Weeding and grooming of pathogens in agricultureby antsCameron R. Currie1,2,3*{ and Alison E. Stuart4{{1Department of Botany, University ofToronto,Toronto, Ontario, Canada M5S 3B22SmithsonianTropical Research Institute, POBox 2027, Balboa, Republic of Panama3Section of Integrative Biology, University ofTexas at Austin, Austin,TX 78712, USA4Department of Zoology, University ofToronto,Toronto, Ontario, Canada M5S 1A2The ancient mutualism between fungus-growing ants and the fungi they cultivate for food is a textbookexample of symbiosis. Fungus-growing ants? ability to cultivate fungi depends on protection of the gardenfrom the aggressive microbes associated with the substrate added to the garden as well as from thespecialized virulent garden parasite Escovopsis. We examined ants? ability to remove alien microbesphysically by infecting Atta colombica gardens with the generalist pathogen Trichoderma viride and thespecialist pathogen Escovopsis. The ants sanitized the garden using two main behaviours: grooming ofalien spores from the garden (fungus grooming) and removal of infected garden substrate (weeding).Unlike previously described hygienic behaviours (e.g. licking and self-grooming), fungus-grooming andgarden-removal behaviours are speci?c responses to the presence of fungal pathogens. In the presence ofpathogens, they are the primary activities performed by workers, but they are uncommon in uninfectedgardens. In fact, workers rapidly eliminate Trichoderma from their gardens by fungus grooming andweeding, suggesting that these behaviours are the primary method of garden defence against generalistpathogens. The same sanitary behaviours were performed in response to the presence of the specialistpathogen Escovopsis. However, the intensity and duration of these behaviours were much greater in thistreatment. Despite the increased e?ort, the ants were unable to eliminate Escovopsis from their gardens,suggesting that this specialized pathogen has evolved counter-adaptations in order to overcome thesanitary defences of the ants.Keywords: behaviour; Escovopsis; fungus-growing ants; mutualism; pathogens; symbiosis
1. INTRODUCTIONFungus-growing ants? (Attini) ability to cultivate fungusfor food is relatively unique, occurring in only a few otherinsects. The attine fungus, which belongs mostly to thefamily Lepiotaceae (Chapela et al. 1994; Mueller et al.1998), serves as the main food source of the ants, whilethe ants provide the fungus with substrate on which togrow and a means of dispersal to new colonies by found-ress queens. The success of this ancient mutualismdepends on the ants? ability to protect their fungal culti-vars from being overgrown by the competitively superiormicrobes associated with the vegetative material that theworkers continuously add to the garden. In addition, thegardens of fungus-growing ants are host to a specializedand virulent fungal parasite in the genus Escovopsis(Ascomycota: anamorphic Hypocreales) (Currie et al.1999a). This fungal pathogen can devastate gardensrapidly, even in the presence of the ants and its persistentpresence signi?cantly reduces the growth rate of colonies,both in terms of garden mass and number of workers(Currie et al. 1999a; Currie 2001a). Knowledge of howfungus-growing ants are able to suppress or eliminate thespecialized pathogen Escovopsis and other potentialmicrobial invaders of the garden is fundamental to under-standing this fascinating mutualism.
Fungus-growing ants are able to suppress or partiallyeliminate Escovopsis through a mutualistic association withantibiotic-producing bacteria (Currie et al. 1999b, 2001).These ?lamentous bacteria are typically carried on thecuticle of fungus-growing ants (often in specialized loca-tions), and they produce antibiotics that speci?callytarget the growth of Escovopsis (Currie et al. 1999b). Theremoval of the bacterium from workers results in anincrease in the abundance and impact of the pathogenwithin infected gardens (Currie et al. 2001). Although thebacterium is important in helping the ants suppressEscovopsis, it does not appear to be the only defencebecause gardens tended by workers with the bacteriumremoved are not completely overwhelmed by this parasite(Currie et al. 2001). Many additional defences have beenproposed and examined, including the production ofantibiotics in the ants? metapleural glands and thedissemination of proteolytic enzymes into critical loca-tions by workers in order to increase the competitiveadvantage of the mutualistic fungus (Martin & Martin1970; Schildknecht & Koob 1970, 1971; Boyd & Martin1975). However, there is currently little empiricalevidence demonstrating the role of other mechanisms inpreventing the invasion of the garden by alien fungi orbacteria (seeWeber 1972; Currie 2000, 2001b).The physical removal of microbial contaminants byworkers, which is generally referred to as `weeding?, hasbeen assumed to be the main defence employed byfungus-growing ants in maintaining healthy gardens (e.g.Ho? lldobler & Wilson 1990; North et al. 1997). Consideringthe fundamental importance of defending the garden
Proc. R. Soc. Lond. B (2001) 268, 1033^1039 1033 ? 2001 The Royal SocietyReceived 20 October 2000 Accepted 2 February 2001
doi 10.1098/rspb.2001.1605
*Author for correspondence (ccurrie@mail.utexas.edu).{Present address: Department of Zoology, University of Otago, PO Box56, Dunedin, New Zealand.{Both authors contributed in equal part to this work.
from microbes, it is somewhat surprising that this rela-tively old idea, which was ?rst suggested by Mo? ller(1893), has received only limited study and has not beenwell described or well de?ned. Weber (1957) reportedobserving weeding of alien fungi in a short publishedabstract. The behavioural details of these observationswere subsequently never published and he later statedthat weeding was of little importance in fungus cultiva-tion by attine ants (see Weber 1972, pp. vii, 1 and 99).Fungus-growing ants minimize exposure of the garden toalien microbes by licking new substrate and all nestsurfaces with their tongues (Stahel & Geijskes 1939;Autuori 1941; Quinlan & Cherrett 1977, 1979) and it hasbeen suggested that this behaviour is important in elimi-nating alien microbes from within gardens (Bass & Cher-rett 1994). In addition, it appears that ants collect sporesin their infrabuccal pockets, which are ?ltering deviceswithin their mouth parts (Eisner & Happ 1962), andsubsequently discard the spores outside the nest in theform of pellets (Eisner & Happ 1962; Febvay &Kermarrec 1981). Nevertheless, the behaviours involved inphysically removing alien fungi have received little study.It is also unclear whether the ants are able to detect andphysically remove the specialized parasite Escovopsis.We video recorded Atta colombica colonies that wereexperimentally infected with one of two di?erent patho-genic fungi (Trichoderma viride and Escovopsis) in order toinvestigate whether fungus-growing ants are able todetect and physically remove alien fungi from theirgardens. Trichoderma spp. are aggressive necrotrophicfungal parasites (Dix & Webster 1995) that rapidly over-grow the fungus cultivated by leaf-cutter ants in pureculture bioassays (C. R. Currie, unpublished data). Yet,Atta spp. workers are able to preventTrichoderma spp. fromproliferating within their fungal gardens (Currie et al.1999a; Currie 2000; C. R. Currie, personal observation).Our experiment examined not only whether A. colombicaworkers are capable of removing general fungal contami-nants (i.e. Trichoderma) and/or the specialized pathogenEscovopsis, but also how the removal is performed for bothtypes of fungal invaders.
2. MATERIAL AND METHODSWe used 15 completely intact colonies of A. colombica in orderto examine the response of workers to alien fungi. Four- to?ve-month-old colonies were collected in Gamboa, Republic ofPanama, between the middle of September and late October.Colonies were kept at the University of Toronto in dualchamber systems with a single fungus garden in a plasticcontainer (diameter of 8 cm and height of 4.5 cm) locatedwithin a larger foraging and dump chamber (height of 10 cm,length of 19.5 cm and width of 14 cm). In order to preventbetween-colony movement of mites, which could serve asvectors for spores of Escovopsis and Trichoderma, colonies werekept on moats of heavy mineral oil and the inside edges ofcontainers were coated with ?uon (Northern Products, Inc.,Woonsocket, RI, USA). Water was added as needed to smalldishes with cotton pads in the outer chamber in order to main-tain an adequate level of humidity within colonies. Colonieswere provided ad libitum access to forage substrate, Quercus spp.and Euonymus spp., both before and during the experiment. Allcolonies used in this study were con?rmed to be free of
Escovopsis infection at the start of the experiment, following themethods of Currie et al. (1999a).The colonies were placed in groups of three according tosimilarity in size (i.e. garden volume), with one colony fromeach block (group of three) randomly receiving one out of threespraying treatments: Escovopsis sp., T. viride, or water (as acontrol). TheT. viride strain was collected from soil in the habitatoccupied by A. colombica in the Republic of Panama. The strainof Escovopsis was isolated from the garden of an A. colombicacolony collected from the same population as the colonies usedin this study. Escovopsis and T. viride were grown on potatodextrose agar medium for ten days prior to the study. Eachspore suspension was prepared by mixing fungal spores in steriledistilled water. A mist inoculator was used for applying ca.30 000^50 000 spores in ca. 0.2^0.4 ml of water to the centre ofthe top surface of the garden. In order to evenly disperse thespores within the water dilution, 0.001ml of Tween 20 (FisherScienti?c, Pittsburg, PA, USA) was added to each solution,including the water control treatment.The responses of the colonies to the three treatments werevideo recorded using a Panasonic Omnimovie1 video camerawith a Sigma1 VT-5 macro lens (Sigma Corporation ofAmerica, Ronkonkoma, NY, USA) mounted on a copy stand. A?bre-optic light source was used for illuminating the gardensurface. Each colony was recorded just prior to spraying, imme-diately following spraying and at 3, 6, 12, 18, 24, 30, 48 and 72 hfollowing the spraying treatment. Each video-recording sessionincluded an initial scan of the entire dorsal surface of the gardenand the lens was then focused on the location with the mostindividuals for 3min.If workers remove alien fungi as hypothesized, we predictedthat workers should move to the location sprayed with fungalspores, perform speci?c behaviours for removing the fungus anddump more or di?erent garden material.We tested whether antsmoved to the location of spraying by counting the number ofants on the top surface of the garden (where the spray wasapplied) at each recording interval (i.e. pre-spraying and 0, 3, 6,12, 18, 24, 30, 48 and 72 h post-spraying). The ants were countedduring the initial video scan of the top surface of the garden.The number of ants on top of the garden was then expressed as apercentage of the total number of ants in the garden (see below).We examined the speci?c behaviours performed by fungus-growing ants by studying the videotapes as well as makingdirect observations with a stereomicroscope. We identi?ed onebehaviour, i.e. `fungus grooming? (see } 3(b) for a description),that appeared to be common in the Trichoderma and Escovopsistreatments and uncommon in the control treatment. The fre-quency of fungus grooming was determined for each treatmentby relating the number of workers exhibiting fungus groomingto the number of workers present in the ?eld of view during the3-min videotaped session. The number of workers was deter-mined by counting all the workers present at the beginning ofthe taping period and all workers subsequently entering the ?eldof view. When it was obvious that an individual had exited andimmediately re-entered the ?eld of view, that individual wascounted only once.The refuse material that accumulated between taping periodswas collected at the end of each videotaping session from theforaging/dumping chamber, placed in 1.5-ml tubes and stored ina freezer for later examination. Additional collections weremade at 100, 125 and 150 h after spraying. The refuse materialwas dried for 48 h at 70 8C and then weighed on an electro-balance. Since dump removal depends upon garden size, we
1034 C. R. Currie and A. E. Stuart Weeding in agriculture by ants
Proc. R. Soc. Lond. B (2001)
calculated the total amount of refuse removed as a proportion ofthe total garden mass (see below). The composition of the dumpmaterial was assessed with a stereomicroscope prior to drying.Speci?cally, we noted whether the refuse had a normal appear-ance (i.e. dried decomposed vegetative material with no visible
signs of the fungal mutualist) or whether the fungal mutualistand fresh (green) leaf material were present in the refuse (and,thus, prematurely dumped). Although a few pellets from theworkers? infrabuccal pockets were noted in all dumps, the refusepiles with large numbers of pellets were noted.Each garden was sampled for the presence of Trichoderma andEscovopsis at the completion of the experiment, 150 h aftergarden treatment. Twenty pieces of garden material werecollected from throughout the garden and placed on nutrientagar in Petri dishes (following Currie et al. 1999a). After thegarden was sampled, all workers were removed, grouped intosize classes (castes) and counted. The garden material was driedand weighed.Our experiments produced three data sets: (i) the number ofants on top of the garden relative to the total number of ants inthe garden, (ii) the number of ants performing fungus groomingrelative to the number of ants in view during the 3-min trial,and (iii) the dump weight per hour relative to the garden mass.The datawere arcsine transformed in order toproduce normality.Pre-spraying and time 0 were used as controls in order to ensurethat there was no di?erence between the treatments before or atthe time of spraying. Statistical di?erences between all threetreatments (i.e. control, Trichoderma and Escovopsis) at allremaining time intervals were analysed using a repeated-measures ANOVA on Data Desk v. 4.1 (Velleman 1993). Eachpair of treatments at all remaining time intervals was alsoanalysed with a repeated-measures ANOVA (i.e. control versusEscovopsis, control versus Trichoderma and Trichoderma versusEscovopsis). In addition, each time interval was analysed sepa-rately using an ANOVA.
3. RESULTS(a) Workers on the top surface of the gardenAtta colombica workers responded to the presence ofalien fungi by moving into the infected locations fromother regions of the garden (repeated-measures ANOVA,d.f. ? 2 and p ? 0.045). On average, a larger proportion ofthe workers were on the top surface of the garden in colo-nies treated with Escovopsis than in colonies sprayed withwater (?gure 1a) (repeated-measures ANOVA, d.f. ?1and p ? 0.035). There was also a trend towards a higherproportion of workers being on the top surface of thegarden inTrichoderma-infected colonies as compared to thecontrol group (repeated-measures ANOVA, d.f. ?1 andp ? 0.08). The movement of workers to infected locationswas rapid in both the Escovopsis and Trichoderma treat-ments, occurring within 3^6 h of spraying (?gure 2a).The number of ants on the top surface of the garden inthe Escovopsis treatment remained elevated throughout the72 h of the experiment, either signi?cantly di?erent fromthe control or approaching signi?cance (p-values rangingfrom 0.033 to 0.074 for 3^72 h post-spraying). However,the number of ants on the top surface of the garden in theTrichoderma treatment was only signi?cantly di?erentfrom the control at 6 h post-spraying (p ? 0.0047).(b) Fungus groomingWe found that workers groomed their fungal mutualistin the presence of Escovopsis and Trichoderma spores.During fungus grooming, the antennae were used tosearch the nearby vicinity of the garden. When the antstopped searching, the maxillae and labium opened,
Weeding in agriculture by ants C. R. Currie and A. E. Stuart 1035
Proc. R. Soc. Lond. B (2001)
0.400.350.300.250.200.150.100.050.00
ants on
 top of g
arden
 (propo
rtional t
otal no.
 of ants
)
A
AB
(a) B
0.350.300.250.200.150.100.050.00
fungus 
groomi
ng frequ
ency 
(per wo
rkers ob
served)
A
B
(b) C
0.450.400.350.300.250.200.150.100.050.00 treatment
refuse d
umped 
from co
lonies
water Trichoderma Escovopsis
(c)
Figure 1. Frequency of di?erent behavioural responses to thepresence of water, T. viride and Escovopsis in intact colonies ofA. colombica. (a) The average proportion of workers located onthe top surface of the garden. (b) The average proportion ofworkers observed during the 3min interval that engaged infungus-grooming behaviour. (c) The average total dry mass ofrefuse material p laced in the dump during the 150 h followingtreatment (proportional to the garden mass) (n ? 5). Barssharing letters are not signi?cantly di?erent and error barsrepresent standard errors.
grasped some fungus, closed and retracted, thus raisingthe fungus o? the substrate and pulling it through theirmouth parts. This pattern was repeated between two andten times, during which the antennae were orientatedparallel to their respective mandibles and held stationary.The ant then moved slightly and resumed grooming.
Parallel, motionless antennae were the most obvious indi-cation that ants were fungus grooming. Even whensearching, the orientation of the antennae remainedunchanged. Generally, the body of the ant remained verystill for long periods during fungus grooming. Based onour observations, fungus grooming was performed exclu-sively by minima workers.Fungus grooming is a behavioural response to thepresence of alien fungi.We found a signi?cant di?erence inthe frequency of fungus grooming between treatments(?gure 1b) (repeated-measures ANOVA, d.f. ? 2 andp ? 0.0001), with a signi?cantly higher frequency of fungusgrooming within both the Escovopsis- and Trichoderma-treated gardens as compared to the controls (?gure 1b)(repeated-measures ANOVA, compared to Escovopsisd.f. ?1 and p ? 0.0005 and compared to T. viride d.f. ?1and p ? 0.0012). In addition, fungus grooming was almostthree times more frequent in colonies sprayed withEscovopsis than colonies sprayed withT. viride (p ? 0.0025).Fungus grooming appeared to be a relatively rare beha-viour in the absence of alien fungi, with an average of lessthan 2% of workers performing fungus grooming in thecolonies sprayed with water.The number of workers performing fungus-groomingbehaviours increased immediately following the sprayingtreatment in both the Trichoderma and Escovopsis treat-ments (?gure 2b). Interestingly, the response in theEscovopsis-treated colonies was more dramatic and moresustained than the response in the Trichoderma-treatedcolonies (?gure 2b). The Escovopsis-treated coloniesgroomed the fungus signi?cantly more than the controlcolonies for 3^30 h post-spraying (p-values ranging from0.001 to 0.05). Conversely, fungus grooming in theTrichoderma-treated colonies was only signi?cantly greaterthan the controls 3^12 h post-spraying. The response inthe Escovopsis-treated colonies was signi?cantly greaterthan that in theTrichoderma-treated colonies at 3, 24 and30 h post-spraying.(c) WeedingWorkers in colonies sprayed with Trichoderma andEscovopsis frequently removed fresh, young, green leafmaterial covered with the hyphae of their cultivar (younggarden material) from the top surface of the garden.Removal of this young garden material involved severalsteps and multiple workers. First, minima workersloosened the piece by chewing at the edges where it wasconnected to the rest of the garden by fungal mycelium.Once the garden piece was loose, the minima worker (orsometimes a larger worker depending on the size of thepiece to be removed) detached it from the garden matrixby holding onto the piece with its mandibles and rockinglaterally, side to side on its legs. Once the piece wasdetached from the garden, a medium or major workerpicked it up and carried it to the dump. We refer to thephysical removal of garden pieces, particularly abnormalpieces, as `weeding? (see } 4).The presence of both Trichoderma and Escovopsis withingardens resulted in an increased rate of refuse removal byleaf-cutting ants. In our experiment, at least one workerremoved a piece of garden during 52.5% (21 out of 40)and 75.0% (30 out of 40) of the observation periods forthe Trichoderma and Escovopsis treatments, respectively.
1036 C. R. Currie and A. E. Stuart Weeding in agriculture by ants
Proc. R. Soc. Lond. B (2001)
0.000
0.004
0.008 + +
+ ++
*
**
+
++ **
**
*
*
*
**
**
** ** ** ******
0.012 (c)
time (h)
Escovopsis
waterTrichodermaEscovopsis
refuse r
emoved
 (dry m
ass in g
 h-1 )
0 70605040302010
0.0
0.8
0.6
0.4
0.2
(b)
frequen
cy of fu
ngus gr
ooming
(per wo
rker)
0 70605040302010
0.40
0.30
0.20
0.10
(a)
number
 of wor
kers on
 top
(per wo
rkers to
tal)
0 70605040302010
Figure 2. Worker behavioural responses over the 72 hfollowing treatment with water, T. viride and Escovopsis inintact colonies of A. colombica. (a) The average proportion ofworkers located on the top surface of the garden. (b) Theaverage proportion of workers observed that engaged infungus-grooming behaviour. (c) The average dry mass ofrefuse material placed in the dump (proportional to gardenmass) (n ? 5). Symbols above the line represent signi?cantdi?erences between the corresponding treatment and control,while those below the line represent signi?cant di?erencesbetween Trichoderma and Escovopsis: p lus symbols, less than0.10 level of signi?cance; single asterisks, less than 0.05 levelof signi?cance; double asterisks, less than 0.01 level ofsigni?cance. Error bars represent standard errors.
Weeding was observed in only 5% (2 out of 40) of thecontrol colonies.The average amount of total refuse removed by colon-ies in the Escovopsis andTrichoderma treatments was almostdouble that of the control treatment (?gure 1c). How-ever, this di?erence was not statistically signi?cant.There was a signi?cant di?erence in the pattern ofrefuse removal over time (?gure 2c) (repeated-measuresANOVA, d.f. ? 2 and p ? 0.0325), with a peak rate ofrefuse removal in both the Escovopsis and Trichodermatreatments. However, refuse removal of the Trichoderma-treated colony peaked at 12 h post-spraying (materialdeposited in the dump between 6 and 12 h post-spraying),whereas the refuse removal in the Escovopsis-treatedcolony peaked at 24 h (material deposited between 18 and24 h post-spraying).The composition of the refuse material removed alsodi?ered between treatments (table 1). The control coloniesonly dumped normal refuse (old decomposed leaf mate-rial with no fungus present) whereas theTrichoderma- andEscovopsis-treated colonies dumped both normal andabnormal refuse (i.e. pieces of green leaves along withthe fungal mutualist). The dumping of normal refusematerial by workers is apparently largely abandoned incolonies infected with the pathogens (table 1); this mayexplain why an overall statistically signi?cant di?erencewas not obtained in our study. Colonies infected withdi?erent fungi initiated abnormal refuse removal atdi?erent times (Trichoderma at 3 h and Escovopsis at 6 hpost-spraying). The refuse removal was also sustained fordi?erent lengths of time. TheTrichoderma-treated coloniescompleted abnormal refuse removal within almost 24 hpost-spraying, whereas the Escovopsis-treated colonieswere still removing some abnormal particles at 150 hpost-spraying (table 1). The Trichoderma- and Escovopsis-infected colonies also dumped a substantially greaternumber of infrabuccal pellets (i.e. approximately ?ve toten times more) during their peak fungus-groomingactivity.(d) Removal of infectionIsolations from the garden at the end of the experimentrevealed that the ants were e?ective at eliminating
T. viride, as the presence of this fungus could only bedetected from one garden piece in one colony (table 2).However, Escovopsis was detected within all coloniestreated with the specialized pathogen and, in fact, wascommon within gardens (41% of pieces). The controlcolonies were con?rmed to be free of both T. viride andEscovopsis.
4. DISCUSSIONOur study indicated that the rare ability of fungus-growing ants to culture fungus for food is possible becauseworkers actively remove potentially devastating microbialpathogens from their gardens. When colonies are sprayedwith spores of the fungal pathogens, workers quicklymove to the infected areas from other locations (?gure 2a).These individuals engage in a behaviour that we term`fungus grooming?, i.e. removing alien spores by cleaningthe garden using their mouth parts (?gure 2b), collectingthe invading spores in their infrabuccal pockets and laterexpelling pellets of spores into the dump. In addition, theants engage in the removal of unhealthy pieces of garden(including fresh leaf material covered in their fungalcultivar) (table 1 and ?gure 2c). These hygienic beha-viours are speci?c to the presence of fungal pathogens asthey are the primary behaviours workers engage in in thepresence of pathogens, while they are uncommon in un-infected colonies. The relative absence of workersperforming these behaviours in uninfected colonies mayhelp explain why the importance and, in fact, occurrenceof these behaviours has mostly been overlooked.Our experiment also revealed that A. colombica workersengage in the same behaviours in attempting to removeboth general fungal contaminants (i.e. Trichoderma) andthe specialist pathogen Escovopsis. However, the intensityand duration of the colony response to Escovopsis wasmuch higher. The number of workers on the top surfaceof the garden and the rate of fungus grooming was alwaysgreater in treatments with Escovopsis than with Tricho-derma, although this was not always statistically signi?-cant. In addition, the Escovopsis-treated colonies had onaverage three times the proportion of ants engaging infungus grooming than the Trichoderma treatments (29.4versus 9.3%, respectively). In fact, this response occurredalmost immediately after spraying (3 h post-spraying)(70.1 versus 27.5%, respectively) (p ? 0.037), suggestingthat the ants can distinguish between the mycelium and/
Weeding in agriculture by ants C. R. Currie and A. E. Stuart 1037
Proc. R. Soc. Lond. B (2001)
Table 1. Presence of the fungal mutualist and green leafmaterial in the refuse material removed from the garden andplaced in the dump by the leaf-cutter A. colombica in coloniesinfected with Trichoderma or Escovopsis(Symbols: minus, no pieces; single plus, one or two pieces;double plus, approximately half the dump pieces; trip le plus,greater than 85% of the dump pieces composed of mutualistand green leaf material.)time post-spraying (h) water Trichoderma Escovopsis0^3 7 7 73^6 7 + + 76^12 7 + + + + +12^24 7 + + + + +24^30 7 + + + +30^72 7 7 + + +72^150 7 7 + +
Table 2. The presence or absence (proportion of coloniesinfected) between colonies and prevalence (proportion of piecesinfected) within gardens of T. viride and Escovopsis after theexperimental period(The values given are percentages.)Trichoderma Escovopsistreatment presence prevalence presence prevalencecontrol 0 0 0 0Trichoderma 20 1 0 0Escovopsis 0 0 100 41
or spores of the generalist and specialist fungal pathogens.In addition, all three behavioural responses (relocation tothe top surface of the garden, fungus grooming andweeding) were maintained for a longer time in theEscovopsis- than in the Trichoderma-treated colonies. Thedecline in these behaviours in theTrichoderma-treated colo-nies appeared to be because the ants were successful atremoving Trichoderma, whereas they were unsuccessful atcompletely removing the Escovopsis contamination(table 2) and, therefore, these behaviours were continued fora longer amount of time in the Escovopsis-treated colonies.As mentioned previously, we suggest that the behaviourof using their mouth parts for cleaning the fungal culti-vars (as described in ? 3(b)) be termed fungus grooming.This terminology is appropriate because the removal ofparasites and pathogens among ants, as well as amongother social animals, is generally referred to as`grooming?, although it is important to di?erentiate thisbehaviour from other forms of grooming. Super?cially,fungus grooming may appear to be the same behaviouras licking of the fungus; however, these are distinctbehaviours (C. R. Currie and A. E. Stuart, unpublisheddata). We observed licking of the fungus immediatelyfollowing the spraying of gardens in our study, withworkers actually lapping up the water droplets and sporesapplied to the garden, and licking of the garden occurred atapproximately the same frequency in all three treatments.Within 3 h following treatment, licking was not acommon behaviour in any treatment (see also Bass &Cherrett 1994). Discovering that fungus-growing antsgroom their garden is not surprising, since they are obli-gately dependent upon their cultivars and grooming as adefence against parasites and pathogens is widespreadamong social animals.Workers also responded to the presence of alien fungiby removing chunks of garden. This is a complex beha-viour, typically involving several minor workers looseninga piece from the matrix of the garden and then medium-and major-sized workers removing and carrying it to thedump. It is clear that the material removed by the antsincluded recently cut leaves and the mycelium of theirfungal cultivars (table 1); thus, this is not the normalprocess of removing vegetative substrate depleted by thefungus. Instead, this behaviour is involved in the removalof garden regions infected by alien fungi.We suggest thatthe speci?c behaviour of discarding garden substrate withthe mycelium of the mutualist fungus be referred to as`weeding?. Mature A. colombica colonies in the ?eldfrequently discard fresh garden pieces composed of freshvegetative material and fungus and this material isfrequently overgrown by the pathogen Escovopsis (C. R.Currie, personal observation), indicating that weeding isvery common in the ?eld. This also explains whyEscovopsis is commonly associated with the dump materialdiscarded by leaf-cutting ants (Bot et al. 2001).Fungus-growing ants, particularly leaf-cutters such asA. colombica, are very e?ective at protecting their gardensfrom generalist fungal pathogens such as Trichoderma(Currie et al. 1999a; Currie 2000; this study). Our ?ndingsindicated that fungus gardens are primarily protected bythe physical removal of these fungi by fungus groomingand weeding. The importance of these behaviours isobvious, as they are the primary tasks performed by
minima workers in colonies treated with the two fungalcontaminants.We suggest that three physical methods areused in defending fungus gardens from alien microbes.First, ants attempt to prevent their gardens from beingexposed to microbial contaminants. This is apparentlyachieved by licking both the nest surfaces upon which thefungus gardens are placed and the substrate added to thegarden (Stahel & Geijskes 1939; Autuori 1941; Quinlan &Cherrett 1977), thereby reducing the abundance anddiversity of microbes that may come into contact with thegarden. Second, colonies attempt to prevent inoculatedmicrobes from growing and establishing an infection;prevention is apparently achieved predominantly byfungus grooming. Finally, once infections are established,workers attempt to suppress and remove the pathogen.Removal of infected garden pieces (weeding) is clearly animportant step in treating infected regions of the garden.Obviously, other mechanisms, some identi?ed and someas yet undiscovered, are employed by the ants in dealingwith alien microbes in some or all of these three lines ofdefence. Experimental work is needed in order to estab-lish the importance of other mechanisms in defending thegardens of fungus-growing ants from alien microbes andthe specialized pathogen Escovopsis. It is interesting that,despite all of these lines of defence, the specialized para-site Escovopsis is still able to establish infections and spreadthrough the garden. An examination of how Escovopsiscircumvents the defences of the ants and their cultivatedfungi would provide interesting insights into the dynamicsof this unique symbiosis.
C.R.C. was supported by a Smithsonian predoctoral award andby Natural Sciences and Engineering Research Council ofCanada (NSERC) postgraduate and postdoctoral awards.A.E.S. was supported by an NSERC postgraduate award. Wethank the Smithsonian Tropical Research Institute andAutoridad Nactional del Ambiente of the Republic of Panamafor facilitating the research and granting collecting permits. Weacknowledge J. Burford, D. Currie, A. Herre, F. Hunter,M. Leone, D. Malloch, S. Margaritescu, S. Rehner, J. Scott,R. Thompson and B. Wcislo for valuable logistic support. Wethank R. Adams, K. Boomsma, G. Currie, U. Mueller andT. Murakami for valuable comments on this manuscript.Additional research support was provided by an NSERCoperating grant awarded to D. Malloch and NSF CAREERgrant DEB-9983879 awarded to U. Mueller.
REFERENCESAutuori, M. 1941 Contribuicao para o conhecimento da sau?va(Atta spp.). I. Evoluca? o do sauveiro (Atta sexdens rubrop ilosaForel, 1908). Arquivos Inst. Biol. Sao Paulo 12, 197^228.Bass, M. & Cherrett, J. M. 1994 The role of leaf-cutting antworkers (Hymenoptera: Formicidae) in fungus garden main-tenance. Ecol. Entomol. 19, 215^220.Bot, A. N. M., Currie, C. R., Hart, A. G. & Boomsma, J. J.2001 Waste management in leaf-cutting ants. Ethol. Ecol. Evol.(In the press.)Boyd, N. D. & Martin, M. M. 1975 Faecal proteinases of thefungus-growing ant Atta texana: properties, signi?cance, andpossible origin. Insect Biochem. 5, 619^635.Chapela, I. H., Rehner, S. A., Schultz, T. R. & Mueller, U. G.1994 Evolutionary history of the symbiosis between fungus-growing ants and their fungi. Science 266, 1691^1694.
1038 C. R. Currie and A. E. Stuart Weeding in agriculture by ants
Proc. R. Soc. Lond. B (2001)
Currie, C. R. 2000 The ecology and evolution of a quadripar-tite symbiosis: examining the interactions among attine ants,fungi, and actinomycetes. PhD thesis, University of Toronto,Canada.Currie, C. R. 2001a Prevalence and impact of a virulent parasiteon a tripartite mutualism. Oecologia. (In the press.)Currie, C. R. 2001b A community of ants, fungi, and bacteria: amultilateral approach to studying symbiosis. A. Rev. Microbiol.(In the press.)Currie, C. R., Mueller, U. G. & Malloch, D. 1999a Theagricultural pathology of ant fungus gardens. Proc. Natl Acad.Sci. USA 96, 7998^8002.Currie, C. R., Scott, J. A., Summerbell, R. C. & Malloch, D.1999b Fungus-growing ants use antibiotic-producing bacteriato control garden parasites. Nature 398, 701^704.Currie, C. R., Bot, A. N. M. & Boomsma, J. J. 2001Experimental evidence of a tripartite mutualism: bacteriaprotect ant fungus gardens from specialized parasites. Ecology.(Submitted.)Dix, N. J. & Webster, J. 1995 Fungal ecology. London: Chapman& Hall.Eisner, T. & Happ, G. M. 1962 The infrabuccal pocket of aformicine ant: a social ?ltration device. Psyche 69, 107^116.Febvay, G. & Kermarrec, A. 1981 Morphologie et fonctionne-ment du ?lter infrabuccal chez une attine Acromyrmexoctospinosus (Reich) (Hymenoptera: Formicidae): ro? le de lapoche infrabuccale. Int. J. Insect Morphol. Emb. 10, 441^449.Ho? lldobler, B. & Wilson, E. O. 1990 The ants. Cambridge, MA:Belknap.Martin, J. S. & Martin, M. M. 1970 The presence of proteaseactivity in the rectal ?uid of attine ants. J. Insect Physiol. 16,227^232.
Mo? ller, A. F. W. 1893 Die Pilzg?rten einiger s?damerikanischerAmeisen. Jena, Germany: Gustav Fischer.Mueller, U. G., Rehner, S. A. & Schultz, T. R. 1998 The evolu-tion of agriculture in ants. Science 281, 2034^2038.North, R. D., Jackson, C. W. & Howse, P. E. 1997 Evolutionaryaspects of ant^fungus interactions in leaf-cutting ants. TrendsEcol. Evol. 12, 386^389.Quinlan, R. J. & Cherrett, J. M. 1977 The role of substratepreparation in the symbiosis between the leaf-cutting antAcromyrmex octospinosus (Reich) and its food fungus. Ecol.Entomol. 2, 161^170.Quinlan, R. J. & Cherrett, J. M. 1979 The role of the fungus inthe diet of the leaf-cutting ant Atta cephalotes. Ecol. Entomol. 4,151^160.Schildknecht, H. & Koob, K. 1970 Plant bioregulators in themetathoracic glands of myrmicine ants. Angew. Chem. Int. Edn9, 173.Schildknecht, H. & Koob, K. 1971 Myrmicacin, the ?rst insectherbicide. Angew. Chem. Int. Edn 10, 124^125.Stahel, G. & Geijskes, D. C. 1939 Ueber den Bau der Nester vonAtta cephalotes (L.) und Atta sexdens (L.) (Hym: Formicidae).Revta Entomol. 10, 27^78.Velleman, P. F. 1993 Data desk, v. 4.1. Ithaca, NY.Weber, N. A. 1957 Weeding as a factor in fungus culture by ants.Anat. Rec. 128, 638.Weber, N. A. 1972 Gardening ants: the attines. Philadelphia, PA:American Philosophical Society.
As this paper exceeds the maximum length normally permitted,the authors have agreed to contribute to production costs.
Weeding in agriculture by ants C. R. Currie and A. E. Stuart 1039
Proc. R. Soc. Lond. B (2001)