SEARCH

SEARCH BY CITATION

Keywords:

  • ant–plant interactions;
  • biotic;
  • chemical and physical resistance traits;
  • defence trade-offs;
  • inducible defences;
  • myrmecophytism;
  • plant defence theory;
  • plant–herbivore interactions

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  1. Ants provide variable protection against herbivores to ant-plants (i.e. myrmecophytes and myrmecophiles). The ways in which ant-plants dynamically adjust both their direct (chemical and physical) and indirect (biotic) defences in response to varying levels of herbivory are not well understood.
  2. We experimentally generated a broad range of ant-attendance levels and herbivory pressures in a tropical myrmecophyte, Cordia nodosa, which allowed exploration of the inducibility of and interactions between direct and indirect resistance traits.
  3. In response to increased herbivory, host plants encouraged indirect (biotic) defence by increasing domatium volume, regardless of whether ants were present on the plant. When ants were present, larger domatia housed more workers, which in turn decreased herbivory on adjacent leaves.
  4. Independent of the presence of ants, plants responded to increased herbivory by inducing both chemical (phenolics) and structural (leaf toughness, trichomes) resistance traits; these traits were associated with reduced palatability to a folivorous beetle.
  5. Synthesis. Our results show that both direct and indirect defences are inducible in C. nodosa, which suggests that C. nodosa may retain direct defences as insurance against varying levels of protection from its ant bodyguards. Thus, the predictions of optimal defence theory are not violated: although C. nodosa invests in multiple forms of defence, they are not redundant.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Plants have evolved a wide variety of direct and indirect defences against herbivory. Direct defences include physical or chemical resistance traits, amongst others, whereas indirect defences depend on traits that attract the natural enemies of herbivores. Optimal defence theory predicts that plants will allocate resources efficiently to defence mechanisms, for example, by investing little in redundant defences (Stamp 2003). Thus, we might expect that plants that are well defended by secondary compounds should not also invest heavily in physical structures like thorns or trichomes, or that plants that engage aggressive ants to deter herbivores should not also make expensive toxins.

However, the search for trade-offs among defences has not produced clear patterns (e.g. Koricheva, Nykänen & Gianoli 2004); in fact, Agrawal (1998) recently described efforts to identify such trade-offs as ‘utter failures.’ To further illustrate this point, consider the literature on trade-offs between direct and indirect defences in ant-plants. Ant-plants make food (i.e. extrafloral nectar or food bodies) or housing (i.e. domatia) to attract ants that protect plants against herbivores. Here, we refer to plants that make domatia as myrmecophytes and plants that make only extrafloral nectar or food bodies as myrmecophiles. Since Janzen (1966) speculated that, ‘the obligate acacia-ants may be regarded as replacing in some sense the physical and chemical protective properties of other acacias’, numerous studies have sought evidence of trade-offs between ant and chemical defences, with mixed results (Rehr, Feeny & Janzen 1973; Seigler & Ebinger 1987; Heil, Fiala & Linsenmair 1999; Heil, Staehelin & McKey 2000; Dyer et al. 2001; Eck et al. 2001; Heil et al. 2002; Koricheva & Romero 2012). Although an early study found reduced chemical defences in a species of ant-acacia compared with some of its non-myrmecophytic relatives (Rehr, Feeny & Janzen 1973), more thorough sampling of acacia taxa subsequently revealed no consistent relationship (Seigler & Ebinger 1987; Heil et al. 2002). Collectively, this body of research has shown that ant-plants commonly possess both direct and indirect defences (e.g. Dyer et al. 2001; Heil et al. 2002; Del Val & Dirzo 2003; but see Koricheva & Romero 2012), despite their potential redundancy and presumed or demonstrated costs (e.g. Stanton & Palmer 2011; Frederickson et al. 2012).

Why do plants frequently make several types of defences? The emerging view is that ‘multiple plant resistance traits in the same species are not likely redundant or wasteful’ (Agrawal 1998). Having multiple resistance traits, even if each one is costly, may be adaptive if diverse defences (i) protect plants against diverse enemies, (ii) act synergistically or (iii) represent a bet-hedging strategy (Koricheva, Nykänen & Gianoli 2004; Agrawal & Fishbein 2006; Rasmann & Agrawal 2009; Agrawal 1998). Thus, instead of trading-off, direct and indirect resistance traits may interact to provide better overall plant defence.

Indirect resistance, by its very nature, may necessitate having a ‘back-up,’ because the natural enemies of herbivores are not always there when plants need them (unlike constitutive chemical or physical resistance traits). Both myrmecophytic and myrmecophilous plants must cope with high spatial and temporal variation in the availability and willingness of ants to engage in plant defence. The number of ants on a myrmecophilous plant often varies with the time of day, season, microhabitat or geographical region (Horvitz & Schemske 1990; Bronstein 1998; de la Fuente & Marquis 1999; Rudgers & Strauss 2004; Ness, Morris & Bronstein 2006; Brenes-Arguedas, Coley & Kursar 2008; Rudgers, Savage & Rúa 2010). Similarly, although ants are usually more consistently present on myrmecophytes than myrmecophiles, most myrmecophytes nonetheless have stages when they lack ants. Most myrmecophytes are not defended by ants as juveniles (Coley 1986; Del Val & Dirzo 2003; Heil & Mckey 2003; Boege & Marquis 2005); they cannot house ants until they make domatia, which may require a minimum size or age, and even once they make domatia, they may not be located and colonized immediately by ants (Heil & Mckey 2003; Frederickson & Gordon 2009). Furthermore, myrmecophytes often outlive their ant colonies (Frederickson & Gordon 2009; Palmer et al. 2010). For example, Cordia nodosa trees live an average of 77 years, but the ant colonies that reside in their domatia live only an average of 8–14 years, depending on the ant species (Frederickson & Gordon 2009). As a result, these plants sometimes persist without ants for months or years in between losing one ant colony and acquiring another. Finally, even when an ant colony is present on a myrmecophyte, it may not provide effective defence against herbivores. Some species of ant bodyguards consistently provide better anti-herbivore protection than others (e.g. Bronstein 1998; Palmer et al. 2010) and larger colonies can provide better protection than smaller colonies (e.g. Duarte Rocha & Godoy Bergallo 1992). Thus, both myrmecophilous and myrmecophytic plants may benefit from having other resistance traits when indirect defence fails.

If some resistance mechanisms function as ‘safety nets,’ it may be adaptive to deploy them only when needed; in other words, back-up defences should be inducible rather than constitutively expressed. Very broadly, natural selection should favour inducible over constitutive resistance mechanisms when (i) making the defence is costly, (ii) herbivore pressure is temporally variable, such that there are times when the defence is not needed, and (iii) herbivore pressure is nonetheless predictable, at least in the short-term, in the sense that damage at one time point provides a reliable indicator of the risk of damage at some later time (Karban 2011). Variation in the effectiveness of biotic defenders may frequently give rise to the second condition, so plants with indirect defences might often experience selection for inducible direct defences.

While there is abundant evidence that both direct and indirect resistance mechanisms can be inducible, research on induced resistance has tended to emphasize direct defences in temperate plants (e.g. Karban & Myers 1989; Nabeshima, Murakami & Hiura 2001; Van Zandt 2007; Karban 2011), but indirect defences in tropical plants (e.g. Agrawal 1998; Agrawal & Rutter 1998; Heil et al. 2001, 2004; Palmer et al. 2008; Bixenmann, Coley & Kursar 2011). Comparatively few studies have examined both at once (but see e.g. Agrawal, Karban & Colfer 2003; Yamawo et al. 2012), especially in tropical species. Here, we explored the inducibility of and interactions among direct and indirect resistance traits in the Amazonian myrmecophyte C. nodosa (Boraginaceae). For nearly a year, we manipulated the presence of Allomerus octoarticulatus ants and most insect herbivores on C. nodosa in a full-factorial experiment. We then measured levels of chemical, physical and biotic resistance traits expressed by the C. nodosa plants in each treatment. We used these data to address the following questions:

  1. How inducible are indirect and direct defences in C. nodosa?
  2. Does C. nodosa have a back-up system that functions to defend tissues when ants are absent or when ants fail to protect their host?
  3. Do plants respond differently to herbivore damage in the presence of ants? For example, does C. nodosa induce indirect defences when ants are present and direct defences when ants are absent?

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Study System and Site

This study was conducted at the Los Amigos Research Center (12°34′ S, 70°05′ W; elevation c. 270 m) in the Department of Madre de Dios, Peru. In this region, C. nodosa associates with several species of ants. The most common is A. octoarticulatus (Myrmicinae), which occupies 40–80% of C. nodosa trees (Yu & Pierce 1998; Frederickson 2009). Other C. nodosa are occupied by Azteca spp. (Dolichoderinae, 10–35% of trees), Myrmelachista schumanni (Formicinae, < 2% of trees), other twig-nesting ant species (very rarely), or else not occupied by ants (10–20% of primarily very young trees; Yu & Pierce 1998; Frederickson 2009). Cordia nodosa produces domatia – hollow stem swellings on otherwise slender branches – whether or not ants are present. When a C. nodosa tree grows a new shoot, it typically produces one domatium together with a whorl of new leaves. If the tree has ants, the colony quickly fills this new domatium with brood and workers. Cordia nodosa trees produce food bodies as a reward for ants on the surfaces of young leaves and shoots (Solano, Belin-Depoux & Dejean 2005). Ant colonies also get additional food from the honeydew-producing scale insects (Hemiptera: Sternorrhyncha: Coccoidea) that live inside domatia.

Previous research has shown that A. octoarticulatus defends C. nodosa against folivores; the presence of A. octoarticulatus ants significantly reduces folivory in ant-exclusion experiments (Yu & Pierce 1998; Frederickson 2005; Frederickson et al. 2012). Some important folivores of C. nodosa are adult and larval leaf beetles (i.e. Chrysomelidae) and orthopterans. Allomerus octoarticulatus workers sterilize C. nodosa inflorescences, reducing C. nodosa fruit production by up to 100%, which benefits ants because sterilized plants produce more domatia and thus house larger, more fecund ant colonies (Yu & Pierce 1998; Dejean et al. 2004; Frederickson 2005, 2009; Szilagyi et al. 2009). In this study, C. nodosa plants were harvested before they reached reproductive maturity, so the sterilization behaviour of A. octoarticulatus was not included as a factor in our experimental design.

Experimental Exclusion or Addition of Ants and Herbivores

More detailed methods can be found in Frederickson et al. (2012). Briefly, we grew 52 C. nodosa from seeds in outdoor cages until they had at least one domatium. We divided the plants into 13 blocks of four plants based on their sizes, and within each block we assigned each plant to one of four treatments at random. The treatments were (i) ants excluded, herbivores present (A−H+) (ii) ants added, herbivores present (A+H+) (iii) ants excluded, herbivores excluded (A−H−) and (iv) ants added, herbivores excluded (A+H−). We planted the saplings in the rain forest understorey such that each sapling in a block formed the corner of a 2 × 2 m square, and blocks were separated by 3–30 m along a trail. We measured the initial height of each sapling, counted its leaves and domatia, and marked each domatium individually with a thin plastic-coated coloured wire.

We added queenright A. octoarticulatus colonies that we collected from small, naturally occurring C. nodosa trees to the saplings in the A+ treatment. To prevent ants from colonizing plants assigned to the A− treatment, we injected 0.2 mL of a dilute, non-systemic, pyrethroid insecticide (Cypermethrin, 0.2 mg mL−1) into each domatium using a syringe at the beginning of the experiment (August 2009) and again in November 2009 and February 2010. All A− plants received insecticide inside domatia only – this treatment did not deter insects from feeding on the associated leaves (Frederickson et al. 2012). To maintain similar light environments among treatments, all plants were planted inside mesh nets. The nets covering plants in the H− treatment were staked securely to the ground. The nets covering plants in the H+ treatment were rolled up so their bottoms hung c. 30 cm off the ground, a height that allowed access by most insect herbivores, but kept leaves similarly shaded underneath the nets as in the H− treatment. The light environment did not differ among treatments (Frederickson et al. 2012). In February 2010, all 52 nets were replaced with larger ones and all new domatia that had been produced since the start of the experiment were marked with additional coloured wires. Plants spent a total of 314–329 days in the experiment.

Damage by Leaf-Chewing Herbivores and Plant Growth

In 2010, we harvested the plants, collecting one block of four plants each day. We first measured the height of each plant, and then cut off, counted and labelled all the leaves, aspirating any ants on the leaves into a plastic vial. We also cut off and counted all the domatia, placing them into individually labelled envelopes that along with the plant stems were sealed into resealable plastic bags to prevent ants from escaping.

In the field, one of us (GB) visually estimated on each leaf the amount of leaf area missing using a score of 0–20. We also digitally photographed each leaf against a white background. On a random subsample of 25 leaves per treatment, we measured percent folivory in the digital images using ImageJ (W.S. Rasband, US National Institutes of Health, Bethesda, MD, USA) and used the least-squares regression equation to convert the visual scores into percentage folivory values (Frederickson et al. 2012).

Because we marked domatia individually at the beginning and in the middle of the experiment and C. nodosa grows six leaves together with each new domatium, we could divide both the leaves and the domatia from each plant into four age-classes (i) leaves and domatia that were already present when we began the experiment (> 11 months old at harvest) (ii) leaves and domatia produced between August 2009 and February 2010 (5–11-month old at harvest) (iii) leaves and domatia produced between February 2010 and June/July 2010 that were fully expanded when we harvested them (< 5 months old at harvest) and (iv) same as age-class iii, but not yet fully expanded (very young leaves and domatia).

Indirect Resistance

To kill the ants, we froze the bags of domatia and stems, as well as the vials of aspirated ants. We then cut open each domatium and counted the worker ants inside. We measured the length and maximum diameter of each domatium using callipers and then calculated the volume of each domatium using the equation for a cone (Edwards et al. 2006).

Direct Resistance

For each plant, we measured leaf toughness and trichome density on three randomly chosen leaves per age-class; note that not all plants had leaves in all four age-classes. We measured leaf toughness using a Mark-10 digital force gauge (CSC Force Measurement, Inc., Agawam, MA, USA), adapting the method of Kursar & Coley (2003). We placed a leaf segment between two Plexiglass plates both drilled with 4-mm diameter holes and then measured the peak force necessary to push a 3-mm diameter probe through the leaf tissue. To measure trichome density, we counted using a microscope the number of trichomes in a 2.25-cm2 area centred at the midpoint of the abaxial surface of each leaf.

We dried leaves in a drying oven for 48–72 h before transporting them to the University of Toronto, where we used the Folin-Ciocalteu method to determine the concentration of total phenolics (Kursar & Coley 2003). For each plant and each age-class for which we had sufficient leaf material, we ground one or two dried leaves using a homogenizer to obtain 0.25 g of tissue. We mixed this with 20 mL of 90% methanol (Caledon) and left the mixture to sit for 20–24 h at room temperature. We then dried the samples in a vacuum concentrator for 2 h at 65 °C. We re-suspended the tissue in 50% methanol/water (v/v), and extracted it with 10 mL of hexane (Sigma-Aldrich, St Louis, MO, USA) for 45 min. We added 60 μL of methanol/water extract to 440 μL of distilled water, followed by 1 mL of Folin–Ciocalteu's phenol reagent (Sigma-Aldrich) and, after 3 min, 1 mL of 0.5 m Na2CO3. We incubated the samples for 60 min at room temperature before reading absorbency at 725 nm on a spectrophotometer. We used tannic acid as a standard to calibrate the absorbency readings. To determine whether heat-drying could have affected our results, we also collected leaves from eight C. nodosa plants not in the experiment, cut them into half, and then either heat-dried or used silica gel to dry the leaf halves; these gave nearly identical total phenolics concentrations (means ± 1 SE: for heat-dried leaves, 0.500 ± 0.079 mg tannic acid equivalents mL−1, for silica gel-dried leaves, 0.505 ± 0.086 mg tannic acid equivalents mL−1; t8 = 0.04, = 0.967).

In the field, we also tested the palatability of the C. nodosa leaves from this experiment to an herbivore in a bioassay. We used adults of a species of beetle, identified as Zepherina defensa (Bechyné) (Chrysomelidae: Galerucinae), that we collected from C. nodosa plants at our field site (Fig. 1). This is the first record of this genus and species for Peru (C. Chaboo, pers. comm.). For each plant, we chose a leaf at random from among those in age-class iii and cut a 4-cm2 square from the leaf's undamaged tissue; one plant had no leaves in this age-class, so we used a leaf from age-class ii instead. We placed four leaf squares, one from each plant in a block, in a lidded container lined with moist paper towels and added five Z. defensa beetles. We began the bioassay the same day we harvested the block of plants and we allowed the beetles to feed for at least 24 h. We then removed and photographed the remnants of the leaf squares and measured percentage herbivory in the digital images using ImageJ.

image

Figure 1. An adult Zepherina defensa beetle on a Cordia nodosa leaf; this species was used in the bioassays. Photo credit: G.A. Miller.

Download figure to PowerPoint

Statistical Analysis

In three blocks, one or both of the ant colonies died before the end of the experiment (two colonies in the H+ treatment and two in the H− treatment) and in a fourth block, one of the plants lost all its leaves, apparently due to a fungal pathogen. We excluded these plants from our data analysis. To improve normality, we square-root transformed all count data (i.e. numbers of ants, domatia, leaves, and trichomes), as well as the folivory and domatia volume data, and log-transformed the height, leaf toughness, and total phenolics data. When we measured several correlated variables, we used principal component analysis to reduce the data; we then used the principal components as response variables in mixed-effect anova or ancova models. Unless otherwise stated, we analysed the response variables (generally direct or indirect resistance trait measurements) with block as a random factor and folivory and/or domatium age-class as fixed factors. When multiple measurements were taken on a single plant, we used either the mean value for the whole plant (folivory, domatium volume) or the mean value for all the leaves in a single domatium age-class for the whole plant (trichome density, leaf toughness). Analyses were carried out in JMP® 9.0.0 (Cary, NC, USA).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Damage by Leaf-Chewing Herbivores and Plant Growth

Elsewhere (Frederickson et al. 2012), we show that ants significantly reduced folivory in this experiment, but they were not as effective as the herbivore exclusion treatment. Briefly, in the H+ treatments, plants with ants received about half as much damage to leaves as plants without ants (means ± 1 SE: 5.05 ± 1.34% for A+H+ plants, compared to 9.88 ± 1.49% for A−H+ plants); herbivore-excluded plants received even less damage (A+H−: 0.84 ± 0.17%, A−H−: 0.84 ± 0.18%). Thus, both our herbivore exclusion and ant exclusion treatments created substantial variation in folivory (Fig. 2).

image

Figure 2. Relationship between final plant size and folivory. Each point is a Cordia nodosa plant. Final plant size was measured as the first principal component of height, number of domatia and number of leaves at the end of the experiment. Folivory was measured as the mean percent of leaf area missing at the end of the experiment for all leaves on each plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between final plant size and folivory from the statistical model (see text). For reference, a plant with a PC1 value of −1 measured 73 cm in height and had 3.5 domatia and 28 leaves at the end of the experiment, while a plant with a PC1 value of +1 measured 125 cm in height and had 23 domatia and 141 leaves.

Download figure to PowerPoint

Plants that experienced more folivory grew less (Fig. 2). We combined our three highly correlated measures of plant growth – namely, height, number of domatia and number of leaves – using principal component analysis. PC1 explained 93% of the variation in final plant size and was the only principal component with an eigenvalue > 1, so we used this principal component in the subsequent ancova model, with PC1 as the response variable, folivory as the covariate, and block as a random factor. There was a strong, negative relationship between folivory and final plant size (F1,33 = 24.77, ≪ 0.001), as well as a significant effect of block (F12,33 = 8.16, ≪ 0.001) in the model (Fig. 2). Because PC1 is a unit-less variable, we provide reference values for plant height, number of domatia, and number of leaves in the legend of Fig. 2.

Indirect Resistance

Plants made larger domatia in response to folivory (Fig. 3) whether or not ants were present. Because larger plants make larger domatia, we included final plant size (PC1) as a covariate in the ancova model, with the mean volume of all domatia produced on each plant since the start of the experiment as the response variable and herbivore and ant treatments as well as their interaction as fixed effects. Only plant size and herbivore treatment explained a significant portion of the variation in mean domatium volume among plants (PC1: F1,46 = 16.13, < 0.001; herbivore treatment: F1,46 = 8.89, P = 0.005; ant treatment: F1,46 = 0.049, P = 0.826; ant × herbivore treatment interaction effect: F1,46 = 0.28, P = 0.602). Replacing the treatment effects with the actual folivory experienced by the plant gave similar results (PC1: F1,46 = 20.36, ≪ 0.001; folivory: F1,46 = 16.07, < 0.001); this is how the data are presented in Fig. 3.

image

Figure 3. Relationship between domatium volume and folivory. Each point is a Cordia nodosa plant. Domatium volume was measured as the mean volume all the domatia produced by a plant since the beginning of the experiment. Folivory was measured as the mean percent of leaf area missing at the end of the experiment for all leaves on each plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between domatium volume and folivory from the statistical model (see text).

Download figure to PowerPoint

For plants with ants, larger domatia contained more worker ants (Fig. 4), regardless of domatium age (domatium volume: F1,322 = 54.62 ≪ 0.001; domatium age-class: F2,322 = 0.44, P = 0.647). Furthermore, when there were more worker ants in a domatium, there was less herbivory on its associated leaves (Fig. 5, F1,140 = 6.36, = 0.013), and here domatium age-class was a significant factor in the ancova model (F2,140 = 9.20, < 0.001) that included number of ants and domatium age-class as fixed factors; the leaves associated with older domatia had a greater percentage of leaf area missing.

image

Figure 4. Relationship between domatium volume and the number of Allomerus octoarticulatus worker ants inside. Each point is a domatium. Most insect herbivores had access to plants symbolized by filled circles, but were excluded from plants symbolized by open circles. The line shows the predicted relationship between domatium volume and number of workers from the statistical model (see text).

Download figure to PowerPoint

image

Figure 5. Relationship between the number of Allomerus octoarticulatus worker ants inside a domatium and folivory on associated leaves. Each point is a domatium. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate domatia and leaves that were < 5 months old at the end of the experiment; triangles indicate domatia and leaves that were 5–11-month old; squares indicate domatia and leaves that were > 11 months old. The line shows the predicted relationship between number of workers and folivory from the statistical model (see text).

Download figure to PowerPoint

Direct Resistance

Plants also invested more in direct resistance in response to herbivore damage. Plants that experienced more folivory made leaves with more trichomes; trichome density was positively associated with folivory for leaves that were < 5 months old (Fig. 6a; F1,46 = 14.82, < 0.001) and 5–11 months old (Fig. 6b; F1,46 = 6.15, P = 0.017), but not leaves that were more than 11 months old (Fig. 6c; F1,45 = 0.475, P = 0.494). Similarly, plants that experienced more folivory made tougher leaves, but only for leaves that were <5 months old (Fig. 7a; F1,45 = 7.85, P = 0.008), and not for leaves that were 5–11 months old (Fig. 7b; F1,46 = 0.860, P = 0.359) or more than 11 months old (Fig. 7c; F1,46 = 0.027, P = 0.871). The concentration of total phenolics in leaves was also positively associated with folivory in all leaf age-classes that we tested (all but very young leaves); there was a significant positive relationship with folivory for leaves < 5 months old (Fig. 8a; F1,40 = 4.65, P = 0.037), between 5 and 11 months old (Fig. 8b; F1,43 = 5.85, P = 0.020), and more than 11 months old (Fig. 8c; F1,43 = 4.08, P = 0.050).

image

Figure 6. Relationship between folivory and the density of trichomes on leaves (a) < 5 months old, (b) between 5 and 11 months old and (c) more than 11 months old. Each point is a Cordia nodosa plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between trichome density and folivory from the regression model (see text).

Download figure to PowerPoint

image

Figure 7. Relationship between folivory and the toughness of leaves (a) < 5 months old, (b) between 5 and 11 months old and (c) more than 11 months old. Each point is a Cordia nodosa plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between leaf toughness and folivory from the regression model (see text).

Download figure to PowerPoint

image

Figure 8. Relationship between folivory and the concentration of total phenolics (expressed as mg tannic acid equivalents mL−1) in leaves (a) < 5 months old, (b) between 5 and 11 months old and (c) more than 11 months old. Each point is a Cordia nodosa plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between total phenolics concentration and folivory from the regression model (see text).

Download figure to PowerPoint

Finally, leaves from plants that experienced little folivory during the experiment were more palatable to Z. defensa in the bioassays [Fig. 9, folivory: F1,46 = 15.01, < 0.001; block effect (random): F12,46 = 3.79, P = 0.001]. We combined our three measures of direct resistance (leaf toughness, trichome density, and total phenolics concentration) using principal component analysis to investigate the relationship between direct resistance and palatability in the bioassays. PC1 had an eigenvalue of 1.51 and explained 50.3% of the variation in the data set; PC2 had an eigenvalue of 0.91 and explained 30.2% of the variation; and PC3 had an eigenvalue of 0.58 and explained 19.5% of the variation. Palatability in the bioassays had a highly negative and significant association with PC1, but not with PC2 or PC3 (PC1: F1,23 = 8.57, P = 0.008; PC2: F1,23 = 0.417, P = 0.525; PC3: F1,23 = 0.032, P = 0.860; block effect (random): F1,23 = 2.91, P = 0.017). The loading matrix showed that leaf toughness contributed more to PC1 than either trichome density or total phenolics concentration, although the loadings were all positive and not markedly dissimilar (leaf toughness: 0.823, trichome density: 0.689, total phenolics; 0.597).

image

Figure 9. Relationship between the palatability of leaves to beetles in the bioassays and folivory. Each point is a Cordia nodosa plant. Palatability was measured as the percent of leaf eaten by beetles during the bioassays. Folivory was measured as the mean percent of leaf area missing at the end of the experiment for all leaves on each plant. Most insect herbivores had access to plants indicated by filled symbols, but were excluded from plants indicated by open symbols. Circles indicate plants with an Allomerus octoarticulatus ant colony; triangles indicate ant-excluded plants. The line shows the predicted relationship between palatability and folivory from the statistical model (see text).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Cordia nodosa appears to have a back-up defence system; plants induce direct resistance when ants fail to defend them against herbivores. In the treatments in which herbivores had access to C. nodosa plants, ant-excluded plants suffered significantly more folivory than plants with A. octoarticulatus colonies. Furthermore, plants that experienced greater herbivory made tougher leaves with more trichomes and more total phenolics; plants in the A−H+ treatment generally had the highest values of these direct resistance traits (Figs 6–8). In bioassays, these leaves were dispreferred by a galerucine leaf beetle that commonly feeds on C. nodosa. These results suggest that this ant-plant has supplemented, rather than replaced, its direct defences with indirect defence.

Our results also suggest that ant-exclusion experiments may frequently underestimate the contribution that ants make to plant defence. Although numerous studies have compared ant-excluded and control treatments, these studies rarely (if ever) account for the fact that any induction of direct resistance in the ant-excluded treatment will tend to minimize differences in herbivory and plant performance between treatments. Therefore, the effect sizes reported in recent meta-analyses of ant-exclusion experiments (Chamberlain & Holland 2009; Rosumek et al. 2009; Trager et al. 2010) probably do not capture the real magnitude of ant benefits to plants and may obscure important patterns in the strength of ant–plant interactions.

Plants made tougher leaves with more trichomes in response to herbivory, but this was true only for leaves produced since the start of the experiment (Figs 6 and 7); in fact, for toughness, it was true only of leaves < 5 months old. Cordia nodosa may not be able to modify these structural properties of leaves, once they are in place. Total phenolics concentration, on the other hand, increased with folivore damage in leaves both young and old (Fig. 8), indicating that C. nodosa can induce chemical resistance in leaves even after they are fully expanded. The negative correlation between these direct resistance mechanisms and palatability to leaf beetles in bioassays suggests that leaf toughness, trichomes, and phenolics (or correlated traits) deter herbivores from feeding on C. nodosa and may therefore be effective direct defences.

Some other ant-plants, such as Acacia drepanolobium, make more domatia after they are browsed by herbivores (e.g. Palmer et al. 2008), but C. nodosa produces one new domatium on every new lateral branch it grows, regardless of herbivore damage. Here, we found that C. nodosa makes larger domatia in response to herbivory (Fig. 3), showing for the first time that domatia size can be induced by herbivory, much like domatia number in other myrmecophytic species. Interestingly, plants made larger domatia in response to herbivory regardless of whether they had ants (this was also true for the physical and chemical resistance traits we measured), suggesting that C. nodosa does not monitor ant presence or activity independently of assessing herbivore damage. It would not be surprising if an ant-plant such as C. nodosa could directly detect the presence of ants from their chemical secretions, in much the same way as plants recognize and respond to compounds in caterpillar saliva (e.g. Alborn et al. 1997). However, C. nodosa plants made larger domatia in response to herbivore damage both in the presence and in the absence of ants (Fig. 3). Our results suggest that making larger domatia following herbivore damage may benefit plants when ants are present, because larger domatia house more workers (Fig. 4), and domatia with more workers experience less herbivory on their associated leaves (Fig. 5). However, it is more difficult to imagine how larger domatia might benefit a plant that does not have ants, unless larger domatia are also better at attracting ant queens looking to found colonies.

Although it is often argued that selection favours inducible over constitutive resistance to save resources when defences are not needed, Agrawal & Karban (1999) suggested many other reasons why induced defences might benefit plants, including the possibility that induced defences reduce the negative effects of plant resistance on the natural enemies of herbivores. We do not know whether or how direct resistance traits – such as leaf toughness, trichomes, or phenolic compounds – affect A. octoarticulatus ants, but it is possible that the interaction is negative. For example, phenols could accumulate in C. nodosa food bodies, in honeydew excreted by the scale insects inside C. nodosa domatia, or in C. nodosa folivores, all of which A. octoarticulatus eats. We are not aware of any studies that have investigated the effects of plant resistance on the performance of phytoecious ants, but this would be an interesting avenue for future work.

In the large literature on inducible defences, studies of temperate plants outnumber those involving tropical species (Boege 2004; Bixenmann, Coley & Kursar 2011). Recently, Bixenmann, Coley & Kursar (2011) proposed that the advantages of inducible defences might be greater in temperate than tropical regions because herbivore pressure is highly variable in temperate ecosystems, but higher and more constant in the tropics. They said that, ‘The lack of evidence supporting induced defences in tropical plants suggests that … selection favours constitutive defences in tropical plants.’ Although this may very well be true of plants that rely primarily on direct defences, plants with indirect defences often experience highly variable rates of herbivory because of the vagaries of biotic defenders such as ants. This is evident in our data set; on average, the presence of A. octoarticulatus reduced folivore damage to C. nodosa in this experiment (see also Yu & Pierce 1998; Frederickson 2005; Frederickson et al. 2012), but the range of folivory values was similar in the A+H+ and A−H+ treatments (e.g. Fig. 2). The negative relationship between folivory and plant size (Fig. 2) suggests that folivory has fitness consequences for C. nodosa, as size appears to be an important fitness component in this species (Frederickson et al. 2012). Although induced defences can be costly and could also have contributed to the pattern we observed in Fig. 2, the slope of the relationship would likely have been even steeper without defence induction (e.g. Baldwin 1998; Cipollini & Sipe 2001). Thus, C. nodosa, and many of the large number of other tropical plants that rely on indirect resistance (e.g. Schupp & Feener 1991; Fiala & Linsenmair 1994), may experience selection for inducible direct defences whenever ants or other natural enemies fail to protect them against herbivores.

We should not, therefore, expect that plant lineages should lose direct defences when they evolve ant associations, but we might predict selection for greater inducibility of direct defences among ant-plants, compared with their relatives (all else being equal). Interestingly, if ant-plants often induce direct defences when ants are absent, this should give the appearance of a within-species trade-off between direct and indirect defences, because the presence of ants will be associated with low levels of direct defences and vice versa. The results of Koricheva & Romero (2012), who meta-analysed studies of within-species variation in plant allocation to direct and ant-mediated defences and found a significant negative correlation, suggest that the defence strategy we observed in C. nodosa may be widespread.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank E. Hodgson, T. Kursar, and R. Sage for chemical ecology assistance or advice; I. Aggarwal for analysing leaf images; A. Coral, L. Flores Quispe, and J. Sanders for help in the field; the staff at the Los Amigos Research Center for logistics; and the Peruvian Ministry of Agriculture for issuing permits (394-2009-AG-DGFFS-DGEFFS and 79-2008-INRENA-IFFS-DCB). We are grateful to C. Chaboo, D. Furth, and especially S. Clark for identifying the beetle we used in the palatability trials; vouchers have been deposited in entomology collections at The University of Kansas, Brigham Young University, and the Museo de Historia Natural, Universidad Nacional Mayor de San Marcos in Lima, Peru. We also thank two anonymous referees for their helpful comments. This research was funded by a NSERC Discovery Grant and a Connaught New Researcher Award to M.E.F. G.A.M. was supported by a Foundational Questions in Evolutionary Biology Postdoctoral Fellowship (funded by the Templeton Foundation), A.R. was supported by a Benjamin A. Trustman Fellowship from Harvard University, G.B. by the Independent Experiential Study Program of the Faculty of Arts and Science and V.A. by the Work-Study Program, both at the University of Toronto.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References