There is increasing evidence that the outcomes of mutualistic interactions between ants and plants bearing extrafloral nectaries (EFNs) is context-dependent. In particular, the total number, density and size of EFNs, as well as the abundance and identity of ants attending host plants, are considered as key factors determining the nature and strength of ant-EFN-bearing plant interaction. Although many previous studies have investigated context-dependency in ant–plant protection mutualisms mediated by EFNs, few have tested whether the protective behavior of ants varies among sympatric plant species. In this study, we performed a field experiment to investigate the predatory behavior of a dominant ant species (Camponotus crassus, Formicinae) among five EFN-bearing plant species growing in the Brazilian savanna. In particular, we studied the variation in the ant abundance, termite predation and time spent to find termites of C. crassus among the plant species, and further analyzed whether this variation could be related to the extrafloral nectar volume and sugar concentration of each plant species. We found that abundance and termite predation of C. crassusmarkedly varied among plant species. Specifically, C. crassus ants were significantly more abundant, active, and protective in Qualea multiflora, the plant species that produced significantly higher volumes and sugar concentrations of nectar. Overall, our results suggest that variation in extrafloral nectar volume and sugar concentration can result in plant species-specificity of defensive behavior of a dominant foliage-dwelling ant in the Brazilian savanna.

1. Study area, plant species and ant species

The fieldwork was conducted in the Ecological Reserve of the Clube de Caça e Pesca Itororó de Uberlândia (18º58’59”S; 48º17’53”W) in Uberlândia, Brazil. The reserve is located in a tropical savanna ecoregion (called Cerrado) (Oliveira & Marquis 2002) and has about 200 ha of cerrado sensu stricto vegetation (specific vegetation of Cerrado). The cerrado sensu stricto vegetation contains palm swamp areas, open areas with shrubs and small trees, and more enclosed areas with trees reaching up to 15 m in height (Del-Claro et al. 2019). In this region, there is a marked rainy season from October to March and a marked dry season from April to September (Alvares et al. 2013, Novaes et al. 2020, Calixto, Novaes, dos Santos, et al. 2021). The mean annual temperature varies from 18 to 28º C and annual rainfall varies from 800 to 2,000 mm (Ferreira & Torezan-Silingardi 2013).

For this study, we selected five of the most abundant EFN-bearing woody species in our study area (Appolinario & Schiavini 2002): Banisteriopsis malifolia Nees & Mart. (Malpighiaceae), Lafoensia pacari (A. St.-Hil.) (Lythraceae), Ouratea spectabilis (Mart.) Engl. (Ochnaceae), Qualea multiflora (Mart.) (Vochysiaceae) and Stryphnodendron polyphyllum (Mart.) (Fabaceae). All the plant species exhibit active EFNs on newly flushed leaves at the beginning of the rainy season (September-October) (Lange et al. 2013, Calixto, Novaes, dos Santos, et al. 2021) and largely vary in production (volume and sugar concentration) and structure (size and type) of EFNs (Table 1). Because plants continuously produce new leaves, EFNs are present on the same individual plants for months. Rainfall events rarely remove extrafloral nectar.

The EFNs of these plant species are frequently visited by Camponotus crassus (Formicinae) ants (Nahas et al.2012, Anjos et al. 2017, Lange et al. 2017, 2019, Pires et al. 2017). This ant species exhibits an aggressive behavior against insect herbivores (Fagundes et al. 2017, Lange et al. 2019, Sousa-Lopes et al. 2020), which usually results in lower levels of herbivory and higher fruit production compared to plants without ants (Calixto et al. in review). Due to its numerical abundance and aggressive behavior, C. crassus is the dominant ant species on EFN-bearing plants in the Cerrado (Nahas et al. 2012, Anjos et al. 2017, Lange et al. 2017, 2019, Pires et al. 2017). Only few other ant species (usually subordinate and docile ants) are allowed to forage on the same plant individuals where C. crassus monopolizes (Calixto et al. in review; Fagundes et al. 2016). Therefore, fitness (e.g. herbivory rate and fruit production) of EFN-bearing plants hosting C. crassus ants would largely depend on the protective behavior of this ant species. 

2. Data collection

In October 2017, i.e., during the rainy season and when most plants had active EFNs (Lange et al. 2013, Calixto, Novaes, dos Santos, et al. 2021), we selected 20 individuals of each plant species (n = 100 individuals) with similar height (1.5-2 m), separated by at least 10 m, and dominated by C. crassus ants. In each individual plant, we selected the first leaf of an apical branch without marks of herbivore damage. Between 09:00 h and 11:00 h, we added an alive worker of Nasutitermes sp. (Isoptera: Termitidae) collected from three nests to conduct a termite predation experiment. Termites were gently captured with the help of tweezers, and without hurting them, they were placed on the plants. The predation of termite baits has been widely used as a proxy of ant effectiveness against herbivores in many plant taxa (Saks & Carroll 1980, Oliveira et al. 1987, Oliveira 1997, Apple & Feener 2001, Anjos et al. 2017, Fagundes et al. 2017, Cruz et al. 2018, Pacelhe et al. 2019, Raupp et al. 2020). In order to avoid biases on encounter time, we only added termites on plants with at least one individual of C. crassus foraging on it, and when all C. crassus individuals patrolling the plant were at least 30 cm away from the selected leaf. 

After adding the termite, we waited for 30 sec until acclimatization and started the observation. We observed termites for a maximum of 15 min (if there was no predation event during that time window), and during this time we counted total C. crassus ant abundance on the whole plant, the time spent by C. crassus ants to find the termite, and the termite survival probability (1 = alive and 0 = preyed) or the termite predation (1 = preyed and 0 = not preyed) by C. crassus ants. We assume that there were no major changes in the total number of ants during the 15 min of experiment. Moreover, C. crassus ants usually do not remove termites from the plant (take and throw the termite out of the plant), but they prey on them (Lange et al. 2019). Therefore, the termite predation parameter was represented by the act of attacking and taking termites to the nest by ants. Because we aimed at evaluating the time spent by ants to find termites and (after finding them) the predation/removal rate, termites have to remain on the plants to be found by ants. If the termite fell from the plant, we discarded the observation and started again.