• In maize (Zea mays), nitrogen (N) remobilization and postflowering N uptake are two processes that provide amino acids for grain protein synthesis.
• To study the way in which N is allocated to the grain and to the stover, two different 15N-labelling techniques were developed. 15NO3− was provided to the soil either at the beginning of stem elongation or after silking. The distribution of 15N in the stover and in the grain was monitored by calculating relative 15N-specific allocation (RSA).
• A nearly linear relationship between the RSA of the kernels and the RSA of the stover was found as a result of two simultaneous N fluxes: N remobilization from the stover to the grain, and N allocation to the stover and to the grain originating from N uptake.
• By modelling the 15N fluxes, it was possible to demonstrate that, as a consequence of protein turnover, a large proportion of the amino acids synthesized from the N taken up after silking were integrated into the proteins of the stover, and these proteins were further hydrolysed to provide N to the grain.
Nitrogen (N) is one of the main limiting factors for plant growth and ultimately for the production of harvestable plant material used for animal and human food. In most plant species examined so far, the plant life cycle with regard to the management of N can be roughly divided into two main phases occurring successively. During the first phase, i.e. the vegetative phase, young developing roots and leaves behave as sink organs for the assimilation of inorganic N and the synthesis of amino acids. These amino acids are further used for the synthesis of enzymes and proteins mainly involved in building up plant architecture and the different components of the photosynthetic machinery. Notably, the enzyme Rubisco can alone account for up to 50% of the total soluble leaf protein content (Mae et al., 1983). Later, at a certain stage of plant development generally starting after anthesis, the remobilization of N takes place. At this stage, shoots and/or roots start to behave as sources of N by providing amino acids released from protein hydrolysis, which are subsequently exported to reproductive and storage organs represented, for example, by seeds, bulbs or trunks (Masclaux et al., 2001). In maize, 45–65% of the grain N is provided from pre-existing N in the stover before silking, a process that is strongly dependent upon the environmental conditions and/or the genotype (Weiland & Ta, 1992; Gallais & Coque, 2005). The remaining 35–55% of the grain N originates from postsilking N uptake (Ta & Weiland, 1992; Bertin & Gallais, 2000; Gallais & Coque, 2005). Therefore, two distinct N fluxes must be considered during the grain-filling period: N translocation from the stover (stalks + leaves + sheaths + cob) and N uptake (Crawford et al., 1982; Rendig & Crawford, 1985; Ta & Weiland, 1992).
However, from the point of view of N management at the whole-plant or organ level, the arbitrary separation of the plant life cycle into two phases is rather simplistic, as it is well known that, for example, N recycling can occur before anthesis for the synthesis of new proteins in developing organs (Lattanzi et al., 2005). In addition, during the assimilatory phase, the ammonium incorporated into free amino acids is subjected to a high turnover, as a result of photorespiratory activity, as it needs to be immediately re-assimilated into glutamine and glutamate (Hirel & Lea, 2001; Novitskaya et al., 2002). Therefore, the photorespiratory flux of ammonium, which, at least in C3 plants, can be 10 times higher than that originating from nitrate reduction, is mixed with that channelled through the inorganic N assimilatory pathway (Novitskaya et al., 2002). Although in C4 plants photorespiration is low, this process is likely to occur in maize (Zea mays) (Taiz & Zeiger, 2002; Ueno et al., 2005). Furthermore, Liu & Jagendorf (1984) and Malek et al. (1984) found that in pea (Pisum sativum) approx. 30% of the labelled amino acids newly incorporated into proteins by isolated chloroplasts were lost during the next 30 min, after a chase with cold amino acids. Therefore, the occurrence of protein turnover concomitantly with the two fluxes of ammonium (from assimilatory and photorespiratory fluxes) introduces another level of complexity in the exchange of N within the pool of free amino acids. The coexistence of these different N fluxes has led us to reconsider the mode by which N is managed from the cellular level to the level of the whole plant (Hirel & Gallais, 2006).
Despite the complexity of the N fluxes, a number of attempts have been made to estimate globally the amount of N remobilized to the grain. To achieve this, methods based on the calculation of N budgets, termed the ‘balance method’, or techniques using 15N-labelling have been used. The ‘balance method’ allows estimation of the amount of N remobilized by calculating the difference between total plant N (stalks + leaves) at silking and total N in the stover at grain maturity. This method is based on the assumption that all the newly synthesized amino acids originating from N uptake are directly allocated to the grain (Moll et al., 1982; Di Fonzo et al., 1993; Rajcan & Tollenaar, 1999; Bertin & Gallais, 2000). However, this method does not take into account the possibility that a significant amount of the newly synthesized amino acids may be allocated to the stover before leaf N remobilization, as already shown by Ta & Weiland (1992). The other techniques, which use 15N-labelled fertilizer, provide an elegant way of investigating how N originating from postsilking uptake is directed towards the kernels in the grain. Methods based on short- or long-term labelling have been developed using plants grown either under hydroponics or in the field. Hydroponics allows 15N application during a well-defined period, either short (with pulse chase experiments) or long (with quasi steady-state labelling), whereas in the field only long-term labelling is possible, with the problem that some residual 15N can remain in the soil long after the labelling period. However, the benefit of field experiments employing long-term labelling near natural abundance is that they allow the separation of newly assimilated N from that taken up earlier, as discussed by Deléens et al. (1994). Furthermore, a large number of genotypes can be studied under agronomic conditions.
In maize, application of 15N-nitrate at the beginning of stem elongation allows the determination of N remobilization (Cliquet et al., 1990b). Indeed, when 15N tracer is applied during the vegetative growth period, just before the beginning of rapid growth resulting from stem elongation, both leaves and stalks are labelled, whereas kernels are labelled only as a result of N remobilization. Furthermore, if the proportion of 15N uptake after silking is negligible (which is easier to arrange in a hydroponic experiment than in the field), N remobilization from the stover to the kernels can then be estimated from the percentage of whole-plant 15N uptake allocated to the kernels. Conversely, when the 15N tracer is applied just after silking, the allocation of recent N uptake to the stover and to the kernels can be estimated. Therefore, the use of two labelling techniques at different times is desirable, in order to obtain a complete picture of N management and recycling during the entire developmental cycle of the maize plant.
With plants grown in the field, the originality of our study was to use two 15N-labelling techniques (labelling at the beginning of stem elongation and after silking), in order to investigate the different N fluxes occurring within the plant. A detailed analysis and exploitation of the 15N-labelling data, taking into account the potential of the technique, were performed by estimating the relative 15N-specific allocation (RSA), in order to follow the fate of N assimilated after silking and the N remobilization flux. A predictive model depicting the 15N fluxes occurring between the stover and the grain was then developed to explain and predict the contribution of the various N fluxes to the amount of N in the stover protein and to grain N protein deposition.
Materials and Methods
Three separate 15N-labelling experiments were performed in the field over a 3-year period from 2001 to 2003. Compared with experiments conducted under controlled conditions (Cliquet et al., 1990a,b), the main difficulty of carrying out 15N labelling during vegetative growth was to avoid 15N uptake after silking. To minimize the uptake of 15N label after silking and to simultaneously favour its distribution within the vegetative organs (leaves and stalks), a temporally discrete application of 15N-labelled fertilizer was made at the beginning of the rapid growth associated with stem elongation (V6 stage). Under our experimental conditions, there was an interval of 30–37 d between the 15N application and the silking period.
Experiment performed in 2001 The experiment was performed to study N partitioning and management in the maize (Zea mays L.) single cross hybrid (Déa) in comparison to its parental lines F2 and Io. These three genotypes are commonly used at Institut National de la Recherche Agronomique for both physiological and molecular genetic studies (Hirel et al., 2005). Each genotype was grown in one plot of approx. 550 m2, at a density of 120 000 plants ha−1, in 80-cm-spaced rows and at two levels of N fertilization: a low level (N0, 30 kg N ha−1) and a high level (N1, 170 kg N ha−1). For each treatment (genotype × level of N fertilization), the 15N-labelled fertilizer was provided at the beginning of stem elongation on three individual microplots containing 12 consecutive plants in a row. Using a syringe, 1.25 mg of 15N was provided to each individual plant by applying 200 ml of a solution of KNO3 at 1.94%15N atom excess. Labelled plants were harvested at three different developmental stages: silking, 35 days after silking (DAS) and grain maturity, corresponding to 60–70 DAS. For each stage of plant development, six plants exhibiting a homogeneous pattern of development were selected from the 12 labelled plants in a microplot. Two plants were pooled, to obtain three replicates. For each of the three replicates, leaves and stalks + sheaths at silking and leaves, stalks + sheaths, husks + cobs and kernels at maturity were separated for the analysis of dry matter content, total N content, and 15N abundance.
Experiment performed in 2002 In this experiment, N partitioning and management were studied using four commercial hybrids: Déa, Anjou 285, Nicco and Tarro (Pommel et al., 2005). 15N-labelled fertilizer was provided at the beginning of stem elongation and at silking. The protocol was similar to that used in 2001, except that the plant density was 100 000 plants ha−1 and only the N1 fertilization regime was used, with four replicates instead of three. For the labelling during stem elongation, in each replicate the 15N-labelled fertilizer was distributed on three noncontiguous microplots containing four plants in a row, each microplot corresponding to a stage of harvest: silking, 35 DAS or maturity. For the labelling at silking, four microplots were used instead of three, as the plants were harvested 15, 25 and 35 DAS and at maturity. Each individual plant was provided with 2.5 mg of 15N by applying 200 ml of a solution of KNO3 at 4.91%15N atom excess. In the two labelling experiments and for each stage of plant development, two labelled plants exhibiting the same phenotype were selected from each microplot. Leaves and stalks + sheaths at silking and leaves, stalks + sheaths, husks + cobs and kernels at maturity were separated for the analysis of dry matter content, total N content, and 15N abundance.
Experiment performed in 2003 The aim of this third experiment was to verify the results obtained in the postsilking 15N-labelling experiment performed in 2002. Two experimental hybrids randomly chosen among the 100 examined by Bertin & Gallais (2000) and Coque (2006) were studied at four stages of plant development including silking, 15 DAS, 25 DAS and maturity. The experimental protocol was similar to that used in 2002 except that the 2.5 mg 15N applied per plant was dissolved in 3 l and three applications of 1 l each were performed every 3 d over a 9-d period to favour distribution within the soil. At silking, 15 DAS and 25 DAS, for each of the three replicates, 15N was provided to microplots of six consecutive plants in a row 5 m long. Three plants exhibiting similar phenotypes were harvested for 15N analysis. For the sampling at maturity, for each of the two replicates 15N was provided to microplots of eight plants, with six plants being harvested (Coque, 2006).
Determination of 15N abundance
In all experiments, after drying and weighing of each plant part, the material was ground to obtain a homogeneous fine powder. A subsample of 2.5 mg was used to determine total N content and 15N abundance using an elemental analyser (N-analyser NA1500; Carlo-Erba, Milan, Italy) coupled to an isotope ratio mass spectrometer (Optima; Micromass, Manchester, UK) calibrated for measuring 15N natural abundance. Using the formula δ15N = (15N/14N − 1) × 1000, the 15N abundance (A) was calculated to determine the 15N content of the sample using the following equation: [15N/(15N + 14N)]. For each plant organ, termed X, the atom percentage excess (EX = AX% − A0X%) was then calculated, AX% representing the 15N abundance percentage in the organ X considered and A0X% representing the natural 15N abundance percentage of the same organ X. A0X% was close to 0.36634, a value which corresponds to the natural abundance of atmospheric dinitrogen (N2). From EX, the relative specific allocation (RSAX) was derived according to the method developed by Cliquet et al. (1990a,b) and Deléens et al. (1994):
or as a percentage RSAX(%) = 100 RSAX. In our study the organ X corresponds to the grain, the stover or the whole plant and AS represents the 15N abundance of the labelled fertilizer. For a given organ, the RSA can be roughly defined as the ratio of the 15N content originating from the labelled fertilizer to the total N content. In order to verify whether or not 15N was taken up after silking, when the labelling was performed at the beginning of stem elongation, the amount of 15N (qX) originating from the labelled fertilizer was calculated at flowering and at maturity using the formula:
qX = QXRSAX
(QX, the total amount of N present in the organ X).
In each of the three experiments, to test whether genotypic effects were statistically significant with respect to the values obtained for both the RSAs and for the ratios RSAgrain/RSAstover and RSAgrain/RSAwhole-plant, analyses of variance were carried out. Either the residual error of such analyses or the coefficient of variation was used to obtain a value for the degree of accuracy associated with individual RSA values or with ratios of RSAs. Furthermore, to study the relationship between RSAgrain and RSAstover or RSAwhole-plant using all samples, we calculated the linear regression of RSAgrain on RSAwhole-plant (with or without the constraint of intercept equal to zero) and the correlation r between these two values. The significance of the correlation coefficient is indicated in the text by *, ** and *** at a threshold of 0.05, 0.01 and 0.001, respectively.
Modelling 15N fluxes
For modelling the 15N fluxes using the results obtained from the different labelling experiments performed before and after silking, it was assumed that: (1) there was no discrimination between 15N and 14N for N distribution within the plant, or if there was some discrimination, the impact was negligible; (2) for labelling at the beginning of the stem elongation, the 15N was homogeneously distributed within the plant, proportionally to the N content of each organ, and isotopic equilibrium was reached (see Appendix); (3) for the labelling just after silking, the 15N was taken up proportionally to the N originating from the soil and from the fertilizer.
The theoretical basis of the model is shown in Fig. 1, when two main plant compartments (stover and grain) are considered. For a given genotype it was assumed (1) that the whole-plant N content at silking (Q0) is known and (2) that at any time t, the amount of unlabelled N taken up after silking (At) and the amount of grain N (Gt) are also known. The parameter x corresponds to the proportion of the newly synthesized amino acids and proteins originating from postsilking N uptake which is allocated to the grain. Thus 1 − x corresponds to the proportion of newly synthesized N-containing molecules (mainly amino acids) allocated to the stover, which are mixed with pre-existing proteins before N remobilization (Rt) as a consequence of protein turnover. To simplify the calculation, we have considered that x remains constant whatever the plant developmental stage. The following equation is then obtained:
Rt = Gt − xAt
The quantity of N present in the stover (Qt) can be calculated as follows:
Qt = (1 − x)At + Q0 − Rt
Let gt and qt be the amount of N arising from the labelled fertilizer and allocated to the grain and to the stover, respectively, at time t. For modelling variation of these amounts during grain filling, it is necessary to distinguish between the labelling during vegetative growth and the labelling at silking. For labelling at the beginning of stem elongation, and assuming that there is no 15N uptake after silking, the increase in the amount of grain 15N during a short interval of time dt originating from the fertilizer must be equal to the decrease in the stover. Thus, we arrive at the following differential equation:
In this equation, qt/Qt represents the relative amount of N originating from the 15N-labelled fertilizer which is allocated to the grain via N-remobilization from the stover (R). This corresponds to the RSA of the stover at time t. Therefore, when Rt, Qt and q0 are known, it becomes possible to derive gt and qt by numerical integration. A 1-d interval was used for the integration. The amount of 15N present in the whole plant (qwp) can then be determined as the sum of qt+ gt. At any date, the (g/Qg) and (qwp/Qwp) ratios correspond to the RSA for the grain and to the RSA for the whole plant, respectively, and therefore the ratio RSAgrain:RSAwhole-plant can be calculated.
For 15N labelling at silking it is necessary to define the dynamics of 15N uptake (at). The simplest model is to assume that 15N originating from labelled fertilizer is taken up proportionally to the 14N present in the soil. Thus at= k At. This proportionality is based on the assumption that (1) 15N is homogeneously distributed in the soil compartment prospected by the roots and (2) there is no isotopic discrimination. Then, the differential equations become:
Thus, if At, Rt and Qt are known it becomes possible to derive gt and qt by numerical integration by considering that q0 = 0. As described above, RSAgrain, RSAwhole-plant and the RSAgrain:RSAwhole-plant ratio can be derived at any time t.
For the application of the model, from our 15N-labelling experiments, we have used curves for At (postsilking N uptake) and Gt (N accumulated into the kernels) in such a way that they fit the 15N-labelling experimental data obtained with the hybrid Déa in 2001 and 2002 (Table 1). In 2001, three harvesting periods were available: silking, 35 DAS and maturity. Therefore, we have linearly interpolated between silking and 35 DAS, and curvilinearly interpolated between 35 DAS and maturity. In 2002, we have only linearly interpolated between each of the five labelling periods. The application of the model was first developed using the Microsoft Excel spreadsheet (Microsoft 98). By assigning different values to parameter x between 0 and 1, it was possible to identify the value that best simulated the observed values for the RSAgrain:RSAwhole-plant ratio. We have considered such a ratio because of its properties, which are described in the Results section (low coefficient of variation and value close to 1 at maturity).
Table 1. Observed values for maize (Zea mays) whole-plant nitrogen (N) uptake and N accumulated in the kernels according to stage of plant developmental and year, utilised in the modelling of 15N fluxes
Data are expressed in grams of nitrogen (N) per plant, for the hybrid Déa grown under a high N fertilization regime.
Variation in RSA values and relationship between RSAgrain and RSAstover or RSAwhole-plant
Labelling during vegetative growth At silking in the 2001 experiment, for the whole-plant RSA the effect of the level of N fertilization (N) was highly significant. The genotype × N interaction was also significant, whereas the genotypic effect was not significant (Table 2). The genotype × N interaction was attributable to the genotype Io under high N fertilization, which had a lower value than expected based on the addition of the effects of the genotype and of the level of N fertilization. In the 2002 experiment, the genotype effect was not significant (Table 3).
Table 2. Analysis of variance of relative 15N-specific allocation (RSA) in maize (Zea mays) for the whole plant and plant parts, and RSA ratios, at silking and maturity from the results obtained in 2001 following 15N labelling at the beginning of stem elongation
At maturity in the 2001 and 2002 experiments, we observed a large environmental variation in RSA between replicates, illustrated by an important residual variation in the analysis of variance, thus leading to nonsignificant genotype, N fertilization, and genotype × N fertilization effects (Tables 2, 3, 4). In contrast, when considering the RSAgrain:RSAwhole-plant ratio, a low standard error and a low coefficient of variation (Tables 3, 4) were obtained with a highly significant genotype effect in 2001 and an effect at the limit of significance in 2002. Furthermore, in the 2001 experiment the influence of the level of N fertilization on the RSAgrain:RSAwhole-plant ratio was highly significant (0.96 at high N input and 0.98 at low N input; Tables 2, 4); the interaction genotype × N fertilization level was only significant at the probability 0.058. When the RSAgrain:RSAstover ratio was considered, it was also found to be estimated with higher accuracy than individual RSAs, but not as accurately as the RSAgrain:RSAwhole-plant ratio (coefficients of variation of 4.8% and 5.6%, compared with 1.9% and 1.8%, respectively, in 2001 and 2002).
Table 4. Relative 15N-specific allocation (RSA) and RSAgrain:RSAwhole-plant ratio in different maize (Zea mays) genotypes grown at low and high nitrogen (N) fertilization inputs in 2001 following 15N labelling at the beginning of stem elongation
A close relationship between RSAstover and RSAgrain was systematically observed whatever the year of experiment. The correlation between the two parameters was 0.96*** in 2001 (18 data points), 0.95*** in 2002 (16 data points) and 0.94*** when the 2001 and 2002 data were pooled (Figs 2, 3). The regression coefficient, for a regression line with zero intercept, was 0.94 in 2001 and 1.04 in 2002. The RSAgrain: RSAstover ratio was therefore always close to 1 (Tables 3, 4). Similarly, the RSAgrain:RSAwhole-plant ratio was always near 1, ranging between 0.95 for the genotype Io in 2001 and 1.03 for the commercial hybrid Nicco in 2002. Consequently, a regression coefficient close to 1 was obtained for the relationship between RSAgrain and RSAwhole-plant by using the 34 data points of both years (Fig. 3), with a strong phenotypic correlation (0.99***) between RSAgrain and RSAwhole-plant.
At 35 DAS in 2001 and 2002, the analysis of variance (ANOVA) showed that the results were similar to those obtained at maturity, with a large coefficient of variation leading to no significant genotype effects for RSAs and a lower coefficient of variation for the ratios RSAgrain:RSAstover and RSAgrain:RSAwhole-plant than with individual RSA values. However, the genotype effect for these ratios was significant (Table 5). In the 2002 experiment, the regression coefficient of RSAgrain onto RSAstover was close to 1 (0.99) at 35 DAS with a correlation coefficient equal to 0.94***. Consequently, the regression coefficient for RSAgrain onto RSAwhole-plant was again close to 1 (Fig. 3). This means that the quasi-equality to 1 observed for the ratio RSAgrain:RSAwhole-plant already existed before maturity when N remobilization and N uptake were still active. Only in the 2001 experiment was the slope of the regression of RSAgrain onto RSAwhole-plant (0.85) significantly lower than 1 (Table 5).
Table 5. Relative 15N-specific allocation (RSA), RSA ratios and relationships between RSAgrain and RSAstover or RSAwhole-plant in maize (Zea mays) at 35 d after silking in 2001 [at low N input (N0) and high N input (N1)] and in 2002, following 15N labelling at the beginning of stem elongation
Labelling after silking When 15N labelling was applied just after silking, the genotype effect was not significant for RSAgrain, RSAwhole-plant and the RSAgrain:RSAwhole-plant ratio, whatever the sampling date (15, 25 or 35 DAS or at maturity). Nevertheless, the coefficient of variation was lower for the RSAgrain:RSAwhole-plant ratio than for RSAgrain and RSAwhole-plant individually (Table 6). The regression coefficients (b) of RSAgrain onto RSAstover were always associated with a high correlation coefficient (varying from 0.77** to 0.96**). They decreased significantly from 15 DAS (b = 2.71) to maturity (b = 1.57). Regression of RSAgrain on RSAwhole-plant gave lower slopes: b = 2.2 for 15 DAS and b = 1.2 at maturity with a zero intercept (Fig. 4), whereas the associated correlation coefficients were higher.
Table 6. Relative 15N-specific allocation (RSA) values and ratios at different stages of maize (Zea mays) plant development in 2002 following 15N labelling just after silking
Significance: *, P < 0.05; ***, P < 0.001; ns, not significant.
Two values with the same superscript letter (A, B or C) are not significantly different at P < 0.05.
The same results were obtained using the data from the 2004 experiment. Pooling the six observations (two genotypes and three replicates) at a given stage of plant development, the average ratio R = RSAgrain/RSAwhole-plant derived from the linear regression of RSAgrain on RSAwhole-plant, with the constraint of a zero intercept, was highest at 15 DAS (R = 1.44, r = 0.91**, r being the coefficient of correlation). R progressively decreased from 25 DAS (R = 1.23, r = 0.93**) to maturity (R = 1.19, r = 0.88*), but it remained significantly higher than 1. A nearly identical R-value (R = 1.20, r = 0.99**) was found with data collected from an experiment performed on 100 different genotypes (data not shown) using the same computing method (Coque, 2006).
Following 15N labelling during stem elongation in 2001, x = 0.55 (the proportion of the newly synthesized amino acids originating from postsilking N uptake allocated to the grain) yielded the best fit to the observed variation of the RSAgrain:RSAwhole-plant ratio during the period spanning 35 DAS to maturity (Fig. 5a). In the 2002 experiment, x-values of approx. 0.20–0.30 fitted more accurately the observed results (Fig. 5b). For labelling after silking in the 2002 experiment, the model clearly predicts a decrease in the RSAgrain:RSAwhole-plant ratio, as observed from silking to maturity. This leads to x-values ranging between 0.30 and 0.55, with x-values increasing from 15 DAS to maturity (Fig. 6).
Explanation for the relationship between RSAgrain and RSAwhole-plant
15N labelling performed either before or after silking always provided a close relationship between RSAgrain and RSAstover or between RSAgrain and RSAwhole-plant, at all stages of development, the latter relationship being slightly stronger than the former because of the contribution of RSAgrain to RSAwhole-plant. The proportionality between RSAstover and RSAgrain at all times could have been caused either by an exchange of N between the stover and the grain or by a common source of variation for N allocation to the two organs. It is logical to find that, for 15N labelling at the beginning of stem elongation, the isotope was also allocated to the grain, as a result of remobilization. However, N remobilization alone cannot explain the finding that RSAstover was very similar to RSAgrain. For 15N labelling after silking, the situation is easier to explain because the 15N tracer appeared to be simultaneously distributed to the stover and to the grain. Nevertheless, neither the linear relationship observed between RSAgrain and RSAstover nor the decrease towards 1.0 of the RSAgrain:RSAwhole-plant ratio from 15 DAS to maturity can be readily explained by 15N distribution between the stover and the grain.
For 15N labelling at the beginning of stem elongation, the strong relationship found between RSAgrain and RSAstover could be a result of significant 15N absorption after silking. In the 2001 and 2002 experiments, the influence of the genotype on postsilking 15N uptake was not significant because of a large environmental effect. On average, in 2001 there were N losses (9%) between silking and maturity, but they were not significant. Therefore, for this year, one can assume that 15N uptake after silking was practically absent. In 2002, on average 23.1% of the 15N present in the whole plant at maturity was taken up after silking. However, this difference was just at the limit of significance at the threshold of 0.05 (data not shown). Even if the contribution of 15N uptake after silking was considered to be significant in 2002, it is not sufficient to explain the strong relationship between RSAgrain and RSAwhole-plant also observed in 2001 in the absence 15N uptake after silking.
Whatever the type of labelling, another important factor explaining the relationship observed between RSAstover and RSAgrain is their large environmental variation, which reflected the large environmental variation in the amount of labelled fertilizer taken up by each plant under field growth conditions. Indeed, in the absence of isotopic discrimination, when environmental variations affect RSAstover they are expected also to affect RSAgrain in the same proportion. Considering only the environmental correlation between RSAgrain and RSAstover, a correlation of 1 may be expected between the two RSA values. Therefore, it appears that the RSAgrain:RSAstover and RSAgrain:RSAwhole-plant ratios are key parameters whatever the plant developmental stage because: (1) they provide a good picture of 15N distribution between the grain and the stover; (2) they can be accurately determined as they are not affected by environmental variation such as 15N availability; and (3) the RSAX:RSAwhole-plant ratio can be used to estimate the rate of N allocation to a particular organ X (Cliquet et al., 1990a,b; Deléens et al., 1994).
It is now necessary to explain why the regression coefficients for both RSAgrain onto RSAstover and RSAgrain onto RSAwhole-plant were always close to 1 from 35 DAS to maturity following 15N labelling at the beginning of stem elongation, whereas for 15N labelling at silking the coefficients were both much higher than 1 and tended to decrease towards 1 from 15 DAS to maturity. It is likely that such changes in the two RSA ratios indicate a movement towards an equilibrium corresponding to an equivalence between RSAgrain and RSAstover, whatever the period of 15N application. With labelling at the beginning of stem elongation, assuming that N remobilization and N uptake occur simultaneously, postsilking allocation to the stover of the N taken up by the plant will dilute the 15N already present in the stover, while the grain will be enriched in 15N by remobilization from the stover. Conversely, when the 15N fertilizer is provided just after silking, N remobilization from the stover to the grain will dilute the 15N allocated to the grain, while at the same time the newly taken up N will be allocated to the stover, thus enriching this organ with 15N. Therefore, such N fluxes contribute to the progression of the 15N:14N isotopic ratio towards similar values in the stover and in the grain.
Considered overall, the proposed model provides a good simulation, fitting the results obtained in the field experiments. In the 2001 and 2002 experiments and with the two types of labelling, there was an allocation of newly synthesized amino acids to the stover (1 − x) ranging between 45 and 70%. The model suggests the hypothesis that, after silking, the amino acids from the newly assimilated N are in part mixed with amino acids from proteolysis for the synthesis of new proteins which are further hydrolysed (Fig. 1). This mixture modifies the RSAgrain:RSAwhole-plant ratio during grain filling in such a way that it reaches a final value close to 1. However, the situation for which RSAgrain∼ RSAstover is easier to reach when labelling during vegetative growth is used, rather than labelling at silking. Indeed, with labelling during vegetative growth, in the absence of N uptake after silking, RSAgrain would strictly be equal to RSAstover. When there is postsilking N uptake and allocation of N originating from the newly synthesized protein to the stover and to the grain, a dilution effect will occur. Therefore, the ratio RSAgrain:RSAwhole-plant will be below 1 at the beginning of grain filling, but thereafter will rapidly increase towards 1. The fact that grain filling begins actively only 10–15 d after ovule fertilization can also quickly lead to a RSAgrain:RSAwhole-plant ratio close to 1. Indeed, at the end of this period, whole-plant N represents c. 85% of the whole-plant N at maturity and c. 70% of the N in the grain. Therefore, the contribution of N uptake will be relatively small in comparison to that of remobilization from the stover to the grain and RSAgrain will be close to RSAstover. When considering labelling at silking, even with a mixture of new (synthesized following N uptake) and pre-existing amino acids (originating from remobilization), RSAgrain will be much higher than RSAstover. Consequently, at the beginning of grain filling, the RSAgrain:RSAstover ratio will be greater than 1, but will decrease towards 1 as a result of a dilution effect occurring in both the stover and the grain. This was verified by our simulation model, which showed an increase in the proportion of newly synthesized amino acids allocated to the kernels (x) from 15 DAS to maturity. From a physiological point of view, it is likely that this proportion increases with time, while leaf senescence is progressing, as already shown by Weiland (1989) and Ta & Weiland (1992) in maize and by Schiltz et al. (2005) in pea.
The allocation of newly synthesized amino acids to the stover must be mainly related to the leaf protein turnover occurring in the stover (Hirel & Gallais, 2006). As there is no increase in the amount of protein in the stover, this allocation is probably occurring to replace the old proteins that have been remobilized. By simulation, we found that a protein lifetime of c. 5 d is sufficient to generate the equivalent of a complete mixture of amino acids from newly assimilated N with amino acids from proteolysis. In fact, our model corresponds to a very quick protein turnover. This finding is consistent with the results obtained by Lattanzi et al. (2005), who showed that during vegetative growth most of the N imported into growing tissue originated from storage pools. Moreover, Liu & Jagendorf (1984), and Malek et al. (1984) showed that, in pea, 20–35% of the proteins newly synthesized by isolated chloroplasts were degraded during the next 30 min. Therefore, the agreement between our model and the experimental results suggests that, after silking, the proteolysis of ageing proteins provides a pool of amino acids which can be either translocated to the grain or recycled for the synthesis of new proteins from a mixture which also contains amino acids recently synthesized following N uptake. Our model is thus close to that proposed by Cooper & Clarkson (1989), in which the amino acids that are cycling between shoots and roots are conceived as a single pool.
Validity of the different assumptions
For 15N labelling at the beginning of stem elongation, it was assumed that 15N uptake after silking is negligible. We have already discussed that the impact of 15N uptake after silking, which was observed in 2002 but not in 2001, was expected to be low. However, experiments are now in progress to evaluate whether it is possible to reduce the risk of significant residual 15N uptake after silking by providing 15N much earlier, i.e. before stem elongation. We expect to find that the kinetics of 15N uptake are similar to that observed by Sheehy et al. (2004a,b), who showed that in rice 90% of the 15N was taken up 20 d after labelling.
The second assumption required for the validity of our experimental design is that 15N is homogeneously distributed within the plant, proportionally to the amount of N in each organ, and that isotopic equilibrium is reached. Isotopic equilibrium is favoured by the time interval between labelling and silking, and also by the low amount of 15N provided per plant. Redistribution of absorbed N among the leaves, as suggested by the results of Simpson et al. (1983) in wheat and those of Schiltz et al. (2005) in pea, would contribute to fulfilling the assumption of homogeneous N distribution among vegetative organs. Similarly, Lattanzi et al. (2005) showed that N tracer entered into storage pools more than once before reaching the growth zone, which is in favour of a homogeneous distribution. An exchange between the phloem and the xylem could also favour homogeneous 15N distribution within the plant (Cooper & Clarkson, 1989). It was in order to favour such a homogeneous distribution of the isotope that we chose to apply 15N just before stem elongation, which corresponds to the beginning of the high growth rate period. If there is a nonhomogeneous N distribution among vegetative organs, but if the same proportion of N is remobilized from the different parts of the plant, the conclusions derived from our model are still valid (see Appendix). Furthermore, the simulation of observed results in our model is a proof that such an assumption tended to be satisfied.
The third assumption in our work concerns the absence of 14N and 15N discrimination by the plant for N distribution and N remobilization. RSAs were higher at low N input than at high N input, suggesting that discrimination may have occurred either at the level of N uptake or during the first steps of inorganic N assimilation (Mariotti et al., 1982; Ledgard et al., 1985; Coque et al., 2005). However, such discrimination does not have any effect on the RSA ratio, because the RSA for each organ will be modified in the same proportion. Only differential discrimination for the synthesis of new amino acids originating from N uptake and for amino acids released following protein hydrolysis could affect the result. Another possible bias could be introduced if there are N losses as a result of NH3 emission during senescence (Schjoerring, 1991; Sharpe & Harper, 1997). However, if there is no discrimination and if there is an equal distribution of the two isotopes among vegetative organs, the RSA value of the stover will not be affected as it is dependent on the 15N:14N ratio. If the stover is considered as a sufficiently homogeneous organ for N distribution, the ratio RSAgrain:RSAstover will not be affected by the losses, whereas the ratio RSAgrain:RSAwhole-plant will be. However, the bias introduced by N losses is expected to be generally low (see Appendix). Furthermore, if N losses did take place, as is probable in the 2001 experiment, it does not appear that they affected either the relationship between RSAgrain and RSAstover or the predictive value of our model.
For a labelling treatment just after silking, the most important assumption is that 15N is homogeneously distributed in the different soil compartments colonized by the roots during the entire period of postsilking N uptake. As discussed before, if new amino acids originating from the N recently taken up are distributed in the different parts of the plant, the heterogeneous N distribution in the soil will not have any significant impact. Concerning N availability during the time span from flowering to maturity, it must be noted that N availability in the first 30 d after silking, during which 90% of the total plant N is absorbed, will be the most important. One can hypothesize that, if the availability of 15N in the soil decreases with time, this will tend to enhance the decrease of RSAgrain during the same period. However, the simulation of observed results by our model showing a decrease in the ratio RSAgrain:RSAwhole-plant from 15 DAS to maturity is a proof that the assumption of a homogeneous 15N distribution in the different soil compartments colonized by the roots tended to be satisfied, at least at the most important phase of postsilking N uptake. Nevertheless, as a whole, the assumptions required for reliable interpretation of the results of our 15N-labelling experiment appear to be more difficult to fulfil for labelling after silking than for labelling at the beginning of stem elongation. However, the value of using 15N labelling after silking was to show that there is direct allocation of N to the stover, originating from postsilking N uptake.
The use of 15N labelling at the beginning of stem elongation and just after silking allowed the demonstration that, in maize, a large proportion (> 50%) of newly synthesized amino acids from postsilking N uptake were first integrated into the stover proteins before translocation to the grain. This finding indicates that protein turnover, which is controlled by an intra-organ regulatory mechanism, is occurring concomitantly with N remobilization to the grain, which represents an inter-organ metabolic process. It remains to be determined whether the proportion of N allocated to the stover and the proportion of N remobilized have a direct link to the protein turnover occurring at the cellular level (Brouquisse et al., 2001; Hörtensteiner & Feller, 2002). The occurrence of such a link would mean that the proteolytic enzymes involved in the control of protein turnover could also be involved in the degradation of leaf protein during senescence. Mae et al. (1983), Huffaker (1990) and Hörtensteiner & Feller (2002) suggested that certain common catabolic enzymes may function during the initial and reversible phase of senescence, while others are specifically induced when senescence is progressing. Therefore, the occurrence of some common and/or distinct mechanisms regulating protein turnover and protein hydrolysis during leaf N remobilization (Irving & Robinson, 2006) must be taken into consideration during the general process of plant N management.
The authors are very grateful to Dr Peter Lea for his careful revision of the English. We also thank very much Dr Gilles Lemaire (INRA, Lusignan, France) and Dr Jacques Le Gouis (INRA, Mons-en-Chaussée, France) for their helpful discussions about the experimental results and the model.
Appendix: Discussion of Some Assumptions for Labelling Performed at the Beginning of Stem Elongation
(1) Homogeneity of 15N:14N ratio in the stover
The model assumes the existence of a stover compartment with a homogeneous 15N distribution. In fact, as demonstrated below, the distribution can be heterogeneous among vegetative organs if the same proportion of N is remobilized to the grain in each organ.
Let T1 and T2 be two stover organs, with Q1 and Q2 total amount of N and q1 and q2 the amount of 15N, respectively, and assuming that the 15N abundance in each organ q1/Q1 is different from q2/Q2. Let g1 and g2 be the 15N quantity in the grain originating from T1 and T2, respectively, and p1 (p2) the 15N amount at silking in the organ. Then, the proportion remobilized from each organ is r1 = g1/p1 and r2 = g2/p2. Now if we assume that the two organs remobilize the same proportion of their N:
(g, the 15N quantity in the grain; wp, the total 15N amount in the whole plant at maturity.)
Therefore, with the assumption of the same proportion remobilized from each organ, despite the heterogeneous distribution of N, the different vegetative compartments can be pooled, which is equivalent to assuming a homogeneous distribution of 15N in the different organs. It is only with a heterogeneous 15N distribution and heterogeneous proportions remobilized among vegetative organs that the estimation of the proportion of remobilized N will be biased.
(2) Effect of N losses on the relationship between RSAs for labelling at the beginning of stem elongation
Consider that the stover (S) consists of leaves (L) and stalks (T), and let RSAL, RSAT and RSAS be the RSAs for leaves, stalks and stover, respectively. QL and QT being the total amount of N in leaves and stalks, without N losses, the stover RSA is:
Assuming that there are losses only affecting the leaves, these losses being Qp, then the new stover RSA, estimated at maturity, will be
which can be reformulated as
[p = Qp/(QL+ QT); R = RSAL/RSAT.]
As an example, if R = 1.20 and p = 0.15 (15% N losses), then RSA′S = 0.96 RSAS. This means there is a low bias. Thus it would be only with large and unlikely losses that the RSAS would be significantly affected.