Austral Ecology (2018) , – Florivorous myrmecophilous caterpillars exploit an ant– plant mutualism and distract ants from extrafloral nectaries ESTEVAO ALVES-SILVA,1,2* ALEXANDRA BACHTOLD€ 1,3 AND KLEBER DEL-CLARO2 1Universidade do Estado de Mato Grosso, Rua Prof. Dr. Renato Figueiro Varella, Caixa Postal 08, CEP: 78690-000, Nova Xavantina, Mato Grosso (Email: [email protected]); 2Instituto de Biologia, Universidade Federal de Uberlandia,^ Uberlandia,^ Minas Gerais; and 3Universidade de Sao~ Paulo, Ribeirao~ Preto, Sao~ Paulo, Brasil Abstract Females of myrmecophilous butterflies tend to oviposit in plants visited by ant species that engage in stable associations with its larvae. In Banisteriopsis malifolia, caterpillars are attended by the same ants that feed on extrafloral nectaries. A conflict may arise when both the plant and caterpillars compete for ant attention, and ants are assumed to forage on the highest quality resource. By attending caterpillars, ants can be indirectly detri- mental to plant fitness because florivorous larvae feed intensively until pupation. In this study, we specifically investigated (i) whether the occurrence of facultative myrmecophilous Synargis calyce (Riodinidae) caterpillars in B. malifolia was based on ant species (Camponotus blandus or Ectatomma tuberculatum) and abundance; (ii) the monopolization of ants by the butterfly larvae and (iii) the florivory rates incurred by the caterpillars on inflores- cences. The abundance of S. calyce was six-fold greater in plants with C. blandus, compared to E. tuberculatum treatments. Caterpillars monopolized up to 50% of C. blandus on the plants, indicating that the resources offered by S. calyce were more attractive to ants than extrafloral nectaries. Florivory by riodinids incurred losses of almost 60% of flower buds. Myrmecophilous riodinids exploited an ant–plant mutualism by attracting aggressive ants that become larvae bodyguards. Thus, this ecological interaction is potentially detrimental to B. malifolia, since the ants, which can provide protection against herbivores, shift to provide defence for one of these herbi- vores. Key words: Banisteriopsis malifolia, Brazilian savanna, Camponotus, Ectatomma, Synargis calyce. INTRODUCTION (B€achtold et al. 2013; Torres & Pomerantz 2016). This timing seems to be important for the natural Many ant species are known for their mutualistic history of myrmecophilous butterflies because it interactions with extrafloral nectary (EFN) plants matches the period when potential specific ant-part- (Del-Claro et al. 2016). These ants, however, not ners occur on plants (Kaminski 2008). Depending only feed on plant-derived resources, but also on the on the ant species, caterpillars can be ignored, secretions of myrmecophilous insects, such as tended or preyed upon (Robbins & Aiello 1982; hemipterans and caterpillars (DeVries 1991; Sendoya DeVries et al. 1992). Therefore, adult females should et al. 2016). For instance, larval stages of Lycaenidae be able to evaluate host plants to maximize larval sur- and Riodinidae butterflies have so-called ant-organs vivorship (Ballmer 2003; Mota & Oliveira 2016) and (DeVries et al. 1992; Campbell & Pierce 2003), thus oviposit in plants with certain ant species to which release a nectar-like solution for ants that, in achieve this goal (Collier 2007; B€achtold et al. 2017). turn, faithfully attend the larvae (Kaminski & Car- A conflict arises when both the plant (with EFNs) valho-Filho 2012). In this mutualistic interaction, and myrmecophilous caterpillars compete for ant ants receive a valuable sugared food source (Cush- attention (DeVries & Baker 1989). Ants can shift man et al. 1994), while butterfly larvae are protected from foraging for plants to animal resources, depend- from natural enemies, especially parasitoids (Stadler ing on quality (Bluthgen€ & Fiedler 2004). Thus, a et al. 2003; Weeks 2003). shift from EFN to insect secretions can occur in In the tropics, florivorous myrmecophilous butter- some scenarios (Katayama et al. 2013). Indeed, it flies occur in a large array of EFN plants, and some- has been recently shown that myrmecophilous cater- times during specific periods of EFN activity pillars can chemically (through secretions from ant- organs) manipulate ant behaviour so that ants remain *Corresponding author. close to larvae (Hojo et al. 2014). In this context, Accepted for publication March 2018. caterpillars can monopolize ant attention (i.e. a high © 2018 Ecological Society of Australia doi:10.1111/aec.12609 2E.ALVES-SILVAET AL. abundance of ants are attracted to caterpillars, in Blooming and EFN activity are markedly seasonal and take comparison to the number of ants that forage on the place from January/February to April. Flower buds are plant) to get them close and serve as bodyguards. 8 mm in diameter, pinkish and grow in inflorescences Most studies on the interaction between ants and located at the apex of stems (Alves-Silva & Del-Claro myrmecophilous larvae have been performed with 2016). This plant is visited by a wide range of arthropods, of which oil-gathering bees and EFN feeding ants have lycaenids (see Fiedler 1991, 2001; Eastwood & Fra- € been the most studied (Rocha-Filho et al. 2008; Vilela et al. ser 1999; Pierce et al. 2002; Bachtold et al. 2014) 2014). Camponotus blandus and E. tuberculatum (Fig. 1a,b) and little is known about ant associations within are frequently found in B. malifolia (B€achtold et al. 2016). myrmecophilous riodinid caterpillars, despite recent These ants are known as plant guards as they engage in efforts in the neotropics (Kaminski et al. 2013). Since mutualistic associations with EFN plants, thereby reducing many riodinid caterpillars have ant-organs, they also herbivory rates (Oliveira et al. 1987). engage in stable associations, either facultative or Banisteriopsis malifolia also supports a rich fauna of obligate, with ants (DeVries 1988; DeVries et al. myrmecophilous caterpillars (Lycaenidae – B€achtold et al. 2004; Kaminski 2008; Kaminski et al. 2013). More- 2016), some of which are specialized in the family over, myrmecophilous riodinids tend to occur in Malpighiaceae (Kaminski & Freitas 2010). Information on the interactions between ants and riodinids are scarce EFN plants, most likely because the same ants that (DeVries & Baker 1989; Kaminski et al. 2013), but in feed on EFNs may also tend the caterpillars (Robbins B. malifolia (and in other Malpighiaceae) it is known that & Aiello 1982; DeVries 1988; DeVries et al. 1992; myrmecophilous metalmarks co-exist with ants (B€achtold Torres & Pomerantz 2016). et al. 2016), what suggests that larvae can be attended by In this study, we investigated some aspects of the the same ants that attend lycaenids and visit EFNs. The interaction between a facultative myrmecophilous rio- myrmecophilous metalmark S. calyce (Riodinidae) spends dinid caterpillar (Synargis calyce (C. Felder and R. its whole immature stage feeding on flower buds of B. mali- Felder, 1862), Riodinidae) and tending ants (Cam- folia (Fig. 1c,d). ponotus blandus (Smith 1858) and Ectatomma tubercu- latum (Olivier, 1792)) in the extrafloral nectaried shrub Banisteriopsis malifolia (Nees and Mart.) B. Sampling and conditions of study Gates (Malpighiaceae). We specifically analysed (i) whether riodinid occurrence was based on ant species A total of 60 B. malifolia individuals were tagged in the end fl (C. blandus or E. tuberculatum) and abundance, (ii) a of January, prior to the onset of the owering season. These shrubs (1.62 Æ 0.02 m tall, mean Æ SE) had the same phe- possible monopolization of ants by riodinid larvae nological conditions, such as presence of young foliage, fl fl and (iii) the orivory rates in in orescences with and absence of flower buds and presence of the target ants. without riodinids. We hypothesized (i) a higher Plants consistently supported either C. blandus or E. tuber- abundance of caterpillars in plants with C. blandus culatum throughout the whole study period (no ant species because ants in this genus are known to engage in overlap), and were tagged to account for two equal groups stable associations with myrmecophilous butterflies, of shrubs (n = 30 shrubs per group) with each ant species. that is, lycaenids (Kaminski et al. 2010), whereas In our study system, Camponotus and Ectatomma occur con- Ectatomma may prey on caterpillars (Robbins 1991); sistently in different individual plants, with no overlap of ii) that ants should be consistently concentrated in different ant species in the same individual plant, a natural € inflorescences with riodinids (ant-monopolization) trend noted in other Malpighiaceae (Bachtold et al. 2017). This permitted us to investigate if riodinids occurred more and (iii) that riodinids themselves should inflict high fl in plants with Camponotus or Ectatomma because no other orivory rates in plants (like their sister group Lycae- ant species were present on tagged plants. Neither ant pres- – nidae Badenes-Perez et al. 2010), compared to ence nor species were manipulated in B. malifolia. other herbivores. In the field, we examined the beha- Synargis calyce caterpillars occurred on the plants from viour of ants towards riodinids to see whether ants February to April. Whenever a larva was found, it was consistently attended caterpillars, what would be evi- monitored in the field until it reached approximately dence of stable ant–larvae relationships (i.e. myrme- 10 mm in length, when part of the study was performed cophily). (objectives ii, iii). This was done to standardize the data,