Nutrient allocation strategies of woody plants: an approach from the scaling of nitrogen and phosphorus between twig stems and leaves

Allocation of limited nutrients, such as nitrogen (N) and phosphorus (P), among plant organs reflects the influences of evolutionary and ecological processes on functional traits of plants, and thus is related to functional groups and environmental conditions. In this study, we tested this hypothesis by exploring the stoichiometric scaling of N and P concentrations between twig stems and leaves of 335 woody species from 12 forest sites across eastern China. Scaling exponents of twig stem N (or P) to leaf N (or P) varied among functional groups. With increasing latitude, these scaling exponents significantly decreased from >1 at low latitude to <1 at high latitude across the study area. These results suggested that, as plant nutrient concentration increased, plants at low latitudes showed a faster increase in twig stem nutrient concentration, whereas plants at high latitudes presented a faster increase in leaf nutrient concentration. Such shifts in nutrient allocation strategy from low to high latitudes may be controlled by temperature. Overall, our findings provide a new approach to explore plant nutrient allocation strategies by analysing the stoichiometric scaling of nutrients among organs, which could broaden our understanding of the interactions between plants and their environments.

reported that leaf biomass scaled as the 3/4 power of stem biomass in seed plants, showing that plants allocate more biomass to stems than to leaves across a large scale. By contrast, the understanding of scaling relationships of nutrients among organs, especially for woody plants, is very limited.
Leaves of vascular plants play a crucial role in conducting photosynthesis, whereas twig stems provide mechanical support and transport water, carbohydrates and nutrients [21][22][23] . N and P allocated to leaves are the key components of metabolic and photosynthetic apparatuses 24 , whereas N and P allocated to twig stems play important roles in respiration 15 , internal nutrient recycling 1,25 , and photosynthate loading and export in the phloem 24 . Owing to the functional linkages between these two organs, nutrient concentrations between leaves and twig stems should also have tight associations via plant nutrient allocation strategies. Two recent studies revealed that scaling exponents of stem N (or P) to leaf N (or P) were > 1, likely because plants required higher nutrient investments in stems (for photosynthate loading and export in the phloem 24 ) than in leaves as plant nutrient increased 2,7 .
Like plant functional traits (e.g., leaf N, leaf P, leaf mass per area or leaf lifespan), the allocation of nutrients among organs can be considered as another plant trait that may reflect the interactions between plants and their environments per se. Previous studies indicate that functional groups, climate and soils have significant influences on plant functional traits, and jointly determine their biogeographic patterns 14,23,[26][27][28][29][30] . Therefore, we hypothesize that nutrient allocation among plant organs varies among functional groups and changes with environmental factors, causing a unique biogeographic pattern. Specifically, we hypothesize that: (1) as plant nutrient concentration increases, there is a faster increase in twig stem nutrient concentration for evergreen plants relative to deciduous plants and for legume relative to non-legume; (2) as plant nutrient concentration increases, plants at low latitudes show a faster increase in twig stem nutrient concentration, whereas plants at high latitudes show a faster increase in leaf nutrient concentration; and (3) both functional group and temperature together drive the latitudinal pattern of nutrient allocation, whereas soil nutrient availabilities and precipitation play modest roles in determining this pattern. In this study, we are to test these hypotheses by examining how functional groups, climate and soils jointly control the stoichiometric scaling of twig stem N (or P) to leaf N (or P) of 335 woody species from 12 forest sites in eastern China.

Results
Patterns of scaling exponents across functional groups. There were significant stoichiometric scaling relationships between twig stem N (or P, or N:P ratio) and leaf N (or P, or N:P ratio), but their scaling exponents varied among functional groups (Table 1; Fig. 1). Scaling exponents of twig stem P to leaf P (α P ) and twig stem N:P ratio to leaf N:P ratio (α N:P ) were highest in evergreen broad-leaved plants, followed by deciduous broadleaved and coniferous plants ( Table S2). Moreover, both α N and α P were significantly correlated with mean annual temperature (MAT) (p < 0.05), but weakly correlated with annual precipitation (AP), soil total N (TN) and total P (TP) concentrations (Fig. 3). Specifically, scaling exponents increased with increasing MAT, Figure 1. Scaling relationships of twig stem N (or P, or N:P ratio) to leaf N (or P, or N:P ratio) for woody plants by functional group (deciduous/evergreen/conifer; legume/non-legume). Reduced major axis (RMA) regression was used to determine the significant line (p < 0.05). All data were log 10 -transformed before analysis.
ranging from 0.74 at − 3.3 °C to 1.45 at 21 °C for α N , and from 0.71 at − 5.7 °C to 1.36 at 21 °C for α P . By contrast, α N:P was significantly correlated with MAT and AP and soil TP, but weakly correlated with soil TN (Fig. 3i-l).

Discussion
The higher α P in evergreen broad-leaved species than in deciduous broad-leaved species (1.26 vs. 0.96 in Table 1) indicates that, as plant P increases, evergreen broad-leaved plants have a faster increase in twig stem P, whereas deciduous broad-leaved plants exhibit a faster increase in leaf P. Evergreen plants, distributed at low and mid-latitudes with low soil nutrient availability (Fig. S1) 31 , have evolved multiple adaptions to their environments, such as longer leaf lifespan, higher leaf mass per area and lower leaf photosynthetic rate than deciduous plants 14,[32][33][34] . Compared with deciduous broad-leaved plants, evergreen broad-leaved plants might require more investments in phloem loading and export apparatus to meet the higher demand for photosynthate transport 1,31 , which is associated with increasing allocation of nutrients to twig stems. Additionally, their demand for nutrient storage via twig stems for a longer leaf lifespan might also induce a higher investment in twig stems than in leaves. In contrast, deciduous broad-leaved plants with short leaf lifespan need to maximize leaf photosynthetic activity to exploit light availabilities during the shorter growing season, resulting in more nutrient allocation to leaves. Furthermore, different from α P , α N is similar between evergreen broad-leaved plants and deciduous broad-leaved plants (1.20 vs 1.19 in Table 1, p > 0.05), which indicates that the two functional types perform the common allocation strategy of N between twig stem and leaf. We infer that, compared with N, plants tend to change their P allocation strategies more easily across the functional group. For both α N and α P , legume species have higher values than non-legumes (1.44 vs. 0.99 for α N ; 1.86 vs. 0.88 for α P ) ( Table 1), meaning that legume species allocate more N and P to twig stems as plant nutrient concentrations increase. Symbiotic N 2 -fixing bacteria in roots of legume species require much carbohydrate from photosynthetic tissues, because N acquisition through biological fixation has a higher energy cost than direct N absorption from the soil 3,35 . Thus, given the higher carbohydrate demand of symbiotic N 2 -fixing bacteria, legumes should increase the photosynthate loading and export rate in the phloem, which is associated with higher N and P investments in twig stems 2 .
Both α N and α P decrease from > 1 to < 1 with increasing latitude (Fig. 2), suggesting that, as plant nutrient concentration increased, plants at low latitudes tended toward a higher increase in twig stem nutrient concentration, whereas plants at high latitudes tended toward a higher increase in leaf nutrient concentration. This indicates a gradual shift in nutrient allocation strategy across a large geographic area. Moreover, plants at high latitudes had higher ratios of leaf N (or P) : twig stem N (or P) (Fig. 4a). These latitudinal patterns of scaling exponents (α N and α P ) and nutrient ratios (leaf N/P : twig stem N/P) may be attributed to the changes in species composition from evergreen broad-leaved plants at low latitudes to deciduous broad-leaved and coniferous plants at high latitudes (Fig. S2) 13 , because α P was highest in evergreen broad-leaved plants, followed by deciduous broad-leaved and coniferous plants (1.26; 0.96; 0.70 in Table 1). However, the latitudinal pattern of α N was hardly explained by the changes in species composition because of similar α N across the three functional types. For evergreen broad-leaved plants, both α N and α P were higher in tropical forests than in temperate forests, and for deciduous broad-leaved plants, both α N and α P were higher in temperate forests than in boreal forests (Fig. 5). Thus, changes in species composition may play a modest role in latitudinal patterns of nutrient allocation. Note that in this study we did not conduct detailed analyses on the scaling exponents of evergreen needle plants at boreal forests and deciduous plants at tropical forests, and variations in the scaling exponents of coniferous plants among biomes because of the paucity of data and less representatives ( Fig. 5; Fig. S2).
Through the exploration of the relationships between scaling exponents and environmental factors, we found that MAT could be a main driver for these latitudinal patterns because of the tightest correlation between MAT and latitude (r 2 = 0.94, p < 0.001 in Table S4). As the temperature increased, scaling exponents increased from < 1 to > 1, meaning that, as plant nutrient concentration increased, plants at low temperature exhibited a higher increase in leaf nutrient concentration, whereas a higher increase in twig stem nutrient concentration occurred for plants at high temperature. Plants at low temperature also had higher ratios of leaf N (or P) : twig stem N (or P) (Fig. 4b). Previous studies have revealed that temperature is the main driver of biogeographic patterns of leaf N and P concentration 27,30,36 , and biomass allocation 37 . According to the "temperature-plant physiological hypothesis" 27 , low temperature induces an increase in leaf nutrient concentrations to offset diminished efficiency of N-rich enzymes and P-rich RNA. However, compared with leaves, twig stems are less influenced by temperature, because their optimal functions are less dependent on nutrient compositions 6 . Thus, more nutrients allocated to leaves at low temperature are probably an adaptive strategy for plants to maintain the functional equilibrium for fitness. Furthermore, temperature largely drives the changes in growing season length and leaf lifespan 34 , which might influence plant nutrient allocation strategies. Decreasing temperature shortens the growing season length and leaf lifespan for deciduous broad-leaved plants 34 , hence more nutrients are allocated to leaves to maximize leaf functions for exploiting light resources during the shorter growing season.
In addition, although we did not detail the allocation of C concentrations, as showed in Fig. S3, the scaling exponents of twig stem C to leaf C concentration increased with MAT, suggesting that, as plant C concentration increased, plants at high temperature tended to a larger increase in twig stem C concentration. This further supports that a higher phloem loading and carbonhydrate transport in twig stems occurred for plants at high temperature, resulting in a higher increase in twig stem nutrient concentration.
Precipitation and soil nutrients might be potential drivers for these biogeographical patterns of nutrient allocation, considering their important roles in plant growth and biogeochemical cycling 5 . However, our study found that there were no significant relationships between scaling exponents and AP, soil TN and soil TP (Fig. 3). Eastern China exhibits a steeper thermal gradient than moisture gradient 13 . Given the high soil heterogeneity, soil TN and soil TP might not directly reflect the real nutrient availability for plants. In addition, these environmental factors vary collaterally with each other (Table S4), and their independent roles might be difficult to detect. Thus, future studies are needed to reveal their single effects on plant nutrient allocation strategies. Additionally, there was little impact of altitude on these patterns of nutrient allocation strategies because all scaling exponents showed no significant relationships with altitude (Fig. S4). Changes in the environmental conditions along the altitudinal gradient relative to latitudinal gradient were much smaller (Table S1), and thus the influence of altitude should be covered by that of latitude.
Plant N:P ratio is widely used as an indicator of N and P limitation in terrestrial ecosystem 38 . Schreeg et al. (2014) 6 found that the scaling exponent of stem N:P ratio to leaf N:P ratio for tropical tree seedlings was > 1,  suggesting that stem N:P ratio might be a better indicator of soil nutrient availability. However, whether this could be applied to other species is little known. In this study, we found that α N:P varied among functional groups and along the latitudinal gradient. Woody plants at tropical forests show a higher variability of twig stem N:P ratios with α N:P > 1, whereas a more constrained twig stem N:P ratios with α N:P < 1 occurred for plants at boreal forests (Fig. 2f). Thus, whether twig N:P or leaf N:P is more sensitive to soil nutrient availability or other factors might depend on functional groups and sites.
Our findings suggest that functional groups and environmental factors jointly influence allocation of nutrients among organs, and result in a gradual shift in nutrient allocation strategy from low to high latitudes across a large geographic scale in eastern China. Variations in the nutrient allocation across functional groups and environmental factors can be explained by the 'optimal partitioning theory' or 'functional equilibrium concept' 22,39 , which means that plants would allocate their nutrients in an optimal way to obtain the 'functional equilibrium' for fitness. This large-scale pattern is more dependent upon temperature than other environmental factors (i.e., AP, soil TN and soil TP) according to our study, indicating that allocation strategy may also follow the "temperature-plant physiological hypothesis" 27 . Our results reveal that the scaling relationship is more "variable" than "invariant", subject to the changes in plant nutrient allocation strategy with ambient environment. In general, by exploring the stoichiometric scaling of nutrients between leaves and twigs, we provide a new dimension to understand how plants regulate nutrient allocation strategies to adapt to ambient environment, and thus would broaden our knowledge about the interactions between plants and their environments from a nutrient allocation perspective.

Materials and Methods
Study sites. This study was conducted at 12 forest sites across eastern China with latitude from 18.7 °N to 50.9 °N (Fig. S1). Eastern China spans a large range of climates from cold and dry in the north to warm and moist in the south 13 , and has diverse vegetation types from boreal coniferous forest to tropical rainforest. Climate data, including mean annual temperature (MAT) and annual precipitation (AP), were obtained from local reports (Table S1). In the study area, MAT and AP ranges from − 5.7 °C to 25.3 °C and from 423 mm to 2031 mm respectively, and soil types shift from nutrient-rich brown soils to nutrient-poor tropical red soils. We classified these 12 sites into three biomes: boreal, temperate and tropical forests 40  Sampling and Measurement. We collected twig stem and leaf samples at 12 forest sites across eastern China, according to a previous protocol 41 . During the growing season (July-August), we chose dominant or common species at each site, and then collected the fully expanded sun leaves from four or five individuals of each species. Accordingly, terminal 10-20 cm of twig stems (i.e., top twig stems) that supported the sun leaves were sampled. In total, we sampled 335 woody species in 198 genera and 73 families. Samples were dried at 60 °C to constant weight and then powdered using a ball mill (NM200, Retsch, Haan, Germany) before measuring N and P concentrations. Twig stem and leaf N concentrations were determined by the Dumas combustion method using an elemental analyser (2400 II CHS/O, Perkin-Elmer, USA). Twig and leaf P concentrations were determined using the molybdate/ascorbic acid method after H 2 SO 4 -HClO 4 digestion 42 . For each site, we also collected three soil samples in three plots where these plants were sampled. A horizon samples of soils were randomly sampled and then thoroughly pooled for each plot to represent one soil sample. Soil samples were air-dried, sieved through a 2-mm mesh, handpicked to remove plant detritus, and ground to pass through a 100-mesh sieve. We measured soil total N (TN) and total P (TP) concentrations using the same method for plant samples. Average values of three soil samples at the same site were calculated to represent the soil TN and TP concentration of each site. Statistical analysis. Stoichiometric scaling relationships of twig stem N (or P, or N:P ratio) and leaf N (or P, or N:P ratio) were analysed using all original data of N and P concentrations and N:P mass ratios from individual plants (see Appendix S1 for details). Data were log 10 -transformed before analysis. RMA regression was used to determine the scaling function 43 , which was expressed by log 10 Y = α*(log 10 X) + β, where X and Y represent leaf N (or P, or N:P ratio) and twig stem N (or P, or N:P ratio), and α and β are the slope (i.e., scaling exponent) and intercept of regression line respectively. We performed these analyses in three ways. First, we divided all original data into respective functional groups (conifer/deciduous broad-leaved/evergreen broad-leaved; legume/ non-legume) and three biomes (tropical/temperate/boreal forests), and then compared their scaling exponents. Second, we performed scaling analyses with original data for each site, and then related scaling exponents to the latitude and environmental factors (MAT, AP, soil TN, soil TP) of each site using linear regressions. Third, we compared scaling exponents of evergreen broad-leaved plants from tropical and temperate forests, and scaling exponents of deciduous broad-leaved plants from temperate and boreal forest. A likelihood ratio test was used to indicate the heterogeneity of RMA regression exponents among groups 43 . In addition, linear regressions were used to explore the relationships between log 10 -transformed ratios of leaf N (or P) : twig stem N (or P) and latitude and environmental factors. All statistical analyses were performed using R 2.15.2 44 .