Anim . Behar ., 1986, 34, 234-240

A novel autostimulatory pheromone regulating transport of leavesin Atta cephalotes

J . W. S . BRADSHAW, P . E . HOWSE* & R . BAKERChemical Entomology Unit, Departments of Biology and Chemistry, Building 3, The Unirersitr,

Southampton S09 5NH, U . K.

Abstract . Workers of the leaf-cutting ant Atta cephalotes mark freshly cut and stored leaves with anabdominal secretion, prior to transport . Such leaves are more readily picked up than unmarked leaves .This effect, and subsequent marking, can be experimentally induced by an extract of Dufour's gland,which contains a number of straight-chain alkanes and an alkene . The components active in releasingpick-up and marking when applied to filter-paper discs are n-tridecane and (Z)-9-nonadecene, and thelatter also affects transport behaviour . These compounds appear to have a role in regulating the chaintransport of leaves from the cutting site to the nest ; the secretion alters the transport behaviour of the antthat applies it, and this pheromone is therefore partly autostimulatory .

Leaf-cutting ants, in common with many othersocial animals, co-operate with one another in theretrieval of food . Freshly-cut leaves are oftendropped to the ground, where they are picked up byother ants for transport to the nest (Mariconi1970) . Piles of leaves may also be made at the sidesof trails, and around the nest entrances. In thestudy to be described, we have investigated howthese processes are regulated by chemical commu-nication .

In the course of cutting a leaf, ants frequentlypivot their abdomens ventrally so that the tip of thegaster brushes over the surface of the leaf fragment ;this may be described as `streaking' behaviour .When cutting is complete, the leaf fragment is heldvertically in the mandibles, and the gaster isbrought round beneath the thorax and the tip isapplied between one and five times to the edge ofthe fragment . This we term `marking' behaviour .The leaf is then manoeuvred into the carryingposition by means of the mandibles and legs, andtransport proper begins . Under the restricted con-ditions of laboratory culture, the initial destinationof the fragment is usually a pile of cut leaves outsidethe nest entrance . Ants moving leaves from onepart of the pile to another, or from the pile to thenest, usually perform marking behaviour in thecourse of picking up the leaf . The source of themarking secretion, and its role in leaf transport,have been investigated .

* To whom reprint requests should be addressed .

234

MATERIALS AND METHODS

Insects

Small colonies of Atta cephalotes were collected inTrinidad, West Indies, and cultured on pairs oftables 530 x 380 mm, in an insectary maintained at28±2°C and 70-90% relative humidity, with a 12 hlight/ 12 h dark cycle . Nest chambers, placed on onetable, consisted initially of plastic boxes, andsubsequently of inverted 1350-m1 museum jars onplaster of Paris bases . These were filled with fungusand brood as the size of the colony increased . Foodwas supplied on the second table, connected to thefirst by a wooden bridge . Forage material wasmainly privet (Ligustrum vulgare), dock (Rumexspp.) and grapefruit albedo . All chemical andbehavioural assays were carried out on coloniesthat had been in culture for less than 12 months .

Gas ChromatographyGas chromatography (GC) was carried out on a

Pye Series 104 Gas Chromatograph, equipped witha glass column, 1 . 5 m by 4 mm internal diameterpacked with polypropyleneglycol adipate (PPGA),l0° on diatomite C (acid-washed, DMCS treated,100/120 mesh) . The carrier gas was nitrogen at aflow rate of 40 ml min - ' . Standard oven conditionswere 80 C for 5 min, 4 C min - ' increase to 170 C .

The solid-sample technique of Morgan & Wad-hams (1972) was used for injection of volatilematerials from single gasters and single or multi-ples of glands . Micropreparative GC was carried

out by the method of Baker et al . (1976), andfractions were collected in redistilled hexane andstored in the dark at -4`C .

Extraction of Ant TissuesWorker ants were killed by plunging them into

dry ice . Extracts of the exocrine structures in thegaster were prepared by rapid dissection underdistilled water, followed by removal of the tissues infine-walled glass capillary tubes which werecrushed under 0 . 1 ml redistilled hexane . Fourextracts, each containing five glands or tissuesamples, were made from the poison gland,Dufour's gland, rectal sac, and fat body taken fromaround the crop . Ten additional Dufour's glandswere fractionated by solid-sample microprepara-tive GC and six fractions collected in 0 . 3 ml hexane ;the constituents of each fraction were measured byGC .

BioassaysIn order to test the effect of marked leaves on the

ants, food was supplied as flushing shoots of privetwith the old leaves removed . In the majority ofcases, ants cut the young leaves at the petiole,marked them and transported them intact . Suchleaves were collected from workers during trans-port by gently pulling them with forceps, and ineach case a leaf corresponding in size and colourwas clipped from a second shoot with scissors ; eachleaf-pair was tested only once. The leaves wereplaced in pairs, circa I mm apart, near to the nestentrance and the time until they were picked up(time to pick-up) was measured . Attraction wastested by placing pairs of marked and controlleaves, obtained in the same way, 5 cm apart oneither side of the main trail to the nest . Positions ofcontrol and marked leaves were reversed in 10successive experiments . Leaves were placed on thetable either 30 s or 300 s after marking, and the antspassing the leaves observed for 120 s . Extracts ofglands, fractions from Dufour's gland . and syn-thetic chemicals were tested on discs, 6 mm indiameter, cut from Whatman no . I filter-paper . Inthe tests of gland extracts the discs were pre-extracted in redistilled hexane : otherwise they wereused as cut . The discs were identified by pencilledsymbols on both sides . Gland extracts were appliedin 0002-m1 lots, equivalent to 0 . 1 glands, and gland

Bradshataw et al . : Transport pheromone in ants

2 3 5

fractions were appplied as 0 . 01-m1 lots, equivalentto 0 . 1 gland contents . a s measured by GC . Syn-thetic chemicals (>99% pure) were applied in0 . 001-m1 lots, containing the quantities indicated inthe Tables. In each experiment one disc was madeup for each treatment; the order in which thesolutions were applied to the discs was determinedby random number sequences . The discs werestored for 120 s to allow evaporation of solvent,and were then placed in a circle on the foragingtable . 5 cm apart . The times of the start and end,and the destination, of each transport . and mark-ing behaviour, were recorded for each disc for amaximum period of 2700 s (gland extracts andfractions) or 1500 s (synthetic chemicals) . Any discsin transit at the end of the period were observeduntil dropped .

RESULTS

Behavioural Effects of Marked LeavesPaired marked and unmarked leaves were pre-

sented to the ants and both attraction and pick-upbehaviour were observed . The median times topick-up of 15 pairs of marked and unmarked privetleaves were 390 and 1215 s respectively (Wilcoxontest, N= 14, T= 14, P < 0 . 01, one-tailed test ; Siegel1956). During the course of this experiment theants were observed to orientate towards marked,but not unmarked, leaves for a few minutes afterpresentation, from distances of up to 2 cm . In asecond experiment, of the ants passing the markedleaves, 21 out of 43 made directed turns, whereasonly seven out of 43 ants passing unmarked leavesmade similar turns (Wilcoxon test, N==8, T=0,P<0 .005, one-tailed) . Marked leaves stored for300 s after collection were not noticeably attractivein 10 trials; 12 out of 53 ants turned towardsmarked leaves, I I out of 56 towards unmarked .The secretion applied to leaves during and imme-diately after cutting is apparently attractive for ashort period, but releases pick-up behaviour whenattraction is no longer apparent .

Glandular Origin of the Leaf-marking SecretionThe behaviour of marking workers indicates that

the secretion is emitted from the tip of the abdo-men. Extracts of the three largest glands in thisregion, the poison gland, Dufour's gland and the

236 Animal Behariotrr, 34, 1

* In 10 replicate trials .t Level of significance in two-tailed Wilcoxon test .$ Level of significance in one-tailed sign test .Ns=not significant, P>0 . 05, compared with fat body extract .

rectal sac, together with an extract of fat body fromthe gaster, and a solvent blank, were tested onfilter-paper discs for their effect on pick-up, and onsubsequent marking and transport (Table I) . Allthree gland extracts decreased the time to pick-up,but in the cases of the poison gland and rectal sacthis was almost entirely due to powerful attractantand arrestant effects, which were not observed inresponse to marked leaves . The Dufour's glandextract decreased time to pick-up compared withthe fat body control and was the only treatment towhich ants responded by marking at first pick-up .The filter-paper discs used in this experiment werenot attractive to the ants after the effects of thesecretions had worn off, and hence the parameterfor transport behaviour used was the time frominitial pick-up to dropping from the edge of the nesttable (dumping). The Dufour's gland extractincreased this time by a factor of more than threecompared with the controls ; none of the other

treatments was significantly different from thecontrols .All three measurements indicate that the

Dufour's gland secretion has the propertiesexpected of the marking secretion, including de-creased time to pick-up as measured in response tomarked leaves (above) . The behaviour of antstowards the poison gland and rectal sac extracts isnot similar to that observed in response to markedleaves, although this does not preclude the possibi-lity that small quantitities of either may be appliedto leaves together with the Dufour's gland secre-tion .

Bioassay of Dufour's Gland Fractions and PureChemicals

Six fractions obtained by micropreparative GCof Dufour's glands were tested in 10 replicate trialsfor pick-up activity, subsequent marking, and

Table 11 . Pick-up, marking and transport measurements for fractions of the Dufour'sgland extract*

* In 10 replicate trials .t Level of significance in one-tailed Wilcoxon test .$ Level of significance in one-tailed sign test .Ns=not significant, P>0 . 05 . All compared with solvent control .

Table I. Pick-up, marking and transport measurements for gland and tissue extracts*

Median time Median time FrequencyExtract to pick-up (s) Pt to dumping$ (s) Pt Pj

Dufour's gland 72 <0-01 937 < 0 . 02 10 < 0 . 005Poison gland 67 < 0 . 05 322 NS I NSRectal sac 80 NS 210 NS I NsFat body 160 255 1Solvent 150 77

Median time Median longest FrequencyFraction to pick-up (s) Pt transport time (s) Pt of marking P$

1 270 <0-01 133 NS 7

< 0 .022 435 NS 90 NS 0 NS3 430 Ns 83 NS 2 NS4 470 NS 73 NS 0 Ns5 293 < 0-05 228 < 0. 02 6

< 0.056 343 NS 103 NS I

NsSolvent 558 95 1

Bradshaw et al. : Transport pheromone in ants

Table III . Pick-up, marking and transport measurements in response to synthetic hydrocarbons

Compound (ng/disc)

n-tridecane (5)n-heptadecane (50)n-nonadecane (12)(Z)-9-nonadecene (12)Solvent

* Level of significance in one-tailed Wilcoxon test .t Level of significance in one-tailed sign test .NS= not significant, P> 0 .05 . All compared with solvent control .

transport (Table 11) . Since in these trials few discs

Table IV. Time to pick-up in responsewere dumped within the maximum period, thecriterion for transport was taken as the longestcontinuous period of carrying by an individual ant .Both fractions I and 5 significantly decreased thetime to pick-up, and increased the frequency ofmarking and transport time compared with thecontrol . A similar effect on transport time wasnoted from the experiments with gland extracts,but the repellent nature of the discs in the series ofexperiments may have interfered with normaltransport behaviour .Gas chromatography of authentic compounds

showed that the main component of fraction I wasn-tridecane, and of fraction 5, n-nonadecane and(Z)-9-nonadecene (identified by Evershed & Mor-gan 1980, 1981). These, together with the maincomponent of the secretion, n-heptadecane, weretested in 10 replicate trials in the same way as thegland fractions, at the levels indicated (Table III) .The time to pick-up was significantly reduced by(Z)-9-nonadecene compared with the control, butno other compound was significantly different fromthe control . Mean longest transport time wasincreased significantly by (Z)-9-nonadecene butnot by any other compound . The reduced effectcompared with fraction 5 in the previous experi-ment may have been due to different activities ofthe ants in the two colonies used .

The apparent lack of activity of n-tridecane inthis last group of tests was further investigated bypresenting this compound at different concentra-tions in 15 min replicate trials . The quantity testedabove (5 ng) is approximately double that found ina single Dufour's gland of a media worker, and itwas anticipated that this might inhibit the pick-upresponse . The time to pick-up of n-tridecane-treated discs was least for an initial application of 2

to different doses of n-tridecane*

* In 15 replicate trials .t Level of significance in one-tailedWilcoxon test, compared with sol-vent control .

NS= not significant, P>0-05 .

and

237

at 0 .5 ng was similar to the controlng/disc,(Table IV) .

The two active synthetic compounds were testedsingly and in combination with each other and it-heptadecane, the major component of the secre-tion, in 20 replicate tests, to determine whetherthere was any additional effect of combination(Table V). All treatments reduced the time to pick-up, although the figures for n-tridecane were notsignificantly different from the control at the 5°,level . This may have been due to differences in theactivity of the ants between this and the previoustrials . or to a preference for the treatments contain-ing (Z)-9-nonadecene, which could interfere insome way with the activity of the alkane . Theactivity of (Z)-9-nonadecene did not appear to bealtered by the addition of n-tridecane, and wasslightly reduced by the further addition of n-heptadecane; in the latter case the rate of volatiliza-tion of the olefin would have been reduced by the

Dose (ng)Median timeto pick-up (s) Pt

5 730 NS

530 <00051 555 < 0 . 0505 1440 NSsolvent 955

Median time Median longest Frequencyto pick-up (s) P* transport time (s) P* of marking Pt

323 NS 240 Ns 8 < 0 . 0-1483 NS 168 NS 4 Ns730 NS 280 NS 6 Ns210 <0025 378 <005 8 <002493 230

238

* In 20 replicate trials .t Level of significance in one-tailed Wilcoxon test .$ Level of significance in one-tailed sign test .NS =not significant, P>0 . 05 . All compared with solvent control .N= number of trials in which pick-up occurred .

comparatively large quantities of alkane . Both n-tridecane and (Z)-9-nonadecene released markingbehaviour as expected (Table V), but the combina-tions of (Z)-9-nonadecene with either alkane didnot significantly increase the occurrence of mark-ing, compared with the control . These results maysuggest that the mixtures of synthetic compoundssimilar to the natural secretion are marked lessfrequently that the 'unnatural' single compound .

The increased number of replicates in this seriesof experiments enabled a more detailed analysis tobe made of the time of transport of discs. In orderto minimize the effects of distance and variations inother chemical signals, such as trail pheromones,only journeys from the presentation site to the nestarea were analysed (Table VI), and were furtherdivided according to whether or not the disc wasmarked immediately before the journey . Log-transformed data were found to give a reasonable

Animal Behaviour, 34, 1

Table VI . Transport times from the foraging area to the nest area

graphical fit to the normal distribution (South-wood 1966), and were used in calculating meansand in analysis of variance (one-way on all means) .

Marking on the control discs increased carryingtime, confirming the effect of marking on transportbehaviour. None of the unmarked discs treatedwith synthetic chemicals was as effective as themarked control, but those treated with (Z)-9-nonadecene and (Z)-9-nonadecene with bothalkanes were significantly more effective thanunmarked controls . Marking on either of thesegave no significant increase in transport times .Thus it appears that (Z)-9-nonadecene increasescarrying time over a standard course, and, at theconcentration used, is not as effective as the wholemarking secretion, and overrides the effect ofmarking. Trace impurities undetected by GC mayhave contributed to the reduction in the carryingtime of marked (Z)-9-nonadecene treated discs

N =number of observations .* Means followed by the same letter in either column are not significantly different at P < 0 .05 byHartley's multiple range test (Snedecor & Cochran 1967) .

Marked UnmarkedTreatment (ng)

(Z)-9-nonadecene n-heptadecane n-tridecane NMeantime(s) N

Meantime(s)

0 0 0 7 220.2a* 8 91-8e10 0 0 15 l09 •l de 5 1382bcd10 40 2 10 142 .6bc 7 123 .2bcd10 0 2 5 155 . 9b 12 118 . 4cde0 0 2 14 137 .7bcd 4 118 .6cde

Table V. Pick-up and marking behaviour in response to synthetic hydrocarbons*

Code Compounds (dose, ng) NMedian timeto pick-up (s) Pt

Frequency ofmarking P$

A n-tridecane (2) 20 260 NS 17 <002B (Z)-9-nonadecene (10) 20 163 <0-025 16 < 0 . 05C A+B 20 183 <0-025 7 NSD C + n-heptadecane (40) 19 230 <0 . 05 12 NSE Solvent (control) 18 383 7

compared with marked control discs or additionaltrace components of the marking secretion may beimportant. Unmarked n-tridecane treated discs aretreated as controls, but the small number ofobservations precludes any definite conclusions asto the role of this compound in transport beha-viour. The figures for discs treated with (Z)-9-nonadecene and n-tridecane are also inconclusive .There was no noticeable decrease in the speed ofmovement of the ants carrying marked control or(7)-9-nonadecene treated discs towards the nest,and it was apparent that both these treatmentsincreased the time spent in movements around thenest area prior to dropping .

DISCUSSION

Two behavioural components in Atta cephalotesworkers, streaking during leaf-cutting and markingprior to leaf transport, suggest that a secretion isapplied to leaves prior to transport to the nest . Leaffragments that have undergone both processes aremore readily picked up by the ants and areattractive for a short period (up to 300 s) aftercutting . Of the exocrine structures that discharge atthe tip of the gaster, Dufour's gland most nearlymimics these effects . Since both this gland and thepoison gland are discharged through the sting, it islikely that, even when the glands themselves aredischarged separately, a small amount of oneinitially contaminates the other. The attractiveproperties of the streaked and marked leaves werenot noticeably present in discs treated withDufour's gland, and may have been due to con-tamination with poison gland secretion, which hasbeen observed to contain powerful attractants(Robinson & Cherrett 1975) .The active components of the Dufour's gland

secretion with respect to pick-up, marking andduration of transport, occur in two fractions, onecontaining volatile, and the other comparativelynon-volatile, components. A synthetic sample of n-tridecane, the major component of the volatilefraction, was not as active in releasing pick-up asthe fraction itself, even at the optimum level of 2 ng/disc (which is approximately equivalent to thewhole gland content of this compound). Thisdisparity could he explained by differences in theactivities of the colonies used, which might have aconsiderable effect on the mean pick-up time of a

Bradshaw et al . : Transport pheromone in ants

239

volatile compound, where the vapour concentra-tion changes rather rapidly, but the possibilityremains that there is an unidentified trace compo-nent of the fraction which synergizes the action ofthe n-tridecane . Marking of n-tridecane treateddiscs is comparable to that of the active volatilefraction, and the apparent doubling of transporttime by fraction I may be due to this rather than anintrinsic property of the components of the frac-tion .

The non-volatile active fraction contained twomain components, identified as n-nonadecane and(Z)-9-nonadecene . No activity in pick-up, markingor transport was detected for n-nonadecane, andthe properties of the fraction could be entirelyascribed to (Z)-9-nonadecene, which releases pick-up and marking, and increases transport time overa fixed route. This last property was not detected indiscs treated with Dufour's gland extract, probablybecause the discs used in that experiment wereslightly repellent but, in the final experiment,marking of untreated discs was shown to increasetransport time over a fixed route .

The precise significance of each of these beha-vioural effects is not easy to determine underlaboratory conditions, but consideration of theforaging strategies of Atta cephalotes may indicatetheir role in the transport of forage material to thenest. Wherever freshly cut leaves are dropped to theground by cutting ants, the two chemical releaserswill increase the probability of the leaf fragmentsbeing picked up and transported towards the nest :the same will probably also be true of leaf frag-ments piled by the sides of trails or around the nestentrances. The release of marking behaviourensures that the secretion applied to the leaf isrenewed virtually every time it is transported, andtherefore remains active for release of subsequenttransport behaviour. However, this would be mostefficiently achieved if the marking took place whenthe fragment was put down, thereby maximizingthe duration of the signal for a given amount ofsecretion : in all the experiments this was neverobserved . This suggests that some behaviouralcomponent(s) released by the secretion are impor-tant to the same ant that deposits the mark_ and itappears that one of these has been measured asduration of transport . In the final experiment, bothmarked untreated and unmarked (Z)-9-nonade-cene treated discs were carried for longer periodsover a fixed route than unmarked untreated discs .The added secretion, or synthetic chemical. was the

240

Animal Behat'iour, 34 . 1

only cue that could have released this, and hence

ACKNOWLEDGMENTSmarking, and specifically (Z)-9-nonadecene in themarking secretion, acts as an 'autostimulatory'chemical signal . The precise role of this effect intransport behaviour has not been determinedalthough, as has been mentioned, the increasedcarrying time was largely taken up in movementaround the nest area . If there is a more-or-lessspecific signal for dropping a marked leaf, theincreased carrying time could represent a period ofsearching for such a signal, while unmarked leavesare dropped at random . Under the laboratoryconditions, marking does not have any obviouseffect on trail-following . Overall, the markingsecretion directs the chain transport of leaf frag-ments to the nest by a succession of worker ants,without the need for direct transfer from one ant toanother . The secretion may have other effects intwo locations not yet examined, at the cutting siteand within the nest. Additionally, trace amounts ofpoison gland secretion that are likely to contami-nate the Dufour's gland secretion, as suggestedabove, may have a role in modifying the responses .

Many animals have the opportunity to perceivetheir own pheromones, either as the secretions areemitted, or in some cases subsequently, but littleattention has been paid to this aspect of chemicalcommunication. In some species of ants, scoutsregularly orientate to their own chemical trails(Holldobler 1978), but the reactions of animals to,for example, their own alarm pheromones, havenot been investigated. The autostimulatory role ofthe leaf marking pheromone appears to supple-ment the ants' short-term memory, but its precisesignificance requires further investigation .

This research was carried out while the first authorwas in receipt of an ICI Research Fellowship . Wewish to thank Dr J. M. Chcrrett of U.C.N.W .,Bangor, and Dr F. D . Bennett of C .I .B .C . . Trini-dad, for supplies of ant colonies .

REFERENCESBaker, R ., Bradshaw, J . W . S ., Evans, D. A ., Higgs, M . D .& Wadhams, L . J . 1976. An efficient all-glass splitterand trapping system for gas chromatography . J. Chro-rnatogr. Sci ., 14, 425-427 .

Evershed, R . P . & Morgan, E .D . 1980 . A chemical studyof the Dufour glands of two attine ants . InsectBiochem ., 10, 81-86 .

Evershed, R . P. & Morgan, E . D . 1981. Chemicalinvestigations of the Dufour gland contents of atoneants . Insect Biochent ., 11, 343-351 .

Holldobler, B . 1978 . Ethological aspects of chemicalcommunication in ants . Adv. Study Behar ., 8, 75 115 .

Mariconi, F . A . M . 1970 . As Saunas. Sao Paulo : EditoraAgronomica `Ceres' .

Morgan, E. D . & Wadhams, L. J . 1972 . Gas chromato-graphy of volatile compounds in small samples ofbiological materials . J. Chromatogr. Sci. . 10, 528-529 .

Robinson, S. W . & Cherrett, J . M. 1975 . Some reactionsof leaf-cutting ants (Attmi : Formicidae) to a syntheticscent-trail pheromone . In : Pheromones and DefensiveSecretions in Social Insects (Ed. by C . Noirot, P . E .Howse & G. Le Masne), pp. 91-97 . Dijon : Internatio-nal Union for the Study of Social Insects .

Siegel, S. 1956 . Nonparametric Statistics for the Beha-rioral Sciences . New York: McGraw-Hill .

Snedecor, G . W. & Cochran, W . G . 1967 . StatisticalMethods . Iowa: Ames .

Southwood, T . R . E . 1966 . Ecological Methods . London :Chapman & Hall .

(Receired 4October 1982; revised 30 November 1984, MS .number : 2302)