Leaf phosphorus fractionation in rice to understand internal phosphorus-use efficiency

Abstract

Background and Aims

Phosphorus (P) availability is often limiting for rice (Oryza sativa) production. Improving internal P-use efficiency (PUE) is crucial to sustainable food production, particularly in low-input systems. A critical aspect of PUE in plants, and one that remains poorly understood, is the investment of leaf P in different chemical P fractions (nucleic acid-P, lipid-P, inorganic-P, metabolite-P and residual-P). The overarching objective of this study was to understand how these key P fractions influence PUE.

Methods

Three high-PUE and two low-PUE rice genotypes were grown in hydroponics with contrasting P supplies. We measured PUE, total P, P fractions, photosynthesis and biomass.

Key Results

Low investment in lipid-P was strongly associated with increased photosynthetic PUE (PPUE), achieved by reducing total leaf P concentration while maintaining rapid photosynthetic rates. All low-P plants exhibited a low investment in inorganic-P and lipid-P, but not nucleic acid-P. In addition, whole-plant PUE was strongly associated with reduced total P concentration, increased biomass and increased preferential allocation of resources to the youngest mature leaves.

Conclusions

Lipid remodelling has been shown in rice before, but we show for the first time that reduced lipid-P investment improves PUE in rice without reducing photosynthesis. This presents a novel pathway for increasing PUE by targeting varieties with reduced lipid-P investment. This will benefit rice production in low-P soils and in areas where fertilizer use is limited, improving global food security by reducing P fertilizer demands and food production costs.

Inorganic P, lipid remodelling, Oryza sativa (rice), phospholipids, phosphorus, phosphorus-use efficiency (PUE), phosphorus fractionation, photosynthetic phosphorus-use efficiency (PPUE)

INTRODUCTION

Phosphorus (P) is an essential plant nutrient required for a range of critical processes, including genetic control, cell structure and energy transfer. Phosphorus is one of the most commonly limiting nutrients for plant productivity, being deficient in over half of the world’s agricultural soils, and is therefore used extensively as a fertilizer (Lynch, 2011; Heuer et al., 2017). The main source of P fertilizers is non-renewable rock phosphate. However, low-cost safe rock phosphate reserves are expected to be exhausted in the next 50–100 years, driving up future costs of food production and impacting global food security (Cooper et al., 2011; Fixen and Johnston, 2012). Furthermore, the irresponsible use of P fertilizer leads to P loss via run-off, causing off-site environmental problems such as eutrophication of waterways (Brinch-Pedersen et al., 2002). One strategy to address these issues is to improve the efficiency of P utilization in agriculture, thus reducing fertilizer requirements. Efforts are underway to develop P-efficient crop cultivars that will maintain yield at reduced fertilizer requirements, thus combining the aims of securing future food security and reducing the environmental footprint of crop production.

There are two main strategies to improve P-use in plants: increased P-acquisition efficiency (PAE) and increased internal P-use efficiency (PUE). Here, we focus on the latter, as research into PUE holds significant potential for advances, particularly in low-input agricultural systems (Wang et al., 2010; Veneklaas et al., 2012; Heuer et al., 2017). A critical aspect of P use in plants, and one that remains poorly understood, is how a plant invests P within its leaves, specifically the allocation of P to different chemical P fractions (Veneklaas et al., 2012; Mo et al., 2019; Yan et al., 2019; Crous and Ellsworth, 2020; McQuillan et al., 2020).

Phosphorus in plants can be separated into five distinct fractions, inorganic-P (P_i, inorganic phosphate), lipid-P (P_L, phospholipids), metabolite-P (P_M, e.g. sugar phosphates and ATP), nucleic acid-P (P_N, RNA and DNA) and residual-P (P_R, e.g. phosphoproteins), with the latter four combining to form organic-P (Kedrowski, 1983; Hidaka and Kitayama, 2011; Veneklaas et al., 2012; Jeong et al., 2017; Yan et al., 2019). It is recognized that to achieve rapid rates of photosynthesis with high photosynthetic PUE (PPUE) requires a fine balance in investment among these five fractions (Heldt et al., 1977; Stitt et al., 2010).

Research into native species adapted to severely P-impoverished ecosystems has revealed that highly P-efficient species alter their allocation of P fractions to improve PUE, a strategy also desirable in agricultural species (Veneklaas et al., 2012). These alterations to P fractions include reductions in lipid P investment through lipid remodelling, reduced investment in P_N and reductions in P_i (Lambers et al., 2012; Sulpice et al., 2014; Yan et al., 2019).

One key strategy to improve PUE is lipid remodelling. This process reduces investment in phospholipids, by substituting P-containing phospholipids with non-P-containing glycolipids. For example, species from P-impoverished south-western Australia replace phospholipids with sulfolipids and galactolipids during leaf development, thus reducing P_L investment in mature leaves (Lambers et al., 2012). Similarly, a low-P-tolerant ancient rice cultivar (Akamai) replaces more phospholipids with non-P-containing glycolipids, compared with a low-P-sensitive cultivar (Koshihikari), suggesting a link between low-P tolerance and lipid remodelling (Tawaraya et al., 2018). Despite lipid remodelling being linked with low-P environments, we still do not know how it changes in relation to other P fractions, how this impacts photosynthesis or its contribution to variation in PPUE among current crop varieties.

Another key strategy utilized by species adapted to P-impoverished landscapes is to reduce investment in P_N (predominantly rRNA) (Sulpice et al., 2014). Nucleic acid-P is generally the largest P fraction in leaves in plants grown under low- to adequate P conditions, making this fraction a prime candidate for economizing on P use (Veneklaas et al., 2012). For example, P-efficient Proteaceae from south-western Australia show very low levels of rRNA; however, whilst this is not associated with slow rates of photosynthesis (Sulpice et al., 2014), these species typically do show slow rates of plant growth (Barrow, 1977; Han et al., 2021). The growth rate hypothesis predicts that faster growing plants probably need relatively more rRNA to support faster rates of protein synthesis, but it remains unknown if faster growing crop plants are able to economize on P through a reduced investment in RNA, without compromising growth or their ability to rapidly change to environmental stresses through faster protein turnover in mature leaves (Matzek and Vitousek, 2009). Nucleic acid-P is the largest P fraction in leaves of rice (40 %) (Jeong et al., 2017) and barley (30 %) (Chapin and Bieleski, 1982) when grown under adequate P supply; however, it remains unclear if rice or other crop plants are able to reduce investment in P_N under low-P conditions, or if it is maintained, possibly to sustain their faster growth rates and rapid responses to environmental stresses (Matzek and Vitousek, 2009; Yan et al., 2019).

The allocation of P fractions other than nucleic acids also changes with external P supply, and this is particularly the case for P_i and P_L, which increase under excess P supply and decrease under P limitation (Veneklaas et al., 2012; Mo et al., 2019). This has been shown in crop and pasture plants such as Brassica napus, Cucurbita maxima (Pant et al., 2008), Oryza sativa (Jeong et al., 2017) and Medicago truncatula (Branscheid et al., 2010), as well as native plants such as Hakea prostrata (Shane et al., 2004). However, very little is known of how the remaining P fractions (P_N, P_M and P_R) change under P stress, or how this impacts physiological processes.

Rice (Oryza sativa) is the second most widely grown cereal crop and a staple for more than half the world’s population, making it a critical resource in future global food security (Van Nguyen and Ferrero, 2006). Phosphorus is often limiting for rice production, particularly in low-input systems, where limited financial and agricultural resources restrict the use of artificial P fertilizers (Saito et al., 2019). Low-input rice-producing regions throughout Asia, as well as sub-Saharan Africa, will benefit from the development of new P-efficient rice varieties that can produce high yields with minimal fertilizer inputs (Saito et al., 2019). There are continued efforts to develop more efficient varieties of rice for use in these low-input regions. Potential donor genotypes with high PUE have been identified (Rose et al., 2016; Watanabe et al., 2020), but it is not known how higher PUE is achieved in these donors.

The overarching objective of this study was to understand how P fraction allocation influences PUE. We grew five contrasting rice genotypes in hydroponics at low and high P supply. This included three P-efficient (high-PUE) and two less P-efficient (low-PUE) genotypes. After approx. 5 weeks under treatment, plants were harvested and the following measured: whole-plant PUE, PPUE, total P, P fractions, photosynthesis and biomass. The aim of this research was to: (1) understand how P fraction allocation patterns change under low-P conditions; (2) explore any differences in P fraction allocation among contrasting rice genotypes; and (3) assess other whole-plant traits contributing to improved PUE.

MATERIALS AND METHODS

Genotype selection

Five rice genotypes of contrasting PUE were selected based on the results of previous experiments (Wissuwa et al., 2015; Rose et al., 2016; Watanabe et al., 2020). These comprised three high-PUE genotypes [DJ123 (DJ), Mudgo (MU) and Yodanya (YO)] and two low-PUE genotypes [Taichung (TA) and IR64 (IR)] (Supplementary data Table S1).

Plant growth

Plants were grown from seed and transferred to hydroponics. Seeds were sterilized (1 % w/v NaClO for 5 min) and germinated in Petri dishes, until initial root emergence (30 °C, dark, 2 d). Germinated seeds were then transferred onto floating plastic mesh in trays containing 8 L of solution (100 µm Ca, CaCl₂; 10 µm Fe, Fe-EDTA; pH 5.8), and placed in the glasshouse. At 6 days after germination (DAG), 5 % Yoshida solution (excluding N and P); (Yoshida et al., 1976) was added to trays. From 8 to 18 DAG, seedlings were placed in larger 45 L tubs for further growth, with each tub containing 10 % Yoshida solution, replenished after 5 d.

At 18 DAG (approx. 3–4 fully matured leaves) 24 seedlings of uniform size were selected for each genotype, transferred to a 1.1 L black bottle and exposed to two treatments. Two seedlings of the same genotype were transferred to each 1.1 L bottle, with plants held in place by a grey-foam strip (n = 6 bottles per treatment × genotype). During the treatment period, each bottle was supplied with the equivalent of a 75 % Yoshida solution each week, given in 2–4 applications, and replenished weekly. The P treatments were applied weekly at each change of nutrient solution, with two P treatments, 0.9 mg (LP) and 10 mg (HP) P in total, as KH₂PO₄, these were given in four doses (LP; 18, 26, 33 and 40 DAG) or five doses (HP; 18, 26, 33, 40 and 47 DAG). To correspond with changes in growth, P treatments were applied in increasing concentrations, ranging from 4 to 8 µm (LP) or from 35 to 70 µm (HP) (see Supplementary data Table S2). Under the LP treatment, all supplied P was taken up between doses (data not shown). All Yoshida solutions excluded P and contained 10–100 µm potassium silicate, used to maintain pH at 5.8. Plants were grown between September and November 2018, in a naturally lit temperature-controlled glasshouse at JIRCAS, Tsukuba, Japan; at a mean temperature of 28 °C (25–35 °C) and a mean relative humidity of 33 % (8–79 %).

Photosynthetic capacity

Light-saturated net photosynthetic rates (A_sat) were measured at 48 and 49 DAG using an infrared gas analysis system (LI-COR 6400, LI-COR Inc., Lincoln, NE, USA). Measurements were made per bottle, by combining the youngest fully expanded leaf (L1) of both plants (n = 6). All measurements were made using a 6 cm² chamber (3 × 2 cm), under a saturating light of 1700 µmol m^–2 s^–1 photosynthetic photon flux density (red–blue light source; 6400-02B) at 400 µmol mol^–1 CO₂. Flow rate was set at 500 µmol s^–1, block temperature at 28 °C and relative humidity maintained at 60–80 %. Where the leaf area in the chamber was <6 cm², it was measured using a ruler (width × length).

Harvest measurements

Plants were harvested at 53–55 DAG (approx. 5 weeks under treatment). Both plants from each bottle were combined for all measurements/sampling. Individual plants were rinsed in deionized water and separated into leaf one (L1; youngest fully expanded leaf blade), leaf two (L2; second youngest fully expanded leaf blade), older leaves (OL; all remaining green mature leaf blades), stem (ST; includes stem, leaf sheaths, immature leaves and leaves with ≥20 % yellowing, senesced leaves) and roots (RO). New leaves were considered fully expanded when they exceeded the size (length) of the next oldest leaf. Both L1 and L2 were identified based on their order down the stem.

All green leaves (L1, L2 and OL) were immediately scanned at 150 dpi (Epson V700, Seiko Epson Corp., Nagano, Japan), with total leaf area later determined for each organ type using Fiji image analysis software (Schindelin et al., 2012). Samples for P fractionation analysis (L1, L2 and a sub-sample of RO, approx. 3 cm from the mid-region of roots) were weighed fresh (f. wt), wrapped in aluminium foil and immediately frozen in liquid nitrogen and kept at –80 °C, before being freeze-dried (5 d; Eyela FDU-2100, Tokyo Rikakikai Co., Ltd, Tokyo, Japan), weighed dry (d. wt) and analysed for P fractions (see below). All other organ types (OL, ST and RO) were oven-dried (70 °C, 72 h) and weighed, before being analysed for total [P] (see below).

Total phosphorus and phosphorus fractionation

Dried sub-samples of OL, ST and RO were ground and acid digested (nitric:perchloric; 3:1) before being analysed for total P using the malachite green colorimetric method (Motomizu et al., 1983).

The total P was divided into five separate P fractions following a modified P fractionation assay, based on Jeong et al. (2017) (Chapin and Kedrowski, 1983; Kedrowski, 1983; Hurley et al., 2010; Hidaka and Kitayama, 2011, 2013; Yan et al., 2019). The five P fractions were P_L (phospholipids), P_M (soluble P-containing small metabolites such as sugar phosphates and ATP), P_i (soluble inorganic phosphate), P_N (RNA and DNA) and P_R (phosphoproteins, and other recalcitrant complexes). The basis of this separation was differential hydrolysis, such that at each stage the desired fraction was completely hydrolysed, whilst the non-desired fractions remained unchanged. First, the P_L, P_M–P_i (P_M and P_i combined), P_N and P_R fractions were extracted and quantified (Kedrowski, 1983; Hidaka and Kitayama, 2011), then the P_i fraction was extracted and quantified alone (Hurley et al., 2010). A detailed description of the P fractionation procedures can be found in Supplementary data Methods S1 and S2; Figs S1 and S2.

In brief, the freeze-dried material (L1, L2 and RO sub-sample) collected for P fractionation was ground using plastic vials and ceramic beads in a vertical ball-mill grinder. A sub-sample of the freeze-dried material (approx. 50 mg) was weighed into a 2 mL tube and extracted three times with 1 mL of cold chloroform:methanol:formic acid (CMF; 12:6:1 v/v/v), followed by three extractions with 1.26 mL of cold chloroform:methanol:water (CMW; 1:2:0.8 v/v/v). Supernatants from these extractions were combined, and 1.9 mL of chloroform-washed water (CWW) was added, resulting in a biphasic solution with an aqueous-based upper layer and a lipid-based lower layer, separated by a thin semi-solid interfacial protein layer. The aqueous and lipid layers were carefully extracted and transferred to 50 mL tubes, labelled P_M–P_i tube (aqueous layer) and P_L tube (lipid layer). The remaining interfacial layer was rinsed with 1.44 mL of CMF:CMW:CWW (1/1.26/0.62 v/v/v) and separated again. The P_L tube represented the P_L fraction (lipid-P) and was set aside.

The CMF/CMW extracted pellet was dried under vacuum and extracted with 1 mL of 85 % methanol, the supernatant was added to the P_M–P_i tube and the pellet dried under vacuum. To this dried pellet we transferred the interfacial layer using 1.3 mL of 5 % trichloroacetic acid (TCA) and extracted this for 1 h at 4 °C, inverting every 10 min, with the supernatant added to the P_M–P_i tube. This was repeated again using 1 mL of 5 % cold TCA. The P_M–P_i tube represented the P_M and P_i fractions (metabolite-P and inorganic-P) and was set aside.

The pellet was further extracted three times with 1 mL of 2.5 % TCA in a heating block at 95 °C for 1 h. The resulting supernatants were added to a new 50 mL tube, labelled P_N tube. The P_N tube represented the P_N fraction (nucleic acid-P) and was set aside. The remaining pellet was then quantitatively transferred using 2 mL of 85 % methanol to a new 50 mL tube, labelled P_R tube. The P_R tube represented the P_R fraction (residual-P) and was set aside.

The four 50 mL digestion tubes containing the four fractions were gently evaporated in a fume-hood at 50 °C; the dried residue was then acid digested (nitric:perchloric; 3:1) before being analysed for P using the malachite green colorimetric method (Motomizu et al., 1983). To reduce conversion of P fractions during extraction, all samples were kept on ice, unless otherwise specified.

The P_M–P_i fraction was further split by separately quantifying P_i and using the difference to determine P_M. The P_i fraction was extracted using an acetic acid extraction (Supplementary data Method S2; Fig. S2) (Hurley et al., 2010). In brief, approx. 10 mg of dried sample was weighed into a 2 mL screw cap tube, along with two small chilled ceramic beads and 1.5 mL of cold 1 % glacial acetic acid. This was homogenized and incubated twice (4 °C, total 0.5 h). The final clear supernatant was transferred and stored at –20 °C, before analysing P using the malachite green colorimetric method (Motomizu et al., 1983). As expected, the P_i fraction was always lower than the combined P_M–P_i fraction.

Total P for L1 and L2 was calculated as the sum of all P fractions. To verify our P fractionation method, the sum of all P fractions was directly compared with total P determined through acid digestion, as described above. Our P fractionation method showed a high recovery of >95 % of total P (Supplementary data Table S3). Internal reference material was included regularly and showed high levels of consistency in our P fractionation method (Supplementary data Table S3).

Calculations and statistics

Three measures of PUE were calculated: PPUE (photosynthetic PUE), PUE_B (biomass PUE) and PUE_P (physiological PUE) (see Table 1 for formulae). Photosynthetic PUE reflects leaf-level PUE efficiency, as the rate of C fixation per unit leaf P. Biomass PUE (PUE_B) and physiological PUE (PUE_P) reflect whole-plant level PUE efficiency, in terms of whole-plant P content (PUE_B) or leaf P concentrations (PUE_P), thus all reflecting different aspects of PUE. Root:shoot ratios were calculated based on total above-ground d. wt (L1, L2, OL and ST) and root d. wt (RO). Resource allocation to younger leaves was determined by the percentage of mature leaf (ML: sum of L1, L2 and OL) d. wt, P content and area allocated to the two youngest fully matured leaves (L1 and L2).

Table 1.

Definition and calculation of leaf-level (PPUE) and plant-level (PUE_B, PUE_P) phosphorus-use efficiencies.

Abbreviation	Definition	Calculation	Unit (short unit)
PPUE	Photosynthetic P-use efficiency:	$\frac{A_{\mathrm{s}\mathrm{a}\mathrm{t}}}{\left[{\mathrm{P}}_{\mathrm{L}1}\right]\times \mathrm{L}\mathrm{M}{\mathrm{A}}_{\mathrm{L}1}}$	µmol CO2 g–1 P s–1 (µmol g–1 s–1)
	Rate of net C fixation per unit leaf P (leaf level)
PUEB	Biomass P-use efficiency:	$\frac{\mathrm{B}}{\mathrm{P}}$	g d. wt mg–1 P (g g–1)
	Total biomass produced per unit total P in the plant (plant level)
PUEP	Physiological P-use efficiency:	$\frac{\mathrm{B}}{[\mathrm{P}]}$	g2 d. wt mg–1 P (g2 mg–1)
	Total biomass produced relative to green leaf P concentration (plant level)

A_sat = light-saturated net photosynthetic rate; [P_L1] = phosphorus (P) concentration per dry weight (d. wt) in the youngest fully matured leaf; LMA_L1 = leaf mass per unit leaf area in the youngest fully matured leaf; B = total plant biomass as d. wt; P = total plant P content; [P] = P concentration per d. wt in all mature green leaves.

Differences in traits across genotypes and treatments, including their interaction, were tested using generalized least squares models (Pinheiro and Bates, 2000). The relationships between PPUE and P fraction allocations, and between A_sat and P fraction concentrations, were inspected through regression analysis. The regression models were selected based on Akaike and Bayesian information criteria (AIC and BIC), and the model’s parameters (P-value and standardized regression coefficient) were presented when significant (P < 0.05). The residuals of each model were visually inspected for heteroskedasticity. In the presence of heteroscedasticity, appropriate variance structures were specified if they significantly improved the model, based on AIC and BIC values (Pinheiro and Bates, 2000). Differences among genotypes and treatments were defined using Tukey HSD post-hoc tests (Hothorn et al., 2008). The relationship between total leaf [P] and P fraction concentrations was inspected using standardized major axis (SMA) regression, as the relationship between these traits was symmetrical, i.e. traits could be placed on either axis (Warton et al., 2006, 2012).

Principal component analysis (PCA) was used to visually demonstrate separation of contrasting genotypes, based on key traits (Clarke, 1993). These key traits were: three measures of PUE (×3), organ-level total [P] (×5), P fraction allocations (×5), total d. wt, root:shoot DW ratio, leaf area, A_sat and percent d.wt allocated to the youngest leaves. In the low-P treatment, there were four missing values in a dataset of 540 (due to low biomass in older leaves). To create a complete dataset, these four values were imputed using an iterative PCA approach (Josse and Husson, 2016). PCA was then performed on the full dataset (540 values, 18 traits and 30 individuals), under either low- or high-P treatment. PCA was conducted using the R function ‘prcomp’; results were extracted using ‘factoextra’ and plotted using ‘ggbiplot’ (Vu, 2011).

All statistical analyses were performed using the R software platform (R Core Team, 2020) with the packages: ‘emmeans’ (Lenth, 2020), ‘factoextra’ (Kassambara and Mundt, 2020), ‘missMDA’ (Josse and Husson, 2016), ‘multcomp’ (Hothorn et al., 2008), ‘multcompview’ (Graves et al., 2015), ‘nlme’ (Pinheiro et al., 2020), ‘smatr’ (Warton et al., 2012), ‘stats’ (R Core Team, 2020) and ‘tidyverse’ (Wickham et al., 2019).

RESULTS

Phosphorus-use efficiency

Photosynthetic PUE (PPUE; rate of net C fixation per unit of P) was higher under low-P treatment, compared with high-P treatment (P < 0.001; Fig. 1A). Under low-P treatment, the high-PUE genotypes tended to show a higher PPUE (588–719 µmol g^–1 s^–1) compared with low-PUE genotypes (497–598 µmol g^–1 s^–1). In contrast, there was far less variation under high-P treatment.

Fig. 1.

Measures of phosphorus (P)-use efficiency (PUE) in contrasting rice genotypes grown under low-P and high-P treatment. (A) Photosynthetic PUE (PPUE), (B) biomass PUE (PUE_B) and (C) physiological PUE (PUE_P); see Table 1 for definitions. Bar heights show means, error bars show standard errors (n = 5–6) and different letters indicate significant differences among genotypes within each treatment (post-hoc Tukey test, P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

Whole-plant PUE was assessed using biomass PUE (PUE_B; total biomass produced per unit total P in the plant) and physiological PUE (PUE_P; total biomass produced relative to mature green leaf P concentration). Both showed similar differences among genotypes and, as expected, were higher under low-P treatment than under high-P treatment (P < 0.001; Fig. 1B, C). For example, genotypes showed a 288–361 % higher PUE_B and 22–88 % higher PUE_P under low-P treatment, compared with high-P treatment (Fig. 1B, C). Under low-P treatment, the high-PUE genotypes all showed a higher whole-plant PUE compared with low-PUE genotypes (P < 0.05). Genotype MU showed the highest whole-plant PUE among genotypes, under both low-P and high-P treatment. In summary, DJ and MU showed both high PPUE and high whole-plant PUE under low-P conditions, whilst under high-P conditions MU had the highest values.

Total phosphorus concentration

Total [P] varied among genotypes and between treatments (Fig. 2), being on average 3.9-fold greater under high-P treatment compared with low-P treatment (P < 0.001). Total [P] was almost always lower in high-PUE genotypes compared with low-PUE genotypes, across all organ types and both treatments (P < 0.05; Fig. 2). Under low-P treatment, L1 had the highest [P]. In L1 under low-P treatment, the high-PUE genotype DJ had the lowest [P] and TA/IR the highest values. Under high-P treatment, MU showed the lowest L1 [P].

Fig. 2.

Total phosphorus (P) concentrations ([P]) of contrasting rice genotypes grown under low-P and high-P treatment, showing the four main organ types. The four main organ types are leaf 1 (youngest fully expanded leaf blade), leaf 2 (second youngest fully expanded leaf blade), older leaves (all remaining green mature leaf blades) and roots (all roots). Bar heights show means, error bars show standard errors (n = 5–6) and different letters indicate significant differences among genotypes (post-hoc Tukey test, P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

Phosphorus fraction allocation

Phosphorus fraction allocation patterns are shown by the percentage of total P allocated to each P fraction for L1 (Fig. 3A), L2 (Supplementary data Fig. S3A) and RO (Fig. 4A). Phosphorus allocation patterns were similar between L1 and L2, and values will therefore only be described in detail for L1. Under low-P treatment, there was a decrease in the percentage allocated to P_i and P_L, and an increase in that allocated to P_N: P_N (50 %), P_i (18 %), P_L (13 %), P_M (10 %) and P_R (9 %; mean values). Under high-P treatment, most leaf P was allocated to the P_N (36 %), followed by P_i (30 %), P_L (19 %), P_N (9 %) and P_R (6 %; mean values, Fig. 3A).

Fig. 3.

(A) Phosphorus (P) fraction allocation (percentage of total P allocated to each fraction) and (B) P fraction concentrations of the youngest fully expanded leaf (leaf 1) of contrasting rice genotypes grown under low-P and high-P treatment, showing all five P fractions. Pie charts show mean P fraction allocations averaged across all genotypes. The five P fractions are nucleic acid-P (P_N, RNA and DNA), inorganic-P (P_i, soluble inorganic phosphate), lipid-P (P_L, phospholipids), metabolite-P (P_M, soluble P-containing small metabolites such as sugar phosphates and ATP) and residual-P (P_R, phosphoproteins and other recalcitrant complexes). Bar heights show means, error bars show standard errors (n = 5–6) and different letters indicate significant differences among genotypes, within each P fraction (post-hoc Tukey test, P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

Fig. 4.

(A) Phosphorus (P) fraction allocation (percentage of total P allocated to each fraction) and (B) P fraction concentrations of the roots (RO) of contrasting rice genotypes grown under low-P and high-P treatment, showing all five P fractions. Pie charts show mean P fraction allocations averaged across all genotypes. The five P fractions are nucleic acid-P (P_N, RNA and DNA), inorganic-P (P_i, soluble inorganic phosphate), lipid-P (P_L, phospholipids), metabolite-P (P_M, soluble P-containing small metabolites such as sugar phosphates and ATP) and residual-P (P_R, phosphoproteins and other recalcitrant complexes). Bar heights show means, error bars show standard errors (n = 5–6) and different letters indicate significant differences among genotypes, within each P fraction (post-hoc Tukey test, P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

Under low-P treatment, the high-PUE genotype DJ showed the most contrasting P allocation pattern among genotypes, followed by MU (Fig. 3A). Both tended to allocate less P to P_L, whilst DJ alone showed a greater relative allocation to P_N.

Differences in P allocation among genotypes were directly linked with differences in PPUE under low-P treatment (Fig. 5). Photosynthetic PUE was negatively correlated with both P_L and P_M allocation (P < 0.001 and P = 0.004, respectively), meaning that a lower allocation to these fractions was associated with a higher PPUE. In contrast, PPUE was positively correlated with P_N and P_R (P = 0.008 and P = 0.015, respectively), and there was no correlation with P_i (P = 0.148). The relationship between PPUE and P_L was strongest, suggesting that a reduced allocation to P_L was tightly associated with improved PPUE.

Fig. 5.

Photosynthetic phosphorus (P)-use efficiency (PPUE) of contrasting rice genotypes in relation to percentage total P allocated to each P fraction, under low-P treatment. Each line represents the line of best fit, derived from a linear model. Only significant relationships are shown (P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR). β, regression coefficient.

Under high-P treatment, MU showed the most contrasting P allocation pattern (Fig. 3A). MU tended to show a lower allocation of P to P_i and relatively more to P_L, compared with other genotypes. In contrast to low-P treatment, PPUE did not correlate with any of the P fraction allocations under high-P treatment, except P_L (P = 0.007; Supplementary data Fig. S4). Furthermore, this was a positive correlation, opposite to that under low-P treatment.

Root P fraction allocation patterns were different from that in leaves but showed less variation among genotypes (Fig. 4A). Most of the P in roots, under both treatments, was allocated to the P_N fraction, the same as in leaves; however, roots showed a much lower allocation to P_i and P_M, and a greater allocation to P_L, resulting in the following order under low-P treatment (decreasing): P_N (51 %), P_L (19 %), P_R (13 %), P_i (12 %) and P_M (5 %); and under high-P treatment: P_N (39 %), P_L (30 %), P_i (14 %), P_M (10 %) and P_R (7 %; mean values).

Phosphorus fraction concentration

It is important to note that P fraction allocations do not always reflect P fraction concentrations (Figs 3B and 4B; Supplementary data Fig S3B). An increase in relative allocation of P to one fraction does not mean that that fraction increases in concentration, only that it increases relative to other fractions.

As expected, all P fraction concentrations decreased under low-P treatment compared with high-P treatment. Under low-P treatment, the P fraction concentrations of high-PUE genotypes were lower than or equal to that of the low-PUE genotypes, reflecting their generally lower total [P] (Figs 3B and 4B; Supplementary data Fig S3B). Under low-P treatment, P_L and P_N showed the most significant differences among genotypes, with [P] as much as 49 µg g^–1 (P_L) and 82 µg g^–1 (P_N) lower in high-PUE genotypes compared with low-PUE genotypes; under high-P treatment, P_i showed the largest difference among genotypes, with up to 307 µg g^–1 lower [P] in high-PUE genotypes.

The concentrations of different P fractions had contrasting influences on photosynthetic capacity under low-P conditions (Fig. 6). For example, although total [P] was positively correlated with A_sat (P = 0.018), only the P_i, P_M and P_R fractions were also positively correlated (P = 0.038, P < 0.001 and P < 0.001, respectively; Fig. 6). Importantly, P_L and P_N showed no correlation with A_sat, and thus reductions in these fractions did not appear to significantly reduce photosynthetic capacity (P = 0.155 and P = 0.155, respectively). In contrast, under high-P treatment, changes in the P fraction and total P concentrations did not influence photosynthetic capacity, with no correlations between them and A_sat (P > 0.05; Supplementary data Fig. S5).

Fig. 6.

Light-saturated net photosynthetic rate (A_sat) of contrasting rice genotypes in relation to the phosphorus (P) concentration ([P]) of each fraction and the total [P], under low-P treatment. Each line represents the line of best fit, derived from a linear model. Only significant relationships are shown (P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR). β, regression coefficient.

As expected, changes in P fraction concentrations were strongly related to total [P], with some P fractions explaining more of the variation in total [P] than others (Supplementary data Table S4). For example, P_L and P_N correlated most strongly with total [P] under low-P treatment (R² = 0.87 and 0.88, respectively), thus explaining most of the variation in total leaf [P]. Under high-P treatment, P_i and P_N correlated most strongly with total [P] (R² = 0.86 and 0.8, respectively). Therefore, reductions in these fractions explain most of the observed reductions in total [P].

Photosynthetic capacity

All genotypes maintained a reasonable photosynthetic capacity, with mean A_sat rates always >21 µmol m^–2 s^–1 (Table 2). Photosynthetic capacity was slightly higher under high-P treatment compared with low-P treatment. There was only a small range in A_sat rates under low-P treatment (21–25 µmol m^–2 s^–1), with the low-PUE genotype TA fastest. In contrast, under high-P treatment, the high-PUE genotypes DJ and MU were fastest (29 and 28 µmol m^–2 s^–1, respectively).

Table 2.

Light-saturated net photosynthetic rates (A_sat) of the youngest fully expanded leaf (Leaf 1) of contrasting rice genotypes grown under low-P and high-P treatment

	Asat (µmol CO2 m–2 s–1)
Genotype	Low P	High P
DJ	21.3 (0.8)a	28.8 (0.3)b*
MU	22.7 (1)ab	28.3 (0.7)ab*
YO	22.5 (0.7)ab	26.0 (0.5)a
TA	25.5 (0.9)b	26.7 (0.7)ab
IR	23.5 (0.4)ab	26.1 (1.1)ab

Values are means (s.e.), n = 6.

Different letters indicate significant differences among genotypes within each treatment, and an asterisk indicates significant difference between treatments within each genotype (post-hoc Tukey test, P < 0.05). High-PUE genotypes are DJ, MU and YO, and low-PUE genotypes are TA and IR.

Biomass and allocation of resources to the youngest leaves

High-PUE genotypes always produced more biomass compared with low-PUE genotypes (Fig. 7A). MU had the greatest biomass, followed by YO and DJ and then IR and TA. DJ and MU showed the highest root:shoot ratios, under both treatments, with YO, IR and TA all substantially lower (Fig. 7B; Supplementary data Fig. S6). DJ and MU produced more root biomass compared with low-PUE genotypes (Supplementary data Fig S6).

Fig. 7.

(A) Total biomass, (B) root:shoot dry weight (d. wt) ratio, percentagr of mature leaf (ML) (C) d. wt, (D) phosphorus (P) content and (E) area allocated to the two youngest fully expanded leaves (L1 and L2) (see the Materials and Methods) and (F) total leaf area. Bar heights show means, error bars show standard errors (n = 5–6) and different letters indicate significant differences among genotypes within each treatment (post-hoc Tukey test, P < 0.05). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

The proportion of resources allocated to younger leaves, estimated as the percentage of mature leaf biomass, P content and area allocated to the two youngest fully matured leaves, increased under low-P treatment and was constitutively highest in genotypes DJ and MU (Fig. 7C–E). Total leaf area was similar among genotypes under low-P treatment, with DJ having a slightly higher leaf area (Fig. 7F).

Principal component analysis

Principal component analysis of key P-related traits under low-P treatment revealed a clear clustering of genotypes into three groups (Fig. 8). First, low-PUE genotypes were separated from high-PUE genotypes along the horizontal axes (PC 1). This was primarily due to differences in PUE_P, PUE_B, biomass (d. wt) and organ [P] (Fig. 8; Supplementary data Fig. S7A). Low-PUE genotypes had lower PUE_P, PUE_B and biomass, and higher organ [P], while high-PUE genotypes were the opposite (Fig. 8). High-PUE genotypes, distributed to the left, were further separated along the vertical axes (PC 2), into two clusters. The high PPUE genotype DJ clustered tightly in the upper left, while the other two genotypes MU and YO were clustered in the lower left. This separation along PC 2 was primarily associated with differences in P fraction allocation, mostly the percentage of total P allocated to P_L, P_R and P_N fractions; combined, they contributed to >46 % of the variation in PC 2 (Fig. 8; Supplementary data Fig. S7C). The high-PPUE genotype DJ was strongly associated with reduced allocation to P_L, and a greater allocation to P_N and P_R (Fig. 8). Therefore, under low-P treatment there was a separation of genotypes into three groups: low PUE (TA and IR), high PUE (MU and YO) and high PUE/PPUE (DJ). Under high-P treatment, these groups showed less clear separation (Supplementary data Fig. S8).

Fig. 8.

Principal component analysis (PCA) biplot of key phosphorus (P)-related traits under low-P treatment. These traits included three measures of P-use efficiency (PPUE, photosynthetic PUE; PUE_B, biomass PUE; PUE_P, physiological PUE; see Table 1), organ-level total P concentrations ([L1], youngest fully expanded leaf; [L2], second youngest fully expanded leaf; [OL], older leaves; [ST], stem; [RO], roots), P fraction allocations (PN%, nucleic acid-P; PI%, inorganic-P; PL%, lipid-P; PM%, metabolite-P; PR%, residual-P), total plant dry weight (d. wt), root:shoot biomass ratio (R:S), leaf area (LA), light-saturated photosynthetic rates (A_sat) and d. wt allocated to the two youngest fully expanded leaves (DW.YL%). High-PUE genotypes are shown in blue (DJ, MU and YO) and low-PUE genotypes are shown in orange (TA and IR).

DISCUSSION

Low investment in phospholipids was associated with high phosphorus-use efficiency

A low investment in phospholipids was strongly associated with a high PPUE under low-P conditions. This was evident across all genotypes, with a strong negative correlation between P_L investment and PPUE (Fig. 5). This was particularly evident in the two most P-efficient genotypes, DJ and MU, which had the lowest investment in and concentration of P_L (Fig. 3).

Why did reduced investment in P_L increase PPUE? First, P_L was the third largest P fraction (approx. 13 %) and explained a large portion of the variation in total [P] (R² = 0.87); thus a low investment in P_L notably reduced total [P] (Supplementary data Table S4). Second, P_L was not related to A_sat, meaning that lowering P_L did not slow photosynthetic rates, unlike most other P fractions, including P_M and P_i. Third, phospholipids (which make up the P_L fraction) can be replaced with glycolipids that do not contain P (Tjellström et al., 2008), unlike nucleic acids which have no non-P substitutes (Raven, 2013). In summary, reduced investment in P_L substantially reduced total leaf [P], without impacting photosynthetic rates, thus improving PPUE. This provides a clear mechanistic understanding of how reduced P_L improved PPUE.

Lipid remodelling probably plays an important role in reducing P_L concentrations. Lipid remodelling involves the degradation and replacement of phospholipids with non-P-containing glycolipids (e.g. sulfolipids or galactolipids), making more P available for other functions and in other parts of the plant (Härtel et al., 2000; Dörmann and Benning, 2002; Frentzen, 2004; Tjellström et al., 2008; Lambers et al., 2012; Pant et al., 2015). This plays an important role in improving low-P tolerance among rice varieties (Jeong et al., 2017; Tawaraya et al., 2018; Adem et al., 2020; Watanabe et al., 2020). However, it was previously unknown if rapid photosynthetic rates could be maintained with reduced levels of phospholipids.

The majority of lipid membranes in the chloroplast do not contain P (i.e. monogalactosyl diacylglycerol, digalactosyl diacylglycerol and sulfoquinovosyl diacylglycerol), and those that do (phosphatidylglycerol) can at least partially be replaced despite being essential in the transport of electrons in photosystem II, and the development of chloroplasts (Dörmann and Benning, 2002; Vance et al., 2003; Frentzen, 2004; Wada and Murata, 2007). We show here, for the first time, that photosynthetic rates were unaffected by phospholipid concentration in rice leaves, meaning that reduced P_L concentrations under low P conditions did not reduce photosynthetic capacity. Similarly, Lambers et al. (2012) showed that in six P-efficient Proteaceae species, replacement of phospholipids with lipids that do not contain P did not compromise photosynthetic capacity (Lambers et al., 2012). We surmise that this is because most phospholipids are found outside of the chloroplast and can be safely remodelled without impacting photosynthesis (Tjellström et al., 2008). It is, however, theoretically possible for photosynthesis to be impacted if the chloroplast phospholipid concentration was sufficiently reduced. Further research is needed to explore other theoretical impacts of reduced phospholipid concentrations under P stress, for example on chilling tolerance.

Phospholipids may be important in early leaf development, and so the rapid replacement of lipids from phospholipid-rich younger leaves to glycolipid-rich older leaves may be an important process in improving P efficiency (Cowan, 2006). A recent study by Adem et al. (2020) highlights the potential importance of a faster response to P starvation, with earlier remodelling of phospholipids from younger leaves.

Reducing investment in other P fractions may not be as effective as reducing P_L. P_N primarily consists of RNA, for which there is no non-P-containing substitute (Raven, 2013). Therefore, although P_N was the largest P fraction in our study and showed no influence on photosynthetic capacity, it is unlikely to be reduced without impacting leaf functioning, i.e. protein synthesis and turnover. This is an effective strategy for some P-efficient slow-growing species (Sulpice et al., 2014), but may be detrimental in faster growing crop plants that require abundant P_N-containing ribosomes to maintain protein turnover and to respond rapidly to changes in environmental conditions (Flis et al., 2016). It is noted that there is potential for improvement of protein synthesis per unit RNA, but this might be a complicated strategy (Veneklaas et al., 2012; Raven, 2013; Sulpice et al., 2014). It is interesting to note that increased investment in P_N was associated with improved PPUE under low P treatment (Fig. 5) and, although this may be largely explained by reductions in other P fractions, it may be worth further investigation to determine the benefits of maintaining investment in P_N.

Inorganic-P is the second largest P fraction and is where most P accumulates when in excess; however, under low P supply, most plants rapidly reduce their P_i pool, and further reductions may negatively impact photosynthetic capacity (Fig. 6) (Veneklaas et al., 2012). Both P_M and P_R are relatively small fractions, and both appear tightly linked with photosynthesis; therefore, reductions in these fractions may also reduce photosynthesis. In contrast, P_L is a substantial P fraction that can be reduced, without reducing A_sat, and can be replaced with non-P-containing substitutes. Therefore, reducing P_L investment is a promising way forward to increase leaf-level PUE, whilst maintaining rapid photosynthesis and growth (Lambers et al., 2012; Veneklaas et al., 2012; Hidaka and Kitayama, 2013).

The ability to manage excess P may be important in maintaining efficiency under high-P conditions. Under these conditions, the most P-efficient genotype, MU, appeared to reduce P investment in P_i and increase that in P_L, relative to low-PUE genotypes (Fig. 3A). Excess Pi may interfere with cellular processes, causing P toxicity (Shane et al., 2004; Hayes et al., 2019; Takagi et al., 2020). Therefore, we speculate that under high-P conditions, the ability to reduce P_i concentrations may increase efficiency by reducing the negative effects of excess P (Takagi et al., 2020).

Unsurprisingly, roots showed a very different P fraction allocation compared with leaves. This reflects differences in functionality between these organs, with photosynthetic leaves being photosynthetically active and roots being involved in water and nutrient uptake. Roots had a much lower proportion of P_i and P_M, even under high-P treatment, reflecting their lower level of metabolic activity and lower capacity for storage, compared with leaves. This difference in P fraction allocation should be considered in future efforts to increase root PUE. However, given that P_L investment is maintained at >19 % in roots, there may also be an opportunity to improve root PUE through promotion of increased lipid remodelling, so long as the root phospholipids can be replaced by glycolipids without impacting functioning.

All low-P plants reduced relative investment in inorganic- and lipid-P, but not nucleic acid-P

As expected, all genotypes reduced their investment in P_i and P_L under low-P conditions (Lambers et al., 2012; Veneklaas et al., 2012; Jeong et al., 2017; Tawaraya et al., 2018). However, it was surprising to find that investment in the largest P fraction, P_N, actually increased under low-P conditions, in contrast to the strategy of many P-efficient native species (Sulpice et al., 2014). P_i and P_L showed the greatest relative reductions under low P, with P_i decreasing from 30 to 18 % of total P, and P_L from 19 to 13 %. The relative investment in P_N increased from 36 to 50 %, but this was because the P_N concentration did not decrease to as low a concentration as other fractions. The P_N concentration remained at approx. 0.4 mg g^–1 under low P, compared with a concentration of approx. 0.1 mg g^–1 in most other P fractions.

An unexpected and novel discovery was that the relative investment in P_N increased in P-stressed plants, remaining higher than the other P fractions. P_N consists primarily of RNA (85 %), with the remainder being DNA (Bieleski, 1968; Veneklaas et al., 2012). The majority of this RNA is rRNA, involved in protein synthesis and growth (Kanda et al., 1994). Therefore, reductions in P_N and consequently rRNA would be likely to reduce metabolic activity, via reduced synthesis of enzymes involved in photosynthesis and protein turnover (e.g. Rubisco and Calvin–Benson cycle enzymes), unless compensated for by increased enzyme substrates (Sulpice et al., 2014; Raven, 2015). We propose that in the case of rice, there is a minimum amount of rRNA, and consequently of P_N required to maintain metabolic activity. The P_N concentration under low-P conditions is similar to that reported by Jeong et al. (2017) in flag leaves under adequate P supply, approx, 0.4 mg g^–1. Other studies report P_N allocations of around 30–40 % in native plants (Hidaka and Kitayama, 2011, 2013; Zhang et al., 2018; Yan et al., 2019) and approx, 30–35 % in rice flag leaves (Jeong et al., 2017). Therefore, the concentration reported here was similar to that reported elsewhere, but the allocation (approx. 50 %) was notably greater. This is likely to reflect the extreme and sustained P stress these plants were under, reducing all non-essential P fractions to their lowest possible levels whilst maintaining sufficient P_N investment. We surmise that P_N investment is maintained above other P fractions in rice leaves under P stress to sustain metabolic activity and protein turnover, and therefore rapid photosynthesis, growth rates and responsiveness to environmental changes. Given this significant investment in P_N, it would be prudent to further investigate the efficiency with which this fraction is recycled from older leaves during development and senescence (Bassham and MacIntosh, 2017). A tight recycling of this significant P fraction may benefit whole-plant P efficiency.

In contrast to P_N, the P_i fraction dramatically declined under low-P conditions (from 30 to 18 %). P_i can be separated into two physiologically distinct pools. The vacuolar P_i pool is found in the cell vacuole and is where excess P_i is stored; it acts as a buffer for the P_i demands of metabolically active cytosolic P_i (Veneklaas et al., 2012). The cytosolic P_i is tightly regulated and maintained within a narrow range, whereas the vacuolar P_i is highly variable and largely dependent on external P supply (Lauer et al., 1989; Mimura et al., 1990). It is expected that the large decline in inorganic P is explained by a reduced vacuolar P_i concentration, whilst the metabolically active cytosolic P_i concentration would be maintained within a narrow range, thus explaining the relatively fast rates of photosynthesis, even under low-P conditions (mean of 23 µmol m^–2 s^–1). Further research into vacuolar and cytosolic P_i dynamics may reveal possible pathways to increase efficiency, by reducing the vacuolar P_i to its lowest possible level, without impacting net growth. This may include studying diurnal fluctuations in these fractions to improve net PUE over time.

Other traits contributing to increased whole-plant PUE

Whole-plant PUE was strongly associated with reduced total [P], increased biomass and increased preferential allocation of resources to the two youngest fully expanded leaves. The genotypes DJ, MU and YO all showed high whole-plant PUE under low-P conditions. Given the same total amount of P, these three genotypes produced more biomass, and achieved similar photosynthetic rates, despite lower leaf P concentrations. These genotypes also tended to allocate more resources (biomass, P and area) to the two youngest fully expanded leaves. This was strongly associated with a reduced number of leaves in high-PUE genotypes, meaning that the two youngest leaves accounted for a greater proportion of the total mature leaf biomass, P and area (data not shown). The youngest fully expanded leaves are generally photosynthetically the most active, and so investing limited resources into these organs, whether by reduced leaf number or increased allocation, can improve overall plant efficiency (Watanabe et al., 2020).

CONCLUSIONS

This study shows for the first time that a low investment in P_L is associated with a high leaf-level efficiency (PPUE) and that it does not lead to a lower photosynthetic capacity in rice. This provides a mechanistic understanding of how reduced P_L concentrations increase PPUE, thus presenting a novel strategy for improving PUE in crop plants. This is the first investigation showing how P fraction investments change under P stress in an important crop species, contributing to our fundamental understanding of P fraction allocation under low-P conditions (Fig. 9).

Fig. 9.

Schematic illustrating key findings of this study. The top box shows key responses in leaf phosphorus (P) fractions under low-P treatment. Pie charts show the percentage of total leaf P allocated to each P fraction, averaged across all genotypes. The five P fractions are nucleic acid-P (P_N, RNA and DNA), inorganic-P (P_i, soluble inorganic phosphate), lipid-P (P_L, phospholipids), metabolite-P (P_M, soluble P-containing small metabolites such as sugar phosphates and ATP) and residual-P (P_R, phosphoproteins and other recalcitrant complexes). The bottom box shows key traits associated with higher P-use efficiency (PUE) under low-P treatment. Reduced investment in lipid-P substantially reduces total leaf P concentration ([P]), without impacting photosynthetic rates, thus improving PUE. Increased allocation to the youngest leaves (L1/L2) and increased biomass are also associated with higher PUE genotypes. L1, youngest fully expanded leaf; L2, second youngest fully expanded leaf; OL, older leaves; ST, stem, immature leaves and senescing leaves; RO, roots.

Targeting P-efficient traits in the development of new varieties will benefit rice production. The genome sequence information for most of the studied genotypes is available, so genomic comparative studies to locate potential variation in quantitative trait loci (QTLs)/genes related to reprogramming lipid allocation and other traits could be achieved and used in breeding efforts to develop modern P-efficient varieties. This could have benefits for global food security by reducing fertilizer demand, decreasing food production costs and reducing negative off-site impacts of excess fertilizers. The next step will be to determine the potential for these traits to improve yields in low-P systems, through predictive modelling and subsequent field trials. Based on our findings, the genotypes DJ and MU showed higher whole-plant PUE and PPUE, making them a potentially valuable resource for improving these traits in other varieties under low-P conditions. The genotype DJ (DJ123) is of particular interest for development of varieties in low-P systems due to its efficient P acquisition traits, having more and longer root hairs (Nestler et al., 2016). In contrast, under higher P systems, only MU (Mudgo) showed both higher whole-plant PUE and PPUE, making it a suitable variety for selection of P-efficient traits in higher-P systems.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: schematic diagram showing the phosphorus fractionation assay. Figure S2: schematic diagram showing the soluble inorganic phosphate assay. Figure S3: phosphorus fraction allocation and concentration in the second youngest fully expanded leaf blades. Figure S4: changes in photosynthetic phosphorus-use efficiency with leaf phosphorus fraction allocations under high-P treatment. Figure S5: changes in light-saturated net photosynthetic rates with leaf phosphorus fraction concentrations and total P concentrations under high-P treatment. Figure S6: leaf one, leaf two, older leaves, stem and root dry weights for each genotype, and their proportion of total biomass. Figure S7: principal component analysis biplot showing the key phosphorus-related traits of five rice genotypes grown under high-P treatment. Figure S8: the percentage contribution of key phosphorus and morphophysiological traits to the first two principal components, derived from principal component analysis of five rice genotypes grown under either low- or high-P treatment. Table S1: details of genotypes used. Table S2: details of phosphorus treatments. Table S3: summary of results from internal reference material used during phosphorus fractionation assay. Table S4: model results for regression analysis between total leaf phosphorus concentration and P fraction concentrations. Method S1: phosphorus fractionation assay method. Method S2: soluble inorganic phosphate assay method.

ACKNOWLEDGEMENTS

We thank T. Matsuda and all the technical staff at the Japan International Research Center for Agricultural Sciences (JIRCAS) for their assistance. We thank Kosala Ranathunge and Li Yan for their feedback in method development, and Hans Lambers for his valuable feedback during writing. P.E.H. and G.D.A. planned and designed the experiment in consultation with J.P.T. and M.W.; P.E.H. and G.D.A. performed the glasshouse experiment; P.E.H. analysed and interpreted the data; P.E.H. wrote the paper with feedback from G.D.A., J.P.T. and M.W.

Funding

P.E.H. and G.D.A. were supported by the Japan Society for the Promotion of Science (JSPS) Postdoctoral Research Fellowship.