Research Insect-mediated reproduction of systemic infections by Puccinia arrhenatheri on Berberis vulgaris Blackwell Science, Ltd Andreas Naef1, Bitty A. Roy2, Roman Kaiser3 and Rosmarie Honegger4 1 Plant Pathology group, Institute of Plant Sciences, Swiss Federal Institute of Technology (ETH) Zürich, CH-8092 Zürich, Switzerland; 2Biology Department, 1210 University of Oregon, Eugene, Oregon 97403 –1210, USA; 3Givaudan Dübendorf Ltd, Fragrance Research, CH-8600 Dübendorf, Switzerland; 4Institute of Plant Biology, University of Zürich, Zollikerstr. 107, CH-8008 Zürich, Switzerland Summary Author for correspondence: Andreas Naef Tel: +41 (1) 632 38 60 Fax: +41 (1) 632 11 08 Email: andreas.naef@ipw.agrl.ethz.ch Received: 26 October 2001 Accepted: 21 January 2002 • Witches’ brooms on Berberis vulgaris are induced by a systemically infecting rust fungus, Puccinia arrhenatheri. These witches’ brooms bear yellow discolored leaves on which the fungus exposes its gametes in a sugary nectar. During the spermatial stage of the fungus the infected leaves emit a strong, flowery scent. • An exclusion-experiment was used to evaluate whether fungal reproductive success, defined by the ability of the fungus to produce aeciospores, depended on gamete transfer by insects. To determine whether insects were attracted to the infected leaves, and if so, why, visitation to infected and uninfected leaves was quantified and volatiles produced by leaves, infected leaves and flowers were analyzed. • The production of aeciospores was significantly higher on witches’ brooms with insect visitation. Visitation rates were higher and visits were longer on witches’ brooms than on uninfected branches. A wide diversity of visitors, mainly Diptera and Hymenoptera, was observed. The volatiles emitted by infected leaves were composed of sweet floral fragrances and insect pheromones. • Our results suggest that sexual reproduction of the pathogen requires out-crossing by insects and that infected leaves attract insects by floral mimicry (bright yellow color and the production of sugary nectar and volatiles). Key words: carvacryl methyl ether, fragrances, indole, jasmin lactone, methyl nicotinate, rust fungi, scanning electron microscopy, spermatia. © New Phytologist (2002) 154: 717– 730 Introduction Berberis vulgaris L., or common Barberry, is a southern European-west Asian species that can grow into a 3 m high shrub. Barberries are most famous for being the alternate host of wheat rust, and for this reason, have been the subjects of eradication efforts for the better part of a century in many countries (Agrios, 1988). Despite the fact that the pathogen can live without the alternate host, the eradication of barberry continues in areas with intensive wheat production. The areas of the alpine belt where cattle-raising predominates provide a refuge for this interesting shrub and another specialist fungus, the witches’ broom forming rust Puccinia arrhenatheri (Kleb.) Erikss. (Basidiomycetes). During the spring, these witches’ brooms are highly attractive to insects. Here we present the © New Phytologist (2002) 154: 717– 730 www.newphytologist.com results of several studies designed to elucidate the natural history of systemic infections of B. vulgaris by P. arrhenatheri and the nature of its interactions with insects. In the early spring, the leaves on witches’ brooms of barberry have sexual structures of the rust fungus on their upper and lower surface (Fig. 1c–d). The infected leaves appear earlier than those on the uninfected part of the bush (Fig. 1a); they are yellow in color (Fig. 1a,c–d) and emit, during a short period of time, a very strong, sweet-smelling scent. In the fungal sexual structures, termed spermogonia, the gametes (spermatia) are produced and released into the sugary nectar that fills the cavity and oozes out of the narrow ostiole (Fig. 2a–b) (Rathay, 1883; Buller, 1950). The minute spermatia most likely need to be transferred from one self-incompatible mating type to another to fertilize the fungus. The sugary 717 718 Research www.newphytologist.com © New Phytologist (2002) 154: 717–730 Research Fig. 2 Scanning electron microscopy of rust spermogonia in systemically infected leaves of Berberis vulgaris (a,b) and of the hind leg of the visiting insect, Scatophaga stercoraria (c–e). (a) Ostiole of a spermogonium of Puccinia arrhenateri with adhering spermatia protruding above the leaf surface (conventionally fixed specimen). (b) Low-temperature scanning electron microscopy (LTSEM) of a cross-fractured spermogonium, with spermatia being released into the mucilage-filled cavity and oozing out between the hair-like hyphae that form the ostiole. (c) Tip of the hind leg of S. stercoraria, with hairy arolium between the tarsal claws. (d) Detail (as marked with asterisk in (c) of the basal part of the arolium, with masses of spermatia of the rust fungus (shriveled due to air drying) sticking between the hairs. (e) Air-dried (and thus shriveled) spermatia adhering to a stiff lateral hair of a tarsal segment. (All micrographs by Rosmarie Honegger) nectar and the volatiles produced by infected leaves might be attractive to insects that may aid in spermatial transfer, very much like pollinators do for flowers. The rust fungus causing the witches’ brooms is macrocy- clic, meaning that it proceeds through five spore stages on two different hosts. After successful outcrossing (dikaryotisation) the fungus produces dikaryotic aeciospores in aecia on the bottom side of infected leaves (Fig. 1f–g). The aeciospores are Fig. 1 Systemic rust infections of Berberis vulgaris. (a) Witches’ broom in spring. (b) Insect exclusion experiment on a Berberis bush installed in spring. (c) Healthy and systemically infected leaves of barberry. (d) Detail of infected leaf surface at the spermatial stage. (e) Lauxania cylindricornis licking an infected leaf. (f) and (g) Witches’ brooms with aecia-bearing leaves (a), (b), (e), (f): photographs by Andreas Naef; (c), (d), (g): photographs by Rosmarie Honegger. © New Phytologist (2002) 154: 717– 730 www.newphytologist.com 719 720 Research distributed by wind to infect the primary host. Several authors have found the alternate host to be a grass species of the genus Arrhenatherum (Urban & Marková, 1994). On the grasses, dikaryotic urediniospores are formed that infect other grass plants. Towards the end of the growing season over-wintering, thick-walled, dikaryotic teliospores are produced. In the spring, the nuclear pairs of teliospores fuse to form diploid nuclei, which immediately undergo meiosis. Subsequently teliospores germinate to form a septate basidium that produces monokaryotic, haploid basidiospores. The basidiospores, upon distribution by wind, infect young leaves of the alternate host, B. vulgaris, where they give rise to haploid hyphae of either one of the two mating types. These grow into the meristem of the lateral branch and thus cause abnormal growth (Fig. 1a,f ) (Gäumann, 1959). After overwintering in the bud meristem the fungus develops within emerging barberry leaves, generating the above mentioned yellowish discoloration, and forms large numbers of spermogonia (Fig. 1c –d). The first experiment we performed was designed to test the role of insects in the fertilization of this fungus. While insect transmission of infectious spores is relatively common (Cartar & Abrahams, 2000; Wilding et al., 1989; Batra, 1991; Antonovics & Alexander, 1992), insect-mediated fertilization outside the seed plants has not often been described. A few studies have shown that insects are required as gamete vectors for some rust fungi (Buller, 1950; Craigie, 1927; Roy, 1993; Pfunder & Roy, 2000; Schürch et al., 2000), for ergot, caused by Claviceps purpurea and other Claviceps spp., all with attractive ‘honey dew’ stage on cereals and grasses (Agrios, 1988), and for some closely related clavicipaceous ascomycetes that cause choke on grasses (e.g. Bultman & White, 1988). Since the witches’ brooms on B. vulgaris are caused by a systemic infection, all leaves on a witches’ broom are likely to be infected by the same fungal mating type. Therefore, spermatia may need to be transferred from another witches’ broom to fertilize the fungus. To our knowledge no fertilization experiment has ever been done with the systemic infection on B. vulgaris to test whether insects are required as gamete vectors. In previous years numerous ants were observed on witches’ brooms (B. A. Roy, personal observation). They seemed to be attracted by the sugary nectar on the leaves. Other authors have mentioned that ants use the sugary nectar of rust fungi as a food source (Buller, 1950; Schürch et al., 2000). We used an insect exclusion experiment (Fig. 1b) and quantified insect visitation to address the following questions. First does the sexual reproduction of the fungus depend on insect visitation? Second, what role do flying insects vs crawling insects, such as ants, play in the reproduction of the fungus? We also performed timed observations of insects on infected and uninfected branches. These observations allowed us to answer the following questions. What kinds of insects visit infected leaves during the spermatial stage of the fungus? Do more insects visit infected leaves than uninfected leaves and do they stay longer? Finally, we trapped the volatiles emitted by infected and uninfected leaves as well as the flowers. The compounds were then identified with a gas chromatograph and mass spectrometry (GC/MS). We used these data to address the following questions: What are the compounds emitted by the infected leaves when the pathogen is sexually receptive? Are the pathogen volatiles similar to those produced by the host flowers? Are the volatiles attractive to insects? Materials and Methods Host and pathogen In Switzerland, Berberis vulgaris grows in sunny and dry sites along forest edges and in scrublands. It prefers calcium rich soils (Hess et al., 1977). Robust fruiting individuals occur from elevations of 300 m to 2000 m, and nonfruiting individuals have been found in the upper Engadin valley even as high as 2660 m above sea level (Hegi, 1974). B. vulgaris shrubs with witches’ broom formation were studied at the following locations in Switzerland: Ardez in the lower Engadin (46°46′25′′ N, 10°12′16′′ W). The bushes were spread over an area of about 15 000 m2 in dry calcareous grassland at 1410 m to 1460 m above sea level oriented at a 20° angle and facing 190° south; in grassland near Trimmis, north of Chur in the Rhine valley (46°53′21′′ N, 9°33′19′′ W, 700 m above sea level); in Felsberg south of Chur in the Rhine valley (46°51′6′′ N, 9°29′22′′ W, 565 m above sea level); in Piotta, in the southern alpine valley Leventina (46°30′52′′ N, 8°40′12′′ W, 1050 m above sea level); in Kippel in the western alpine valley Lötschental (46°23′47′′ N, 7°45′46′′ W, 1400 m above sea level); and in the botanical gardens of the University of Zürich in the city of Zürich (47°21′39′′ N, 8°33′36′′ W, 435 m above sea level). Conspicuous witches’ brooms on B. vulgaris, induced by a systemic infection by Puccinia arrhenatheri are relatively common in the Swiss montane and subalpine areas (Gäumann, 1959). The rust fungus was named P. arrhenatheri according to its ability to infect Arrhenatherum species (Gäumann, 1959). In 1966 Cummins reduced the number of species of Barberry-infecting rust fungi because of morphological variability in the telial stage and consequently named the same fungus Puccinia brachypodii f. sp. arrhenatheri Kleb. (see Table 1 for synonyms of this rust fungus). The full host range was summarized by Gäumann (1959). Insect exclusion experiment To test whether insects fertilize the fungus, we conducted an insect exclusion experiment during spring 2000. At the Ardez site, we chose 25 bushes of B. vulgaris, each of which had at least four witches’ brooms. The four witches’ brooms on each www.newphytologist.com © New Phytologist (2002) 154: 717–730 Research Table 1 Synonyms for systemic rust infections on Berberis vulgaris used in cited literature (the name used in this study is given in bold face) Name of grass rust species First and other authors Puccinia arrhenatheri Syn. Aecidium Maelhaenicum Syn. Puccinia magelhaenica Eriksson (1898), Gäumann (1959) Magnus (1875) Peyritsch in (Magnus, 1894), Urban & Marková (1994) Cummins & Greene (1966) Syn. Puccinia brachypodii f. sp. arrhenatheri bush were randomly assigned to one of the following treatments: (1) no visitors – the witches’ broom was caged with small-meshed florist gauze net (Kleen Test Products, division of Meridian Industries, Inc., Milwaukee, WI, USA). The bags allow light, wind and water to penetrate but exclude even the smallest insects. To prevent the cage from collapsing on to the infected leaves, we put a frame of chicken wire inside the cages (Fig. 1b). (2) crawling insects only – we used the cages described above but left an opening of about 1 cm around the stem to allow insects to crawl in. (3) flying insects only – to prevent crawling insects from visiting we put a 5-cm wide tape around the stem of the witches’ broom and smeared it with an insect trapping adhesive (Tangle-trap from The Tanglefoot Company, Grand Rapids, MI, USA). (4) all visitors – uncaged control with natural insect visitation. All treatments were installed from 24 March 2000–10 April 2000 before buds on the witches’ brooms opened (Fig. 1a). For the next month the site was visited once a week to record the stage of the leaves and to check whether the treatments excluded the proper insects. Replicates that allowed improper visitors were not included in the analysis. In addition, we lost some treated witches’ brooms because they were broken by grazing sheep or appeared to be dead. Not all leaves on the witches’ brooms showed signs of infection. The first leaves appeared completely yellow and covered with spermogonia (Fig. 1c–e), the organs producing the fungal gametes (spermatia), whereas older leaves were yellowish and bore spermogonia only along the veins or were apparently healthy. This phenomenon was also observed by Gäumann (1959). To quantify fungal reproductive success, all infected leaves were harvested from the treated witches’ brooms on 17 and 18 June 2000 when they started to dry up. The reproductive success of the pathogen was defined by the fungus’ ability to produce aeciospores as a result of successful dikaryotisation. Aeciospores are heterokaryotic spores that are distributed by wind to the grass host. We based our analysis on the proportion of aecia-bearing leaves from the total of infected leaves. To determine the effect of our treatments, this proportion was analyzed with a factorial analysis of variance (ANOVA), without replication, with treatments as a fixed and bushes as a random factor. From our total of 100 treated witches’ brooms, 80 were included in the analysis. Before analysis, the proportion was arcsine square root transformed to obtain normality and homogeneity of variance. The statistical analyses were performed with the program JMP, Version 3.2.2 (SAS, 1994). © New Phytologist (2002) 154: 717– 730 www.newphytologist.com Scanning electron microscopy SEM techniques were used to explore the dimension of spermatia and their location on visiting insects. Conventional preparative procedures included chemical fixation of infected leaf fragments in the vapor of a 4% solution of osmium tetroxide, followed by dehydration in a graded series of acetone and critical point drying. Arthropod visitors of witches’ brooms were air dried. Both types of specimen were mounted with conductive silver print paint and sputtercoated with gold before examination in a Cambridge Stereoscan scanning electron microscope at 20 kV. For lowtemperature scanning electron microscopy (LTSEM) studies infected leaf fragments were mounted with commercially available white glue on the specimen holder of a Bio-Rad SP 2000 A cryotrans system, then cryoimmobilized by slushfreezing in subcooled liquid nitrogen (LN2). Frozen-hydrated samples were freeze-fractured, sputter-coated with an alloy of gold and palladium and examined on the stage of a Hitachi Stereoscan field emission microscope at 20 kV. Insect observations To examine the attractiveness of the witches’ brooms to insects during the stage when infected leaves emit a sweet-smelling scent, we recorded insect visitation for 20 min observation periods on a witches’ broom and simultaneously on leaves on a similar sized healthy branch of B. vulgaris. All observations were made by the same observer at four sites in alpine valleys in Switzerland. The four sites will be referred to as Trimmis, Piotta, Ardez and Kippel. In Trimmis we tallied the visits on 26 April 2000 during four observation periods of 20 min, each made on a different bush. In Piotta we also counted visits during four observation periods of 20 min made on four different bushes on 1 May 2000. At the two other sites, in Ardez, at the same site as the exclusion experiment, and in Kippel, we made six replicates of 20 min observation periods on three randomly chosen bushes per site. Each set of observations on the three bushes was made within a time block of about 70 min (20 min × 3 plus about 5 min time to move between bushes). To remove bias due to observing in chronological order, we systemically changed the order of the observed bushes after each time block. In Ardez, we observed for four time blocks on 3 May 2000, and two time blocks on 4 May 2000, and at the same time of day in Kippel for four time blocks on 9 May 2000, and two time 721 722 Research blocks on 10 May 2000. All observations were made between 1030 h and 1530 h under clear to partly cloudy skies. In addition to the number of visits, we measured the duration of each visit. The insect or spider visitors were identified as well as possible during the observations and identifications were verified after the experiment by catching representatives, which were identified by B. Merz (Diptera), N. Neumeyer (ants) or A. Naef (other insects and spiders). The insect observations in Ardez and Kippel were analyzed by mixed-model analyses of variance (ANOVA) on the number of visits per observation period, and the mean duration per visit. Observation site, time block (replicates) and bush were included as random factors with bush nested within site, and the kind of branch (systemically infected vs uninfected) was included as a fixed factor. Nonsignificant interactions were removed from the models one by one starting with the highest order interaction as suggested by Underwood (1981). The interaction between time blocks and branch types was included into the final model for duration per visits because it was significant in some models (F5,295 = 2.288, P = 0.046). The number of visits and the duration per visit were natural log transformed to meet the normality and homogeneity assumptions of ANOVA. Synthetic denominators, as calculated in JMP, Version 3.2.2 (SAS, 1994) were used to calculate the F-ratios. Fragrance collection and analysis Five infected branches, two uninfected branches and one flowering branch were collected and their cut ends kept in flasks with water in them for transport back to the laboratory. Headspace collection was performed as described in Kaiser (1993) and Kaiser & Tollsten (1995). Briefly, a clean glass flask that had been baked at 250°C to remove contaminating volatiles was placed over the stem, a battery-operated portable pump (SKC personal air sampler) was attached to a Teflon line holding at its end a fragrance trap (a small glass tube filled with 3 mg Poropak). Fragrance-laden air was drawn over the trap for 6 h. Volatiles were recovered by pressing 40 µl of hexane:acetone (5 : 1) through the trap. The eluate was analyzed directly by injecting 1.5 µl into the GC (Carlo Erba Fractovap 4160) or GC-MS (Carlo-Erba Mega 5160 coupled to a Finnigan MAT 212 instrument with INCOS computer system). The analyses were made on a DB-WAX column ( J & W Scientific) 30 m × 0.32 mm i.d., film thickness 0.25 µm; splitless injection; temperature program: 40°C-3′iso-2.5°C/ min-230°C. Compounds were identified by comparison of their mass spectra and retention times with authenticated reference samples. Results Insect exclusion experiment The insect exclusion treatments significantly affected aeciospore production on witches’ brooms of Berberis vulgaris (ANOVA F3,52 = 87.06, P < 0.0001; Table 2). Figure 3 shows the mean proportion of aecia-bearing leaves from the total of infected leaves on treated witches’ brooms (11% without insect visitation, 39% for crawling insects only, 79% for flying Table 2 Factorial ANOVA on the proportion of aecia-bearing leaves from the total of infected leaves on differently treated witches’ brooms of Berberis vulgaris in an insect exclusion experiment in Ardez (data arcsine-transformed). P-values ≤ 0.05 are given in bold face Source of variation DF MS F P Bush Treatment Error 24 3 52 0.069 3.038 0.035 1.974 87.061 0.0205 < 0.0001 Fig. 3 Mean proportion of aecia-bearing leaves from the total of infected leaves (± SE) on differently treated witches’ brooms of Berberis vulgaris in an insect exclusion experiment. Treatments with different letters differ significantly at P ≤ 0.05. www.newphytologist.com © New Phytologist (2002) 154: 717–730 Research Table 3 Number of visitors observed in Trimmis and Piotta during 4 observation periods of 20 min on uninfected and systemically infected branches (witches’brooms) of Berberis vulgaris Trimmis Class/Order Arachnida Araneae Insecta Coleoptera Diptera Witches’ broom Uninfected branch Witches’ broom Family Species Araneidae indet. 1 1 indet. Sciaridae Sepsidae Syrphidae Drosophilidae indet. indet. Sepsis sp. Platycheirus albimanus Müller Drosophila cameraria Halliday indet. Agromyza sp. Eudasyphora zimi Hennig Helina reversio Harris Azelia sp. indet. indet. Formica pratensis Retzius Formica sp. indet. 1 1 6 Agromyzidae Muscidae Hymenoptera Uninfected branch Piotta Lauxanidae Formicidae Ichneumonidae Total of visits insects only, and 85% for natural visitation). A Tukey test at P < 0.05 showed that any insect visitation significantly increased aeciospore production. Compared with natural insect visitation, the exclusion of flying insects significantly reduced the proportion of aecia-bearing leaves, whereas the exclusion of crawling insects did not differ significantly from natural insect visitation. Not surprisingly, there was significant variation in the proportion of aecia produced by individual bushes (ANOVA F24,52 = 1.97, P = 0.0205; Table 2), suggesting both genetic and environmental influences on aecia formation. Scanning electron microscopy The hind leg of a Scatophaga stercoraria Linné, which was captured during its visit of a witches’ broom, was ultrastructurally examined (Fig. 2c–e). Masses of spermatia were found sticking between the fine hairs of the arolium (the median, pad-like lobe that lies between the tarsal claws of a pretarsus; Fig. 2c). Spermatia adhered also to the long, stiff, laterally protruding hairs (Fig. 2e), which easily come into contact with the spermatia-covered ostiole of the innumerable spermogonia on the leaf surfaces. Insect observations During a total observation time of 80 min per site, 27 visitors were observed in Piotta (Table 3) and 30 visitors in Trimmis (Table 3). In both sites, more than 85% of the visits were observed on witches’ brooms. The main visitors were insects © New Phytologist (2002) 154: 717– 730 www.newphytologist.com 1 1 1 3 1 1 3 11 2 2 2 2 1 13 2 1 3 24 3 27 of the classes Diptera and Hymenoptera. The ants counted in Trimmis were all tallied during one observation period of 20 min. They were densely patrolling on a witches’ broom, but no ants were present on the other observed witches’ brooms at this site. In addition to the insects at each site, one web spider was observed. At the Ardez and Kippel localities we made more extensive observations. In Ardez 148 visitors were counted during 360 min split into 18 observation periods of 20 min each. 81% of the visits were counted on witches’ brooms. Table 4 shows the high diversity of visitors at this site. Diptera from 11 different families accounted for 81% of all visits, Hymenoptera (mostly Formicidae) for 7%, and spiders (mostly Salticidae) for 10%. The remaining 2% of visitors belonged to the insect orders Coleoptera, Heteroptera and Homoptera. Table 4 also shows the mean time spent per visit for different species. Since the sample size for single species was small, the difference in duration per visit between infected and uninfected branches was significant only for the unidentified anthomyid flies (t-test, P < 0.05). Many visitors were observed licking the fungal nectar on the upper side of infected leaves (Fig. 1e) and moving within the witches’ broom. Such movements within a branch, either infected or uninfected, were considered to be one visit. By contrast to this behavior, the sarcophagus Scatophagidae remained immobile up to the whole 20 min observation time unless a prey approached. Visiting insects were hunted by spiders (Salticus scenicus Clerck), which jumped back and forth between the infected and uninfected branches; such changes between branches were counted as separate visits. In 723 724 Research Table 4 Visitors observed in Ardez during 18 observation periods of 20 min on uninfected and systemically infected branches (witches’ brooms) of Berberis vulgaris: Number of visits and mean time spent per visit in seconds (grand means ± SE) Class/Order Arachnida Araneida Insecta Coleoptera Diptera Heteroptera Homoptera Hymenoptera Uninfected branch Witches’ broom Number Number Family Species Araneidae Salticidae indet. Salticus scenicus Clerck Coccinellidae Nitidulidae indet. indet. Epurea depressa Ill. indet. Chironomidae Bibionidae Syrphidae Sepsidae Lauxanidae Drosophilidae Acartophthalmidae Scatophagidae Anthomyiidae 2 4±2 1 1 3 5 15 4±1 3 8 4±1 17 ± 5 Muscidae indet. indet. indet. indet. Sepsis sp. Lauxania cylindricornis Fab. Drosophila nigrosparsa Strobl indet. Scatophaga stercoraria Linné Anthomyia liturata Rob.-Desv. indet. indet. indet. indet. indet. Formicidae Ichneumonidae indet. indet. Formica rufa L. indet. 1 1 7 Total of visits Time per visit 1200 109 ± 48 Time per visit 7 293 ± 106 2 2 10 1200 ± 0 638 ± 443 334 ± 147 5 1 1 7 1 25 4 8 5 29 1 1 29 ± 18 10 2 99 ± 42 793 42 ± 17 26 ± 16 743 ± 143 95 ± 63 165 ± 31 14 10 1 6 1 117 10 121 ± 33 5 204 ± 29 80 4 92 ± 36 31 84 ± 40 Table 5 Visitors observed in Kippel during 18 observation periods of 20 min on uninfected and systemically infected branches (witches’ brooms) of Berberis vulgaris: Number of visits and mean time spent per visit in seconds (grand means ± SE) Class/Order Insecta Diptera Family Species Chironomidae Bibionidae Sepsidae Lauxanidae Scatophagidae Anthomyiidae Tachinidae indet. indet. Bibio sp. Sepsis sp. Lauxania minor Martinek Cnemacantha muscaria Fallén Drosophila nigrosparsa Strobl Scatomyza flava Fallén Acartophthalmus nigrinus Zett. Acartophthalmus pusio Frey Scatophaga stercoraria Linné Hylemya variata Fallén indet. indet. Argidae Ichneumonidae Arge sp. indet. Drosophilidae Acartophthalmidae Hymenoptera Total of visits Kippel, during the same amount of observation time as in Ardez, 195 visitors were observed (Table 5). Similar to Ardez (Table 4), 82% of the visits were counted on witches’ brooms. The diversity of the visitors was slightly lower than in Uninfected branch Witches’ broom Number Number Time per visit 1 2 1 273 69 ± 35 16 3 101 ± 73 2 5 11 7 104 ± 13 338 ± 114 105 ± 70 13 ± 7 3 1 36 86 ± 63 3 115 ± 31 Time per visit 3 1 3 2 4 22 3 1 53 13 44 9 1 49 ± 32 67 23 ± 14 118 ± 48 603 ± 246 344 ± 86 768 ± 276 320 148 ± 26 686 ± 105 240 ± 38 200 ± 122 62 159 266 ± 26 Ardez. 98% of the visitors were Diptera from nine different families. Hymenoptera (Argidae and Ichneumonidae) accounted for the remaining 2%. In Kippel no ants were observed during the whole observation time, whereas in www.newphytologist.com © New Phytologist (2002) 154: 717–730 Research Fig. 4 Observed visitors on uninfected branches and systemically infected branches (witches’ brooms) of Berberis vulgaris. (a) Mean number of visits during observation periods of 20 min (±1 SE) in Ardez and in Kippel. (b) Mean number of visits during observation periods of 20 min (±1 SE) for the different time blocks. (c) Mean time spent per visit (±1 SE) in Ardez and in Kippel. (d) Mean time spent per visit (±1 SE) for the different time blocks. Time blocks 1– 4 were on day 1 at 1030 h, 1150 h, 1310 h, and 1420 h; Time blocks 5 –6 were on day 2 at 1220 h and 1340 h. Ardez 10 individuals of Formica rufa L. were observed on one bush, four on the uninfected branch and six on the witches’ broom. For several insect species some differences were observed in the mean time spent per visit on infected and uninfected branches (Table 5). None of these differences were significant with t-tests, most likely because the sample size was too small. The untransformed number of visits per observation period and the time spent on uninfected branches and witches’ brooms are shown in Fig. 4. In Ardez and Kippel the visitation rates on witches’ brooms were significantly higher than on similar sized uninfected branches (ANOVA, F1,60 = 146.194, P < 0.0001; Table 6). The visitation rates were also significantly affected by the bushes (ANOVA, F4,60 = 5.672, P = 0.0006; Table 6) and the timeblocks (ANOVA, F5,60 = 2.338, P = 0.0525; Table 6). The mean time spent per visit was significantly longer on witches’ brooms than on uninfected branches (ANOVA, F1,330 = 22.01, P = 0.005; Table 7). On both branch types, the visitors spent significantly more time in Ardez, than in Kippel (ANOVA, F1,330 = 46.73, P = 0.0005; Table 7). The interaction between time blocks and branch types was included in the model because it was significant in some analyses. In Fig. 4(d) we show the untransformed data for this interaction. One can see that the duration of visits on witches’ brooms is longer than on uninfected branches for all time blocks, except for time block six. In this time block, during six observation periods of 20 min, only eight visitors were counted on uninfected branches. One of them was a spider that stayed for the © New Phytologist (2002) 154: 717– 730 www.newphytologist.com Table 6 ANOVA on the number of visits by insects and spiders during replicate 20 min observation periods on uninfected and infected branches of Berberis vulgaris (data log-transformed). Site, bush (nested within site), and time block (= 3 consecutive observation periods) are included as random factors and branch type as a fixed factor. P-values ≤ 0.05 are given in bold face Source of variation DF MS F P Site Bush [site] Time block Branch type Error 1 4 5 1 60 0.446 0.868 0.358 22.373 0.153 0.514 5.672 2.338 146.194 0.5132 0.0006 0.0525 < 0.0001 whole 20 min observation time, causing the high mean and the large standard error of the duration per visit in this time block. Fragrance analysis The volatiles recovered from headspace collections from infected and uninfected leaves as well as the flowers of B. vulgaris are listed in Table 8. The compounds that are produced primarily by spermatia are listed alphabetically first, followed by those that were produced only or primarily by normal green tissues (the ‘green leaf ’ volatiles), and finally those that were primarily produced by the flowers. The ‘green leaf ’ volatiles were produced in quantities at least 50 times less than those of either infected tissues or the flowers. 725 726 Research Table 7 ANOVA on the time spent per visit on uninfected and infected branches of Berberis vulgaris (data log-transformed). Site, bush (nested within site), and time block are included as random factors and branch type as a fixed factor. P-values ≤ 0.05 are given in bold face Source of variation DF MS F P Site Bush [site] Time block Branch type Time block * branch type Error 1 4 5 1 5 330 32.630 0.607 4.123 99.625 4.603 2.414 46.728 0.252 0.897 22.005 1.907 0.0005 0.9086 0.5457 0.0049 0.0927 The most prevalent volatiles (as a percentage of total) emitted from the spermatia-covered leaves were indole (11.5 – 26.2%), methyl nicotinate (5.7– 25.9%), α-phellandrene (1.2–22.5%), carvacryl methyl ether (0.8 –11.5%), and Jasmin lactone (1.0 –6.8%). Of these, only a trace of indole was found in the control, uninfected leaf sample (Table 8), the rest were only found in the samples with spermatia. The characteristic and intense scent of infected leaves as the human nose perceives it is mainly due to indole, methyl nicotinate, Jasmin lactone and carvacryl methyl ether. Depending on the sample, one or the other of these compounds may dominate olfactorily. We made multiple collections of infected branches: one from Trimmis, three from Felsberg and one from Zürich Botanic Gardens ( Table 8). The major volatile compounds were shared by all collections, however, there was variation in the presence and absence of some of the minor compounds, and in the proportion of the total sample each compound represented. The host flowers open later in the year, after the fungus is finished reproducing. Not surprisingly, then, there was little overlap of compounds between them. Only two minor constituents were shared: 6-methyl-5-heptene-2-one and 2-phenylethanol. By contrast to the flowery-fruity fragrance of witches’ brooms, B. vulgaris flowers emit a rather unpleasant, spermy odor. This is due to a yet unidentified volatile compound which is perceived below the threshold of the mass spectrometer, but which humans have sensitive receptors for. This unknown compound is also present in some orchids and other flowers ( R. Kaiser, personal observations). Discussion Insect exclusion experiment The question of whether insects are required for sexual reproduction of the rust fungus causing the witches’ brooms on Berberis vulgaris, was answered with the exclusion experiment. We found that successful reproduction of the fungus Puccinia arrhenatheri likely depends on gamete transfer by insects between witches’ brooms of different mating types. Our ultrastructural data show that masses of spermatia adhere to the extremities of arthropod visitors of witches’ brooms. A certain degree of fungal reproduction on heterothallic rust fungi when visitors were excluded can be explained by a low level of self-compatibility, multiple infections of different mating types, or errors in the experimental work, such as undiscovered visitors. A small amount of fungal reproduction has been found in other exclusion studies as well (3– 19%) (Craigie, 1927; Roy, 1993; Pfunder & Roy, 2000; Schürch et al., 2000). Compared with natural insect visitation, the exclusion of flying visitors significantly reduced the reproductive success of the fungus, whereas exclusion of crawling insects did not have a significant effect. The results suggest that flying insects, mainly Diptera (see below), are the most important gamete vectors, and crawling insects, such as ants, are of minor importance. Scanning electron microscopy Our ultrastructural data show that masses of spermatia adhere to the extremities of arthropod visitors of witches’ brooms. Low-temperature scanning electron microscopy (LTSEM) facilitated the visualization of pycnidial mucilage and spermatia in the near natural state. Insect observations At all sites the diversity of observed insects was high. The most frequently recorded taxa belonged to the Diptera and Hymenoptera, but depending on the locality, different genera and species dominated. In Kippel we made an extraordinary observation – the most tallied species was Acartophthalamus pusio Frey, a small fly that has not been collected in Switzerland before (B. Merz, pers. comm.). Diptera of the family Drosophilidae (mostly Drosophila nigrosparsa Strobl) were recorded at all four sites, whereas anthomyid and scatophagid flies were common visitors only in Ardez and in Kippel. Species of these two families were also found to be visitors of pseudoflowers induced by rust fungi on crucifers (Roy, 1994; Roy, 1996). Halictid bees, also frequent visitors of pathogen-induced pseudoflowers on crucifers, were never observed on witches’ brooms of B. vulgaris, although they occur in Switzerland. The most common observed Hymenoptera were ants. Different species of Formica ants were observed in Ardez, Piotta and Trimmis but not in Kippel. In previous years numerous ants were seen on witches’ brooms in Ardez. In the spring of 2000, ants were crowded on a witches’ broom on only one bush in Trimmis. The ants were usually present in similar low quantities on uninfected branches as on witches’ brooms. However, we do not know what caused the higher attractiveness of witches’ brooms to ants in previous years; different meteorological conditions leading to differences in food sources might have played a role. www.newphytologist.com © New Phytologist (2002) 154: 717–730 © New Phytologist (2002) 154: 717–730 www.newphytologist.com Table 8 Volatiles emitted from Berberis vulgaris tissues covered in spermatia of Puccinia arrhenatheri, leaf controls and the flowers of B. vulgaris. The volatiles are expressed as a percentage of the total sample. Sperm = spermatia, Trim = the Trimmis locality, Fels = the Felsberg locality, BotGdn = the Zürich Botanical Garden 1° emitted from Compound Type R304 Spermatia R3918 Spermatia Isoprenoid Monoterpene “ R518 R8 R9 R3 R520 R10 R567 Spermatia Spermatia Spermatia Spermatia Spermatia Spermatia Spermatia R695 R410 R699 R678 Spermatia Spermatia Spermatia Spermatia “ “ “ “ “ “ Isoprenoid Bisnormonoterpene Lipid Metabolite “ “ “ R1675 R11884 Spermatia Spermatia “ “ R676 R2049 N10 Spermatia Spermatia Spermatia R1803 R249 Spermatia Spermatia “ “ Amino acid Metabolite Benzenoid Benzenoid N0107 R4101 Spermatia Spermatia R11 R7 RU64 R6 R25 R26 R2 R103 Leaves Leaves Leaves Leaves Leaves Leaves Leaves Leaves R190 Leaves Nitrogen containing Isoprenoid Monoterpene “ “ “ “ “ “ Isoprenoid Sesquiterpene “ Compound Sperm. Trim Sperm. Fels1 Sperm. Fels2 Sperm. Fels3 Camphene 0.30 0.50 0.10 0.10 Carvacryl methyl ether (= Thymyl methyl ether) Fenchone α-Phellandrene β-Phellandrene β-Pinene Piperitone γ-Terpinene 6-Methyl-5-hepten-2-one 3.60 11.50 4.00 0.90 22.50 2.60 0.20 2.20 0.60 0.10 1.20 2.20 0.10 1.00 0.10 3.40 2.00 0.10 0.80 0.05 2.90 1.70 0.10 0.80 0.30 0.05 0.10 0.10 1.00 0.04 0.01 0.20 0.30 4.60 0.20 6.80 0.80 5.00 0.10 0.10 0.10 0.30 0.20 0.60 0.40 0.10 0.60 10.20 0.10 0.30 21.20 0.40 26.20 1.00 0.60 0.80 0.30 1.00 0.50 5.70 0.20 13.20 7.20 25.90 6.40 20.40 6.00 14.50 10.80 1.10 30.10 4.10 13.10 3.60 18.10 4.90 0.60 0.20 4.90 1.10 0.10 0.10 0.30 0.20 0.05 0.10 0.10 0.60 0.05 0.10 0.10 1.70 0.10 0.10 0.10 γ-Caprolactone Decanal δ-Decalactone Jasmine lactone (= (Z)-7-Decen5-olide) γ-Jasmolacton (= (Z)-7-Decen-4-olide) Methyl palmitate (= Methyl hexadecanoate) δ-Octalactone Pentadecane Indole Methyl benzoate Phenylethyl alcohol (= 2-Phenylethanol) Methyl nicotinate unknown p-Cymene Limonene A isomer of p-Menthatriene Myrcene (E)-Ocimene (Z)-Ocimene α-Pinene Caryophyllene (E,E)-α-Farnesene 0.20 11.50 0.10 Sperm. Bot.Gdn 14.4 Leaves Fels. Leaves Bot.Gdn Flowers Bot.Gdn 1.2 0.80 15.00 2.80 0.20 1.40 1.8 0.50 11.10 23.40 2.90 0.40 13.60 2.00 0.30 4.60 1.10 1.40 19.5 14.40 15.00 3.40 0.50 7.80 3.30 1.40 0.60 9.6 3.50 0.20 17.4 21.0 4.5 0.05 0.30 0.05 Research Kaiser Ref# 727 728 Research Table 8 continued www.newphytologist.com © New Phytologist (2002) 154: 717–730 Kaiser Ref# 1° emitted from Compound Type R2037 Leaves R256 R250 R2797 R5714 R7736 R409 R2048 R2047 R466 R10225 Leaves Leaves Leaves Leaves Leaves Leaves Leaves Leaves Leaves Flowers R4284A Flowers Isoprenoid Tetranorsesquiterpene Lipid Metabolite “ “ “ “ “ “ “ Benzenoid Isoprenoid Monoterpene “ R4284B R4284C R4284D R1422B R1422C Flowers Flowers Flowers Flowers Flowers “ “ “ “ “ R232 R7219 R1279 R778 R304 Flowers Flowers Flowers Flowers Flowers R570 Flowers R774 R257 R1242 R1885 Flowers Flowers Flowers Flowers “ “ Isoprenoid “ Isoprenoid Sesquiterpene Isoprenoid BisnorSesquiterpene Lipid Metabolite “ Benzenoid “ Compound Sperm. Trim Sperm. Fels1 Sperm. Fels2 Sperm. Fels3 Sperm. Bot.Gdn Leaves Fels. (E)-4,8-Dimethyl-1,3,7-nonatriene 0.50 0.40 0.10 0.20 0.10 5.1 1-Hexanol (Z)-3-Hexenol (Z)-3-Hexenyl acetate (Z)-3-Hexenyl butyrate (Z)-3-Hexenyl isovalerate Nonanal Tetradecane Tridecane Benzaldehyde 2,6-Dimethyl-3,7-octadien-2,6-diol Lilac alcohol A (= 2-(5-Ethenyl-5methyltetrahydrofuran-2yl)propan-1-ol) Lilac alcohol B Lilac alcohol C Lilac alcohol D Lilac aldehyde B Lilac aldehyde C (= 2-(5-Ethenylmethyltetrahydrofuran-2yl)propanal) Linalool cis-Linalool oxide (furanoid) Prenol Prenyl acetate (E)-Nerolidol 0.10 0.05 0.02 0.03 0.50 0.50 0.30 0.40 0.60 0.10 0.30 0.20 0.40 1.2 1.8 Leaves Bot.Gdn Flowers Bot.Gdn 1.20 4.90 1.00 9.00 24.60 3.60 1.00 1.00 0.50 0.90 0.10 1.40 0.20 0.10 0.9 2.1 1.2 0.20 0.10 0.10 0.10 0.10 0.10 0.60 78.90 0.40 0.50 0.10 0.30 (E)-Geranylacetone (= (E)-6,10dimethyl-5,9-undecadien-2-one) 0.05 Hexyl acetate 1-Octanol Benzyl alcohol Methyl salicylate (= Methyl 2hydroxybenzoate) 0.60 0.10 0.05 0.10 Research It has been shown that the attraction of insects by pathogen-induced pseudoflowers can affect the insect visitation on co-occurring flowers of the host and other plants (Roy, 1994; Roy, 1996; Pfunder & Roy, 2000). In this case, B. vulgaris shrubs do not flower at the same time that the fungus is in the spermatial phase, so there is no influence of the pathogen’s attractiveness on the reproduction of the hosts. There may well be other fitness consequences of being infected. Given that most large shrubs bear witches’ brooms, but also flower and have high between year survival (B. A. Roy, personal observations), it is unlikely that moderate infection has severe fitness consequences in mature plants. The insect observations suggest that insects are attracted by the leaves on witches’ brooms, probably by the combination of visual and olfactory cues such as yellow discoloration of infected leaves, nectar secretion and emission of sweetsmelling volatiles. In the spring, during the spermatial phase of the fungus, the leaves on witches’ brooms appear earlier than leaves on uninfected branches, and they are yellow to orange in color. The dense growth structure of witches’ brooms increases the spatial concentration of infected leaves, thus forming a conspicuous yellow patch, which may work as a visual cue for insects, comparable with pseudoflowers induced by some crucifer rusts (Roy, 1993). Roy & Raguso (1997) showed that odor may also be an important attractant for insects on pseudoflower-forming rust fungi. Similar to pseudoflowers, infected leaves of B. vulgaris emit a sweet-smelling scent. The fragrances are distinct from those of uninfected leaves (see below) and therefore are assumed to be produced by the fungus, most likely to attract insects. Fragrances The broad spectrum of volatiles emitted from infected leaves undoubtedly played a role in attracting the high diversity of insects observed in this study. The compounds found have at least three biological roles. First, there were the floral fragrances, such as Jasmine lactone, methyl benzoate, phenylethyl alcohol, methyl palmitate and camphene, which are often recovered from flower fragrances and are known to be attractive to pollinators (Knudsen, 1993). Second, there were pheromone-like substances, such as γ-caprolactone, which is attractive to several insects including hornets and dermestid beetles (Naoshima et al., 1983), and 6-methyl-5hepten-2-one, which is an attractant for a parasitoid of aphids (Du et al., 1998). Third, there were compounds that may play a defensive role, such as β-pinene (Heldt, 1997), carvacryl methyl ether, fenchone (Rice & Coats, 1994) and indole (Gloer, 1995). Since these defensive compounds were not recovered from uninfected leaves (or in the case of indole and carvacryl methyl ether were recovered in tiny amounts in one leaf sample), it is reasonable to conclude that they were produced either by host tissue in response to the attack of the pathogen, or by the pathogen itself. © New Phytologist (2002) 154: 717– 730 www.newphytologist.com All of the compounds recovered from the spermatiacovered leaves are found in true flowers of other species and likely play a role in attracting flower-visiting insects. Fewer of these compounds were shared with other insect-attracting fungi, although this is more difficult to evaluate as there are fewer studies to compare with. Nonetheless, it is interesting that only two compounds found in this study, 2-phenylethanol and indole, have been found to be emitted from other rust infected plants (Connick & French, 1991; Raguso & Roy, 1998). 2-phenylethanol is commonly found in flower volatile emissions (Knudsen, 1993), is known to be attractive to noctuid moths and is even a mating-pheromone constituent of some species (Birch et al., 1989). Indole is one of the dominant compounds found in rust-infected B. vulgaris, and is also the major component of the sweet fragrance emitted by Puccinia punctiformis Diet et Holw on Cirsium arvense (L) Scop. (Canada thistle) (Connick & French, 1991). Indole is also found in the scents of many flowers (Knudsen, 1993; Kite, 1995). However, mechanical injury and feeding by insects are also known to release indole from plants (Pare & Tumlinson, 1996). Thus it is possible that the indole recorded from these samples is the result of the stems having been cut or from other mechanical injury incurred while the sample was being collected. However, in our study, the control uninfected leaf sample had only a trace of indole (10 –20 times smaller than that on infected plants). Further, the flower samples, whose stems were also cut, did not emit this compound. Indole is also known to be an antifungal agent (Gloer, 1995); it is thus possible that this volatile may be a defensive response of the plant. Since indole has been reported to have many biological roles, and since it is such a prevalent compound in these rust fungus–host interactions, further experimentation should be performed to determine its exact role in these systems. There was variation in the proportion of the total sample represented by each volatile, and in the presence and absence of some of the minor compounds (Table 8). Several compounds were only found in the Trimmis sample, whereas the one from a cultivated bush in the botanic garden showed lower diversity. We wondered whether these differences might have arisen as a result of the fungi being of different races or species. We tested this idea by examining the ITS sequences of all the fungi from which fragrances were detected. There were no differences in the sequences and they all matched the sequences for P. arrhenatheri (A. Naef and B. A. Roy unpublished). The differences in the compounds among samples are thus likely to have arisen as a result of differences in how long the fungus had been receptive (age), and from individual differences in the hosts such as genotype and overall condition. Acknowledgements The authors thank D. Siemens for his comments on the manuscript, B. Merz and R. Neumeyer for insect identifications, K. Ullrich and R. Langenauer for help with 729 730 Research grass identifications and S. Züllig for help in the field. Special thanks are due to F. von Planta and the community of Ardez for permission to work on their properties, to E. Steinmann for showing us the Felsberg and Trimmis localities, and to J.-J. Pittet for competent help with the artwork. References Agrios GN. 1988. Plant pathology. San Diego, CA, USA: Academic Press. Antonovics J, Alexander HM. 1992. Epidemiology of anther-smut infection of Silene alba (= S. latifolia) caused by Ustilago violacea: patterns of spore deposition in experimental populations. Proceedings of the Royal Society of London. B 250: 157–163. Batra SWT. 1991. Floral mimicry and insects as vectors of conidia. In: Batra LR, ed. World species of Monilinia ( fungi): their ecology, biosystematics and control. Berlin, Germany: J. Cramer. 93– 97. Birch M, Lucas D, White P. 1989. The courtship behavior of the cabbage moth, Memestra brassicae (Lepidoptera: Noctuidae) and the role of male hair pencils. Journal of Insect Behavior 2: 227– 239. Buller AHR. 1950. Researches on fungi, VII: the sexual process in the uredinales. Toronto, Canada: University of Toronto. Bultman TL, White JF. 1988. ‘Pollination’ of a fungus by a fly. Oecologia 75: 317– 319. Cartar RV, Abrahams MV. 2000. The infection of pyrola (Pyrola asarifolia ; Pyrolaceae) by spruce cone rust (Chrysomyxa pirolata ; Uredinales): Morphological correlates in the host and consequences for spore dispersal. Ecoscience 7: 357–364. Connick WJ, French RC. 1991. Volatiles emitted during the sexual stage of the Canada thistle rust fungus and by thistle flowers. Journal of Agricultural and Food Chemistry 39: 185–188. Craigie JH. 1927. Discovery of the function of pycnia of the rust fungi. Nature 120: 765–767. Cummins GB, Greene HC. 1966. A review of the grass rust fungi that have uredial paraphyses and aecia on Berberis – Mahonia Mycologia 58: 702–721. Du YJ, Poppy GM, Powell W, Prickett JA, Wadhams LJ, Woodcock CM. 1998. Identification of semiochemicals released during aphid feeding that attract parasitoid Aphidius ervi. Journal of Chemical Ecology 24: 1355 –1368. Eriksson J. 1898. Studien über den Hexenbesenrost der Berberitze, Puccinia arrhenatheri Kleb. Cohns Beiträge zur Biologie der Pflanzen 8: 1–16. Gäumann E. 1959. Die Rostpilze Mitteleuropas mit besonderer Berücksichtigung der Schweiz. Bern, Switzerland: Buchdruckerei Buchler. Gloer JB. 1995. The chemistry of fungal antagonism and defense. Canadian Journal of Botany 73: 1265 –1274. Hegi G. 1974. Illustrierte Flora von Mitteleuropa. Berlin, Germany: Parey. Heldt HW. 1997. Plant biochemistry and molecular biology. Oxford, UK: Oxford University Press. Hess HE, Landolt E, Hirzel R. 1977. Flora der Schweiz. Basel, Switzerland: Birkhauser. Kaiser R. 1993. The scent of orchids. Amsterdam, The Netherlands: Elsevier. Kaiser R, Tollsten L. 1995. An introduction to the scent of cacti. Flavour and Fragrance Journal 10: 153–164. Kite GC. 1995. The floral odour of Arum maculatum. Biochemical Systematics and Ecology 23: 343–354. Knudsen JT. 1993. Floral scents – a checklist of volatile compounds isolated by head-space techniques. Phytochemistry 33: 253–280. Magnus P. 1875. Aecidium Berberidis und Aecidium magelhaenicum. Verhandlungen des Botanischen Vereins der Provinz Brandenburg 17: 76–87. Magnus P. 1894. Die von J. Peyritsch in Tirol gesammelten und im Herbarium der K. K. Universität zu Innsbruck aufbewahrten Pilze. Berichte des Naturwissenschaftlich-Medizinischen Vereines in Innsbruck 21: 25 –73. Naoshima Y, Ozawa H, Kondo H, Hayashi S. 1983. Synthesis of racemates (+)-enantiomers of gamma-caprolactone, gamma-dodecanolactone and delta-hexadecanolactone, lactonic sex-pheromones of the dermestid beetle, rove beetle and oriental hornet. Agricultural and Biological Chemistry 47: 1431–1434. Pare PW, Tumlinson JH. 1996. Plant volatile signals in response to herbivore feeding. Florida Entomologist 79: 93 –103. Pfunder M, Roy BA. 2000. Pollinator–mediated interactions between a pathogenic fungus, Uromyces pisi (Pucciniaceae) and its host plant, Euphorbia cyparissias (Euphorbiaceae). American Journal of Botany 87: 48 –55. Raguso RA, Roy BA. 1998. ‘Floral’ scent production by Puccinia rust fungi that mimic flowers. Molecular Ecology 7: 1127–1136. Rathay E. 1883. Untersuchungen über die Spermogonien der Rostpilze. Denkschrift der Kaiserlichen Akademie der Wissenschaften. XLVI: 1–51. Rice PJ, Coats JR. 1994. Structural requirements for monoterpinoid activity against insects. Bioregulators for Crop Protection and Pest Contrology 557: 92 –108. Roy BA. 1993. Floral mimicry by a plant pathogen. Nature 362: 56 –58. Roy BA. 1994. The effects of pathogen-induced pseudoflowers and buttercups on each other’s insect visitation. Ecology 75: 352–358. Roy BA. 1996. A plant pathogen influences pollinator behavior and may influence reproduction of non hosts. Ecology 77: 2445–2457. Roy BA, Raguso R. 1997. Olfactory versus visual cues in a floral mimicry system. Oecologia 109: 414–426. SAS. 1994. JMP version 3.1. Cary, NC, USA: SAS Institute. Schürch S, Pfunder M, Roy BA. 2000. Effects of ants on the reproductive success of Euphorbia cyparissias and associated pathogenic rust fungi. Oikos 88: 6–12. Underwood AJ. 1981. Techniques of analysis of variance in experimental marine biology and ecology. Oceanography and Marine Biology Annual Reviews 19: 513–605. Urban Z, Marková J. 1994. The rust fungi of grasses in Europe. 2. Puccinia brachypodii Otth and its allies. Acta Universitatia Carolinae Biologica 38: 13 –57. Wilding N, Collins NM, Hammond PM, Webber JF. 1989. Insect–fungus interactions. London, UK: Academic Press. www.newphytologist.com © New Phytologist (2002) 154: 717–730