Field Crops Research

Transcription

1 Field Crops Research 140 (2013) Contents lists available at SciVerse ScienceDirect Field Crops Research jou rnal h om epage: Maize root growth angles become steeper under low N conditions S. Trachsel a, S.M. Kaeppler b, K.M. Brown a, J.P. Lynch a, a Department of Plant Science, The Pennsylvania State University, University Park, PA 16802, USA b Department of Agronomy, University of Wisconsin, Madison, WI 53706, USA a r t i c l e i n f o Article history: Received 2 May 2012 Received in revised form 12 September 2012 Accepted 16 September 2012 Keywords: Zea mays L. Root architecture Crown root Brace root Nitrogen a b s t r a c t Root traits that increase the speed and effectiveness of subsoil foraging may enhance nitrogen acquisition in leaching environments. We investigated root depth distribution of maize genotypes across the cropping cycle, effects of root angles on plant performance and potential plastic responses of root growth angles to nitrogen fertilization. We focus on genetic variation for growth angles of crown and brace roots among 108 inbred lines of maize in high and low nitrogen field environments in the USA and South Africa. Root angles of crown roots were significantly associated with rooting depth calculated as the depth containing 95% of the root mass (D 95 ). The number of brace roots as well as rooting depth (D 95 ) increased between 43 days after planting (DAP) and flowering, but did not show any major changes between flowering and physiological maturity. Brace root branching increased between 43 DAP and flowering and showed reductions between flowering and physiological maturity. Under well-fertilized conditions genotypes initially selected as steep and shallow did not alter their root angles. Brace and crown root angles became up to 18 steeper under nitrogen deficient conditions. Increases in root angles under nitrogen deficient conditions were more accentuated for shallow genotypes, resulting in root angles and rooting depths similar to the ones measured for steep genotypes. Steeper root angles enabled plastic genotypes to potentially explore similar soil volumes under nitrogen deficient conditions as steep genotypes, thereby not incurring any reductions in grain yield compared to genotypes constitutively forming steep root angles. Additive main and multiplicative interaction effects (AMMI) analysis revealed that out of 29 genotypes best adapted to 4 different nitrogen fertilizer treatment-by-location combinations, 11 were steep, 11 were plastic and 7 were shallow genotypes. The number of plastic genotypes among the adapted entries was disproportionately high compared to 6 that could be anticipated based on the distribution in the entire genotypic set. We postulate that modulation of rooting depth by root growth angles is important for nitrogen acquisition by positioning roots in soil domains with the greatest nitrogen availability. Genotypic variation in root growth angles and the plasticity of root growth angles in response to nitrogen may be useful in breeding crops with improved nitrogen acquisition Elsevier B.V. All rights reserved. 1. Introduction Suboptimal soil nitrogen availability is a primary constraint to crop production in developing countries (Hu et al., 2008). In Abbreviations: AMMI, additive main and multiplicative interaction effects; BA, angles of brace roots ( ); BB, branching of brace roots (# of lateral roots cm 1 ); BN, number of brace roots (count; BW, number of brace root whorls (count); CA, angles of crown roots ( ); CB, branching of crown roots (# of lateral roots cm 1 ); CN, number of crown roots (count); D 95, root depth above which 95% of the root mass is located (cm); DAP, days after planting (days); GDD, growing degree days (GDD); GY, grain yield (kg ha 2 ); K, potassium; N, nitrogen; P, phosphorus; PHT, plant height (cm); QTL, quantitative trait loci; SD, stem diameter (mm); SPAD, single photon avalanche diode. Corresponding author at: Department of Plant Sciences, Penn State, Tyson 221, University Park, PA 16802, USA. Tel.: address: jpl4@psu.edu (J.P. Lynch). wealthy nations, intensive nitrogen fertilization is the primary energy cost and economic cost of the production of key crops, and a primary source of air and water pollution (Mishima et al., 2011). It is estimated that less than half of nitrogen fertilizer applied to maize in the USA is actually taken up by the crop, the remainder largely being leached resulting in pollution (McIsaac and Hu, 2004). Nitrogen fertilizer costs have been increasing, driven by the increasing cost of fossil fuels. Crop genotypes with enhanced ability to acquire soil nitrogen would therefore have utility in enhancing food production in developing nations, and in enhancing the profitability and sustainability of crop production in wealthy nations. The predominant form of nitrogen in agronomic systems is nitrate, which is highly soluble in water and leaches with precipitation into deeper soil strata (Thorup-Kristensen et al., 2009). Rapid leaching of nitrate below the effective rooting zone of the developing crop is a primary source of nitrogen loss in agricultural systems (Raun and Johnson, 1999; Kristensen and Thorup-Kristensen, /$ see front matter 2012 Elsevier B.V. All rights reserved.

2 S. Trachsel et al. / Field Crops Research 140 (2013) ). Root traits that permit rapid exploitation of deeper soil strata may therefore improve nitrogen acquisition. The steep, cheap, and deep ideotype consists of several root traits that may enhance nitrogen and water acquisition by maize root systems by improving the speed of subsoil exploitation (Lynch, 2009). Steep refers to architectural phenes such as growth angles of crown and brace roots that increase the depth of soil exploration. Cheap refers to architectural and anatomical phenes that reduce the metabolic cost of soil exploration, which is an important limitation to root growth and development (Lynch and Ho, 2005). Our focus here is on the growth angles of crown and brace roots as important determinants of effective rooting depth. Root architecture has a key influence on nutrient and water acquisition by positioning root foraging activity in specific soil domains in time and space (Lynch, 1995, 2011; Forde, 2009). Modeling suggested that topsoil root density plays the biggest role in reducing nitrate losses under conditions where the nitrate source is in the topsoil and leaching is rapid (Dunbabin et al., 2003). Exploration of larger soil volumes at greater depth has been predicted to be better enhanced by a more sparsely branched herringbone like root system (Dunbabin et al., 2004). The main factors determining root depth distribution and rooting depth are the length and number of crown roots (Abe and Morita, 1994) and root angle (Araki et al., 2000; Singh et al., 2010). Oyanagi (1994) found that set point angles of seminal roots in wheat were strongly correlated with depth reached by 50% of the seminal roots. It is to be expected that the formation of a root architecture with steep root angles will be beneficial under conditions where nitrate solubilized in water is being leached or both water and nitrate can only be found at greater soil depth resulting in greater nutrient acquisition efficiency. We therefore hypothesize that, in maize, a plastic root system forming shallow or steep root angles depending on nutrient availability and distribution in the soil would be beneficial for nutrient acquisition and plant performance. A shallow root system would be beneficial under well-fertilized conditions, when nutrients predominantly originate from the decomposition of organic matter on the soil surface or when phosphorus or potassium (which are predominantly found in the topsoil) are limiting. A steep root system would be favorable under conditions in which water and therein solubilized nutrients can predominantly be found in deeper soil layers i.e. under terminal drought or under conditions with a high potential for nutrient leaching (high precipitation, soils with low nutrient retention capacity). The gravitropic setpoint angle and gravitropic response of roots are affected by genetic, physiological and environmental factors (Oyanagi et al., 1993). Temperature in maize (Nagel et al., 2009) and phosphorus availability in beans (Bonser et al., 1996; Liao et al., 2001) have been found to affect root angles, while interactions between light, ph, water and nitrate possibly interacting with auxin (Vidal et al., 2010) and abscisic acid modulate root angles in Arabidopsis (LaMotte and Pickard, 2004). Detection of QTL for root angles in plants grown in containers (Norton and Price, 2009; Liao et al., 2004) and under field conditions in maize (Trachsel et al., unpublished) as well as gene expression networks (Vidal et al., 2010) indicate a strong genetic component influencing root angles. Root angles and their response to nutrient availability have been investigated in sand-filled pots in beans (Liao et al., 2001, 2004), on agar plates in Arabidopsis (Wolverton et al., 2011), in perspex growth chamber in rice (Norton and Price, 2009) as well as for their response to different temperature regimes in growth pouches in maize (Hund et al., 2009). These artificial systems do not mimic highly heterogeneous natural growth conditions as encountered in soils. Root growth might be affected by seed reserves in studies carried out at the seedling stage and might be constrained by the size of the volume/area of the growth media or container in studies carried out at later stages. Relative to undisturbed soil, differences in soil bulk density as a result of sieving and recompacting as is the case for column studies can equally affect elongation rate and trajectory of growing roots. Moreover the root system is buffered from the atmospheric environment in a completely different way when grown in a small container compared to the field. Hence, there is a high risk for artifacts of root growth or of root-shoot-interaction in such investigations, when aiming to simulate field-situations (Walter et al., 2009). Nakamoto et al. (1991) circumvented these issues by investigating wheat root angles under field conditions on volcanic ash soils. Evaluation of root angles in the field allows for direct examination of root architecture on plant performance. To date it remains to be elucidated if and how root angles in maize respond to differences in fertilization and resulting nitrogen availability and whether those changes have an effect on nitrogen acquisition and plant performance under field conditions. Evaluation of roots in the field is laborious and time consuming. Trachsel et al. (2011) proposed a simple, cheap high-throughput method that allows the investigation of root architecture in general and more specifically root angles in a natural environment under field conditions in a few minutes: shovelomics. By this process, maize plants are excavated at flowering, root crowns are soaked and cleaned from surrounding soil. Traits affecting root architecture such as the number, angles, branching of crown and brace roots can then be evaluated. The objectives of this study were to (i) investigate root depth distribution of steep and shallow maize genotypes across the cropping season under different levels of nitrogen fertilization in the field, (ii) investigate the influence of nitrogen fertilization on root angles and (iii) quantify the effect of root angles on plant performance under different levels of nitrogen fertilization. 2. Materials and methods 2.1. Experimental site Experiments were carried out in 2009 at the Russell Larson Research and Education Center of the Pennsylvania State University in Rock Springs, PA, USA ( N, W, 366 masl), in 2010 at the Hancock Agricultural research station of the University of Wisconsin in Hancock, WI, USA ( N, W, 331 masl) and in the 2010/2011 season in Alma, Limpopo, South Africa ( S, E, 1235 masl). The experiments were conducted on a Hagerstown silt loam (fine, mixed, semiactive, mesic Typic Hapludalf) in Rock Springs, a Plainfield loamy sand (mixed, mesic Typic Udipsamment) in Hancock and a Clovelly loamy sand (Typic Ustipsamment) in Alma Field management In 2009, genotypes were planted on June 1 in Rock Springs (Table 1) and in 2010 on May 28 in Hancock. In 2009 in Rock Table 1 Duration of the growing period, growing degrees days (GDD), mean temperature and precipitations at root harvest and physiological maturity during the growing season of 2009 in Rock Springs, PA, 2010 in Hancock, WI and Alma, LP, ZA in C was used as a base temperature for the calculation of GDD. Rock Spring 2009 Hancock 2010 Alma 2011 Planting date 1 June 28 May 14 January Harvesting date 10 October 5 October 27 April GDD shovelomics/maturity 774/ / Average T ( C) 20.3/ / shovelomics/maturity Precipitation (mm) shovelomics/maturity 254/ /

3 20 S. Trachsel et al. / Field Crops Research 140 (2013) Springs plants were evaluated 43 days after planting (DAP), at flowering (76 DAP) and at physiological maturity (112 DAP). In 2010 and 2011 plants were evaluated for root architecture at flowering only. Yield was measured in 2009 at Rock Springs and in 2010 at Hancock. No yield data were collected in 2011 at Alma. In 2009, plants were sampled on August 11 at Rock Springs and in 2010 at Hancock on August In 2011 at Alma genotypes were planted on January 14 and harvested on May 2 6. At root sampling, plants had accumulated 774 (physiological maturity: 1054) growing degree days (GDD) at Rock Springs in 2009, 982 (1371) GDD at Hancock in 2010 and 1081 in 2011 at Alma. GDD were calculated from air temperature with a base temperature of 10 C and a maximum of 28 C. The mean temperature during the 2009 season was 17.9 C (root sampling 20.3 C) at Rock Springs and 20.6 C (21.6 C) at Hancock in 2010 and 19.5 C during 2011 at Alma. Precipitation within the experimental period was 321 mm (root harvest 254 mm) at Rock Springs in 2009, 513 mm (481 mm) at Hancock in 2010, and 151 mm in 2011 at Alma. Row width was 75 cm, and distance between plants within a row was 23 cm, resulting in an overall planting density of 6 plants m 2. Based on soil analysis at the beginning of the cropping season, all plots under optimum nitrogen conditions at Rock Springs were fertilized with urea at a rate of 100 kg N ha 1, while all plots were fertilized with 33.6 kg K 2 O ha 1. The well fertilized plots at Hancock were amended with 222 kg N ha 1 while the low N treatments were amended with 71 kg N ha 1. The plots at Alma were amended with 35 kg K 2 SO 4 ha 1 and 28 kg KH 2 PO 4 ha 1 split over the cropping season, the well-fertilized plots were fertilized with a combination of (NH 4 ) 2 SO 4 and NH 4 NO 3 resulting in 100 kg N ha 1 split over the course of the cropping season while the plots assigned to the nitrogen deficiency treatment were fertilized with (NH 4 ) 2 SO 4 at a rate of 26.5 kg ha 1 at the beginning of the cropping season only. In all environments, pest control and irrigation were carried out as needed. Two days prior to root sampling, the fields were irrigated using an irrigation cannon at Hancock and Rock Springs or an irrigation rig at Alma with 13 mm of water to soften the soil in order to facilitate excavation of root crowns. Throughout the vegetation period irrigation was used to compensate for water lost by evaporation and transpiration on a weekly basis. This generally resulted in the irrigation of fields every 1 2 weeks in Alma while it was applied at irregular intervals in Hancock and Rock Springs as weekly precipitations were sufficient to sustain plant growth Plant material Based on previous experiments carried out under well-fertilized conditions (Trachsel et al., 2011), 8 genotypes with contrasting root angles (subsequently referred to as the group of 8) were grown in 2009 in Rock Springs, 2010 in Hancock and 2011 in Alma. Genotypes classified as steep were MO368, MO79, OH21 and OH48, genotypes classified shallow were NyH227, NyH180, MO85 and OH206. Lines with the MO designation are publicly available lines from the IBM population, and lines with the OH and NyH designation are recombinant-inbred lines that we have developed from crosses of Oh43xW64A and Ny821xH99, respectively. 100 Recombinant inbred lines (RILs) from the cross of B73 with 25 diverse parents were selected from the NAM (McMullen et al., 2009) based on contrasting root architecture. Selection for root angles was based on crown and brace root angles measured in the NAM grown in 2010 in Alma (Trachsel et al., unpublished). Half of the RILs had been selected for shallow root angles while 50 were selected based on steep root angles (Appendix 1). 3. Measurements at sampling 3.1. Shovelomics Phenotypic evaluation of root crowns was carried out based on shovelomics (Trachsel et al., 2011) with minor modifications. Selection of three plants to be excavated within a plot was based on plant height and general appearance so that the selected plants were representative of the individuals in the plot and were bordered on both sides by a plant 23 cm distant. Since selected root crowns within a plot were expected to be homogeneous, root crowns of the three plants per plot were bulked for evaluation and only a single rating was recorded. At harvest, roots were excavated by removing a soil cylinder of 40 cm diameter and a depth of 25 cm with the plant base as the horizontal center of the soil cylinder. Excavation was carried out using spades with a width of 30 cm and a depth of 45 cm. The excavated root crowns were shaken briefly to remove a large fraction of the soil adhering to the root crown. Most of the remaining soil was then removed by soaking and agitating the root crown in mild detergent at a concentration of 0.5% (containing sodium laureth sulphate, cocamidophorol betaine, cocamide DEA, Styrene acrylate copolymer, chlorhexidine gluconate and sodium chloride). In a third step remaining soil particles were removed from the root crown by vigorous rinsing at low pressure with water. The clean roots were subsequently used to measure the following traits: Numbers of crown (CN) and brace roots (BN) were counted, a protractor was used to measure root angles (CA, BA) in degrees from horizontal: Horizontal roots were classified as 0, vertical roots as 90. On a root crown angles of four individual roots were measured for each crown and brace root. Mean values per root crown were recorded. Branching of crown (CB) and brace roots (BB) was based on counts taken on corresponding roots on a 1 cm segment originating 5 cm below the base. All roots emerging below ground were classified as crown roots while roots emerging aboveground were classified as brace roots Soil coring Soil cores of a diameter of 5 cm and length of 60 cm were taken within the planted row midway between two plants. After excavation, soil cores were stored at 4 C until further processing. Each soil core was then subdivided into segments 10 cm long and roots extracted from each segment. Subsequently extracted root samples were weighed. In the next step, a cumulative root depthdistribution specific to each soil core was calculated by summing up the total root mass reached at a specific soil depth: RLCum Seg i = i k=1 RL Seg k where RLCum Seg i is the cumulative root weight in the ith segment (start counting from the top segment) and RL Seg k the root weight in segment k. These cumulative root weights were then expressed as percent of the total root weight. The depth reached by 95% of the roots (D 95 ) was determined by linear interpolation between the cumulative root weights. 4. Experimental design 4.1. Evaluation of the group of 8 for root angles across 3 harvest stages In 2009 eight genotypes contrasting for root angles were grown in Rocks Springs under optimum and nitrogen deficient conditions. Roots were harvested at three harvest stages: 43 DAP, flowering

4 S. Trachsel et al. / Field Crops Research 140 (2013) (76 DAP) and physiological maturity (112 DAP). Grain yield was measured at the end of the cropping season. Fertilization treatment levels were randomly assigned to main plots, within main plots genotypes were randomly assigned to sub-plots, and within sub-plots harvest stages were randomly assigned to sub-sub-plots. One sub-plot was 5 rows wide and 6 m long. Sub sub-plots were assigned to half of the rows 2 and 4 (sub sub-plots thus fully bordered) and were 3 m long. Shovelomics (Trachsel et al., 2011) was carried out at all three harvest stages, concurrently plant height and leaf greenness were measured. To quantify rooting depth one soil core per sub sub-plot was taken within the planted row in the middle between two plants. Soil cores were taken down to a depth of 60 cm (see more detailed description in Section 2) Evaluation of eight genotypes contrasting for root angles across three environments In 2009 (Rock Springs, PA, USA), 2010 (Hancock, WI, USA) and 2011 (Alma, LP, ZA) eight genotypes contrasting for root angles were grown under optimum and N deficient conditions. Roots were harvested at flowering. Concurrently plant height and leaf greenness were measured and soil cores were taken (see detailed description in Section 2). In 2009 in Rock Springs and 2010 in Hancock yield was measured at the end of the cropping season. In each location the experiment was laid out as a split-plot design replicated three times. In all locations, four sections with optimal and nitrogen deficient fertilization blocks that were adjacent to each other were assigned to sections of the experimental fields. Due to fertilization errors only data acquired in three replications are used in the analysis. Within fertilization treatment blocks, genotypes were randomly assigned to single row plots 6 m long Evaluation of 100 genotypes contrasting for root angles across 2 environments In 2010 (Hancock, WI, USA) and 2011 (Alma, LP, ZA) 100 genotypes contrasting for root angles were grown under optimum and nitrogen deficient conditions. Roots were harvested at flowering and concurrently plant height and leaf greenness (SPAD) were measured. To estimate relative plant performance under fertility levels (many genotypes did not mature in Alma before frost) the product of SPAD reading x plant height was calculated. The index of SPAD plant height was chosen because this combination showed greater correlations with grain yield than any other trait individually (r 2 = 0.6; n = 600). Correlation between SPAD x plant height and grain yield was established in 2010 in Hancock, WI. In 2010 in Hancock grain yield was measured at the end of the cropping season. In both locations the experiment was laid out as a split-plot design replicated three times. In both locations 3 blocks with optimal and N deficient fertilization adjacent to each other were assigned to sections of the experimental fields. Within fertilization treatment blocks genotypes were randomly assigned to single row plots 6 m long. The mixed linear model effect used for the analysis of traits measured was: Y ijklm =+ i + ˇj + m + ( ˇ) ij + ( im ) + (ˇ jm ) + r k + b jkl + ε ijklm where Y ijklm is the trait value of the ith genotype (i = 2, steep or shallow) within the jth treatment (j = 2, optimum fertilization, N deficient), within the mth harvest stage (m = 43, 76, 112 DAP), within the kth replication (k = 1,...,4) and the lth block (l = 1,...,4); the main effect of the genotype, ˇ the main effect of the treatment, the main effect of the harvest stage, ˇ the genotype-by-treatment interaction, the genotype-by-harvest stage interaction, ˇ the treatment-by-harvest stage interaction, r the effect of the replication and b the block effect and ε ijk the sub sub-plot error (residual). i, ˇj m + ( ˇ) ij, ( im ), (ˇ jm ) were considered to be fixed effects, while replication and block were considered to be random effects. For the analysis of the experiment with eight genotypes (4 steep and 4 shallow) grown across 3 environments the effect of the harvest stage and associated interactions were dropped from the analysis. Instead a random factor for the environment with n levels (n = 1, 2, 3) was added. For the analysis of 100 entries (20 very shallow, 20 shallow, 20 medium, 20 steep, 20 very steep) grown in Hancock and Alma n was reduced to 2 and i increased to 5. Data were fitted using linear mixed effect model in package lme4( ) in R (Bates and Sarkar, 2007). Initially three-way interactions between factors were included in the model, but were dropped as none were significant. For all experiments data for BW was log10 transformed for the analysis while the root fresh weight measured in the harvest stage experiment was square-root transformed. Comparisons among genotypes within harvest stages were carried out using the Tukey Kramer multiple comparison test. Correlation analysis was carried out using the cor.test( ) function in R. In order to identify genotypes best adapted across environments, an additive main multiplicative interactive effects analysis (AMMI) was carried out. For this analysis four environments were defined according to the fertility treatment-by-location combinations used in this experiment. The mixed linear model effect used for the analysis of traits measured was: Y ijk = + i + ˇj + ( ˇ) ij + r k + ε ijk where Y ijk is the product of SPAD and plant height at flowering approximating overall biomass of the ith genotype (i = 100, 1,...,100) within the jth environment (j = 1,...,4) the kth replication (l = 1,...,3); the main effect of the genotype, ˇ the main effect of the environment, ˇ the genotype-by-environment interaction, r the effect of the replication within environment and ε ijk the error. Genotype, environment and genotype-by-environment interaction were considered to be fixed effects, while replication within environment was considered to be a random effect. The genotypeby-environment interaction explained 8.8% of the phenotypic variance observed. The R package agricolae (de Mendiburu et al., 2010) was then used to calculate a PCA on the interaction effects of individual genotypes. Scores for both genotypes and environments from the two-component interaction component model were computed. The two components (PC1 and PC2) explaining the largest portion of the genotype-by-environment interaction were plotted for genotype-by-environment combinations (Appendix Figure). The values plotted for each genotype-by-environment group are deviations from main effect predictions of genotype and environment effects. Genotypes with a proportional distance (in relation to the genotype with the longest distance) from 0.2 to 0.2 from the origin were considered to be best adapted across environments, while genotypes with a distance < 0.5 and >+0.5 were considered to be poorly adapted across environments. The package agricolae( ) in R was used for all computations and graphing (de Mendiburu, 2010). 5. Results 5.1. Rooting depth and root angles under HN and LN at three different harvest stages at one location in the group of eight The analysis of variance for eight genotypes contrasting for root architecture grown under optimum and nitrogen deficient fertilizations levels harvested at three different stages showed that all traits but CA and BW were significantly affected by the time of harvest (Table 2). The nitrogen fertilization treatment had a significant effect on plant height, the number and angles of brace

5 22 S. Trachsel et al. / Field Crops Research 140 (2013) Table 2 ANOVA table and (in the lower part of the table) means and standard errors (SE) for trait values measured at different harvest stages (H1, H2, H3) in genotypes with shallow root angles, effects and standard errors of nitrogen deficiency (low N), steep genotypes (steep) and differences in interactions (slope) for the various factor combinations (H2:steep, H3:steep, H2:low N, H3:low N, Low N:steep) on root and shoot traits. Traits displayed are: Brace (BA) and crown (CA) root angles, rooting depth (D 95), grain yield (GY), the number of whorls occupied by brace roots (BW), the number of brace roots (BN), the number of brace root laterals (BB), the number of crown roots (CN), the number of crown root laterals (CB), plant height (PHT), SPAD and root fresh weight (RFW). BA ( ) CA ( ) D 95 (cm) BW (#) BN (#) BB (# cm 1 ) Harvest *** ns *** ns *** *** Treatment * ns * ns ** ns Genotype *** *** ** ns ns ns Harvest:Genotype # ns ns ns ns ns Treatment:Genotype # ns ns ns ns ns Harvest:Treatment ns ns ns ns ns ns Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean of shallow genotype across harvest stages Mean of H a a a a Mean of H b b b b Mean of H b b c b Effect of low N treatment (low N) # Effect of steep rooted genotypes (steep) * *** Effect of H2:steep interaction * * Effect of H3:steep interaction * Effect of low N:steep interaction ** Effect of H2:low N interaction *** Effect of H3:low N interaction * * * CN (#) CB (# cm 1 ) PHT (cm) SPAD RFW (g) Harvest *** *** *** *** *** Treatment ns ns * *** ns Genotype * *** *** *** ns Harvest:Genotype * * * *** ns Treatment:Genotype ns ns ns *** ns Harvest:Treatment ns ns ns ns * Mean SE Mean SE Mean SE Mean SE Mean SE Mean of shallow genotype across harvest stages Mean of H a a a Mean of H b b b Mean of H b b a Effect of low N treatment (low N) Effect of steep rooted genotypes (steep) ** Effect of H2:steep interaction * Effect of H3:steep interaction * Effect of low N:steep interaction Effect of H2:low N interaction Effect of H3:low N interaction * Different letters indicate significant differences between treatment levels level. # Significance at p-levels of 0.1. * Significance at p-levels of ** Significance at p-levels of *** Significance at p-levels of roots and rooting depth (D 95 ). Genotype significantly affected plant height, angles and branching of crown, the number of crown roots as well as the rooting depth. Significant harvest stage-by-genotype interactions were found for plant height and the number of crown roots. Across all genotypes and treatments angles of crown and brace roots as well as the number and branching density of crown roots did not change across growth stages, indicating that the crown root system was fully developed at the time of the first harvest at 43 days after planting (DAP; Table 2). Angles of brace (BA) and crown (CA; ns) roots of shallow genotypes responded in a more plastic way to nitrogen deficiency than did steep genotypes, as indicated by the negative interaction observed between steep genotypes and the low nitrogen treatment (Table 2, Fig. 1). Increases in crown and brace root angles under nitrogen deficiency became more obvious at flowering (for brace roots) and physiological maturity, as indicated by the positive significant interaction terms between harvest stage 3 and the nitrogen deficiency treatment for both brace (BA) and crown (CA) root angles (Fig. 1, Table 2). At flowering, steep genotypes were taller (13%) under optimum conditions than shallow genotypes. Shallow genotypes had brace root angles of 47 and crown root angles in the range from 52 to 62 and on average formed 7.2 lateral roots cm 1 of crown root. Genotypes that had been selected as steep consistently had steeper crown (13 ) and brace root angles (16 ) and less crown root branching across all harvest stages compared to shallow genotypes (Fig. 1). Under nitrogen deficiency brace root angles increased by 11 on average (Table 2). The increase in brace root angles was more pronounced for shallow genotypes under nitrogen deficiency resulting in similar brace root angles for both steep and shallow genotypes as classified initially. SPAD, plant height, the number of brace roots as well as rooting depth increased between 43 DAP and flowering, but did not show any major changes between flowering and physiological maturity. Brace root

6 S. Trachsel et al. / Field Crops Research 140 (2013) allowing similar rooting depths as genotypes initially classified steep (Fig. 2) Correlation of rooting depth (D 95 ) with crown and brace root angles Root angles of crown roots were significantly associated with rooting depth calculated as D 95 (Fig. 3). Correlation coefficients were stronger under well fertilized conditions, while slopes were steeper under nitrogen deficiency. Under well fertilized conditions brace roots originating from aboveground showed angles similar to crown roots originating from belowground as indicated by strong positive correlations (Table 3; r = 0.95). Crown (r = 0.85) and brace root (r = 0.79) angles had moderate to strongly positive effects on rooting depth (D 95 ). Branching of brace roots had a negative relationship with D 95 (r = 0.68) consistent with the negative interrelationship between branching and angles (e.g. BB vs BA; r = 0.90, BB vs CA; r = 0.95). Under non-stressed conditions, the number of brace roots was negatively associated with grain yield (r = 0.69). Under nitrogen deficiency branching of brace roots (r = 0.73) had a moderately positive association with grain yield, highlighting the importance of the formation of lateral roots for nitrogen acquisition. Crown root angles were positively associated with the number of crown roots (r = 0.76), while a negative relationship with crown root branching (r = 0.79) was observed as under optimum conditions Group of 8 genotypes contrasting for root angles across environments Fig. 1. Rooting depth (A: D 95), brace (B) and crown root (C) angles as affected by optimum (circle) and nitrogen deficient (triangle) fertilization at 43 days after planting, flowering and physiological maturity for four steep (open symbols) and four shallow (full symbols) angled genotypes. Different letters represent significant differences within a harvest stage. branching increased between 43 DAP and flowering and showed reductions between flowering and physiological maturity (Table 2). Under optimum nitrogen fertilization shallow genotypes had more roots in shallow soil while steep genotypes had greater root biomass in deep soil (Fig. 2). Under nitrogen deficiency shallow genotypes still had greater root mass in the topsoil but had similar root mass below 40 cm as steep genotypes (Fig. 2). Root depth distribution largely remained the same between flowering and physiological maturity (Fig. 2). Rooting depth calculated as D 95 largely confirmed these findings (Fig. 1): at the first harvest stage shallow genotypes had significantly greater D 95 under nitrogen deficiency compared to optimum conditions. At flowering and physiological maturity D 95 for shallow and steep genotypes under nitrogen deficiency as well as for steep genotypes under well fertilized conditions was greater than D 95 for shallow genotypes under optimal fertilization. It is noteworthy that under nitrogen deficiency brace root angles (Fig. 1b) and to some extent crown root angles (Fig. 1c) of shallow genotypes had increased to a degree A set of eight genotypes was evaluated in three different environments at flowering under optimum and nitrogen deficient conditions (Table 4). Significant treatment effects were observed for plant height, SPAD, crown and brace root angles, number and diameter of brace roots, CB and grain yield. Significant effects for the factor genotype were observed for BA, D 95, BN and PHT. A genotypeby-treatment interaction was identified for BA (p = ). Under nitrogen deficiency both crown and brace root angles became steeper, crown roots by 10 and brace roots by 12 (Fig. 4; Table 4). Brace root angles of shallow genotypes increased more strongly than steep genotypes, which is in accordance with results measured across different harvest stages. Steep genotypes constitutively formed steeper brace root angles under optimum nitrogen fertilization, while the difference between shallow and steep genotypes was much smaller under nitrogen deficiency (7 ), as shallow genotypes increased rooting angles more strongly than did steep genotypes. Crown roots for steep genotypes tended to be steeper than for shallow genotypes but the difference was not significant. Rooting depth (D 95 ) was greater for steep genotypes compared to shallow genotypes under optimum fertilization. Rather surprisingly, D 95 (Fig. 4c) of shallow genotypes did not increase as strongly as was expected based on increases in brace root angles. This is possibly related to the fact that D 95 is more strongly related to crown (rather than brace) root angles, and that crown root angles did not show significant increases under nitrogen deficiency compared to optimum conditions. Moreover overall root growth might have been reduced by N deficiency as indicated by smaller D 95 values (significant at p = 0.1) observed for steep genotypes which did not have steeper brace root angles. Positive correlations between the product of SPAD and PHT (as a proxy for general plant vigor and performance) with D 95 under nitrogen deficient conditions (r 2 = 0.3 under LN compared to r 2 = under HN) indicates the influence of general plant vigor on rooting depth under LN, potentially explaining these discrepancies. Nitrogen deficiency reduced plant height (13%), SPAD (16%), the number of brace

7 24 S. Trachsel et al. / Field Crops Research 140 (2013) Fig. 2. Root depth distribution as affected by root architecture and nitrogen fertilization level for four steep (open symbols) and four shallow (full symbols) angled genotypes at 43 days after planting (Harvest 1), flowering (Harvest 2) and physiological maturity (Harvest 3). Mean standard error of the difference (MSED) are shown for each harvest stage-by treatment combination. Table 3 Spearman-rank correlation coefficients among traits measured for eight inbred lines contrasting for root architecture. Traits displayed are the number of whorls occupied by brace roots (BW), number of brace roots (BN), angles of brace (BA) and crown (CA) roots, brace (BB) and crown (CB) root branching, grain yield and rooting depth (D 95) under well fertilized and nitrogen deficient conditions. BW BN BA BB CN CA CB Yield High N BN 0.29 BA BB * CN # 0.82 * CA ** 0.95 ** 0.69 # CB Yield # D * 0.68 # * Low N BN 0.82 * BA BB CN CA * CB * Yield # D * * 0.32 # Significance at p-levels of 0.1. * Significance at p-levels of ** Significance at p-levels of *** Significance at p-levels of

8 S. Trachsel et al. / Field Crops Research 140 (2013) Table 4 ANOVA table and (in the lower part of the table) mean values and standard errors (SE) for shallow genotypes grown under well fertilized (HN) and nitrogen deficient conditions (LN), and effects of steep genotypes (steep) and the interaction (slope) between steep genotypes and low nitrogen (LN:steep) on root and shoot traits displayed for 4 shallow and 4 steep genotypes measured across three environments. Traits displayed are: Brace (BA) and crown (CA) root angles, rooting depth (D 95), grain yield (GY), the number of whorls occupied by brace roots (BW), the number of brace roots (BN), the number of brace root laterals (BB), the number of crown roots (CN), the number of crown root laterals (CB), plant height (PHT) and SPAD. BA ( ) CA ( ) D95 (cm) GY (g plant 1 ) BW (#) BN (#) Genotype (G) *** ns *** ns ns *** Treatment (T) ** ** ns * *** ** G:T * ns ns ns ns ns Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean of shallow genotypes under HN and LN HN a a a a a a LN b b b b b b Effect of steep genotypes (steep) *** *** Effect of LN:steep interaction * # BB (# cm 1 ) CN (#) CB (# cm 1 ) PHT (cm) SPAD Genotype (G) ns ns ns *** ns Treatment (T) ns ns *** *** *** G:T ns ns ns ns ns Mean SE Mean SE Mean SE Mean SE Mean SE Mean of shallow genotypes under HN and LN HN a a a LN b b b Effect of steep genotypes (steep) # Effect of LN:steep interaction Different letters indicate significant differences between treatment levels. # Significances at p-levels of 0.1. * Significances at p-levels of ** Significances at p-levels of *** Significances at p-levels of roots (22%) and grain yield (28%) compared to the well fertilized treatment (Table 4) Evaluation of 100 RILs in 2 environments The analysis of variance for the 100 RILS grouped into quintiles according to their root angle phenotypes (very shallow, shallow, medium, steep, very steep) showed that all traits but grain yield, BW and BN were significantly affected by phenotype and treatment (Table 5). Significant phenotype-by-treatment interactions were observed for the angles of brace and crown roots. The evaluation of the 100 RILs grouped into different quintiles according to brace root angles confirmed differences for brace and crown root angles among the different quintiles across two environments. Among quintiles differences were observed for brace and crown root angles under optimum nitrogen fertilization (very steep > steep > medium > shallow > very shallow; Fig. 5) as well as for the number of brace roots where the shallow phenotypes had slightly fewer roots compared to the other phenotypic classes. Compared to growth under optimum nitrogen fertilization, genotypes with very shallow, shallow and medium root angles showed strong increases in root angles when grown under nitrogen deficiency, resulting in much smaller differences among genotypes. Nitrogen deficiency reduced plant height (15%), the number of brace roots (27%), branching of crown roots (11%) and brace roots (25%) and SPAD values (27%). Brace root angles (2.9 ) and crown root angles (3.7 ) became steeper under nitrogen deficiency compared to optimum conditions (Table 5). Overall 34 genotypes showed increases in brace root angles of more than 5 (29 for CA; data not shown) under nitrogen deficiency, while only 13 (10) genotypes showed more shallow root angles. 53 genotypes (61) did not show accentuated plastic effects as they had already been relatively steep under well-fertilized conditions Effect of nitrogen deficiency on root architecture and resulting grain yield Fig. 3. Association between brace and crown root angles and root depth distribution under well fertilized and nitrogen deficient conditions. Genotype and treatment both significantly affected yield in the group of 8 genotypes (Table 4; Fig. 4) as well as in the experiment consisting of 100 RILs (Table 5). Yield was reduced by 31% (100 RILs and 20% subset of 8 genotypes) under nitrogen deficient conditions compared to well-fertilized conditions. Differences between steep and shallow genotypes were neither observed under conditions of optimum nitrogen availability nor under nitrogen deficient conditions. As a result of increases in brace and crown root angles for genotypes classified as shallow, rooting depths between shallow

9 26 S. Trachsel et al. / Field Crops Research 140 (2013) Table 5 ANOVA and (in the lower part of the table) mean values for 20 genotypes with intermediate root angles grown under well fertilized (HN) and nitrogen deficient (LN) conditions, as well as effects on root and shoot traits evaluated across two environments by 20 very shallow, 20 shallow, 20 steep and 20 very steep genotypes with more shallow (very shallow, shallow) and steeper (steep, very steep) root angles as well as interactions (slopes) between factor combinations (LN:very shallow, LN:shallow, LN:steep, LN:very steep). Traits displayed are: Brace (BA) and crown (CA) root angles, grain yield (GY), the number of whorls occupied by brace roots (BW), the number of brace roots (BN), the number of brace root laterals (BB), the number of crown roots (CN), the number of crown root laterals (CB), plant height (PHT) and SPAD. BA ( ) CA ( ) GY (g plant 1 ) BW (#) BN (#) Genotype (G) *** *** ns ns *** Treatment (T) *** * *** *** *** G:T *** ** ns ns ns Mean SE Mean SE Mean SE Mean SE Mean SE Mean of genotypes with intermediate root angles under HN and LN HN a a a a a LN b b b b b Effect of very shallow phenotypes e f e Effect of shallow phenotypes f e e Effect of steep phenotypes g g f Effect of very steep phenotypes h h f Effect of LN:very shallow interaction * # Effect of LN:shallow interaction # # Effect of LN:steep interaction ns * Effect of LN:very steep interaction # ** BB (# cm 1 ) CN (#) CB (# cm 1 ) PHT (cm) SPAD Genotype (G) ns *** *** *** *** Treatment (T) *** * ** *** *** G:T ns ns ns ns ns Mean SE Mean SE Mean SE Mean SE Mean SE Mean of genotypes with intermediate root angles under HN and LN HN a a a a LN b b b b Effect of very shallow phenotypes e e Effect of shallow phenotypes e e e Effect of steep phenotypes f f f Effect of very steep phenotypes f f Effect of LN:very shallow interaction Effect of LN:shallow interaction # Effect of LN:steep interaction Effect of LN:very steep interaction Different letters indicate significant differences between treatment levels. # Significances at p-levels of 0.1. * Significances at p-levels of ** Significances at p-levels of *** Significances at p-levels of and steep genotypes aligned under nitrogen deficiency, potentially allowing all plants to explore a similar soil volume. Additive Main and Multiplicative Interaction Effects (AMMI)As yield had not been collected in all environments the product of SPAD and plant height at flowering was used as indicator for adaptation to specific environments. In Hancock and Rock Springs, where grain yield had been collected, the product of SPAD and plant height showed a correlation with yield of r 2 = 0.6 (data not shown). The AMMI model indicated that the genotype-byenvironment interaction accounted for 8.8% of the total sum of squares (Table 6). The IPCA1 and IPCA2 accounted for 49.9% and 31.3% of the genotype-by-interaction observed. Twenty-nine genotypes were identified with a proportional distance from the origin in the range from 0.2 to 0.2 and were considered to be the ones best adapted (although this is not an indication of their overall performance) to the environmental conditions applied in the present study. Out of the 29 genotypes adapted to the various environmental conditions 11 were genotypes that had been classified as steep, 11 genotypes were shallow genotypes that had become steep under nitrogen deficient conditions thus plastic genotypes and seven genotypes that had been classified as shallow and did not change root angles under nitrogen deficient conditions. Table 6 ANOVA table showing mean sum of squares, the proportion of variance explained and significance for factors environment (fertilization level-by-location combinations), and genotypes and the genotype by environment interaction affecting the product of SPAD and plant height; and influence of principal components on the genotype-by-environment interaction for 100 genotypes contrasting for root angles grown under different levels of fertilization in Hancock and Alma. ANOVA Total sum of squares Proportion of variance explained (%) ENV 2,089,842, ** REP(ENV) 414,016, *** GEN 502,582, *** ENV:GEN 327,422, *** Residuals 386,829, PCA % Mean.Sq F value PC *** PC *** PC PC