Nitrogen and phosphorus interact to control tropical symbiotic N2 fixation: a test in Inga punctata



  1.  Symbiotic di-nitrogen (N2) fixation is a critical biogeochemical process in tropical forests, yet it remains unresolved how fixation is controlled by the availability of soil nitrogen and phosphorus, two nutrients often considered limiting in terrestrial ecosystems.
  2.  We examine whether individual N2-fixing trees can overcome nitrogen and phosphorus constraints by employing different strategies of nutrient acquisition and use: N2 fixation, phosphatase exudation, mycorrhizal symbiosis and changes in root–shoot ratio or tissue stoichiometry.
  3.  We grew a common and widespread N2 fixer, Inga punctata, in a full factorial nitrogen and phosphorus addition experiment (each nutrient at three levels) and evaluated whether trees adjusted their strategies of nutrient acquisition to overcome limitation.
  4.  N2 fixation was controlled by nitrogen availability in phosphorus-sufficient soils, but both fixation and plant growth were constrained by phosphorus in the unamended native phosphorus-poor soils. Despite the investment in both extracellular phosphatases and mycorrhizal symbionts, plants were unable to overcome phosphorus limitation.
  5.  Our findings support the hypotheses that: (i) N2 fixation is proximately controlled by nitrogen availability, consistent with a facultative fixation strategy, and (ii) N2 fixation and N2 fixer biomass growth are ultimately constrained by soil phosphorus. We found no support for the hypothesis that fixers can overcome phosphorus limitation by trading fixed N2 for soil phosphorus.
  6.  Synthesis. This study provides new knowledge about how nitrogen and phosphorus interact to regulate tropical N2 fixation by examining a suite of strategies that plants may employ to overcome nutrient limitation. These findings, focused at the organismal level, have broader implications for biogeochemical controls at the ecosystem level in tropical forests.


Di-nitrogen (N2) fixation is a critical process that determines the biogeochemical function of tropical forests (e.g. Högberg 1986; Vitousek & Howarth 1991; Galloway et al. 2004; Houlton et al. 2008; Gerber et al. 2010; Reed, Cleveland & Townsend 2011), but opinions differ as to the factors that govern this mechanism. On the one hand, fixation may occur in response to the availability of soil nitrogen, with fixation up-regulated when soil nitrogen is low and down-regulated when nitrogen becomes abundant. On the other hand, fixation may impose a high phosphorus demand and may therefore occur only when soil phosphorus is sufficient, and thus fixation may be constrained in tropical forests where phosphorus is often considered the limiting element (Vitousek & Howarth 1991; Binkley, Senock & Cromack 2003). Recently, a third alternative perspective has been proposed that fixation is favoured in phosphorus-poor tropical forests because fixation allows fixers to harvest additional soil phosphorus via the investment of fixed nitrogen in extracellular phosphatases (Houlton et al. 2008; Baribault, Kobe & Finley 2012).

Resolving these hypotheses is critical since fixation is central to understanding the dynamics of the land carbon sink (Gerber et al. 2010), and since coupled land carbon models increasingly include nitrogen and/or phosphorus feedbacks (IPCC 2007; Thornton et al. 2007; Yang et al. 2009; Gerber et al. 2010; Goll et al. 2012). Of particular importance is fixation by trees in symbiosis with rhizobia (‘symbiotic fixation’) as these fixers are abundant in many tropical forests, have access to substantial reducing power (light) and can catalyse high rates of ecosystem N2 input (Hedin et al. 2009; Sprent 2009; Barron, Purves & Hedin 2011).

The distinction between the three alternative models depends upon the properties of individual trees and the strategies of nutrient acquisition that they employ in response to nutrient deficiencies. These strategies involve either the alteration in pathways and rates of external nutrient uptake (N2 fixation for acquiring nitrogen and extracellular phosphatase production or symbiosis with mycorrhizae for acquiring phosphorus) or the adjustment of the efficiency of internal nutrient use (by changing root–shoot biomass or tissue carbon:nitrogen:phosphorus stoichiometry) (e.g. Chapin 1980, 1980; Tilman 1982; Vitousek 1982; Bloom, Chapin & Mooney 1985; McGroddy, Daufresne & Hedin 2004; Reich & Oleksyn 2004; Yavitt et al. 2011). At heart, understanding fixation in tropical forests may depend on resolving how individual plant strategies of fixation and nutrient acquisition respond to, and feed back upon, nutrient cycles that emerge at the ecosystem level.

The first hypothesis suggests that fixation only occurs when soil nitrogen is scarce (e.g. Högberg 1986; Vitousek & Howarth 1991; Hartwig 1998; McHargue 1999; Barron, Purves & Hedin 2011; Reed, Cleveland & Townsend 2011). At the individual level, we would expect that if individual fixers employ a facultative fixation strategy (Hedin et al. 2009; Menge, Levin & Hedin 2009; Barron, Purves & Hedin 2011), fixers will minimize energetic costs (Gutschick 1981; McKey 1994; Vance 2008) by down-regulating fixation as nitrogen availability increases (Barron, Purves & Hedin 2011). Alternatively, if fixers employ an obligate fixation strategy, they will be incapable of adjusting fixation rates, regardless of soil nitrogen (Menge & Hedin 2009). At the ecosystem level, forests with trees that employ either strategy will experience a loss of fixation as soil nitrogen increases; however, as fixation becomes less beneficial, facultative fixers will remain present (although not fixing), while obligate fixers would disappear completely through competitive exclusion.

The second hypothesis assumes that individual fixers will only grow and fix N2 when supplied with sufficient phosphorus (e.g. Cole & Heil 1981; Högberg 1986; Vitousek & Howarth 1991). This phosphorus constraint can occur either because fixation uniquely requires phosphorus-rich metabolites (Gutschick 1981; Israel 1987; Gentili & Huss-Danell 2003) or because fixation allows fixers to grow and thus creates phosphorus demand (Robson, O'hara & Abbott 1981). Such a phosphorus constraint would translate into low fixer abundances and depressed ecosystem fixation rates in phosphorus-poor forests.

A third and contrasting hypothesis holds that fixers will have an advantage over non-fixers because they can fix and invest in extracellular phosphatases and mycorrhizal fungi when phosphorus-deprived. This mechanism depends on the investment of fixed N2 in the production of extracellular phosphatase enzymes or through the use of fixed N2 for enhanced photosynthesis that, in turn, supplies carbon for mycorrhizal fungi. Both mechanisms could allow the plant to access additional phosphorus (Robson et al. 1981; Joner et al. 2000; Wang et al. 2007; Houlton et al. 2008; Turner 2008; Baribault et al. 2012). We would expect that since fixation allows fixers greater access to fixed N2, increased fixation rates would correlate with greater investment in extracellular phosphatases and mycorrhizal colonization. Fixers should also continue to grow, and ecosystem fixation should remain high in phosphorus-poor soils. This mechanism could explain the build-up of nitrogen relative to phosphorus commonly observed in tropical soils (Houlton et al. 2008).

The response mechanisms invoked in these three hypotheses do not, however, constitute the only ways by which plants could adjust to soil nutrient conditions. Two commonly proposed additional mechanisms could apply to any of the three hypotheses and would influence a plant's ability to avoid nutrient limitation in general: (i) an increase in root–shoot allocation or (ii) a stoichiometric decrease in tissue content of a potentially limiting nutrient. We therefore evaluate each of these two mechanisms as specific cases of potential plant response within each hypothesis.

While tropical N2 fixers may respond to additions of nitrogen or phosphorus (Van Kessel & Roskoski 1981; Ribet & Drevon 1996; McHargue 1999; Thomas, Bashkin & Richter 2000; Binkley, Senock & Cromack 2003; Patreze & Cordeiro 2004; Pons et al. 2007), our experimental approach is to examine the integrated response of a suite of plant strategies that potentially allow N2 fixers to overcome varying soil constraints and that have direct implications for major hypotheses of how fixation functions in tropical forests.

Specifically, we evaluate at the scale of individual N2-fixing plants strategies of nutrient acquisition and use that may or may not respond to soil nutrient conditions. We manipulated nitrogen and phosphorus and evaluated the ability of N2 fixers to employ five different mechanisms (each central to one or more of the hypotheses discussed above): (i) symbiotic N2 fixation, (ii) extracellular phosphatase production, (iii) symbiosis with mycorrhizal fungi, (iv) root–shoot allocation and (v) tissue stoichiometry. We interpret these mechanisms in the light of whether they support one or a combination of the three hypotheses.

We conducted a full factorial experiment in a greenhouse setting; this allowed us to control effects caused by light and soil moisture and to exclude herbivory. While it is difficult to generalize across the diverse and widely distributed taxa that make up N2-fixing plants, we focused our study on Inga punctata Willd., a tree species that is common in Panamanian forests and widespread across the Neotropics. This species belongs to one of the most common N2-fixing genera in tropical forests: Inga comprised 1.6% of individual trees and 20% of known fixers in 50 ha of mature forest on Barro Colorado Island, and > 11% of trees and > 60% of fixers in a nearby young secondary forest in Panama (Batterman et al. 2013).

Materials and methods

Experimental Set-up

We conducted the open-air greenhouse study on Barro Colorado Island (BCI) of the Smithsonian Tropical Research Institute (STRI), Panama. We grew the native N2-fixing tree, Inga punctata (Fabaceae subfamily Mimosoideae), in phosphorus-poor soils from the Wetmore soil type on BCI. In this soil, levels of total (0.12–0.24 g kg−1) and organic (15–47 mg kg−1) phosphorus and of total nitrogen (0.38–1.52 g kg−1) (Dieter, Elsenbeer & Turner 2010) are relatively low when compared to tropical soils world-wide (Dieter, Elsenbeer & Turner 2010; Quesada et al. 2010; Turner & Engelbrecht 2011). Soil bulk density on BCI is 0.85 ± 0.05 Mg m−3 (Schwendenmann, Pendall & Potvin 2007). Inga punctata is common in primary and secondary forests throughout Panama and belongs to one of the most widespread, diverse and abundant taxa of symbiotic fixers across the Neotropics (Barron, Purves & Hedin 2011). The species has evolved in taxonomically diverse forests (c. 300 tree species in 50 ha plot of BCI) where fixers are abundant (8% of stems; Hedin et al. 2009) and adapted to warm (annual average of 26 °C) and wet conditions (2600 mm year−1; Leigh 1999).

We germinated I. punctata seeds from nearby forests and grew them for 4 months in homogenized soil to expend seed nutrient reserves and acclimatize plants. Once transplanted into 7.5 L pots, fixers received a homogenized inoculum of fine roots and nodules from native forests to ensure access to a full suite of bacterial and fungal symbionts. Nutrients were added weekly as three nitrogen levels (0N, +N and ++N, corresponding to 0, 30 and 60 g N m−2 year−1 as NH4NO3) and three phosphorus levels (0P, +P and ++P, corresponding to 0, 20 and 40 g P m−2 year−1 as KH2PO4) in full factorial combination (nine treatments total) with 10 replicates per treatment. All treatments were supplemented with an addition of nitrogen and phosphorus-free Hoagland's solution to ensure that all other macro- and micronutrients were balanced and non-limiting. In the case of potassium, which has been shown to influence some plant processes in Panamanian forests (Santiago et al. 2011; Wright et al. 2011), we adjusted the Hoagland's solution to account for the potassium added with the phosphorus solution. We did not examine the potential role of the trace element molybdenum in fixation (Barron et al. 2009; Wurzburger et al. 2012), as we focus on the two nutrients that have historically been considered the most critical to fixation. All plants received two small applications of organic pesticides directly to the leaves and were watered regularly.

Plant Responses

We harvested plants after 10 weeks before plants were pot-bound (c. 0.5 m in height for the tallest plants). We quantified plant biomass growth, root nodule biomass, fixation activity, phosphatase activity, mycorrhizal colonization, root–shoot allocation and tissue carbon:nitrogen:phosphorus content. We divided tissues into leaves, stem, fine roots 0–1 mm, fine roots 1–2 mm, coarse roots > 2 mm and nodules. We weighed biomass after drying at 65 °C for 24–96 h, pulverized and homogenized tissues to a fine powder and subsequently analysed tissues for carbon and nitrogen by a NC2500 Elemental Analyzer (CE Elantech, Lakewood, NJ, USA) in our laboratory at Princeton University and for phosphorus by digestion followed by flow injection analysis at the University of Nebraska (QuikChem 8500; Lachat Instruments, Loveland, CO, USA). We calculated plant growth from the individual plant biomass at the experiment end (sum of all tissue pools) minus the individual biomass at the experiment start (estimated from allometric equations developed from plants harvested at experiment start; r2 > 0.92; 187 leaves and 10 stems, roots and nodules from 10 plants).

Nutrient Acquisition Strategies

We quantified nodule biomass and measured fixation using acetylene reduction assay (ARA; Barron, Purves & Hedin 2011) with fresh nodules during the plant harvest. C2H4 was measured on a gas chromatograph (GC8610; SRI Instruments, Torrance, CA, USA) at BCI. All nodules, including those used for ARA, were cut open to confirm activity. Initial nodule biomass was determined from plants harvested at experiment start (n = 9), and investment in nodulation was calculated as the ratio of per plant nodule biomass to whole-plant biomass.

We measured phosphomonoesterase and phosphodiesterase activities with fluorometric analysis in Dr. Ben Turner's laboratory at STRI (Turner 2010). Fresh soil from the rooting zone was stored in a cold room for up to 2 weeks and analysed using 4-methylumbelliferyl phosphate (MU; for phosphomonoesterase) and bis-(4-methylumbelliferyl) phosphate (bis-MU; for phosphodiesterase) with excitation filter at 360 nm and emission filter at 460 nm.

We preserved fine roots (0–1 mm in diameter) from each plant in 95% ethanol. For mycorrhizal colonization estimates, we subsequently rinsed roots in DI water, cleared root tissues in 10% KOH (3–5 h in a 70 °C water bath) and stained tissues with trypan blue (0.05% in a 1:1:1 mixture of lactic acid, glycerol and DI water). We quantified fraction of root length associated with mycorrhizal structures (arbuscules, coils, hyphae and vesicles) using a random intercept technique (McGonigle et al. 1990). Biomass-specific colonization (m mycorrhizal root g plant−1) was calculated as a function of total plant root length (RL):

display math

Plant root–shoot ratio was calculated as total below-ground biomass (roots and nodules) divided by total above-ground biomass (leaves and stem) and whole-plant carbon:nitrogen and carbon:phosphorus by summing total nutrient content across all tissues. Finally, we defined the phosphorus cost of fixation as the percent of total plant phosphorus that occurred in nodules.

Statistical Analyses

Data were log-transformed to achieve normality when necessary and analysed using R2.6.0 (R Development Core Team). anova and Tukey's honest significant difference post hoc comparison tests were used to test for differences between categorical variables; general linear models and Pearson's correlations were used to test for relationships between continuous variables.



We determined the ability of I. punctata to use strategies to overcome nutrient constraints by first examining growth as a function of nutrient conditions. Growth was strongly influenced by soil phosphorus but not nitrogen conditions (Fig. 1; phosphorus effect: < 0.001, n = 88, anova) following two patterns: First, trees were largely unresponsive to differences in nitrogen (n.s., Tukey's post hoc) when phosphorus was sufficient (+P, ++P levels in Fig. 1), such that fixers could overcome any local nitrogen deficiency. Second, I. punctata showed strong sensitivity to soil phosphorus, with a substantial depression in growth in the lowest phosphorus treatment relative to the +P and ++P treatments (although all plants in the 0P treatment survived and grew an average of 35%; 0P in Fig. 1; < 0.001, Tukey's post hoc). In addition, leaves of plants in the low-phosphorus treatment displayed increased necrosis, as is diagnostic for phosphorus-deficient plants, and leaves of plants in the 0P0N treatment displayed chlorosis, indicating that low phosphorus inhibited the ability of the plants to gain nitrogen via N2 fixation. Two lesser responses were visible but not statistically significant: a growth decrease with added nitrogen at the lowest phosphorus level (0P) and an apparent increase in biomass with added nitrogen at the highest phosphorus level (++P++N).

Figure 1.

Biomass growth (g) response of Inga punctata to nitrogen and phosphorus additions over 10 weeks. Biomass growth was calculated as the difference between final and initial biomass and is represented as the geometric mean ± SEM across three levels of phosphorus addition (0P, +P, ++P) and three levels of nitrogen addition (0N = white, +N = light grey, ++N = dark grey) (= 10 plants per treatment). Nitrogen effect: n.s., phosphorus effect: F = 71, < 0.001, interaction between nitrogen and phosphorus: n.s., n = 88, anova.

N2 Fixation

Given the lack of nitrogen effect and strong phosphorus effect on growth, we next determined the nitrogen and phosphorus conditions that allowed plants to use fixation to overcome nitrogen limitation. Additionally, we determined whether the fixation response was facultative or obligate by evaluating the ability of plants to adjust fixation rates across nutrient treatments. Whole-plant fixation rates, determined as per plant nodule biomass*per nodule biomass fixation rate, were highly responsive to both nitrogen and phosphorus (Fig. 2a; < 0.001 for both nitrogen effect and phosphorus effect, n = 83, anova). When phosphorus-sufficient (+P, ++P) but nitrogen-starved (0N), trees were highly effective at producing nodules and maintaining high fixation (Fig. 2a). When phosphorus-sufficient (+P, ++P), trees suppressed three different measures of fixation as a function of increasing nitrogen addition: per plant fixation rate (Fig. 2a; r = −0.68, < 0.001, n = 83, Pearson's correlation), per plant nodule density (data not shown; r = −0.64, < 0.001, n = 83) and per nodule mass fixation rate (data not shown; r = −0.55, < 0.001, n = 83). In contrast, at our lowest phosphorus level (0P in Fig. 2a), plants fixed at exceptionally low rates (<8% of the rate observed in the +P0N treatment) but still progressively suppressed fixation with increasing nitrogen (r = −0.61, < 0.01, n = 26, Pearson's correlation). Because plants adjusted fixation in response to their environment, we determined that I. punctata employs a facultative fixation strategy.

Figure 2.

N2 fixation response of Inga punctata to nitrogen and phosphorus additions. (a) N2 fixation rate (μmol C2H4 fixed h−1) represented as the geometric mean ± SEM across three levels of phosphorus addition (0P, +P, ++P) and three levels of nitrogen addition (0N = white, +N = light grey, ++N = dark grey) (= 10 plants per treatment). Nitrogen effect: F = 24, < 0.001, phosphorus effect: F = 64, < 0.001, interaction nitrogen and phosphorus: n.s., n = 83, anova. (b) Change in plant investment in nodulation relative to total plant biomass (g nodule g plant−1) over the 10-week experiment. Boxes represent the means (± SEM) of plants growing prior to nutrient additions at experiment start (Time = 0, n = 10) and plants growing with high phosphorus (++P) with and without nitrogen additions (as indicated by the arrows) at experiment end (Time = 10, n = 10 for each treatment).

Time-Scale and Magnitude of Fixation Response

We next evaluated the temporal scale of the facultative fixation response in I. punctata. Individual plants down-regulated fixation within 10 weeks when switched from insufficient nitrogen to sufficient nitrogen (Fig. 2b). Per plant fixation rate (F) was calculated as:

display math

where Nm is nodule mass per plant mass (extensive response; Fig. 2b), Fm is fixation rate per nodule mass (intensive response; data not shown), and B is biomass per individual plant. On average, plants reduced Nm by 85% and Fm by 58% compared to plants that continued to fix over the 10-week period (0N vs. ++N in Fig. 2b).


We next examined the ability of I. punctata to overcome phosphorus constraints by producing extracellular phosphatases. While phosphatase activity occurred in the rhizosphere across all treatments, any phosphorus that plants acquired through this pathway was insufficient to support growth in low-phosphorus soils. Moreover, we did not find any evidence that plants used fixation to increase phosphatase activity in phosphorus-poor conditions. First, both phosphatase forms displayed slightly negative (rather than positive, as expected if plants traded N2 for phosphorus) relationships with whole-plant fixation (Fig. 3a; < 0.05; n = 81; linear regression). Second, when averaged across all nitrogen additions, both phosphatases were slightly elevated (inline image = 10%) in the 0P treatment where fixation (and growth) was negligible (Fig. 3a inset) (Fig. S1a in the Supporting Information; < 0.05 for difference between 0P vs. ++P; Tukey's post hoc). Third, within each phosphorus treatment, phosphatase increased with fertilizer nitrogen addition, indicating a general response to soil inorganic nitrogen rather than individual plant fixation (Fig. S1a; nitrogen effect: < 0.05, n = 90, anova). Both phosphomonoesterase and phosphodiesterase responded similarly across all soils (r = 0.74, < 0.001, n = 90 plants; Pearson's correlation).

Figure 3.

Relationship between N2 fixation activity (μmol C2H4 fixed h−1) and phosphorus acquisition strategies. White points in main graphs represent all nitrogen and phosphorus treatments, while black points in insets represent the low-phosphorus (0P) level for (a) phosphatase activity (nmol MU g soil−1 min−1) (R2 = 0.05, d.f. = 81, < 0.05, phosphomonoesterase activity = 7.72–0.125*F) and (b) biomass-specific mycorrhizal colonization (m root colonized g plant−1) (R2 = 0.09, d.f. = 80, < 0.001, mycorrhizal colonization = 213.5–6.904*F).


We next examined whether I. punctata used mycorrhizal symbiosis as a strategy to overcome phosphorus constraints. As with phosphatases, trees could not use mycorrhizae to support growth in low-phosphorus soils (0P). While plants increased mycorrhizal colonization per plant biomass by inline image = 20% across all nitrogen addition levels in the 0P treatment, individual plants nevertheless barely grew or survived (0P in Fig. S1b) (< 0.05 for difference between 0P vs. ++P; Tukey's post hoc). Across all treatments, investment in mycorrhizal colonization illustrated a slightly negative relationship with fixation (Fig. 3b; < 0.001; n = 80; linear regression). Overall, we observed no consistent effect of nutrient treatment on mycorrhizae at any of the higher phosphorus treatments (+P, ++P) or at any nitrogen level (interaction between nitrogen and phosphorus: < 0.1, n = 89, anova).

Tissue Stoichiometry and Root–Shoot Allocation

We evaluated whether I. punctata could adjust tissue stoichiometry and root–shoot ratios in response to soil nutrient conditions. In the 0P treatment, plants increased the root–shoot ratio by inline image = 40% in all nitrogen treatments (0P effect: < 0.001, n = 89, Tukey's post hoc) and tissue carbon:phosphorus ratios by inline image = 65% (0P effect: < 0.001, n = 89, Tukey's post hoc), but neither strategy was sufficient to overcome the depression of plant growth (Fig. 4a,c). While plant growth was substantially reduced in the 0P treatment, it nevertheless remained positive (+35% biomass increase) over the course of the experiment. We observed isolated examples of leaf necrosis across all 0P treatments and widespread chlorosis in the 0P0N treatment.

Figure 4.

The response of biomass allocation and tissue stoichiometry of Inga punctata to nitrogen and phosphorus additions. Bars represented as the geometric mean ± SEM across three levels of phosphorus addition (0P, +P, ++P) and three levels of nitrogen addition (0N = white, +N = light grey, ++N = dark grey) (= 10 plants per treatment) for (a) root–shoot ratios (phosphorus effect: < 0.001, nitrogen effect: ns., n = 89, anova), (b) whole-plant C:N (interaction between nitrogen and phosphorus: F = 39, < 0.001, n = 90, anova), (c) whole-plant C:P (nitrogen effect: F = 4, < 0.05, phosphorus effect: F = 610, < 0001, interaction between nitrogen and phosphorus: n.s., n = 89, anova) and (d) fine root (1–2 mm diameter) C:N (nitrogen effect: F = 22, < 0.001, phosphorus effect: n.s., interaction between nitrogen and phosphorus: n.s., n = 63, anova).

Although the carbon:phosphorus ratio differed across tissue types, the ratio of each individual tissue (leaves, fine roots 0–1 mm, fine roots 1–2 mm) showed the identical response as whole plants (Table S1). In addition, in the 0P0N treatment, we observed a significant increase in carbon:nitrogen ratio of whole plants (inline image = 70%) and in all comparisons between leaf and fine root classes (inline image = 18–43%) (< 0.001 for all comparisons; Tukey's post hoc; Table S1). Fine roots 1–2 mm displayed an additional negative response to nitrogen addition across all phosphorus addition levels (nitrogen effect: < 0.001, n = 64, anova), implying that plants consistently decreased nitrogen (but not carbon) investment in fine roots to satisfy demands by photosynthetic tissues.

Phosphorus Cost of N2 Fixation

Phosphorus allocation to nodules varied as a function of nodulation, from a low of 0.7% in nitrogen-replete conditions (where nodules were few) to a high of 6.4% in treatments with no nitrogen addition but sufficient phosphorus (i.e. +P0N and ++P0N).


Individual Plant Strategies

Most broadly, our results show that I. punctata was highly flexible in its ability to adjust fixation to local soil nitrogen conditions as long as soil phosphorus was sufficient. Plants adjusted fixation to meet the demands of biomass growth when soil nitrogen was insufficient, but down-regulated fixation within weeks once nitrogen supply became sufficient. This flexibility indicates a facultative fixation strategy (cf. Barron, Purves & Hedin 2011) and implies that I. punctata has evolved to employ fixation in nitrogen-poor conditions while avoiding the costs of fixation in nitrogen-replete conditions (cf. Hedin et al. 2009). In addition, the ability to fix N2 when phosphorus was sufficient allowed plants to maintain tissue nitrogen content and root–shoot allocation independent of variations in soil nitrogen levels. These findings support our first hypothesis that the growth of fixers and their strategy of fixation are strongly linked to local soil nitrogen conditions.

In contrast to nitrogen, however, I. punctata did not possess any strategy that could successfully overcome phosphorus limitation. None of the mechanisms that we evaluated – phosphatase production, mycorrhizal symbiosis, root–shoot biomass investment or tissue carbon:phosphorus investment – were sufficient to support plant growth in the low-phosphorus conditions of our native soils. While plants were able to adjust root–shoot and carbon:phosphorus ratios, both responses occurred only in our most phosphorus-poor conditions in which plant growth was substantially depressed. All 0P treatments displayed low plant growth and isolated leaf necrosis, consistent with effects expected for phosphorus deficiency. However, plants in the 0P0N treatment also displayed widespread chlorosis, consistent with nitrogen deficiency and indicating that soil phosphorus was too low to sustain N2 fixation (which would remedy nitrogen deficiency, as observed in the +P and ++P treatments). We interpret this response as a consequence of severe stress and phosphorus limitation rather than as any successful strategy for acquiring soil phosphorus. As a result, I. punctata displayed a sharp phosphorus growth threshold: plants grew and fixed N2 when phosphorus-sufficient but grew little and fixed little N2 when phosphorus-deficient. These results support our second hypothesis that fixation can be limited by soil phosphorus conditions.

Our findings do not, however, support the idea that the fixation process itself confers upon fixers an additional phosphorus cost (Högberg 1986; Israel 1987; Vitousek & Howarth 1991; Gentili & Huss-Danell 2003). First, even in conditions where highest fixation occurred, nodules contained only a minor fraction (2–7%) of total plant phosphorus. Second, plant carbon:phosphorus ratios and root–shoot ratios were insensitive to whether or not I. punctata fixed, indicating that tissue phosphorus and plant biomass allocation did not depend on whether or not fixation was expressed. Limitation instead emerged as a consequence of the general demand for phosphorus imposed by plant biomass growth (as proposed by Robson, O'hara & Abbott 1981). The common appearance that fixers are uniquely sensitive to soil phosphorus may therefore stem from the ability of fixers to rapidly grow in nitrogen-poor soils rather than any disproportionate phosphorus demand caused by fixation per se.

We were surprised to find no support for the hypothesis that I. punctata can directly trade fixed N2 for phosphorus gain in low-phosphorus environments (Wang, Houlton & Field 2007; Houlton et al. 2008). We found no evidence that either of the two ‘mining for phosphorus’ mechanisms could provide enough new phosphorus to support plant growth in phosphorus-limited conditions. First, contrary to the expectation that plants trade fixed N2 for soil phosphorus, we found no positive relationship between soil phosphatases and fixation across treatments, including the most phosphorus-limited conditions. Second, phosphatase activity was only slightly higher (10%) in the low-phosphorus treatment (even though plants barely fixed) and was insufficient to sustain plant growth. Third, while plants supported greater (c. 37%) mycorrhizal colonization in our most phosphorus-poor conditions, this response was also inadequate for overcoming phosphorus limitation on growth. These results are consistent with model-based predictions of Rastetter & Shaver (1992): While phosphatases can increase the local availability of phosphorus, they cannot increase the total phosphorus in an ecosystem, and so their ability to alleviate limitation is limited.

Our lack of evidence for a N2-for-phosphorus trading strategy in I. punctata implies that this species is not uniquely advantaged relative to non-fixers in phosphorus-poor soils. This observation does not preclude the idea that conditions of high soil nitrogen can stimulate phosphatase production in a more general plant–soil phosphorus feedback (reviewed in Marklein & Houlton 2012), but this phosphatase effect could be caused by either plants or microbes, and benefits would apply to any organism (fixer, non-fixer or microbes) rather than uniquely to fixers.

Ecosystem Scale Implications

Our findings raise questions about how the interaction of N2 fixation with nitrogen and phosphorus scales up from individual trees to whole ecosystems for I. punctata. Our findings could apply to other species within the same genus and possibly more generally to other taxa of tropical N2 fixers. All species of Inga that we have examined in Panama appear to employ a facultative strategy similar to I. punctata (Barron, Purves & Hedin 2011; Batterman et al. 2013). From this context, we next examine three fundamental questions that arise from our findings.

First, does a facultative fixation strategy mean that soil nitrogen governs rates of fixation and abundance of fixers at the ecosystem scale? Our observation of highly flexible adjustment of fixation to soil nitrogen conditions implies that, at the ecosystem scale, fixation should not be assumed to be a simple linear function of the density of fixers. Instead, ecosystem fixation may be better predicted by combining the density and taxonomic identity of fixers with an understanding of the processes that generate and maintain nitrogen limitation in forests (e.g. forest treefall gaps, landslides, flooding, nitrogen leaching; Batterman et al. 2013).

Perhaps even more vexing are the mechanisms that maintain fixers and fixation within a forest community over time. First, our lack of evidence for a successful N2-for-phosphorus trading strategy implies that fixers may not be competitively favoured in phosphorus-poor conditions. Second, since the process of fixation will over time cause the disappearance of the conditions that favour fixers over non-fixers, the persistence of fixers must depend on the regeneration of nitrogen-poor conditions. Third, while facultative fixation allows fixers to avoid the energetic penalty of fixation when nitrogen is abundant, it does not explain how the strategy of fixation is selected for in the absence of nitrogen-poor conditions. These considerations imply that, over time, fixation has evolved in close association with processes that generate nitrogen-poor conditions in tropical forests (e.g. landslides, flooding and treefall gaps; Barron, Purves & Hedin 2011).

Second, does our finding that fixers grew little and barely fixed nitrogen in the low-phosphorus treatment mean that fixers commonly experience phosphorus limitation in natural tropical forests? While our results indicate a strong influence of phosphorus on fixation, observations from intact forests suggest that such limitation may in fact be rare. Fixers are common even in forests with very low soil phosphorus, including the Rio Negro region and the white sand soils of the Amazon (Coomes 1997; Pons et al. 2007; Fine et al. 2010). Fixers nodulate and presumably fix N2 in highly phosphorus-poor soils on the Guyana shield (Pons et al. 2007). In our Panamanian forests, N2-fixing taxa (including Inga) are common even in soils with phosphorus as low or lower than those used in this experiment (Wurzburger, Hedin, unpublished work.). These observations, while limited, illustrate a need for in-depth field data on fixers and fixation to better evaluate the mechanisms that may promote or inhibit phosphorus constraints on fixation.

How, then, are fixers able to fix nitrogen and grow in the natural ecosystem, but not in the 0P0N experiment? This question is of fundamental importance as it identifies the need to resolve whether fixer phosphorus demand can be met by the ecosystem phosphorus cycle when soils alone (as in our experiment) are incapable of supplying sufficient phosphorus. Plausible mechanisms include: (i) that build-up of a well-developed plant–soil phosphorus cycle combined with low net forest phosphorus demand (as in old-growth forests; e.g. Newbery et al. 2002) recycles enough phosphorus to allow N2 fixer growth and fixation. In contrast, a poorly developed phosphorus cycle in which net forest demand exceeds the ability of soils to supply phosphorus (as in forests recovering from disturbance) would instead promote phosphorus limitation. (ii) Mineralization and loss mechanisms tend to promote the availability of phosphorus over nitrogen in the soil organic horizon (Wurzburger et al. 2012). Such a general plant–soil feedback could, in effect, bring phosphorus supply above the threshold identified in our experiment.

Conclusions and implications for global models

Our findings for I. punctata are consistent with the idea that fixers and fixation may respond in a complex manner to both nitrogen and phosphorus across tropical forests. While strategies of fixation must be understood more generally across N2-fixing taxa, our findings can begin to inform us about how to incorporate N2 fixation as an essential dynamical process in biogeochemical models of tropical forests. These effects on N2 fixation may be important for understanding the tropical carbon sink, as illustrated by Gerber et al. (2010) who reported a large difference in predicted carbon uptake by tropical forests depending on how fixation is incorporated in Earth System Models (ESMs).

As a first approximation, our results indicate that tropical fixation ought to be incorporated as an ecologically facultative (rather than obligate) process that is highly responsive to spatial and temporal variations in soil nitrogen availability, and subject to a soil phosphorus threshold. The utility of including an ultimate constraint by phosphorus is less clear (cf. Goll et al. 2012) and probably depends on the degree to which the internal forest phosphorus cycle is balanced relative to local soil phosphorus availability. The prospect of global changes in climate, disturbance and nutrient deposition may, in turn, further alter where, how and why fixation is constrained by nitrogen and/or phosphorus at scales of individuals as well as ecosystems.


We thank Terri Shirshac, Helmut Elsenbeer, Jefferson Hall, Allen Herre, Katherine Li, Scott Mangan, Laura Morales, Ben Turner, Jill Urriola and Lydia Wheeler for various aspects of the conception and implementation of this project. This work was supported by grants to L.O.H. from the NSF (DEB-0614116), NOAA (grant NA17RJ262 - 344), and the Cooperative Institute for Climate Science of Princeton University, and to S.A.B. and N.W. from the Smithsonian Tropical Research Institute. The Smithsonian Tropical Research Institute provided facilities and support. The authors declare no conflict of interests.