Relationship Between Belowground Carbon Allocation and Nitrogen Uptake in Saplings Varies by Plant Mycorrhizal Type

While it has long been hypothesized that belowground carbon (C) allocation in plants is tightly coupled to nutrient uptake, empirical tests of this are rare, especially for woody plants. We grew tree saplings of nine species in soils enriched in isotopically-labeled nitrogen (N) and after several months, pulse-labeled trees with 13CO2. This approach allowed us to track how 13C allocation from foliage to absorptive root tissue related to 15N movement from soil to plant tissues as a measure of each species’ N return on C investment. We hypothesized that tree species known to associate with ectomycorrhizal (ECM) fungi would have greater belowground C fluxes than those that associate with arbuscular mycorrhizal (AM) fungi, and that species with greater belowground C allocation would acquire the most soil N. Overall, we found large interspecific differences in both the amount of recently-fixed C allocated belowground and plant N uptake, yet no differences in either flux between AM and ECM trees (P > 0.05). Moreover, we found no differences between mycorrhizal groups in terms of their N return on C investment. However, mycorrhizal type influenced the relationship between belowground C allocation and N uptake, which was positive in AM species (r = 0.42; P =0.001) and negative in ECM species (r = -0.44; P =0.003), suggesting that the relationship between C allocation and plant nutrition is more complex than theory predicts. Collectively, our results suggest that tree species’ nutrient return on C investment can differ greatly among species, and that efforts to model these dynamics should consider the traits and tradeoffs that underlie these dynamics.


INTRODUCTION
Plant productivity is frequently limited by nutrients (Elser et al., 2007;LaBauer and Treseder, 2008) and consequently a large portion of primary production is allocated belowground to acquire nutrients from soil (Gill and Finzi, 2016). While numerous studies have highlighted the importance of C allocation and fine root production in forests (Raich and Nadelhoffer, 1989;Litton et al., 2007;McCormack et al., 2015), our understanding of the factors that control belowground C fluxes remains limited (Grayston et al., 1997;Epron et al., 2012). This results from the fact that dynamic C allocation belowground (as opposed to static measurements of C pools such as fine root biomass and root: shoot relationships) is challenging to study in situ and is likely to be highly variable within and among plant species (Grayston et al., 1997). Given that variation in belowground C allocation can affect soil C storage, nutrient retention and plant uptake (Pendall et al., 2004;Cheng et al., 2014), an improved understanding of species-specific effects of belowground C dynamics is needed to accurately predict how future species shifts will affect ecosystem functioning (Nguyen, 2003;Heimann and Reichstein, 2008;Cheng et al., 2014).
Trees can allocate anywhere from 5 to 25% of recently fixed C belowground (Grayston et al., 1997;Phillips and Fahey, 2005;Drake et al., 2011), yet the impacts of belowground C fluxes on nutrient availability in forests are not well-established. Theory predicts trees will preferentially allocate energy (i.e., C) to acquire the most limiting resource (Bloom et al., 1985;Chapin et al., 1990;Franklin et al., 2012). In most temperate forests, nitrogen (N) limits net primary production (LaBauer and Treseder, 2008), suggesting that belowground C allocation and N acquisition should be coupled processes. Using a meta-analytical and modeling approach, Finzi et al. (2015) found support for such C-for-N "return on investment" in forests. They estimated that up to one-third of total soil N mineralization may be fueled by C allocated belowground and delivered to soil via rhizodeposition. In a separate meta-analysis, Gill and Finzi (2016) reported that forests with the largest belowground C fluxes have the fastest rates of N mineralization-consistent with the idea that more C allocated belowground should result in more plant N uptake, at least in N-limited forests. However, greater belowground C allocation under low soil N availability could have the opposite effect on plant uptake. If belowground C inputs to soil increase plant-microbe competition for N, increases in N immobilization could reduce plant N uptake (Kuzyakov and Xu, 2013). Such a scenario would be likely in plants that produce C that is of low chemical quality as a microbial substrate.
Theory also predicts that the C cost of N uptake is a function of not only soil N availability but also the type of mycorrhizal fungi a plant associates with . Recent work shows plant mycorrhizal association can be considered an integrative plant functional trait, with plants from distinct mycorrhizal groups displaying differences in other plant nutrient traits such as leaf and root N (Averill et al., 2019), leaf and root litter decay rates (Keller and Phillips, 2019;See et al., 2019) and foliar resorption . Moreover, trees associated with ectomycorrhizal (ECM) fungi are believed to allocate more C to their symbionts, which in turn have the capacity to mine N from SOM. In contrast, trees associated with arbuscular mycorrhizal (AM) fungi, typically lacking such capabilities, are thought to allocate less C to their symbionts as they commonly grow in more N-rich soils Lin et al., 2016). Several studies have shown that ECM trees may release more C to soil (Phillips and Fahey, 2005;Yin et al., 2014) and that such inputs may lead to root-accelerated N cycling in ECM-dominated stands (Brzostek et al., 2012;Yin et al., 2014; but see Wurzburger and Brookshire, 2017). However, the return on investment of AM and ECM trees may differ depending on soil N availability. In a modeling analysis, Brzostek et al. (2015) reported that while ECM trees have a greater N return on C investment in low N soils, AM trees may have a greater N return on C investment in high N soils. This may explain, in part, why AM trees are becoming increasingly abundant (relative to ECM trees) in areas that receive the highest levels of N deposition (Jo et al., 2019).
Improved quantification of interspecific differences in belowground C allocation, N uptake, and N return on C investment may facilitate the inclusion of belowground traits into other plant resource use theory (e.g., the plant economic spectrum). Numerous studies have reported weak links between species' aboveground and belowground traits (Mommer and Weemstra, 2012;Weemstra et al., 2016), a pattern that has been attributed to the multidimensionality of roots (Mccormack and Iversen, 2019) and the lack of appropriate belowground trait data. If belowground C allocation and N uptake can be predicted by other more easily measured belowground traits, improved parameterizations can be used to model these critical plant-soil fluxes at large spatial scales in the wake of global changes (McCormack et al., 2017).
The goals of this study were to quantify interspecific differences in belowground C allocation and N uptake, as a means of estimating interspecific variation in N return on C investment. To this end, we conducted a dual 13 C and 15 N pulse labeling greenhouse experiment with nine tree species (5 AM, 4 ECM species) where we simultaneously traced the flux of recently photosynthesized 13 C belowground and the uptake of 15 N into tree saplings. We hypothesized that ECM trees would allocate more C belowground (relative to AM trees) resulting in greater N uptake from soil. Accordingly, we predicted that relative to AM trees, ECM trees would allocate a greater percentage of fixed 13 C to their absorptive roots, resulting in greater soil-derived 15 N recovered in plant tissues. Moreover, across all species we expected to find a positive relationship between the amount of 13 C allocated belowground and plant 15 N.

Experimental Set-Up
We conducted a dual ( 13 C and 15 N) isotopic labeling study of nine tree species, comprising two types of mycorrhizal associations (AM and ECM) under greenhouse conditions. We selected five AM-associated tree species: Acer saccharum (sugar maple), Juglans nigra (black walnut), Liriodendron tulipifera (tulip poplar), Prunus serotina (black cherry), Sassafras albidum (sassafras), and four ECM-associated tree species: Betula nigra (river birch), Quercus alba (white oak) Quercus rubra (red oak), and Tilia americana (American basswood). At the start of the study, saplings were 2 years old (obtained from Vallonia State Nursery, Indiana, USA) except for S. albidum and T. americana (obtained from Cold Stream Farm Nursery, Michigan, USA) which were 1 year old. All trees were "bare root" (i.e., obtained without soil) and consequently likely not colonized by mycorrhizal fungi.
Ten experimental saplings of each species were grown in a forest soil-sand mix (due to some sapling death, the final number of replicates varied by species slightly, Table 1; additional 'unlabeled control' saplings of each species were also grown in the same conditions). We collected all soil from Lilly-Dickie Woods in Brown County, Indiana, USA (39 • 14 N, 86 • 12 W; altitude 290 m). For detailed description of the soil and microbial community properties of these soils see Cheeke et al. (2017) and Craig et al. (2018). Within this forest, we selected from two stands with either >90% basal area of AM-associated or ECMassociated tree species. At each stand, we removed the standing litter layer and collected soil from the top 10 cm (including Oe and Oa, where present, and A horizons). We removed large clumps of roots and large rocks and then mixed each soil type with equal parts sand that had been steamed at 80 • C for 4 h. Saplings were planted in individual pots (19.7 cm width × 45.7 cm height, 9.63 liters volume) in an AM or ECM soil: sand matrix in a fully reciprocal species × soil type design. However, using two-way analysis of variance (ANOVA) we found no soil type or species × soil type effect for any variable measured (P > 0.10 in all cases) and subsequently grouped all replicates of each species together for data analysis. We allowed all saplings to grow in pots for ∼2 months to develop their rooting systems.
After 2 months, we added 15 N-labeled plant leaf material to each pot (hereafter referred to as particulate organic N, PO 15 N). The addition of an organic source of 15 N (as opposed to just adding inorganic 15 N) allowed us to investigate the degree to which trees can shift their belowground C allocation to mine N from organic matter. To do this, we removed saplings from their pots, mixed in 0.75 g/1 L soil of PON 15 to the bulk soil, and repotted the saplings. The same procedure was performed for saplings that did not receive PO 15 N, to serve as unlabeled controls (one unlabeled sapling per soil type × species combination). Six soil-only controls were also included. PO 15 N was produced by growing a different set of saplings (same tree species, obtained from the same respective nurseries) in a 50:50 sand: Turface R fritted clay mixture amended with 15 Nlabeled Hoagland's solutions wherein 30% of the N was 98% 15 N ( 15 NH 15 4 NO 3 and K 15 NO 3 ). After several months, we collected foliage from all nine species, dried it at 60 • C for 48 h and ground it into a coarse powder using a mortar and pestle. We combined equal masses of foliage from each species to make a homogeneous composite source of PO 15 N.
All saplings were grown for an additional 2 months before pulse labeling with 13 C. This allowed time for the PO 15 N to decompose and plants to take it up. Sapling height was measured twice over the course of the experiment, once immediately after repotting and once immediately prior the start of the 13 C pulse labeling period.

C Pulse Labeling of Saplings
To track the flux of recently photosynthesized C belowground, we used a custom-made gas-tight chamber, following the protocol of Kannenberg and Phillips (2017). In brief, we pulse-labeled four saplings at a given time (groupings were selected using a randomized block design with species as the blocking factor), with the stem and foliage of each sapling housed within the gastight, transparent chamber while the pot (and therefore all soil and belowground tissues) remained outside the chamber. Thus, all 13 C recovered in the roots or soils resulted from belowground C allocation dynamics within each plant. Labeling occurred on sunny or partly sunny days under full greenhouse lighting and lasted for 4 h during mid-day. After flushing the chamber with pure N 2 gas, we injected 50 mL of 99% 13 CO 2 into the chamber at the beginning of each labeling period. We injected a second 50 mL of 13 CO 2 after 2 h to prevent significant drawdown of CO 2 inside the chamber over the course of the labeling period. The control saplings, as described above, were placed in the chamber and exposed to the same conditions as the labeled trees with the exception of using standard CO 2 gas rather than 99% 13 CO 2.

Harvesting and Laboratory Analyses
We sampled foliage from each pot immediately following pulse labeling and soil from each pot 1, 3, and 5 days after pulse labeling. We also sampled foliage immediately prior to labeling to provide a baseline from which to calculate total 13 CO 2 fixation by each sapling during the 4 h labeling period. To sample foliage, we cut or punched holes in two leaves from each sapling and composited these subsamples into one sample per sapling per time point. We then oven-dried foliage samples at 60 • C for 48 h, ground the tissue to a fine powder, and weighed 3 mg of each sample into tin capsules for 13 C and 15 N analysis (see below).
We sampled soil from each pot by taking two soil cores (2.5 cm diameter) to the depth of the pot and compositing the cores into one soil sample per pot per time point. The soil was immediately transferred to the lab, where we sieved and picked out all roots. For soil samples with sufficient root biomass, we processed the root tissue as described above for foliage tissue and analyzed root 13 C and 15 N (see below). Due to insufficient root biomass in some cores, the sample size for absorptive root measurements varied slightly among species.
Five days post labeling, we harvested each entire pot. We separated plant biomass into foliage, stem, and transport and absorptive root pools, and collected a representative subsample from the homogenized soil. We also collected two absorptive root (1st-3rd) samples from each sapling. Root samples were thoroughly cleaned of soil with water and immediately scanned on an Epson GT 20000 scanner. Mass-specific root length (SRL) and average root diameter measurements were obtained from the scanned images using WinRhizo software (Regent Instruments, Quebec City, Quebec, Canada). Plant tissues, including scanned root samples, were oven-dried at 60 • C for 48 h and weighed. Foliage and root tissues (not including scanned roots) were processed as described above. Transport and absorptive roots were separated into distinct pools from the intact oven-dried root systems of each sapling following Kong and Ma (2014).

Isotopic Analyses
We determined δ 13 C and δ 15 N, as well as elemental C and N, for plant pools using a Costech ECS 4010 Elemental Analyzer (Costech Analytical Technologies, Valencia, CA) coupled to a ThermoFinnigan DELTA plus XP isotope ratio mass spectrometer (Thermo Fischer Scientific, San Jose, CA). Carbon isotopic ratios were converted to the Vienna Pee Dee Belemite (VPDB) international scale by calibration with Acetanilide 1 and EDTA-1 and N isotopic ratios were compared to atmospheric N 2 (Schimmelmann et al., 2009). Analyses were conducted at the TABLE 1 | Species-specific differences in root mass: shoot mass, absorptive root mass: transport root mass, foliar and root % N, absorptive root diameter, and specific root length (SRL).

Species (n)
Root: We quantified the incorporation of newly photosynthesized 13 C and 15 N into distinct plant tissue pools (foliar, stem, transport and absorptive roots) by first calculating atom % for C and N separately as: Atom % = (100 * STND * (δ/1000 + 1))/(1 + STND * ( δ/1000 + 1)) where STND represents the standard isotopic value for VPDB (0.0111802) and atmospheric N 2 (0.0036764) for 13 C and 15 N, respectively. Excess δ 13 C and δ 15 N for each pool were then calculated as: Excess 13 C or 15 N (ng) = (atom % L − atom % UL )/100 * DW * (% C or % N)/100 * 10 6 where atom % L is atom % of labeled material and atom % UL is atom % unlabeled material, DW is the dry weight of the plant (mg), and % C or % N represents the elemental composition of the tissue. Isotopic enrichment of foliage, stem and absorptive root tissue was measured for each individual sapling. Isotopic enrichment of transport roots was measured on two randomly selected experimental saplings per species and one control sapling per species. Whole plant 15 N excess (ng N) was calculated by adding foliar, stem, transport and absorptive root 15 N excess pools together. We then calculated the N return on C investment in two ways: (1) plant 15 N (ng N) / absorptive root 13 C (ng C/ ng foliar 13 C) and (2) whole plant N content (g N)/absorptive root biomass (g C).

Statistical Analysis
We first tested if foliar atom % 13 C was significantly different in labeled compared to unlabeled samples using a Student's ttest. We used a one-way analysis of variance (ANOVA) and Tukey's HSD post-hoc test to assess species differences in sapling growth (% change in height), SRL, absorptive root diameter, and tissue 13 C and 15 N content. We then used step-wise multiple linear regression to assess if there were significant relationships between 13 C and 15 N in plant tissues and if such relationships varied by mycorrhizal group. We subsequently explored how belowground C allocation and N uptake relate to other plant traits in a multidimensional trait space using PCA. All analyses were conducted in R 3.3.2 (R Development Core Team).

C Labeling of Plants
We found a significant increase in foliar atom % 13 C in labeled compared to unlabeled plants after the 4 h 13 CO 2 pulse labeling treatment (t = 13.598, P < 0.001), indicating successful labeling of foliar tissue in our experimental saplings (foliar atom % 13 C labeled mean = 1.25, standard error = 0.13; unlabeled mean = 1.08, standard error = 0.00046). While there was significant variation in leaf mass [F (8, 54) = 7.581, P < 0.001] and tree height [F (8, 54) = 98.8, P < 0.001] {and change in height over time [F (8, 54) = 10.7, P < 0.001]} among species, there was no difference in these metrics between mycorrhizal groups (mean tree height across species at end of experiment: 83.11 cm ± 5.13; mean increase in height across species: 28.22 cm ± 3.74). Species-specific plant traits are presented in Table 1.
On average, ∼10% of fixed 13 CO 2 was transferred from the foliage to the roots within 5 days post-labeling (estimated as the ratio of foliar 13 C enrichment immediately after labeling to absorptive root 13 C 5 days post-labeling), indicating significant movement of newly photosynthesized C from aboveground to belowground tissues. There was no time point × species interaction for root 13 C enrichment, and therefore peak 13 C FIGURE 1 | Species patterns in allocation of recently-fixed C allocated to absorptive roots. Total 13 C absorptive root enrichment (A) and relative 13 C absorptive root enrichment per unit 13 C fixed (B) shown. Species means are show with standard error bars. Light bars indicate AM-associated tree species, dark bars indicated ECM-associated species. Letters indicate significant pairwise differences (P < 0.05). enrichment across time points for each sapling was used in subsequent analyses. Root 13 C enrichment (as calculated in Equation 2) varied significantly among species [F (8, 54) = 2.3, P < 0.001, Figure 1A; F (8, 54) = 3.575, P < 0.001, Figure 1B] but not between mycorrhizal types. There was a 16-fold difference in root 13 C enrichment (ng C/ ng foliar 13 C) among species; Acer saccharum (sugar maple; AM-associated) had significantly higher root 13 C enrichment compared to T. americana (American basswood; ECM-associated).

N Labeling of Plants
The addition of PO 15 N to each pot resulted in significantly enriched 15 N-soil in the experimental vs. control pots (P < 0.001). As evidenced by 15 N-enriched plant tissue in our experimental saplings, a detectable proportion of this soil-derived 15 N pool was subsequently decomposed and taken up by plants. Similar to belowground C allocation, plant 15 N uptake varied among species but not between mycorrhizal types [F (8, 54) = 2.58, P = 0.015, Figure 2]. Among species, there was a 4-fold difference in plant 15 N enrichment, with A. saccharum (sugar maple; AM-associated) having the greatest PO 15 N uptake among the nine species.

Plant N Return on C Investment
By comparing root 13 C tissue enrichment to plant 15 N tissue enrichment (quantifying new plant N uptake), we can assess the relationship between plant N demand and dynamic belowground C allocation. Among species, there was a more than 10fold difference in mean plant 15 N enrichment: absorptive root 13 C enrichment (P = 0.012, SI Figure 1). We found a significant positive relationship between plant 15 N enrichment and absorptive root 13 C for AM species (r 2 = 0.18, P = 0.001) but a significant negative relationship for ECM species (r 2 = 0.19, P = 0.003; Figure 3). When accounting for variation in plant N content among species, we found similar but weaker relationships with a negative relationship for ECM species (r 2 = 0.15, P = 0.017) but no significant relationship among AM species (SI Figure 2). However, there were no mycorrhizal type differences in N return on belowground C investment when calculating this ratio either by 13 C and 15 N tissue enrichment or total C and N tissue concentrations (as described in methods; SI Figure 1). Using PCA, we found variation in increased root 13 C aligned with increased allocation to absorptive compared to transport roots. Meanwhile, mass-specific root length (SRL) and root diameter were orthogonal to root 13 C enrichment. The first two orthogonal PCA axes explained 31.5 and 20.9%, respectively, of the variation in our dataset (Figure 4).

DISCUSSION
Trees allocate a significant amount of C belowground to acquire limiting resources, yet we lack a predictive understanding of what controls belowground C flux and how this relates to plant nutrient uptake. Using a dual 13 C and 15 N isotopic labeling approach under controlled greenhouse conditions, we examined how transfer of recently fixed C belowground varied across nine different tree species that are known to associate with either AM or ECM fungi. We then assessed how variation in belowground C allocation related to plant N uptake. We found that fixed 13 CO 2 was rapidly transferred from sapling foliage to absorptive roots, and this rate varied across species but not between mycorrhizal groups (Figure 1). Notably, we found that the percentage of 13 C allocated to absorptive roots was related to plant N uptake, but this relationship depended on plant mycorrhizal type: AM trees showed a positive relationship between root 13 C and plant 15 N, while this relationship was negative for ECM trees (Figure 3). This suggests that the N return on C invested belowground may be greater for AM tree species relative to ECM tree species and that plants' return on belowground investment may be mediated by distinct plant functional group traits.

Belowground C Dynamics
Our results highlight the importance of species identity in driving belowground C allocation patterns. Most evidence of species differences to date comes from studies comparing only a few species (e.g., Phillips and Fahey, 2005;Yin et al., 2018), mainly non-woody plants (e.g., Cheng, 2009) and often from broadly distinct plant functional groups (e.g., Dijkstra et al., 2009). Here, we expand on this work and report evidence of species-specific patterns across nine temperate deciduous tree species.
Notably, principal component analysis revealed close alignment between the vectors describing the variation in absorptive: transport root biomass and those for root 13 C enrichment (Figure 4). Why were species with the most absorptive roots the strongest sinks for recently-fixed C? One hypothesis is that absorptive roots contain the greatest density of rhizosphere microbes, which increase C sink strength by assimilating C released from roots as rhizodeposits. Other pulse-labeling studies have found a correlation between 13 C recovered in roots and microbially processed 13 C recovered in FIGURE 2 | Species patterns in sapling plant tissue 15 N enrichment, compared to saplings grown without 15 N-labeled particulate organic matter. Species means are shown with standard error bars. Light bars indicate AM-associated tree species, dark bars indicated ECM-associated species. Letters indicate significant pairwise differences (P < 0.10).
FIGURE 3 | Variation in belowground allocation of recently-fixed C (as measured by absorptive root 13 C enrichment) in relation to plant N uptake (as measured by whole plant 15 N enrichment). Individual species are shown with unique symbols, with filled shaped representing ECM species. Linear regression lines of best fit for each mycorrhizal type are shown with 95% confidence interval individually. For AM species, r 2 = 0.18, P = 0.001; for ECM species, r 2 = 0.19, P = 0.003. the soil or as CO 2 efflux (SI Figure 3; Dilkes et al., 2004;Phillips and Fahey, 2005;Pausch et al., 2012;Karst et al., 2016), indicating that much of the recently-fixed C allocated to roots ultimately ends up in soil (Karst et al., 2016). Alternatively, interspecific differences in belowground C allocation may relate to differences in C surplus and transport. Under nutrient limiting conditions, plants export C from leaves to other tissues to avoid biochemical inhibition of photosynthesis (Sharkey et al., 1986). In a study using three of the nine species used in this study (sugar maple, tulip poplar and white oak), export of recently-fixed C out of leaves was greater for sugar maple and tulip poplar relative to white oak (Kannenberg and Phillips, 2017), which aligns with our findings (Figure 1).
We did not find any evidence to support our hypothesis that ECM trees allocate more C belowground than AM trees. This contrasts with both theory and field studies of mature trees that have reported that ECM-associated tree species allocate more C to absorptive roots, mycorrhizal hyphae and root exudates compared to AM-associated species (Brzostek et al., 2012(Brzostek et al., , 2014Yin et al., 2014, but see Liese et al., 2017. One possible explanation for our finding of no mycorrhizal group differences is that our roots were likely only weakly colonized by mycorrhizal fungi. While we did not measure mycorrhizal colonization of root tissues in this study, a previous investigation using a subset of species (maple, tulip and white oak) and similar growing conditions found that few of the roots contained mycorrhizal structures (Kannenberg and Phillips, 2017). Thus, the lack of a mycorrhizal type effect on rhizosphere C flux in our study may point to the importance of mycorrhizal fungal partners as key drivers of root C allocation and movement of C from plant to soil pools Högberg et al., 2008). To the extent that plants are able to actively regulate C transfer to mycorrhizal partners, and given some evidence that ECM fungi typically have a higher plant C cost compared to AM fungi (Field et al., 2017), AM and ECM plant functional type differences in rhizosphere C fluxes may be more pronounced when mycorrhizal colonization rates are high. This warrants future field studies quantifying AM and ECM differences in root C fluxes into the soil and assessing the degree to which % mycorrhizal colonization drives these patterns.

Plant N Uptake and N Return on C Investment
Our results also reveal significant species variation in plant 15 N uptake, with evidence that more acquisitive root traits (e.g., specific root length) enable greater N uptake (Figure 4). While we initially hypothesized greater plant 15 N uptake would be positively related to root 13 C across all species, only AM species showed a significant positive relationship. In contrast, root 13 C and plant 15 N were negatively related across ECM species (Figure 3). These distinct patterns of C-N coupling between mycorrhizal types may be due, in part, to differences in rhizodeposition stoichiometry. Stoichiometric variation in root-derived compounds could mediate opposite patterns in microbially-mediated soil N cycling and plant N uptake (Drake et al., 2013). For example, release of highquality substrates into the rhizosphere should stimulate microbial activity, induce a positive priming effect and increase plantavailable inorganic N. In contrast, low quality rhizodeposits may induce greater microbial community N limitation, enhanced microbial N immobilization and lower plant-available N (Cheng and Kuzyakov, 2005;Kuzyakov and Xu, 2013). Many ECM plants contain higher concentrations of condensed tannins and phenolics than AM plants (Midgley et al., 2015;Fox et al., manuscript in preparation), and these compounds can slow the decay of root-derived inputs (Kraus et al., 2003;Sun et al., 2018). To the extent that root chemistry reflects rhizodeposit chemical quality, rhizodeposition stoichiometry could play a key role in driving distinct patterns of soil C-N coupling between mycorrhizal types.
Overall, the positive relationship between belowground C allocation and plant N uptake observed among AM species aligns with theory suggesting plants allocate more energy belowground to acquire limiting resources and subsequently benefit from this preferential allocation. Why then did ECM trees show the opposite pattern (i.e., a negative relationship between belowground C allocation and N uptake)? Plant N uptake capacity also mediates the nature of this "C for N" exchange. That is, plants with a poor ability to take up N (per unit C) may need to allocate substantially more C to absorptive roots to compensate for the low N uptake capacity, while plants with high N uptake capacity per unit C would be expected to allocate less C to absorptive roots. Such a pattern would result in a negative relationship between absorptive root 13 C and plant 15 N, as observed among ECM species.
Importantly, allocation of newly fixed C to absorptive roots is only one of a suite of possible nutrient uptake strategies and our results highlight the degree to which complex suites of morphological and physiochemical traits control plant-soil C and N dynamics. As such, a simple "C for N" economic investment framework may not fully capture how plant C allocation relates to resource uptake across plant functional types. For example, two morphological root traits (i.e., SRL and absorptive root diameter) were orthogonal to root 13 C, while absorptive: transport root mass aligned closely with root 13 C, suggesting important but complex links between plant morphological traits and their effects on element cycling. This builds on previous work that has shown such traits to be an important factor driving species-specific patterns of rhizosphere priming effects, given that root morphology influences the rhizosphere surface area: root biomass ratio (Cheng et al., 2014). Additionally, root maintenance incurs a C cost that is not captured by commonly measured root traits. Differences between mycorrhizal groups in this maintenance cost could also help explain distinct C-N coupling patterns between AM and ECM plant species. Finally, roots are tasked with acquiring diverse resources belowground, and different suites of traits may be advantageous for different resource environments. For example, plants in water-limited environments may prioritize deeper, woody roots while nutrient-limited plants may allocate more energy toward shallow, absorptive roots. Taken together, our work suggests that evidence of a well-coordinated root economic spectrum may remain elusive (Mommer and Weemstra, 2012;Mccormack and Iversen, 2019) due to the multi-functional nature of roots.
Our greenhouse results motivate future field studies of species effects on rhizosphere C-N coupling. Additional studies are needed to understand how effects of saplings may differ from those of mature trees. While saplings may have higher rhizodeposition (Fu and Cheng, 2002), mature trees may exert stronger rhizosphere effects given increased total root mass (Kuzyakov, 2002). Additionally, although biotic and abiotic soil properties are recognized as important factors driving rhizosphere dynamics , how these factors interact with plant species effects under field conditions to influence C-N coupling in forests remains poorly quantified. For example, while N-limitation is common in temperate forests, other nutrients (e.g., Ca, P) as well as water can also control plant productivity and C allocation. Additionally, while N stocks tend to be greater in AM than ECM temperate forest soils (Craig et al., 2018), more of this N may be bound up in protected forms in AM soils (Craig et al., 2019) and therefore not as susceptible to microbial priming. Future research should consider how such soil physiochemical variables may mediate belowground C dynamics and plant nutrition.
Overall, we found species-specific differences in belowground C allocation across nine tree species, indicating the importance of species identity in driving rhizosphere C dynamics. Additionally, we found the relationship between belowground C allocation and plant N uptake depended on plant mycorrhizal type. While roots and root-microbe interactions are increasingly being recognized as key components of ecosystem and earth system models (Ostle et al., 2009;McCormack et al., 2017), the sensitivity of these models to species identity remains poorly understood. Moreover, species effects on rhizosphere dynamics appear to be sensitive to global change (Cheng et al., 2012;Sulman et al., 2017). Thus, future data-fusion efforts combining greenhouse and field studies with modeling experiments are called for to develop accurate and predictive carbon-climate products.

AUTHOR CONTRIBUTIONS
AK and RP conceived and designed the experiment, revised, and wrote the final manuscript. AK carried out the experiment, performed the data analysis, and wrote the first draft.

FUNDING
We acknowledge the support of the U.S. Department of Energy Office of Biological and Environmental Research, Terrestrial Ecosystem Science Program (Award# DESC0016188) and the US National Science Foundation Ecosystem Studies Program (1153401).