Nutrient resorption tightens plant nitrogen and phosphorus coupling and decreases with sulfur deposition as mediated by interannual precipitation in a meadow

Purpose Sulfur (S) deposition as a global change issue causes worldwide soil acidication, nutrient mobilization and marked changes in plant nutrition. Here, we investigated how S deposition would affect leaf nutrient resorption and how this effect varies with yearly uctuations in precipitation. Methods In a semiarid meadow exposed to S addition, we measured nitrogen (N), phosphorus (P) and S concentrations in green and senescent leaves of a grass and a sedge and calculated nutrient resorption eciencies (NuRE) across two years with contrasting precipitation (13% higher and 27% lower than long-term mean annual precipitation). Two-way ANOVAs used to explore the main and interactive effects of S and sampling year on nutrient in and NuRE of two dominate with block as a random factor. Regression models were used to quantify relationships between S addition rates and leaf nutrient characteristics with the best curve-tting results chosen based on coecient of correlation. Pearson correlation analysis was used to examine the relationships between soil nutrient availability and leaf and NuRE. All these analyses were performed using


Introduction
Nutrient resorption is a key physiological process for senescing plants to conserve their nutrients, particularly in nutrient-poor habitats (Aerts 1996;Brant and Chen 2015;Drenovsky et al. 2019;Lü et al. 2012). Based on global estimates, about half of foliar N and P are resorbed during leaf senescence (Aerts 1996;Killingbeck 1996). The resorbed nutrients are readily available for subsequent plant growth, which makes a species less dependent on soil nutrient supply and thus weakens nutrient competition among plant species (Killingbeck 1996;van Heerwaarden et al. 2003). Moreover, this nutrient resorption process can improve plant nutrient-use e ciency, reduce nutrient loss with litterfall decomposition, and eventually increase plant tness in nutrient-poor environments (Vergutz et al. 2012). Therefore, nutrient resorption is an essential trait that provides an alternative strategy for plants to adapt to drier and warmer climates that substantially impoverish global ecosystems (Berdugo et al. 2020;Ren et al. 2018).
Atmosphere sulfur (S) deposition is a main driver of soil acidi cation which has been regarded as a global environmental issue (Sullivan and Gadd 2019;Vet et al. 2014). Despite large reductions in S deposition during the last few decades across the world (Dentener et al. 2006), chronic deposition continues to acidify terrestrial surface soils and its legacy effects are substantially in uencing ecosystem nutrient cycling (Xiao et al. 2020;Yang et al. 2012). Sulfur-induced soil acidi cation could inhibit soil nitri cation and reduce the ratio of soil NO 3 − to NH 4 + by favouring NH 4 + accumulation, thus unbalancing soil mineral N pool (Chen et al. 2013;Kemmitt et al. 2005;Pan et al. 2020). Furthermore, soil acidi cation could increase soil P availability via exchanging PO 4 3− from soil minerals and promoting activity of acid phosphatase (Jaggi et al. 2005). However, this direct S supply would promote soil S availability in a greater degree than that of N and P. As a result, chronic but continuous S deposition translates into asynchronous increases in soil N, P, and S availability to differentially promote plant nutrient uptake (Brown et al. 2000;Jaggi et al. 2005;Stutter et al. 2004;Wang et al. 2002).
Previous studies simulating S deposition with manipulative S addition found that S addition could enhance N concentrations in both green leaf and plant litterfall (Wang et al. 2002, because of synergistic interactions between N and S during plant assimilation (Li et al. 2019). Similarly, leaf P concentration increased with S addition as a response to soil P mobilization with acidi cation (Sherman et al. 2006;Singh et al. 2012). Therefore, leaf nutrients (such as N, P, and S) are tightly coupled in nature ecosystems (Ågren et al. 2012;Nazar et al. 2011;Sardans et al. 2012;Tallec et al. 2009). In stressful environmental conditions, however, plants would excessively store some of these nutrients (Chapin et al. 1990;, somewhat leading to the decoupling of plant nutrients (He et al. 2008;Yuan and Chen 2015). However, the N and P resorption process can drive these nutrients to be re-coupled in plants (Lü et al. 2016). As such, stoichiometric N:P ratios strongly vary with leaf physiological status and environmental stresses, which may be not feasible to be universally used as a reliable indicator for plant N and P limitation (Yan et al. 2017). However, the role of nutrient resorption in driving nutrient coupling under scenarios of S deposition still remains elusive.
As soil nutrients become more available with atmospheric S deposition, plants may reduce their dependence on nutrient resorption concurred with a decrease in nutrient resorption e ciency (NuRE, Lü et al. 2020;Wright and Westoby 2003;Yuan et al. 2015). This negative link between NuRE and soil nutrient availability as a paradigm has been found in many ecosystems (Lü et al. 2013;Ren et al. 2018;Zong et al. 2018). While these studies on plant NuRE mainly focused on the key growth-limiting nutrients of N and P (Su et al. 2021), resorption e ciency of another macronutrient S (SRE) is rarely studied and if SRE ts in this 'plant NuRE-soil nutrient availability' paradigm under S-deposition scenarios remains large unknown.
Plant NuRE is an integrator of the effects from various factors, such as climatic conditions, plant physiological status, and soil resources (Suseela et al. 2015;Yuan et al. 2005). For instance, the negative effects of soil nutrient availability on N and P resorption e ciencies were only shown in the wet years instead of the dry years (Ren et al. 2018). This is because drought may slow down internal nutrient transportation within plants or decouple plant-soil interactions , thus cutting off the above-mentioned 'plant NuRE-soil nutrient availability' feedback loop. Moreover, drought can release plant nutrient-acquisition intensity via shortening plant lifespan under water stress and therefore shows no impact on NuRE (Drenovsky et al. 2019). Because of these complex interactions, studies on temporal variability of nutrient resorption with interannual precipitation are need to achieve a mechanistic understanding of how plant NuRE responding to S deposition.
The meadow steppe in northern China is sensitive to global change (Yang et al. 2012). In the past two decades, atmospheric S deposition rate has doubled in the northeast areas of China due to economic development and energy consumption, even though the average rate across the country decreases (Yu et al. 2017). Sulfur deposition contributed to soil acidi cation across grassland ecosystems in northern China, with the largest decrease of 0.80 pH units in the meadow steppe (Yang et al. 2012). The in uence caused by the S deposition on leaf nutrient resorption is rarely considered in this grassland area. But such information is critical to develop a more comprehensive understanding of the factors regulating nutrient conservation strategies in these grasslands. Therefore, our main aim was to investigate the responses of leaf nutrient concentrations and resorption e ciency of two dominant species to S addition in two contrasting wet and dry years in a meadow steppe. We hypothesized that (1) S addition would increase N, P and S concentrations in both green and senesced leaves with stronger coupling relationships among nutrients in senesced leaves than in green leaves as driven by nutrient resorption; and (2) leaf NuRE would decrease with S addition due to increased soil nutrient availability, but this would mainly show in the wet year instead of the dry year.

Materials And Methods
Site description and experimental design The S addition experiment is located in the Erguna Forest-Steppe Ecotone Research Station, Inner Mongolia, China (50°10′ N, 119°22′ E; elevation 550-600 m). Mean annual precipitation (MAP) is 363 mm with about 70% falling between May and September. Mean annual temperature is -2.45°C. The grassland is dominated by Leymus chinensis (perennial grass), Carex duriuscula (sedge), Stipa baicalensis (perennial bunchgrass), which account for almost 75% of total aboveground biomass. The soil is classi ed as haplic chernozem according to the IUSS Working Group WRB (2015). The pH of topsoil was 6.8-7.0. No grazing and fertilizer were received prior to this experiment.
The experiment was established in a homogeneous and at eld following a randomized block design, with eight S addition rates (0, 1, 2, 5, 10, 15, 20 and 50 g S m −2 year −1 ) randomly assigned into each block. Each treatment had ve replicates. Elemental S was added once a year in mid-May since 2017 and continued each year. Puri ed S powder fertilizer (elemental S > 99%) was weighed and mixed with 200 g soil, and then spread evenly on the surface of the soil in each plot. Atmospheric S deposition at the site is approximately below 3 g S m −2 yr −1 (Yu et al. 2017) but is expected to increase due to industrial and transportation development (Yu et al. 2017). The high doses of S addition were much higher than the actual local S deposition level, and the aim is to simulate the long-term and accumulative effects of ecosystem S enrichment as caused by anthropogenic activities.
Field sampling and laboratory measurements Plant and soil samples were collected in the second (410 mm, 13% higher than the MAP) and third year (266 mm, 27% lower than the MAP; Fig. S1) of S treatments. At least 20 healthy plant individuals of L. chinensis and C. duriuscula with mature and fully extended green leaves were randomly selected in each plot and homogenized into one composite sample in early August. The senesced leaves were sampled in the same way in early October for the two sampling years. All the collected leaves were oven-dried at 65°C for 48 hr to constant weight and then ground with a ball mill for chemical analyses (Retsch M400, Retsch GmbH, Haan, Germany). Leaf N concentration was analyzed with an automatic element analyzer (Vario MACRO cube, Elementar Analysensysteme GmbH, Germany). For total P and S concentrations, 300 mg leaf samples were digested with HNO 3 -HClO 4 solution and then determined with inductively coupled plasma mass spectrometer (5100 ICP-OES, Perkin Elmer, America).

Calculations and statistical analyses
Nutrient resorption e ciency (NuRE) was quanti ed as the percentage change of a nutrient in senesced leaves relative to the green leaves using the following equation: where Nu g and Nu s are the N, P or S concentration in green and senesced leaves; MLCF is the mass loss correction factor with a value of 0.64 and 0.713 for forbs and graminoids respectively as reported by Vergutz et al. (2012).
Data were tested for normality using the Kolmogorov-Smirnov test before performing ANOVA. Two-way ANOVAs were used to explore the main and interactive effects of S addition and sampling year on nutrient concentrations in green and senesced leaves and NuRE of two dominate species, with block included as a random factor. Regression models were used to quantify relationships between S addition rates and leaf nutrient characteristics with the best curve-tting results chosen based on coe cient of correlation. Pearson correlation analysis was used to examine the relationships between soil nutrient availability and leaf nutrient concentrations and NuRE. All these analyses were performed using SPSS16.0 (SPSS Inc., Chicago, USA).

Results
Responses of plant available soil nutrients and leaf nutrient concentrations to S addition Sulfur addition signi cant decreased soil pH in the two sampling years (Table S1). Soil NH 4 + was the dominant inorganic N form compared with NO 3 − , especially with the increasing S-addition rates. Effects of S addition on plant available nutrients were similar in the two year, where soil NH 4 + , available P, and available S increased but NO 3 − decreased with S addition (Table S1).
For L. chinensis, green leaf N concentration was unaltered in the wet year (i.e., 2018), while it signi cantly decreased with S addition in the dry year (i.e., 2019) (Fig. 1a). Green leaf P concentration was unaltered in the two years (Fig. 1b). Green leaf S concentration increased signi cantly with S treatments only in the wet year (Fig. 1c). For C. duriuscula, green leaf N concentration signi cantly increased by S addition only in the wet year (Fig. 1d). Green leaf P and S concentrations of C. duriuscula increased with S addition in two years (Fig. 1e,f).
Senesced leaf N, P, and S concentrations for both L. chinensis and C. duriuscula increased nonlinearly with S addition across the two sampling years, except for senescent leaf N concentration of C. duriuscula showing a linear increase in the dry year ( Fig. 1g-l).

Coupling relationships between nutrients in green and senesced leaves
The coupling relationships between leaf N and P concentrations in senescent leaves of two species were higher than that in green leaves when two years were pooled (Fig. 2a,b vs. Fig. 2c,d). In the wet year, the relationships between leaf N and P in senescent leaves were more tightly than that in green leaves (Fig. 2). However, the coupling relationship of N and P in senescent leaves was weaker than that in green leaves in the dry year (Fig. 2).
Responses of nutrient resorption e ciency to S addition For L. chinensis, the reductions in leaf NRE, PRE, and SRE were noticeably similar in response to S addition in the dry and wet years (Fig. 3a-c; Table S2). For C. duriuscula, leaf NRE and PRE increased and then decreased with S addition in the wet year (Fig. 3d,e). Leaf NRE in the dry year and SRE in the two years decreased linearly with S addition (Fig. 3d,f; Table S2). For both species, NuRE was signi cantly higher in the wet year than that in the dry year ( Fig. 3; Table S2).

Correlation analyses
Only for L. chinensis, green leaf N concentration positively correlated with soil NO 3 − in the dry year 2019.
Only for C. duriuscula, green leaf P concentration positively correlated with plant available P in the wet year 2018. However, green leaf S positively related to plant available S for L. chinensis in the wet year and for C. duriuscula in the two years (Table S3).
For both species, leaf NRE and PRE were mostly independent of plant available N and P, except for NRE of C. duriuscula in year of 2018 (Table S4). However, leaf SRE and plant available S concentration was negatively correlated for both species in the two sampling years (Table S4).

Discussion
Nutrient resorption can decrease plant reliance on soil nutrient pool and substantially in uence plant growth, survival and reproduction (Yuan and Chen 2015). Given by the fact that atmospheric S deposition and its related legacy effects (e.g., soil acidi cation) are still an important environmental issue in the grassland ecosystems of northern China (Yu et al. 2017), information on how S deposition impacts plant nutrient resorption would be helpful in understanding plant community assembly and plant species adaption to global change factors. However, no studies as we know have yet investigated S-deposition effects on leaf NuRE, especially SRE in meadow grasslands, as mediated by interannual precipitation. Therefore this research was the rst to concern the coupling of N, P, and S concentrations in green versus senescent leaves and their resorption e ciency as affected by S addition and natural precipitation. As expected, we found that N and P concentrations were tightly coupled in the senescent leaves in the wet year as mainly driven by higher NuRE. However, NuRE decreased more sharply with S-addition gradient in the dry year than the wet year (Fig. 4), which was contrary to our hypothesis.

Leaf nutrient concentrations in response to S addition with stronger coupling of N and P in senescent versus green leaves
Partially consistent with our rst hypothesis, nutrient concentrations mostly increased with S addition for both green and senescent leaves with the exception for green leaf of L. chinensis (Fig. 1). Indeed, L. chinensis was suggested to stay more homoeostatic than other species (such as Carex korshinskyi) in response to exogenous nutrient addition (Yu et al. 2010). Green leaf N concentration even decreased with S addition for L. chinensis in the dry year (year 2019), (Fig. 1), which was mainly driven by soil NO 3 − concentration as evidenced by the positive correlation between the two parameters (Table S3). In contrast to L. chinensis, green leaf N concentration of C. duriuscula showed an increasing trend with S treatment, suggesting that the two species possibly have contrasting N acquisition strategy (Legay et al. 2014).
An increase in soil available P under S-addition induced acidi cation (Xiao et al. 2020; Table S1) might have accounted for the higher leaf P concentration herein. This was further evidenced by a positive correlation between green leaf P concentration and plant available P concentration for C. duriuscula in 2018 (Table S3). As expected, leaf S concentration (except for green leaf S concentration of L. chinensis in 2019) increased with the increasing S addition and positively correlated with plant available S (Table  S3) due to plant luxury absorption of S (Wang et al. 2002). Furthermore, we found that the leaf S concentration in wet year was signi cantly lower than that in dry year (Fig. 1i,l), possibly because of higher dilution effects on plant nutrient concentrations or soil SO 4 2− leaching in the wet year (Blake-Kalff et al. 2000;Li et al. 2019). Moreover, we found N, P and S in senescent leaves of the two species consistently increased with S addition which would result in higher litter quality, subsequently increasing litter decomposition rates.
Consistent with our rst hypothesis, the positive relationship between leaf N and P was much stronger in senescent leaves than that in green leaves across the two sampling years when data were pooled (Figs. 2 and 4). However, this stronger relationship in senescent relative to green leaves was only found in the wet year instead of the dry year (Fig. 2a,b vs. Fig. 2c,d). Possibly, variations in interannual environmental conditions, e.g. water availability could show remarkable in uences on the coupling of N and P in leaves (You et al. 2018;Yuan and Chen 2015). This suggested that nutrient-resorption process could tighten the relationships of leaf N and P upon the alleviation of water limitation in this semi-arid ecosystem ( Fig. 4; Lü et al. 2016;You et al. 2018). In the dry year, the coupling pattern was inversed as shown by a stronger relationship between N and P in the green leaves than that in the senescent ones (Fig. 2). This might be a result of stronger effects from biochemical and metabolic processes in green leaves to couple N and P than that from nutrient-resorption process in senescent leaves under plant water stress (Duarte 1992;Rentería et al. 2011). Previous studies also found that drought triggered leaf senescence as a physiological response (Munné-Bosch and Alegre 2004) and resulted in nutrient imbalance and decoupling . Overall, our results provided new evidence for the role of hydrologic conditions in mediating the coupling relationships between N and P in leaves with different physiological conditions under S addition.

Decreases in NuRE with S addition vary with interannual precipitation
Consistent with our second hypothesis, leaf N, P and S resorption e ciency decreased with S addition (Figs. 3 and 4). Evidences suggested that soil nutrient availability modulated plant nutrient conservation strategies so that plants in nutrient-rich environments tended to have lower resorption e ciency (Yuan and Chen 2015). Along with elemental S gradient, soil available N (mainly NH 4 + ) increased through inhibiting nitri cation (Chen et al. 2013;De Boer and Kowalchuk 2001;Xiao et al. 2020) and available P increased through sulphate replacing phosphate ions from the colloidal surface and/or via organic P mineralization (Jaggi et al. 2005). For N and P, insigni cant correlations between nutrient resorption e ciency and soil nutrient availability for both species (Table S4) suggested that leaf NuRE might be also controlled by other soil characteristics, such as soil moisture and soil pH (Yuan et al. 2005). Similarly, inconsistent relationships between plant nutrient resorption and soil nutrient availability were also found at global scale (Vergutz et al. 2012). This inconsistency could be partly caused by the divergent response of N and P resorption e ciencies to nutrient addition in nutrient-rich versus nutrientpoor habitats (Wright and Westoby 2003). Compared to NRE and PRE, SRE was more clearly correlated with plant available S in two species (Table S4), which was directly due to the exogenous supply of plant available S (Kobe et al. 2005). Therefore, our results suggest that plants tend to rely less on nutrient internal cycling when soil nutrient availability increases (Wright and Westoby 2003), but the degree of this decrease of reliance varies among different nutrients. Furthermore, S-induced decreases in nutrient resorption would substantially alter chemical composition of plant tissues and eventually the litter decomposition processes and the coupling relationship of above-and belowground nutrient cycling (Suseela and Tharayil 2018).
We also found that the response of NuRE to S addition varied with interannual precipitation with a higher NuRE in the wet year but sharper decreases with increasing S rates in the dry year (Figs. 3 and 4). This may be due to the longer reproductive stage and stand age under lower water stress, when plants have to invest more nutrient and energy on both vegetative growth and reproductive efforts (Brant and Chen 2015). Additionally, enhanced leaching of nutrients from leaves and soils could result in higher NuRE (Lu et al. 2019). Consistently, Ren et al. (2018) suggested that N and P resorption e ciency decreased with nutrient addition in the wet year, but that these effects attened in the dry year in a desert grassland. However, we can not rule out the possibility that other factors, i.e., plant species characteristics, soil nutrient pool size (Lü et al. 2012), type and amount of S additions (Lü et al. 2013) and habitat type (Kobe et al. 2005) can in uence the response of NuRE to S addition. Therefore, the single-dimensional 'plant NuRE-soil nutrient availability' paradigm should be reconsidered as a multifaceted feedback network that nutrient availability is coordinated by other factors (e.g., precipitation in this study) to in uence plant NuRE (Fig. 4).
Nevertheless, this exibility of NuRE in response to temporal variations in precipitation and nutrient availability can shed light on the spatially resource-dependent strategies of plant nutrient use in microscale soil fertility patchiness (Lü et al. 2012). Therefore, the exible and species-speci c NuRE can help explain heterogeneity in species distribution and species turnover with exotic species colonization and native species extinction within plant communities. To our knowledge, the linkages of NuRE with plant species turnover and the consequential community assembly have rarely been tested (Lü et al. 2019). Our study calls for future working investigating the in uences of temporal and spatial variations in resources on plant NuRE at both species and community level in order to understand the role of NuRE for evidencing plant community dynamics.

Conclusion
The study found that two dominant species from the semiarid meadow grassland tended to increase nutrient (N, P, and S) concentrations in leaves but to reduce nutrient resorption e ciency with S addition during two consecutive years. Leaf N and P were more tightly coupled in senescent leaves than that in green leaves averaging across the two years, which could be mainly driven by nutrient resorption. The decrease of leaf NuRE with S addition was possibly due to mobilization of plant available nutrients in soil under acidi cation. Nutrient resorption was also regulated by interannual precipitation as evidenced by the fact of higher NuRE in the wet year and that only leaf SRE was closely correlated with plant available S. Higher NuRE in the wet year suggested enhancement of plant nutrient requirements in less waterlimited conditions, likely derived from longer plant reproductive stage and stand age in such conditions. To our knowledge, the study was the rst to reveal the distinct coupling relationships between N and P concentrations in green versus senescent leaves and that interannual precipitation substantially modulate the response of nutrient resorption to S deposition. These ndings imply the important role of nutrient resorption, as a plant-nutrition integrator comprising the in uences from multifaceted ecological processes, in affecting plant species competition, plant community assembly, and the associations between above-and below-ground nutrient cycling. Further work is clearly required to establish a linkage between nutrient resorption and plant community dynamics under global change scenarios and to verify the multidimensional feedback network of NuRE responding to global change factors as proposed in the current study.

Declarations
Author contribution YJ designed the study. RW, TL, and HL set up the eld experiment and applied fertilizer every year. XF, TL, and JC performed eld and laboratory works and data statistical analyses. RW and XL contributed to the interpretation and discussion of the results. XF and RW drafted the manuscript with suggestions from all the co-authors.

Figure 4
A multidimensional framework of the effects of sulfur (S) deposition on coupling of nitrogen (N) and phosphorus (P) and nutrient (i.e., N, P, and S) resorption e ciency (NuRE) as mediated by precipitation. Positive coupling relationships between N and P were tightened by nutrient resorption process, as shown by blue dots with plus symbol '+' showing positive relationships. NuRE decreased with increasing S addition rates as shown by red dots with a minus symbol '-' showing negative relationships. NuRE was higher in the wet year but showing a sharper decrease with S-addition rates in the dry year.