Biology of Lissoderes Champion (Coleoptera, Curculionidae) in Cecropia saplings inhabited by Azteca ants

This paper provides an account of the biology of Lissoderes (Coleoptera, Curculionidae, Conoderinae) focusing on L. pusillus Hespenheide and L. subnudus Champion. The eggs, larvae, and pupae live inside the hollow stems of Cecropia saplings. Adult weevils chew through the stem and deposit eggs on the inner surface of the internode. The larvae feed on the parenchyma lining the hollow internodes and pupate inside the internode. Emerging adults chew their way out of the stem. Two hymenopteran parasitoids were reared from larvae and prepupae of L. pusillus: Neocatolaccus sp. (Pteromalidae) and Heterospilus sp. (Braconidae). Menozziola sp. (Diptera, Phoridae) and Conoaxima sp. (Hymenoptera, Eurytomidae) were observed parasitizing Azteca queens. Parasitism by these species may explain part of the high mortality observed in colonizing Azteca queens. Direct competition with L. pusillus and L. subnudus appears not to be a major cause of queen mortality, although possible indirect effects of the weevils remain unknown.


Introduction
Species of Cecropia Loefling (Cecropiaceae) are some of the most common and characteristic trees of early successional vegetation in the Neotropics (Berg and Franco Rosselli 2005). Moreover, the mutualism between Cecropia and certain species of Azteca ants (Hymenoptera, Formicidae) is one of the best known ant-plant associations (Wheeler from the dense pad of trichomes (the trichilium) located at the base of the petioles and pearl bodies from the undersides of the leaves; in turn the worker ants cut encroaching vegetation (Janzen 1969), remove insect herbivores (Schupp 1986;Agrawal 1998), and provide the plant with nitrogen (Sagers et al. 2000). The hollow stems of Cecropia are divided into a series of closed chambers (internodes) which are separated from one another by solid partitions (septa) at each node. In young stems the upper part of each internode develops an oval depression (prostoma) where the wall is thinner and through which the colonizing Azteca queen chews a hole in order to enter. Once inside the internode the queen scrapes plant material over the entrance hole, which eventually closes even more tightly as callus tissue grows over it. The queen thus remains sealed inside the internode and apparently relies primarily on her own body reserves to rear the first brood of workers.
Cecropia saplings are colonized by numerous Azteca queens, although mature trees are usually dominated by just a single colony of ants (Longino 1989(Longino , 1991a. This results in competition for nesting sites among queens, especially since they often prefer the upper internodes, and different species of Azteca vary in their colonization strategy. For example, queens of A. constructor Emery are more successful when they found colonies together (pleometrosis) in the same internode, whereas queens of A. xanthochroa Roger survive equally well in both haplometrotic (founding alone) and pleometrotic associations, including those containing both species (Choe and Perlman 1997). Once the first workers emerge and reopen the prostoma, they begin foraging for Mü llerian bodies, storing them in the internodes presumably as food for larvae. In pleometrotic associations each queen produces workers; in A. xanthochroa only a single queen survives past the emergence of the first workers whereas in A. constructor cofoundresses may cooperate for a year or more (Choe and Perlman 1997). As the colony grows the workers chew holes in the septa that separate the internodes. Saplings frequently contain numerous incipient colonies in separate internodes and at some point the strongest colony presumably eliminates all others (either directly and/or indirectly by harvesting most of the Mü llerian bodies ;Longino 1991a).
In Costa Rica there are four species of Cecropia (excluding a species restricted to Cocos Island), three of which are colonized by Azteca ants: C. insignis Liebmann, C. obtusifolia Bertoloni, and C. peltata L. (primarily on the Pacific side); the fourth species, C. angustifolia Trécul [formerly C. polyphlebia Donnell Smith (Berg and Franco Rosselli, 2005)], is a higher elevation (1400-2400 m) species devoid of Azteca colonies, although founding queens often attempt colonization (Burger 1977;Longino 1991a). There are more than 20 species of Azteca in Costa Rica, five of which are associated with Cecropia: A. alfari Emery, A. ovaticeps Emery, A. coeruleipennis Emery (primarily in C. peltata on the Pacific side), A. constructor, and A. xanthochroa (Longino 1991b). The only other ant that is a regular associate of Cecropia in Costa Rica is an undescribed species of Pachycondyla. Unlike the Azteca species mentioned above, queens of this species keep the prostoma open (they do not seal themselves inside the internode) and their colonies, which are much smaller, mature and reproduce very quickly before Azteca take over the tree. Other ants in various genera (e.g. Camponotus, Crematogaster, and Pseudomyrmex; L. M. LaPierre, personal observation) occasionally colonize young Cecropia stems, but in Costa Rica none of these is known to be a Cecropia specialist.
Even before they begin producing workers, Azteca queens suffer a high rate of mortality (Longino 1989). Because the causes of this mortality are largely unknown, more knowledge of the other inhabitants of young Cecropia stems may be relevant to a better understanding of the colonization process. In general, these other inhabitants have received very little attention. In his revision of the genus Lissoderes Champion (Curculionidae, Conoderinae, formerly Zygopinae), Hespenheide (1987) noted that the larvae and adults of these weevils have been observed inside the internodes of Cecropia saplings, an observation that was confirmed by LaPierre (2002). However, as far as we are aware no further details are known of their biology. Indeed, the biology of Neotropical Conoderinae is in general poorly known (Lyal 1986;Jordal and Kirkendall 1998;Hespenheide 2005 and references cited therein; Hespenheide and LaPierre forthcoming; H. A. Hespenheide and L. M. LaPierre, unpublished). The genus Lissoderes currently comprises five described species, all of which are confined to the Neotropical region, and four of which occur in Costa Rica. Here we describe the biology of Lissoderes, based on data collected from L. pusillus Hespenheide and L. subnudus Champion, and compare its colonization of Cecropia saplings with that of Azteca queens. We also discuss to what extent the weevils are a parasite of the mutualism between Cecropia and Azteca and whether it contributes to the high mortality rate observed in colonizing Azteca queens. Finally, we provide some previously unreported observations on other sources of this high mortality rate.

Alberto Manuel Brenes Biological Reserve
Most of the research was carried out by Weng, Nishida, and Hanson in Alberto Manuel Brenes Biological Reserve (AMBBR; 10u139120N, 84u36960W; 850 m altitude) located on the Caribbean side of the Tilarán mountains, in the San Ramó n protected zone, Alajuela Province, Costa Rica. The site is in premontane rainforest, with an average daytime temperature of 21uC and yearly precipitation of 4000-5000 mm (Ortiz 1996). Three visits were made to the field site: March 2002, October 2002, and April 2003. Although more than one Lissoderes spp. can occur at a given site (Hespenheide 1987), only one species of Lissoderes, L. pusillus, was observed and reared at this site during the three visits. Two species of Cecropia grow at the site, C. insignis (the dominant species in the area) and C. obtusifolia (which reaches the upper altitudinal limits of its distribution in this area). Saplings (,1.5 m) of both species, which were growing along the dirt road leading into the station, were cut at the base and brought back to the laboratory. These plants all lacked patrolling ants. Samples for the quantitative studies of colonization were taken in October and included 35 saplings of C. insignis and 44 of C. obtusifolia (two very small saplings of the latter species were excluded because they lacked internodes).
At AMBBR, stems were cut transversely at each node; each internode was dissected longitudinally and examined under a dissecting microscope. The number of internodes occupied by ants, weevils (including oviposition holes), and other insects was recorded. Photographs of each life stage were taken with a digital camera (Nikon Coolpix 990; Nikon Corp., Tokyo, Japan). Some larvae were boiled and preserved in 75% ethanol. At the entomological laboratory of the University of Costa Rica (UCR; average room temperature 23-24uC), the remaining larvae and pupae were reared in freshly cut internodes, which were placed in plastic bags for further observation. Larvae and pupae of weevil parasitoids were also reared in the internodes under well-ventilated conditions to avoid infestation by fungi. Voucher specimens of all insects are deposited in the Museo de Zoología, Escuela de Biología, Universidad de Costa Rica.
To compare resource quantity between the two Cecropia species, the length of internodes occupied by larvae and the thickness of its parenchyma (at the midpoint of the internode) were measured. Interactions between weevil larvae and ants were observed by splitting open internodes containing one or more queens, placing a weevil larva within, and then either watching the ensuing interaction under the dissecting microscope, or attaching the two halves of the internode with masking tape and reopening it about an hour later.
The Mann-Whitney U test was used to compare the frequency of living and dead Azteca queens found in internodes harboring solely dead and living queens in each Cecropia species. Pearson correlation analyses were used to determine the relation between number of internodes occupied by weevils and Azteca ants (live and dead ants) of each plant species. Simple linear regression analyses were used to determine the effect of total number of internodes of each plant type (independent variable) on the number of internodes colonized by Azteca ants and those occupied by weevils, excluding saplings that harbored neither Azteca ants nor weevils. The Student's t test was used to compare the average length and parenchymal thickness of occupied internodes between two Cecropia species.

La Selva Biological Station
Data from a second study site, La Selva Biological Station (''La Selva''), visited by LaPierre during September 1996 are included in order to make comparisons with some of the data gathered from AMBBR. La Selva (10u269N, 83u599W; 50-150 m altitude) is located on the Caribbean slope of the Cordillera Central at the confluence of the Río Puerto Viejo and Río Sarapiquí, Heredia Province, Costa Rica. La Selva is in lowland rainforest with an average daytime temperature of 26uC and yearly precipitation of 3962 mm (McDade et al. 1994). Two Lissoderes species, L. pusillus and L. subnudus, are known to occur at La Selva (H. A. Hespenheide, personal communication). Similar to AMBBR, only C. insignis and C. obtusifolia occur at this site but with the latter species being more abundant at La Selva. Twenty-two C. obtusifolia and 12 C. insignis saplings (0.5-2.0 m) lacking patrolling ants were harvested and brought into the laboratory for dissection. The data recorded from the La Selva samples were limited to the numbers of ants, weevils, and other insects encountered in each internode.
Two-by-two contingency table analyses applying G tests were used to compare the internode contents between the two Cecropia species at La Selva.
About 40% of the 639 internodes examined were used by Lissoderes or Azteca: 19.2 and 20.4%, respectively. In only about 1% of the internodes did we find evidence of Lissoderes and Azteca living together. In the laboratory, the results of placing Lissoderes larvae in the same internode as Azteca queens were variable; the ants ignored the weevil larva, cared for it (one queen repeatedly carried the weevil larva back to her pile of eggs as it attempted to crawl away), or killed it. In contrast Pachycondyla queens and Azteca workers always killed the weevil larva.
Lissoderes colonization of Cecropia saplings begins very early, when the plant has only three or four available internodes ( Figure 1). In contrast, colonization by Azteca ants differs between the two species of Cecropia-in C. obtusifolia colonization begins simultaneously with that of the weevils (when the plant has four internodes), but in C. insignis it does not begin until the plant has eight internodes ( Figure 1). Thus the ants colonize a greater range of sapling sizes in C. obtusifolia than in C. insignis, and in the former species ants occasionally colonize every available internode (e.g. one plant with eight internodes), which never occurred in C. insignis. As a result, ants occupied a greater proportion of internodes in C. obtusifolia than in C. insignis (Table I), and more C. obtusifolia saplings (27.3%) contained only ants, compared with C. insignis (5.7%). In contrast, the proportion of internodes occupied by live weevils in the two Cecropia species was quite similar (Table I). Fewer C. insignis saplings (33%) contained weevils and ants, compared with 73% for C. obtusifolia. A higher percentage of C. insignis saplings (28.6%) contained only weevils compared with C. obtusifolia (15.9%). Despite these differences between Cecropia species, we found no evidence of any particular sequence of colonization by weevils and ants; weevils were found just as often in internodes above and below those harboring ants.

Colonization of Cecropia saplings by ants and Lissoderes spp. at La Selva
A total of 422 internodes from 34 saplings were dissected (236 C. obtusifolia and 186 C. insignis). Although all of the saplings showed evidence of colonization by Azteca, Lissoderes, or other ants, more C. obtusifolia internodes were empty (51.7% compared to 29.0% for C. insignis; G test P,0.0001; Table I). Weevils colonized twice as many internodes of C. insignis as C. obtusifolia, and of the occupied internodes none contained ants and weevil larvae together. Ant species other than obligate Azteca species were relatively common in C. insignis internodes (Table I). Live Azteca queens were encountered in the same proportion of internodes between the two Cecropia species, but success in founding a colony was higher in C. obtusifolia by a factor of nearly three while the number of C. insignis internodes with dead queens was nearly double that of C. obtusifolia (Table I). So while the number of internodes with Azteca queens did not differ significantly between C. insignis and C. obtusifolia (27.4% compared to 30.5%, respectively; G test P50.49), the outcome for the queens did differ with the majority of these internodes containing dead queens in C. insignis and incipient colonies in C. obtusifolia (Table I; G test P50.05).
The egg (Figure 2) is white, ovoid, and measures 0.49¡0.02 mm60.63¡0.03 mm (n54; mean¡SD). Under the dissecting microscope the chorion appears to be smooth and shiny. Sixteen eggs were observed in dissected internodes (seven in C. insignis and nine in C. obtusifolia). The egg is laid singly per internode; however, in three cases we observed two eggs in the same internode of C. insignis; in two of these the eggs were laid adjacent to each other and only one perforation scar was found (suggesting that they were laid by the same female), and in the third case, two perforations were observed about 6 mm apart. The eggs were nearly always located on the inner wall of the internode chamber, next to the ovipositional perforation (mean¡SD51.18¡1.07 mm from the hole; Figure 2). These perforations are visible from the outside of the stem as a light brown scar measuring approximately 1 mm in diameter (Figure 3), and usually located in the upper one-third of the internode (50%), infrequently in the lower third (31%) and middle third (19%). About a third of the ovipositional perforations located in the upper third were in the prostoma (Figure 3).
Although oviposition behavior was not observed, the female weevil probably utilizes her mandibles to perforate the stem and oviposit through this hole, as has been reported in other weevils (Howden 1995). In one instance we observed an egg that was located inside an incomplete perforation and in this case the stem wall appeared to be thicker than the mean length of the female rostrum. The rostrum of female L. pusillus (Figure 13) is about 2 mm long and 0.4 mm wide at its broadest point, the apex.
No internode contained more than one living weevil larva, but one larva was found in an internode together with a head capsule of another larva of the same age. When two larvae were placed in the same chamber for 24 h, only one survived (n57 pairs). When two eggs were placed together, the larva that hatched first killed the other larva upon hatching (n53 pairs). In both cases the live larva had not eaten the dead larva after 42 h (at which point it was beginning to decompose). Thus they do not appear to be cannibalistic, but rather very aggressive toward competitors.
The body of the larva (Figure 4-7) is subcylindrical and fleshy, with dorsal plicae (enlarged swellings or folds) that bear a few setae. The thoracic segments have a greater diameter than the abdominal segments and the lateral swellings distinguish them from abdominal segments. The first instar larva (Figure 4) is creamy white, 1.86-2.46 mm in length. The head is pale yellowish brown with reddish brown mouth parts, 0.28-0.32 mm in width, and retracted up to two-thirds of its length into the prothorax (n54). The second instar (Figures 5, 6) is 3.12-3.54 mm, with a head width of 0.40-0.46 mm (n54). The third instar (Figure 7) is 4.02-5.76 mm, head width 0.58-0.64 mm (n56). The prepupa (Figure 8) is creamy white and the head is protruded. The pupa (Figures 9-11) is white with dark red eyes and moveable abdominal segments. The weevil larvae feed on the parenchyma (Figures 4-6) that lines the inner wall of the internode. First instar larvae (Figure 4) were usually found close to or on the upper septum. Some of the sites where first instar larvae had been feeding on the septum developed callus tissue, but larvae were never observed feeding on this tissue. Second and third instar larvae (Figures 5-7) were observed feeding on the spongy white parenchyma that lines the inner walls of the hollow internode. Often, especially when the parenchyma is very spongy, the first and sometimes the second instar larvae burrowed into the tissue upon which they were feeding. Places where the larvae had been feeding on the internode wall (and surrounding tissues) exhibited a light brownish (oxidized) colour.
In the case of 14 larvae in 10 saplings of C. insignis (having 5-18 internodes), the larvae had chewed through the septum (which is still tender in young internodes) and were found feeding on the parenchyma of an adjacent internode (number of internodes utilized by a single larva varied from two to six). In 64% of these cases (nine of 14) the larvae had moved to internodes above the original one (younger internodes, with more succulent parenchyma); four larvae moved to lower internodes, and one larva had perforated the septa on both sides and was found in a middle internode. This behavior of penetrating adjacent internodes was observed in only one larva in C. obtusifolia (in a plant having seven internodes). This difference in larval behavior in the two species of Cecropia could be a result of differences in the parenchymal thickness of occupied internodes in C. insignis (mean¡SD57.0¡2.0 mm) and C. obtusifolia (mean¡SD58.4¡2.8 mm) (t522.78, 86 df, P50.007). The mean length of occupied internodes did not differ between Cecropia species (t521.02, 86 df, P50.31). Nonetheless, among all the larvae observed (including those that had crossed over to another internode), the vast majority consumed less than half of the available parenchyma present in the internode. The maximum amount consumed was about 80% of the available surface area in an internode of C. obtusifolia harboring a second instar larva.
All larval stages were always observed in an ''upside-down'' position, that is, in contact with the plant surface with their dorsum (Figures 5-7, 15a). Larvae appeared to move about by means of peristaltic motion of the fleshy swellings located on the dorsum, perhaps helped by setae located on those swellings. Feeding also occurs in an ''upside-down'' position (Figures 6,15b) and is accomplished by curving the head backwards and pushing it against the parenchyma. Larvae feed in one location and then crawl to another location in which to feed. Feeding scars were more or less scattered, but were more concentrated in the upper portion of the internode. Larvae in all instars appear to adhere to the vertical surface of the internode wall by means of their dorsal projections (setae) and by surface tension (sometimes they were coated with a liquid that appears to be secreted from the posterior end). When disturbed with forceps, second and third instar larvae regurgitated a translucent brown liquid. Larvae were seen defecating a few minutes after feeding; the fecal pellets (see Figures 4-11) were often attached to the larvae, connected together like a pearl necklace extruding from the anus that opens ventrally. Strings of fecal pellets were found scattered throughout the inner walls of the hollow internode, usually around feeding scars.
The larva pupates in the upper end of the internode by hanging itself freely inside, attaching either to the vertical internal walls of the internode or to the upper septum (Figures 8, 9). The last segment of the pupa is bifurcate and attached to the exuvia of the last instar larva (Figure 10), which in turn adheres to parenchyma.
Teneral adult weevils that had recently eclosed ( Figure 12) were occasionally found in the dissected internodes (never more than one per chamber). About 2-3 days (n53) after eclosing, the hardened adult chews through the wall of the stem, making a circular to ovalshaped hole, 1.3-1.5 mm in diameter. We observed two exit holes and in both cases the hole was in the middle portion of the internode.
In the field, adult weevils were observed resting or walking on undersides of Cecropia leaves. Most were observed on the upper leaves of saplings, usually on plants with no or few worker ants on the surface (which react aggressively toward adult weevils). Upon disturbance, weevils displayed thanatosis (feigned death; Figure 13). Hespenheide (1987) also noted the occurrence of adult Lissoderes on the undersides of leaves and commented that this is rather unusual in this subfamily. He also noted sexual dimorphism in the rostrum (Figures 12, 13) and that in Lissoderes the males are larger than females, which is the reverse of most other Conoderinae.

Parasitoids of Lissoderes and Azteca
Two species of parasitoids were found associated with larvae of L. pusillus at AMBBR: Neocatolaccus sp. (Hymenoptera: Pteromalidae) ( Figure 14) and Heterospilus sp. (Hymenoptera: Braconidae), both probably undescribed species. These parasitoids presumably oviposit from the outer surface of the stem (there was no evidence that they entered the internode). Both species appear to feed as ectoparasitoids of mature weevil larvae or prepupae, and the adult female probably paralyzes the host prior to oviposition (based on the known behavior of closely related species; Hanson and Gauld 1995). We were unable to determine the range of host stages that are attacked or the site of egg deposition on the body of the host. The mature larva of Neocatolaccus sp. has enlarged dorsal projections and several of the abdominal projections are setose. The pupa hangs from the inner wall of the internode (Figure 14), apparently using its larval skin to attach itself, whereas the pupa of Heterospilus is enclosed in a cocoon. Neocatolaccus was only found in C. obtusifolia and Heterospilus only in C. insignis, although this may be an artifact of the small sample size. Comparing the numbers of parasitoids with the numbers of available hosts (excluding eggs and first instar larvae) in each of these plants, the rates of parasitism were 8% for Neocatolaccus and 14% for Heterospilus. No parasitoids were observed associated with Lissoderes at La Selva Biological Station.
Among the internodes that harbored Azteca ants, 69% contained dead queens at AMBBR and 43% contained dead queens at La Selva. The only parasitoid of Azteca queens that we found at AMBBR was an undescribed species of Menozziola (Diptera, Phoridae), which was observed attacking both A. constructor and A. xanthochroa, and in both species of Cecropia (but primarily in C. obtusifolia). The larvae were found in the gaster of live (but dying) queens, and in several instances we observed dead adult flies in the internodes. The latter presumably died because they cannot chew their way out and are dependent upon the prostoma remaining open or the workers reopening it (one internode of C. obtusifolia with an open prostoma contained a dead queen with phorid larvae and a live worker). In total, 16% of the internodes with dead Azteca queens contained phorids. At La Selva the only parasitoid of Azteca queens observed was Conoaxima affinis Brues (Hymenoptera, Eurytomidae; Figure 16). Here 13% of the internodes with dead Azteca queens contained C. affinis. These percentages are probably underestimated since we did not dissect queens in order to detect earlier stages of the parasitoids. Several of the dead queens were covered with nematodes and infected with entomophagous fungi, although we did not quantify their presence or determine whether the nematodes are parasites or scavengers.

Discussion
Interactions among Lissoderes, Azteca, and Cecropia Two principal questions regarding the biology of Lissoderes are: to what extent is it a parasite of the mutualism between Cecropia and Azteca, and whether these weevils contribute to the high mortality rate observed among colonizing Azteca queens. There are at least three plant characteristics/resources that are utilized by Azteca ants: (1) the prostoma, a soft spot in the stem, where the queens enter; (2) the hollow internodes, where the ants nest, together with the spongy parenchyma that lines the inner walls of the internodes (see below); and (3) the Mü llerian bodies and pearl bodies which the worker ants harvest. The results of our study suggest that L. pusillus depends only upon the second of these. Adult females of L. pusillus do not seem to be dependent upon the prostoma for oviposition, although they do sometimes utilize this spot for making their ovipositional puncture. The weevil larvae never come into contact with the Mü llerian bodies or pearl bodies, but adults have been observed feeding on these food bodies in the field along with other weevils including Pseudolechriops spp. (Conoderinae; Hespenheide and LaPierre forthcoming). More observations are required to determine whether or not adult Lissoderes are dependent on the food bodies.
The larvae of Lissoderes feed almost exclusively upon the spongy parenchyma that lines the inner walls of the internodes; the only exception is that first instar larvae appear to occasionally feed on the upper septum, which in young saplings is soft and spongy. Previous observations (Benson 1985), as well as those from the present study, suggest that Azteca queens also feed on this spongy parenchyma. Internodes occupied by queens show large areas where the parencyhyma has been scraped away and the detritus at the bottom of the internode appears to consist of chewed parenchyma. We were unable to determine whether the ants simply scrape away this parenchyma without ingesting it (many stemnesting ants scrape away the inner walls of the stems in which they nest; J. Longino, personal communication), or whether they actually ingest it. Nor are we aware of any previous studies that have demonstrated that Azteca queens actually ingest parenchyma tissue, thus further research is required to resolve this question. Nonetheless, it seems likely that the parenchyma layer is being utilized by Azteca queens, at least to some extent. Because they seal themselves inside the internode, their only other food resource is the stored reserves in their bodies.
It has been noted previously that among the species of Cecropia that harbor ants, the parenchyma tends to be best developed in those species that regularly grow under conditions of high light intensity (Davidson and Fisher 1991). Of the two species included in this study, C. obtusifolia is more common in secondary vegetation (Burger 1977), and indeed we noted that the parenchyma of this species is thicker, whiter, and more spongy than that found in C. insignis. Our results also provide evidence that these differences in the parenchyma affect both ants and weevils. Azteca queens begin colonizing C. obtusifolia at an earlier stage of plant development than they do in C. insignis (Figure 1), which might be due to an earlier development of parenchyma in the former than in the latter. However, this remains to be demonstrated and the stage at which the prostomata and trichilia begin developing also needs to be considered. Larvae of L. pusillus are more likely to cross over into an adjacent internode in C. insignis than in C. obtusifolia, which could be due to the poorer quality parenchyma present in the former. Although not included in this study, preliminary observations suggest that the ant-free, cloud forest species, C. angustifolia, has even thinner parenchyma than C. insignis. Lissoderes cecropiae Hespenheide larvae were found in the stems of this Cecropia, and here the larvae appear to regularly cross over into several adjacent internodes. Thus at least one congener of Lissoderes is not dependent upon the Cecropia-Azteca mutualism (i.e. the well-developed parenchyma present in Cecropia species that are inhabited by ants). It would be interesting to compare the chemical composition of the parenchyma in these three species of Cecropia.
Assuming that Azteca queens ingest parenchyma, Lissoderes is utilizing a resource that is presumably being provided by the plant for its mutualistic ant. In this sense it is a parasite of the mutualism. However, there appears to be relatively little direct competition between the weevils and the ants. Approximately 20% of all the Cecropia internodes we examined at AMBBR contained only weevils, another 20% contained only Azteca queens, and the other 60% were empty. Only once did we observe a live weevil larva in the same internode as live Azteca queens. Preliminary results from placing weevil larvae in the same internode as ants suggest that the former is often killed. It is likely that female weevils avoid ovipositing in internodes that contain other weevils or live ants, although more research is required to confirm this and determine if female weevils distinguish between occupied and unoccupied internodes through substrate vibrations by active larvae or ants, or through pheromone signals deposited by conspecifics (Ferguson et al. 1999). One would expect selection to favor female weevils that avoid ovipositing in occupied internodes, since there appears to be a high probability that the offspring will be killed if the internode contains another weevil or live ants. It is less clear whether there is an advantage for Azteca queens to avoid colonizing internodes previously occupied by L. pusillus, i.e. whether an internode that has had much of its parenchyma removed is deleterious to the queen(s). Once they become sealed inside the internode, Azteca queens are apparently unable to cut through the septum and move to an adjacent internode, as occurs in Lissoderes. Although we found little evidence to suggest that Azteca queens (live or dead) occupied internodes previously occupied by Lissoderes, this subject requires further investigation. In addition, studies are required to determine whether feeding by weevil larvae has any indirect effects on Azteca queens (e.g. effects on the parenchyma in other internodes or on the production of Mü llerian bodies).
When Azteca workers emerge, they chew their way out of the stem, begin patrolling the plant, and eventually chew through the septa that separate adjacent internodes. Thus it is likely that Cecropia plants become increasingly difficult for Lissoderes to colonize as the worker population grows. We did find weevil larvae in the lower internodes of two saplings of C. obtusifolia in which the upper internodes were inhabited by queens and workers of A. constructor, although the workers had not yet begun chewing through the lower septa. We also found L. pusillus larvae in unoccupied internodes of mature trees, although this has not been observed in lowland sites (L. M. LaPierre, personal observation) where Azteca colonies more completely dominate mature trees (at higher altitudes, such as in AMBBR, Azteca colonies are more feeble; J. Longino, personal communication). Nonetheless, this study found larvae of Lissoderes to be slightly more common in saplings of the same two Cecropia species in lowland sites (Table I).

Azteca mortality
In agreement with previous studies (Longino 1989), we observed a high incidence of internodes containing dead queens, although the causes of this mortality have never been explained. The results of our study suggest that neither direct competition with weevil larvae nor hymenopteran parasitoids are major causes of this mortality. Azteca queens are known to be parasitized by species of Conoaxima (Brues 1922;Wheeler 1942), but we observed none in the stems that we dissected at AMBBR and few from stems at La Selva, and it is generally quite uncommon (J. Longino, personal communication). The only parasitoid of Azteca queens that we found at AMBBR was a phorid fly, Menozziola sp., which appears to account for at least 16% of the mortality. The only previous host records for this genus are from queens of Camponotus and Crematogaster in Europe (Brown et al. 1991;B. Brown, personal communication).
Other possible causes of mortality include quality and quantity of parenchyma in the internode (which in turn could be affected by the presence of other queens in the same internode) and non-insect parasites (e.g. nematodes and fungi). In C. obtusifolia we found greater mortality in internodes containing numerous queens. At this early stage of colonization the queens apparently do not kill each other, but it is possible that they compete for parenchyma, although this remains to be shown and is contradicted by the increased success of pleometrotic associations of A. constructor (Choe and Perlman 1997). In C. insignis, where the queens colonized fewer internodes, the aggregation of numerous queens in a single internode was less pronounced; nonetheless, the majority of internodes that were colonized contained dead queens. Because C. insignis has a thinner layer of parenchyma it is tempting to speculate that queens require a larger internode than they do in C. obtusifolia, but data are currently lacking. The role of nematodes and/or fungi in queen mortality, if any, is also unknown. In South America Jolivet (1991) observed that egg masses of a galerucine chrysomelid oviposited into the internodes of Cecropia had a lethal affect upon ant queens inhabiting the same internodes. However, we did not encounter any egg masses in our samples.
Although there is a high mortality rate among colonizing Azteca queens, this probably has little effect on the mutualism between Azteca and Cecropia (assuming that mortaility is never 100%) since mature trees are inhabited by one colony. In mature colonies it has been shown that nitrogen from ant residues becomes incorporated into the plant (Sagers et al. 2000). Thus, it is possible that the high rate of colonization and mortality benefit a young plant by supplementing nitrogen.

Lissoderes larvae
In addition to its association with Cecropia saplings that also harbor mutualistic ants, another notable aspect of the biology of Lissoderes is the unusual position of the larva, which crawls about on its back and feeds by moving its head backwards. Although we have not undertaken an exhaustive review of the literature, this position appears to be unusual in weevils. The only reference to back-crawling weevil larvae that we are aware of is in the Nemonychidae (May 1994;Kuschel and May 1997). The larvae of Lissoderes have fleshy swellings and setae on the dorsum, which appear to be used in moving about in the interior of the internode (in nemonychids these structures are termed ''ambulatory ampulae''). Setose dorsal swellings have been reported in the final instar larva of Conoaxima, a parasitoid of Azteca queens (Figure 16; Brues 1922), and we observed them in the final instar larva of Neocatolaccus, a weevil parasitoid. More study is required to determine the extent and functional significance of this apparent convergence among the inhabitants of Cecropia stems. biological station at AMBBR, especially the multi-talented Hugo Perez, and La Selva Biological Station, especially the crew at Proyecto ALAS (Danilo Brenes, Ronald Vargas, and Maylin Paniagua). Fieldwork by LaPierre was supported by fellowships from the Department of Ecology and Evolutionary Biology, UCLA, and by personal funds.