JournalofResearchon theLepidoptera 31(3-4):213-232, 1992 The influence oflarval age and ant number on myrmecophilous interactions ofthe African Grass Blue butterfly, Zizeeria knysna (Lepidoptera: Lycaenidae) Konrad Fiedler and Dorthe Hagemann Zoologie II, BiozentrumderUniversitat, AmHubland, D-97074Wurzburg, Germany Abstract.InteractionsbetweenmyrmecophilousZizeeriaknysnalar- vae andLasiusflavus ants were quantitativelystudied inlaboratory experiments. Larvae delivered secretions from their dorsal nectar organ (DNO) more frequently in the initial 3-minute interval ofan interaction than later on. Tentacle eversions were likewise more commonatthebeginningofinteractions.Non-feedingprepupallarvae secretedsignificantlymoredropletsthanfeedingfourthinstars.Actual tendinglevels differedbetweenlarvae testedwith 5 (2.6-3.1 ants per larva) or 15 ants (5.3 ants/larva), respectively. Secretion rates in- creased with tending level (5-ant trials: feeding L4 larvae 5.5 DNO droplets/h, prepupae 16.4 droplets/h; 15-ant trials: feeding L4 9.5 droplets/h,prepupae25.5droplets/h). Secretiondropletsaveraged0.2 mm in diameter (volume 0.004 pi). From these data, a model is developed to estimate lifetime DNO secretion amounts ofindividual larvae. Estimatesrange from 1.3-4.7 pi perlarvain 5 d, representing approximately 0.2-0.7 mg carbohydrates with a physiological energy equivalentof3.4-12J.Hence,Z.knysnalarvaeprovideonlyamarginal foodrewardforattendantants,suggestingthatmyrmecophilyisalow- costlife-historystrategyin thatbutterflyspecies. KeyWords: mutualism - symbiosis - caterpillars - strategicbehavior- energetic investment - Formicidae - Lycaenidae Introduction Interactions between immatures of butterflies and ants, termed myrmecophily, are widely known from a broad range oflycaenid and riodinidspecies(seereviewsbyMalicky1969;Cottrell1984;Pierce1989; Fiedler 1991). Much work has concentrated on the description ofindi- m vidual life-cycles, on the structure and function of rrmecophilous 3 organs, and on the ecological outcome of myrmecophily (mutualism, parasitism). In contrast, fewerstudieshave focused on detailed quanti- tativestudiesoftherelevantbehaviors.Suchstudieseitherattemptedto quantify myrmecophilous interactions in the light ofinterspecific com- parisons (Fiedler 1991, Ballmer & Pratt 1991), or they experimentally elucidated phenomena like the release offood recruitment in tending ants (Fiedler & Maschwitz 1989), the secretory capacity ofindividual larvae (Fiedler & Maschwitz 1988a), the influence oflarval food on the Papersubmitted 25July 1994; revised manuscriptaccepted 15 November 1994. 214 J. Res. Lepid. expression ofmyrmecophily(Fiedler 1990, Baylis & Pierce 1991), or on conditionalfactorsregulatingsecretorybehavior(Leimar&Axen 1993). Withtheexceptionofthelastmentionedwork, allstudies assumedthat secretionrates observedinexperiments aremore orlessrepresentative for the populations or species under investigation. The energetic investment oflycaenid larvae in their symbiosis with ants may, however, be plastic in response to the actual needs. For example, Leimar &Axen (1993) showedthatlarvae ofthefacultatively myrmecophilous speciesPolyommatusicarus (Rottemburg, 1775)deliv- ered more secretion droplets from theirnectar organwhen subjectedto a simulated attack or when tended bytwo instead ofoneLasiusflavus (Fabricius, 1781)workerants.Furtherincreaseinthenumberoftending ants did not add to secretion rates. In addition, the intensity of myrmecophilyoftenincreaseswithprogressivelarvaldevelopment(e.g. Malicky1969,Fiedler1989),althoughLeimar&Axen(1993)observedno significant correlation between body mass and secretion rates among fourth instars of P. icarus. Wagner (1993) demonstrated significant weight losses in non-feeding prepupal larvae ofa Nearctic facultative m 3rrmecophile, Hemiargus isola (Reakirt, [1867]). Her result points to particularlyintensive and energeticallycostlyinteractionswithants in the prepupal phase. We here present a quantitative laboratory study on myrmecophilous interactions between larvae ofZizeeria knsyna (Trimen, 1862) and the scatLasiusflavus.Specifically,weaddressthefollowingquestions: 1)Are thesecretionsfromthedorsalnectarorgan(DNO)deliveredataconstant rate oris there atemporal patterninthe secretorybehavior? 2)Arethe secretion rates offeeding mature larvae equal to those ofnon-feeding prepupal larvae? 3) Does the number of tending ants influence the outcome oflarva-antinteractions? 4) Is larvalmyrmecophilycorrelated with body mass and how are the various myrmecophilous behaviors correlated with each other? 5) Finally, we try to estimate the lifetime investment in nectar-like secretions ofindividual larvae ofZ. knysna. Material and methods Study organisms TheAfHcanGrass Blue,Zizeeriaknysnaisasmallbutterflydistributedfrom the Canaryislands and the Iberian Peninsula southwards throughout most of Afiica, includingMadagascarandthe Mascareneislands, eastwards extending to Arabia. The species is polyvoltine. It occurs in open, xeric habitats, and the larvae feedon avarietyofplants, notablyvarious generaofFabaceae, but also on members ofAmaranthaceae, Chenopodiaceae, Oxalidaceae, Zygophyllaceae and Euphorbiaceae. In addition, oviposition has been observed on Malvaceae (Schurian1994).Thereare4larvalinstars,witholderlarvaefacultativelytended by ants (see Clark & Dickson 1971 for a detailed illustrated description ofthe basic life cycle). Tending ants have rarely been specified. Schurian (1994) recorded a Pheidole species (Myrmicinae) from the Canary islands. Further records, such as Tapinoma melanocephalum (Dolichoderinae: Warnecke 1932/ 31(3-4):213-232, 1992 215 33),refertotherelatedbutterfly,Z. karsandra(Moore, 1865),whosestatusasa distinctspecies hasbeen subjectto controversyuntilrecently. Forourexperiments, weusedlaboratory-bredoffspringoffemales caughton Gran Canaria. Butterflies were kept in plastic cages in a greenhouse (see Schurian 1989, for details on the breeding method). The laboratory stock was maintainedfor5generationsthroughouttheyear1993usinginflorescencesand youngfoliageofMedicagosativa(Fabaceae)asthemainlarvalfood.Experiments wereconductedwithmembersofthe4thand5thgenerationbetweenSeptember andNovember 1993. To control forpossible effects oflarval diet (Fiedler 1990, Baylis & Pierce 1991), we fed all experimental animals invariablywith young foliageofM. sativa.Larvaewererearedinanenvironmentalchamberat25.5°C and 15:9hL:Dcycle. Theywerekeptintransparentplasticvials (250 ml)lined withmoistfilterpaper.Adlibitum amountsoffreshlycutterminalfoliageofM. sativawereprovideddaily,andthelarvaeweretransferredtoanewrearingvial everydaytominimizethe riskofdiseases. Lasius flavus (Formicinae) is a common subterranean ant species of the Palearcticregion.Itmostlyoccursinopengrasslandorheaths,butalsocolonizes forests(Kutter1977).L.flavusavoidstrulyxerichabitatsandthereforeprobably rarely,ifever,co-occurswithZ.knysna,althoughthedistributionsofbothspecies overlapontheIberianPeninsulaandinnorthwesternAfrica.ThedietofL.flavus ants mainly consists ofthe honeydew ofroot aphids. Furthermore, aphids are eaten in large quantities to obtain proteins (Pontin 1978). Due to their food specialization, L. flavus ants show intensive trophobiotic behavioreven under laboratory conditions and avidly tend lycaenid larvae and pupae (e.g. Fiedler 1990, 1991,Leimar&Axen 1993).Therefore,thisantspeciesisverysuitablefor laboratorystudiesonlycaenidmyrmecophily, althoughassociationsbetweenL. flavusworkerantsandlycaenidimmatureshaverarelybeenobservedinnature (Fiedler 1991). Antcolonieswerekeptatlaboratorytemperatures ofapprox. 20-23 under ambientlightconditionsinlargeearthnests,whichweremaintainedinplastic arenas (size64cmx44cmx 12cm)withabottomofplasterofParis. Sidewalls weresmearedwithFluontopreventantsfromescaping.Thenestsweresprayed dailywithwatertoadjusthumidity,andfood(honey-waterandcutcockroaches) was provided as needed. For our experiments we used three ant colonies originatingfrom northern Bavaria. Experimental procedure Experiments were conductedin plasticarenas (10 cmx 10 cmx 6cm)with a bottomofplaster.Thebottomwaskeptmoistduringalltrials.Forexperiments, either 5 or 15 foragingworkers ofL. flavus were takenwhile ontheirwayto a feeding place in the foraging area ofthe nest arenas and were carefully trans- ferred into a test arena with the help ofa brush. Disturbance ofants due to handlingwasminimizedandanother5minallowedbeforeasingletestlarvawas placedinthecenterofthetestarena.Afterthattimeperiodthealarmbehavior oftheantshadsubsided.Beginningwiththefirstencounterbetweenanantand thelarva,werecordedthebehavioralinteractionsfor15min.Observationswere made under a Zeiss stereomicoscope at ten-fold magnification with normal daylightbetween9:00hand 15:00hlocaltime.Thearenawasrotatedfromtime to timeto eliminate possible effects ofdirectional illuminationon ant activity. Thefollowingeventswerecountedevery30seconds: a)thenumberofantsin 216 J. Res. Lepid. immediate physical contact with the larva; b) the number of DNO secretion dropletsdeliveredduringthattimeinterval;andc)thenumberofeversionsofthe tentacleorgans(TOs).Totaldurationofcontactsbetweenantsandthetestlarva were recorded with a stop-watch to the nearest second. At the end of each experiment,thelarvawasweighedtothenearest0.1mg(SartoriusBasicBA61 balance). Each set of worker ants was used for a maximum of three subsequent experiments to avoid possible effects due to a drop in ant activity ifkept in isolation from their nestmates for longer periods. In a large series ofearlier experiments(Burghardt&Fiedler,unpubl.)wehaveestablishedthatnoadverse effects occurifL.flavusarekeptawayfromtheircolonyforupto 1h.Atleast5 min elapsedbetweenthe experiments. Two classes oflarvae were used in experiments. “Feeding larvae” refers to animalsthatwerewellwithinthefourth(=final)instarandhadnotyetleftthe foodplanttosettledownforpupation.FeedingL4larvaeinourtestsrangedfrom 25.7-53.0mg(wetweight)andwereallneartheirlarvalpeakbodymass.“Non- feedingprepupallarvae”denotesthosewhichhadstoppedfeeding. Suchlarvae hadmostlyleftthehostplanttosettledownamongthefilterpapers,buthadnot yetspun asilkgirdle. Theyall showed acharacteristictransformationofcolor: theirmarkings became indistinct and the overall appearance was transparent and “glossy”. Non-feeding prepupal larvae are still able to crawl and their myrmecophilousorgansremainfunctional.Wetweightsofnon-feedingprepupae tested ranged from 26.1-51.2 mg. After one day, the true immobile girdled prepupais formed, whichis no longerableto evertthe tentacle organs. Larvalsexdiscriminationwasnotattempted,sincesexualweightdimorphism inourlaboratorycultureswasgenerallylow.Anylarvawastestedonlyonceper dayandatmosttwiceperlifetime(onceasafeedinglarva,againasanon-feeding prepupa).Afewlarvaeineachserieswereexaminedinonlyoneofthesephases. Quantitative evaluation ofresults Attractiveness, orac^altendinglevel, wascalculatedfromdatarecordedfor eachindividuallarva(definedasthe arithmeticmeanofthe numberoftending ants throughout the experiment, i.e. across 30 census points). In addition, we calculatedthe total numberofsecretion droplets delivered perexperiment and the sumoftentacle eversions. To examine the time course oflarva-antinterac- tionswesubdividedeachexperimentalperiodintofive3-minintervals.Sincethis analysis revealed adistinctdifferencebetweenthefirst 3-mininterval andthe subsequentintervals(seebelow),wealsocalculatedthenumberofsecretionacts andtentacleeversions ofeachexperimentallarvaforthefinal 12minofatrial. Alldatawerethensubjectedtostatisticalanalysis. Comparisonsbetweenthe larvalageclassesorbetweentheexperimentalserieswithdifferentantnumbers were computed using the non-parametric U-test of Mann & Whitney, while comparisons between the time intervals within experiments were made using Wilcoxon’s matched-pairs signed-ranks test. Spearman rank correlations be- m tween mnecophily parameters and larval weight were likewise calculated 3 (Sachs 1992). Results Temporal patterns oflarva-ant interactions Regardlessoflarvalageorantnumber,allexperimentswithZ. knysna 31(3-4):213-232, 1992 217 larvaerevealedsimilartemporalpatternsofmyrmecophilousbehaviors andinteractions.DNOsecretionsoccurredsignificantlymoreofteninthe first 3-minintervalthaninthe four subsequentintervals ofthe experi- ments. This was true for feeding larvae (Fig. lA) and non-feeding prepupae (Fig. IB) in experiments with either 5 or 15 L, fLavus worker ants(Wilcoxon-test,p<0.02forallcomparisonsbetweenfirstandsecond 3-minute experimental interval). On average, 1-2 droplets were deliv- ered by feeding larvae, and 2-3 by non-feeding prepupae, in the initial threeminutes.Thiscomparesto 1-2droplets(feedingL4)or3-5droplets (prepupae) in the subsequent 12 min. Virtually the same pattern occurred with the TO eversions. Feeding larvae evertedtheirTOs significantlymoreofteninthefirst3 minthan later (Wilcoxon-test, p < 0.01 for experiments withboth 5 and 15 ants), andinthefinal9experimentalminutesTOeversionswereveryrare(Fig. 2A).Thesamewasobservedwithnon-feedingprepupae(Wilcoxon-Test, p < 0.01), butthe effectwas delayed inthe experiments with 15 ants to the third 3-mininterval (Fig. 2B). Overall, TO eversions occurred more frequentlyin the prepupae duringthe final 9 experimental minutes. The attractiveness oflarvae to ants remained stable throughout the courseoftheexperiments.Alllarvaewerealmostconstantlytendedfrom theirfirstencounterwithants.Totaltendingtimeswere12:45-15:00min in experiments with five L. fLavus ants (only five feeding L4 and four prepupaehadassociationtimesshorterthan15:00min),and13:27-15:00 min in trials with 15 ants (three prepupae had association times lower than 15:00min).Within 1-2minafterthefirstencounter,thenumberof tendingantsin all experimentsreachedthe average level. Rarelythere was afurtherslightincrease,butneveradistinctdrop, inthenumberof tending ants withtime. Comparison between feedingmature larvae and non-feeding prepupae TherewasadistinctdifferenceinDNOsecretionratesbetweenfeeding larvae and prepupae (Fig. 3). During both experimental series with either 5 or 15 ants, prepupae produced much more secretion droplets thanfeedingfourthinstars(U-test,p< 0.002, withorwithoutthefirst3 min ofeach experiment beingincluded). DuetothenumericalpreponderanceofTOeversionsintheinitial3min ofeach experiment, the total frequencyofTO eversions throughout the 15-min trials showed no significant differences between the two larval ageclasses(p>0.20foreversionratesin15min,withboth5and15ants). When the initial 3 minwere deducted, a significant difference emerged in the 15-ants series: prepupal larvae everted their TOs significantly more often thanfeedingfourthinstars (U-test, = 114, p < 0.05). In the5-antsseries,asimilar,albeitnon-significantdifferencebetweenthe two age classes occurred. In experimentswith 5 ants, the actualtendinglevel(meannumberof tending ants per larva) increased slightly, but significantly, from the 218 J. Res. Lepid. A DNO secretion droplets 20- . 12-15 time (min) B DNO secretion droplets time (min) Fig. 1:TemporalpatternofDNOsecretionactsobservedinexperimentswith larvae ofZizeeriaknysna. Given aremeans+standarderrorsforfivesuccessive 3-min time intervals. Hatched bars: with 5 Lasius flavusants; white bars: with15ants.A):feedingmaturefourthinstars(n=19with5ants;n=20with 15ants); B): non-feedingprepupae(n=18with5ants; n=20with 15ants). Initial secretion rates are significantly higher than in subsequent 3-min intervals Intervals (Wilcoxon-test, p < 0.05). 31(3-4):213-232, 1992 219 TO eversions 8- 0-3 3-6 6-9 9-12 12-15 time (min) B TO eversions Fig.2:TimecourseofTOeversionrates(means+S.E.)inZknysnalarvae.Hatched bars:with5Lflavusanis;whitebars:with15ants.A):feedingmaturefourth instars (n= 19with 5ants; n =20with 15ants); B): non-feedingprepupae (n=18with5ants;n=20with15ants).Initialeversionratesaresignificantly higherthan in subsequent3-min intervals (Wilcoxon-test). 220 J.Res. Lepid. DNO secretion droplets i0n 8- 6 - I feeding L4 larvae non-feeding prepupae Fig. 3: Total numberof DNOsecretion droplets (means+ S.E.) delivered in 15-min experimentalintervalsbylarvaeofZknysna.Hatchedbars:with5Lflavus ants(n=19);whitebars:with 15ants(n=20).Differencesbetweenfeeding larvaeandnon-feedingpupae,aswellasbetween5-antsand15-antstrials are all statistically significant (U-test, p < 0.05). feeding (2.63 ants/larva) to the non-feedingprepupal phase (3.13 ants/ larva; U = 100, p < 0.05). In the parallel series with 15 ants, larvae received an actual tending level of5.35 ants/larva already during the feeding phase and this did not change with the transition into the prepupalstage(5.27ants/larva).Hence,prepupallarvaeattractalarger group ofworker ants than still feeding mature larvae; however, under our experimental conditions an upper physiological limit (“saturation”) is reached at an average ofroughly 5 ants per larva. The influence ofant number DNOsecretionsoccurredmorefrequentlyamongexperimentswith 15 ants, but this difference was only marginally significant for feeding larvae (U^g.^g = 133; p (1-tailed) < 0.10) ornon-feedingprepupae (U^g.^g = 127;p(1-tailed)< 0.10)(Fig. 3). Thesamestatisticaltrendwasobserved when the secretion events ofthe initial 3-min intervals were removed. No consistent result was obtained with respect to TO eversions. Feeding larvae everted their TOs more frequentlyin experiments with fewer ants present(p < 0.02), butthis difference largelydisappearedin the prepupal stage (p > 0.20). The actual mean tendinglevel increased from 2.63-3.13 ants/larva in 31(3-4):213-232, 1992 221 A the 5-ant trials to 5.27-5.35 ants/larva in the 15-ant experiments. threefold increase in the number ofavailable mutualists thus resulted onlyinanincreaseofthetendinglevelbyafactorof1.7-2.0. Inthe5-ant series,larvaeorprepupaeattractedonaverage52-62%oftheiravailable mutualists, whereas in the 15-ant trials only 35% ofthe ants actually tendedthelycaenidimmatures.Maximumtendinglevelswere,however, muchhigher. Twoprepupallarvae attracted8.29 and9.63 ants, respec- tively(averagedoverthe 15minperiod).Thesetwoanimalsweretended by 10-11 ants over several minutes and were then literally covered. Rank correlations Neither at the feeding stage nor in the prepupal phase was the DNO secretionrate significantlycorrelatedwithbodymass (rgvalues ranged from -0.006 to 0.318, p > 0.10). We also failed to detect significant correlations between the frequency ofTO eversions and DNO secretion rates (rgrangingfrom -0.244to 0.017, p > 0.17), orbetweenTO eversion rates and actual tendinglevels (rg between -0.304 and 0.154, p > 0.12). Correlationsdid,however, occurbetween actualtendinglevel andDNO secretion rates (feeding larvae: rg = 0.336, p = 0.093 (with 5 ants); rg = 0.381; p = 0.066 (with 15 ants); prepupae (with 15 ants): rg = 0.535, p < 0.01). Similarcorrelationswereobtained,whentheDNO secretiondata from the initial 3 min ofeach trialwere removed. These results suggest that DNO secretion and TO eversion rates are independentfromoneanotherandthatbodymassplaysatmostaminor role inthe m3n*mecophilyofZ. knysna immatures. Alarger antguardis somewhatmoreeffectivein stimulatingmorefrequentDNO secretions, but this relationship is far from being close. DNO Estimates ofindividual lifetime production of secretions Based on our experimentally established figures for average DNO secretion rates ofZizeeria knysna larvae, we here develop a model to estimate the total lifetime investment of individual larvae in these secretions. Forthispurpose,weassumethata)ourexperimentalvalues of secretion rates are representative, and b) secretion rates remain largely constant once a larva-ant association is established. Therefore, we only use the average secretion rates from the final 12 min of our experiments because at the beginning oflarva-ant interactions secre- tions occur more frequently for a short period of time (see above). DNO Acceptingthese premises, hourly secretion rates are as follows: • with5 antspertrial (i.e. actualtendinglevel 2.63-3.13 ants/larva): feeding L4 1.1 droplets/12 min = 5.5 droplets/h; prepupae 3.3 droplets/12 min = 16.5 droplets/h; • with 15 ants(i.e. actualtendinglevel5.3 ants/larva):feedingL4 1.9 droplets/12 min = 9.5 droplets/h; prepupae 5.1 droplets/12 min = 25.5 droplets/h. In our laboratory culture the active feeding period offourth instars 8 222 J. Res. Lepid. lasted 4 days andthelarvaeremained aboutonedayinthenon-feeding prepupalphase. Clark& Dickson(1971)recordedadevelopmentaltime of 6-7 days for the entire fourth instar in South Africa, hence our laboratory animals grew somewhat faster than under subtropical field conditions. Furthermore, we assume that aZ. knysna larvais tended by ants for atleast 8h dailythroughoutthefourthinstar. Forcomparison, we also calculate secretion rates under the assumption of a permanent (24 h daily)ant-association.Weassumethattheperiodofincreasedsecretion rates within the prepupal phase does not exceed 8 h because the non- feeding prepupa then becomes fully immobile and the DNO non-func- tional.FielddataontendinglevelsofZ. knysnaarenotyetavailable,but observations on many other facultatively myrmecophilous lycaenids suggestthatitisrealistictopostulate8-24hdailytendingby2-5antsper larva.OurmodelhenceprovidesupperandlowerlimitsforlifetimeDNO secretion amounts. Under these assumptions, a Z. knysna L4 in our 5-ant trials would secrete 308 (8-hour ant association) to 660 (permanently ant-tended) dropletsfromitsDNO.Therespectivevaluesforthe15-anttrialsare508 (8 h) to 1116 droplets (24 h). The diameterofsecretion dropletswas determinedusing acalibrated DNO eye-piece on the stereomicroscope. droplets ofZ. knysna larvae mm measured 0.233 ± 0.061 in diameter (n = 6, range 0.15-0.30 mm), correspondingto ameandropletvolumeof0.00662 pi. Forthefollowing mm calculations, we used an average droplet diameter of0.2 (volume 0.00419 pi) to avoid overestimation. The lifetime secretion volumes of individualZ. knysna larvae canthusbe estimatedtorangefrom 1.3-2. pi in 5-anttrials and from 2.1-4.7 pi in 15-ant trials. Data on the energy content ofDNO secretions are unavailable forZ. knysna or any closely related lycaenid butterflies. In the facultatively myrmecophilous European species Polyommatus (Lysandra) hispanus (Herrich-Schaffer, 1852) and P. icarus, the secretions contain approxi- mately 15 % carbohydrates (Maschwitz et al. 1975). If we assume a DNO similarcompositionof secretionsforZ. knysna,thentheindividual lifetimesecretionvolumes areequivalentto0.2-0.42mg(5 ants)or0.32- 0.71 mg carbohydrates (15 ants). The mean dry weight ± SD of adult specimens (males and females pooled)fromourlaboratoryculturewas2.78±0.71mg(range1.2-4.5mg, n = 43). In relationtothe average adultweight, the estimated carbohy- DNO % dratecontentoflarval secretionsisequivalentto7.2-15.1 (5-ant % trials) or 11.5-25.5 (15-ant trials). Discussion Temporal patterns ofinteractions Interactions between Zizeeria knysna larvae and ants show a clear