Annals of Botany 92: 807-834, 2003
© 2003 Annals of Botany Company
The Pollination Ecology of an Assemblage of Grassland Asclepiads in South Africa
1 Landscape and Biodiversity Research Group, School of Environmental Science, University College Northampton, Park Campus, Northampton NN2 7AL, UK and 2 School of Botany and Zoology, University of Natal, Private Bag X0I, Scottsville, Pietermaritzburg 3209, Republic of South Africa
* For correspondence. E-mail jeff.ollerton{at}northampton.ac.uk
Received: 14 March 2003; Returned for revision: 17 July 2003; Accepted: 10 September 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
The KwaZulu-Natal region of South Africa hosts a large diversity of asclepiads (Apocynaceae: Asclepiadoideae), many of which are endemic to the area. The asclepiads are of particular interest because of their characteristically highly evolved floral morphology. During 3 months of fieldwork (November 2000 to January 2001) the flower visitors and pollinators to an assemblage of nine asclepiads at an upland grassland site were studied. These observations were augmented by laboratory studies of flower morphology (including scanning electron microscopy) and flower colour (using a spectrometer). Two of the specialized pollination systems that were documented are new to the asclepiads: fruit chafer pollination and pompilid wasp pollination. The latter is almost unique in the angiosperms. Taxa possessing these specific pollination systems cluster together in multidimensional phenotype space, suggesting that there has been convergent evolution in response to similar selection to attract identical pollinators. Pollination niche breadth varied from the very specialized species, with only one pollinator, to the more generalized, with up to ten pollinators. Pollinator sharing by the specialized taxa does not appear to have resulted in niche differentiation in terms of the temporal or spatial dimensions, or with regards to placement of pollinaria. Nestedness analysis of the data set showed that there was predictability and structure to the pattern of plant-pollinator interactions, with generalist insects visiting specialized plants and vice versa. The research has shown that there is still much to be learned about plant–pollinator interactions in areas of high plant diversity such as South Africa.
Key words: Apocynaceae, Asclepiadaceae, plant assemblage, community structure, floral morphology, grassland, mutualism, pollination, nestedness, niche, species interactions, SEM, South Africa.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
Niche-partitioning between co-occurring species of plants may facilitate the development of species-rich plant communities. Most studies of niche-breadth and overlap in plant communities have been focused on the physiology and growth form of plants (e.g. Tofts and Silvertown, 2000; Silvertown et al., 2001). Much less attention has been paid to the possibility that high diversity within plant communities is facilitated by niche-partitioning in the reproductive biology of plants (Parrish and Bazzaz, 1979; Armbruster, 1995). Some attempts were made to ascribe differences in flowering times between plants to niche partitioning (e.g. Robertson, 1895; Heithaus, 1974; Rathcke, 1988), although it is possible that much of the between-taxa variation in phenology is a result of phylogenetically constrained flowering times (Kochmer and Handel, 1986; Ollerton and Lack, 1992; Johnson, 1993). Recently, there has been renewed interest in partitioning of pollination systems in the context of food-web theory (Memmott, 1999; Dicks et al., 2002). It is hypothesized that plant species sharing pollinators may experience inter-specific competition resulting from diminished rates of pollinator visitation, pollen wastage and stigma clogging (Waser, 1978). In general, reproductive interference is expected to be greatest among closely related species pollinated by the same vectors (Armbruster et al., 1994; though see Borba and Semir, 2001).
Studies of plant–pollinator interactions involving assemblages of phylogenetically related, co-occurring plant species can therefore potentially yield insights into the ecology of niche overlap, competition for resources (pollinators), adaptation and the evolution of convergent suites of floral traits (‘syndromes’) in species sharing common descent (Sakai et al., 1999; Kessler and Krömer, 2000; Borba and Semir, 2001; Johnson et al., 2002). The grasslands of KwaZulu-Natal offer excellent opportunities for such studies as they are ancient, relatively intact, botanically rich and characterized by high levels of endemism and local radiation of genera (Cowling et al., 1991; O’Connor and Bredenkamp, 2000). Alpha diversity of grassland at the 1000 m2 plot level exceeds all other biomes in southern Africa, including the renowned Cape fynbos, and the biome as a whole contains almost 4000 plant species (Cowling et al., 1991).
In this paper we present the results of a study of the pollination ecology of an assemblage of asclepiads (Apocynaceae subfamily Asclepiadoideae sensu Endress & Bruyns, 2000) co-occurring and co-flowering in an upland grassland in Kwazulu-Natal, South Africa. The asclepiads are of particular interest because of their specialized floral morphology, comprising three highly evolved key features: (1) the gynostegium—congenitally fused and highly synorganized gynoecium and androecium; (2) pollinaria—pollen aggregated into masses (pollinia) with mechanical clips (translators) which attach them to pollinators; and (3) the corona—more or less well-developed outgrowths of the corolla and/or stamens which serve a number of possible functions (Brown, 1811; Kunze, 1990, 1997; Liede and Kunze, 1993; Fishbein, 2001; Ollerton and Liede, 2003). Although the pollination biology of the asclepiads has received quite a lot of attention (for a review, see Ollerton and Liede, 1997) there are rather few comparative studies of co-occurring species assemblages (Kephart, 1983, 1987; Vieira and Shepherd, 1999).
South Africa is one of the main centres of diversity and endemism for the asclepiads; there are about 90 genera and 700 species of Apocynaceae in southern Africa (Victor et al., 2000) and the KwaZulu-Natal grasslands are particularly rich. There are few data on the floristics of grasslands in southern Africa, but Hilliard and Burtt (1987) list the asclepiads with 234 species as the fifth largest plant family in the province of KwaZulu-Natal. Cowling and Hilton-Taylor (1997) show that the asclepiads are the eighth largest family in southern Africa, with a remarkable 87% endemism.
Despite this high level of diversity and endemism, there has been a paucity of studies of the pollination ecology of asclepiads in southern Africa, although see Weale (1873), Scott-Elliot (1891) and Liede and Whitehead (1991). The majority of fieldwork has concentrated on the succulent stapeliads (tribe Ceropegiae; for a review, see Meve and Liede, 1994) but for most asclepiad genera absolutely nothing is known of their pollinator requirements, levels of specialization, breeding systems, etc. Our study addressed the following questions.
(1) How ecologically specialized are these asclepiads in their pollinator requirements?
(2) To what extent do co-occurring species share pollinators and therefore overlap in their pollination niche? Is the likelihood of pollinator-sharing linked to phylogenetic relatedness, i.e. the possession of shared floral traits?
(3) Are the patterns of flower visitation random with respect to the overall diversity of flower visiting insects in the community? Or is there filtration of visitors by plants that would lead to predictable structure in the patterns of interaction?
(4) Do non-related species which share pollinators show evidence of convergent evolution of floral traits, and therefore can specific features of the floral morphology of the studied asclepiads be ascribed to adaptations for particular specialized pollinators?
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
The study site was an 8-ha area of moist montane grassland on the farm Wahroonga (30°08'E, 29°37'S, altitude approx. 1200 m) that is formally protected as a South African Natural Heritage site. The grasslands at this site have never been ploughed and are not grazed by domestic livestock.
Fieldwork was carried out between 15 Nov. 2000 and 16 Jan. 2001. In total, 63 man-hours were spent observing and collecting flower visitors of nine species of asclepiads. Most of the observation was diurnal, with limited nocturnal surveillance. Samples of insect visitors were returned to the laboratory for analysis of pollinaria loads. Voucher specimens are kept in the university collection of S.D.J.
Flowers of all species were bagged with fine mesh cloth bags to determine nectar volume and concentration following 24 h accumulation. Volume was determined using microcapillary pipettes and sugar concentration (as sucrose equivalents) using Bellingham and Stanley sugar refractometers (for details of these techniques, see Dafni, 1992; Kearns and Inouye, 1993).
Flowers of all species were collected for closer observation in the laboratory, including scanning electron microscopy and reflectance spectrophotometry. Spectral reflectance of flower parts was measured with an Ocean Optics S2000 spectrometer and fibre optic reflection probe (UV/VIS 400 µm) held at 45° to the petal surface. An Ocean Optics DT-mini deuterium tungsten halogen light source, with an approx. 200–1100 nm spectral range, and an Ocean Optics WS-1 diffuse reflectance standard to calibrate the spectrometer were used.
In addition, pollinaria removal and pollinia insertion were recorded from the collected flower material (sample range 11–35 flowers per species). Vouchers of all plants are kept in the spirit collection of the Natal University Herbarium (see Table 1 for collection numbers).
|
On the 16 Dec. 2000, two 1-h surveys of the insect visitors present on flowers of all plant species encountered within a 2-m-wide strip, in two circuits that covered the whole of the extent of the field site, were carried out. These surveys were undertaken simultaneously by two groups of observers and insects were identified to broad taxonomic group as ‘fly’, ‘hoverfly’, ‘bee’, ‘butterfly’, etc., except in those cases in which we were positive of the identity of the insect (usually because they were important visitors to asclepiad flowers).
Statistical analysis
The data collected on floral traits were entered into a species-trait matrix (see the Appendix) and subjected to multidimensional scaling analysis (e.g. Ollerton and Watts, 2000) This and all other statistical analyses were performed using SPSS 8.0 for Windows (1997, SPSS Inc. Chicago, IL, USA). All means are presented ± 1 standard deviation.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
The study species
Our survey of the Wahroonga site during November 2000 to January 2001 yielded over 20 species of asclepiads, although the actual total for the site is probably higher. Many of these species flowered too early or too late, or had too low a population density, to be included in our comparative study of asclepiad pollination biology. This study therefore focused on nine abundant, co-flowering species. Detailed descriptions of the growth forms and distribution of these species will be presented elsewhere by the authors.
These species varied in their flower size and colour (Figs 1 and 2) and in other floral traits such as morphology, longevity, scent and nectar characteristics (Table 1 and Figs 6–9). The nine species represent six genera belonging to two tribes of the asclepiads, the Ceropegieae [Sisyranthus trichostomus (Harv.) K. Schum.] and the Asclepiadeae [Xysmalobium involucratum Decne., Xysmalobium gerrardii Scott Elliot, Miraglossum verticillare (Schltr.) Kupicha, Miraglossum pilosum (Schltr.) Kupicha, Pachycarpus natalensis N.E. Br., Aspidonepsis diploglossa (Turcz.) A. Nicholas & D.J. Goyder, Asclepias woodii (Schltr.) Schltr., Asclepias cucullata (Schltr.) Schltr.]. They therefore represent different levels of phylogenetic relatedness; however, the exact relationships between these taxa is unclear as a preliminary molecular phylogenetic study has shown that a number of genera (e.g. Xysmalobium and Asclepias) are not monophyletic (D. J. Goyder, A. Nicholas and S. Liede, unpubl. res.). Conclusions that rely on assumptions about the phylogeny of these taxa have therefore been avoided.
|
|
|
Photography using a UV filter indicated that for most species there was very little floral ultraviolet reflectance (not shown) and this was confirmed by spectrophotometric analysis (Fig. 2). Contrast in spectral reflectance between the corona and corolla was evident only in Pachycarpus natalensis and Asclepias cucullata. In general, the spectra of the study species were characterized by lack of spectral purity (chroma), hence the predominance of ‘muddy’ brown, purple and creamy colours to the human observer. This may, in turn, reflect the importance of scent, rather than visual, advertising for the attraction of beetles and wasps (see Pollinators, below).
The nine species varied considerably in the size and shape of their pollinaria (Fig. 12).
|
Pollinators
In total we recorded over 60 species of insect flower visitor to the nine asclepiads (Table 2; for some examples, see Fig. 1), excluding ants and other very small insects which were considered to be too small to remove pollinaria. The majority of visitors, however, were uncommon and usually did not pick up pollinaria (Table 2). They must therefore be considered as opportunistic nectar thieves rather than legitimate pollinators.
|
The numbers of pollinators recorded for the nine asclepiads ranged from 1 to 10 (mean = 3·1 ± 3·1, median = 2·0, n = 7 asclepiad species, excluding the two species for which pollinators were not determined exactly—see below). There was a strong correlation between number of flower visitors and number of pollinators per plant species (Pearson’s r = 0·93, d.f. = 5, P = 0·003; once again excluding the two species for which pollinators were inferred but not determined). The mean ratio of flower visitors to pollinators was 6·0 (± 3·4, range 2·7–12·0).
The species Pachycarpus natalensis and Miraglossum verticillare were both pollinated exclusively by male and female individuals of Hemipepsis hilaris (Hymenoptera: Pompilidae) (Fig. 1B). Interestingly, these were also the only two species that have flower parts that reflect ultraviolet light, i.e. wavelengths below about 400 nm (Fig. 2). This spider-hunting wasp was the sole visitor to the less common species Miraglossum pilosum and, whilst none of the pompilids collected on M. pilosum carried pollinaria, it is thought that this is the legitimate pollinator, and that the absence of pollinaria-carrying individuals of Hemipepsis hilaris is due to the low sample size (Table 2). Unlike the other two pompilid wasp-pollinated taxa, the flowers of M. pilosum do not reflect in the ultraviolet, and neither do the other three asclepiads for which H. hilaris is a minor pollinator or flower visitor (Table 2). This suggests that UV reflectance per se is not an important attraction trait for flowers visited by this spider-hunting wasp, but its role in the pollination biology of P. natalensis and M. verticillare remains unclear and requires further study.
The pollinators of Asclepias woodii and Sisyranthus trichostomus were chafers (Coleoptera: Scarabaeidae: Cetoniini) belonging to two species, Atrichelaphinis tigrina and Cyrtothyrea marginalis. The single record of Apis mellifera carrying pollinaria of S. trichostomus must be considered unusual in relation to the large number of visits by the chafers. Atrichelaphinis tigrina appears also to be the main pollinator of Xysmalobium involucratum, although visitation rates to this asclepiad were rather low, and H. hilaris was also captured carrying pollinaria on one occasion.
Of the remaining three species, Xysmalobium gerrardii appears to be the most generalist, with H. hilaris, Atrichelaphinis tigrina, Aspilocoryphus fasciativentris (Heteroptera: Lygaeidae: Lygaeoidea) and Sarcophaga sp. (Diptera: Sarcophagidae) all recorded as major pollinators, in addition to a range of less common pollinators. The lygaeid bug Aspilocoryphus fasciativentris was a very common and distinctive component of the pollinator fauna of this species and individuals regularly carried pollinaria on the tarsi of one or more legs. These lygaeids were very active within plants, but their flying abilities and the extent of movement between plants is unknown.
Low visitation rates make it difficult to be certain, but Aspidonepsis diploglossa appears to be primarily bee pollinated, with occasional visits by Diptera which also do carry pollinaria. The final species, Asclepias cucullata, is infrequently visited by Diptera, bees and other insects (Table 2). None of these visitors carried pollinaria, and so it can only tentatively be classified as possibly mainly bee pollinated.
An analysis of the cumulative observation of new flower visitors over time (Fig. 3) indicates that, by day 23, an asymptote had reached, beyond which no new flower visitors were recorded. This masks a great deal of individual variation between the asclepiad species, with some very specialized species reaching an asymptote following as little as 2–5 d of sampling (data not shown). However, these results suggest that, in this assemblage at this site during this flowering season, the majority of the flower visitor diversity for the asclepiads had been sampled.
|
Pollination niche breadth and overlap
Pollination niche breadth was calculated for each of the eight asclepiads for which pollinators were identified, following the approach of Levins (1968), as presented in Waite (2000). With reference to Fig. 11A, niche breadth of the generalist X. gerrardii and the bee (?)-pollinated A. diploglossa were the highest of the eight species, with the three chafer-pollinated species having a rather narrower niche breadth. The three pompilid wasp-pollinated taxa each had the smallest possible niche breadth (0·0), as only one species of pollinator was identified. It appears that the chafer-pollinated species, although exhibiting a number of specific adaptations to pollination by cetoniid beetles (see below), are rather less specialized than the pompilid wasp-pollinated taxa.
|
Pollination niche overlap was calculated for all pairs of species, again following the approach by Levins (1968), as presented in Waite (2000) (Fig. 11B). Comparison of mean pollination niche overlap between taxa with different pollination systems reveals some interesting patterns (Fig. 11C). The pompilid wasp-pollinated taxa show complete pollination niche overlap, whilst the chafer-pollinated taxa show rather less overlap, though the average is still high. Niche overlap between the generalist X. gerrardii and the pompilid wasp-pollinated taxa is higher than that with the chafer-pollinated taxa. Overlap between the chafer- and pompilid wasp-pollinated taxa is low, though not zero (because the pompilid wasp Hemipepsis hilaris was an occasional visitor to the otherwise chafer-pollinated X. involucratum). The bee(?)-pollinated A. diploglossa showed almost zero overlap with the other species, though Apis mellifera was an occasional visitor to both it and the otherwise chafer-pollinated S. trichostomus.
Assemblage structure of the flower visitors and pollinators
The flower visitor and pollinator data sets were subjected to nestedness analysis using the AICS Research Inc. software (http://www.aics-research.com/research/index. html) (Atmar and Patterson, 1993). The degree of nestedness within a data set refers to the extent to which ‘smaller’ samples are subsets of ‘larger’ samples. This type of analysis has classically been performed for island biotas, to calculate the extent to which species present on small islands comprise a subset of those on larger islands (Patterson, 1990).
In the present analysis, the approach of Bascompte et al. (2003) and Dupont et al. (2003) is followed. The asclepiad species are analogous to islands, ranked by degree of specialization (number of flower visitors or pollinators) rather than ‘island size’. The flower visitors/pollinators are in turn ranked according to the number of plant species that they visit. The data can be considered as a matrix of presence–absence values for interactions between each plant and each insect (Fig. 4A and B). A perfectly nested data set would fill the matrix with a triangle of positive data running bottom left to top right of the matrix. In both the flower visitor matrix (Fig. 4A) and the pollinator matrix (Fig. 4B) that triangular pattern is strongly suggested, though not perfectly realized. The descriptive statistic used to describe the degree of fit to the nested ideal is the matrix ‘temperature’, defined as the ‘heat of disorder’, i.e. the degree of order or disorder in the matrix, relative to the idealized, perfectly nested pattern (Atmar and Patterson, 1993). A matrix temperature of 0° indicates a perfectly nested pattern, whilst one of 100° indicates a completely random pattern (Atmar and Patterson, 1993). The probability that this pattern, and the resultant matrix temperature, are non-random can be calculated using Monte-Carlo simulations of the original data set. The flower visitor and the pollinator data sets both show a significantly nested pattern of insect use amongst these asclepiads: flower visitor matrix temperature = 18·02°, P = 0·00007 based on 1000 Monte-Carlo iterations. Pollinator matrix temperature = 14·34°, P = 0·05 based on 1000 Monte-Carlo iterations.
|
As a refinement of this approach we have overlain insect abundance values (from Table 2) on the nestedness matrix for flower visitors (Fig. 4C). This shows that those insect species which visit a wide range of asclepiad species (i.e. those which cluster in the upper left of the matrix) are also those species that are in greatest abundance. This suggests that insect population size may be one factor determining the relative breadth of generalization of flower visiting insects, in terms of numbers of plant species visited.
Community flower visitor survey
The combined total for the two 1-h surveys was 381 individual insects in 14 broad taxonomic groups (Table 3). The purpose of this survey was to ascertain if insects were visiting flowers of the asclepiad species in proportion to their abundance at the site or, conversely, whether the asclepiad flower visitors were a selected subset of the flower-visiting insects in the community. The ideal statistical test for such data would be a test of frequencies (such as the 
2- or G-tests). However, the large number of zero observations for many of our categories of insect flower visitor renders these tests invalid. As an alternative approach, the Spearman rank correlation of proportional abundance of each visitor group in the community survey was calculated against the proportion of that insect group on each asclepiad species. In all cases there was no statistically significant relationship; the asclepiad flower visitors do not visit those flowers in proportion to their overall abundance (Table 3). Clearly, the specific floral traits of the asclepiads filter particular flower visitors. Some groups of flower visitors are notably rare in the community (e.g. Hemipepsis hilaris, Cyrtothyrea marginalis and Heteroptera) but common as pollinators of asclepiads. Conversely, there are some insect groups which are rare as asclepiad flower visitors in relation to their community abundance (Apis mellifera and butterflies). All of the asclepiads that were studied overlapped broadly in their flowering phenologies, although they did not necessarily coincide in their peak flowering times. This survey was conducted almost exactly mid-way through the study period. Hence the difference found between asclepiad pollinator abundance and the abundances of flower visitors in the community is not simply a reflection of a seasonally changing insect fauna, but is a consequence of the filtering effect that specific sets of floral traits have on the attraction of insects.
|
Pollinaria removal and pollinia insertion, and reproductive success of the asclepiads
The removal and insertion of pollinia were scored in a random sample of flowers from all asclepiad taxa. Each of the nine species of asclepiad was found to have had pollinaria removed from flowers, and all except Xysmalobium involucratum had pollinia inserted between the guide rails, indicating successful pollination (Table 4). In almost all cases there is evidence that the pollinators which were carrying pollinaria were also carrying out pollination (see also Pollinaria placement on pollinators, below). Later in the season, fruit-set was apparent on most of the asclepiads that were studied, although this was not quantified. Insertion rates were generally one or two orders of magnitude smaller than removal rates, which is commonly documented in asclepiads (Wyatt, 1978; Ali and Ali, 1989) despite the two-fold greater opportunity for insertion (i.e. the paired pollinia must be removed together [as a single pollinarium], but may be inserted separately).
|
What is the relationship between site of pollinaria placement on an insect’s body and the likelihood of pollination occurring? It may be expected that those asclepiads with ‘messy’ pollination systems (e.g. chafer-pollinated Xysmalobium involucratum, which places pollinaria all over the bodies of the pollinators) would have a lower number of pollinia insertions relative to pollinaria removals, compared with taxa which place pollinaria on specific body areas (e.g. pompilid wasp-pollinated Pachycarpus natalensis). However, examination of the mean ratio of insertions to removals in Table 4 suggests that there is no such relationship. The probability of pollination is more likely to be governed by other factors such as the rate of insect visitation.
Relationship between floral traits and pollination system
The multidimensional scaling analysis of floral traits (see the Appendix) indicates that species possessing the same specialized pollination systems (pompilid wasp and chafer) cluster closely in phenotype space (Fig. 5). Traits shared in common by the chafer-pollinated species include dense, mechanically strong inflorescences, yellow or cream petal and corona coloration, and moderate to concentrated nectar in small to moderate quantities. The generalist-pollinated Xysmalobium gerrardi clusters nearby and shares a number of traits in common with the chafer-pollinated group. This shows that floral traits alone may not always be a good predictor of pollinators and may instead reflect phylogenetic relatedness, in this case of the two Xysmalobium species (although see comments in the Introduction). Taxa pollinated by the pompilid wasp shared the following traits: medium to large flowers in sparse inflorescences, with fewer than 15 flowers open per plant at any one time, a green corolla with purple spotting, quite drab in appearance, and possessing concentrated nectar The two species which are inferred to be bee-pollinated (Aspidonepsis diploglossa and Asclepias cucullata) form a separate, although mutually distant cluster, sharing traits such as medium-sized, open flowers of short life-span, arranged in sparse inflorescences, with fewer than 15 flowers open at any one time, and a discrete nectar cup containing small amounts of moderately concentrated nectar.
|
Pollinator behaviour in relation to flower morphology
The scanning electron microscope (SEM) study of the floral morphology of these nine asclepiads revealed a number of specific adaptations for pollination which can be directly correlated with insect behaviour on the flowers. This is particularly true of the shape, size and function of the corona elements, and is detailed below for each taxon.
Sisyranthus trichostomus was the only species of asclepiad with conspicuously hairy floral parts (Fig. 6A). The dense frill of hairs originates around the rim of the mouth of the short, wide corolla tube and effectively obscures the gynostegium (Fig. 6B). The hairs are multicellular and tapered; along their length are many short, raised ridges, each orientated parallel with the long axis of the hair (Fig. 6C). The innermost ranks of hairs point downwards and into the (presumably nectar-filled) cups formed by the corona (Fig. 6B). The body and legs of chafers feeding at these flowers are excluded from the gynostegium by the short corolla tube and the corolla hairs (Fig. 1D). All pollinaria placement was therefore on the mouthparts of the chafers (Fig. 1E and Table 5) as would be expected from this combination of flower morphology and beetle behaviour. The exact function of these hairs is obscure, but it is hypothesized that they may be multifunctional and: (a) exclude small nectar-robbing insects such as ants; (b) act as a wick to draw up nectar from the nectar cups, encouraging the chafers to probe more deeply with their mouthparts; (c) serve as a visual contrast, highlighting the presence of the opening of the corolla tube (note the darker colouration of the hairs, compared with the corolla, Fig. 1A). An alternative hypothesis is that the hairs may function as food rewards for the beetles, as occurs in some cetoniid-pollinated South African orchids (S. D. Johnson, unpub. res.), although no evidence (i.e. damaged, eaten hairs) was found that would support this.
|
Kunze (1991) investigated the functional morphology of an undetermined Sisyranthus sp. and concluded that the dense hairs blocking the corolla tube served to guide the proboscis of a foraging insect towards the nectary and guide rails. That particular Sisyranthus sp. possesses a much deeper tubular corolla and Kunze’s interpretation is reasonable. It is felt that this is a less likely interpretation for S. trichostomus, however, given its much shallower tube. This illustrates the fact that homologous structures in flowers, even when they appear to serve the same purpose, can in fact be fulfilling entirely different functions, or may even have no function at all if they are present only because of common descent.
Asclepias woodii is also a specialized chafer-pollinated species. The fat, fleshy corona lobes are prominent and clearly have an attractive function (Figs 1A and 9A). They also function as nectar secretion and storage organs; the inside surfaces of these ‘cups’ are strikingly papillate, probably indicating the site of nectar secretion (Fig. 9B). The opening to these cups is unusually narrow for the genus Asclepias and its relatives; e.g. the more usual form taken by Asclepias cucullata shown in Fig. 6D–F. The odd nectar cup morphology becomes understandable when one observes the feeding behaviour of the chafers Atrichelaphinis tigrina and Cyrtothyrea marginalis: their mouthparts are exactly the right proportions to penetrate the partially closed cups and access the nectar. To do so they must move radially around the flower, at a tangent to the circumference, accessing each cup in turn. It follows, therefore, that most pollinaria placement was on the legs of the chafers (Table 5).
|
Xysmalobium involucratum is the final specialized chafer-pollinated species, but the fine structure of flower morphology does not appear to show any particular adaptations to chafer pollination, and indeed a range of other insect species do visit the flowers, and even occasionally pollinate them in the case of Hemipepsis hilaris (Table 2). Nectar is secreted by the inside surface of the corona lobe, where it presses against the gynostegium (Fig. 7A). The long, narrow cleft at the tip of the corona lobe (Fig. 7B) is superficially similar (although smaller) to the opening of the nectar cup in Asclepias woodii (Fig. 9A and B) but we did not observe chafers feeding specifically at this cleft; rather, they appeared to be feeding all over the sides of the gynostegium and corona. The small flowers of this species meant that chafers feeding on one flower were being supported by other flowers and picking up pollinaria from them (Fig. 1C). Consequently, pollinaria were placed all over the chafers’ bodies, although concentrated on their legs (Table 5), as they scrambled over the densely packed inflorescences.
|
The corona elements of the putatively bee-pollinated Asclepias cucullata are shallow and laterally compressed, with slightly arching, raised hoods (Fig. 6D). They too function as nectar cups and the inner epidermis of the cups is presumably the site of nectar secretion, judging by its densely papillate appearance (Fig. 6E and F). The corona of the bee-pollinated Aspidonepsis diploglossa also functions to present nectar to pollinating insects (Fig. 8E). In this case, there is no evidence of morphologically distinct nectar-secreting tissue (Fig. 8F), although it is possible that the nectaries are more deeply situated in the nectar cup and therefore are not immediately apparent from SEM views. The tongue of tissue within the nectar cup is curious and is found in a number of species of Aspidonepsis (Nicholas and Goyder, 1992). If it has any function it may be to effectively raise the nectar level (and therefore its appearance and/or accessibility to pollinators?) without incurring the greater resource cost of actually producing more nectar. Prof. Sigrid Liede (University of Bayreuth) has suggested that the tongue may also increase brilliance effects (pers. comm., April 2003).
|
The specialized pompilid wasp-pollinated taxa each show rather different adaptations to pollination by these insects, even between the closely related Miraglossum spp. The elongated corona lobes of M. verticillare are laterally bent through 90° relative to the main floral axis before turning through a second 90° angle to turn upwards (Fig. 8A). Nectar is secreted at the base of the guide rails, in the space between the broad proximal lobes of the corona (Fig. 8B). A pompilid wasp has its head braced downwards by the lateral coronal ‘bar’ as it approaches the position of nectar secretion. This presumably encourages the wasp to force its mouthparts deeper into the nectar chamber, increasing the likelihood of pollinarium removal or pollinium deposition. This contrasts with the situation in M. pilosum where the elongated corona lobes extend straight upwards (Fig. 8C, note that the re-curving of the corona lobes to form a cage is an artefact of the SEM preparation, cf. Fig. 1A). In M. pilosum nectar is accessed from the ‘back’ of the nectar chamber, rather than the front as in M. verticillare. A wasp approaching the flower braces its front legs against two adjacent corona lobes and accesses the nectar in the chamber diametrically opposite, not the one in the space between its braced legs. This distinction between pollination mode of the two Miraglossum spp. may give a clue as to how pollinia from the two species do not become lodged in the ‘wrong’ species. In M. verticillare, pollinaria will mainly become attached to the top of the mouthparts, whereas in M. pilosum, pollinaria should attach to the underside of the mouthparts. It is in fact rather difficult to observe this distinction in pinned specimens.
Our original idea, that the extended corona lobes of these Miraglossum spp. may be acting as osmophores (specialized scent-producing organs) appears to be unfounded; a close examination of the epidermis of the corona lobes (Fig. 8D) revealed that there were none of the papillae or dense stomata that one might expect to find associated with osmophore tissue (e.g. Vogel, 1990; Curry et al., 1991; Stpiczyñska, 1993).
The third specialized pompilid wasp-pollinated species was Pachycarpus natalensis. In this species the overall flower dimensions are much larger than in the Miraglossum spp. (Fig. 1A and Table 1) and the corona lobes are large, flat structures which over-arch the whole of the gynostegium (Fig. 9C). Nectar is secreted by the corona lobes where they fold together to form a groove (Fig. 9C) and at the point where the corona folds against the gynostegium (Fig. 9D). Copious nectar runs either side of the guide rails and the epidermis morphology of the inner surface of each corona is consistent with its function as a nectary (Fig. 9D and E). Within the guide rails there are very distinct upward-pointing teeth half-hidden behind the outermost flap of the rails (Fig. 9F). The function of these teeth is probably to prevent the pollinator from backing out of the guide rails and the pollinium from being pulled backwards following insertion. Variations on this feature are common amongst asclepiads (e.g. Kunze, 1991; Dr Ulrich Meve, University of Bayreuth, pers. comm. 2001). Pompilid wasps forage around the circumference of the corona lobes (Fig. 1B), feeding on the nectar at the bases of the lobes and picking up pollinaria on their mouthparts (Table 5).
Xysmalobium gerrardii is the most ecologically generalized of the nine species under consideration. It possesses the rather globular corona lobes characteristic of this genus (Fig. 7C and D, cf. X. involucratum, Fig. 7A) as taxonomically circumscribed. Nectar appears to be secreted by the inner surface of the lobes where they press against the gynostegium. Close examination of this area has revealed no evidence of morphologically specialized nectary tissue (Fig. 7E and F).
Pollinaria placement on pollinators
Pollinaria size and shape varied considerably between taxa (Fig. 12). Insects visiting the nine species of asclepiads picked up pollinaria on their mouthparts, each of the three pairs of legs and/or the underside of the body (Table 5 and Fig. 10). However, distinct patterns of pollinaria placement were apparent, which in part related to the pollination system and in part to the phylogenetic identity of the plant. The data in Table 5 have been converted to proportions of pollinaria attached to either mouthparts, legs or body. These data have been plotted onto a triangular graph (Fig. 13). Each data point represents a species of Wahroonga asclepiad and plots the proportion of pollinaria that were attached to one of the three areas of the insect. For example, of the chafer-pollinated guild, Sisyranthus trichostomus placed pollinaria only on the mouthparts of the beetles, while Xysmalobium involucratum and Asclepias woodii placed pollinaria on more or less all of the possible positions (Table 4 and Fig. 13). Conversely, the wasp-pollinated taxa placed pollinaria only on the mouthparts of the insects.
|
|
The advantage of using a triangular graph to represent these data is that pollination systems can be overlain onto the graph to assess whether pollinator identity determines the site of pollinaria placement, or whether other factors such as insect body size and the phylogenetic identity of the plant also have an influence (see Discussion).
Pollinaria number and pollinia size in relation to pollinator size
Pollinator size ought to affect two aspects of the biomechanics of pollinaria dispersal: (1) the probability of pollinaria being removed, because an insect must be large and strong enough to pull the pollinaria out of the flower; and (2) the number of pollinaria carried by the insect, because larger insects have more area over which pollinaria can attach. Using insect length as a measure of insect ‘size’, there is a weak, marginally statistically significant positive relationship between the mean length of the flower visitor and the mean number of pollinaria carried by those insects (Fig. 14; Spearman rank correlation, rs = 0·26, n = 46, P = 0·085). However, this relationship disappears when flower visitors which carried no pollinaria are removed from the analysis (Pearson product moment correlation, r = 0·11, d.f. = 12, P = 0·70). This suggests that there is not a simple linear relationship between insect size and pollinaria load; below a certain size threshold, most insects will not be large enough to remove pollinaria from flowers, but above that threshold there is no relationship between insect size and number of pollinaria carried. With reference to Fig. 14, this threshold appears to be about 5 mm. However, we note that this threshold is not absolute because one very small insect (Coleoptera sp. 3, 2·1 mm in length) is able to remove pollinaria of Xysmalobium gerrardii. Other aspects of insect biology, such as behaviour at the flowers and specificity of pollinaria placement, are likely to be much more important in determining pollinaria loads. For example, chafers visiting inflorescences of Xysmalobium involucratum are large in relation to the size of the flowers and may be simultaneously picking up pollinaria from a number of different flowers (Fig. 1C). There is no relationship between pollinarium size and the size of the insect bearing it (Fig. 15; Spearman rank correlation, rs = 0·24, n = 22, P = 0·27). However, the triangular shape of the distribution of data points in Fig. 15 suggests that the relationship between pollinarium size and insect size is not random: large insects can transport both large and small pollinaria, but small insects are only strong enough to remove the smaller pollinaria.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
Specialization and generalization amongst the Wahroonga asclepiads
The average number of pollinators per asclepiad (mean = 3·1, median = 2·0) is low compared with the averages for the family as a whole (mean = 11·8, median = 3·0) which were published by Ollerton and Liede (1997). As far as can be ascertained from the data presented here, given that they relate to only one site in one season, the Wahroonga asclepiads are on average rather more ecologically specialized (sensu Waser et al., 1996) than most other species of asclepiad. This may be evidence to support the assertion of Johnson and Steiner (2000) that areas of high plant diversity such as South Africa accommodate many more highly specialized plant–pollinator interactions compared with areas of lower diversity. This is certainly true on a case-by-case basis (there are a large number of specialized interactions in the South African flora), but whether the plants in South Africa are, on average, more ecologically specialized is unknown [though see Ollerton and Cranmer (2002) in contrast to Olesen and Jordano (2002)].
Three very distinct pollination systems can be recognized within this assemblage of nine asclepiad species. (1) Specialized pompilid wasp pollination: Miraglossum pilosum, Miraglossum verticillare, Pachycarpus natalensis; (2) specialized fruit chafer pollination: Asclepias woodii, Sisyranthus trichostomus, Xysmalobium involucratum; and (3) generalized insect pollination: Xysmalobium gerrardii. In addition, it is suspected that Asclepias cucullata and Aspidonepsis diploglossa may be mainly bee pollinated, but this requires further observation.
Pollination by pompilid wasps has been documented for a number of North American Asclepias species (e.g. Robertson, 1928; Kephart, 1979) and South American Oxypetalum species (Vieira and Shepherd, 1999) but only as a component of (sometimes extremely) generalized pollination systems, never as a specialized interaction. A Pepsis sp. was recorded as a pollinator of Morrenia odorata and Philibertia gilliesii in Argentina (Sigrid Liede, University of Bayreuth, unpub. res., see the ASCLEPOL online database of plant–pollinator interactions in the asclepiads at: http://www.uni-bayreuth.de/departments/planta2/research_wgl/pollina/as_pol_t.html) but it is not known if these asclepiads possess specialized pompilid wasp pollination systems. In South Africa, flowers of Pachycarpus asperifolius are pollinated by a large number of wasps, including Sphex sp. (Sphecidae) and Hemipepsis dedjas (Pompilidae), but this appears to be an example of a ‘general’ wasp-pollinated system (Craig Peter, University of Natal, unpub. res.). Specialized vespid wasp (Hymenoptera: Vespidae) pollination is known in South American Oxypetalum (Vieira and Shepherd, 1999; J. Ollerton, unpubl. res.) and Blepharodon nitidum (J. Ollerton, unpubl. res.), as well as in South African Gomphocarpus (S. D. Johnson, unpub. res.). These results have therefore confirmed that pompilid wasp pollination joins vespid wasp pollination as another specialized pollination system in the asclepiads. Floral traits shared in common by these specialized wasp-pollinated taxa include green flowers, often with purple colouration or spotting, and sometimes a spicy-sour odour.
Specialized pompilid wasp pollination appears to be virtually unknown in the rest of the flowering plants. The only other known case is pollination of a South African orchid, Disa bivalvata, by precisely the same pompilid wasp, Hemipepsis hilaris (Steiner et al., 1994). The Disa system is based on the sexual deception of male wasps by the orchid, and so differs to that of the asclepiads under consideration in this work, which is probably a more straightforward attraction–reward situation, as wasps of both sexes were attracted and all of the asclepiad taxa produce nectar. However, the flower coloration of D. bivalvata is rather similar to that of Pachycarpus natalensis (Fig. 1), with dark purple patterning on a light green background, suggesting that both species are employing similar visual attraction cues. The same pompilid wasp has also been observed visiting flowers of the related Pachycarpus appendiculatus (another species with green and purple flowers) elsewhere in South Africa (Dr Peter Bruyns, University of Cape Town, pers. comm.). Adult pompilid wasps are commonly known to visit flowers to feed on nectar, so it should not be surprising if other examples of specialized pompilid wasp pollination are discovered in the future.
Specialized beetle pollination is rare in the asclepiads (Ollerton and Liede, 1997) and, once again, most records are of beetles as occasional pollinators of highly generalized North American Asclepias species (e.g. Robertson, 1928; Willson and Price, 1977; Queller, 1985). Forster (1989) documented what may be specialized beetle pollination in the asclepiad Marsdenia fraseri, involving the lycid Metriorrhynchus lateralis (Coleoptera: Lycidae) in Australia, though this was based on a single day of observations and therefore it is unknown whether other insects may be involved in pollination of this species. The only other record of cetoniid chafers visiting asclepiad flowers are those of Woodell (1979), who recorded Mausoleopsis aldabrensis as a common visitor to three asclepiads on Aldabra but did not record their status as pollinators. Interestingly, these chafers were also important flower visitors to other plants on the island, which has intriguing parallels with our findings in KwaZulu-Natal (see below).
Ollerton and Liede (1997) predicted that specialized beetle pollination systems may prove to be more common as research progresses, and that certainly has been shown to be the situation in KwaZulu-Natal. The three species of chafer-pollinated asclepiads share a range of floral traits in common, despite their diverse phylogenetic histories, suggesting that convergent evolution towards a specialized chafer pollination system has occurred.
The two species of chafers found at Wahroonga were common flower visitors to a wide range of other plants in grasslands around the Natal Midlands area during our fieldwork period (J. Ollerton et al., pers. obs.; S. D. Johnson and P. Neal, unpubl. res.; C. Peter, pers. comm.). Bees are relatively uncommon in these grasslands and the chafers appear to be fulfilling the role of the large bees (e.g. Bombus spp.) which would commonly be found in, for instance, European grasslands. These chafers are agile, fast flying, hairy and appear to mainly feed on nectar. Importantly, they do not damage flowers when they are feeding, which other beetles often do (Proctor et al., 1996). At Wahroonga and other sites, the chafer-pollinated asclepiads appear to form a guild with plants from other families, such as Helichrysum spp. and Berkheya setifera (Asteraceae) and Zantedeschia bimaculata (Araceae), although this needs confirmation (J. Ollerton et al., unpubl. res.). Holm and Marais (1992) note that many cetoniids can be collected from flowers, as well as the fruit which gives them their vernacular name of ‘fruit chafers’. Whitehead et al. (1987) point out the importance of cetoniid beetles (and other Scarabaeidae) as pollinators of plants in the Cape flora and the same may well be true in other parts of South Africa. In addition, Liede (1989) and Struck (1995) have shown that pollination by Scarabaeidae is important in some South African genera of Mesembryanthemaceae. Chafer pollination may therefore be much more common than previously realized and, on the basis of the observations made during summer 2000–2001, fruit chafers may prove to be one of the dominant pollinators in KwaZulu-Natal grasslands, functionally replacing large bees. A similar argument has been made by Steiner (1998) regarding the importance of monkey beetles (Scarabaeidae: Hopliini) as dominant pollinators in the south-western Cape region of South Africa.
Heteropteran bugs frequently visit flowers with easily accessible nectar or pollen, but are usually not considered to be efficient pollinators (Proctor et al., 1996). There are no known examples of specialized heteropteran- pollinated plants (reviewed by Ollerton, 1999). The most abundant pollinator of the generalist Xysmalobium gerrardii was the lygaeid bug Aspilocoryphus fasciativentris, which carried abundant pollinaria and was a very active insect. Heteroptera are known to visit flowers and carry pollinaria of North American Asclepias species, and are probably minor pollinators within these generalist pollination systems (Robertson, 1928; Willson and Bertin, 1979; Fishbein and Venable, 1996). More work is required to understand the importance of this lygaeid in the pollination ecology of X. gerrardi (e.g. quantifying bug movements between plants) but it is suspected that it may prove to be a significant, and possibly the major, pollinator of this species.
In a theoretical analysis of classical pollination syndromes by Ollerton and Watts (2000) and J. Ollerton and R. Alarcon (unpub. res.), pollination by wasps and beetles falls relatively close to one another in phenotype space. This suggests that transitions between these pollination systems during speciation events are more likely than, for instance, transitions between wasp and moth pollination or beetle and bee pollination. Moderately large genera of southern African asclepiads such as Pachycarpus (approx. 30 species) and Xysmalobium (approx. 40 species) would provide an excellent test of these theoretical expectations, by combining phylogenetic analyses with field surveys of pollination systems. It is hoped this will be made a focus of future work on the pollination biology of southern African asclepiads.
Pollinator and flower visitor diversity
Species accumulation curves are rarely presented in studies of flower–visitor interactions (Ollerton and Cranmer, 2002; though see Petanidou and Ellis, 1993; Minckley et al., 1999) yet they are a useful tool for determining how comprehensive the sampling of a pollinator or flower-visitor fauna has been. In this study, species accumulation curves were used to gauge how accurate was the assessment of flower-visitor diversity for these species. There was considerable variation between asclepiad species: for the specialists, with only one or two pollinators, the full diversity of flower visitors had been sampled within 2 d. For the more generalist species, new species were still being added after 20 (non-consecutive) field days. For the combined assemblage, an asymptote was reached after 23 d. There are few published data with which to compare our results but, clearly, highly generalist taxa such as many Asteraceae and Apiaceae (or even some North American Asclepias spp.) are going to show a rather different pattern, with the cumulative curve possibly never reaching an asymptote within practical sampling time. Similarly, studies that extend over multiple seasons may reach a temporary asymptote within each season, only to accumulate further species in the subsequent season, as Petanidou and Ellis (1993) found. It would be very useful if data were presented in this way more often by pollination ecologists.
Floral traits and pollinator identity
Multidimensional analysis of floral traits showed that the functionally specialized taxa largely grouped together, but that there may also be some phylogenetic conservatism with regard to floral traits which may result in a species grouping with the ‘wrong’ taxa, i.e. the generalist Xysmalobium gerrardii clustering with the chafer-pollinated species, and specifically it’s congener, X. involucratum, on the basis of floral characteristics that may be phylogenetically conservative. Clearly, multivariate floral trait analysis can be a useful tool for exploring the relationship between flower phenotype and pollination system, and is starting to become widely used (Ollerton and Dafni, 2003) but the results should be treated cautiously, particularly in the absence of information regarding the identity of a plant’s pollinator(s) or an explicit phylogeny. Recent examples confirm this: Borba et al. (2002) analysed floral traits in populations of five species of fly-pollinated Pleurothallis (Orchidaceae). Principal component analysis separated out the populations into three groups, depending upon the exact identity of the pollinating flies. However, the floral phenotypes associated with two of the pollination systems (involving the fly genera Tricimba and Megaselia) were in close proximity in phenotype space, suggesting that it is not always possible to predict the precise pollinator purely from the observed floral traits. Similarly, research on the relationship between pollinator identity and floral traits in Bornean gingers by Sakai et al. (1999) resolved 44 species of Zingiberaceae and Costaceae into three pollination guilds. The three guilds were moderately discrete in their relative positions in phenotype space, but some species did not cluster within the main groupings and defining the groupings themselves was a little arbitrary (see their Fig. 2). Of the 29 species for which pollinators were known, two were positioned in the ‘wrong’ area for their pollinator. Finally, in a multivariate analysis of ten species of Labiatae in Greece, there was no correspon