1 Introduction

As an essential micronutrient, zinc (Zn) is required for the health of both crops and humans, but Zn deficiency is currently a widespread problem in human nutrition (Stein 2010; Cakmak and Hoffland 2012; Gibson 2012). This is especially true for the populations in developing countries, who rely on cereal grains as staple foods (Rengel et al. 1999; Palmgren et al. 2008; Cakmak 2009). Achieving a sufficient Zn intake from consumption of cereal grains is therefore important. Phosphorus (P) fertilizer application has increased cereal grain yield in the last few decades (Roy et al. 2016), but has globally reduced grain Zn concentration irrespective of crop species (Christensen and Jackson 1981; Verma and Minhas 1987; Biswapati and Mandal 1990; Abbas et al. 2007; Amanullah 2016; Chen et al. 2017; Zhang et al. 2017a). As a consequence, any attempts to increase grain Zn availability should consider the potentially negative effect of P fertilizer.

Previous studies have evaluated the influence of P fertilizer application on cereal grain Zn concentrations, usually with specific conditions and single experiments (Friesen et al. 1980; Debnath et al. 2015; Drissi et al. 2015; Dang et al. 2016; Coetzee et al. 2017; Iqbal et al. 2017; Sánchez-Rodríguez et al. 2017). The conclusions regarding the P–Zn relationship have sometimes been quite different. Besides finding an antagonistic effect of P on cereal grain Zn concentrations, researchers have found a non-effect (Su et al. 2018) or a slightly synergistic effect (Iqbal et al. 2017; Naeem et al. 2018), and these differences might be due to variation in soil texture, pH, or other soil properties (Haldar and Mandal 1981; Goh et al. 1997; Grant et al. 2002; Gao et al. 2011; Ghasemi-Fasaei and Mayel 2012; Hagh et al. 2016). In recent years, a number of studies on the effects of P application on cereal grain Zn have greatly increased. This increase has made it feasible to use a meta-analysis to quantify the negative/positive effect of P application on grain Zn concentration and to identify the underlying mechanisms.

Many studies have focused on the P–Zn antagonism and have proposed various mechanisms to explain the antagonism (Ragab 1980; Singh et al. 1986; Totawat and Saeed 1990; Zhao et al. 2007; Zan 2012; Vafaei and Sarraf 2014; Smith et al. 2017). These proposed mechanisms include reductions in Zn availability in soil, Zn concentration by roots, Zn translocation from roots to other tissues, and Zn remobilization from vegetative organs to grain (Fig. 1a). As the main cereal crops, wheat, maize, and rice play an important role in providing dietary Zn intake for human (Fig. 1 b, c, and d). Some studies found that Zn availability in soil was reduced by P application (Adnan 2016) because P fertilizer additions enhanced Zn adsorption by increasing the negative charges on the surface of the iron and aluminum oxides (Saeed and Fox 1979). Other reports showed, however, that Zn availability in soil was not affected or was even slightly enhanced by P fertilizer application (Takkar et al. 1976; Reddy and Yadav 1994; Zhang et al. 2012). In a pot experiment, the rate of Zn uptake per unit fresh weight of maize roots was reduced by P application (Safaya 1976). Grain Zn accumulation is also affected by the translocation and remobilization of Zn within the plant (Haslett et al. 2001; Pearson and Rengel 1994). Our previous field studies with wheat and maize on the North China Plain indicated that root-to-shoot Zn translocation and shoot-to-grain Zn remobilization efficiency from source to sink tissues were not affected by P application (Zhang et al. 2015, 2016). Overall, these results suggest that a clear understanding of the effects of P application on Zn uptake, translocation, and remobilization is still missing.

Fig. 1
figure 1

The possible mechanism of P affecting Zn transportation from soil to grain (a) in cereal crops: winter wheat (b), summer maize (c), and rice (d). The crops of winter wheat and summer maize are planted in Quzhou county in China, and rice is planted in Chongqing in China. Photographs by Wei Zhang

Full size image

Root concentration of Zn has been considered a key process determining the Zn concentration in aboveground plant parts (Zhang et al. 2017b). Root morphology and rhizosphere processes can affect Zn acquisition. Zinc uptake efficiency in rice cultivars, for example, is closely associated with root length, root volume, and root surface area (Chen et al. 2009). In addition, a quantitative meta-analysis clearly showed that root colonization by arbuscular mycorrhizal fungi (AMF) can greatly increase crop Zn concentrations (Lehmann et al. 2014). The mycorrhizal pathway of Zn uptake by roots contributes to Zn accumulation in wheat grain (Coccina et al. 2019; Watts-Williams et al. 2015; Watts-Williams et al. 2014). Under P deficiency in maize, root exudation of organic acids increased (Gaume et al. 2001; Hinsinger et al. 2003), which can increase soil Zn availability (Duffner et al. 2012). In contrast, a higher P fertilizer application, which exceeded the critical P application for maximal value of root morphology, can decrease root dry weight, root length, and root surface area (Wen et al. 2017), as well as rhizosphere processes including root colonization by AMF (Deng et al. 2017) and root exudation (Shen et al. 2002; Gaume et al. 2001). It remains unclear whether P application induces changes in root morphology and physiology that further influence the uptake and accumulation of Zn in the grain of cereal crops.

Crop species markedly differ in their capacity for Zn uptake in response to P fertilizer addition perhaps because the effect of P fertilization on root and rhizosphere properties differs among crops. On a P-deficient soil, for example, a high rate of root colonization by AMF increased P uptake by maize (Itoh & Barber 1983; Jemo et al. 2014; Lyu et al. 2016; Deng et al. 2017; Wen et al. 2017), whereas a large amount of root dry weight was more important for P uptake by rice (Chin et al. 2011). These root and rhizosphere traits in different crops may further influence P and Zn uptake by roots. A quantitative evaluation of the effect of P application on the Zn nutrition of the major cereal crops is therefore necessary.

In the current study, we conducted a meta-analysis to test two hypotheses: (1) P fertilizer application does not affect Zn translocation efficiency or remobilization efficiency but reduces Zn uptake efficiency by roots and (2) especially in wheat and maize, P fertilizer application reduces root Zn uptake efficiency in part by reducing AMF colonization of roots. To test these hypotheses, we conducted a meta-analysis in order to quantitatively evaluate the effects of P fertilizer application on grain Zn of the main cereal crops (wheat, maize, and rice). We also used the analysis to explore the possible mechanisms controlling Zn mobilization from soil to grain and to explore the possible reasons for differences among the three crops in P–Zn antagonism.

2 Materials and methods

2.1 Literature search

We conducted a literature search using the ISI Web of Knowledge (http://apps.webofknowledge.com) and the China National Knowledge Infrastructure database (CNKI, http://www.cnki.net) to collect the peer-reviewed journal articles published before March 2019. By using search terms phosphorus* AND zinc* AND (wheat OR maize OR corn OR rice), we collected a total of 51 articles (46 in English and 5 in Chinese), which were conducted at 43 locations globally. Among the 51 articles, 25, 16, and 6 solely concerned wheat, maize, and rice crops, respectively; 3 considered both wheat and maize crops; and 1 considered both rice and wheat crops. Our analysis included reports concerning all types of wheat (spring wheat, winter wheat, and durum wheat), maize, and rice and reports from both field and reports greenhouse experiments. Among the 51 articles, 28 described field studies, 22 described greenhouse studies, and 1 described a field and greenhouse study. Data were further scrutinized and extracted using the following inclusion criteria: (1) studies should include pair-wise control (no P fertilizer application) and P treatments (P fertilizer added) such that the P treatments have the same indicators as the control; (2) crop species were wheat, maize, or rice; (3) if one paper reported multiple independent experiments (e.g., two experiments at separate locations, years, and crops), each was considered an individual study and was incorporated as an independent observation in our dataset; and (4) Zn concentration in grain, straw, shoots, or roots was reported. Zn and P concentrations in grain, straw, and roots were measured as mg Zn kg−1 and g P kg−1, respectively. In this report, Zn “content” is equal to Zn concentration multiplied by the dry weight of the indicated plant part. Data, i.e., means, standard deviations, standard errors, and number of replicates were collected from tables and figures by using GetData Graph Digitizer (version 2.25).

The following information was documented for each study: crop; Zn concentrations and dry weights in grain, straw, and roots; Zn content in grain and straw; AMF colonization; organic acid exudation; experiment site; experiment type; year; soil type; soil P concentration (Olsen-P, Bray-P, total P); soil Zn concentration (DTPA-Zn, total Zn, and water soluble Zn); and Zn application. The term “shoot” refers to aboveground plant parts. The term “shoot Zn content” refers to the sum of the aboveground Zn content in grain and straw at crop maturity.

In this study, root Zn concentration was considered an indicator of Zn uptake efficiency, and the shoot-to-root Zn ratio and grain-to-straw Zn ratio are used to calculate the translocation efficiency of Zn from roots to shoots and the remobilization efficiency of Zn from straw to grain as shown below:

$$ \mathrm{shoot}-\mathrm{to}-\mathrm{root}\ \mathrm{Zn}\ \mathrm{ratio}=\mathrm{shoot}\ \mathrm{Zn}\ \mathrm{concentration}/\mathrm{root}\ \mathrm{Zn}\ \mathrm{concentration} $$
$$ \mathrm{grain}-\mathrm{to}-\mathrm{straw}\ \mathrm{Zn}\ \mathrm{ratio}=\mathrm{grain}\ \mathrm{Zn}\ \mathrm{concentration}/\mathrm{straw}\ \mathrm{Zn}\ \mathrm{concentration} $$

ΔGrain Zn concentration and Δsoil Zn concentration (DTPA-Zn, total Zn, and water soluble Zn) are calculated by the following equations in order to analyze the change in values caused by P application compared to the control:

$$ \Delta \mathrm{grain}\ \mathrm{Zn}\ \mathrm{concentration}=\mathrm{grain}\ \mathrm{Zn}\ \mathrm{concentration}\ \mathrm{with}\ \mathrm{P}\ \mathrm{application}-\mathrm{grain}\ \mathrm{Zn}\ \mathrm{concentration}\ \mathrm{with}\mathrm{out}\ \mathrm{P}\ \mathrm{application} $$
$$ \Delta \mathrm{soil}\ \mathrm{available}\ \mathrm{Zn}\ \mathrm{concentration}=\mathrm{soil}\ \mathrm{available}\ \mathrm{Zn}\ \mathrm{concentration}\ \mathrm{with}\ \mathrm{P}\ \mathrm{application}-\mathrm{soil}\ \mathrm{available}\ \mathrm{Zn}\ \mathrm{concentration}\ \mathrm{with}\mathrm{out}\ \mathrm{P}\ \mathrm{application} $$

2.2 Data analysis

In the meta-analysis, the natural logarithm of the response ratio (lnR) is calculated as the effect size (Hedges et al., 1999), i.e., the effect of P treatments on Zn concentration in crops, using the following equations:

$$ \mathrm{R}=\frac{\mathrm{Xt}}{\mathrm{Xc}} $$
$$ \mathrm{lnR}=\ln \left(\frac{\mathrm{Xt}}{\mathrm{Xc}}\right)=\ln \left(\mathrm{Xt}\right)-\ln \left(\mathrm{Xc}\right) $$

where Xt is the mean of Zn concentration in grain, straw, or roots for the P treatment and Xc is the mean of those concentrations for the no P control. Mean effect sizes and bias-corrected 95% confidence intervals (CIs) were generated using a bootstrapping procedure (4999 iterations). To facilitate the interpretation, the percentage of change in Zn concentration of crops in the P treatment relative to no P control is calculated by the following equation:

$$ \mathrm{Change}\ \mathrm{in}\ \mathrm{Zn}\ \mathrm{concentration}\ \left(\%\right)=\left(\mathrm{R}-1\right)\times 100\%=\left(\frac{\mathrm{Xt}-\mathrm{Xc}}{\mathrm{Xc}}\right)\times 100\% $$

A positive value indicated an increase in Zn concentration of crops under P treatment relative to the control, while a negative value indicated a decrease. The mean percentage change was considered significantly positive or negative when the 95% CI did not overlap with zero. In addition, the frequency distributions of effect sizes were plotted to reflect the distribution regularities of individual studies. The frequencies of effect sizes were also fitted to a Gaussian distribution function to test the homogeneity of observations. The effect of P fertilizer application was considered significant if the 95% CI did not overlap with zero. Means of categorical variables were considered significantly different if their 95% CIs did not overlap with each other.

2.3 Statistical analysis

The effect sizes fit a normal distribution (range from − 1.28 to 0.44, P > 0.05) according to Kolmogorov-Smirnov analysis, suggesting that the data were suitable for a meta-analysis. Mean effect size, changes in Zn concentration, and corrected bias (i.e., the 95% confidence intervals [CI]) for each category generated using bootstrapping (10,000 iterations) were calculated using a mixed-effect model with SPSS 13.0 (SPSS Inc., Chicago, IL, USA) and SigmaPlot 12.5 (Systat, San Jose, CA, USA) software. One-way analyses of variance (ANOVAs) were conducted to evaluate the treatment effect in each figure, and means were compared using the least significant difference (LSD) at a 5% level of probability, using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).

3 Results and discussion

3.1 Effect of P application on dry weight and Zn concentration in grain, straw, and roots

Phosphorus application increased the dry weight of grain, straw, and roots (Fig. 2). Averaged across the three crops, P application increased the dry weight relative to the control by 32.8% for grain, 39.2% for straw, and 49.5% for roots. For grain dry weight, the increase caused by P application was 35.7% for wheat, 25.8% for maize, and 14.7% for rice (Fig. 2a). The increase in straw and root dry weight was greater for wheat and maize than for rice (Fig. 2 b and c).

Fig. 2
figure 2

Changes in the dry weight of grain (a), straw (b), and roots (c) of three cereal crops in response to P application compared to the control. The values are means with 95% bootstrap confidence intervals, and sample sizes are in parentheses. The zero point indicates no effect in dry weight of grain, straw, and root of crops under P treatment relative to the control. Means with confidence intervals that do not overlap zero indicate that the dry weight was significantly affected by P application

Full size image

Table 1 showed that P fertilizer application decreased Zn concentration in all tissues. The grain Zn concentrations were reducing from 37.7, 36.3, and 22.0 mg kg−1 under no P application to 30.8, 25.2, and 20.6 mg kg−1 under P application of wheat, maize, and rice, respectively. In terms of the Zn reduction degree, across all 51 studies, we found an overall negative effect of P application on grain Zn concentration (− 16.0%, CI = − 17.8% to − 14.2%; Fig. 3a). The reduction in grain Zn concentration caused by P application was 16.6% for wheat and 20.2% for maize; P application, however, did not significantly affect the grain Zn concentration in rice. Phosphorus application reduced straw Zn concentration by an average of 21.9% across the three crops; the reduction was 24.2% for wheat and 28.2% for maize (Fig. 3b), but the P application did not significantly affect straw Zn concentration in rice. The reduction in root Zn concentration caused by P application was 15.8% for wheat and 8.22% for maize (Fig. 3c), and once again, P application did not significantly affect the root Zn concentration for rice. P application increased Zn content in grain, straw, and root of wheat and rice, but had no effect or slightly decreasing tendency on Zn content of maize (Fig. 3 d, e, and f). Compared with the control, P application resulted in an overall increase of 6.75% in shoot Zn content for all crops; the increase was 12.5% for wheat and 10.8% for rice (Fig. 3g); P application had no effect on shoot Zn content for maize.

Table 1 Effects of No P and P application on Zn concentrations in grain, straw, and root of wheat, maize, and rice
Full size table
Fig. 3
figure 3

Changes in grain Zn concentration (a) and content (d), straw Zn concentration (b) and content (e), root Zn concentration (c) and content (f), and shoot Zn content (g) of the three cereal crops in response to P application compared to the control. The shoot Zn content includes the total Zn content of grain and straw at crop maturity; as explained in the text, Zn content is determined by multiplying the dry weight of the plant component by the concentration of Zn in the plant component. Values are means with 95% bootstrap confidence intervals, and sample sizes are in parentheses. The zero point indicates no effect in Zn concentration and content of crops under P treatment relative to the control. Means with confidence intervals that do not overlap zero indicate that the Zn concentration or content was significantly affected by P application. No data is available for rice in (f)

Full size image

This global-scale meta-analysis provided quantitative evidence that despite increasing grain yield, P application reduced cereal grain Zn concentration. This negative effect of P fertilizer application potentially hinders Zn biofortification progress worldwide, especially for wheat and maize, for which the reduction in grain Zn was greater than for rice. The difference between rice and the other two crops has several potential explanations. First, the mobilization of P and Zn is greater in paddy soil than in dryland soil (Faye et al. 2006). Second, the decrease in AMF colonization due to P application was greater for maize and wheat roots than for rice roots. A previous study indicated that colonization of rice roots by AMF is rare due to the anoxic environment under the flooded conditions (Ilag et al. 1987).

To our knowledge, this is the first study to use meta-analysis method to quantitatively evaluate Zn concentration reduction extent due to P fertilizer application in grain of wheat, maize, and rice crops. We speculate that the reduction in grain Zn concentration for wheat, maize, and rice by P application may be influencing human health because high inputs of P fertilizer have been extensively reported for the intensive production of these three crops (Vitousek et al. 2009; MacDonald et al. 2011; Conley and Likens 2009). This concern with human health is particularly relevant for people who rely on wheat or maize as staple crops. In the case of wheat, it is reported that an adult human could consume 300 g grain d−1, and that only 20% of whole grain Zn will be finally absorbed by the human intestine (Rosado et al., 2009). On the basis of grain Zn concentration in the present study (Table 1), 300 g d−1 of wheat grain with no P application treatment could provide 2.26 mg Zn d−1; however, humans who eat wheat grain from crops receiving P fertilizer as main food only obtain 1.84 mg Zn d−1. This suggests that P fertilizer application indirectly decreases human Zn intake by 19% compared to no P application. The study therefore indicates that production areas with high rates of P fertilizer application should pay close attention to P–Zn antagonism. With respect to human nutrition and health, the consumption of wheat flour as a staple food may result in higher risk of low Zn intake than the consumption of rice.

3.2 Which factors determine grain Zn concentration?

Phosphorus application had no effect on the grain-to-straw Zn ratio (Fig. 4 a, b, c, and d) or shoot-to-root Zn ratio (Fig. 4e, f, g, and h) for any of the three crops. Compared to the control, P application also had no effect on the soil available Zn concentration for any of the three crops (Fig. 4i, j, k, and l). P application, however, reduced root colonization by AMF by 31.5% for wheat and by 41.5% for maize, but did not affect AMF colonization of rice (Fig. 5a). Phosphorus application generally did not affect root exudation of organic acids (Fig. 5b).

Fig. 4
figure 4

Effects of P application on grain-to-straw Zn ratio (a overall; b wheat; c maize; d rice), shoot-to-root Zn ratio (e overall; f wheat; g maize; h rice), and soil available Zn (i overall; j wheat; k maize; l rice). Sample sizes are in parentheses. The solid line through the box indicates the median, and the dotted line indicates the mean. Box boundaries indicate upper and lower quartiles, whisker caps indicate 95th and 5th percentiles, and black circles indicate outliers

Full size image
Fig. 5
figure 5

The effects of P application relative to the control on changes in AMF colonization (a) and on changes in root exudation of organic acid (b) for wheat, maize, and rice. Values are means with 95% bootstrap confidence intervals, and sample sizes are in parentheses. The zero point indicates no effect in AMF colonization and organic acid of crops under P treatment relative to the control. Means with confidence intervals that do not overlap zero indicate that the changes in AMF colonization or changes in root exudation of organic acid were significantly affected by P application

Full size image

For all crops, Δgrain Zn concentration was not related to Δsoil available Zn concentration (Fig. 6a), Δshoot-to-root Zn ratio (Fig. 6c), or Δgrain-to-straw Zn ratio (Fig. 6d). In contrast, Δgrain Zn concentration in wheat and maize increased linearly with Δroot Zn concentration (Fig. 6b).

Fig. 6
figure 6

The relationship between Δ grain Zn concentration and a Δ soil available Zn concentration, b Δ root Zn concentration, c Δ shoot-to-root Zn ratio, and d grain-to-straw Zn ratio for wheat (circles), maize (triangles), and rice (squares). In (b), the regression equation is y = − 7.26 + 2.20*x (R2 = 0.96) for wheat and y = − 2.59 + 0.32*x (R2 = 0.25) for maize. No data is available for rice in (a), (b), and (c)

Full size image

Previous attempts to explain the P–Zn relationship focused on four processes affecting the movement of Zn from soil to grain: availability of soil Zn, Zn uptake efficiency by roots, Zn translocation from roots to shoots, and Zn remobilization from vegetative organs to grain (Lambert et al. 1979; Lu and Miller 1989; Jain and Dahama 2006; Kizilgoz and Sakin 2010; Jin et al. 2014; Khan et al. 2014; Imran et al. 2015; Li et al. 2015). The interaction between P and Zn in the soil did not explain the negative effect of P fertilizer on cereal grain Zn concentration in the current global-scale meta-analysis. The level of soil available Zn was not affected by P application (Fig. 4 i, j, k, and l), which was consistent with previous individual study (Bogdanovic et al. 1999). In fact, the results indicated that the soil P levels in the present agricultural systems were far lower than the level required to decrease soil available Zn (Chen et al. 2019). Consistent with results of Zhang et al. (2016), our meta-analysis revealed that P fertilizer application did not affect the root-to-shoot transport of Zn, i.e., Zn transport from roots to shoots does not appear to contribute to the decline of grain Zn concentration in response to P application. The grain-to-straw ratio of Zn was also not affected by P application, indicating that P application did not affect Zn remobilization. That result was consistent with our previous finding that Zn remobilization efficiency was not affected by P application rate (Zhang et al. 2015).

In addition, in regarding to the reasons of P–Zn antagonistic, the “dilution effect” also causes decrease in Zn concentration (Racz and Haluschak 1974; Moraghan 1984; Orabi et al. 1985; Maftoun and Moshiri. 2010; Lu et al. 2011; Mai et al. 2011a; Mai et al. 2011b; Zhang et al. 2017c). This study showed that the dry weight and Zn contents were slightly increasing with increasing P application, while Zn concentration was continually decreasing (Fig. 3). Under the low available P in soil, increasing P fertilizer rapidly enhances crops’ biomass or yield, and the “dilution effect” maybe plays an important role in reduction of Zn concentrations, especially when the amount of P fertilizer increases from less to appropriate, whereas increasing P fertilizer from optimal to excessive has little or no effect on further improvement of crops biomass or yield, but Zn concentration is continuous declination (Zhang et al. 2012). Meanwhile, the increased yield was associated with different trends in the concentrations of Fe, Cu, and Mn in grain compared to Zn concentrations (Zhang et al. 2012). This meant that it could not be only explained by the “dilution effect” but the limitation of Zn acquisition capability by root. Together, these results indicate that the negative effect of P application on grain Zn concentration is not due to reductions in available soil Zn or to translocation or remobilization of Zn within plants.

Zinc concentrations in roots of wheat and maize were significantly decreased by P application. Compared to the no P treatment, application of P fertilizer increased wheat grain dry weight by 35.7% and wheat shoot Zn content by 12.5%. Together, the decrease in Zn concentration in wheat roots, the substantial increase in wheat grain weight, and the smaller increase in wheat shoot Zn content suggest that Zn uptake efficiency by wheat roots is limited by P fertilizer application. The results confirm that the reduction in root Zn concentration caused by P application cannot be attributed to a restriction of root growth but instead can be attributed to the “dilution effect” (an increase in root biomass without a concomitant increase in root Zn content) and to a reduction in Zn uptake efficiency. A previous study also indicated that P application could restrict Zn concentration by roots (Zhang et al. 2016). Some physiological factors, including root dry weight, root colonization by AMF, and root exudation of organic acids, have been reported to indirectly affect Zn concentrations following P application (Hoffland et al. 2006; Lambers et al. 2006; Cavagnaro et al. 2010; Tian et al. 2012). The current global-scale meta-analysis is consistent with these previous reports.

Phosphorus application could potentially decrease root Zn concentration by decreasing root colonization by AMF and by decreasing root exudation of organic acids. Our meta-analysis indicated that P application reduced AMF root colonization, which is consistent with previous reports (Teng et al. 2013; Thompson et al. 2013). According to previous reports, AMF increase root uptake of P and metal elements (e.g., Zn, Cu, and Fe), which indicated that an increase in AMF root colonization can increase root uptake of Zn and P (Kothari et al. 1991; Smith 2003; Cavagnaro 2008; Ryan et al. 2008; Smith et al. 2011). A previous meta-analysis also reported that AMF increased Zn concentrations in various crops under P-deficient conditions (Lehmann et al. 2014). In wheat, P-deficient conditions caused the roots to increase their exudation of organic acids, which can increase P and Zn availability in the rhizosphere (Khademi et al. 2010). It is well indicated that organic acid exudation is able to mobilize Zn in soil. For instance, Rose et al. (2011) proposed that for rice, enhanced malate exudation was a response of Zn efficient rice genotypes to Zn deficiency. Another study also proved that carboxylates exuded by root increased Zn mobilization from a calcareous soil (Degryse et al., 2008), whereas the present study showed that P application reduced total organic acid exudation which was consistent with Shen et al. (2002). The results therefore indicated that P application could potentially affect rhizosphere processes and thereby decrease root Zn concentration.

3.3 Attaining both high grain yields and high grain Zn concentrations

We therefore suggest that the negative effect of P application on grain Zn concentration might be reduced by field management practices that alter root and rhizosphere properties, i.e., that increase root dry weight, AMF colonization, and organic acid exudation. In intensive agricultural production systems, the quantities of P fertilizer applied have frequently exceeded crop requirements in some regions (Vitousek et al. 2009). The quantity of P fertilizer applied, however, should be one that achieves both high grain yield and high grain Zn concentration. An optimal level of P fertilizer should be established based on soil P tests, crop requirements, and on data concerning the effects of P supply on crop roots and rhizosphere properties (Shen et al. 2013; Teng et al. 2013; Deng et al. 2017; Zhang et al. 2018). Teng et al. (2013) found that the optimal rate of P application for maximum grain yield was at or near the rate that began to negatively affect root and rhizosphere properties. In addition, crop breeding programs should attempt to improve root and rhizosphere properties (e.g., AMF colonization and organic acid exudation) so as to achieve both a high P-use efficiency (Hinsinger et al. 2005; Veneklaas et al. 2012) and a high grain Zn concentration. For instance, over-expression of PSTOL1 (phosphorus-starvation tolerance 1), which increases early root growth in rice, significantly increases grain yield in P-deficient soil by enabling plants to acquire more P and other nutrients (Gamuyao et al. 2012). In another example, the genes STOP1 (a transcription factor) and ALMT1 (a malate transporter) underlie a malate exudation-dependent mechanism of Fe relocation in the root apical meristem and are essential for reprogramming root growth under low-P conditions (Mora-Macías et al. 2017). The current meta-analysis clearly shows that biofortification for increasing grain Zn concentration, whether via agronomic management or breeding, should account for the effects of P application on root and rhizosphere properties.

4 Conclusion

According to our meta-analysis, P fertilizer application resulted in an overall reduction of 16.0% in grain Zn concentration of wheat, maize, and rice. Zn content in grain, straw, and root of wheat and rice increased with increasing P application except for maize. Zinc mobilization processes, including soil available Zn, Zn translocation, and Zn remobilization, were not affected by P fertilizer supply and did not contribute to the reduction of grain Zn concentration. Root Zn concentration was substantially reduced by P fertilizer application and played an important role in decreasing grain Zn concentration. Our analysis indicated that P application decreased root Zn concentration by the “dilution effect” due to increasing root dry weight, and by altering root and rhizosphere properties including root colonization by AMF. P application increased wheat and maize biomass more than rice biomass, partly explaining why grain Zn concentration was reduced more in wheat and maize than in rice. Our meta-analysis is the first to assemble and assess quantitative data demonstrating a negative effect of P application on grain Zn concentration for wheat, maize, and rice. Our analysis is also the first to use meta-analysis to indicate that antagonism between P and Zn can be explained by the “dilution effect” and root and rhizosphere properties.