Rhizosphere stoichiometry: are C : N : P ratios of plants, soils, and enzymes conserved at the plant species-level?



  • As a consequence of the tight linkages among soils, plants and microbes inhabiting the rhizosphere, we hypothesized that soil nutrient and microbial stoichiometry would differ among plant species and be correlated within plant rhizospheres.
  • We assessed plant tissue carbon (C) : nitrogen (N) : phosphorus (P) ratios for eight species representing four different plant functional groups in a semiarid grassland during near-peak biomass. Using intact plant species-specific rhizospheres, we examined soil C : N : P, microbial biomass C : N, and soil enzyme C : N : P nutrient acquisition activities.
  • We found that few of the plant species' rhizospheres demonstrated distinct stoichiometric properties from other plant species and unvegetated soil. Plant tissue nutrient ratios and components of below-ground rhizosphere stoichiometry predominantly differed between the C4 plant species Buchloe dactyloides and the legume Astragalus laxmannii. The rhizospheres under the C4 grass B. dactyloides exhibited relatively higher microbial C and lower soil N, indicative of distinct soil organic matter (SOM) decomposition and nutrient mineralization activities.
  • Assessing the ecological stoichiometry among plant species' rhizospheres is a high-resolution tool useful for linking plant community composition to below-ground soil microbial and nutrient characteristics. By identifying how rhizospheres differ among plant species, we can better assess how plant–microbial interactions associated with ecosystem-level processes may be influenced by plant community shifts.


The discovery of consistent elemental ratios among marine plankton and the ocean environment by Redfield (1958) spurred decades of research into ecological stoichiometry. The Redfield ratio is not universal, but rather represents the average stoichiometry of the phytoplankton species inhabiting the modern ocean (Klausmeier et al., 2004), and it could change in response to evolution or environmental changes (Quigg et al., 2003; Konneke et al., 2005; Finkel et al., 2010). Unlike the well-mixed surface oceans, terrestrial environments are characterized by relatively high heterogeneity in resources, community composition and ecosystem properties (Green et al., 2008). Similar to aquatic systems, however, sunlight and nutrient availability also constrain survival of terrestrial plants (Geider et al., 2001). There is evidence that the fundamental evolutionary and biochemical constraints operating in marine systems also affect the stoichiometry of organisms and resources in terrestrial ecosystems (Sinsabaugh et al., 2009; Finzi et al., 2011).

At the global scale, soils and microbial biomass carbon (C) : nitrogen (N) : phosphorus (P) ratios are constrained to a fairly narrow range across ecosystems that differ widely in climate, geology and plant community composition (Cleveland & Liptzin, 2007). Within that narrow range, soil and microbial elemental stoichiometry have been shown to differ among biomes (Cleveland & Liptzin, 2007) and vary across latitudinal climate gradients (McGroddy et al., 2004). On the other hand, the stoichiometry of terrestrial plant communities varies across a broad range, and depends in part on the relative abundance of structural woody tissue with broad nutrient ratios (i.e. C : P, c. 1100; N : P, c. 50) compared with metabolic tissue with more narrow nutrient ratios (i.e. C : P, c. 230; N : P, c. 20) (Reiners, 1986). Most previous studies have focused on global relationships, and it is less clear if these patterns are exhibited at smaller scales.

At the ecosystem scale, soil microbial properties, including microbial community composition, plant litter decomposition rates and microbial biomass stoichiometry, can vary widely depending on plant species' tissue stoichiometry (Hobbie et al., 2006; Martiny et al., 2006). As a result, the relative demand for N and P can span a large range among plant species within a single ecosystem (Güsewell, 2004). Nutrient requirements are critical determinants of plant competition and affect species assemblages within plant communities. There is ample evidence that plants are able to exert some control over nutrient availability through interactions with the soil microbial community that produces enzymes to degrade organic matter and subsequently mineralize soil nutrients for microbial and plant uptake (Richardson et al., 2009; Bezemer et al., 2010; Eisenhauer et al., 2010). There is further evidence suggesting that some plants can modify their soil environment in a manner that favors their persistence. For example, when the C3 annual grass species Bromus tectorum invades, the soil communities in its rooting zone change relative to those beneath native plants (Belnap & Phillips, 2001; Kuske et al., 2002; Belnap et al., 2005). These soil communities tend to have higher N mineralization rates, which is thought to be integral to the persistence of B. tectorum (Schaeffer et al., 2011). Stoichiometric relationships that exhibit a relatively higher change in consumer stoichiometry (y-axis) compared with resource stoichiometry (x-axis) are indicative of positive nutrient feedbacks (no homeostasis), promoting competitive nutrient exclusions among plant species (Bever, 2003; Hawkes et al., 2005; Levine et al., 2006). By contrast, a negative nutrient feedback (homeostasis) supports plant species coexistence with lower proportional changes in consumer stoichiometry compared with resource stoichiometry (Bever, 2002; Sterner & Elser, 2002b; Diez et al., 2010). Evaluating homeostatic relationships can provide valuable insight into assessing plant competition (Klironomos, 2002; Elgersma et al., 2012) or plant coexistence (Schnitzer et al., 2011) within the context of plant–microbe feedbacks (Sterner & Elser, 2002c), particularly in highly diverse grassland ecosystems.

The stoichiometric relationships among plant, microbial, and soil components within the soil zone directly impacted by roots, known as the rhizosphere (de Graaff et al., 2010; Orwin et al., 2010), have not been previously examined. In the rhizosphere, the cycling of nutrients between soils and plants is mediated by microbes and the enzymes that they produce to depolymerize organic substrates. Thus, there is a strong theoretical basis for predicting that the stoichiometry of plants, microbes, enzymes and soil should be coupled (Sterner & Elser, 2002b). For example, plant litter N : P was shown to correlate with the N : P biomass ratio of the microbes decomposing the litter, suggesting that microbial biomass stoichiometry is affected by the stoichiometry of their substrates (Fanin et al., 2013). Plants also affect the activity and composition of soil microbial communities through alteration of the physical environment during root growth; moisture regimes through shading, mulching, and transpiration; nutrient availability through nutrient uptake; and substrate availability through root exudation (Zak et al., 2003; Johnson et al., 2004; Bird et al., 2011).

Microbes can also control the stoichiometry of assimilable resources in their local environment by altering the allocation of resources towards production of enzymes that degrade substrates rich in C, N, P, and other nutrients (Sinsabaugh & Moorhead, 1994; Allison et al., 2011; Hall et al., 2011). Phosphatase enzyme activities have been shown to increase with N fertilization (in relatively P-limited systems) and decrease with P fertilization (Treseder & Vitousek, 2001). Thus, this general pattern of enzyme allocation has been suggested to follow the trajectory of ecosystem development. As P becomes limiting in well-developed ecosystems, microbes will potentially increase the production of phosphatase enzymes (Allison et al., 2007); at which point higher P enzyme activities would negatively correlate with microbial biomass P concentrations. At the global scale, P-degrading enzyme activity increases as climate becomes warmer and wetter, consistent with increasing P limitation and narrowing soil enzyme C : P acquisition activities (Sinsabaugh et al., 2009). While enzyme nutrient acquisition ratios are consistent indicators of resource limitations relative to the stoichiometric demands of microbial biomass at large scales, we do not yet know if these relationships also apply to the rhizosphere scale.

In this study, we examined the relationships among the stoichiometry of plant tissues, microbial biomass, enzymes, and soils within the rhizosphere of individual plants in a semiarid grassland ecosystem. We collected plants with intact root systems and rhizosphere soil for eight different species in a semiarid grassland in Wyoming, USA. We examined nutrient stoichiometry of soil enzyme acquisition activities, microbial biomass, mineral soil, and plant roots and shoots to address three fundamental questions: does rhizosphere stoichiometry differ among plant species; if so, is the stoichiometry of plant, microbial, and soil components correlated within plant rhizospheres; and, lastly, how does the stoichiometry of plant rhizospheres compare with the range found within plant-free soil? We hypothesized that the stoichiometry of plants, microbes, and soils is strongly correlated within individual rhizospheres. Furthermore, we hypothesized that rhizosphere stoichiometry varies more among species than among individuals of the same species because of coevolved plant–microbe interactions (McGroddy et al., 2004; Frank & Groffman, 2009), implying that plant community assemblages directly influence below-ground ecosystem functioning. We also expected that the stoichiometry of exoenzyme activities would negatively correlate with microbial biomass stoichiometry. Alternatively, rhizosphere stoichiometry may be more strongly regulated by the variation among individual plants, and not conserved at the plant species level. This study is among the first to explore the utility of rhizosphere stoichiometry for relating plant community composition to below-ground C, N and P dynamics as an index of overall ecosystem functional traits.

Materials and Methods

Site description

The study site is a northern semiarid grassland dominated by the perennial C4 grass Bouteloua gracilis (H.B.K) Lag., and two C3 grasses, Hesperostipa comata Trin and Rupr. and Pascopyrum smithii (Rydb.); c. 20% of the plant biomass is composed of sedges and forbs. The site is located at the USDA-ARS High Plains Grassland Research Station (1930 m above sea level (asl)), 15 km west of Cheyenne, WY, USA (41°11′N, 104°54′W). Annual precipitation is 384 mm, with c. 60% falling during the growing season (March–September). Mean air temperatures are 17.5°C in summer and −2.5°C in winter. Soils are Mollisols (fine-loamy, mesic Aridic Argiustoll, mixed Ascalon and Altvan series), with an average pH of 7.0 and organic soil C content of 1.9% (SD = 0.3) at 0–5 cm.

Experimental design

Our research questions focused on the strong influence of actively growing plant species relationships on rhizosphere stoichiometry components. Therefore, in May 2011, we identified plant species isolates or homogeneous patches (in the case of the grasses) for eight plant species as well as for unvegetated soil patches (visually devoid of plant growth) to represent nonrhizosphere soil. Four replicates for each plant species and unvegetated soil cores were established within a sample area of c. 1000 m2. The plants selected for observation represented all plant functional groups occurring at the study site, including seven native species and one exotic forb species, Linaria dalmatica (Table 1). After identifying all plant species and unvegetated soil areas, we installed 10-cm-diameter PVC collars around individual plant replicates (and in the bare soil areas) to a depth of 10 cm. We have very high confidence that the plant tissue and rhizosphere soil samples that were collected for analysis were associated with specific plant species chosen for observation. Likewise, during sampling (upon visual inspection) we were able to verify that no roots from other plant species grew into the collars after they were inserted into the soil. Lastly, to avoid excessive soil drying within the cores throughout the growing season, we added water to all cores on four different occasions, totaling 20% of the annual mean precipitation. Water additions (before the 21 July 2011 harvest) were applied on 7 June, 27 June, 1 July, and 7 July (2011).

Table 1. Plant species names (with abbreviation), plant functional groups, and common names for the plants selected to examine rhizosphere stoichiometry
Species name (code)Functional groupCommon name (characteristics)
  1. All plants were sampled in July 2011 at the experimental semiarid grassland site in southern Wyoming (USA).

Astragalus laxmannii (ASLA)LegumeAstragalus (tap-root system)
Bouteloua gracilis (BOGR)C4Blue Grama (grass)
Buchloe dactyloides (BUDA)C4Buffalo grass (grass)
Koeleria cristata (KOCR)C3June grass (bunchgrass)
Hesperostipa comata (HECO)C3Needle and thread (bunchgrass)
Linaria dalmatica (LIDA)ForbToadflax (noxious weed)
Chrysopsis villosa (CHVI)ForbGolden aster (native)
Artemisia frigida (ARFR)ForbFringed sage (native)

Sampling and processing

We harvested soil and plant material (or soil samples devoid of any plant growth) as single, intact plant/soil or soil-only cores (in the case of unvegetated soils) within each PVC collar down to 10 cm depth on 21 July, coinciding approximately with peak biomass. Immediately after removal, we divided each plant rhizosphere core into bulk soil, rhizosphere soil and plant material. The rhizosphere soil was operationally defined as the soil that remained attached to the roots after loosening and shaking the core. Rhizosphere soil was carefully removed from the roots using a 2 mm sieve in the field and subsequently stored at 4°C. Within 24 h of collection, rhizosphere soil was extracted with K2SO4 (15 g of soil in 35 ml of 0.05 M K2SO4) and processed using the chloroform fumigation extraction method for microbial biomass C and N (Vance et al., 1987) (k = 0.45). Fumigated and unfumigated extracts were analyzed for total extractable organic C and total extractable N using a total organic carbon analyzer with an N measuring unit (Shimadzu TOC-VCPN; Shimadzu Scientific Instruments, Wood Dale, IL, USA). Unfumigated extracts were analyzed for inorganic N (NH4+ and NO3) using an O I Analytical Flow Solution IV (OI Analytical, College Station, TX 77842-9010, USA). We combined extractable NO3–N and NH4–N to represent the total available N pools. Likewise, we felt that using both forms of inorganic N (NO3 or NH4) was a relatively robust representation of available soil N (as a result of the highly variable in vivo dynamics characteristic of these inorganic N forms) to assess stoichiometric relationships among plant species' rhizosphere components. Rhizosphere soil subsamples were collected to assess extractable phosphates using 5 g soil in 50 ml −0.5 M NaHCO3 for colorimetric P analysis using the ammonium molybdate-ascorbic acid reagent as described by Olsen & Sommers (1982). Units for microbial biomass carbon (MBC), microbial biomass nitrogen (MBN), soil organic C, soil PO4, and total soil N are expressed as μg−1 g−1soil. The rhizosphere soil pH was measured with an ion-specific probe using a 2 : 1 soil – DI H2O paste extract (Robertson et al., 1999).

Plant material was sorted into coarse roots (> 2 mm), fine roots (< 2 mm), stems and leaves and dried (60ºC) within 48 h, then ground to a fine powder for chemical analyses. Percentage C and % N of fine roots and leaves were measured on a Finnigan DeltaPlus XP connected to a Carlo Erba NC-2500 elemental analyzer via a Finnigan ConFlo III open-split interface. Plant leaf P was extracted from 0.5 g ground tissue ash (550°C) by digesting the ash in 5 ml of 6 N HCl at 80°C for 30 min. Extracts were diluted to 100 ml and analyzed on a Perkin Elmer inductively coupled plasma optical emission spectrometer calibrated against four standards (Soltanpour et al., 1996).Units for root and leaf C and N and leaf P are expressed as % DW.

We measured the potential activity of seven hydrolytic soil enzymes that degrade a range of substrates that are common constituents of organic matter. These enzymes were selected to represent the degradation of C-rich substrates (β-1,4-glucosidase, β-d-cellubiosidase, α-glucosidase, and β-xylosidase), N-rich substrates (β-1,4-N-acetylglucosaminidase and leucine aminopeptidase) and P (phosphatase) (Sinsabaugh et al., 2009). Enzyme assays were conducted using standard fluorimetric techniques (Saiya-Cork et al., 2002; Wallenstein et al., 2009, 2012; Bell et al., 2013). In brief, assays were conducted by homogenizing 2.75 g of soil in 91 ml of 50 mm sodium acetate buffer (pH 6.8) in a Waring blender for 1 min. The soil slurries were then added to a 96-deep-well (2 ml) microplate using an eight-channel repeat pipettor. Additional quench control replicates of soil slurry and 4-methylumbellfferone (MUB) or 7-amino-4-methylcoumarin (MUC) standard curves (0–100 μM concentrations) were included with each sample. Soil slurries with fluorometric substrates were incubated for 3 h at 25°C. After the incubation period, plates were centrifuged for 3 min at 2900 g, after which 250 μl of soil slurry was transferred from each well into a black Greiner flat-bottomed 96-well plate (into corresponding wells) and then scanned on a TECAN Infinite M200 microplate reader using excitation at 365 nm and emission at 450 nm. Units for all enzyme nutrient acquisition activities are expressed as nmol activity g−1 dry soil h−1.

Statistical methods

The Shapiro–Wilk test of normality and Levene's test of equal variances was performed in R: Language for Statistical Computing using R ‘stats’ and ‘car’ packages, respectively (Fox & Weisberg, 2011; R Core Team, 2013), to assess if any univariate parameter distributions among plant species groups significantly deviated from normal. Following Box-Cox transformations using the R: MASS package (Box & Cox, 1964; Venables & Ripley, 2002), we selected natural-log data transformations (as needed) to improve the assumption of normality and homoscedasticity for all subsequent statistical analyses. Means and standard errors were calculated for all parameters assayed using SPSS v 20.0 (IBM SPSS Statistics for Windows, ver. 20.0; IBM Corp., Armonk, NY, USA).

Distance-based redundancy analysis (dbRDA) was used as the statistical criterion to assess overall differences in rhizosphere stoichiometry among plant species and plant functional groups using the R: vegan package (Oksanen et al., 2013). Stoichiometry components used in the dbRDA model included total C : N : P enzyme acquisition activities, soil extractable C : N : P, and soil microbial biomass C : N. We chose dbRDA over other multivariate statistical approaches because it has nonlinear distance-metric options with robust multidimensional resolution to assess categorical variables, which is a well acknowledged approach for ecological studies (Legendre & Anderson, 1999). Distance-based RDA is a three-step ordination technique that tests the effects of response parameters (i.e. stoichiometry) on defined ecological groups (i.e. plant species). First, a dissimilarity or distance matrix is calculated for the community. We selected the Bray–Curtis dissimilarity (nonlinear) measure to model the species matrix as suggested by Legendre & Anderson (1999). In steps 2 and 3 of the dbRDA, a principal coordinate analysis (PCoA) is calculated based on the distance matrix, from which the eigenvalues (obtained in the PCoA) were applied to an RDA.

One-way analysis of variance (ANOVA) was used to determine differences in univariate components of rhizosphere stoichiometry (i.e. enzyme C : N : P nutrient activities, soil C : N : P, plant leaf C : N : P, microbial biomass C : N and plant root C : N ratios) among plant species rhizosphere and bare soil samples using SPSS v. 20.0 (SPSS: IBM Corp). Nonratio univariate parameters were also analyzed using ANOVA to assess significant below-ground differences among plant species and bare soil samples. Tukey post hoc comparisons were used for all univariate analyses.

Pearson correlations were calculated to determine how stoichiometry of plant, microbial, and soil components correlated within plant rhizospheres using SPSS v 20.0 (SPSS: IBM Corp). Furthermore, homeostasis (H) indices with corresponding r2 were calculated to evaluate the relationship between rhizosphere stoichiometry (i.e. soil, microbial biomass, or enzyme activity) and species-specific plant tissue stoichiometry. Nutrient stoichiometry relationships between rhizosphere components (y ; consumer) and plant tissue (x ; resource) were evaluated by calculating the homeostatic regulation coefficient (H) using the regression equation (y = cx1/H) to solve for H using log transformed values for x and y (Sterner & Elser, 2002b). H coefficients between c. 1 and 0 demonstrate nonhomeostatic relationships between consumer and resource stoichiometry as there is a constant proportional change in resource and consumer stoichiometry. H coefficients progressively > 1 (i.e. low slopes approaching 0) indicate varying degrees of homeostatic regulation (Cleveland & Liptzin, 2007), with lower changes in consumer stoichiometry compared to resource stoichiometry (Sterner & Elser, 2002b). Using this approach, rhizosphere stoichiometry components (i.e. soil enzymatic C : N, C : P, or N : P, microbial C : N, and soil nutrient C : N, C : P, or N : P) were reported in terms of their relative homeostasis relationships with plant tissue stoichiometry (i.e. C : N, C : P, or N : P).


Below-ground nutrient stoichiometry (as reflected in nutrient ratios of soil extracts, enzymes, and microbial biomass) significantly differed among plant species, with separation along axes 1 (= 0.005) and 2 (= 0.01) accounting for 87% of the total variance among plant species (Fig. 1a). B. dactyloides (BUDA; C4 grass) differed from the majority of plant species along axis 1 (representing 48.9% of the variance explained). Wider soil C : N and MBC : N drove plant species differences horizontally (Fig. 1a,b). Soil C : P and MBC : N appeared to influence differences vertically along axis 2 (representing 37.9% of the variance explained; Fig. 1b). Stoichiometry components categorized by plant functional group significantly differed in below-ground nutrient stoichiometry along axis 1 (= 0.02; Fig. 1c), also influenced by wider soil C : N and MBC : N (Fig. 1d).

Figure 1.

Overall differences among rhizosphere stoichiometry components demonstrated by distance-based redundancy analysis (dbRDA) ordination among (a, b) plant species-specific rhizospheres and (c, d) plant species categorized into their respective plant functional group. The stoichiometry parameter scores for plant species-specific rhizospheres (b) or plant functional groups (d) include: enzyme carbon (C) : nitrogen (N), C :  phosphorus (P), and N : P nutrient acquisition activities, soil C : N, C : P, and N : P, and microbial biomass C : N. Abbreviations for plant species (a) and plant functional group (c) are positioned as multivariate centroids surrounded by 95% confidence interval (CI) ellipsoids. The scatterplots positioned to the right of the ordination plots for (b) plant species and (d) plant functional groups represent species–stoichiometry scores (i.e. coordinates) along axis 1 and 2 for the corresponding plant species and plant functional group ordination plots. These scores can be interpreted as the strength of influence that each stoichiometric response variable has in separating plant species or plant functional group along axis 1 (left–right) and/or along axis 2 (top–bottom). In brief, the stoichiometric response scores (i.e. coordinates) further from the origin (0.0, 0.0) are more influential in distinguishing rhizosphere stoichiometry among plant species (or functional groups). Overall (multivariate) differences in rhizosphere stoichiometry were considered significant at  0.05. The sample size for all plant species is = 4. Plant species abbreviations include: one legume species (pink): Astragalus laxmannii (ASLA); two C3 grass species (blue): Koeleria cristata and Hesperostipa comata (KOCR and HECO, respectively); two C4 grass species (gray): Bouteloua gracilis and Buchloe dactyloides (BOGR and BUDA, respectively); and three Forb species (yellow): Linaria dalmatica, Chrysopsis villosa, and Artemisia frigida (LIDA, CHVI, and ARFR, respectively). Bare soil samples (clear) were plotted only in (c) (to demonstrate bare soil stoichiometric relationships to below-ground rhizosphere stoichiometry dynamics).

Univariate analysis revealed that B. dactyloides exhibited significantly wider soil C : N (when compared with A. laxmannii;= 0.01; Fig. 2a), and significantly lower soil N : P compared with all plant species and bare soils (P < 0.001; Fig. 2c). B. dactyloides also demonstrated trends suggesting wider microbial biomass C : N ratios; although this trend was only significant at = 0.06 (Fig. 3). All plant species (with exception of the forb Chrysopsis villosa) exhibited significantly wider leaf C : N compared with the legume species (= 0.002; Fig. 2g). Likewise, almost all plant species demonstrated significantly wider root C : N when compared with A. laxmannii (P < 0.001; Fig. 3b).

Figure 2.

Nutrient ratios for rhizosphere components and plant tissue for eight plant species and bare soil samples. Letters above the mean ± SE indicate significant differences at  0.05 using Tukey post hoc tests. Nutrient ratios among eight different plant species and bare soils collected at the experimental semiarid grassland site in southern Wyoming (USA) for (a) soil extractable organic carbon (SOC) : total inorganic soil nitrogen (TN); (b) SOC : soil PO4; (c) TN: soil PO4; (d) ∑ enzyme C : N acquisition activities; (e) ∑ enzyme C : P acquisition activities; (f) ∑ enzyme N :  P enzyme acquisition activities; (g) leaf C : N; (h) leaf C : P; and (i) leaf N : P. BG, β-glucosidase; CB, β-d-cellubiosidase; XYL, β-xylosidase; AG, α-glucosidase; LAP, leucine aminopeptidase; NAG, N-acetyl-β-glucosaminidase; PHOS, phosphatase. Plant species include: one legume species (pink) – Astragalus laxmannii; two C3 species (blue) – Koeleria cristata and Hesperostipa comata; two C4 species (gray) – Bouteloua gracilis and Buchloe dactyloides; and three Forb species (yellow) – Linaria dalmatica, Chrysopsis villosa, and Artemisia frigida. All ratio values are presented as means ± SE;= 4. Letters above error bars indicate significant differences at  0.05 using Tukey post hoc tests. Note that the ‘overall’ (in black) demonstrates the combined means among all plants and bare soil (where applicable), included as a qualitative reference for potential ‘ecosystem-level’ stoichiometry estimates. The ‘overall’ category was not included in the statistical analysis.

Figure 3.

Carbon (C) : nitrogen (N) ratios for microbial biomass carbon (MBC) : N (a) and root tissue (b) for eight plant species and bare soil samples. Values are presented as means ± SE (= 4). Significant differences were considered at  0.05 using Tukey post hoc tests. Plant species abbreviations include: one legume species (pink) – Astragalus laxmannii (ASLA); two C3 species (blue) – Koeleria cristata and Hesperostipa comata (KOCR and HECO, respectively); two C4 species (gray) – Bouteloua gracilis and Buchloe dactyloides (BOGR and BUDA, respectively); and three Forb species (yellow) – Linaria dalmatica, Chrysopsis villosa, and Artemisia frigida (LIDA, CHVI, and ARFR, respectively). Letters above error bars indicate significant differences at  0.05 using Tukey post hoc tests. All ratio values are presented as means ± SE (= 4). Note that the ‘overall’ (in black) demonstrates the combined means among all plants and bare soil, and is included as a qualitative reference for potential ‘ecosystem-level’ stoichiometry estimates. The ‘overall’ category was not included in the statistical analysis.

In several cases, microbial, soil, and plant stoichiometry components significantly correlated within plant tissue stoichiometry (Table 2; Fig. 4). For example, leaf C : N negatively correlated with soil N : P (Table 2; = −0.64; < 0.001), but demonstrated a positive relationship with soil C : N (Fig. 4a; = 0.21; = 0.2). Furthermore, leaf N : P positively correlated with soil N : P (Fig. 4c; = 0.64; < 0.001). Microbial biomass C : N demonstrated a weak positive trend with leaf C : N (Fig. 4b; = 0.22; = 0.2) and was negatively correlated with soil N : P (Fig. 4d; = −0.47; = 0.007), as well as with soil C : P (Table 2; = −0.46; = 0.008). Overall, soil enzyme activities did not strongly correlate with microbial biomass or soil stoichiometry (Table 2; Fig. 4e). As the exception, enzyme N : P activities positively correlated with soil C : P (Table 2; = 0.40; = 0.02). Overall, these findings suggest that enzyme activities do not correspond to environmental stoichiometric characteristics. Conversely, soil C : N and N : P stoichiometry appears to constrain microbial and plant stoichiometric ranges in some cases at the plant species level (i.e. A. laxmannii when compared with B. dactyloides); suggesting that soil stoichiometry is strongly related to plant spatial distributions in this grassland.

Table 2. Pearson correlation matrix featuring stoichiometry relationships of plant, microbial, and soil components among the eight plant species' rhizospheres (overall) at this experimental grassland
 Enzyme C : PEnzyme N : PMBC : NSoil C : NSoil C : PSoil N : PRoot C : NLeaf C : NLeaf C : PLeaf N : P
  1. Significant correlations ( 0.05) are indicated in bold. The sample size for each nutrient variable is = 32. MBC, microbial biomass carbon.

Enzyme C : N 0.45 −0.340.13−0.04−0.35−0.310.340.250.07−0.27
 Enzyme C : P 0.69 −−0.07 0.38 −0.12−0.060.10
    Enzyme N : P−0.260.22 0.40 0.180.13−0.33−0.110.32
   MBC : N0.070.460.47−−0.20
    Soil C : N 0.35 0.66−−0.24
     Soil C : P0.380.02−0.310.23 0.55
      Soil N : P0.160.500.09 0.64
       Root C : N0.48−0.200.22
        Leaf C : N 0.47 0.65
         Leaf C : P0.31
Figure 4.

Microbial and soil stoichiometry components correlated within plant species' rhizospheres (= 4) (per plant species). Significant differences were considered at  0.05. Data were ln-transformed to improve the linear fit in the regression scatterplots. Nutrient ratios for plant species collected at the experimental semiarid grassland site in southern Wyoming (USA) for: (a) leaf carbon (C) : nitrogen (N)/soil C : N; (b) microbial biomass carbon (MBC)  :  N/leaf C : N; (c) leaf N :  phosphorus (P)/soil N : P; (d) MBC : N/soil N : P; (e) ∑ enzyme N : P acquisition activities/soil N : P. Plant species abbreviations include: one legume species (pink) – Astragalus laxmannii (ASLA); two C3 species (blue) – Koeleria cristata and Hesperostipa comata (KOCR and HECO, respectively); two C4 species (gray) – Bouteloua gracilis and Buchloe dactyloides (BOGR and BUDA, respectively); and three Forb species (yellow) – Linaria dalmatica, Chrysopsis villosa, and Artemisia frigida (LIDA, CHVI, and ARFR, respectively). The sample size for all ratio values per species was = 4. Significant correlations were considered at  0.05; ns, not significant.

Enzyme, soil, and plant C : N stoichiometry strongly correlated with relative C : P and N : P stoichiometry components (i.e. relative enzyme, microbial biomass, soil, and plant stoichiometry; Table 2). For example, enzyme C : N activities were positively correlated with enzyme C : P activities (Table 2; = 0.45; = 0.01); and negatively correlated with enzyme N : P activities (Table 2; = −0.34; at = 0.06). Enzyme C : P activities were positively correlated with enzyme N : P activities (Table 2; = 0.69; < 0.001). Soil C : N was positively correlated with soil C : P (Table 2; = 0.35; = 0.05) and negatively correlated with soil N : P (Table 2; = −0.66; < 0.001). Lastly, leaf C : N was positively correlated with leaf C : P (Table 2; = 0.48; = 0.006) and negatively correlated with leaf N : P (Table 2; = −0.65; < 0.001).

The majority of C3 and forb species in this study had similar microbial, soil, and edaphic characteristics (Table 3). Differences in microbial, soil, and/or nutrient components among plant species were predominantly associated with three plant species: the C4 grass B. dactyloides, the forb C. villosa, and the legume A. laxmannii (Table 3). For example, the C4 species B. dactyloides exhibited higher MBC than all other plant species and bare soil (ANOVA: = 0.002; Table 3). B. dactyloides rhizospheres also demonstrated lower soil N compared with the majority of plants, including bare soil (ANOVA:  0.001, Table 3). The legume A. laxmannii demonstrated significantly higher rhizosphere soil N along with higher root and leaf N concentrations than many of the plant species selected for this study (ANOVA: P ≤ 0.001, Table 3). The forb C. villosa demonstrated significantly higher soil pH levels (within its rhizosphere zone) than other species and bare soils (ANOVA:  0.001, Table 3). Bare soils demonstrated similar nutrient, microbial and edaphic properties to those exhibited by the majority of plant species measured (Table 3; Figs 1c, 2). For example, C, N, and P enzyme activities, MBN, soil organic C and soil PO4 concentrations did not significantly differ between plant species' rhizospheres and unvegetated soils (Table 3). Total soil N concentrations were higher in bare soil than in B. dactyloides (C4), K. cristata (C3), and A. frigida (forb) (Table 3; < 0.001).

Table 3. Descriptive statistics table displaying mean ± SE for parameters used to assess stoichiometric relationships and edaphic properties among the rhizosphere from eight different plant species at the experimental semiarid grassland site in southern Wyoming (USA)
 ∑ C enzymes∑ N enzymes∑ P enzymesMBCMBN
= 0.9= 0.35= 0.91= 0.002= 0.8
Overall173.70 ± 14.55136.59 ± 7.23178.99 ± 8.63442.28 ± 25.3035.66 ± 2.21
Astragalus laxmannii 133.19 ± 11.05150.77 ± 3.49178.96 ± 12.07401.02 ± 55.89 b36.98 ± 9.67
Bare soil130.31 ± 12.2699.60 ± 4.21151.99 ± 1.85361.78 ± 85.80 b33.70 ± 11.52
Bouteloua gracilis 187.20 ± 51.88141.10 ± 16.59180.94 ± 33.02492.47 ± 51.41 b41.93 ± 8.62
Buchloe dactyloides 162.66 ± 14.94139.50 ± 6.01159.10 ± 9.30743.40 ± 36.68 a43.33 ± 0.74
Artemisia frigida 185.72 ± 50.46125.78 ± 29.17180.23 ± 28.55377.47 ± 55.84 b28.26 ± 5.59
Chrysopsis villosa 222.97 ± 40.24181.50 ± 13.56176.23 ± 38.92406.59 ± 14.53 b40.03 ± 3.94
Koeleria cristata 173.49 ± 53.81140.08 ± 32.12210.63 ± 28.67470.95 ± 32.91 b32.63 ± 2.91
Hesperostipa comata 173.82 ± 59.89136.94 ± 28.01198.10 ± 40.04311.62 ± 23.74 b34.54 ± 11.29
Linaria dalmatica 172.95 ± 65.28108.32 ± 17.92167.99 ± 11.70384.80 ± 60.34 b29.38 ± 3.17
 Soil organic CTotal soil NSoil PO4Soil moistureSoil pH
= 0.36P < 0.001= 0.32= 0.12P < 0.001
Overall31.72 ± 2.265.01 ± 0.623.22 ± 0.176.38 ± 0.446.74 ± 0.05
A. laxmannii 29.65 ± 4.267.94 ± 1.06 a2.54 ± 0.206.58 ± 0.886.77 ± 0.09 b
Bare soil42.63 ± 7.178.85 ± 2.28 a3.85 ± 0.595.16 ± 1.246.70 ± 0.10 b
B. gracilis 29.86 ± 2.488.23 ± 4.09 abc3.24 ± 0.667.53 ± 1.686.53 ± 0.13 b
B. dactyloides 29.16 ± 3.471.91 ± 0.36 c3.83 ± 1.0810.35 ± 0.536.60 ± 0.10 b
A. frigida 21.66 ± 5.353.06 ± 0.55 bc2.92 ± 0.206.28 ± 1.086.79 ± 0.09 b
C. villosa 50.06 ± 8.324.54 ± 0.42 ab3.65 ± 0.524.73 ± 0.607.32 ± 0.14 a
K. cristata 23.15 ± 3.203.29 ± 0.55 bc3.10 ± 0.176.33 ± 1.756.67 ± 0.09 b
H. comata 33.50 ± 7.025.04 ± 0.28 ab3.06 ± 0.204.66 ± 0.956.75 ± 0.07 b
L. dalmatica 24.82 ± 3.893.95 ± 0.44 abc2.76 ± 0.105.53 ± 0.696.57 ± 0.10 b
 Root CRoot NLeaf CLeaf NLeaf P
= 0.001P < 0.001= 0.07P < 0.001= 0.29
  1. All values are presented as means ± SE (= 4). Letters next to the means ± SE indicate significant differences at  0.05 using Tukey post hoc tests following ANOVA. Parameters include: ∑C cycling enzymes = [(BG, β-glucosidase) + (CB, β-d-cellubiosidase) + (XYL, β-xylosidase) + (AG, α-glucosidase)]; ∑N cycling enzymes = [(LAP, leucine aminopeptidase) + (NAG, N-acetyl-β-glucosaminidase); ∑P cycling enzymes = [PHOS, phosphatase]; MBC, microbial biomass carbon; MBN, microbial biomass nitrogen. Units for all enzyme nutrient acquisition activities are expressed as nmol activity g−1 dry soil h−1). Units for MBC, MBN, extractable soil organic C, soil PO4, and total extractable inorganic soil N are expressed as μg−1 g−1 soil. Units for root and leaf C and N and leaf P are expressed as % DW.

Overall39.42 ± 0.390.83 ± 0.1239.23 ± 0.311.42 ± 0.130.12 ± 0.01
A. laxmannii 40.33 ± 0.41 ab2.23 ± 0.50 a38.85 ± 1.462.63 ± 0.15 a0.14 ± 0.02
B. gracilis 39.18 ± 0.49 bc0.53 ± 0.03 b38.78 ± 0.181.23 ± 0.33 bc0.10 ± 0.02
B. dactyloides 37.95 ± 0.70 bc0.53 ± 0.03 b38.55 ± 0.740.95 ± 0.05 c0.14 ± 0.02
A. frigida 41.98 ± 0.98 ab0.78 ± 0.05 b40.73 ± 0.261.20 ± 0.11 bc0.13 ± 0.03
C. villosa 41.15 ± 0.44 ab0.58 ± 0.05 b40.03 ± 0.691.93 ± 0.29 ab0.14 ± 0.01
K. cristata 37.03 ± 0.94 c0.63 ± 0.03 b38.13 ± 0.370.87 ± 0.27 c0.09 ± 0.00
H. comata 37.65 ± 0.47 bc0.55 ± 0.03 b38.13 ± 0.671.00 ± 0.06 c0.09 ± 0.01
L. dalmatica 39.55 ± 2.65 bc0.70 ± 0.00 b41.55 ± 0.051.40 ± 0.10 bc0.12 ± 0.02

Rhizosphere stoichiometry (i.e. soil, microbial biomass, or enzyme activity) predominantly exhibited nonhomeostatic relationships with corresponding plant tissue stoichiometry. Soil N : P across all species (H = 0.98, r2 = 0.5), along with enzyme N : P under the forb A. frigida (H = 0.43, r2 = 0.82) and C3 grass K. cristata (H = 0.5, r2 = 0.8) demonstrated strong nonhomeostatic relationships with leaf N : P (Table 4c). However, many H coefficients among plant species demonstrated a poor linear fit (i.e. low r2), so we are unable to evaluate homeostasis relationships among all rhizosphere components. Regardless, these relationships (Table 4), along with correlation results (Table 2; Fig. 4), suggest that plant tissue and soil stoichiometry strongly influence nutrient stoichiometry constraints in this grassland.

Table 4. Homeostasis relationships between leaf carbon : nitrogen : phosphorus (C : N : P) and corresponding consumer (i.e. enzyme, microbial and soil) stoichiometric counterparts among eight plant species using H coefficients (using y = cx1/H to define homeostasis) and associated r2
Resource (x- axis)(a) Leaf C : N(b) Leaf C : P(c) Leaf N : P
Consumer (y- axis)Enzyme C : NMBC : NSoil C : NEnzyme C : PSoil C : PEnzyme N : PSoil N : P
Plant species H r 2 H r 2 H r 2 H r 2 H r 2 H r 2 H r 2
  1. To better identify meaningful homeostasis vs nonhomeostatic relationships, only H values with corresponding r2 ≥ 0.40 are listed; significant linear relationships are indicated in bold. Negative H values are by definition homeostatic. The sample size for each nutrient variable among species is = 4, and for overall relationships is = 32. H coefficients between 0 and c. 1 demonstrate varying degrees of nonhomeostatic relationships between consumer and resource stoichiometry. MBC, microbial biomass carbon.

Overall0. 1.02 0.5
Astragalus laxmannii −1.860.540.350.020.370.040.020.36
Bouteloua gracilis 0.380.540.530.
Buchloe dactyloides −1.440.60.080.06−0.390.590.490.470.0050.640.65
Artemisia frigida 0.43 0.82 0.47
Chrysopsis villosa 0.760.770.003−0.740.720.35−0.260.561.320.510.01
Koeleria cristata 0.5 0.8 0.04
Hesperostipa comata 0.470.130.391.820.660.010.40.3
Linaria dalmatica 0.54 0.94 0.31.330.62


Ecological stoichiometry is typically used to explore relationships between terrestrial components such as soils, microbes, and plants at the ecosystem scale using composited bulk soil samples to overcome small-scale heterogeneity associated with individual plants (Elser et al., 2000; McGroddy et al., 2004; Sinsabaugh et al., 2008). Here, we assessed ecological stoichiometry at a finer spatial scale within individual plant rhizospheres to provide additional insight into the role of plant community composition in driving ecosystem function (Sterner & Elser, 2002c; Weidenhamer & Callaway, 2010; Morgan et al., 2011). Our central hypothesis was that the stoichiometry of plants, microbes, and soils is strongly correlated within individual rhizospheres, varying more among plant species than among individuals, because of coevolved plant–microbe interactions. Overall, we found strong positive correlations between soil and plant tissue stoichiometry, but only weak relationships between these components and microbial or enzyme stoichiometry. However, for the majority of plant species, plant tissue and rhizosphere stoichiometry does not seem to be strongly conserved at the plant species level, as we found an overlapping range of plant tissue, microbial, and soil nutrient stoichiometry among most of the plant species regardless of plant functional type. We expected that the stoichiometry of exoenzyme activities would negatively correlate with microbial biomass stoichiometry. On the contrary, soil microbial enzyme activities did not strongly correlate with microbial biomass stoichiometry. Overall, our findings indicate that there are tight linkages between soil and plant nutrient stoichiometry that do not appear to be strongly influenced by microbial stoichiometry or soil enzyme activities during peak biomass in this semiarid grassland ecosystem.

Plant tissue stoichiometry may be tightly constrained by below-ground stoichiometry in this grassland, as there were many nonhomeostatic relationships observed among plants and rhizosphere components (i.e. slopes > 1) during peak biomass. However, the high variability among plant rhizospheres in this study (i.e. exhibiting a weak linear fit; Table 4) probably reflects a strong individual physiological influence (rather than species-level influences) regulating nutrient uptake, incorporation, and release (Frost et al., 2005), which is nonetheless known to be facilitated by soil microbial elemental regulation (Cleveland & Liptzin, 2007; Yu et al., 2010, 2011). Furthermore, the majority of significant homeostatic relationships in this study were observed between leaf N : P and rhizosphere N : P (soil and enzyme) stoichiometry, which may reflect temporary declines in plant N and P concentrations during peak biomass (Reich et al., 1995; He et al., 2006).

Compared with the other plant species in this study, the microbial and soil rhizosphere characteristics of the C4 grass B. dactyloides demonstrated relatively higher microbial C and lower soil N, while inverse patterns occurred under the legume A. laxmannii at peak plant biomass. These differences in microbial biomass stoichiometry among plants may suggest a strong species-level influence on below-ground C and N dynamics, indicative of SOM decomposition and nutrient mineralization activities. Furthermore, B. dactyloides (C4 grass) had significantly wider soil and leaf C : N (c. 45) and significantly narrower soil and leaf N : P (c. 8) than the N-fixing legume A. laxmannii (i.e. C : N c. 15; N : P c. 20), consistent with high N concentrations common in legume species (Fox et al., 1990; Vitousek et al., 2002; Sadras, 2006; Lott et al., 2009). Regardless, the distribution of these two plant species may be tightly constrained by soil stoichiometry spatial distributions (i.e. Fig. 4c), as spatial distributions of soil nutrient properties can have a strong influence on plant fitness and distributions (Williams et al., 2007; Condit et al., 2013).

Plant stoichiometry has been suggested to be inherently linked to soil ecosystem functional properties. The temporal legacies of plant and microbial turnover may homogenize below-ground stoichiometry among vegetated and unvegetated soil patches in this grassland. Many studies suggest that any shift in plant species' abundances able to alter soil nutrient stoichiometry (Sterner & Elser, 2002a; de Graaff et al., 2010; Li et al., 2010) can influence rhizosphere microbial and soil enzyme activities (Sinsabaugh et al., 2009; Bever et al., 2010), which can further induce plant community species shifts (Schwinning & Parsons, 1996; Abbas et al., 2013) to ultimately alter ecosystem function (Kardol et al., 2010). Plant–microbe nutrient feedbacks within the rhizosphere can strongly influence the coexistence of competing plant species (Bever, 2002; Hawkes et al., 2005; Diez et al., 2010) and undoubtedly influence soil stoichiometry via nutrient mineralization and immobilization throughout the growing season (Sterner & Elser, 2002c; Hessen et al., 2004; Frost et al., 2005). However, plant tissue stoichiometry inputs from the previous year's growth may ultimately have a stronger influence on soil nutrients and soil microbial activities under plants and unvegetated soil patches alike at this site. For example, although plant root C exudate during the growing season can initiate a microbial reaction within minutes to rapidly alter rhizosphere nutrient availability (Cheng, 2009; Phillips et al., 2011; Drake et al., 2013), structural and microbial processed plant tissues can persist in soils for years (Kögel-Knabner, 2002; Schmidt et al., 2011). Likewise, microbial biomass may turn over seasonally (Bell et al., 2010; Treseder et al., 2010), while soil enzymes may turn over within days. In our study, the relationships between the stoichiometry of plant tissues and corresponding soil or microbial components measured at peak biomass in most cases closely resemble the unvegetated soil stoichiometry characteristics. To some degree, this finding may simply reflect the temporal legacy of plant and microbial turnover inherent in this mixed plant community in buffering the soil stoichiometry in this grassland.

For this study, we chose to sample at a single time point corresponding with peak biomass. Although recent studies have shown relatively consistent stoichiometry among plant species in grassland ecosystems during peak biomass (Niklas et al., 2007; Zhang et al., 2013), a better account of temporal variation throughout a growing season would probably provide additional insights by incorporating differences among individual plant species and stoichiometric soil interactions (Zhang et al., 2013). Observing multiple time points throughout the growing season would potentially allow us to observe how differences in plant phenology and fitness (i.e. occurrences) influence below-ground stoichiometry (Fujita et al., 2010; Isbell et al., 2011). Regardless of the limited differences observed in this study at a single time point, we assert that assessing rhizosphere stoichiometry among plant species' rhizosphere zones has offered a useful insight beyond what can be achieved with plant community-level sampling approaches.

Rhizosphere stoichiometry is a useful approach for assessing plant and microbial functional variability within a single ecosystem. Rhizosphere stoichiometry provides a high-resolution assessment of below-ground functional relationships to provide ecological insights linking species distributions and ecosystem functioning. Fanin et al. (2013) experimentally evaluated plant species-specific litter stoichiometry to assess microbial community structural and functional responses; demonstrating that wider litter N : P ratios corresponded with wider fungal : bacterial ratios (r2 = 0.68), wider microbial biomass N : P (H = 0.26; r2 = 0.51), and higher rates of microbial biomass P immobilization (relative to N uptake). As plant tissue stoichiometry changes, microbial communities may shift to favor taxa with stoichiometry more similar to the surrounding substrate (Moe et al., 2005; Sistla & Schimel, 2012). Thus, plant stoichiometry inputs may strongly influence microbial traits related to soil biogeochemical cycling processes in favor of plant species-specific stoichiometric requirements (Sterner & Elser, 2002b). In this study, the stoichiometry of different plant functional groups closely resembled one another (Figs 1, 2), suggesting that photosynthesis pathway alone is not a strong determinant of plant stoichiometry properties during peak biomass. Our findings likely reflect how individual- and species-level plant growth patterns unrelated to photosynthetic pathways, such as phenology (Nicotra et al., 2002; Rivas-Ubach et al., 2012) and plant ontogenies (Niklas et al., 2007; Agren, 2008; Bever et al., 2010), could affect spatial distributions of nutrients and plant–microbe interactions.

We also suggest that rhizosphere stoichiometry provides a useful framework to assess how potential shifts in plant community species' assemblages may influence future below-ground ecosystem processes. Many studies suggest that any shift in plant abundances that elicit changes to soil and microbial stoichiometry and physiology may eventually initiate a shift in ecosystem function by altering soil decomposition and mineralization rates (Cardinale et al., 2006; Kardol et al., 2010; Isbell et al., 2011). For example, shifts in plant species' abundances has been shown to alter soil nutrient stoichiometry (Sterner & Elser, 2002a; de Graaff et al., 2010; Li et al., 2010) and rhizosphere microbial and soil enzyme activities (Sinsabaugh et al., 2009; Bever et al., 2010). Our findings suggest that, although plant tissue stoichiometry is constrained by soil stoichiometry in this grassland, a plant community shift favoring certain species may impact ecosystem function. For example, an increase in C. villosa could lead to increased enzyme N : P ratios (i.e. Fig. 3f) and, as a consequence, decrease P availability (relative to N). Thus, an increase in C. villosa abundance could promote competitive exclusion, owing to the wide enzyme N : P activities within its rhizosphere (Schwinning & Parsons, 1996; Abbas et al., 2013), which, if persistent, could ultimately cause a shift in ecosystem function (Kardol et al., 2010). Plant–microbe feedbacks within the rhizosphere can strongly influence the coexistence of competing plant species (Bever, 2002; Hawkes et al., 2005; Diez et al., 2010).

In conclusion, assessing rhizosphere stoichiometry at the plant species level provided a valuable insight into the link between ecosystem function and plant community composition, microbial community traits, soil nutrient availability and potential nutrient feedbacks. Rhizosphere stoichiometry is a unique approach to evaluating plant–microbial interactions within a single ecosystem, which integrates information related to soil ecology and plant species' abundances. This approach could be more broadly applied to examine how plants influence rhizosphere microbes to alter their nutrient acquisition potential depending on the stoichiometry of available resources, soil ecological impacts of invasive plant species, and whether plant–microbe interactions are a fundamental determinant of plant fitness in competitive environments.


This research was supported by the US National Science Foundation (DEB# 1021559), and the US Department of Energy's Office of Science (Biological and Environmental Research). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the US NSF. A special thanks to Akihiro Koyama for his highly valued intellectual and technical contributions towards this effort. Last but not least, we thank the student assistants who contributed to this experiment in the laboratory and field, including Jennifer Bell, Lana MacDonald, Swastika Raut, and Shanker Tamang.