Field Observations on Milichia patrizii Ant-Mugging Flies (Diptera: Milichiidae: Milichiinae) in KwaZulu-Natal, South Africa

ABSTRACT Kleptoparasitic behaviour of a myrmecophilous fly, Milichia patrizii Hennig, 1952, towards the ant Crematogaster castanea tricolor Gerstaecker, 1859 in South Africa is documented, illustrated by photographs, and discussed.


INTRODUCTION
Myrmecophily, or ant-loving behaviour, has been known in Milichiidae for more than a hundred years. Most cases involve larvae that live in ant nests (Donisthorpe 1927;Sabrosky 1959;Moser & Neff 1971;Waller 1980), but a group of short-faced Milichia species exhibit myrmecophilous behaviour in the adult stage, mostly in connection with myrmicine ants in the genus Crematogaster. Milichia dectes Collin, 1922, M. proectes Collin, 1922, M. prosaetes Collin, 1922 and probably also M. farquharsoni Collin, 1922 solicit regurgitated food from Crematogaster (Farquharson 1919(Farquharson , 1922. In one case involving Milichia dectes, the ant was identified as Crematogaster buchneri Forel, 1894. Two species have been observed licking the anal secretions of ants: Milichia myrmecophila de Meijere, 1909, those of Crematogaster difformis Smith, 1857, and Milichia brevirostris (de Meijere, 1910, those of Dolichoderus bituberculatus Mayr, 1862. Milichia patrizii Hennig, 1952 has not been previously known as a myrmecophile. However, the holotype label states that the species is a commensal of Crematogaster, and an unidentified Crematogaster worker specimen (possibly C. schultzei Forel, 1910) is mounted on the same pin as the fly. Milichia patrizii belongs to the myrmecophila species-group within the short-faced group of Milichia species (Brake 1999). The purpose of the present paper is to report a recent observation on the myrmecophilous behaviour of Milichia patrizii in South Africa.

MATERIAL AND METHODS
On 5 July 2008, AW encountered by chance numerous individuals of Milichia patrizii patrolling Crematogaster castanea tricolor Gerstaecker, 1859 trails on two different Acacia trees. Interactions between the flies and the ants were observed and photographed for two hours between 11:00 and 13:00 hr. The habitat was along the edge of a clearing in a coastal forest in the St Lucia Estuary, KwaZulu-Natal, South Africa (28.3872°S: 32.405°E). Voucher specimens (two female flies and three ants) were collected into 95 % ethanol and have been deposited at BMNH. The ants were identified by AW and the flies by IB.
Photographs were taken with a Canon EOS 20D digital SLR using an MP-E 65 mm 1-5× macro lens and an MT-24EX twin flash. Images were processed using Adobe Photoshop CS2 for minor adjustments to colour balance, brightness and contrast.
In describing the milichiid antenna we prefer to use the term 'basoflagellomere' (Thompson 1999) for the structure that Stuckenberg (1999) called 'postpedicel'.

RESULTS
Flies patrolled both vertical ant trails on Acacia trunks and horizontal trails along the upper surfaces of branches (Fig. 1). The behaviour of the ants and the flies did not appear to show differences in either situation. Flies were spaced along the trails at least 10 cm apart and usually considerably longer. On the first Acacia tree, between two and four flies were observed along a vertical trail, while on the second tree more than a dozen flies were active along a single horizontal trail. The sex of all the observed flies was not determined in the field, nor was mating observed, but based on the sample of collected and photographed individuals we believe all were likely female.
To attack an ant, a fly moves into the ants' trail and attempts to isolate a single individual (as in Fig. 2). Successful attacks generally result from a fly being able to immobilize an ant either from the front or the side; once an ant passes, the fly is normally unable to catch it even though it may continue to follow the ant for up to a few centimetres. The flies use their cup-shaped basoflagellomeres to grasp the terminal antennomeres of an ant (Fig. 3). The ant responds by standing still and crouching down (Fig. 4), allowing the fly to extend its proboscis into the ant's mouthparts to trigger the regurgitation response (Figs 4-6). Either before or after the initiation of the food exchange, the terminal segments of the ant's antenna are released from the basoflagellomeres and enfolded instead by the pilose clypeal membrane of the proboscis (Fig. 6). Disengagement is apparently initiated by the fly, which withdraws its proboscis (Fig. 7). The ant remains relatively motionless for a brief period following the interaction, while the fly usually moves back to patrolling the trail. In one timed interaction, the sequence of events from first contact to disengagement lasted 23 seconds, with the food exchange itself occurring in the final 10 seconds.
Most kleptoparasitic attempts (roughly 80-90 %) were unsuccessful, the ants either outrunning or outmanoeuvring the flies. A fly would pursue a particular ant for a few seconds, and if unsuccessful would disengage and return to face the ant trail again. Interestingly, the flies did not pursue other ant species present on the same branches (Tetraponera spp., Cataulacus brevisetosus Forel, 1901), including Camponotus troglodytes that apparently mimics Crematogaster. Throughout the observation period, the Crematogaster engaged in regular trophallaxis behaviour with nestmates along the trail (Fig. 8). DISCUSSION A similar behaviour to the one observed here was described by Farquharson (1919) for M. dectes, M. proectes and M. prosaetes, though he did not provide as much detail. It is not possible from the publication to verify whether the behaviour is present in all three species or just one, but Farquharson classified all three as "mendicants". Farquharson wrote that the fly approaches the ant from the side, the ant either continues on or stops and partly rotates its head. The fly drinks a little at one side, then moves around to the other side before the ant has time to move away and drinks a little more. The use of the basoflagellomeres in M. patrizii to grasp the ants' antennae is probably restricted to this species, as the scape (first antennal segment) is enlarged and strongly sclerotized in M. patrizii (Hennig 1952, fig. 17) compared to the other short-faced species. From the photographs (Figs 4-6), the palpus does not appear to have any function in the myrmecophilous behaviour, instead the clypeal membrane between the palpi, which is covered in a short microtomentum, seems to surround the tip of the ant's antenna after the basoflagellomeres let go. Dissection of the proboscis did not reveal any sclerotized structures between the palpi which could be used to grasp or hold the ant's antenna. The shape and setation of the proboscis might trigger the trophallaxis-like behaviour. Both Milichia patrizii and its sister species M. farquharsoni are characterized by an unusual shape of the palpus and proboscis (Hennig 1952, figs 18, 19). At the tip of each labellum there are three or four setae about 1.5 times as long as the labellum itself. These likely serve to capture liquids from the mouth of the ant. M. farquharsoni was observed to 'haunt' the nest of Crematogaster ants (Farquharson 1922), although its feeding behaviour is unknown. The other species of the short-faced Milichia group have enlarged, flat palpi and flat spear-head shaped labella that are quite different from the homologous structures in M. patrizii and M. farquharsoni.
Trophallaxis food exchange among social insects presents an opportunity for exploitation. Milichia patrizii is one of many myrmecophiles to have independently acquired an ability to co-opt the food-sharing behaviour. Trophallaxis-based parasitism appears to evolve readily, as similar behaviour is sporadically present in the insect orders Thysanura, Orthoptera, Coleoptera, Lepidoptera, Diptera, and Hymenoptera (Hölldobler & Wilson 1990;Sivinski et al. 1999). The other dipterans known to engage in kleptobiosis with ants, mosquitoes of the genus Malaya and at least two phorid species (Sivinski et al. 1999), and some shore flies of the genus Rhynchopsilopa (Wirth 1968), indicate convergent evolution as they are not closely related to Milichiidae. Additionally, Crematogaster ants seem vulnerable to the attention of kleptobionts; in addition to the solicitation of liquid foods from the mouth of the ants, thievery of their anal exudates may also occur (Wirth 1968;Freidberg & Mathis 1984).
The novelty of the M. patrizii/Crematogaster interaction seems to lie in the unique antennal-grasping mechanism used to initiate the exchange. Additional research is needed to reveal the specificity and extent of the interaction.