Introduction

Plants and ants maintain a great diversity of mutualistic and antagonistic interactions with each other (Hölldobler and Wilson 2008). From the perspective of the plant, these interactions are usually associated with the indirect defense of vegetative and reproductive organs against herbivores (Rosumek et al. 2009), although there is a great debate about the effects of ants on plants (Oliveira 1997; Ohm and Miller 2014; Ibarra-Isassi and Oliveira 2018). The protective action of ants, however, can simultaneously involve several types of conflicts for plants in terms of pollination (Ness et al. 2006; Sendoya et al. 2009; Assunção et al. 2014; Ibarra-Isassi and Oliveira 2018; Villamil et al. 2019), exerting an important effect on the adaptive fitness of plants. Indeed, plant visitors may differentiate plants by the presence of ants, telling them apart from other organisms or even differentiating among ant species (Sendoya et al. 2009; Ibarra-Isassi and Oliveira 2018). It is often reported that ants negatively affect plants when interacting with flowers by reducing the frequency, duration, or diversity of flower visitors, especially other insects (Lach 2008; Hansen and Müller 2009; Junker et al. 2010; Gonzálvez et al. 2012; Souza et al. 2020). This may be attributed to factors such as their mere presence and aggressive behavior combined with chemical traces (Cembrowski et al. 2014; Li et al. 2014) and may lead to less effective pollination (Tsuji et al. 2004). However, there is evidence that the presence of ants on flowers can also benefit the plant if these insects repel less effective pollinators (Gonzálvez et al. 2012), floral herbivores (Oliveira 1997; Bleil et al. 2011), and seed predators (Higginson et al. 2010) or even act as primary pollinators in a few situations (Ibarra-Isassi and Sendoya 2016).

In recent decades, many studies on the interaction between ants and plants have been carried out in tropical forests and savannas (Rico-Gray and Oliveira 2007; Cembrowski et al. 2014; Belém et al. 2020; De Sousa-Lopes et al. 2020). Such studies have demonstrated that these organisms can form intricate networks that can be influenced by climate, latitude, forest cover, and landscape heterogeneity (Rico-Gray and Oliveira 2007). Conversely, temperate grasslands, spanning about 35% of our planet's surface, have been notably overlooked in understanding ant–plant interactions (Shantz 1954; Juárez-Juárez et al. 2023). An example of such landscapes is the Pampa, a prevalent vegetative cover extending across southern Brazil, Uruguay, and parts of Argentina, characterized by grasses, a few trees, and herbs (Fidelis et al. 2009; Pillar et al. 2009; Andrade et al. 2023). Eryngium (Apiaceae) is a genus of rosette-shaped plants that are particularly rich and abundant in temperate regions, comprising a total of 220 species, 100 of which are native to the Americas (Calviño et al. 2008). Due to their prominent presence in the Pampa’s landscape, these plants are used as food and shelter for a wide range of animals. Their complex architectural structure and dense inflorescences benefit a large number of arthropods (Campos 2010; Campos and Fernández 2011; Oleques et al. 2021). The overlapping leaves of some Eryngium species can form a water-harvesting rosette known to amplify arthropod diversity within ecosystems (Gonçalves-Souza et al. 2010; Fernandez-Barrancos et al. 2017). Our field observations indicate a large number of ants foraging on these plants. In particular, we have found several nests of Camponotus termitarius (Emery) ants close to patches of Eryngium chamissonis Urb. in the extreme south of the state of Rio Grande do Sul, Brazil. Ant visitation to this plant is likely driven by multiple factors, such as the presence of floral resources, insect trophobionts (e.g., treehoppers, whiteflies, or aphids), and the plant’s suitability as a microhabitat. Therefore, it is expected that ant presence or foraging behavior has consequences for other organisms associated with the plant.

Our study aimed to evaluate whether the presence of C. termitarius ants on the inflorescences of E. chamissonis affects the number of insect visitation on flowers. We also determined whether the ant’s effect on floral visitors may be derived from their physical presence only (in contrast to the presence of organisms other than ants) or from their behavior and mobility. We predicted that (I) plants with ants on their inflorescences would show a lower number of floral visitors when compared to plants without ants; (II) only the presence of ants and not other organisms on the inflorescences would have a negative effect on visitation. In other words, we expected the number of visits in plants with an organism other than ants to be higher when compared to plants with ants; and (III) the number of floral visitation in plants with active ants would be lower than in plants with dead pinned ants, suggesting that movement is an important visual cue for flower-visiting insects.

Materials and methods

Study site

This study was carried out between December 2022 and February 2023 in an area belonging to the Faculty of Veterinary Medicine of the Federal University of Pelotas, located in the city of Capão do Leão, Rio Grande do Sul (31° 80′ S–52° 41′ W). The predominant vegetation in this area is typical of the Pampa biome, including E. chamissonis and large nests of C. termitarius. This is a flat region with well-defined seasons and temperatures that can reach approximately 40 °C in the summer, as well as low rainfall, which is characteristic of this time of year. The study area is an occasionally flooded field, which is waterlogged in the winter and drier in the summer, typical of the regions south of the Coastal Plain of Rio Grande do Sul (Becker et al. 2007).

Study system

The plant E. chamissonis (Fig. 1a), known as Gravatá-do-banhado or Caraguatá (Fidelis et al. 2009), is restricted to flooded fields, usually forming dense clusters and conferring a typical appearance in moist areas (Joly 2005; Pillar et al. 2009). It is an erect rosette-shaped herb that stands out from other plants of genus Eryngium due to its larger and more robust size (Wanderley et al. 2005). It has rigid leaves with aculeate margins, and during the reproductive months (December to February), it produces dense inflorescences at the top of a rachis that can reach 4 m in height (Wanderley et al. 2005), which contain globular capitula of yellowish to light green colors with inconspicuous flowers (Cardozo et al. 2021; Oleques et al. 2021) (Fig. 1a, c, d). These plants in our study area are associated with C. termitarius ants (= C. punctulatus subsp. termitarius), which use the base of E. chamissonis leaves for building satellite nests, most likely as an adaptive strategy to environmental fluctuations caused by flooding (pers. obs.). These ants forage throughout the plant, including the inflorescences during the reproductive period, sometimes interacting with aphids (Fig. 1b). The pollinators of E. chamissonis remain unknown. Although Camponotus ants are often seen walking over its inflorescences, our observations suggest that they do not serve as primary pollinators of this plant. In fact, plants in the reproductive phase without ants were commonly observed in the study area, indicating that this interaction is facultative and further studies will test if C. termitarius can effectively pollinate E. chamissonis.

Fig. 1
figure 1

Study system in Brazilian Pampa. a General view of the vegetation at the study site with the presence of Eryngium chamissonis in the reproductive period; arrows indicate their inflorescences. b Camponotus termitarius workers tending hemipterans on the inflorescences of E. chamissonis. c Workers of C. termitarius foraging on E. chamissonis inflorescences. d Globular flower head of E. chamissonis

Full size image

Experimental design

First, we established a total of 17 buffers, each with a radius of 10 m. These buffers were used in this study as experimental blocks, which were distanced from each other at intervals of no less than 30 m. In each block, we documented the occurrence of E. chamissonis, with and without active inflorescences (indicated by the presence of the inflorescence scape). Additionally, we recorded the number of E. chamissonis individuals with satellite C. termitarius nests in the vegetative sector of the plant. To test our hypotheses, we selected four individuals of E. chamissonis per experimental block, spaced at least 1 m apart, with inflorescences at the same phenological stage and a minimum height of 1.7 m. We defined as the same phenological stage those plants that presented open globular capitula. Any plant that was serving as a contact bridge for ants to reach the inflorescences of the selected plants in each block was pruned to avoid bridges for the ants to pass from the surrounding vegetation to the inflorescences of the selected plant. The selected individuals were randomly assigned to one of the following treatment groups: (I) inflorescence without ant presence, (II) inflorescence with active C. termitarius ants (i.e., plants where ants were foraging on the inflorescences at the time of selection), (III) inflorescence with dry dead C. termitarius ants that had been pinned, and (IV) inflorescence with another dry dead and pinned organism (Ulomoides dermestoides Fairmaire, Coleoptera: Tenebrionidae). For treatments III and IV, we attached eight C. termitarius workers and eight U. dermestoides individuals, respectively, on to the inflorescences for each observation session, ensuring that they were positioned in all cardinal and collateral directions. To prevent ant access to the inflorescences in treatments I, III, and IV, we applied the sticky resin Tanglefoot® (Tanglefoot Co., Grand Rapids, MI, USA) at the base of the inflorescences 24 h before the start of the experiment. Throughout this period, we also removed the ants in treatment I and any other associated insects from the selected plants' inflorescences. We applied a small dab of Tanglefoot® resin in treatment II, which did not inhibit the passage of ants, but controlled potential chemical or manipulative effects of the resin on floral visitors.

Observation and evaluation of floral visitors

In each block, the mounted treatments were observed by the same observer from December to February, with only one observation carried out per block during this period. For each plant, we counted the number of branches present in each inflorescence. Subsequently, we began the experiments by observing the floral visitation of insects for a period of 15 min per plant and registering the number of insect visitations on the inflorescences. During this time, we observed the insects from a distance of 1 m for 10 min, and in the remaining 5 min, we approached the plant to observe smaller and/or associated insects. Associated insects largely refer to small coleoptera and cockroaches that during our observations fed on the inflorescences and were not easily identified at a distance of 1 m. We considered any insect that touched the inflorescence to be a floral visitor (hence, it counts as a visit). When a given insect touched/landed on another branch of the same plant, we counted this interaction as a new visit. The floral visiting insects were grouped into the following major taxonomic groups: bees (Hymenoptera: Anthophila), flies (Diptera), wasps (Hymenoptera: Apocrita, excluding Formicidae and Anthophila), beetles (Coleoptera), cockroaches (Blattodea), and Other groups during the observation period. Due to the immense morphological variation of Diptera, in our observations, we counted those organisms with a morphology similar to the main representatives of the suborder Brachycera (Schiner) as flies. Those insects that could not be identified due to their fast flight and/or small size were allocated as ‘Other groups’. All observations took place on sunny days between 10 a.m. and 3 p.m., totaling 17 h of observation. All treatments in the same experimental block were observed only once on the same day.

We performed three experimental comparisons within each block using different combinations of the selected plants. Our first experiment aimed to assess how the presence of any organism on the inflorescences influenced the floral visiting insect fauna. This was carried out by comparing the number of visits in treatment I with the number of visits in treatments II, III, and IV pooled together. In the second experimental comparison, we compared visitation between treatments III and IV to determine whether only ants (but not other insects of similar size) could influence the frequency of floral visitation by insects. The third experimental comparison aimed to determine whether ant mobility and foraging behavior affected the number of visits. In this case, we compared visitation between treatments II and III.

Statistical analyses

Premises

Considering the complexity of the study system, we verified whether the observed visitation to flowers was influenced by the number of plants in the reproductive stage in the block prior to the experimental comparisons. Generalized linear models assuming a negative binomial distribution of error were constructed for this purpose, using the number of floral visitations as a response variable and the total number of plants in the reproductive stage as a predictor variable. Since we did not identify significant relationships between these variables (Z = 0.23, P = 0.817, additional data are given in Online Resource Table 1), we did not consider the number of plants in the reproductive stage as a covariate in the subsequent models.

Testing our hypotheses

Experimental comparison data were analyzed using generalized linear mixed models (GLMMs). We used the Poisson distribution model, which was chosen based on the error dispersion and adequacy of the data in the models using the DHARma package (Hartig 2020). In these models, we compared the observed number of visitations of each group of insects between treatments. We used the visitation frequency of each group of visiting insects as the response variable and the treatments as fixed variables. The experimental block was taken as a random variable.

To facilitate the specific planned comparisons, we structured contrasts among the categorical predictor levels (treatment) before model fitting, following the methodology outlined by Crawley (2007). Three contrasts were constructed to encompass our specific planned experimental comparisons. The first contrast aimed to evaluate whether any organism attached to the inflorescences affected floral visitors (i.e., a contrast assuming positive values for treatment I without ant access and negative values for all other treatments). The second contrast aimed to assess the effect of C. termitarius on floral visitation against any other organism of similar size (i.e., a contrast assuming positive values for treatment IV with pinned U. dermestoides beetles and negative values for treatment III with pinned ants). The last contrast determined whether the behavior of live ants affected insect visitation (i.e., a contrast assuming positive values for treatment III with pinned ants and negative values for treatment II with live ants).

We built models for each category of the observed insect groups. We also built a model for the entire floral visitor dataset. All models followed the comparisons specified in the contrasts. We calculated the effect size for each comparison using the effect size package (Ben-Shachar et al. 2020). The modeling process was performed using glmer from the lme4 package (Bates et al. 2015). For other statistical inferences, we used likelihood comparisons via the Anova function from the Car package (Fox and Weisberg 2019) and graphical visualization with the ggplot2 package (Wickham 2016), all executed in R software version 4.2.2 (R Development Core Team 2023).

Results

After 1020 min of observation in 68 plant individuals, 2791 floral visits were recorded. The most frequently observed taxa were bees (42.8%), flies (37.9%), wasps (7%), and beetles (5.3%) (see Table 1 for more information).

Table 1 Number of visits to flowers of Eryngium chamissonis by each group of insects per experimental group
Full size table

Manipulation experiment

Our results demonstrated that plants without ants had a higher number of floral visits when compared to plants with any experimental organism (live ants, pinned ants, or pinned U. dermestoides). This result held for flies (Z = 4.04, P < 0.01, Fig. 2a), bees (Z = 9.26, P < 0.01, Fig. 2c), and cockroaches (Z = 3.78, P < 0.01, Fig. 2e), as well as when considering the whole group of visitors (Z = 9.42, P < 0.01, Fig. 2g; for additional details, see Online Resource Table 2). The remaining insect groups did not exhibit differences among the experimental treatments.

Fig. 2
figure 2

Results presented by the models built to investigate the effects of the proposed treatments on Eryngium chamissonis floral visitors. Each colored circle represents one of the 17 experimental blocks. The upper bars refer to the following experiments carried out: (I) comparison between plants without ants with the other treatments (the first contrast—Any organism), (II) comparison between plants with pinned ants against those with the pinned beetle Ulomoides dermestoides (the second contrast—Ant presence), and (III) comparison between plants with live, pinned ants (the third contrast—Ant mobility). The variation in gray, throughout the plots, represents the probability density estimated via Kernel, that is, the highest frequency of results predicted by the model, for each treatment and the asterisks refer to the significance values presented by the model (***P < 0.01; *P = 0.05)

Full size image

Regarding the specific presence of ants (compared with pinned U. dermestoides), we identified a lower number of visits to plants with pinned ants for bees (Z = 6.06, P < 0.01, Fig. 2c), flies (Z = 4.00, P < 0.01, Fig. 2a), and the whole insect community model (Fig. 2g). This result was not observed for the other insect groups. We also observed a negative effect on the total number of insect visits caused by ant activity (i.e., live ants) for the bee fauna (Z = 6.30, P < 0.01, Fig. 2c), beetles (Z = 1.95, P = 0.05, Fig. 2d), and other visitors (Z = 1.94, P = 0.05, Fig. 2g).

When examining the magnitude of effects observed in each experiment, the fly fauna was prominently affected by the presence of insects on inflorescences, especially by pinned ants (Fig. 3a). Conversely, bees showed a more pronounced response to the presence of live ants on plants compared to other contrasts (Fig. 3c), a trend also evident within the overall floral visitor community (Fig. 3g). In the case of cockroaches, the presence of any organism on the plants had a greater effect on this group of insects, causing them to visit the inflorescences less (Fig. 3d). Similarly, the same occurred for beetles and other groups, where the presence of live ants was the only factor that most influenced these groups of insects to visit the inflorescences less (Fig. 3e, f).

Fig. 3
figure 3

Standardized coefficients (x-axis) for each constructed contrast used for hypothesis testing (y-axis) in GLMMs. Any organism—comparison between inflorescences without ant access versus inflorescences with live ants, pinned ants, and pinned Ulomoides dermestoides. Ant presence—comparison between inflorescences with ants and pinned U. dermestoides. Ant activity—comparison between inflorescences with live ants and pinned ants. The asterisks refer to the significance values presented by the model (***P < 0.01; *P = 0.05) and 95% confidence intervals are shown by blue bars (absent when intervals are extremely low)

Full size image

Discussion

Our study revealed that the presence of the experimental insects on E. chamissonis inflorescences had a detrimental impact on floral visitors. Specifically, we observed a prominent negative effect of ant presence on the floral visitation by bees and flies, compared to the presence of the beetle U. dermestoides. The interactions between flowers and floral visitors may depend on the presence of potential antagonists on the flowers (Romero et al. 2011; Benoit and Kalisz 2020). These antagonists can alter visitor behavior, either through aggressiveness or resource competition (Grether et al. 2009; Benoit and Kalisz 2020). In this context, we highlight that potential competition between C. termitarius ants and floral visitors results in a decrease in the interaction of these organisms with the inflorescences of E. chamissonis.

The influence of C. termitarius on E. chamissonis floral visitors was not similar for all visiting insects. Wasps, for example, were not influenced by the experiments proposed in this work. These predatory insects are highly aggressive and as such may not feel threatened by the presence of other insects on E. chamissonis inflorescences. However, the influence of C. termitarius on the other visitors may be derived from the behavior of these ants, since the presence of live ants negatively affected bee visitation. This effect was also observed in the visitation by beetles and other groups with no identification in the field, extending to the entire community of visiting insects. This effect is likely to be linked to the foraging and aggressive behavior of C. termitarius, which has been observed to consume floral nectar, attacking visitors and intimidating other visiting insects. Furthermore, when we observe the results of the effects of ants on the community, this directly reflects the essentially concentrated effect on bees and dipterans that together represented a large part of the floral visitor fauna of E. chamissonis. Therefore, these two groups may represent the insects that are most affected by the behavior and presence of C. termitarius.

Although the number of studies on ant–flower–floral visitor interactions in temperate regions is limited, previous research has demonstrated the impact of various ant species on flower-visiting insects in other geographical locations, specifically tropical and subtropical areas (Gonzálvez et al. 2012; Villamil et al. 2019; Belém et al. 2020; Souza et al. 2020). Ants can compete with floral visitors through exploratory competition (consuming floral rewards and reducing resource availability) and interference competition (directly excluding pollinators from flowers) (Cembrowski et al. 2014). Our findings indicate that C. termitarius influences floral visitors to E. chamissonis through both mechanisms, particularly affecting bees and flies, which accounted for over 80% of the visits.

Exploring the effects of Camponotus termitarius on floral visits to insects

An extremely generalist pollination is reported for the Apiaceae family (Lindsey 1984), including other Eryngium species that occur in grasslands in southern Brazil (see Oleques et al. 2021). We suggest that bees and flies might be the primary pollinators of E. chamissonis, mainly based on their high frequencies and behavioral strategies (see Herrera 1987; Rodríguez-Rodríguez et al. 2013). The small and numerous flowers in the inflorescences of this plant with exposed nectar and pollen facilitate its use by these organisms, which are hampered by the presence of the ant C. termitarius.

For bees, both the presence and activity of ants were factors in these insects reducing the number of visits to inflorescences. In contrast, fly visitation was only reduced by the presence of ants on flowers, but not by their activity. Although few studies have directly assessed the role of ant behavior in flower visitation, decreased visits by bees and flies have been frequently observed (Romero et al. 2011). Ants are considered important predators in many cases (Cuny et al. 2021), which even generate negative effects on the reproductive fitness of plants by driving away or preying on pollinators (Ohm and Miller 2014; Benoit and Kalisz 2020). The perceived risk of predation is well known to influence wildlife behavior (Palmer et al. 2017; Prugh et al. 2019), creating a “landscape of fear” where species alter their behavior or avoid areas with a higher perceived risk (Laundré et al. 2010). Some studies indicate that pollinators can identify the presence of a predator on flowers (e.g., Romero and Koricheva 2011; Assunção et al. 2014) and the bees and flies observed here might interpret the presence of ants as a threat.

In the case of flies, viewing an ant's silhouette appears to be the key to modifying their behavior when trying to interact with flowers. We observed that this group showed less interactions with plants that had ants attached than with U. dermestoides, although they also exhibited less interactions when the plants had any organism on the inflorescences. As for bees, one of their foraging strategies is the chemical detection of occupied flowers, both to avoid predators and to identify a potential available resource (Leonard et al. 2011; Li et al. 2014; Su et al. 2022). In this sense, we demonstrate that bees can also be deterred from visiting flowers due not only to the silhouette of ants but also to the presence of live ants, which can chemically make flowers less attractive. Ants have glands in their bodies that secrete substances detectable by bees (Cembrowski et al. 2014). Furthermore, ants can reduce nectar availability by acting as thieves while occupying inflorescences (Lach 2008; Irwin et al. 2010). Miner and Wilson Rankin (2023) demonstrated that bees modify their foraging behavior due to the presence of ants, but mainly because they are capable of chemically detecting the presence of these insects. Thus, our results might reflect a difference between bees and flies regarding the identification of potential predators in inflorescences. Bees reduce their visits to inflorescences with live ants because they quickly perceive these insects, both by chemical and visual signaling, with the latter perception being the dominant cue for flies (Agrawal et al. 2014).

Final considerations

Our data demonstrated that individuals of E. chamissonis with ants had less floral insect visitation, potentially arising from the visual or chemical detection of ant presence by visitors or their active behavior. Considering the overall impact observed, future studies should investigate whether the presence of ants reduces the pollination and reproductive success of E. chamissonis. In a broader context, the interaction between these plants and C. termitarius could affect not only plant reproduction but also the acquisition of food resources by floral visitors. Given the study site's characteristics, inflorescences may be among the few available floral resources for visiting insect fauna. It is worth noting that the similar floral morphology of Eryngium species suggests that associations with ants in other contexts may yield comparable results (Cardozo et al. 2021).

Furthermore, understanding that the effects of C. termitarius on E. chamissonis inflorescences result from competition through interference or exploitation (Cembrowski et al. 2014), limited to two groups of floral visitors, prompts the question of whether the presence of this species and its active behavior on flowers influence the structure of the floral visiting insect community. Examining how groups of organisms interacting with E. chamissonis inflorescences are morphologically organized for ant identification is essential. For instance, does the morphology found in flies and bees contribute to the accurate identification of ants on plants, leading them to visit other inflorescences in the absence of ants? Therefore, potential studies that evaluate this interaction from a community perspective may provide crucial insights into the potential effects of ants on the community of floral visiting insects and the possible adaptations of these organisms in their interaction with E. chamissonis and C. termitarius.