Introduction

A critical question for understanding temperate savannas is whether phosphorus (P) constrains the ability of the ecosystem to recover the nitrogen (N) lost by recurring fires. Plants that associate with symbiotic N2-fixing bacteria (hereafter referred to as N2-fixers) are central to resolving this potential P–N interaction as they can bring in atmospheric N to balance fire losses; however, P may limit the ability of N2-fixing plants to fix (Dixon and Wheeler 1983; Vitousek and Howarth 1991). The question is particularly relevant for the longleaf pine ecosystem, which persists on weathered and P-poor soils and once extended across most of the Coastal Plain of the southeastern US. Moreover, the longleaf pine ecosystem possesses a unique and highly diversified group of herbaceous N2-fixing legumes (Hainds et al. 1999; Norden and Kirkman 2006) that have evolved traits to persist with a frequent fire cycle (i.e., every 1–5 years; Van Lear et al. 2005), such as re-sprouting and fire-induced germination (Wiggers et al. 2017).

While rare, fertilization experiments have confirmed that P, more so than N or K, can limit the growth of longleaf pine trees in these P-poor soils (Lewis 1977; Dickens 2001). Whether this limitation extends to leguminous N2 fixers in these ecosystems is less known, however, and there are no experiments that have evaluated whether nutrients may influence the ability of N2-fixers to bring new N to the ecosystem. In addition to soil P, recent findings point to the potential influence of molybdenum (Mo) on fixation (Wurzburger and Hedin 2016), which can be in low supply in highly weathered soils (Karimian and Cox 1979). Other findings have shown that taxonomic differences can be as important as the effect of P (or Mo) in determining the rate and strategy (i.e., when fixation occurs) by which legume species fix N2 (Wurzburger and Hedin 2016). Some N2-fixers appear to have evolved an obligate strategy of fixing N2 independent of variations in local soil N condition, while other N2-fixers facultatively downregulate fixation in soils with abundant bioavailable soil N (Sheffer et al. 2015).

A consideration of fire may help resolve the potentially complex dynamics among N, P and fixation, since longleaf pine ecosystems are dependent on a prescribed fire interval of 2–3 years. Fire depletes N in the ecosystem while temporarily increasing P availability through thermal mineralization (Boring et al. 2004). However, fire can also increase P losses from ash or in hydrological leaching (Raison et al. 1984; Lavoie et al. 2010). Sandy soils may be particularly vulnerable to the hydrological export of P (and possibly of Mo), with their low contents of P-sorbing clay minerals and rapid rates of soil water infiltration. Such losses may have detrimental effects on N2-fixers, since P is needed to fuel the extra growth that N2-fixers accomplish in N-poor conditions, and is also needed for the ATP-rich energy transformations that support the N2 fixation process (Batterman et al. 2013). In addition, leguminous N2-fixers have a fundamental requirement for Mo as a cofactor in the nitrogenase enzyme that is responsible for breaking the atmospheric N2 bond (Fenner and Lee 1989; Alberty 2005).

An entirely different dimension of this problem is introduced by inter-species differences in rates and strategies of fixation. First, N2 fixation can vary greatly among species (Hiers et al. 2003; Cathey et al. 2010; Wurzburger and Hedin 2016), possibly due to local environmental conditions that influence N2-fixer growth rate, but also due to inherent differences in growth rate potential (Pate 1996; Sprent 2009; Wurzburger and Hedin 2016). That the N2-fixing symbiosis has arisen multiple times within the Leguminosae (Doyle 2011) adds the further possibility that taxonomic differences in fixation capacity may be constrained by evolutionary history.

Whether individual species employ an obligate or facultative fixation strategy may also influence the magnitude and timing of N inputs to longleaf pine savannas. N2-fixers that possess a facultative capacity (i.e., can adjust fixation to meet plant N demand) may track fire-induced N dynamics, such that fixation is upregulated after a fire event to facilitate the regrowth of aboveground biomass, and downregulated after sufficient accumulation of N. Obligate N2-fixers, however, may be largely insensitive to the timing of fire, causing fixation to persist without any feedback to the internal ecosystem nitrogen cycle. In nature, N2-fixers likely employ a range of fixation strategies, across the spectrum of facultative to obligate (Menge et al. 2015). Theoretical analysis suggests, however, that the obligate strategy is most favored in temperate climates where plants experience extended periods of N limitation following a disturbance event (Sheffer et al. 2015).

Here, we experimentally evaluated how soil nutrients and species identity influence fixation by seven herbaceous legumes, which commonly occur in frequently burned longleaf pine savannas. Although we did not directly manipulate fire in our study, we used soils from longleaf pine savannas with a prescribed fire interval of 2–4 years, which capture the long-term effect of fire on N, P and Mo soil stocks and availability. Based on the evidence for an influence of P on longleaf pine growth and that P can limit legume N2 fixation in other ecosystems, we hypothesized that P would also limit fixation and growth of herbaceous legumes. Second, because Mo is low in Coastal Plain soils (Karimian and Cox 1979), we also examined the alternative hypothesis that Mo alone or in combination with P limits fixation. Our third hypothesis was that fixation is primarily determined by the ability of N2-fixers to either maintain or downregulate fixation, which may depend on soil N and species identity. To test these hypotheses, we grew seven species from two field sites in different regions of the longleaf pine ecosystem: a Fall-Line sandhill location in Georgia, and a Coastal Plain sandhill location in Florida. From the soils at each location, we determined the individual and interactive effects of P and Mo addition on N2 fixation. To resolve species-specific fixation strategies, we used a 15N pool dilution approach to determine the responsiveness of fixation across a range of N input rates.

Methods

Experimental design

We conducted two concurrent greenhouse experiments at the University of Georgia from March to September 2015. In experiment 1, we determined how P and Mo regulate legume fixation and in experiment 2 we assessed legume responses to N addition and the importance of species identity. We collected field soil at the end of summer from longleaf pine savannas at Fort Benning Army Base, Georgia, and Eglin Air Force Base, Florida (hereafter Benning and Eglin). Benning is located at the transition of the Piedmont and Coastal Plain geomorphic provinces (i.e., the Fall-Line) (Dilustro et al. 2002) and has a mean annual average precipitation and temperature of 1260 mm and 18.0 °C (NCDC 2017). In contrast, Eglin is located on the Gulf Coastal Plain with a mean annual average precipitation and temperature of 1802 mm and 18.7 °C (NCDC 2017). At both sites, we collected Lakeland sands (Thermic, coated Typic Quartzipsamments). We collected soil to a depth of 20 cm from four mature (i.e., > 80 years old) longleaf stands at each site. We sieved soils to remove coarse organic debris, homogenized them separately and filled 1 L experimental pots (7 cm pot diameter). From the site-level soil mixtures, we determined a suite of soil chemical and physical properties including total C and N (by combustion) (Robertson et al. 1999), total digestible P (Nelson 1987), wet-digestible Mo (Pequerul et al. 1993), Mehlich III-extractable P (Mehlich 1984), resin-extractable P (Robertson et al. 1999) and soil texture (percent sand, silt and clay) (Carter 1993; Bindu and Ramabhadran 2010). We germinated legumes from a southeastern seed source (Roundstone Native Seed Company, Upton, Kentucky) in sand and transplanted seedlings into the pots after 2 weeks of growth. We allowed seedlings to acclimate to pots for 3 weeks, assigned each plant a nutrient treatment and ensured that each treatment contained approximately the same distribution of plant sizes.

Study legumes

We selected seven herbaceous legume species that were among the ten most abundant species at our field sites at Benning and Eglin: Centrosema virginianum (L.) Benth, Chamaecrista fasciculata (Michx.), Desmodium floridanum Chapm., Lespedeza hirta (L.) Hornem., Lespedeza virginica (L.) Britton, Mimosa quadrivalvis L., and Tephrosia virginiana (L.) Pers. These species differ in size, growth form, taxonomy, and nodule morphology and reflect the diversity of the legume traits observed in savanna legume communities (Table 1). Due to low germination rates, L. virginiana and D. floridanum were only used in experiment 2. We also used the native non-N2-fixing forb Asclepias tuberosa (L.) as a reference plant to determine background levels of soil 15N, which we used to calculate N2 fixation. This species has a growth form and rooting depth that are similar to those of the legumes in our experiments (Hiers et al. 2003).

Table 1 Herbaceous legume species native to longleaf pine savannas selected for experiments, including subfamily in Leguminosae, life history, growth form and nodule morphology
Full size table

Nutrient treatments

We added P, Mo and N to plants on a biweekly basis for 16 weeks in both experiments. For experiment 1, we supplied P at two levels (−P and +P, with a total application of 0 and 15.38 mg P per plant, which corresponds to 0 and 5 g P m−2 year−1 as NaPO4) and Mo at two levels (−Mo and +Mo, or 0 and 153 µg Mo, corresponding to 0 and 50 mg Mo m−2 year−1 as Na2MoO4) in a complete factorial design for a total of four treatments and ten replicates per treatment by species combination. We selected the P application rate based on prior studies of longleaf pine growth on Coastal Plain soils (Lewis 1977; Dickens 2001) and estimates of legume P demand (Hiers and Mitchell 2007). We selected the Mo application rate from studies on Mo limitation of N-fixing plant growth (Becking 1961; Brkic et al. 2004; Finzi and Rodgers 2009). For Experiment 2, we supplied N at three levels (−N, +N and +++N, or 0, 6.15, and 30.7 mg N per plant, and corresponding to 0, 2 and 10 g N m−2 year−1 as 5 atom % labeled 15N (NH4)2 SO4) with eight replicates per treatment by species combination. Nitrogen addition rates were selected from field estimates of net N mineralization rates in longleaf pine savannas (Wilson et al. 1999), where the +N treatment was intended to isotopically enrich the available soil N pool, but not stimulate plant growth or downregulate fixation, while the +++N treatment was intended to at least double N inputs from N mineralization and deposition. We made stock nutrient solutions at the beginning of the experiment, maintained them at 4 °C and applied treatments in 50 ml of DI water. In between nutrient treatments, plants were regularly watered with tap water, which had an undetectable amount of N and P.

Biomass growth, tissue nutrients and N2 fixation measurements

We destructively harvested all plants to quantify leaf, stem, root, nodule and reproductive biomass 105–115 days after initiating nutrient treatments, at which point all plants received eight treatment doses. We harvested at least one individual per species by treatment combination on each day of the harvest to minimize the confounding effects of weather variation on fixation (Lee and Son 2005). We also restricted our harvest interval to the hours between 9 am and 3 pm to minimize diurnal patterns in fixation rates (Lee and Son 2005). For each plant, we carefully separated roots and nodules from sandy soils over a sieve and separated leaves and seeds or flowers from stems. In the laboratory, root tissues were rinsed with DI water to remove any adhering soil. All tissues were oven dried at 70 °C for 48 h and weighed to the nearest milligram. In parallel, we harvested eight control individuals per species to determine allometric relationships across the range of biomass sizes. We measured the height and leaf number of all seedlings before the first nutrient treatment and used our allometric equations to account for pre-treatment biomass. To determine total plant biomass, we summed tissue mass from all compartments. To determine foliar C, N, P and Mo, we ground foliar tissues to a find powder and quantified C and N content through combustion, and P content from acid digestion (Greweling 1976; D’Angelo et al. 2001) and Mo through wet digestion (Pequerul et al. 1993) and analysis with inductively coupled plasma mass spectroscopy.

During harvest, we quantified whole-plant N2 fixation (henceforth referred to as instantaneous fixation) by performing assays of acetylene reduction activity (ARA) on all the root nodules from each plant. We also calculated fixation investment, which was the rate of fixation per unit mass of plant. For each plant, we removed all live nodules from the root system and immediately placed them into sealed 250 ml glass jars fitted with a rubber septum. We replaced 10% of the jar headspace with acetylene (C2H2) and collected gas samples twice over 30 min. We conducted blank incubations without nodules to account for background levels of ethylene (C2H4) in the acetylene gas. All gas samples were analyzed for C2H4 production using a gas chromatograph. We empirically determined C2H4 production to N2 fixation ratios for each species by simultaneously performing ARAs and 99 atom  % 15N2 incubations on the nodules from a subset of individual plants (n = 17) (Robertson et al. 1999). To calculate these ratios, we ground nodule tissues to a fine powder and determined δ15N of enriched and unenriched tissues with isotope ratio mass spectroscopy. Our mean and (SE) for the C2H4:N2 conversion factor across species was 4.86 (0.54).

Total N2 fixation

For plants in experiment 2, we used the 15N dilution method (Robertson et al. 1999) to calculate the percentage of plant N derived from the atmosphere (%Ndfa) and total N2 fixation over the lifetime of the plants (henceforth referred to as total fixation). To obtain the 15N content of our legume and reference plant tissues, we mixed and pulverized all tissue pools for each plant (n = 6 plants per species by treatment combination) and analyzed a subsample of those tissues for δ 15N using a mass spectrometer as above. To calculate  %Ndfa, the soil δ15N pool must be resolved from the atmospheric signal, thus we used the low (+N) and high N (+++N) treatments with the following equation (Robertson et al. 1999):

$$\% {\text{N}}_{\text{dfa}} \, = \,\left( { 1 - \left( {{\text{AE}}_{\text{f}} /{\text{AE}}_{\text{r}} } \right)} \right)\, \times \, 100,$$

where %Ndfa is a percentage of plant nitrogen derived from the atmosphere, AEf is a difference in atom % 15N between the labeled fixer and the control fixer and AEr is the difference in atom % 15N between the labeled reference and the control reference.

We calculated total fixation by multiplying the fraction of Ndfa values by total plant N. We calculated total fixation investment by dividing total fixation by total plant biomass.

Statistical analysis

To determine the differences in soil chemistry and texture between our two sites, we used two-tailed t tests with equal variances. To determine if the addition of nutrients (i.e., +P, +N, and +++N) enhanced nutrient availability to plants and affected biomass growth relative to controls, we fit linear models and used the glht function in the multcomp package to perform a priori contrasts with Bonferroni-corrected P values (Hothorn et al. 2008). In the case of Mo, we conducted t tests to compare the foliar Mo concentration between +Mo and the control.

To analyze the effect of P and/or Mo on biomass growth (g), nodule investment (g nodule g−1 plant), nitrogenase activity (µmol N g−1 nodule h−1), instantaneous fixation (µmol N plant−1 h−1), and fixation investment (µmol N g−1 plant h−1), we fit linear models where we treated P, Mo, species and soil origin as categorical variables. We analyzed factor effects and their interactions using a three-way ANOVA and removed variables that were not significant. To assess significant interactions, we used the glht function in the multcomp package (Hothorn et al. 2008) to perform post hoc linear contrasts with Tukey correction on the interactions.

To further isolate the mechanisms that control fixation, we fit a linear model testing the effect of biomass growth (g) on instantaneous fixation (µmol N plant−1 h−1) of plants from experiment 1, using species identity and nutrient treatment as fixed effects. We analyzed categorical effects with an ANCOVA and post hoc linear contrasts as described above.

To determine species-specific fixation strategies, we assessed how N addition affected total fixation investment (mg N g−1 plant) with a three-way ANOVA with N treatment (+N, +++N), species and soil origin as categorical variables, and we removed interaction terms that were not significant. Where we found a significant three-way interaction, we analyzed the effects of N treatment and soil origin for each species separately using a linear model and a two-way ANOVA. We analyzed the effect of N on plant tissue chemistry with three-way ANOVAs as described above.

To determine the effects of species identity on our response variables (biomass growth, nodule biomass, total fixation and total fixation investment), we performed a priori contrasts between the different species used in experiment 2. As an independent test of fixation differences among species, we also determined species differences in instantaneous fixation for plants exposed to the − N (control) treatment using a linear model as above. All statistical analyses were conducted in R (version 3.3.2, R Development Core Team, Vienna, Austria).

Results

P and Mo effects

We first determined how P and Mo addition affected legume growth and fixation. The strongest response was a stimulatory effect of P addition on both variables, supporting our hypothesis. P addition stimulated biomass growth, nodule investment, instantaneous fixation and fixation investment, but had no effect on nitrogenase activity (Fig. 1, Table 3, Online Resource 1). P addition had a stronger positive effect on biomass growth for legumes in Eglin versus Benning soil (Online Resource 3), indicating a contributing effect of soil origin (see below). In support of the finding that P addition enhanced the rate and investment of fixation by legumes, plant tissue P and N concentrations increased, while tissue C:N decreased in response to P addition (Table 3).

Fig. 1
figure 1

Growth and fixation responses of legumes to P addition, including a nodule investment, b nitrogenase activity, c instantaneous fixation and d fixation investment. Fixation rates were determined via ARA and conversions to N2 and expressed on b a nodule mass, c total plant or d per plant mass basis. Bars represent the mean ± SE of all species across two levels of P (−P, +P), aggregated across both soil types, and asterisks denote significant response to P (α = 0.05)

Full size image

To understand how P regulates growth and fixation in herbaceous legumes, we analyzed instantaneous fixation rates across plants as a function of the biomass growth rate of individual plants, with or without P addition. The fixation rate increased exponentially with biomass growth for both control and P addition plants (F1,180 = 182.9, P < 0.001, R2 = 0.58), but P addition alone increased instantaneous fixation only modestly (F1,180 = 4.48, P = 0.04) (Fig. 2). These results support the idea that P chiefly limits biomass growth, and that the larger P-supplemented plants then required higher rates of instantaneous fixation to support their growth.

Fig. 2
figure 2

Instantaneous fixation rate of herbaceous legumes as a function of biomass growth over the experimental period. Linear models for control plants (circles and solid line) and P addition plants (triangles and dashed line) aggregated from both soil types were back-transformed. Fixation rates were determined via ARA and conversions to N2. Species codes: CEVI Centrosema virginianum, CHFA Chamaecrista fasciculata, LEHI Lespedeza hirta, MIQU Mimosa quadrivalvis, TEVI Tephrosia virginiana. This figure is available in color in the online version of the journal

Full size image

We found no support for the alternative hypothesis that Mo would constrain legume fixation, either alone or in combination with P. Although our Mo treatment increased foliar Mo concentrations relative to the control (Table 3), Mo addition had no effect alone or in interaction with P on biomass growth, instantaneous fixation, nodule biomass or any other response variables in our experiment (P > 0.05 for all variables).

Soil origin effect

We sampled soils from two sites (Benning and Eglin), and although both sites are classified as sandhills, they possessed distinct chemical and physical properties which affected legume growth and fixation. Relative to Eglin, soils from Benning stimulated legume biomass growth, instantaneous fixation and nodule biomass (Online Resource 1). We determined the chemical and physical properties of these soils at the beginning of the experiment. While Eglin soils contained slightly higher amounts of total soil N, Benning soils contained higher available P (Mehlich III- and resin-extractable P) and a higher pH, but a lower sand content (Table 2). In contrast, wet-digestible soil Mo did not differ between sites. Differences in soil properties, combined with the P growth response, suggest that the differences in observed legume growth and fixation between soil types may largely be due to higher P availability in Benning relative to Eglin soils.

Table 2 Chemistry and texture of experimental soils from longleaf pine sandhills at Fort Benning and Eglin Air Force Base
Full size table

Effects of species and N addition

We next determined how species identity and N addition affected fixation among the herbaceous legumes of our study. To understand species-specific differences, we analyzed instantaneous fixation rates across control plants from both experiments (across seven species), and found that Mimosa quadrivalvis and Tephrosia virginiana supported the highest rates of fixation, followed by Centrosema virginianum, Desmodium floridanum, Lespedeza virginica, Chamaecrista fasciculata and L. hirta possessed the lowest rate (F6,104 = 10.74, P < 0.001) (Fig. 3, Online Resource 4). Species also differed in their total investment in fixation and in their response to N addition (Fig. 4, Online Resource 2 and 4). Four species (Centrosema virginianum, Desmodium floridanum, Lespedeza hirta and L. virginica) reduced their total fixation investment when exposed to +++N treatment, and this response was significantly more pronounced for C. virginianum and D. floridanum when grown in Eglin soils (Online Resource 2). However, three other species (Chamaecrista fasciculata, Mimosa quadrivalvis and Tephrosia virginiana) maintained fixation investment, even when exposed to higher availability of N (Fig. 4). Total plant N (%) increased and tissue C:N decreased across all species in response to +++N but not in response to +N, indicating our high N treatment increased plant N uptake (Table 3).

Fig. 3
figure 3

Species differences in instantaneous fixation rate at the end of the experiment. Box plots represent median and quartiles, and letters indicate Tukey’s post hoc comparisons (α = 0.05). Fixation rates were determined via ARA and conversions to N2. Species codes: CEVI Centrosema virginianum, CHFA Chamaecrista fasciculata, DEFL Desmodium floridanum, LEHI Lespedeza hirta, LEVI Lespedeza virginica, MIQU Mimosa quadrivalvis, TEVI Tephrosia virginiana

Full size image
Fig. 4
figure 4

Species differences in total fixation investment in response to N addition. Box plots depict median values and quartiles, and letters indicate Tukey’s post hoc comparisons (α = 0.05). Fixation rates were quantified from 15N dilution, where plants received either low or high additions of 15N. Species codes: CEVI Centrosema virginianum, CHFA Chamaecrista fasciculata, DEFL Desmodium floridanum, LEHI Lespedeza hirta, LEVI Lespedeza virginica, MIQU Mimosa quadrivalvis, TEVI Tephrosia virginiana

Full size image
Table 3 Nutrient treatment effects on foliar tissue chemistry and biomass growth
Full size table

Discussion

Longleaf pine savannas experience chronic N loss from frequent fire, yet we lack an understanding of what constrains herbaceous legumes from replenishing these losses through fixation. In a greenhouse experiment with field soils, we found that species identity and soil P were both strong determinants of N2 fixation among legumes native to longleaf pine savannas. These results suggest that in natural ecosystems, variation in N2 fixation rates may result from the diverse assemblages of herbaceous legumes across the longleaf pine landscape (Hainds et al. 1999) and how these assemblages respond to differences in soil P. Perhaps most importantly, our results suggest that soil P could constrain the ability of legumes to replace N lost from fire.

P constrains N2 fixation

Our findings reveal soil P as a strong constraint on N2 fixation for all legume species in this study. The addition of P stimulated growth, instantaneous fixation, and how much plants invested in nodule biomass and fixation. The strong P effect on plant biomass growth and fixation (Table 3 and Fig. 1) suggests that P increased growth and thus also increased plant N demand (Perreijn 2002; Batterman et al. 2013). Further analysis of instantaneous fixation supports such a physiological response—instantaneous fixation increased as a function of plant growth, whereas the addition of P alone had only a modest independent effect on fixation (Fig. 2).

Soil origin influenced the strength of P limitation in our study, with differences in legume growth and fixation rate between soils collected from our Fall-Line (Benning) versus Coastal Plain (Eglin) sites. Since soil origin and P addition similarly affected growth, fixation and nodule biomass, it is reasonable to infer that the growth differences we observed resulted from differences in P availability between the two sites (Table 2). Indeed, Mehlich-extractable P was ~ 40% higher and resin-extractable P was tenfold higher in Benning than in Eglin soils. While both soils were sufficiently P poor to induce P limitation, the relative response to P on growth and fixation was nevertheless threefold greater for plants grown in Eglin versus Benning soils. Our findings therefore suggest that P availability is likely to constrain fixation in longleaf pine sandhill ecosystems, but that variability in underlying parent material (Peet 2007) and soil weathering (Craft and Chiang 2002) may determine the magnitude of this P constraint across the longleaf pine landscape.

The addition of Mo increased foliar Mo concentration in our experiment, but contrary to our expectation, Mo addition had no individual or interactive (with P) effect on growth or fixation. These results imply that supplies of Mo are low, but sufficient for the growth of herbaceous legumes at our study sites. While it is possible that our treatment did not sufficiently alleviate Mo limitation on growth and fixation, two lines of evidence suggest that this is not the case. First, foliar Mo concentrations increased by over an order of magnitude with the addition of Mo, with no concurrent changes in fixation or biomass growth. Second, foliar Mo concentrations in our plants were high (control) or exceedingly high (+Mo treatment) relative to those of a related N2-fixing species in a nearby ecosystem where Mo has been implicated in constraining N2 fixation (Hungate et al. 2004).

Species identity and fixation strategy

Our results indicate that herbaceous legume species vary widely in their ability to supply new N to the ecosystem via N2 fixation, corroborating previous studies (Hiers et al. 2003; Cathey et al. 2010). Plant growth form, evolutionary history, and nodule morphology may contribute to such differences in fixation among species. In our study, a trailing vine (Mimosa quadrivalvis) supported high rates of instantaneous fixation relative to several prostrate species, which may reflect critical trade-offs as trailing vines do not invest in structural support for vertical growth (Pate 1996). Interestingly, Chamaecrista fasciculata, the only annual species in our study, supported the least amount of growth and fixation, which may be attributed to high investment in reproduction (instead of growth) and an early onset of senescence at the end of our study period (Antonovics 1980; Pate 1996). Evolutionary history and nodule morphology may also contribute to interspecific differences in fixation (Sprent 2009; Doyle 2011). In our study, the three species with determinate nodule growth, Desmodium floridanum, Lespedeza hirta and L. virginica (Desmodieae tribe), had lower nodule biomass and less total fixation compared to Tephrosia virginiana and Mimosa quadrivalvis (Millettieae and Mimosae tribes, respectively), which possess indeterminate nodule growth forms (Sprent 2009).

We observed two different N2 fixation strategies among the seven species studied. Centrosema virginianum, Desmodium floridanum, Lespedeza hirta, and L. virginica demonstrated a facultative response to N addition and downregulated their investment in fixation, while Chamaecrista fasciculata, Mimosa quadrivalvis and Tephrosia virginiana demonstrated an obligate response. Downregulation was particularly strong for Centrosema virginianum and Desmodium floridanum in Eglin soils, likely due to P constraints on growth (see below) and hence a lower demand for N relative to Benning soils. It remains a possibility that the N addition levels of our experiment were not high enough to trigger downregulation, especially for fast-growing species like Mimosa quadrivalvis and Tephrosia virginiana, though plants received an amount of N (30 mg N total over 16 weeks) comparable to that of foliar plant N (37.7 mg N average value) at the end of the experiment. Interspecific differences in fixation strategy observed in this study support recent findings that temperate herbaceous legumes differ in fixation strategy (Menge et al. 2015) and point to the various roles that individual species play in the timing of fixation in longleaf pine savannas.

While Chamaecrista fasciculata, Mimosa quadrivalvis and Tephrosia virginiana all emerged as obligate N2-fixers in our study, and M. quadrivalvis and T. virginiana fixed nearly four times more N2 than did C. fasciculata. One explanation for such differences is that species-specific differences in growth rates may influence plant N demand and thus also fixation rates (Wurzburger and Hedin 2016). Our findings support this idea, as these species were among those that accrued the most (Tephrosia virginiana and Mimosa quadrivalvis) and the least (Chamaecrista fasciculata) biomass over the experimental period. However, species also varied in the amount of N they fixed on a relative basis, suggesting that factors beyond growth determine the intrinsic N demand of a species.

Our findings suggest that herbaceous legume species differ in the amount and timing of fixation. The supply of new N to the ecosystem may thus depend on the identity of species, the biomass of individuals and the supply of N relative to demand. However, certain species, such as Tephrosia virginiana and Mimosa quadrivalvis, may contribute disproportionately to supplying new N because of their high rates of fixation and growth. Interestingly, these species also demonstrate an obligate fixation strategy, which suggests that they may continue to invest in fixation independent of fire-induced changes in N supply and demand. Because our analysis of species differences was conducted under low and high levels of N addition (Fig. 4), species with an obligate strategy may be more likely to emerge as higher contributors to total fixation. However, we observed similar patterns among species in instantaneous fixation under control conditions (–N) (Fig. 3), which substantiates the role of Tephrosia virginiana and Mimosa quadrivalvis as major contributors to fixation in longleaf pine ecosystems.

P and fixation strategy interactions

Our findings raise questions about how soil P regulates fixation among species with differing fixation strategies. If obligate N2-fixers fix at relatively constant rates proportionate to plant size, then P availability may ultimately control fixation by regulating growth (Augusto et al. 2013). For facultative N2-fixers, a sufficient N supply may only downregulate fixation in severely P-poor conditions, since N demand due to growth will be constrained by soil P. Our findings support this idea, as we observed greater downregulation in P-poor Eglin soils than Benning soils when Centrosema virginianum and Desmodium floridanum received high doses of N. Consequently, the availability of soil P appears to emerge as the governing and most fundamental constraint on fixation, despite a diverse leguminous community comprising differing growth and fixation strategies.

Conclusion

Our results demonstrate that species identity and nutrient availability exert substantial control on fixation in legumes native to longleaf pine savannas. Species varied nearly fourfold in fixation capacity, which may result from patterns in growth, taxonomy and fixation strategy. Fixation responded strongly to P addition across all species in our study, suggesting that the P cycle controls N inputs by herbaceous legumes. Field-based nutrient manipulations are needed to help translate our greenhouse study to a field setting, but our findings suggest that the potential for savanna legumes to offset N losses from fire depends on both the composition of legume communities and gradients in soil P across the landscape of longleaf pine ecosystems (Craft and Chiang 2002; Boring et al. 2017).