Herbivores and mutualistic ants interact to modify tree photosynthesis



In the arms race between plants and herbivores, plants have evolved a variety of strategies to reduce the fitness costs of herbivore damage by investing resources in defense compounds and structures. Because both herbivore damage and defense investments are costly in terms of plant’s resources, tradeoffs have emerged as a central theme in the evolutionary ecology of plant responses to herbivory, shaping the extent and the forms of defense that plants evolve (Coley et al., 1985; van der Meijden et al., 1988). While understanding allocation tradeoffs has fuelled prolific research and debate (Nuñez-Farfan et al., 2007), empirical studies have not generated consistent support for tradeoff-based arguments. Plant physiology research, which focuses on resource acquisition and metabolic processes upstream of phenotypic plant defense traits, offers an under-explored perspective that may yield deeper insights into costs and allocation patterns of plant defense. Several studies have demonstrated that herbivory can trigger increases in leaf-level photosynthetic rates (Welter, 1989; Trumble et al., 1993; Thomson et al., 2003). For plants that utilize carbon-based defense strategies, this photosynthetic up-regulation has the important consequence of increasing the available pool of carbohydrates that can potentially be allocated to defense. Such varying rates of carbon assimilation fundamentally change the playing field for evaluating the costs of defense investments.

Plants employ three main strategies to mitigate the costs of herbivory: resistance, indirect defense and tolerance (Karban & Baldwin, 1997). Resistance strategies, such as thorns and noxious chemicals, and indirect defenses, whereby plants attract enemies of herbivores, both reduce damage to plants by decreasing the rates of herbivore attack. Tolerance strategies, by contrast, reduce the fitness cost of herbivory by allocating more resources to growth in order to replace lost tissue. Acacia drepanolobium, a dominant tree through much of East Africa’s savannas, exhibits inducible carbon-based investments in all three classes of defense strategies. In response to herbivory, trees allocate photosynthate to produce longer and more numerous spines, which deter mammalian browsers (resistance strategy) (Young, 1987; Young et al., 2003). Herbivory can also stimulate increased rates of leaf and shoot growth (tolerance strategy) (Huntzinger et al., 2004). Finally, A. drepanolobium trees provide inducible extrafloral nectar and bulbous hollow thorns as food and housing for resident ants, which in turn help to protect trees from herbivory (indirect defense strategy) (Madden & Young, 1992; Stapley, 1998).

In the savannas of Laikipia District in northern Kenya, nearly every A. drepanolobium tree (c. 100% of trees > 1 m tall) is occupied by one of four mutually exclusive ant species, with ant species turnover rates of c. 15% yr−1 (Palmer et al., 2000). Each ant species exhibits its own suite of distinctive behaviours towards its tree host (Young et al., 1997; Palmer et al., 2000). Crematogaster mimosae behaves as a faithful mutualist, forming dense populations, recruiting aggressively to attack browsers and patrolling to remove insect herbivores (Young et al., 1997). Crematogaster nigriceps also defends strongly against herbivory, with dense colonies and an aggressive response to browsers and insects (Palmer & Brody, 2007), but these ants eat and destroy most of the host trees’ axillary and terminal shoots (Stanton et al., 1999). The two other ant species, Crematogaster sjostedti and Tetraponera penzigi, offer little, if any, mutualistic benefit to their host trees; they defend weakly against mammalian browsers and ignore or even promote insect damage to host trees (Palmer & Brody, 2007; Palmer et al., 2008).

While numerous previous studies in this system have shown that resident ant species differentially affect the extent to which various defensive traits are induced or relaxed in response to changing browsing pressure (Gadd et al., 2001; Young et al., 2003; Huntzinger et al., 2004; Palmer et al., 2008), here we investigate for the first time whether interactions with browsers and resident ants also modulate leaf-level photosynthetic rates, thereby affecting available carbon pools and metabolic pathways to induced defenses.

Materials and Methods

We conducted the study at Mpala Research Center, Kenya (36°52′E, 0°17′N) within the Kenya Long-term Exclosure Experiment (KLEE), a replicated array of 4-ha plots maintained with access to all wildlife (hereafter ‘browsed’) or electric-fenced to allow no wildlife access (hereafter ‘unbrowsed’) since 1995, where A. drepanolobium (Sjostedt) trees constitute 95% of the tree canopy (Young et al., 1998). In three browsed and three unbrowsed plots, we marked 12 similarly sized trees, three occupied by each ant species. In January 2009 (growing season) and April 2009 (dry season), we measured gas-exchange rates on one unshaded, fully expanded leaf per tree, using an LI-6400XT photosynthesis system (LI-COR Biosciences, Lincoln, NE, USA) at fixed photon flux density, CO2 concentration, humidity and leaf temperature, from which net photosynthesis (Pn), transpiration (Tr) and water use efficiency (WUE) were calculated. In January, we collected 10 unshaded, fully expanded leaves from one tree per ant species per plot for total nitrogen and δ15N analysis (University of California, Davis Stable Isotope Facility). Using JMP Statistical Software (SAS, Inc., Cary, NC, USA), we analyzed the effects of browsing and occupying ant species on Pn, Tr and WUE for each season using two-way ANOVA, and planned Student’s pairwise contrasts between browsing levels within each ant species. Relationships between metabolic parameters and tissue chemistry in January were analyzed using ANOVA techniques. Distributions met assumptions of normality and homoscedasticity.

Results and Discussion

During the active growing season, photosynthetic rates per unit area (Pn) in the browsed plots were c. 50% greater on trees occupied by C. mimosae and C. nigriceps than on trees occupied by C. sjostedti and T. penzigi. In unbrowsed plots, all trees exhibited similar Pn regardless of ant occupant (Fig. 1a). From unbrowsed to browsed treatments, Pn increased significantly for C. mimosae and C. nigriceps trees (contrasts: = 0.040 and = 0.006, respectively), while the slight decreases in Pn for trees occupied by C. sjostedti and T. penzigi were not significant (contrasts: = 0.081 and = 0.149, respectively). Transpiration rates showed nearly identical patterns of variation in response to ants and browsing as Pn (Fig. 1b). As a consequence, there were no significant differences in WUE (the ratio of Pn : Tr) between treatments or ant occupants (Fig. 1c), meaning that trees with elevated photosynthesis were therefore using larger absolute quantities of water to support those rates.

Figure 1.

 Means (± 1 SE) of leaf-level net photosynthesis (Pn), transpiration (Tr) and water use efficiency (WUE) of Acacia drepanolobium trees located in browsed plots and plots from which browsers had been excluded for 15 yr, grouped by resident ant species. Measurements were performed during the growing season in January 2009 (a–c) and during the dry season in April 2009 (d–f). Results of two-way ANOVAs are given for each variable. In January, 66 trees were measured, = 8 or 9 trees in each ant × browsing treatment combination. In April, 60 trees were measured, = 7, 8, or 9 trees in each treatment combination.

In the dry season, Pn was generally lower and all metabolic rates showed greater variation, which made ant-specific effects harder to detect. A. drepanolobium is deciduous, and trees were in variable stages of leaf senescence by the April census. Despite that variability, we again found that browsed trees occupied by C. mimosae and C. nigriceps had significantly higher Pn than trees occupied by C. sjostedti or T. penzigi, while unbrowsed trees showed no ant-specific differences (Fig. 1d, contrasts C. mimosae, C. nigriceps vs C. sjostedti, T. penzigi: browsed < 0.0001, unbrowsed = 0.061). Unlike the growing season, dry season Tr was significantly higher in browsed trees, with no significant ant or interaction effects (Fig. 1e). This translated into an ant-specific effect on WUE, namely that unbrowsed trees with C. mimosae and C. nigriceps had greater WUE than other ant-browsing combinations (Fig. 1f). While all browsed trees showed higher rates of Tr, only the trees occupied by C. mimosae and C. nigriceps achieved higher Pn by doing so.

The key finding of this study is that A. drepanolobium trees exhibit elevated photosynthetic rates in response to browsing only when occupied by strongly mutualistic ants. We discuss key implications of this novel result in the context of previous studies of defense expression in this system, as well as studies of induced variation in photosynthetic rates in other systems.

We compared our findings with previously documented effects of browsing and resident ant species on defense traits in this system. The strongest parallel was between Pn and tolerance, assessed as new leaf growth (Gadd et al., 2001) (Table 1). Photosynthetic up-regulation has been postulated to enable tolerance-related growth (Trumble et al., 1993; Karban & Baldwin, 1997; Strauss & Agrawal, 1999). A few recent studies have tested for correlated changes in Pn and tolerance by crossing herbivory treatments with another environmental variable known to affect Pn and/or tolerance (Gonzales et al., 2008; Stevens et al., 2008; Suwa & Maherali, 2008). Yet, to our knowledge, this is the first study that has in fact found congruent increases in Pn and tolerance when tested across another environmental variable, which in our case was the species of occupying ant. The ant-specific photosynthetic responses we observed were not congruent, however, with patterns of resistance or indirect defense, assessed via thorn length (Young et al., 2003), extrafloral nectary and swollen thorn production (Palmer et al., 2008), (Table 1). Instead, each ant species is associated with a distinct pattern of host tree photosynthetic rates and subsequent allocation to defense in response to herbivory. The environmentally variable correlations between Pn, tolerance and other defense traits highlight the need to understand, in more detail, external drivers of metabolic resource assimilation rates and subsequent allocation before assessing the costs of defense.

Table 1.   Comparison of effects of browsing on Acacia drepanolobium tree traits, which vary depending on the species of ant occupying trees
Tree traitOccupying ant species
Crematogaster mimosaeCrematogaster nigricepsCrematogaster sjostedtiTetraponera penzigi
  1. Tissue-level photosynthesis responses were measured in this study; responses of tolerance, resistance and indirect defense traits were drawn from cited studies.

  2. ↑, Significant increase in browsed trees; ○, no significant change.

  3. 1, From Gadd et al. (2001); 2, from Huntzinger et al. (2004); 3, from Palmer et al. (2008).

Primary metabolism: net photosynthesis
Tolerance: new leaves per shoot1
Resistance: thorn length2  
Indirect defense: nectaries3 
Indirect defense: swollen thorns3

Current research has focused on herbivore-induced tissue loss and damage as the cueing mechanism for photosynthetic up-regulation, usually through the creation of carbon sinks and gradients (Schwachtje & Baldwin, 2008). In our study system, the behaviours of the two strong mutualist ant species cause very different degrees of tissue damage on their host trees. C. nigriceps ants cause continual damage by chewing and destroying nearly all axillary buds, while C. mimosae ants do not display this behaviour (Young et al., 1997; Stanton et al., 1999). The steep carbon gradient resulting from C. nigriceps’ damage-causing behaviour is thought to explain why their host trees continue to produce more nectar and longer thorns even when herbivores are absent, whereas C. mimosae trees only elevate nectar production and thorn elongation in the presence of browsers (Young et al., 2003; Palmer et al., 2008). In contrast to their differential expression of defense traits, C. nigriceps and C. mimosae trees both show the same reduction in photosynthesis when browsers are excluded. Thus, damage-based mechanisms can adequately explain documented patterns of allocation of photosynthate to defense traits, but it appears that some additional mechanism, driven by the interaction between herbivores and occupying ants, is responsible for regulating the rate of photosynthesis itself.

In addition to intrinsic cueing mechanisms, Pn up-regulation can depend on resource availability, such as nitrogen (Ripley et al., 2006) or water (Gonzales et al., 2008). We investigated whether the quantity of nitrogen per leaf area, as tissue nitrogen concentration and as leaf thickness, differed between treatments that exhibited elevated vs ‘baseline’Pn (i.e. browsed C. mimosae and C. nigriceps trees, vs all other ant-browsing treatments) in the growing season. The percentage of nitrogen in leaves did not vary based on treatment Pn elevation (ANOVA, degrees of freedom (df) = 1, = 0.04, = 0.52). nor did specific leaf area (cm2 g−1) (ANOVA, df = 1, = 2.51, = 0.13). These results imply that the photosynthetic advantage of browsed C. mimosae and C. nigriceps trees was not the result of greater nitrogen allocation per leaf area. Given the observed trends in WUE, water availability may be a factor impacting the costs of herbivory and photosynthetic up-regulation. However, we are unable to propose a plausible way that C. mimosae and C. nigriceps and browsing could interact to increase a tree’s water availability as a mechanism of Pn up-regulation. Nevertheless, most studies in this system point to the complex, multivariate nature of interactions, in terms of both upstream cues and downstream responses (Young et al., 1997; Palmer, 2003; Palmer et al., 2008). Further studies that manipulate resources may help to determine whether, and how, resource availability plays a mechanistic role in Pn modulation in response to ants and browsers.

Our findings suggest that the strong mutualist ant species C. mimosae and C. nigriceps may reduce the costs of herbivory to their host trees in two distinct ways. First, given that C. mimosae and C. nigriceps ants are known to swarm much more aggressively to disturbed branches than C. sjostedti or T. penzigi (Palmer & Brody, 2007), and browser deterrence is proportional to the abundance of swarming ants (Madden & Young, 1992; Stapley, 1998), the aggressive behaviour of these species should directly reduce the rates of herbivore damage. Second, by enabling photosynthetic up-regulation, the presence of the strong mutualist ants can further mitigate costs of herbivory by increasing pools of photosynthate available for additional defense or for regrowth of lost tissue. To our knowledge, this is the first study to identify a metabolic benefit of indirect defenders via increased photosynthesis.

The tritrophic structure of this plant–herbivore system, with multiple indirect defenders and all three inducible defense strategies, presents a particularly powerful lens for seeing how biotic interactions modulate plant responses to herbivory in terms of cueing mechanisms, metabolic response and allocation to induced defense (Heil & McKey, 2003; Heil, 2008). The potential for metabolic variation, however, hypothetically exists in every plant–herbivore system. Despite growing attention to herbivory-induced metabolic variation in ecophysiological research (Schwachtje & Baldwin, 2008), most evolutionary ecology studies of defense do not consider photosynthetic up-regulation as an integral factor affecting resource availability for defense, or that it can respond to biotic and abiotic environmental conditions independently of defense trait expression (Fig. 2). Tritrophic regulation of photosynthesis provocatively illustrates the need to consider primary metabolism in order to unravel the costs of plant defense.

Figure 2.

 (a) Schematic diagram of the prevailing paradigm for evaluating biotic interactions and plant defense. Effects of herbivores and indirect defenders are recognized to affect plant allocation of resources to defense, but not primary metabolic acquisition of resources. (b) Revision of the paradigm to include critical effects of herbivores and indirect defenders on rates of carbon assimilation, which subsequently alters pools of resources available for allocation to defense traits.


The research was supported by NSF grants EAR-0847368 and DEB-0816453, and by internal grants from Princeton University through the Grand Challenges Initiative and the Princeton Institute for International and Regional Studies. Equipment was generously provided through a grant from the LI-COR Environmental Education Fund (LEEF). The exclosure plots were built and maintained by grants from the James Smithson Fund of the Smithsonian Institution (to A. P. Smith), The National Geographic Society (4691-91), The National Science Foundation (LTREB BSR-97-07477, 03-16402, and 08-16453) and the African Elephant Program of the US Fish and Wildlife Service (98210-0-G563) (to T. P. Young). We thank A. Lekutaas, C. Riginos, and D. Rubenstein for their help and comments.