Ants contribute to pollination but not to reproduction in a rare calcareous grassland forb

Study system

Siberian spurge, Euphorbia seguieriana subsp. seguieriana Necker (Malpighiales: Euphorbiaceae), is a perennial forb, native to Eurasia. In Germany the species inhabits dry, calcareous or sandy grasslands, mainly in the basins of the rivers Rhine, Main, Ems, Nahe, Mosel and Unstrut (Senghas & Seybold, 2003). The plant is considered a threatened species and listed in the Red List of the German Federal Agency for Nature Conservation. Siberian spurge typically grows 15–60 cm tall with numerous stems branched from its base. It flowers from June until August and is protogynous, in which the stigma is receptive before the pollen is shed. The inflorescence is a typical cyathium, consisting of strongly reduced male and female flowers, bearing sickle shaped nectaries that offer easily accessible nectar. Functionally the cyathium can be considered as hermaphroditic flower and for simplicity will be termed “flower” here. The capsule fruit is on average 2.3 mm long and bears three seeds which are shed when the capsule dehisces (retrieved from http://www.floraweb.de, 2017).

Two ant species were observed to visit the nectaries of E. seguieriana. Formica cunicularia Latreille (Hymenoptera: Formicidae; Fig. 1) is a moderately thermophilic species and common throughout Europe where it prefers semidry habitats with a developed herb layer (Seifert & Schultz, 2009). Workers are 4.0–6.5 mm long and forage for nectar and honeydew; the species is also zoophagous (Collingwood, 1979; Novgorodova, 2015). Tapinoma erraticum Latreille (Hymenoptera: Formicidae) is a small black ant (2.6–4.2 mm) which lives in diverse habitats, including xerothermic grassland. The species is present in most of Europe; it has a carnivorous and aphidicolous lifestyle and feeds on nectar (Collingwood, 1979). The clypeal cleft is considered a morphological adaptation for exploiting liquids hidden in narrow spaces such as floral nectars (Seifert, 2016).

The field site was a dry calcareous grassland (480 m²) characteristic for the region of Lower Franconia. The site was situated within the boundaries of the botanical garden of the University of Würzburg, Germany (49°45′54.34″N, 9°55′48.28″E, 214 m above sea level), located on a slope of the Main river valley. The study population of E. seguieriana consisted of ca. 300 individuals and was chosen for its ease of access and to limit any negative impact by the experimenter on protected areas. Further populations of E. seguieriana exist on calcareous grassland sites along the Main river, which are integral nature reserves (e.g., Grainberg-Kalbenstein, Höhfeldplatte and Scharlachberg) of international significance with a high diversity of endemic and endangered species. Experiments were carried out from 2006–2008.

Frequency of ants as flower visitors

The frequency of ants and other insects visiting E. seguieriana flowers was assessed during a period of 10 observation days in August 2007. Each observation unit consisted of a bundle of rays with approximately 125 cyathia. Twenty of these floral units were observed twice daily (10:00–11:00 and 15:00–16:00) for 1 min per unit and all insects drinking nectar or carrying pollen were recorded. Observations were made at a distance of 80 cm from the inflorescences to avoid disturbance. Specimens of flower visitors were caught with a net after the observation period. Ants were determined to species level (Seifert, 1984; Seifert, 1996) while flying insects were identified to the level of family (Schaefer, 2000). Voucher specimens of flower visiting ant species were conserved in 70% ethanol and stored in the collection of the Department of Botany, University of Würzburg. The weather during each observation was categorized as sunny, cloudy or rainy.

Foraging behaviour of ants

The behaviour of individual T. erraticum and F. cunicularia ants on E. seguieriana flowers was compared to obtain more information about their role as potential pollinators. Two to three neighbouring stems of a single plant were marked with a small piece of tape to form one observation unit. Each observation unit consisted of 100 cyathia. Depending on plant size, one or two observation units per plant were designated. Individual ants were observed on 20 units distributed over 13 randomly selected plants for a period of 5 min and at a distance of 80 cm to avoid disturbance. The software The Observer 5.0.25 (Noldus Information Technology, Leesburg, VA, USA) was used to record numbers of visited cyathia, time spent consuming nectar and frequency of moving between stems of the same plant.

In a second experiment, the potential for outcrossing was assessed by recording whether individual ants exclusively foraged on flowers of the same individual E. seguieriana plant or whether other E. seguieriana in the vicinity were also visited. Six plants were randomly selected and all foraging ants were carefully marked (paint marker, Edding 750) with a colour dot on the thorax while walking on the flower (n = 290 T. erraticum, n = 120 F. cunicularia). A different colour was used for each of the six plants. The following five days, all E. seguieriana plants at the field site (approx. 300) were checked daily for the presence of marked ants during 10:00–14:00. The presence of colour-coded ants on the six experimental spurge plants and on other, unmarked E. seguieriana was recorded.

Effect of ants on pollen quality

The quality of pollen attached to the cuticles of foraging ants was assessed using the fluorochromatic procedure (FCR) according to Heslop-Harrison, Heslop-Harrison & Shivanna (1984). Ants foraging on E. seguieriana and carrying pollen were allowed to walk onto a petri dish with polytetrafluoroethylene (Fluon; Asahi Glass Co., Ltd., Tokyo, Japan) treated sidewalls to prevent escape and a microscope slide with a drop of FCR solution placed in the centre of the dish. The FCR solution was freshly prepared each day by dropwise adding fluorescein diacetate in acetone (2 mg ml⁻¹) to 2 ml of a 0.6 M sucrose solution until the solution became slightly cloudy. The trapped ant was held by the legs with insect handling tweezers, placed immediately into the solution with her dorsal side to dislodge all pollen, and was then released. A new microscope slide was used for each ant. Four to five slides were prepared at a time and immediately taken to the laboratory for testing of fluorescent and non-fluorescent pollen numbers using a fluorescence microscope (Leica DMR). The pollen load of 30 ants per species was assessed. As a control, pollen was collected directly from E. seguieriana anthers and tested as described above (n = 23 plants).

Effect of ants on plant reproductive success

An exclusion experiment similar to Schürch, Pfunder & Roy (2000) was carried out to assess the role of ants as pollinators. Plants in the bud stage were randomly selected and each plant was assigned to one of the following treatments (n = 10): (i) All visitors—plants were not manipulated, therefore all insects could visit flowers, (ii) No visitors—Flowers were enclosed by tying mosquito netting fabric around the stems, (iii) Winged only—flying insects could visit the flowers but crawling insects were excluded by driving a plastic cylinder (diameter: 24 cm, height: 24 cm) into the ground that surrounded the plant. The outer cylinder wall was painted with an insect trapping adhesive (Raupenleim grün; Schacht GmbH and Co. KG, Braunschweig), (iv) Ants only—flying insects were excluded by enclosing plants with cages made of mosquito netting fabric and bamboo stick frames. The cage was attached around a cylinder as in (iii) but a 5 cm gap between ground and cylinder rim was maintained to allow ants to access the flowers. Experimental plants were checked daily to ensure that the correct insect groups were excluded. In rare cases where non-target species were found, these were removed immediately. A subsample of five stems per plant (randomly selected and marked before the experiment) was assessed by counting the numbers of formed fruit capsules and unfertilized female flowers.

The following year the experiment was repeated with the treatments ‘Winged only’ and ‘Ants only’ (n = 9–10). Five stems per plant with approximately similar numbers of cyathia were marked. As soon as the capsules started to ripen and dry out, large empty tea filters (Cilia; Melitta Group Bentz KG, Minden) were tied around the stems. Fruit capsules were left to dehisce and the harvested seeds were then stored for four weeks at 10 °C in a sealed paper bag. Thirty seeds from each plant were sown into trays with seed mix and kept in an open greenhouse tunnel to germinate. After 10 weeks the number of established seedlings was counted.

Statistical analyses

For each of the individual plants, the numbers of visits by each ‘species’ (two ant species, F. cunicularia and T. erraticum, both individually and as a total of all ants, and Winged insects) were totalled over the 10 sampling days and over both morning and afternoon samplings. This total number of visits for each plant and species (and ants as a whole) was then square root transformed to convert it from a Poisson-distributed variable to a variable with an approximate Normal distribution. Four differencing variables were calculated for each plant: (a) visits by F. cunicularia vs. T. erraticum, (b) visits by F. cunicularia vs. Winged insects, (c) visits by T. erraticum vs. Winged insects, (d) visits by ants (in total) vs. Winged insects. For each of the four variables, a 95% confidence interval (CI) for the true difference in (square root of) number of visits between the two species was calculated. If this 95% CI did not include zero, this meant there was a 5% significant difference between the two species in visitation rate. In such a case, the level of significance of the difference was determined by calculating the 99% CI, 99.9% CI, and so on. Furthermore, for each of the 20 half-days (10 sampling days × morning/afternoon), the number of visits for each half-day and species (and the three species as a whole) was totalled over the 20 plants and square root transformed. These variables were then input into an analysis of covariance (ANCOVA) routine, with treatment terms being sampling day + morning/afternoon (main effects only) and covariate weather (with 0 = rainy, 3 = cloudy and 4 = sunny). The above four differencing variables were also calculated for each half-day and subjected to the same statistical analysis.

The time ants spent collecting nectar, the frequency of switching between stems, the viability of pollen and the germination rate of seeds were analysed by Mann–Whitney U tests. A Kruskal-Wallis ANOVA was used to determine differences in viability of pollen recovered from both ant species and the control.

For the first exclusion experiment overall analysis of variance was considered unsuitable due to the observed data distribution. Instead, the percentage of formed fruit capsules was calculated for each experimental plant; then, the 95% confidence interval for the true mean percentage was calculated for each treatment separately. The data were treated separately since the variability varied widely between the four treatments (two treatments were close to 100%, one treatment was close to 0%, while the fourth treatment had relatively high variability).

Ants as flower visitors

The two ant species that foraged on E. seguieriana accounted for 65% (F. cunicularia: 36%, T. erraticum: 29%) of all flower visitors and were the most frequent visitors (√(no. of ants) vs. √(no. of winged insects): 95% CI [1.867 ± 0.595], p < 0.001, Fig. 2). The remaining 35% of nectar foraging insects were winged insects of the orders Hymenoptera (Sphecoidea, Halictus) and Diptera (Conopidae, Tachinidae, Syrphidae, Ceratopogonidae). A comparison of flower visitation rates showed no significant differences between the three visitor groups (on the scale of √(total no. over 10 days): F. cunicularia vs. T. erraticum: 95% CI [1.100 ± 1.399]; F. cunicularia vs. winged insects: 95% CI [0.203 ± 0.643]; T. erraticum vs. winged insects: 95% CI [−0.896 ± 0.971]). No significant differences were also found between morning (10:00–11:00) and afternoon (15:00–16:00) visitation rates in the ANCOVAs for any of the eight variables (all species, individual species and pairwise combinations of species including ants vs winged insects, Fig. 3A). However, flower visits correlated positively with weather for the sum of the three species (p < 0.05), for F. cunicularia (p < 0.10) and for winged insects (p < 0.01), Fig. 3B. Lower numbers of insects were observed on rainy days and more were observed on sunny days.

Foraging behaviour of ants

Ants of the larger species F. cunicularia visited more than three times as many flowers during each observation period (Mann–Whitney U test, U = 25.5, p < 0.001, Fig. 4A) but spent only half as much time collecting nectar per flower compared to individuals of T. erraticum (Mann–Whitney U test, U = 154, p < 0.001, Fig. 4B). F. cunicularia was also observed to switch more frequently (1.0 ± 0.2 switches per observation) from one stem to another within the same plant than T. erraticum (0.2 ± 0.1 switches per observation; Mann–Whitney U test, U = 245, p < 0.001).

The marking experiment demonstrated that during a five-day observation period both ant species consistently returned to the same individual E. seguierana plant on which they had been marked at the start of the experiment. None of the marked ants was found foraging on any other spurge plant in the study population during the daily 4-hour assessments, therefore no statistical tests were carried out. Overall more F. cunicularia than T. erraticum ants were observed foraging on the plants on which they had been first caught, despite the fact that more T. erraticum had been marked initially.

Effect of ants on pollen quality

Ants captured on E. seguieriana differed significantly in the number of pollen attached to their cuticle (Mann–Whitney U test, U = 147.0, p < 0.001, Fig. 5A). On average, individuals of F. cunicularia had pollen loads six times higher than T. erraticum. However, in most cases (87%) the pollen sampled from F. cunicularia ants did not fluoresce and was considered unviable. The average viability of pollen recovered from all F. cunicularia workers was therefore 8 ± 4%. In cases where recovered pollen did show fluorescence (i.e., excluding zero values), the viability rate was 59 ±7%. In T. erraticum, 43% of captured ants carried viable pollen with an average viability rate of 30 ± 7%. In cases where recovered pollen did show fluorescence, the viability rate was 69 ± 8%. A Mann–Whitney U test indicated that numbers of viable pollen carried by ants was significantly greater for T. erraticum than for F. cunicularia (U = 321, p = 0.016, Fig. 5B). The fluorescence test further showed that pollen recovered from ant cuticles was less viable compared to pollen collected directly from unmanipulated male flowers (viability rate: 72 ± 4%, Table S4) (Kruskal-Wallis test followed by Dunn’s tests, d.f. = 2, H = 50.5, p < 0.001, F. cunicularia vs plant: p < 0.001, T. erraticum vs plant: p < 0.001).

Effect of ants on plant reproductive success

Fruit set in E. seguieriana depended on the presence and type of insect visitors (Fig. 6). When access to plants was unrestricted for all visitors, nearly 100% of flowers (95% CI [96.6 ± 3.5]) set fruit. Likewise this was found in plants pollinated exclusively by winged insects (95% CI [98.4 ± 2.5]) and therefore no significant difference (α = 0.05) in fruit set was observed between treatments ‘All visitors’ and ‘Winged only’. Where access was allowed to ants only, the percentage of flowers turning into seed bearing fruit was 37.2 ± 18.8% (mean and 95% CI) and therefore significantly lower compared with E. seguieriana that received flying visitors. Two ‘Ants only’ plants failed to produce seeds. When all insects were excluded, only 1.6 ± 1.5% (mean and 95% CI) of flowers set seed, significantly less than with all other treatments.

Significant differences were also found in the number of fruit capsules in the second year. ‘Winged only’ treatment resulted in an average (±SE) of 97.8 ± 0.2%, ‘Ants only’ in 22.7 ± 1.7% fruit set per stem (Mann–Whitney U test, U = 0.0, p < 0.001). Seeds from ‘Winged only’ also had a higher germination rate compared to ‘Ants only’ plants (Mann–Whitney U test, U = 0.0, p < 0.001, Fig. 7). All of the ‘Winged only’ plants produced viable offspring with approximately one third of seeds developing into seedlings. In contrast, only one of the ‘Ants only’ plant produced offspring (germination rate 6.7%). Seeds from all other plants failed to germinate.