Two summaries for two ecology papers- for phyllis young

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

rspb.royalsocietypublishing.org

Research Cite this article: Mathis KA, Tsutsui ND. 2016 Dead ant walking: a myrmecophilous beetle

predator uses parasitoid host location cues to

selectively prey on parasitized ants.

Proc. R. Soc. B 283: 20161281. http://dx.doi.org/10.1098/rspb.2016.1281

Received: 8 June 2016

Accepted: 20 July 2016

Subject Areas: behaviour, ecology

Keywords: myrmecophily, phorid fly parasitism,

Azteca sericeasur, Staphylinidae,

predation strategy, complex interactions

Author for correspondence: Kaitlyn A. Mathis

e-mail: [email protected]

†Present address: Department of Ecology

and Evolutionary Biology, University of

Arizona, 1041 East Lowell Street, Tucson,

AZ 85721, USA.

Electronic supplementary material is available

at http://dx.doi.org/10.1098/rspb.2016.1281 or

via http://rspb.royalsocietypublishing.org.

& 2016 The Author(s) Published by the Royal Society. All rights reserved.

Dead ant walking: a myrmecophilous beetle predator uses parasitoid host location cues to selectively prey on parasitized ants

Kaitlyn A. Mathis† and Neil D. Tsutsui

Department of Environmental Science, Policy, and Management, University of California, 130 Mulford Hall, Berkeley, CA 94702-3114, USA

KAM, 0000-0002-3809-9128; NDT, 0000-0002-1868-3941

Myrmecophiles (i.e. organisms that associate with ants) use a variety of

ecological niches and employ different strategies to survive encounters with

ants. Because ants are typically excellent defenders, myrmecophiles may

choose moments of weakness to take advantage of their ant associates. This

hypothesis was studied in the rove beetle, Myrmedonota xipe, which associates with Azteca sericeasur ants in the presence of parasitoid flies. A combination of laboratory and field experiments show that M. xipe beetles selectively locate and prey upon parasitized ants. These parasitized ants are less aggressive

towards beetles than healthy ants, allowing beetles to eat the parasitized

ants alive without interruption. Moreover, behavioural assays and chemical

analysis reveal that M. xipe are attracted to the ant’s alarm pheromone, the same secretion used by the phorid fly parasitoids in host location. This strategy

allows beetles access to an abundant but otherwise inaccessible resource, as

A. sericeasur ants are typically highly aggressive. These results are the first, to our knowledge, to demonstrate a predator sharing cues with a parasitoid

to gain access to an otherwise unavailable prey item. Furthermore, this work

highlights the importance of studying ant – myrmecophile interactions

beyond just their pairwise context.

1. Introduction Ant societies attract a suite of symbiotic organisms that take advantage of a col-

ony’s abundant resources. Beetles in the family Staphylinidae are common ant

associates, yet relatively little is known about the role of these beetles in their

host colonies. Owing to the formidable chemical and behavioural defences of

ants, beetles often possess complex strategies to safely interact with their ant

symbionts [1,2]. Many beetles act as scavengers that hide in refuse piles, or as

ant-mimicking social parasites within a host colony, draining them of resources.

Others are predators of the ants themselves, locating their prey by ‘eavesdrop-

ping’ on the ants’ communication system [3]. While commensal, parasitic and

predatory beetles are common within ant societies, little is known about how bee-

tles might benefit their ant associates [4]. Furthermore, although most biological

communities function as complex networks of context-dependent interactions,

ant – beetle associations are rarely observed outside of their pairwise context.

While often studied in isolation, predator – prey and host – parasitoid inter-

actions have a wide array of effects within food webs and are thus increasingly

approached from a community perspective [5]. Natural enemies can directly

influence hosts or prey by reducing population size or inducing changes in

phenotype (e.g. behaviour, morphology). These interactions often have cascading

effects on other species within the community [6 – 8]. Cascading (or indirect)

effects can be mediated by both changes in host/prey density (density-mediated

indirect effects) and changes in the traits of host/prey species (trait-mediated

indirect effects) (reviewed in [9]).

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

2

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

Trait-mediated interactions may be particularly relevant to

host – parasitoid systems because host behaviour and/or physi-

ology is frequently modified as a result of parasitism, with

subsequent consequences for structuring biological communities

[10]. Trait modifications can occur before or after parasitism and

are either adaptive or non-adaptive to the parasitoid and/or host

[11 – 13]. For example, in the presence of parasitoids, hosts will

often suspend normal activity in order to implement chemical

or behavioural defensive strategies [14 – 16]. After parasitism,

immature parasitoids within the host may alter the host’s physi-

ology to encourage behaviours that optimize conditions for

parasitoid development [17 – 19]. Alternatively, host immune

response or even self-sacrifice behaviour may increase to prevent

parasitoids from developing [20,21].

Here, we experimentally demonstrate a context-dependent

predation strategy involving: (i) phorid fly parasitoids, where

parasitism reduces ant aggression; and (ii) predatory beetles,

where beetles selectively prey upon parasitized ants, thus

potentially reducing phorid fly populations without impacting

ant populations. We also demonstrate that this interaction

is mediated by the alarm pheromone of the ant, which both

natural enemies use in host/prey location.

Azteca sericeasur (referred to as Azteca instabilis in prior pub- lications, but recently identified as A. sericeasur; J. Longino 2014, personal communication) is a highly aggressive and territorial

arboreal ant species that lives in large polydomous carton

nests [8]. This species frequently nests in the shade trees of

coffee plantations and forages in the coffee below, preying on

and removing herbivores [22]. Ants in the genus Azteca are known for their pungent alarm pheromone that they disperse

liberally from large pygidial gland sacs when disturbed.

A suite of Pseudacteon phorid fly species uses this alarm phero- mone to locate and parasitize A. sericeasur [23]. Pseudacteon lasciniosus is the largest of the three species and the most abun- dant at our field sites and Pseudacteon planidorsalis is a smaller species and the second most abundant [24]. The presence of

phorid fly parasitoids not only reduces the ants’ ability to

forage by as much as 50% [16,25], but also indirectly affects inter-

actions between the ants and a wide range of competitors and

mutualistic partners [8]. The newly described species of rove

beetle, Myrmedonota xipe (Staphylinidae), has been observed in association with A. sericeasur ants. Myrmedonota xipe beetles are found near disturbed A. sericeasur, often mating or preying on A. sericeasur ants after the arrival of phorid flies [26]. These interactions suggest two questions. First, because A. sericeasur workers are notoriously aggressive, how are the beetles able

to prey upon the ants? Second, are the beetles using the same

alarm pheromone as the phorid flies to locate the ants?

We test the hypothesis that parasitism by phorid flies

reduces ant aggression, allowing beetles to gain access to the

ants as a prey item. Furthermore, we hypothesize that this

context-dependent interaction is induced by the ant’s alarm

pheromone, which is released during phorid attack.

2. Material and methods (a) Study site We conducted all fieldwork on a shaded coffee plantation, Finca

Irlanda, in the Soconusco region of Chiapas, Mexico (158110 N, 928200 W) between July 2012 and March 2013. Finca Irlanda is approximately 280 ha in size, located between 950 and 1150 m

elevation, and receives approximately 4500 mm of precipitation

per year. Azteca sericeasur is the most abundant ant of the approximately 60 species of arboreal ants on the farm [25].

Azteca sericeasur builds carton nests on the trunks of shade trees within the coffee plantation, where their colonies tend to

be distributed in patches [27]. Ants, beetles and phorid flies

were all collected for laboratory experiments at five different

A. sericeasur nests within the field site using an aspirator.

(b) Arena experiment To determine whether M. xipe selectively attack parasitized ants, and whether parasitized and unparasitized ants respond to

beetle attacks differently, we placed parasitized and unparasitized

ants in an arena with the beetles, filmed their interactions, and ana-

lysed the resulting footage (electronic supplementary material, S1).

We placed A. sericeasur ants (n ¼ 5) in small plastic container with an individual phorid fly until the ants were parasitized

(approx. 1 h). We then chilled the containers in the 2208C freezer for 2 min until the fly and ants were anaesthetized. While the ants

were anaesthetized, we confirmed that ants were successfully

parasitized by inspecting for oviposition wounds from the

phorid fly. If the ant was successfully parasitized, we added a

single dot of paint (green, white or blue chosen at random for

each observation) on the head of each ant and identified the anaes-

thetized phorid flies to species under a microscope (electronic

supplementary material, S2). New phorid flies were used for

each parasitism event. We also anaesthetized and painted five

unparasitized ants using the same method. We confirmed the

ants to be unparasitized by inspecting them for a lack of ovipos-

ition wounds. We allowed the ants a 1 h recovery period, which

was enough time for ants to resume normal activities. Parasitized

ants were used within 4 h of the initial parasitism event. We placed

the five unparasitized ants and five parasitized ants (by either

P. lasciniosus or P. planidorsalis) with two M. xipe beetles in a plastic container (arena) coated with fluon (Northern Products, Rhode

Island, USA) to prevent the insects from climbing the walls of

the arena. We sealed the arena with a transparent glass lid. We

then filmed the arena for 15 min. A total of 82 arena experiments

were conducted (50 P. lasciniosus-parasitized versus healthy and 32 P. planidorsalis-parasitized versus healthy treatments). We later analysed video footage of the arena experiments using

OBSERVER XT software (v. 11, Noldus Information Technology,

Wageningen, The Netherlands), by recording the duration of

each behaviour. The behaviours we included were: (i) attack (one

animal bites another animal; and (ii) mandible flare (an ant

opens and closes her mandibles repeatedly in the direction of

another animal) [28]. We recorded behaviours with the OBSERVER

Software, blind to the treatment type. We calculated the total

duration of each behaviour by organism type ( parasitized ant,

unparasitized ant or M. xipe) and the corresponding target organism type for each observation.

(c) Beetle trap experiment Previous studies have shown that dying ants will leave their nests

or are restricted entry to their nests [29]. To determine whether

M. xipe beetles are able to selectively locate individual parasitized ants within the coffee plantation, we constructed beetle traps from

small plastic cups with plaster of Paris on the bottom to retain

moisture and lids with holes large enough for the beetles to

enter the traps (electronic supplementary material, S3). Each cup

contained an A. sericeasur worker randomly assigned to one of four treatment types. The four treatments included: (i) A. sericeasur parasitized by P. lasciniosus, (ii) A. sericeasur parasitized by P. planidorsalis, (iii) A. sericeasur manually injured by puncturing their mesothorax with a Minuten pin (0.20 mm diameter, Bioquip

Products, Rancho Dominguez, CA, USA) to simulate a phorid

attack wound, and (iv) A. sericeasur untreated as a control. Ants in parasitism treatments were parasitized according to the same

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

3

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

methods in the arena experiment. We placed a total of 624 traps,

divided evenly by treatment. We chose 39 sites in both high

shade (10 sites with focal trees containing A. sericeasur nests and 10 sites without A. sericeasur nests) and low shade (10 sites with focal trees containing A. sericeasur nests and nine sites without A. sericeasur nests). We placed 16 traps at each site: one cup of each treatment type 0 m and 5 m from each sites’ focal tree, both

on the ground and suspended 1 m in the nearest coffee plant.

We placed the cups for each site in the field on the same day,

and retrieved them 2 days later (electronic supplementary

material, S4). We identified M. xipe beetles in each cup, and recorded whether the ant was living or completely consumed.

(d) Extraction and analysis of Azteca alarm pheromone To determine the primary components of A. sericeasur alarm phero- mone, we collected A. sericeasur volatile alarm pheromone compounds by first placing an A. sericeasur worker in a 50 ml glass beaker. Then we disturbed the worker by pinching her tibia

and covered the beaker with aluminium foil. We then inserted a

solid phase microextraction (SPME) fibre into the beaker through

the aluminium foil for 10 min to adsorb the headspace volatiles.

We conducted five ant volatile collections. SPME fibres were

immediately inserted into a Finnigan Trace MSþ gas chromato- graph/mass spectrometer equipped with a DB-5 capillary column

(30 m � 0.32 mm � 0.25 mm, Agilent Technologies, CA, USA). Extracts were analysed in splitless mode, with a temperature

programme that started at 1008C for 1 min, which then increased by 208C min21 until it reached 1508C, and then increased by 58C min21 until it reached 3258C where it stayed for 5 min. Injector and transfer line temperatures were kept at 3258C and 2808C, respectively.

(e) Alarm pheromone bioassays To determine whether M. xipe are attracted by the alarm pheromone of A. sericeasur, we obtained commercially availa- ble synthetic 2-heptanone and 2-pentanone (Sigma-Aldrich,

St Louis, USA), the two primary compounds released by disturbed

ants. We prepared four treatment solutions: (i) 1 ml of pesticide-

grade hexanes as a negative solvent control, (ii) 20 crushed

A. sericeasur pygidial glands in 1 ml of hexane, (iii) 5 ml 2-heptanone in 1 ml of hexane, and (iv) 5 ml of 2-pentanone in 1 ml of hexane. We

then placed treatment solutions in 2 dram glass vials, open, with a

filter paper wick at 20 field sites [23]. All field sites were at least

25 m apart, at the base of trees within the coffee farm that contain

an A. sericeasur nest. At each site, we placed the treatment solution vial on the ground with leaf litter removed from the surrounding

area. Once we opened the vial with treatment solution, we

observed a 10 cm2 area surrounding the vial for 15 min, and col-

lected and identified beetles attracted to the area using an

aspirator. We quantified the total number of beetles from each

species collected at each site with each treatment type.

( f ) Data analysis All statistical analyses were performed in R (v. 3.1.2) [30]. We

tested for significant differences in prey choice ( parasitized or

healthy ants) and ant response (attack and mandible flare beha-

viours) in the arena experiments (n ¼ 83 observations) using separate two-way analysis of variance (ANOVA) followed by

Tukey HSD means comparison tests. To examine whether ant

removal via beetle predation differed with treatment or habitat

types in the beetle trap experiments, we compared ant removal

by beetles with generalized linear mixed models (GLMMs) in

the ‘lme4’ package. For all models we used the binomial error dis-

tribution with the probit link to improve fit. We compared three

models. In the first model, we included treatments ( parasitized

by P. lasciniosus, parasitized by P. planidorsalis, manually injured,

or healthy control), sites (1 – 39), distance from focal tree (0 m or

5 m), the presence or absence of an A. sericeasur nest at focal tree, trap height placement (0 m or 1 m) and shade cover (high or

low) as fixed effects. In the second model, we added interactions

between treatment and height, and between treatment and

shade. In the third model, we added all interactions between treat-

ment and each habitat variable. To select the best model, we used

the Akaike’s information criterion (AIC) provided by the ‘mass’

package and after we fitted the models we checked homoscedasti-

city of the residuals [31]. The model that best predicted ant removal

by beetle predators was the first model, thus including interactions

between habitat variable and treatments did not improve fit

of the model. For the alarm pheromone bioassays, we tested

for significant differences between treatments (2-heptanone,

2-pentanone, pygidial gland extract positive control or hexane

negative control) using a one-way ANOVA followed by Tukey

HSD mean comparison tests.

3. Results (a) Do beetles prefer to prey on parasitized ants? In the arena experiment, beetles attacked ants that were

parasitized by P. lasciniosus more often than they attacked unparasitized ants (figure 1a; ANOVA; F1,48 ¼ 24.59, p , 0.001). Ants parasitized by P. planidorsalis were not attacked more than unparasitized ants (figure 1b; ‘% of observation spent attacking’: 11.21 + 5.67% spent attacking parasitized ants, 8.97 + 6.23% spent attacking unparasitized ants; ANOVA; F1,31 ¼ 0.81, p ¼ 0.37). In the beetle trap exper- iment, beetles were found consuming only ants parasitized

by either P. lasciniosus (14.7% of ants) or P. planidorsalis (2.5%) and did not consume control or injured (sham-parasitized)

ants. Consumption of P. lasciniosus-parasitized ants was significantly higher than consumption of control or injured

ants (GLMM, x23 ¼ 50.33, p , 0.0001; figure 2), and consump- tion of P. planidorsalis-parasitized ants was not significantly higher than control or injured ants (GLMM; x23 ¼ 50.33,

p ¼ 0.20).

(b) Do parasitized and unparasitized ants display different levels of aggression towards predatory beetles?

In the arena experiment, parasitized ants were less aggressive

than healthy ants. During the observations, parasitized ants

attacked beetles significantly less often than did unparasitized

ants, regardless of which phorid species parasitized them

(figure 1c,d; P. lasiniosus-parasitized ants � unparasitized ants: F1,34 ¼ 4.124, p , 0.05; P. planidorsalis-parasitized ants � unparasitized ants: F1,26 ¼ 4.495, p , 0.05). However, ants parasitized by the two species of phorid fly differed in their

mandible flare performance when compared with unparasi-

tized ants. Ants parasitized by P. lasciniosus flared their mandibles less often than did unparasitized ants (figure 1e; F1,40 ¼ 11.38, p , 0.002), but ants parasitized by P. planidorsalis did not (figure 1f; F1,28 ¼ 0, p ¼ 0.99).

(c) Are beetles able to successfully locate parasitized ants in different habitat types?

In the beetle trap experiment, ants were completely consumed

in all traps where M. xipe were collected after 2 days. Beetles

ant mandible flare type

ant attack type

beetle attack type

100

pe rc

en ta

ge o

f ob

se rv

at io

n sp

en t

pe rc

en ta

ge o

f ob

se rv

at io

n sp

en t80

60

40

20

0

100 15

0

12

8

6

0

15

10

5

0

5

10

15

0

5

10

80

***

**

*

**

*

60

40

20

0

(a) (b)

(c) (d)

(e) ( f )

Figure 1. Box plots of arena experiment behaviours. (a) Beetles attacking ants parasitized by P. lasciniosus (left column) and healthy ants (right column). (b) Beetles attacking ants parasitized by P. planidorsalis (left column) and healthy ants (right column). (c) Pseudacteon lasciniosus-parasitized ants (left column) and by healthy ants (right column) attacking beetles. (d ) Pseudacteon planidorsalis-parasitized ants (left column) and by healthy ants (right column) attacking beetles. (e) Mandible flares by P. lasciniosus-parasitized ants (left column) and by healthy ants (right column). ( f ) Mandible flares by P. planidorsalis-parasitized ants (left column) and by healthy ants (right column). The box plots indicate the median (large horizontal bars), the 25th and 75th percentiles (white rectangles), the minimum and maximum values (whiskers) and outliers (white circles). Asterisks represent significance level (*�0.05, **�0.01, ***�0.001). (Online version in colour.)

control

0.4

injure planidorsalis lasciniosus

***

0

0.1

0.2

0.3

pr op

or ti

on o

f an

ts c

on su

m ed

b y

M . xi

p e

Figure 2. Proportion of ants consumed by M. xipe beetles in beetle trap experiments. Bars represent the four treatment types: controls are traps baited with healthy ants, injure treatments are traps baited with manually injured ants, P. planidorsalis treatments are traps baited with ants parasitized by P. planidorsalis and P. lasciniosus treatments are traps baited with ants parasitized by P. lasciniosus. Asterisks indicate significance level (***�0.001).

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

4

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

successfully located and consumed ants parasitized by

P. lasciniosus in both habitat types. Attacks on ants in high shade were significantly more frequent than attacks in low

shade (GLMM; figure 3a, P. lasciniosus � shade, x21 ¼ 15.41, p , 0.0001) and significantly more frequent on the ground than above ground (GLMM; figure 3b; x21 ¼ 9.21, p , 0.003). Furthermore, there were no differences based on distance

from focal tree (GLMM; figure 3d; x21 ¼ 2.65, p ¼ 0.10), or with and without A. sericeasur presence (GLMM; figure 3c; x

2 1 ¼ 0.06, p ¼ 0.80).

(d) How do beetles locate ants? Analysis of SPME fibres containing volatiles from disturbed A. sericeasur workers showed that workers have two primary components present within their alarm pheromone, 2-penta-

none and 2-heptanone, in a roughly 3 : 8 ratio (electronic

supplementary material, S5; 0.374 + 0.02 2-pentanone : 2-hep- tanone ratio; n ¼ 5). In alarm pheromone bioassays, M. xipe was attracted to both pygidial gland extracts and synthetic

2-heptanone (the most abundant compound in the pheromone

blend) significantly more than the hexane control (ANOVA;

F3,76 ¼ 7.072, p , 0.03). Myrmedonota xipe beetles were not attracted to 2-pentanone (the secondary component of the

blend) more than the hexane control (ANOVA; F3,76 ¼ 7.072, p ¼ 1.0; figure 4).

4. Discussion Our results show that M. xipe selectively preys on parasitized ants, particularly those parasitized by P. lasciniosus. These para- sitized ants display reduced aggression (including a reduced

frequency of mandible flaring), which may allow beetles to

more easily gain access to them as a prey item. Furthermore,

in the field, beetles were able to consume up to 14.7% of ants

parasitized by P. lasciniosus, which suggests that these beetles may have an important role in reducing P. lasciniosus popu- lations. This preliminary estimate of beetle predation rates

should be tested in further studies. However, it is probably con-

servative, given that M. xipe are generally already present just after phorid parasitism and are frequently observed preying

on ants near disturbed colonies shortly after parasitism takes

place (KA Mathis 2014, personal observation). Myrmedonota

0.30

no . a

nt s

co ns

um ed

p er

t re

at m

en t

no . a

nt s

co ns

um ed

p er

t re

at m

en t

0.25

0.20

0.15

0.10

0.05

0

0.30

0.25*** **

0.20

0.15

0.10

0.05

low shade

Azteca present Azteca absent

high shade 0 m 1 m

0 m 5 m

0

0.30

0.25

0.20

0.15

0.10

0.05

0

0.30

0.25

0.20

0.15

0.10

0.05

0

(a) (b)

(c) (d)

Figure 3. The proportion of P. lasciniosus-parasitized ants consumed by M. xipe beetles in beetle trap experiments, broken down by habitat type. (a) Average number of beetles found in traps with ants parasitized by P. lasciniosus in low shade (left bar) and high shade (right bar) habitats. (b) Average number of beetles found in traps with ants parasitized by P. lasciniosus on the ground (left bar) and 1 m above the ground (right bar). (c) Average number of beetles found in traps with ants parasitized by P. lasciniosus at sites with A. sericeasur nests (left bar) and without A. sericeasur nests (right bar). (d ) Average number of beetles found in traps with ants parasitized by P. lasciniosus at 0 m (left bar) and 5 m (right bar) from the focal tree. Asterisks represent significance level (*�0.05, **�0.01, ***�0.001).

2-heptanone

Myrmedonota xipe attraction to Azteca sericeasur alarm pheromone

no . b

ee tl

es t

o ar

ri ve

a

6

5

4

3

2

1

0

a

b b

2-pentanone

treatment

hexane pygidial

Figure 4. Box plots of the average number of M. xipe beetles to arrive at alarm pheromone bioassays. Each box represents a treatment type, where 2-heptanone is synthetic 2-heptanone, 2-pentanone is synthetic 2-pentanone, hexane is a solvent control and pygidial is an extract from crushed pygidial glands of A. sericeasur. The box plots indicate the median (large horizontal bars), the 25th and 75th percentiles (white rectangles), the minimum and maximum values (whiskers) and outliers (white circles). Lowercase letters show significant differences between treatment types.

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

5

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

xipe preyed on ants parasitized by P. planidorsalis significantly less often than it preyed ants parasitized by P. lasciniosus, probably because ants parasitized by P. planidorsalis were more aggressive than ants parasitized by P. lasciniosus. These experiments indicate that beetles probably have a stronger

impact on ant – P. lasciniosus interactions than ant – P. planidorsalis interactions.

This study also demonstrates that M. xipe uses A. sericeasur alarm pheromone as a cue to locate prey. Although phorid flies

also use the alarm pheromone to locate A. sericeasur hosts, they are attracted to 1-acetyl-2-methylcyclopentane, a less

abundant compound within the alarm pheromone blend

that is only found in Azteca species ants [23,32,33]. The com- pound that attracts M. xipe, 2-heptanone, is relatively common in the alarm pheromone of dolichoderine ants. The

use of 2-heptanone by M. xipe suggests that these beetles may be less selective and may prey on other dolichoderine

ant species in addition to A. sericeasur. However, of the approximately 15 species at our study sites, A. sericeasur is over- whelmingly the most abundant, and are probably the ant that

these beetles encounter most frequently.

Several other studies have shown intraguild predation of

parasitoids, including examples where: (i) a predator preys

on the adult parasitoid as well as the host, (ii) a predator

preys on the unparasitized hosts as well as hosts with an ecto-

parasitoid, (iii) a predator preys more on parasitized hosts

than healthy hosts owing to host vulnerability after parasit-

ism, and (iv) a predator preys more on unparasitized hosts

owing to the more advantageous location of parasitized

hosts in a nest (reviewed in [34]). These studies differ from

ours in that M. xipe appears to almost exclusively prey on parasitized ants, as they are unable to access unparasitized

ants as a resource.

The unique strategy of M. xipe may have adaptive impli- cations for both A. sericeasur and the phorid parasitoids. First, as beetles are only consuming individuals that harbour

phorid fly eggs, beetle predation may, counter-intuitively,

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

6

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

benefit its prey by reducing the phorid fly population and

thus predation may be adaptive to A. sericeasur. Adaptive suicide in social insects is relatively common as a pre-emptive

strategy (e.g. honeybees using their stings to defend the

colony), although it is notably difficult to evaluate in the con-

text of host – parasitoid interactions [29,35 – 39]. The host

suicide hypothesis postulates that mature parasitoids emer-

ging from hosts are more likely to infect host’s kin than

non-kin. Therefore, when maturation of the parasitoid is pre-

vented, the inclusive fitness of the host should be increased.

Even a very small increase in inclusive fitness will be

enough to drive the system, and favourable situations for

adaptive suicide include systems where the host is a social

insect, when there is high host inbreeding, and when parasi-

toids have small search ranges [20]. This system appears to be

a good candidate to meet these criteria, as A. sericeasur is not only a social insect, but is also polygynous and polydomous,

forming colonies that can have territories spanning several

hectares [40]. Although the lifetime dispersal distances of

these phorid flies are currently unknown, P. lasciniosus and P. planidorsalis are rarely found farther than two meters from any given A. sericeasur nest [41]. Therefore, it is unlikely that phorid flies are dispersing beyond the boundaries of a

single A. sericeasur colony prior to oviposition, although future work on the precise dispersal distance of these

phorid flies is needed.

However, despite the potential benefits, selective predation

on parasitized ants may still have costs for the ant colony, and

therefore, not support the host suicide hypothesis. First, it is

possible that some small percentage of phorid fly larvae

do not successfully mature to adulthood. In these cases,

ants that survive parasitism and go on to benefit the colony

would be at a greater risk for predation, which would ulti-

mately cost the colony as a whole. Second, if the parasitized

workers are active colony members during the phorid’s devel-

opment period, the benefit of eliminating the phorid fly larvae

may be offset by the costs incurred from losing productive

parasitized ant workers. Future work is needed to confirm

true costs and benefits of beetle predation on parasitized

ants. Additionally, it will be worthwhile to determine the phys-

iological mechanism by which this behavioural switch in

aggression occurs in A. sericeasur.

Nonetheless, our study shows that the effects of the M. xipe association with A. sericeasur is dependent upon phorid fly presence and that these beetles may be indirectly beneficial to

A. sericeasur, by reducing the number of developing P. lascinosus and P. planidorsalis parasitoids by approximately 14.7% and 2.6%, respectively. To our knowledge, ours is one of the few

studies that document the role of ant-associated beetles outside

of a pairwise context [2,42 – 46], and the first study to demon-

strate a predator sharing host/prey cues with a parasitoid to

gain access to a prey item that would otherwise be unavailable.

Increasingly, it is becoming apparent that investigating the

ecological complexity within a system provides instructive

examples of how organisms can change their behaviour or

morphology in response to challenges from other organisms,

and subsequently, how these changes can have cascading

effects throughout a network of interacting species [8,9].

Current literature on the role of beetles within ant societies

tells us that these beetles are exceedingly common and behav-

iourally diverse, but few ant – beetle associations have been

examined in depth. Further investigation into both the roles

of these beetles and their ant hosts, as well as within the

network of organisms surrounding ant societies is crucial to

understanding how social insect colonies function within the

community as a whole.

Data accessibility. All data are available from the Dryad data repository: http://dx.doi.org/10.5061/dryad.g7599.

Authors’ contributions. K.A.M. conceived, designed and performed the experiments. K.A.M. and N.D.T. analysed the data and wrote the manuscript.

Competing interests. We declare we have no competing interests. Funding. We acknowledge USDA National Institute of Food and Agri- culture and Hatch project CA-B-INS-0087-H for support of the Tsutsui Laboratory. This material is based upon work supported by the National Science Foundation Graduate Research Fellowship pro- gramme (DGE 1106400), National Institutes of Health (award no. K12GM000708), the Robert Van den Bosch Fellowship, the UC Mexus Research Fellowship, the Edna & Yoshinori Tanada Fellowship and the Margaret C. Walker fund.

Acknowledgements. We thank Finca Irlanda for allowing us to conduct research on the farm. We thank SEMARNAT (Secretaria de Medio Ambiente y Recursos Naturales) for permission to collect and export samples, and J. Rojas and E. Chamé Vasquez for facilitating the process of acquiring permits.

References

1. Kistner DH. 1979 Social and evolutionary significance of social insect symbionts. Social Insects 1, 339 – 413.

2. Stoeffler M, Tolasch T, Steidle JLM. 2011 Three beetles—three concepts. Different defensive strategies of congeneric myrmecophilous beetles. Behav. Ecol. Sociobiol. 65, 1605 – 1613. (doi:10. 1007/s00265-011-1171-9)

3. Holldobler B, Wilson EO. 1990 The ants. Cambridge, MA: Harvard University Press.

4. Hughes DP, Pierce NE, Boomsma JJ. 2008 Social insect symbionts: evolution in homeostatic fortresses. Trends Ecol. Evol. 23, 672 – 677. (doi:10. 1016/j.tree.2008.07.011)

5. Orlofske SA, Jadin RC, Hoverman JT, Johnson PT. 2014 Predation and disease: understanding the

effects of predators at several trophic levels on pathogen transmission. Freshw. Biol. 59, 1064 – 1075. (doi:10.1111/fwb.12329)

6. Bruno JF, Cardinale BJ. 2008 Cascading effects of predator richness. Front. Ecol. Environ. 6, 539 – 546. (doi:10.1890/070136)

7. Liere H, Larsen A. 2010 Cascading trait-mediation: disruption of a trait-mediated mutualism by parasite- induced behavioral modification. Oikos 119, 1394 – 1400. (doi:10.1111/j.1600-0706.2010.17985.x)

8. Perfecto I, Vandermeer J, Philpott SM. 2014 Complex ecological interactions in the coffee agroecosystem. Annu. Rev. Ecol. Evol. Syst. 45, 137 – 158. (doi:10. 1146/annurev-ecolsys-120213-091923)

9. Werner EE, Peacor SD. 2003 A review of trait- mediated indirect interactions in ecological

communities. Ecology 84, 1083 – 1100. (doi:10. 1890/0012-9658(2003)084[1083:AROTII]2.0.CO;2)

10. Hatcher MJ, Dick JT, Dunn AM. 2014 Parasites that change predator or prey behaviour can have keystone effects on community composition. Biol. Lett. 10, 20130879. (doi:10.1098/rsbl. 2013.0879)

11. Poulin R. 1992 Altered behaviour in parasitized bumblebees: parasite manipulation or adaptive suicide? Anim. Behav. 44, 174 – 176. (doi:10.1016/ S0003-3472(05)80769-9)

12. Poulin R, Brodeur J, Moore J. 1994 Manipulation of host behaviour: should hosts always lose? Oikos 70, 479 – 484. (doi:10.2307/3545788)

13. Shorter JR, Rueppell O. 2011 A review on self- destructive defense behaviors in social insects.

rspb.royalsocietypublishing.org Proc.

R. Soc.

B 283:

20161281

7

on August 10, 2016http://rspb.royalsocietypublishing.org/Downloaded from

Insectes Soc. 59, 1 – 10. (doi:10.1007/s00040- 011-0210-x)

14. Gross P. 1993 Insect behavioral and morphological defenses against parasitoids. Annu. Rev. Entomol. 38, 251 – 273. (doi:10.1146/annurev.en.38. 010193.001343)

15. Mehdiabadi NJ, Kawazoe EA, Gilbert LE. 2004 Parasitoids and competitors influence colony-level responses in the red imported fire ant, Solenopsis invicta. Naturwissenschaften 91, 539 – 543. (doi:10. 1007/s00114-004-0561-5)

16. Philpott SM, Maldonado J, Vandermeer J, Perfecto I. 2004 Taking trophic cascades up a level: behaviorally-modified effects of phorid flies on ants and ant prey in coffee agroecosystems. Oikos 105, 141 – 147. (doi:10.1111/j.0030-1299. 2004.12889.x)

17. Vinson SB. 1977 Insect host responses against parasitoids and the parasitoids’ resistance: with emphasis on Lepidoptera-Hymenoptera association. Comp. Pathobiol. 3, 103 – 125. (doi:10.1007/978-1- 4615-7299-2_6)

18. Fritz RS. 1982 Selection for host modification by insect parasitoids. Evolution 36, 283 – 288. (doi:10. 2307/2408046)

19. Henne DC, Johnson SJ. 2007 Zombie fire ant workers: behavior controlled by decapitating fly parasitoids. Insectes Soc. 54, 150 – 153. (doi:10. 1007/s00040-007-0924-y)

20. Smith Trail DR. 1980 Behavioral interactions between parasites and hosts: host suicide and the evolution of complex life cycles. Am. Nat. 116, 77 – 91. (doi:10.1086/283612)

21. Horton DR, Moore J. 1993 Behavioral effects of parasites and pathogens in insect hosts. Parasites Pathogens Insects 1, 107 – 124. (doi:10.1016/B978- 0-08-091649-1.50010-4)

22. Jiménez-Soto E, Cruz-Rodrı́guez JA, Vandermeer J. 2013 Hypothenemus hampei (Coleoptera: Curculionidae) and its interactions with Azteca instabilis and Pheidole synanthropica (Hymenoptera : Formicidae) in a shade coffee agroecosystem. Environ. Entomol. 5, 915 – 924. (doi:10.1603/ EN12202)

23. Mathis KA, Philpott SM, Moreira RF. 2011 Parasite lost: chemical and visual cues used by Pseudacteon in search of Azteca instabilis. J. Insect Behav. 24, 186 – 199. (doi:10.1007/s10905-010-9247-3)

24. Reese KM, Philpott SM. 2012 Environmental and habitat drivers of relative abundance for a suite of Azteca-attacking Pseudacteon phorid flies. Environ. Entomol. 41, 1107 – 1114. (doi:10.1603/EN12014)

25. Philpott SM. 2005 Trait-mediated effects of parasitic phorid flies (Diptera: Phoridae) on ant (Hymenoptera: Formicidae) competition and resource access in coffee agro-ecosystems. Environ. Entomol. 34, 1089 – 1094. (doi:10.1093/ee/34.5.1089)

26. Mathis KA, Eldredge KT. 2014 Descriptions of two new species of Myrmedonota Cameron (Staphylinidae: Aleocharinae) from Mexico with comments on the genus taxonomy and behavior. Zootaxa 3768, 95 – 100. (doi:10.11646/zootaxa. 3768.1.7)

27. Vandermeer J, Perfecto I, Philpott SM. 2008 Clusters of ant colonies and robust criticality in a tropical agroecosystem. Nature 451, 457 – 459. (doi:10. 1038/nature06477)

28. Torres CW, Brandt M, Tsutsui ND. 2007 The role of cuticular hydrocarbons as chemical cues for nestmate recognition in the invasive Argentine ant (Linepithema humile). Insectes Soc. 54, 363 – 373. (doi:10.1007/s00040-007-0954-5)

29. Heinze J, Walter B. 2010 Moribund ants leave their nests to die in social isolation. Curr. Biol. 20, 249 – 252. (doi:10.1016/j.cub.2009.12.031)

30. R Development Core Team. 2015 R: a language and environment for statistical computing (v 3.1.2). Vienna, Austria: R Foundation for Statistical Computing.

31. Vanables WN, Ripley BD. 2002 Modern applied statistics with S, 4th edn. New York, NY: Springer.

32. Mathis KA, Philpott SM. 2012 Current understanding and future prospects of host selection, acceptance, discrimination, and regulation of phorid fly parasitoids that attack ants. Psyche 2002, 1 – 9. (doi:10.1155/2012/895424)

33. Wheeler JW, Evans SL, Blum MS, Torgerson RL. 1975 Cyclopentyl ketones: identification and function in Azteca ants. Science 187, 254 – 255. (doi:10.1126/science.1111099)

34. Rosenheim JA, Kaya H, Lester E, Marois J, Jaffee B. 1995 Intraguild predation among biological-control agents: theory and evidence. Biol. Control 5, 303 – 335. (doi:10.1006/bcon.1995.1038)

35. Tofilski A, Couvillon MJ, Evison SEF, Helanterä H, Robinson EJH, Ratnieks FLW. 2008 Preemptive

defensive self-sacrifice by ant workers. Am. Nat. 172, E239 – E243. (doi:10.1086/591688)

36. Chapuisat M. 2010 Social evolution: sick ants face death alone. Curr. Biol. 20, 104 – 105. (doi:10.1016/ j.cub.2009.12.037)

37. Tomlinson I, Latta B. 1987 Adaptive and non- adaptive suicide in aphids. Nature 330, 701. (doi:10.1038/330701a0)

38. Stamp NE. 1981 Behavior of parasitized aposematic caterpillars: advantageous to the parastioid or the host? Am. Nat. 118, 715 – 725. (doi:10.1086/ 283863)

39. McAllister MK, Roitberg BD. 1987 Adaptive suicidal behaviour in pea aphids. Nature 328, 797 – 799. (doi:10.1038/328797b0)

40. Remfert J. 2012 Genetic variation and cluster formation in the ant Azteca in coffee agroecosystems. Master’s thesis. (http://www.snre. umich.edu/current_students/forms_policies/theses)

41. Philpott SM, Perfecto I, Vandermeer J, Uno S. 2009 Spatial scale and density dependence in a host parasitoid system: an arboreal ant, Azteca instabilis, and its Pseudacteon phorid parasitoid. Environ. Entomol. 38, 790 – 796. (doi:10.1603/022.038.0331)

42. Mynhardt G. 2013 Declassifying myrmecophily in the Coleoptera to promote the study of ant-beetle symbioses. Psyche 2013, 1 – 8. (doi:10.1155/2013/ 696401)

43. Rettenmeyer CW, Rettenmeyer ME, Joseph J, Berghoff SM. 2011 The largest animal association centered on one species: the army ant Eciton burchellii and its more than 300 associates. Insectes Soc. 58, 281 – 292. (doi:10.1007/s00040- 010-0128-8)

44. O’Keefe ST. 2000 Ant-like stone beetles, ants, and their associations (Coleoptera: Scydmaenidae; Hymenoptera: Formicidae; Isoptera). J. NY Entomol. Soc. 108, 273 – 303. (doi:10.1664/0028- 7199(2000)108[0273:ALSBAA]2.0.CO;2)

45. Parmentier T, Dekoninckn W, Wenseleers T. 2014 A highly diverse microcosm in a hostile world: a review on the associates of red wood ants (Formica rufa group). Insectes Soc. 61, 229 – 237. (doi:10.1007/s00040-014-0357-3)

46. Hsieh HY, Liere H, Jiménez-Soto E, Perfecto I. 2012 Cascading trait-mediated interactions induced by ant pheromones. Ecol. Evol. 2, 2181 – 2191. (doi:10. 1002/ece3.322)