Variation in Acacia drepanolobium defence by Crematogaster mimosae and a previously undescribed acacia ant at Hell’s Gate National Park, Kenya

Variation in Acacia drepanolobium defence by Crematogaster mimosae and a previously undescribed acacia ant at Hell’s Gate National Park, Kenya

Clare Elliott, University of Nottingham, United Kingdom
Christina Ieronymidou, Imperial College London, Cyprus
Evelyn Fosuah, Kwame Nkrumah University, Ghana

Abstract
An as yet unidentified brown ant species (right in photo above) associated with Acacia drepanolobium, found in Hell’s Gate National Park, Kenya, was compared to the well known acacia ant Crematogaster mimosae (left in photo above) in terms of tree guarding behaviour against herbivory. Baseline information was also collected on its activity patterns and pseudogall occupancy. The brown ants were found to be significantly less aggressive than C. mimosae, and were also present at a lower abundance on the acacia trees. The location at which the trees were artificially ‘browsed’ affected the level of recruitment of ant guards for both species, with more ants being recruited at basal locations rather than at shoot tips, suggesting an effect of proximity to inhabited pseudogalls. Brown ants were found to occupy trees that were taller and healthier than C. mimosae trees, but the nectaries of brown ant acacias were significantly smaller that those of trees hosting C. mimosae. The brown ants appear to be poor defenders of A. drepanolobium against megaherbivores, and their place within the ant succession on A. drepanolobium is still unclear, as are their effects on tree trade-offs in plant resource allocation to defences.

Introduction
Acacia drepanolobium, locally known as whistling thorn, is one of the dominant tree species found in the climax vegetation of the savannah ecosystem in the Kenyan Rift Valley, here studied in Hell’s Gate National Park, Naivasha. A. drepanolobium is an important source of food for many browsing megaherbivores. It has developed a number of strategies for defence against such damage from herbivory; one is to intersperse its leaves with long spines, and another is to play host to ants who defend the tree by attacking and irritating any herbivores who try to eat the leaves. The tree produces domatia, also termed pseudogalls, which provide a home for the ants, and nectaries on the leaves, which provide a source of food (Coe and Beentje, 1991).

Four species of obligate acacia-ants have been studied on A. drepanolobium (Palmer et al., 2000). Crematogaster sjostedti is completely black in appearance. C. mimosae has a red head and thorax and a black abdomen (RRB). C. nigriceps is black except for a red abdomen. The Crematogaster ants are often characterised by raising their abdomens when alarmed, and are also known as the ‘cock-tail’ ants. Tetraponera penzigi is the fourth species of ant to be studied in association with A. drepanolobium and is also completely black but is much longer and thinner than the Crematogaster ants. Previous investigations into the interactions between A. drepanolobium and ants carried out in Hell’s Gate National Park on past Tropical Biology Association (TBA) courses observed only three ant species in 1997, C. mimosae, C. nigriceps, and Tetraponera penzigi (Ahmed and Leturque, unpublished data; Burston and Jelnes, unpublished data) and in 2000 four ant species were observed (the aforementioned and C. sjostedti) (Swainson and Mucha, unpublished data).

These ants are mutually exclusive, and as such only one species of ant occupies a tree at any one time. There is variation in the ecologies of these different ant species; some colonies live on more than one tree at a time, whereas other ant colonies, such as T. penzigi, complete their entire ecology on the tree and never need to leave it. There is considerable variation in the aggressiveness and therefore the effectiveness in tree defence between the different ant species. It has been suggested that there is a succession of ant species on acacia trees, with the relatively subordinate ant T. penzigi being an early successional ant, and is evicted by the more dominant Crematogaster spp. Young et al. (1997) observed the eviction of T. penzigi and also noted a significant relationship between host tree height and resident ant species, with taller and presumably older trees hosting later successional ants.

The findings of Young et al. (1997) showed that different acacia ant species have different effects and activities on A. drepanolobium trees. C. nigriceps is noted for its pruning behaviour, as it is the only ant associated with eating axillary shoots on the host tree. As such it is also thought to be an early successional ant, and it has been hypothesised that this pruning behaviour may avoid the host tree growing into contact with other acacia trees potentially hosting rival ant species. Nectaries are destroyed by T. penzigi, which may discourage the colonisation of the tree by the dominant Crematogaster spp. C. sjostedti and C. mimosae both tend scale insects on A. drepanolobium as a source of food. This variety of relationships between the different ants and the acacia trees hosting them is an example of coexisting diversity on an apparently uniform resource (Young et al., 1997).

The relationship between these obligate plant-ants and their host acacia is considered to be a mutualistic one, however the degree to which each of the parties benefits is subject to variability, especially considering the variation in the behaviour of the different ant species and their effects on A. drepanolobium. This can be seen where C. nigriceps prunes the tree to avoid conflicts with other acacia ants. Another studied example of the tree overcoming the apparent conflict between the ant guards and the trees own interests is that of acacia flowers repelling the resident ants for the period of pollen release to permit pollinators to visit (Willmer and Stone, 1997).

It has been hypothesised that a tree may alter its resource allocation to other forms of herbivory defence such as mechanical defence according to the effectiveness of the ants it is hosting. It has also been suggested that a tree could control the concentration of ants on particularly vulnerable parts of the plant by altering the nectar production at the nectaries around the tree.

At Hell’s Gate National Park another ant has been observed living on A. drepanolobium around the gorge entrance. This brown ant appears to be distinctly different from any of the four species mentioned above and, as such, is suspected to be a different species altogether. It is larger than the Crematogaster spp.; the abdomen is similar in shape but it does not cock its tail in the characteristic way of the Crematogaster ants.

In this project brown ants were studied to obtain some baseline data, since as far as can be seen in the literature, they have not been previously studied in their relationship with A. drepanolobium. A comparative study with C. mimosae was also carried out and the following hypotheses were considered:

Hypothesis 1: Crematogaster mimosae are more aggressive in tree defence than brown ants.
Hypothesis 2: Tree resource allocation to mechanical defence and ant benefits varies with ant species.

Materials and Methods
Location
Thirty-five Acacia drepanolobium trees were selected in the area around the gorge entrance at Hell’s Gate National Park, Kenya. Twenty trees inhabited by brown ants and fifteen trees inhabited by C. mimosae.

Ant behaviour
Ant aggression was measured and compared for both the brown ants and C. mimosae (15 trees for each ant species) by attacking each tree in two places, a branch tip and the base of a branch, using forceps to simulate browsing. Each tree was ‘browsed’ at a regular rate using forceps for 3 minutes. The number of ants within a distance of 10cm from the point of ‘browsing’ was counted before and after attacking the plant as a measure of ant recruitment. The time taken for the ants to attack the forceps by walking onto them was also recorded as a measure of ant aggression.

Brown ant activity on nine of the A. drepanolobium trees was assessed during the course of one day. Three branches were selected on each tree and a count of the total number of ants present on the distal 30cm was taken every half hour from 0900h to 1230h, and from 1430h to 1800h.

On a further five A. drepanolobium trees, three pseudogalls inhabited by brown ants were selected by observing ant activity around the entrance hole. The holes were plugged with modelling clay, and detached from the tree. Pseudogall contents were killed using ethyl acetate and opened up to count the number of ants in each.

Tree status
Data were collected from 30 trees, inhabited by each of the two ant species being studied, to ascertain the general status of the trees hosting the two ant species by assessing tree height and health (good, moderate or poor) based on number of growing shoots and general appearance. The density of leaves and spines was measured by counting the number present in a 20cm length of two randomly selected branches of each tree. Leaf and spine length were measured for the ten terminal spines and leaves of the two selected branches. Ten leaves were randomly selected from each tree and the length of the basal nectaries was measured along the midrib.

Analysis
The data were tested for normality using the Anderson-Darling normality test and transformations were used to achieve normality where possible. The standard error of the means was calculated throughout and is shown in the relevant figures. Appropriate statistical tests were conducted using Minitab and R statistical software.

Results
Ant behaviour

There was no significant difference between the time taken for the ants to attack at the tip of a branch or at a branch base within each ant species (brown - B or C. mimosae - RRB). Wilcoxon signed-rank test, for B W = 6, n = 15, p = 0.181, RRB W = 61, n = 15, p = 0.977. Therefore data were combined for the two ‘browsing’ locations on the tree (Figure 1). There was a significant difference between brown ant time to attack (median = 180s, n = 15) and C. mimosae time to attack (median = 96.5s, n = 15); Mann-Whitney test, W = 328.5, p<0.001. Hence, C. mimosae were more aggressive defenders of A. drepanolobium.




Figure 1; Time to attack as a measure of ant aggression (means; B 165.4s ± 7.82, RRB 91.1s ± 10.54, n=15).

Recruitment of ants to the point of attack differed significantly between the two ant species (analysis of deviance with quasi-Poisson error structure, t = 5.135, p < 0.001), with C. mimosae showing more recruitment than brown ants (Figure 2). The location of attack also affected recruitment, with more ants aggregating at basal locations (analysis of deviance with quasi-Poisson error structure, t = 3.472, p < 0.001) and more ants vacating shoot tips (analysis of deviance with quasi-Poisson error structure, t = 2.15, p = 0.036).



Figure 2; Difference in the number of ants before and after ‘browsing’ as a measure of ant recruitment (means; B tip 1.47 ± 1.10, B base 0.60 ± 0.45, R tip 1.33 ± 1.14, R base 6.73 ± 1.53, n=15).

Figure 3 shows mean brown ant activity on all trees surveyed. The numbers of ants on the branches surveyed varied appreciably from 0 to 38 and variation of numbers within branches was low, with standard errors ranging from 0 to 1.79. There was an overall trend of increased activity as the day progressed.




Figure 3; Mean number of Brown ants recorded on branches between 9:00h and 18:00h.

The mean number of ants found in the inhabited pseudogalls sampled from brown ant trees was 28.8 ± 5.69 (Figure 4).


Figure 4; Box and whisker plot showing number of brown ants found in 15 pseudogalls (mean 28.8 ± 5.69).

Tree Status
Tree height data were log10 transformed to give a normal distribution (B mean = 2.22m ± 0.35m, n = 15, RRB mean = 1.47m ± 0.09m, n =15). Trees hosting brown ants were found to be significantly taller than trees hosting C. mimosae (Figure 5), independent samples t-test, t = 2.160, d.f. = 19, p = 0.043.




Figure 5; Height of trees hosting Brown ants (B) and C. mimosae (RRB) (means; B 2.22m ± 0.35m, RRB 1.47m ± 0.09m, n=15).

Crematogaster mimosae tends to occupy trees of poorer health than brown ants (Figure 6; Fisher’s exact test, p = 0.040).




Figure 6; Health status of trees hosting Brown ants (B) and C. mimosae.

There is a small but significant difference in the size of the nectaries on trees hosted by brown ants or C. mimosae (Figure 7) B mean = 0.61mm ± 0.03mm, n = 15, RRB mean = 0.77mm ± 0.04mm, n =15, independent samples t-test, t = 3.130, d.f. = 28, p = 0.004.




Figure 7; Mean length of nectaries on trees hosting each ant species (means; B 0.61mm ± 0.03mm, RRB 0.77mm ± 0.04mm, n=15).

Leaf length, spine length and leaf density were not significantly different between trees hosting brown ants or C. mimosae, however there was a significant difference in spine density between ant species. Trees hosting C. mimosae have a significantly higher density of spines, B mean = 1.64cm ± 0.07cm, n = 15, RRB mean = 1.92cm ± 0.07cm, n =15, independent samples t-test, t = 2.910, d.f. = 28, p = 0.007.

Discussion
The brown ants were observed in considerably lower abundance on A. drepanolobium than C. mimosae and they were frequently absent from large parts of their tree, but relatively concentrated in other places such as on growing shoots or with scale insects. Observations of the pseudogalls and their contents confirmed the suspicion that these brown ants are living and reproducing on the tree, rather than just visiting. These results show that the brown ants are less aggressive against herbivory than C. mimosae, but both ant species seem to have similar strategies for recruitment to defence depending on proximity to ants on other branches and inhabited pseudogalls. However, there are differences in the characteristics of the host acacia trees between the two ant species.

Ant behaviour
According to Young et al. (1997), C. mimosae is the most aggressive ant guard of the four species found in Laikipia, Kenya, and the hypothesis that C. mimosae are more aggressive in tree defence than brown ants is strongly supported by these results. The brown ants almost always failed to attack the forceps throughout the 3 minutes the tree was ‘browsed’. C. mimosae were much more reliable in attacking the ‘browsing’ forceps (Figure 1). They were also the more numerous ant species as they showed greater recruitment of ants to the point of ‘browsing’ to defend the tree (Figure 2). More ants were recruited at basal ‘browse’ locations than at the tips of the branches (Figure 2). This is somewhat counter intuitive as it was expected that a more efficient ant guard would defend the more delicate growing tips more aggressively. The reason for this apparent paradox could be proximity to the rest of the colony, in terms of access to other pseudogalls; however, data relating to pseudogall location were not collected.

A single day of ant counts was not sufficient to elucidate the exact activity patterns of the brown ant and this was further compromised by the low abundance of brown ants on the acacia trees. Branches selected randomly on the tree were more likely to carry no or very few ants, and the numbers of ants on each branch remained relatively constant for the entire survey period. Therefore, selection of those branches carrying the most ants may have produced more informative results. There was a tendency for increasing ant activity during the course of the day, which is most likely due to increasing temperature. This was particularly evident between 9:00am and 10:30am (Figure 3). A large number of replications of such surveys, covering a range of weather conditions, would be necessary to determine the activity patterns of this ant species and the factors which affect them.

Pseudogall inspection confirmed that brown ants are an A. drepanolobium resident species (Figure 4), since larvae and eggs as well as adult ants were found inside the domatia. Inhabited galls were selected for this survey, as the brown ant abundance on the acacia trees is quite low. An estimate of colony size on a tree could be achieved by determining the proportion of inhabited galls however, considerable variation in the numbers of ants was found between pseudogalls. Consequently, the accuracy of a colony size estimate obtained in this way would be limited.

Scale insects were found inside the dissected pseudogalls and were also observed on the branches of trees hosting brown ants. The brown ants were seen tending the scale insects, presumably as a source of food. The effect of scale insect tending by ants on the tree could alter the cost benefit ratio for the ant-acacia association, as it poses an extra cost to the tree of supporting ant guards. Scale insects may be directly harmful to the acacia or may require a larger investment of resources. It would be interesting to compare the trade offs for the two different ant species since C. mimosae also tend scale insects inside their domatia. It is expected that A. drepanolobium trees will tolerate the higher cost of scale insect tending within the pseudogalls when the quality of defence against herbivory by the ant guards is high. Therefore, the better defender is more likely to be tending scale insects within domatia and is also expected to tend larger numbers of these.

Tree Status
Brown ants were generally found on taller trees with better health than the trees hosting C. mimosae ants (Figure 5; Figure 6). Young et al. (1997) suggest that the height of a tree can be an indicator of the age of a tree and therefore the height of the acacia hosting a particular ant species could be suggestive of the position the ant takes in the succession of ant species inhabiting acacia over the tree lifetime. Although less aggressive than C. mimosae in defending the tree against herbivory brown ants may be more successful in interspecific conflicts. Alternatively, the ants may take advantage of trees vacated by the more dominant Crematogaster spp. thus avoiding interspecific conflict.

Despite the better health of trees hosting brown ants, the nectaries were found to be smaller than those on trees hosting C. mimosae (Figure 7). This may be indicative of the quality of guarding afforded by hosting C. mimosae who were found to be much more aggressive against herbivory. An alternative explanation could be that nectary size is dependent on the levels of ant activity; since brown ant are less abundant than C. mimosae the average nectary size across a brown ant acacia would be lower. To test this, nectaries on occupied branches should be compared with those on unoccupied branches.

Stapley (1998) points out that there is an important interaction between thorns and ants as two methods of herbivore defence and that the combination of the two methods may be the most effective use of the trees resources. Since the succession of ants on acacia trees leads to relatively short term occupancy of a tree by any one species of ant (Young et al., 1997) the tree is unlikely to perceivably alter its resource allocation to mechanical defence mechanisms to match the guarding abilities of the different ant species in the time available. This may explain the lack of difference in leaf and spine length and leaf density between trees hosting the two ant species. However, trees hosting the less aggressive brown ants had a significantly lower density of spines suggesting less intensive herbivory. It would be necessary to study the differential levels of herbivory on the trees hosting the two species of ants and any effect this may have on the trees resource allocation. The hypothesis that ‘tree resource allocation to mechanical defence and ant rewards varies with ant species,’ therefore cannot be fully supported.

The brown ants studied here are suggested to be poor defenders against herbivory as they are sparsely distributed on A. drepanolobium and exhibit low aggressiveness. More research needs to be carried out to determine, more accurately, the ecology and biology of these brown ants to provide further insights into the mutualism between ants and acacias, and to identify and place this species more firmly in the succession of ants already described on A. drepanolobium in Kenya. More information on these relationships would help to elucidate the cost-benefit ratios and tradeoffs involved in the ant-acacia mutualism.

Acknowledgements
Thanks are due to Clive Nuttman and Graham Stone for their valuable advice and assistance, to Rosie Trevelyan, Kizito Masinde for the impeccable organisation of this TBA course and to Moses Kamoga and the staff at the Elsamere Field Studies Centre, and the Kenya Wildlife Service for logistical support and access to the National Park. Financial support was gratefully received from the TBA and the British Ecological Society.

References
Ahmed, H.M. and Leturque, H. (1997) Protection of A. drepanolobium from mammalian browsing by the association of its galls and ants. The Darwin Courses in Tropical Biology, Project Reports, Naivasha and Hell’s Gate, Kenya. TBA 97/3: 64-71.

Burson, P. and Jelnes, I.K.S. (1997) The relationship between A. drepanolobium and ants associated with the galls in Hell’s Gate National Park, Kenya. The Darwin Courses in Tropical Biology, Project Reports, Naivasha and Hell’s Gate, Kenya. TBA 97/3: 72-80.

Coe, M. and Beentje, H. (1991) A Field Guide to the Acacias of Kenya. Oxford University Press, Oxford.

Palmer, T.M., Young, T.P., Stanton, M.L. and Wenk, E. (2000) Short-term dynamics of an acacia ant community in Laikipia, Kenya. Oecologia 123: 425-435.

Stapley, L. (1998) The interaction of thorns and symbiotic ants as an effective defence mechanism of swollen-thorn acacias. Oecologia 115: 401-405.

Swainson, C. and Mucha, A. (2000) Aspects of ant-acacia interactions on Acacia drepanolobium in the Lake Naivasha area. Tropical Biology Association, Field Course Project Reports. TBA 00/1: 111-125.

Willmer P.G. and Stone, G.N. (1997) How aggressive ant-guards assist seed-set in Acacia flowers. Nature 388:165-167.

Young, T.P., Stubblefield, C.H. and Isbell, L.A. (1997) Ants on swollen –thorn acacias: species coexistence in a simple system. Oecologia 109: 98-107.