Reassessment of the Contribution of Belowground N from Soybean after Testing Different 15N Leaf-Labelling Strategies

Purpose: Soybean is the most important grain crop in Brazil with a mean N accumulation of over 250 kg N ha -1 , principally from biological N 2 xation. The residual N benet depends heavily on the quantity of the belowground N at harvest, much which cannot be directly recovered in roots. The purpose of this study was to investigate different aspects of the 15 N leaf-labelling technique to quantify non-recoverable root N (NRRN) derived from senescent roots and nodules (rhizodeposits). Methods: Soybean plants were grown in pots of soil and at 27 days after planting (vegetative stage V4) cut or whole leaves were exposed to highly enriched 15 N-labelled urea or glutamine. Seven sequential harvests of the plants and soil were taken until the nal grain harvest at 70 days after labelling. Results: After only 48 h, the plants labelled with 15 N urea transferred approximately 5% to the soil, while only 1% was found in the roots. Leakage of 15 N label was even more pronounced when the leaves were labelled with 15 N glutamine. After this initial leakage, the excess 15 N deposited in the soil only increased by a further 2.6% of applied label, which suggested that only part of this N represented senescence of roots or nodules. Conclusions At the nal harvest, N in roots separated from the soil amounted to 6.4% of total plant N. Discounting the early rapid deposition of 15 N-enriched N to the soil, our calculations indicated that at nal harvest the total NRRN was 2.8% of total plant N.


Introduction
Soybean is quantitatively the fourth most important grain crop produced in the world after maize, wheat and rice when ranked by grain production (FAO-STAT-2020). However, if ranked by protein production it is the world´s foremost crop. Brazil, the USA, Argentina and China are the main producers in that order. Herridge et al. (2008) estimated that globally soybean was responsible for over 77 % of all BNF contributions of all grain legumes. Soybean production has increased by over 60 % since the publication of that paper, far more than the production of other grain legumes, such that the proportion today must be over 80% (FAO-STAT, 2020). The quanti cation of the residual N left in the soil after harvest is essential to quantify the long-term sustainability of soybean crop rotations and also for the assessment of global uxes of reactive nitrogen (Fowler et al. 2013).
In Brazil the proportion of N derived from air (%Ndfa) via BNF by soybean is estimated at approximately 80 % (Zotarelli et al. 2012). The harvest index of the cultivars at present planted in Brazil is estimated to be approximately 40 %. The grain is rich in nitrogen (approximately 6.5 % N -37 to 41 % protein) and the remaining vegetative shoot tissue (stems, leaves and empty pods) is low in N content such that the harvest index (shoot tissue only) is estimated to be close to 80 %, as physically recovered roots at harvest contribute little to plant N (Alves et al. 2006;Zotarelli et al. 2012). It follows that even though the BNF inputs are large, with such a high harvest index there may be little overall addition of N to the soil for the next crop. This conclusion ignores the input of N derived from the roots present at time of harvest as well as other N deposited from senescent roots and nodules, exudates, lysates and sloughed off cells, collectively known as "rhizodeposits". We use the term non-recoverable root nitrogen (NRRN) in this paper to include all N deposited in the soil which is not recoverable as roots by sieving the soil at harvest.
The amount of N deposited into the soil from roots until the time of harvest has generally been estimated with the use of the 15 N isotope (Wichern et al, 2008). Janzen and Bruisma (1989) were among the rst to use the technique and labelled the shoots of wheat with 15 N-labelled ammonia gas. Since that time there have been many studies to estimate the quantity of non-recoverable root N (NRRN) of legumes and non-legumes using 15 N labelling of leaves, petioles, stems, "split-root technique" and other techniques (see Wichern et al. 2008). The split-root technique, and those that use gaseous forms of technique to use on eld-grown soybeans, the labelling of leaves, petioles or stems were possible choices.
As the petiole technique has been found less effective quantitatively for labelling the plants ( McNeill et al. 1997; Yasmin et al. 2006), and the stem-feeding (cotton-wick) technique (Russell and Fillery 1996a) requires thick stems, the most frequently used technique is leaf-labelling generally following the "leaf-ap" procedure used by Khan et al. (2002). The technique relies on estimating the added labelled N (excess of 15 N above the natural abundance of the soil) and the 15 N enrichment of the tissue or compounds being lost by the plant to the soil. The roots are assumed to be the source of N lost to the soil so the calculation of the proportion (%) of N derived from the roots (%NdfR) becomes (Janzen and Bruisma, 1989): %NdfR = (Atom%excess soil/Atom%excess root) x 100 To calculate the total N in the soil derived from rhizodeposition (NRRN), %NdfR must be multiplied by the total N in the soil (TN soil ). NRRN = (%NdfR x TN soil )/100 As Janzen and Bruisma (1989) and several subsequent authors have stated (e.g. Rasmussen, 2011;Hupe et al. 2016), for the technique to estimate accurately the %NdfR, the following conditions or basic assumptions need to be met: 1. The N deposited in the soil has the same 15 N enrichment as the roots, 2. The added excess 15 N is evenly distributed in the root system, 3. The 15 N enrichment of the roots is constant over the growth period of the plants.
Most authors have labelled plants just once, so it is surprising that there have been so few studies on the uniformity of 15 N enrichment in the roots and changes in enrichment with time.
If plants are labelled with one pulse of a 15 N-labelled substrate at a relatively early stage of growth, as the plants grow and accumulate more nitrogen from the soil and/or, in the case of legumes, from BNF, it is to be expected that the 15 N enrichment of plant tissues will decline with time. This is generally observed for the shoot tissues of plants labelled in this manner, but surprisingly this does not always seem to be true for root tissues (McNeill et al. 1997;Gasser et Gasser et al. (2015) investigated the use of the leaf labelling technique to estimate the N derived from roots of red clover (Trifolium pratense). In this study leaf labelling of the plants was conducted at 83 days after germination and soil and plants were sampled starting 1 day after labelling and then at 14-day intervals for further 42 days. The 15 N enrichment of the root did not change signi cantly over the 42-day period. A further important result of this study was that the authors found that within one day of labelling a pulse of 15 N-enriched N was leaked into the growth medium (bentonite/sand).
If the 15 N enrichment of the roots remains approximately stable with time from leaf-labelling until nal harvest, then the third basic assumption of Janzen and Bruisma (1989) that "The 15 N enrichment of the roots is constant over the growth period of the plants" would seem to be ful lled. However, further studies are required to con rm this.
In preliminary studies by our team at Embrapa Agrobiologia with soybean we also noted that there was a considerable accumulation of 15 N-labelled N in the soil only three days after leaf-labelling with 15 N-enriched urea (Paredes et al. 2007).
Root exudation induced by application of organic compounds to plant shoots was 'hot topic' many years ago, and exudation of amino compounds through roots was a known effect of spraying urea on plant shoot (Rovira et al. 1969). In addition, urea may freely cross plasmatic membranes owing to its size and nonpolar nature (Canarine et al. 2019), and then it can move from aerial tissues to the rhizosphere. It was thought that urea may be a substrate that caused considerable trauma to the plant metabolism owing to its rapid hydrolysis to ammonia which is toxic to the plant by several mechanisms (Gerendás et al. 2001), and that a compound such as glutamine might be a suitable alternative.
We hypothesised that: 1. If labelling plants with urea induces a short-term transfer (leakage) of enriched N into the soil, then 15 N-labelled glutamine will be less prone to cause this effect.
2. The estimates of "non-recoverable root nitrogen" (NRRN) will be independent of the source of 15 N-enriched N used for leaf labelling.
The objective of this study was to investigate the change in 15  The pots were laid out in randomized complete blocks with four leaf-labelling treatments, seven harvests and ve replicates (blocks). The leaf-labelling treatments were: 15 N-labelled urea (ULL), 15 N-labelled glutamine (GLL) and 15 N-labelled urea + glucose (USLL) using the leaf-ap protocol and a further treatment with 15 N-labelled urea where an entire leaf was immersed in the labelled solution (UEL). The treatment with glucose was included as it was thought that its presence may enhance urea uptake and assimilation by the leaves (Borkowska and Szczerba, 1991 The leaf labelling was performed at growth stage V4, 27 days after planting. For the leaf-ap protocol the central leaf of the last trifoliate, with leaves already expanded, was cut parallel to the central nerve permitting the insertion of the "lap" into an Between 7 and 10 DAL, shoot N accumulation ceased or was negative, but N accumulation subsequently continued. This interruption of N accumulation may be attributed to the trauma induced by the leaf cutting and labelling with an exogenous source of N seven days earlier. There was no evidence that the labelling with glutamine was any less harmful than urea to N uptake in this 3-day period. The 15 N enrichment of the rhizosphere soil was between 12 and 53 times higher than that of the bulk soil. Compared to the approximately 5 kg of bulk soil, the mass of rhizosphere soil was small. As root mass increased, the amount of rhizosphere soil increased, starting at 0.75 to 0.97 g plant − 1 at 2 DAL, rising to between 4 and 7 g at the nal harvest (70 DAL). The total enriched N deposited in the rhizosphere soil was on average only 2.2 % of that found in the bulk soil and less than 0.28 % of the total excess 15 N applied (Table 1). The different methods of leaf labelling with 15  Two strategies were adopted to calculate NRRN. For the rst (NRRN-A) it was assumed that all N deposited in the soil was derived from the ne roots and the 15 N enrichment of the ne roots was used as 'Atom%excess root' in Eq. 02 (Table 2). For the second strategy (NRRN-B) it was assumed that N deposited in the soil was derived from both the ne roots and the nodules and the weighted mean 15 N enrichment of roots and nodules was utilized. The estimates of NRRN for the harvests until 14 DAL are not reliable, because at this time the 15 N enrichment of the ne roots was changing rapidly (Fig. 4C) such that the enrichment of the N being deposited is unknown. However, the results do show that deposition of labelled N began very soon after leaf labelling and could not be considered to be gradual deposition of N from senescing roots and/or nodules. Table 2 Total N in recovered roots and nodules and estimates of non-recoverable root N (NRRN mg plant − 1 ) utilizing A) the 15 N enrichment of ne roots, or B) the weighted mean 15 N enrichment of the ne roots + nodules as the sources of N deposited in the soil.  (Fig. 4). Likewise, the weighted mean 15 N enrichment of the ne roots + nodules varied from 0.246 to 0.228 and then to 0.202 atom % 15 N excess for the same three harvests. The lower value of the 15 N enrichment of the ne roots + nodules compared to ne roots alone led to higher estimates of the NRRN based on roots + nodules at 70 DAL (Table 2). At this nal harvest as well as those at 7, 14 and 25 DAL the estimate of NRRN was higher for the plants labelled with glutamine than in any of the urea labelling treatments. However, owing to the low proportion of labelled N in the glutamine, the 15 N enrichment of the bulk soil was very low (0.0008 atom % 15 N excess) and these estimates cannot be considered to be reliable.
When the N in the recovered roots and nodules was added to the estimate of the NRRN based on the 15  Nodule total N showed a similar behaviour (Fig. 2D). As the 15  N deposited in the soil in the rst 24 h after leaf-labelling with 15 N-enriched urea was equivalent to up to 5 % of unrecovered roots of white and red clover (T. repens and T. pratense, respectively). Gardner et al. (2012) suggested that the presence of 15 N label in the rst day after labelling was due to unrecovered roots which were impossible to separate from the soil, but even after 32 days there was no signi cant increase in the excess 15 N recovered, which begs the question as to why there was no further increase in the quantity of unrecovered roots in the soil during this period. Rasmussen et al. (2019) also suggested that much of this short-term deposition of excess 15 N in the soil could be due to unrecovered roots, but they recognised that unrecovered labelled roots could not explain the rapid appearance of 15 N in neighbouring grass which corresponded to 6-8 % and 12 to 16 % of the excess 15 N recovered outside the labelled plant (soil + grass). In contrast the rapid loss of a small proportion of excess 15 N from red clover was attributed to leakage of soluble forms of 15 N by Gasser et al. (2015).
In our study on soybean in the treatments leaf labelled with urea, by 2 DAL approximately 5 % of the 15 N label had been transferred from the labelled leaf to the shoot and less than 1% was found in the roots. However, in this 48-h period the mean "leakage" to the soil was between 4.0 and 5.9 % of all labelled N, thus amounting to almost half of the N exported from the labelled leaf to the shoot, roots and soil (Table 1). There was a similar behaviour of the excess 15 N deposited in the soil derived from labelled glutamine. At 2 DAP, only 2.8 % of the applied N was recovered in the roots while 11.6 % was released into the soil. As the amount of 15 N excess in the roots of all treatments at 2 DAL was much lower than that found in the soil, the roots must be considered to be a conduit for enriched N rather than a source. This supports the hypothesis of It is logical to expect that the source of "rhizodeposits" should mostly be root exudates and products of root and nodule turnover and senescence. The expected chronological pattern would thus be a gradually increasing rate of loss of N from roots with time and especially considerable losses from senescent nodules. This was not the pattern observed in this study.
The mean quantity of excess 15 N deposited in the soil in the rst 7 DAL from the urea labelled plants was 5.5 % of that fed to the leaves. Subsequently, over the next 63 days, this only increased further by 1.8 %, amounting to a total on average of 7.2 % (Table 1). Thus, the results suggested that most of the total excess 15 N deposited in the soil at the nal harvest came from initial leakage from the root, although some of this soluble labelled N could have been reabsorbed.
Labelling with 15 N-enriched glutamine appeared to have no advantages over 15 N-enriched urea. Initial uptake by the plant was faster but short-term deposition in the soil seemed more severe than was the case with the urea-labelling treatments.
For the urea-labelling treatments, an average 4.6 % of the 15 N label was deposited in the soil in the rst two days; for the glutamine, the amount was estimated to be 11.6 % and, unlike the urea, some of the label appeared to be re-absorbed over the next eight days such that at 10 DAP the amount of excess 15 N in the soil decreased to 7.2%.
Initially we utilized two strategies to calculate the NRRN both assuming that the sole sources of 15 N excess in the soil were ne roots and nodules. In the rst strategy, A, the ne roots were assumed to be the sole source of the N loss and in the second strategy, B, ne roots and nodules were the source.

Conclusions
With respect to the hypotheses outlined in the Introduction, the labelling of the plants with 15  In this present study correcting for initial leakage reduced the estimate of NRRN by 74 % (equivalent to a 290 % overestimate of uncorrected value). Even the uncorrected value is considerably lower than from earlier studies on soybean and many estimates of NRRN from other legumes.
However, from the results obtained, it is reasonable to assume that the initial stimulation of the deposition of 15 N-enriched N within the rst two days after labelling would cause an overestimation of NRRN. The proportion of the excess 15 N that was leaked is considerably higher than that experienced by Gasser et al (2015). The difference could be due to the different species studied or the fact that in our study the plants were grown in soil where a larger microbial biomass may have enhanced N leakage by rapid assimilation.
We conclude that almost all estimates of non-recoverable belowground N published to date, are large overestimates and we echo the recommendations of Gasser (2015) that the 15