at SciVerse ScienceDirect

Animal Behaviour 83 (2012) 737e745

Contents lists available

Animal Behaviour

journal homepage: www.elsevier .com/locate/anbehav

Hidden aggression in termite workers: plastic defensive behaviour dependentupon social context

Yuki Ishikawa 1, Toru Miura*

Graduate School of Environmental Science, Hokkaido University, Sapporo, Japan

a r t i c l e i n f o

Article history:Received 9 May 2011Initial acceptance 4 July 2011Final acceptance 30 November 2011Available online 20 January 2012MS. number: 11-00378R

Keywords:aggressionbehavioural plasticitycolony defenceHodotermopsis sjostedtisocial interactionsoldiertermite

* Correspondence: T. Miura, Graduate School of EnvUniversity, N10W5, Kitaku, Sapporo, 060-0810, Japan

E-mail address: miu@ees.hokudai.ac.jp (T. Miura).1 Present address: Graduate School of Life Science

Japan.

0003-3472/$38.00 � 2011 The Association for the Studoi:10.1016/j.anbehav.2011.12.022

Social insect colonies are self-organized systems that respond to changes in environmental conditionsby altering the relative proportions of certain castes or individuals engaged in specific tasks. Whiletermites are known to regulate the ratio of morphologically specialized soldiers in response to theproportion of nestmate castes, the soldier differentiation process requires a relatively long time to beaccomplished. Consequently, termites are thought to employ other plastic and flexible mechanisms forcolony defence. We quantified the defensive behaviours of soldiers, pseudergates (workers) andreproductives (neotenics) of the damp-wood termite Hodotermopsis sjostedti. When individuals of thecaste were exposed to an intruder, soldiers exhibited the most vigorous defensive behaviour, followedby pseudergates and then reproductives. While the aggression levels of soldiers and reproductiveswere independent of the accompanying castes, pseudergate aggression was more plastic and dependedon the other castes present. When paired with reproductives, pseudergates exhibited high levels ofaggression towards enemies. However, pseudergate aggression levels remained low when they werepaired with soldiers, suggesting that pseudergates moderated their defensive behaviour depending onsocial context. Plasticity with respect to social behaviours may facilitate rapid and flexible responsesrequired for colony defence.� 2011 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

A variety of animal taxa modulate their behaviour and/ormorphology in response to rapid changes in their environment.By changing the ratios of castes or individuals that performspecific functions within the colony, social insects exhibit plasticresponses at the colony level (Wilson 1971). The regulatorymechanisms employed by a colony to respond to environmentalchanges without any centralized method of control are one of themajor themes in sociobiology (Gordon 1996; Beshers & Fewell2001).

The importance of colony defence in social insects is clearlyapparent from the numerous defensive mechanisms that haveevolved in eusocial taxa (Wilson 1975). These defence mecha-nisms include the use of cryptic habitats, the construction ofstrong and robust nests, nestmate aggregation, and the produc-tion of morphologically and/or behaviourally specialized castes.The roles of the defensive castes, such as soldiers or guards, arewell defined and have been acquired independently in numerouseusocial taxa (Hermann 1984). Despite their potential contribu-tion to colony survival, the relatively high costs associated with

ironmental Science, Hokkaido.

s, Tohoku University, Sendai,

dy of Animal Behaviour. Published

soldier production may result in the production of fewer workers,which take care of the soldiers in addition to their other functionssuch as foraging and colony maintenance (Rivera-Marchand et al.2008). Thus, the proportion of soldiers or guards must be opti-mized in response to the defensive demands of the surroundingenvironment.

Termites are highly vulnerable to attacks by a wide variety ofpredators (Deligne et al. 1981). The soldier caste of termitescontributes directly to colony defence by aggressively attackingintruders using specialized morphological adaptations, such asmandibles or defensive glands (Stuart 1967; Howse 1970;Deligne et al. 1981; Prestwich 1984). Optimization of soldierratios is also expected in termites because, while they areeffective for colony defence, having too many soldiers placesa burden on the colony as mentioned above. The mechanismsresponsible for controlling the proportion of soldiers withina colony have been investigated previously in classic studies,which suggested that the presence of soldiers inhibits thedifferentiation of new soldiers from workers (or pseudergates),while the presence of reproductive castes or nymphs acceleratesit (Castle 1934; Miller 1942; Springhetti 1970; Nagin 1972; Lenz1976; Haverty & Howard 1981; Lefeuve & Bordereau 1984).Although encounters with enemies can be sudden and unpre-dictable, the response of the colony with respect to defensive

by Elsevier Ltd. All rights reserved.

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745738

task allocation through controlling caste ratios is limited by thefact that soldier differentiation requires 15e30 days to complete(Ogino et al. 1993; Cornette et al. 2008). Consequently, thedevelopment of defence systems that are more flexible and rapidwould be expected.

Workers of social insects are the primary workforce in a colony,performing tasks such as foraging, brood care, nest maintenanceand sometimes colony defence. These workers can respond tochanges in the surrounding environment independently of theirage or morphology (Gordon 1996). Indeed, in some ant species,workers modify their aggression without undergoing anymorphological modifications depending on the presence of othernestmates (Tanner 2006, 2008; Tanner & Adler 2009). Therefore,we particularly focused on termite workers as the potentialdefensive force within a colony when soldiers were absent.Although worker termites are known to defend their colonies,primarily through actions involving the construction and repair ofnest structures, direct defensive behaviours (e.g. biting) have only

Figure 1. (a) Caste differentiation in Hodotermopsis sjostedti. Larvae undergo six moults toreproductive castes, such as alates and neotenics. (bed) Experimental designs for the quanintrusion (Figs 2, 3), biting frequency and duration of attack posture under (b) isolated or ((d) To quantify the behavioural response to nest opening (Fig. 4), head banging, guarding, popening of the gate.

been reported in soldier-less species or in species with soldiersthat have well-developed frontal glands and reduced mandibles(Stuart 1967; Eisner et al. 1976; Deligne et al. 1981; Traniello &Beshers 1985).

In this study, we investigated the context-dependent plasticityof defensive behaviours in soldiers, pseudergates and reproduc-tives of the damp-wood termite Hodotermopsis sjostedti. Thistermite species is considered to retain several primitive eusocialcharacteristics (Thorne & Traniello 2003), such as biting-typesoldiers and a linear caste-differentiation pathway (Fig. 1a;Miura et al. 2000, 2004). In this species, true workers are absent,and older juveniles (referred to as pseudergates; etymologically‘false workers’) perform worker tasks. Pseudergates are sociallyactive and potentially capable of developing into imagos (Grassé& Noirot 1947; Noirot 1985; Thorne 1996; Parmentier & Roisin2003). In H. sjostedti, after hatching from eggs, larvae undergosix moults before developing into pseudergates (Miura et al.2000, 2004). While pseudergates are totipotent and capable of

develop into pseudergates, some of which then differentiate into sterile soldiers ortification of defensive behaviour. (b, c) To quantify the behavioural response to enemyc) paired conditions were recorded for 3 min while the termite was exposed to an ant.atrolling, walking, gnawing and grooming were recorded for a 10 min period following

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745 739

differentiating into reproductives or sterile soldiers, most remainpseudergates through stationary moults (Fig. 1a; Miura et al.2000, 2004).

We initially examined differences in defensive responses amongindividual soldiers, pseudergates and reproductives when exposedto the ant Formica japonica. We then examined whether the caste-specific tendencies of the defensive response were altered by thepresence of nestmates. Since defensive tasks (e.g. encounteringenemies) are usually restricted to the peripheral regions of a nest,caste localization within the nest is also important for defensivetask allocation. Therefore, we also examined the positioning andbehavioural responses of the castes at the time of nest opening inthe presence of other nestmates.

METHODS

Insects

Colonies of H. sjostedti were sampled from decaying wood inevergreen forests on Yakushima Island in Kagoshima Prefecture,Japan. Colonies were maintained in the laboratory as stock atapproximately 25 �C under constant darkness. Soldiers,neotenics and pseudergates were used for the behaviouralexperiments. As in previous studies (Miura et al. 2000, 2004),seventh-instar larvae were regarded as pseudergates in thisstudy. This species is considered to have two reproductive castes,primary reproductives and neotenics (supplementary orsecondary reproductives; Miura et al. 2000). As is common inlower termites, most mature H. sjostedti colonies in nature donot contain a pair of primary reproductives, but have dozens ofneotenics instead (Matsumoto & Hirono 1985). In this study,therefore, neotenics were regarded as being representatives ofthe reproductive caste.

Experimental Set-up

To compare defensive behaviours among castes, we introducedindividual termites into a plastic petri dish (diameter: 60 mm)which was placed in an arena (14 � 10 cm) in a plastic enclosure.The dish, which was considered to represent an artificial nest, waslined with moist filter paper and had an opening representing themain entrance (hereafter referred to as the gate), which couldeasily be opened or closed by rotating the lid of the dish. Onceinside the artificial nest, the termite was exposed to an ant(F. japonica), which was placed on the end of a toothpick andpositioned inside the gate (Fig. 1b, c). The tip of the toothpick wasattached to the dorsal thorax of the ant with adhesive, so that itcould move its legs freely.

Table 1The effects of defensive behaviour on the enemy ants

Castes Accompanying castes No. of experiments No.

Soldier e 16 3Soldier 13 3Pseudergate 15 2Neotenic 15 3

Pseudergate e 17 3Soldier 15 2Pseudergate 18 3Neotenic 16 2

Neotenic e 17 3Soldier 15 3Pseudergate 16 2Neotenic 15 3

* Number of killed/injured/uninjured ants after the behavioural analyses.

Caste Responses to Enemy Invasion

We compared aggressive responses among castes in responseto enemy invasion under isolated conditions in which individualtermites were introduced into a nest (Fig. 1b). After the termitehad been allowed to acclimatize in the nest for 15 min, the gatewas opened and a live ant, placed on the tip of a wooden tooth-pick, was positioned inside the open gate. The behaviour of thetermite was then recorded using a digital video camera recorder(Handycam DCR-PC5, Sony Corporation, Japan) for 3 min afterencountering the ant. Defensive behaviours were defined as(1) biting, if the termite bit the ant, or (2) assuming an attackposture, if the termite directed its head towards the ant in the areadelimited by the dashed line in Fig. 1b. The number of times eachtermite opened and closed its mandibles, and the amount of timethe termite assumed an attack posture were measured and used asindexes of biting and attack posture, respectively. Both sexes ofeach caste were used in the behavioural assays. However, since thesex of the termite did not affect the extent of the behaviouralresponse (ManneWhitney U test: biting: U ¼ 288, Nmale ¼ 35,Nfemale ¼ 11, P > 0.05; attack posture: U ¼ 218.5, Nmale ¼ 35,Nfemale ¼ 11, P > 0.05), we did not discriminate between the sexeswhen analysing the data. The number of experiments and coloniesanalysed are shown in Table 1.

Next, to clarify whether any caste-specific tendencies of thedefensive response existed with respect to the presence ofnestmates, behavioural observations were conducted underpaired conditions (Fig. 1c). Under paired conditions, thebehavioural responses of two termite individuals that had beenplaced inside the nest for 15 min and exposed to an ant at thegate were recorded for 3 min after one of the termites hadencountered the ant intruder. The paired individuals wereselected from the same colony. Numbers of trials and coloniesare described in Table 1.

Differences in the biting frequency or the duration of an attackposture were assessed statistically using the KruskaleWallis test.When the overall KruskaleWallis tests were significant,SteeleDwass tests were employed to evaluate the significantdifferences between each individual category. Because the anal-yses of the defensive behaviours included multiple comparisons,the cutoff point for rejection of the null hypothesis in each testwas determined by the procedure of Guo & Rao (2008) to keepthe false discovery rate below 5%. The approach is a modified‘step-down’ procedure of Benjamini & Hochberg (1995), and canbe applied irrespective of the strength of correlation among thetests. The control of P values by the original method of Benjamini& Hochberg (1995), which can be applied under the assumptionof independence or positive correlation, gave the same results.

of colonies Killed ants* Injured ants* Uninjured ants*

15/16 1/16 0/1612/15 3/15 1/1512/15 3/15 0/1514/15 1/15 0/154/17 8/17 5/17

12/15 3/15 0/156/18 8/18 4/186/16 8/16 2/160/17 0/17 17/17

14/15 1/15 0/156/16 8/16 2/160/15 0/15 15/15

Table 2Statistical values and results of Benjamini & Hochberg’s (1995) false discovery rate (FDR) control

Method Behaviour Pairwise df c2/t value Observed P Rank of P FDR alphalevel

Isolated conditionsKruskaleWallis Biting 2 34.74 2.9E-08 1/2 0.025

Attack position 2 34.44 3.3E-08 2/2 0.050SteeleDwass Biting N PE 4.815 4.4E-06 3/6 0.025

N S 4.927 2.5E-06 1/6 0.008PE S 2.469 0.036 6/6 0.050

Attack position N PE 3.550 0.001 5/6 0.042N S 4.866 3.4E-06 2/6 0.017PE S 4.217 7.4E-05 4/6 0.033

Paired conditionsKruskaleWallis Biting 8 144.56 1.1E-28 1/2 0.025

Attack position 8 123.49 1.1E-24 2/2 0.050SteeleDwass Biting S with N S with PE 0.830 0.996 65/72 0.045

S with S 1.556 0.828 62/72 0.043PE with N 0.198 1.000 70/72 0.049PE with PE 2.739 0.134 52/72 0.036PE with S 4.295 0.001 38/72 0.026N with N 5.881 1.5E-07 5/72 0.004N with PE 4.928 2.9E-05 22/72 0.016N with S 4.987 2.2E-05 20/72 0.014

S with PE S with S 2.639 0.170 53/72 0.037PE with N 0.613 1.000 67/72 0.047PE with PE 3.783 0.005 44/72 0.031PE with S 4.542 1.9E-04 34/72 0.024N with N 5.881 1.5E-07 5/72 0.004N with PE 4.928 2.9E-05 22/72 0.016N with S 4.987 2.2E-05 20/72 0.014

S with S PE with N 1.981 0.556 55/72 0.038PE with PE 1.726 0.730 60/72 0.042PE with S 4.602 1.5E-04 30/72 0.021N with N 6.675 8.9E-10 2/72 0.001N with PE 5.469 1.6E-06 9/72 0.006N with S 5.412 2.2E-06 11/72 0.008

PE with N PE with PE 3.173 0.040 48/72 0.033PE with S 4.487 2.5E-04 35/72 0.024N with N 5.975 8.3E-08 4/72 0.003N with PE 4.994 2.1E-05 18/72 0.013N with S 5.036 1.7E-05 17/72 0.012

PE with PE PE with S 3.815 0.004 43/72 0.030N with N 6.996 9.5E-11 1/72 0.001N with PE 5.685 4.7E-07 7/72 0.005N with S 5.654 5.6E-07 8/72 0.006

PE with S N with N 5.228 6.1E-06 16/72 0.011N with PE 4.785 6.0E-05 26/72 0.018N with S 4.988 2.2E-05 19/72 0.013

N with N N with PE 0.216 1.000 69/72 0.048N with S 1.834 0.660 58/72 0.040

N with PE N with S 1.733 0.726 59/72 0.041Attack position S with N S with PE 0.568 1.000 68/72 0.047

S with S 0.793 0.997 66/72 0.046PE with N 2.493 0.235 54/72 0.038PE with PE 4.593 1.5E-04 31/72 0.022PE with S 4.464 2.8E-04 36/72 0.025N with N 5.348 3.2E-06 12/72 0.008N with PE 4.748 7.2E-05 28/72 0.019N with S 4.547 1.9E-04 33/72 0.023

S with PE S with S 0.055 1.000 72/72 0.050PE with N 1.684 0.757 61/72 0.042PE with PE 3.871 0.003 42/72 0.029PE with S 3.970 0.002 41/72 0.028N with N 5.326 3.6E-06 14/72 0.010N with PE 4.713 8.5E-05 29/72 0.020N with S 4.178 0.001 40/72 0.028

S with S PE with N 1.947 0.580 56/72 0.039PE with PE 4.865 4.0E-05 25/72 0.017PE with S 4.586 1.6E-04 32/72 0.022N with N 6.266 1.3E-08 3/72 0.002N with PE 5.322 3.7E-06 15/72 0.010N with S 4.884 3.6E-05 24/72 0.017

PE with N PE with PE 3.192 0.038 47/72 0.033PE with S 3.677 0.007 45/72 0.031N with N 5.329 3.5E-06 13/72 0.009N with PE 4.749 7.1E-05 27/72 0.019

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745740

Table 2 (continued )

Method Behaviour Pairwise df c2/t value Observed P Rank of P FDR alphalevel

N with S 4.190 0.001 39/72 0.027PE with PE PE with S 1.530 0.842 63/72 0.044

N with N 5.454 1.8E-06 10/72 0.007N with PE 4.372 4.2E-04 37/72 0.026N with S 1.871 0.634 57/72 0.040

PE with S N with N 3.540 0.012 46/72 0.032N with PE 2.985 0.070 50/72 0.035N with S 0.187 1.000 71/72 0.049

N with N N with PE 1.142 0.968 64/72 0.044N with S 3.107 0.049 49/72 0.034

N with PE N with S 2.827 0.108 51/72 0.035Nest openingKruskaleWallis Head banging 3 30.87 9.1E-07 1/6 0.008

Walking 3 6.85 0.077 5/6 0.042Grooming 3 17.42 0.001 3/6 0.025Gnawing 3 13.36 0.004 4/6 0.033Guarding 3 28.57 2.8E-06 2/6 0.017Patrolling 3 3.06 0.382 6/6 0.050

SteeleDwass Head banging N with PE PE with N 3.832 0.001 2/12 0.008PE with N PE with S 2.858 0.022 6/12 0.025PE with S S with PE 3.844 0.001 11/12 0.004

Grooming N with PE PE with N 1.734 0.306 10/12 0.042PE with N PE with S 0.296 0.991 12/12 0.050PE with S S with PE 3.647 0.002 4/12 0.017

Gnawing N with PE PE with N 2.147 0.138 8/12 0.033PE with N PE with S 0.789 0.860 11/12 0.046PE with S S with PE 2.761 0.029 7/12 0.029

Guarding N with PE PE with N 3.250 0.006 5/12 0.021PE with N PE with S 1.938 0.212 9/12 0.038PE with S S with PE 3.777 0.001 3/12 0.013

S: Soldier; PE: Pseudergate; N: Neotenic. Observed P values are ranked in descending order of significance. Observed P values that remained significant after correction arehighlighted in bold.

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745 741

Alpha levels were calculated for each statistical test of eachexperiment (Table 2).

Responses to Nest Opening

Under both natural and laboratory conditions, when termitenests are exposed to open air, all colony members respond to thedisturbance (some escape to the centre of the nest while othersexhibit an aggressive response; Eisner et al. 1976). We thereforeexamined the responses of the termite castes to nest opening interms of the positions they assume and their behaviours.

Five pseudergates were introduced into a nest together witheither five nestmate soldiers or five nestmate neotenics (N ¼ 11groups, Fig. 1d). Five colonies were used for groups of pseudergatesand soldiers, and two colonies were used for groups of pseu-dergates and neotenics. In preliminary studies, 15 min of acclima-tization was not sufficient to keep the termites inside the openednest. Consequently, we left the 10 individuals to be tested in thenest overnight before initiating the experiments. All of the termiteswere marked with oil paint to facilitate individual identification.After placing the nest in the arena for 15 min, we opened the gateand recorded the behaviours of all of the termites in the dish withthe digital video camera for 10 min. In this experiment, theobserved defensive behaviours were defined as (1) head banging,considered as an alarm signal (Howse 1964), if the termiterepeatedly beat its head on the ground, (2) guarding, if the termitedirected its head towards the gate in the region delineated by thedashed line in Fig. 1d, and (3) patrolling, if the termite walkedaround the outside of the petri dish in the arena. In addition,examples of nondefensive behaviours were also defined, including(4) walking, if termites walked around in the dish, (5) gnawing, iftermites gnawed the filter paper or dish wall, and (6) grooming, ifthe termites groomed other individuals.

Differences in each defensive and nondefensive behaviourbetween all individual categories (neotenics with pseudergates,pseudergates with neotenics, pseudergates with soldiers, andsoldiers with pseudergates) were assessed statistically using theKruskaleWallis test. When the KruskaleWallis test for thebehaviour was significant, SteeleDwass tests were employed toevaluate the significant differences between the individual cate-gories. Because the analyses included multiple comparisons, thecutoff point for rejection of the null hypothesis in each test wasdetermined by the procedure of Guo & Rao (2008) to keep thefalse discovery rate below 5%. Alpha levels were calculatedindependently for each statistical test of each experiment(Table 2).

RESULTS

Responses towards Intruders when Isolated

As in nature, under isolated conditions, soldiers exhibiteddefensive behaviours more frequently than the other castes (Fig. 2,Table 2). After encountering an ant, the soldiers bit it repeatedlyuntil it was dead; the frequency of this defensive behaviour thendecreased. Soldiers rarely exhibited an escape response afterencountering an ant. Pseudergates also exhibited several defensivebehaviours under isolated conditions (Fig. 2). Upon encounteringan ant, pseudergates often bit the ant several times beforeescaping to take up an attack posture that was shorter in durationthan that observed in soldiers (Fig. 2b); although less effectivethan soldiers, most of the pseudergates that attacked the antseither killed or injured the intruder (Table 1). Neotenics exhibitedthe weakest defensive responses of all of the castes tested, interms of both biting and the adoption of an attack posture (Fig. 2,Table 2).

30

25

20

Bit

ing/

min

(fr

equ

ency

)15

10a

b

c

a

b

c

5

0Neotenic

(a)

Pseudergate Soldier

60

50

40

Att

ack

pos

ture

/min

(s)

30

20

10

0Neotenic

(b)

Pseudergate Soldier

Figure 2. Differences in the defensive behaviour of castes in response to enemy intrusion under isolated conditions: (a) biting and (b) attack posture. Box plots show the median,upper and lower quartiles and range; circles indicate outliers. Different letters indicate statistical significance (KruskaleWallis test followed by SteeleDwass test: P < 0.05).

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745742

Responses towards Intruders when Paired

In the presence of nestmates under paired conditions, soldierswere consistently aggressive while neotenics were seldomaggressive, and these levels of aggression were generally main-tained regardless of which castes were placed together (Fig. 3).Duration of attack postures of neotenics paired with soldiers ten-ded to be higher than when neotenics were paired with neotenics(Fig. 3, Table 2), because neotenics spent more time with the antkilled by soldiers, assuming an attack posture.

However, pseudergates dramatically altered their defensivebehaviour depending on which caste they were paired with (Fig. 3,Table 2). When paired with soldiers, the frequency of defensivebehaviours exhibited by pseudergates was lower than when pairedwith pseudergates or neotenics, and was similar in intensity to thatexhibited by neotenics (Fig. 3, Table 2). On the other hand, the

30

25

20

15

10

Bit

ing/

min

(fr

equ

ency

)A

ttac

k p

ostu

re/m

in (

s)

5

0 ab

(a) (b)

a

a a

ab

ab

ab

a

With N With PE With S

60

50

40

30

20

10

0With N With PE

Soldier PseuWith S

60

50

40

30

20

10

0With N Wi

30

25

20

15

10

5

0With N W

(d)

Figure 3. Differences in the defensive behaviour of castes in response to enemy intrusiondergates (PE) and (c, f) neotenics (N). Box plots show the median, upper and lower quartil(KruskaleWallis test followed by SteeleDwass test: P < 0.05).

frequency of defensive responses in pseudergates paired withneotenics was higher than when pseudergates were paired withsoldiers or pseudergates (Fig. 3, Table 2). Indeed, when paired withneotenics, the levels of aggression among pseudergates did notdiffer significantly from that exhibited by soldiers (Fig. 3, Table 2).The escape behaviour of pseudergates decreased when paired withneotenics (data not shown).

Responses to Nest Opening

In an experimental group consisting of five neotenics and fivepseudergates, the pseudergates positioned themselves near thegate and exhibited more aggression than neotenics (Fig. 4,Appendix Fig. A1, Table 2). In an experimental group consisting offive pseudergates and five soldiers, soldiers positioned themselvesnear the gate and exhibited higher frequencies of defensive

(c)

bccd

ede

cde

b

c

d d d

dergate Neotenicth PE With S

60

50

40

30

20

10

0With N With PE With S

ith PE With S

30

25

20

15

10

5

0With N With PE With S

(e) (f)

under paired conditions: (aec) biting and (def) attack. (a, d) Soldiers (S), (b, e) pseu-es and range; circles indicate outliers. Different letters indicate statistical significance

500 (a)

(b)

(c)

400 ****

**

300

200

100

0

2500

2000

**

**

1500

1000

500

0

800

600

400

Patr

olli

ng/

10 m

in (

s)G

uar

din

g/10

min

(s)

Hea

d b

angi

ng/

10 m

in(f

req

uen

cy)

200

Neotenicwith PE

Pseudergatewith N

Pseudergatewith S

Soldierwith PE

0

Figure 4. Differences in the defensive behaviour of soldiers (S), pseudergates (PE) andneotenics (N) in response to nest disturbance (nest opening): (a) head banging,(b) guarding and (c) patrolling. Box plots show the median, upper and lower quartilesand range; circles indicate outliers. Asterisks indicate statistical significance (Krus-kaleWallis test followed by SteeleDwass test: **P < 0.01).

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745 743

behaviour than the pseudergates (Fig. 4, Appendix Fig. A1, Table 2).Similar to the direct behavioural responses against an intruder(Fig. 3), the positioning and defensive responses of pseudergates tonest disturbance seemed to be affected by the coexisting castes. Thefrequency of the head-banging (alarm) behaviour and the locali-zation (guarding) in pseudergates grouped with neotenics tended

to be higher than when they were grouped with soldiers (Fig. 4,Appendix Fig. A1, Table 2).

DISCUSSION

Our results showed that pseudergates possess the potential toattack enemies by altering their behaviour in a context-dependentmanner. When exposed to an intruding ant, pseudergatesexhibited relatively low levels of aggression in the presence ofsoldiers, but very high levels of aggression in the presence ofneotenics (Figs 3, 5). A similar result was observed in the mannerin which pseudergates located themselves within the nest inresponse to a nest disturbance (Figs 4, 5). Based on these findings,pseudergates under natural conditions are thought to retreat deepinside the nest if there are soldiers and no reproductives aroundthem, or to remain on site and attack intruders if there arereproductives and no soldiers. A similar finding was suggested byRoisin et al. (1990), who found that defensive responses inworkers of Nasutitermes princeps also decreased in the presence ofsoldiers. Since these behavioural changes can occur withinminutes after the defence demands of a colony have changed,a flexible and rapid defensive strategy can be achieved through thebehavioural plasticity of workers (or pseudergates). Contextdependence was also observed in the alarm response, that is, morefrequent head banging of pseudergates in the absence of soldiers.Although a behavioural response to such an alarm signal has notyet clearly been demonstrated, such signals are considered to actas a trigger for the recruitment of additional soldiers (and/orworkers and pseudergates) to the sites of reproductives. Thus, theplastic defensive response of workers or pseudergates may playessential roles in the optimization of colony defence by compen-sating for the relatively slow responsiveness in caste differentia-tion (Fig. 5).

The findings of this study also suggest that pseudergates possessthe ability to recognize the castes in their vicinity. Since thecompound eyes and optic lobes of workers and pseudergates aremarkedly reduced, it is unlikely that they discriminate castes byvisual cues. Recent studies on caste-specific chemicals secretedfrom exocrine glands or epithelia (Miura et al. 1999; Hanus et al.2009; Korb et al. 2009; Liebig et al. 2009; Weil et al. 2009)suggest that either odour or contact signals act as caste recognitioncues. Sound is another potential candidate signal for distinguishingbetween castes, as soldiers head banged more frequently thanpseudergates and neotenics (Fig. 4), and the actual soundsproduced by soldiers may also differ from those produced bypseudergates and neotenics (Howse 1964). To clarify these differentcaste recognition cues more accurately, experiments using chem-icals or audio stimuli in isolation will be required.

Since the presence of castes affects both the behaviours anddevelopment of pseudergates (Castle 1934; Miller 1942;Springhetti 1970; Nagin 1972; Lenz 1976; Haverty & Howard1981; Lefeuve & Bordereau 1984; present study), it is likely thatsome physiological links exist between changes in aggression andthe regulation of soldier differentiation (Fig. 5). It is well knownthat juvenile hormone (JH) plays an important role in the castedetermination of termites, with exposure to high JH levels duringintermoult periods triggering the differentiation of workers intosoldiers, while exposure to low JH levels is considered to inducealate differentiation (Nijhout &Wheeler 1982; Cornette et al. 2008).Recent studies have suggested that insect aggression is also affectedby JH; for example, high JH titres cause increased male aggressionin lobster cockroaches, Nauphoeta cinerea (Kou et al. 2009). Thissuggests that in termites, the recognition of nestmates maypossibly alter JH levels in pseudergates, resulting in both changes inaggression and caste determination (Fig. 5).

Figure 5. Regulation mechanism of colony defence in the focal termite species based on the results of this study. Pseudergates are plastic with respect to both aggression (defensivebehaviour) and developmental fates, while soldiers and neotenics exhibit stable behaviour and are developmentally terminal. The presence of soldiers represses both soldierdifferentiation and worker aggression, while the presence of neotenics promotes them.

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745744

Since almost all extant termite species possess distinctivesoldier castes (the loss of a soldier caste in some termites isconsidered to have occurred secondarily), the evolutionary processby which the morphologically and behaviourally unique soldiercaste has been originated (as an evolutionary novelty) is difficult topostulate. Deligne et al. (1981) suggested that the soldier caste wasderived fromworkers, but the evolutionary process through whichhighly differentiated soldiers have originated from less aggressiveand undifferentiated workers remains unclear. In this study,pseudergates exhibited aggression comparable to soldiers incertain situations. If the workers or pseudergates (i.e. helpers) ofancestral termite species exhibited a similar behavioural plasticity,then it can be hypothesized that the plastic nature of aggressivebehaviour may have varied among genetically distinct colonies orpopulations. This plastic responsemay then gradually have becomefixed as morphological specialization occurred, either subsequentlyor concurrently, in response to predation pressure. This wouldsatisfy the classic idea that behavioural plasticity is the first stepinvolved in evolutionary change, and that this change is then suc-ceeded by morphological specialization (Baldwin 1896, 1902;Wcislo 1989). To verify this scenario, further studies into thegenetic variation underlying plastic characters need to beconducted.

Acknowledgments

We are grateful to K. Maeno, Y. Hayashi, M. K. Hojo, A. Ishikawaand K. Ogawa for their valuable comments on this study. We thankW. M. Iwasaki, A. S. Tanabe and M. P. Sato for their support in thestatistical analysis. We also thank Y. Hayashi, M. K. Hojo, T. Yodoiand K. Tanabe for their assistance in field sampling and laboratoryexperiments. This work was supported by a Grant-in-Aid for Young

Scientists (No. 21677001) from the Ministry of Education, Culture,Sports, Science and Technology of Japan. Y.I. was supported byJapan Society for the Promotion of Science (JSPS) ResearchFellowship for Young Scientists.

References

Baldwin, J. M. 1896. A new factor in evolution. American Naturalist, 30, 441e451.Baldwin, J. M. 1902. Development and Evolution. New York: Macmillan.Benjamini, Y. & Hochberg, Y. 1995. Controlling the false discovery rate: a practical

and powerful approach to multiple testing. Journal of the Royal Statistical SocietyB, 57, 289e300.

Beshers, S. N. & Fewell, J. H. 2001. Models of division of labor in social insects.Annual Review of Entomology, 46, 413e440.

Castle, G. 1934. The damp-wood termites of the western United States, genusZootermopsis (formerly, Termopsis). In: Termites and Termite Control (Ed. byC. Kofoid), pp. 273e310. Berkeley: University of California Press.

Cornette, R., Gotoh, H., Koshikawa, S. & Miura, T. 2008. Juvenile hormone titersand caste differentiation in the damp-wood termite Hodotermopsis sjostedti(Isoptera, Termopsidae). Journal of Insect Physiology, 54, 922e930.

Deligne, J., Quennedey, A. & Blum, M. 1981. The enemies and defence mechanismsof termites. In: Social Insects (Ed. by H. Hermann), pp. 1e76. London: AcademicPress.

Eisner, T., Kriston, I. & Aneshansley, D. J. 1976. Defensive behavior of a termite(Nasutitermes exitiosus). Behavioral Ecology and Sociobiology, 1, 83e125.

Grassé, P. P. & Noirot, C. 1947. Le polymorphisme social du termite á cou jaune(Calotermes flavicollis F.). Les faux-ouvriers ou pseudergates et les muesrégressives. Les Comptes Rendus de l’Académie des Sciences de Paris, 224,219e221.

Gordon, D. M. 1996. The organization of work in social insect colonies. Nature, 380,121e124.

Guo, W. G. & Rao, M. B. 2008. On control of the false discovery rate under noassumption of dependency. Journal of Statistical Planning and Inference, 138,3176e3188.

Hanus, R., Vrkoslav, V., Hrdý, I., Cvacka, J. & Sobotnik, J. 2009. Beyond cuticularhydrocarbons: evidence of proteinaceous secretion specific to termite kings andqueens. Proceedings of the Royal Society B, 277, 995e1002.

Haverty, M. & Howard, R. 1981. Production of soldier proportions by laboratoryexperimental groups of Reticulitermes flavipes (Koller) and Reticulitermes vir-ginicus (Banks) (Isoptera: Rhinotermitidae). Insectes Sociaux, 28, 32e39.

Appendix

2000 (a)

(b)

**

(c)

1500

1000

500

0

300

250

200

150

100

50

0

400

300

200

Gn

awin

g/10

min

(s)

Gro

omin

g/10

min

(s)

Wal

kin

g/10

min

(s)

100

Neotenicwith PE

Pseudergatewith N

Pseudergatewith S

Soldierwith PE

0

Figure A1. Difference in nondefensive behaviours of soldiers (S), pseudergates (PE)and neotenics (N) in response to nest opening: (a) walking, (b) grooming and (c)gnawing. Box plots show the median, upper and lower quartiles and range; circlesindicate outliers. Asterisks indicate statistical significance (KruskaleWallis test fol-lowed by SteeleDwass test: **P < 0.01).

Y. Ishikawa, T. Miura / Animal Behaviour 83 (2012) 737e745 745

Hermann, H. 1984. Defensive Mechanisms in Social Insects. New York: Praeger.Howse, P. E. 1964. The significance of the sound produced by the termite

Zootermopsis angusticollis (Hagen). Animal Behaviour, 12, 284e300.Howse, P. E. 1970. Termites: a Study in Social Behavior. London: Hutchinson

University Library.Korb, J., Weil, T., Hoffmann, K., Foster, K. R. & Rehli, M. 2009. A gene necessary for

reproductive suppression in termites. Science, 324, 758.Kou, R., Chou, S. Y., Chen, S. C. & Huang, Z. Y. 2009. Juvenile hormone and the

ontogeny of cockroach aggression. Hormones and Behavior, 56, 332e338.Lefeuve, P. & Bordereau, C. 1984. Soldier formation regulated by a primer

pheromone from the soldier frontal gland in a higher termite, Nasuti-termes-Lujae. Proceedings of the National Academy of Sciences, U.S.A., 81,7665e7668.

Lenz, M. 1976. The dependence of hormone effects in termite caste determinationon external factors. In: Phase and Caste Determination in Insects (Ed. byM. Lüscher), pp. 73e89. Oxford: Pergamon Press.

Liebig, J., Eliyahu, D. & Brent, C. S. 2009. Cuticular hydrocarbon profiles indicatereproductive status in the termite Zootermopsis nevadensis. Behavioral Ecologyand Sociobiology, 63, 1799e1807.

Matsumoto, T. & Hirono, Y. 1985. On the caste composition of a primitive termiteHodotermopsis japonicus Holmgren (Isoptera, Termopsidae). Scientific Papers ofthe College of Arts and Sciences, The University of Tokyo, 35, 211e216.

Miller, E. 1942. The problem of castes and caste differentiation in Prorhinotermessimplex Hagen. Bulletin of the University of Miami, 15, 1e27.

Miura, T., Kamikouchi, A., Sawata, M., Takeuchi, H., Natori, S., Kubo, T. &Matsumoto, T. 1999. Soldier caste-specific gene expression in the mandibularglands of Hodotermopsis japonica (Isoptera: Termopsidae). Proceedings of theNational Academy of Sciences, U.S.A., 96, 13874e13879.

Miura, T., Hirono, Y., Machida, M., Kitade, O. & Matsumoto, T. 2000. Castedevelopmental system of the Japanese damp-wood termite Hodotermopsisjaponica (Isoptera: Termopsidae). Ecological Research, 15, 83e92.

Miura, T., Koshikawa, S., Machida, M. & Matsumoto, T. 2004. Comparative studieson alate wing formation in two related species of rotten-wood termites:Hodotermopsis sjostedti and Zootermopsis nevadensis (Isoptera, Termopsidae).Insectes Sociaux, 51, 247e252.

Nagin, R. 1972. Caste determination in Neotermes jouteli (Banks). Insectes Sociaux,19, 39e61.

Nijhout, H. F. & Wheeler, D. E. 1982. Juvenile hormone and the physiological basisof insect polymorphisms. Quarterly Review of Biology, 57, 109e133.

Noirot, C. 1985. Pathways of caste development in the lower termites. In: CasteDifferentiation in Social Insects (Ed. by J. Watson, B. Okot-Kotber & C. Noirot), pp.41e58. Oxford: Pergamon Press.

Ogino, K., Hirono, Y., Matsumoto, T. & Ishikawa, H. 1993. Juvenile hormoneanalog, S-31183, causes a high-level induction of presoldier differentiation inthe Japanese damp-wood termite. Zoological Science, 10, 361e366.

Parmentier, D. & Roisin, Y. 2003. Caste morphology and development in Termi-togeton nr. planus (Insects, Isoptera, Rhinotermitidae). Journal of Morphology,255, 69e79.

Prestwich, G. D. 1984. Defense mechanisms of termites. Annual Review of Ento-mology, 29, 201e232.

Rivera-Marchand, B., Giray, T. & Guzman-Novoa, E. 2008. The cost of defense insocial insects: insights from the honey bee. Entomologia Experimentalis etApplicata, 129, 1e10.

Roisin, Y., Everaerts, C., Pasteels, J. M. & Bonnard, O. 1990. Caste-dependentreactions to soldier defense secretion and chiral alarm/recruitment pheromonein Nasutitermes princeps. Journal of Chemical Ecology, 16, 2865e2875.

Springhetti, A. 1970. Influence of the king and the queen on the differentiation ofsoldiers in Kalotermes flavicollis Fabr. (Isoptera). Monitore Zoologico Italiano, 4,99e105.

Stuart, A. M. 1967. Alarm, defense, and construction behavior relationships intermites (Isoptera). Science, 156, 1123e1125.

Tanner, C. J. 2006. Numerical assessment affects aggression and competitive ability:a team-fighting strategy for the ant Formica xerophila. Proceedings of the RoyalSociety B, 273, 2737e2742.

Tanner, C. J. 2008. Aggressive group behaviour in the ant Formica xerophila iscoordinated by direct nestmate contact. Animal Behaviour, 76, 1335e1341.

Tanner, C. J. & Adler, F. R. 2009. To fight or not to fight: context-dependentinterspecific aggression in competing ants. Animal Behaviour, 77, 297e305.

Thorne, B. L. 1996. Termite terminology. Sociobiology, 28, 253e263.Thorne, B. L. & Traniello, J. F. A. 2003. Comparative social biology of basal taxa of

ants and termites. Annual Review of Entomology, 48, 283e306.Traniello, J. F. A. & Beshers, S. N. 1985. Species-specific alarm recruitment

responses in a neotropical termite. Naturwissenschaften, 72, 491e492.Wcislo, W. T. 1989. Behavioral environments and evolutionary change. Annual

Review of Ecology and Systematics, 20, 137e169.Weil, T., Hoffmann, K., Kroiss, J., Strohm, E. & Korb, J. 2009. Scent of a queen-

cuticular hydrocarbons specific for female reproductives in lower termites.Naturwissenschaften, 96, 315e319.

Wilson, E. O. 1971. The Insect Societies. Cambridge, Massachusetts: The BelknapPress of Harvard University Press.

Wilson, E. O. 1975. Sociobiology: The New Synthesis. Cambridge, Massachusetts:Harvard University Press.