Biological di-nitrogen (N₂) fixation is the major natural source of new N to tropical forests [1] and is thought to influence whether N or phosphorus (P) limits the ability of these forests to respond to CO₂ fertilization [2], [3]. The macronutrient P has for a long time been thought to limit N₂ fixation in tropical forests, but some observations question how generalizable this might be. From previous work in Hawaii, Vitousek [4] noted that trace metals might together with P influence asymbiotic N₂ fixation in soils of a montane tropical forest. Our own studies have demonstrated that molybdenum (Mo) but not P limits asymbiotic fixation in a Panamanian lowland forest [5]. Mo functions as a co-factor in the nitrogenase enzyme (which converts N₂ to ammonia) while P is needed for ATP and cell growth [6].

Here we examine the influence of both Mo and P on asymbiotic N₂ fixation in mature lowland tropical forests of Panama. Asymbiotic fixers (bacteria that live freely in soils) are ubiquitous and may contribute substantial atmospheric N₂ in conditions where symbiotic N₂-fixers (bacteria that live in association with plant roots) are absent, rare or not fixing (as observed in some mature forests) [7]–[9]. Both asymbiotic and symbiotic fixation contribute to the common pattern of abundance of N relative to P or other plant resources in many tropical forests [7], [10].

Limitation by Mo alone has also been reported for asymbiotic fixers in a highly leached temperate forest soil [11] as well as for symbiotic fixers in some managed pastures and agricultural systems [12], [13]. Responses to trace-metal cocktails combined with P fertilizer in Hawaiian soils [4], [14], [15] also raise the possibility of a Mo influence. Positive responses to P fertilizer [4], [5], [14]–[16] are, however, complicated by the observation that superphosphate fertilizer often contains trace amounts of Mo [5]. In our previous study of a Panamanian forest we found that asymbiotic fixation responded positively to Mo alone, to superphosphate fertilizer in which Mo was a “hidden” contaminant, but not to additions of trace-metal free P [5].

These findings raise fundamental questions about the role of Mo and P in constraining N₂ fixation: a) Are fixers sensitive to variations in these soil nutrients across landscapes; b) Do fixers respond to both Mo and P in the same location; and c) What conditions and what mechanism(s) favor limitation by one element over the other? These questions identify the need to understand how Mo, P, or both elements together act to constrain N₂ fixation in nature.

Molybdenum and P are both rock-derived and oxyanions in their bioavailable forms (molybdate, MoO₄²⁻ and phosphate, PO₄³⁻ respectively). It is plausible that Mo limitation emerges because molybdate is susceptible to leaching, complexation by organic matter and adsorption to iron (Fe) oxides in highly weathered soils [13], [17]. However, weathered tropical forest soils can also be substantially poor in P [18] and phosphate availability is reduced by sorption and immobilization processes [19]. The remarkable heterogeneity that exists in weathering conditions and distributions of soil nutrients across tropical forests [20] offers further motivation for resolving the dependence of N₂ fixation on Mo and P across tropical landscapes.

We chose six forest sites distributed along a steep soil P∶Mo gradient in Panamanian lowland forests, across which total soil P spans the range reported from tropical forests world-wide while Mo remains relatively constant. The gradient derives from local variations in geology where andesitic and basaltic lithologies give rise to P-rich soils while marine sediments and rhyolitic tuff produce P-poor soils [21], [22]. The forests are within 8 km of each other, share a similar pool of tree species, and experience a similar climate.

We examined whether limitation on N₂ fixation depended on landscape-scale variations in the relative abundances of soil Mo and P. We predicted that Mo would limit fixation in P-rich soils while P would limit in P-poor soils. In each forest, we evaluated the response of N₂ fixation to four nutrient addition treatments: control (C), Mo only (+Mo), trace-metal-free P only (+P) and both elements in combination (+Mo+P). To further understand the mechanism by which nutrient limitation differed across sites we intensively sampled vertical distributions of fixation, Mo and P within each soil and conducted laboratory experiments on the ability of organic matter to interact with bioavailable Mo and P.

Soils from our sites in lowland Panama span the range of total P content reported from tropical forests on well-developed soils world-wide (Fig. 1a): from P-rich montane soils [31] (similar to our AVA site) to large areas of P-deficient soils in lowland Amazonia [20] (similar to our Rio Paja site). In contrast, total soil Mo was low compared to the few measures available from natural soils [32] and did not vary systematically across our sites. Therefore, our gradient primarily captures variations in P relative to Mo and provides a localized experimental system for understanding how P and Mo might interact to constrain N₂ fixation in tropical landscapes.

Our vertical sampling of soil profiles revealed that asymbiotic fixation primarily occurred in the O-horizon across all our sites [4], [5], [15], [33]. We infer that asymbiotic N₂-fixers are stimulated in the O horizon by the continual input of energy-rich plant litter that is poor in N (as indicated by high C∶N ratios) relative to the physiological demand of soil microbes. Although a suite of environmental factors (e.g., moisture, temperature) can influence the rate of fixation, our goal here was to examine the specific role of Mo and P in limiting fixation in the O horizon of a given soil.

We predicted that Mo would limit N₂ fixation in P-rich soils while P would limit in P-poor soils. Nutrient amendments stimulated N₂ fixation substantially in four of our six sites following a clear pattern: fixation was limited by Mo alone in P-rich soils but co-limited by both Mo and P in P-poor soils. Contrary to our expectation, we never found exclusive P limitation at any of our sites. The influence of Mo extended even to the most P-poor sites, where additions of both Mo and P were necessary to stimulate rates of fixation (Zetek and Rio Paja in Fig. 1b). We did not observe statistically significant responses to nutrient additions in two locations, perhaps due to the influence of edaphic factors other than Mo and P [34] as suggested by the low fixation rates observed at the AVA site. The slightly weaker response to combined Mo and P addition at Gigante was possibly caused by interference of the macronutrient phosphate on the availability of the micronutrient molybdate.

Our results offer a direct experimental demonstration that asymbiotic N₂-fixers can be influenced by broad variations in Mo and P availability across forests, and that the influence of Mo can extend (either alone in P-rich soils or in combination with P in P-poor soils) across soil P levels. These findings, together with the lack of P-only limitation across our P gradient raise important questions about how Mo and P interact to constrain fixation in tropical soils.

We next examined why Mo was such a prevalent constraint on fixation in the organic soil layer across our sites. It may be reasonable to expect that molybdate and phosphate behave similarly in soils as they are both oxyanions. However, we found that available forms of Mo and P possessed dramatically different trends with soil depth. Most strikingly, the organic horizon possessed high levels of available P relative to Mo and was characterized by available P∶Mo ratios that were up to five orders of magnitude greater than those of deeper soil layers. These contrasting distributions of Mo and P with soil depth create a localized biogeochemical niche (the O horizon) in which high organic matter, high fixation rates, and high available P to Mo ratios could lead to Mo limitation on N₂ fixation.

We next examined whether differences in how Mo and P interact with the plant-soil organic matter cycle can explain whether Mo and/or P emerge as primary constraints on N₂ fixation. Mo is one of the least abundant plant essential elements [35] and is generally scarce relative to P in the organic horizon. For example the mass ratio of total digestible P∶Mo averaged ∼5000 in O horizons across our study sites, which is 6–7 times greater than the optimal P∶Mo ratio required by free-living N₂-fixing microorganisms in our laboratory cultures (680–850 for Azotobacter vinelandii; [27]). This discrepancy was even further magnified when considering available forms of P and Mo, such that P∶Mo ratios exceeded the cellular ratio of N₂-fixers by up to 200 times in the O horizon. These observations offer direct evidence that the availability of P can be abundant relative to Mo in the organic soil layer.

We considered the possibility that the high ratio of available P∶Mo in the O horizon is caused by differences in how each element interacts with organic matter. Mo is held strongly by polyphenolic complexes that can diminish bioavailability in soils [17], while most organic P can be hydrolyzed by extracellular phosphatases produced by plants and microbes [36], [37]. We found support for the hypothesis that plant organic matter preferentially reduces the availability of Mo relative to P. Not only does a large fraction of P exist in highly available forms relative to Mo, but additions of Mo and P showed preferential binding of molybdate to natural organic matter and to experimental additions of tannic acid. These results indicate that relative to phosphate, free molybdate is almost entirely consumed by soil even at concentrations much higher than ambient levels.

We speculate that what little molybdate is free in solution in the soil organic layer is likely consumed by biological uptake or lost by leaching to deeper soil horizons. In addition, some organically-complexed Mo can be accessed by free-living N₂-fixers that have evolved high-affinity systems for Mo uptake [17], [38]. The extent of such mining for Mo is poorly known and may depend on the energetic cost and effectiveness of metallophore production in soil environments. The presence of such a specialized uptake mechanism, however, indicates that Mo-poor environments have acted as a strong constraint on asymbiotic N₂-fixing organisms over evolutionary time.

We identify the existence of broad gradients in nutrient limitation on N₂ fixation across tropical forests. Our findings are consistent with the following set of interacting mechanisms: 1) N₂ fixation is favored in the organic soil horizon because the plant-soil cycle creates a vertical gradient in energy-rich and N-poor organic matter; 2) In this specialized niche Mo is complexed strongly while most organic P is readily turned over, favoring the availability of P over Mo and making possible the localized emergence of Mo over P limitation on N₂-fixers; and 3) N₂-fixers respond to broad trends in soil P availability, such that Mo constrains fixation in P-rich soils while Mo and P together constrain fixation in P-poor soils.

Our results offer a link between Mo and P stoichiometry from the scale of molecular biochemistry to that of landscape nutrient cycles. Sole Mo-limitation indicates deficiency in the FeMo co-factor of nitrogenase [39], while Mo+P co-limitation suggests deficiency in both the co-factor and the P-rich molecules (e.g., ATP, RNA and phospholipids) that support N₂ fixation and cell growth [6]. These constraints may, in turn, scale up to larger levels of observation if Mo and P interact across soils in general (as we found in lowland Panama).

These results lend support to the idea of an emerging paradigm shift in the way we view nutrient limitation in tropical forests. Our findings suggest that N₂ fixation should not be considered solely constrained by P in conceptual theories or global biogeochemical models (e.g., [40]). The idea of uniform single-element limitation is increasingly challenged as more complex relationships between nutrient cycles emerge. In these Panamanian forests, this complexity is a consequence of the presence of N₂ fixing organisms, inputs of P and Mo (from weathering and dust), and interactions of C, N, P and Mo within different soil layers. It is critical to determine if the soil-based mechanisms observed in this study are pervasive across land ecosystems where N₂ fixation is an important source of N.

Our conceptual model may also apply to N-poor ecosystems in temperate and boreal biomes that have highly leached soils (thereby low Mo availability), distinct organic horizons (thereby high Mo complexation) and where asymbiotic N₂ fixation can be the major vector of new N inputs. A previous experiment in temperate rainforests identified a direct influence of Mo on fixation [11], but there exists little to no information on how Mo and P interact across soils of extratropical as well as tropical biomes.

A critical issue is whether our results extend also to symbiotic N₂-fixers. Limitation by Mo has to our knowledge not been experimentally evaluated for symbiotic fixers in natural communities including tropical forests. It is plausible that the new concept of Mo+P co- limitation also applies to symbiotic fixers that primarily access nutrients from organic-rich soil horizons. Our findings therefore raise the question of how symbiotic N₂-fixing plants relate to variations in Mo and P distributions in soils.

Here we offer a new mechanistic model against which the diverse responses of tropical N₂ fixation to nutrient additions (including superphosphate fertilizer) can be interpreted [4], [5], [14]–[16]. We demonstrate that Mo and P act together to limit fixation across a range of soil P concentrations, from among the highest to the lowest reported for tropical forest ecosystems. In no case did we observe sole P limitation, but this constraint might emerge in soils where Mo is more abundant than it is across our gradient. Re-examining previous studies with our current hindsight is challenging, but we note that the common use of superphosphate fertilizer might confound the very interactions of P and Mo that our present study has uncovered. Since tropical soils vary greatly in distributions of macro- and micronutrients [20] it is likely that the land N cycle depends on P and Mo in ways that are more complex than we have previously considered.