Attachment Forces of Ants Measured With a Centrifuge: Better ‘Wax-Runners’ Have a Poorer Attachment to a Smooth Surface

The surfaces of many plant species are glaucous (‘waxy’), i.e. they are covered by a bluish layer of epicuticular wax crystals. Such wax blooms are known to have anti-adhesive properties and to be highly slippery for insects (e.g. Knoll, 1914; Stork, 1980b; Eigenbrode, 1996). We recently discovered some species of ant that are specialised ‘wax-runners’. These ants (members of the genera Crematogaster, Technomyrmex and Camponotus) live in association with Macaranga (Euphorbiaceae) host plants that have glaucous stems (Federle et al., 1997). In contrast to most other ant species, the mutualistic ant partners of these trees are capable of moving around on the slippery, glaucous twig surfaces with no difficulty (Fig. 1).

Thus far, nothing is known about the mechanism of this striking ‘wax-running’ capacity. Federle et al. (1997) found a tendency for smaller ants to have a better wax-running performance, an effect easily explained by scaling laws (lower weight/surface area ratio in smaller animals). In other words, a high ratio of adhesive area to body weight (e.g. smaller ants with relatively larger attachment organs) could provide a simple explanation for the observed ‘wax-running’ capacity. If this explanation were true, however, good ‘wax-runners’ should also attach well to a smooth surface. Here, we test this hypothesis by comparing the attachment forces to a smooth Perspex surface of wax-runners and closely related ants that are unable to climb up waxy twigs.

The capacity of many insects to hold on to smooth surfaces has fascinated biologists for centuries (e.g. Hooke, 1665; Rombouts, 1883; Dahl, 1884; Dewitz, 1884; Simmermacher, 1884; Knoll, 1914; Nachtigall, 1974; Stork, 1983a,b; Walker et al., 1985; Wigglesworth, 1987; Lees and Hardie, 1988; Dixon et al., 1990; Gorb, 1998). When surfaces are rough, insects can hook their claws onto small projections. Attachment to smooth surfaces, however, is accomplished by a variety of specialised adhesive organs on the tarsus or pretarsus, which differ considerably among insect orders. The structures involved are either pads of numerous spatulate hairs (e.g. in the Diptera and Coleoptera) or smooth appressoria, which can often be inflated by haemolymph pressure (e.g. in many Thysanoptera, Orthoptera and Hymenoptera).

We selected 11 tropical arboreal ant species for the measurements (see Table 1). Ant genera were identified according to Bolton (1994). Crematogaster (Decacrema) mspp. 1, 3, 4 and 6 (Myrmicinae) are obligate ant partners of Macaranga trees (morphospecies numbers follow Fiala et al., 1999). Colonies of Technomyrmex sp.A (Dolichoderinae) regularly inhabit the stipule domatia of the glaucous Macaranga pruinosa; Technomyrmex spp.B and C are similar morphospecies. Despite thorough examination, we could detect no morphological difference between Technomyrmex sp.A and sp.C. However, the different nest location (sp.C nests under tree bark or leaf undersides) and the strikingly different wax-running capacity (see Results) suggest that they are indeed different taxa. We use morphospecies names corresponding to Federle et al. (1997). Reference ant specimens are in the collection of W. Federle.

A simple field experiment was conducted in Peninsular Malaysia to determine the ‘wax-running’ capacity of the 11 selected ant species (Federle et al., 1997). We used vertical, glaucous stems (diameter 15 mm) of the ant-plant Macaranga pruinosa (Euphorbiaceae). With the aid of a paint-brush, we placed ants individually on the small wax-free scars left by the abscission of the leaves. Of each species, we tested 20 workers of similar size. We determined the proportion of ants that managed to walk past one of two markers located 5 cm above and below the release point within 10 min (for details, see Federle et al., 1997).

We used a centrifuge technique that allows measurements with no prior treatment of the insects and which is applicable to very small animals. The experimental arrangement is shown in Fig. 2. Centrifuge methods have previously been used by Dixon et al. (1990) and Brainerd (1994).

With a paint-brush or a paper strip, ants were carefully placed onto the outside of a Perspex (polymethylmethacrylate, PMMA) cylinder (diameter 80 mm) in the rotor of a centrifuge. The centrifuge could be gradually accelerated from 0 to 6000 revs min⁻¹. The Perspex cylinder was manufactured to be as smooth as possible so that its surface was glossy and free of visible scratches. We quantified its roughness using a surface roughness detector (Mitutuyo Surftest 211). The roughness index R_a (ISO, DIN) was determined to be 0.054±0.010 μm (mean ± S.D., N=20 vertical lines along the cylinder measured). PMMA has a critical surface tension (γ_c) of 39 mN m⁻¹ at 20 °C (Baier et al., 1968). Before each experiment, the cylinder was carefully cleaned with a lens cloth and 25 % ethanol. When we pulled ants resting on such a smooth Perspex surface in the horizontal direction, we could make them slide in a perfectly even movement (W. Federle, unpublished results), suggesting that the microscopic irregularities on the Perspex surface were not large enough for their claws to interlock with.

For each of the 11 ant species, two colonies were collected in Peninsular Malaysia between March and April 1997; they were then maintained in the laboratory in Germany. We took similar numbers of workers from the two different colonies for all the ant species tested.

To compare the results of the centrifuge method with a conventional force-measuring technique, we used a strain gauge force transducer (see, for example, Walker et al., 1985). Oecophylla smaragdina ants were anaesthetised by cooling on ice for 2 min. A fine thread was tied to the thorax between the front and middle legs, so that the loop did not touch either the legs or coxae. The ants were allowed to recover for more than 1 h. The other end of the thread was attached to the strain gauge. The voltage output was recorded on a DAT recorder (Biologic DTR-1801). Before each measurement, the transducer was calibrated by loading it with weights (in steps of 100 mg). Ants were placed onto a horizontal smooth Perspex plate (with material and surface properties identical to those of the cylinder used in the centrifuge) and pulled away very slowly using the hand-held strain gauge until they fell off. For each ant, we recorded five detachment events within a period of 1–3 min. The recordings were digitised using a 1401 A/D converter (CED) and then analysed using the Program Spike2 (V2.24, CED; sampling frequency 10 Hz). To exclude short force peaks caused by vibrations of the hand-held strain gauge, we calculated mean forces and averaged them over 1 s. From these mean forces, we determined the maximum force that could be applied before the ant became detached.

Ant species differ strikingly in their running performance on slippery wax crystal surfaces (Fig. 1; Federle et al., 1997). In a previous study, we quantified ‘wax-running’ capacity by placing ants on vertical, glaucous Macaranga twigs of 15 mm diameter and determining the proportion of ants that succeeded in walking 5 cm upwards or downwards (for details, see Federle et al., 1997). Using the same method, we quantified the running performance on glaucous stems of the 11 ant species selected for the present study. The results are shown in Fig. 3 (some of the data are taken from Federle et al., 1997). Within the Crematogaster (Decacrema) and the Technomyrmex group, the running performance on waxy Macaranga stems differed considerably. The ant species that are regularly associated with glaucous host-plants, Crematogaster (Decacrema) mspp. 1 and 6 and Technomyrmex sp.A, performed much better than their congeners.

In all the 11 ant species investigated, we observed a characteristic behaviour when centrifugal accelerations were applied. We could easily watch the ants during the centrifuge runs because of the 200 Hz stroboscopic flash illumination (at circulation periods near multiples of 0.005 s, images of subsequent turns become superimposed by the human eye). At low accelerations most of the ants were still running around on the Perspex surface. When we gradually speeded up the centrifuge, however, the ants at first walked much more slowly and finally stood motionless. In this stereotyped ‘freezing’ position, the ants kept all their legs spread out and in contact with the surface. We could elicit the same freezing behaviour when we directed strong puffs of air at running ants. Most of these ants also ‘froze’ immediately and spread out their legs.

These observations indicate that the detachments in the centrifuge were not caused by spontaneous, erratic movements of the ants. On the contrary, we assume that, by ‘freezing’, the ants bring themselves into a body position in which attachment is particularly strong. In the relatively large workers of Oecophylla smaragdina, we observed that the freezing reflex on a smooth glass surface was accompanied by a conspicuous inflation of the attachment organs (‘arolia’; Snodgrass, 1956). It is likely that a similar reaction also occurs in other ants.

Fig. 4 shows the results of the centrifuge measurements. For each individual, we counted only the maximum attachment force from three measurements. As a rule, forces decreased slightly over the course of the three trials (mean of second measurement, 84.5 % of first; mean of third, 75.1 % of first; Friedman rank analysis of variance, ANOVA: d.f.=2, 𝒳²=28.3, P<0.001).

The ant species tested differed strongly with regard to the measured forces (F in newtons) (data with heterogeneous variances; Kruskal–Wallis test: H=99.64; d.f.=10; P<0.001) and attachment force per body weight (F/W, where W is body weight in newtons) (H=69.19; d.f.=10; P<0.001). When ants fall off a slippery waxy plant stem, their own body weight causes them to detach. Therefore, we compared F/W values for ‘wax-runners’ and related ants. The ant species tested included two groups of morphologically similar, congeneric ants, Technomyrmex and Crematogaster (Decacrema), comprising both wax-runners and non-wax-runners. Within each of these groups, F/W varied significantly between species [analysis of covariance, ANCOVA, log-transformed data, body mass as the covariate, Crematogaster (Decacrema) spp., d.f.=3, F=9.557, P<0.001; Technomyrmex spp., d.f.=2, F=4.692, P=0.016]. However, the wax-runners did not show higher attachment forces as predicted by our hypothesis. Instead, our data provide significant evidence for the opposite effect (Fig. 4). Crematogaster (Decacrema) mspp. 3 and 4 are associated with glossy Macaranga host plants and have a strikingly poorer attachment to crystalline epicuticular wax surfaces than the wax-runners Crematogaster (Decacrema) mspp. 1 and 6 (Fig. 3; see also Federle et al., 1997). However, they attached best to the smooth surface (F/W means of mspp.3 and 4: 146.4 and 142.1, respectively). The difference between Crematogaster (Decacrema) wax-runners and non-wax-runners was highly significant [ANCOVA, nested design, log-transformed data, mass as the covariate, Crematogaster (Decacrema), d.f.=1, F=19.649, P<0.001].

We obtained an analogous result in the three morphologically similar Technomyrmex species investigated. Technomyrmex sp.A, which is much better at running on a crystalline wax surface than her congeners Technomyrmex spp.B and C (Fig. 3), attached significantly less well to the smooth surface (ANCOVA, nested design, log-transformed data, mass as the covariate, Technomyrmex, d.f.=1, F=8.869, P=0.0054).

The weight (W) of an animal is proportional to the cube of its body length, but the body surface area (A) only to the square. If animals had identical body shapes and densities, A would scale with W^0.66 (see, for example, Pedley, 1977; McMahon and Bonner, 1983).

Fig. 5A shows our data as a logarithmic plot of F/W against m, where m is body mass. Pooled across all the species investigated, we found a positive scaling coefficient (slope of the regression line, b=0.101; r2=0.032; P<0.05). Within species, however, most correlations with body mass were negative (r_s<0 in nine out of the 11 species investigated, where r_s is the Spearman rank correlation coefficient; see legend to Fig. 5A). Because body shapes probably vary more strongly among species than within species, the positive interspecific scaling coefficient (Fig. 5A) may be explained by different body shapes. To correct for the apparent large interspecific differences (see, for example, Oecophylla smaragdina in Fig. 5A), we standardised our data by calculating ‘relative’ F/W and m values (=relative proportion of the species means). Fig. 5B gives a plot of the corrected values (b=−0.331; r²=0.038; P<0.05). The intraspecific, negative scaling coefficient is therefore consistent with the predicted value of −0.333, although there is considerable scatter in the data.

In the large ant Oecophylla smaragdina, we compared the forces measured using the centrifuge method with those measurements using a strain gauge. In contrast to the ‘freezing behaviour’ observed in the centrifuge (see above), the ants tied to the thread generally continued to move during the measurements and only rarely were all six legs simultaneously in contact with the surface. Fig. 6 shows that the vertical attachment forces measured in the centrifuge were of the same order of magnitude, but significantly higher (ANCOVA, body mass as the covariate, d.f.=1, F=28.8, P<0.001). We conclude that this difference may be because the ants are more strongly manipulated in the strain gauge method than in the centrifuge method (anaesthetization, uneven pull of the thread, etc.).

Our findings clearly reject the hypothesis that the ‘wax-running’ behaviour of some Macaranga-associated ants is caused by a general, superior attachment capacity. There was no positive correlation between running performance on slippery plant wax crystals and clinging capacity to a smooth surface. On the contrary, the attachment force to smooth Perspex was found to be inversely related to the capacity to walk on waxy stems. This suggests that an unknown trade-off is involved in the specialisation of ants for wax-running. We assume that the effect is not caused by an alternative use of either claws or arolia depending on surface roughness. Crematogaster (Decacrema) msp.1 ‘wax-runners’ deprived of their arolia were unable to climb up waxy Macaranga stems (W. Federle, unpublished results). Thus, the micro-roughness of plant wax crystals is probably not sufficient for claws to hook onto (see also Knoll, 1914). Moreover, scanning electron microscope examination of the pretarsi of Crematogaster (Decacrema) mspp.1, 3, 4 and 6 and of Technomyrmex spp.A, B and C failed to detect any clear morphological differences in the claws or the arolium between ‘wax-runners’ and closely related ants incapable of climbing up glaucous stems W. Federle, unpublished results). We are currently investigating whether the properties of an adhesive secretion and/or behavioural mechanisms could provide an explanation for the wax-running capacity and the paradoxical results found in this study.

The freezing behaviour described in this study appears to be a reflex advantageous to attachment that has not been described previously. When insects walk (typically using a tripod gait), the legs have to be lifted and their attachment potential is lost. Walker et al. (1985) have shown that there is an approximately linear relationship between leg number and attachment force. Freezing behaviour may be identical to the reaction to vibrations reported for ants and honeybees (Fuchs, 1976; Little, 1962; Michelsen et al., 1986). Freezing as a response to puffs of air or to accelerations appears to be a general phenomenon widespread among insects. We have observed similar behaviour in members of the Diptera (Musca domestica, Muscidae), Coleoptera (Clytra quadripunctata, Chrysomelidae) and Heteroptera (Graphosoma italicum, Pentatomidae), even though these taxa have completely different attachment organs (see Bauchhenß, 1979; Ghazi-Bayat and Hasenfuss, 1980; Stork, 1980c). However, we were unable to provoke a visible freezing reflex in some ground-living ant species (e.g. Pheidologeton sp., Lasius flavus), even though these ants possess arolia. When exposed to strong air puffs, they continued to run and were easily detached from smooth surfaces (as a consequence, centrifuge force measurement cannot be applied in these species). We assume that freezing is adaptive to arboreal ants, but more data are needed to test this hypothesis. The intraspecific scaling coefficient of attachment found in this study (F∝W^0.66) is consistent with the hypothesis that attachment force is a linear function of the adhesive area. In their study of four aphid species, Dixon et al. (1990) found similar scaling coefficients (interspecific relationship F∝W^0.62). Proportionality of force to contact area is in agreement with the widely accepted view that most insects attach to smooth surfaces using wet adhesion (for a review, see Walker, 1992). However, it is beyond the scope of the present study to investigate the physical mechanism of adhesion in ants.

Our results show that attachment capacity differs considerably among ant species. Table 1 compares our results with other known records of insect attachment to smooth surfaces. The clinging capacities of the 11 species of arboreal ants selected for the present study are outstanding. With respect to detachment force per body weight, they far exceed the values known for other insects.

For wingless insects such as ants, survival in arboreal habitats requires the capacity to get there ‘on foot’. Moreover, when subordinate arboreal ants with restricted territories fall off a tree where they nest or forage, they will almost certainly die, since no odour trails lead them back to their nest and they may be confronted with superior numbers of other aggressive ant species. This is especially true for host-specific plant-ants, e.g. the Crematogaster (Decacrema) partner ants of Macaranga trees investigated in this study, but also for generalist colonisers of extreme habitats such as mangrove forests. The exceptional clinging capacities of the Crematogaster (Decacrema) plant-ants and of some other arboreal ant species measured in the present study thus appear to be adaptive to their particular habitat. In accordance with this, some ground-living ponerine and leptanilline ant species lack an adhesive arolium (Freeland et al., 1982).

As mentioned by Stork (1980a), wind and vibrations of a twig can strongly increase detachment forces, so that an insect must resist forces higher than those caused by gravity alone. However, the capacity of some ants to sustain attachment forces of more than 100 times their own body weight may be more than sufficient to prevent them from falling off their ‘home’ trees. Other ecological factors may be important in this respect. Many ant species transport prey items that are much heavier than their own body or tear their prey to pieces during group hunting (Hölldobler and Wilson, 1990). An extreme example is the Asian weaver ant Oecophylla smaragdina, which is known to transport very large vertebrate prey (Wojtusiak et al., 1995). In this ant, exceptional attachment forces are also needed during the construction of their leaf tent nests. Before leaves can be connected using larval silk, many workers (and often chains of workers) have to pull the leaves together. On the back of these living ‘clamps’, the last one or two workers are attached to the smooth upper side of a leaf and must sustain the entire force.

Our data show that the considerable forces measured in this study may be traced back in part to the centrifuge method, which probably yields higher values than other force-measuring techniques. Thus far, measurements of adhesive forces have been conducted by connecting insects with thread to a force transducer (Stork, 1980a; Walker et al., 1985; Dixon et al., 1990) or by loading them with weights (Ishii, 1987; Lees and Hardie, 1988). These experimental approaches required a hook or a thread to be fixed to the insect body. For these treatments, insects had to be anaesthetised and, glue or wax droplets were usually applied to the insect cuticle. Obviously, very small insects were difficult to measure using this method. Even in larger insects, the attachment forces may have been been weakened by the manipulations. Moreover, when ants are pulled by a thread, the freezing reflex is much less pronounced, and all six legs do not generally touch the surface. We therefore think that the centrifuge technique is a more appropriate method. It requires less manipulation, can be applied to very small insects and allows rapid measurements under a variety of experimental conditions.

We would like to thank Gerhard Eisenmann for the construction of the centrifuge. We are especially grateful to Thomas A. McMahon, Elizabeth Brainerd and Wulfila Gronenberg, who helped improve the manuscript with productive discussions. We appreciate the critical comments on the manuscript by two anonymous referees. This study was financially supported by a grant of the Deutsche Forschungsgemeinschaft (DFG-SFB 251).