Left Out in the Cold: Thermal Variation in an Ant-Plant Defense

Material Information

Left Out in the Cold: Thermal Variation in an Ant-Plant Defense
Tamashiro, Ryan Akio
Publication Date:


Despite the vast research on the thermal ecology of ants, few studies have examined the role that temperature plays in mediating protective ant-plant interactions. These mutualisms are conditional interactions that depend on the ants' ability to repel herbivory and improve the fitness of their host. To assess for temperature dependence, we observed baseline activity and the aggressive response against a simulated herbivore in four symbiotic ant species on their myrmecophyte host (Acacia drepanolobium) across a thermal gradient. Furthermore, we experimental removed the most aggressive ant species (C. mimosae) from trees to determine if thermal variation in activity led to a functional response in a model herbivore (goats). Aggressive responses to the simulated herbivore increased with surface temperature for each species, but the two most aggressive species had a stronger response to temperature. As surface temperature increased, goats took fewer bites and spent less time feeding from A. drepanolobium as a direct result of increased C. mimosae aggression. However, we did not detect any thermal dependence in baseline activity among the ant species. Overall, our study suggests that thermal variation plays a role in mediating species interactions in a mutualism network. ( en )
General Note:
Awarded Bachelor of Science, summa cum laude, on May 8, 2018. Major: Zoology
General Note:
College or School: College of Liberal Arts and Sciences
General Note:
Advisor: Todd Palmer. Advisor Department or School: Biology

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Ryan Akio Tamashiro. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

UFDC Membership

UF Undergraduate Honors Theses


This item is only available as the following downloads:

Full Text


Title: Left Out in t he Cold: Thermal Variation in an Ant Plant Defense 1 2 Ryan A. Tamashiro 1,2 and Todd M. Palmer 1,2 3 1 Department of Biology, University of Florida, Gainesville, Florida 32611 USA 4 2 Mpala Research Centre, Box 555, Nanyuki, Kenya 5 6 Abstract 7 Despite the vast research on the thermal ecology of ants, few studies have examined the 8 role that temperature plays in mediating protective ant plant interactions. These mutualisms are 9 conditional interaction s that depend on the ants' ability to repel herb ivory and improve the 10 fitness of their host. To assess for temperature dependence we observed baseline activity and the 11 aggressive response against a simulated herbivore in four symbiotic ant species on their 12 myrmecophyte host ( Acacia drepanolobium ) acros s a thermal gradient Furthermore, we 13 experimental removed th e most aggressive ant species ( C. mimosae ) from trees to determine if 14 thermal variation in activity led to a functional r esponse in a model herbivore ( goat s ). Aggressive 15 responses to the simulated herbivore increased with surface temperature for each species, but the 16 two most aggressive species had a stronger response to temperature As surface temperature 17 increased, goats took fewer bites and spent less time fe eding from A. drepanolobium as a direct 18 result of increased C. mimosae aggression However, we did not detect any thermal dependence 19 in baseline activity among the ant species. Overall, our study suggests that thermal variation 20 plays a role in mediating sp ecies interactions in a mutualism network 21 Keywords: plant defense, ant plant mutualism, thermal ecology, herbivory, abiotic 22 23


Introduction 24 Plants have evolved a wide variety of defenses in response to herbivory including 25 chemical meta bolites physical armature and protein based defenses ( reviewed in Freeman and 26 Beattie 2008 ). One unique defensive adaptation involves a mutualis tic relationship with 27 protective symbiotic ants Such ant plant protection mutualisms have repeatedly evolved 28 throughout the tropics in volving over 100 species of flowering plants and 40 genera of ants 29 (Davidson and McKey 1993). A nts provide a number of services for their host plants such as 30 defense against herbivory, pathogens, and encr oaching plants receiving nesting space and food 31 (i.e. nectar and/or food bodies) in return (reviewed in Heil and McKey 2003 ) 32 The ant defenders are considered an inducible defense as they can vary the intensity of 33 their response depending on several biotic factors ( i.e. plant disturbance, leaf damage, extrafloral 34 nectar secretion; reviewed in Agrawal and Rutter 1998 ) Comparatively few er studies have 35 considered abiotic factors that effect this plant defense ( but see Kersch and Fonseca 2008 ; 36 Fitzpatrick et al 2013, 2014) Because ants are poikilotherms, their activity levels tend to 37 increase with temperature (Hulbert et al. 2008) with peak activity occurr ing in each species' 38 optimal thermal range after which activity decreases due to risk of overheating or desiccation 39 ( C erd‡ et al. 1998, Bucy and Breed 2006 Azacarte et al. 2007 ). Since the efficacy of ants in 40 defending against herbivory is tied to their activity level s (e.g., Palmer and Brody 2007), 41 d ifferences in thermal tolerance among ant species may t herefore influence the defensive 42 efficacy of different ant species For example, in the Sonoran desert, barrel cacti ( Ferocactus 43 wislizeni ) are defended by four ant species the most effective of which has a narrower thermal 44 tolerance than the least effect ive mutualist ( Fitzpatrick et al. (2014) As a consequence, as 45 temperatures peak during the hottest part of the day, the most effective defenders are replaced by 46


less effective defenders, suggesting that temperature can strongly influence the quality of 47 pr otection offered by ant mutualists. However, despite the potential for temperature dependence 48 of ant defense of plants, to our knowledge there are no experimental studies evaluating how 49 temperature influences the ability of ants to defend plants from herbi vore attack. 50 In this study, we examined interactions among the East African ant plant Acacia 51 ( Vachellia ) drepanolobium and its ant symbionts across the natural temperature gradient that 52 occurs from sunrise to sunset. Acacia drepanolobium is a myrmecophytic tree that is broadly 53 distributed and locally abundant on clay rich "black cotton" savannas throughout East Africa 54 (Herlocker 1974, Stapley 1998), frequently comprising >90% of total woody plant cover (Young 55 et al. 199 7, 199 8). Across th is range, trees host single colonies (at any given time) of between 56 one and four sympatric symbiotic ant species, each of which imposes different costs and confers 57 different benefits to its host plant (Palmer and Brody 2007, Palmer et al. 2010). At our stu dy site, 58 ants are exposed to average daily temperature cycles that vary by 16¡ C over the course of the 59 day with even higher variation in surface temperatures, which allow ed us to address three main 60 questions: (1) How does temperature affect ant activity o n their host tree throughout the day ? (2) 61 How does temperature affect the strength of a defensive response against simulated herbivory 62 throughout the day ? ( 3 ) Does the efficacy of ant defense against a browsing herbivore (goats) 63 depend on temperature? 64 65 Me thods 66 Study System 67 We conducted our study at Mpala Research Centre (0¡17'N, 36¡52'E, 1700 2000m 68 elevation) in Laikipia, Kenya. The lowest average monthly minimum and highest average 69


monthly maximum ambient temperatures are 11.0¡C and 29.5¡C, respectively The study site is 70 characterized by a heavy "black cotton" vertisol (Ahn and Geiger 1987) dominated by several 71 grasses and a single tree species, A drepanolobium (Young et al. 1997 1998 ). In an ecosystem 72 with a broad range of large browsing herbivores, A drepanolobium is well defended by a 73 combination of spines and symbiotic ants (Hocking 1970). While all spines function as 74 mechanical defenses, a subset is swollen to provide a nesting cavity for the protective ants. 75 A cacia drepanolobium also provide a food source for the ants by secreting a carbohydrate rich 76 nectar from extrafloral nectaries located on the rachis of each leaf Almost every tree is occupied 77 by a single colony of one of four symbiotic ant species: Crematogaster sjostedti, C. mim osae, C. 78 nigriceps and T. penzigi While each tree only host s one colony, a single colony can occupy 79 several neighboring trees. The defensive capabilities among ant species differ ; C. mimosae has 80 been shown to defend their hosts the most aggressively, foll owed by C. nigricep s T. penzigi and 81 C. sjostedti (Palmer et al. 2010). Removal of C. mimosae dramatically increased elephant 82 damage compared to controls, and removal of both aggressive ants ( C. mimosae and C. 83 nigriceps ) increased the proportion of browsed branch apices (Stanton and Palmer 2011). 84 Ant Defensive B ehavioral Observations 85 To observe the relationship between ant activity and temperature, we surveyed 60 trees 86 within a 0.1 km 2 study area. We identified 15 trees occupied by each ant species that were 1.5 2 87 meters tall an d not in contact with the canopies of surrounding trees. Selected trees occupied by 88 conspecifics were separated by at least 15 meters to ensure each replicate was an independent 89 colony We s eparated the 60 trees int o 12 spatial blocks and randomly sampled the five trees 90 within each spatial block using a random number generator Every spatial block contained at 91


least one tree occupied by each species ( except for one lacking a tree hosting C. mimosae ) to 92 minimize a bloc k effect 93 For every tree we measured ant behaviors from before sunrise to after sundown 94 span ning a range of surface temperatures from 9¡C (lowest recorded) to 49¡C (highest recorded) 95 Each tree was observed at least once within the following time windows : dawn ( before 6:31 am), 96 dusk (af ter 6:30pm), and at two hour intervals in between Before behavioral observations were 97 taken, we recorded the time of day and surface temperature of an outer branch The average 98 temperature was recorded by holding the Fluke 62 Max IR Thermometer (Fluke Inc., Everett, 99 Washington, USA) about 15 cm above a branch on the sunned side of the tree for five seconds 100 We chose to measure surface temperature rather than ambient temperature because ants exist in 101 superheated bounda ry layers than can be considerable hotter than the air a few more millimeters 102 above the surface (Kaspari et al. 2015). 103 Two behavior s were observed: (1) baseline activity along undisturbed branches and (2) 104 swarming an artificial herbivore. First, we measured baseline activity by haphazardly selecting 105 healthy bra nches on each tree and counting the number of ants that pass ed into a 5 cm segment 106 of the branch over 30 second s The focal segment was the closest 5 cm section to an apical bud 107 free of any swollen thorns. We performed this procedure on two branches simultaneously to 108 produce a sum baseline activity level. Second, we simulated herbivory by rapidly sliding a 109 gloved hand 15cm down the branch towards the tip three times and then grasping t he branch 110 within the 15 cm region for 30 s careful not to directly grab a swollen thorn (Palmer and Brody 111 2013) A fter the 30s period, we released the glove counted the number of ants that had crawled 112 onto the glove. The sum of the ants on both gloves was used as an index for the intensity of anti 113


herbivory defense. Given that the ant colony sizes are in the thousands to tens of thousands we 114 assumed that each glove test did not influence successive observations ( Palmer 2004) 115 Ant Defense against a Model Herbivore 116 In our second experiment, we sought t o determine whether temperature dependent 117 variation in ant defensive level influenced herbivore behavior We used g oats (which widely 118 occur throughout our study region) as model browsers and examined their feeding behavior at 119 three distinct times on host plants with ants, and on host plants from which ants had been 120 removed The goats were all female approximately the same size, and hunger motivated, as food 121 was withheld beginning the night before eac h experiment First, we selected fourteen 1 meter tall 122 trees with similar canopy volumes that were occupied by C. mimosae We then paired them 123 sp atially and randomly selected a control (ant occupied) and experimental ( removal ) tree with a 124 coin toss The removal tree controlled for temporal differences in hunger within the goats. Five 125 days before conducting the feeding trials, w e removed the ants from experimental trees by 126 disturbing the trees and misting them with short lived, pyrethrin based insecticide (0.6% a lpha 127 cyp ermethrine) and wrapped the trun ks in duct tape and Tanglefoot ( Contech, Spartanburg, 128 South Carolina, USA ) to prevent ground recolonization by ants 129 Over two days, we brought goats to the site and allowed them to feed on experimental 130 trees at thre e times that spanned a range of temperatures : 6:30am, 9:30am, and 12:30pm. A pair 131 of goats was assigned to each pair of trees (removal and occupied) Both goats fed from the same 132 tree simultaneously so they experienced the same level of ant aggression. The first tree the pair 133 of goat s fed from was random ized by another coin flip We allowed the goats to feed from the 134 trees for a maximum of five minutes and counted the bites taken and total time spent feeding. 135 Total feeding time was defined as the time from the first bite to the last bite before the goat 136


refused to continue feeding. For each goat, we calculated the difference in total number of bites 137 taken and total feeding time for ant removal and ant occupied trees and used this di fference for 138 our analyses. 139 Statistical Analyses 140 Behavioral observations To assess for temperature dependence in baseline ant activity 141 and response to a simulated herbivore we performed a generalized linear mixed model (GLMM) 142 with a negative binomial distribution due to over dispersion We constructed models using 143 temperature, species, and the interaction of the two as fixed effects Each model included the 144 individual tree identity and time block as random effect s to account for repe ated measures and 145 variation between individual trees. The models were compared using Akaike Information 146 Criterion (AIC c ) score s We designated the best model as the one with the lowest AICc that did 147 not produce a convergence error 148 Goat experiment We used a linear mixed model to assess the effect of temperature on 149 C. mimosae 's efficacy in deterring goat herbivory We modeled the difference in bites taken and 150 time spent feeding between the two treatments (removal tree occupied tree) with temperatur e as 151 the only fixed effec t ; the random effects were tree identity and goat identity, nested within tr ee 152 identity 153 All s tatistical analyses were conduc ted in R 3.4.2 (R Core Team 2017 ). We used the lme4 154 package ( Bates et al. 2015 ) to perform the generalized linear mixed models the lmerTest 155 package to calculate p values for parameter estimates (Kuznetsova et al. 2017), and the bbmle 156 packaged was used to calculate and compare AICc scores (Bolker 2017) 157 158


Results 159 Surface temperature and b ehavio ral observations 160 Over the course of the day, surface temperatures ranged from 9¡C to 49¡C displaying a 161 strong parabolic relationship with time (Figure 1) The average surface temperature was highest 162 during the 10:31 12:30 period. The variance in surface temperature between trees increased with 163 time, peaking at the 12:30 to 14:30 time period, and declining thereafter Scattered cloud cover 164 during the middle of the day appeared to cause the variance in surface temperature during the 165 middle of the day. For baseline activity our model with the lowest AICc included species 166 temperature but not an interaction term ( Figure 2 ) Temperature was positively associated with 167 baseline activity but not significant ly so ( Temperature : b = 0.0067 S E = 0.0084 p = 0.423 ) and 168 the more aggressive ants ( C. mimosae and C. nigriceps ) had higher levels of activity compared to 169 C. sjostedti ( C. mimosae : b = 1.6138, SE = 0.197346, p << 0.01; C. nigriceps : b = 1.5564, SE = 170 0.1977, p << 0.01). Tetraponera penzigi had the lowest level of baseline activity measured on the 171 branches (b = 0.5445, SE = 0.2167, p = 0.012). While temperature was not associated with the 172 number of ants moving through the branch segment the speed of the ants appeared to greatly 173 inc rease with surface temperature. 174 In examining the relationship between defensive response and temperature, our model 175 including temperature and species had the lowest AICc (Figure 2) Temperature was positively 176 correlated with the number of ants respond ing to simulated herbivor y (Temperature: b = 0.0251, 177 SE = 0.0068, p << 0.01). Again, C. mimosae and C. nigriceps were more aggressive than C. 178 sjostedti ( C. mimosae : b = 3.1151, SE = 0.2088, p << 0.01; C. nigriceps : b = 2.7252, SE = 179 0.2094, p << 0.01), but T. penzigi was also a more aggressive defender (b = 0.9894, SE = 0.2133, 180 p << 0.01). Across all temperature, C. mimosae and C. nigriceps released an acrid o do r when 181


responding to the disturbance At the lowest temperatures, many of the ants fell off of the tr ee 182 when responding to simulated herbivory. 183 Defense against the Model Herbivore 184 Surface temperature was a significant predictor of the time spent feeding and bites taken 185 by goats (Figure 3) The difference in bites that the goats took from the ant removal trees 186 compared to the ant occupied tree increased with branch surface temperature (b = 1.554, SE = 187 0.3489, p <<0.01). Similarly, goats spent more time feeding from the removal trees compared to 188 the occupied trees as branch surface temperature incre ased (b = 5.320, SE = 1.351, p = 0.0017). 189 The time spent feeding and total number of bites that each goat took from ant removal tree s did 190 not significantly differ across different temperatures. When introduced to experimental plants, 191 goats rapidly began fe eding during all time intervals each day, suggesting that did not learn 192 avoidance of ant occupied trees, consistent with an earlier experimental study of goat browsing 193 on A. drepanolobium by Stapely (1998) As the goats began feeding from the occupied tree s, the 194 ants crawled onto the goat's head and began biting near the eyes and nose. The goats responding 195 by scratching at their face and snorting violently. These behaviors were not observed while the 196 goat fed from the removal trees. 197 Discussion 198 Our results d emonstrate that surface temperature plays a key role in mediating the 199 intensity of the ant defenders to both simulated herbivory and a model herbivore increases with 200 branch surface temperature The increase in defensive activity was observed across the four ant 201 species, but the differences in defensive intensity between the species was accentuated by 202 increasing temperature as the more aggressive ant species ( C. mimosae and C. nigriceps ) 203 increased their swarming response to a greater extent Within C. mimosae we found that 204


increased surface temperature produced a functional defensive response, as goats became 205 increasingly hesitant to feed from ant occupied trees than ant removal trees with increasing 206 temperature 207 When the thermal niche of the ant defenders d oes not align with that of the herbivores, 208 host plants are at increased risk of damage (Fitzpatrick et al. 2013 2014 ). In sub Saharan Africa, 209 many browsers including those that feed on A. drepanolobium preferentially feed during cooler 210 periods of the da y such as dawn and dusk (i.e. steenbok impala, greater kudu, giraffe, Du toit 211 and Yetman 2005; african elephant, Shannon et al. 2008) the same periods with lower levels of 212 ant activity. Unlike the barrel cactus system, the four A. drepanolobium mutualist are obligate 213 mutualist rather than facultative (Stanton et al. 2002) where the survival of the colony closely 214 intertwines with the fitness of their host (Agrawal and Rutter 1998) 215 Thus, the discordance between the temperature preferences of th e ants and vertebrate 216 herbivores is surprising, though several hypothes e s may explain our observations. First, goats 217 may serve as poor model browsers in a landscape dominated by mega herbivores (Goheen and 218 Palmer 2010). They have not evolved alongside the myrmecophyte and may not respond to the 219 visual and chemical cues of the ants (Wood et al. 2002 and 2006). Ant pheromones have been 220 shown to repel insect herbivores ( reviewed in Offenberg 2014), and Goheen and Palmer (2010) 221 have suggested that the smell of the aggressive acacia ants is enough to deter elephants. Goat 222 feeding also varies markedly from elephant or giraffe feeding and may not elicit the proper 223 response from the ants early in the morning. For instance, we observed the goats pick at 224 individual le afs between the thorns. Conversely, elephants destroy large sections of the canopy 225 or knock the tree over entirely (Stanton and Palmer 2011) and giraffes strip entire branches of the 226 leaves, thorns, and bark (Milewski et al. 1991). By engaging larger areas of the tree, the larger 227


herbivores potentially cause more ants may to respond to the disturbance. Second, the seemingly 228 redundant spines could function as the plant s primary defense during times when ant activity is 229 depressed reducing selective pressure for ants active at lower temperatures The spines have 230 been shown to reduce or at least slow herbivory in A. drepanolobium and other Acacia species 231 (Milewski et al. 199 1 Stapley 1998 ) though this effect is countered by longer feeding time 232 Lastly, the daily pattern of ant activity and our short term observations may not capture the long 233 term consequences of the mutualism For example, Stanton and Palmer (2011) have argued that 234 short term experiments will underestimate the benefits of the ant mutualists as the costs of 235 hosting ants are relatively continuous and the benefits are sporadic Here, we examined only a 236 single aspect of the mutualism that contributes to the plant s fitness over a few days but 237 neglected other aspects such as the costs as sociated with hosting the ants (i.e. branc h pruning, 238 Stanton et al. 1999; nectar induction and tending scale insects, P rior et al. 2018 ; but see 239 Bronstein 2001 ) The plant s may tolerate the cyclical pattern of activity because it also reduces 240 the costs associated with the mutualism. 241 As global temperatures continue to rise scientists continue to predict how climate change 242 will affect the fate of species (Thomas et al. 2004, Keith et al. 2008, Urban 2015 ) and species 243 interactions (Walther et al. 2002, Wong and Candoli n 2015). Unlike other studies ( Cerd‡ et al. 244 1998, Fitzpatrick et al. 201 3 and 201 4), we did not find that higher temperatures limited ant 245 activity Thus, a 1 3¡C rise ( Collins et al. 2013) may reinforce the ant Acacia mutualism rather 246 than disrupt it. Higher temperatures, especially during the early periods of the day, increase d the 247 quality of defense that the mutualists are supposed to provide and may improve the fitness of the 248 host tree as a result. However, th e fitn ess improvement likely depends on the physiological 249 constraints of the ants and a complicated network of interactions between the different members 250


of the mutualism network (Berg et al. 2010) In order to predict the future of keystone species 251 interactions (such as mutualisms) in the face of climate change, a strong understand ing in the 252 role that temperature plays in each aspect of the interaction s is vital 253


Literature Cited 254 Agrawal, A.A. and M.T. Rutter 1998. Dynamic anti herbivore defense in ant plants: the role of 255 induced responses. Oikos 83: 227 236. 256 Ahn, P.M. and Geiger, L.C. 1987. Soils of Laikipia District. 257 Azc‡rate, F.M., E. Kovacs and B. Peco. 2007. Microclimatic conditions regulate surface activity 258 in harvester ants Messor barbarus. Journal of Insect Behavior 20 : 315 329. 259 Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015 Fitting Linear Mixed Effects Models 260 Using lme4. Journal of Statistical Software 67: 1 48. 261 Berg, M.P., E. Kiers G. Driessen M. Van Der Heijden B.W. Kooi F. Kuenen M. Liefting 262 H.A. Verhoef, and J. Ellers 2010. Adapt or disperse: understanding species persistence in 263 a changing world. Global Change Biology 16 : 587 598. 264 Bolker, B. and R Development Core Team. 2017 bbmle: Tools for General Maximum 265 Likelihood Estimation. R package version 1.0.20. 266 Bronstein, J.L. 2001. The costs of mutualism. American Zoologist 41 : 825 839. 267 Bucy A. and M.D. Breed 2006. Thermoregulatory trade offs result from vegetation removal by 268 a harvester ant. Ecological entomology 31 : 423 429. 269 Cerd‡, X., J. Retana and S. Cros 19 98. Critical thermal limits in Mediterranean ant species: 270 trade off between mortality risk and foraging performance. Functional Ecology 12 : 45 55. 271 Collins, M., R. Knutti, J. Arblaster, J. L. Dufresne, T. Fichefet, P. Friedlingstein, X. Gao, W.J. 272 Gutowski, T. Johns, G. Krinner, M. Shongwe, C. Tebaldi, A.J. Weaver and M. Wehner, 273 2013: Long term Climate Change: Projections, Commitments and Irreversibili ty. Climate 274 Change 2013: The Physical Science Basis 1029 1136 275


Davidson, D. W., and D. McKey. 1993. The evolutionary ecology of symbiotic ant plant 276 relationships. Jou rnal of Hymenoptera Research 2 :13 83. 277 Du Toit, J.T. and C.A. Yetman. 2005. Effects of body size on the diurnal activity budgets of 278 African browsing ruminants. Oecologia 143 : 317 325. 279 Fitzpatrick, G., G. Davidowitz, and J.L. Bronstein. 2013. An herbivore's thermal tolerance is 280 higher than that of the ant d efenders in a desert protection mutualism. Sociobiology 60 : 281 252 258. 282 Fitzpatrick, G., M. C Lanan, and J.L. Bronstein 2014. Thermal tolerance affects mutualist 283 attendance in an ant plant protection mutualism. Oecologia 176 : 129 138. 284 Freeman, B.C. and G.A. Beattie. 2008. An overview of plant defenses against pathogens and 285 herbivores. The Plant Health Instructor 286 Goheen, J. R. and T. M. Palmer 2010. Defensive plant ants stabilize megaherbivore driven 287 landscape change in an African savanna. Current Biology 20:1768 1772. 288 Heil, M. and D. McKey. 2003. Protective ant plant interactions as model systems in ecological 289 and evolutionary research. Annual Review of Ecology, Evolution, and S ystematics 34 : 290 425 553. 291 Herlocker, D.J. 1974. Woody vegetation of the Serengeti National Park. 292 Hocking, B. 1970. Insect associations with the swollen thorn acacias. Ecological Entomology 293 122 : 211 255. 294 Hurlbert, A.H., f. Ballantyne, and S. Powe ll 2008. Shaking a leg and hot to trot: the effects of 295 body size and temperature on running speed in ants. Ecological Entomology 33 : 144 154. 296


Kaspari, M., N.A. Clay J. Lucas S.P. Yanoviak, and A. Kay. 2015. Thermal adaptation 297 generates a d iversity of thermal limits in a rainforest ant community. Global change 298 biology 21 : 1092 1102. 299 Keith, D.A., H.R. Akakaya W. Thuiller G.F. Midgley R.G. Pearson S.J. Phillips H.M 300 Regan M.B. Araœjo, and T.G. Rebelo. 2008. Predicting extinction risks under climate 301 change: coupling stochastic population models with dynamic bioclimatic habitat 302 models. Biology letters 4 : 560 563. 303 Kersch, M.F. and C.R. Fonseca. 2005. Abiotic factors and the conditional outcome of an ant 304 plant mutua lism. Ecology 86 : 2117 2126. 305 Kuznetsova A P.B. Brockhoff and R.H.B. Christensen (2017). lmerTest Package: Tests in 306 Linear Mixed Effects Models. Journal o f Statistical Software 82: 1 26. 307 Milewski, A.V., T.P. Young, and D. Madden. 1991. Thorns as induced defenses: experimental 308 evidence. Oecologia 86 : 70 75. 309 Ness, J. H. 2006. A mutualism's indirect costs: the most aggressive plant bodyguards also deter 310 pollinators. Oikos 113:506 514. 311 Offenberg, J. 2014. Pest repelling properties of ant pheromones. IOBC WPRS Bulletin 9 9: 173 312 176. 313 Palmer, T.M. 2004. Wars of attrition: colony size determines competitive o utcomes in a guild of 314 African acacia ants. Animal Behaviour 68 : 993 1004. 315 Palmer, T.M. and A.K. Brody 2007. Mutualism as reciprocal exploitation: African plant ants 316 defend foliar but not reproductive structures. Ecology 88 : 3004 3011. 317 Palmer, T. M., D. F. Doak, M. L. Stanton, T. P. Young, J. L. Bronstein, E. T. Kiers, J. R. Goheen 318 and R. M. Pringle. 2010. Synergy of multiple partners, including freeloaders, increases 319


fitness in a multi species mutualism. Proceedings of the National Acad emy of 320 Sciences 107:17234 17239. 321 Palmer, T.M. and A.K. Brody. 2013. Enough is enough: the effects of symbiotic ant abundance 322 on herbivory, growth, and reproduction in an African acacia. Ecology 94 : 683 691. 323 Prior, K. M. and T. M. Palmer. Economy of scale: third partner strengthens a keystone ant plant 324 mutualism. Ecology 1 12 325 R Core Team. 2017 R: A language and env ironment for statistical computing. R Foundation for 326 Statistical Computing, Vienna, Austria. 327 Shannon, G., B.R. Page R.L Mackey K.J. Duffy, and R. Slotow. 2008. Activity budgets and 328 sexual segregation in African elephants (Loxodonta africana). Journal of 329 Mammalogy 89 : 467 476. 330 Stanton, M.L., T.M. Palmer T.P. Young A. Evans, and M.L. Turner. 1999. Sterilization and 331 canopy modification of a swollen thorn acacia tree by a plant ant. Nature 401 : 578. 332 Stanton, M. L., T. M. Palmer and T. P. Young. 2002. Competition colonization trade offs in a 333 guild of African Acacia ants. Ecological Monographs 72:347 363. 334 Stanton, M.L. and T.M. Palmer. 2011. The high cost of mutualism: effects of four species of 335 East African ant symbionts on their myrmecophyte host tree. Ecology 92 : 1073 1082. 336 Stapley, L. 1998. The interaction of thorns and symbiotic ants as an effective defence mechanism 337 of swollen thorn acacias. Oecologia 115 : 401 405. 338 Thomas, C.D., A. Cameron R.E. Green, M. Bakkenes L.J. Beaumont Y.C. Collingham B.F. 339 Erasmus M.F. De Siqueira A. Grainger L. Hannah, and L. Hughes. 2004. Extinction 340 risk from climate change. Nature 427 : 145. 341


Urban, M.C., G. Bocedi A.P. Hendry J.B. Mihoub G. Pe'er A. Singer J.R. Bridle L.G. 342 Crozier L. De Meester W. Godsoe and A. Gonzalez. 2016. Improving the forecast for 343 biodiversity under climate change. Science 353 : aad 8466. 344 Walther, G.R., E. Post P. Convey A. Menzel C. Parmesan T.J. Beebee J.M. Fromentin O. 345 Hoegh Guldberg, and F. Bairlein. 2002. Ecological responses to recent climate 346 change. Nature 416 : 389. 347 Wong, B. and U. Candolin. 2015. Behavioral responses to cha nging environments. Behavioral 348 Ecology 26 : 665 673. 349 Wood, W. F., T. M. Palmer and M. L. Stanton. 2002. A comparison of volatiles in mandibular 350 glands from three Crematogaster ant s ymbionts of the whistling thorn 351 acacia. Biochemical Systematics and E cology 30:217 222. 352 Wood, W. F., T. M. Palmer and M. L. Stanton. 2006. Volatiles in the mandibular gland of 353 Tetraponera penzigi: a plant ant of the whistling thorn acacia. Biochemical Systematics 354 and Ecology 34:536 538. 355 Young, T. P., C. H. Stubblefield, and L. A. Isbell. 1997. Ants on swollen thorn acacias: Species 356 coexistence in a simple system. Oecologia 109:98 107. 357 Young, T. P., B. D. Okello, D. Kinyua, and T. M. Palmer. 1998. KLEE: a long term multi 358 species herbivore exclusion experiment in Laikipia, Ken ya. African Journal of Range 359 Forage Science 14:94 102. 360


361 Figure 1 Outer canopy s urface temperatures (¡C) throughout the day on A drepanolobium 362 occupied by any of the four ant symbionts Temperatures were measured to the nearest degree, 363 so each point in the figure can represent several observations. Boxplots repres ent the interquartile 364 range s of temperature during each measurement period. 365 10 20 30 40 50 Time Surface Temperature (¡C) 4:306:30 6:318:30 8:3110:30 10:3112:30 12:3114:30 14:3116:30 16:3118:30 18:3120:30


366 Figure 2 : Predictions of o ur best model for the r elationship between surface temperature and 367 baseline activity of an ant colony (left) and defensive response to a simulated herbivore (right). 368 Baseline activ ity was measured by the numbe r of ants passing in or out of two 5cm segment s of a 369 branch for 30 seconds The positive relationship between baseline and activity was not 370 significant once we controlled for colony identity and measurement periods. The defensive 371 response was measured by the number of ants swarming onto two glove s after the tree was 372 disturbed The positive relationship between temperature and defensive activity was significant 373 while controlling for colony identity and meas urement periods. 374 0 50 100 150 10 20 30 40 50 Temperature (¡C) Total Ants Attacking Species_Name C. mimosae C. nigriceps C. sjostedti T.penzigi Predicted incident rates 0 10 20 30 10 20 30 40 50 Temperature (¡C) Total Ants Species_Name C. mimosae C. nigriceps C. sjostedti T.penzigi Predicted incident rates Baseline Ant Activity (Total Ants)


375 Figure 3 Predictions for the relationship between surface temperature and goat feeding behavior. 376 Goats fed for up to five minutes from trees with ants (occupied trees) and trees where the ants 377 were removed (removal trees) at three different times Goats took more bites from removal trees 378 compared to occupied trees as temperature increased (left) measured by the difference of bites 379 taken from rem oval trees and occupied trees Goats also spent more tim e feeding from removal 380 trees compared to occupied trees as temperature increased (right), measured by the the difference 381 in feeding time of removal trees and occupied trees Feeding time was measured as the time from 382 the first bite to the time when the goat refused to continue feeding. Both relationships were 383 significantly positive when controlling for the individual goat and tree pair. 384 0 50 100 150 200 250 15 20 25 30 Temperature (¡C) Difference in Feeding Time (Removal-Occupied) Predicted values 0 25 50 75 100 125 15 20 25 30 Temperature (¡C) Difference in Bites Taken (Removal-Occupied) Predicted values