Two hundred and seventy-four soybean genotypes (Glycine max L.) (Table S1) were collected from Sichuan Province, Yunnan Province, Guizhou Province, Hubei Province, Chongqin City, and Guangxi Zhuang Autonomous Region in the southwest of China (Figure 1). Most of these soybeans are traditional genotype and the gene pool was not domesticated by artificial direct or indirect purpose.

The field experiment was conducted in 2015 at the experimental station of Sichuan Agricultural University in Ya'an, Sichuan Province, China (29°58′N, 102°58′E) with an altitude of 600 m above sea level. Annual mean temperature is 15.4°C with a maximum and minimum temperature of 25.4° and 6.1°C, respectively. The frost-free period is 294 days, annual precipitation is 1500 mm and potential evaporation is 838 mm. Annual sunshine is 1019 h and total solar radiation averages is 3750 MJ m⁻² year⁻¹. The experimental soil is classified as Purple soil (Luvic Xerosols).

The 274 soybean genotypes were planted in a randomized block design with 3 replicates. Each block consisted of 274 plots with the dimension of 1.5 × 4.0 m². Every plot consisted of three plant rows, and the rows spacing is 50 cm. Density of soybean was about 11 plants m⁻². Soybean was sown on early June and harvested on late October in 2015. In the last season, maize had been sown in this field, and the soil properties at the start of this study were pH (water) 6.4, organic matter concentration 30.1 g kg⁻¹, total N 1.80 g kg⁻¹, available N 110 mg kg⁻¹, Olsen-P 17.4 mg kg⁻¹, exchangeable K 91 mg kg⁻¹, and Cation Exchange Capacity 22.0 cmol kg⁻¹ of dry soil in the top 20 cm soil layer. During the growth period, all the plots were well irrigated and weeded manually with no fertilizer input.

Shoot dry matter (DM) of soybean was measured at maturity. 10 plants of soybean were sampled from the middle row of each plot. All samples were dried in an oven at 70°C and then ground for further chemical measurements. The samples were wet-digested with concentrated H₂SO₄ and H₂O₂ (30%) for P determination following the vanadomolybdate method (Page, 1982). Shoot P uptake was calculated as P concentration multiplying with shoot biomass. The harvest index (HI) was calculated as dividing the yield by the shoot biomass. Grain yield of soybean came from harvesting the remaining parts of the plot after shoot sampling.

The linear- plateau model was used to analyze the relationship between yield and HI. The linear-plateau model is defined by Equations (1) and (2) as: (1)y=a+bxif(x)<c (2)y=Ypif(x)≥c where y is HI; a is the intercept parameter; b is the slope parameter; x is the yield (g/plant); c is the critical yield, which is the interception point of the two linear segments; and Yp is the plateau value which is often 90% of the maximum HI. Equation (1) can be interpreted as the region during which the HI responds to yield increasing and Equation (2) to the plateau region where increase of yield does not come from the increase of HI.

Based on yield and P accumulation of soybean genotypes in the field trial (Table S2), E14, E64, E141, E295, E311 and D55, E108, E150, E277-1, E283 representing the P-efficient, P-inefficient genotypes were used to evaluate tolerance of P deficiency in hydroponic experiment, respectively. Seeds were surface-sterilized as described by Vincent (1970), placed on sterile Whatman filter paper, and germinated in sterile water in pot filled with quartz sand until taproots were 3-4 cm long. The seedlings were transferred to a 24 liter container (0.58 × 0.28 × 0.15 m³) filled with half-strength modified Hoagland solution (0.75 mM K₂SO₄, 2 mM Ca (NO₃)₂, 0.65 mM MgSO₄, 0.1 mM KCl, 0.2 mM KH₂PO₄, 0.1 mM Fe(III)NaEDTA, 10 uM H₃BO₃, 1 uM MnSO₄, 0.1 uM CuSO₄, 1 uM ZnSO₄, 0.09 uM (NH₄)₆Mo₇O₂₄), each container with 25 soybean plants. The experiment was designed as a complete randomized with two P treatment levels (0.05 and 0.25 mM P L⁻¹) and six replicates. Plants were grown in a greenhouse from 14th February to 16th March in 2016 with an average temperature of 25/20°C (day/night), relative humidity 75%, average daytime photosynthetically active radiation between 800 and 1000 mmol m⁻² s⁻¹ and photoperiod of 14 h day/10 h night. The solution was well aerated and renewed every 5 days and pH maintained at 5.4–5.5 with daily regulation. The plants were harvested at stem elongation stage (30 days after sowing) for testing the physiological and root morphology index below. Plants show higher root secretion rate at stem elongation stage than later in growth period, when the amount of fixed C allocated to roots and rhizosphere (Gregory and Atwell, 1991).

For root exudation collection, three replicates of each genotype were washed absolutely with running tap water followed by distilled water and then placed separately in 100 ml glass test tube filled with deionized water and covered with black plastic to prevent light degradation of exudations. The exudate was collected for 6 h, and the details before high performance liquid chromatography (HPLC) analysis refers the method of Dong et al. (2004). Organic acids were analyzed by HPLC (Agilent 1100, Agilent, USA) after Libert and Franceschi (1987) with modifications (Yu et al., 2002). A Hypsil (Hypsil, Dalian, China) C₁₈ column (5 uM, 4.6 × 250 mm) was used as the static phase and the mobile phase was a solution containing 0.5% KH₂PO₄ and 0.5 mM tetrabutylammonium hydrogen sulfate (TBA) buffered at pH 2.0 with orthophosphoric acid. The flow rate was 1 ml min⁻¹ and detection wavelength was 220 nm.

After exudate collection, soybean plants immediately divided into shoots and roots. Shoots were dried in an oven at 70°C until constant weight. Roots were placed in clear plastic bags filled with 50% ethanol and stored in a refrigerator. Washed roots were gently arranged with minimum overlap using tweezers in a plexiglass tray full with water, and scanned using an Epson perfection V700 photo, Japan. Images were analyzed using WinRhizo (Version 2007d, Regent Instrument Inc., Canada) to estimate total root length, root surface area, and root volume. Several images of one root were analyzed in more detail to determine the root morphology. Roots were dried in an oven at 70°C until constant weight after imaging. Specific root (cm g⁻¹ DW) length was referring to the method of Pang et al. (2010).

After harvest, fresh roots of three replicates were washed absolutely with running tap water and distilled water, and froze immediately in liquid nitrogen, then stored at −80°C. 0.3 g root tissues were ground with a mortar and pestled with 5 mL of 15 mM 2-morpholinoethanesulfonic acid, monohydrate (MES) buffer (pH 5.5, 0.5 mM CaCl₂ H2O, 1 mM EDTA). The extracts were centrifuged at 4°C for 20 min at 10,000 rpm to obtain the supernatants which were used for the determination of enzyme activity.

P-nitrophenyl phosphate disodium salt hexahydrate (pNPP) was used as substrate to determinate APase activity. At first, 0.5 ml enzyme extract with a total volume of 4 ml containing 15 mM MES buffer and 10 mM pNPP was incubated at 37°C for 30 min. Subsequently, an equal volume of 0.25 M NaOH was added to terminate reaction immediately. APase activity was measured from the release of p-nitrophenol (pNP) and expressed as pNP μg g⁻¹ fresh weight (FW) min⁻¹, and pNP was determined spectrophotometrically using a UV spectrophotometer (model UN-2600A, UNICO) at 412 nm relatively to standard solutions (Sharma and Sahi, 2005).

P concentration of root and shoot was analyzed following the vanadomolybdate method (Page, 1982). Shoot and root P accumulation was calculated by multiplying P concentration with the DW, respectively.

The ratio of root P accumulation: shoot P accumulation is a typical index for plant response to P deficiency. Generally, plant response to P deficiency is the increase in root: shoot ratio, which might be due to preferential assimilate P distribution to the roots (Vance et al., 2003).

Data from 3 replicates were sorted out by Excel (Microsoft) software packages. Regression equations were developed for the relationships between yield and seeds P concentration, shoot P uptake, HI. The liner-plateau model was analyzed by the SAS 9.1.3 software (SAS Institute Inc., USA). The liner model was analyzed by the SPSS 19.0 software. Significant difference of shoot, root dry matter and P accumulation, root length, root surface area, root volume, specific root length, activities of APase and organic acid exudation rate between soybean genotypes and P treatments were analyzed by analysis of variance (ANOVA) and mean values were compared by least significant difference (LSD) multiple comparison using the SPSS19.0 software (SPSS Institute Inc., USA).

To screen P-efficient soybean genotypes for achieving high yield and improving P use efficiency, 274 soybean genotypes were collected from Southwest of China (Figure 1). Besides low availability of soil P, low P use efficiency partly resulted from the behavior of fertilizer P application by farmers, who often input more P than crops need (Vitousek et al., 2009). The remaining P accumulation in the soil is easily eroded with soil by rainfall (Zhang et al., 2012). To sustainably improve crop production and fertilizer P used efficiency through exploiting the biological P potential in the soil, employing the P-efficient genotypes is an efficacious strategy (Li et al., 2011). In this study, grain yield, shoot P accumulation, seed P concentration and harvest index (HI) of 274 soybean genotypes were analyzed in Purple soil with a suitable soil P concentration (initial Olsen-P concentration was 17.4 mg kg⁻¹). The results showed that seed P concentration was diluted and shoot P accumulation improved following increasing yield (Figure 2). The 274 soybean genotypes were originated from a large span area with low availability of soil P (Figure 1), which differed substantially in grain yield and P accumulation potentials (Figure 2). The yield of 274 soybean genotypes ranged from 5.5 to 36.0 g plant⁻¹. The seed P concentration and shoot P accumulation ranged from 0.045 to 0.93% and 0.065 mg plant⁻¹ to 0.278 mg plant⁻¹, respectably (Figure 2). It suggested that genetic variations existing in the 274 soybean genotypes, and which provide a potential for assessment P-efficient soybean genotypes (Pan et al., 2008). Such substantial genotypic variation in response to different P use efficiency was also detected for a considerable number of soybean genotypes in South and Northeast of China (Zhao et al., 2004; Pan et al., 2008). In the field study, based on the stand enumerated before 5 P-efficient and 5 P-inefficient soybean genotypes were chosen (Figure 2). Obviously, the evidence mentioned above to proving the soybean genotypes with different P use efficiency is not enough, much more work should be provided to support the results as root physiological, chemical characteristics. Hydroponic study is a fine pathway to check plant root characteristics as it is pellucid and efficient by short plant growth period.

Crop P efficiency was defined as the ability to produce biomass or yield under certain available P supply conditions (Wissuwa et al., 2009), and further explained the capability of P uptake from the media and tolerance in P insufficient conditions (Wissuwa and Ae, 2001; Wissuwa et al., 2005; Wang et al., 2010). In the hydroponic experiment, P-efficient genotypes exhibited better adaptability and tolerance than the P-inefficient genotypes grown in low P solution (Figure 3C). Especially E141 and E311, showed significantly greater biomass than that of P-inefficient genotypes in low P level (Figure 3). These results corresponded well with previous studies which showed that the P-efficient plant genotypes demonstrated greater biomass compared to the P-inefficient when grown in low P condition. Some scholars had reported that P-efficient plants of soybean (Zhao et al., 2004), rice (Oryza sativa L.) (Wissuwa and Ae, 2001; Mori et al., 2016), maize (Zea mays L.) (Zhang, 2012), and Brassica napus (Zhang et al., 2009) showed significantly biomass ascendancy compared with P-inefficient plants in P shortage conditions.

Early studies suggested that P-efficient plants preferential distribution P to roots for improving the tolerance to P deficiency and stimulate root growth and P uptake (Wissuwa et al., 2005). In present study, much more P distribution to roots of P-efficient soybean genotypes in low P condition may be an adaptation involved in increasing the tolerance to P deficiency (Table 1). Tolerant plants have been more efficient in P uptake per root size and the additional P then drove further root growth, assuming that low P availability affected root biomass accumulation directly (Wissuwa et al., 2005). In this study, P accumulation of E141 and E311 was higher than the other genotypes in low P solution (Figure 4C). Some scholars reported that P-efficient soybean genotypes are able to obtain sufficient P from acid red soil and alkaline soil under P lack conditions (Zhao et al., 2004; Pan et al., 2008), and other P-efficient plant species also showed positively P accumulation grown in P lack conditions, like maize (Liu et al., 2004), wheat (Triticum aestivum L.) (Fageria and Baligar, 1999) and rice (Mori et al., 2016). It indicated that the P-efficient soybean genotypes (especially E141 and E311) demonstrate great capability on P absorption and accumulation potentials.

Root physiological adaptations play important roles in enhancing soil P bioavailability and crop P use efficiency (Shen et al., 2011). These adaptation mechanisms mainly include altering root morphology to enhance P absorption (Wissuwa, 2003), facilitating organic acids exudation into the rhizosphere to increase P availability by mobilizing sparingly soluble mineral P and organic P sources (Johnson et al., 1996), and promoting phosphatase exudation to mineralize Po (Li et al., 2012). Root morphological traits closely linked with P acquisition ability of plants (Pang et al., 2010), and fine root morphology is propitious to maximize P assimilation. P-efficient soybeans showed high parameters of root length, root surface area and root volume than the P-inefficient in present study (Table 2). The specific root length of E311 increased by 48.6% compared with that in high P level (Table 2). It indicated that penurious P availability has negative effects on root morphology, however the less reduction of those root morphology parameters of P-efficient genotypes seems to be the results of excellent tolerance and adaptation. It reported that P-efficient plant genotypes altered root morphology to adapt low P conditions, and then assimilated more P than P-inefficient genotypes (Fita et al., 2011). There was great possibility for plants in improving soil P excavation and utilization by increasing root length, root surface area, root volume and specific root length(Vance et al., 2003), which result in a large total amount of P and dry matter accumulation (Zhang et al., 2013).

Besides root morphology, exudation of APase to improve P bioavailability is another important root physiological adaptation. As we know, exudation of APase by roots increased under P deficiency condition, and plant roots with high APase activities have great potential to utilize soil Po (Hayes et al., 2000; Lung and Lim, 2006). In this study, the results corresponded well with earlier researches that P shortage conditions stimulate the APase activity of soybean genotypes, especially the P-efficient soybean genotypes (Figure 5). Po comprises 30–80% of the total P in most agricultural soils (Dalai, 1977), it can be released through mineralization processes mediated by enzymes activity. The P-efficient plant species generally had high enzymes activity in root extracts when grown in P lack or high Po media. For instance, maize, lupin and chickpea showed high APase activities when grown in P deficiency conditions (Wasaki et al., 2003; Li et al., 2004), Gulf, Marshall Ryegrass and P. hydropiper had greater APase activities when grown in phytate-sufficient media (Sharma and Sahi, 2011; Huang et al., 2012). In our results, the APase activities of E311 and E141 were significantly higher (1.2~2.3-folds) than others in low P level (Figure 5C), suggesting that they demonstrated great capabilities of activation and utilization soil P. Besides, transgenic expression of APase gene in P-efficient genotypes was another path for great dry matter and P accumulation. It reported that over-expressed AtPAP15 in soybean and GmPAP4 in Arabidopsis significantly increased dry matter and P accumulation compared to the wild type when grown in media with Po (Wang et al., 2009; Kong et al., 2014). Above all, APase activity was a symbol of efficient mineralization and utilization soil P by plants.

Root exudation of organic acids into the rhizosphere had been proposed to improve soil P availability and plant P accumulation (Dinkelaker et al., 1989; Johnson et al., 1996). Exudation of organic acids is also an important root physiological adaptation to P deficiency. A 2-fold increase in exudation of citrate was observed under P starvation in alfalfa (Lipton et al., 1987). Dong et al. (2004) also reported that oxalate and malate exudation of soybean plants was found to markedly increase in response to P deficiency. In present study, low P condition stimulates exudation of malate, citrate, tartrate and oxalate acid, especially in the P-efficient soybean genotypes (Table 3). The average exudation rate of malate, citrate, tartrate and oxalate acid in the P-efficient soybean genotypes increased by 2.24, 1.40, 2.98, and 2.54 folds compared to the P-inefficient genotypes under P starvation, respectively. Particularly, E141 and E311 exhibited high organic acid exudation potentials (Table 3). These results were in agreement with Dong et al. (2004), who reported that a considerable amount of organic acid exudation in P-efficient soybean genotypes contributed to P accumulation under P starvation condition. Adequate evidences demonstrated that root exudation of organic acids into the soil rhizosphere contributed to higher P acquisition. Exudation of malate in P-efficient faba bean enhanced P acquisition in the alkaline soil (Rose et al., 2010). Barley (Hordeum vulgare L.) genotypes expressing the TaALMT1 gene from wheat improved P uptake per unit root length from an acid soil (Delhaize et al., 2009).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.