Oecologia (2002) 132:369Ð373 DOI 10.1007/s00442-002-0998-1

PLANT ANIMAL INTERACTIONS

B. Anderson á J. J. Midgley It takes two to tango but three is a tangle: mutualists and cheaters on the carnivorous

Received: 5 September 2001 / Accepted: 10 June 2002 / Published online: 3 July 2002 © Springer-Verlag 2002

Abstract Roridula dentata is associated with hemipter- wards (Soberon and Martinez del Rio 1995). The result ans, which facilitate nitrogen assimmilation from insects. may be mutualism collapse, which may trigger the ex- R. dentata is also associated with spiders and their role tinction of obligate partners (Boucher et al. 1982). This in digestion is unknown. We quantify approximately fluctuation of reward quality (e.g. availability) has how much nitrogen Roridula assimilates from insects been identified as one of the major factors constraining through “indirect digestion.” Using δ15N we then deter- the evolution of specialist or one-on-one, obligate mutu- mine whether nitrogen absorption from hemipteran in- alisms (Thompson 1994). Howe (1984) suggests that di- sects differs with varying spider densities. In this way, verse faunas frequently affect reward quality negatively we are able to determine their nutritional role. At low by competing with mutualists. As a result, diverse faunas spider densities, indirect digestion of prey accounts for are sometimes viewed as the source of unpredictable re- approximately 70% of plant nitrogen. These values are ward quality. comparable to methods of direct prey digestion found in Carnivorous have a plethora of closely associ- other carnivorous plants. However spiders decrease the ated invertebrate fauna with both facultative and obli- numbers of hemipteran individuals inhabiting Roridula gate relationships that range from parasitisms to mutual- plants and also decrease efficiency of indirect prey di- isms (Beaver 1983, 1985; Juniper et al. 1989). Because gestion by up to 30%. We deduce that spiders are cheat- of their diversity, these systems may be excellent for ers as they exploit plant rewards without offering any re- exploring the evolution and breakdown of mutualisms wards in return. However, indirect carnivory is still effi- and the role of cheating. However, very few studies cient enough when hemipteran densities are at their low- have examined relationships in detail: est, ensuring that the mutualism does not break down. Zamora (1990) observed that ants, which are kleptopar- asites of , steal a large proportion of prey. Keywords Mutualism á Evolution á Cheating á Context Clarke and Kitching (1995) noted that pitch- dependency á Carnivorous plant ers may capture an excess of prey, which often causes putrification of the pitcher if not removed. But specialist ants living on the Nepenthes plants remove excess prey, Introduction a process that reduces the incidence of putrification by up to 12-fold. Bradshaw and Creelman (1984) showed Thieves or cheaters are an inevitable part of almost every that pitcher fauna may increase the rate of digestion in mutualistic system (Springer and Smith-Vaniz 1972; Jan- purpurea and thus could be considered mu- zen 1975; Thompson 1982). The invasion of mutualisms tualists. by cheaters may cause variation in the reward quality of The plant Planch. catches insects mutualisms (e.g. Pellmyr et al. 1996). For example, using sticky traps but has no digestive enzymes to utilise cheaters such as nectar thieves may leave little rewards the prey (Marloth 1925). Using an isotopic labelling for mutualistic pollinators. As a result, these pollinators technique, Ellis and Midgley (1996) determined that R. may be forced to seek flowers with more reliable re- gorgonias has a digestive mutualism with a specialist he- mipteran (Dolling and Palmer 1991) that consumes the captured prey. These hemipterans defecate on Roridulas’ B. Anderson (✉) á J.J. Midgley and the plants absorb nitrogen directly through Botany Department, University Of Cape Town, P Bag Rondebosch, 7701 South Africa their cuticle (Ellis and Midgley 1996). However, the e-mail: [email protected] amount of nitrogen gained in this manner remains to be Fax: +27-21-6504041 quantified. 370 Using isotopic methods, we aim to examine and quan- rhizal (VAM) association, reference plants were chosen that also tify the effects of two different relationships with had this association (based on Allsopp and Stock 1993). Pamer- idea (n=5 from each population) and live prey items were also col- Roridula. We examine an unstudied of Roridula lected from each population (prey items were pooled by grinding (R. dentata L.) with a unique species of associated he- all insects together). Prey consisted of four Hymenoptera, three mipteran (Pameridea marlothi Poppius). A monophylet- Coleoptera, three Diptera and one Hemipteran. At these sites we ic group of hemipterans (one , two species) is asso- felt it unnecessary to control for rooting depth as R. dentata has a shallower rooting depth than all the other plants sampled. Since ciated with every known Roridula population and each δ15N typically increases with rooting depth (Gebauer and Schulze of the two Roridula species is associated with a different 1991; Hogberg 1997; Schmidt and Stewart 1997) any differences species of Pameridea (Dolling and Palmer 1991). In ad- between Roridula and reference plant δ15N values will thus be dition, a specialist spider (Synaema marlothi Dahl., conservative. Thomsidae) is also associated exclusively with R. denta- In a separate experiment, two more sites were used to compare levels of insect-derived nitrogen in R. dentata with and without ta. These spiders occur on most R. dentata populations spiders. These sites are geographically distant from Pop6, Pop7 in the southern and central part of their range. Where and Pop8. The first population (Pop5) is a small R. dentata popu- Synaema occurs, it is usually very common and builds its lation with no spider inhabitants and is in the Tulbagh region nests in the axils of Roridula’s branches and dead leaves (33¡21′ S, 19¡05′ E). The closest known R. dentata population to this is situated approximately 17 km away in the Ceres area (Marloth 1925; personal observation). The spiders are (33¡22′ S, 19¡15′ E). This site (Pop9) is dry with sandy soil and carnivores and presumably compete with hemipterans has several plant species (e.g. Leucadendron rubrum Burm.f., and for resources (Marloth 1925). In addition it has been re- Protea nana (P.J. Bergius) Thunb., both Proteaceae) in common corded that Synaema also consumes Pameridea (Lloyd with Pop5. In contrast to Pop5, the spider Synaema marlothi is abundant at Pop9. Leaves were bagged before sampling as above. 1934; personal observation). Spiders may thus disrupt At the time of sampling, five Roridula and five P. nana (reference the mutualism between Roridula and Pameridea if they plants) were sampled per population. decrease the numbers of hemipterans but do not supply plants with nitrogen. Alternatively spiders may also fa- cilitate nitrogen absorption in a similar way to Pamer- Mass spectrometer methods idea. Until now, no studies have been performed on the The relative contribution (%NCI) by insects to the nitrogen content spiders and their role in Roridula nutrition. Spiders are of a carnivorous plant was calculated by Schulze et al. (1991) us- also very commonly associated with other carnivorous ing methods modified from Peterson and Fry (1987). Where δ15 δ15 δ15 plants but their roles in other systems also remain un- NCAR is the N value for the carnivorous plant, NI is the δ15 δ15 δ15 studied. In this study, we quantify the nutritional benefits N value for insect prey and NREF is the N value for refer- ence plants (see Eq.1). of the hemipteran mutualist and determine the nutritional role played by spiders. (1)

The equation used to determine the percentage of nitrogen that Roridula gains from hemipteran faeces is derived from Schulze et Materials and methods al. (1991) (see Eq. 2).

The densities of hemipterans and spiders may be related to nitro- (2) gen uptake in plants; therefore, these were the first data collected. δ15 In addition to the five sites mentioned below, density data were Two methods were used to calculate NFAECES. The first is an in- collected from another four sites. These sites were Pop1 (32¡07′ S, direct method while the second is direct. 19¡00′ E), Pop2 (32¡40′ S, 19¡11′ E), Pop3 (32¡50′ S, 19¡12′ E) and Pop4 (32¡49′ S, 19¡13′ E). Synaema marlothi occurred at all except three study sites (Pop1, Pop8, Pop3) used in this study. At Method 1 every R. dentata population, 30 rosettes (approximately 20 leaves each) from random plants were rapidly bagged and the number of Since insects discriminate against 14N during excretion (Peterson adult hemipterans and spiders were counted. This enabled us to and Fry 1987) and retain 15N, their faeces will be depleted of 15N, observe the densities of hemipterans and spiders relative to each relative to their food source. The value of δ15N retained by an in- other. It also enabled us to relate levels of insect-derived nitrogen sect is: in Roridula to hemipteran or spider density. Three geographically close sites (Pop6: 32¡39′ S, 19¡11′ E, (3) Pop7: 32¡40′ S, 19¡11′ E, and Pop9: 32¡40′ S, 19¡11′ E) were used to investigate the nutritional impacts of hemipterans and the The δ15N value for faeces will be δ15N of the prey minus the δ15N role played by spiders. These sites were chosen because they retained by the insects: shared very similar vegetation, are geographically close and dif- fered in the density of spiders and Pameridea. Young samples (4) (n=5 per species) were collected from Roridula and reference plants at each site for isotopic analysis. Approximately 1 week be- fore leaf harvesting, all leaves were bagged in transparent nylon Method 2 bags while still in bud so as to exclude prey items and Pameridea. The following reference plants were collected at each site: Ten hemipterans from each population were kept in separate petri plumosa (L.) Thunb. (), densa (Lam.) Karis dishes for three days. After this period, all hemipterans were dried (Asteraceae) and Diosma sp. (Rutaceae). Reference plants were and homogenised together. Their faeces were also left to dry and chosen according to their mycorrhizal associations as this is were scraped off the petri dishes. Both faeces and hemipterans thought to influence δ15N values (Hobbie et al. 2000; Schmidt and were run on the mass spectrometer to determine average δ15N val- Stewart 1997). Since Roridula has a vesicular-arbuscular mycor- ues for faeces and hemipterans. 371 Results

There is a significant (r2=0.57 and P=0.018) negative re- lationship between hemipteran density and spider density (Fig. 1). The highest hemipteran densities were found at the study site Pop1 that had no spiders. The site Pop5 also had no spiders but hemipteran densities were not as high as two other sites with very low spider densities. The highest spider and lowest hemipteran densities were found in the site Pop8. At sites Pop6, 7 and 8, there was no significant differ- Fig. 1 The densities of the hemipteran, Pameridea and spider, ence (Kruskal-Wallis, Fig. 2) between the δ15N values Synaema marlothi at a range of different study sites for Metalasia (χ2=3.75, P=0.1534), Diosma (χ2=1.33, P=0.5134) and Stoebe (χ2=1.33, P=0.5134). In contrast, the δ15N values of Roridula plants at the three sites were significantly different from each other (χ2=6.92, P=0.0308). Hemipteran densities at the three sites also differed significantly (χ2=18.61, P=0.0001). The lowest mean hemipteran densities (SE) were from Pop8 (0.20±0.70). Hemipteran densities on Pop6 and Pop7 were 1.23 (0.22) and 1.46 (0.21) respectively. Using EQ. 2, the levels of insect nitrogen (%NRI) at each study site were calculated (using the indirect method Ð 1) as Pop6=70%, Pop7=71% and Pop8=41% (Table 1). The δ15 NFAECES values were extremely similar using the di- rect and indirect approaches (Table 1). Plants from the study site Pop9 had a mean of 0.3 ± Fig. 2 Delta 15N values for R. dentata, reference plants and Pa- 0.2 (SE) adult spiders per rosette. These plants had sig- meridea at three study sites (Pop6, Pop7 and Pop8) within close nificantly fewer hemipterans (Mann-Whitney U, proximity of each other. Mean, box = SE and error bars =1.96×SE Z=Ð2.04, P=0.042) than plants at the Pop5 site, which had no spiders. The δ15N signatures of P. nana from the two sites were similar (Fig. 3). At Pop9, R. dentata had similar δ15N values to the reference plant but at Pop5, R. dentata had very high δ15N values (Fig. 3). Roridula plants from Pop9 had higher δ15N values than all other plants tested in the study (Kruskall-Wallis, χ2=10.30, P=0.016). Due to the lack of an appropriate (VAM) ref- erence plant, we weren’t able to calculate the percentage of nitrogen derived from insects at Pop9 and Pop5. Al- though the comparison between Pop9 and Pop5 cannot be used on its own to show the role of spiders in Roridula nutrition, it nevertheless supports the other re- sults in this study.

Discussion Fig. 3 Delta 15N values for R. dentata and reference plant at two different study sites: (Pop9 and Pop5). Sites differed in spider and Spiders negatively influence the strength of the plant- hemipteran densities (see Fig. 3). Mean, box = SE and error bars hemipteran mutualism and thus the amount of insect- =1.96×SE derived nitrogen in Roridula varied according to spider

Table 1 The percentage of ni- δ15 δ15 δ15 trogen derived from hemipteran Study site N Roridula Faecal N Faecal N Faecal N Faecal N faeces for Roridula at three (SE) (method 1) (method 1) (method 2) (method 2) study sites, calculated using two methods. δ15N value for Pop6 3.12 (0.31) 5.19 70% 5.32 68% prey was 6.57‰ Pop7 2.86 (0.21) 4.90 71% 4.98 70% Pop8 1.03 (0.13) 5.34 41% 4.98 43% 372 densities. The majority of Roridula plants analysed fell The question may be partly answered by comparing within the range of insect nitrogen found by Schulze et levels of insect-derived nitrogen in Roridula populations al. (1991), where δ15N=0.819–3.302 ‰. However, at one (with high rates of kleptoparasitism) against levels of in- site (Pop9), the δ15N values were much lower (0.35±0.13 sect nitrogen in other carnivorous plants. Although the ‰) and δ15N was the same as for the non-mycorrhizal benefits of prey capture fluctuate, Roridula still receives reference plant. These comparatively low δ15N values a substantial amount of insect nitrogen (40%) even when suggest that Roridula does not obtain much nitrogen spider densities are at their highest. This is still far high- from insects at this site. er than the levels of insect nitrogen found in At two sites (Pop6 and Pop7) Roridula was deriving erythrorhiza (Dixon et al. 1980). Furthermore, it is com- exceptionally large amounts of nitrogen (%NRI=70%) parable with the average percentage of insect nitrogen from hemipteran faeces and therefore very little from the (50%) found in upright Drosera (Schulze et al. 1991). soil. Insect dependence in Roridula is much higher than We speculate that for Roridula, the gains of mutualism the average for upright Drosera (%NDI=50%) calculated outweigh the costs, even in the face of heavy cheating. by Schulze et al. (1991) and is one of the highest values Thus it is unlikely that the present levels of cheating of insect nitrogen uptake recorded. Insect dependence in should cause mutualism collapse in Roridula. Roridula is comparable to the most insect dependent car- This study identifies a close one-on-one mutualism nivorous plants such as Dionaea (Schulze et al. 2001) between a carnivorous plant and a carnivorous hemip- and pitcher plants (Schulze et al. 1997). After fire, Dio- teran making it one of the few close, one-on-one mutual- naea obtains approximately 75% of its nitrogen from in- isms known. We also find that spiders have invaded the sects. Pitcher plants such as Nepenthes and Darlingtonia mutualism, causing reward fluctuations that may con- can also be extremely insect dependent (61.5% and strain the evolution or maintenance of mutualisms. We 76.4% respectively). Insect nitrogen in Roridula exceeds believe that other carnivorous plants such as Sarracenia insect nitrogen in the (26%). and Nepenthes may also provide similar insights into However, at a single site, (Pop8) Roridula obtained very mutualisms and cheating. Like Roridula, many Sarrace- little nitrogen from insects, although it was still similar nia species do not have digestive enzymes (Juniper et al. to the average for Drosera calculated by Schulze et al. 1989). In addition, there is a large, unstudied inverte- (1991). brate fauna associated with these plants and we predict 15 Both %NRI and N values were variable, indicating that many may have digestive mutualisms similar to that the nutritional benefits to Roridula are also variable. those found in Roridula 15 At the sites Pop9 and Pop8, N and %NRI respectively were exceptionally low. These were also the only two Acknowledgements We thank Louis Hanekom, Willem Hane- sites analysed with high spider densities and low hemip- kom and Theunus Hanekom for their hospitality and for sharing teran densities. The negative correlation between spider their field knowledge. We would also like to thank Cape Nature and hemipteran densities indicates that spiders have a Conservation, in particular Marius Brand, Pieter Viljoen and Koos δ15 for their help in the field. Thanks to the Ceres Municipality for al- negative effect on hemipteran abundance. The low N lowing us access to the Ceresdam site. Thanks to all those who values and low percentages of insect nitrogen at sites spent time in the field with us, especially Amrei. Thanks to Isa- with high spider densities suggest that spiders do not con- belle Olivieri, the NRF and CNRS. Lastly thank you to two anon- tribute significant amounts of nitrogen to Roridula plants. ymous reviewers. Spiders can thus be regarded as kleptoparasites or cheat- ers. Differences in the behaviour of spiders and hemipter- ans may account for the differing effects of these two in- References vertebrate taxa on plant nutrition. For example, spiders spend large amounts of time in nests and presumably this Allsopp N, Stock WD (1993) Mycorrhizal status of plants growing is where they defecate. Alternatively, they allow their fae- in the , South Africa. Bothalia 23:91Ð104 ces to fall to the ground. In both cases, nitrogen is not Beaver RA (1983) The communities living in Nepenthes pitcher easily accessible to Roridula. This contrasts with hemip- plants: fauna and food webs. Phytotelmata: In: Frank JH, terans that defecate on the leaves of Roridula. Lounibos P (eds) Terrestrial plants as hosts of aquatic insect communities. Plexus, Marlton, N.J. pp 129Ð159 Insect-derived nitrogen may be crucial for the surviv- Beaver RA (1985) Geographical variation in food web structure in al of Roridula in the wild. In addition, hemipterans are Nepenthes pitcher plants. Ecol Entomol 10:241Ð248 major pollinators of Roridula and are responsible for up Boucher DH, James S, Keeler KH (1982) The Ecology of mutual- to 65% of seed set in R. dentata (unpublished data). The ism. Annu Rev Ecol Syst 13:315Ð347 combination of nutritional and pollination benefits of Bradshaw WE, Creelman RA (1984) Mutualism between the car- nivorous purple pitcher plant and its inhabitants. Am Midl Nat this mutualism indicates that hemipterans may be crucial 112:294Ð304 for plant survival. We tentatively classify the relationship Clarke CM, Kitching RL (1995) Swimming ants and pitcher as a one-to-one obligate mutualism, which Howe (1984) plants: a unique ant-plant interaction from Borneo. J Trop Biol describes as “rare in practice and anomalous in theory.” 11:589Ð602 Dixon KW, Pate JS, Baily WJ (1980) Nitrogen nutrition of the tu- This raises the important question of how such a close, berous sundew Drosera erythrorhiza Lindl. with special refer- one-on-one mutualism evolved and how it persists in the ence to catch of arthropod fauna by its glandular leaves. Aust face of cheating? J Bot 28:283Ð297 373 Dolling WR, Palmer JM (1991) Pameridea (: ): Peterson BJ, Fry B (1987) Stable isotopes in ecosystem studies. predaceous bugs specific to the highly viscid plant genus Annu Rev Ecol Syst 18:293Ð320 Roridula. Syst Entomol 16:319Ð328 Schmidt S, Stewart GR (1997) Waterlogging and fire impacts on Ellis AG, Midgley JJ (1996) A new plant-animal mutualism in- nitrogen availability and utilisation in a subtropical wet heath- volving a plant with sticky leaves and a resident hemipteran. land (wallum). Plant Cell Environ 20:1231Ð1241 Oecologia 106:478Ð481 Schulze ED, Gebauer G, Schulze W, Pate JS (1991) The utiliza- Gebauer G, Schultze E-D (1991) Carbon and nitrogen isotope ra- tion of nitrogen from insect capture by different growth forms tios in different compartments of a healthy and a declining of Drosera from Southwest Australia. Oecologia 87:240Ð246 Picea abies forest in the Fichtelgeberge, NE Bavaria. Oecolo- Schulze W, Schulze ED, Pate JS, Gillison AN (1997) The nitrogen gia 87: 198Ð207 supply from soils and insects during growth of the pitcher Hobbie EA, Macko SA, Williams M (2000) Correlations between plants Nepenthes mirabilis, Cephalotus follicularis and Dar- foliar 15N and nitrogen concentrations may indicate plant-my- lingtonia californica. Oecologia 112:464Ð471 corrhizal interactions. Oecologia 122:273Ð283 Schulze W, Shulze ED, Schulze I, Oren R (2001) Quantification of Hogberg P (1997) Tansley review no. 95. 15N natural abundance in insect nitrogen utilization by the venus fly trap Dionaea mus- soil-plant systems. New Phytol 137:179Ð203 cipula catching prey with highly variable isotope signatures. Howe HF (1984) Constraints on the evolution of mutualisms. Am J Exp Bot 52:1041Ð1049 Nat 123:764Ð777 Soberon J, Martinez del Rio C (1985) Cheating and taking advan- Janzen DH (1975) Pseudomyrmex nigropilosa: a parasite of a mu- tage in mutualistic associations. In: Boucher DH, et al (ed) tualism. Science 188:936Ð938 The biology of mutualism. Croom Helm, London, pp 192Ð216 Juniper BE, Robins RJ, Joel DM (1989) The carnivorous plants. Springer VG, Smith-Vaniz WF (1972) Mimetic relationships in- Academic Press, London volving the fishes of the family Bleniidae. Smithson Contrib Lloyd FE (1934) Is Roridula a carnivorous plant? Can J Res Zool 112:1Ð36 10:780Ð786 Thompson JN (1982) Interaction and evolution. Wiley, New York Marloth R (1925) The flora of South Africa. Darter, Cape Town Thompson JN (1994) The coevolutionary process. The University Pellmyr O, Leebens-Mack J, Huth CJ (1996) Non-mutualistic of Chicago Press, Chicago yucca moths and their evolutionary consequences. Nature Zamora R (1990) Observational and experimental study of a car- 380:155Ð156 nivorous plant-ant kleptobiotic interaction. Oikos 58:368Ð372