Introduction

Considerable research has documented a variety of global environmental changes resulting from human activities. Of particular interest are the steady increase in atmospheric carbon dioxide (CO2) concentration over the past two centuries (Vitousek 1994), increases in nitrogen (N) inputs to ecosystems from fertilization and atmospheric deposition (Sala et al. 2000; Norby 1998), and the rapid loss of biodiversity (Sala et al. 2000). The simultaneous occurrence of these global changes raises questions about how their interactions will influence ecosystems and individual species in the coming decades.

Increasing atmospheric CO2 concentrations have been shown to affect many ecosystem processes. In particular, plant growth rate and net biomass accumulation increase under elevated CO2 concentrations in many ecosystems (Niklaus et al. 1998; King et al. 2004; Torbert et al. 2004), although this response may only be observed after several years of CO2 enrichment (Winkler and Herbst 2004). The extent and nature of ecosystem response to elevated CO2 may be influenced by N availability, with ecosystems with low N showing a more modest response to elevated CO2 than those with high N (Cannell and Thornley 1998). Since productivity in many terrestrial and aquatic ecosystems is limited by N (Elser et al. 1990; Vitousek and Howarth 1991), changes in CO2 and N must be considered together to understand their effects on plant communities and ecosystem processes. Local biodiversity and plant functional type (e.g., N-fixing legumes vs. non-leguminous forbs) may also be involved in mediating how CO2 and N perturbations affect ecosystems. For example, studies have shown that enhancement of atmospheric CO2 and N deposition separately leads to increased biomass production but that this response is lower in species-poor areas (Reich et al. 2001b). In addition, legumes tend to be more productive than non-N fixing forbs under elevated CO2 conditions because of the stimulation of N2 fixation (Lüscher et al. 1998; Reich et al. 2001b). These studies highlight the interactive roles that biodiversity, plant functional type, and species identity play in ecosystem-level responses to changes in CO2 and N deposition.

Because global environmental changes clearly influence multiple plant physiological processes involved with the acquisition of energy and nutrients, we would expect to see a corresponding change in the ratios of major elements [i.e., C, N, and phosphorus (P)] in plant tissue (Marschner 1995; Aerts and Chapin 2000; Sterner and Elser 2002). These stoichiometric ratios are important from a functional perspective as indicators of substrate quality for herbivores (Elser et al. 2000) and decomposers (King et al. 2004), which in turn influences the fate of the production of organic matter (Cebrian 1999). For these reasons, it is important to understand how changes in CO2, N, and biodiversity jointly affect this aspect of plant response. Elevated CO2 treatments often lead to decreases in the percentage of N (%N) or increases in the C:N ratio in plant leaf tissue in a variety of species (Gifford et al. 2000; Reich et al. 2001a; Billings et al. 2003; King et al. 2004). Plant functional type also tends to modify the way C:N ratios respond to elevated CO2, as legumes may show less increase in C:N ratios (due to simultaneous increases in atmospheric N2 fixation) than non-legumes (Torbert et al. 2004). Elevated N deposition has also been shown to reduce plant C:N ratios in a variety of ecosystems (Norby 1998; Rueth and Baron 2002).

While C:N ratios are commonly used indicators of physiological responses to elevated CO2, other potential stoichiometric changes are relatively unstudied. For example, few studies have examined how increased CO2 affects plant biomass N:P or C:P ratios. Of these, one by Gifford et al. (2000) showed that elevated CO2 levels induce an increase in C:P ratios, while that of Niklaus et al. (1998) found no change in the N:P ratios. Furthermore, the interactive effects of CO2 and N on plant C:P and N:P ratios and the role of biodiversity and plant functional type in modifying their effects remain to be studied. A focus on C:P and N:P ratios is important because of the previously unappreciated role of P intake as an aspect of food quality for herbivores (Elser et al. 2000; Perkins et al. 2004) and because P can be a critical limiting nutrient for microbial processes, even if plant production is limited by other factors (Sundareshwar et al. 2003).

In the study reported here, we examined changes in foliar stoichiometric ratios under manipulations of CO2 levels, N addition, and plant species diversity (richness). We sampled five commonly found dicot plant species in a grassland ecosystem in Minnesota where plants are grown experimentally in a full factorial combination of CO2, N addition, and diversity levels. Because some studies show functional type differences in C:N response to elevated CO2 and nitrogen, we selected species that were representative of both N2-fixing legumes and non-N2 fixing forbs. Our results indicate a variety of direct and interactive effects of CO2, N, and biodiversity on plant C:N:P stoichiometry and have significant potential implications for terrestrial ecosystems and food webs.

Material and methods

Experimental design

This study was conducted as part of the BioCON (Biodiversity, CO2, and N) experimental setup performed at the Cedar Creek Natural History Area in Minnesota, USA (Reich et al. 2001a, b). The design consists of six 20-m diameter rings distributed across a grassland field, with each ring containing sets of 2×2-m plots representing fully crossed treatments of species diversity (1, 4, 9, 16 species), and N addition (no addition vs. +4 g N m−2 year−1). Three rings use a free air CO2 enrichment system (FACE) that supplies air with an elevated CO2 concentration (approx. 560 ppm) to the plots, while three control plots receive ambient levels of CO2 (approx. 370 ppm). One-half of the plots within the rings were selected randomly to receive N fertilizer three times each year for a total N addition of 4 g N m−2 year−1 as NH4NO3. Each treatment combination was replicated at least two times within a ring. Species selected for this experiment were native to neighboring prairie and grassland areas or have been naturalized to the area (Reich et al. 2001a, b). Additional details and background information about BioCON and the experimental setup can be found in Reich et al. (2001a, b).

While plots within the main BioCON experiment contained up to 16 species, only monocultures and four-species plots were used in this study because of time constraints. During the second week of July 2002, we collected leaf samples from five plant species: two non-legume forbs (Solidago rigida and Achillea millefolium) and three legumes (Amorpha canescens, Lespedeza capitata, and Lupinus perennis). We sampled from monocultures and from four-species plots that represented one functional type, either legume or non-legume forb. This allowed us to examine the effects of species diversity while holding functional group diversity constant (as in Reich et al. 2004). Shoots from five plants were sampled from each of at least two replicate plots from each of the CO2 and N treatment combinations. Samples were dried at 60°C for 72 h, then milled and analyzed for C, N, and P content. Carbon and N analyses were completed on a PE-2400 CHN analyzer (Perkin-Elmer, Foster City, Calif.) at Arizona State University. Leaf P content was measured by persulfate oxidation followed by colorimetric analysis (Schade et al. 2003). All stoichiometric ratios (C:N, C:P, N:P) are reported as molar ratios. Ratios were averaged per plot prior to statistical analyses.

Statistical analysis

An analysis of variance (ANOVA) was performed on the data since the assumptions of normality and equal variance were upheld based on examination of plots of the expected value for normal distribution versus residuals. The main effects of diversity, CO2, and N manipulations along with functional type and species (nested in functional type) and their interactions were tested for each stoichiometric ratio. Because this was a split-plot design, the “ring” was tested as a random effect nested within the CO2 treatment. Significant differences among treatment means were then analyzed using the post-hoc Tukey’s multiple comparison test. Least squares means plots are shown only for significant effects and interactions. Note that the scale of the BioCON experiment implies that the level of replication of the treatment combinations (n=2) and the number of treatment levels [CO2 and N enrichment have only two levels (enriched or control) and in our analysis, only two diversity levels are considered] are limited. Thus, our analysis may be unable to detect all actual effects. Nevertheless, various manipulations resulted in statistically significant differences among the treatments, especially for the C:N ratio, and we confine our discussion and interpretation to these patterns.

Results

Foliar C:N:P ratios of the five plant species considered in our study responded significantly to manipulations of CO2 and N supply. However, these responses depended on local species diversity and the identity or functional type of the species involved. In the following sections we summarize the main effects followed by the interactions for each of the dependent variables examined in the study.

C:N ratios

The C:N ratio (hereafter, C:N) responded significantly to all of the independent factors tested in this experiment: CO2 and N enrichment, species identity, diversity, and plant functional type. The five species differed significantly in C:N on average across all treatment variables (P<0.001; Fig. 1a; Table 1). The range in C:N was substantial, with Lupinus and Amorpha averaging as low as approximately 21 and Solidago averaging approximately 33. Despite these chemical differences at the species level, on average across all five species, CO2 enrichment produced a significant increase in the C:N (P<0.001; Fig. 1b), with C:N increasing from approximately 24 to approximately 28 under enriched CO2. Nitrogen addition caused a statistically significant decline in C:N (P=0.036; Fig. 1c; Table 1); however, the absolute magnitude of this effect was modest (C:N decreased from 26.6 to 25.6 under conditions of enriched N). Increased local plant diversity also resulted in lower C:N (P<0.001; Fig. 1d; Table 1). Furthermore, functional type had a significant effect, as non-leguminous forbs had a significantly higher C:N than legumes (approx. 29.4 for non-legumes compared to approx. 22.7 for legumes) (P<0.001; Table 1; see Fig. 1a).

Fig. 1
figure 1

Least squares means plots of the main effects of species identity (a), of CO2 (b), of N (c), and of diversity (d) on plant C:N ratios. Effects of plant functional type can be seen by inspection of a, as Amorpha, Lespedeza, and Lupinus are N-fixing legumes (left-hand side of a), while Achillea and Solidago are non-fixing forbs (right hand side of a). Letters above each column differ when the column means are significantly different from each other according to a Tukey’s multiple comparisons test. Error bars indicate ±1 SE

Full size image
Table 1 Results of analysis of variance (ANOVA) for plant C:N, C:P, and N:P ratios indicating main effects and interactions of carbon (CO2 of approx. 360 ppm vs. approx. 550 ppm), diversity (mono-specific vs. four species), N (no addition vs. +4 g N m− 2 year− 1), plant functional type (legume vs. forb), and species (Solidago rigida, Achillea millefolium, Amorpha canescens, Lespedeza capitata, and Lupinus perennis)
Full size table

For each of the manipulated factors there were multiple significant interactions that also contributed to C:N response. The effect of CO2 enhancement interacted with plant species identity (P=0.001; Table 1), as only three species (Lespedeza, Achillea, and Solidago) had significant increases in C:N under elevated CO2 (Fig. 2a). Thus, plant C:N response to CO2 was somewhat idiosyncratic among the taxa. Nitrogen addition had a significant negative effect on C:N under ambient CO2 levels, but no effect on C:N under elevated CO2 (significant CO2-N interaction, P=0.015; Fig. 2b; Table 1). However, there was no significant N-species interaction (P>0.05; Table 1).

Fig. 2
figure 2

a–d Least squares means plots of the two-way interactions of CO2 (Amb C ambient CO2, Enr C enriched CO2) and species (a), of CO2 and N (Amb N ambient N, Enr N enriched N) (b), of CO2 and diversity (1 sp monospecific plot, 4 spp four-species plot) (c), and of diversity and functional type (Legume N-fixer, Forbs non-N-fixer) (d) on C:N ratios. The two-way interaction between CO2 and functional type on the C:N ratio can be seen following an inspection of a as Amorpha, Lespedeza, and Lupinus are N-fixing legumes (left-hand side of a), while Achillea and Solidago are non-fixing forbs (right-hand side of a). (e) Least squares means plot of the three-way interaction of CO2, diversity, and type on C:N ratios. Letters above each column differ when the column means are significantly different from each other according to a Tukey’s multiple comparisons test. Error bars indicate ±1 SE

Full size image

Diversity also affected how C:N responded to CO2 enrichment. Under low diversity, CO2 enrichment led to a larger increase in plant C:N than it did under high diversity (P=0.006; Table 1; Fig. 2c). The interactive effect of diversity and N on the C:N ratio was not significant (P>0.05; Table 1). However, the effect of diversity did depend on plant functional type, as non-legumes had a larger decrease in C:N with increased diversity than did legumes (Fig. 2d), leading to a significant diversity-type interaction (P=0.032; Table 1).

C:N also responded differently to CO2 addition in a plant functional-type manner. While plants of both functional types had higher C:N with CO2 addition (see Fig. 2a), the response of the non-leguminous forbs was greater (significant CO2-type interaction; P<0.001; Table 1). However, there was no statistical interaction between N enrichment and functional type on the plant C:N ratio (P>0.05; Table 1).

Finally, plant functional type appeared to modulate the response of plant C:N to CO2 and diversity, as indicated by a significant three-way CO2-diversity-type interaction (P=0.004; Table 1; Fig. 2e). Both legumes and non-legumes responded to CO2 enrichment with an increase in C:N under both low and high diversity. While the magnitude of the increase was the same under both diversity levels for the legumes, in the non-legumes C:N increased more in response to CO2 in the low diversity treatment than in the high diversity treatment (Table 1; Fig. 2e). There were no other significant three-way interactions (P>0.05; Table 1).

C:P ratios

All five species in our study had significantly different C:P ratios (hereafter, C:P) on average across all treatments (P<0.001; Table 1), covering a twofold range (approx. 357 for non-leguminous Achillea to approx. 860 for the leguminous Lupinus; Fig. 3a). Carbon dioxide enrichment and N manipulation had no significant effects on C:P (Table 1). However, C:P responded to local plant diversity as well as plant functional type. Overall, plant C:P, like C:N, decreased significantly from approximately 613 in the monocultures to approximately 563 in the four-species plots (P=0.003; Table 1; Fig. 3b). Furthermore, legumes had significantly higher C:P than the non-legumes (approx. 763 vs. approx. 412) when species were pooled across all treatments (P<0.001; Table 1; see Fig. 3a).

Fig. 3
figure 3

a, b Least squares means plots of the main effects of species identity (a) and of diversity (b) on C:P ratios. As in Fig. 1, effects of plant functional type can be seen by inspection of a, as Amorpha, Lespedeza, and Lupinus are N-fixing legumes (left-hand side of a) while Achillea and Solidago are non-fixing forbs (right-hand side of a). (c, d) Least squares means plots of the two-way interactions of species and CO2 (Amb C ambient CO2, Enr C enriched CO2) (c) and of species and diversity (1 sp monospecific plot, 4 spp four-species plot) (d) on C:P ratios. Letters above each column differ when the column means are significantly different from each other according to a Tukey’s multiple comparisons test. Error bars indicate ±1 SE

Full size image

The C:P was affected by several interactions between factors. While there was no significant CO2-N interaction for C:P, there was a significant species-CO2 interaction (P=0.002; Table 1), which was driven largely by the strong increase in Lespedeza C:P in response to CO2 enrichment (Fig. 3c). The species-diversity interaction was also significant (P<0.001; Fig. 3d; Table 1) as three species (Lespedeza, Lupinus, and Solidago) had C:P ratios that decreased under high diversity (only that of Lupinus was significant), while the C:P ratios of Amorpha and Achillea increased slightly (but not significantly) under high diversity (Table 1; Fig. 3b). No other interactions significantly influenced C:P (Table 1).

N:P ratios

Species differed significantly in N:P ratios (N:P) on average across all treatments (P<0.001; Table 1), exhibiting a threefold range in N:P across the species from approximately 14 (Achillea and Solidago) to approximately 42 (Lupinus). However, not all species were significantly different from each other (Fig. 4a). The N:P of Lupinus was significantly greater than those of the remaining four species. Amorpha and Lespedeza had higher N:P than the non-legumes, but the ratios of Amorpha and Lespedeza were not significantly different from each other. Likewise, Achillea and Solidago N:P were significantly lower than those of the legumes but not significantly different from each other (Fig. 4a). None of the primary independent variables (CO2 or N manipulations, diversity) had a significant main effect on the N:P of these five species (Table 1). However, legumes did have significantly higher N:P (approx. 34) than non-leguminous forbs (approx. 14), resulting in a significant functional type effect (P<0.001; Table 1; see Fig. 4a).

Fig. 4
figure 4

Least squares means plot of the main effect of species (a) on N:P ratios. As in Fig. 1, effects of plant functional type can be seen by noting that Amorpha, Lespedeza, and Lupinus are N-fixing legumes (left-hand side) while Achillea and Solidago are non-fixing forbs (right-hand side). Least squares means plot of the two-way interaction of species and diversity (1 sp monospecific plot, 4 spp four-species plot) (b) on N:P ratios. Letters above each column differ when the column means are significantly different from each other according to a Tukey’s multiple comparisons test. Error bars indicate ±1 SE

Full size image

CO2 and N manipulations had no interactive effects on N:P (Table 1), but the species-diversity interaction was significant (P=0.002; Table 1; Fig. 4b), indicating that the response of N:P to diversity manipulation depended on the identity of the species. Lespedeza, Lupinus, and Solidago all had lower N:P with increasing diversity, while Amorpha and Achillea had higher N:P when diversity was high. However, none of the diversity effects within a species were statistically significant (Fig. 4b). No other interaction terms were statistically significant for N:P (Table 1).

Discussion

Our data show that foliar stoichiometric ratios responded to changes in CO2, N, and diversity despite overall differences in tissue chemistry among species. Enriched CO2 treatments alone affected elemental ratios, but these responses to elevated CO2 were also mediated by N and diversity treatments. Plant functional type also influenced how stoichiometric ratios responded to elevated CO2. It is important to note that our data pertain only to foliar nutrient contents on a single date during the growing season and thus may not ramify to ecosystem-level nutrient cycling effects such as nutrient use efficiency (Vitousek 1982) if observed ratios are not correlated with overall plant nutrient use during the growing season or with the nutrient content of senesced tissue (Killingbeck 1996). In general, foliar nutrient content correlates closely with nutrient content in other plant tissues (Kerkhoff et al. 2006), and thus foliar data are relevant to whole-plant nutrient use. Furthermore, foliar nutrient content is positively correlated with that of senesced leaves (Kobe et al. 2005). Consequently, our data set, while somewhat limited, should nevertheless have relevance in gaining an understanding of how plant physiological response to CO2, nitrogen, and diversity may affect nutrient cycling processes.

The effects of elevated CO2 on plant stoichiometric (C:N:P) ratios depended on both species identity and N fertilization levels. Of the five species investigated, the average increase in C:N in response to elevated CO2 was similar to increases found in other studies (Niklaus et al. 1998; Gifford et al. 2000; Reich et al. 2001a), but the magnitude of this response in our study differed among species. Three of the five species, Lespedeza, Achillea, and Solidago, responded with substantial increases in C:N (Fig. 2a). Similarly, in terms of changes in C:P, species responded idiosyncratically to elevated CO2, with the only significant increase in C:P observed in Lespedeza (Fig. 3c). Stoichiometric response to elevated CO2, however, also depended on N enrichment. While there was an overall increase in C:N in response to CO2 enrichment, this CO2 response was larger under enriched N (Fig. 2b). Furthermore, N addition resulted in lower C:N, but only under ambient CO2 (Fig. 2b). This contrasts with the findings of Gifford et al. (2000) who showed a stronger C:N response to elevated CO2 under low levels of N deposition; however, that study focused on only a single species. Importantly, our data demonstrate over multiple plant species that atmospheric changes in CO2 and N deposition can affect C:N:P stoichiometry directly but also that these changes have inter-dependent effects.

While stoichiometric responses to CO2 and nutrient manipulations have been examined previously, a more unique, and potentially important, outcome of our study is the documentation that biodiversity modifies plant stoichiometric response to CO2 (Fig. 2c). In general, elevated levels of diversity in terms of species richness lowered both C:N and C:P (Figs. 1d, 3b), but diversity also significantly affected how plants responded to elevated CO2. In the monocultures, increasing CO2 led to an amplified increase in C:N relative to the four-species plots (Fig. 2c). A possible explanation for this difference is that, in monocultures, individuals likely experience a more severe competition for nutrients due to an increased overlap of resource use strategies with conspecific neighbors. However, in high diversity plots, neighbors will often be of a different species, which could reduce potential resource use overlap and increase nutrient availability for individual plants. This hypothesis could link these stoichiometric responses to diversity to the findings of Reich et al. (2001a), who argued that overall biomass increases due to elevated CO2 or N were higher in species-rich than species-poor treatments because of interspecific niche partitioning. Furthermore, such a variation in the severity of intra- and interspecific competition may underlie the species-specific variation in C:P and N:P response to diversity that we observed.

Plant functional types also differed in stoichiometric ratios and in their response to elevated CO2. Legumes generally had lower C:N (Fig. 1a), higher C:P (Fig. 3a), and higher N:P (Fig. 4a) than non-legumes, thereby reflecting that legumes are more likely to be P-limited than non-legumes. In response to elevated CO2, however, C:N increased more in non-legumes than in legumes (Fig. 2a), presumably because N2 fixation by legumes allows them to match enhanced C assimilation due to increased CO2 concentration with additional N drawn from N fixation (see also Reich et al. 2001a,b; Lee et al. 2003; Torbert et al. 2004).

Our data show that foliar stoichiometric ratios are impacted by changes in CO2, N, and plant diversity. Such changes are relevant ecologically because plant elemental composition potentially affects important ecosystem processes and ecological interactions, such as ecosystem nutrient use efficiency (Vitousek 1982), litter decomposition, and herbivore performance. Previous studies have shown that the nutrient content of decomposing biomass can influence decomposition rates (Enriquez et al. 1993). Thus, changes in plant stoichiometric ratios due to CO2, N, or diversity changes could also imply changes in litter decomposition rates, if litter and plant elemental composition are functionally correlated (Hobbie 1992). In our study, elevated CO2 led to higher plant C:N ratios, possibly leading to slower decomposition rates, as observed in some studies (De Angelis et al. 2000; Liao et al. 2002; Schadler et al. 2003) but not in others (Norby et al. 2001; Dukes and Hungate 2002). For species that do experience such changes in elemental composition, this stoichiometric shift could buffer global CO2 increases, as plant nitrogen-use efficiency (NUE) increases, while slower decomposition rates lead to an increase in the pool size of soil C. Importantly, diversity modified these responses. High diversity significantly dampened the increase of biomass C:N in response to elevated CO2, indicating that high diversity will tend to offset potential stoichiometric effects on C storage in plant standing stocks and soils. Our data also have implications for the possible impacts of altered CO2, nutrients, and diversity on terrestrial food webs, since an important aspect of the quality of plant biomass for herbivores is its nutrient (N, P) content (White 1993; Elser et al. 2000). Previous studies have shown that growing plants under elevated CO2 levels can impair herbivore performance (Stiling et al. 2003; Hamilton et al. 2004). Our data add to these findings in two ways. First, the moderating effect of plant diversity on plant C:N increase in response to atmospheric CO2 suggests that maintenance of plant diversity may be important in maintaining the production of higher trophic levels under conditions of increasing CO2. Secondly, P availability can limit herbivore growth in nature (Elser et al. 2000; Schade et al. 2003; Perkins et al. 2004), an effect that is more likely when plant N:P exceeds the N:P of herbivore biomass (Sterner and Elser 2002). In our study, the combination of increased CO2 and N inputs along with low diversity tended to increase plant biomass C:P and N:P. In such a situation, we predict an increased prevalence of herbivore P-limitation, especially for P-rich herbivore taxa (Sterner and Elser 2002), with further potential impacts on detrital processing of organic matter as unconsumed production enters detrital pathways and affects long-term storage pools (Cebrian 1999).

Conclusion

Given the multivariate global environmental changes that are occurring, we should expect stoichiometric ratios of plant biomass to respond as well. The results of our study suggest that there are significant interactions between CO2, N, and diversity that modify how plant elemental ratios respond to important environmental drivers. Importantly, differences in local biodiversity may affect ecosystem adjustments to biogeochemical perturbations, such as increased CO2 or N deposition. Given that plant elemental composition responds to various environmental changes, a stoichiometric perspective may help elucidate some of the mechanisms by which global environmental changes will affect plant physiology and plant-mediated above- and belowground processes. While the immediate effects of global change may be easily seen on the plant level, higher trophic levels may also be affected via indirect mechanisms such as those we document here.