Do Biochemical Exaptations Link Evolution of Defense and Pollination Systems? Historical Hypotheses and Experimental Tests with Dalechampia Vines

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by University of New Orleans

University of New Orleans ScholarWorks@UNO

Biological Sciences Faculty Publications Department of Biological Sciences

3-1997

Do Biochemical Exaptations Link Evolution of Defense and Pollination Systems? Historical Hypotheses and Experimental Tests with Dalechampia Vines

Jerome J. Howard University of New Orleans, [email protected]

Follow this and additional works at: https://scholarworks.uno.edu/biosciences_facpubs

Part of the Biology Commons

Recommended Citation Armbruster, W.S., J.J. Howard, T.P. Clausen, E. Debevec, J. Loquvam, M. Matsuki, B. Cerendolo, and F. Andel. 1997. Do biochemical exaptations link evolution of defense and pollination systems? Historical hypotheses and experimental tests with Dalechampia vines. American Naturalist 149 (3): 461-484.

This Article is brought to you for free and open access by the Department of Biological Sciences at ScholarWorks@UNO. It has been accepted for inclusion in Biological Sciences Faculty Publications by an authorized administrator of ScholarWorks@UNO. For more information, please contact [email protected]. The University of Chicago

Do Biochemical Exaptations Link Evolution of Plant Defense and Pollination Systems? Historical Hypotheses and Experimental Tests with Dalechampia Vines Author(s): W. Scott Armbruster, Jerome J. Howard, Thomas P. Clausen, Edward M. Debevec, John C. Loquvam, Mamoru Matsuki, Bianca Cerendolo and Frank Andel Reviewed work(s): Source: The American Naturalist, Vol. 149, No. 3 (Mar., 1997), pp. 461-484 Published by: The University of Chicago Press for The American Society of Naturalists Stable URL: http://www.jstor.org/stable/2463378 . Accessed: 08/08/2012 12:07

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press, The American Society of Naturalists, The University of Chicago are collaborating with JSTOR to digitize, preserve and extend access to The American Naturalist.

http://www.jstor.org Vol. 149,No. 3 The AmericanNaturalist March1997

DO BIOCHEMICAL EXAPTATIONS LINK EVOLUTION OF PLANT DEFENSE AND POLLINATION SYSTEMS? HISTORICAL HYPOTHESES AND EXPERIMENTAL TESTS WITH DALECHAMPIA VINES

W. SCOTT ARMBRUSTER,"*JEROME J. HOWARD,1'tTHOMAS P. CLAUSEN,2EDWARD M. DEBEVEC,1 JOHNC. LOQUVAM,1MAMORU MATSUKI,1't BIANCA CERENDOLO,2 AND FRANKANDEL2 'Instituteof ArcticBiology and 2Departmentof Chemistry,University of Alaska, Fairbanks,Alaska 99775-7000 SubmittedOctober 16, 1995;Revised May 30, 1996;Accepted June 4, 1996

Abstract.-Mappingresin secretion and pollinationecology onto the estimated phylogeny of speciesof theeuphorb vine Dalechampia generated two historicalhypotheses: resin rewards offeredto pollinatorsby Dalechampia flowers evolved by minormodification of a preexisting, resin-based,floral defense system, and resindefense of leaves in advancedspecies evolved by modificationof thepreexisting resin-reward system. From these hypotheses, we derivedtwo predictions:floral reward resins are chemicallysimilar to putativefloral defense resins and exhibitantiherbivore activities, and foliarresins are chemicallysimilar to rewardresins and also exhibitantiherbivore activities. We testedthese predictions by chemicalanalyses and by usinga broadsample of Neotropical herbivorous insects in a seriesof bioassays. All floraland foliarresins were chemically similar. Tests with two generalist (Orophus tesselatus, Orthoptera: Tettigoniidae;Atta colombica, Hymenoptera: Formicidae) and four Dalechampia-specialist herbivores(Syphraea sp., Coleoptera:Chrysomelidae; Ectima rectifasciata, Hamadryas ipthime, andHamadryas amphinome, Lepidoptera: Nymphalidae) showed floral and foliar resin to deter significantlyfeeding or leaf cutting.These resultssupport our two hypothesesand indicate that,in thissystem, biochemical exaptations have playeda majorrole in the evolutionof plant-insectrelationships, adaptations reducing herbivory have affected the evolution of plant- pollinatorrelationships, and adaptationsfor pollination have affectedthe evolutionof plant- herbivorerelationships.

Mutualismsand antagonismsbetween plants and animalsplay majorroles in the functionof biologicalcommunities (Gilbert and Raven 1975;Gilbert 1980; Thompson1982; Futuyma and Slatkin1983; Strong et al. 1984;Boucher 1985). Whilethe short-term ecological dynamics of plant-animal interactions have been reasonablywell studiedin a numberof systems,we knowrelatively little about how theserelationships originate and evolve (Thompson1982, 1994; Futuyma 1983;Schemske 1983; Boucher 1985; Mitter et al. 1988;Farrell et al. 1992;Farrell

* To whomcorrespondence should be addressed;E-mail: [email protected]. t Presentaddress: Department of BiologicalSciences, University of New Orleans,New Orleans, Louisiana70148. t Presentaddress: Research School of BiologicalSciences, Australian National University, Can- berraACT 2601,Australia.

Am. Nat. 1997. Vol. 149, pp. 461-484. ? 1997 by The Universityof Chicago. 0003-0147/97/4903-0003$02.00.All rightsreserved. 462 THE AMERICANNATURALIST and Mitter1994). For example,how do completelynew mutualistic or antagonis- ticinteractions come into existence? What constrains the evolution of interactions amongspecies? Do evolutionaryresponses of a populationto one set of inter- actingorganisms (e.g., herbivores)affect interactions with other organisms (e.g., pollinators)? Integrationof modernphylogenetic methods with ecological data is allowing these questionsto be addressedfrom a long-termevolutionary perspective, throughthe generationof explicithistorical hypotheses (see Mitteret al. 1988; Brooksand McLennan1991; Harvey and Pagel 1991;Armbruster 1992; Farrell et al. 1992; Farrelland Mitter1994; Thompson 1994). However, the historical hypotheseshave onlyrarely been testedwith independent data or experiments (see Armbruster1993; Pellmyr and Huth 1994; Pellmyr et al. 1996).In thepresent study,we synthesizedphylogenetic and ecological information to generate historical hypothesesabout the evolutionary relationship between plant pollination and antiherbivoredefense systems. We thendevised several ecological experiments to testthese hypotheses.

EVOLUTION OF PLANT-ANIMAL RELATIONSHIPS The interactionsbetween plants and otherorganisms are commonlymediated by plantsecondary chemistry. Two featuresindicate that chemically mediated interactionsmay evolve rapidly. First, organisms often have highlyspecific re- sponsesto compounds,and minorstructural changes in a compoundmay result in majorchanges in biologicalactivity (e.g., Dodson et al. 1969;Farentinos et al. 1981;Reichardt et al. 1984;Bryant et al. 1992).Second, a singleplant compound mayaffect more than one interactingspecies (e.g., Clausenet al. 1986;Pellmyr and Thien1986; Bryant et al. 1991),and theevolution of plant secondary chemis- trymay be influencedby simultaneousinteractions with different functional groupsof organisms(e.g., pathogensvs. herbivoresvs. mutualists). Whilethe role of secondarycompounds in plantdefense (see reviewin Rosen- thaland Berenbaum 1992) might appear unrelated to a rolein attracting mutualists (see reviewin Simpsonand Neff1981), chemical similarities indicate that these tworoles may sometimes be linked.For example,1, 8-cineole,a monoterpene, is a strongdeterrent to snowshoehares and apparentlyfunctions as a defense compoundin balsampoplars (Populus balsamifera) in Alaska (Mattes et al. 1987; Reichardtet al. 1990).The same compoundis secretedby tropicalorchids and servesas an attractantand rewardto pollinatingmale-euglossine bees (Dodson et al. 1969).Thus, the possibilityexists that a singlecompound or mixtureof compoundsin a plantcould simultaneously, or sequentially(in developmentalor evolutionarytime), play several very different ecological roles. Pellmyr and Thien (1986) pointedout thatmany fragrances that attract pollinators to flowersare biochemicallysimilar to compoundsthat deter herbivores, and the same or similar compoundsmay play the role of attractionin theflowers and deterrencein the leaves. If thisis common,relatively small genetic changes could cause rapid shiftsin ecologicalrelationships between plants and theirmutualists and antagonists.This raises some interestingquestions. Do complexplant-animal interac- EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 463 tionsnecessarily reflect a longhistory of coevolution, as has oftenbeen thought? Or are ecologicalrelationships between plants and animalsoften evolutionarily labile,changing rapidly in responseto subtleselective pressures or stochastic events(see Pellmyrand Thompson1992)? Are biochemicalexaptations (sensu Gouldand Vrba 1982;i.e., preadaptations)important, and can newrelationships evolve suddenlyin species withpreaptations in place? Do plantrelationships withherbivores influence the evolution of plant-pollinatorinteractions and vice versa?

EVOLUTION OF EUPHORB VINES To beginaddressing these questions, we investigatedthe chemical ecology of a modelplant-animal system: species of the tropicalplant genus Dalechampia (Euphorbiaceae)and their mutualists and antagonists. Most species of Dalecham- pia offeran unusualreward to pollinators:terpenoid resins. The resinis secreted byblossom glands and is collectedby pollinating bees foruse in nestconstruction (Armbruster1984, 1986, 1988). This rewardsystem has been observedin only one othergenus (Clusia, Clusiaceae; Armbruster 1984). How did suchan unusual mutualismarise? The widespreadrole of resin as a defensesystem in otherplants and phylogeneticstudies (see below) indicatesthat resin secretion may have originatedin Dalechampia as a defensesystem and secondarilyassumed a reward functionafter resin-collecting bees beganvisiting the flowers incidentally to steal resin(Armbruster 1984, 1993). If thishypothesis is correct,the initial functional shift from a resindefense to a resinreward required little or no geneticchange. Yet theresin reward repre- senteda novelmutualistic relationship with resin-collecting bees, led to a radical shiftin the natureof pollinationof Dalechampia,and appearsto have affected thecourse of evolutionof mostof this large and widespreadgenus (about 100of the 120extant species are pollinatedby resin-collectingbees; Pax and Hoffmann 1919;Armbruster 1984, 1988; Webster and Armbruster1991). Thus an ecological coincidencemay have resultedultimately in a majorchange in the subsequent evolutionand ecologyof thisgroup of plants.

STUDY SYSTEM Dalechampia(Euphorbiaceae) is a genusof about 120 tropical,mostly viny species. It has a nearlypantropical distribution, although the greatestdiversity (ca. 90 spp.) is foundin theNeotropics (Webster and Armbruster1991). Pollination by Resin-CollectingBees Severalhundred species of vinesand shrubs(Dalechampia) and of treesand epiphytes(Clusia) produceterpenoid resins as rewardsfor pollinators (Armbrus- ter 1984). These plantsoccur in mostlowland tropical regions of the world. Species offeringresin rewards are pollinatedby femaleeuglossine (Apidae), fe- maleanthidiine (Megachilidae), or workermeliponine (Apidae) bees thatcollect floralresins for use in nestconstruction (Skutch 1971; Armbruster and Webster 1979;Armbruster 1984). The floralresins in Dalechampia are mixturesof oxygen- 464 THE AMERICANNATURALIST atedtriterpenes. In bothDalechampia and Clusiathe floral resins are waterproof, slow to harden,and producedon a regular,predictable basis. For thesereasons, floralresins are probablyvaluable resources for the hundreds of speciesof tropical bees thatuse resinin nestconstruction (Michener 1974, 1979; Jayasingh and Freeman1980; Cane et al. 1983;Roubik 1983, 1989; Armbruster 1984; Howard 1985; Messer 1985).

Evolutionof Pollination Systemsin Dalechampia The unisexualflowers of Dalechampiaare unitedinto functionally bisexual, pseudanthial(blossom) inflorescences (Webster and Webster1972). Associated withthe staminate flowers in most species is a glandlikestructure. In all buta few speciesthe gland secretes a resinousmixture of oxygenated triterpenes. Cladistic analyses,based on morphologicaldata not directlyrelated to pollination,have been used to estimatethe phylogeny of Dalechampia. Most of thephylogenetic topologyis stablein theface of addition and removalof taxa and characters(see Armbruster1993, figs. 4, 5) and thereforecan probablybe trusted(see Donoghue 1983; Felsenstein1985; Armbruster1992). By mappingthe knownpollination systemsof Dalechampia species onto the phylogenetic tree, it has beenpossible to estimatethe evolutionary history of the pollinator-reward system (Armbruster 1993; see also Donoghue1989). These resultsindicate that the mostprimitive speciesof Dalechampia lacked the large resin gland and werepollinated by pollen-collectingor fragrance-collectingbees. Earlyin the evolutionof the genus, resinbecame the pollinator reward. This event was followedby extensiveradia- tion(Armbruster 1993). Herbivoryon Dalechampia NeotropicalDalechampia species are hostto severaloligotrophic nymphalid lepidopteranherbivores including, in approximateorder of abundance,Hama- dryas spp., Dynamine spp., Ectima spp., Catonephele spp., Myscelia spp., Biblis spp.,and Mestraspp. (Armbruster1983; Jenkins 1983, 1984, 1985a, 1985b; Arm- brusterand Mziray1987; DeVries 1987;Otero 1988). Dynamine species tend to feed on staminateflowers, developing seeds, and otherinflorescence parts, whereasmembers of the othergenera usually feed on the leaves (Armbruster 1982,1983; DeVries 1987).Paleotropical Dalechampia are hostto membersof severalother genera of oligotrophicnymphalid butterfly larvae, including Byblia spp. and Neptidopsisspp. (Armbrusterand Mziray1987; Armbruster 1994). An- othercommon specialist in theNeotropics is theeuphorb flea beetle (Syphraea spp.; Coleoptera:Chrysomelidae: Alticinae). Otherinsects that commonly feed on Dalechampiaspecies include leaf-mining and stem-boringbuprestid beetles and various generalist orthopterans (esp. Tetti- goniidae;Armbruster 1983; Armbruster and Mziray1987; H. Hespenheide,per- sonal communication).Another generalist herbivore that attacks both inflorescences and leaves of some species is the leaf-cuttingant, Atta (Hymenoptera: Formicidae;W. S. Armbruster,unpublished data). Because mostspecies of Da- lechampiaoccur in habitatswhere Atta are abundant,these plants are likelyto experienceoccasionally heavy defoliation by leaf-cuttingants. EVOLUTIONOF PLANTDEFENSE AND REWARDSYSTEMS 465

TerpenoidResin as a Defense Compound Most plantspecies thatsecrete terpenoid resins are thoughtto do so as a meansof defendingtheir vegetative structures from attack by herbivoresand/or microorganisms.Conifers and treesin the Leguminosaeand Burseraceae,for example,exude large amounts of terpenoidresin in responseto woundsin their stemsand trunks(Langenheim 1969). Some plants,such as Hymenaeaand Co- paifera(Leguminosae), sequester in theirleaves terpenoidresins that are effec- tive in deterringherbivores and microbes(Stubblebine and Langenheim1977; Arrheniusand Langenheim1983; Langenheim and Hall 1983;Langenheim and Stubblebine1983; Langenheim et al. 1986;Feibert and Langenheim 1988; Macedo and Langenheim1989). Terpenoid constituents of leaves of a varietyof tropical plantsreduce the acceptability of the leaves to Atta,a majorgeneralist herbivore in the Neotropics(Hubbell et al. 1984)and oftenhave negativeeffects on the ants' fungalmutualist (Chen et al. 1983;Howard 1987; Howard et al. 1988).

CLADISTIC AND HISTORICAL HYPOTHESES The phylogeneticrelationship among a taxonomicallybroad sample of 42 species ofDalechampia (about half the New Worldspecies) was estimatedby con- ductingcladistic analysis on 54 morphologicalcharacters (see Armbruster1993, 1997for methodological details and the species-by-charactermatrix). We then used MacClade3.0 (Maddisonand Maddison 1992) to map,under the assumption of parsimony,the following traits onto the phylogenetic tree: resin secretion by floralbractlets, pollination by resin-collecting bees, and resinsecretion by leaves and stipules.To preserveindependence of thetwo steps of theanalysis (estima- tionand mapping),the mapped characters were not used inphylogeny estimation (see Armbruster1992 for discussion). Figure1 showsthe strict consensus of the 968 maximallyparsimonious phylo- geneticrelationships among the Dalechampia species under study. From mapping ecologicaltraits onto the tree (using the parsimonycriterion), we hypothesized thatresin secretion by floralbractlets originated prior to its role as a pollinator reward.To checkthe robustness of thishypothesis, we specifiedthe alternative treethat allows pollinationby resin-collectingbees to originatesimultaneously withresin secretion, using the topological constraint option in PAUP (Swofford 1993).The shortesttree under this constraint was six steps(4.1%) longer,indicat- ingsupport for our phylogenetichypothesis. It also requiredtwo to threeextra steps(200%-300% increase) in mapping pollination traits to achievesimultaneous originof resinsecretion and pollinationby resin-collectingbees. Thus resinmay have initiallyplayed no role in pollinationbut, instead, func- tionedto defendstaminate flowers from attack by florivorousinsects. This hy- pothesisis also supportedby blossommorphology. In themost primitive Dale- champia species, the resiniferousbractlets are dispersed throughoutthe staminateinflorescence and cover the staminatefloral buds. Only in moread- vancedspecies pollinated by resin-collectingbees are the homologousbractlets aggregatedinto a glandto facilitateresin collected by pollinators(fig. 2). -~~~~~~~~~~~ c

t _ r l | | a,| | ~~~~RESINSECR. STAMINATEBRACTLETS

FIG. 1.-Tree showingconsensus of 968 mostparsimonious phylogenetic relationships among42 Neotropicalspecies of Dalechampia and three candidate sister taxa: Tragiaspp., Plukenetiasubgenus Cylindrophora spp., and Plukenetia subgenus "Euplukenetia" spp. (see Armbruster1993 for details). The shadingindicates the most parsimonious mapping of resin secretionby floral bractlets; the straight slash mark indicates the inferred origin of pollination by resin-collectingbees; and the wavyslash markindicates the inferredorigin of resin- secretingleaf and stipuleglands. The paraphyleticsection, Rhopalostylis, includes all Da- lechampiabelow the straight slash mark.

A B FIG. 2.-Arrangementof resiniferousbracelets. Circles, staminateflower; curved lines, resiniferousbractlet. A, PrimitiveDalechampia not pollinated by resin-collecting bees (e.g., D. fragrans);resin has a presumeddefense function and is diffuselydeployed. B, More advancedDalechampia pollinated by resin-collectingbees (e.g., D. websteri);resin acts as a pollinatorreward and is presentedby a unifiedgland, formed by an aggregationof resiniferous bractlets. EVOLUTIONOF PLANTDEFENSE AND REWARDSYSTEMS 467

The last twistin thishistorical scenario is theevolution of resinsecretion by vegetativestructures. One smalladvanced clade secretesresin at themargins of stipulesand leaves (fig.1), indicatinga recent resumption of the defense function of terpenoidresin, perhaps to decreaseattack by thenearly ubiquitous specialist folivore,Hamadryas, and/or equally ubiquitous leaf-cutting ants. This historicalreconstruction was used to generatetwo predictions.First, if floralreward resins are derivedfrom floral defense resins, they should be chemicallysimilar to floraldefense resins and shouldstill inhibit feeding and/or growth ofinsect herbivores. Second, if foliar resins evolved for leaf defense by modifica- tion of the resin-rewardsystem, they should be chemicallysimilar to reward resinsand shouldinhibit growth and/or feeding of insectherbivores. Tests of thesepredictions are presentedbelow.

EXPERIMENTAL METHODS

Study Sites and Choice of Assay Material The bioassayswere conducted in the Gamboa Laboratory facility of the Smith- sonianTropical Research Institute, Balboa, Panama,and in thelab facilitiesat theBiology Department, University of New Orleans(Atta trials on Viburnum). Plantmaterials and insectswere collected in ParqueNacional Soberania, along PipelineRoad, nearGamboa. To achievea strongtest of the prediction that Dalechampia resin has antiherbi- vore properties,we conducteda varietyof choice testsand growthtrials on a taxonomicallybroad sample of herbivore species. These included two generalists, Orophus tesselatus (Orthoptera:Tettigoniidae) and Atta colombica (Hymenop- tera:Formicidae) and four euphorb specialists, Syphraea sp. (Coleoptera:Chrys- omelidae: Alticinae),Ectima rectifasciata,Hamadryas ipthime,and Hamadryas amphinome(Lepidoptera: Nymphalidae). For choiceand growth trials we usedleaves from the two commonest species of Dalechampia along PipelineRoad: Dalechampia heteromorphaand Dalechampia tiliifolia(app. A). These are bothadvanced species thatoffer resin rewards in theirblossoms but lack resin secretion from their leaves. Resin used in the experi- mentswas collectedfrom the floralglands of threespecies withcopious resin secretion:Dalechampia dioscoreifoliaand D. tiliifoila(both growingalong Pipe- lineRoad) andDalechampia armbrusteri cultivated in a greenhousein Fairbanks, Alaska (originalstock collected near Itabuna, Bahia, Brazil).All threeare pollinated by resin-collectingbees. Dalechampia tiliifoliaand D. armbrusteriare advancedspecies and fairlyclosely related, whereas D. dioscoreifoliais a derived speciesin a distantclade. We also collectedresin from secretary glands on vege- tativeparts, primarily stipules and leaves,of the advanced species Dalechampia stipulacea(in the same clade as D. armbrusteri),cultivated in a greenhousein Fairbanks(original stock was collectednear Campinas, Sao Paulo,Brazil; app. A). Chemical Methods Twentygrams of floral glands from greenhouse-cultivated D. armbrusteri were extractedwith diethyl ether by soakingfor 12 h. The plantmaterial was removed, 468 THE AMERICANNATURALIST andthe ether was decolorizedwith activated cabon, dried with anhydrous magne- siumsulfate, and flashevaporated to yield0.55 g ofresidue (2.75%). The residue was thenflash chromatographed (Still et al. 1978)using silica gel (40 pLm)as absorbentand methylenechloride/acetone (99.5: 0.5) as theeluting solvent. Two fractionswere isolated. The firsteluted fraction (0.33 g) containedprimarily four triterpeneketones, which were later identified using nuclear magnetic resonance (NMR) (table 1). The secondfraction (0.22 g) containedtwo majortriterpene alcohols.The ketoneand alcoholfractions were used in bioassays. Stipuleswere cut from whole vines of greenhouse-cultivated D. stipulacea and subsequentlyextracted and chromatographedaccording to theabove schemeto yieldtwo fractions.The firstfraction was determinedby gas chromatography/ mass spectrometryto be composedprimarily of mono-and sesquiterpenes.The secondfraction was foundto containseveral triterpene alcohols (table 1). These twofractions were used in bioassays. Crudefloral resin was collectedfrom fresh field-collected floral material for use in additionalbioassays. Resin was scrapedoff floral glands using forceps and dissolved in methylenechloride (CH2Cl2; Dalechampia scandens, D. tiliifolia)or diethylether (D. dioscoreifolia). Identificationof the smallamounts of floralresins secreted by the primitive species, Dalechampia heterobractea and Dalechampia attenuistylus,was not possiblewith NMR; we insteadmatched mass spectraproduced by integrated gas chromatography/massspectrometry. Statistical Analyses All proportionaldata were arcsinesquare-root transformed prior to analysis. A varietyof one-way factorial, two-way factorial, and randomized-block analyses of variancewere performed using methods described in Sokal and Rohlf(1981) and Neteret al. (1985).Post hoc multiplecomparison tests followed the Tukey HSD procedure.Feeding choice trials were analyzed with the paired-sample t-test (Sokal and Rohlf1981). The SYSTAT package(Wilkinson 1988) was used for mostcomputations. Feeding Choice Trials with0. tesselatus For each trial10 leaves of D. heteromorphawere collectedfrom about five plantsand randomized.The terminalhalf of the right and leftleaflets (the leaves are ternate)were cut off to be used in paired-choicetrials. One leafletpiece from a leafwas used forthe controltreatment and the otherwas used forthe resin treatment,in a randomizedblock design, with each individualleaf-nymph combi- nationas a block.The resinwas appliedas a solutionin methylenechloride to achieve2% resin/leaffresh weight (resin weights of freshD. stipulacealeaves varybetween 2% and 10%). The sameprocedure was followedon controlleaves exceptthat the methylene chloride solution lacked resin. Four test solutions were used (app. A). The matchedtreatment and controlleaflets were placed in 150-mLplastic vials withplaster-of-paris bottoms, each leafletleaning against opposite walls of the vial. In each vial,we placeda second-to-fourth-instarOrophus nymph that was (N + + +~~~~~~~~~~~~~~~~~~~- +~~~~~~

CA Q

o ?~~~~~

Cd - - C s

+ +~~+

00 + +~~~~

0 ~~~) ~) u S

z~~~~~~l a

0 ~~~~~

cn ~ ~ ~ ~ ~ ~~~ 470 THE AMERICANNATURALIST

alreadyacclimated to feedingon D. heteromorphaleaves. Feedingtrials lasted 40-53 h, generallyuntil about half of at leastone leafhad been eaten. The amountof leafeaten was determinedby comparingposttrial leaf-outline traceswith pretrial leaf-outline traces. The papertracings were weighed to obtain percentagesconsumed. Tests on A. colombica Severalcrude resins and resinfractions were tested for behaviorally repellent or deterrentactivity toward leaf-cutter ants in "pickup" tests(Howard 1987). Resinwas dissolvedin diethyl ether or methylenechloride and coated onto small, steel-cutwheat flakes to yielda concentrationof 1%-2% by weight.Control flakeswere preparedby brieflysoaking in solvent.Six ant coloniesin Parque NacionalSoberanfa were tested by simultaneouslyplacing single test and control flakeson trailsand replacingthem as theywere harvested. Flakes wereoffered for 15 min,and the averagenumber of test and controlflakes harvested per minutewas calculatedfor statistical analysis. The effectof Dalechampia resin on Attacutting performance was evaluatedin two ways. We used field"cutting" bioassays (Howard et al. 1988)to studythe effectof applyinga thin,even coat of resinto the leaf surfaces.Dalechampia tilifolialeaves were traced and rinsed with methylene chloride to removenonpo- lar secretions.An aliquotof resin equal to 2% offresh weight was appliedto half theleaves. Controlswere prepared by applying solvent alone. Leaf petioleswere insertedinto water-filled vials, and leaves wereplaced by the side of activeant trails.Four antcolonies were presented with three control and threetest leaves for2 h, afterwhich leaves and discardedfragments were recovered. The amount harvestedfrom each leafwas estimatedby comparingthe mass of leaftracings beforeand aftereach experiment. Resinis secretedas dropletson marginsof leaves and stipulesof some Dalech- ampiaspecies. We askedhow resin deployment in droplets affects cutting perfor- mancein a secondset of experimentsconducted at the Universityof New Or- leans.Laboratory colonies of A. colombicaexcavated from the vicinity of Parque NacionalSoberania were permitted to forageon familiarfood, Viburnum rufidu- lumleaves, to themargins of whichdroplets of D. tilifoliafloral resin had been applied. Controlleaves were preparedby applyingsmall droplets of etherto the margins.We observedand recordedthe activityof individualworker ants continuouslyfrom the time they encountered the leaves untilthey left. Toxicityof severalresin constituents (app. A) to leaf-cutterant workerswas investigatedby measuringsurvivorship on artificialdiets with and withoutresin. Experimentswere conductedin an un-air-conditionedlaboratory in Panama. Workersfrom a singlecolony in ParqueNacional Soberania were randomly assignedto 100 x 15mm petri dishes (n = 5 perdish) and 10petri dishes randomly assignedto each treatment.Ants were fed a liquiddiet (Boyd and Martin1975) in a smalldish containing a 4.25-cm glass fiber disk. Resin constituents in 400 [IL of diethylether solvent were applied to disks;control treatment was 400 [IL of solventalone. Each daythe diet and diskswere changed, and thenumber of ants remainingalive was recorded. EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 471

Feeding Choice Trials withSyphraea sp. The identicalprocedure was followedfor the Syphraea feeding choice trials as was used in theOrophus feeding choice trials, except that the resin was applied as dropletsof a concentratedviscous solution, and D. tiliifolialeaves wereused (thebeetles were originally collected on D. tiliifolia).The resinused in thetest solutionwas crudeextract of D. tiliifoliafloral glands. Twelve beetle-leaf combi- nations(one perchamber) were used. Feeding Choice Trials withH. ipthime The identicalprocedure was followedfor the Hamadryas feeding choice trials as was used in theOrophus feeding choice trials, except that D. tiliifolialeaves wereused. The resinused in thetest solution was crudeextract (containing both triterpenealcohols and ketones)of D. scandensfloral glands. GrowthTrials withE. rectifasciata Growthtrials were conducted as randomizedblock designs using leaves ofD. tiliifolia,following general procedures outlined in Ayresand MacLean (1987). Each leafwas cutinto four sections and systematicallyassigned to treatmentand control;right and leftlobes wereused fortreatments, the center lobe was used forthe control, and the basal portionwas used as thematched control to estimate watercontent. Treatment compounds were dilutedin methylenechloride and appliedto theleaf pieces to achievea 2% resin:wet-weight application. Ectima eggs were collected fromleaves of D. heteromorpha,D. dioscoreifolia, and D. scandensalong PipeleneRoad, Parque Nacional Soberania.The eggs werelaid singlyand camefrom several to manyfamilies. Larvae werereared at roomtemperature on D. tiliifoliaand D. heteromorphaleaves untilthey had grownto thefourth instar. Each was weighedjust beforethe trial. Treatmentand controlleaves and larvae were placed in 150-mLvials with moistenedplaster bottoms. Vials were placed in an un-air-conditionedlab in Gamboafor the duration of the growth trial (ca. 24 h). The lengthof the trial was themaximum possible without the larva running out of food or beginningto molt. At theend of thetrial the larvae were reweighed. Frass and leaves werecol- lected,oven dried, and weighed. Sample larvae were also weighedand thenoven driedand reweighedto determinedry- to wet-weightratios. Initialdry weights of trialleaves wereestimated from the dry- to wet-weight ratiosof the correspondingmatched control for each leaf. However,we later foundthat water content varied among leaf sections, so we estimatedthe average differencesand used thesevalues to correctthe estimatedinitial dry weight of theexperimental leaves. Two separatetrials were run. In one trial,the performanceof larvae on triterpenealcohols versusmono- and sesquiterpeneprecursors (both from D. stipulaceastipules) versus the controlwas compared.In the second trial,the performanceof larvae on triterpenealcohols versus triterpene ketones (from floral resinof D. armbrusteri)versus the control was compared. The relativegrowth rate (RGR), relativeconsumption rate (RCR), and effi- 472 THE AMERICANNATURALIST ciencyof conversionof ingested material (ECI) werecalculated according to the formulasin appendixB.

GrowthTrials withH. amphinome The proceduresdescribed above forEctima were used in the Hamadryas growthtrials, except that eggs were collected from and all larvaewere reared on leaves of D. tiliiolia. Also, experimentalleaves were dividedinto six sections and randomlyassigned to treatments,and all larvae(which in thisspecies hatch synchronouslyfrom large clutches) were begunsimultaneously for each trial. Each trialwas conductedon one maternalhalf-sib family. Threeseparate trials were run. In two trialsthe performance of two families of fourth-instarlarvae was evaluated.In a thirdtrial the performanceof third- instarlarvae from a thirdfamily was evaluated(app. A). The values forRGR, RCR, and ECI werecalculated as forEctima.

EXPERIMENTAL RESULTS

Chemical Analyses The bractletssubtending the staminateflowers in mostspecies of the basal, paraphyleticsection Rhopalostylis secrete small droplets of resin(fig. 1). The resinssecreted by twomembers of this section, Dalechampia heterobractea and Dalecahmpiaattenuistylus, are chemicallysimilar to thefloral resin of mostad- vanced species (table 1). However,the resinsecreted by the bractletsof the staminateinflorescence of speciesin sectionRhopalostylis plays no rolein pollination,because these species are pollinatedby fragrance-or pollen-collecting bees (Armbrusteret al. 1992;Armbruster 1993). The leaf blades and stipulesof D. stipulaceaand its close relativessecrete resinin smalldroplets from tentacular processes along the margins.This foliar resincomprises the same basic mixtureof compoundsas thefloral resin of advanced species,except that the ketoneconstituents are minoror absent(table 1). The chemicalsimilarity of floral"defense" resins, floral reward resins, and foliardefense resins is consistentwith our predictions.

Tests withGeneralists Feedingchoice trials with Orophus tesselatus.-Orophus nymphs ate an average of 21.0% of theleaves treatedwith the triterpene alcohol mixture extracted fromthe leaves and stipulesof D. stipulaceaversus 43.6% ofthe control leaves; this differencewas significant(t = 3.13, df = 9, P = .012). The nymphsate an averageof 28.3% ofthe leaves treatedwith the triterpene ketones from the floral resinof Dalechampiaarmbrusteri versus 42.4% of the controls;this difference was also significant(t = 3.24, df = 8, P = .012). Orophus nymphsate an average of 28.6% of theleaves treatedwith the triterpene alcohols from the floral resin ofD. armbrusteriversus 31.5% of the controls; this difference was notsignificant (t = 1.51, df = 9, P = .165). The nymphsate an average of 33.1% of the leaves treatedwith the crudeextract from Dalechampia tiliifolia floral resin glands (a EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 473

TABLE 2

MEAN RATES OF PICKUP OF RESIN-TREATED WHEAT FLAKES BY ATTA WORKERS FROM SEVEN COLONIES ALONG PIPELINE ROAD, PARQUE NACIONAL SOBERANiA, PANAMA

Treatment Rate F (df) P

Floral triterpenealcohols (from Dalechampia armbrusteri) .49 49.7 (1, 5) .001 Control 1 .87 Floral triterpene ketones (from D. armbrusteri) .49 81.5 (1, 5) <.001 Control 2 .92 Floral triterpene mixture (from Dalechampia dioscoreifolia) .77 11.4 (1, 9) .008 Control 3 1.16 Foliar triterpene alcohols (from Dalechampia stipulacea) .25 51.2 (1, 5) .001 Control 4 .92 Foliar mono- and sesquiterpenes (from D. stipulacea) .54 31.1 (1, 5) .003 Control 5 .82 Crude floral resin (from Dalechampia tiliifolia) .45 25.0 (1, 5) .004 Control 6 .56

NOTE.-Controls are paired with each treatment. Rates are in number of flakes picked up per minute. F and P values refer to test for significance of treatment effect with effect of colony estimated and removed (randomized-block ANOVA). mixtureof triterpeneketones and alcohols)versus 49.2% of the controls;this differencewas significant(t = 2.70, df = 8, P = .027). A weak pattern of reducedfeeding inhibition in laterinstars was observed(M. Matsukiand W. S. Armbruster,unpublished data); thisindicates that the inhibitoryeffect of resin maybe primarilymechanical (larger instars are strongerand less affectedby the stickinessof theresin). Atta pickup and toxicitytrials.-All resin componentsand mixturesshowed some inhibitoryeffect on Attacollection of wheat-flakebaits. Bait pickupwas decreasedto 27% ofthe control rate by the D. stipulaceafoliar triterpene alcohol and to 80% of the controlrate by the D. tiliifoliafloral resin. The inhibitory activityof D. stipulaceafoliar mono- and sesquiterpenes,D. armbrusterifloral resins,and Dalechampiadioscoreifolia floral resins were somewherebetween theabove values(table 2). The inhibitoryeffects were all highlysignificant (P < .008 in all cases; table2). Therewere no significanteffects of Dalechampiaresin on ant survivorshipat anyconcentration tested. Dalechampia tiliifolia floral resins at concentrationsup to 1 mg/g,D. dioscoreifoliafloral resin up to 0.1 mg/g,and D. stipulaceafoliar resinup to 0.1 mg/gproduced no significantdifferences in survivorshipcompared withcontrols. Atta cuttingtrials. -Atta in the fieldcut a mean of 9.1% of controlD. tiliifolia leaves and a meanof 11.7% of the resinfilm-treated leaves. Althoughcolony had a significanteffect on proportionof leaves cut (F = 21.5, df = 1, 21, P < .001),coating the leaves withresin did not(F = 0.13, df = 1, 21, P = .72). 474 THE AMERICANNATURALIST

In thelaboratory, the presence of resindroplets did notdiscourage ants from investigatingleaves and biting their edges. In fact,ants encountering resin-treated leaves were significantlymore likely to explorethe leaves by bitingthan were antsencountering control leaves (49 out of 124vs. 35 out of 131; Pearsonx2 = 4.72, df = 1, P = .03). Antsencountering resin-treated leaves ran a significant riskof becomingmired in a resindroplet. Approximately 28% (35 out of 124)of all antsinvestigating resin-treated leaves becamestuck in resinfor at leasta few seconds.One ant was unableto freeitself during 2 h of observationand would presumablyhave diedunder natural conditions. A similarproportion of ants encountering both resin-treated and controlleaves attemptedto cut pieces (18 out of 131vs. 16 out of 124). However,ants cutting resin-treatedleaves were significantly less likelyto finishthe cut than ants cutting controlleaves (seven out of 16vs. 14out of 18;X2 = 4.15,df = 1,P = .042).Ants encounteringresin droplets typically stopped cutting and groomed themselves. A majorityof theseants (nineout of 16) did not resumecutting (vs. 14 out of 18 completingcuts on controlleaves). Aftercutting resin-treated leaves, ants had greatdifficulty in raising and carrying leaf pieces because resin caused leaf pieces to adhereto theirmandibles and antennaeor to otherobjects. Overall,ants attemptingto cutpieces from resin-treated leaves were significantly less success- fulat cuttingand carryingoff leaf pieces thanwere ants encounteringcontrol leaves (only two out of 16 vs. 14 out of 18; x2 = 14.49,df =1, P < .001). These observationsindicate that the inhibitory effect on antsof resindeployed in dropletsis primarilya mechanical one. Tests withSpecialists Syphraeafeeding choice trials.-Syphraea beetles (adults) consumed an average of 1.21% (SE = 2.43%) ofD. tiliifolialeaves treatedwith droplets of crude D. tiliifoliafloral resin. The beetlesconsumed an averageof 10.87%(SE = 7.51%) ofthe control leaves. This difference was highlysignificant (ANOVA; F = 16.46, df = 1, 11, P = .001). Hamadryas ipthimefeeding choice trials.-Hamadryas ipthimelarvae con- sumeda meanof 22% of theD. tiliifolialeaves coatedwith the crude floral resin fromblossoms of D. scandens(n = 10),and a meanof 58% ofthe control leaves (n = 10). The ANOVA showedboth the block (leaf-larva combination) and the treatmenteffects to be significant(F = 3.99,df = 9, 9, P = .026, and F = 13.76, df = 1, 9, P -- .005, respectively). Ectimarectifascia growth trials. -In thefirst growth trial, relative consumption rate(RCR) was significantlydepressed by the application of triterpene alcohol and mono/sesquiterpenemixtures from D. stipulaceafoliage (table 3). The ANOVA indicated significanttreatment (F = 36.90, df = 2, 18, P < .0001) and block effects(F = 3.20, df = 9, 18, P = .017). A multiplecomparison test showed thatthe mean RCRs of bothtriterpene alcohol and mono/sesquiterpenetreat- mentswere significantlylower than the mean of thecontrol (P = .0001,and P = .0002,respectively) but not different from each other(P = .44). The relative growthrate (RGR) showeda similardepression by the resintreatment. While ANOVA indicatedno significantblock (leaf-larva) effect (F = 1.17,df = 9, 18, EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 475

TABLE 3

MEAN VALUES (+ SE) IN mg/mg/dFOR RELATIVE CONSUMPTION RATE (RCR), RELATIVE GROWTH RATE (RGR), AND EFFICIENCY OF CONVERSION OF INGESTED MATTER (ECI) IN Two GROWTH TRIALS WITH FOURTH-INSTAR ECTJMA RECTIFASCIA ON DALECHAMPIA LEAVES TO WHICH VARIOUS COMPOUNDS WERE ADDED

COMPOUND ADDED

DALECHAMPIA DALECHAMPIA STIPULACEA FOLIAR ARMBRUSTERI FLORAL

Triterpene Mono/ Triterpene Triterpene Alcohol sesquiterpene Alcohol Ketone CONTROL

Trial 1: RCR - .002a 494a . .. . 3.19 (.466) (.389) (.212) RGR .239a .335a ...... 746' (.050) (.051) (.020) ECI . . . .678 ...... 234 Trial 2: RCR ... ..1l.72a 1.13a 4.49' (.378) (.548) (.446) RGR ...... 259a .187a .653' (.074) (.086) (.090) ECI ...... 150 .165 .145

NOTE.-Values with differentletters in each row are significantlydifferent from one another at P < .001 (all of trial 1 and trial2 RCR) and P < .05 (trial2 RGR) (Tukey's HSD multiplecomparison test).

P = .366),there was a highlysignificant treatment effect (F = 42.08, df = 2, 18,P < .0001).A multiplecomparison test showed that the mean RGRs of both triterpenealcohol and mono/sesquiterpenetreatments were significantlylower thanthe mean of the control (P = .0001,P = .0002,respectively) but not different fromeach other(P = .26). The decreasedRGR in thetreatments appears to be primarilycaused by RCR, becausethe mean efficiency of conversion of ingested material(ECI) was actuallyhigher in themono/sesquiterpene treatment than in thecontrol (table 3). In a secondgrowth trial with Ectima, using tritepene alcohols and ketones from blossomsof D. armbrusteri,we foundthat the mean RCR offourth-instar larvae was significantlylower on treatmentthan control leaves (table3). The ANOVA indicated a significanttreatment effect (F = 14.73, df = 2, 13, P = .0005) but no block effect(F = 1.55, df = 9, 13, P = .23). A multiplecomparison test showedthat the meanRCRs of bothresin alcohol and ketonetreatments were significantlylower than the mean of thecontrol (P = .003,P = .0006,respectively)but not differentthan each other(P = .53). The meanRGR showeda similarpattern. The ANOVA indicateda significanttreatment effect (F = 6.91, df = 2, 13, P = .009) but no block effect(F = 1.27, df = 9, 13, P = .34). A multiplecomparison test showed that the mean RGRs of bothresin alcohol and ketonetreatments were significantlylower than the mean of the control(P = .03, and P = .009, respectively)but not differentthan each other(P = .76). 476 THE AMERICAN NATURALIST

TABLE 4

ANOVA OF RELATIVE CONSUMPTION RATE (RCR) AND RELATIVE GROWTH RATE (RGR) FOR ALL HAMADRYAS GROWTH TRIALS

Sums Mean Source of Squares df Square F P

RCR: Treatment 58.20 4 14.55 24.65 <.0001 Family 21.15 2 10.57 17.91 <.0001 Treatmentx family 28.33 8 3.54 6.00 <.0001 Error 63.74 108 .032 RGR: Treatment 1.577 4 .394 12.25 <.0001 Family 2.251 2 1.125 34.95 <.0001 Treatmentx family .958 8 .120 3.72 .0007 Error 3.478 108 .032

NOTE.-Treatments are explained in table 5.

The decreasedRGR in thetreatments again appeared to be caused primarilyby decreasedRCR, because the mean ECIs were slightlyhigher in the triterpene alcoholand ketonetreatments than in thecontrol (table 3). Hamadryasamphinome growth trials.-Three growth trials were conducted on H. amphinome.In all threetrials larvae were fed leaves subjectedto the same combinationof treatments(app. A). The trialswere initially analyzed in a single ANOVA. Therewere significant treatment, family, and instareffects, as well as significanttreatment-family and treatment-instarinteractions for both RGR and RCR (table4). The significantinteraction terms indicated that each trialshould be analyzedand interpretedseparately. Fourth-instarH. amphinomelarvae were used in thefirst trial. ANOVA indicated a significanttreatment effect on RCR (F = 13.27, df = 4, 35, P < .0001) but no significantblock effect(F = 1.74, df = 9, 35, P = .12). A multiple comparisontest showed that means of foliarand floralresin alcohol treatments weresignificantly lower than means of the control. Means of the mono/sesquiterpene and floralresin ketone treatments were not significantly different from the control(table 5). A similarpattern was seen in RGR. ANOVA indicatedsignifi- cant treatmentand significantblock effects(F = 8.06, df = 4, 35, P = .0001; F = 4.59, df = 9, 35, P = .0005, respectively).A multiplecomparison test showedthat the means of foliarand floralresin alcohol treatments were significantlylower than the means of the control. The meansof the mono/sesquiterpene and floralresin ketone treatments, however, were not significantly different from thecontrol (table 5). The decreasedRGR inthe treatments appears to be primarily caused by decreasedRCR, because the mean ECIs were higherin the resin treatmentsthan in thecontrol (table 5). Third-instarH. amphinomelarvae were used in the second trial.ANOVA indicateda significanttreatment effect on theRCR (F = 13.97,df = 4, 29, P < .0001) but no block effect(F = 1.53, df = 5, 20, P = .22). A multiplecomparison testshowed that means of foliarresin treatments were significantlylower than EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 477

TABLE 5

MEAN VALUES (? SE) IN mg/mg/dFOR RELATIVE CONSUMPTION RATE (RCR), RELATIVE GROWTH RATE (RGR), AND EFFICIENCY OF CONVERSION OF INGESTED MATTER (ECI) IN THREE GROWTH TRIALS WITH THIRD- OR FOURTH-INSTAR HAMADRYAS AMPHINOME ON DALECHAMPIA LEAVES TO WHICH VARIOUS COMPOUNDS WERE ADDED

COMPOUND ADDED

DALECHAMPIA DALECHAMPIA STIPULACEA FOLIAR ARMBRUSTERI FLORAL

Triterpene Mono/ Triterpene Triterpene Alcohol sesquiterpene Alcohol Ketone CONTROL

Trial 1 (family 1, fourth instar): RCR .823a 2.1701 1.062a 2.3331 2.600' (.310) (.227) (.203) (.150) (.212) RGR .568a .706b .592a .784b .830b (.042) (.056) (.053) (.040) (.058) ECI .702 .325 .558 .336 .319 Trial 2 (family 2, third instar): RCR 1.086a 2.014al 3.842c 2.966& 4.236c (.533) (.542) (.144) (.176) (.190) RGR .687a .726a 1.073b .gg1b 1.1631 (.092) (.085) (.031) (.030) (.046) ECI .663 .360 .279 .334 .274 Trial 3 (family 3, fourth instar): RCR 1.579a 1.639a 1.709a 1.7OOa 3.2141 (.318) (.266) (.232) (.192) (.151) RGR .549a5 .533a .518a .422a .843' (.076) (.063) (.046) (.112) (.062) ECI .347 .325 .303 .248 .262

NOTE.-Values with different letters in each row are significantly different from one another at P < .001 (trial 3 RCR), P < .01 (trial 1 RCR and RGR, trial 2 RCR), and P < .05 (trial 2 RGR and trial 3 RGR) (Tukey's HSD multiple comparison test).

meansof the control and, with one exception,floral resin treatments. The means of the floralresin treatments were not significantlydifferent from the control (table 5). A similarpattern was seen in RGR. ANOVA indicateda significant treatmenteffect (F = 14.25,df = 4, 20,P < .0001)but no significantblock effect (F = 2.29,df = 5, 20, P = .08). A multiplecomparison test showed that means of foliarresin treatments were significantly lower than means of thecontrol and floralresin treatments. The meansof thefloral resin treatments were not significantlydifferent from the control (table 5). The decreasedRGR in thetreatments appearsto be primarilycaused by decreasedRCR because themean ECIs were higherin theresin treatments than in thecontrol (table 5). Fourth-instarH. amphinomelarvae were used in thethird trial. ANOVA indicated significanttreatment and blockeffects on RCR (F = 15.40,df = 4, 30, P < .0001;F = 3.60,df = 9, 30,P = .004,respectively). A multiplecomparison testshowed that the means of all resintreatments were significantly lower than 478 THE AMERICANNATURALIST themean of thecontrol. The effectsof thetreatments did notdiffer significantly fromone another(table 5). A similarpattern was seenin RGR. ANOVA indicated a significanttreatment effect (F = 4.70, df = 4, 30, P = .005) but no block effect (F = 0.86, df = 9, 30, P = .57). A multiplecomparison test showed that the meansof all resintreatments were significantly, or almost significantly (P = .059 forD. stipulaceatritepene alcohol), lower than the means of the control(table 5). The decreasedRGR in the treatmentsappears to be primarilycaused by decreasedRCR because themean ECIs werehigher in mostof the resintreat- mentsthan in thecontrol (table 5).

DISCUSSION

General Results Feedingchoice trials with the tettigoniid indicated that both floral and foliage resinsgenerally deterred feeding. Similarly, both floral and foliageresins de- creasedthe pickup of baitby leaf-cuttingants. The fieldleaf-cutting trials with Attashowed no negativeeffects of resin spread as a thinfilm on treatmentleaves. However,laboratory tests with resin placed in dropletson leaves, muchas it is deployedin nature,showed significant inhibition of Atta leaf cutting. All theeffects described above could be theresult of either biochemical activity or the mechanicaleffects of the stickyresin. The resinappeared, however, to have no biochemicalactivity (toxicity) in theAtta survivaltrials. The behavior of theants in thetwo leaf-cuttingtrials indicated likely mechanical deterrence. Thus,the resin treatments had predominantly negative effects on feedingby gen- eralistherbivores, and thereis a strongindication that the effects were primarily mechanical. Similareffects were seen on specialistDalechampia herbivores. Feeding by fleabeetles was significantlyinhibited by the floralresin deployed in droplets on Dalechampia tiliifolialeaves. Similarly,Hamadryas ipthimelarvae preferred controlover floral resin-treated leaves. Bothof these results could be caused by eithermechanical or biochemicalrepellent properties of theresin. Both specialists, Ectima rectifasciata and Hamadryas amphinome, showed reducedrelative growth rates (RGRs) whenfed Dalechampia leaves treatedwith floralor foliage resin. The effectmay have been mechanical rather than biochemical becausethe relative consumption rate was reducedas muchas theRGR and because the meanefficiency of conversionof ingestedmaterial was about the samefor insects fed treatment leaves and insectsfed control leaves.

Evolutionof Plant-InsectRelationships in Dalechampia Floralreward resins were biochemically similar to nonreward(defense?) resins and had inhibitoryeffects on severalgeneralist herbivores, including some of the mostcommon florivores on Dalechampia. These observations are consistentwith theidea thatthe resin originated as a floraldefense and secondarilyassumed a roleas a rewardfor pollinators. Primitive Dalechampia species still secrete resin fromfloral bracts, apparently for defense.(They are pollinatedby pollen- EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 479 collectingor fragrance-collectingbees, notresin-collecting bees; fig.1 [Armbrus- teret al. 1992;Armbruster 1993].) These resins are chemicallysimilar to thefloral rewardresins on whichwe conductedthe bioassays. In theseprimitive species, theresin is deployedon bractletsscattered throughout the staminate inflorescence (fig.2), and maleflowers in bud are coveredby resiniferousbractlets. In contrast,species thatuse resinas a pollinatorreward have evolvedresin deploymentin a localizedglobule that is easy forthe bees to collect.This has beenachieved by elaborationand aggregationof the resiniferous bractlets on the upperside ofthe staminate cymule and theirloss elsewhere(fig. 2). Thisform of deploymentappears to haveimproved the precision of match between flower and pollinator(Armbruster 1988, 1990), much as zygomorphy(bilateral symmetry) has overactinomorphy (radial symmetry) in otherplant species (Faegri and van derPijl 1971).In thesespecies the staminate flowers are notcovered by resiniferous bractletswhen in bud and wouldbe completelyexposed to attackby flori- voreswere it not for a "compensatory"defense system; involucral bracts in most of thesemore advanced species envelope both staminate and pistillateflowers at night(Armbruster and Mziray1987; Armbruster 1997). The findingthat foliage resins greatly inhibit feeding or leafcutting by specialist and generalistfolivores is consistentwith the hypothesis that resin was deployed in leaves and stipulesof a fewhighly advanced species of Dalechamnpia to reduce leafloss to thevoracious nymphalid butterfly larvae or leaf-cutter ants that proba- bly attackmost Dalechampia species. The nextstep is to conductfield experi- mentswith these plant species to see ifthey have escaped, to someextent, attack by specialistlepidopterans and leaf-cutterants.

CONCLUSIONS Thisstudy illustrates the value of generating explicit historical hypotheses from phylogeneticand ecologicaldata and thentesting these hypotheses with ecological experiments.This protocolyielded several unexpected insights. Chance and exaptationappear to have playedmajor roles in the evolutionof defenseand rewardsystems. Especially important were biochemical preaptations. The resinreward system in Dalechampiaappears to have originatedas a defenseand secondarilyassumed its role in attractingpollinators. The physical propertiesof thefloral defense resin mixture (liquid, sticky, and slow to harden) appearto have madeit an effectivedeterrent against herbivorous insects. At the same time,being liquid, viscous, waterproof, and slow to hardenallowed the resinmixture to be a valuablenest-building material (e.g., cementor sealant)for resin-collectingbees, whichcollect and use resinonly when liquid (Armbruster 1984;Roubik 1989). The evolutionof resin defense of leaves was also apparently facilitatedby the physical properties of the reward resin, the biochemical machin- eryfor which was alreadyin place in leafcells. Thusit appearsthat in Dalechampia, at least,defensive traits have influenced, by exaptation,the evolutionof plant-pollinatorrelationships. Pollination traits have similarlyinfluenced the evolution of plant-herbivorerelationships. 480 THE AMERICAN NATURALIST

ACKNOWLEDGMENTS We thankB. Bechyneand V. Savinifor determining alticid beetles, D. A. Nicklefor determining the tettigoniid, S. F. MacLean foradvice on growthtrials, and R. Staffordfor help in the lab. Laboratoryspace and logisticalassistance wereprovided by the Smithsonian Tropical Research Institute, and financial sup- portwas providedby theNational Science Foundation (DEB-9020265).

APPENDIX A

TABLE Al

SOURCES OF LEAVES AND RESINS USED IN BIOASSAYS ON DIFFERENT INSECT SPECIES

Bioassay Species and Type of Trial Leaf Source Resin Source Orophustesselatus: Feedingchoice Dalechampiaheteromorpha Dalechampiaarmbrusteri, floral alcohols and ketones; Dalecham- pia tiliifolia,crude floralmixture;Dalechampia stipulacea, vegetativealcohols Attacolombica: Wheat-flakepickup ... D. armbrusteri,floral alcohols and ketones; D. tiliifolia,crude floral mixture;Dalechampia dioscorei- foila, crude floralmixture; D. stipulacea, vegetativealcohols and mono/sesquiterpenemixture Leaf cutting D. tiliifolia,Viburnum sp. D. tiliifolia,crude floralmixture Toxicity ... D. tiliifolia,crude floralmixture; D. dioscoreifolia,crude floral mixture;D. stipulacea, vegetative mono/sesquiterpenemixture Syphraeasp. colombica: Feeding choice D. tiliifolia D. tiliifolia,crude floralmixture Ectimarectifasciata: Growthrate D. heteromorpha D. armbrusteri,floral alcohols and ketones; D. stipulacea, vegetative alcohols and mono/sesquiterpene mixture Hamadryasipthime: Feeding choice D. tilifolia Dalechampia scandens, crude floral mixture Hamadryasamphinome: Growthrate D. tiliifolia D. armbrusteri,floral alcohols and ketones; D. stipulacea, vegetative alcohols and mono/sesquiterpene mixture

NOTE.-All bioassay compounds are triterpenesunless otherwisenoted. EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 481

APPENDIX B FORMULAE USED FOR CALCULATION OF RELATIVE GROWTH RATE (RGR), RELATIVE CONSUMP- TION RATE (RCR), AND MEAN EFFICIENCY OF CONVERSION OF INGESTED MATTER (ECI) RGR = [ln(WfW)- ln(Wiw)]ITtot; RCR = C/(W *Ttt); and ECI = RGR/RCR, where WfW= finalwet mass; WiW= initialwet mass; Ttot total elapsed time; C = Lid - Lfd; Lfd = finalleaf dry mass; Lid = estimatedinitial leaf dry mass = (Cid * Liw)/Ciw; Cid = initialdry mass of the matched"control" (not used in feedingexperiment); Liw= initialwet mass of the experimentalleaf piece; C = wet mass of the "control" leaf piece; W = mean larval mass = (Wfd - Wid)/ln (Wfd/Wid); Wfd= finaldry mass of the larva; and Wid = initialdry mass of the larva.

LITERATURE CITED Armbruster,W. S. 1982.Seed productionand dispersalin Dalechampia(Euphorbiaceae): divergent patternsand ecologicalconsequences. American Journal of Botany69:1429-1440. . 1983.Dalechampia scandens. Pages 230-233 in D. H. Janzen,ed. CostaRican natural history. Universityof Chicago Press, Chicago. . 1984.The role of resinin angiospermpollination: ecological and chemicalconsiderations. AmericanJournal of Botany71:1149-1160. . 1986.Reproductive interactions between sympatric Dalechampia species: are naturalassem- blages"random" or organized?Ecology 67:522-533. . 1988.Hierarchical comparisons of the morphology, function, and evolutionof Dalechampia blossoms.Ecology 69:1746-1761. . 1990.Estimating and testing the shapes of adaptive surfaces: the morphology and pollination ofDalechampia blossoms. American Naturalist 135:14-31. 1992.Phylogeny and theevolution of plant-animal interactions. BioScience 42:12-20. 1993.Evolution of plant pollination systems: hypotheses and testswith the Neotropical vine Dalechampia.Evolution 47:1480-1505. 1994.Early evolution of Dalechampia (Euphorbiaceae): insight from phylogeny, biogeography,and comparativeecology. Annals of the Missouri Botanical Garden 81:302-316. . 1997.Exaptations link the evolution of plant-herbivoreand plant-pollinatorinteractions: a phylogeneticinquiry. Ecology (in press). Armbruster,W. S., andW. R. Mziray.1987. Pollination and herbivore ecology of an AfricanDalech- ampia(Euphorbiaceae): comparisons with New Worldspecies. Biotropica 19:64-73. Armbruster,W. S., and G. L. Webster.1979. Pollination of two speciesof Dalechampia(Euphor- biaceae)in Mexicoby euglossinebees. Biotropica11:278-283. Armbruster,W. S., A. L. Herzig,and T. P. Clausen. 1992.Pollination of two sympatricspecies of Dalechampiain Surinameby male euglossinebees. AmericanJournal of Botany79: 1374-1381. Arrhenius,S. P., and J. H. Langenheim.1983. Inhibitory effects of Hymenaeaand Copaiferaleaf resinson theleaf fungus, Pestalotia subcuticularis. Biochemical Systematics and Evolution 11:361-366. 482 THE AMERICAN NATURALIST

Ayres,M. P., and S. F. MacLean. 1987.Development of birch leaves and thegrowth energetics of Epirritaautumnata (Geometridae). Ecology 68:558-568. Boucher,D. H. 1985.The biologyof mutualism. Oxford University Press, New York. Boyd,N. D., and M. M. Martin.1975. Faecal proteinasesof thefungus-growing ant Attatexana: theirfungal origin and ecologicalsignificance. Journal of InsectPhysiology 21:1815-1820. Brooks,D. R., andD. H. McLennan.1991. Phylogeny, ecology, and behavior. University of Chicago Press,Chicago. Bryant,J. P., F. D. Provenza,J. Pastor,P. B. Reichardt,T. P. Clausen,and J. T. du Toit. 1991. Interactionsbetween woody plants and browsing mammals mediated by secondarymetabo- lites.Annual Review of Ecologyand Systematics22:431-436. Bryant,J. P., P. B. Reichardt,and T. P. Clausen.1992. Chemically mediated interactions between woodyplants and browsingmammals. Journal of Range Management 45:18-24. Cane, J. H., S. Gerdin,and G. Wife.1983. Mandibular gland secretions of solitarybees (Hymenop- tera:Apoidea): potential for cell disinfection.Journal of the Kansas EntomologicalSociety 56:199-204. Chen,T., A. C. Ales, N. C. Baenzitten,and D. F. Wiemer.1983. Ant-repellent terpenoids from Cordiaalliodora. Journal of OrganicChemistry 48:3525. Clausen,T. P., J. P. Bryant,and P. B. Reichardt.1986. Pinosylvin and pinosylvinmethyl ether as feedingdeterrents in greenalder. Journal of Chemical Ecology 12:2117-2131. DeVries,P. J. 1987.The butterfliesof Costa Rica and theirnatural history. Princeton University Press,Princeton, N.J. Dodson,C. H., R. L. Dressler,H. G. Hills,R. M. Adams,and N. H. Williams.1969. Biologically activecompounds in orchidfragrances. Science (Washington, D.C.) 164:1243-1249. Donoghue,M. J. 1983.A preliminaryanalysis of phylogenetic relationships in Viburnum(Caprifolia- ceae, S.L.). SystematicBotany 8:45-58. 1989.Phylogenies and the analysisof evolutionarysequences, with examples from seed plants.Evolution 43:1137-1156. Faegri,K., andL. vander Pijl. 1971.The principlesof pollination ecology. 2d ed. Pergamon,Oxford. Farentinos,R. C., P. J. Capretta,R. E. Kepner,and V. M. Littlefield.1981. Selective herbivory in tassel-earedsquirrels: role of monoterpenesin ponderosapines chosen as feedingtrees. Science(Washington, D.C.) 213:1273-1275. Farrell,B. D., and C. Mitter.1994. Adaptive radiation in insectsand plants:time and opportunity. AmericanZoologist 34:57-70. Farrell,B. D., C. Mitter,and D. J. Futuyma.1992. Diversification at the insect-plantinterface. BioScience42:34-42. Feibert,E. B., and J. H. Langenheim.1988. Leaf resinvariation in Copaiferalangsdorfii: relation to irradianceand herbivory.Phytochemistry 8:2527-2532. Felsenstein,J. 1985.Confidence limits on phylogenies:an approachusing the bootstrap. Evolution 39:783-791. Futuyma,D. J. 1983.Evolutionary interactions among herbivorous insects and plants. Pages 207-231 in D. J. Futuymaand M. Slatkin,eds. Coevolution.Sinauer, Sunderland, Mass. Futuyma,D. J.,and M. Slatkin,eds. 1983.Coevolution. Sinauer, Sunderland, Mass. Gilbert,L. E. 1980.Food weborganization and the conservation of Neotropical diversity. Pages 11-33 in M. E. Soule and B. A. Wilcox,eds. Conservationbiology: an evolutionary-ecological perspective.Sinauer, Sunderland, Mass. Gilbert,L. E., andP. H. Raven.1975. Coevolution of animals and plants. University of Texas Press, Austin. Gould,S. J.,and E. S. Vrba.1982. Exaptation-a missing term in thescience of form. Paleobiology 8:4-15. Harvey,P. H., and M. D. Pagel. 1991.The comparativemethod in evolutionarybiology. Oxford UniversityPress, Oxford. Howard,J. J. 1985.Observations on resincollecting by six interactingspecies of stinglessbees (Apidae:Meliponinae). Journal of the Kansas EntomologicalSociety 58:337-345. 1987.Leafcutting ant diet selection: the role of nutrients,water, and secondarychemistry. Ecology68:503-515. EVOLUTION OF PLANT DEFENSE AND REWARD SYSTEMS 483

Howard,J. J.,J. Cazin,and D. F. Wiemer.1988. Toxicity of terpenoid deterrents to theleafcutting antAtta cephalotes and itsmutualistic fungus. Journal of ChemicalEcology 14:59-69. Hubbell,S. P., J. J. Howard,and D. F. Wiemer.1984. Chemical leaf repellency to an attineant: seasonaldistribution among potential host plant species. Ecology 65:1067-1076. Jayasingh,D. B., and B. E. Freeman.1980. The comparativepopulation dynamics of eightsolitary bees and wasps (Aculeata:Apocrita: Hymenoptera) trap-nested in Jamaica.Biotropica 12: 214-219. Jenkins,D. W. 1983.Neotropical Nymphalidae. I. Revisionof Hamadryas.Bulletin of the Allyn Museum81:1-146. 1984.Neotropical Nymphalidae. II. Revisionof Myscelia. Bulletin of the Allyn Museum 87:1-64. . 1985a.Neotropical Nymphalidae. III. Revisionof Catonephele. Bulletin of the Allyn Museum 92: 1-65. . 1985b.Neotropical Nymphalidae. IV. Revisionof Ectima. Bulletin of the Allyn Museum 95: 1-30. Langenheim,J. H. 1969.Amber: a botanicalinquiry. Science (Washington, D.C.) 163:1157-1169. Langenheim,J. H., and G. D. Hall. 1983.Sesquiterpene deterrence of a leaf-tyinglepidopteran, Stenomaferrocanella, on Hymenaea stigonocarpa in centralBrazil. Biochemical Systemat- ics and Ecology11:29-36. Langenheim,J. H., andW. H. Stubblebine.1983. Variation in leaf resin composition between parent treeand progeny in Hymenaea: implications for herbivory inthe humid tropics. Biochemical Systematicsand Ecology11:97-106. Langenheim,J. H., C. L. Conovis,C. A. Macedo,and W. H. Stubblebine.1986. Hymenaea and Copaiferaleaf sesquiterpenes in relationto lepidopteranherbivory in southeasternBrazil. BiochemicalSystematics and Ecology14:41-49. Macedo,C. A., andJ. H. Langenheim.1989. Microlepidopteran herbivory inrelation to leaf sesquiter- penesin Copaifera langsdotfii adult trees and their seedling progeny in a Brazilianwoodland. BiochemicalSystematics and Ecology17:217-224. Maddison,W. P., and D. R. Maddison.1992. MacClade. Version 3. Sinauer,Sunderland, Mass. Mattes,B. R., T. P. Clausen,and P. B. Reichardt.1987. Volatile constituents of balsam poplar: the phenolglycoside connection. Phytochemistry 26:1361-1366. Messer,A. C. 1985.Fresh dipterocarp resins gathered by megachilidbees inhibitgrowth of pollen- associatedfungi. Biotropica 17:175-176. Michener,C. D. 1974.The socialbehavior of bees. Belknap,Cambridge, Mass. . 1979.Biogeography of the bees. Annalsof the Missouri Botanical Garden 66:277-347. Mitter,C., B. Farrell,and B. Wiegman.1988. The phylogenetic study of adaptive zones: has phytoph- agypromoted insect diversification? American Naturalist 132:107-128. Neter,J., W. Wasserman,and M. H. Kutner.1985. Applied linear statistical models. Irwin, Home- wood,Ill. Otero,L. D. 1988.Contribuci6n a la historianatural del generoHamadryas (Lepidoptera: Nymphali- dae). Ph.D. diss. UniversidadCentral de Venezuela,Maracay. Pax, F., and K. Hoffmann.1919. Euphorbiaceae-Dalechampieae. Das PflanzenreichIV. 147. XII (Heft 68):1-59. Pellmyr,O., and C. J. Huth.1994. Evolutionary stability of mutualismbetween yuccas and yucca moths.Nature (London) 372:257-260. Pellmyr,O., and L. B. Thien.1986. Insect reproduction and floralfragrances: key to theevolution of theangiosperms. Taxon 35:76-85. Pellmyr,O., andJ. N. Thompson.1992. Multiple occurrences of mutualism in the yucca moth lineage. Proceedingsof the National Academy of Sciencesof the USA 89:2927-2929. Pellmyr,O., J. Leebens-Mack,and C. J. Huth.1996. Non-mutualistic yucca moths and theirevolu- tionaryconsequences. Nature (London) 380:155-156. Reichardt,P. B., J. P. Bryant,T. P. Clausen,and G. D. Wieland.1984. Defense of winter-dormant Alaskapaper birch against snowshoe hares. Oecologia (Berlin) 65:58-69. Reichardt,P. B., J. P. Bryant,B. R. Mattes,T. P. Clausen,and F. S. ChapinIII. 1990.The winter chemicaldefense of Alaskanbalsam poplar against snowshoe hares. Journal of Chemical Ecology16:1941-1959. 484 THE AMERICAN NATURALIST

Rosenthal,G. A., and M. R. Berenbaum.1992. Herbivores: their interactions with secondary plant metabolites.Vol. 2. Ecologicaland evolutionaryprocesses. 2d ed. AcademicPress, San Diego,Calif. Roubik,D. W. 1983.Nest and colonycharacteristics of stinglessbees fromPanama (Hymenoptera: Apidae).Journal of the Kansas EntomologicalSociety 56:327-355. 1989.Ecology and naturalhistory of tropical bees. CambridgeUniversity Press, New York. Schemske,D. W. 1983.Limits to specializationand coevolutionin plant-animalmutualisms. Pages 67-109in M. H. Nitecki,ed. Coevolution.University of Chicago Press, Chicago. Simpson,B. B., and J. Neff.1981. Floral rewards: alternatives to pollenand nectar.Annals of the MissouriBotanical Garden 68:301-322. Skutch,A. F. 1971.A naturalistin Costa Rica. Universityof FloridaPress, Gainesville. Sokal,R. R., and F. J. Rohlf.1981. Biometry. 2d ed. W. H. Freeman,San Francisco. Still,W. C., M. Kahn,and A. Mitra.1978. Rapid chromatographic technique for preparative separa- tionswith moderate resolution. Journal of OrganicChemistry 43:2923-2925. Strong,D. R., J. H. Lawton,and R. Southwood.1984. Insects on plants.Harvard University Press, Cambridge,Mass. Stubblebine,W. H., and J. H. Langenheim.1977. Effects of Hymenaeacourbaril leaf resin on the generalistherbivore Spodoptera exigua (beet armyworm). Journal of ChemicalEcology 3: 633-647. Swofford,D. L. 1993.PAUP: phylogeneticanalysis using parsimony. Version 3.1.1. IllinoisNatural HistorySurvey, Champaign. Thompson,J. N. 1982.Interaction and coevolution.Wiley, New York. . 1994.The coevolutionaryprocess. University of ChicagoPress, Chicago. Webster,G. L., and W. S. Armbruster.1991. A synopsisof the Neotropical species of Dalechampia (Euphorbiaceae).Botanical Journal of the Linnean Society 105:137-177. Webster,G. L., and B. D. Webster.1972. The morphologyand relationshipsof Dalechampia scandens(Euphorbiaceae). American Journal of Botany 59:573-586. Wilkinson,L. 1988.SYSTAT: thesystem for statistics. SYSTAT, Inc., Evanston,Ill.

Associate Editor: Nancy A. Moran