Reproductive biology of Davilla kunthii A . St-Hil . ( Dilleniaceae ) in Central Amazonia

(Reproductive biology of Davilla kunthii A. St-Hil. (Dilleniaceae) in Central Amazonia). Th is survey aimed at describing the interactions of fl oral visitors and Davilla kunthii A. St.-Hil. as well as characteristics of its reproductive biology in Itacoatiara, state of Amazonas, Brazil. Tests of the breeding system were performed. Th e guild of visitors was described according to richness, abundance, relative frequency and constancy. Th e breeding system tests indicated that D. kunthii is self-compatible. Th e pollination system was characterized as generalist, with 39 visitor species, from three diff erent orders. Bees were the main group of pollinators, thus some behavioural aspects were described. Th e period of highest foraging activity was between 7 and 10 am. Some species presented agonistic and monopolistic behaviour. Given the behaviour and destructive potential, the Curculionidae seem to have a greater impact as seed predators than pollinators.


Introduction
Considering its geographic area and biological diversity, it is acceptable that the Amazon rainforest harbours many biological and ecological mechanisms still unknown to science.Plant-pollinator interactions are among the processes that characterise the complex dynamics of this ecosystem.More than 90% of all fl owering plants of the tropical lowland rainforest have biotic vector-mediated pollination (Bawa 1990).Among the biotic agents, insects predominate over other animals, as they are abundant and have vision capable of diff erentiating chromatic patterns and forms and have a large energetic demand due to their high metabolic activity (Ollerton 1999).
Among the insects, bees form the principal group of pollinators due to their diversity and potential generalism (Michener 2000).Th ey permanently demand food harvest in great quantity.Moreover, the potential polytrophism of bees is caused by nutritional dependence upon pollen and nectar throughout life (Corbet et al. 1991;Free 1993).Many bees also have a good system of communication and are capable of specialising on food resources for some periods of time, thereby granting a certain degree of fi delity to them (Ollerton 1999;Ramalho 2004).
Interaction networks of diff erent pollinators and plant species with shared pollen vectors may cause competition either for the food resource, such as that among fl oral visitors that share the same trophic niche (Ramalho 2004), or, among plants, for pollination agents (Bell 2005;Caruso 2000).Concerning competition for food sources, there may be behavioural and physiological alterations among species: diff erences in foraging time, varying distribution through resource patches and even agonistic interactions that may lead to food source monopolisation (Roubik 1989;Herrera 1995;1997;Ramalho et al. 2004).
On the other hand, there is competition for pollination agents when two or more sympatric plant species have a concomitant fl owering period and similar fl oral characteristics (Caruso 1999;2000).In this case, that competition may occur due to the preferences of the pollinator or inappropriate pollen transfer (Bell et al. 2005).Th e fi rst case occurs when a more attractive plant draws visitors away from other plants, whose reproductive successive is consequently negatively aff ected (Waser 1983;Sih & Baltus 1987;Brown et al. 2002).Th e inappropriate transfer of pollen occurs when heterospecifi c pollen is deposited on the stigma of one or more competitors, similarly decreasing seed set (Rathcke 1983;Brown & Mitchell 2001).
Th ere may also be inter-population variations in the visitor-guilds' structures as well as the manner in which fl oral rewards are exploited.Th e causes that govern interpopulation variation of pollinators for a certain plant species, which consequently promote generalised pollination systems, are diverse.Th ese causes may be nested according to the ecophysiological requirements of the pollinator (Ramalho et a.l 1994;1998;2004;Herrera 1995;1997;Hilário et al. 2000;2001;Kajobe et al. 2007), plant population size (Eckhart 1991;Conner & Neumeier 1995;Roll et al. 1997), and competition with more sympatric species (Caruso 2000).
Contrary to the discussion of generalisation-specialisation of pollination mechanisms, Armbruster et al. (2002) and Fenster et al. (2004) point out that a plant with an assembly of phylogenetically unrelated fl oral visitors does not necessarily indicate a generalised system.According to these authors, many visitors can be included in functional groups, which permit the understanding of specialisation as adaptive evolution.Other ways to understand the same phenomenon refer to time of anthesis (Armbruster 1985;Stone et al. 1998), location of pollen deposition (Nilsson 1987;Armbruster et al. 1994), or even homoplasies of fl oral traits (Schemske 1981;Nilsson 1983;Temeles & Kress 2003).
Davilla Vand. is one of the most diversifi ed genera within the Dilleniaceae Salisb.and comprises about 30 neotropical species of lianas and erect or climbing shrubs.Its distribution ranges from eastern Mexico to Bolivia, Paraguay and southern Brazil.Th is genus is represented by commonly occurring species and is indicative of forest borders and environments of high insolation, typical of fragmented landscapes, ecotones, or clearings (Armelin & Mantovani 2001;Nascimento & Laurence 2006).Th e reproductive biology of the genus Davilla is scarcely known.Regarding fl oral visitors, among reports concerning the genus Davilla, Ducke (1902) reported Halictus Latreille 1804 visiting fl owers of D. rugosa Poir.For the same species, Kuhlmann & Kühn (1947) indicated bees and other insects as pollinators.Croat (1978) verifi ed fl owers of D. nitida (Vahl) Kubitzki being visited by Partamona cupira (Smith 1863).Th is survey aimed to describe the reproductive biology of Davilla kunthii A. St.-Hil, in the municipality of Itacoatiara (Amazonas).

Materials and methods
Th is survey was carried out in the municipality of Itacoatiara, in Central Amazonia (58° 45' 41''W and 03° 02' 44'' S), on the borders of native forest patches ("terra fi rme" forest) that form a mosaic inside an area of Brazil nut cultivation (Bertolletia excelsa Humb.& Bonpl.).
Morphology and fl oral biology such as fl owering rhythm (number of fl owers opening per day) and time of odour liberation were assessed in situ in 10 plants.Th e fl owering rhythm was characterised by marking 10 infl orescences (two per plant) and conducting continuous observations from anthesis of the fi rst fl ower until complete senescence of all fl owers of each infl orescence.We monitored 30 fl owers from fi ve individuals for pollen availability (anther dehiscence) as well as for stigmatic receptivity, which was assessed by means of the activity of peroxidase, using hydrogen peroxide (10 volumes) dropped directly on the stigma in order to observe bubble formation (Kearns & Inouye, 1993).Tests of the breeding systems were performed with fl oral buds prior to anthesis in at least 10 diff erent individuals and always on the base of the infl orescence.Th e buds were bagged in order to avoid any contact with possible pollinators.Five treatments were performed: spontaneous self-pollination (n: 1.633), manual self-pollination (n: 337), geitonogamy (n: 296), xenogamy (n: 298) and agamospermy (n: 77).For similar fructifi cation frequencies we use a statistical test χ 2 (p:0,05).Not all treatments co-occurred in the same individual.With the exception of the self-pollination treatment, all fl owers were emasculated between 6:00 and 6:30 a.m.before anther dehiscence.In total, 524 developing fruits were marked (seven infl orescences, one per plant) to evaluate natural pollination (pollen-vector-driven fruit-set) and another 950 (twelve infl orescences, one per plant) were marked and collected aft er 5 months to verify the presence of Curculionidae larvae.
Field work consisted of direct observation of four sample groups (A, B, C and D) of D. kunthii.Each group was observed during three consecutive days in August 2007.Although it was diffi cult to determine the number of individuals in each sample group (vine), all were of similar area (about 100 m 2 ) and showed similar fl oral displays.Hence, our total observation time consisted of 12 days (three days per plant group) and 84 hours.Th e sample groups were more than 4 km away from each other so as to assure no overlapping of bee foraging area (Araujo et al. 2004).
Observations started at 5:00 a.m. and lasted until the sepals closed and the bees' foraging period ended, which was around 12:00.Observations were conducted during one hour intervals.During the fi rst 20 minutes of each hour, fl oral visitors were quantifi ed.In the remaining 40 minutes, we analysed details of visitor fl oral behaviour.Th e structure of the bee guild was characterised by richness, abundance and relative frequency of species.To characterise constancy, we utilised the classifi cation of Bödenheimer (1955): species with a relative frequency ≥ 50% were considered constant species (W); accessory species (Y) had a relative frequency between 25 and 50%; and rare species (Z) had a relative frequency lower than 25%.
To verify the existence of diff erences between foraging intensities or foraging times, the abundance of visitors in each hour was considered a dependent variable.Th e experimental design was organized by way of sub-divided parcels, with the sample groups (A, B, C, D) being considered parcels and the visitation times (hours) being the sub-parcels in a variance analyses.Variance analyses and a regression were calculated using SISVAR soft ware.Th e species accumulation curve (collector's curve) was based upon the relation of new species per collected individual and it was calculated using ECOSIM soft ware (Gotelli & Entsminger 2009).A fi gure describing the interactions between the four plant sample groups and bees was prepared using PAJEK soft ware.
Th e voucher of D. kunthii was deposited in the herbarium of the Instituto Nacional de Pesquisas da Amazônia (INPA) under number 221995, and the bees were deposited in the Hymenoptera collection of the Universidade de São Paulo (USP), campus of Ribeirão Preto ("Coleção Camargo") and the authors' names follow Camargo & Pedro (2008).

Results
Th e infl orescence of Davilla kunthii (Fig. 1) is terminal or axillary, paniculate (4-8 cm long) with bracts 2-4 mm wide.Th e fl owers opened between 6:00 and 6:30 a.m.; pollen liberation occurred 30 minutes aft er anthesis, and the sweet fl oral odour was most intense between 7:00 and 9:00 a.m.Th e fl ower is nectarless, plate-like, and actinomorphous.Th e corolla is yellow and dialypetalous, with four to six elliptical to obovate petals that senesce three to four hours aft er the beginning of anthesis.Th e calyx (5-sepals) is asymmetric, three outer sepals smaller than inner acrescent ones, and around 11:00 a.m. the larger sepals close and remain covering the entire developing fruit until maturation.Th e androecium is dialystaminous with 50-70 stamens, located in concentric verticils.Th e anthers (0.25 mm long) are basifi xed and rimose.Th e ovary is hypogynous, unilocular with two ovules, and the stigma is peltate.Stigmatic receptivity progressively increases until around 9:00 a.m.Th e fruit is globose with a membranous pericarp.
Th e fl owering pattern was classifi ed as mass annual and asynchronic, with fl owers opening non-concomitantly between adjacent sample groups.Th e fi rst bee visitors appeared aft er the partial opening of fl owers (5:30 a.m.).All buds from each infl orescence opened in a interval of six or seven days, and the number of open fl owers per day (n=10 infl orescences) indicated that two fl owering peaks occurred on the fourth and sixth days, with an average of 161.1 and 50.7 fl owers, respectively (Fig. 2).Th e breeding system tests indicate that Davilla kunthii is self-compatible, and data from cross and pollinator-driven natural pollination were very similar (χ2: 0.5843, p: 0.4446; Tab. 1).
We collected 771 bee visitors harvesting pollen, divided into 35 species (Tab.2).Th e species accumulation curve did not reach total saturation (Fig. 3), although it is believed that species important for the pollination of D. kunthii have been captured.Bee species found in all four sample groups are presented in Table 1.A Diptera (Syrphidae) was also observed feeding on pollen, which, due to its harvesting behaviour, may be considered a pollinator.Moreover, two species of Vespidae were observed consuming pollen, and one species of Curculionidae was observed walking on the fl owers and laying eggs on them.Th e body surface of the Curculionidae was covered with pollen grains.However they hardly migrated between plants.Th is probably prevents them from being eff ective pollinators, yet they may serve as occasional pollinators.Categorisation of constancy revealed the absence of constant species (W).Accessory species (Y) were: Tetragona essequiboensis (sample group C), Camargoia camargoi (sample group B), Trigona group fuscipennis (sample group D) and Augochloropsis sp.3 and Plebeia sp.(sample group A).Th e other species were classifi ed as rare (Z).
Many bee species occurred in only one sampling site, as can be seen in Figure 4, which shows shared and exclusive bee visitors.Variance analysis showed a signifi cant eff ect (p<0.01) of foraging intensity in relation to time (hours), but isolated eff ects in site (sample group) or the interaction of both factors were not seen (p>0.05 for both).Th e time of highest foraging activity was in the interval between 7:00 and 10:00 a.m.(Tab.3 and Fig. 5).
Comparing the abundance and distribution of pollinator species within individual sample groups, it is noted that even though Te. essequiboensis and Tr.gr.fuscipennis were the most abundant, they did not appear in all sample groups.Although Trigona branneri, Ca. camargoi, Cephalotrigona femorata, Melipona fulva and Melipona amazonica were of lower abundance, they were better distributed within the sample groups, as seen in the interaction web (Fig. 4).We noted dominance behaviour and intense harvesting by T. branneri and T. essequiboensis, with the former behaving agonistically against Apis mellifera.Th is behaviour was also seen in Augochloropsis sp.3, which attacked Frieseomelitta trichocerata, in this case, the latter bee fl ew away.Similarly, Ca. camargoi presented aggressive behaviour towards M.    amazonica and also against smaller bees, chasing them away while it harvested part of the reward (individuals group of D. kunthii) in a monopolistic way.
Early in the morning, larger bees (such as A. mellifera) harvested more intensely in the canopy extensions of D. kunthii, where it was not possible to collect them.Th ese bees began to collect in lower extensions of the vines between 9:00 and 10:00 a.m., which is probably related to the decrease in the more aggressive species that are concentrated in the early morning hours.Smaller species such as Schwarzula coccidophila, Trigonisca vitrifrons, Plebeia sp. and Augochloropsis sp.3 did not contact the stigma when walking or collecting pollen and instead walked only on the petals.
We also verifi ed the predation of fl oral visitors by ants that constantly scouted the entire infl orescence in plant individuals group A and B. When bees noted the presence of ants (fi ve to eight on each infl orescence), they tended to avoid them or fl y away.One Curculionidae species, which is also predated by ants, fl ew in front of a senescent fl ower and oviposited between the stamens or near the ovary, which are subsequently covered by closing acrescent sepals.When Curculionidae larvae hatch inside the chamber of acrescent sepals, they feed on the seed and developing fruit.Of the 937 sepal chambers opened and counted, 311 had Curculionidae larvae (34%) and 626 did not (66%).

Discussion
Th e mass-fl owering pattern found in D. kunthii, which was asyncronic among sample groups and peaking during  2.  the fl owering period, corroborates the observations for D. elliptica A. St-Hil.(Rocha & Lomônaco 2006).Also in common with D. elliptica is the 6:20 h opening time (Gottsberger 1977).However, in the same study, which was conducted in Cerrado vegetation, D. elliptica was fi rst visited by A. mellifera and later by Meliponini.Considering the high harvesting activity frequency of A. mellifera, we believe this behavior may interfere with the later harvest activity of Meliponini .In our study, A. mellifera visited infrequently, and the Meliponini began to harvest even before the petals opened entirely.We believe that intercalated fl owering peaks among individuals (Fig. 2) is a strategy which forces bees to forage over proximally located individuals during the fl owering period.
A simple fl ower of low restrictive morphology (plate-like), like that of Davilla kunthii, permits that diff erent visitor species deposit a suffi cient quantity of pollen grains per visit, independent of their morphological characteristics (Gómez 2002).Th e facility of accessing reproductive parts is even more capable of promoting fertilization in plants with few ovules (Johnson et al. 1995), like D. kunthii.Nevertheless, taking into consideration that Davilla is preferentially xenogamic, it is necessary that the visitor transports pollen among individuals.Since natural fructifi cation (pollinator driven) was very similar to that of xenogamy, it seems visitors are acting as pollen vectors, which leads us to conclude that there is no pollen limitation in D. kunthii.Th e short period between fl ower anthesis and petal dehiscence is already known in Davilla spp. in the Cerrado (Gottsberger 1977).Moreover, petals fall three hours aft er opening, remaining open only during the highest foraging period (between 7:00 and 10:00 a.m).Th is leads us to think that petals perhaps serve as pollen source indicators to visitors, also evidenced by the visible decrease of pollen collected following senescence.
In the analysis of the collector's curve (Fig. 3), it should be noted that the sampling eff ort was not enough to reach the entire saturation of the curve.However, we feel that it is suffi cient to support the fact that D. kunthi is generalist in a taxonomic point of view.Furthermore, we support Herrera's (2005) view that one should be focused on the visit event itself, as less frequent species have a lower importance for pollination.Hence, a much higher sampling eff ort would be necessary to saturate the curve, and this eff ort may not reveal more species that are important for the plant's reproductive success.Th is previous statement is based upon D. kunthii having been visited not only by 35 bee species but also by a Syrphidae, two Vespidae and a Curculionidae species.
However, from a diff erent perspective, it is noted that all visitors, except beetles, present similar functionality, harvesting behaviour and are motivated to visit the plant by the collection of pollen, which gathers all of them into the same functional group.According to Fenster et al. (2004) and Armbruster et al. (2002), these functional groups of visitors may collectively drive the evolution of plants, so a similar process could have occurred in the case of D. kunthii.In our study, pollinators have no morphological characters substantially diff erent to mediate any modifi cation in the fl oral traits.Th erefore the plant has no possibility of establishing a hierarchy of pollinators according to the diff erential importance of them (Schemske & Horvitz 1984;Stanton et al. 1991;Galen 1999;Zamora 2000).In such a process fl owers would be adapted to be pollinated by the "most effective group of pollinators" rather than the "most eff ective pollinator" species (Stebbins 1970).
Th e absence of constant fl oral visitor species in this study (according to classifi cation of Bödenheimer 1955) may be related to the great availability of sympatric polliniferous species in the study area.An example of the latter is B. chrysophyla ("murici" with oil and pollen as reward), on whose fl owers we observed bees not found visiting D. kunthii, but many visitors of Davilla were seen foraging on B. chrysophyla.Furthermore, several authors point out that inter-plant competition may diminish the quantity and quality of pollen deposited on co-specifi c stigmas, as constancy of visitors is aff ected (Harder & Barret 1996;Caruso 1999;Brown et al. 2002).Working on mono-specifi c and mixed environments, Bell et al. (2005) found 42% plant-plant and 10% inter-specifi cally fl ower-fl ower movement in mixed environments, which greatly decreased the pollination effi ciency at this site.
Th e web diagram (Fig. 4) illustrates the interactions between sample groups and visiting bees, clearly showing the shared and the exclusive bee species in the four sample groups of D. kunthii.Th e existence of unique species (visitors of only one sample group) is in agreement with the perspective that oscillations in the composition of the visiting guild depends more upon the characteristics of each species of visitor (Ramalho et a.l 1994(Ramalho et a.l , 1998;;2004;Herrera 1995;1997) and especially upon agonistic interactions and the proximity of the fl oral source to the nesting sites of the bees (Hilário et al~. 2000;2001;Kajobe et al 2007).From our observations, we conclude that distances over 4 km between sample groups of D. kunthii explain the presence of unique species in each sample site, considering that stingless bees normally fl y no more than 2 km (Araújo et al. 2004).Th erefore the common species found in more than one sampling site probably came from diff erent nests.Some surveys regard beetles (Scarabeidae, Chrysomelidae and Curculionidae) as the main pollinators of Hibbertia hypericoides (DC.)Benth., H. scandens (Willd.)Gilg, and other species of Dilleniaceae, whereas bees and fl ies had secondary importance (Keighery 1975).Later, Bernhardt (1984) stated that fl owers of H. stricta are visited by beetles and bees.Despite this reference to beetles, Armstrong (1979), in his review, added eight genera of Australian bees (Apidae) collected on the fl owers of Hibbertia spp.and characterised these plants as bee-attractive.Recently, a review of studies on Hibbertia concluded that species within this genus are pollinated primarily by Apidae and pollen-collecting Syrphidae, claiming that bees are more effi cient than beetles for pollination (Tucker & Bernhardt 2000).Although D. kunthii was visited occasionally pollinated by beetles, we agree with Gottsberger (1977), who pointed out that there is a destruction of 20 to 30% of fruits in D. elliptica and up to 100% in D. rugosa.Th us, we believe that beetles have a greater impact as seed predators than as pollinators, since they are found in a large proportion of fruits (34%).Hence, aft er adding our data to those of Gottsberger (1977) and Tucker and Bernhardt (2000), we come to the conclusion that bees may be the main group of pollinators of Dilleniaceae genera, although studies with other genera and more observation would be required to confi rm this assertion.

Figure 1 .
Figure 1.Illustration of Davilla kunthii A. St.-Hil., in detail the fl oral bud and the open fl ower.

Figure 3 .
Figure 3. Collector's curve; number of individuals trapped on the X-axis and cumulative species richness on the Y-axis.Points above and below the curve indicate the confi dence interval (95% probability).

Figure 4 .
Figure 4. Interaction web representing relations between bee species and sample groups of Davilla kunthii, bee numbers refer to Table2.

Table 1 .
Result of the breeding system tests indicating that Davilla kunthii is possibly self-incompatible.Natural pollination refers to pollination made by fl oral visitors.

Table 2 .
Total abundance and frequency at the four plant sample group of insects captured on fl owers of Davilla kunthii in Itacoatiara-AM in August 2007.

Table 3 .
Results of variance analysis for foraging intensity.We tested the local infl uence, time (h) and the interaction local*time.