein and oil concentration than previous studies (Below, et al., 2004, Uribelarrea et al., 2007).

Hybrid Characterization
The large genetic variance for yield response to N fertilization and plant density highlights the importance of hybrid characterization to identify proper agronomic management.There was no difference between different seed brands among maize hybrids' responses to N fertilizer and plant density (data not shown).At high N, the majority of the 10 highest-yielding and the 10 lowest-yielding maize hybrids were observed at the higher plant density (Table 5); indicating that hybrids with tolerance to crowding stress under high N conditions can produce high yields, while hybrids that are susceptible to crowding stress have a more limited yield, regardless of N fertilizer availability.While full-season hybrids can exhibit greater biomass plasticity and partitioning to the grain in response to plant Table 3. Pearson's pairwise correlation coefficients (r) between maize yield and yield components (kernel number and individual kernel weight) at different N fertilizer rates and plant densities.Values are averaged across 101 hybrids grown at three locations (DeKalb, Champaign, and Harrisburg, IL) for three years (2011, 2013, and 2014).Table 4. Maize grain oil and protein concentration and content responses to N fertilizer supply and plant density.Values are averaged across 101 hybrids grown at three locations (DeKalb, Champaign, and Harrisburg, IL) for three years (2011, 2013, and 2014).
N fertilizer rate (kg N ha -1 ) Kernel parameter 0 † 67 252 --------------------------------110,000 plant ha -1 --------------------------------Oil concentration, g kg density compared with short-season hybrids (Sarlangue et al., 2007), there was no significant relationship for yield found in this study between hybrid maturity (CRM) and tolerance to crowding stress (data not shown).Contrasting results obtained by Sarlangue et al. (2007) may be attributed to the genetic improvement of current early maturity hybrids with increased reproductive sink capacity or to different agronomic conditions.Although the 10 highest-yielding hybrids achieved similar yields under high N conditions, these hybrids exhibited differential NUE and yield component proportions across N rates and plant densities (Table 5).The 10 highest-yielding hybrids exhibited, on average, higher NUE (P ≤ 0.01) under high N (252 kg N ha -1 ) than the 10 lowest-yielding hybrids.Similarly, the 10 highest-yielding hybrids had greater KN at high N fertility compared to the 10 lowest-yielding hybrids, but there were no significant differences in KN at low N, or in KW between the groups.Differences in NUE between hybrids when grown under high N are attributed to both the difference in a hybrid's yield performance when unfertilized and its yield response to N fertility.
To identify hybrids with good agronomic performance under conditions associated with lower soil N availability (i.e., lower rates of N fertilizer application and/or weather conditions leading to N loss), check-plot (0 kg N ha -1 fertilization) yields were compared with the response to low N (Fig. 3A).Yields without fertilizer ranged from 5.87 to 6.78 Mg ha -1 and the yield response to low N ranged from +2.40 to +3.14 Mg ha -1 among all hybrids when averaged across all environments.Subsequently, hybrids were divided into four groups: group 1A (42% of the hybrids) included those that yielded below average when unfertilized but had an above-average yield response to low N; group 2A (9% of the hybrids) included those with above-average yield both when unfertilized and in response to low N; group 3A (4% of the hybrids) included those with below-average yield when unfertilized and in response to low N; and group 4A (45% of the hybrids) included those with above-average yield when unfertilized but below-average yield response to low N. Newer cultivars with transgenic traits have been found to have greater NUE and yield at low N, with a corresponding lower N rate to maximize yield (Haegele et al., 2013) and are more represented in groups 1A and Table 5. Plant density effects on yield, N-use efficiency (NUE), kernel number (KN), and kernel weight (KW) from the 10 highest and lowest-yielding maize hybrids grown under high N conditions (252 kg N ha -1 ).Values are the average of three locations (DeKalb, Champaign, and Harrisburg, IL) and three years (2011, 2013, and 2014).Low and high N for NUE indicate the values with N fertilizer supplied at 67 and 252 kg N ha -1 , respectively.Low and high N for ΔKN and ΔKW indicate the change in kernel number and kernel weight between 67-0 and 252-67 kg N ha -1 , respectively.2A, possibly because of greater root mass (Head et al., 2017)(Fig.3A).Because yield from unfertilized plots and the yield response to low N were negatively correlated (r = -0.35,P < 0.05), selecting hybrids with above-average performance for both traits (Group 2A) may be challenging in a maize-breeding program.Hybrids adaptable to intensive crop management were identified as those with high yield response to both high N fertility and increased plant density (Fig. 3B).The yield response of hybrids, averaged across environments, to high N fertilizer (252 kg N ha -1 ) ranged from +4.47 to +5.64 Mg ha -1 and the yield response to increased plant density ranged from -0.60 to +0.58 Mg ha -1 .Compared with results from previous studies under similar agronomic conditions, the findings reported here identified substantially greater yield ranges for both response to high N fertilizer (Haegele et al., 2013) and increased plant density (Ruffo et al., 2015), demonstrating greater genetic variability for these traits in current elite maize hybrids.Moreover, four groups of hybrids were identified: group 1B included 25% of the hybrids with below-average yield response to high N but above-average response to increased plant density; group 2B included 24% of the hybrids with above-average yield responses to both high N and increased plant density; group 3B contained 25% of the hybrids with below-average yield response to both high N and plant density; and group 4B consisted of 26% of the hybrids with above-average yield response to high N but below-average response to increased plant density.Under standard plant density and high N, groups 2B and 3B obtained the same average yield of 11.64 Mg ha -1 , which was significantly higher (P < 0.001) than those of groups 1B (11.32 Mg ha -1 ) and 4B (11.19 Mg ha -1 ).However, under high N and high plant density, group 2B (11.99 Mg ha -1 ) had a significantly higher yield (P < 0.001) than groups 3B (11.15 Mg ha -1 ), 4B (11.59 Mg ha -1 ), and 1B (11.67 Mg ha -1 ).Maize hybrids that were adapted to increased N fertilizer and plant density exhibited higher than average yield and would be more suitable for intensive crop management practices.

Hybrid Stability Analysis
Yield stability was evaluated across different N rates and plant densities for 61 maize hybrids that were grown in at least three environments.Hybrids exhibited similar yield stability correlations between N levels across different plant densities (Table 6).Overall, phenotypic correlations for hybrid stability (b-values) across N treatments and plant densities for yield ranged from nonsignificant to 0.85 (P ≤ 0.001).Correlation coefficients tended to be larger at standard plant density compared with high plant density.In addition, hybrid yield stability appeared to be more associated with a hybrid's tolerance to N stress when unfertilized than with a hybrid's response to N fertilizer or tolerance to crowding stress.Increased s 2 G × E at high N and high plant density may have contributed to decreased correlation coefficients for yield stability.
Hybrids characterized with different crop management responses also exhibited different characteristics of yield stability across N and plant density treatments (Table 7).Group 2A (hybrids with above-average yield when unfertilized and in response to 67 kg N ha -1 ) had better Fig. 3. Characterization of 101 elite maize hybrids under different agronomic conditions: a) relationship between yield when unfertilized (0 kg N ha -1 ) and in response to low N (67-0 kg N ha -1 ) and b) relationship between yield response to greater plant density (110,000-79,000 plt ha -1 ) at high N conditions and yield response to high N (252-0 kg N ha -1 ).Hybrids were grouped based on below or above average performance for each phenotypic trait.Values are averaged across three locations (DeKalb, Champaign, and Harrisburg, IL) and three years (2011, 2013, and 2014).Dashed lines represent the average performance from all hybrids within each phenotypic trait.
yield stability (smaller b-value) than groups 1A, 3A and 4A at 67 kg N ha -1 under both plant densities and higher yield stability than group 4A at 252 kg N ha -1 at standard plant density.Yield stability of hybrid group 2A tended to increase with additional N fertilizer, regardless of the plant density, indicating that hybrids with above-average tolerance to N loss will provide consistent performance across high-yield environments and that this stability is plant density-independent.On the other hand, hybrid group 2B (hybrids with above-average yield response to both high N and increased plant density) exhibited average yield stability across N and plant density treatments.Hybrids with above-average response to high N and below-average response to increased plant density (group 4B) had higher yield stability than did hybrids with below-average yield response to high N and aboveaverage response to increased plant density (group 1B).
In previous studies, reduced response to selection for yield at high plant density has been attributed to reduced stand uniformity, increased plant-to-plant variability, and reduced plant prolificacy (Hallauer and Sears, 1969;de Leon and Coors, 2002;Fasoula and Tollenaar, 2005).However, the improvement of maize agronomic management across time has increased stand uniformity and reduced plant-to-plant variability.In addition, current maize hybrids appeared to be more density-dependent, Table 6.Pearson's pairwise correlation coefficients (r) for yield stability (b-values) from sixty-one maize hybrids grown at different N rates and plant densities and averaged across three locations (DeKalb, Champaign, and Harrisburg, IL) and three years (2011, 2013, and 2014).
Plant density N rate 79,000 plant ha -1 110,000 plant ha -1 0 67 252 0 67  0.01b 0.03b 0.41a 0.02b 0.06b 0.27a † Hybrid groups: 1A (hybrids with below average yield under unfertilized conditions and above average yield response to low N), 2A (hybrids with above average yield both unfertilized conditions and in response to low N), 3A (hybrids with below average yield under unfertilized conditions and in response to low N), 4A (hybrids with above average yield under unfertilized conditions and below average yield response to low N), 1B (hybrids with below average yield response to high N and above average response to increased plant density), 2B (hybrids with above average yield responses to both high N and increased plant density), 3B (hybrids with below average yield response to both high N and plant density), and 4B (hybrids with above average yield response to high N and below average response to increased plant density).‡ Least square difference for yield stability between hybrids within N and plant density treatments (P ≤ 0.05); NS, Nonsignificant.§ Within a plant density level, variances followed by the same letter are not significant different according to the Brown-Forsythe method (P ≤ 0.05).
exhibit reduced barrenness, and are mostly single-ear hybrids.Hybrids with a large yield response to high N fertilizer rates and that are also plant density-independent (produce less-than-average yield decreases from increased plant density) may have greater yield stability under standard plant density.Yield stability variance significantly increased (P ≤ 0.05) with additional N fertilizer at both plant densities (Table 7).Under high N conditions, yield stability variance was notably higher under standard plant density than under the high plant density condition.Possibly, the plasticity of the typically greater number of kernels per plant and kernel weight produced at lower densities led to this higher yield stability variance (Boomsma et al., 2009).

CONCLUSIONS
Current elite maize hybrids expressed large genetic variation for yield in tolerance to N deficiency, N fertilizer response, and tolerance to crowding stress when grown across different environments in Illinois.Yield stability was more associated with a hybrid's ability to tolerate N stress and respond to N fertilizer than with a hybrid's ability to tolerate high plant density conditions.Selecting hybrids with above-average yield performance under both unfertilized and low N conditions may be challenging in a maize breeding program, as less than 10% of the hybrids evaluated in this study were characterized as such.These hybrids achieved greater than average yield stability under high-yield environments.Although hybrids with an above-average yield response to high N fertilizer supply and increased plant density produced higher yields, these hybrids also had larger-than-average stability variance, suggesting that hybrid selection will be more advantageous under specific growing environments.Future research evaluating hybrids' responses to agronomic management (e.g., N fertilizer, plant density, and row spacing) may be integrated with different hybrid selection methods for the development of maize hybrids with improved agronomic performance.

Fig. 1 .
Fig. 1.Influence of location, year, and plant population on grain yield, arranged by increasing average yield (Environmental index) of the normal and high population plots for 101 maize hybrids grown at DeKalb, IL (DK), Champaign, IL (CH), and Harrisburg, IL (HB) under high N conditions (252 kg N ha -1 ) in 2011, 2013, and 2014.Dots represent yield estimates for individual hybrids within each environment and population.Horizontal lines in the box plot indicate the median, top and bottom edges of the box refer to 75th and 25th percentiles, and whiskers extend to the 10th and 90th percentiles.

Fig. 2 .
Fig. 2. Plant density and N fertilizer rate effects on average kernel number, kernel weight, and broad-sense heritability (estimated on hybrid-mean basis) for 101 maize hybrids grown at three locations (DeKalb, Champaign, and Harrisburg, IL) for three years (2011, 2013, and 2014).Bars extending from data points indicate ± 1 standard error for 95% significance level.

Table 7 .
Plant density and N fertilizer effects on yield stability (b-value) of 61 maize hybrids averaged across eight environments and three years.Hybrids were grouped based on their average yield response to N fertilizer and plant density.Yield stability variance was calculated from all hybrids tested within each N fertilizer and plant density treatment.