Functional aspects of floral nectar secretion of Ananas ananassoides, an ornithophilous bromeliad from the Brazilian savanna

Abstract

Background and Aims

Several members of Bromeliaceae show adaptations for hummingbird pollination in the Neotropics; however, the relationships between floral structure, nectar production, pollination and pollinators are poorly understood. The main goal of this study was to analyse the functional aspects of nectar secretion related to interaction with pollinators by evaluating floral biology, cellular and sub-cellular anatomy of the septal nectary and nectar composition of Ananas ananassoides, including an experimental approach to nectar dynamics.

Methods

Observations on floral anthesis and visitors were conducted in a population of A. ananassoides in the Brazilian savanna. Nectary samples were processed using standard methods for light and transmission electron microscopy. The main metabolites in nectary tissue were detected via histochemistry. Sugar composition was analysed by high-performance liquid chromatography (HPLC). The accumulated nectar was determined from bagged flowers (‘unvisited’), and floral response to repeated nectar removal was evaluated in an experimental design simulating multiple visits by pollinators to the same flowers (‘visited’) over the course of anthesis.

Key Results

The hummingbirds Hylocharis chrysura and Thalurania glaucopis were the most frequent pollinators. The interlocular septal nectary, composed of three lenticular canals, extends from the ovary base to the style base. It consists of a secretory epithelium and nectary parenchyma rich in starch grains, which are hydrolysed during nectar secretion. The median volume of nectar in recently opened ‘unvisited’ flowers was 27·0 µL, with a mean (sucrose-dominated) sugar concentration of 30·5 %. Anthesis lasts approx. 11 h, and nectar secretion begins before sunrise. In ‘visited’ flowers (experimentally emptied every hour) the nectar total production per flower was significantly higher than in the ‘unvisited’ flowers (control) in terms of volume (t = 4·94, P = 0·0001) and mass of sugar (t = 2·95, P = 0·007), and the concentration was significantly lower (t = 8·04, P = 0·0001).

Conclusions

The data suggest that the total production of floral nectar in A. ananassoides is linked to the pollinators' activity and that the rapid renewal of nectar is related to the nectary morphological features.

INTRODUCTION

Species of Bromeliaceae are mainly ornithophilous and represent one of the most important energy sources for hummingbirds in Neotropical regions (Snow and Snow, 1986; Bernardello et al., 1991; Sazima et al., 1996; Buzato et al., 2000; Krömer et al., 2006). Nectar is the floral resource for hummingbirds and, in Bromeliaceae, it is produced by septal nectaries whose structure was described by Bernardello et al. (1991) and Sajo et al. (2004).

It has been suggested that the characteristics of Bromeliaceae nectar are predominantly determined by putative adaptations of nectar sugars to preferences of pollinators, rather than by phylogenetic relationships (Krömer et al., 2006). Additionally, Stiles and Freeman (1993) verified that flowers associated with hummingbirds from distinct geographic regions shared a common sugar composition, indicating an adaptive convergence that reflects the taste preferences and/or the digestive physiology restrictions of hummingbirds.

As noted by McDade and Weeks (2004, p. 197), ‘Despite the central role that nectar plays in mediating plant–pollinator interactions, for most plant species, we know little more than that nectar exists’. The authors follow by saying ‘Clearly, we are far from having a complete understanding of the role of nectar in plant–pollinator interactions and of the evolution of nectar traits’.

Given these considerations, Bromeliaceae–hummingbird interactions represent a good model by which we may increase our knowledge of features of both the plant and the pollinator in the nectar-mediated interaction. Our focus was on the processes of nectar secretion including nectary characteristics, which could be informative in regard to the connection between nectar produced per flower and pollinator feeding behaviour. Additionally, we connected these results with a comparison of total nectar produced between ‘visited’ (i.e. nectar experimentally emptied) and ‘unvisited’ (control) flowers with the aim of identifying the effect of pollinator activity on plant physiological mechanisms related to both energy investment and saving.

MATERIALS AND METHODS

Study site and study organism

The study was conducted in a population of Ananas ananassoides (Baker) L.B.Sm. (Bromeliaceae) occurring at Reserva Particular Palmeira da Serra (22°48′50″S, 48°44′35″W), Pratânia municipality, São Paulo State, Brazil, in a cerrado sensu stricto phytophysiognomy (Brazilian savanna). The climate is characterized as Cwa (Köppen, 1948) and mesothermal, with rains in the summer and drought in the winter, and the median temperature of the hottest month is >22 °C (Cunha and Martins, 2009). Fieldwork was conducted during two consecutive flowering seasons of A. ananassoides. Flowering occurred at the beginning of the rainy season, spanning September to November 2008 and from September to October 2009.

This terrestrial bromeliad has leathery leaves and spiral phyllotaxis; a floral scape of approx. 1·3 m in length, with large, coloured bracts; a spike with many densely arranged flowers; sessile, tubular, trimerous flowers having lilac petals with stamens included and attached to the petal bases and a syncarpic gynoecium and inferior ovary with developed septal nectaries; and syncarpic, fleshy fruit (Wanderley and Martins, 2007). Voucher specimens were deposited in the Herbarium of the Department of Botany (BOTU) of the Institute of Biosciences, UNESP–Univ Estadual Paulista, Brazil, under numbers 24198 and 24193.

Plant–pollinator interactions

Floral morphology and events of anthesis were observed in ten plants, with emphasis on the opening time, colour of the floral elements, presence of nectar and floral longevity. Flowers were monitored to check for visitors at different times of day throughout the season totalling 60 h of observation, from dawn (0300 h) to evening (1700 h). Visitor behaviour was described based on field observations, analysis of photographs and video, recording the time, duration and frequency of visits, body regions that come into contact with the anther/stigma and the type of resources collected. Some visitors were captured for identification, and others, such as hummingbirds, were identified by photographs and video.

Structural and ultrastructural studies

For anatomical studies, samples were fixed in FAA 50 (formaldehyde, acetic acid, 50 % ethanol 1:1:18 v/v/v) (Johansen, 1940) for 24 h, followed by gradual dehydration in an ethanol series; the samples were then embedded in hydroxyethyl-methacrylate (Leica Microsystems Inc., Heidelberger, Germany). Transverse and longitudinal sections (6–8 µm) were cut with a rotary microtome and stained with 0·05 % toluidine blue (pH 4·3) (O'Brien et al., 1964). Histochemical analyses were performed according to the references in Table 1.

For scanning electron microscopy (SEM), fragments of ovary were isolated and fixed in 2·5 % glutaraldehyde (0·1 m phosphate buffer, pH 7·2), dehydrated in an ethanol series, dried to critical point and subsequently sputter-coated with approx. 10 nm gold as described by Robards (1978). The samples were examined in a scanning electron microscope, model Quanta 200 (Fei Company, Hillsboro, OR, USA), and all images were processed digitally.

For transmission electron microscopy (TEM), samples of nectaries obtained from the basal region of the ovary of functional flowers were fixed in glutaraldehyde (2·5 % with 0·1 m phosphate buffer, at pH 7·3) and left overnight at 4 °C. They were then post-fixed with 1 % osmium tetroxide (OsO₄) in the same buffer for 2 h at room temperature, dehydrated in a graded series of acetone solutions and embedded in Araldite resin. Ultrathin sections were stained with uranyl acetate and lead citrate (Reynolds, 1963) and observed under a Philips TEM 100 microscope at 80 kV (Philips, Czech Republic).

Process of nectar secretion during anthesis and nectar sugar composition

The process of nectar secretion and the effect of nectar removal on the total energy and water content secreted during the lifetime of the flowers were investigated using all of the flowering individuals available in the population. Inflorescences with labelled pre-anthesis buds (n = 50 flowers, ten plants) were protected with bridal veil bags to prevent nectar depletion by visitors during the experiments, as recommended by Corbet (2003). The volume (μL) of nectar from open flowers was measured immediately after collection using graded syringes (Hamilton, USA). The sugar concentration was measured with a digital refractometer (Callmex, Brazil) as per cent weight/weight sucrose (g sucrose per 100 g solution). These nectar data (volume and concentration) were used to estimate the total milligrams of sugar produced per flower using the exponential regression proposed by Galetto and Bernadello (2005).

To determine the pattern of nectar secretion throughout anthesis in the absence of nectarivores (i.e. accumulated nectar per flower), 11 groups of flowers were used, each one with three bagged flowers, referred to here as ‘unvisited’ flowers (n = 33 flowers, seven plants). Every hour from 0600 h to 1600 h, the nectar accumulated in each flower was withdrawn, and the volume and concentration were measured each time in a group of three flowers, which were then discarded, so that each of the 33 flowers was evaluated only once.

To evaluate the floral response to repeated nectar removals, one group of 17 flowers was used in an experimental design simulating multiple visits by pollinators (11 visits) to the same flowers over the course of anthesis (i.e. nectar experimentally emptied). This group was referred to here as ‘visited’ flowers (n = 17 flowers, three plants). In each flower of the ‘visited’ group the accumulated nectar was also withdrawn every hour from 0600 h to 1600 h, but in this treatment the nectar of each flower was drained 11 times. We summed the partial amounts for each flower (volume and milligrams of sugar obtained each hour) and averaged them to calculate the mean total production per flower during anthesis. Nectar measurements at 1700 h, in both treatments, ‘visited’ and ‘unvisited’, were not possible because the corolla began to wilt, preventing the entry of the syringe into the tube. We compared nectar production of ‘unvisited’ (control) with ‘visited’ (nectar experimentally emptied) flowers using volume, concentration and solute mass obtained from the following sets of data: for ‘unvisited’ flowers we used the values obtained from five groups of three flowers drained during the period comprised between 0900 h and 1300 h (n = 15 flowers, three plants), during which time nectar volume reached the maximum and remained almost constant; for ‘visited’ flowers we used values of total nectar produced per flower during the whole of anthesis (n = 17 flowers, three plants). The differences were evaluated by a t-test. The data of volume and mass of sugar showed different standard deviations, so the Welch correction was applied to perform the test.

For sugar composition analysis, samples of nectar (2 µL) were collected from open flowers of three plants at approx. 1000 h during October 2008. These samples were stored at –20 °C as dried spots on Whatman No. 1 filter paper as described by Galetto and Bernardello (2005). After the samples were thawed to ambient temperature, nectar was recovered from the filter paper by static elution with 100 µL of distilled water for 3–4 min, followed by centrifugation for 5 min at 11 000 g. The supernatant was analysed by isocratic high-performance liquid chromatography (HPLC) with the LC1 Waters system. A 20 µL aliquot of the sample and standard solution was injected. Water (MilliQ, pH 7) with a flow rate of 0·5 mL min⁻¹ was used as the mobile phase. Sugars were separated in a Waters Sugar-Pack I column (6·5–300 mm), maintained at 90 °C, and were identified by a refractive index detector (Waters 2410).

RESULTS

Plant–pollinator interactions

Inflorescences are indeterminate, exhibiting 1–10 flowers in anthesis per day. Flower opening began at approx. 0300 h and was completed at 0600 h, when the corolla lobes were completely recurved and the nectar accumulated inside the corolla tube was available to floral visitors. At the end of each day, at approx. 1700 h, the lilac petals presented a pink tonality, and the apices were curled into the corolla centre, preventing access by visitors to the tube. Anthesis lasted approx. 11 h.

Relatively small hummingbirds of two species, Thalurania glaucopis (Fig. 1A) and Hylocharis chrysura (Fig. 1B), performed several visit sequences between 0700 and 0900 h, acting as trapliners. They foraged on 5–10 flowers per inflorescence (i.e. the majority of open flowers of each inflorescence) and usually visited each flower once or twice during the same sequence. After 1000 h, visits to the inflorescences were less frequent but continued intermittently. Hummingbirds flew over the inflorescences for approx. 1 min. They then hovered in front of a flower and introduced their beak into the floral tube for 1–3 s. After that, they flew toward other flowers of the same plant or of neighbouring plants, or they landed on a nearby branch to perform systematic beak cleaning.

Individuals of two butterfly species, Hamadryas februa (Fig. 1C) and Phoebis sennae (Fig. 1D), also visited the flowers, landing outside the corolla and introducing their proboscides for 1–3 min into the floral tubes. Next, they would fly toward other flowers on the same plant or on neighbouring plants.

Large bees, Bombus morio (Fig. 1E), visited A. ananassoides flowers as legitimate pollinators, but they visited less frequently than hummingbirds and butterflies. Bombus morio individuals hovered in front of flowers and placed part of their heads into the floral tubes, coming into contact with the reproductive structures during apparent nectar collection. Small-bodied individuals of three bee species, Trigona spinipes, Plebeia droryana and one unidentified species, visited A. ananassoides flowers for pollen. These species seemed to act primarily as pollen thieves and may have caused self-pollination.

Flower visitors, feeding behaviours (legitimate or illegitimate), nature of the reward apparently collected and frequency of visits are summarized in Table 2. We assumed legitimate floral visitors to be those who, during resource collection, were able to perform cross-pollination through contact with the anthers, which were filled with pollen, and the receptive stigma, and that visited different individuals of A. ananassoides sequentially.

Nectary structure and ultrastructure

The septal nectary in A. ananassoides extends from the base of the ovary locules to the base of the style, where it opens to the base of the corolla tube. The nectar-secreting channels exhibit a convoluted and undulating outline (Fig. 2A) that is more developed at the basal region of the ovary (Fig. 2B). At or above the ovule attachment region, the channels present a linear outline (Fig. 2E) and have progressively less secretory tissue (Fig. 2F), ending in three apical pores through which the nectar flows.

In the secretory phase, the convolute region of the septal nectary (Fig. 2B) comprises two well-delimited regions in cross-section: an epithelium composed of 1–3 layers of juxtaposed, columnar cells that are disposed perpendicular to the septal nectary surface, and a differentiated nectary parenchyma composed of 3–6 layers of smaller, isodiametric cells (Fig. 2G). Cuticle was indistinguishable in this nectary region but it was visible on the non-secretory surface (Fig. 2F). The cells of the nectary parenchyma present a denser cytoplasm than the ground parenchyma cells (Fig. 2F). The septal nectaries lack an individual vascular supply, but vascular bundles (Fig. 2C) composed of phloem and primarily xylem elements occur near the nectary parenchyma tissue (Fig. 2B) without ramifying into it. The vascular parenchyma cells of both xylem and phloem contain dense cytoplasm, well-developed plastids with prominent starch grains, previously detected with the use of Lugol reagent, and numerous, small vacuoles (Fig. 2D). The presence of idioblasts with raphides (Fig. 2E, F, H) of calcium oxalate, confirmed with the use of sulfuric acid (5 %), is typical in the neighbouring nectary parenchyma.

Histochemical analyses of the nectary during the secretory stage were positive for lipophilic and hydrophilic substances (Table 1). Staining with Sudan IV revealed the presence of small oil droplets dispersed in the protoplast of epithelial and parenchyma secretory cells. Treatment with NADI reagent clearly showed the presence of terpenes, which were observed as densely stained droplets both inside the epithelial and nectary parenchyma cells and in the periplasmic space (Fig. 2I); bodies stained with NADI that are larger in size and spherical or ellipsoid in shape also occur on the surface of the epithelial cells and inside the nectary channels. All regions of the nectary showed strong positive reactions for polysaccharides because of the abundance of starch grains. Treatment with both Lugol's reagent and Dragendorff's reagent confirmed the presence of darkly stained starch grains in the epithelial and nectary parenchyma cells at anthesis, and it was possible to detect clearly a gradual reduction in the size and abundance of starch grains toward the epithelial cells (Fig. 2J). Samples of the nectary that were treated with Fehling's reagent exhibited a positive reaction, indicating the presence of reducing sugars. Phenolic compounds, mucilage and proteins were absent in the epithelial and parenchyma cells.

The epithelial cells of secreting nectaries exhibit thin radial and tangential walls, large ellipsoidal nuclei, abundant cytoplasm and small vacuoles (Fig. 3A). The surface of the outer tangential wall facing the channel is electron dense and presents deposits of osmiophilic material intermixed with fibrilar wall material (Fig. 3C, E) that are released by the disintegration of the cell wall during the channel development that occurs by schizogenesis (S. R. Machado et al., unpubl. res.). Ultrastructural analysis confirmed that epithelial cells in this region do not have cuticle (Fig. 3C, E, F). The plasma membrane has an irregular outline (Fig. 3B, C) and gives rise to the periplasmic space, which is more developed in the apical pole of the epithelial cell (Fig. 3D, F). These spaces contain paramural bodies (Fig 3E) and large lipophilic drops (Fig. 3F). Vesicles (Fig. 3C) and portions of endoplasmic reticulum (Fig. 3E) close to the plasma membrane are visible.

The cytoplasm of the epithelial cells stains densely and contains large drops of lipophilic material (Fig. 3B), mitochondria (Fig. 3B, E, I), polyribosomes (Fig. 3B, E), endoplasmic reticulum (Fig. 3B, E), dictyosomes with adjacent secretory vesicles (Fig. 3E, G) and plastids (Fig. 3E, I). Some plastids are devoid of thylakoid membranes, and contain a homogeneous stroma and 1–2 ovoid starch grains (Fig. 3E); other plastids, mainly in the subjacent layer, are well developed and filled with prominent starch grains (Fig. 3I). The vacuoles are variable in size and are electron lucent (Fig. 3A, E).

Plasmodesmata are very common in the anticlinal (Fig. 3G) and periclinal walls (Fig. 3H) of the epithelial and parenchyma cells. The nectary parenchyma cells (Fig. 3J) contain slightly lobed nuclei; the cytoplasm is less abundant than in epithelial cells and contains plentiful mitochondria and globe-shaped plastids that are filled with conspicuous starch grains; the vacuome is constituted by one central vacuole and numerous small ones in the periphery, and images suggesting the occurrence of fusion of vacuoles are observed in these cells.

Process of nectar secretion during anthesis and nectar sugar composition

‘Visited’ and ‘unvisited’ flowers presented a volume of nectar accumulated at 0600 h, of 23·47 ± 11·84 µL and 27 µl (median), respectively. In each group, ‘visited’ and ‘unvisited’, we registered just one flower with no nectar at 0600 h. Nectar volume in ‘unvisited’ flowers appeared to increase until 0800–0900 h, reaching a total of approx. 60 µL per flower (Fig. 4A). Between 0900 and 1300 h, nectar volume reached the maximum and remained almost constant (Fig. 4A). Thereafter, there was a continuous decrease in the total volume per flower, with a low amount of nectar per flower registered at 1600 h (Fig. 4A). For the ‘visited’ flowers there was a decrease in the nectar accumulated per hour during the afternoon in terms of both volume (Fig. 5A) and mass of sugar (Fig. 5B). Nectar concentration in ‘visited’ and ‘unvisited flowers varied very little during anthesis (Fig. 6); thus, nectar solutes (milligrams of sugar, Figs 4B and 5B) showed the same pattern described for nectar volume, for both groups.

We found a general effect of repeated nectar removals on accumulated total production during a flower's life (Table 3). The mean total accumulated nectar volume of ‘visited’ flowers was higher than the mean volume in ‘unvisited’ flowers, and an inverse pattern was registered for nectar concentration values (Table 3). Thus, volume differences between flower groups can disappear in terms of nectar solutes since mean nectar concentration in the ‘visited’ and ‘unvisited’ flowers differed significantly. However, the differences were confirmed, because a higher quantity of nectar (in milligrams of sugar) was produced in the ‘visited’ flowers compared with ‘unvisited’ ones (Table 3). The nectar carbohydrate composition for this species was 138·9 ± 27·3 mg mL⁻¹ of sucrose, 61·6 ± 9·9 mg mL⁻¹ of glucose and 52·7 ± 3·5 mg mL⁻¹ of fructose, indicating a sucrose-dominated nectar.

DISCUSSION

Plant–pollinator interactions

The two hummingbird species that pollinate A. ananassoides are relatively small bodied and short beaked (Mendonça and Anjos, 2005), and they are important visitors of other Bromeliaceae species (Araújo and Sazima, 2003; Canela and Sazima, 2003; Kaehler et al., 2005; Machado and Semir, 2006). Individuals of Hylocharis chrysura possess beaks approx. 19 mm long (Mendonça and Anjos, 2005), which is comparable with the dimensions of A. ananassoides flowers and with the distance between the base of the corolla tube and the floral reproductive structures. Machado and Semir (2006) showed that in dense populations of two bromeliads, Aechmea nudicaulis and Vriesea philippocoburgii, individuals of one hummingbird species, T. glaucopis, were the only visitors and exhibited territorial behaviour. In contrast, we observed that in sparser populations of A. ananassoides, T. glaucopis acted as a trapliner, exhibiting feeding behaviour similar to that observed by Canela and Sazima (2005) in Bromelia antiacantha (Bromeliaceae). Trapline foraging has been reported for various animal species collecting food from renewable resource patches (Ohashi and Thomson, 2009, and references therein). Considering that A. ananassoides is predominantly allogamous, with low self-fertility (0–8 %) (Coppens et al., 1993), such trapline feeding behaviour favours outcrossing, which is particularly important for this bromeliad species.

Nectary structure and ultrastructure

Based on position, the septal nectaries of A. ananassoides are considered to be interlocular (sensuSimpson, 1998), as in other epigynous Bromelioideae investigated (Sajo et al., 2004). The anatomy of the septal nectary of A. ananassoides agrees generally with previous reports for other Bromeliaceae species (Daumann, 1970; Cecchi Fiordi and Palandri, 1982; Varadarajan and Brown, 1988; Bernardello et al., 1991), in which it was found that the septal nectaries are structural (cf. Fahn, 1979) because they are histologically differentiated into an epithelium and nectariferous parenchyma.

Our results showed that the septal nectary of A. ananassoides lacks a cuticle in the convolute region, which could be related to the mechanism of the channel development that occurs by schizogenesis (S. R. Machado et al., unpubl. res.). The absence of cuticle in this region could not be considered as an artefact since we have registered the presence of intact cuticle in the nectary of other species processed by the same methods for optical and TEM analysis (Machado et al., 2006, 2008; Paiva and Machado, 2008). Absence of cuticle in septal nectaries was also reported by Bernardello et al. (1991) for the bromeliad Tillandsia tenuifolia. Therefore, the secretion of the nectar in A. ananassoides seems to occur through the cell wall of epithelial cells. The overall ultrastructure of the septal nectary examined here is similar to that previously described in the nectar-producing tissues (Fahn, 1988; Nepi, 2007; Paiva and Machado, 2008). Abundance of mitochondria, primarily in the sub-epithelial layers of parenchyma cells, is consistent with elevated energetic demands due to secretory processes, and shows the involvement of this tissue in the secretion of nectar, as reported for other nectaries (Nepi, 2007). The septal nectary of A. ananassoides is not vascularized, but it is associated with numerous vascular bundles of the ovary; in addition, the nectary and vascular parenchyma cells contain large starch reserves that can used as energy in the pre-secretory and secretory phases, as commonly seen in the nectaries of other plants (Paiva and Machado, 2008). The presence of endoplasmic reticulum and vesicles close to the plasma membrane, and formation of ample periplasmatic spaces in the epithelial cells of A. ananassoides, suggest that the elimination of secretions from the protoplast is by exocytosis. This process of secretion has been observed in other glands of different species (Fahn, 1979, Nepi, 2007).

According to Ren et al. (2007), the sugar in nectar is supplied from at least two sources: the hydrolysis of nectary starch in the late phase of nectary development and the flux of photosynthates into the nectary. When a large amount of nectar is produced in a short time, it is generally produced from the hydrolysis of starch stored in the parenchyma (Pacini et al., 2003), and this seems to be the case for A. ananassoides. Starch stored in nectary parenchyma could act as a finite resource in the short-lived flowers of A. ananassoides and could thus explain the lower concentration of nectar in repeatedly emptied flowers, which suggests a depletion of available carbohydrates in nectary parenchyma.

Process of nectar secretion during anthesis and nectar sugar composition

Our results for volume and concentration of floral nectar in A. ananassoides agree with the findings for other ornithophilous Bromeliaceae species reported by Galetto and Bernardello (2003), Canela and Sazima (2005) and Machado and Semir (2006). Hovering hummingbird pollinators have particular requirements as they have higher energy needs than bees, butterflies and other insects (Cruden et al., 1983; Nicolson, 2006) and more need of water shunting than other species of birds (Nicolson, 2006), and their associated flowers usually produce copious nectar that is more dilute than that produced by insect-pollinated flowers (Baker and Baker, 1983). Mean nectar volume produced by a flower throughout its lifetime in the six Bromeliaceae species studied by Galetto and Bernardello (1992) was 16·4 ± 14·23 µL, and the mean amount of sugar produced per flower was 7·36 ± 6·93 mg. Flowers of these species lasted from 15 h to 6 d, exhibiting a much longer life span in general than A. ananassoides flowers. Nevertheless, despite their short life, A. ananassoides flowers presented a greater accumulated volume and mass of sugar during their life span, resulting in a considerable energy source for pollinators.

Nectar in ornithophilous flowers presents an average dilution of 75–80 % water (Nicolson and Fleming, 2003), which is comparable with the values verified in A. ananassoides floral nectar. Maintenance of the sugar concentration, even in dry warm environments, ensures adequate viscosity to allow nectar consumption by pollinators, which is especially important for nectarivorous birds (Baker, 1975; Nicolson, 1995; Nicolson and Nepi, 2005). In A. ananassoides flowers, nectar concentration, varying from approx. 26 % to 30 %, remained almost constant throughout anthesis, favouring collection and possibly fidelity by hummingbird pollinators. The concentration of nectar can be much more important for pollinators than the volume because it greatly affects nectar energy intake and osmotic balance, as well as their ability to collect it from the nectary or from the nectar chamber (Roberts, 1996). The sucrose-dominant nectar of A. ananassoides matches hummingbird preferences, as previously described for other ornithophilous species by Baker and Baker (1983), Stiles (1976) and Galetto and Bernardello (2003). A prevalence of sucrose in Bromeliaceae nectar and a convergence of nectar features in hummingbird-visited species growing in different biogeographic regions were found by Galetto and Bernardello (2003).

In A. ananassoides, flowers repeatedly emptied of nectar clearly showed an increase in both the total volume and total milligrams of sugar produced. This result differs from that seen in other species of Bromeliaceae, in which the total sugar production was not affected by repeated nectar removal or was diminished, as in Puya spathacea (Galetto and Bernardello, 1992). The A. ananassoides septal nectary corresponds to the ‘labyrinthine common nectarial cavity’ type, sensuSchimd (1985), which increases the nectary surface by undulation and convolution as observed by Bernardello et al. (1991) in other Bromeliaceae species. The continuous nectar production throughout the day in repeatedly ‘visited’, i.e. experimentally emptied, flowers of A. ananassoides may be associated with the structure of the nectary that has a great longitudinal size, a labyrinthine surface, phloem and abundant xylem. Therefore, these nectary features could allow a rapid nectar renewal after a visit and a relatively constant supply of water and solutes for traplining hummingbirds throughout the day.

Commonly, hummingbird-pollinated flowers begin to secrete nectar 1–4 h prior to the activity of its pollinators, and the rate of secretion continues until some critical amount has accumulated, and then nectar secretion ceases (Stiles, 1976; Cruden et al., 1983; Castellanos et al., 2002), similar to the pattern verified in A. ananassoides flowers. Based on our observations, three different phases of the process of secretion throughout anthesis could be hypothesized, comprising periods of active secretion, cessation and reabsorption of nectar. A short initial secretion period was identified in ‘unvisited’ flowers of A. ananassoides, as well as a clear decrease in the water and sugar content after 1400 h, suggesting the occurrence of active nectar reabsorption. This nectar decrease was not due to evaporation, because nectar concentration was maintained during the whole of anthesis. Usually, the concentration of nectar increases and its volume decreases, as a consequence of evaporation (Cruden et al., 1983). Reabsorption of nectar has been assumed to have the function of retrieving energetically valuable sugars that are not utilized by pollinators (Nicolson, 1995). Considering that nectar production can require a high energy investment (Southwick, 1984), the recycling of sugar not collected by floral visitors could represent an important mechanism of energy saving by the plant, as the expenditure of energy in the reabsorption process could be lower than the energy recovered by sugar influx, resulting in an overall net gain, as reported by Stipczynska and Nepi (2006). Nectar reabsorption in A. ananassoides flowers may be related to the large surface area of the septal nectary. It is important to note that the contact of nectary tissue with the secreted nectar is a prerequisite for nectar reabsorption (Bonnier, 1878; Búrquez and Corbet, 1991; Nepi et al., 1996). For A. ananassoides, in the absence of pollinator visits, the reabsorption of water and sugar could represent an important mode of energy saving, as this species produces a fleshy fruit clustered into a highly hydrated, sweet infructescence. Nevertheless, available nectar may be close to zero if visitation by pollinators to open flowers is intense; then reabsorption may not be energetically significant to the plant. However, considering that visitation rates of pollinators can vary between days, populations and among plants of a population, this ability of A. ananassoides could represent an advantage in some situations.

In contrast, the process of nectar secretion was maintained until 1600 h if flowers were repeatedly ‘experimentally emptied’ of nectar every hour (‘visited’ flowers), indicating that secretion ability is preserved throughout the flower's life span if visits of pollinators take place. These results suggest that if flowers of A. ananassoides are not visited, a homeostatic mechanism drives active nectar reabsorption; this physiological mechanism that adjusts nectar solutes and volume was previously reported for other plant species (Galetto et al., 1994; Nepi et al., 2011). Thus, the process of nectar secretion can be adjusted, favouring successive visits, i.e. renewal of nectar after a visit; or a nectar-saving mechanism can be activated if flowers are not visited, i.e. active secretion is diminished and nectar reabsorption occurs at the end of the flower's lifetime.

ACKNOWLEDGEMENTS

We are indebted to the reviewers for useful and constructive comments on our manuscript, to M. G. Sajo for her suggestions about septal nectaries in Bromeliaceae, to S. Matheus for the identification of the bees, to I. Sazima for the identification of the hummingbirds, to L. A. Kaminski and C. A. Iserhard for the identification of butterflies, to T. M. Rodrigues for the informatics support, to the technicians of the Electron Microscopy Centre (CME) IBB, UNESP for their assistance, and to M. Guarnieri (Department of Environmental Sciences, University of Siena) for performing HPLC analysis. L.G. is a researcher from CONICET and thanks SECyT (Universidad Nacional de Córdoba) and CONICET for support. This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo–FAPESP (Thematic Project–Biota, Process number 08/55434 and grant number IC 08/55433 awarded to J.M.S.) and Conselho Nacional de Desenvolvimento e Pesquisa–CNPq (Edital Universal 470649/2008 and grant number PQ 301464/2008) awarded to S.R.M. Both E.G and S.R.M contributed equally to the supervision of this study.

LITERATURE CITED