Mycocepurus castrator Rabeling & Bacci

Rabeling, Ch. & Bacci, M., 2010, A new workerless inquiline in the Lower Attini (Hymenoptera: Formicidae), with a discussion of social parasitism in fungus-growing ants., Systematic Entomology 35, pp. 379-392 : 382-391

Mycocepurus castrator Rabeling & Bacci , sp.n.

(Figs 1A, C, E, G; 2A, C, E, G)

Holotype, ♀, BRAZIL: São Paulo, Rio Claro, Campus of São Paulo State University ( UNESP), 22.3955º S, 047.5424ºW, elevation 608 m, 29.ix.2006, C. Rabeling acc. no. CR 060929 - 14 , ex Mycocepurus goeldii nest. Holotype deposited at MZSP. Measurements (in millimetres): HW 0.6, HL 0.64, SL 0.76, WL 1.07, PPW 0.62, PW 0.21, PL 0.24, PPL 0.19, CI 94, SI 127 .

Paratypes, 104 ♀, 78 ♂, BRAZIL: same nest as holotype, 29.ix.2006 - 02.x.2006, col. C. Rabeling. Paratypes deposited at: AMNH, BMEL, CRC, MCZC, MZSP, UCDC, USNM .

Holotype, ♀ (queen). Diagnosis. Small species (WL 1.07) with a unique morphology reflecting the parasitic life history. In full face view, head rectangular (CI 94); sides approximately parallel, slightly tapering above mandibular insertions; head widest directly above the eyes; posterior margin of the head heart shaped, with a slight but distinct median concavity; posterolateral corners rounded, in lateral view drawn out to form a short, rounded lobe forming the ventrolateral corner of the head. Antennae with 11 segments; antennal scapes extremely long (SL 0.76), surpassing the posterior margin of the head by nearly half their length (SI 127). In full face view, frontal carinae and antennal scrobes absent. Frontal lobes small and rounded, barely covering the antennal sockets in frontal view. Median triangular portion of clypeus raised

between the antennal insertions. Mandibles reduced, narrow, elongate, blade-like terminating in a pointed tooth; otherwise lacking teeth except for a small basal denticle. Maxillary palps reduced, with only three segments, labial palps with two segments. Ocelli slightly raised above the surface of vertex. Mesosoma with characteristic morphology related to wing bearing. Pronotal spines absent; propodeal spines well developed, stout, as wide as long at the base and sharply pointed; metapleural gland orifice very large and circular in oblique view, ventral margin forming small, vertical tooth. Petiole with a short peduncle; node triangular in side view, with sharp crest terminating in two thick pointed teeth. In dorsal view, postpetiole approximately 3× as wide as long (PPL 0.19, PPW 0.62); lateral borders tapering into pointed angles; translucent area near posterior margin forming broad u-shaped invagination. First gastric tergite strikingly concave in lateral view. Entire body surface more or less smooth and shiny, in most areas with hexagonal microsculpture resembling a honeycomb. Body sparsely covered with stiff setae; setae erect on vertex and frontal lobes, sub-decumbent on mesoscutum and scutellum, and appressed on postpetiole and metasoma. Wings infuscated with reduced venation, densely covered with setae; clear spot or fenestra in apical part of forewing absent; rsf1 faint, hardly visible. Colour: light to dark reddishbrown. - Paratype ♀♀. Measurements (n = 15). HW 0.6-0.65, HL 0.63-0.64, SL 0.73-0.8, WL 1.07-1.23, PPW 0.62-0.65, PW 0.21-0.25, PL 0.24-0.28, PPL 0.18-0.2, CI 94-104, SI 115-128.

Paratype ♂♂ (males). Diagnosis. Remarkably similar to female, not resembling any other Mycocepurus male; characters as in female diagnosis with the following exceptions: head size of males smaller (HL 0.58-0.6, HW 0.58-0.6), whereas body length similar (WL 1.1-1.2). Mandibles reduced, narrow, elongate, blade-like, which do not terminate in a pointed tooth; otherwise lacking any teeth or denticles. Number of antennal segments reduced to 11; funicular segments approximately as long as broad, slowly increasing in length towards apex to 1.5× their width, only apical segment 5× as long as wide. Mesosoma lower and narrower; tiny opening present at the metapleuron, corresponding to the position of a metapleural gland opening in the female. First gastric tergite flat to slightly concave; male genitalia projecting forward from tip of metasoma. Basal apodeme lobed, separated from aedeagus by a deep constriction; ventral border lacking serration. Wing colour: medium to dark brown. Measurements (n = 15). HW 0.58-0.6, HL 0.58-0.6, SL 0.73-0.75, WL 1.1-1.2, PPW 0.63-0.65, PW 0.23-0.3, PL 0.25-0.28, PPL 0.18-0.2, CI 96-104, SI 121-126.

Worker. The worker caste is unknown and probably nonexistent.

Additional material examined. BRAZIL: Sa˜o Paulo, Rio Claro, Campus of São Paulo State University ( UNESP), 22.3955◦S, 047.5424◦W, elevation 608 m, 03.x.2008, C. Rabeling acc. no. CR 081003 -01, CR 081003 -02, CR 081003 -03, CR 081003 -04, CR 081003 -05; ex Mycocepurus goeldii nest.

Comments. Mycocepurus castrator is an obligate, workerless social parasite of M. goeldii and is so far known only from Rio Claro, Sa˜o Paulo State, Brazil. Mycocepurus castrator occurs sympatrically with M. smithii and M. obsoletus , but cannot be confounded with any other Mycocepurus species because of its multiple morphological adaptations for a parasitic lifestyle (Table 3). Mycocepurus castrator can be recognized by the following characteristics: (i) the long antennal scapes surpassing the posterior margin of the head by half their length; (ii) reduced, blade-like mandibles lacking dentition of masticatory margin; (iii) concave shape of first gastric tergite; (iv) smooth and shiny body sculpture with hexagonal microsculpture; (v) reduced palpal formula (3,2); (vi) females and males with 11 antennal segments; (vii) males lacking serrated ventral border of aedeagus; (viii) absence of clear fenestra from forewings of queens and males; (ix) worker caste presumably absent; (x) metapleural gland orifice enlarged in females, and potentially present in males. Mycocepurus castrator males and females look extremely similar to each other, and males are distinguished most easily from the females by the genitalia protruding from the tip of the metasoma and their darker brown colour (vs reddish brown in the queens).

Etymology. During collections of M. castrator , the host colonies were not observed to produce any alate queens and males, although sympatrically nesting M. goeldii colonies released alates. Therefore, we assume that the inquiline inhibits the host queens ' production of sexual offspring, allowing only for the production of the sterile worker caste. This is essentially ‘social castration ', hence the specific name ‘ castrator '.

Host species. Mycocepurus castrator has been found only in nests of M. goeldii and is so far only known from the type locality (Rio Claro, SP). Mycocepurus goeldii is a conspicuous, widely distributed species ranging approximately from the 40th to the 67th meridian west and from the 2nd to the 31st latitude south, an area covering most of Brazil, parts of Bolivia, Paraguay and northern Argentina . The range of habitats occupied by M. goeldii is remarkably diverse and ranges from Amazon rainforest, savannahs (Cerrado) to the fertile South American lowlands (Pampas), and secondary habitats disturbed by human activities. It does not occur in elevated sites of the South American Cordilleras. Mycocepurus goeldii workers can be distinguished clearly from its congeners based on the size and spine pattern of the mesosoma: it is the largest species in the genus and has the most complete set of spine pairs on the mesosoma ( Kempf, 1963: figs 2, 3). The natural history of this species has been studied near Sa˜o Paulo City ( Luederwaldt, 1918, 1926) and in the Manaus region of the Amazon Basin ( Rabeling et al., 2007b), but these studies do not report the presence of a social parasite attacking M. goeldii . Like most inquilines for which we have data, M. castrator probably has a patchy and locally restricted distribution. In addition, it is probable that M. castrator is host specific, occurring only in nests of M. goeldii . Despite extensive excavation of nests of sympatrically occurring Mycocepurus species, the parasite was never encountered in the nests of M. smithii ( Rabeling et al., 2009) or any other Mycocepurus species in Latin America (Rabeling, unpublished).

Natural history and nest biology. Mycocepurus castrator has been found twice in adjacent nests of M. goeldii . The two host nests had five and eight chambers, respectively, which were distributed between 5 and 190 cm depth (Table 1). The colony studied in 2006 contained 105 alate queens and 78 alate males of M. castrator , and 771 workers of M. goeldii (Table 1). Dealate queens of either species could not be encountered, suggesting that the queenright chamber was either missed during the excavation or that the queens escaped into adjacent tunnels.

The 2008 colony contained 15 dealate and 66 alate M. castrator queens, only six alate males, 1034 M. goeldii workers, a single dealate M. goeldii queen and worker pupae (Table 1). The parasite 's numerical male/female sex ratio was strongly female biased (6/66 = 0.09). Twelve of the 15 dealate M. castrator queens were encountered in the same fungus garden chamber as the reproductively active female of M. goeldii . Thus, M. castrator is host-queen tolerant (Table 1). The other three dealate M. castrator queens were found together in a separate fungus chamber (chamber 1; Table 1). The 12 queens encountered with the M. goeldii queen showed different reproductive activities: three were active egg layers, showing developed ovaries, yellow bodies and sperm-filled spermathecae. Thus, the parasite can be polygynous. In contrast, the remaining nine queens were prereproductive with filled spermathecae, but the ovaries were still developing, and yellow bodies were absent. The three dealate queens from chamber 1 were also prereproductive. The single M. goeldii queen was reproductively active.

The unparasitized M. goeldii colony studied in 2008 contained a single reproductively active queen, 33 alate queens,

Entomology, 35, 379-392

All four species originated independently, but evolved similar traits convergently, allowing for classifying them as incipient and evolutionarily derived social parasites. A second, hitherto undescribed Pseudoatta species is not considered in this table, because only few morphological characteristics are described in the literature ( Delabie et al., 1993; see also Schultz et al., 1998). Some life history information summarized here is derived only from single observations or stems from indirect evidence, and therefore should be considered tentative.

Numbered references in this table refer to the following sources (please see reference list for complete citations): [1] Schultz et al. (1998); [2] Souza et al. (2007); [3] this study; [4] S.H. Yek & U.G. Mueller (personal communication); [5] Rabeling (personal observation); [6] Bekkevold & Boomsma (2000); [7] Gallardo (1929); [8] Bruch (1928); [9] Gallardo (1916); [10] Della Lucia & Vilela (1986); [11] Delabie (1989); [12] Bekkevold et al. (1999); [13] Rabeling et al. (2007b); [14] Kusnezov (1951); [15] Kusnezov (1954); [16] J. J. Boomsma & V. Nehring (personal communication); [17] T.R. Schultz (personal communication).

496 workers and no males (Table 2). During the excavation, males and queens were leaving the maternal colony for their nuptial flight, which started on 7 October.

A natural history study of M. goeldii in the Amazon Basin ( Rabeling et al., 2007b) showed that some colonies had a single queen, whereas others were occupied by as many as four queens. Dissection of eight individuals from three separate colonies revealed that all of them were inseminated and had fully developed ovaries, demonstrating that these colonies were functionally polygynous.

The observations on nesting biology and colony counts suggest that M. castrator is polygynous, host tolerant and allows for the production of sterile M. goeldii workers, whereas the production of host sexual offspring is suppressed in the presence of the parasite. The host, M. goeldii , appears to be monogynous in the Rio Claro population, but both mono- and polygynous colonies co-occur in the Brazilian Amazon.

Behaviour. In the late afternoon of 29 September 2006, M. castrator was discovered when 31 queens and a single male left the host colony to aggregate on the nest mound. The dispersal activity was interrupted by rain, but continued on 2 October, when 24 queens and 72 males emerged. No further behavioural observations were made that year.

In 2008, M. goeldii colonies were excavated at the end of the dry season in order to study parasitized colonies before the nuptial flight. An approaching mating flight is easily identified in M. goeldii colonies, because the workers increase the number of nest entrances per soil mound to maximally 30 entrances, giving the nest mound a sponge-like appearance ( Rabeling et al., 2009). Until 3 October, when a M. goeldii colony parasitized by M. castrator was encountered, the M. goeldii workers did not modify the nest mounds for mating flights. Upon excavating the parasitized colony, all individuals from a total of five nest chambers were transferred to artificial nest chambers for behavioural studies.

Parasite mating behaviour. As soon as the uppermost chamber(CR 081003 -01, Table 1) was opened during excavation, and the ants were transferred to the artificial nest chambers, M. castrator males started copulating with M. castrator females inside the artificial chamber ( M. goeldii alates from an adjacent nest, placed into a laboratory nest, were never observed to copulate). During this time of ongoing mating activities, females and males ran erratically in jerky movements, and males mounted females seemingly at random. Observed copulations lasted between 18 and 27 s (n = 4). Single males attempted to copulate more than once. It is unknown whether repeated copulations resulted in successful transfer of sperm. Within 3 h after transfer to the artificial nest, three queens shed their wings, and subsequent dissections demonstrated that these females had been inseminated. However, their ovaries were still developing (i.e. ripe oocytes and yellow bodies were absent), indicating recent insemination (note: queens were preserved for dissection2 weeks after copulation). Postcopulatory females did not tolerate mating attempts of males, and walked faster to outdistance their pursuer. After copulation, and wing shedding,

the recently mated queens gathered and engaged in allogrooming, frequently licked each others meso- and metasomas, and wings for extensive periods of time. The first dead males were found 12 h after the mating event. Alates from other chambers did not copulate after transfer to the artificial nest. Potentially, the individuals in the topmost chamber were anticipating the upcoming nuptial flight and the opening of their nest chamber triggered the mating behaviour.

Host worker -parasite interactions. Host workers and parasite alates frequently antennated and interacted nonaggressively. Mycocepurus castrator alates did not require grooming by host workers because individuals cleaned themselves (i.e. licking appendages, cleaning antennae), and females groomed each other. Dealate M. castrator queens groomed M. goeldii workers, and were groomed by them also. On several occasions, M. goeldii workers licked the tip of a M. castrator metasoma for several minutes; it is not clear if the workers removed fecal droplets, or M. castrator queens laid either fertile or trophic eggs. Mycocepurus goeldii workers fed the parasite queens via trophallaxis. To be fed, M. castrator females frequently climbed onto the host workers ' backs, antennated the host 's antennae and head, until it bent its head backwards, regurgitated liquid, which was then consumed by the parasite. In addition to being fed, M. castrator males and females actively licked the fungus garden.

Three days after insemination, the host workers aggressively attacked one dealate queen from the topmost chamber (CR 081003 -01); six to eight workers secured her by the antennae, legs, head and petiole, until she died. Approximately 24 h after her death, three workers continued to carry around her corpse in the nest chamber. Six days after insemination, the host workers had attacked and killed several M. castrator queens, and had placed them on the refuse dump. Three dealate queens remained unmolested by hiding together in the fungus garden.

Host queen -parasite interactions. To observe the interactions between the host queen and the dealate M. castrator queens (n = 12), we placed the queens in a smaller nest chamber, after M. goeldii workers had arranged the fungus garden. The M. castrator queens were much more agile than the M. goeldii queen and initially walked around the nest chamber until they encountered a suitable spot; there they aggregated and started licking each other. When first placed in the chamber, the M. goeldii queen crawled under an adjacent piece of fungus garden and remained motionless; a worker then picked her up by the metasoma and moved the queen to a different position. During the carrying, the M. goeldii queen remained motionless. After several minutes, one M. castrator queen left the aggregation, ‘searching ' for the M. goeldii queen. When the host queen was encountered, she was surrounded by host workers antennating her. Regardless, the parasite climbed on the host queen 's back (Fig. 3), and started licking her mesosoma, petiole, postpetiole and metasoma. Shortly afterwards, a second M. castrator queen joined the first; the M. goeldii queen continued to remain motionless. The remaining M. castrator queens eventually joined the grooming cluster, and alternated grooming themselves by pulling their legs and antennae through the tibio-tarsal cleaning apparatus of the foreleg, with grooming the host queen. When the first M. castrator queen climbed on the host queen 's back, the attending workers left and resumed fungus-gardening activities. They did not react aggressively to the parasite queens and often returned to antennate and to feed the host queen via trophallaxis. Once a worker carried the host queen to a different part of the fungus garden, and a M. castrator queen rode on her back during the location, licking her, and was not chased away. Either workers or M. castrator queens attended the host queen for most of time. Rarely and then for very short periods of time, she sat by herself. The M. castrator queens attempted constantly to climb on the backs of either other M. castrator queens, the host queen or host workers.

Introduction of parasite queen into a field colony. Tw o inseminated M. castrator queens from the topmost chamber (CR 081003 -01) were introduced to a M. goeldii colony, which opened its nest mound in preparation for the nuptial flight the previous day. The M. castrator queen was placed next to the nest mound. After orienting briefly, she immediately walked towards one of the entrance holes, and within a few seconds she disappeared into one of the entrances. The M. goeldii workers, which guarded the entrances, were not seen to attack, catch or struggle with the invading parasite. After 3 h the observation was stopped, and until then, M. goeldii workers had not expelled the M. castrator queen.

A second parasite queen was placed next to a M. goeldii colony, which had closed the supernumerary nest entrances after the nuptial flight. In contrast, the parasite did not start searching for the nest entrance and we repeatedly (five times) placed her on the side of the nest mound before she finally, perhaps by chance, walked over the nest entrance. When crossing the entrance, M. goeldii workers attacked the parasite immediately. We collected the parasite queen and a dissection identified her as recently inseminated with developing ovaries.

Introduction of parasite queen into a laboratory colony. To observe how M. castrator queens invade a M. goeldii colony, we maintained a nonparasitized M. goeldii colony in the laboratory. Upon transfer to the artificial nest, M. goeldii workers immediately covered the host queen with mycelial tufts, until she was completely hidden some 5 min later. To introduce the parasite queen, she was placed in a tube, which was connected to the fungus chamber. Quickly, she found her way out, headed directly towards the fungus garden, and immediately encountered the host queen. Then she started running in circles on the piece of fungus garden, under which the host queen was hidden. At that point, the host workers started chasing her, until one worker got hold of her petiole, and a second worker grabbed an antenna. The trio remained motionless for about 10 min, until both workers suddenly released the captive. Immediately, the M. castrator queen made a beeline for the host queen 's hiding place, where she was captured again. When being captured, the M. castrator queen assumed a characteristic position: she tucked her metasoma under the mesosoma, with the first gastric sternite touching the coxae. The combination of smooth body surface and broad, concave first gastric tergite provided little contact surface for attacking host workers, and their mandibles repeatedly slipped, upon which the host workers turned around and left. While being held by workers, the parasite queen was often antennated and licked at the base of the mandibles. After she was ‘dumped ' on the refuse pile, she promptly aimed for the host queen. This cycle of capture and release was repeated overnight. Meanwhile, the M. goeldii worker removed the fungus garden fragments from the host queen, releasing her from her mycelium prison. Unfortunately, the first contact of host and parasite queen was not observed, but early the next morning, the M. castrator queen was ‘riding ' on the M. goeldii queen, licking her mesosoma, petiole, postpetiole and metasoma. The M. goeldii workers frequently antennated and licked both queens, and the parasite queen was not attacked anymore. Subsequent observations revealed that the M. castrator queen was always ‘riding ' on top of the M. goeldii queen, and M. goeldii workers seemed to attend both queens equally.

Discussion

Several lines of evidence from natural history, behaviour and morphology identify M. castrator as an evolutionarily derived inquiline parasite of M. goeldii . Mycocepurus castrator is functionally polygynous and host-queen tolerant. It does not seem to produce a worker caste. Parasite queens apparently suppress the production of host sexual offspring, effectively castrating the infested host colony. Mating takes place inside the host nest and mating flights of the parasite have not been observed and probably do not occur. Alates of M. castrator , although fully alate and seemingly capable, do not fly. After mating, we opened artificial laboratory nests to test if queens and/or males disperse via flight and none did. Alates were encouraged to climb objects that could serve as potential launch pads (i.e. pencils), but they plunged down after reaching the tip, without becoming airborne. Thus, dispersing queens must walk to new host nests, severely limiting their dispersal.

Recently inseminated M. castrator queens found new colonies by invading host nests and showed elaborate behaviours related to securing adoption by the hosts. Compared with the host, parasite queens and males are reduced in body size and exhibit several morphological specializations known as the inquiline or anatomical parasite syndrome ( Kutter, 1969; Wilson, 1971, 1984; Holldobler & Wilson, 1990). The new parasite appears to have only one host, M. goeldii , even though other free-living congeners are present in the type locality. Lastly, even though the host ant is widespread and abundant throughout much of southern South America, M. castrator has been collected only twice, both times at the type locality. This suggests that parasite populations are probably few in number, small in size and patchily distributed. Intranidal mating, limited dispersal and small, isolated populations imply minimal gene flow between populations, and high levels of inbreeding within populations, conditions that have been postulated to exist for a number of other inquilines ( Kutter, 1969; Wilson, 1971; Zamora-Munoz et al., 2003; Buschinger, 2009). Inbreeding, however, has not yet been documented genetically in M. castrator (Rabeling, in preparation). Alternatively, it might be possible that the queens from a single nest are not necessarily closely related, because M. castrator is functionally polygynous, and parasite queens seem to invade host colonies independently.

Mycocepurus castrator shows several additional morphological features related to its parasitic lifestyle (Table 3). Mycocepurus castrator is the only inquiline known from the Lower Attini and, interestingly, it shares convergently evolved derived morphological characters with another distantly related attine workerless inquiline, P. argentina (Table 3). In both species, males and females exhibit a reduced palp formula of 3,2 (Figs 1G, 2G), whereas the plesiotypic attine palp formula is 4,2 (Figs 1H, 2H) ( Gallardo, 1916; Schultz et al., 1998). Elsewhere in the Attini , the reduced palp formula is found only in the free-living genus Apterostigma Mayr 1865 ( Kusnezov, 1951, 1954; Schultz, 2007), where it evolved independently. In addition, in M. castrator the number of antennal segments is reduced from the plesiotypic attine condition of 11 segments in the females and 13 in the males to 11 segments in both sexes. The reduction of segments in M. castrator and P. argentina suggests that both species are evolutionarily derived inquilines. According to Wilson 's (1984) character analysis of nine presumably independently evolved inquilines in the genus Pheidole Westwood , the reduction of antennal segmentation occurred secondarily during parasite evolution, preceded by the loss of the worker caste and other morphological reductions, like the reduction in size and body sculpture.

The wing morphology of M. castrator provides indirect support for the functional wing morphology hypothesis of the

Paleoattini (sensu Kusnezov, 1963; Schultz & Meier, 1995) ( Fernandez-Marin et al., 2005). Emery (1913) first noted that the forewings of queens of Myrmicocrypta F. Smith bear a small clear spot, or fenestra, which is not covered with fine setae and lacks pigmentation. This fenestra is also present in other paleoattines ( Fernandez-Marin et al., 2005; Schultz, 2007). During colony founding, queens of Paleoattini inoculate their fungus garden on their shed forewings, and fix the forewing under the ceiling of the new nest chamber to grow the fungus cultivar ( Fernandez-Marin et al., 2004). The fenestra was hypothesized to serve some unspecified function during nest founding ( Fernandez-Marin et al., 2005). Interestingly, the clear spot is absent in M. castrator wings, whereas it is present in all Mycocepurus species. The loss of the fenestra in obligate social parasites, which do not found colonies independently, suggests that it indeed has an adaptive function during independent nest founding, because otherwise the fenestra would be maintained by natural selection in obligate inquilines. Potentially, the clear area is easier to clean before inoculation with a novel cultivar, or it may provide a tactile or even visual cue to the queen where to place the inoculum. It would be interesting to test if parasite queens contribute a fungal inoculum to the new host colony [suggested by Schultz et al. (1998)], and if the infrabuccal cavity of socially parasitic attines experienced morphological modifications or reductions.

Comparing M. castrator with other attine inquilines is instructive and provides insight into the life history evolution of social parasites (Table 3). Like M. castrator , P. argentina appears to be a phylogenetically derived, workerless inquiline and shares some morphological and life history traits with M. castrator (see Table 3). Acromyrmex ameliae and A. insinuator , however, are much different. Like M. castrator and P. argentina , the Acromyrmex inquilines are each associated with a single host species. However, they show much less morphological and life history specialization ( Schultz et al., 1998; de Souza et al., 2007). For example, A. ameliae and A. insinuator lack the palpal and antennal segment reductions and other morphological features defining the inquiline syndrome(Table 3). Both species have mating flights, sometimes synchronously with host alates. Both species produce a worker caste, and parasitize host colonies at higher frequencies than most inquilines for which we have data ( Bekkevold & Boomsma, 2000; de Souza et al., 2007). Because mating flights are retained, dispersal is probably much less restricted, and inbreeding may be slight or nonexistent, as was demonstrated for A. insinuator ( Sumner et al., 2004a) . These traits and other aspects of their biology (summarized in Table 3) strongly suggest that these species are recently evolved inquiline social parasites that have yet to develop the morphological inquiline syndrome and full workerless parasitism.

It should be stressed that the attine inquilines are not closely related to each other ( Sumner et al., 2004b; Schultz & Brady, 2008) and what is known of their life histories provides additional evidence of separate and idiosyncratic evolutionary trajectories. For instance, limited evidence suggests that the derived inquiline P. argentina may be host-queen intolerant( Bruch, 1928). In addition, there is some indication that the fungus garden in P. argentina colonies breaks down after a parasite brood has been reared by the host workers ( Bruch, 1928). Thus, Pseudoatta colonies may be quite short lived, perhaps even semelparous. Curiously, evidence has accumulated that the recently evolved inquiline A. insinuator may also be semelparous ( Bekkevold & Boomsma, 2000), even though parasitized colonies retain a fertile host queen and could theoretically last as long as she lives and lays eggs. Nonetheless, Bekkevold & Boomsma (2000) provided evidence that the colony 's fungus garden breaks down after a generation of parasite sexuals is produced. In contrast, in M. castrator colonies the host queen is retained and the fungus garden appears to remain healthy after the production of parasite sexuals. Thus, the colonies may last for several parasite reproductive cycles. Although one might expect that colony longevity should be favoured by natural selection under a broad range of conditions, such counterintuitive examples remind us that we are far from fully understanding the complex and diverse life history phenomena grouped under the term ‘inquiline social parasitism '.

In conclusion, obligate social parasites are prime examples for the study of convergent evolution. Over 80 inquiline species are known currently ( Holldobler & Wilson, 1990; Huang & Dornhaus, 2008), and mapping parasite lineages onto recent phylogenies ( Brady et al., 2006; Moreau et al., 2006; Rabeling et al., 2008) illustrates that inquilines evolved at least 30 times convergently. A higher taxonomic resolution of speciose groups will probably increase the number of independent parasite origins. In fungus-gardening ants, two groups of inquilines evolved: incipient and evolutionarily derived social parasites. Mycocepurus castrator is a derived parasite, showing multiple morphological and life history traits of the inquiline syndrome ( Kutter, 1969; Wilson, 1971, 1984). Mycocepurus castrator is the first social parasite of the Lower Attini and an additional social parasite from a subtropical habitat, a geographical region from which social parasites are little known. A phylogenetic and population genetic study is underway and will answer additional questions about the evolutionary origin of M. castrator (Rabeling, in preparation).