Decline of seedling phosphorus use efficiency in the heterotic pool of flint maize breeding lines since the onset of hybrid breeding

Improved management and breeding increased maize ( Zea mays L.) yields over the last century, but nutritional efficiency was usually not the focus. This study inves-tigates whether old and recently released flint and dent maize seedlings vary in the phosphorus (P) acquisition and utilization. P use efficiency (PUE) and related traits were measured and compared at two P levels in a calcareous soil. PUE and P acquisition efficiency (PAE) from founder flints to elite flints declined over the last decades. This was associated with smaller root systems, reduced ability to exploit external P, decreased rhizosphere pH and shorter root hairs in low P. Comparing flints with doubled haploid landraces (DH_LR), old and elite dents and hybrids revealed that dents started to acquire exogenous P earlier and had improved PUE. Most DH_LRs had similar PUE as elite flints. When evaluating root traits associated with P efficiency, seed P was also critical, and it is important to stack different root traits to optimize PUE, P utilization efficiency (PUtE) and PAE in breeding programmes. The root hair length, the ability to acidify the rhizosphere and the root diameter in flint and dent pools may be utilized to improve P use in maize germplasm.


| INTRODUC TI ON
Of all essential macro-elements for plants, phosphorus (P) is typically least bioavailable in soils, despite high total P amounts within soils. Almost all P in soils is found in fully oxidized form as orthophosphate (P i ), but due to slow soil processes such as adsorption, precipitation and immobilization (often in organic molecules), the free P i concentration even in agricultural soil solution is rarely above 10 µM and sparingly soluble P-forms are converted into insoluble forms with time (Bieleski, 1973). This low P i concentration is more typical for micro-nutrients that are required by orders of magnitude less. If available, P is readily taken up by the roots as (di-)hydrogen phosphate (H 2 PO 4 or HPO 4 2-) and the P i concentrations within plant cells are approximately 5 to 20 mM (Raghothama & Karthikeyan, 2005). Although many soils have large reserves of total P, the inorganic P (P i ) and organic P (P o ) pools of most soils are sparingly soluble and little mobile, especially at acidic and slightly basic pH (Akhtar et al., 2008).
Ensuring stable food and feed production even with dwindling productive agricultural zones with P as a finite, but an essential resource, requires resource-saving production and nutrient efficient crops that cope with these soil properties (Wang et al., 2010;Wissuwa et al., 2009). Several definitions and calculation methods are common for crop P use efficiency (PUE), an indicator of the ability of plants to produce yield or biomass under particular P conditions (Batten, 1992;Lynch, 1998;Moll et al., 1982). PUE is often separated into P acquisition efficiency (PAE) and P utilization efficiency (PUtE). PAE is the ability of plants to capture P from soils; while PUtE is the ability to produce biomass or yield using this acquired P. The P efficiency of the plant can be enhanced by improving the acquisition and/or utilization of P (Manske et al., 2001;Parentoni & Souza Júnior, 2008;Shenoy & Kalagudi, 2005).
Plants use diverse strategies to increase P acquisition and/or improve internal P utilization. The below-ground strategies to increase P acquisition can be divided into morphological traits, chemical modifications of the rhizosphere and microbial interactions. Root morphological strategies include vigorous root branching, increased frequency and length of root hairs, and shallow root systems with preferential root development in nutrient-rich topsoil, to access a large soil area (Haling et al., 2018;Postma et al., 2014;Vance et al., 2003). Additionally, plant roots can acquire P by chemical modification of rhizospheres, such as exudation of H + , sugars, organic anions, amino acids and secondary metabolites (Hinsinger, 2001;Pang et al., 2018;Richardson & Simpson, 2011). Furthermore, P deficiency can induce the establishment of symbioses with arbuscular mycorrhizal fungi to help access P from the labile soil P pool (Irshad et al., 2012;Smith et al., 2011;van der Heijden et al., 2015).
There is substantial evidence for coordination and trade-offs among root functional traits among species and environments (Lynch, 2011;Ma et al., 2018;Reinhart et al., 2012;Wen et al., 2019).
Among the major crops, maize and wheat with fibrous root systems respond to low soil P availability by altering root size and architecture. Meanwhile, legumes tend to modify the chemistry of the rhizosphere (Lyu et al., 2016). Rapeseed, barley and potato have evolved specific traits of root morphology and root exudation that enhance their P uptake capacity under low P conditions (Wang et al., 2015).
The root diameter exerts a strong influence on root trait variation across plant species, growth forms and biomes (Lyu et al., 2016;Ma et al., 2018;Wen et al., 2019). Maize seedlings performed differently in different soil substrates (Liu et al., 2017). Erel et al. (2017) compared maize hybrids grown in neutral and alkaline soil and found that the soil type had a strong effect on PAE. In the neutral soil, the best genotypes were characterized by topsoil exploration, but in the alkaline soil, the 'mining strategy' was optimal.
Recently, there is also increasing evidence that root functional traits changed along with the evolution history (Ma et al., 2018;Reinhart et al., 2012). Due to long-term agricultural practices, especially high fertilization and long-term organic residue accumulation, the total P pool per hectare increased steeply between 1900 and 2010 (Sattari et al., 2012;Zhang et al., 2017). As modern varieties have often been selected under high-nutrient input conditions to obtain a high yield, there is reasonable concern that beneficial traits for uptake under limited P supply are getting lost in elite varieties or even already got lost. Genes or traits related to the efficient nutrient acquisition at low P availability may provide a trade-off when all nutrients are amply available, as plant adaptive traits to nutrient deficiency often result in additional carbon costs (Wang et al., 2010;Wissuwa et al., 2009). By comparing older varieties or landraces with modern lines, loss of root traits contributing to P uptake was already reported in wheat cultivars (Zhu et al., 2001). However, modern cultivars that were selected through conventional breeding approaches for better yield may alternatively unintentionally be better adapted to soil P fluctuations and acquire more P from P-rich zones. For example, modern soya beans have shallow root architecture and higher PAE because they can acquire more P from the soil P-rich surface zone. In contrast, wild soya bean genotypes have deep root architectures and lower PAE, and the root architecture and PAE of semiwild soya bean are intermediate between modern cultivars and wild soya bean (Zhao, 2004). Furthermore, different maize genotypes (bred in P-rich and P-poor environments, respectively) performed differently in homogeneous and heterogeneous P soils and exhibited different root traits, as well P uptake strategies. The genotypes bred in the P-poor environments showed higher phosphatase activity as well higher mycorrhizal colonization within homogenous P soils compared with genotypes bred under P-rich environments .
In maize, efficient P use is critical between planting and the sixleaf stage, when the young root is still too small to sustain shoot nutrient supply. P deficiency during that stage reduced grain yield (Barry & Miller, 1989). Already after the first week of seedling development, most stored P in the form of phytate was completely hydrolysed and was remobilized for P nutrition, while the uptake of exogenous P in maize was delayed and started only at around the fifth day (Nadeem et al., 2011). Maize seedling development was reported to be initially independent of exogenous P supply; within the first three weeks of growth, losses of up to 40% of seed P were reported (Nadeem et al., 2012).
In this study, we initially selected 16 representative maize genotypes from the flint heterotic pool of a public temperate maize breeding programme that spans the breeding progress since the onset of hybrid breeding from the 1950s to 2010s (Hölker et al., 2019). We hypothesized that the increase in biomass in modern elite varieties is associated with increased PUE, PAE and/or PUtE, compared to old genotypes and landraces. We additionally hypothesized that modern elite genotypes lost some beneficial root-related traits, especially under limited P supply, because of selection under ample P supply. Plants were grown in a calcareous soil-sand mixture with two P application levels and were harvested after three weeks.
Moreover, rhizoboxes with observation windows enabled the collection of root physiological traits, such as exudates, and the analysis of root morphology crucial for P acquisition and P utilization. Our analysis was then expanded to doubled haploid landraces and dent lines from different decades for a broad overview of major germplasm pools (flint and dent), as well as three hybrids (derived from crosses of studied flint and dent lines). We finally hypothesized that high PUE under high P supply is associated with trade-offs in root traits that are beneficial for P acquisition under P limitation, with the root length and root hair length exerting a major influence on other root traits across genotypes.

| Plant materials
Thirty four maize genotypes were tested in this study, including

| Plant growth conditions
A long-term stored nutrient-poor subsoil (pH (CaCl 2 ) 7.6, C org <0.3%, CaCO 3 30%) was used, of which information is found in Supplementary for HP, respectively) and bicarbonate extraction (Olsen-P, 6.6 mg· kg −1 for LP and 60.2 mg· kg −1 for HP, respectively). After adding the nutrient solution, the soil-sand substrate was sieved through 5 mm before potting. Soil moisture was set to 70% of the soil water holding capacity and gravimetrically adjusted every 2 days.  Figure S4).

| Harvest and measured traits
Maize P (both shoot and root) and the root traits related to P acquisition were determined. Root traits were classified in root morphological traits and P mobilizing traits (Table S4). Firstly, root exudates were collected from 1 cm subapical root zones of roots by application of strong sorption filter paper (MN156050, Macherey-Nagel, Munich, Germany) and measured as described by Ma et al. (2021).
After sampling the root exudates, 5 pictures were taken from the root hair zone of each rhizotron under a Stemi 2000-C microscope (Carl Zeiss, Germany). From each picture, 5 root hairs' length was measured using Axio Vision 3.1 software (Carl Zeiss, Germany). The average length of 25 root hairs was taken as one replicate.
After taking photographs of root hairs, roots were carefully lifted out of the soil and shaken to remove bulk soil. Afterwards, the whole root was soaked into 50-100 ml 0.2 mM CaCl 2 for a while to collect rhizosphere soil (Pearse et al., 2007). Rhizosphere soil pH was measured by a pH Meter (SevenCompact pH meter S220, Mettler Toledo), and dent rhizosphere pH was additionally measured by soil pH meter (HANNA Instrument) directly placed next to the root in the rhizotron. After collecting rhizosphere soil, the roots were carefully washed with deionized water. Acid phosphatase activity at the root surface was assayed using pnitrophenyl phosphate as the substrate (MP Biomedicals). Two fresh roots were carefully placed into 40 ml of acetate buffer (0.1 mmol/L, pH 5.2) containing 0.5 g·L -1 disodium p-nitrophenyl phosphate. The spectrophotometric assay was carried out at pH 5.2 and 25℃, as recommended by the manufacturer, for one hour.
Activity was quantified by comparing the absorption at 405 nm (Hitachi U-3300, Hitachi Ltd. Corporation) to a standard curve of diluted p-nitrophenyl solution and NaOH.
After measuring the acid phosphatase activity, the root systems were stored in 70% (v/v) ethanol. After that, single roots were submerged and separated in a water film on transparent perspex trays and subsequently scanned using a flat-bed scanner at 400 dpi resolution (Epson Expression 1,000 XL). The root length and average root diameter of the scanned images were measured by using the WinRHIZO root analysis system (Reagent Instruments).
The plant materials (shoots, roots and seeds) were oven-dried at 60℃ until constant weight for determining dry weight. The dried material was ground to a fine powder. 250 mg shoot dry matter was incubated with HNO 3 , H 2 O 2, and distilled water in a microwave (MLS Maxi 44) at a maximum of 210℃ and 1,400 W for 65 min. This solution was adjusted to 20 ml and filtered through 90μm mesh filter paper. P was measured spectrophotometrically via orthophosphate determination after the addition of molybdate-vanadate reagent (Gericke & Kurmies, 1952). Based on the P concentration, we calculated P efficiency-related indexes as detailed below.

| Definitions and abbreviations of phosphorus use efficiency
P use efficiency (PUE) here refers to the ability of plants to produce biomass under particular P conditions (Batten, 1992;Lynch, 1998;Moll et al., 1982). It was calculated by the following formula: PUE (g DW mg -1 P s ) = TDW/P substrate .
where TDW is the total dry weight of the corresponding plants, P substrate is the P amount within the fertilized soil-sand mixture (equal to CAL-P in the soil plus added P per rhizotron), DW is short for dry weight, and P s is short for P substrate .
PAE refers to the ability of plants to capture P from soils, while PUtE is the ability to produce biomass using this acquired P. They were calculated by the following formulas: PAE (mg P mg -1 P s ) = (P content − P seed ) / P substrate .
with P content is the P amount within the plants, P seed is the P amount within the corresponding seeds. For calculating PAE, we assumed that P seed had been used up at harvest.
where [P] is the plant P concentration with the unit of mg/g DW.
We also calculated the PUE w/o seed representing only the external P acquired from the soil assuming that P seed had been completely used up by the growing seedlings at harvest. It was calculated by the following formula: PUE w/o seed (g DW mg -1 P s ) = [(P content − P seed ) × PUtE] / P substrate .

| Statistical analyses
All data were analysed by the JMP Pro 15 statistical package (SAS Institute Inc.). In general, the experiment was composed of the two key factors: P level and maize genotypes (or groups from different breeding periods).
Data validity was explored by plotting the response on the x-axis and the genotype or P level on the y-axis. Next, a mixed model with the P level, maize genotype and their interaction as fixed effects was fit, and the assumption of residual normality as well as homogeneity of within-group variances was checked by using diagnostic plots. Data showing irregular or untypical residual plots (residual by predicted plot, residual by row plot and residual quantile plot) were transformed by the logarithm of the raw data (Kozak & Piepho, 2018;Piepho, 2009

| Seedling biomass and P use efficiencies of groups of flint lines, DH_LRs, dent lines and hybrids
The average shoot dry weight of three-week-old seedlings de-

| Biomass, PUE, PUtE, PAE in the flint heterotic pool and comparison with DH_LRs, dent lines and hybrids
The total seedling dry weight, as well as their root and shoot dry weight, varied, but generally increased with higher P for the 34 genotypes ( Figure 2). As expected, most dent seedlings and hybrids had larger biomass than flint seedlings and most of the DH_LRs. A strong growth response to HP compared with LP was found in one DH_LR line (CG1), in several older flints, and some, but not all modern dent lines. The response to P was highly variable among genotypes that were released in the same period ( Figure 2). The P levels and the genotypes had significant effects on the shoot and total dry weights, but the root dry weight was little affected by P application.
Individual PUEs and associated efficiencies were first studied in the flint pool. Within these 16 lines, a strong decline of PUE and PAE both in LP and HP was observed, while a moderate decline of PUtE was found significant only at HP level (Figure 3a-c). The decline in PUE was paralleled by a decline in root dry weight in more modern lines, and this was more significant in HP than in LP conditions (Figure 3d). This decline in PUE was not as evident when the dent lines were included in F I G U R E 1 Average shoot dry weight, shoot P concentration, leaf numbers and average phosphate efficiencies of different genotype groups. Mean performance ±SE for shoot dry weight (a), shoot P concentrations (b), leaf number at harvest (c), phosphate use efficiency (d), phosphate use efficiency without seed P (e), phosphate utilization efficiency (f) and phosphate acquisition efficiency (g) were shown. Different letters represent significant differences at p < .05 according to the result of Student's t test all pairwise comparisons (logtransformed data were used to ensure data normality) the graph ( Figure S1). PUE and PUtE were always higher in LP than in HP (Figure 3a,b), while PAE was higher in HP ( Figure 3c). The decline of the dry weight of the roots (and total biomass) of seedlings (Figure 3d) was not seen when landraces, dents and hybrids were included in the graph ( Figure S1). The DH_LR lines were overall similar to recent elite flint lines concerning PUE and PAE (Figure 2, Figure S1), despite some variation between individual lines (Figure 2, Figure S1).
We furthermore measured PUE, PUtE and PAE in eight dent lines, as well as in three hybrids that were crosses of the studied flint and dent lines ( Figure S1). The classical dent lines that are often used in genetic studies (Mo17 and B73) were intermediate with respect to PUE and PAE. The PUtE of old dents was similar in to elite flint lines, while modern dent lines and especially hybrids had increased PUtE in LP ( Figure S1). This elevated PUtE caused higher PUEs of dent lines and hybrids in LP. These lines and hybrids also had, on average, increased PUE in HP that was caused by increased PAE ( Figure S1).

| Seed P content and the origin of P in seedlings under LP and HP conditions
The P content per seed substantially varied, between 0.38 and 1.16 mg P seed -1 , but without apparent trends related to breeding history or heterotic group origin (Figure 4a). The seed P concentrations were, however, relatively stable, around 3 mg/g ( Figure S2a).
Due to efficient remobilization during germination, approximately 1 mg of stored P is sufficient to support seedling growth for around two weeks (Nadeem et al., 2011;White & Veneklaas, 2012).
Moreover, the total seedling P content of many genotypes in LP was at seedling harvest in the same range, indicating that several genotypes did not efficiently start to acquire extra P from the soil.
In HP, the majority of P in the seedlings invariably came from soil P (Figure 4b). The P from seeds contributed only ~10% of total P in HP in modern dents and hybrids, while the P fraction derived from the seed P was larger in several flint lines. It was highest in F030 (~45%) and the SF2 DH landrace (~50%) (Figure 4b). Interestingly, under LP, the seedlings of more recently released flint lines (1,278, 5,267, 5,271, F012, F142) and some DH landraces (CG1 and SF2) had even less P than was initially stored in the seeds, suggesting that these genotypes lost P during seedling establishment to the environment (Figure 4c). This loss was as large as 50% of seed P in SF2 (DH_LR) (Figure 4c). Modern dent lines, however, accumulated even in LP a large fraction of their total P (~60%) from exogenous P, while the exogenous P fraction was between 37% and 50% in the hybrids in low P. There was only a weak and nonsignificant trend for an increase in SDW of flints with seed P for F I G U R E 2 Mean performance for total dry weight (TDW) of 34 maize genotypes (N = 5) displayed in groups. Significant differences between P treatments and genotype tested based on Student's t test pairwise analysis using log-transformed data. Asterisk means differences between P treatment. *, 0.01 < p ≤.05; **, 0.001 < p ≤.01; ***, p ≤.001. Different letters mean significant difference among genotypes at p < .05 LP and HP, respectively (Figure 4d). Seed P did not significantly affect seedling growth at LP or HP, when analysed for all genotypes together ( Figure S2b). Likewise, RDW was little affected by seed P (Figure 4e).
There was, however, a robust relationship between the logarithmic TDW of all flints with the P uptake from the soil, both at LP and at HP; this substantial correlation was still observed when DH_LRs, dents and hybrids were included ( Figure S2c). Overall, the seed P content was weakly, but significantly (p < .001) correlated (0.29) with TDW in LP, whereas the correlation was even higher in HP (0.36, Table S7). In HP, the P concentration was not significantly affected by the seed P, but in LP the P concentration even negatively correlated (−0.43) with seed P, whereas the P content and seed P content were positively correlated mostly in HP, due to increasing TDW (Table S7).

| Root traits affected by the breeding progress
We next considered root traits in HP and LP. Root morphological traits included total root length (TRL), specific root length (SRL), average root diameter (RootDiam) and average root hair length (RootHairLen). P mobilizing traits included the amounts of organic acid anions released from root tips (Carboxylates), rhizosphere soil pH (Rhizos-pH) and acid phosphatase activity at the root surface (APase). The proportional origin of seedling P (red = stored seed P; blue = external P; shaded red: stored seed P lost to the soil) of different lines under high P (b) and low P conditions (c). (d) Relation between shoot dry weight (SDW) and seed P in the LP and HP treatments for 16 flints. (e) Relation between root dry weight (RDW) and seed P in the LP and HP treatments for 16 flints There was a significant effect of genotypes on TRL and SRL, but there was no P treatment effect for these two measures (Tables S5). Genotypes, treatments and their interaction significantly influenced other root morphological traits, such as RootDiam, root-to-shoot ratio (Root/Shoot) and RootHairLen.

P mobilizing traits, which include Carboxylates, Rhizos-pH and
APase, were also significantly affected by the genotype, treatment and the interaction (Table S5). Although the internal remobilization of stored P was the crucial P source in LP at the investigated juvenile stage, all root traits except APase were significantly correlated with PAE in LP (Table 1) patterns for founder and elite flints. The first two components together explained more than 58% of the total variation in both HP and LP treatments. Mostly in LP, founder flint genotypes were separated from the other lines ( Figure 6). PUE and RootHairLen were mainly represented by PCA1, but interestingly, RootHairLen contributed oppositely in LP and HP ( Figure 6).

| D ISCUSS I ON
The application of mineral P fertilizers increases plant-available P in the soil and improves crop growth and yield, but its success dependents on many different factors such as type of fertilizer, application form, the timing of fertilization, weather conditions, various soil properties and plant factors (Holford, 1997;Shen et al., 2011).
As P is a finite resource, breeding of major crops with improved P efficiency may have large effects on the sustainability of the agroecosystems (Ludewig et al., 2019). Efforts in breeding tropical maize for P-deficient, acid soils with high loads of toxic aluminium have already identified candidate loci for important root traits (Azevedo et al., 2015), an example of positive breeding impact on root traits.

| The decline of PUE in flint lines along with the breeding history
In this study, there was substantial variation in maize seedling growth of elite varieties under low and high P conditions. The biomass of modern flint seedlings decreased compared with founder flints, which was inconsistent with our expectations, as modern varieties are selected for higher final yields (Figures 1a and 2). In elite flints released from the 1950s to 2010s, PUE, PUtE and PAE decreased dramatically under both low P and high P (Figure 3a-c). Surprisingly, most DH_LRs performed similar to moderately modern elite flints (Figures 1a and 2). Apparently, there is no increase in P utilization  (Table 1 and Figure 5h). In addition, elite flint lines were apparently on average not capable to acquire external P under LP conditions, in contrast to founder flint lines (Figure 4c).
In the conditions tested, most dent elite varieties performed superior than the flints. Those selected in the same public breeding programme released from the 1990s to 2010s generally had higher biomass, PUE, PUtE and PAE. These values were comparable to some flint founder lines (Figures 1d-g and 3a-c), which directly verified the hypothesis that increased biomass of modern elite varieties is associated with increased PUE, PAE and/or PUtE compared with old genotypes and landraces. PAE of all the genotypes was higher in HP than in LP, pointing to the importance of seed P under LP and probably reflecting the selection process with high P fertilization (Figure 1g). It is noteworthy that the founder flint EP1 was superior to most other flint lines in LP and similar to the modern elite dent P415. Moreover, EP1 was the only flint line that exhibited a similar P uptake as the dent lines,
of nearly 70% in LP. P415 was more responding to P fertilization than EP1 and doubled its biomass, while EP1 did not differ in TDW under LP and HP (Figure 2). Both lines were apparently P efficient in LP, but only P415 showed additional P responsiveness.

| Different P efficiency and root traits between flint and dent panels
Many examples have shown that P efficiency mechanisms are different from one genotype to another within a given plant species (Bayuelo-Jiménez et al., 2011;Colombi et al., 2019;Li et al., 2019;Ozturk et al., 2005;Tang et al., 2020;York et al., 2015). There are substantial variability and distinct genetic architectures in dent and flint maize diversity panels (Cartea et al., 1999;Messmer et al., 1992;Rincent et al., 2014), which is consistent with different P efficiency trends of the two panels. Due to different geographic separation and contrasting environmental conditions, phenotypic differences between these two germplasm pools are expected (Brown & Anderson, 1947;Unterseer et al., 2016). Much of the large variation in seedling biomass between these two germplasm pools was probably due to different P remobilization from seeds, and a F I G U R E 5 Root morphological and rhizosphere traits of genotypes and relation to their year of release. Mean performance of (a) rootto-shoot ratio, (b) total root length, (c) specific root length, (d) average root diameter, (e) acid phosphatase activity, (f) Carboxylates, (g) rhizosphere pH and ( delayed switch to exogenous P acquisition in modern flints in LP. Especially in LP, seedlings of several genotypes (especially some in the Elite Flint2 and Elite Flint3 groups) even contained less P than originally stored in the seed (Figure 4c). This suggests that during germination these were not able to utilize the remobilized seed P and failed to import it into the seedling, resulting in a net seed P F I G U R E 6 Principal component analysis (PCA) of root functional traits and phosphate efficiency for 16 genotypes in different classes with high (a) and low (b) soil phosphorus availability. Root trait abbreviations in figures: TRL, total length of the whole root system; SRL, length per dry mass of the whole root system; RootDiam, average diameter of the whole root system; Carboxylates, the total amount of organic acid anions collected by filter papers on the surface of root tips; APase, acid phosphatase activity in the rhizosheath. Rhizos-pH, the pH of the rhizosphere soil solution or pH in the rhizosheath; Root/Shoot, the ratio of root dry weight to shoot dry weight; PUE, phosphorus use efficiency; PAE, phosphorus acquisition efficiency; PUtE, phosphorus utilization efficiency F I G U R E 7 Schematic model summarizing the major observations of traits contributing to PUE, PAE and PUtE of founder flint and most modern flint EL3. Schematic plant appearance for LP and HP is shown loss. Massive losses of macronutrients and micro-nutrients during maize germination have been reported and can be prevented by silicon treatments (Moradtalab et al., 2018). In juvenile maize seedlings, re-uptake of transiently lost seed P from internal sources appears thus a major function of the root system. Previous research showed that the measurable P uptake by maize roots begins with a delay of around five days after germination, but in the variety of that study, the P uptake rate was independent of the initial seed P content and was little affected by external P (Nadeem et al., 2011;White & Veneklaas, 2012). For the different genetic backgrounds analysed here, there was a strong correlation of total dry mass with exogenous P taken up, especially in HP, but not with the seed P content (Table S7). Some dent elite lines (and hybrids) apparently switched earlier or more robustly to acquisition from exogenous P pools, so that they had already acquired substantial external P at the time of harvest (Figure 4). These consequently produced more seedling biomass.
Several root traits differed between flint and dent panels, and between different lines. On average, dent lines had increased RootDiam and released more Carboxylates, while TRL decreased in modern dents compared with old dents in HP ( Figure 5). However, comparisons of the elite dents with these old dent lines must be taken with care, as their genetic origin is different, and also the sample size within these groups differed. In addition to the genetic differences, the temperature setting of the climate chamber may have been more favourable for the dent lines compared with the flints, the latter are cold tolerant and this may explain part of the differences between the flint and dent lines. While the different morphological characteristics between the flint and dent panels can be beneficial in view of hybrid breeding and the associated heterosis effect, it is important to note that beneficial traits in hybrids were more similar to dents, but were overall very little related to their flint or dent parents. The comparison of each hybrid with its parents indicated that all PUE traits were a complex additive combination of their parents that was dominated by the more efficient dent parent and was often, but not always, superior to its parents.

| Trade-offs of root traits related to PUE under LP
Several root traits were found to be associated with PUE in the breeding progress, but this was different in LP and HP. In HP, only TRL was positively correlated with PUE, implying that breeding with ample P resulted in an unintentional selection for TRL.
In LP, TRL, RootDiam, Rhizos-pH, Carboxylates and RootHairLen were positively correlated with PUE, suggesting that genotypes bred under ample P remained some relationships with root traits important for limited P (Table 1). Moreover, RootHairLen and Carboxylates exhibited the highest positive contribution to PUE in LP (Table 1). Within the flint pool, several root traits were associated with the decrease in PUE in the breeding process: Rhizos-pH and RootHairLen in LP, and to a lesser extent RootDiam in HP ( Figure 5). Founder flints had a weaker ability to acidify the rhizosphere and grew longer root hairs in limiting P conditions, which verified the hypothesis that modern elite genotypes lost beneficial root-related traits under limited P supply, because of selection under ample P. Interestingly, RootHairLen was the only root functional trait where Pearson's correlation of PAE significantly shifted from a negative to positive value with different P treatments (Table 1).
A wide variation and co-variation of key root and rhizosphere traits was found. This creates multiple P acquisition strategies that may be similarly efficient in an ecosystem (Brundrett & Tedersoo, 2018;Lambers et al., 2018;Zemunik et al., 2015).
Previous studies showed that in controlled experimental conditions, different plant species showed different P acquisition strategies with different root and rhizosphere trait trade-offs.
Maize with its fibrous root system invests relatively little into rhizosphere traits (Lyu et al., 2016;Wen et al., 2019). In our study, only TRL and SRL (morphological traits) were not significantly affected by P treatment, and all the traits (morphological traits and rhizosphere traits) showed variation among genotypes (Table S5).
Meanwhile, the effect of the breeding process on TRL and SRL was not different among the P treatments (Figure 5b,c). A high co-variation of root functional traits was found in our 34 genotypes. In low P, the correlation between SRL and investigated root traits except TRL was negative ( Table 2). The negative correlations between SRL and the RootDiam were already reported in other studies dealing with P responsive root traits within maize seedlings (Zhu & Lynch, 2004). This is due to its definition (SRL = TRL/ RDW). Postma and Lynch (2011) claimed that with low P supply, maize roots increased cortical aerenchyma, which may lead to a decrease in root biomass and an increase in SRL, without changes in root diameter. RootHairLen was positively correlated with TRL, Carboxylates and Rhizos-pH under LP, but negatively correlated with RootDiam (Table 2), showing that in LP, genotypes with longer root hairs released more organic acids and had more fine roots. Many studies demonstrated the importance of fine roots in P uptake (Lynch, 2011). Therefore, it is assumed that the positive contribution of RootDiam to PUE is mainly based on the high correlation between RootDiam and the biomass. This may also explain the correlation between Root/Shoot and PAE in LP compared with HP.
As we have mentioned, TRL was important in both LP and HP for PUE; RootDiam, Rhizos-pH, Carboxylates and RootHairLen were more important in LP for PUE. In HP, TRL was negatively correlated with RootHairLen (r = −.32), and in LP, TRL was positively correlated with RootHairLen (r = .37) ( Table 2). Moreover, the correlation between PUE and TRL decreased from 0.50 in HP to 0.39 in LP. The correlation between PUE and RootHairLen increased from −0.19 to 0.45 (Table 1). Altogether, this indicates trade-offs among the root traits, which verified our ultimate hypothesis. The variation and covariation still existed when only considering exclusively the flint pool. In HP, there was a large variation of all these eight root functional traits; in LP, there was co-variation among TRL, RootHairLen, Root/Shoot, Rhizos-pH and Carboxylates ( Figure 6). Trade-offs between TRL and RootHairLen are found in the flint pool ( Figure 6), which confirmed that breeding selection can have negative impacts under different environmental conditions.
Meanwhile, maize genotypes released in different breeding periods showed different P acquisition strategies. A schematic summary of the different strategies of old and modern elite flint lines to acquire and utilize P efficiently in LP and HP is given in Figure 7 and for dent lines in Figure S3. Dent lines exhibited the lowest SRL values, which means they produced thicker roots than flint lines.
The high RootDiam within the dent lines is indicative of a different root system with long, thick seminal and crown roots (Figure 5h). On the other hand, the production of dent root biomass required less P (Table S6). Modern dent lines exhibited the highest PUE and had the highest average root diameter. The weak Root/Shoot and RootDiam correlation and that with several other traits that were different in LP and HP, potentially indicates a trade-off for acclimation to contrasting P conditions, as the increased contribution of one trait resulted in the decreased contribution of another trait ( as the presence of cortical root aerenchyma (Postma & Lynch, 2011) or the association with mycorrhiza .

| CON CLUS ION
When evaluating root traits associated with high P efficiency, seed P should also be considered. A decline in PUE in modern maize elite flint material was associated with smaller roots, shorter root hairs and less rhizosphere acidification, while high PUE was identified in elite dents and hybrids from the same breeding programme. Some preselected doubled haploid landraces were similar to modern elite varieties. Although the relationships of root traits and their contribution to optimal PUE, PUtE and PAE in juvenile maize flint and dent pools need to be further verified in different soils, growth conditions and in mature plants, our study may help to focus on improving P use in maize agroecosystems to contribute to more sustainable agriculture.

ACK N OWLED G EM ENTS
This study was supported by the German Research Foundation

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N
We thank H. Ochott for help with nutrient analyses and Prof. Dr. Günter