Physiological and genetic analysis for maize root characters and yield in response to low phosphorus stress.

Root system architecture (RSA) plays an important role in the acquisition of mineral nutrients. Nevertheless, RSA has seldom been selected as an important agricultural trait in conventional breeding programs. Little is known about the response of RSA and phosphorus use efficiency (PUE) in regards to phosphorus (P) availability between parental inbred lines and their F1. In this study, 6 maize inbred lines and their 15 F1 generated by a diallel mating system, were used to analyze the genetic associations between RSA and PUE. Heterosis for PUE-related traits were comparatively greater under high P condition and reduced significantly under low P condition. Relative mid-parent heterosis for root traits were expressed more under the low P condition. Low P supply had a significant effect on heterosis, GCA and SCA of RSA- and PUE- related traits. The hybrid C3 (7922 × 8703-2), which had the highest PUE, showed an average yield with a lower P uptake under the both P conditions. Results from this study suggested breeding for a relatively high grain yield with reducing aboveground P demand and grain P concentration should be sufficient to reduce P fertilizer input and improve P efficiency.

Introduction

Phosphorus (P) is an essential and irreplaceable macronutrient for plant growth and yield productivity (Manschadi , Marschner 1995, Shen ). Consequently, crop production relies on a continuous input of P fertilizers. However, phosphorus fertilizers are produced from nonrenewable rock-P sources which will be depleted in 50 to 400 years (Cordell , Dawson and Hilton 2011). In contest to other macronutrients, P is less available in soils due to the easy bound with soil organic matter or chemicals (Al, Fe, etc.). Concentration of available P is often very low (about 2 μM) and seldom exceeds 10 μM even in well-fertile soils.

Maize (Zea mays L.) is a vital and widely cultivated cereal crop necessary for global food security, as its importance is seen in the economic uses ranging from direct human consumption to processing into other food materials, animal feeds, and industrial products (Ishola 2016). However, more than 30% of maize productivity in the world is affected by lack of P in the soil (Vance ). Therefore, in order to avert a future P deficiency crisis, we need to significantly reduce P losses. Developing P efficient maize genotypes that can increase yield with low P supply should be a key breeding target to sustainable use of P resources and agricultural production (Lynch 2007, Manschadi ).

Phosphorus use efficiency (PUE) is commonly defined as the crop yield per unit of P supply from soil and added fertilizer. PUE can be considered as the product of two subcomponents: (i) Phosphorus uptake efficiency (PupE; total P in the above-ground plant tissue at maturity per unit of P supply), and (ii) Phosphorus utilization efficiency (PutE; crop grain yield per unit of P taken up) (Barraclough , Manschadi , Moll , Wang ). Improving PUE in plants can be achieved by employed three strategies: (i) root foraging, (ii) soil P mining to increase P uptake, and (iii) enhancing internal PutE to produce more biomass and yield per unit of P uptake (Lynch 2011, Manschadi , Richardson ). However, PupE could be more important than PutE in wheat and soybean studies under P stress (Manske , Zhao ). Hence, enhancing PupE appears to be an effective method for increasing the PUE and reducing the critical soil-P requirements of crop varieties (Lynch 2007, Manschadi , Ozturk ).

It is widely recognized that root morphological and physiological traits play a crucial role in P acquisition in crop species (de Dorlodot , Den Herder ). Due to the immobility of P in soil, topsoil generally have greater P availability than the subsoil. Therefore, a root system architecture (RSA) that increases topsoil foraging is important for P acquisition. In maize, bean and soybean, shallow root growth angles of axial roots enhance topsoil foraging and thereby phosphorus acquisition (Lynch 2011). Recombinant inbred line (RIL) population in common bean with three basal root whorls accumulated 60% greater dry matter than related RIL population with two whorls in the field research under low P level (Miguel 2011). Increasing lateral root branching density (Postma ), root hair growth (Haling ), and root to shoot ratio (Hermans ) and enhancing lateral root over primary root growth (Wang , Williamson ) leads to higher P uptake. By selecting RSA traits via a conventional breeding program, several soybean varieties that show substantial yield gains in low P soils have been officially certified and released to farmers in southern China (Wang ). In addition, through molecular breeding, the PHOSPHORUS STARVATION-TOLERANCE 1 (OsPSTOL1) gene in rice, was found to enhance root length and surface area at the seedling stage. Lines overexpressing OsPSTOL1 showed an increase of P uptake and approximately a 30% increase in grain yield under low P conditions (Gamuyao ).

When breeding for improved genotypes, breeders are primarily concerned with traits selection for increased grain yield and are aware if a trait from either parent can be reflected in their progeny (Chun ). Bolanos and Edmeades (1993) found that the use of secondary traits that are correlated with yield will help to increase selection efficiency. RSA-related traits in the different developmental stages showed significant correlation with grain yield (Cai ). Tuberosa also reported significant positive correlations between root traits and grain yield under water stress conditions. In the production of maize hybrid cultivars, it is important to evaluate an inbred lines combining ability (Ishola 2016). Chun reported that heterosis of root traits was significant at high and low nitrogen (N) levels and was attributed to both general combining ability (GCA) and specific combining ability (SCA). Betrán also found a significant positive correlation between the hybrid and mid-parent grain yield under different N conditions. However, little was known about the heterosis of RSA- and PUE-related traits in relation to P supply. A better understanding of the heterosis of root characteristics would provide a more targeted approach to improving P efficiency and grain yield within a breeding program.

In this study, 6 maize inbred lines and their 15 F1 created by a diallel mating system were used to: (1) evaluate RSA-and PUE-related traits of inbred lines and their F1 under low P stress; (2) determine the GCA and SCA from the F1 of 6 maize inbred lines; and (3) investigate heterosis of RSA traits and yield under low and high P levels.

Materials and Methods

Plant materials

Using 6 maize inbred lines, 7922, YuanYin1, Chen94-11, 8703-2, Zao27 and 428, a complete diallel mating was designed to produce 15 F1. A complete list of the maize lines examined in this study is shown in Supplemental Fig. 1. In a previous study, three P efficient maize lines (Zao27, 428, and YuanYin1) and three P inefficient lines (7922, Chen94-11 and 8703-2) were selected based on the phenotypic evaluation of a set of 100 inbred lines under two contrasted P conditions by two years (Liu 2003).

Plant growth conditions

All the tested lines were grown in a long-term P fertilized field located in the Changping District, Beijing, China (40°02′ N, 116°10′ E). The experimental design was a split plot with 3 replications. The variable between plots was two levels of P fertilizer, high P (135 P2O5 kg/ha) and low P (no P supplied) since 1986. P fertilizer was applied before sowing. N fertilizer was applied at 270 kg N/ha each year in all the plots, with 40% applied at sowing, and remaining 60% at stem jointing stage. The sub-plots consisted of five 5 m long rows with 50 cm inter-row spacing. Each row contained 13 plants. And the distance between plants within a row was 33 cm. Six inbred lines and their 15 F1 were overseeded with hand planters on May 5th 2002. Plots were thinned at the seedling stage to a final stand of 60,000 plants/ha. Herbicide and irrigation were applied as necessary.

At the 6-leaf stage, a soil volume of 33 cm in width × 50 cm in length × 60 cm in depth surrounding a plant was excavated for 3 plants per sub-plot. All visible roots were picked out by hand. The roots were washed free of soil using a 0.4 mm mesh. The number of seminal roots (SRN) and crown roots (CRN) were counted. The roots were then floated in a transparent plastic tray with water to be scanned by a scanner (EPSON V700, China). Using the software, WinRHIZO (Pro 2004b, Canada), the total root length (TRL) of the scanned images was calculated. Primary root within 15 cm was scanned by scanner. The total lateral root length (TLRL) was calculated based on the scanned images using the same software. Lateral root number (LRN) was counted as the number of lateral roots within 5 cm from the first emerged lateral root on the primary root. Shoots and roots were dried at 65°C for moisture removal, and shoot dry weight (SDW), root dry weight (RDW), and root to shoot ratio (RS) were evaluated. Shoot samples were then analyzed for P concentrations according to the molybdate-blue colorimetric method (Murphy and Riley 1962). The P content of shoot was calculated by multiplying the shoot weight by the P concentration (Table 1).

Table 1

Summary of the investigated traits in this study and the measurements

Trait	Abbreviation	Unit	Trait measurement
RSA-related traits in the seedling stage
The number of seminal roots	SRN	number	Counted manually
The number of crown roots	CRN	number	Counted manually
Total root length	TRL	cm	Root was scanned by scanner and TRL was determined by WinRHIZO
Total lateral root length	TLRL	cm	Primary root within 15 cm was scanned by scanner and TLRL was determined by WinRHIZO
Lateral root number	LRN	number	Count number of lateral root within 5 cm from the first emerged lateral root in the primary root
Shoot dry weight	SDW	g/plant	Dried and weighted using a balance (1/1000 g)
Root dry weight	RDW	g/plant	Dried and weighted using a balance (1/1000 g)
Root to shoot ratio	RS	g/g	Root dry weight/shoot dry weight
P acquisition per plant	–	mg/plant	Total amount of shoot P content per plant
PUE-related traits in the mature stage
Grain yield	GY	kg/ha	–
Total P uptake	Pup	kg/ha	Total amount of P content in plants per hectare
P use efficiency	PUE	kg/kg	Grain yield/total P available in soil
Harvest index	HI	kg/kg	Grain yield/(grain yield + shoot biomass)

At maturity, two middle rows with 26 plants were harvested for determination of grain yield. Stover tissues were harvested, and weighed after oven-drying at 65°C. The grain yield and shoot biomass were calculated according to the harvest area. The grain and shoot samples were then analyzed for P concentrations. The P content of the grain and shoot was calculated by multiplying their weights by their P concentration, respectively. Total P content of the plants was then calculated as the sum of both the grain and shoot. Harvest index (HI) and P use efficiency (PUE) were calculated as described in Table 1.

Statistical analysis

Data was analyzed using PROC GLM in SAS (SAS Institute Inc., 1989). The general combining ability (GCA) and specific combining ability (SCA) of each trait were computed from the mean square of the treatments based on Griffing’s (1956) diallel method I model I where the genotype was considered as a fixed effect and replication was considered as a random effect as shown below:

Where μ is the overall mean, gi is the general combining ability effect of the ith parent, gj is the general combining ability of the jth parent. Sij is the specific combining ability effect (interaction effect) of the ith and jth parents, rij is the reciprocal effect, replication, bk is the effect of kth replications and eijk is the error.

The relative mid-parent heterosis (MPH%) was calculated as follows: MPH% = (hybrid phenotypic performance – mid-parent phenotypic performance)/mid-parent phenotypic performance × 100%. Pearson correlation coefficient was calculated using SPSS Statistics 17.0 (SPSS, Inc., Chicago, IL, USA).

Results

Evaluation of RSA- and PUE-related traits performance

The investigated RSA-related traits, included: root number (SRN, CRN and LRN), root length (TRL and TLRL), shoot (SDW) and root (RDW) biomass, root to shoot ratio (RS), and P acquisition per plant at the seedling stage. PUE-related traits consisted of grain yield (GY), total P uptake (Pup), P use efficiency (PUE), and harvest index (HI) at the mature stage (Table 1). Significant (p < 0.01) differences were observed in the performance of the F1 and the parental inbred lines for all the traits except for LRN, where significant (p < 0.05) differences were also found (Table 2). In the parental inbred lines, the P treatment effects were significant for RSA-related traits, such as CRN, TRL, TLRL, SDW, RDW and P acquisition per plant. PUE-related traits, such as GY, Pup and HI also yielded significant P treatment effects. For the F1, P treatment had a significant effect on all traits except for SRN.

Table 2

Variance analysis of genotypic differences of RSA- and PUE-related traits of 6 maize inbred lines and their 15 F1 at low and high P levels

Source	df	RSA-related traits in the seedling stage									PUE-related traits in the mature stage
		SRN	CRN	TRL	TLRL	LRN	SDW	RDW	RS	P acquisition per plant	GY	Pup	PUE	HI
Inbred lines
Block	2	0.1336	10.2*	3608	369**	43.9	0.1216	0.0023	0.0013	0.0667	58407	1.3	1613	0.0035
P levels (P)	1	0.0711	35.6**	376434**	6097**	29.2	9.7**	0.1495**	0.0011	192.7**	7025150**	335.9**	3251	0.1736**
Inbred (I)	5	6.2**	48.7**	179008**	2182**	393.4*	1.3**	0.0369**	0.0024*	33.1**	4978150**	73.4**	33614**	0.2182**
I × P	5	0.7584	4.4	182716**	2632**	39.8	1.5**	0.0197**	0.0058**	25.5**	462254*	10.1*	8558**	0.0607**
Error	22	0.6594	2.8	4453	81	40.4	0.1311	0.0014	0.0007	1.04	142557	3.7	846	0.1359
F1
Block	2	0.2653	3.8	54158*	138	8.2	2.0*	0.0146	0.0002	31.4*	250644	0.3	1135	0.0002
P levels (P)	1	0.0284	257.0**	4229898**	11657**	145.4*	274.6**	3.0581**	0.0021*	5214.7**	158308494**	3580.9**	141868**	0.2801**
Crosses (C)	14	2.0952**	21.2**	171475**	2377**	80.8**	1.3**	0.0245**	0.0008*	23.8**	1819854**	40.3**	8136**	0.0303**
C × P	14	0.359	2.6	113839**	1428**	33.8	0.8*	0.0185**	0.0004	16.1*	649361.5	28.8**	2154**	0.0047
Error	58	0.5573	2.4	11920	331	23.1	0.3582	0.0051	0.0004	7.2	424992	4.4	798	0.0028

Values are the mean squares from ANOVA. F1 were derived from 6 inbred lines by diallel mating.

indicate significance at 0.05 and 0.01 probability levels, respectively.

For the RSA-related traits at the seedling stage, low P greatly reduced the SDW in all the parental inbred lines and their F1. RDW and P acquisition per plant showed a similar response to low P related to SDW (Fig. 1, Supplemental Table 1). Generally, TRL was decreased under low P condition. However, for some inbred lines and their F1, the TRL did not change, or even increased in some cases, e.g. parental line P3, and cross line C2. For the trait TLRL, except for parental inbred line, P1, which showed a slight increase in low P stress, all other parental lines showed a decrease. The TLRL increased for 7 F1 in the low P level. For root number traits, SRN was not affected by P treatment, and there was no difference between high and low P treatment. The mean CRN observed was 3, which was reduced under the low P stress in the parental inbred lines and the F1. The LRN trait, under the low P level, increased in some inbred lines and the F1, e.g. P1, P3 and P4 in the inbred lines, and C2, C6, C7, C10 and C11 for the F1 (Fig. 1, Supplemental Table 1).

Fig. 1

Influence of different P supply on RSA-related traits in the seedling stage. A, C, E, G, I, K, M, O and Q representing parental inbred lines; B, D, F, H, J, L, N, P and R representing F1. Abbreviations for root traits are as follows: SRN, seminal root number; CRN, crown root number; TRL, total root length; TLRL, total lateral root length; LRN: lateral root number; SDW, shoot dry weight; RDW, root dry weight; RS: root to shoot ratio. HP: high P condition; LP: low P condition.

For the PUE-related traits at the mature stage, low P greatly reduced the GY of all the parental inbred lines and their F1 (Fig. 2, Supplemental Table 2). The mean GY observed was 1629 kg/ha for parental inbred lines and 3940 kg/ha for the F1 in low P level, and 3433 kg/ha for parental inbred lines and 6592 kg/ha for the F1 in high P level, respectively. Pup and HI displayed the same response to low P as that of GY. For the trait PUE, parental inbred lines P2, P5 and P6 showed a decrease at the low P level, all F1 (C1–C15) showed an increase at the low P level (Fig. 2, Supplemental Table 2).

Fig. 2

Influence of different P supply on PUE-related traits in the mature stage. A, C, E and G representing parental inbred lines; B, D, F and H representing F1. Abbreviations for PUE-related traits are as follows: GY: grain yield; Pup: total P uptake; PUE: P use efficiency; HI: harvest index. HP: high P condition; LP: low P condition.

Heterosis of RSA- and PUE-related traits under low P and high P levels

Mid-parent heterosis explains the performance of a hybrid relative to the mid-performance of the parental lines. In this study, the effect of mid-parent heterosis was significant for all the investigated traits except for CRN (Table 3). In addition, large variations were observed among the crosses between low and high P treatments. The mid-parent heterosis estimates ranged from approximately –20.3% to 446% under low P stress, and –7.73% to 2308% under high P level, respectively. Compared with high P condition, the effect of mid-parent heterosis for the RSA-related traits except for RS were more apparent at the low P level. However, for the PUE-related traits, high P level showed a greater effect of mid-parent heterosis compared with low P stress (Table 3). Under both P conditions, significant correlations between parental lines and their hybrid progenies were found for PUE. The root traits CRN, and the SDW also showed significant correlations in the low P stress (Table 4).

Table 3

Relative mid-parent heterosis for RSA-related traits and PUE-related traits of 15 maize F1 under low and high P conditions

Traits	Relative mid-parent heterosis (%)								Significance between LP and HP
	LP				HP
	Mean	Max	Min	LSD0.05	Mean	Max	Min	LSD0.05
RSA-related traits in the seedling stage
SRN	80.2	158	12.1	104	−7.73	17.3	−36.4	27.2	*
CRN	13.8	32.3	−17.1	33.9	9.26	44.5	−18.1	28	NS
TRL	72.8	134	−50.5	55.5	5.34	81.5	−50.5	31.1	*
TLRL	104	192	40.2	65	9.33	96.8	−49.9	45.5	*
LRN	24.3	50.6	−5.4	30.4	−0.27	26.9	−24.8	19.2	*
SDW	174	454	75.2	98.7	−1.89	19.8	−36.5	30.9	*
RDW	112	230	56	93.9	5.97	61.1	−22.1	34.5	*
RS	−20.3	2.79	−41.1	30.9	9.49	33.5	−13.1	30.9	*
P acquisition per plant	167	448	58.3	113	−1.22	41.3	−35.5	39.4	*
PUE-related traits in the mature stage
GY	446	975	115	346	2308	21318	136	4599	*
Pup	143	237	55.8	84.4	174	386	60.3	88.5	*
PUE	108	214	8.99	46.7	726	7934	46.8	2211	*
HI	77.8	211	11.3	39.3	337	3395	30.1	903	*

indicates significance at 0.05 probability levels; NS: not significant

Table 4

Apparent correlation of biomass and root morphological traits between the mid-parent and hybrid performance

Traits	LP	HP
RSA-related traits in the seedling stage
SRN	0.288	−0.301
CRN	0.692**	−0.393
TRL	−0.044	0.014
TLRL	0.26	−0.186
LRN	0.272	−0.09
SDW	−0.516*	−0.079
RDW	0.219	0.134
RS	−0.15	−0.067
P acquisition per plant	0.049	−0.264
PUE-related traits in the mature stage
GY	0.161	0.151
Pup	0.567*	0.647**
PUE	−0.01	0.235
HI	0.437	0.253

indicate significance at 0.05 and 0.01 probability levels, respectively.

Combining ability of analysis of RSA- and PUE-related traits across P treatments

Variance analysis indicated that the parental inbred lines and their F1 were significantly different in both P treatments (Table 2), and so the combining ability of these traits was analyzed further. Analysis of variance for combining ability revealed GCA and SCA had significant contributions (p < 0.05, 0.01) for most traits under low and high P levels (Table 5). For the RSA-related traits, the variance of GCA was significant for CRN, TRL, TLRL, and P acquisition per plant in the low P level, and the variance of GCA for all traits except for SDW was significant in the high P level. The variance of SCA was significant for TLRL in low P level, and significant for all traits except for CRN in high P level. For the PUE-related traits, the variance of GCA was significant for all traits except for GY in the low P level. The variance of SCA showed a similar pattern except that SCA was not significant for GY in the high P level.

Table 5

Variance analysis of general combining ability (GCA) and specific combining ability (SCA) of RSA- and PUE-related traits of 15 maize F1 under low and high P levels

	RSA-related traits										PUE-related traits
	df	SRN	CRN	TRL	TLRL	LRN	SDW	RDW	RS	P acquisition per plant	GY	Pup	PUE	HI
LP
GCA	6	0.57	10.6**	8630**	407**	3.02	0.06	2.64E-04	1.85E-04	0.47*	163366	3.29**	3449**	1.10E-02**
SCA	14	0.3	0.9	3692	168*	3.7	0.06	3.73E-04	1.34E-04	0.28	287367**	1.42**	290	2.11E-03
Error	40	0.24	0.69	2116	57	8.17	0.03	5.94E-04	1.67E-04	0.16	82219	0.37	395	1.22E-03
HP
GCA	6	0.55**	7.16**	144372**	1522**	56.2**	0.31	0.014**	4.49E-04**	13.4*	996352**	39.0**	2528**	1.26E-02**
SCA	14	0.34*	1.59	59247**	734**	22.8*	0.82**	0.014**	9.58E-05	12.7*	348686	10.9**	1725**	2.94E-03**
Error	40	0.13	0.91	5554	153	7.64	0.2	0.003	8.90E-05	4.47	198642	2.44	146	6.99E-04

Values are the mean squares from ANOVA. F1 were derived from 6 inbred lines by diallel mating.

indicate significance at 0.05 and 0.01 probability levels, respectively.

GCA effects for the tested parental inbred lines showed large variations and was influenced by P level (Table 6). Inbred line P1 had positive GCA estimates for TRL, and P acquisition per plant in the low P level, and SRN, TLRL, LRN and P acquisition per plant in the high P level. P2 and P6 had similar GCA estimates at different P treatments as both had consistent positive GCA estimates for root traits indicating that they would be well suited to be combined. P3 and P4 had a negative GCA for most root traits at both P treatments indicating that they would be possibly poorly combined. P5 had negative GCA estimates for SRN, CRN and TLRL with others traits being influenced by positive GCA estimate under high P level. For PUE-related traits, P1 and P5 had a significant positive GCA estimate for GY in high P level indicating the superiority of these two lines in terms of general combining ability. However, P2, P3, P4 and P6 had negative GCA estimates. Pup, PUE and HI had similar GCA estimates, with some parental lines showing a positive GCA and others negative.

Table 6

Effects of general combining ability (GCA) for RSA- and PUE-related traits of 6 maize parents inbred lines under low P and high P levels

Treatment	RSA-related traits								PUE-related traits
	SRN	CRN	TRL	TLRL	LRN	RDW	RS	P acquisition per plant	GY	Pup	PUE	HI
LP
P1	–b	−1.89*	56.7*	−8.94*	–	–	–	0.32	–	−0.74*	22.4*	−0.01
P2	–	0.64	50.8*	17.1*	–	–	–	−0.34	–	0.33	12.1	0.01
P3	–	−1.31*	−48.9*	−8.17*	–	–	–	0.03	–	0.54	−37.8*	−0.06*
P4	–	2.68*	−3.3	−6.36	–	–	–	−0.34	–	−1.40*	39.5*	0.10*
P5	–	−0.51	−50.8*	0.91	–	–	–	−0.15	–	1.09*	−15	−0.01
P6	–	0.39	−4.44	5.5	–	–	–	0.47*	–	0.19	−21.2*	−0.01
LSD0.05 a	–	1.21	66.6	10.9	–	–	–	0.59	–	0.88	28.8	0.05
HP
P1	0.41*	−1.44*	−2.6	0.36	1.1	−0.04	−0.011*	1.3	569*	1.46	4.47	−0.06*
P2	0.16	1.03*	213*	33.8*	−1.67	0.06*	0.009*	−0.91	−345	−2.67*	12.1*	0.06*
P3	0.26	−1.42*	−247*	−18.9*	−5.00*	−0.01	−0.009*	0.56	−116	1.78*	−22.9*	−0.04*
P4	−0.29	1.60*	−176*	−18.2*	−2.43	−0.08*	−0.005	−3.30*	−511*	−5.08*	40.8*	0.08*
P5	−0.59*	−0.67	4.58	−4.19	2.68*	0.01	0	0.81	674*	2.75*	−6.97	−0.03*
P6	0.06	0.90*	208*	7.14	5.31*	0.06*	0.016*	1.54	−271	1.75*	−27.5*	−0.01
LSD0.05a	0.52	1.38	108	17.9	4	0.07	0.014	3.06	645	2.26	17.5	0.04

GCA was significantly different from zero at p = 0.05.

For comparison between the GCA of inbred lines at the same P level.

These traits were not analyzed further.

The performance for SCA estimates in the F1 also varied in magnitude among the RSA- and PUE-related traits, heavily influenced by P levels (Table 7). For TLRL under the low P level, C6 had a significant positive SCA estimate, while C7 and C11 had a significant negative SCA estimates. In high P level, a significant positive SCA effect for SRN was found in C9, with a significant negative SCA effects found in C1 and C14. For TRL, significant positive SCA effects were found in C5 and C13, with significant negative SCA effects found in C2, C4 and C14. The GY performance for SCA estimates showed that all hybrid F1 with positive SCA values had greater yield than the mean except for hybrid F1 C10, C12 and C14 under low P level (Fig. 2, Table 7). For example, the GY of C4 which had a positive SCA effect showed a much higher GY than C15 which had a negative SCA effect under low and high P levels. For Pup, C9 and C13 had significant positive SCA estimates, while C7 and C15 had significant negative SCA estimates under low P level. C9, C11 and C15 had significant (either positive or negative) SCA estimates for Pup, PUE and HI at the high P level.

Table 7

Effects of specific combining ability (SCA) for RSA- and PUE-related traits of 15 maize F1 under low and high P levels. F1 (C1–C15) were derived from 6 inbred lines by diallel mating

Lines	RSA-related traits								PUE-related traits
	LP	HP							LP		HP
	TLRL	SRN	TRL	TLRL	LRN	SDW	RDW	P acquisition per plant	GY	Pup	Pup	PUE	HI
C1	−10.3	−0.92*	−68.1	9.8	−0.88	−1.18*	−0.16*	−4.21*	74.7	0.88	1.93	−5.07	−0.08*
C2	10.1	−0.12	−170*	−16.1	1.23	−0.4	−0.03	−1.94	−65.5	−0. 50	2.28	−41.3*	−0.05*
C3	1.14	0.43	45.7	13.8	1.32	0.92*	0.18*	2.29	−259	−0.27	−0.86	24.5*	0.03
C4	4.74	0.29	−250*	−24.6*	0.31	0.06	−0.03	1.46	314	0.32	−0.49	27.3*	0.05*
C5	−5.71	0.32	443*	17.1	−1.98	0.58	0.04	2.39	−63.5	−0.45	−2.86*	−5.39	0.05*
C6	18.6*	0.34	116	−29.1*	4.44*	−0.07	−0.05	0.09	59.8	−0.4	0.84	−13.7	0.04*
C7	−14.4*	−0.08	57.7	8.63	−0.38	−0.13	0	−0.4	−754*	−1.87*	2.23	−50.5*	0
C8	6.84	−0.02	−77.9	25.7*	3.12	0.07	0.06	−0.6	−128	−0.38	−2	12	−0.01
C9	−0.79	0.67*	−28	−15.1	−6.31*	1.29*	0.16*	5.12*	747*	1.76*	−3.00*	57.2*	0.05*
C10	−5.84	−0.11	22.6	−8.96	0.97	−0.06	−0.01	−0.16	123	0.6	−3.62*	36.3*	−0.02
C11	−20.6*	−0.35	70.3	9.75	−8.68*	0.2	0.08	0.46	−98	0.31	−3.55*	35.9*	0.04*
C12	−2.32	0.24	−39	44.4*	2.03	0.34	0.02	1.56	19	−0.02	4.05*	−17.1	−0.02
C13	9.62	0.53	254*	11	−1.46	0.58	−0.03	3.02	733*	1.28*	3.25*	−25.4*	−0.01
C14	9.47	−0.78*	−380*	−24.5*	−0.45	−1.31*	−0.14*	−4.74*	157	0.25	−0.99	15.1	0
C15	−0.64	−0.45	4	−21.9*	6.71*	−0.9	−0.08	−4.34*	−821*	−1.54*	2.77*	−49.9*	−0.07*
LSD0.05a	18.9	0.89	187	31	6.94	1.12	0.13	5.3	719	1.53	3.92	30.3	0.07
LSD0.05b	15.5	0.73	153	25.3	5.66	0.92	0.11	4.33	587	1.25	3.2	24.7	0.05

SCA was significantly different from zero at p = 0.05.

For comparison between the F1 of which one of parents was same.

For comparison between the F1 of which neither of parents was same.

Discussion

Crop production relies on large inputs of P fertilizers. Despite being a finite natural resource, P use is very inefficient. The environmental and financial costs of using P fertilizers are high. Developing crop varieties with enhanced P efficiency will make a key contribution to sustainable use of P resources. Due to the significant role of roots in the acquisition of water and mineral nutrients, plant breeders are paying more attention to roots to increase PUE and crop yield in agriculture. Previous research to enhance PUE and grain yield by genetic improvements of RSA have been reported (Chin , Gamuyao , Gu ). However, RSA has seldom been selected as the important agricultural trait in conventional breeding programs. In this study, 6 maize inbred lines and their 15 F1 created by a diallel mating system were used to analyze the genetic associations between RSA and PUE. These findings can provide a successful case to apply to genetically improving PUE in maize.

Response of RSA-related traits to low P stress

Low P stress significantly affected all RSA-related traits in the seedling stage, except for SRN in the case of both the parental inbred lines and their F1 (Fig. 1, Supplemental Table 1). In this study, all genotypes showed a decrease of SDW and RDW at low P level, and most genotypes had higher root to shoot ratios. A similar study was reported by Liu , in which low P availability increased root to shoot ratio, lateral root length, and acid phosphatase activity in roots and on the root surface. Similar to P, low N stress increased the root to shoot ratio as an adaptive response for a plant to conserve the ability to acquire more N from the soil (Chun ).

Through a simulated three-dimensional development of maize root architecture varying across six phosphorus levels, Postma found that reduced leaf area expansion and increased root to shoot ratio improve the P status of the maize plant under low P stress. Simultaneously, the improved P status increases the relative allocation of carbohydrates to the shoot, reducing the availability of carbohydrates for root growth. The post-embryonic shoot-borne roots play an important role in maize for the later growth and development stages (Hochholdinger and Tuberosa 2009). The result in our work showed the CRN was reduced under low P conditions for all the genotypes. This may reduce root metabolic costs by reducing competition for P within and among plants. Lateral roots have a strong influence on root system architecture (Lynch 1995) and are responsible for the majority of water and nutrient uptake of the maize plant (McCully and Canny 1988). To overcome P-starvation, plants RSA have evolved many adaptive strategies, some of which involve dramatic changes in lateral root growth, such as increases in lateral root number, length and lateral root branch density (Lynch 2011, Postma ). In our experiment, some genotypes exhibit increases in TLRL and LRN when grown under low P conditions, others displayed little or no response (Fig. 1, Supplemental Table 1). Similar results were observed by Chin and Shimizu in rice root system studies.

Yield and PUE performance under low P conditions

Large variation in grain yield commonly exists among the maize genotypes under different P treatments (Fig. 2, Supplemental Table 2). The P stress level chosen for this study resulted in yield loss ranging from 35.6%–99.2% with an average of 57.3% for the parental inbred lines, and 25.7%–49.0% with an average of 40.2% for the F1. Liu and Plenet also reported the similar result that maize grain yield was significantly decreased under low P stress. In this study, among six inbred lines P1 and P3 were extremely sensitive to low P stress. Their reduction of GY under low P was mainly resulted from the significant decreases of kernel size and grain number per kernel. In addition, P stress leads to an increased anthesis-silking interval, which often negatively affect the pollination and kernel development.

Under low P conditions, the highest yielding F1 were C9 and C13. They displayed a relatively higher CRN, TRL, TLRL and PUE in the low P level. The hybrids performed higher GY under stress conditions, showing that it is possible to combine stress tolerance and yield potential in maize hybrids (Betrán ). This implies that RSA-related trait responsiveness to P stress could be a potential criterion for current commercial breeding programs. A major component of this is likely to be breeding for increased yield, which is a significant component of early root establishment in the field. Increasing total root length and lateral root length, whilst maintaining or decreasing P acquisition of shoot, will lead to increased PUE and then increase the yield. Similar results have been reported with evaluating genotypes for PUE as a potential criterion for breeding higher yielding varieties (Romer and Schenk 1998).

Heterosis, GCA and SCA of RSA- and PUE-related traits as affected by low P stress

Heterosis is the major factor responsible for the better adaptability of hybrids to stress conditions as compared to the parental inbred lines (Chun , Ishola 2016). In this study, heterosis effects for RSA- and PUE-related traits were significant between low and high P conditions (Table 3). Relative mid-parent heterosis for root traits were expressed more under the P stress condition compared with the high P condition. It is indicated that the root size of the F1 was improved by utilizing heterosis at low P level (Fig. 1). Chun also observed that low N significantly increased the heterosis of total and lateral root length. For above-ground, heterosis for PUE-relative traits were comparatively greater under high P conditions and reduced significantly under low P stress. This phenomenon indicates root traits are relatively better adapted to low P stress than PUE-related traits. Significant correlations between the mid-parent and their hybrid progenies were only found in CRN, SDW in the low P stress condition. Pup showed significant correlations between the mid-parent and their F1 at both P treatments (Table 4). Similar weak correlations were observed for different agriculture traits between the parental lines and their F1 (Balko and Russell 1980, Lafitte and Edmeades 1995).

The effect of SCA was significant for most root traits and PUE related traits in low P stress condition. With respect to both P conditions observed in this study, C1, C4 and C13 showed stable yield performance. C3 (P1 × P4) and C13 (P4 × P5) presented quite different responses to low P for grain yield and root traits (TRL and TLRL) despite having one common parent (Figs. 2, 3). C13 showed much higher grain yield than C3 in low P condition because the P5 parental line of C13, was insensitive to low P stress. This may be due to its SCA which enables more prevalent genes in the roots into the F1 during hybridization. The results also showed that C3 had the highest PUE with a lower Pup in the low P stress. This indicates that C3 can achieve an average yield with 19% (data not shown) less P fertilizer input and reduce the grain P concentration by 12.8% (data not shown) in low P condition. In grain crops, phytic acid is the major organic storage form of P, which is non-available for humans and monogastric animals (Raboy 2001). The amount of phytic acid-P stored in the seed is equal to 65% of the global annual P fertilizer input (Lott ). Therefore, breeding genotypes with low seed P concentration may offer health and environmental benefits. Liu identified quantitative trait loci (QTLs) associated with grain phytic acid-P concentration which can be improved without affecting inorganic P concentration and grain nitrogen concentration.

Fig. 3

Picture representing the response of grain yield to low P condition in maize genotypes P1, P4, P5, C3 (P1 × P4) and C13 (P4 × P5). HP: high P condition; LP: low P condition.

In conclusion, low P supply had a significant effect on heterosis, GCA and SCA of RSA- and PUE-related traits. Maize hybrids respond to low P stress by reducing aboveground P demand and grain P concentration to enhance the PUE and achieve stable grain yield. Thus, breeding for a relatively high yield hybrids with reducing above-ground P demand and grain P concentration should be sufficient to reduce P fertilizer input and improve P efficiency.