southern Central America and adjacent northwestern South America (1.0 PP, 88% BS). Another clade includes P. euryphyllum (= P. longispicum C.DC.) and P. begoniicolor (1.0 PP, 98% BS). These two species, along with several others (including at least one species that occurs at high altitudes in Costa Rica), have been informally recognized as an Andean lineage within sect. Macrostachys by R. Callejas (pers. comm.). This Andean subclade is morphologically distinct and is resolved by our data; thus it appears, pending the inclusion of additional species in future analyses, that Callejas' informal rank is supported. However, the small clade that includes P archeri, P. obtusilimbum, and P. marsupiatum, which are all South American species, is not strongly supported (0.74 PP, 20% BS).
The major clade within sect. Macrostachys (Fig. 9.5, clade C), although not present in all MP trees, is present in both ML and Bayesian analyses. Within clade C, two sister subclades (D and E) are present. The unresolved placements of P fimbriulatum and P gigantifolium are sister to clade C. In several of the MP trees, P. fimbriulatum and P gigantifolium are sister to each other, but are placed in a large, unresolved portion of the tree. Subclade D includes rare species (except P. obliquum, which is widespread) that have distributions restricted to Costa Rica and adjacent Panama. Subclade E is internally unresolved by the ITS, with the exception of P. cernuum, which is resolved as basal to the remainder of sect. Macrostachys s.s. It includes species of Costa Rican endemics, as well as South American species.
Many of the species included in our analyses were represented by more than one individual from more than one location (asterisks, Fig. 9.5). In all cases except for P. imperiale, all replicates of a given species are resolved as closest relatives and were therefore excluded in final analyses. In contrast, different collections of P. imperiale are placed distant to each other: in clade E, and at the base of the tree as unresolved (Fig. 9.5). Taxonomic determinations for some collections from Costa Rica are difficult because they tend to have traits that are characteristic of both P. imperiale and P. biseriatum. Furthermore, these ambiguous specimens are unequivocally resolved among individuals of P. biseriatum and P. imperiale in our ITS trees. It is possible that these problematic specimens represent hybrids, or species not yet identified. Although the collections included in this study were unambiguously identified, the placement of P. imperiale collections in two distant parts of the tree may indicate possible gene flow between P. biseriatum and P. imperiale. If hybridization is in fact implicated, this would represent the first report of hybridization within Piper in the wild. Thus far, the only report of hybridization in Piper is of artificial crosses between the cultivated pepper, P. nigrum, and several of its close relatives (Sasikumar etal. 1999).
9.5. BURGER'S HYPOTHESES REVISITED 9.5.1. Systematic Relationships
The P. obliquum complex (sensu Burger 1971) comprises eight Costa Rican species that Burger recognized as being closely related (Fig. 9.1). In addition to this complex, he posited that six other Costa Rican species were more closely allied to this complex than to any other Piper species that occur in Costa Rica (Fig. 9.1). The P. obliquum complex along with the allied species constitute all of the Costa Rican members of sect. Macrostachys, and, although these 14 species are each other's closest relatives among Costa Rican Piper species (the geographic extent of Burger's study), based on the ITS phylogeny, neither the P. obliquum complex nor the allied species constitute a monophyletic group when analyzed with additional species of sect. Macrostachys that occur outside of Costa Rica (Figs. 9.5 and 9.6). To further test the nonmonophyly, we constrained our ML analysis to force monophyly of the P. obliquum complex. The resulting likelihood scores of the trees were significantly higher than that of the most likely tree (P = 0.001, Shimodaira-Hasegawa test; Shimodaira and Hasegawa 1999). This result is consistent with all analyses of the ITS. In addition, it is clear that Burger did not believe the P. obliquum complex to be monophyletic because he proposed several close relationships between species in the P. obliquum complex with members of sect. Macrostachys not present in Costa Rica.
Although Burger (1971) proposed few systematic relationships within the P. obliquum complex itself, he did suggest that several groups of species were closely related to it (Fig. 9.1). He had doubts regarding the affinities of P. gibbosum. In some instances, he included it in the P. obliquum complex, most closely related to P. euryphyllum, but later in the text he included it in a group with P. aereum and P. melanocladum (Fig. 9.1). Indeed, P. gibbosum has characters in common with both groups. It shares leaf shape, habitat, and the presence of styles (rather than sessile stigmas) with P. euryphyllum, and smaller leaf size, leaf texture, and venation patterns with P. aereum and P. melanocladum. Our analyses do not support a close relationship between P. gibbosum and either one of these two species groups but, instead, place P. gibbosum in the clade containing P. obliquum and several other species (Fig. 9.5, clade D). Although the positions of P. euryphyllum and P. aereum/P. melanocladum are not fully resolved, they do not appear to be closely related to P. gibbosum.
Secondly, Burger (1971) was the first to suggest that P. arboreum and P. tuberculatum were closely related to species of sect. Macrostachys (Fig. 9.1) on the basis of anther form and the lack of an apically developed prophyll. This conclusion has been strongly supported by morphological (Callejas 1986) and molecular data (Jaramillo and Manos 2001; Chapter 10). Again, Burger believed that P. aereum and P. melanocladum are closely related to each other (Fig. 9.1), and that these two species, along with P. gibbosum, form a link between P. arboreum and P. tuberculatum and the remainder of the P. obliquum complex. This conclusion is logical since these five species (i.e., Fig. 9.1(B-C)) are relatively slender and small-leaved plants when compared to the rest of sect. Macrostachys. Indeed, P. arboreum does represent a group of species in our analyses that includes P. tuberculatum and possibly P. cordulatum (a Panama endemic, not sampled) that is placed outside of the core Macrostachys s.s. clade (i.e., Fig. 9.5, clade B), and P. aereum and P. melanocladum are supported as closely related, as Burger suggested. However, the affinities of P. aereum and P. melanocladum to the remainder of the Macrostachys species cannot be determined since the placement of this clade is unresolved in the ITS trees. Piper caracasanum is placed between P. arboreum and the rest of Macrostachys (clade B), but its position is not well supported. Increased sampling and the addition of more variable markers may resolve these nodes, and P. aereum and P. melanocladum may indeed turn out to represent the link that Burger suggested (Fig. 9.1). Again, P. gibbosum is resolved in clade D (Fig. 9.5) and thus is apparently not closely related to these species.
Lastly, Burger suggested that, although P. sagittifolium is a morphologically unusual species, unlike any other in Costa Rica (Fig. 9.2), it is allied to the P. obliquum complex through P. hebetifolium and P. gibbosum (Fig. 9.1). Our data strongly support a close relationship for these three species, with the addition of P. obliquum, and the more recently described P. calcariformis (Fig. 9.5, clade D). However, rather than being distantly related to the P. obliquum complex, these species form a clade that is nested fully within the core of sect. Macrostachys. The placement of P. sagittifolium relative to the remainder of the section is the most apparent inconsistency between our analyses and Burger's hypotheses. Burger believed P. sagittifolium to be isolated and primitive among the Costa Rican Macrostachys species. However, a number of morphological characters appear to support clade D (Fig. 9.5), including flowers with styles (most members of sect. Macrostachys outside of clade D have sessile stigmas), long and recurved stigma lobes, apiculate anthers, and a number of vegetative traits that support obligate associations with ants. In fact, with the exceptions of P. hebetifolium and P. gibbosum, all members of this clade are involved in obligate associations with Pheidole bicornis, the obligate plant ant. The most recent circumscription of Piper obliquum R&P (Tebbs 1989) suggests that it is an extremely variable and widespread species. However, this broad circumscription includes a number of well-defined taxa that warrant specific status (Callejas 2001). The form of P. obliquum represented in this clade (Fig. 9.6, Clade D) appears to be a unique form of the species that is restricted to southern Central America. Of the species included in our analysis, P nobile, P. pseudonobile, P. archeri, and P. caracasanum have all been considered synonymous with P obliquum at one time or another. This topological distribution does not support the broad concept of P. obliquum adopted by some authors.
In addition to proposing systematic relationships among species of Costa Rican Piper, Burger (1972) postulated a number of evolutionary trends. This discussion, however, is limited to those trends related to Macrostachys specifically. Burger considered the flowers of P. sagittifolium to represent the primitive condition in Piper. This conclusion was based on the idea that Piperaceae is derived from an ancestor similar to Saururus Mill. (Saururaceae; Burger 1972), the well-supported sister group to the Piperaceae (Tucker et al. 1993, Savolainen et al. 2000). It follows then, that the flowers of primitive Piper species should be plesiomorphic and resemble those of Saururus. Indeed, similarities in the flowers of P. sagittifolium and Saururus include relatively large parts, long styles with long, divergent stigmas, and unusually large anthers (~1 mm!) on long filaments (to ~2 mm). Furthermore, the inflorescences of P. sagittifolium are not as tightly packed as is typical of many Piper species but, instead, are more loosely associated, similar to those of Saururus. Our analyses strongly support the placement of P. sagittifolium within sect. Macrostachys. If our phylogeny is accurate, parsimony then suggests that any resemblance between the flowers of P. sagittifolium and Saururus are homoplasies rather than plesiomorphies.
Anthers of the majority of Piper have latrorse dehiscence through longitudinal slits. A tendency toward upward dehiscence of anthers is found in several groups of Neotropical Piper species, including sect. Macrostachys. This shift from lateral to upward dehiscence is presumably due to the tight, cylindrical packing of flowers in several species groups (Burger 1972, Jaramillo and Manos 2001). Upward dehiscence among Neotropical Piper species is achieved in two distinct ways. Anthers with parallel thecae and apical dehiscence are found exclusively in a small group of scandent and climbing species (e.g., P. xanthostachyum C.DC., sect. Churumayu; Burger 1972). In contrast, upward dehiscence has also been attained through broadening of the lower part of the anther connective in species of sections Radula and Macrostachys (Burger 1972). This results, in the most extreme cases, in the anther thecae being oriented end to end, 180° to each other. Dehiscence is technically lateral, but because of the expanded connective and the altered orientation of the thecae, pollen is effectively released apically. Sections Radula (e.g., P. aduncum and P. hispidum) and Macrostachys are exceedingly dissimilar morphologically; thus it is not surprising that Burger (1972) proposed that upward dehiscence via expanded connective evolved independently in these two lines. Molecular evidence, however, strongly supports a sister group relationship between Radula and Macrostachys (Jaramillo and Manos 2001; Chapter 10) suggesting that a tendency toward upward anther dehiscence through expanded connectives may be a synapomorphy that unites these two sections.
9.6. ANT-PLANT ASSOCIATIONS IN Piper sect. Macrostachys
Several recent studies have used phylogenetic analyses to address the evolution of ant-plant mutualisms (Michelangeli 2000, Blattner et al. 2001, Brouat et al. 2001, Davies et al. 2001). Interestingly, each study has revealed a unique pattern of evolution of ant-plant associations including numbers of origins and losses of ant associations, trends and correlations in associated plant morphologies, and whether or not plants with intermediate morphologies and facultative associations with ants represent plant species that are ancestrally intermediate between plant species with obligate associations and those with none. Our use of the term intermediate throughout this paper is not with regard to a stage in the evolutionary process, but rather to describe species that are morphologically intermediate and that have facultative associations with ants, i.e., the associations are neither obligate nor random and ephemeral.
Numerous origins of myrmecophytes have been reported in the Melastomataceae, with representatives in nine genera from several tribes (Gleason 1931, Whiffin 1972, Vasconcelos 1991, Morawetz et al. 1992, Michelangeli 2000). A cladistic analysis of morphology for Tococa revealed at least two origins of associations with ants with a minimum of one loss (Michelangeli 2000). Furthermore, the presence, location, and morphology of the domatia appear to be rather plastic at the species level and above (Michelangeli 2000). In Leonardoxa africana (Fabaceae), a single origin of ant interactions among four subspecies is supported, each representing a different degree of association with ants (McKey 1984, 1991, 2000, Chenuil and McKey 1996, Brouat et al. 2001). Brouat et al. (2001) sequenced several chloroplast markers for the four subspecies of L. africana. The subspecies are not resolved as monophyletic groups, but rather are intermixed. Although evolutionary hypotheses could not be determined for L. africana, the study elegantly illustrates how apparent gene flow or insufficient molecular divergence between taxa can obscure evolutionary interpretations (Brouat et al. 2001). In the Asian genus Macaranga (Euphorbiaceae), up to four independent origins of ant-plant associations and numerous reversals are supported through a molecular phylogenetic analysis (Blattner et al. 2001, Davies et al. 2001). In addition, several species of Macaranga exhibit traits previously thought to be intermediate in the evolution of ant mutualisms (Fiala and Maschwitz 1992a). However, these species appear to represent a distinct clade, separate from other myrmecophytes, with an independent origin of myrmecophytism rather than evolutionary intermediates (Blattner et al. 2001, Davies et al. 2001). Phylogenetic studies of ants have also been used to study ant-plant associations with similarly diverse results (Ward 1991, 1999, Ayala et al. 1996, Chenuil and McKey 1996).
The somewhat unresolved nature of portions of the ITS tree (Figs. 9.5 and 9.6), and, to some degree, the absence of natural history information for a number of South American species do not permit us to make unambiguous hypotheses regarding the number of origins of obligate and facultative myrmecophytism in species of Piper sect. Macrostachys. Mapping known ant-plant associations onto the ITS trees, however, suggests that obligate associations evolved independently at least twice, but not more than four times (Fig. 9.6). Again, although the ML (= Bayesian) tree is presented in this paper (Fig. 9.6), none of the MP trees contradict the hypotheses presented here (see Fig. 9.6). It is possible to detect reliably the presence of obligate ant-plant associations from herbarium specimens owing to the presence of persistent, conspicuous pearl bodies in the petiole chambers, and, in most specimens, the remains of numerous ants or antparts. Examination of herbarium specimens does not suggest the presence of any obligately myrmecophytic species in sect. Macrostachys additional to those included in this study.
The obligate myrmecophyte P. cenocladum, placed in clade E (Fig. 9.5) is isolated from the other four species of obligate myrmecophytes and, thus, presumably represents an independent origin. Three of the remaining myrmecophytes are placed in clade D. Two equally parsimonious reconstructions of obligate myrmecophytism are possible in this clade (Fig. 9.6). One reconstruction has a single gain of obligate associations with two independent and complete losses. Alternately, obligate associations could have evolved twice independently within this clade: once in the P. obliquum/P. sagittifolium clade (0.96 PP, 96% BS) and again in P. calcariformis.
The position of the obligate myrmecophyte, P. fimbriulatum, is unresolved. Thus it is unclear whether this species represents an additional origin of obligate myrmecophytism. Piper fimbriulatum resembles the rest of the species in clade D (Fig. 9.5) in having long, recurved stigma lobes, but differs markedly in leaf texture and absence of a style (i.e., sessile stigmas). Additional species and more informative markers are required to further evaluate the hypotheses proposed here, and to resolve the ambiguities in our analyses.
Facultative associations are more ephemeral and inconspicuous than obligate associations. Often only one to several petioles on a given plant are occupied by ants (E. Tepe, pers. obs.). On the basis of species that we have actually observed in the field with resident ants, our analyses suggest that facultative associations evolved independently two to three times (Fig. 9.6). However, large sheathing petioles with persistent margins are typical of a majority of Macrostachys species, and it is likely that the petioles of many more species than we have observed in the field form closed shelters and are therefore possibly inhabited by ants. Detection of facultative myrmecophytes in herbarium specimens is more ambiguous than it is for obligate associations. The degree of petiole closure is not preserved in dried specimens, but petiole morphology can suggest that facultative associations are likely to occur in a given species. Examination of herbarium specimens has led us to predict that the phenomenon of facultative associations is probably much more common and widespread than we have observed in the field. When predicted facultative associations are mapped onto the tree, a single origin is supported within the Macrostachys clade (Fig. 9.6). Consequently, the hypothesis of the number of origins of facultative associations based only on plants observed in situ with resident ants is the most conservative, and it is undoubtedly an underestimation.
9.6.2. Evolution of the Mutualism
All obligate myrmecophytes in sect. Macrostachys are characterized by tightly closed petioles with pearl body production localized on the inner surface of the petiolar tube. Accordingly, the distribution of these two characters parallels obligate myrmeco-phytism (Fig. 9.6). Similarly, hollow stems are restricted to the plants with obligate associations. But because the stems never become hollow in P. calcariformis, a minimum of two and a maximum of three independent origins are required to explain the distribution of hollow stems, or stems that the ants are capable of or inclined to excavate, as the case may be. None of the facultative ant-plants studied thus far have hollow stems; however nonhomologous cavities are occasionally formed by stem-boring insects (E. Tepe, pers. obs.).
According to our current phylogenetic hypothesis based on the ITS data, the most parsimonious reconstructions of the evolution of obligate associations with ants is three to four independent gains, with up to two losses (Fig. 9.6). The number of gains depends on the placement of P. fimbriulatum and the resolution of the clade containing P. obliquum; the number of losses depends entirely upon the resolution of the latter. It seems almost certain, however, that the origin of myrmecophytism in P. cenocladum is independent of that for the P. obliquum clade (Fig. 9.5, clade E).
The hypothesis that obligate associations evolved twice in Clade D (Fig. 9.5), once for P. obliquum/P. sagittifolium, and independently in P. calcariformis, is intriguing because P. calcariformis is unique among the obligate ant-plants in that it consistently has solid stems (Fig. 9.3). The stem anatomy of P. calcariformis is different from the other four obligate myrmecophytes in that the medulary vascular bundles are scattered throughout the pith (which is the most common arrangement among species of sect. Macrostachys). In contrast, the medulary bundles of the obligate species with hollow stems are arranged in a second ring just interior to the primary ring (Fig. 9.2(C)). In these plants, the pith is reportedly excavated by the resident ants (Risch et al. 1977, Letourneau 1998, Dyer and Letourneau 1999). It is unknown whether bundles throughout the pith of P. calcariformis and the rest of the large-stemmed species of Macrostachys (Fig. 9.5) would preclude excavation by ants, but the bundles of the other four obligate myrmecophyte species with hollow stems do not extend into the area of pith that becomes hollow (Fig. 9.2). Alternatively, if the obligate myrmecophytes in the P. obliquum clade (Fig. 9.5, clade D) are the result of a single origin, the loss of the associations in P. gibbosum and P. hebetifolium is not surprising, since neither of these species is morphologically suited to support ant residents; both of these species have small petioles that do not close tightly, and stems that are more slender than any of its close relatives (Fig. 9.5). Furthermore, the petiole margins of P. gibbosum are caducous, leaving behind a broadly U-shaped petiole (Fig. 9.4(G)). It appears that the slender stem morphology is derived in these two species, perhaps in response to their mid- to high elevational habitats. It is possible that the mutualism between Piper and Pheidole bicornis is not as stable at higher elevations as at lower elevations. Myrmecophytic species of Cecropia are more frequently found with ant inhabitants at lower elevations, and increasingly less so as altitude increases (Wheeler 1942, Janzen 1973). It is possible that the same phenomenon is responsible for the lack or loss of associations in P. gibbosum and P. hebetifolium.
Multiple origins and losses of obligate myrmecophytism appear to be common in the ant-plant associations that have been studied in other genera thus far. For example, two to four origins, with numerous losses, are supported in Macaranga (Euphorbiaceae; Blattner et al. 2001, Davies et al. 2001), and a minimum of two gains and one loss are supported in Tococa (Melastomataceae; Michelangeli 2000, also see Davidson and McKey 1993). Davies et al. (2001) proposed that multiple origins of myrmecophytism in Macaranga may be the result of the combination of certain morphological traits and a specific ecological and biogeographical setting. For example, the ancestors of the Malesian species of Macaranga that have given rise to myrmecophytes appear to have had large-diameter stems with soft pith and food bodies, and they occurred in a climate that allowed uninterrupted food body production (Davies et al. 2001). In other words, it appears that myrmecophytism has most likely evolved in plants that had a morphological predisposition for supporting such associations, and that were located in constant, tropical environments. According to this hypothesis, the combination of such characters as large, sheathing petioles, large-diameter stems, possibly widespread pearl body production, and a climatologically constant habitat that also includes Pheidole bicornis have contributed to the development of ant-plant associations in Piper.
A number of studies have demonstrated that ant partners of obligate associations are rarely species-specific, and that no ant-plant association studied thus far is the result of parallel cladogenesis (Mitter and Brooks 1983, Ward 1991, Ayala et al. 1996, Chenuil and McKey 1996). The fact that a single ant species, Ph. bicornis, is associated with all five obligate species of sect. Macrostachys, excluding the possibility of cryptic species of ants, precludes the possibility that species-for-species coevolution between ant and plant species has taken place. In fact, that one ant species is obligately associated with several plant species is not surprising. Frequent host switching of an ant species among host plant species appears to be common among many groups of plant ants, namely Azteca (Ayala et al. 1996), Crematogaster (Blattner et al. 2001), and Pseudomyrmex (Ward 1991). Thus, it appears that multiple origins of obligate ant-plants, and host switching by obligate plant ants, is common and that Ph. bicornis is capable of switching between the different obligate myrmecophyte species in Piper as well.
The mutualism with ants in obligate myrmecophytes is not maintained by a single plant character, but rather a suite of characters that is implicated in the associations. As the number of concurrent characters increases among species (i.e., tightly closed petiole sheaths, obligate-type pearl bodies, restricted areas of pearl body production, hollow stems in four of the five species), it becomes increasingly unlikely that these characters evolve in parallel or through convergence. In order to test the possibility that this suite of characters, and thus obligate myrmecophytism, evolved only once, we constrained our analysis to force the monophyly of the obligate myrmecophytes. The MP analysis produced the shortest trees, which were 11 steps longer than the shortest tree in our unconstrained analysis (125 vs. 114 steps). The ML analysis resulted in trees with likelihood values significantly higher than the most likely tree's from the unconstrained analysis when they were compared with the Shimodaira-Hasegawa test (P = 0.001). When the ML analysis was constrained to force the obligates into a monophyletic clade with the facultative myrmecophytes as a paraphyletic clade basal to them, so as to suggest that the facultatives are evolutionary transitions between species without associations and with obligate associations, the resulting trees were also longer (MP: 129 vs. 114 steps, ML: P = 0.002). Thus, according to our current analyses, neither the obligate nor the facultative myrmecophytes in Macrostachys can be explained by a single evolutionary origin, and the facultative myrmecophytes are not transitional between obligate myrmecophytes and species that lack associations with ants.
9.6.2b. Petiolar domatia and facultative associations
Given that petioles form the primary domatia in Piper ant-plants, evolution of facultative myrmecophytism cannot be discussed separately from petiolar morphology. In fact, no close, species-specific relationship has developed in any ant-plant system studied thus far that only provides food for the ant partners, but no shelter (Fiala and Maschwitz 1992b). The petiole cavity of the obligate species is tightly closed throughout its length, and the petiole margins are pressed tightly to the stem such that little, if any, water running down the stem enters the cavity. This morphology is exceptionally constant among the five species of obligate ant-plants in Piper. The degree of petiole closure of the facultative myrmecophytes is more variable, but several petioles on a given plant are often closed enough so as to provide sufficient shelter for ant colonies.
The ITS phylogeny supports two to three independent origins of facultative myrmecophytism, based solely on the species that we have observed in the field with ants nesting in the petioles. However, we have only had the opportunity to study 14 species of Macrostachys in the field (Table 9.1). On the basis of the examination of herbarium specimens of species that we have not studied in vivo, we have observed that most species of sect. Macrostachys have large, sheathing petioles; consistent with the observed correlation of plant morphology and ant occupancy of species that we have observed in the field, we believe that facultative mutualisms are much more taxonomically and geographically widespread than we have reported here. If these potentially facultative ant-plants are mapped onto the phylogeny, then a single origin of facultative associations near the base of sect. Macrostachys is supported, with several independent losses, and with the obligate myrmecophytes derived from the facultative species (Fig. 9.6). Under this scenario, facultative myrmecophytes may represent evolutionary precursors to the obligate myrmecophytes. In fact, Risch et al. (1977) suggested that the mutualism between ants and Piper might have originated with the evolution of large, sheathing petioles. This suggestion and our predictive tree corroborate the findings of Fiala and Maschwitz (1992b) that domatia are the most important plant trait for the development of myrmecophytism. However, additional data and taxa are needed before this hypothesis can be more fully tested.
The ant genera that have been found nesting in the petioles of P. biseriatum and P. imperiale (e.g., Crematogaster spp., Solenopsis spp., Wassmannia spp., and other species of Pheidole) are opportunistically nesting, arboreal ants (Holldobler and Wilson 1990, Orivel and Dejean 1999). These ants apparently nest in petioles of Piper species whenever they encounter one that provides sufficient protection from the environment.
Arboreal ants occasionally nest in cavities formed by other stem-boring insects in stems, petioles, and even the leaf midveins (E. Tepe, pers. obs.). Ward (1991) noted that a number of arboreal pseudomyrmecine ants have a tendency to nest in cavities in living plant parts, as opposed to dead, hollow twigs, as is typical of most opportunistic, arboreal pseudomyrmecines. He suggested that ant species that nest in living plant parts might lend insights into the evolution of obligate ant-plant associations (Ward 1991). However, phy-logenetic studies of Pheidole are currently unavailable that would allow us to determine the nesting habits of species related to Ph. bicornis. Fiala et al. (1994) found that the presence of facultative ants in Macaranga can dramatically reduce damage by herbivores, and therefore may be important in driving plants toward more complex and mutually beneficial associations. This is likely the case in Piper as well.
Fischer et al. (2002) studied the chemical composition of pearl bodies of the four hollow-stemmed species of obligate myrmecophytes in Piper and found that, with the exception of slightly different levels of soluble carbohydrates and proteinaceous nitrogen in P. sagittifolium, pearl body composition did not vary significantly between species. This similarity in pearl body composition could be explained by common ancestry, but may also be explained by selective pressures exerted by the nutritional requirements of the ants. The ants derive the majority of their sustenance from the pearl bodies (Fischer et al. 2002). Food bodies are undeniably important, but in Piper, as in Macaranga, they appear to be second to domatia as the most important factor in the development of obligate ant-plant associations (Fiala and Maschwitz 1992b).
In all plant genera that have obligate mutualisms with ants, very few, if any, of the plant parts implicated in the associations evolved completely de novo. With the possible exceptions of the Beltian bodies in Acacia and the collagen-containing Mullerian bodies of Cecropia, all ant-associated plant traits are modifications of preexisting structures (Janzen 1966, Rickson 1973). In Piper, the petioles of the obligate species are not fundamentally different from those of many species of sect. Macrostachys, except that they are more tightly folded and more consistently closed. In fact, Fiala and Maschwitz (1992b) noted that only Macaranga species with a predisposition for domatia developed into obligate myrmecophytes, and this appears to hold true for Piper as well. Stem anatomy of hollow-stemmed species appears to differ in that the medulary vascular bundles do not extend as far into the pith as in species with solid stems (Fig. 9.2). The arrangement of vascular bundles is novel, but again, are no more than modifications of preexisting structures.
Although no pearl bodies were found in the petioles of the facultative ant-plant species, we have frequently observed structures resembling pearl bodies on leaf and young shoot surfaces of P. aduncum, P. nigrum, and P. tuberculatum growing in greenhouses and even on leaves of P. auritum for sale in a Mexican market in Nashville, TN (E. Tepe, pers. obs.). Furthermore, they have been observed on young shoots and leaves on a number of Piper species in the field (L. Dyer, pers. comm.). It is possible that these structures function as generalized ant attractants in some species of Piper, in lieu of extrafloral nectaries, but may not have been reported from plants in the field because they are removed continuously by ants and leave no macroscopically visible trace. The difficulty of detecting food bodies on exposed plant surfaces has also been reported in Macaranga (Fiala and Maschwitz 1992b); however, microscopically visible traces of food bodies have been observed (Hatada et al. 2001). These extrapetiolar bodies have a somewhat different appearance than those found in the petioles of obligate myrmecophytes in that they are recognizably larger and more translucent. However, they also contain lipids and proteins as do the pearl bodies of obligate myrmecophytes (Sudan IV and Bromphenol Blue spot staining respectively; methods from Baker and Baker 1975), and around 2% carbohydrates when freshly extracted contents were measured with a refractometer (E. Tepe, unpubl. data; method from Kearns and Inouye 1993). If, in fact, these bodies are homologous with the pearl bodies found in the petioles of the obligate species, then pearl bodies, which are so important to the maintenance of ant-plant mutualisms (Fiala and Maschwitz 1992b), are also preexisting structures modified only in size and location.
Was this article helpful?