Brian G. Gall1, Kari L. Spivey2, Trevor L. Chapman3, Robert J. Delph4, Edmund D. Brodie

1 The Indestructible Insect: Velvet Ants from Across the United States Avoid Predation by

2 Representatives from all Major Tetrapod Clades.

4 Brian G. Gall1, Kari L. Spivey2, Trevor L. Chapman3, Robert J. Delph4, Edmund D. Brodie,

5 Jr.5, Joseph S. Wilson6

7 1Department of Biology, Hanover College, 517 Ball Drive, Hanover, IN 47243,

8 [email protected]

9 2Department of Biology, Missouri State University, 901 S. National Ave., Springfield, MO

10 65897

11 3Department of Biology, East Tennessee State University, 325 Treasure Lane, Johnson City, TN

12 37604

13 4Department of Natural Resources, U.S. Army Dugway Proving Ground, 5330 Valdez Circle

14 Dr., Dugway, UT 84022

15 5Department of Biology, Utah State University, 5305 Old Main Hill, Logan, UT 84322-5305 16

17 6Department of Biology, Utah State University-Tooele, 1021 West Vine Street, Tooele, UT

18 84074

26 Abstract

27 To determine whether birds might prey on velvet ants native to the eastern United States

28 (Dasymutilla vesta & Dasymutilla occidentalis), we trained birds to eat wax-worms from

29 feeders. We then exposed these birds to mealworms painted to resemble velvet ants, clay models,

30 or live velvet ants, and recorded their behavior. Eastern bluebirds and mockingbirds readily

31 learned to eat from the feeders and both species consumed mealworms painted tan but avoided

32 aposematically painted mealworms. Mockingbirds also attacked clay models painted black but

33 would not strike at models painted to resemble a velvet ant. Trials with live velvet ants elicited

34 strong avoidance behavior by mockingbirds and no individuals would approach the dish.

35 Bluebirds that were presumably naïve did strike at live velvet ants, however none were

36 consumed and all velvet ants were unharmed. The results of these experiments suggest that birds

37 may be a limited threat to these extremely well-defended organisms.

40 Introduction

41 Predation is an extremely powerful selective force driving the evolution of morphology,

42 physiology, and behavior among animals (Brodie et al., 1991; Endler, 1986; Lima and Dill,

43 1990). Because of the intense nature of the interaction (prey either escape to live another day or

44 die), it has resulted in a bewildering array of defensive structures and strategies to mitigate this

45 risk. Extreme examples include venomous frogs (Jared et al., 2015), salamanders with ribs that

46 pierce their skin (Brodie et al., 1984; Nowak and Brodie, 1978), beetles with rear rotary turrets

47 ejecting toxins at 100°C (Aneshansley et al., 1969; Arndt et al., 2015), and ouabain resistant

2 48 rodents with skeletons designed to take a punch (Kingdon et al., 2011). Regardless of the

49 defensive strategies utilized by prey, each is used during one of two distinct stages along the

50 chain of a predatory interaction (Endler, 1986; Hopkins et al., 2011); either before a predation

51 event has been initiated (predator-avoidance behavior) or after a predator has detected the

52 presence of its prey (antipredator mechanisms) (Brodie et al., 1991).

53 Despite prey being well-defended, predators must eat, and a similar diversity of

54 mechanisms have evolved to help predators acquire their prey. For example, the terminal scales

55 on the spider-tailed viper (Pseudocerastes urarachnoides) have evolved to be flexible and is uses

56 caudal luring to attract insectivorous birds which it envenomates and eats (Fathinia et al., 2015).

57 The lower jaw of dragonfly naiads has evolved into a protrudable grasping mouthpart allowing

58 the sit-and-wait predators to strike at prey with amazing speed (Needham and Westfall, 1954).

59 Apart from some apex predators, few organisms are likely to completely escape predation, and

60 even those prey which are so well-defended that it seems nothing can eat them have been found

61 to have a specialized predator (e.g. Brodie, 1968). One organism that possesses a myriad of

62 defensive structures and behaviors and for which its risk to potential predators is largely

63 unknown are velvet ants (order: Hymenoptera; family Mutillidae). Velvet ants are members of a

64 unique wasp family, Mutillidae. Their common name stems from the extensive setae that can

65 cover their entire body (fig 1) and the fact that the females are wingless, making them look like

66 ants. Although the taxonomic relationships within this group are beginning to be unraveled

67 (Williams, 2012), very little is known about their ecology (cite). Velvet ant females spend much

68 of their time actively searching for the nests of ground-nesting bees and wasps (cite). After

69 finding a host’s nest, the female velvet ant deposits an egg on or near the host pupae, which the

70 larvae consume after hatching (cite).

3 71 Given their flightlessness, one would expect diurnal females of this group to be highly

72 susceptible to predation. Yet, velvet ants have a number of defenses at their disposal to thwart

73 potential predators. Like other Aculate wasps, females are armed with a venomous sting, though

74 unlike most other wasps, velvet ants sting can be nearly half the length of their body. Although

75 the composition of the venom is unknown, it can be extremely painful (Schmidt, 1990; Schmidt

76 et al., 1984; Starr, 1985), which is often evident in their common names (e.g. cow killer). On a

77 human pain index, at least one velvet ant (Dasymutilla klugii) outscored 58 species of wasps and

78 bees in the painfulness of its sting, falling short of only the bullet ant, warrior wasp, and tarantula

79 hawk in the amount of pain induced (Starr, 1985). The venomous nature of the females is

80 complemented by the striking aposematic coloration present of almost all diurnal species (Fig.

81 1). This coloration comes in various shades of white, orange, yellow, or red. Different

82 colors/patterns corresponds to a specific Müllerian mimicry ring consisting of dozens of species

83 (Wilson et al., 2012). These rings are extensive, with eight distinct rings making up one of the

84 largest Müllerian mimicry complexes on earth (Wilson et al., 2015)

85 Although being dangerous and advertising this danger with bright colors is an obvious

86 antipredator mechanism, velvet ants possess several other defensive structures and behaviors.

87 When distressed, a stridulatory organ on their abdomen is contracted which produces audible

88 squeaking (Schmidt and Blum, 1977; Tschuch, 1993), and an alarm secretion may be released

89 from the mandibular gland (Fales et al., 1980; Schmidt and Blum, 1977). These function as

90 auditory and chemosensory aposematism, warning potential predators that if they continue an

91 attack, a sting is imminent. The exoskeleton of velvet ants possesses two properties that make it

92 an effective defense against predators. First, the exoskeleton is remarkably strong. Using a force

93 transducer, Schmidt and Blum (1977) calculated 11 times more force was needed to crush the

4 94 exoskeleton of a velvet ant as opposed to a honeybee (Apis). The rounded shape of the

95 exoskeleton also renders attacks more difficult as attempted stings or bites glance off the

96 abdomen instead of piercing it (Schmidt and Blum, 1977).

97 Despite the suite of defenses possessed by velvet ants (primarily females), relatively little

98 is known about their relationships with potential predators or the pressures that may have driven

99 the evolution of these defenses. Schmidt and Blum (1977) conducted a series of studies with

100 Dasymutilla occidentalis and various potential predators. In this seminal work, ants, spiders,

101 lizards, and gerbils were presented velvet ants. Yet, only 2 out of 59 presentations resulted in the

102 consumption of a velvet ant by any predator; once by a tarantula and another by a gerbil. In most

103 cases, the velvet ants were either ignored from the start, or, were attacked, released, and

104 eventually left unscathed.

105 Given the limited information on potential predators of velvet ants, we conducted a series

106 of observational and experimental studies with a host of potential invertebrate and vertebrate

107 predators. The predators were selected based on the potential for natural interactions, the high

108 probability of contact with diurnal velvet ants, and dietary overlap (i.e. insectivorous). The

109 predators include: a praying mantis, toads, lizards, birds, shrews, and a mole. Experiments were

110 conducted with velvet ants from both the eastern and western United States (i.e. multiple

111 mimicry rings), with predators selected that are representative of the appropriate region. Because

112 the predators and velvet ants used in all experiments were wild-caught, the experience of each is

113 generally unknown.

115 Methods

116 Birds

5 117 General Description

118 All experiments took place in a manicured yard (0.4 ha) located in a rural setting near

119 Hanover, Indiana. Two feeding stations were attached to previously established bluebird nesting

120 boxes. Each feeding station consisted of a 15 cm diameter petri dish glued to the box such that

121 birds naturally perching on top of the box would see the dish and inspect the contents. Two 2 MP

122 digital trail cameras (Wildgame Innovations) were either affixed to a post approximately 1 m

123 from the feeding station or were mounted directly to the box.

124 Birds were initially trained to forage at the feeders by placing four wax moth larvae

125 (Galleria mellonella) in each petri dish at 0700 hrs each day for one week until testing began.

126 Days in which the birds were fed wax moth larvae will henceforth be called “training” days. The

127 photos from the trail cameras were obtained on the third day of training and reviewed to ensure

128 that the birds were feeding from the dishes.

129 On “test” days, behavioral observations were conducted from a deck located 10 m and 20

130 m from each of the respective feeding stations. A Nikon spotting scope (15-45x) and Bushnell

131 7x50 handheld binoculars were used to observe the feeding stations. The general procedure on

132 test days consisted of placing the appropriate experimental subject (see below) in the petri dishes

133 at 0700 hrs and recording observations for 40 minutes. We recorded the species of bird visiting

134 the feeder, the general behavior of the bird toward the subjects in the dish, the type (control or

135 experimental animal) and number of experimental subjects struck at, and the type and number of

136 experimental subjects consumed. A minimum of two training days followed each test day. We

137 exposed wild-birds to the following treatments to determine if birds are potential predators of

138 velvet ants (experiments were conducted in the order presented): (1) mealworms (Tenebrio

139 molitor) or preserved female velvet ants (Dasymutilla vesta), (2) mealworms painted tan or with

6 140 the aposematic coloration of Dasymutilla occidentalis, (3) clay models painted black or with the

141 aposematic coloration of Dasymutilla occidentalis, or (4) live velvet ants (Dasymutilla

142 occidentalis). Dasymutilla occidentalis and Dasymutilla vesta are both members of the Eastern

143 mimicry ring, and therefore, have very similar coloration patterns.

145 Preserved Velvet Ants

146 To determine if birds are potential predators of velvet ants, we exposed wild-birds to

147 either mealworms (Tenebrio molitor) or preserved female velvet ants (Dasymutilla vesta).

148 Pinned velvet ants were collected between 1951-1970 and were provided by the Utah State

149 University Insect Collection. The velvet ants were rehydrated by placing them in a sealed plastic

150 container on paper towels moistened with tap water. After 48 hrs, the velvet ants were removed

151 from the containers and the limbs, head, and antennae were repositioned so that the velvet ant

152 appeared in a normal crawling posture. After re-positioning the velvet ants, they were re-pinned

153 to polystyrene foam and left to dry for several days.

154 On test days, a feeding station was randomly assigned to receive either four velvet ants or

155 four mealworms. Pins were removed from the velvet ants before placing them in the feeder. Only

156 complete specimens (i.e. not missing appendages or antennae) were used during the experiment.

157 A total of four replicates were conducted on separate test days.

159 Painted Mealworms

160 Four mealworms were painted tan and four mealworms were painted red and black to

161 simulate the aposematic coloration pattern of the velvet ant, Dasymutilla occidentalis (figure 2).

162 We used a non-toxic and water-soluble acrylic paint that did not prevent the mealworms from

7 163 moving normally. Two mealworms of each color pattern were added to each of the feeding

164 stations and observations recorded for 40 minutes. A total of two replicates on separate test days

165 were conducted.

167 Live Velvet Ants

168 The final experiment involved testing the responses of birds to live velvet ants. Two

169 female velvet ants (D. occidentalis) were collected near Hanover, IN and were housed in a 7.5 L

170 plastic tank with a screen lid. A 2 cm layer of sand was placed in the bottom of the tank. A dried

171 leaf (Magnolia grandifloria) was placed in the tank to provide cover. Fresh grapes were cut in

172 half and placed in the tank to provide a source of water and sugar until testing (1 week). To

173 ensure the velvet ants did not escape from the feeding dishes, we attached glass preparation

174 dishes (11 cm diameter x 4 cm deep) to the feeding stations. The velvet ants were removed from

175 the holding container by pushing them into a 25ml centrifuge tube and dumping them directly

176 into the glass dish; this procedure was used to ensure the velvet ants were not exposed to a

177 simulated predation event (i.e. grasping with forceps). Each feeding station had one live velvet

178 ant. At the completion of testing, mealworms were placed in the glass dishes to ensure the birds

179 were hungry. All mealworms were consumed shortly after being presented. One replicate was

180 conducted.

182 Mole

183 A mole (Scalopus aquaticus) was collected in 2014 in a field on the Hanover College

184 campus. Fresh burrows were monitored during the morning and evening, and a dog (Canis lupus;

185 terrier) was used to initially locate moles in their burrows. Upon detecting a mole, a researcher

8 186 removed the mole with a shovel and transported it to the lab in a 19-l container. Housing

187 chambers were designed so that there was a designated feeding area adjacent to a burrowing

188 chamber. The large section of the housing unit consisted of a container (55 × 35 × 30 cm) filled

189 with 20 cm of dry soil for burrowing. A feeding chamber (18 × 10 × 10 cm) was attached to the

190 burrowing chamber with a PVC tunnel (20 cm long, 5 cm diameter). The feeding chamber did

191 not contain soil, and any soil displaced into it by the mole was removed and placed into the

192 burrowing section. Moles were fed moist cat food every 24 hours.

193 For testing, the mole was transferred to a test arena consisting of two chambers (11 × 11

194 × 16 cm) connected by a clear tube (30 cm long, 6 cm diameter). The arena was left empty. After

195 transferring the mole to the test arena, a 5-minute acclimation period was initiated. Following the

196 acclimation period, a velvet ant was introduced into the arena and observations were recorded for

197 25 minutes. At the conclusion of the trial, a control cricket was introduced and immediately

198 consumed.

200 Shrews

201 Shrews (Blarina brevicauda, n = 4) were collected on 17 November 2014 in a wooded

202 area on the Hanover College campus. Sherman live traps (HB Sherman Traps, Inc.) were baited

203 with wet cat food and were placed in close proximity to burrows and logs throughout the forest.

204 Traps were checked every three hours and captured shrews were placed in a 19-l container and

205 transported to the lab. Individuals were housed in 38 L chambers with a 2-inch layer of dry soil

206 (collected from campus), strips of cotton cloth, and a water dish. Shrews were maintained on a

207 diet of moist cat food and fed every 24 hours. A single shrew (Notiosorex crawfordi; henceforth

9 208 “UT shrew”) was collected from Dugway Proving Ground, Utah, and housed under similar

209 conditions.

210 For experimental trials, the shrews were placed in 38 L aquaria that were completely

211 empty. The shrews were allowed to acclimate for 5 minutes, after which a velvet ant was

212 introduced. Detailed observations were then recorded for approximately 20 minutes, after which

213 the velvet ant was removed and a control cricket was introduced into the chamber. For the

214 experimental trial with the UT shrew, the velvet ant’s stinger was removed with forceps. Shrews

215 were tested only once and were given a control cricket at the completion of the trials. All control

216 crickets were immediately consumed.

218 Toads

219 A single American toad (Anaxyrus americanus) was collected on Hanover College’s

220 campus and housed in a 38 L aquaria with damp sphagnum moss. The toad was not fed until

221 testing (2 days). For testing, the American toad was transferred to an empty 38 L tank and

222 presented a velvet ant for 20 minutes. At the conclusion of testing, the toad was presented, and

223 immediately consumed a cricket.

224 Two Great Basin spadefoot toads (Spea intermontana) were collected from Dugway

225 Proving Ground and housed individually in 150 L tanks. Each toad was presented (in its home

226 tank) a velvet ant (either Sphaeropthalma mendica or Dasymutilla scitula) on two separate

227 occasions. The testing days were separated by at least 3 days. After each trial the toads each

228 consumed a cricket.

230 Lizards

10 231 Between 2013 and 2015, we collected lizards [Aspidoscelis tigris (n = 6), Uta

232 stansburiana (n = 3), Gambelia wilzenii (n = 2) from Dugway Proving Grounds, UT, to test the

233 antipredator defenses of various species of velvet ant. Lizards were collected with pitfall traps

234 and housed in 227 L tanks with sand substrate, a water dish, and various natural elements (sticks,

235 rocks, etc.). Each tank had a UVB daytime heat lamp (Exo Terra) and a heat rock (24 hrs).

236 Lizards were fed commercially available crickets (Acheta domestica) and mealworms (Tenebrio

237 molitor) ad libitum. Prior to testing, lizards were in captivity between 2 weeks to 2 years, with

238 most between 4-12 months. Food was withheld from each lizard for 3 days prior to testing. The

239 responses of each lizard to velvet ants were conducted in the lizard’s home tank to reduce

240 handling effects. On test days, trials were conducted at 0800 hrs and consisted of a single velvet

241 ant (various species, Table 2) being dropped into the tank. Observations were recorded for 5 min,

242 after which the velvet ant was removed and a control cricket was introduced. Each lizard quickly

243 consumed a control cricket at the completion of the trial. In addition to each initial trial with a

244 lizard, a series of “secondary” trials were also conducted with various species of velvet ants.

245 These trials were conducted at least one day following each primary trial. Results of the

246 secondary trials are discussed separately from the initial trials (see below).

247 In addition to the predation trials conducted in captivity, two semi-natural trials were

248 conducted. In the first, a velvet ant (Dasymutilla scitula) was placed in a glass dish (with lid) and

249 set in the open in a sandy area LOCATION. In a second trial, a velvet ant (Dasymutilla foxi) was

250 tied to a small thread and staked in the ground in an open area LOCATION. Observations were

251 recorded for 1.5 hrs from approximately 10 m away.

253 Praying Mantis

11 254 A European mantis (Mantis religiosa) was collected on 30 September 2013 on the Utah

255 State University – Tooele campus and housed in a 3.8L glass aquarium with a screen lid. A 2 cm

256 layer of sand was placed in the bottom of the tank and small rock and some twigs were added to

257 provide a place for the mantis to perch. A single velvet ant (Dasymutilla vestita) was placed into

258 the tank and no other food source was provided. The tank was sprayed with water every day to

259 provide water and the enclosure was placed near an east facing window so the insects would

260 receive morning sun. Observations were made daily to determine if the velvet ant had been

261 consumed.

263 Results

264 Birds

265 Preserved Velvet Ants

266 Mockingbirds were the only species to visit the feeding station during observations. At

267 least 4, and likely 5, separate mockingbirds were seen foraging at the stations throughout the

268 experiment (i.e. multiple birds were visible in the same field of view). Mockingbirds exhibited

269 significantly more strikes at mealworms (n = 16) than preserved velvet ants (n = 1; χ2 = 13.2, P <

270 0.001). A single mockingbird did exhibit one strike at a preserved velvet ant, however it was

271 immediately dropped and not consumed. All strikes on the mealworms were immediately

272 followed by consumption (n = 6), or the mealworm was held in the beak and carried away from

273 the feeder (n = 10); in these cases the birds flew out of view and, although they were likely

274 consumed, their fate is unknown. If these mealworms are categorized as consumed, the

275 mockingbirds consumed significantly more mealworms (n = 16) than preserved velvet ants (n =

276 0; χ 2 = 16.0, P < 0.001).

12 277

278 Painted mealworms

279 The mockingbirds consumed more tan painted mealworms (n = 4) than mealworms

280 painted with the Dasymutilla aposematic color pattern (n = 0; χ 2 = 4.0; P = 0.045). However, the

281 mockingbirds exhibited more strikes at aposematically painted mealworms (n = 13) than tan

282 painted mealworms (n = 5; χ 2 = 3.55; P = 0.06). Three of the four tan-colored mealworms were

283 consumed immediately by the mockingbirds. One mealworm was struck and dropped before

284 being picked up and consumed. Despite receiving more strikes than neutrally colored

285 mealworms, the mockingbirds appeared hesitant to feed on the aposematically painted

286 mealworms and none were consumed over the course of the trials. One bird tilted its head so as

287 to visually inspect the dish, got approximately 15 cm from the mealworm, and retained this

288 posture for 30 seconds. The bird then struck at an aposematic mealworm and carried it to the

289 ground 6 m from the feeding station. Later inspection found a damaged, but uneaten, aposematic

290 mealworm at this location. The mealworm had an “open wound” on the dorsal side of where the

291 head would normally be on a live velvet ant/mealworm. Another aposematic mealworm was

292 inspected, struck, and dropped a total of six times before being carried to the ground

293 approximately 20 m from the feeding station. The bird then appeared to peck vigorously at the

294 worm for several seconds before flying away. Later inspection of the site discovered a

295 mealworm that had been “decapitated” but which was otherwise unharmed and uneaten (Fig X).

296 No aposematically colored mealworms were consumed during any trial.

298 Live Velvet Ants

299 Birds

13 300 During trials with live velvet ants, mockingbirds appeared hesitant to visit the feeders.

301 The birds landed on top of the feeding station, glanced at the dish, and flew away. This behavior

302 had not been observed with any other trials; mockingbirds typically landed next to the dish and

303 inspected the contents before ignoring or striking the available prey. No strikes were exhibited

304 toward the live velvet ants by mockingbirds, however, control mealworms were immediately

305 consumed at the conclusion of the trial.

306 In addition to mockingbirds, at least 5 separate juvenile bluebirds also visited one of the

307 feeding stations during the trial. On four occasions the birds landed on top of the station,

308 inspected the dish, but flew away without approaching. On one occasion, a bird landed next to

309 the dish, inspected the velvet ant, and flew away. A fifth bluebird landed on the edge of the dish

310 and struck a live velvet ant twice on the thorax. The velvet ant was visibly struck because it

311 became flattened against the bottom of the glass dish. However, the bird did not grasp the velvet

312 ant in its beak and, given the lack of visual distress, it is doubtful whether the bird was stung

313 during the interaction; it is unknown whether the velvet ant stridulated during the interaction.

315 Mole

316 The mole attacked the velvet ant once during the interaction. After the initial attack, the

317 velvet ant appeared to escape unharmed and the mole did not appear to be stung by the velvet

318 ant. Shortly after, the velvet ant and mole passed through the central tube simultaneously and got

319 “wedged” together inside the tube. After a few seconds, the mole began thrashing wildly and

320 appeared to be stung by the velvet ant. After retreating to opposite chambers, the mole began

321 rubbing the area where the velvet ant had previously been wedged and where the mole had

14 322 presumably been stung. After these initial interactions, the mole and velvet ant came in contact 4

323 separate times. Each time, the mole recoiled and rapidly retreated from the velvet ant.

325 Shrew

326 After introducing a Dasymutilla vesta to a short-tailed shrew, the shrew vigorously

327 sniffed the velvet ant and struck it. However, the velvet ant was rejected. It is unknown if the

328 shrew was stung. The shrew rapidly moved about the chamber exhibiting escape behavior until

329 the end of the trial. During an interaction between a short-tailed shrew and a Dasymutilla

330 occidentalis, the velvet ant stridulated upon contact with the shrew 5 separate times. The velvet

331 ant was never attacked. In another trial with a short-tailed shrew, the velvet ant was attacked 7

332 separate times in the first five minutes of the trial. Each time the velvet ant stridulated and was

333 released. On the eighth attack, an audible crack was heard after which the velvet ant was flung

334 across the chamber and repeatedly attacked. After a series of attacks, the shrew paused and

335 appeared irritated. The right front paw was enlarged and the shrew continually licked and

336 chewed at this paw (presumably stung). At the completion of the trial the velvet ant was still

337 alive and was inspected for injuries. A small patch of setae was discolored on the abdomen, but

338 no puncture in the exoskeleton was visible. During the final shrew-velvet ant trial, the velvet ant

339 was bitten on the posterior portion of the thorax. An audible crack was heard during this strike.

340 The velvet ant immediately stridulated and the velvet ant was dropped; the shrew did not appear

341 to be stung. Shortly after, the velvet ant was struck again, during which the shrew was stung in

342 the mouth and the velvet ant was dropped. The velvet ant was attacked 6 separate times after this

343 event. After these attacks, the velvet ant’s stridulations became inconsistent and could not walk.

344 The shrew began itching its head and side of the neck vigorously, as well as biting its right-front

15 345 paw. Any further contact between the velvet ant and shrew resulted in avoidance. All velvet ants

346 survived the interactions with shrews.

347 When a velvet ant (Dasymutilla bioculata – sting removed) was introduced to the UT

348 shrew, it immediately attacked the velvet ant, dropped it after the velvet ant stridulated, and ran

349 to the opposite side of the test chamber. The velvet ant’s exoskeleton was slightly cracked, but

350 the velvet ant survived the interaction.

352 Toad

353 When presented with a velvet ant (Dasymutilla occidentalis), the American toad hoped

354 toward the velvet ant and upon contact inflated its lungs, dropped a shoulder, and closed the eye

355 closest to the velvet ant. The toad remained in this position until the velvet ant was removed (~20

356 min).

357 Upon the initial interactions with a velvet ant, each Spadefoot toad attacked and

358 swallowed a velvet ant. However, in each case the velvet ant was quickly regurgitated, which

359 was followed by the toad wiping its hands over its tongue multiple times. Both velvet ants were

360 unharmed. During the second set of interactions, both toads avoided the velvet ants completely.

362 Lizards

363 Among the three species of lizards, and 12 independent trials, only two lizards (one

364 whiptail, one side-blotched lizard) attacked a velvet ant (Table 3). In each case the lizard was

365 stung in the face and quickly dropped the velvet ant, after which it avoided the velvet ant. The

366 velvet ants were unharmed in each case. Twenty-four hours following the initial trial with the

367 side-blotched lizard described above, the animal was found dead in its tank with a noticeable

16 368 discoloration on the head where it had been stung. The remaining lizards either ignored the

369 velvet ant completely (n = 4), or approached the velvet ant (n = 6). Approaching the velvet ant

370 was followed by avoidance (n = 2), tongue flicking (n = 1), or nudging the velvet ant with their

371 snout (n = 3). In 59 secondary trials with these same lizards, only 4 strikes were exhibited. In

372 each case, the lizard was one that had not previously struck a velvet ant (i.e. had not been stung).

373 One strike by a leopard lizard resulted in the lizard swallowing the velvet ant. However, the

374 lizard immediately regurgitated the velvet ant and exhibited avoidance; it is unknown if the

375 lizard was stung on the inside of the mouth. Across all 71 trials, no velvet ant was injured or

376 killed during an interaction with a lizard.

377 While most secondary trials (where lizards that had previously been exposed to velvet

378 ants) took place within a week of the initial trial, the one whiptail lizard that attacked the velvet

379 ant and was stung in the face was re-exposed to a velvet ant 15 months later. This whiptail

380 closely watched the velvet ant, but did not attempt to attack it.

381 In the semi-natural trials, a single lizard (Aspidoscelis tigris) approached the glass dish,

382 nudged the lid off the dish and grabbed the velvet ant. It then immediately ran under a nearby

383 bush, dropped the velvet ant, and ran away. The velvet ant was observed crawling into a burrow

384 under the bush and neither the velvet ant or lizard were recovered. In the second trial, a single

385 lizard approached the snared velvet ant, tongue flicked it several times, and then avoided the

386 velvet ant.

388 Praying Mantis

389 No observed attacks on the velvet ant were made until October 10, 2013. At

390 approximately 1100 hrs the velvet ant was observed sitting still on a rock sunning itself when the

17 391 mantis approached slowly and grabbed it. The mantis immediately attempted to bite the velvet

392 ant on the thorax but it looked like the mantis was unable to bite through the velvet ants cuticle

393 on the edge of its thorax (this was later confirmed by examining the velvet ant under a

394 microscope and there was no visible sign of damage of any kind). As soon as the mantis grabbed

395 the velvet ant, the velvet ant began attempting to sting the mantis but it looked like the sting

396 couldn’t easily penetrate the cuticle of the mantis' forelimbs. The velvet ant also began

397 stridulating and continued to do so even after it was released. The mantis held the velvet ant for

398 about 5 seconds then suddenly released it. It was unclear if velvet ant was able to sting the

399 mantis but following the release of the velvet ant the mantis began grooming itself for about 5

400 seconds, focusing primarily on the arm that was closest to the velvet ant’s sting. About 15 min

401 after the encounter with the velvet ant the mantis was offered a moth, which it immediately

402 consumed. Approximately 24 hours after the trial with the velvet ant, the mantis was offered a

403 yellow jacket (Vespula vulgaris), which it immediately consumed.

405 Discussion

406 The results of this study indicate that velvet ants from both the Eastern and Western

407 United States possess a myriad of defenses that render them almost invulnerable to a suite of

408 potential predators including an insect, amphibians, reptiles, birds, and small mammals. The

409 predators selected were chosen based on a high probability of interaction and dietary overlap that

410 would make interactions between these species highly likely in the wild. Nevertheless, out of

411 over 100 interactions between potential predators and various species of velvet ant, velvet ants

412 were struck at only 15 separate times and the three velvet ants that were eaten were immediately

413 regurgitated.

18 414 The birds that visited our feeders during this study forage heavily on insects (Beal, 1915;

415 Cottam and Knappen, 1939), yet all birds appeared extremely wary around both live and dead

416 velvet ants, as well as models and mealworms painted to resemble velvet ants. These same birds

417 foraged immediately upon control mealworms. A similar avoidance response was observed by a

418 single starling (Sturnus vularis) in a trial by Schmidt and Blum (1977). While the experience of

419 our birds is unknown, work with the aposematic color patterns of snakes indicates that these

420 patterns (red/yellow/black) are avoided by avian predators (Brodie, 1993; Brodie and Janzen,

421 1995) and that this avoidance is innate in at least one species of neotropical bird (Smith, 1975).

422 Studies with invertebrate prey are more ambiguous and both innate and learned avoidance of

423 aposematic patterns has been observed (Coppinger, 1970; Exnerová et al., 2006; Svádová et al.,

424 2009). The bluebirds visiting our feeders had recently fledged (juvenile plumage; likely the same

425 birds that had recently fledged from the box making up the feeding station). Yet, with the

426 exception of one strike, even these young birds avoided the velvet ants. Interestingly, mealworms

427 painted with aposematic coloration matching velvet ants did receive more strikes than plain

428 mealworms and two were decapitated but left uneaten. Partially eating or seizing and pecking at

429 newly discovered distasteful prey occurs in some birds (Wiklund and Järvi, 1982), and these

430 results suggest the mockingbirds were experienced with insect warning coloration but may not

431 have had prior experience with velvet ants.

432 Similarly to birds, various species of lizards were wary around the velvet ants and no

433 velvet ant was injured or eaten by these lizards out of 71 total interactions. Even in field trials

434 with tethered velvet ants, none were consumed. These results were surprising given the diurnal

435 activity patterns, stout head and jaws, and insectivorous nature of the lizards tested. The natural

436 history of both predator and prey in this case likely brings both species into contact frequently,

19 437 yet lizards do not appear to be predators of velvet ants. Schmidt and Blum (1977) tested lizards

438 from Florida with local velvet ants and while some did attack, all velvet ants were released

439 unharmed. Similarly, two horned lizards (Phrynosoma cornutum), which regularly prey upon

440 unpalatable ants, consumed a cryptically colored Dasymutilla dilucida but ignored three species

441 of aposematic velvet ants (Manley and Sherbrooke, 2001). The broadhead skink (Plestiodon

442 laticeps) is the only lizard to have successfully consumed velvet ants during experimental trials.

443 These occurred after repeated failed attacks (up to 23), during which an interaction in the wild

444 would have likely resulted in the velvet ants successful escape (Vitt and Cooper, 1988).

445 Predator avoidance and antipredator defenses are used at different points during

446 interactions between predators and prey. This sequence occurs from approach and identification

447 to the eventual subjugation and consumption of the prey (Endler, 1986; Hopkins et al., 2011). Of

448 the specific defenses present in velvet ants, each can function at different stages of the predator-

449 prey interaction, thus maximizing the probability of surviving the interaction (as prey move

450 further along in the interaction the probability of survival decreases). In addition, the role of a

451 particular defense is also dependent on the particular predator type. For example, almost all the

452 birds and many of the lizards tested avoided the velvet ants immediately upon sight of the

453 warning coloration; birds and lizards are visually oriented predators (cite). While shrews are

454 well-known to be voracious predators (e.g. Brodie et al., 1979), they have poor vision (cite) and

455 all but one shrew attacked the velvet ants, many multiple times. In some of these cases,

456 stridulation (and possibly release of a chemical signal) was enough to cause the release of the

457 prey. However, in most cases the interaction escalated and envenomation was required to prevent

458 predation; all shrews eventually exhibited avoidance.

20 459 Velvet ants appear to possess an effective suite of defense mechanisms; a hard and

460 slippery exoskeleton, venom, warning chemicals and sounds, rapid escape behavior, and bright

461 coloration. While these are common defenses among animals (Endler, 1986), this combination

462 appears to make velvet ants almost immune to predation. The pressure to evolve this suite of

463 defenses was likely intense, and the diurnal and flightless nature of the females may have played

464 a role in this evolution. While the observations presented here provide strong evidence that these

465 adaptations function in defense, function is not always responsible for the form, and the

466 dangerous nature of their hosts must not be overlooked (Deyrup, 1988). Female velvet ants

467 parasitize ground dwelling bees and wasps (best citation?), and the size and strength of their

468 exoskeleton also prevents penetration by the biting and stinging insects they parasitize (Brothers,

469 1972). Further, the relatively rare and scattered nature of the host nests requires females to spend

470 extensive time searching for hosts, leaving females vulnerable to predation throughout this

471 duration and possibly leading the evolution of some of these defenses (e.g. stridulations, venom)

472 (Deyrup, 1988).

473 Schmidt and Blum (1977) suggest that velvet ants may have evolved different defenses in

474 response to different predators. While that may be true, our observations indicate that it is the

475 combination of these defenses that enable velvet ants to be so successful. For example, we find

476 that when an inexperienced lizard first encounters a velvet ant and attacks it, the hard slippery

477 cuticle of the velvet ant stops the lizard from immediately crushing its intended prey. The lizard

478 then attempts to manipulate the velvet ant in its mouth, which gives the velvet ant time to sting

479 the lizard in the mouth. This painful sting causes the lizard to release the velvet ant, where it is

480 immediately exposed to both the aposematic colors and the striduations. The sting, accompanied

481 by the warning coloration and sounds appear to provide an effective deterrent to future predation

21 482 events, in our trials after one failed predation attempt a lizard still avoided the velvet ant even

483 after 15 months with no reinforcement of the signal.

484 The extreme effectiveness of the velvet ant defensive suite has likely led to the evolution

485 of the large velvet ant mimicry complexes found throughout North America (cite). These

486 complexes include both harmless Batesian mimics as well as Müllerian mimics (cite). The

487 Batesian mimics include those of spiders (Edwards, 1984; Nentwig, 1985), antlion larvae (Brach,

488 1978), and several beetles (Acorn, 1988; Mawdsley, 1994; Lanteri and Del Rio, 2005). Velvet

489 ants, along with some other wasps, form the largest known Müllerian mimicry complex

490 worldwide, with over 350 species from 25 genera and two families participating in eight distinct

491 mimicry rings (Wilson et al. 2012; 2015; Rodriguez et al. 2014).

494 It has been suggested that because velvet ants are similar in appearance and behavior of ants,

495 predators that often eat ants could pose a threat to velvet ants (Pan et al. 2017). The evolution of

496 aposematism and long setae seen on many velvet ants, therefore, likely developed as a way to

497 differentiate themselves from true ants to reduce attacks from ant specialist predators. Many

498 diurnal velvet ants have bright contrasting colors and relatively long setae, which are not seen in

499 true ants.

503 Summarize our results

504 Discuss in relation to other studies on predators of velvet ants

22 505 Suggest velvet ants are essentially invincible

506 Evolution of these defenses

507 combination of selection from predators (slow moving, flightless, diurnal)

508 can’t rule out selection from host D – what paper was that – Michigan lake?

509 any link to parasite host interactions?

510 Likely why mimicry complexes so large and so many different ones exist

511 Talk about joes other two papers?

512 What limits populations – distribution and abundance of host

513 Any other examples of this in nature?

514 Surprising because studies show parasites are often responsible for regulating host

516 Acorn, J. H. Mimetic tiger beetles and the puzzle of cicindelid coloration (Coleoptera:

517 Cicindelidae). Coleopterists Bull. 42, 28-33 (1988).

518 Mawdsley, J. R. Mimicry in Cleridae (Coleoptera). Coleopterists Bull. 48, 115-125 (1994).

519 Lanteri, A. A. & Del Rio, M. G. Taxonomy of the monotypic genus Trichaptus Pascoe

520 (Coleoptera: Curculionidae: Entiminae), a potential weevil mimic of Mutillidae.

521 Coleopterists Bull. 59, 47-54 (2005).

522 Edwards, G. Mimicry of velvet ants (Hymenoptera: Mutillidae) by jumping spiders

523 (Araneae: Salticidae). Peckhamia 2, 46-49 (1984).

524 Nentwig, W. A mimicry complex between mutillid wasps (Hymenoptera: Mutillidae) and

525 spiders (Araneae). Stud. Neotrop. Fauna Envir. 20, 113-116 (1985).

23 526 Brach, V. Brachynemurus nebulosus (Neuroptera: Myrmeleontidae): a possible Batesian

527 mimic of Florida mutillid wasps (Hymenoptera: Mutillidae). Entomol. News 89, 153-156

528 (1978).

529 Rodriguez, J., J.P. Pitts, C.D. von Dohlen, and J.S. Wilson. 2014. Müllerian mimicry as a result

530 of codivergence between velvet ants and spider wasps. PLoS ONE. 9: e112942.

531 Pan, A.D., K.A. Williams, and J.S. Wilson. 2016. Are diurnal Iguanian lizards the evolutionary

532 drivers of New World female velvet ant (Hymenoptera: Mutillidae) Müllerian mimicry rings?

533 Biological Journal of the Linnean Society. DOI: 10.1111/bij.12894

536 Acknowledgements

537 All animals from Utah were collected and housed under COR permit #4COLL8467. The

538 shrews and mole were collected under Indiana Scientific Purposes License #14-040.

540 Tables

541 Table 1. Species of velvet ants tested with lizard predators, number of trials conducted for each

542 species, and the number of instances that species was attacked.

# of # of Velvet ant species trials attacks Dasymutilla arenivaga 2 1 Dasymutilla bioculata 18 0 Dasymutilla foxi 8 1 Dasymutilla gloriosa 10 0 Dasymutilla gorgon 14 2 Dasymutilla scitula 9 1 Dasymutilla vestita 5 1 Sphaeropthalma mendica 4 1 543

24 544 Table 2. Summary of the outcomes from initial and secondary trials with 4 species of lizards and

545 various velvet ants. Number in parentheses are ?.

# of primary # of # of # of ants # of # of ants injured Lizard species trials investigations strikes consumed stings or killed Aspidoscelis tigris 6 5 (4) 2 (1) 0 2 (1) 0 Eublepharis macularius 1 0 0 0 0 0 Gambelia wislizenii 2 1 0 0 0 0 Uta stansburiana 3 3 1 0 1 0

# of secondary # of # of # of ants # of # of ants injured Lizard species trials investigations strikes consumed stings or killed Aspidoscelis tigris 36 21 (19) 3 0 2 0 Eublepharis macularius 2 0 0 0 0 0 Gambelia wislizenii 15 4 1 1 0 0 Uta stansburiana 6 1 0 0 0 0 546

547 Figures

549 Figure 1. Painted clay models (left) and painted mealworms (right) used to test the role of

550 aposematic coloration found in Dasymutilla occidentalis during interactions with wild birds.

25 552

553 Figure 2. Photograph of an aposematically painted mealworm that was struck at by a

554 mockingbird and “decapitated” but not consumed.

26 556

557 Figure 3. Photograph of the feeding station with a mockingbird perched on top. Photograph by

558 Richard Vaupel.

560 Aneshansley, D.J., Eisner, T., Widom, J.M., Widom, B., 1969. Biochemistry at 100°C: Explosive Secretory 561 Discharge of Bombardier Beetles (Brachinus). Science 165, 61-63. 562 Arndt, E.M., Moore, W., Lee, W., Ortiz, C., 2015. Mechanistic origins of bombardier beetle (Brachinini) 563 explosion-induced defensive spray pulsation. Science 348, 563-567. 564 Beal, F., 1915. Food habits of the thrushes of the United States. US Department of Agriculture. 565 Brodie, E.D., III., 1993. Differential avoidance of coral snake banded patterns by free‐ranging avian 566 predators in Costa Rica. Evolution 47, 227-235. 567 Brodie, E.D., III., Janzen, F.J., 1995. Experimental studies of coral snake mimicry: generalized avoidance 568 of ringed snake patterns by free-ranging avian predators. Funct Ecol 9, 186-190. 569 Brodie, E.D., Jr., 1968. Investigations on the skin toxin of the adult rough-skinned newt, Taricha 570 granulosa. Copeia 1968, 307-313. 571 Brodie, E.D., Jr., Formanowicz, D.R., Jr., Brodie, E.D., III., 1991. Predator avoidance and antipredator 572 mechanisms: distinct pathways to survival. Ethology Ecology & Evolution 3, 73-77. 573 Brodie, E.D., Jr., Nowak, R.T., Harvey, W.R., 1979. The effectiveness of antipredator secretions and 574 behavior of selected salamanders against shrews. Copeia 1979, 270-274.

27 575 Brodie, E.D., Nussbaum, R.A., Digiovanni, M., 1984. Antipredator adaptations of Asian salamanders 576 (Salamandridae). Herpetologica 40, 56-68. 577 Brothers, D.J., 1972. Biology and immature stages of Pseudomethoca f. frigida, with nots on other 578 species (Hymenoptera: Mutillidae). University of Kansas Science Bulletin 50, 1-38. 579 Coppinger, R.P., 1970. The effect of experience and novelty on avian feeding behavior with reference to 580 the evolution of warning coloration in butterflies. II. Reactions of naive birds to novel insects. The 581 American Naturalist 104, 323-335. 582 Cottam, C., Knappen, P., 1939. Food of some uncommon North American birds. The Auk, 138-169. 583 Deyrup, M., 1988. Review of adaptations of velvet ants (Hymenoptera: Multillidae). Great Lakes 584 Entomologist 21, 1-4. 585 Endler, J.A., 1986. Defense against predators, in: Feder, M.E., Lauder, G.V. (Eds.), Predator-Prey 586 Relationships. University of Chicago Press, Chicago, pp. 109-134. 587 Exnerová, A., Štys, P., Fučíková, E., Veselá, S., Svádová, K., Prokopová, M., Jarošík, V., Fuchs, R., Landová, 588 E., 2006. Avoidance of aposematic prey in European tits (Paridae): learned or innate? Behav Ecol 18, 589 148-156. 590 Fales, H.M., Jaouni, T.M., Schmidt, J.O., Blum, M.S., 1980. Mandibular gland allomones of Dasymutilla 591 occidentalis and other mutillid wasps. J Chem Ecol 6, 895-903. 592 Fathinia, B., Rastegar-Pouyani, N., Rastegar-Pouyani, E., Todehdehghan, F., Amiri, F., 2015. Avian 593 deception using an elaborate caudal lure in (Serpentes: Viperidae). Amphibia-Reptilia 36, 223-231. 594 Hopkins, G.R., Gall, B.G., Brodie, E.D., Jr., 2011. Ontogenetic shift in efficacy of antipredator mechanisms 595 in a top aquatic predator, Anax junius (Odonata: Aeshnidae). Ethology 117, 1093-1100. 596 Jared, C., Mailho-Fontana, P.L., Antoniazzi, M.M., Mendes, V.A., Barbaro, K.C., Rodrigues, M.T., Brodie, 597 E.D., Jr., 2015. Venomous frogs use heads as weapons. Curr Biol 25, 2166-2170. 598 Kingdon, J., Agwanda, B., Kinnaird, M., O'Brien, T.O., Holland, C., Gheysens, T., Boulet-Audet, M., 599 Vollrath, F., 2011. A poisonous surprise under the coat of the African crested rat. P R Soc B. 600 Lima, S.L., Dill, L.M., 1990. Behavioral decisions made under the risk of predation: A review and 601 prospectus. Canadian Journal of Zoology 68, 619-640. 602 Manley, D.G., Sherbrooke, W.C., 2001. Predation on velvet ants (Hymenoptera: Mutillidae) by Texas 603 horned lizards (Phrynosoma cornutum). Southwestern Naturalist 46, 221-222. 604 Needham, J.G., Westfall, M.J., Jr., 1954. The manual of the dragonflies of North America (Anisoptera). 605 University of California Press, Los Angeles. 606 Nowak, R.T., Brodie, E.D., Jr., 1978. Rib penetration and associated antipredator adaptations in the 607 salamander Pleurodeles waltl (Salamandridae). Copeia 1978, 424-429. 608 Schmidt, J.O., 1990. Hymenopteran venoms: striving toward the ultimate defense against vertebrates, 609 in: Evans, D.L., Schmidt, J.O. (Eds.), Insect defenses: adaptive mechanisms and strategies of prey and 610 predators. State University of New York Press, Albany, NY, pp. 387-419. 611 Schmidt, J.O., Blum, M.S., 1977. Adaptations and responses of Dasymutilla occidentalis (Hymenoptera: 612 Mutillidae) to predators. Entomologia experimentalis et applicata 21, 99-111. 613 Schmidt, J.O., Blum, M.S., Overal, W.L., 1984. Hemolytic activities of stinging insect venoms. Archives of 614 Insect Biochemistry and Physiology 1, 155-160. 615 Smith, S.M., 1975. Innate recognition of coral snake pattern by a possible avian predator. Science 187, 616 759-760. 617 Starr, C.K., 1985. A simple pain scale for field comparison of hymenopteran stings. Journal of 618 Entomological Science 20, 225-232. 619 Svádová, K., Exnerova, A., Štys, P., Landova, E., Valenta, J., Fučíková, A., Socha, R., 2009. Role of different 620 colours of aposematic insects in learning, memory and generalization of naive bird predators. Anim 621 Behav 77, 327-336.

28 622 Tschuch, G., 1993. Sound production in mutillid wasps (Mutillidae, Hymenoptera). Bioacoustics 5, 123- 623 129. 624 Vitt, L.J., Cooper, W.E., 1988. Feeding responses of skinks (Eumeces laticeps) to velvet ants (Dasymutilla 625 occidentalis). Journal of Herpetology 22, 485-488. 626 Wiklund, C., Järvi, T., 1982. Survival of distasteful insects after being attacked by naive birds: a 627 reappraisal of the theory of aposematic coloration evolving through individual selection. Evolution 36, 628 998-1002. 629 Williams, K.A., 2012. Systematics of Multillidae (Hymenoptera) With Special Emphasis on Dasymutilla 630 and Their Allies, Biology. Utah State University, p. 342. 631 Wilson, J.S., Jahner, J.P., Forister, M.L., Sheehan, E.S., Williams, K.A., Pitts, J.P., 2015. North American 632 velvet ants form one of the world's largest known mullerian mimicry complexes. Curr Biol 25, R704- 633 R706. 634 Wilson, J.S., Williams, K.A., Forister, M.L., Von Dohlen, C.D., Pitts, J.P., 2012. Repeated evolution in 635 overlapping mimicry rings among North American velvet ants. Nature Communications 3, 1272.