Pervasive interactions between foliar microbes and soil nutrients mediate leaf production and herbivore damage in a tropical forest Eric A. Griffin1,2, S. Joseph Wright3, Peter J. Morin4 and Walter P. Carson2 1Smithsonian Environmental Research Center, 647 Contees Wharf Rd, Edgewater, MD 21037, USA; 2Department of Biological Sciences, University of Pittsburgh, A234 Langley Hall, 4249 Fifth Avenue, Pittsburgh, PA 15260, USA; 3Smithsonian Tropical Research Institute, Apartado 0843-03092, Balboa, Panama; 4Department of Ecology, Evolution, and Natural Resources, Rutgers University, 14 College Farm Rd, New Brunswick, NJ 08901, USA Author for correspondence: Eric A. Griffin Tel: +1 770 561 5731 Email: eag46@pitt.edu Received: 8 April 2017 Accepted: 21 June 2017 New Phytologist (2017) doi: 10.1111/nph.14716 Key words: leaf production, niche differentiation, plant microbiome, plant– microbe interactions, plant–microbe– herbivore interactions, soil resource availability, tropical forests. Summary  Producing and retaining leaves underlie the performance and survivorship of seedlings in deeply shaded tropical forests. These habitats are characterized by conditions ideal for foliar bacteria, which can be potent plant pathogens. Leaf production, retention and susceptibility to enemies may ultimately depend upon interactions among soil nutrients and foliar microbes, yet this has never been tested.  We experimentally evaluated the degree that foliar bacteria and soil resource supply medi- ate leaf dynamics for five common tree species (five different families) in a Panamanian forest. We reduced foliar bacteria with antibiotics for 29 months and measured leaf production, retention and damage for seedlings nested within a replicated 15-yr factorial nutrient enrich- ment experiment (nitrogen, N; phosphorus, P; potassium, K).  Our results demonstrate that when we applied antibiotics, soil nutrients – particularly N – always regulated seedling leaf production (and to a lesser extent herbivore damage) for all five tree species. In addition, it was common for two macronutrients together to negate or completely reverse the impact of applying either one alone.  Our findings of frequent plant–microbe–nutrient interactions are novel and suggest that these interactions may reinforce plant species–environment associations, thereby creating a fairly cryptic and fine-scale dimension of niche differentiation for coexisting tree species. Introduction The production, retention and defense of leaves underlie the per- formance of seedlings and saplings that live for years in deeply shaded forest understories where soil resources are scarce and ene- mies cause significant mortality. In tropical forests, the amount of damage on seedlings from both pathogens and herbivores is much higher than in temperate forests (Coley & Aide, 1991; reviewed by Coley & Barone, 1996; Gilbert, 2002; but cf. Moles, 2013). Consequently, leaf production is costly for shade-tolerant species that inhabit tropical forest understories because their seedlings invest heavily in structural and chemical defenses (Coley, 1983; Coley & Aide, 1991; Reich et al., 1992; Wright & Cannon, 2001; Wright et al., 2004; Gilbert, 2005). Moreover, tropical leaves in resource-poor habitats persist up to nine times longer and take up to three orders of magnitude longer (in days) to payback returns on carbon investments vs leaves in high- resource habitats (Chabot & Hicks, 1982; Chazdon & Fetcher, 1984; Coley, 1988; Williams et al., 1989; Kikuzawa, 1991; Sobrado, 1991; Coley & Barone, 1996; Westoby et al., 2000; Wright et al., 2004). Thus, because leaf production is costly and retention is essential, understanding the mechanisms that under- lie leaf dynamics is critical to understanding seedling perfor- mance. In addition to contending with nutrient-poor soils and deep shade, seedlings in tropical forests also occur in a habitat rich in foliar microbes (Chazdon & Fetcher, 1984; Vitousek & Sanford, 1986; Wright & Van Schaik, 1994; Wright et al., 2004; Poorter & Bongers, 2006; Poorter et al., 2009; Griffin et al., 2016; reviewed by Griffin & Carson, 2015). Because UV radiation is low, and temperature and humidity are high, foliar microbes, particularly bacteria, are likely to be diverse and abundant, and thus potentially key regulators of leaf dynamics (Griffin & Car- son, 2015; Griffin et al., 2016). Indeed, leaves are one of the world’s largest microbial habitats occupying an area twice the size of the Earth’s land area (Vorholt, 2012). On average, foliar bacte- ria occur in densities of 1–10 million cells cm2 and, moreover, an average of over 500 taxa occurred on single trees in a tropical forest in Panama (Lindow & Brandl, 2003; Delmotte et al., 2009; Vorholt, 2012; Kembel et al., 2014). Leaves of seedlings in the understory are almost certainly teaming with bacteria because of the numerous ways bacteria are able to colonize leaf surfaces  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) 1 www.newphytologist.com Research and subsequently access leaf interiors (reviewed by Griffin & Car- son, 2015). In a recent study, Griffin et al. (2016) demonstrated that foliar bacteria in a tropical forest caused up to a 49% reduc- tion in growth rates for seedlings of three common tree species. These results suggest that foliar bacteria are commonly pathogenic and may typically regulate leaf dynamics, although data are nearly nonexistent (but cf. Griffin et al., 2016). Leaf dynamics may be strongly dependent on soil nutrient availability, because the low availability of macronutrients com- monly reduces plant performance even in shaded habitats (Wright et al., 2011; Pasquini & Santiago, 2012; Santiago et al., 2012; Pasquini et al., 2015). For example, Santiago et al. (2012) demonstrated that 2 years of soil nutrient additions increased seedling height growth by up to 24% (see also Pasquini et al., 2015). Furthermore, Griffin et al. (2016) found that the degree to which foliar bacteria were harmful to seedlings varied among tree species and was often ameliorated by the greater availability of soil nutrients, particularly potassium. This suggests that the impact of foliar bacteria may typically depend on the availability of macronutrients (Griffin et al., 2016). Oddly, the degree to which soil nutrients interact with plant-associated microbes to regulate plant performance in situ, however, remains little stud- ied. This is surprising, because plant-associated microbes are criti- cal mediators of plant functional traits and trophic interactions (Friesen et al., 2011; van der Putten et al., 2013; Turner et al., 2013; Averill et al., 2014; van der Heijden et al., 2015). Foliar bacteria and soil nutrients may commonly interact to regulate leaf dynamics. These interactions could alter host plant nutrient status and plant defenses, thereby altering the rate of leaf loss or gain, and changing the amount and type of damage on small seedlings and saplings (Coley, 1983; Coley et al., 1985; Bazzaz et al., 1987; reviewed by Coley & Barone, 1996). In addi- tion, pathogenic bacteria may, like fungal pathogens, cause pre- mature leaf abscission when plants shed leaves in response to infection (Ostry, 1987; Eyal et al., 1993; Patterson, 2001; David- son et al., 2011). Thus, interactions between foliar bacteria and soil nutrient availability may commonly regulate patterns of enemy damage. For example, studies among agricultural crop species have shown that nutrient availability may mediate the negative impacts of foliar pathogens (Dordas, 2008; Johnson et al., 2010). Potassium in particular tends to mitigate the severity of pathogen damage, likely because potassium fortifies plant cell walls to confer protection from pathogen entry (Dordas, 2008). Ultimately, variation in nutrient availability may interact with bacterial communities to favor some plant species over others as resources vary patchily across the landscape (Griffin et al., 2016). If so, then interactions among foliar bacteria and soil nutrient availability could reinforce plant species–environment associa- tions, thereby creating a fairly cryptic and fine-scale dimension of niche differentiation (Griffin et al., 2016). In the present study we assess whether foliar bacteria interact with soil nutrients to govern leaf dynamics for seedlings of co- occurring tree species in the shaded understory. We tested the fol- lowing hypotheses within a tropical forest in Panama: the degree to which soil nutrients mediate leaf dynamics (production, reten- tion, and enemy damage) varies substantially among co-occurring tree species; the degree to which foliar bacteria mediate leaf dynamics varies substantially among tree species; and interactions among soil nutrients and foliar bacteria will be frequent, thereby mediating leaf dynamics among co-occurring host plant species. To address these hypotheses, we experimentally reduced foliar bacteria for 29 months for seedlings of five common tree species. These seedlings were nested within a fully factorial, well- replicated nutrient enrichment experiment (nitrogen (N), phos- phorus (P), and potassium (K) and all combinations) that com- menced in 1998 (Yavitt & Wright, 2008; Yavitt et al., 2009, 2011; Wright et al., 2011; Pasquini & Santiago, 2012; Santiago et al., 2012; Pasquini et al., 2015; Griffin et al., 2016). Materials and Methods Study site and fertilization experiment We conducted this study in a mature, seasonally moist, semi- deciduous tropical secondary forest (c. 200 years old) on the Gigante Peninsula in Panama, which is part of the Barro Colorado Nature Monument. The soils are oxisols, alfisols, inceptisols and acric nitisols (Turner et al., 2016). The site receives 2600 mm of rainfall annually, though < 10% of this falls during the four month dry season (January–April). Beginning in 1998, we applied nitrogen (N), phosphorus (P) and potassium (K) by hand four times a year between June and November in a 29 29 2 factorial design (Supporting Information Fig. S1). We replicated each treatment four times along a mild elevational and soil gradient, using 32 plots, measuring 409 40 m separated by at least 40 m (Yavitt et al., 2009; Turner et al., 2016). We applied 125 kg N ha1 yr1 as urea, 50 kg P ha1 yr1 as triple super-phosphate, and 50 kg K ha1 yr1 as KCl. Study species We selected five common and relatively shade tolerant woody species from five different families; these species vary in life his- tory traits and spanned a wide range of maximum adult heights. Alseis blackiana Hems. (Rubiaceae) is a mid-canopy tree, Desmopsis panamensis Saff. (Annonaceae) and Heisteria concinna Standl. (Olacaceae) are understory treelets, Sorocea affinis Hemsl. (Moraceae) is a small tree, and Tetragastris panamensis Kunze. (Burseraceae) is a canopy tree (Dalling et al., 2001; Wright et al., 2003, 2010; Gilbert et al., 2006; nomenclature follows Garwood, 2009). Hereafter, we refer to each species by genus or by four- letter abbreviations in figures (ALBL, Alseis; DEPA, Desmopsis; HECO, Heisteria; SOAF, Sorocea; TEPA, Tetragastris). Antibiotic applications Within the inner 309 30 m of each fertilization plot, we ran- domly assigned three c. 20–30 cm tall seedlings of each species for antibiotic treatment and another three for control treatment (sterile water; n = 941 seedlings). Beginning in January 2010, we carefully sprayed antibiotics or sterile water to all seedlings to New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist2 saturation every 10–15 d for 29 months. We placed a plastic sheet around the base of each seedling to prevent exposing soil microbes to either treatment; soil samples verified that neither the antibiotic, nor water, altered soil bacterial abundance or rich- ness (see Griffin et al., 2016). We alternated the antibiotic treatments between streptomycin (up to 100 ppm of Agri-mycin 17; Hummert International #02- 0150; Earth City, MO, USA) or a joint treatment of oxytetracy- cline and gentamicin (up to 1752 ppm of Agry-Gent Plus 800; Quımica Agronomica de Mexico, Chihuahua, Mexico). These are the three most commonly used broad-spectrum antibiotics in temperate and tropical agricultural crops and they can reduce bacterial abundance on the leaf surface as well as inside leaves by up to 85% (McManus et al., 2002; Vidaver, 2002; Traw et al., 2007; Griffin et al., 2016). Streptomycin (Agri-mycin) and gen- tamycin (Agry-gent) inhibit protein synthesis for Gram-negative bacteria, and oxytetracycline (Agry-gent) inhibits both Gram- positive and Gram-negative bacteria (Chopra & Roberts, 2001; McManus et al., 2002; Ding & He, 2010; Nelson & Levy, 2011). All products have limited nontarget effects, including those on fungi (Ingham & Coleman, 1984; Colinas et al., 1994; Chopra & Roberts, 2001; Thiele-Bruhn & Beck, 2005). Leaf production and retention We recorded the total number of leaves on each seedling at the beginning of the experiment and after 29 months of antibi- otic or control treatments. Thus, we define the rate of leaf change for each seedling after 29 months as ‘leaf production’ (see Statistical analyses). In addition, we randomly selected and marked four leaves from each seedling and estimated enemy damage (see Enemy damage) at the outset as well as enemy damage and how many of the original leaves remained after 14 months of applications (retention). We recorded leaf reten- tion and enemy damage after only 14 months because almost all leaves (c. 98%) had fallen before the end of the experiment (29 months). We excluded Alseis from the herbivore, pathogen and retention analyses at month 14 because all leaves turnover each dry season (Dalling et al., 2001). Finally, all data were collected blindly so the observer was unaware of antibiotic treatment and soil nutrient additions. Enemy damage We estimated percentage of leaf area removed by leaf-chewing herbivores and percentage damage by pathogens (chlorosis or lesions) for four randomly selected leaves from each seedling fol- lowing the protocols of Schnitzer et al. (2002) and Mangan et al. (2010). As stated above, we estimated damage at the beginning of the experiment and then 14 months after either antibiotic or control applications. We based percentage loss estimates on a template of artificial (paper) leaves with 24 levels of damage: 0%, 1%, 2.5%, 5%, 7.5%, 10% and in 5% increments up to 100% area removed (Carson & Root, 2000; Schnitzer et al., 2002). Although some insect damage may cause lesions and chlorosis (e.g. Miller & Davidson, 2005), the primary causes of this type of damage are fungi, bacteria and viruses (e.g. Garcia-Guzman & Dirzo, 2001; Myster, 2002; Mangan et al., 2010; E. Griffin, pers. obs.). Hereafter, we will refer to percentage area loss by leaf- chewing insects as ‘herbivore damage’ and chlorosis and lesions as ‘pathogen damage’. It is important to note that we also measured the mean canopy openness (0–100%) above each seedling with a concave densitometer at beast height at each time point (0 months, 14 months, 29 months). Canopy openness had no effect on any of the models. Statistical analyses We performed MANOVAs to evaluate whether nutrient addi- tions caused changes in plant performance metrics. We chose to use MANOVAs because all five species were nested within the N, P and K treatments, and MANOVAs allowed us to avoid pseudo- factorialism by adjusting for correlated response variables among species (Morrison, 1976; Morin, 1983; Winer et al., 1991; Hurl- bert, 2013). We assessed the significance of nutrient additions using Wilks’ Criterion, one of four standard test statistics com- monly used to evaluate the MANOVA (Morrison, 1976), defined by (determinant (E))(determinant (E+ H))1, where H is the matrix of sums-of-squares and cross products calculated among treatment means (e.g. +N and N), and E is the ele- ment-wise squared difference between each observation and the mean vectors for that group (Morin, 1983). We used the method of linear discriminant functions, some- times called canonical analysis of discriminance (CAD), to iden- tify which species contributed to significant differences among our treatments for each response (Fisher, 1936; Legendre & Legendre, 1998). CAD follows directly from the calculations used to determine test statistics in the multivariate analysis of variance (MANOVA). Species that are significantly correlated with the values of the discriminant function scores are those that contribute to significant differences among the treatments exam- ined in the MANOVA. Although single-species ANOVAs often yield similar results to those detected by CAD, the ANOVAs can miss correlated responses among the five variables (species responses; see Methods S1), and their significance levels are not adjusted for multiple tests on potentially correlated variables (Morin, 1983). For a full explanation of the discriminant func- tion analysis, see Methods S1. MANOVA models for control individuals We performed a first set of MANOVAs to evaluate seedling leaf production rate (L), percentage leaf area attacked by herbivores and pathogens (H and P), and leaf retention (R). We did this to test differences among nutrient addition treatments (N, P, K and all combinations) for control seedlings only. We calculated leaf production rates (leaves/leaf month1) for each seedling as: L ¼ ðloge L1  loge L0Þ=ðt1  t0Þ (L0 and L1, initial and final leaf number; t1 t0, time period (29 months; Santiago et al., 2012)).  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 3 We estimated herbivore and pathogen damage among all seedlings before treatments began (Time = 0 months). Estimating damage at the beginning of the experiment allowed us to evaluate how over 15 years of soil nutrient additions impacted leaf damage among all seedlings. We calculated leaf retention R by determin- ing the proportion of four randomly selected leaves from each seedling that remained after 14 months. The average for each species in a plot (e.g. L for leaf produc- tion) for species i was LðiÞ ¼ Lði; controlÞ Because all five species were nested within treatment plots (and nonindependent), our response vectors for each metric (e.g. L) in plot j was: Lj ¼ ðLð1Þ; Lð2Þ; Lð3Þ; Lð4Þ; Lð5ÞÞj where numbers 1 through 5 represent each plant species. Alseis was not included in the retention analysis because their leaves had flushed within two months of the beginning of the experiment. MANOVA models for differences among antibiotic and control individuals We performed a second set of MANOVAs to evalu- ate the differences in leaf production rates (dL) after 29 months of applications and herbivore damage (dH), pathogen damage (dP) and leaf retentions (dR) after 14 months of applications. We calculated the difference in plot mean values (e.g. dL, d H , dP and dR) for replicated individuals of each species with or without antibiotic applications in each plot. Thus, for leaf production, dLðiÞ ¼ Lði; antibioticÞ  Lði; controlÞ For herbivore or pathogen damage rates, we only included sur- viving leaves after 14 months for species (i) and evaluated as: H ðiÞ ¼ ðH14; i H0; iÞ=ð1H0; iÞ or PðiÞ ¼ ðP14; i  P0; iÞ=ð1 P0; iÞ Therefore, our response vector for each performance metric (e.g. dLj ) in plot j was: dLj ¼ ðdLð1Þ; dLð2Þ; dLð3Þ; dLð4Þ; dLð5ÞÞj where the numbers 1 through 5 refer to the five plant species. A MANOVA of this response vector tests whether dLj differed across nutrient treatments. Alseis was not included in the herbi- vore, pathogen, or retention MANOVA analyses because its leaves flushed two months after the experiment began. We logit transformed all proportional data (Warton & Hui, 2011). We used SAS 9.4 (SAS Institute, Cary, NC, USA) to run MANOVAs. Results Antibiotic efficacy We previously demonstrated that Agry-gent and Agri-mycin sig- nificantly decreased mean abundance of epiphytic and Table 1 MANOVA results for the effects of nitrogen (N), phosphorus (P) and potassium (K) on plant performance metrics for control seedlings treated with sterile water of five species (Alseis blackiana, Desmopsis panamensis, Heisteria concinna, Sorocea affinis and Tetragastris panamensis) Effect Leaf production Leaf retention Herbivore damage Pathogen damage N 1.64 8.71*** 1.79 0.87 P 1.47 1.80 4.31** 1.63 K 6.90*** 2.28 2.67 1.22 N9 P 2.36 5.71** 3.69* 0.62 N9 K 3.38* 2.58 1.54 1.03 P9 K 2.62 2.80 1.10 0.48 N9 P9 K 1.76 2.26 1.80 0.17 We did not include Alseis in the leaf retention analysis because their leaves turnover annually. Entries are F-values determined fromWilks’ Criterion. Degrees of freedom are 5,20 for the leaf production, herbivore damage and pathogen damage analyses, and 4,20 for leaf retention. Detection of a significant treatment effect indicates that the mean values of the leaf met- ric differ for the corresponding nutrient addition. Significance: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. Table 2 Pearson correlation coefficient values (r) for correlations between the discriminant function scores of significant nutrient effects compared to the original variables (species) in MANOVAs for control seedlings treated with sterile water Metric Nutrient Alseis Desmopsis Heisteria Sorocea Tetragastris Production K 0.84*** 0.61*** 0.052*** 0.28 0.08 Production N9 K 0.73*** 0.62*** 0.50*** 0.23 0.33 Retention N NA 0.39* 0.34 0.77*** 0.55*** Retention N9 P NA 0.48*** 0.48*** 0.84*** 0.15 Herbivory P 0.02 0.12 0.32 0.56*** 0.56*** Herbivory N9 P 0.17 0.18 0.26 0.81*** 0.44* Significant r correlations indicate a nutrient9 species interaction for the corresponding leaf metric and corresponding nutrient addition (N, nitrogen; P, phosphorus; K, potassium). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist4 endophytic bacteria by over 50% on our focal species and also decreased microbial morphotype richness by over 20% (Griffin et al., 2016). Each of these antibiotic treatments were equally effective and remained so over time; also, their impact was very similar among nutrient treatments and plant species (Griffin et al., 2016). Thus, our antibiotics worked in the sense that they reduced foliar bacterial loads, although our results are likely con- servative because we only reduced bacterial abundance and rich- ness by c. 50%. At the beginning of the experiment, mean leaf numbers were 7.53 for Alseis (n = 181), 10.59 for Desmopsis (n = 188), 5.81 for Heisteria (n = 191), 5.32 for Sorocea (n = 187) and 6.70 for Tetragastris (n = 193). Below, if we do not report the impact of the antibiotic or specific nutrient or nutrient combination, it is because it did not have a significant impact on leaf dynamics. It is important to note that c. 10% seedlings died during the entirety of the experiment, and mortality did not differ among tree species or nutrient or antibiotic applications. The impacts of macronutrients on leaf dynamics separate from the effects of antibiotics Leaf production For three species, N and K interacted to cause increases or decreases in leaf production by 55–88% (Tables 1, 2; Fig. 1). K addition caused Alseis to produce far fewer leaves, but adding K and N together significantly reduced the negative effects of K alone (Tables 1, 2; Fig. 1b). For two other species, K addition alone caused the production of significantly more leaves 0.04 * * * 0.03 0.02 0.01 0.00 –0.01 ALBL DEPA HECO SOAF TEPA –N +N –N +N–N +N Species Alseis Desmopsis Heisteria Treatment Treatment 0.04 –K +K –K +K –K +K–K +K 0.03 0.02 0.01 0.00 –0.01 0.04 0.03 0.02 0.01 0.00 –0.01 0.04 0.03 0.02 0.01 Le af p ro du ct io n ra te (le av es p er le af m on th ) Le af p ro du ct io n ra te (le av es p er le af m on th ) 0.00 –0.01 (a) (b) (c) (d) Fig. 1 Significant effects of potassium (K) and nitrogen (N)9 K on leaf production rates among control seedlings of Alseis blackiana (ALBL), Desmopsis panamensis (DEPA), Heisteria concinna (HECO), Sorocea affinis (SOAF) and Tetragastris panamensis (TEPA) after 29months of nutrient enrichment. (a) A significant K9 species interaction effects on leaf production for Alseis, Desmopsis and Heisteria (K effect: F5,20 = 6.90, P = 0.001; Alseis: P < 0.0001; Desmopsis: P = 0.0002; Heisteria: P = 0.002). (b–d) N9 K interaction effects on leaf production (F5,20 = 3.38, P = 0.023; (b) Alseis: P < 0.0001; (c) Desmopsis: P = 0.0002; (d) Heisteria: P = 0.004). Bars represent mean values  SE.  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 5 (Fig. 1a), as did adding N alone, but adding both K and N together did not cause an additive increase in leaf production (significant N9K interaction, Fig. 1c,d). Leaf retention We found a very consistent and significant effect of adding N alone, or adding N and P together on leaf retention. N addition alone decreased leaf retention for all four species (sig- nificant for three or four) but this decrease disappeared or was even reversed (Desmopsis) when N and P were added together (N9 P interactions, Tables 1, 2; Fig. 2a–d; Alseis was not included). These findings demonstrate that a strong effect of adding one soil resource could be entirely offset by the simultane- ous addition of another. Herbivore and pathogen damage For two species, significant N9 P interactions mediated herbivore damage (Tables 1, 2; Fig. 3). For Sorocea, adding P reduced damage, however, adding N and P together increased damage (Tables 1, 2; Fig. 3b). Con- versely, for Tetragastris, adding P more than doubled herbivore damage but this effect disappeared when we added N and P together (Tables 1, 2; Fig. 3c). Nutrient additions never signifi- cantly altered pathogen damage. Overall, mean herbivore damage across all tree species was 8.5% ( 0.35 SE) and pathogen dam- age was 6.1% ( 0.26 SE; Fig. 4). The impact of antibiotics and macronutrients on leaf dynamics Leaf production The regular application of antibiotics and soil nutrients over 29 months caused increases or decreases in leaf production for all five species ranging from 38% to 140% (Tables 3, 4; Figs 5, 6). However, the effects of antibiotics were 1.0 0.8 0.6 0.4 0.2 0.0 DEPA * * * Le af re te nt io n ra te (r at e of re m ai ni ng le av es a fte r 1 4 m on th s) Le af re te nt io n ra te (r at e of re m ai ni ng le av es a fte r 1 4 m on th s) HECO SOAF TEPA Species Treatment Desmopsis Heisteria Sorocea –N +N –N +N –P +P –P +P –P +P Treatment –N +N Treatment –N +N 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 (a) (b) (c) (d) Fig. 2 Significant effects of nitrogen (N) and N9 phosphorus (P) on leaf retention rates among control seedlings of Desmopsis panamensis (DEPA), Heisteria concinna (HECO), Sorocea affinis (SOAF) and Tetragastris panamensis (TEPA) after 14months of nutrient enrichment. (a) A significant N9 species interaction effect on leaf retention for Desmopsis, Sorocea and Tetragastris (N effect: F4,20 = 8.71, P = 0.0003; Desmopsis: P = 0.027; Sorocea: P < 0.0001; Tetragastris: P = 0.001). (b–d) N9 P interaction effects on leaf retention (F4,20 = 5.71, P = 0.003; (b) Desmopsis: P = 0.005; (c) Heisteria: P = 0.005; (d) Sorocea: P = 0.001). Bars represent mean values  SE. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist6 always governed by N9 K interactions, and to a lesser extent N9 P interactions (Table 3). For example, for Desmopsis, antibi- otics increased leaf production in plots fertilized with P but this was reversed in plots fertilized with both P and N (Table 4; Fig. 5b). For Sorocea, antibiotics increased leaf production but only in plots fertilized with both N and P (Fig. 5c). For four of five species, the impact of antibiotics was strongly governed by N9K interactions (Table 4; Fig. 6). For three of five species, the response was consistent: adding antibiotics and K together increased leaf production; however, antibiotics had no effect when K and N were added together (Figs 6a,c,d, 3a). For a fourth species, Heisteria, applying antibiotics and K together decreased leaf production unless N and K were applied together 14 12 10 8 6 4 2 0 ALBL M ea n le af h er bi vo re d am ag e (% le af re m ov ed ) M ea n le af h er bi vo re d am ag e (% le af re m ov ed ) M ea n le af h er bi vo re d am ag e (% le af re m ov ed ) DEPA HECO Species –N +N –P +P –P +P –P +P * * –N +N Treatment Sorocea Tetragastris SOAF TEPA 5 4 3 2 1 0 5 4 3 2 1 0 (a) (b) (c) Fig. 3 Significant effects of phosphorus (P) and nitrogen (N)9 P on leaf herbivore damage rates among control seedlings of Alseis blackiana (ALBL), Desmopsis panamensis (DEPA), Heisteria concinna (HECO), Sorocea affinis (SOAF) and Tetragastris panamensis (TEPA) at the beginning of the experiment (Time = 0months). (a) A significant P9 species interaction effect on herbivore damage for Sorocea and Tetragastris (P effect: F5,20 = 4.31, P = 0.01; Sorocea: *P = 0.001; Tetragastris: *P = 0.001). (b, c) N9 P interaction effects on herbivore damage (F5,20 = 3.69, P = 0.016; (b) Sorocea: P < 0.0001; (c) Tetragastris: P = 0.013). Bars represent mean values  SE. 25 20 15 10 5 0 ALBL DEPA HECO SOAF TEPA Species Herbivore damage Pathogen damage M ea n le af d am ag e (% le af re m ov ed o r d is ea se d) Fig. 4 Mean insect herbivore and pathogen damage (chlorosis or lesions) among leaves of Alseis blackiana (ALBL), Desmopsis panamensis (DEPA), Heisteria concinna (HECO), Sorocea affinis (SOAF) and Tetragastris panamensis (TEPA) at the beginning of the experiment. We estimated percentage leaf area removed by leaf-chewing herbivores and percentage damage by pathogens for four randomly selected leaves from each seedling before we began applying antibiotics (n = 3572 leaves). Bars represent mean values  SE. Table 3 MANOVA results for the effects of nitrogen (N), phosphorus (P) and potassium (K) on plant performance metric differences between antibiotic and control seedlings of five species (Alseis blackiana, Desmopsis panamensis, Heisteria concinna, Sorocea affinis and Tetragastris panamensis) Effect Leaf production Leaf retention Herbivore damage Pathogen damage N 1.97 1.76 1.51 2.48 P 1.49 0.90 0.19 1.13 K 3.52* 0.25 0.20 0.97 N9 P 3.39* 0.37 0.93 2.12 N9 K 4.45** 2.41 3.19* 1.20 P9 K 0.56 1.13 1.94 0.92 N9 P9 K 1.16 1.18 2.58 1.05 Detection of a significant treatment effect indicates that the mean values of the leaf metric differ among antibiotic and control individuals for the corresponding nutrient addition. We did not include Alseis in the leaf retention analysis because their leaves turnover annually. Entries are F- values determined fromWilks’ Criterion. Degrees of freedom are 5,20 for leaf production, and 4,20 for the leaf retention, herbivore and pathogen damage analyses. Significance: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 7 (Fig. 6b). Overall, the effects of the antibiotics were often reversed when two macronutrients were added together. Leaf retention We never detected any impact of antibiotics on leaf retention under any combination of our nutrient treatments (Table 3). Note that in the second set of analyses we evaluated the difference in means between the antibiotic and control indi- viduals. This simply means that there were no differences in how antibiotic individuals responded to nutrient enrichment vs con- trol individuals, not that there were no nutrient effects on reten- tion (e.g. Tables 1, 2; Fig. 2a–d). Herbivore damage For three species, the impact of antibiotics was consistent and regulated entirely by N9K interactions (Tables 3, 4; Fig. 7). Specifically, antibiotics decreased damage when neither N nor K was applied, or when both were applied together; conversely, antibiotics increased damage when just N was applied (Table 4; Fig. 7). This was most pronounced for Heisteria and Sorocea (Fig. 7b,c). These results mirror the strong N9K interactions that regulated leaf production (Fig. 6). To our surprise, we never detected any impact of the antibiotic on pathogen damage under any combination of nutrient treatments (Table 3). Discussion Our results demonstrate that soil nutrients and broad-spectrum antibiotics regulate leaf dynamics among five species in a deeply shaded forest understory. Specifically, potassium (K) was a key resource in regulating leaf production (K and N9 K interac- tions), nitrogen (N) was key for leaf retention (N and N9 P interactions), and phosphorus (P) was key for regulating herbi- vore damage (P and N9 P interactions; Tables 1, 2; Figs 1–3). Notably, antibiotics caused increases in leaf production for four of five species (Fig. S2). More importantly, the impact of broad-spectrum antibiotics and soil nutrients interact to regulate leaf dynamics and, to a lesser extent, herbivore damage. This was true for all five of our focal tree species, which came from five different plant families, and vary in adult stature and other life history traits. The direction and magnitude of the impact of antibiotics were entirely dependent upon interactions with soil nutrients for all five species (Table 4). N was a key resource involved in each of these interactions for both leaf production and herbivore damage (i.e. significant N9 K and N9 P inter- actions). For example, it was striking that applying antibiotics caused a dramatic increase (up to 140%) in leaf production in plots enriched with K, but this increase was entirely reversed if N was applied together with K (e.g. Alseis (Figs 5a, 6a) showed highly significant N9 K interactions). These types of reversals occurred for two other species (N9 K interactions for leaf pro- duction; Fig. 5c,d) and also for three species for the effects of antibiotics on herbivore damage (N9 K interactions; Fig. 7a– c). Our results provide compelling evidence that the impact of foliar microbial communities depends upon soil resource avail- ability and, in particular on two macronutrients, N and K. Fur- thermore, our results strongly suggest that the application of antibiotics in combination with a single macronutrient (typi- cally K) decreases the abundance of pathogenic bacteria, often increasing leaf production. However, the benefits of reducing bacteria, or the indirect effects of reducing bacteria (see Potential mechanisms underlying plant–microbe–nutrient interactions), were often negated when N and another macronu- trient were applied together. Overall, our findings support recent studies that challenge the conventional wisdom that light is only a limiting resource for tropical seedlings and moreover demonstrate that soil nutrient availability and foliar bacteria are major drivers of plant performance (e.g. Pasquini & Santiago, 2012; Santiago et al., 2012; Pasquini et al., 2015; Griffin et al., 2016). Our results support all three hypotheses and demonstrate that antibiotics alter foliar bacteria to such a degree that it can substan- tially mediate leaf production and herbivore damage. Moreover, the magnitude of the impacts of nutrients on leaf production was host specific, and N and K availability regulated these responses. Thus, we suggest that without considering microbial communi- ties, it is difficult to understand the mechanisms that regulate seedling performance and dynamics among co-occurring species and thus key aspects of forest regeneration (e.g. Mangan et al., 2010). Moreover, these interactions appear strong enough to fre- quently change the rank-order performance of seedlings of differ- ent species across forest understories because soil resources and most likely bacteria as well, are patchily distributed at both large and small spatial scales (e.g. John et al., 2007; Baldeck et al., 2013; Condit et al., 2013; Kembel et al., 2014). For example, for our five species, antibiotic applications changed the rank-order performance for two of five species for leaf-production and Table 4 Pearson correlation coefficient (r) for correlations between the discriminant function scores of significant nutrient effects compared to the original variables (species) in MANOVAs of antibiotic and control differences Metric Nutrient Alseis Desmopsis Heisteria Sorocea Tetragastris Production K 0.93*** 0.01 0.32 0.31 0.36* Production N9 P 0.15 0.86*** 0.04 0.46* 0.02 Production N9 K 0.47*** 0.26 0.52* 0.57** 0.78*** Herbivory N9 K NA 0.50** 0.84* 0.59*** 0.25 Significant r correlations indicate nutrient9 species9 antibiotic interactions for the corresponding leaf metric and nutrient addition (N, nitrogen; P, phos- phorus; K, potassium). Alseiswas not included in the herbivore, pathogen or retention analyses because its leaves turnover annually. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist8 herbivore damage, and four of five species for seedling growth (Fig. S2; growth data from Griffin et al., 2016). Notably, how- ever, Griffin et al. (2016) only detailed the interplay between soil nutrients and foliar bacteria and their impacts on seedling growth among tree species. Here, we went one step further to measure the impacts of these interactions on leaf production and retention, two key underlying mechanisms for seedling growth. Addition- ally, we demonstrate that interactions between antibiotics and soil nutrients impacted plant growth and leaf production differently under different experimental conditions and sometimes in oppos- ing directions. For example, Griffin et al. (2016) demonstrated that antibiotics mediated seedling growth when P, K and N9 P were added; however, here we demonstrated that antibiotics mediated leaf production when K, N9 P and N9K were added. In another example, Griffin et al. (2016) demonstrated that antibiotics increased Desmopsis growth rates when N and P were added; however, in this study antibiotics decreased Desmopsis leaf production rates when N and P were added. Thus, species-specific impacts of foliar bacteria vary among nutrient additions for differ- ent performance metrics, even if those metrics correlate with one another to a certain degree (e.g. growth and leaf production). For these reasons, we argue that foliar bacteria should be considered an entirely independent yet cryptic plant functional trait that reg- ulates plant phenotypes and performance metrics among species. Ultimately, we hypothesize that these interactions may well repre- sent a novel dimension of fine-scale niche differentiation, thus promoting the long-term coexistence of plant species (Griffin et al., 2016). Although speculative, we argue that combinations of two nutrients together (particularly N) make leaves more vulnerable to enemies, thus negating or reversing the effects of enrichment on performance. These results are consistent with findings by both Andersen et al. (2010) and Santiago et al. (2012), who sug- gested that enhanced herbivory caused by nutrient enrichment could mask growth responses to fertilization. This conclusion, however, runs counter to our result where the effects of nutrients, the application of antibiotics, or their interaction never signifi- cantly altered pathogen damage, and had significant but variable effects on herbivore damage for three species. It is possible that we did not pick up many significant herbivore or pathogen effects because the majority of leaves we selected were fully expanded and potentially already accrued most of their damage. For example, almost 70% of lifetime enemy damage to tropical leaves occur within the first few weeks of leaf expansion (Coley & Aide, 1989; Kursar & Coley, 2003; reviewed by Coley & Barone, 1996). Moreover, it is important to point out that plants infected with foliar pathogens may suffer reduced performance yet often show no visible symptoms (reviewed by Griffin & Car- son, 2015). For example, Pseudomonas tomato, a plant pathogen, can decrease tomato leaf production by as much as 30% without showing any sign of infection (Bashan & Okon, 1981). Indeed, we found that antibiotics or their interactions with nutrients caused no differences in foliar pathogen damage yet caused increases in leaf production by up to 140%. Thus, our results lead us to suggest that high nutrient tissue concentrations, partic- ularly for N and K together and N and P together, may com- monly make host plants more vulnerable to plant enemies. Indeed, some authors have argued that increased vulnerability to enemies may select against luxury consumption under low light and nutrients, particularly for N (Ostertag, 2010; Sayer & Banin, 2016). Nevertheless, our study demonstrates there were pervasive interactions between antibiotic and macronutrient treatments, 0.03 * * 0.02 0.01 0.00 –0.01 –0.02 0.03 0.02 0.01 0.00 –0.01 –0.02 0.03 –N +N –N +N Treatment –P +P –P +P –K +K Sorocea Desmopsis ALBL DEPA HECO Species L (a nt ib io tic ) – L (c on tro l) L (a nt ib io tic ) – L (c on tro l) L (a nt ib io tic ) – L (c on tro l) SOAF TEPA 0.02 0.01 0.00 –0.01 –0.02 (a) (b) (c) Fig. 5 Significant effects of antibiotic applications and potassium (K) and nitrogen (N)9 phosphorus (P) on leaf production rates (L) among seedlings of Alseis blackiana (ALBL), Desmopsis panamensis (DEPA), Heisteria concinna (HECO), Sorocea affinis (SOAF) and Tetragastris panamensis (TEPA) after 29months of applications and nutrient enrichment. When bars are above the line, antibiotic applications increased leaf production and when below the line, antibiotic applications decreased production. (a) For Alseis and Tetragastris, applying antibiotics increased leaf production when K was added (K effect: F5,20 = 3.52, P = 0.0192; Alseis: *P < 0.0001; Tetragastris: P = 0.0427). (b, c) For Desmopsis and Sorocea, antibiotics and N9 P regulated leaf production (significant antibiotics9N9 P interaction, F5,20 = 3.35, P = 0.0233; Desmopsis: P < 0.0001; Sorocea: P = 0.0141). Bars represent mean values  SE.  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 9 particularly N treatments, on leaf production for all five tree species. Ultimately, foliar bacteria appear to be a highly cryptic component of plant performance, and future studies should con- sider the impacts of pathogens even when there are no visible symptoms on plant hosts. Potential mechanisms underlying plant–microbe–nutrient interactions Although our antibiotic treatments reduced bacterial abundance and richness (Griffin et al., 2016), it remains unclear whether the impacts of antibiotics are due to bacterial reduction, changes in the species composition of bacterial communities, indirect effects (e.g. changes in fungal pathogens due to reductions in bacteria), or a combination of these. Indeed, whereas many species of bacteria are plant pathogens, others are mutualists, and thus like some fungal endophytes, produce compounds such as growth hormones that increase plant resistance to pathogens and herbivores (see Arnold et al., 2003; Herre et al., 2007; Mejia et al., 2008, 2014 for leaf fungal endophytes; mutualist bacteria reviewed by Griffin & Carson, 2015). Griffin et al. (2016) recently demonstrated that antibiotics in general increased tropical seedling growth and here, antibiotics increased leaf production for four of five species (Fig. S2). Moreover, we found that in four cases, antibiotics, in combina- tion with the addition of a single macronutrient, increased leaf production (Figs 5, 6). Thus, our findings suggest that bacteria typically function as plant pathogens. In one case, however, antibiotics in combination with a single macronutrient decreased performance (Figs 5, 6). Because the mechanisms are 0.04 Alseis L L (a nt ib io tic ) – L (c on tro l) L (a nt ib io tic ) – L (c on tro l) Heisteria Sorocea Tetragastris 0.03 0.02 0.01 0.00 –0.01 –0.02 0.04 0.03 0.02 0.01 0.00 –0.01 –0.02 0.04 –K +K –N +N –N +N Treatment Treatment –K +K –K +K–K +K 0.03 0.02 0.01 0.00 –0.01 –0.02 0.04 0.03 0.02 0.01 0.00 –0.01 –0.02 (a) (b) (c) (d) Fig. 6 Significant effects of antibiotic applications and nitrogen (N)9 potassium (K) on leaf production rates (L) among seedlings of Alseis blackiana, Heisteria concinna, Sorocea affinis and Tetragastris panamensis after 29months of applications and nutrient enrichment. When bars are above the line, antibiotic applications increased leaf production and when below the line, antibiotic applications decreased production. All panels illustrate significant effects of antibiotics and N9 K on leaf production rates (significant antibiotics9N9 K interaction; F5,20 = 4.45, P = 0.0069). (a, c, d) For Alseis, Sorocea and Tetragastris, applying antibiotics increased leaf production when K was added; however, antibiotics had no effect on leaf production when K and N were added together (Alseis: P = 0.0064; Sorocea: P = 0.0006; Tetragastris: P < 0.0001). (b) For Heisteria, applying antibiotics decreased leaf production when K was added, however, antibiotics had no effect on leaf production when K and N were added together (P = 0.0022). Bars represent mean values  SE. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist10 still unclear, future experiments are needed using genomic tech- niques to critically evaluate how antibiotics alter bacterial communities to allow for a deeper understanding of how particu- lar bacterial taxa drive leaf dynamics. Moreover, it is possible that antibiotic applications cause direct effects (e.g. toxic) on nontarget organisms (e.g. fungi); however, empirical studies have looked for this and found little evidence (Ingham & Cole- man, 1984; Colinas et al., 1994; Chopra & Roberts, 2001; Thiele-Bruhn & Beck, 2005). Nevertheless, more reductionist approaches are imperative to link bacterial communities to plant physiology and plant performance. We suggest that future studies address the mechanisms and pathways by which foliar bacteria and other enemies impact plant hosts in situ. On the one hand, enemy attacks often trigger plants to produce proteins, phenolics and alkaloids for protection from both pathogens and insects; thus, bacterial infection may induce plant defenses against insect enemies (Tierens et al., 2001; Thomma et al., 2002; Wittstock & Gershenzon, 2002; Haq et al., 2004; Kniskern et al., 2007). On the other hand, pathogen infec- tion may upregulate the salicylic acid pathway. If so, this directly inhibits the jasmonic acid pathway, which is a key pathway that enhances plant defense against herbivores (e.g. Traw et al., 2003; reviewed by Stout et al., 2006; Bari & Jones, 2009; Pieterse et al., 2009; Thaler et al., 2012). Thus, upregulation of microbial defenses in response to attack may leave the host plants more vulnerable to herbivore attack (Glazebrook, 2005; Koornneef & Pieterse, 2008). Studies addressing the mechanisms underlying enemy attack, however, have almost exclusively focused on Arabidopsis thaliana or agricultural crop species. Thus, future studies should address the interactions among microbes, herbivores and plant hosts, in species- rich ecosystems where enemy pressure is substantial. Implications for plant diversity maintenance and niche differentiation Studies have demonstrated that increased soil nutrient availability decreases realized niche space of co-occurring species, yet it remains uncertain whether soil nutrients alone can maintain hyper-diversity of plant species in tropical forests (e.g. Hubbell et al., 1999; Hubbell, 2001; reviewed by Wright, 2002; Silver- town, 2004; Kitajima & Poorter, 2008). To date, the mechanism underlying this niche dimension is often assumed to be direct resource competition, where different plant species are better or worse competitors along resource gradients (Tilman, 1982). Thus, spatial variation in soil nutrient availability favors different species in different places. Here, we demonstrate that foliar bacte- ria are critical mediators of interactions between plants and soil resource availability, supporting other studies that demonstrate the importance of plant–microbial interactions in diversity main- tenance (e.g. Mangan et al., 2010; Schnitzer et al., 2011; Pender- gast et al., 2013). In fact, some have proposed that plant- associated microbes can act as stabilizing factors to increase differ- ences in species’ performance outcomes along resource gradients or among interactions with other trophic levels (e.g. Chesson, 2000; reviewed by Bever et al., 2010; Mordecai, 2011). In this framework, such stabilizing processes cause intraspecific effects to be more negative than interspecific differences (Chesson, 2000). Thus, when any single species increases in abundance, its per cap- ita growth rate slows relative to growth rates of other species, which aids in species coexistence (Chesson, 2000). Indeed, we found that foliar bacteria caused co-occurring plant species to 0.06 Desmopsis Heisteria Sorocea –N +N Treatment –K +K –K +K –K +K0.04 0.02 0.00 –0.02 –0.04 –0.06 0.06 0.04 0.02 0.00 –0.02 –0.04 –0.06 0.06 0.04 0.02 0.00 H (a nt ib io tic ) – H (c on tro l) H (a nt ib io tic ) – H (c on tro l) H (a nt ib io tic ) – H (c on tro l) –0.02 –0.04 –0.06 (a) (b) (c) Fig. 7 Significant effects of antibiotic applications and nitrogen (N)9 potassium (K) on herbivore damage rates (H) among seedlings of Desmopsis panamensis, Heisteria concinna, and Sorocea affinis after 14months of applications and nutrient enrichment. When bars are above the line, antibiotic applications increased herbivore damage and when below the line, antibiotic applications decreased herbivore damage. (a–c) For Desmopsis, Heisteria and Sorocea, applying antibiotics decreased damage when neither N nor K was applied, or when both were applied together; conversely, antibiotics increased damage when just N was applied (significant antibiotics9N9 K interaction; F4,20 = 5.05, P = 0.0353; Desmopsis: P = 0.0041; Heisteria: P < 0.0001; Sorocea: P = 0.0005). Bars represent mean values  SE.  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 11 perform in some cases better or in other cases worse among dif- ferent soil resource treatments. This suggests that bacteria interact with a familiar and long-standing key niche axis (soil nutrient availability and its degree of patchiness). Although speculative, we argue that plant–bacterial interactions more finely divide niche differences among coexisting plant species and thus func- tion to promote plant diversity. Acknowledgements We thank Jonathan Pruitt, Brian Traw, Sarah Pasquini, Michelle Spicer, Nick Keiser, Nathan Brouwer and two anonymous review- ers for helpful comments and discussions that greatly improved the quality of the manuscript. We thank Omar Hernandez, Rufino Gonzalez and Severino Fernandez for plant identification and help with antibiotic applications in the field. We acknowledge financial support from a National Science Foundation Graduate Research Fellowship, a Smithsonian Tropical Research Institute Predoctoral Fellowship, a Lewis and Clark Fund for Exploration and Research, two Sigma Xi Grants-in-Aid of Research, and the University of Pittsburgh, all of which supported E.A.G.’s research. A Central Research Development Fund from the University of Pittsburgh provided funds for W.P.C. and E.A.G. Author contributions E.A.G., S.J.W. and W.P.C. planned and designed the experi- ment; E.A.G. performed the experiment and conducted field- work; P.J.M. analyzed the data; and E.A.G., S.J.W. and W.P.C. wrote the manuscript. References Andersen KM, Corre MD, Turner BL, Dalling JW. 2010. Plant–soil associations in a lower montane tropical forest: physiological acclimation and herbivore-mediated responses to nitrogen addition. Functional Ecology 24: 1171–1180. Arnold AE, Mejia LC, Kyllo D, Rojas EI, Maynard Z, Robbins N, Herre EA. 2003. Fungal endophytes limit pathogen damage in a tropical tree. Proceedings of the National Academy of Sciences, USA 100: 15649–15654. Averill C, Turner BL, Finzi AC. 2014.Mycorrhiza-mediated competition between plants and decomposers drives soil carbon storage. Nature 505: 543– 545. Baldeck CA, Harms KE, Yavitt JB, John R, Turner BL, Valencia R, Vavarrete H, Davies SJ, Chuyong GB, Kenfack D et al. 2013. Soil resources and topography shape local tree community structure in tropical forests. Proceedings of the Royal Society B 280: 20122532. Bari R, Jones JDG. 2009. Role of plant hormones in plant defence responses. Plant Molecular Biology 69: 473–488. Bashan Y, Okon Y. 1981. Inhibition of seed germination and development of tomato plants in soil infested with Pseudomonas tomato. Annals of Applied Biology 98: 413–417. Bazzaz FA, Chiariello NR, Coley PD, Pitelka LF. 1987. Allocating resources to reproduction and defense. BioScience 37: 58–67. Bever JD, Dickie IA, Facelli E, Facelli JM, Klironomos J, Moora M, Rillig MC, Stock WD, Tibbett M, Zobel M. 2010. Rooting theories of plant community ecology in microbial interactions. Trends in Ecology & Evolution 25: 468–478. Carson WP, Root RB. 2000.Herbivory and plant species coexistence: community regulation by an outbreaking phytophagous insect. Ecological Monographs 70: 73–99. Chabot BF, Hicks DJ. 1982. The ecology of leaf life spans. Annual Review of Ecology and Systematics 13: 229–259. Chazdon RL, Fetcher N. 1984. Light environments of tropical forests. In: Medina E, Mooney HA, Vazquez-Yanes C, eds. Physiological ecology of plants of the wet tropics. The Hague, the Netherlands: Dr W. Junk, 553– 564. Chesson P. 2000.Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics 31: 343–366. Chopra I, Roberts M. 2001. Tetracycline antibiotics: mode of action, applications, molecular biology, and epidemiology of bacterial resistance. Microbiology and Molecular Biology Reviews 65: 232–260. Coley PD. 1983.Herbivory and defensive characteristics of tree species in a lowland tropical forest. Ecological Monographs 53: 209–234. Coley PD. 1988. Effects of plant growth rate and leaf lifetime on the amount and type of anti-herbivore defense. Oecologia 74: 531–536. Coley PD, Aide TM. 1989. Red coloration of tropical young leaves: a possible antifungal defence? Journal of Tropical Ecology 5: 293–300. Coley PD, Aide TM. 1991. A comparison of herbivory and plant defenses in temperate and tropical broad-leaved forests. In: Price PW, Lewinsohn TM, Fernandes GW, Benson WW, eds. Plant–animal interactions: evolutionary ecology in tropical and temperate regions. New York, NY, USA: John Wiley & Sons, 25–49. Coley PD, Barone JA. 1996.Herbivory and plant defenses in tropical forests. Annual Review of Ecology and Systematics 27: 305–335. Coley PD, Bryant JP, Chapin FS III. 1985. Resource availability and plant anti- herbivore defense. Science 230: 895–899. Colinas C, Ingham E, Molina R. 1994. Population responses of target and non- target forest soil organisms to selected biocides. Soil Biology and Biochemistry 26: 41–47. Condit R, Engelbrecht BM, Pino D, Perez R, Turner BL. 2013. Species distributions in response to individual soil nutrients and seasonal drought across a community of tropical trees. Proceedings of the National Academy of Sciences, USA 110: 5064–5068. Dalling JW, Winter K, Nason JD, Hubbell SP, Murawski DA, Hamrick JT. 2001. The unusual life history of Alseis blackiana: a shade-persistent pioneer tree? Ecology 82: 933–945. Davidson JM, Patterson HA, Wickland AC, Fichtner EJ, Rizzo DM. 2011. Forest type influences transmission of Phytophthora ramorum in California oak woodlands. Phytopathology 101: 492–501. Delmotte N, Knief C, Chaffron S, Innerebner G, Roschitzki B, Schlapbach R, von Mering C, Vorholt JA. 2009. Community proteogenomics reveals insights into the physiology of phyllosphere bacteria. Proceedings of the National Academy of Sciences, USA 106: 16428–16433. Ding C, He J. 2010. Effect of antibiotics in the environment on microbial populations. Applied Microbiology and Biotechnology 87: 925–941. Dordas C. 2008. Role of nutrients in controlling plant diseases in sustainable agriculture. Agronomy for Sustainable Development 28: 33–46. Eyal Y, Meller Y, Lev-Yadun S, Fluhr R. 1993. A basic-type PR-1 promoter directs ethylene responsiveness, vascular and abscission zone-specific expression. Plant Journal 4: 225–234. Fisher RA. 1936. The use of multiple measurements in taxonomic problems. Annals of Eugenics 7: 179–188. Friesen ML, Porter SS, Stark SC, von Wettberg EJ, Sachs JL, Martinez-Romero E. 2011.Microbially mediated plant functional traits. Annual Review of Ecology, Evolution, and Systematics 42: 23–46. Garcia-Guzman G, Dirzo R. 2001. Patterns of leaf-pathogen infection in the understory of a Mexican rain forest: incidence, spatiotemporal variation, and mechanisms of infection. American Journal of Botany 88: 634–645. Garwood NC. 2009. Seedlings of Barro Colorado Island and the neotropics. Ithaca, NY, USA: Cornell University Press. Gilbert B, Wright SJ, Muller-Landau HC, Kitajima K, Hernandez A. 2006. Life history trade-offs in tropical trees and lianas. Ecology 87: 1281–1288. Gilbert GS. 2002. Evolutionary ecology of plant diseases in natural ecosystems. Annual Review of Phytopathology 40: 13–43. Gilbert GS. 2005. Biotic interactions in the tropics. Cambridge, UK: Cambridge University Press. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist12 Glazebrook J. 2005. Contrasting mechanisms of defense against biotrophic and necrotrophic pathogens. Annual Review of Phytopathology 43: 205–227. Griffin EA, Carson WP. 2015. The ecology and natural history of foliar bacteria with a focus on tropical forests and agroecosystems. The Botanical Review 81: 105–149. Griffin EA, Traw MB, Morin PJ, Pruitt JN, Wright SJ, Carson WP. 2016. Foliar bacteria and soil fertility mediate seedling performance: a new and cryptic dimension of niche differentiation. Ecology 97: 2998–3008. Haq SK, Atif SM, Khan RH. 2004. Protein proteinase inhibitor genes in combat against insects, pests, and pathogens: natural and engineered phytoprotection. Archives of Biochemistry and Biophysics 431: 145–159. van der Heijden MGA, Martin FA, Selosse MA, Sanders IR. 2015.Mycorrhizal ecology and evolution: the past, the present, and the future. New Phytologist 205: 1406–1423. Herre EA, Mejia LC, Hyllo DA, Rojas E, Maynard Z, Butler A, Van Bael SA. 2007. Ecological implications of anti-pathogen effects of tropical fungal endophytes and mycorrhizae. Ecology 88: 550–558. Hubbell SP. 2001. The unified neutral theory of biodiversity and biogeography. Princeton, NJ, USA: Princeton University Press. Hubbell SP, Foster RB, O’Brien ST, Harms KE, Condit R, Wechsler B, Wright SJ, De Lao SL. 1999. Light-gap disturbance, recruitment limitation, and tree diversity in a neotropical forest. Science 283: 554–557. Hurlbert SH. 2013. Pseudofactorialism, response structures and collective responsibility. Austral Ecology 38: 646–663. Ingham ER, Coleman DC. 1984. Effects of streptomycin, cycloheximide, Fungizone, captan, and PCNB on soil microorganisms.Microbial Ecology 10: 345–358. John R, Dalling JW, Harms KE, Yavitt JB, Stallard RF, Mirabello M, Hubbell SP, Valencia R, Navarrete H, Vallejo M et al. 2007. Soil nutrients influence spatial distributions of tropical tree species. Proceedings of the National Academy of Sciences, USA 104: 864–869. Johnson PTJ, Townsend AR, Cleveland CC, Gilbert PM, Howarth RW, McKenzie VJ, Rejmankova E, Ward MH. 2010. Linking environmental nutrient enrichment and disease emergence in humans and wildlife. Ecological Applications 20: 16–29. Kembel SW, O’Connor TK, Arnold HA, Hubbell SP, Wright SJ, Green JL. 2014. Relationships between phyllosphere bacterial communities and plant functional traits in a neotropical forest. Proceedings of the National Academy of Sciences, USA 111: 13715–13720. Kikuzawa K. 1991. A cost-benefit analysis of leaf habit and leaf longevity of trees and their geographical pattern. The American Naturalist 138: 1250–1263. Kitajima K, Poorter L. 2008. Functional basis for resource niche partitioning by tropical trees. In: Carson WP, Schnitzer SA, eds. Tropical forest community ecology. Oxford, UK: Wiley-Blackwell, 160–181. Kniskern JM, Traw MB, Bergelson J. 2007. Salicylic acid and jasmonic acid signaling defense pathways reduce natural bacterial diversity on Arabidopsis thaliana.Molecular Plant-Microbe Interactions 20: 1512–1522. Koornneef A, Pieterse CMJ. 2008. Cross talk in defense signaling. Plant Physiology 146: 839–844. Kursar TA, Coley PD. 2003. Convergence in defense syndromes of young leaves in tropical rainforests. Biochemical Systematics and Ecology 31: 929–949. Legendre P, Legendre L. 1998. Numerical ecology, 2nd edn. Amsterdam, the Netherlands: Elsevier. Lindow SE, Brandl MT. 2003.Microbiology of the phyllosphere. Applied and Environmental Microbiology 69: 1875–1883. Mangan SA, Schnitzer SA, Herre EA, Mack KML, Valencia MC, Valencia EI, Bever JD. 2010. Negative plant-soil feedback predicts tree-species relative abundance in a tropical forest. Nature 466: 752–755. McManus PS, Stockwell VO, Sundin GW, Jones AL. 2002. Antibiotic use in plant agriculture. Annual Review of Phytopathology 40: 443–465. Mejia LC, Herre EA, Sparks JP, Winter K, Garcia MN, Van Bael SA, Stitt J, Shi Z, Zhang Y, Guiltinan MJ et al. 2014. Pervasive effects of a dominant foliar endophytic fungus on host genetic and phenotypic expression in a tropical tree. Frontiers in Microbiology 5: 479. Mejia LC, Rojas EI, Maynard Z, Van Bael S, Arnold AE, Hebbar P, Samuels GJ, Robbins N, Herre EA. 2008. Endophytic fungi as biocontrol agents of Theobroma cacao pathogens. Biological Control 46: 4–14. Miller DR, Davidson JA. 2005. Armored scale insect pests of trees and shrubs (Hemiptera: Diaspididae). Ithaca, NY, USA: Cornell University Press. Moles AT. 2013. Dogmatic is problematic: interpreting evidence for latitudinal gradients in herbivory and defense. Ideas in Ecology and Evolution 6: 1–4. Mordecai E. 2011. Pathogen impacts on plant communities: unifying theory, concepts, and empirical work. Ecological Monographs 81: 429–441. Morin PJ. 1983. Predation, competition, and the composition of larval anuran guilds. Ecological Monographs 53: 119–138. Morrison DF. 1976.Multivariate statistical methods. New York, NY, USA: McGraw-Hill. Myster RW. 2002. Foliar pathogen and insect herbivore effects on two landslide tree species in Puerto Rico. Forest Ecology and Management 169: 231–242. Nelson ML, Levy SB. 2011. The history of the tetracyclines. Annals of the New York Academy of Sciences 1241: 17–32. Ostertag R. 2010. Foliar nitrogen and phosphorus accumulation responses after fertilization: an example from nutrient-limited Hawaiian forests. Plant and Soil 334: 85–98. Ostry M. 1987. Biology of Septoria musiva andMarssonina brunnea in hybrid Populus plantations and control of Septoria canker in nurseries. European Journal of Forest Pathology 17: 158–165. Pasquini SC, Santiago LS. 2012. Nutrients limit photosynthesis in seedlings of a Lowland tropical forest tree species. Oecologia 168: 311–319. Pasquini SC, Wright SJ, Santiago LS. 2015. Lianas always outperform tree seedlings regardless of soil nutrients: results from a long-term fertilization experiment. Ecology 96: 1866–1876. Patterson SE. 2001. Cutting loose. Abscission and dehiscence in Arabidopsis. Plant Physiology 126: 494–500. Pendergast TH, Burke DJ, Carson WP. 2013. Belowground biotic complexity drives aboveground dynamics: a test of the soil community feedback model. New Phytologist 197: 1300–1310. Pieterse CMJ, Leon-Reyes A, Van der Ent S, Van Wees CM. 2009. Networking by small-molecule hormones in plant immunity. Nature Chemical Biology 5: 308–316. Poorter H, Niinemets U, Poorter L, Wright IJ, Villar R. 2009. Causes and consequences of variation in leaf mass per area (LMA): a meta-analysis. New Phytologist 182: 565–588. Poorter L, Bongers F. 2006. Leaf traits are good predictors of plant performance across 53 rain forest species. Ecology 87: 1733–1743. van der Putten WH, Bardgett RD, Bever JD, Bezemer TM, Casper BB, Fukami T, Kardol P, Klironomos JN, Kulmatiski A, Schweitzer JA et al. 2013. Plant– soil feedbacks: the past, the present and future challenges. Journal of Ecology 101: 265–276. Reich PB, Walters MB, Ellsworth DS. 1992. Leaf life-span in relation to leaf, plant, and stand characteristics among diverse ecosystems. Ecological Monographs 62: 365–392. Santiago LS, Wright SJ, Harms KE, Yavitt JB, Korine C, Garcia MN, Turner BL. 2012. Tropical tree seedling growth responses to nitrogen, phosphorus, and potassium addition. Journal of Ecology 100: 309–316. Sayer EJ, Banin LF. 2016. Tree nutrient status and nutrient cycling in tropical forest – lessons from fertilization experiments. In: Goldstein G, Santiago LS, eds. Tropical tree physiology. Basel, Switzerland: Springer International, 275– 297. Schnitzer SA, Klironomos JN, HilleRisLambers J, Kinkel LL, Reich PB, Xiao X, Rillig MC, Sikes BA, Callaway RM, Mangan SA et al. 2011. Soil microbes drive the class plant diversity–productivity pattern. Ecology 92: 296–303. Schnitzer SA, Reich PB, Bergner B, Carson WP. 2002.Herbivore and pathogen damage on grass and woodland plants: a test of the herbivore uncertainty principle. Ecology Letters 5: 531–539. Silvertown J. 2004. Plant coexistence and the niche. TRENDS in Ecology and Evolution 19: 605–611. Sobrado MA. 1991. Cost–benefit relationships in deciduous and evergreen leaves of tropical dry forest species. Functional Ecology 5: 608–616. Stout MJ, Thaler JS, Thomma BPHJ. 2006. Plant-mediated interactions between pathogenic microorganisms and herbivorous arthropods. Annual Review of Entomology 51: 663–689.  2017 The Authors New Phytologist 2017 New Phytologist Trust New Phytologist (2017) www.newphytologist.com New Phytologist Research 13 Thaler JS, Humphrey PT, Whiteman NK. 2012. Evolution of jasmonic and salicylate signal crosstalk. Trends in Plant Science 17: 260–270. Thiele-Bruhn S, Beck I-C. 2005. Effects of sulfonamide and tetracycline antibiotics on soil microbial activity and microbial biomass. Chemosphere 59: 457–465. Thomma BP, Cammune BP, Thevissen K. 2002. Plant defenses. Planta 216: 193–202. Tierens KFJ, Thomma BP, Brouwer M, Schmidt J, Kistner K, Porzel A, Mauch-Mani B, Cammue BP, Broekaert WF. 2001. Study of the role of antimicrobial glucosinolate-derived isothiocyanates in resistance of Arabidopsis to microbial pathogens. Plant Physiology 125: 1688–1699. Tilman D. 1982. Resource competition and community structure. Princeton, NJ, USA: Princeton University Press. Traw MB, Kim J, Enright S, Cipollini DF, Bergelson J. 2003. Negative cross- talk between salicylate-and jasmonate-mediated pathways in the Wassilewskija ecotype of Arabidopsis thaliana.Molecular Ecology 12: 1125–1135. Traw MB, Kniskern JM, Bergelson J. 2007. SAR increases fitness of Arabidopsis thaliana in the presence of natural bacterial pathogens. Evolution 61: 2444– 2449. Turner BL, Yavitt JB, Harms KE, Garcia MN, Wright SJ. 2016. Seasonal changes in soil organic matter after a decade of nutrient addition in a lowland tropical forest. Biogeochemistry 123: 221–235. Turner TR, James EK, Poole PS. 2013. The plant microbiome. Genome Biology 14: 209. Vidaver AK. 2002. Uses of antimicrobials in plant agriculture. Clinical Infectious Diseases 34: S107–S110. Vitousek PM, Sanford RL. 1986. Nutrient cycling in moist tropical forest. Annual Review of Ecology and Systematics 17: 137–167. Vorholt JA. 2012.Microbial life on the phyllosphere. Nature Reviews Microbiology 10: 828–840. Warton DI, Hui FKC. 2011. The arcsine is asinine: the analysis of proportions in ecology. Ecology 92: 3–10. Westoby M, Warton D, Reich PB. 2000. The time value of leaf area. The American Naturalist 155: 649–656. Williams K, Field CB, Mooney HA. 1989. Relationships among leaf construction cost, leaf longevity, and light environment in rain-forest plants of the genus Piper. The American Naturalist 133: 198–211. Winer BJ, Brown DR, Michels KM. 1991. Statistical principles in experimental design, 3rd edn. New York, NY, USA: McGraw-Hill. Wittstock U, Gershenzon J. 2002. Constitutive plant toxins and their role in defense against herbivores and pathogens. Current Opinion in Plant Biology 5: 300–307. Wright IJ, Cannon K. 2001. Relationships between leaf lifespan and structural defenses in a low-nutrient, sclerophyll flora. Functional Ecology 15: 351–359. Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z, Bongers F, Cavender-Bares J, Chapin T, Cornelissen JH, Diemer M et al. 2004. The worldwide leaf economics spectrum. Nature 428: 821–827. Wright SJ. 2002. Plant diversity in tropical forests: a review of mechanisms of species coexistence. Oecologia 130: 1–14. Wright SJ, Kitajima K, Kraft NJ, Reich PB, Wright IJ, Bunker DE, Condit R, Dalling JW, Davies SJ, Diaz S et al. 2010. Functional traits and the growth- mortality trade-off in tropical trees. Ecology 91: 3664–3674. Wright SJ, Muller-Landau HC, Condit R, Hubbell SP. 2003. Gap-dependent recruitment, realized vital rates, and size distributions of tropical trees. Ecology 84: 3174–3185. Wright SJ, Van Schaik CP. 1994. Light and the phenology of tropical trees. The American Naturalist 143: 192–199. Wright SJ, Yavitt JB, Wurzburger N, Turner BL, Tanner EVJ, Sayer EJ, Santiago LS, Kaspari M, Hedin LO, Harms KE et al. 2011. Potassium, phosphorus, or nitrogen limit root allocation, tree growth, and litter production in a lowland tropical forest. Ecology 92: 1616–1625. Yavitt JB, Harms KE, Garcia MN, Mirabello MJ, Wright SJ. 2011. Soil fertility and fine root dynamics in response to 4 years of nutrient (N, P, K) fertilization in a lowland tropical moist forest, Panama. Austral Ecology 36: 433–445. Yavitt JB, Harms KE, Garcia MN, Wright SJ, He F, Mirabello MJ. 2009. Spatial heterogeneity of soil chemical properties in a lowland tropical moist forest, Panama. Australian Journal of Soil Research 47: 674–687. Yavitt JB, Wright SJ. 2008. Seedling growth responses to water and nutrient augmentation in the understorey of a lowland moist forest, Panama. Journal of Tropical Ecology 24: 19–26. Supporting Information Additional Supporting Information may be found online in the Supporting Information tab for this article: Fig. S1 Map of study area showing the placement of nutrient treatments within the study site. Fig. S2 The impacts of antibiotic applications on rank-order per- formance among plant performance metrics (leaf production, growth rate and herbivore damage). Methods S1 Full description of linear discriminant function analysis. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. New Phytologist (2017)  2017 The Authors New Phytologist 2017 New Phytologist Trustwww.newphytologist.com Research New Phytologist14