Nitrogen and phosphorus interactions at a 21 nitrogen:1 phosphorus Redfield‐like ratio impact growth and seed yield in wheat (Triticum aestivum L.)

Abstract

Efficient use of nitrogen (N) and phosphorus (P) is essential to reduce fertilizer costs and nutrient pollution and to lower the carbon footprint of agriculture. This requires a better understanding of N and P limitations on photosynthesis and biomass generation in one of the world's most important crops, wheat (Triticum aestivum). In a fully factorial nutrient experiment, we find that the limiting nutrient, N or P, at a Redfield‐like N:P ratio (21N:1P molar) produces the ‘effective dose of fertilizer’ for the generation of above‐ground biomass (seed or vegetative). Best‐fit structural equation models showed that neither N nor P applied in excess of 21N:1P led to any increase in biomass. The light (qP, NPQ) and dark (A_max, V_cmax, J_max) reactions of photosynthesis are also constrained by the limiting nutrient at the 21N:1P Redfield‐like ratio. Beyond this ratio, ‘excess N’ or ‘excess P’ had no effects. The direct effects of the ‘effective dose of fertilizer’ on the accumulation of biomass were stronger than its indirect effects via photosynthesis, likely driven by the N and P costs in building nucleic acids needed for cell division, cell maintenance and RNA transcription. Modern composite fertilizers are typically provided at a much higher P content than a 21N:1P ratio, potentially resulting in a huge global wastage of P, a finite resource, with all the concomitant costs to the farmer, consumer and the environment.

INTRODUCTION

While the intensification of agriculture through the application of increasing amounts of fertilizer has typically led to improved yields per hectare for many crops (e.g. wheat, rice, maize and soybean) across the world, in recent years, yields have been slowing, stagnating or even declining for some crops depending on where they are grown (Ray et al., 2012). Such observations are perhaps surprising given that fertilizer application has increased fourfold since 1965, with 109 million metric tonnes of nitrogen (N) and 48 million metric tonnes of phosphorus (P) used globally in 2021 (International Fertilizer Industry Association, 2023). In addition, it is notable that the typical mass ratio at which the main nutrients of nitrogen (N) and phosphorus (P) are applied has also risen from an N:P mass ratio of ~1:2 in 1965 to ~2:3 in 2021 (International Fertilizer Industry Association, 2023) despite both ratios being very P rich compared with plant tissues (see below). Given that N and P fertilizers are a major and growing cost for farmers and consumers, contribute to greenhouse gases (both in their manufacture and in reducing soil carbon) and lead to run-off that damages ecosystems, it is essential that they are applied at an appropriate level and N:P ratio. This will optimize their impact on crop yield and ensure food security while minimizing their detrimental impact on the environment (Chowdhury et al., 2017).

Wheat (Triticum aestivum L.) is one of the world's most important crop species by yield, being used for both human consumption and livestock feed. The huge global production of wheat has been achieved through the application of fertilizers, such that wheat production alone uses 18.2% of the global supply of N (Heffer et al., 2017). Hitherto, most research aiming to improve N and P nutrient use efficiency has focussed on, for example, breeding to increase root/shoot ratios (Heinemann et al., 2023), adjusting the type and method of application of fertilizers (e.g. Kabato et al., 2022; Yokamo et al., 2023) or enhancing N or P acquisition, by, for example, increasing expression of phosphate transporters (e.g. Lynch, 2019; van de Wiel et al., 2016). Here, we take an alternative strategy, by focussing attention on how the ratio of N and P applied plays a role in influencing efficient biomass and seed production.

The N and P ‘Redfield ratio’, which typically occurs at a molar ratio of 106 Carbon (C):16N:1P, has ‘come to define our understanding of ocean biogeochemical cycling’ (Nature Geoscience Editorial, 2014). The N:P ratio inherent to the Redfield ratio is found in much marine phytoplankton and typical seawater, and while variants of the 106C:16N:1P ratio have been reported in individual species and some aquatic systems (Sterner & Elser, 2002), an overall tight nutrient stoichiometry between sea water and phytoplankton is probably maintained by feedback loops involving available P, the balance between bacterial N fixation and N denitrification and the N:P requirements of phytoplankton for growth (Gruber & Deutsch, 2014). This intriguing coupling of N to P in the generation of biomass has led to work examining N:P ratios in terrestrial systems. A meta-analysis revealed constrained ‘Redfield-like’ ratios in soils and microbial biomass, suggestive of similar processes operating in soils as reported in the oceans (Cleveland & Liptzin, 2007). In terrestrial plants, N:P ratios have been shown to vary depending on available soil nutrients, life form and the tissue analysed (Hobbie & Gough, 2002; Sterner & Elser, 2002; Tian et al., 2017; Wang et al., 2024), the latter arising, for example, from the occurrence of stored nutrients (e.g. N as proteins and pigments and P as phytate) and the different functions of organs (leaves, tubers, etc.) (Tang et al., 2018). Nevertheless, an analysis of leaves across 1280 seed plant species revealed only limited variation in N:P stoichiometry between plants across different latitudes and environments, ranging from a reported mean mass ratio of 11N:1 P in Arctic plants to 23N:1P in tropical plants (note: molar N:P ratios are 2.21 greater than the mass ratio) (Reich & Oleksyn, 2004). Potentially the similarity of N:P ratios in leaves, despite widely different ecologies, indicates that seed plants are similarly constrained by N and P, as occurs in marine phytoplankton and soil microbes. If so, it can be envisaged that N and P fertilizers should be supplied at, or near Redfield ratio of the crop, to best sustain yield, minimize waste and costs and reduce the environmental damage of fertilizer runoff. However, fertilizers are not typically supplied at these ratios. To achieve optimal fertilizer use, we need to better understand how the application of different N and P treatments might affect biomass production.

In considering the application of N and P for plant growth, it is possible that maximum biomass production will occur when N and P are supplied at or near the optimal Redfield ratio (i.e. when plant growth is limited equally by both N and P). Assuming so, should the supply depart from such a ratio, then growth will be limited by the scarcer of the two elements. This hypothesis is supported in a review of short-term wetland and grassland fertilizer experiments, which suggested that N and P applied in N:P mass ratios of <10 typically (but not always) caused plant biomass to be constrained by N, while ratios >20 resulted in growth constrained by P. Between these ratios, it was assumed that the plants were N and P co-limited (Güsewell, 2004). Similarly, a comparison of data from numerous plant systems (including conifer forests, chalk grasslands and estuaries; Tessier & Raynal, 2003) indicated N limitation in the mass ratio range <6.7–16N:P and P limitation in the range > 12.5–26.3.

Biomass N:P molar ratios can be poor indicators of available N and P in the soil due to plant homeostatic regulation. For instance, the biomass N:P ratio of Carex curta ranged only 2-fold despite N and P applications that ranged more than 600-fold (Güsewell, 2004, 2005). Moreover, no correlation was found between soil and leaf N:P ratios in plants growing along a latitudinal gradient across the Loess Plateau in China (Fang et al., 2019) or within tropical forests (Townsend et al., 2007). Such data suggest that plants regulate N and P uptake within narrow limits, consistent with a hypothesis that plant growth is co-limited by N and P within ranges of the N:P ratio.

Here, we address the following hypotheses that wheat (i) vegetative biomass (i.e. above-ground biomass minus seed biomass) and (ii) seed biomass are co-limited by N and P because the light and dark reactions of photosynthesis are N and P demanding (Figure 1). We also address the hypotheses that (iii) vegetative and (iv) seed biomass are constrained by the limiting nutrient, N or P, at a Redfield-like N:P ratio.

[IMAGE OMITTED. SEE PDF]

MATERIALS AND METHODS

Source and growth of plant material

A total of 108 seeds of Triticum aestivum ‘USU-Apogee’ were used in two N and P treatment experiments. Experiments 1 and 2 analysed young plants in their 3rd week of growth (between days 14 and 22), and in addition, Experiment 2 also analysed the same plants again at maturity (~3 months old) when they were green but setting seed (see below for parameters measured).

Experiments 1 and 2 each comprised 18 seeds sourced from Utah University (USA) (= S1A, Type A), 18 seeds from the University of Bristol (UK) (= S2X, type X) and 18 seeds obtained by selfing plants from the University of Bristol–sourced material (= S2Y, Type Y). In each experiment, we grew three seeds (one seed from each source) in 1 of 15 varying N and P treatments. In a further control treatment, with no added N or P, we grew nine seeds, comprising three seeds from each source. Thus, in total, there were 16 experimental treatments, n = 15 × 3 seeds = 45 seeds and 1 × 9 seeds (control) = 54 seeds for each of Experiments 1 and 2.

Preparation of soil mixtures corresponding to the 16 N and P experimental treatments investigated

To prepare 16 nutrient treatments that vary in N and P content, first, we made four nutrient solutions comprising: (1) high N/High P treatment – 4.33 g/L MS medium (Murashige & Skoog, 1962). (2) High N/Zero P treatment – 0.61 g/L modified MS medium (lacking N and P), plus 1650 mg/L NH₄NO₃ and 1900 mg/L KNO₃. (3) Zero N/high P treatment – 0.61 g/L modified MS medium, plus 170 mg/L KH₂PO₄ and 710 mg/L KCl. (4) Zero N/zero P treatment (control) – 0.61 g/L modified MS medium, plus 710 mg/L KCl (see Table S1). A further 12 experimental treatment solutions were prepared by mixing these four solutions in appropriate proportions to derive the 16 experimental nutrient treatment solutions outlined in Table S1, comprising zero N, low N (25% of full strength), mid N (50% of full strength) and high N (full strength) each with zero P, low P (25% of full strength), mid P (50% of full strength) and high P (full strength). The 16 nutrient solutions were then mixed with soil in a 4:1 volume ratio of soil-to-nutrient solution (the soil comprised Westland® perlite, Westland® vermiculite, Westland® horticultural washed sand and Canna® coir (coco natural plant medium) in a 3:2:1:1 ratio by volume).

The soil/nutrient mix was autoclaved to kill microorganisms prior to use, placed in the dark at 20°C and turned over every day for 8 days to allow complete homogenization of the nutrients with the soil. The soil/nutrient mix was then added to the 7.5 × 7.5 × 5 cm (= 200 mL) plant pots to give one of four N treatments (0, 10.3, 20.6 and 41.2 mg N pot⁻¹) with one of four P treatments (0, 0.5, 1 and 2 mg P pot⁻¹). Table 1 shows how these nutrient treatments translate into the amount N and P added per pot surface area (kg ha⁻¹) and mmol N and P supplied per pot.

TABLE 1 Amount of nitrogen (N) and phosphorus (P) applied in the 16 experimental treatments used to investigate the impact of nutrient limitation on a range of photosynthetic, growth and seed parameters of Triticum aestivum ‘USU-Apogee’.

Treatment	Zero N	Low N	Mid N	High N
N	P	N	P	N	P	N	P	N	P	N	P	N	P	N	P
mg pot−1 (kg ha−1)	mmol	mg pot−1 (kg ha−1)	mmol	mg pot−1 (kg ha−1)	mmol	mg pot−1 (kg ha−1)	mmol
Zero P	0 (0)	0 (0)	0	0	10.3 (23.3)	0 (0)	3	0	20.6 (46.6)	0 (0)	6	0	41.2 (93.2)	0 (0)	12	0
Low P	0 (0)	0.5 (1.1)	0	0.0625	10.3 (23.3)	0.5 (1.1)	3	0.0625	20.6 (46.6)	0.5 (1.1)	6	0.0625	41.2 (93.2)	0.5 (1.1)	12	0.0625
Mid P	0 (0)	1 (2.2)	0	0.125	10.3 (23.3)	1 (2.2)	3	0.125	20.6 (46.6)	1 (2.2)	6	0.125	41.2 (93.2)	1 (2.2)	12	0.125
High P	0 (0)	2 (4.4)	0	0.25	10.3 (23.3)	2 (4.4)	3	0.25	20.6 (46.6)	2 (4.4)	6	0.25	41.2 (93.2)	2 (4.4)	12	0.25

Seed germination and growth

All seeds were first surface sterilized with a solution of 10% (v/v) sodium hypochlorite and 0.1% (v/v) Tween-20 in deionized water to reduce or eliminate bacterial contamination. The seeds were then left to dry for 1 day before sowing in the pots containing the appropriate soil nutrient mixture (see above) and watered with deionized water (Elgar pure lab). The pots were then left at 4°C in a dark room for 7 days to synchronize seed germination.

The pots were then placed in a dish in a growth room and plants were grown at a constant temperature (25°C) with a 16-h day and an 8-h night cycle. The plants were grown under 300 μmol photons m⁻² s⁻¹ (canopy height and photosynthetic active radiation). All plants were watered with deionized water three or four times a week in such a way as to minimize loss of nutrients through runoff (any liquid loss that did occur into the pot's dish was returned back to the pot). Pots were randomly arranged in groups of five on circular plates under the light source and their positions changed every 2nd or 3rd day by rotating the plates and changing the plate position on the shelf, using a movement design that minimized potential position effects (e.g. edge effects).

Assessing biomass

Above-ground biomass (including any seed biomass if present) was measured for Experiments 1 and 2 by weighing the material after drying at 40°C for 8 days. For Experiment 1, measurements of above-ground biomass included biomass removed for measuring leaf chlorophyll and C, N and P content. This was achieved by partitioning the removed biomass into two, taking the wet weight of both fractions, drying one fraction to obtain dry weight and using that wet weight:dry weight ratio to estimate the dry weight of material used for measuring leaf chlorophyll and C, N and P. Plants from Experiment 2 were left to grow until set seeds (~3 months) for seed biomass, number and C, N and P content measurements.

Chlorophyll content

Chlorophyll content was estimated from the first leaf of each plant in Experiment 1. Fresh leaves were weighed and stored at −80°C in 2 mL tubes and ~200 mg of leaf material was placed in 2 mL Eppendorf tubes, ground in a TissueLyser LT QIAGEN with one or two metal beads. Immediately after grinding, samples were placed on ice to minimize tissue degradation. Chlorophyll extraction, following Porra et al. (1989), was performed by adding 3 mL of 80% (v/v) acetone in 2.5 mM sodium phosphate buffer pH 7.8 to 100–200 mg of leaf material to minimize conversion of chlorophylls to phaeophytins.

Stomatal measurements

Stomatal length and density were measured from fully emerged third leaves from three plants in Experiment 1 that were growing in zero N with zero P, or high N with high P. At the point of harvest, stomatal length and density were assessed using the method described in Hilu and Randall (1984). The mid-section of each abaxial leaf surface was painted with clear nail varnish and after drying, the nail varnish was removed with clear tape, leaving the epidermis impression on the tape. This impression was photographed using an ORCA ER camera mounted on a Leica CTR MIC microscope. Stomatal length was measured from 33 stomata of each leaf from the abaxial surface using ImageJ. Stomatal density was calculated as the number of stomata per square millimetre of the epidermis, measured in 23 different fields of view of the epidermis of each leaf.

Chlorophyll fluorescence to assess the performance of photosystem II (PSII)

Photosynthetic efficiency was measured using fully emerged leaves from plants of Experiment 2 that were 2 to 3 weeks old after seed germination when they had just three to four leaves (the age of plants was used as a fixed effect in linear models). Plants were dark adapted for 30 min prior to analysis.

The data were obtained by analysing rapid light curves (RLC) and induction light curves (ILC) providing information on (i) the relative electron transport rate (rETR) through photosystem II (PSII); (ii) the amount of light energy used for photosynthesis (= qP); and (iii) the amount of light energy dissipated as heat (= non-photochemical quenching; NPQ) or fluorescence (Maxwell & Johnson, 2000). All chlorophyll fluorescence measures were made using Junior-PAM (Walz, Germany) and WinControl-3.24 software to generate RLC and ILC.

RLCs were generated using actinic light (wavelength range from 400 to 700 nm) that increased in intensity in eight steps from 0 to 65, 90, 125, 190, 285, 420, 625 and 820 μmol photons m⁻² s⁻¹ for a duration of 10 s each (Ralph & Gademann, 2005). Curve fitting using methods given in Platt et al. (1980) was used to give estimates of the following parameters: (i) maximum electron transport rate (ETR_max) which is an approximation of the maximum rate of electron flow through PSII and is defined as the effective quantum yield of PSII (ϕPSII) multiplied by the photosynthetically active radiation (PAR); (ii) F_v/F_m is the maximum quantum efficiency of PSII photochemistry derived from (F_m−F_o)/F_m, where F_m is the maximum fluorescence level after the first saturating pulse given to dark-adapted material, F_o is the basal fluorescence level from weak light (<1 μmol photons m⁻² s⁻¹) and F_v is F_m−F_o (Maxwell & Johnson, 2000). Given that ILC and RLC both provide an estimate for F_v/F_m, the value reported in the results is the average of the F_v/F_m values obtained from both approaches.

ILCs were generated using a different part of the same leaf used for RLC following the approaches described in Murchie and Lawson (2013). They were obtained over ~27 min in two cycles, each divided into a light and a dark phase. In both cycles, the leaf material was given a pulse of intense, saturating light to determine the maximum quantum efficiency of PSII (i.e. F_v/F_m). Then, actinic light (420 μmol photons m⁻² s⁻¹ intensity) was applied for 5 min and, during this light phase, 0.8 s of saturating light pulses (10,000 μmol photons m⁻² s⁻¹) was applied every min to determine the level of maximum fluorescence under actinic light (Fm′). The remaining period of the ILC was conducted without actinic light, with saturating light pulses applied every 2 min for ~7 min in total. Then, in the second cycle, the same procedure was repeated but with a weak light (<1 μmol photons m⁻² s⁻¹), the ‘dark’ phase, to keep PSII reaction centres open, enabling measurement of Fo. ILCs were generated for the following parameters: (i) NPQ was calculated as (F_m−$$$ {F}_{\mathrm{m}}^{\prime } $$$)/$$$ {F}_{\mathrm{m}}^{\prime } $$$. NPQ values were taken from the last pulse of the last light phase of the ILC with actinic light; (ii) qP was derived from ($$$ {F}_{\mathrm{m}}^{\prime } $$$−F′)/($$$ {F}_{\mathrm{m}}^{\prime }-{F}_{\mathrm{o}}^{\prime } $$$), where F′ is the minimum fluorescence under actinic light and $$$ {F}_{\mathrm{o}}^{\prime } $$$ is the minimum fluorescence without light. qP values were taken from the last light pulse in the cycle of the ILC with actinic light.

A/C_i photosynthetic gas exchange analysis to assess carbon assimilation

A/Ci photosynthetic gas exchange analysis was conducted as outlined in Johnson and Murchie (2011) using a CIRAS-1 gas exchange system (PP Systems, Amesbury, MA, USA). A light source (Schott halogen cold light KL 1500) with a saturating irradiance of 1500 μmol m⁻² s⁻¹ was installed over the leaf chamber, and a LI-190R (Li-Cor Biosciences, Lincoln, NE, USA) measured the irradiance. The analysis was conducted on intact leaves that were clamped into an airtight 2.5 cm² cuvette at a vapour pressure deficit of 1.3 kPa and acclimatized for 30 min at a CO₂ concentration of 400 μmol.mol⁻¹ until a steady state of CO₂ uptake (A) was reached. The temperature in the cuvette varied between 21.03°C and 25.92°C. The CO₂ concentration in the cuvette was decreased in seven steps (300, 200, 120, 100, 80, 60 and 40 μmol mol⁻¹) before returning to the initial concentration, and then increased in four steps (600, 800, 1200 and 2000 μmol mol⁻¹) with ~5 min acclimation at each step. The internal leaf CO₂ concentration (C_i) and (A) are given by the CIRAS-1 software.

Curve fitting and modelling were performed using the Plantecophys package in R (R Core Team, 2016; Duursma (2015)) in the fitaci function. The A/C_i curves under saturating light provided information on (i) the maximum rate of CO₂ uptake (A_max) under saturating light and CO₂ (2000 μmol mol⁻¹), (ii) the maximum rate of carboxylation of ribulose-1,5-bisphosphate (RuBP) by RuBisCO (V_cmax) and (iii) the maximum rate of electron transport (J_max) used to regenerate RuBP (Farquhar et al., 1980).

Measurement of C, N and P content

Carbon, N and P content and their ratios were measured in (i) the second leaf of each plant in Experiment 1; (ii) up to three seeds from each plant in Experiment 2; and (iii) parental seeds (three replicates of each type of seeds – S1A, S2X and S2Y – Table S2). Leaf material was dried at 40°C for 8 days and ground to a powder. For P determination, each sample was subjected to acid-peroxide digestion to allow the conversion of organic P-containing compounds to soluble reactive P. For this, potassium persulphate (0.15 g) and 10 N sulphuric acid (1 mL) were added to each sample and homogenized, followed by autoclaving for 40 min at 120°C. The samples were then filtered using 0.2 μm filters and analysed by a Continuous Flow Analyser (San++, Skalar Analytical B.V., Breda, The Netherlands). For estimating the amount of C and N in (i) the soil mix (before adding any N or P, Table S3) and (ii) plant biomass (from liquid nitrogen ground tissue), we used a combined elemental analyser and mass spectrometer (Integra 2, Sercon, UK) by analysing 1 mg ground material per sample and using Casein powder as a standard.

Statistical analyses

Data (Table S4) were analysed using linear models in R version 3.3.1 (R Core Team, 2016). Experimental N and P soil concentrations were treated as continuous independent variables. The best-fit models were found by comparing R² and AIC values and thorough scrutiny of diagnostic plots. When necessary to generate a normal distribution from the data to meet statistical assumptions, data were natural log or square root transformed. When generalized linear models (GLMs) were used, McFadden's pseudo-R² was calculated. N and P interaction terms were always tested in the models, but the interaction was removed if non-significant and model performance was improved with its exclusion.

Structural equation models (SEM) were conducted using directed separated (d-sep) conditional independence claims using the package ‘piecewiseSEM’ in R version 4.1.1 (R Core Team, 2020). The d-sep tests the conditional independence between parameters in a path diagram. The causal links are specified between variables in a directed acyclic graph (Guignard et al., 2019), such that the direction of arrows represents the edges (A) while the variables represent nodes/vertices (V). d-sep tests evaluate the missing paths directly, first by fitting the missing relationship between the variables and then by testing whether the path coefficients are significantly different from zero and we are justified in excluding them.

RESULTS

All the data obtained in Experiments 1 and 2 were fitted into linear models and are given in Table S4. For Experiments 1 and 2, ‘seed source’ and ‘age’ of plants at the time of photosynthetic data collection were used as categorical and numerical variables in the models, respectively. The variable plant ‘age’ had no significant effect and did not improve model performance when analysing photosynthetic data. ‘Age’ was therefore removed from the models. In contrast, the source of seeds was shown to have a significant effect in some models and so was included, revealing that the seed quality varied between the three sources, as also seen in the contrasting amounts of C, N and P measured in the seeds (Table S2). The full outputs from the linear models are given in Table S5 (Experiment 1) and Tables S6, S7 and S10 (Experiment 2).

Generation of above-ground biomass under N and P limitation

An analysis of young plants in their 3rd week of growth (Experiment 1) showed that both N (p < 0.05) and P treatments (p < 0.001) were associated with increases in above-ground biomass at this early stage of growth (linear model adj. R² = 0.30, Table S5A), probably because there was some available P and especially N inherent to the soil mix (see below). The effects of both nutrients on above-ground biomass were stronger for the mature plants of Experiment 2, where P (p < 0.001) and N:P interactions (p < 0.001) significantly promoted the accumulation of biomass (linear model adj. R² = 0.79, Table S6A).

The effect of varying N and P on leaf N:P ratios

We used linear models to test whether varying N and P treatments impacted leaf N and P content. The P content of young leaves (Experiment 1) increased significantly with the application of P (p < 0.001) and reduced with the application of N (via direct effects of N (p < 0.05) and N:P interactions (p < 0.01, linear model adj. R² = 0.70, Table S5B)).

We observed that leaf N content increased with the addition of N (p < 0.001, linear model adj. R² = 0.54, Table S5C; Figure 2a) resulting in increased leaf N:P ratios (p < 0.001, Figure 2b), while the addition of P decreased N:P ratios (p < 0.001, linear model adj. R² = 0.65, Table S5D). The multiple linear model shows that increasing P has the effect of reducing the N content in leaves. The addition of N (p < 0.05) was also associated with increased C:P ratios while the addition of P (p < 0.001) resulted in decreased C:P ratios (linear model adj. R² = 0.74, Table S5E). This effect is also apparent in comparisons of the regression line from a simple linear regression of N dose against N or N:P and a multiple linear regression, which considers also the effects of P and seed source (Figure 2).

[IMAGE OMITTED. SEE PDF]

The effect of varying N and P on photosynthetic parameters

We analysed the young plants of Experiment 1 and showed that the interaction of N and P led to increased chlorophyll content (p < 0.01; linear model adj. R² = 0.21; Table S5F). We also analysed wheat plants (3 weeks post-germination, Experiment 2) and showed that there were no significant effects of N and P treatments on the parameters reflecting the efficiency of the light reaction, that is, F_v/F_m (the maximum quantum efficiency of PSII; Table S6B) or ETR_max (the maximum rate of electron flow through PSII; Table S6C). However, we did observe that the addition of P significantly decreased NPQ (a measure of the amount of light energy dissipated as heat (p < 0.001, linear model adj. R² = 0.38; Figure S1A, Table S6D)) and significantly increased qP (the amount of light energy used for photosynthesis (p < 0.01, linear model adj. R² = 0.24, Table S6E)). There was also a strong positive correlation between qP (p < 0.05) and a negative correlation with NPQ (p < 0.01) on above-ground biomass (linear model adj. R² = 0.40, Table S6F, Figure S1B,C). qP and NPQ values were negatively correlated (p < 0.001, R² = 0.37, Figure S1D).

Stomatal length and density were measured in plants that received the most contrasting N and P treatments (i.e. zero N and P and high N and P – Experiment 1). While stomatal length increased significantly under high N and P (p < 0.05, linear model adj. R² = 0.60), the addition of N and P did not influence stomatal density (Table S5G,H). Given that increased stomatal size without changes in density should increase CO₂ conductance (Farquhar & Sharkey, 1982), we also analysed A/Ci curves to quantify how N and P treatments affected the dark reaction of photosynthesis, namely the maximum rate of carboxylation of RuBP by RuBisCO (V_cmax), the maximum rate of electron transport used to regenerate RuBP (J_max) and the maximum rate of CO₂ uptake used as a measure of net photosynthesis (A_max).

We observed a significant positive effect of N on V_cmax (p < 0.05; linear model adj. R² = 0.12; Table S6G), a significant positive effect of P on J_max (p < 0.01; linear model adj. R² = 0.18; Table S6H) and significant positive effects of N and P on A_max (p < 0.05; linear model adj. R² = 0.18; Table S6I). We also showed that A_max, J_max and V_cmax were significantly associated with above-ground biomass (p < 0.001, linear model adj. R² = 0.30, 0.24 and 0.19, respectively, Table S6J–L).

The effect of varying N and P on seed number and quality

We examined the mature, but still green wheat plants at seed set (Experiment 2, Table S7). Total seed biomass per plant increased significantly with the addition of P and with interactions between N and P (p < 0.05, linear model adj. R² = 0.74; Table S7A), varying approximately sevenfold across the range of N and P treatments. There was a significant increase in the number of seeds per plant with increasing N (p = 0.05) and P (p < 0.001, linear model adj. R² = 0.34, Table S7B) treatment. Depending on the N and P treatment, the number of seeds per plant ranged from 0 (zero N with zero P, i.e. the control) to 30 (high N with high P). Nevertheless, several plants growing in the zero N with zero P control soil did produce some seeds, with the mean number of seeds across the replicates being nearly three seeds per plant (Table S8). The growth of these control plants presumably reflects the presence of N and P in the seeds from which the plants were derived and any bioavailable nutrients in the soil mix (i.e. Tables S2 and S3).

In contrast to the total seed biomass and number, individual seed weight decreased significantly with the addition of N (p < 0.05) and increased with the addition of P (p < 0.05, linear model adj. R² = 0.36, Table S7C). Thus, N treatment resulted in more numerous, but lighter seeds while the addition of P resulted in plants generating more numerous, heavier seeds.

The amount of N in seeds (mg/g seed) was significantly influenced by the addition of N and P, with N application increasing the N content of seeds (p < 0.01) and P application reducing the N content (p < 0.001, linear model linear model adj. R² = 0.54, Table S7D). Similarly, the amount of P in seeds (mg/g seed) increased with the application of P (p < 0.05) and decreased with N addition (p < 0.05, linear model adj. R² = 0.27, Table S7E) all influencing N:P ratios (Table S7F).

Photosynthesis and the generation of vegetative and seed biomass are co-limited by N and P

We tested the hypothesis that the generation of vegetative biomass (above-ground biomass minus seed biomass) is co-limited by available N and P because the light and dark reactions of photosynthesis are N and P demanding (Figure 1). This was achieved using structural equation models (SEM) to determine the most likely relationships between N and P treatments, the photosynthetic parameters (qP, NPQ and one of V_cmax, A_max or J_max) and vegetative biomass (Figure 3, Figures S2 and S3). We compared five hypothesized path models (Figure S2A) with each model considering V_cmax, A_max or J_max. In each case, SEM model 5 (Table S9) best explained how N and P treatment influenced vegetative and seed biomass (for paths and summary statistics, see Figure 3, Figure S3A,B).

[IMAGE OMITTED. SEE PDF]

The models suggested that the addition of P leads to a significant increase in qP and a decrease in NPQ. Thus, the addition of P enables more light energy to be used in photosynthesis, with less being wasted as heat (NPQ). qP was also found to have a significant direct role in heat dissipation by decreasing the amount of light energy lost via NPQ. The models suggest that P has no direct effect on the parameters that measure carbon assimilation, V_cmax, A_max and J_max, (i.e. the dark reaction) but there were significant positive indirect effects of P via qP.

In contrast to the effects of P, the addition of N did have significant direct effects on V_cmax, A_max or J_max. These parameters of the dark reaction of photosynthesis were also significantly correlated with vegetative biomass (Figure 3a,c, Figure S3A), but only V_cmax was significantly correlated with seed biomass (Figure 3b). We note, however, that in simple regression analyses, A_max and qP are significantly correlated with above-ground biomass (Figure S1C,E).

N and P utilization at Redfield-like ratios

To examine if above-ground biomass production in wheat is correlated with the utilization of N and P at a Redfield-like ratio, we first analysed the effects of N without P, and of P without N on the generation of above-ground biomass (Figure 4a, Experiment 2). To conveniently visualize and calculate the effects of the N and P treatments in Figure 4a, we multiplied the mmol of P added by a factor (modified P (P_m)) to enable the two nutrients to be plotted on the same scale as N (here, the factor is ‘21’ to reflect the 21 N:1 P Redfield-like ratio that our analysis below suggests constrains wheat growth). We observed that when wheat was grown without any N or P added, there was nevertheless some biomass production (0.05–0.5 g of above-ground biomass, Figure 4a), growth that was dependent on seed resources (Table S2) and any bioavailable nutrients inherent to the soil mix before adding nutrients (Table S3). The addition of only N to the soil mix led to no further increase in above-ground biomass (see green circles, Figure 4a), but the addition of only P (see blue triangles) did result in some increase in biomass (Figure 4a). This suggests an excess of bioavailable N relative to P in the soil mix. A visual inspection of Figure 4a suggests that there is about 1.25 mmol excess N relative to P in the soil mix because adding P supported additional growth.

[IMAGE OMITTED. SEE PDF]

We used linear models to investigate the hypothesis that vegetative biomass production in wheat is constrained by the limiting nutrient, N or P, at a ‘Redfield-like’ N:P ratio. We analysed a range of plausible N:P ratios by considering the effect of the smallest value of added N (mmol +1.25 mmol N inherent in the soil mix) and P_m (mmol, where m = 15–30) on vegetative biomass. The best model (lowest AIC = −31.395 and highest R² = 0.882) was obtained by assuming a limiting nutrient at 21 N:1 P (we present results for 16 N:1 P, 20 N:1 P, 21 N:1 P, 22 N:1 P and 26 N:1 P, Table S10). All tested models that assumed a limiting nutrient at Redfield-like ratios performed better than an alternative model of the ‘actual dose’ of fertilizer (Table S10). A regression analysis plotting the ‘effective dose of N and P’ at a 21N:1P Redfield-like ratio against above-ground biomass revealed a strong positive relationship that explained much of the data (p < 0.001, regression analysis R² = 0.89, Figure 4b).

The generation of above-ground biomass is limited by N and P at a 21N:1P Redfield-like ratio

We used SEM to test the hypothesis that vegetative biomass is constrained by the limiting nutrient, N or P, at a 21N:1P Redfield-like ratio. We compared six hypothesized path models (see Figure S2A) that explored the relationships between the three variables ‘effective dose of N and P’ (= ‘effective dose’ mmol of either N or P), the effects of N applied in excess of the 21N:1P Redfield-like ratio (= ‘excess N’ (mmol N)) and P applied in excess of the 21N:1P Redfield-like ratio (= ‘excess P’ (mmol P)), on vegetative and seed biomass (for paths and summary statistics, see Figure 5 and Figure S3C,D; for demonstration of how ‘effective dose of N or P’, ‘excess N’ and ‘excess P’ are calculated, see Table S11).

[IMAGE OMITTED. SEE PDF]

The best-fit models (model 5 for vegetative biomass and model 6 for seed biomass, Table S12) showed that the ‘effective dose of N and P’ was significantly and positively correlated with qP and negatively correlated with NPQ. qP in turn was significantly and positively correlated with V_cmax, A_max and J_max. These measures of photosynthesis provided links that improved the performance of the path models, but were not significantly correlated with vegetative biomass, and only V_cmax was significantly positively correlated with seed biomass. The key difference between model 5 and model 6 is the absence or presence, respectively, of a link in the path between qP and biomass. That link is significant in models considering V_cmax (involving vegetative biomass or seeds, Figure 5a,b), A_max (involving vegetative biomass, Figure 5c) and J_max (involving vegetative biomass, Figure S3C). Potentially, this link is providing ATP and NADP to drive metabolism. Critical to all models examining ‘effective dose of N and P’ is that none showed any significant role for ‘excess N’ or ‘excess P’, that is, there were no significant effects on qP, NPQ, V_cmax, A_max, J_max, vegetative or seed biomass.

DISCUSSION

While many fertilizer studies on crops have focussed on the effect of limiting N (e.g. Cui et al., 2010; Gao et al., 2009; Reeves et al., 1993; Shangguan et al., 2000) and P (Batten, 1992; Deng et al., 2018; Wang et al., 2010), fewer have considered N, P and their interactions (but see, e.g. Güsewell, 2004; Reich et al., 2009) and none, as far as we are aware, have explored the influence of the limiting nutrient, N or P, with respect to Redfield-like ratios on above-ground and seed biomass.

Growth, biomass production and 21N:1P Redfield ratio

Our data provide support for the hypothesis that above-ground biomass production in wheat is constrained by nutrient utilization at a 21N:1P Redfield-like ratio. Indeed, we were surprised at how tightly the data fitted a regression analysis of the limiting nutrient at this ratio and above-ground biomass (Figure 4b). SEMs suggest that the limiting nutrient, N or P (‘effective dose’), manifests directly on vegetative and seed biomass and indirectly via photosynthesis. The direct effects likely arise because of the N and P demands of metabolism and growth (i.e. cell division, transcription, translation, cell membranes and cell walls). The indirect effects arising from the N and P demands of the light and dark reactions of photosynthesis (Figure 1, see also ‘The roles of N and P in photosynthesis’ below).

Despite the limitations of a 21N:1P Redfield-like ratio on the accumulation of biomass, both leaf N and P content do change with N and P treatments, probably through the accumulation of ‘excess N’ or ‘excess P’ as storage products. Thus, it is possible that the 2.21-fold range in N:P ratios reported in leaves of species growing in a diversity of environments (Townsend et al., 2007) arises from variations in the amount of N and P storage products, along with species-specific characteristics and the habitat and latitude in which the species grows (Reich & Oleksyn, 2004; Tian et al., 2017).

The roles of N and P in photosynthesis

Our data have provided support for the hypothesis that partitioning of N and P to different components of the photosynthetic pathway leads to N and P co-limitation in the generation of vegetative and seed biomass. We observed that N and P interactions led to increased chlorophyll content (Table S5F). Given that each chlorophyll molecule requires four atoms of N and no P, such a result nevertheless indicates a significant role of P in chlorophyll synthesis. It is clear that some components of photosynthesis are impacted by deficiencies in P alone (e.g. NPQ, qP, J_max, Table S6D,E,H respectively), while other components are impacted by both nutrients (e.g. V_Cmax, A_max and J_max, qP, NPQ), relationships that influence the generation of vegetative and seed biomass (see Figure 3 and Figure S3A,B). These results are consistent with Xu et al. (2007) and Rodríguez et al. (1999) who reported that A_max declined in P-deficient rice (Oryza sativa) and wheat, since in these scenarios, P would become the limiting nutrient. However, in a large meta-analysis using published data from 314 plant species, Reich et al. (2009) reported that A_max was influenced by both leaf N and interactions between leaf N and P, but not by leaf P alone. However, as stated above leaf N and P levels may be influenced by multiple factors. Lu et al. (2001) reported significant effects of N limitation on qP in maize, which appears to be inconsistent with our analysis of the dose of N and P in treatments (Figure 3b, Figure S3A,B). However, Lu et al.'s (2001) N-limitation experiments were performed under high P, and if maize is also regulating photosynthesis at N:P Redfield-like ratios as in wheat, then the results may be consistent if recalculated to express the amount of N and P added in terms of the ‘Effective dose’ based on a Redfield-like ratio.

We were surprised that F_v/F_m (a measure of functional photosystem II (PSII) complexes of the light reaction of photosynthesis) and rETR (a measure of electron transport rate through PSII) were unaffected by N and P availability. Probably the plants are modulating these and other aspects of their photosynthetic machinery to ensure reactions are optimized, irrespective of nutrient availability. While similar results have been reported for some species such as Oryza sativa (Veronica et al., 2017) and some climbers (Xing & Wu, 2014), studies of other species have reported F_v/F_m to be reduced under N and/or P deficiency, including in Eustoma grandiflorum (Chen et al., 2018) and Arachis hypogaea (Shi et al., 2020). Thus, the responses of these photosynthetic measures to N and P treatments may be species specific. Harnessing intra- and inter-specific genetic variation to enhance photosynthesis is likely to become important in crop breeding aiming to maximize productivity with lower N and P applications (Kromdijk & McCormick, 2022).

Experimental studies have previously shown that P deficiency limits activation of RuBisCO, ribulose-1,5-bisphosphate (RuBP) regeneration by RuBisCO and photosynthetic rates (Brooks et al., 1988). The inverse relationship shown here between qP (increases) and NPQ (decreases) when P is added (Figure 3) likely arises because P-limited plants have perturbations in ATP and NADP cycling, which renders photosynthesis less efficient in the conversion of light energy into chemical energy (Campbell & Sage, 2006). A consequence of P limitation is that excess light energy is shunted to NPQ, to protect cells from potential photo-oxidative damage. We observed elevated NPQ under P limitation in our experiments despite the wheat plants being grown under lower light conditions than would be experienced in the field, and we might expect in natural light that NPQ/qP interactions would be greater still. In support of these interpretations, Carstensen et al. (2018) showed that P-deficient barley (Hordeum vulgare) plants had increased levels of NADPH relative to ATP generation, resulting in a build-up of protons in the thylakoid lumen, which in turn signalled an increase in NPQ. Other studies have also shown that chlorophyll content and qP decreased in soybean (Glycine max) leaves when P limited (Singh & Reddy, 2016), consistent with the results here. Reducing chlorophyll content and qP and increasing NPQ would act to protect PSII from excess light damage when photosynthesis is compromised by nutrient limitation.

Best-fit SEMs, considering the actual dose of N and P added, showed direct effects of N limitation on V_cmax, A_max and J_max (Figure 3, Figure S3A,B), probably because high levels of N are needed for RuBisCO. In SEMs considering the ‘effective dose of N and P’, ‘Excess N’ and ‘Excess P’ beyond 21 N:1 P had no additional effects on the photosynthetic parameters qP, NPQ, A_max, V_cmax or J_max on vegetative or seed biomass (Figure 5 and Figure S3C,D). In these models, there was a significant direct effect of ‘Effective dose of N and P’ on biomass accumulation, which was not driven by N and P demands of photosynthesis. Indeed, these links were stronger than those via photosynthesis, indicative of substantial N and P demands of general metabolism and cell division. In these models, we also detected a significant positive correlation between qP and vegetative biomass, and in some models, also with seed biomass. It is likely this arises because qP provides ATP and NADP for general metabolism and cell division to facilitate biomass accumulation.

What is in a seed?

Wheat, like many other plants, reallocates nutrients from vegetative biomass to the seed. Seed production is therefore directly affected by the amount and quality of plant biomass produced. Our results confirm the hypothesis that N and P co-limitation influences the accumulation of vegetative and seed biomass. We demonstrate that the strength of the links is high, and is underpinned by N and P usage at the 21N:1P Redfield-like ratio (Figure 5b,d, Figure S3B). The role of N (e.g. Cui et al., 2010) and P (e.g. Grant et al., 2001) in the generation of seed biomass is well known, and previous research has shown that N input as fertilizer is important for grain yield (Edmeades, 2003) and seed composition. Nevertheless, the results here also show that ‘excess N’ and ‘excess P’ – beyond a 21N:1P Redfield-like ratio – have no significant role in seed biomass (Figure 5b,d, Figure S3D). The significant negative relationship we found between the carbon content in wheat seeds and increasing P treatment (Table S7G) is consistent with previous studies that have measured starch in soybean (Glycine max) and sugar beet (Beta vulgaris) grown under limiting P (Fredeen et al., 1989; Qiu & Israel, 1992; Rao & Terry, 1989). It has also been shown in wheat grown under reduced N fertilizer application that the amount of starch increased while the protein content decreased (Kindred et al., 2008). Such results suggest that nutrient availability can play a role in influencing the fate of the metabolic products of photosynthesis, such that under nutrient limitation, the photosynthetic products may be utilized for biomass production by cell division (which are N and P demanding) or moved to storage products (such as starch). It is perhaps the different allocation of nutrients to seeds depending on N and P treatment that rendered the link between photosynthesis and seed biomass accumulation weak.

CONCLUSIONS

This study highlights the importance of understanding how the application of N and P fertilizers at specific N:P ratios could impact above-ground biomass and seed biomass. We show that a 21N:1P molar ratio supports plant biomass production and suggests that the application of nutrients at this ratio may be important in wheat agriculture. However, compound fertilizers (or complex fertilizers) which are widely used globally vary considerably in their N:P ratios, although there are minimum nutrient requirements, for example, they should contain ≥3% N and ≥5% P₂O₅ and be ≥18% (by weight) of total nutrients (Dittmar et al., 2009). In widely used modern compound fertilizers, the N and P content is generally similar (e.g. N:P per cent mass ratios of 15%:15%–20%:10%, Dittmar et al., 2009), equivalent to N:P molar ratios of ≈2.1:1–4.3:1, which are much higher in relative P levels than the 21N:1P effective dose indicated by our analysis. Farmers are also offered a range of fertilizer N:P ratios, depending on the growth stage through the season and the crop being grown, with recommendations of, for example, plentiful P early in the season, and more N as the season progresses. It is potentially better to provide the plants with an optimal ratio of 21N:1P throughout the growth phase of the plants. Such a regime would reduce run-off and potentially save considerable P use and cost.

Total fertilizer consumption around the world has increased by ~10% from 177.2 million metric tonnes (Mt) in 2011/12 to 195.4 Mt in 2023/24 (Statistica, 2024b). The demand for N, P and K is broken down as 111.6 Mt, 47.1 Mt and 36.7 Mt, respectively, in 2023/24. Focusing on the United Kingdom specifically, fertilizer use in 2020 was 1.4 Mt (Statistica, 2024a) with N being the most commonly used element, making up almost 1 Mt of the total fertilizer use. This is compared to 174 thousand tonnes (kt) and 253 kt for P and K respectively. In 2022, UK farmers spent £1.62 billion on chemical fertilizers, which was £1.17 billion more than they had spent in 2020, making fertilizers the highest-growing agricultural input by price. This was the main driver of growing agricultural price inflation, with consumers paying the price. Our results suggest that globally we are using far too much P in comparison to N, and probably this is not being noticed because we are using too much of both leading to runoff, pollution and wastage.

We recommend that field trials with hexaploid wheat, of the sort performed by Mussarat et al. (2021), are conducted using applications of fertilizers at 21N:1P molar ratio. Ideally, N and P (with potassium) are applied together in the form of slow-release fertilizers, which increases N transporter gene activity (Li et al., 2023), to continually deliver the effective N and P dose (Ghosh et al., 2024). We recognize that the optimal Redfield-like N:P ratio may differ with species and be impacted by soil conditions (Koerselman & Meuleman, 1996). Phosphorus in particular has different mobilities in different soils, with different adsorptions of P to soil particles depending on soil type, especially when iron rich, such that the bio-available N:P may differ from a 21N:1P application ratio.

We recommend that similar analyses are conducted to those performed here on other crops. Different species may differ in the optimal Redfield-like N:P ratio that should be applied, not least because they show different nutrient-use efficiencies (Yu et al., 2021). Other aspects of the species physiology may also impact the optimal Redfield-like ratio for the species, for example, plant genome size may have an impact given that DNA is rich in P and N and must be packed with histone proteins that are also rich in N (Hessen et al., 2010).

Of course, hydroponic culture of plants is ideally placed to take advantage of efficiency gains in applying fertilizers at Redfield-like ratios. Field experiments with N and P applied at Redfield-like ratios will need to consider, in addition to soil type and climatic variables, such as temperature and growth, which interact with fertilizers to affect yield (Zhang et al., 2024). In field experiments, the complexity of the soil matters too, and yields are likely to be enhanced by applying fertilizers at appropriate Redfield-like ratios when used with, for example, biochar or with the application of composts (Kabato et al., 2022) or manures, which increases soil enzymatic activities and microbe abundance and improves soil structure and aeration (Du et al., 2020). Furthermore, excess N and P applications or imbalanced N and P are also likely to detrimentally impact mycorrhizal diversity (Liu et al., 2024).

If the data here are applicable to crop growth in the field, then the application of effective doses of fertilizers has the potential to dramatically reduce the input of P in particular. This is important because there are finite reserves of mineable P, which with current usage may become limiting in ~50 years' time (Hollas et al., 2021). In addition, N and/or P fertilizers used in excess are increasingly expensive to farmers and consumers, are environmentally damaging through run-off (Guignard et al., 2017) and contribute to greenhouse gas emissions (Ma et al., 2025).

AUTHOR CONTRIBUTIONS

ARL, IJL and MT conceived the experiment with advice from PJR, AVR and RAN. SSMS conducted the experiment with assistance from MSG and data were analysed by LF under the supervision of ARL and IJL with advice from MT and RAN. ARL led paper writing, with original material and editorial comments from IJL, AVR, LF, NO, PJR, MSG, MT and RAN.

ACKNOWLEDGEMENTS

This work was funded by PhD studentships for SS (Science without Borders, Brazil) and LF (Overseas Faculty Development Program, University of Swat, Pakistan). We thank Keith Edwards (University of Bristol) for the supply of seeds, Erik Murchie (University of Nottingham) for the use of CIRAS-1 and Phillip Davey (University of Essex) for its maintenance. We thank undergraduate students JJ Wolfensohn and CH Fantecelle for their help. We thank Felicity Shelley and Ian Sanders for their assistance with the Skalar and elemental analysis. We thank Nick Smith for his helpful comments in refereeing.

FUNDING INFORMATION

No funding was received to support this research or manuscript.

CONFLICT OF INTEREST STATEMENT

The authors have stated explicitly that there are no conflicts of interest in connection with this article.

DATA AVAILABILITY STATEMENT

The data that supports the findings of this study are available in the Supporting Information of this article.

References

Batten, G. D. (1992). A review of phosphorus efficiency in wheat. Plant and Soil, 146(1), 163–168.

Brooks, A., Woo, K. C., & Wong, S. C. (1988). Effects of phosphorus nutrition on the response of photosynthesis to CO₂ and O₂, activation of ribulose bisphosphate carboxylase and amounts of ribulose bisphosphate and 3‐phosphoglycerate in spinach leaves. Photosynthesis Research, 15(2), 133–141.

Campbell, C. D., & Sage, R. F. (2006). Interactions between the effects of atmospheric CO2 content and P nutrition on photosynthesis in white lupin (Lupinus albus L.). Plant, Cell & Environment, 29(5), 844–853.

Carstensen, A., Herdean, A., Schmidt, S. B., Sharma, A., Spetea, C., Pribil, M., & Husted, S. (2018). The impacts of phosphorus deficiency on the photosynthetic electron transport chain. Plant Physiology, 177(1), 271–284.

Chen, C.‐T., Lee, C.‐L., & Yeh, D.‐M. (2018). Effects of nitrogen, phosphorus, potassium, calcium, or magnesium deficiency on growth and photosynthesis of Eustoma. HortScience Horts, 53(6), 795–798.

Chowdhury, R. B., Moore, G. A., Weatherley, A. J., & Arora, M. (2017). Key sustainability challenges for the global phosphorus resource, their implications for global food security, and options for mitigation. Journal of Cleaner Production, 140, 945–963.

Cleveland, C. C., & Liptzin, D. (2007). C:N:P stoichiometry in soil: Is there a “Redfield ratio” for the microbial biomass? Biogeochemistry, 85(3), 235–252.

Cui, Z., Zhang, F., Chen, X., Dou, Z., & Li, J. (2010). In‐season nitrogen management strategy for winter wheat: Maximizing yields, minimizing environmental impact in an over‐fertilization context. Field Crops Research, 116(1), 140–146.

Deng, Y., Teng, W., Tong, Y.‐P., Chen, X.‐P., & Zou, C.‐Q. (2018). Phosphorus efficiency mechanisms of two wheat cultivars as affected by a range of phosphorus levels in the field. Frontiers in Plant Science, 9, 1614.

Dittmar, H., Drach, M., Vosskamp, R., Trenkel, M. E., Gutser, R., & Steffens, G. (2009). Fertilizers, 2. Types. Ullmann's Encyclopedia of Industrial Chemistry.

Du, Y., Cui, B., zhang, Q., Wang, Z., Sun, J., & Niu, W. (2020). Effects of manure fertilizer on crop yield and soil properties in China: A meta‐analysis. Catena, 193, [eLocator: 104617].

Duursma, R. A. (2015). Plantecophys ‐ an R package for analysing and modelling leaf gas exchange data. PLoS One, 10(11), [eLocator: e0143346].

Edmeades, D. C. (2003). The long‐term effects of manures and fertilisers on soil productivity and quality: A review. Nutrient Cycling in Agroecosystems, 66(2), 165–180.

Fang, Z., Li, D.‐D., Jiao, F., Yao, J., & Du, H.‐T. (2019). The latitudinal patterns of leaf and soil C:N:P stoichiometry in the Loess Plateau of China. Frontiers in Plant Science, 10, 85.

Farquhar, G. D., von Caemmerer, S., & Berry, J. A. (1980). A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta, 149(1), 78–90.

Farquhar, G., & Sharkey, T. D. (1982). Stomatal conductance and photosynthesis. Annual Review of Plant Physiology, 33(1), 317–345.

Fredeen, A. L., Rao, I. M., & Terry, N. (1989). Influence of phosphorus nutrition on growth and carbon partitioning in Glycine max. Plant Physiology, 89(1), 225–230.

Gao, Y., Li, Y., Zhang, J., Liu, W., Dang, Z., Cao, W., & Qiang, Q. (2009). Effects of mulch, N fertilizer, and plant density on wheat yield, wheat nitrogen uptake, and residual soil nitrate in a dryland area of China. Nutrient Cycling in Agroecosystems, 85(2), 109–121.

Ghosh, A., Singh, A. K., Kumar, R. V., Singh, P. D., Misra, S., Ahamed, S., Ojha, D., Chandra, A., & Bhattacharyya, R. (2024). Silica and polymer coated controlled release nitrogen‐phosphorus fertilizer for improving nutrient and water use efficiency in semi‐arid India. Journal of Environmental Chemical Engineering, 12(3), [eLocator: 112737].

Grant, C. A., Flaten, D. N., Tomasiewicz, D. J., & Sheppard, S. C. (2001). The importance of early season phosphorus nutrition. Canadian Journal of Plant Science, 81(2), 211–224.

Gruber, N., & Deutsch, C. A. (2014). Redfield's evolving legacy. Nature Geoscience, 7(12), 853–855.

Guignard, M. S., Crawley, M. J., Kovalenko, D., Nichols, R. A., Trimmer, M., Leitch, A. R., & Leitch, I. J. (2019). Interactions between plant genome size, nutrients and herbivory by rabbits, molluscs and insects on a temperate grassland. Proceedings of the Royal Society B: Biological Sciences, 286(1899), [eLocator: 20182619].

Guignard, M. S., Leitch, A. R., Acquisti, C., Eizaguirre, C., Elser, J. J., Hessen, D. O., Jeyasingh, P. D., Neiman, M., Richardson, A. E., Soltis, P. S., Soltis, D. E., Stevens, C. J., Trimmer, M., Weider, L. J., Woodward, G., & Leitch, I. J. (2017). Impacts of nitrogen and phosphorus: From genomes to natural ecosystems and agriculture. Frontiers in Ecology and Evolution, 5, [eLocator: 70]. [DOI: https://dx.doi.org/10.3389/fevo.2017.00070]

Sterner, R. W., & Elser, J. J. (2002). Ecological stoichiometry: The biology of elements from molecules to the biosphere. Princeton University Press.

Tang, Z., Xu, W., Zhou, G., Bai, Y., Li, J., Tang, X., Chen, D., Liu, Q., Ma, W., Xiong, G., He, H., He, N., Guo, Y., Guo, Q., Zhu, J., Han, W., Hu, H., Fang, J., & Xie, Z. (2018). Patterns of plant carbon, nitrogen, and phosphorus concentration in relation to productivity in China's terrestrial ecosystems. Proceedings of the National Academy of Sciences of the United States of America, 115(16), 4033–4038.

Tessier, J. T., & Raynal, D. J. (2003). Use of nitrogen to phosphorus ratios in plant tissue as an indicator of nutrient limitation and nitrogen saturation. Journal of Applied Ecology, 40(3), 523–534.

Tian, D., Yan, Z., Niklas, K. J., Han, W., Kattge, J., Reich, P. B., Luo, Y., Chen, Y., Tang, Z., Hu, H., Wright, I. J., Schmid, B., & Fang, J. (2017). Global leaf nitrogen and phosphorus stoichiometry and their scaling exponent. National Science Review, 5(5), 728–739.

Townsend, A. R., Cleveland, C. C., Asner, G. P., & Bustamante, M. M. C. (2007). Controls over foliar N:P ratios on tropical rain forests. Ecology, 88(1), 107–118.

van de Wiel, C. C. M., van der Linden, C. G., & Scholten, O. E. (2016). Improving phosphorus use efficiency in agriculture: Opportunities for breeding. Euphytica, 207(1), 1–22.

Veronica, N., Subrahmanyam, D., Vishnu Kiran, T., Yugandhar, P., Bhadana, V. P., Padma, V., Jayasree, G., & Voleti, S. R. (2017). Influence of low phosphorus concentration on leaf photosynthetic characteristics and antioxidant response of rice genotypes. Photosynthetica, 55(2), 285–293.

Wang, L., Arif, M., Zheng, J., & Li, C. (2024). Patterns and drivers of plant carbon, nitrogen, and phosphorus stoichiometry in a novel riparian ecosystem. Frontiers in Plant Science, 15, [eLocator: 1354222].

Wang, Y. P., Law, R. M., & Pak, B. (2010). A global model of carbon, nitrogen and phosphorus cycles for the terrestrial biosphere. Biogeosciences, 7(7), 2261–2282.

Xing, D., & Wu, Y. (2014). Effect of phosphorus deficiency on photosynthetic inorganic carbon assimilation of three climber plant species. Botanical Studies, 55(1), 60.

Xu, H. X., Weng, X. Y., & Yang, Y. (2007). Effect of phosphorus deficiency on the photosynthetic characteristics of rice plants. Russian Journal of Plant Physiology, 54(6), 741–748.

Yokamo, S., Irfan, M., Huan, W., Wang, B., Wang, Y., Ishfaq, M., Lu, D., Chen, X., Cai, Q., & Wang, H. (2023). Global evaluation of key factors influencing nitrogen fertilization efficiency in wheat: A recent meta‐analysis (2000–2022). Frontiers in Plant Science, 14, [eLocator: 1272098].

Yu, X., Keitel, C., & Dijkstra, F. A. (2021). Global analysis of phosphorus fertilizer use efficiency in cereal crops. Global Food Security, 29, [eLocator: 100545].

Zhang, P., Sadras, V. O., Zhang, R., Liu, L., Yang, X., Sun, B., Hu, C., Xu, H., & Zhang, S. (2024). Responses of winter wheat yield and soil organic carbon to long‐term (1990–2021) fertilization regimes under inter‐annual weather variation in the Loess Plateau. European Journal of Agronomy, 157, [eLocator: 127189].