Are crops different from other plants in their P concentration?
Phosphorus use and photosynthesis
Crop development and canopy P distribution
Internal redistribution of P in a growing vegetative plant
Allocation of P to reproductive structures
Constraints to P remobilisation
Do physiological or phylogenetic trade-offs constrain traits that could improve PUE?
Identifying genetic loci associated with PUE
Limitation of grain crop productivity by phosphorus (P) is widespread and will probably increase in the future. Enhanced P efficiency can be achieved by improved uptake of phosphate from soil (P-acquisition efficiency) and by improved productivity per unit P taken up (P-use efficiency). This review focuses on improved P-use efficiency, which can be achieved by plants that have overall lower P concentrations, and by optimal distribution and redistribution of P in the plant allowing maximum growth and biomass allocation to harvestable plant parts. Significant decreases in plant P pools may be possible, for example, through reductions of superfluous ribosomal RNA and replacement of phospholipids by sulfolipids and galactolipids. Improvements in P distribution within the plant may be possible by increased remobilization from tissues that no longer need it (e.g. senescing leaves) and reduced partitioning of P to developing grains. Such changes would prolong and enhance the productive use of P in photosynthesis and have nutritional and environmental benefits. Research considering physiological, metabolic, molecular biological, genetic and phylogenetic aspects of P-use efficiency is urgently needed to allow significant progress to be made in our understanding of this complex trait.
There is an increasing awareness that there are limits to global rock phosphate reserves, and that increasing the efficiency with which these reserves are used to produce crops is vital to maintain current agricultural productivity, or to even increase it (Cordell et al., 2009). Annual grain crops (cereals, oil seeds and pulses), which are the focus of this review, provide 58% of the dietary energy for the world’s growing population (FAO, 2010). Improving the efficiency of phosphorus (P) fertilizer use for crop growth requires enhanced P acquisition by plants from the soil (P-acquisition efficiency) and enhanced use of P in processes that lead to faster growth and greater allocation of biomass to the harvestable parts (internal P-use efficiency (PUE), further defined below in Section II). As only 15–30% of applied fertilizer P is taken up by crops in the year of its application (Syers et al., 2008), potentially large gains in efficiency can be made by improving P acquisition. This aspect of P efficiency has received significant attention and has been reviewed recently (White & Hammond, 2008; Ramaekers et al., 2010; Wang et al., 2010; Richardson et al., 2011), identifying promising opportunities for improved crop traits and agronomic measures. By contrast, much less attention has been paid to physiological PUE, and comparative studies of PUE have often suffered from confounding effects of variation in P-acquisition efficiency (Rose et al., 2011). Aiming at increasing P uptake alone will benefit yields, but will also increase total amounts of P exported from the field. Increased P exports can cause considerable off-site environmental problems (Tiessen, 2008; Childers et al., 2011), and will need to be replaced with additional fertilizer to avoid soil P depletion in the long term. The most sustainable and productive agricultural systems will be those where P exports are balanced by P inputs, and which have high yields per unit P taken up, that is, they have high PUE. In this review we identify plant traits, from the whole plant to the biochemical level, that contribute to efficient use of internal P after it has been acquired. We also assess how efficient crop plants are in a wider context, and evaluate options for improvement.
II. P-use efficiency and P dynamics in a growing crop
PUE is the amount of total biomass, or yield, that is produced per unit of P taken up (Hammond et al., 2009), distinguished when relevant by subscripts, PUEt and PUEy, respectively. In the case of biomass, measurements are often restricted to aboveground plant parts. In grain crops, PUEy is the grain yield per unit of maximum aboveground plant P. In vegetative plants, the productive use of a unit of P taken up is determined by (a) the efficiency with which it is used in metabolism and growth, and (b) the duration of its presence in living parts of the plant where it contributes to these processes. Following Berendse & Aerts (1987), who formalized this concept,
where P residence time represents the time that a unit of P remains in living parts of the plant. Even from this simplified equation, which is most useful for steady-state vegetative growth, it is clear that PUE encompasses a wide range of physiological, structural and developmental traits as they determine tissue-level use of P and allocation and reallocation of P among plant parts with different functions and efficiencies. It is therefore useful to first examine how P requirements change during the different developmental stages of plants. Key aspects of PUE will then be discussed in detail in the following sections.
Once seed P reserves are depleted, continued growth depends upon P uptake by roots from the soil. This point in development typically coincides with exponential growth in crop plants when availability of soil P is likely to be highest, because fertilizers are freshly applied and there is limited resource competition between roots. Inevitably, crop relative growth rate (RGR) declines as a result of self-shading, increasing respiratory costs, and senescence of older plant tissues and organs, even if soil water and nutrients are not depleted. At this stage, P uptake by roots continues, but remobilization of P from senescing tissues can become a very significant internal source of P. At the time of reproductive growth, remobilization of P from senescing vegetative tissues is typically the main source of P for sink tissues, as P uptake by roots decreases, as a result of declining root growth and depletion of available soil P. In crops, a very large fraction of the P present in vegetative parts is remobilized to the grain, and soil P availability at this stage has a relatively small effect on grain yield (Römer & Schilling, 1986; Rose et al., 2008).
The whole-plant changes in P economy during crop development highlight the fact that, besides an efficient use of P in processes that lead to accumulation of biomass, optimal allocation and efficient reallocation of P will also contribute to high PUE. In this review we explore opportunities for improvement of PUE by analysing the size and identity of P pools in plant tissues and their functional importance, as well as their dynamics, including re-use elsewhere in the plant.
III. P pools in plants
Phosphorus in plants exists either as the free inorganic orthophosphate form (Pi) or as organic phosphate esters. The Pi concentration of tissues generally reflects the Pi supply (White & Hammond, 2008). This is true in crop plants, such as Brassica napus and Cucurbita maxima (Pant et al., 2008), pasture plants, such as Medicago truncatula (Branscheid et al., 2010), and undomesticated species, such as Hakea prostrata (Shane et al., 2004). Tissue Pi is separated into two physiologically separate pools. The metabolically active Pi pool, of the order of 0.1–0.8 mg P g−1 dry weight (DW) (Bieleski, 1968; Rébeilléet al., 1983; Rouached et al., 2011), is located in the cytoplasm and is kept within fairly narrow limits (Bieleski, 1968; Mimura et al., 1996; Mimura, 1999). Cellular Pi in excess of current cytoplasmic need is stored in the vacuole and is used to buffer the Pi demands of the cytoplasm, and is accordingly the most variable P fraction. For example, barley (Hordeum vulgare) grown in the absence of added Pi had no detectable vacuolar P, as measured using 31P-NMR (Foyer & Spencer, 1986). As Pi was withheld from barley, pre-grown in nutrient solution containing 1.0 mM Pi, the vacuolar Pi concentration decreased sharply, while the cytosolic Pi concentration was less affected. Using protoplasts and vacuoles isolated from intact barley leaves, it was shown that the vacuolar Pi concentration declined from 145 to 5 mM, when P supply declined from 40 to 0 mM, whereas that of the cytoplasm (cytosol plus mitochondria and plastids) declined from 35 to 26 mM (Mimura et al., 1990). Similarly, for soybean (Glycine max) having just reached full-flowering development, the vacuolar Pi concentration decreased from 8.0 to < 0.05 mM as P supply declined, whereas the cytosolic Pi concentration declined considerably less markedly, from 8.3 to 0.23 mM; the hexose monophosphate concentration declined less than that of Pi, from 7.7 to 0.54 mM (Lauer et al., 1989). Pratt et al. (2009) reported that Pi concentrations in the cytosol of sycamore (Acer pseudoplatanus) and Arabidopsis suspension cells were much lower than those in the cytoplasm. In summary, there is significant compartmentalization of Pi within cells, and Pi concentrations are most variable in the vacuole, showing its buffering function. Differences in P uptake by plants result in large variability of Pi concentration, making it a sensitive indicator of plant P status (Bollons & Barraclough, 1999).
The main pools for esterified P are nucleic acids, phospholipids, phosphorylated water-soluble low relative molecular mass (Mr) metabolites (commonly referred to as P-esters) and phosphorylated proteins. Phosphorylation of proteins serves a regulatory function, so phosphorylated proteins will not be considered further. In the following sections, opportunities for increased P efficiencies in the most significant P pools will be explored.
Fig. 1 summarizes available data on inorganic and organic P fractions in photosynthetic tissue of a range of species. Observations of total P concentrations ([P]) > 5 mg g−1 DW are generally from plants growing at very high P supply. Optimal concentrations for most crops are < 4 mg P g−1 DW (McLachlan et al., 1987; Föhse et al., 1988; Rodríguez et al., 2000). For wheat (Triticum aestivum), Bollons & Barraclough (1999) reported critical concentrations for total P and Pi of approx. 3 and 0.6 mg g−1 DW, respectively. In order of size, and ignoring stored (vacuolar) Pi, P pools are usually RNA > P-lipid > P-ester > DNA > metabolically active Pi. Averaged for all cases where total [P] < 4 mg g−1 DW (n =21 species), and taking P-ester = 1, there are 1.3 units of lipid P, 2.0 units of nucleic acid P (mainly RNA) and 1.4 units of inorganic P.
1. Nucleic acids – role and minimum requirements
The nucleic acid pool is typically the largest organic P pool in a plant (Fig. 1), generally containing 40–60% of the P found in the combined organic P pool (i.e. total P minus Pi). This equates to c. 0.3–2 mg P g−1 DW, depending on the species, tissue and P supply (Smillie & Krotkov, 1960; Bieleski, 1968; Tachibana, 1987). The nucleic acid pool generally contains at least 85% RNA, with the remainder being DNA (Bieleski, 1968; Tachibana, 1987). Most RNA is ribosomal RNA (rRNA): Cucurbita ficifolia seedling roots were found to contain c. 94% rRNA, 4% transfer RNA and 2% messenger RNA (Kanda et al., 1994).
There is generally a positive correlation between rRNA content, and therefore ribosome number and protein synthesis capacity, and growth rate over a range of taxa (Elser et al., 2000, 2010). As leaves develop, rRNA levels increase to accommodate the rapid synthesis of proteins, for example, Rubisco (Suzuki et al., 2010), that are needed for photosynthesis. At full expansion or shortly thereafter, net leaf protein synthesis ceases and RNA levels decrease by up to 75% (Hensel et al., 1993; Suzuki et al., 2010). It is not clear how much of the leaf RNA remains at the end of senescence. The RNA pool generally adjusts in accordance with the growing conditions, for example, RNA concentrations increased in trees in which N concentration ([N]), [P] and growth rate were enhanced through fertilization (Reef et al., 2010).
Opportunities for increasing the use efficiency of RNA and the P it contains include optimizing the protein synthesis system during plant development. The aim would be to deploy many ribosomes, fully engaged in protein synthesis, when growth needs to be rapid, but only enough ribosomes for maintenance processes when growth is complete, and no ribosomes upon senescent death. Presently, a fraction of the 40 S and 60 S ribosomal subunits that are present at any given time is not actively synthesizing protein (Mustroph et al., 2009). The role, and thus the importance, of these subunits need to be evaluated so that the pool size could be targeted for optimization. Moreover, Rubisco large and small subunit mRNA levels can decline by > 90% in fully expanded leaves, but cytosolic rRNA levels drop < 75% during this period (Hensel et al., 1993; Suzuki et al., 2010). This suggests an increased disjunction between the actual amount of protein synthesis that takes place and the protein synthetic capacity that is available. However, ribosomes and tRNA are still needed in mature leaves for protein turnover, including replacement of damaged protein, and for acclimation and programmed cell death during senescence.
The costs of P, as RNA, of protein turnover and repair are substantial (Raven, 2011, 2012). Quigg & Beardall (2003) compiled data from vascular land plants showing protein turnover at 0.04–0.42 d−1. For a relative growth rate of a crop plant of 0.2 g g−1 d−1 this would mean an increase in RNA content relative to what would be needed with no protein turnover of 20–210% in an optimally allocating plant. With RNA representing 47% of the organic P (Fig. 1, total [P] < 4 mg g−1), this would mean that an increase of 9–99% in the organic P content of the plants is needed to support protein turnover. Assuming that the protein turnover estimates are accurate (the turnover rates are not easy to measure), protein turnover would seem to be a target for saving on P. For the specific cases of photodamage to the D1 protein of photosystem II and O2 damage to nitrogenase, it is, respectively, known and probable that these proteins turn over significantly more rapidly than most other proteins (Raven, 2011, 2012). D1 protein is ubiquitous among oxygenic organisms, and the relatively small additional P cost for ribosomes used in repairing photodamage in an optimally allocating organism is already included in PUE estimates (Raven, 2011, 2012). The greater P costs of growth often found for symbiotic and free-living diazotrophically growing photosynthetic organisms than when they are growing on combined nitrogen (Vitousek et al., 2002) might be explained, at least in part, by the need for more ribosomes for synthesizing nitrogenase to replace O2-damaged nitrogenase (Raven, 2011, 2012). This possibility needs further investigation.
The amount of P in the DNA pool is considerably smaller than that in the RNA pool, and reductions of DNA content would seem difficult and undesirable. However, there may be useful long-term optimization strategies. A pivotal question here is whether genome size and gene copy number can be optimized. The answer will require a better understanding of the function of the so-called ‘noncoding’ regions within genomes and of the ways that genes are controlled, as well as the development of new technologies to change genome structure rationally. The common occurrence of polyploidy in agricultural and invasive species (Soltis & Soltis, 2000), suggesting competitive benefits in productive and disturbed environments, also requires further research. A high copy number of rRNA genes (a few hundred to several thousand rRNA genes per genome; Rogers & Bendich, 1987) is thought to help the plant meet the demands to produce large amounts of rRNA to support protein synthesis and growth (Elser et al., 2010). However, the actual number of rRNA genes shows considerable variation between species, between individuals of the same species and even within individuals of some species (Rogers & Bendich, 1987). The question of genome size and gene copy number applies to both nuclear and organellar genomes. As an example, there can be > 10 000 copies of chloroplast DNA per cell, depending on species, tissue, developmental stage and environmental conditions (Bendich, 1987). If such high copy numbers reflect the demands for high levels of certain transcripts, such as plastid rRNA or the transcripts for the large subunit of Rubisco (Bendich, 1987), this is in stark contrast with the low gene copy number for the small subunit of Rubisco in the nuclear genome. In future, it may be possible to engineer plant genomes in all three DNA-containing compartments to optimize gene copy numbers for minimal P content while maintaining gene function.
2. Other P pools – roles and requirements
Phospholipids are important in membranes. While it may not be feasible to reduce the total area of membranes, there may be possible efficiency gains in their synthesis and composition. Thomas & Sadras (2001) characterized the assembly process of thylakoid membrane complexes as having large excess production and subsequent breakdown of components which is energetically inefficient but allows high responsiveness to changing internal and external conditions. If the implied trade-off between efficiency and responsiveness exists, tighter regulation of biosynthesis may be a trait worth investigating for increased resource efficiency in predictable productive environments.
Sulfolipids and galactolipids, rather than phospholipids, are the major lipids in the thylakoid membrane, and to a lesser extent in the plastid envelope membranes. Phospholipids in other membranes can also be replaced by sulfolipids and/or galactolipids; either constitutively or in response to P deficiency. Data (see Supporting Information Notes S1), particularly on cyanobacteria and algae, show that this replacement can be almost complete (Van Mooy et al., 2006, 2009), and that replacement can also occur as a result of mutations. However, the information on the functional consequences of such replacement is limited to the tentative conclusion that there is no effect on proton permeability (Notes S1, Table S1), but that there may be increased membrane leakage of electrolytes which constrains chill tolerance (for Arabidopsis, see Hurry et al., 2000).
Tissue P concentration may also be decreased by lowering the content of low-Mr water-soluble Pi-esters, and the nonstorage, metabolically active Pi. Notes S2 discusses three ways by which this could be achieved for the photosynthetic CO2-assimilation pathway involving Rubisco and the photosynthetic carbon reduction cycle (Rubisco–PCRC). The conclusion is that even the most radical of the three proposed interventions, that is, replacing the Rubisco–PCRC with some other pathway involving fewer phosphorylated intermediates, is unlikely to decrease the content of the low-Mr water-soluble Pi-esters by > 10%, with an overall saving based on the allocations in Fig. 1 of 1.2%. It is important to note that phloem is cytosolic and that any decrease in Pi-esters and Pi in its metabolically active cytosol, plastid stroma, mitochondrial matrix and nucleoplasm might have impacts on P (and N) transport in phloem which is particularly relevant for remobilization from senescing tissues. Notes S3 and Table S2 deal with the phloem in more detail, showing that it is unlikely that P translocation out of senescing (or any) tissues is limited by the availability of solutes to generate turgor for phloem transport or other constraints on the composition of phloem sap.
IV. Phosphorus pools and growth rates
When plants are subjected to external P deprivation, shoot growth is often rapidly inhibited, well before plant P status is reduced. For example, in Spirodela oligorrhiza (Bieleski, 1968) shoot growth declines within hours upon a shift to a medium lacking P, while shoot Pi concentration is not greatly affected during this time. Similarly, barley leaf growth is severely inhibited when the plants are subjected to Pi deprivation, although leaf Pi concentrations remain adequate (Mimura et al., 1996). Shoot growth also declines in Arabidopsis plants grown with reduced Pi supply, despite adequate concentrations of Pi in the leaf vacuoles (Rouached et al., 2011). These observations indicate that plants start to economize on Pi, by reducing shoot growth, well before the leaf vacuolar storage pool of Pi is depleted. This was explored by Wareing’s group using a range of species. They showed that a limiting supply of P caused a decline in export of cytokinins from roots and in cytokinin concentrations in roots and leaves (El-D et al., 1979; Horgan & Wareing, 1980). Using Plantago major, Kuiper et al. (1989) showed that the decline in shoot growth upon a decrease in nutrient supply could be restored by supplying kinetin to the roots; this increased the endogenous concentration of cytokinins. The effect of kinetin on shoot growth was transient, as expected, but it does illustrate that roots sense that nutrients (N or P) are limiting well before leaves experience deficiency symptoms and that shoot growth is regulated in a feed-forward manner (Lambers et al., 2008).
As reduced shoot growth is a major concern for agricultural systems, it would be desirable to maintain maximal shoot growth as long as possible under P-limiting conditions, that is, until vacuolar Pi contents are almost depleted, which would ultimately reduce the amount of fertilizer needed for these managed systems. The uncoupling of P limitation from its effect on growth is in fact possible, as demonstrated in a recent study (Rouached et al., 2011). The Arabidopsis PHOSPHATE1 (AtPHO1) gene has previously been shown to be necessary for Pi transport from root to shoot (Poirier et al., 1991; Hamburger et al., 2002). The shoots of Atpho1 mutants grown under Pi-sufficient conditions have all the typical symptoms of P deficiency, including severely reduced shoot growth and expression of Pi-starvation inducible (PSI) genes (Morcuende et al., 2007). By contrast, reduction of AtPHO1 expression by co-suppression, or expression of the rice (Oryza sativa) PHO1 orthologue (OsPHO1;2; Secco et al., 2010) in an Arabidopsis Atpho1 null mutant resulted in shoot growth, fresh weight, seed yield and lipid profiles under high Pi supply that were more similar to those of wild-type plants than to those of Atpho1 null mutants, despite the depleted shoot vacuolar Pi in the transgenics (Rouached et al., 2011). Moreover, the AtPHO1 suppressed and OsPHO1;2 expressor lines also largely lacked the Pi-starvation response typical of pho1 mutants (i.e. the expression of a large set of PSI gene transcripts). Notably, the same conclusion was reached when PHO1 under-expressors and wild-type plants with similar root-to-shoot Pi-transport rates were compared (Rouached et al., 2011). While the shoot fresh weight of the transgenics remained high with low vacuolar Pi content, the shoot fresh weight of the wild-type decreased to almost half, despite a much higher vacuolar Pi concentration than in the PHO1 under-expressor. Importantly, the size of the leaf cytoplasmic Pi pool was similar for all the plants. These data illustrate that the reduction of growth, changes in lipid profiles or changes in gene expression that are associated with the Pi-deprivation response are not a direct consequence of low shoot P status, but the result of signalling events that can be genetically controlled. This gives hope that uncoupling growth from Pi status will eventually also be feasible in crop species, and will potentially become an important approach to ensure biomass production at lower Pi input.
V. Are crops different from other plants in their P concentration?
Crops have high tissue concentrations of P and N, as expected for fast-growing plants (Poorter & Bergkotte, 1992). In a global compilation of leaf traits known to be associated with the fast growth–slow growth continuum in plants (Reich et al., 1999; Wright et al., 2004), crop species form a very nonrandom subset of plants as a whole (Fig. 2a). The high concentrations of both leaf N and P may result from one of two processes: (1) preferential selection of fast-growing species for our crops, or (2) artificial selection shifting trait values within crop species towards the fast end of the continuum. Of course these processes are not mutually exclusive. The first process is probably the dominant factor for N and P concentrations, as it is not uncommon to find lower concentrations in modern crops than in their wild progenitors (Evans, 1993).
In plants with higher concentrations of N and P, the ratio N : P is lower (Elser et al., 2010); however, crops have relatively high N : P compared with noncrops at this end of the spectrum (Fig. 2a), partly as a result of the predominance of legume crops, which tend to have high N concentrations. N : P ratios are related to the cellular content of (N-rich) protein relative to the content of (P-rich) RNA (Elser et al., 2010). Crops tend to have lower and less variable protein : RNA ratios (i.e. a greater investment in RNA relative to protein), compared with tree species (Fig. 2b). Woody species generally have much longer leaf life-spans and invest more in lignified tissue to resist mechanical damage. Crops have more similar protein : RNA ratios to algae and microbes than to woody species (e.g. Loladze & Elser (2011)). Protein : RNA and N : P ratios are sensitive to nutrient availability (e.g. Close & Beadle (2004)) and thus the observed ratio of c. 15 in crop plants (Fig. 2b) may reflect the nutrient-replete conditions generally found in agricultural systems, or is the signature of selection for growth in high-nutrient soils (or both). Sadras (2005) showed that N : P variation in crops is better explained by differences in P concentration than in N concentration, and suggested that closer regulation of [N] and more variable remobilization of P may contribute to this pattern. The variation in protein : RNA ratios in trees and even among crops (Fig. 2b) may indicate the potential for selection to reduce investment in RNA relative to protein in order to increase PUE.
VI. Phosphorus use and photosynthesis
Plant productivity relies on photosynthesis, and the photosynthetic process relies on P-containing compounds. Thus, an efficient use of P in photosynthesis is a potentially important determinant of crop PUE. Photosynthetic phosphorus-use efficiency (PPUEmax) is defined as the instantaneous light-saturated rate of leaf photosynthesis expressed per unit leaf P. Photosynthetic capacities (Amax, mass-based) of crop species are found at the high end of the leaf trait spectrum, associated with fast growth and low leaf mass per area (LMA) (Lambers & Poorter, 1992). As a result of similar increases of Amax and P concentrations with decreasing LMA, however, PPUE is not strongly correlated with LMA (Fig. 3), so fast-growing species, which tend to have low LMA, do not necessarily have a high PPUE. For example, thin leaves of barley have a similarly high PPUE to the sclerophylls of the genus Banksia (Lambers et al., 2011). Barley achieves this with a high [P], whereas Banksia has very low [P], but barley apparently uses P sparingly in structural tissue and very efficiently in photosynthetic machinery (Lambers et al., 2011). Interestingly, PPUE varies by an order of magnitude at any LMA, and part of this variation can be attributed to leaf N concentration (cf. Reich et al.(2009): mean PPUE is 59 nmol g−1 s−1 for leaves with N : P < 15, whereas PPUE is 129 nmol g−1 s−1 for leaves with N : P > 15 (n =74 and 138, Glopnet data set; Wright et al., 2004). In the Glopnet data set, high N : P is associated with low [P] rather than high [N] (cf. Fig. 2a), and low [P] is clearly not conducive to high rates of photosynthesis (Fig. 3b). Crops generally have low LMA but not particularly high N : P (12.3 for the 17 grain crops included in Fig. 2a), suggesting that a closer look at variation in PPUE among and within crop species may be worthwhile and may give insights into factors underlying high PPUE.
It is important to note that the above data refer to maximum photosynthetic rates, rather than actual photosynthesis of leaves during growth. In dense canopies, only a small proportion of leaves operate at full photosynthetic capacity, for part of the day. Nevertheless, it is fair to assume that mean daily rates of actual photosynthesis, and even night-time respiration, scale with photosynthetic capacity, as has been shown for several species in a number of environments (Zotz & Winter, 1993; Reich et al., 2009).
Highest PPUE tends to be achieved by leaves with high rates of photosynthesis, high P concentration and low LMA, but such traits are associated with short leaf life-spans (Wright et al., 2004), implying a trade-off that will reduce the C return per unit P over the life of a leaf (PPUEleaf,life). For an appraisal of PPUEleaf,life it is also important to consider that a large fraction of P in leaves is remobilized during senescence and re-used elsewhere in the plant (cf. Small, 1972; Aerts & Chapin, 2000). A simple model of PPUEleaf,life taking into account the mentioned factors is given by
where Amax is the maximum mass-based photosynthetic rate (nmol g−1 s−1), β converts Amax to a daily total C exchange rate, LL is the life-span of a leaf (d), CC is the carbon cost for leaf structure (g g−1, taking into account any C remobilization during senescence), rP is the fraction of P remobilized during senescence, and [P] is the concentration of P in the mature leaf (g g−1). In crops that have a high relative growth rate (growth rate per mass present), it is reasonable to assume that the carbon cost (CC) of a leaf represents a minor fraction of its lifetime carbon gain (Amax × β × LL) (Poorter et al., 1991). The extent of variation in β is poorly known, but the proportionality between photosynthetic capacity and daily net C assimilation (Zotz & Winter, 1993; Koyama & Kikuzawa, 2009; Reich et al., 2009) indicates that such variation is low among species in the same environment. Global leaf trait relationships (Wright et al., 2004) show that Amax is approximately inversely proportional to LL, which makes it likely, according to Eqn 2, that high PPUEleaf,life in nature is mostly achieved by leaves with low concentrations of P and high remobilization of P upon senescence. Such traits are to be expected in high-LMA leaves with long life-spans and low photosynthetic capacities. In conclusion, the near-proportionality of Amax and leaf [P] and the inverse proportionality of these traits with leaf life-span lead to the absence of general trends in C return per unit P along the leaf economics spectrum. This is in contrast to the instantaneous PPUEmax. It is important to note, however, that there is considerable variation between species in both indices, indicating that other leaf traits influence the efficiency of P use in photosynthesis, such as shown for leaf N. Zhu et al. (2010) argue that future increases in crop yield potential will come from increased photosynthetic capacity. It will be important to achieve this along with increased PPUE.
VII. Crop development and canopy P distribution
The general pattern of P accumulation is remarkably similar among monocarpic crops (Figs 4, 5a). Early growth of a seedling depends to a large extent on P reserves in the seed, that is, remobilization of stored P (Bolland & Baker, 1988; Thomson et al., 1991; White & Veneklaas, 2012). Duration of seed P reserve dependence during early growth varies with seed mass, seed [P], soil P availability and seedling P requirement, but is unlikely to last longer than a few weeks. In crop plants, P concentrations of seeds and seedlings are quite similar (e.g. 2.4–6.4 mg g−1 in seeds and 2.9–4.5 mg g−1 in young plants of eight crops including cereals, legumes and oil seeds (Schultz & French, 1978)), so growth beyond the seedling stage is soon dependent on P uptake from the soil. Maize (Zea mays) seedlings started taking up significant amounts of 32P from soil c. 5 d after sowing, but continued to use seed P reserves for c. 2 wk (Nadeem et al., 2011). While P derived from seeds may be a minor component of whole-plant P at maturity, it does have a lasting influence on plant development (Bolland & Baker, 1988), especially if early vigour provides better access to potentially growth-limiting resources such as soil moisture, nutrients and light, in the absolute sense or relative to competitors.
As crop canopies develop and leaf area indices increase, photosynthetic rates of the lower shaded leaves decrease. Plants redistribute N from older shaded leaves to younger well-lit leaves in a pattern that approaches an optimal distribution of total canopy N (Field, 1983; Hirose & Werger, 1987; Franklin & Ågren, 2002). Redistribution of P for optimal plant C gain has been little studied, but the same principles are expected to apply as for N. Adequate concentrations of P are important to maintain high rates of photosynthesis (Rychter & Rao, 2005), and rates of P remobilization from senescing leaves are at least as high as those for N (Aerts, 1996). With near-optimal nutrition, shoot [N] and [P] decrease in a similar way as plants grow (Greenwood et al., 2008), suggesting similar patterns in uptake and redistribution of N and P under such conditions; however, when one of these elements is limiting, shifts in the N : P ratio may occur as a result of reduced uptake of N or P which cannot be fully compensated by internal redistribution. The N : P ratio of shoots of six cereal and oilseed crops changed only slightly from approx. 11 to 9 between the early vegetative and mid-flowering stages, while concentrations of both elements decreased by > 60% (Schultz & French, 1978).
In densely planted, fast-growing crops, the transition from young leaves to senescing leaves happens rapidly (Niinemets, 2007), and such leaves do not reach their potential life-span. This may cause a reduction in PPUEleaf,life but enhances growth and whole-plant PUE if the proportional gain in whole-plant photosynthesis is greater than the proportional loss of P as a result of leaf senescence. Thomas & Sadras (2001), however, point out that leaves are often retained when they have long ceased to make a positive contribution to plant C balance. These authors argued that these ‘unproductive’ leaves may act as storage organs for C and N needed at a later stage for reproductive growth. The importance of such storage and its effect on whole-plant C balance depend on developmental stage and nutrient availability. In dense stands of Xanthium canadense, leaf retention at high N availability was longer than expected based on optimal allocation for whole-plant C gain (Oikawa et al., 2008). The selective pressure for retention of N and P in unproductive tissue may be primarily related to allocation of these reserves to reproductive output in nature, or to yield in crops; constraints on the N : P ratio in sieve tube sap are considered in Notes S3. There is only limited research into the effect of N or P redistribution on C balance in the reproductive stage, but Schieving et al. (1992) showed that leaf shedding and associated remobilization of N in flowering Solidago altissima helped maintain high rates of canopy photosynthesis. In contrast to the ‘parasitic’ old leaves described by Thomas & Sadras (2001) for species at high nutrient availability, Reich et al. (2009) found that leaves of ten evergreen woody species still had a positive C balance when they reached their average life-span. Enhancing PUE of crops will require a better understanding of the distribution and redistribution of P among leaves in the canopy and its effect on C gain.
VIII. Internal redistribution of P in a growing vegetative plant
As plants tend to remobilize at least 50% of P from senescing leaves, and often much more (Aerts, 1996), redistributed P is a quantitatively important P source for growth, especially at later stages of plant development and in situations where soil P availability is low. In vegetative plants in the exponential growth phase, amounts of senescing tissues are much smaller than amounts of growing tissue, such that the potential P benefit from remobilization is very small, probably only a fraction of the total P required (Fig. 6). At this stage, P uptake by roots is by far the most important P source, and roots would indeed be expected to be exploring the most P-rich soil. As growth rates decline, usually when the older foliage experiences lower light conditions and starts senescing, and expansion and P uptake by root systems decrease, remobilization can provide significant amounts of P to new growth. This shift from uptake-dominated P supply to remobilization-dominated P supply probably happens in the late vegetative or early reproductive stage (Fig. 5). Remobilized P is also particularly important in P-deficient plants. Under P starvation, plants display several transcriptional and post-transcriptional adaptive responses (Plaxton & Tran, 2011; Notes S4). P deficiency accelerates senescence in some but not all cases (Crafts-Brandner, 1992; Lynch & Brown, 2006), but suppression of new growth in P-deficient plants is always likely to be larger than any effect on senescence, such that the relative contribution of remobilized P for growth becomes greater.
Between early and late vegetative stages, crop growth is faster than P uptake, causing a considerable decrease of shoot [P] (Fig. 7); however, remobilization of P from senescing tissues helps maintain high [P] in young leaves which fuels this growth. The importance of P remobilization during vegetative growth is expected to be greatest in crops that have long growing seasons and short leaf life-spans, and grow in dense stands.
IX. Allocation of P to reproductive structures
Large changes occur in monocarpic plants, such as annual grain crops, when they enter into the reproductive phase. The massive translocation of C and mineral nutrients into seeds during this last phase of the life cycle enhances resource availability for rapid early growth of the next generation (Sklensky & Davies, 1993). Domestication and breeding of crops for high yields has achieved further increases in the proportion of dry mass allocated to seeds (harvest index (HI); Sinclair, 1998), and the allocation of P and N has increased similarly. High grain N concentrations are favoured because of the nutritional value of protein; however, high P concentrations are not necessarily desirable (Batten, 1986; Raboy, 2009). First, most P in cereal grains and legume seeds is in the form of phytate, which is not assimilated by humans or monogastric livestock and restricts the bioavailability of iron and zinc in their diets (White & Broadley, 2009). Secondly, most P contained in produce that is exported from crop land is released to the environment in the form of waste streams (e.g. manure and sewage) and needs to be replaced with new P-fertilizer to maintain soil P status (Raboy, 2009). There are therefore nutritional as well as environmental reasons for reducing the allocation of P to grain (the phosphorus harvest index (PHI)) while maintaining or increasing HI. However, mutants with reduced seed phytate concentrations often, although not always, have reduced vigour (Raboy, 2007). Crop management options may be able to deal with this by separate production of seeds under higher P fertilization rates, and/or by ensuring high P availability for seedling growth through seed coating technology or fertilizer placement (Pariasca-Tanaka et al., 2009; Burns et al., 2010; Sekiya & Yano, 2010). But is it feasible to decrease grain P while increasing yield? Before this is discussed, it is important to examine the processes that determine grain yield and its P content.
Patterns of P accumulation and partitioning in a number of crop species (Figs 4, 5) reveal that most of the increase in seed P content is synchronous with a decrease in the P of senescing leaves and stems, although there can still be some increase in whole-plant P content (i.e. net P uptake) during early reproductive growth, particularly in indeterminate crops. As P concentrations are typically much higher in grain than in vegetative material at maturity, the proportion of total plant P that is found in grains (PHI) is higher than the proportion of dry mass (HI) (Fig. 8a). Breeding for high yield in the past has increased HI in all crops (Sinclair, 1998), but PHI has not increased to the same extent, which means that a decrease in P concentrations in the grain has occurred. Batten (1986) reported a decrease of c. 27% in grain [P] from diploid to hexaploid wheat grown in pots with adequate P fertilization, while HI more than doubled and PHI increased by only 15%. This trend was also shown in a field study by Calderini et al. (1995) for wheat cultivars released between 1920 and 1990, but a pot study of cultivars released between 1840 and 1983 showed increased HI but not decreased grain [P] (Jones et al., 1989). In 37 varieties of rice, with high average HI and PHI, correlations between grain [P] and these allocation indices were weak (Rose et al., 2010). P fertilization causing a six-fold increase in wheat grain yield did not affect PHI and HI significantly (Fig. 8b; Jones et al., 1989).
The above indicates that modern crop varieties use P more efficiently than older varieties (higher PUEy), as a result mainly of improvements in HI which are related to plant structural and C-allocation traits. How much further can these traits be improved? Can key physiological traits such as high photosynthetic rates (enabling a small plant to produce a large amount of photosynthates to fill grain) and efficient translocation of carbohydrates to growing seeds be maintained in crops that have low leaf [P]? The breeding goals for efficient use of P are quite similar to those for N, with the important exception that high grain [N] can be a positive trait and high grain [P] is not. As shown above, efficient use of N and P requires both elements to be present at high concentrations. Similarly, high percentage remobilization from senescing to young vegetative tissue is also favourable for both N and P. Therefore, the key question for P is whether mobilization of P into grains can be reduced without reducing grain growth or grain [N]. Does grain-filling require high [P]? Are the fluxes of carbohydrates, N and P into grain functionally linked? Aspects of phloem function that relate to these issues are discussed in Notes S3.
Marshall & Wardlaw (1973) observed a close correspondence between transport of P and that of sugars from source leaves to grain in wheat and concluded that the movement of P from leaves is largely determined by the movement and demand for carbohydrate within the plant and not by the P requirement of the sink. Similarly, Pugnaire & Chapin (1992) found that P and N remobilization from leaves to ears in barley increased in plants with high sink strength, but also those with low source activity. Peng & Li (2005) confirmed this in wheat. Batten & Wardlaw (1987) found no positive effect on grain growth after adding P to P-deficient wheat ears. Movement of P into developing wheat grains is faster than that of carbohydrates, and apparently regulated independently (Batten & Wardlaw, 1987; Rodriguez & Goudriaan, 1995; Peng & Li, 2005). P arriving in the grain tissue is quickly converted to phytate to keep Pi concentrations low. High Pi concentrations would inhibit starch formation (Smidansky et al., 2002, 2003). Genetic transformation of the key regulatory enzyme of starch biosynthesis, ADP-glucose pyrophosphorylase, to reduce its allosteric inhibition by Pi leads to greater yield in rice and wheat, not by improving allocation of C into seeds (improved HI), but by enhancing sink strength which indirectly increases plant growth (Smidansky et al., 2002, 2003). As P import in early grain growth seems more than adequate, even under P deficiency, it would seem worth exploring ways to reduce P allocation to grain.
The N : P ratio in grain is lower than that in the vegetative plant. Greenwood et al. (2008) modelled the decrease of N : P with development in horticultural crops as a transition from growth-related tissue with an N : P ratio of approx. 12 to storage-related tissue with an N : P of approx. 6. Sadras (2005) found that adequately fertilized cereal and oilseed crops attaining maximum yields had N : P ratios at final harvest between 4 and 6. In wheat, grain N : P was unchanged from old to modern wheat cultivars and averaged approx. 6 (Calderini et al., 1995). It appears that N : P ratios are quite conserved in productive, adequately fertilized crops. In noncrop species, seed N : P ratios are also lower than whole-plant N : P ratios (Güsewell, 2004). The larger allocation of P than N to seeds agrees with the greater degree of remobilization of P from leaves (e.g. 22% more remobilization of P than N in graminoids; Aerts, 1996). Greater remobilization of P than N from senescing tissue may partly explain low seed N : P, but there may also have been stronger selection for high seed P, perhaps because acquiring poorly mobile P is more challenging than acquiring N for small root systems.
X. Constraints to P remobilization
Remobilization of P from senescing leaves is more variable than that of N. P-resorption efficiency (PRE; the percentage of mature leaf P that has been exported before death) can reach values as high as 90% (Aerts & Chapin, 2000). In wild plants PRE is not significantly different among plant groups; for example, woody species and forbs have similar means and variability (Aerts & Chapin, 2000). Nevertheless, there is evidence that PRE is under selection, giving rise to ‘ecotypes’ that have higher PRE in P-limited sites (e.g. Güsewell, 2005; Lovelock et al., 2007), suggesting that selective breeding for high PRE in crops could be successful. The high variability in P remobilization is determined by its sensitivity to a range of environmental factors, including nutrient availability, water availability, and sink strength (Pugnaire & Chapin, 1992; Güsewell, 2005; Fife et al., 2008). When conditions favour remobilization, little P is left behind in senesced tissue (Fig. 6; Schultz & French, 1978). Güsewell (2005) suggests that higher PRE is achieved by lengthening of the senescent phase of leaf life times and by increasing the sink strength for P (e.g. underground storage organs, or seeds). Plant breeding has so far focused mainly on lengthening or shortening leaf life time before senescence rather than on modifying the duration of the senescent phase (Gregersen et al., 2008), yet this may be a trait that if altered could improve PRE in crops. The possibility that PRE may be related to the relative sizes of the different P pools has not been explored in crops or other plants.
It was shown earlier that low leaf [P] and high PRE are important factors for PPUEleaf,life, in a global leaf trait data set where leaf longevity and photosynthesis are inversely proportional. Aerts & Chapin (2000) came to a similar conclusion for forbs and graminoids, from which many crops are derived. This seems to offer little hope for increased crop PUE, as low [P] and high PRE are not common in fast-growing plants with high levels of metabolic activity. However, this does not mean that no improvement is possible. There is significant variation among and within species in these traits, and it is important to assess which traits can be improved without secondary undesirable effects (Lambers et al., 2011).
Insight into the key enzymes and transporters involved in the breakdown of P compounds, their export from senescing tissue and transport to growing tissue is increasing (Notes S4). Considering the large pool of nucleic acid P in cells, it is not surprising that RNases become more abundant during senescence, as well as under P deprivation (Taylor et al., 1993; Bariola et al., 1999; Morcuende et al., 2007). Similarly, purple acid phosphatases (PAPs) play an important role in the hydrolysis of Pi from a wide range of organic P compounds including mononucleotides derived from nucleic acid breakdown (Plaxton & Tran, 2011). Transcription of the phosphatase genes AtPAP17 and AtPAP26 is greatly increased in senescing Arabidopsis leaves (Gepstein et al., 2003), suggesting an important role for the encoded enzymes in Pi scavenging during senescence. Long-distance transport of nutrients from senescing leaves to the seeds or other parts of the plant requires Pi transporters. Reports are now emerging of individual Pi transporter genes that are up-regulated at the transcript level in senescent tissues; this suggests that they have a role in Pi remobilization from old source to new sink tissues and are linked to ethylene signalling (Chapin & Jones, 2009; Nagarajan et al., 2011). Further research on the molecular physiology of P remobilization may provide valuable insights leading to enhanced P re-use in crop plants.
XI. Do physiological or phylogenetic trade-offs constrain traits that could improve PUE?
Correlations among traits of diverse land plants provide valuable insights into trade-offs in the use and allocation of P. In some cases it might be desirable to breed crop plants that have combinations of traits that are rarely observed in nature. It would therefore be useful to know whether the observed correlations among species are imposed by physical or genetic constraints, or if they are the result of natural selection in which certain trait combinations are disadvantageous, but not impossible. A recent analysis (Donovan et al., 2011) suggests that strong correlations among species in leaf traits (photosynthesis, N concentration, leaf mass per unit area and leaf life-span) are the result of natural selection, rather than being attributable to unavoidable physical or genetic constraints. This was shown by compiling information from studies of quantitative genetic variation in leaf traits. Specifically, it was observed that individuals from the same species typically exhibit little genetic correlation in leaf traits (e.g. LMA and photosynthesis per area), suggesting that only a small proportion of variation in trait values (mean heritability values < 0.3; Donovan et al. (2011) can be attributed to genetic similarity. In the future, it might be informative to apply similar analyses to measures of intra-specific variation in P-use traits. This could help us understand mechanisms behind the evolution of inter-specific correlations between P use and other traits (e.g. the relationship between leaf P and N concentrations; Fig. 2), and provide insights into the potential to decouple these correlations through artificial selection.
XII. Identifying genetic loci associated with PUE
There is substantial genetic variation in traits associated with PUE within the crop plants that have been examined (see references in White et al. (2012) and within Arabidopsis, see Prinzenberg et al. (2010)). Analysis of this variation has led to the identification of numerous genetic loci that influence PUE. The ability to identify these quantitative trait loci (QTL) suggests that improvements in PUE may be gained through conventional or marker-assisted breeding programmes, directed gene identification and genetic engineering, or a combination of these approaches.
Nearly all QTL mapping has focused on traits associated with efficient P acquisition rather than efficient internal use of P (see references in White et al. (2012)). These traits include high total plant P content, large root systems, improved root architecture (increased lateral root production, improved topsoil foraging, greater root surface to volume ratio, greater root hair production, and greater root:shoot ratio), and the exudation of phosphatases and organic acids into the rhizosphere. Nevertheless, QTL that have the potential to influence internal PUE have been found in several crop species (see references in Vance (2010); Rose et al. (2011); White et al. (2012)). However, internal PUE is generally lower in plants with high P-acquisition efficiency as a result of higher tissue P concentrations, making it difficult to disentangle QTL that affect agronomic PUE generally from QTL that may specifically influence internal PUE (Rose et al., 2011). Rose et al. (2011) proposed that identifying QTL for internal PUE requires studies where P acquisition is equal and metabolically nonsaturating among cultivars. As an alternative, explicit quantification of tissue P pools would allow a more specific evaluation of genotypes and identification of QTL that are related to the efficiency of P use in nucleic acids, phospholipids and P-esters. As shown in Fig. 1, plants with high P concentrations store large amounts of Pi in vacuoles. This large Pi pool does not contribute to metabolism and growth and therefore has a net negative effect on internal PUE. Calculation of PUE indices based on the main metabolically active P pools (i.e. growth or photosynthesis expressed per unit of nucleic acid, phospholipid, or P-ester), might yield better insights into the efficiency with which P is used at a cell physiological level. This approach might also help us to understand the processes that lead to accumulation of Pi in existing tissues rather than its incorporation in metabolically active P of new tissues. The P-pool-specific PUE indices are likely to be better suited to QTL mapping studies, and should be targeted if the goal is to identify QTL and genes related specifically to internal PUE.
There is much scope for future mapping of QTL that influence internal PUE. Advanced methods of genetic analysis have not yet been used extensively in the area of plant P relations. Approaches such as genome-wide expression (transcriptome) QTL analyses (Hammond et al., 2011) and genome-wide association studies (Zhao et al., 2007; George et al., 2011) supported by such emergent technologies as next generation sequencing hold much promise to identify loci related to and controlling plant P relations, including internal PUE (Vance, 2010; Morrell et al., 2012).
A greater effort to enhance the efficiency of P use for crop production will reduce environmental impacts associated with P fertilizers and P waste, increase the nutritional value of grains, and improve farm economies. We have identified developmental and physiological aspects of PUE, underpinned by genetics and molecular biology, which suggest that significant improvement in internal PUE is possible. The magnitude of PUE gains that may be obtained through different mechanisms, and their variation associated with genetic and environmental factors, should be quantified through targeted research. More efficient use of P within the plant adds to the gains that can be made by improving P-acquisition efficiency, but also reduces P fluxes on crop land and in the environment. The largest yield benefits of improved PUE are expected for crops growing in soils that have very low P content and where little or no P fertilizer is applied. The largest savings in P fertilizer are expected on productive land where conditions for crop growth are near optimal.
This paper is a result of the workshop ‘Phosphorus, the inside story’, held at the School of Plant Biology, The University of Western Australia. Funding contributions from the School of Plant Biology, the Faculty of Natural and Agricultural Sciences, the Pro Vice-Chancellor (Research) and the Institute of Agriculture of The University of Western Australia are gratefully acknowledged. The University of Dundee is a registered Scottish charity, No SC10596. Constructive comments from three reviewers are gratefully acknowledged.