The Effect of Temperature Increases on an Ant-Hemiptera-Plant Interaction

Global temperature increases are significantly altering species distributions and the structure of ecological communities. However, the impact of temperature increases on multi- species interactions is poorly understood. We used an ant-Hemiptera-plant interaction to examine the potential outcomes of predicted temperature increases for each partner and for the availability of honeydew, a keystone resource in many forest ecosystems. We re-created this interaction in growth cabinets using predicted mean summer temperatures for Melbourne, Australia, for the years 2011 (23°C), 2050 (25°C) and 2100 (29°C), respectively, under an unmitigated greenhouse gas emission scenario. Plant growth and ant foraging activities increased, while scale insect growth, abundance and size, honeydew standing crop per tree and harvesting by ants decreased at 29°C, relative to lower temperatures (23 and 25°C). This led to decreased scale insect infestations of plants and reduced honeydew standing crop per tree at the highest temperature. At all temperatures, honeydew standing crop was lower when ants harvested the honeydew from scale insects, but the impact of ant harvesting was particularly significant at 29°C, where combined effects of temperature and ants reduced honeydew standing crop to below detectable levels. Although temperature increases in the next 35 years will have limited effects on this system, by the end of this century, warmer temperatures may cause the availability of honeydew to decline. Decline of honeydew may have far-reaching trophic effects on honeydew and ant-mediated interactions. However, field-based studies that consider the full complexity of ecosystems may be required to elucidate these impacts.


Introduction
Global surface temperature has increased by approximately 0.6°C in the past century [1] and this change is beginning to have a significant impacts on biological systems [2]. Specific mechanisms through which climate change, particularly temperature, affects species are complex. Temperature may influence species by directly altering population densities, survival and fecundity [3][4][5] or indirectly through its effect on interacting species, food sources, natural enemies, competitors and species traits [6][7][8][9]. Insects are likely to respond quickly to increasing

Study system
Temperature in Australia is predicted to increase significantly over the next 100 years under a range of different greenhouse gas emission levels and mitigation scenarios [33]. Here, we used the high greenhouse gas emission and no mitigation scenario (A1F1). Under this scenario, temperatures in Australia are predicted to increase approximately 3 and 6°C for the years 2050, 2100, respectively. We added these temperature increases to the 2011 mean maximum (24°C) for spring and summer (September-February) for Victoria (http://www.bom.gov.au/climate/ data/), to which the study species (described below) are native. In 2011, 2050, and 2100, the mean maximum expected temperatures for spring to summer in Victoria were predicted to be 24, 27 and 30°C respectively ( Table 1).
River red gum (E. camaldulensis) and sap-sucking scale insect (E. coriaceus) were selected as the model system based on their host relationship, geographical distribution and easy access to sampling honeydew. The river red gum is native to Australia and is widely distributed (http:// chah.gov.au/avh/public_query.jsp). The scale insect is found mostly in cooler regions of Australia where annual mean maximum temperature ranges from 12-24°C, although it has been reported from the tablelands in Queensland, where the annual mean maximum temperature is 27°C [34]. Field populations were collected from La Trobe University campus and from La Trobe University Wildlife Sanctuary. Eucalyptus camaldulensis seedlings were bought from Wimmera Native Nursery (36°26'16.21"S, 142°1'2.69"E) and grown for 3.5 months in 15 cm diameter plastic pots with native potting mix. During the first two months plants were infested with E. coriaceus by tying infested twigs and leaves onto the stems. Water was provided via the trays in which the plants were grown. Liquid fertilizer was applied to the plants during re-potting and two weeks before placing them in the growth cabinets (see experimental design).
Iridomyrmex rufoniger is native ant common throughout Australia, with colonies containing thousands of workers in interconnected nests [35]. Field colonies of I. rufoniger were collected from La Trobe University Wildlife Sanctuary using a battery-driven hand held vacuum cleaner and an aspirator. All nests were within an area of about 0.5 ha and were separated by approximately 20 to 160 m.
The ants were kept in the laboratory for 2.5 months prior to using them in the experiment. These colonies were kept in plastic containers (15 x 10 x 7 cm, l x w x h), with three to five 10 ml plastic centrifuge tubes as nest sites, each of which was one-third filled with water and stopped with cotton balls. The ants were incubated in a growth cabinet at 24°C (day) and 10°C (night) temperatures and a 12:12 hour light-dark cycle and were fed 25% sugar water and 2-3 freshly squashed crickets (Teleogryllus oceanicus) three times per week. Colonies from different nests did not interact aggressively with each other, which suggests that we sampled from a large, polydomous colony [35] or a supercolony.

Experimental design
To simulate expected temperature increases, this study was conducted in three growth cabinets varying in size, with each cabinet set to specified maximum (day) and minimum (night) temperatures (Table 1). Each cabinet was divided into treatments with ants and without ants to test the effects of ants and temperature on honeydew standing crop. Each ant treatment had eight potted E. camaldulensis placed in a tray, with one plant per pot (15 cm diameter). All plants had similar height across three temperature treatments (ANOVA; F 2, 45 = 1.891, P = 0.163) which ranged from 100-190 cm and had 1-182 E. coriaceus adults and instars. Temperature inside the growth cabinets was monitored with ibuttons set to record every 20 minutes. The experiment was run for 62 days. Ant treatment consisted of five I. rufoniger colonies per growth cabinet, with each colony consisting of 500 workers. However, there was shortage of queens so only three colonies had a queen each while the other two colonies had 200 brood each initially and were supplemented with 200 brood twice during the course of experiment to maintain similar brood levels amongst colonies. Placing brood in queenless colonies may allow workers to perform normal colony functions [36]. The colonies were placed in smaller plastic containers of the same size and treatment described earlier, except that they had up to six 1 mm exit holes on one of the sides. The holes were large enough to allow access of workers but not the queens. These small containers were placed inside a larger plastic container (18 x 11.5 x 8.5 cm, l x w x h) and placed next to the growth trays. Ant access to trees was facilitated by connecting the larger plastic containers with plastic tubing (4.9 mm diameter). One end of the tube was tied to the plants~3-5 cm from the pot surface. Wooden spatulas were used to bridge the pots so that all the five colonies had access to all the plants as different colonies were observed not to interact aggressively with one another.
A cardboard barrier with holes that fitted securely around each pot was fitted to each tray to prevent ants from drowning in the water. Then the sides of the trays were sealed with duct tape and greased with tanglefoot to prevent ants from escaping. A small plastic container with the bottom removed was fitted to one of corners of the tray for watering. Four freshly squashed crickets and 4-6 g of scrambled egg were provided twice per week as protein source for the ants.

Observations
Eucalyptus growth and primary productivity. Effects of temperature and scale insects on the growth and primary productivity of E. camaldulensis were examined by measuring plant height and biomass. Height was determined by measuring the growth of main stems and branches. Regions close to the meristem were marked with a permanent marker and the height to the meristem was measured and counted before and after the experiment. The difference in height indicated growth of the trees. The average height gained across a single plant was used as the response variable in the analysis. The plants were harvested at the end of the experiment and oven dried at 55-60°C for 48 hours to determine their biomass. The roots were thoroughly washed with water and pre-dried at 30°C for 30-60 minutes to remove excess moisture before including them with rest of the plant material. It is difficult to measure plant biomass before the experiment. However, gain in biomass is related to growth in many Eucalyptus species [37]. Therefore, we assumed that height gained provides an indication of biomass gained at the end of the experiment.
Scale insect population dynamics. Per capita growth rate, per capita instar production and size (head body length of adults) of the scale insect were used as measures of population dynamics. The effect of temperature on the growth rate of scale insects was examined by measuring per capita instar production and per capita growth rate. Per capita instar production was the final number of instars divided by number of adult scale insects at the end of the experiment. Per capita growth rate was the final number of adult scale insects divided by the initial number of adult scale insects. Instars and adult scale insects were easily distinguished as felted sacs are present only in adults [28]. The effect of temperature on scale insect size was determined by measuring body length (tip of the head to end of anus ventrally) of the adults. The number of individuals used to obtain mean measures varied from 2 to 5 individuals, depending on the number of scale insects available per tree. The measurements were taken at the end of the experiment to avoid altering feeding activity and reproduction. We used a stereo microscope (Leica 1 M165C, Switzerland) with camera attached to take live measurements using Leica Application Suite (LAS) imaging software (Version 3.4.0). Felt sacs were removed before taking the measurements.
Honeydew standing crop. To test for the effect of temperature and ants on honeydew standing crop per tree and scale insects, we sampled honeydew present on the anus of the scale insects using filter papers [38]. Honeydew standing crop is the mass of honeydew available at any given point in time which can be affected by consumers and the environment [22,31]. Honeydew was removed from all scale insects one hour prior to sampling to allow us to determine the mass of honeydew available on a tree per hour. Honeydew was sampled at 15 minute intervals for 60 minutes per growth cabinet once or twice per week for eight weeks. Honeydew on the anus of the scale insects was absorbed onto small pieces of cut filter papers that had been pre-weighed and oven dried at~45°C for 24 hours. Each piece of filter paper was kept separate to avoid fluids being exchanged and care was taken to reduce smearing on storage containers. Filter papers were again oven dried at~45°C for 24 hours and re-weighed using microbalance (Mettler Toledo 1 XS3DU, Switzerland, accurate to 0.0001 mg). The difference in the mass of the filter paper provided a measure of honeydew (dry mass) available per hour. Honeydew standing crop per insect was determined by dividing honeydew dry mass per plant by the total number of scale insects on that plant.
Honeydew harvesting and ant activity. To test for the effect of temperature on the mass of honeydew harvested, we weighed six ants (per cabinet) observed tending the scale insects or feeding on honeydew that had distended gasters and another six ants exiting the nest or moving around on the surface of the plant pots without distended gasters using a Mettler Toledo 1 microbalance. The difference in mean weights between these groups provided a measure of the honeydew harvested by an average worker. The ants were aspirated into a small pre-weighed plastic vial with the top half of the vial painted with fluon (to prevent ants escaping) and the weight of the pre-weighed vial was subtracted from the weight of the vial with ants to determine the weight of the ants. Aspirating the ants may have provoked them to release defensive or alarm chemicals, potentially reducing their weight. However, we have no reason to expect differences in this response between ants with empty or distended gasters. All ants were kept in vials to avoid re-sampling until sampling was completed and then returned to their cabinets. Observations were made within 30-60 minutes per cabinet once or twice weekly for eight weeks. However, in the cabinets at 23 and 25°C, less than six or no ants were observed harvesting honeydew on any one tree, which limited the number of ants sampled per tree. Therefore ants were sampled across the trees rather than per tree in all the cabinets.
The effect of temperature on ant activity was measured by counting the number of ants ascending/descending trees per minute for 60 minutes per cabinet once per week for five weeks. The number of ants was counted across all trees because in growth cabinet 29°C there were many ants (78-238) that moved very fast and took more than one minute to count ants per tree. This made it difficult for one observer to count ants on individual trees. Observations were made by one person spending an hour at each cabinet each sampling period.
This study was conducted with permission from Department of Environment and Sustainability, Victoria under license number 10005518 and with permission from La Trobe University Wildlife Sanctuary. For E. camaldulensis no specific permissions are required as it is widely sold in native nurseries, while I. rufoniger is not endangered or protected species.

Statistical analysis
A log-linear model based on the Poisson link function was used to test for the effect of temperature, scale insect abundance and ants on the growth of Eucalyptus followed by Tukey contrast test. Scale insect abundance was included as another predictor in the model because scale infestation can reduce the growth of Eucalyptus [39]. Eucalyptus camaldulensis biomass and E. coriaceus growth rate and brood production data were analysed with two-way ANOVA, followed by post-hoc Tukey tests. Honeydew standing crop data were analysed with linear mixed-effects regression (LMER) based on a REML (residual maximum likelihood estimation) approach. Some trees had no scale insects thus no honeydew. Such unbalanced data with repeated measures are appropriately analysed with REML than traditional ordinary least square methods [40,41]. Ant attendance was used as another predictor because attending ants can have positive effects on the survival of sap-sucking insects [42] and honeydew production [43]. Data were log, square root or fourth root-transformed to meet the assumptions of normality and homogeneity of variances for ANOVA tests. All analyses were conducted using R 2.15.2 statistical software [44]. Linear mixed-effects regression was implemented with lme function in nlme package. Ant harvesting of honeydew and activity data were not analysed statistically because observations were not replicated across experimental units (i.e. plants). Daily mean temperature in each growth cabinet was slightly lower than the predicted temperatures (see S1 Predicted Temperatures). Therefore, all the analyses were based on actual temperatures in the growth cabinets rather than the predicted temperatures.

Eucalyptus and scale insect performance
Temperature had a significant effect on the growth of trees but not scale insects, ants or their interactions (Table 2A). The significant effect of temperature on trees was mainly due to trees gaining more height at 29°C than 23 and 25°C according to Tukey contrast test (Fig 1A). The mean dry weight of trees at 29°C was slightly higher (21.48 ± 2.26) than 23 (19.09 ± 1.26) or 25°C (17.39 ± 1.02) but these differences were not statistically significant (Table 2A). Temperature again had a significant effect on scale insect growth rate, instar production (Table 2B). A post-hoc Tukey test showed that these effects were due to a low growth rate at 29°C compared with 23 and 25°C Fig 1B). There was no effect of ants or combine effect of ants and temperature on scale insect growth rate and instar production. However, interaction between ants and temperature had a significant effect on the size of scale insects (Table 2B). Scale insects with ants at 29°C were significantly smaller than those with no ants and those in both ant treatments at 23 or 25°C according to post-hoc Tukey test (Fig 2). In general, all the indicators of scale insect population growth such as per capita growth rate (Fig 1B), showed reduced performance with increasing temperature.

Honeydew standing crop
Honeydew standing crop per insect was significantly affected by ants (df = 42, t = -2.570, P = 0.012, Fig 3A), but not temperature (df = 2, t = 0.166, P = 0.884, Fig 3B) or their interaction (df = 40, t = -1.444, P = 0.156). Ants removed most of the honeydew available per scale insect in comparison to treatments with no ants (Fig 3A). However, there was a significant main effect of both ants and temperature on honeydew standing crop per tree (ants, df = 42, t = -3.328, P = 0.002; temperature, 42, t = -4. 868, P<0.001). Again, there was no significant interaction effect (df = 42, t = 0.770, P = 0.445). That is, less honeydew was present in the treatment with ants than in the treatment with no ants (Fig 3C), indicating that ants were removing honeydew. Along the temperature gradient, honeydew standing crop per tree at 29°C differed significantly (df = 42, t = -4. 868, P<0.001) from both 23°C and 25°C, but there was no difference between 23 and 25°C (df = 42, t = -0.891, P = 0.378). The lower honeydew crop per tree at 29°C is the result of negative effect of elevated temperature on size and number of scale insects per tree (Table 2B).    (Fig 4A). We also observed higher ant activity (ascending/ descending) at 29°C than 23 and 25°C (Fig 4B).

Discussion
We manipulated one climate variable: temperature, and observed responses of each partner in an ant-scale insect-plant interaction. The results, summarized in Fig 5, show strong effect of increasing temperature on the growth of plants and scale insects, honeydew standing crop per tree and harvesting by ants but not on honeydew standing crop per scale insect. These findings allow us to predict how each partner in the ant-scale insect-plant interaction may respond to increasing temperature. We also discuss how increasing temperature may affect honeydew production and availability and flow-on effects on ants that primarily feed on liquid carbohydrates.

Eucalyptus and scale insect performance
Elevated temperature (29°C) positively influenced the growth of E. camaldulensis, which is consistent with findings of other studies [45,46]. The high growth of E. camaldulensis at elevated temperature is not surprising because this species thrives in many habitats and temperature ranges. E. camaldulensis is thus likely to be adapted to cope with a broad range of temperatures. Such wide ranging species provide a readily available host for sap-suckers such as E. coriaceus. However, when temperatures are warmer during spring and summer, an increase in E. coriaceus infestation levels can affect the growth of plants [38]. Although we found no significant effect of scale insect infestation, we observed one plant heavily infested with scale insects dying at 25°C, suggesting negative impacts of infestation. However, lethal effects of elevated temperature on E. coriaceus annulled the negative impacts of scale insects on E. camaldulensis growth at 29°C, which on average had higher growth (31 cm) than at 23°C (13 cm) or 25°C (13 cm). An increase in growth with temperature should result in more C being stored as biomass [45], while scale insect infestations would reduce biomass. This means that plants growing at 29°C should have gained more biomass than plants growing at 23 and 25°C. However, the increase in biomass gain at 29°C was not significant, possibly because this temperature is close to the physiological limits of E. camaldulensis [47,48]. Biomass in leaves, stems and buds could be lost through the scale insects as these are the parts of the plant where sap-sucking insects commonly feed [38,45]. However, low scale insect density at elevated temperatures means biomass loss through herbivory by scale insects is probably not significant. Among many factors that interact to regulate the population dynamics of sap-sucking insects, seasonal temperature variation is a key factor. Erioccocus coriaceus density is highest when the temperature is lower (late winter and spring) [29]. The low abundance of E. coriaceus during summer is linked to low fecundity of females and high predation rates [29]. We found that elevated temperature led to reduced population density of E. coriaceus and slowed its growth rate (Fig 1B) while combine effect of temperature and ants reduced E. coriaceus body size. The body size-fecundity relationship was observed in earlier studies, where larger female E. coriaceus produced more instars than smaller females [28,29]. Small size and low growth rate at elevated temperature indicates poor nutrition or stressful physiological conditions. Ants can encourage sap-sucking insects to increase honeydew production [43,49]. With less honeydew available for the ants at 29°C, ants are likely to encourage E. coriaceus to produce honeydew more frequently than they should. Although not significant this, high honeydew availability per scale insect at 29°C compared to 23 and 25 suggest influence of ants on honeydew production (Fig 3B). If ants are encouraging E. coriaceus to produce honeydew more frequently, E. coriaceus may be able to afford less time to assimilate nutrients from the sap. Besides, climate warming is associated with higher CO 2 , which is likely to affect nutritional value of the sap [47,48]. However, we did not measure CO 2 in this study and the combined effects of higher temperature and CO 2 on scale insects and honeydew production may differ from their individual effects. Instead, we found that elevated temperature increased plant biomass (Fig 1A), which we would expect to increase scale insect feeding rates, resulting in more honeydew being produced at 29°C. However, this was not the case (Fig 3), suggesting that scale insects did not feed well at elevated temperatures. Elevated temperature beyond the optimum Effects of Temperature on an Ant Interaction range negatively affects insect physiological functions [50][51][52][53]. Therefore, at 29°C, temperatures may have been beyond the limits for the scale insects to function normally, directly affecting their size, growth and feeding behavior.

Honeydew standing crop, harvesting and ant activity
We expected temperature and the presence of ants to increase honeydew standing crop per insect. However, we found no effect of temperature, which can be partly explained by the feeding behavior of the scale insects. In the aphid Tuberolachnus salignus, honeydew excretion at 20°C was close to rate at which sap was forced through the stylets by the turgor pressure of the plants [54]. Scale insects are therefore likely to be feeding at their maximum rate at 23°C, such that increasing temperatures to 25 and 29°C does not allow greater ingestion of sap by scale insects. When we considered the total honeydew standing crop per plant, we found negative effects of elevated temperatures (Fig 3C). This was expected, since scale insect density and growth were higher at lower temperatures so the mass of honeydew available was directly related to scale insect density. It is likely that high instar production and population density at lower temperatures resulted in greater honeydew production than elevated temperature. Favourable temperatures may have also increased the flow of phloem sap by reducing sap viscosity or by increasing the turgor pressure of plants [48]. Since ants attending sap-sucking insects can enhance honeydew production by rapidly stroking the posterior end of the sapsucking insects with their antennae [43,49], honeydew production might have been higher in treatments with ants. However, we did not measure this, so can only comment on the observed standing crop per tree.
Our data suggest a strong effect of elevated temperature on the mass of honeydew harvested by individual ants, but a lack of independent replication for this component of the study limits interpretation. At elevated temperatures, honeydew availability was low, so less honeydew was harvested per ant than at lower temperatures ( Fig 4A). Low honeydew availability at 29°C meant individual ants carried very low loads on average (0.09 mg), compared with ants at lower temperatures (23°C = 0.25 mg, 25°C = 0.15 mg). When resources are scarce, colonies must somehow maintain their energy intake. This might have been achieved by colonies increasing their foraging activities by recruiting more foragers to the resource. We observed high ant activity (ascending/descending) per minute at 29°C, suggesting that ants were removing most of the honeydew. However, at lower temperatures, where honeydew availability was high and ant activity was low, individual ants carried greater loads on average than ants at 29°C. It seems that ants were behaving in such a way to maximize their energy intake per unit time [55]. We suggest that it may have been less energetically expensive for ants at elevated temperature to forage.
Although honeydew availability depended on ant presence and temperature, it is possible that honeydew quality may also have been affected. High temperature [56], ants and host plants [57] can alter the nutritional content of honeydew. Changes in quantity and quality of honeydew may also affect ant colony nutritional requirements and ant preference for sap-sucking insects [58]. Honeydew quantity and quality may play an important role in defining ant community organization with dominant ant species dominating nutritious honeydew at the expense of less competitive species [25,59].
If climate change in Australia follows current predictions, our study suggests that E. coriaceus will do well and honeydew availability will remain high between 23 and 25°C (Fig 5). However, before or by end of this century (2100), we predict that the warmer climate ( 29°C) may lower the availability of honeydew, affecting liquid feeding ants such as Iridomyrmex and their interactions. However, we focussed on a scale insect species that is limited to cooler regions and was negatively affected by increasing temperature. The loss or decline of sap-sucking insects in cooler regions may reduce the amount of honeydew available to tending ants, unless these losses are supplemented by other species better adapted to warmer climates. Fieldbased studies that consider the full complexity of ecosystems may be required to elucidate these potential impacts.