Floral biology and a pollinator eff ectiveness test of the diurnal fl oral visitors of Tabernaemontana undulata Vahl . ( Apocynaceae ) in the understory of Amazon Rainforest , Brazil

(Floral biology and a pollinator eff ectiveness test of the diurnal fl oral visitors of Tabernaemontana undulata Vahl. (Apocynaceae) in the understory of Amazon Rainforest, Brazil). In this paper we examined the fl oral biology, per-visit eff ectiveness, frequency of visits and relative abundance of the diurnal fl oral visitor taxa of T. undulata (Apocynaceae) at two populations located in the primary forest and in a disturbed area connected to the continuous forest. Its hermaphrodite fl owers show a long and narrow fl oral tube and introrse anthers form a cone around the apex of the style head where the pollen is deposited. Th e stigmatic head has three functional regions, in a complex pollination mechanism favoring cross-pollination. In the pollinator censuses conducted in the primary forest population we registered Eulaema bombiformis. In the disturbed area Euglossa sp. was responsible for visits on T. undulata fl owers. Eulaema bombiformis was not absent in the disturbed area, but preferred to forage on pollen and nectar-rich fl owers of two neighboring fl owering species. In the primary forest, T. undulata was the only fl owering treelet at the end of the dry season. Signifi cant diff erences were not observed regarding the frequency of visits and relative abundance among the diurnal fl ower visitors at both sites. In the per-visit eff ectiveness experiment among fl ower visitors, we observed one fruit produced aft er E. bombiformis’ visits. Due to the complex breeding system of T. undulata in that it is characterized by very low fruit production even under natural conditions, an indirect measure of pollinator eff ectiveness is needed in at a future stage of this study in order to clarify the role of these bee species as pollinators of T. undulata.


Introduction
Th ere is a traditional notion that the evolution of fl oral diversity is based upon specialized interactions of plant species and pollinators, which led to the classifi cation of fl owering plants into pollination syndromes (Faegri & Pijl 1979).However, many studies have subsequently shown that even plants which exhibit characteristics of one pollination syndrome may be visited by other organisms which may contribute in both positive and negative ways to pollination dynamics and ultimately fi tness at some level (Spears 1983;Waser et al. 1996).Ollerton (1996) pointed out an apparent paradox between the ecological processes and phylogenetic constraints as many species produce specialized fl owers which receive a diverse range of fl ower visitors suggesting that among all the fl ower visitors of a plant species only a small proportion can eff ectively pollinate it.Although diff erences in relation to pollination ability among fl ower visitors are probably quite common in nature, they have been poorly studied (Ollerton 1996;Johnson & Steiner 2000;Ollerton et al. 2007).
Simple lists of fl ower visitors received by a plant provide little useful information because there is no distinction between visitors and eff ective pollinators (Johnson & Steiner 2000).Identifying the pollen vectors among the fl ower visitors represents an important requirement for any pollination system investigation (Waser et al. 1996;Johnson & Steiner 2000;Sahli & Conner 2006).
Th e main tropical Apocynaceae s.s. are characterized by a complex pollination mechanism favoring cross pollination.Th is may have evolved in order to attract insects with long, strong mouth parts able to access the nectar in the tubular fl owers (Darrault & Schlindwein 2005).Besides the complex fl oral morphology, the existence of secondary pollen presentation co-occurring with herkogamy (Lopes & Machado 1999) and the recent report of functional dioecy in members of the family (Koch et al. 2002), also indicate a specialized pollination mechanism in Apocynaceae, preventing self pollination.Th e genus Tabernaemontana L. belongs to the subfamily Rauvolfi oideae in the Apocynaceae s.s.(Simões et al. 2006) where information regarding the breeding system and pollination biology are scarce.Few registers on pollination biology for representatives of the genus have been made in Brazil, with the exception of Gottsberger & Gottsberger (2006) for Tabernaemontana hystrix Steud.and T. solanifolia DC., both with nocturnal fl owers and classifi ed as moth-pollinated by the authors.
Th e Amazon Rainforest, the largest tropical forest in the world, harbors a striking richness of animals and plants species (Whitmore 1998).Plant-animal systems, as all other kinds of species interactions, are also a critical aspect of this biodiversity (Ollerton & Cranmer 2002;Ollerton et al. 2006).Most plant species in this ecosystem, especially those from the understory, are dependent on biotic pollination (Janzen 1975).Nevertheless, basic data on pollination biology are limited, probably as a consequence of the various logistical diffi culties imposed by the complex structure of tropical rainforests (Bawa 1990;Kay & Schemske 2003;Ollerton et al. 2006).
To our knowledge there are no studies on Tabernaemontana undulata Vahl., an understory treelet found in the Central Amazonia rainforest.Despite the fact that its fl owers show morphological features compatible with the moth-pollination syndrome, they are frequently visited by Euglossine bees at the same time as they emit a strong sweet fl oral scent and off er nectar in the morning.In this study we examine some aspects of the fl oral biology and breeding system of T. undulata, and the per-visit eff ectiveness, frequency of visits and relative abundance of its diurnal fl oral visitors (Euglossine bees) at two populations in Central Amazonia.Our main question is: Do the Euglossine bees contribute to the pollination of the moth-fl owers of T. undulata?

Study area
Th e study was conducted at Fazenda Experimental da Universidade Federal do Amazonas (2º 38' 57,6"S; 60º 3' 11"W, 96 m altitude), located at km 38 of the federal road BR-174, in Manaus, Amazonas, northern Brazil.Th e area has various vegetation types, from which primary rainforest represents the largest portion of the total area.Th e study site is characterized by a hot, wet climate with distinct rainy and dry seasons.Th e rainy season is from December to May and the dry season from June to November.
We investigated two Tabernaemontana undulata populations at two diff erent areas (Site 1 and Site 2) 2 km from each other.Site 1 represents the primary, continuous forest where ten individuals were marked.Site 2 is located in a secondary vegetation area where a plantation of Brazil nuts (Bertholletia excelsa Humb.& Ponpl.) was made almost 40 years ago aft er deforestation.Th e plantation was abandoned and the area is now re-connected to the continuous forest, showing a mix of late successional and primary forest species.We also marked ten individuals at Site 2. Th e distance between individuals in the two populations ranges from 2-50 m.We conducted additional observations at a third area (Site 3) in the primary forest distant one kilometer from Site 1 and almost 500 m from Site 2 in order to verify possible microspatial variation in relation to the fl ower visitors' identity.

Data collection
Observations were carried out on the morphology, size and longevity of fl owers (11 fl owers from 10 individuals) and time of anthesis.Location of the stigmatic area of the style head was determined by the peroxidase technique (Kearns & Inouye 1993).Five isolated fl owers (fi ve individuals) were used to measure nectar volume during anthesis aft er com-pleting a pollinator census, with a micro-syringe, and sugar concentration, with a portable refractometer.Pollen viability was tested in 10 fl owers (10 individuals) with acetocarmine (Radford et al. 1974).Th e number of ovules and pollen grains was determined from 10 fl oral buds.Compatibility system was tested by the fruit set resulting from spontaneous self pollination (40 fl owers from 20 individuals) and hand self pollination (40 fl owers from 20 individuals), following the methods described by Darrault & Schlindwein (2005).Furthermore, 40 exposed fl owers (20 individuals from the two populations) were marked to verify fruit set under natural conditions.
In order to investigate the role of Euglossine bees in T. undulata pollination, we conducted diurnal pollinator censuses at Sites 1, 2 and 3 between October 19 and November 14, 2007, covering the total T. undulata fl owering season.We conducted 10 censuses (a total of 30 hours) at Site 1 and another 10 (30 hours) at Site 2 on alternate days.At Site 3, four additional censuses (a total of 12 hours) were undertaken along the fl owering period.In order to standardize data for statistical analysis, we defi ned a standard daily census period from 9am to noon when visitation is concentrated.Additional observations of the fl ower visitors' activity were conducted sporadically along the day (10 hours) and we also made nocturnal observations twice at Sites 1 (4 hours) and 2 (4 hours).During each 3h-census, we carefully watched the activity of fl ower visitors at one marked treelet, recording the total number of fl owers visited and the total time spent in fl owers by each visitor species.We then calculated the visitation rate as the average of the total number of fl owers visited by each fl oral visitor species in each census divided by the total time of a census (3 hours).To test the pollinator eff ectiveness, we marked and isolated with pollinator exclusion bags (voile) two unopened fl owers per plant in both Sites 1 and 2 populations (N=40) 1 day before a pollinator eff ectiveness experiment.Next day, we exposed unvisited, fi rst day fl owers, to visits of a single pollinator species which were bagged aft er each census to verify posterior fruit set.Each marked treelet in both populations was censused once.We recorded the number of buds and open fl owers of the marked treelets and quantifi ed the fruit production of censused treelets at the two populations in order to obtain an indicator of pollination success under natural conditions.
A majority of the visitation sequences were recorded with a video camera and fl ower visitors were identifi ed through these video observations, following the methods described by Kay & Schemske (2003).Identifi cation of all the fl ower visitors to species would have required collecting, which was not possible for video observations.
We performed the Mann-Whitney U Test for non parametric data using Systat 12.For statistical analyses, we considered signifi cant results if P ≤ 0.05.

Flower morphology, fl ower biology and compatibility system
Tabernaemontana undulata is a treelet which grows in the understory of terra fi rme forest.Flowering occurs once in a year, at the end of the dry season from October to November.Th e treelets, including some individuals which were not marked for the pollinator censuses, produced between 45-78 fl oral buds and their infl orescences usually open four fl owers a day, following the steady state pattern described by Gentry (1974).
Th e fl owers are hermaphrodite, salverform, presenting a long (N=11; mean= 29 mm; range= 24-33) and narrow (N=11; mean diameter= 3.2 mm; range= 2.8-3.6)fl oral tube.Th e corolla tube is pale red-white and the lobes are white fl ushed with red.Th e stamens are inserted on the corolla tube and the sessile, introrse anthers form a cone around the apex of the style head, occupying the entire inner part of the fl ower tube.Th e anthers shed longitudinally their pollen grains onto the apex of the style head, forming a pollen chamber characteristic of the tribe Tabernaemontaneae.In accordance with Fallen (1986), the style head (Fig. 1) shows three functional regions: (1) an apical, sterile region where pollen is deposited (secondary pollen presentation), (2) a median region which produces a sticky substance, and (3) a lower receptive region beneath a basal membranous ring.Upon contact with H 2 O 2 this lower region reacted by liberating oxygen.
Th e fl owers opened around 5am and last for one and a half day.Nectar is available aft er 8am.Mean nectar volume was 3μl.Sugar concentration varied from 16-23%.Only fi rst-day fl owers have a sweet scent (personal observation) which is perceptible only at a very close distance.However, aft er 9am the fl oral fragrance intensifi ed and we could perceive it ca. 1 m distant from the plant.
In the compatibility system experiment, both hand and spontaneous self-pollination tests resulted in a total failure of fruit production, suggesting that Tabernaemontana undulata is self-incompatible.From the 40 exposed flowers (control), none produced fruits under natural conditions.

Pollinator censuses
Site 1: In the understory of the primary forest, T. undulata is one of the very few species fl owering at the end of the dry season.We found only two herbaceous species simultaneously fl owering: Heliconia sp.(Heliconiaceae), which was visited by hummingbirds and Monotagma spicatum (J.B.Aublet) Macbride (Marantaceae) visited by an un-identifi ed bee.In the 10 censuses conducted, we registered Eulaema bombiformis Packard (1869) visiting T. undulata fl owers at Site 1.On average, E. bombiformis visited twice (Tab. 1) each censused treelet and the interval between visits was 15-30 min.From the few E. bombiformis visits received by a plant, most are concentrated in the morning (9am-midday), coinciding with the time when nectar was fully available and the sweet fl oral odor intensifi ed.Aft er midday, E. bombiformis visited T. undulata fl owers sporadically, but we never observed visits aft er 2pm.Aft er visiting all fi rst-day fl owers on the same plant, we oft en observed E. bombiformis visiting other T. undulata plants in sequence.Despite the fact that we did not mark the bees, we never saw two E. bombiformis individuals foraging simultaneously, which may indicate that a majority of visits were performed by a single bee individual.We also made nocturnal observations (4 hours) at Site 1, when no nocturnal fl ower visitors were registered.
Site 2: This disturbed area now connected to the continuous forest has late successional species growing together with some mature forest species.At the period of T. undulata flowering season, two arboreal species were simultaneously flowering at Site 2: the late successional species Bellucia cf.dichotoma Cogn.(Melastomataceae), which was in the flowering peak and the Brazil nut Bertholletia excelsa (Lecythidaceae), starting its flowering period.During all 10 censuses, we registered Euglossa sp.dominating the visits to T. undulata flowers.We also observed in just two censuses an unidentified butterfly species visiting two flowers per individual.Along each census, we observed Euglossa sp.performing a single visit on each censused treelet (Tab.1), also visiting only firstday flowers.Similar to E. bombiformis behavior at Site 1, Euglossa sp.concentrated its visits in the morning (9ammidday), with rare visits between midday-2pm, which suggests that the sweet floral fragrance plays an important role in attracting T. undulata diurnal flower visitors.We registered that Euglossa sp.frequently visited only few opened flowers on the same plant and then moved to another T. undulata treelet.The fact that we could not observe E. bombiformis visiting T. undulata flowers does not mean that this bee species was absent at Site 2. On the contrary, before each census (7am-9am), we could observe E. bombiformis visiting frequently the pollenrich flowers of B. cf.dichotoma.Although we could not observe directly E. bombiformis visiting B. excelsa because of tree height (30 m), there is a great possibility since the Brazil nut species represents an important nectar source for Euglossini bees and depends on them as its pollinators.In nocturnal observations at Site 2 (4 hours), no nocturnal flower visitors were observed.
Site 3: In four additional censuses conducted in a primary forest area distant 500 m from Site 2, we registered E. bombiformis at T. undulata fl owers.Th e fl oral visitor behavior was similar to that described for Site 1.In this area, we also found Heliconia sp. and Monotagma spicatum (J.B.Aublet) Macbride fl owering.

Frequency of visits and pollination eff ectiveness
We did not fi nd signifi cant diff erences on visitation rates of E. bombiformis and Euglossa sp.(Tab. 1) between the two populations at Sites 1 and 2 (Mann-Whitney U Test: 45, df = 1, P= 0.70).Th e number of visits performed by each pollinator taxa (Tab. 1) was not signifi cantly diff erent (Mann-Whitney U Test: 66, df = 1, P= 0.170), as well as the number of opened fl owers in the day when each treelet was censused (Tab. 1) in the two populations (Mann-Whitney U Test: 48, df = 1, P= 0.877).
In the per-visit eff ectiveness test, from 20 marked fl owers visited by E. bombiformis at Site 1 population, we registered one fruit produced.At site 2, there was no fruit production.Whereas T. undulata seems to be self-incompatible and, in fact, shows complex morphology preventing selfpollination, we consider that any fruit produced by a treelet must be a consequence of the pollinators' visits.
We quantifi ed the number of buds and fruit production under natural conditions on each censused treelet in both populations (Tab.2).Th ere was higher fruit production (Tab. 1) in Site 2 population compared to Site 1 (Mann-Whitney U Test: 15.5, df = 1, P= 0.008).

Discussion
Th e arrangement of the fl ower parts in Tabernaemontana undulata makes autogamy diffi cult and thus favors cross pollination in this species (Fallen 1986).Th e fl oral structure ensures that in a single visit the fl ower visitor fi rst deposits outcrossed pollen on the stigmatic surface and then transfers the self pollen.To self pollinate a fl ower, the fl ower visitor would have to visit the same fl ower at least twice in order to remove self pollen and then deposit it onto the style head.However, we never observed E. bombiformis and Euglossa sp.visiting the same fl ower consecutively.
Despite the fact that T. undulata pollination mechanism avoids self pollination, this does not prevent geitonogamy.Th e pollinator eff ectiveness test failure may be explained by at least two reasons: geitonogamy and a low fruit set pattern for this species.Th e deposition of self pollen onto the stigma of a self-incompatible species may interfere with the growth of pollen tubes from outcross pollination, resulting in a clogging eff ect.Observations of the foraging behavior of the fl ower visitors indicated that, while E. bombiformis visited all opened fl owers of the same treelet before moving to another T. undulata individual, Euglossa sp.visited few opened fl owers on the same plant and then moved to another T. undulata treelet.Diff erences in relation to the visitation pattern of the two bee species could have been responsible for the higher fruit production detected in Site 2 population, where Euglossa sp. was the unique fl ower visitor observed.Darrault & Schlindwein (2005) also considered that the best pollinator of Hancornia speciosa Gomez (Apocynaceae) would visit one or a few fl owers of the same plant individual before switching to fl owers of another treelet, avoiding geitonogamy in this way.However, the clogging eff ect through geitonogamy is unlikely to be the only reason for the very low fruit set observed in the pollination eff ectiveness test and under natural conditions.Low levels of fruit set are reported by a number of studies in Apocynaceae (Lopes & Machado 1999 and references therein) which may represent a widespread feature for this family.Herrera (1991) and Lin & Bernardello (1999) report very low fruit : fl ower ratios for Nerium oleander L. and Aspidosperma quebracho-blanco Schlecht.(Apocynaceae), respectively.Th ey suggest that an overproduction of fl owers in the two species could be explained in terms of increased pollinator attraction and pollen donation in order to ensure a small amount of fruit and seed set.
Euglossini bees are well-known for their trapline foraging on steady-state understory plants, being considered important long fl ight distance pollinators (Dressler 1982;Endress 1994;Borrell 2005).Th ese bees show some fl ower constancy, visiting the same individual plants daily along their feeding route (Janzen 1971).Flowers of species of Apocynaceae s.s. are considered important nectar sources for Euglossine bees and, because they have long proboscides compatible with the fl oral morphology commonly found in this family, these bees are generally considered to be pollinators (Tostes et al. 2003).
Th e salverform fl owers of T. undulata show some features which allow a classifi cation into the moth-pollination syndrome such as fl ower color and presence of a narrow fl oral tube (Endress 1994).In fact, the observed diff erence in fruit set between the two populations may also be related to visits from nocturnal pollinators.However, the fact that fl owers are open by the morning and that they emit an intense sweet fl oral scent, together with nectar availability at this time, indicate that diurnal fl ower visitors, especially long-tongued ones, may contribute at some level to T. undulata reproduction.
To test the per-visit eff ectiveness of the Euglossine bees we used a direct measure of pollinator eff ectiveness by measuring the fruit set in response to pollinator visits, especially because it requires fewer assumptions and gives clearer results (Spears 1983).Th rough this method, we observed that only E. bombiformis was able to transfer pollen and pollinate one fl ower.Such as result comes as surprising considering the fact that the two bee species have similar morphology and behavior.Due to the complex breeding system of T. undulata, with very low fruit production even under natural conditions, an indirect measure of pollinator eff ectiveness, such as the measure of the amount of pollen transferred to stigmas, would be very important at a future stage of the present study in order to clarify the role of these bee species as pollinators of T. undulata.

Figure. 1 .
Figure. 1. Tabernaemontana undulata Vahl.(Apocynaceae) in Central Amazonia, Brazil.Longitudinal section of the fl oral tube showing the style with the stigmatic head.

Table 1 .
Visitation rates of the fl oral visitors of Tabernaemontana undulata Vahl. in Central Amazonia, Brazil, registered along 20 censuses (Sites 1 and 2) at Fazenda Experimental UFAM in 2007 (19 October -14 November).Rates in bold represent the total number of visited fl owers by each fl oral visitor species in each census divided by the total time of a census (3 hours).Total number of legitimate visits is represented in parenthesis under each rate.(F) number of opened fl owers per individual plant in each census.

Table 2 .
Th e number of buds (NOF) on each censused treelet and number of fruits produced under natural conditions by all censused plants in the two populations of Tabernaemontana undulata Vahl. in Central Amazonia.Th ere are no signifi cant diff erences in the number of buds between the two populations (Mann-Whitney U Test: 35, df = 1, P= 0.256).Th e fruit set in the two populations is signifi cant diff erent (Mann-Whitney U Test: 15.5, df = 1, P= 0.008).