Pollination biology of four sympatric species of Habenaria (Orchidaceae: Orchidinae) from southern Brazil

Study site

Studies were performed in two neighbouring municipalities (Cambará do Sul, approximately 29°10′S, 50°19′W, and São Francisco de Paula, approximately 29°25′S, 50°23′W) in the Atlantic Rain Forest domain, in the State of Rio Grande do Sul, Southern Brazil. For conservation reasons, we cite here only the coordinates of both municipalities and omit the exact location details of the populations. These data, however, are available on request. This particular region is locally known as Campos de Cima da Serra and consists of a mosaic of native grasslands and forest patches dominated by Araucaria angustifolia (Bertol.) Kuntze (Araucariaceae), at an altitude of c. 900–930 m. Average annual rainfall is c. 2468 mm and annual average temperature is approximately 14.5 °C (Moreno, 1961). The climate is characterized by a cool summer, a cold winter and the lack of a well-defined dry period (Nimer, 1989; Almeida, 2009). All species studied inhabit swamps dominated by Eryngium pandanifolium Cham. & Schltdl. (Apiaceae).

Studied species

Habenaria johannensis and H. macronectar are currently placed in section Macroceratitae Kraenzl. (Kränzlin, 1892; Batista et al., 2006), an easily identifiable assemblage of species with proportionally large, greenish–white, long-spurred (spur length ≥ 6 cm) flowers.

Habenaria megapotamensis and H. montevidensis were never assigned formal sectional classification. According to an unpublished phylogenetic analysis based on DNA sequence data (J. A. N. Batista, pers. comm.), these two species belong in the same clade, which also includes H. exaltata Barb. Rodr., H. henscheniana Barb. Rodr. and H. ekmaniana Kraenzl.

Of the species studied, H. johannensis has the widest geographical distribution, ranging from Bolivia and Paraguay to north-eastern, central western, south-eastern and southern Brazil. It is typical of wet and marshy areas, but occasionally can occupy human-disturbed habitats, growing on damp roadsides or in ditches. Habenaria macronectar is restricted to south-eastern and southern Brazil and Uruguay. This species occurs in wet or marshy areas. Habenaria megapotamensis is known only from southern Brazil, whereas H. montevidensis also occurs in southern Brazil and Uruguay. These last two species are typical of marshy areas or wetlands, and most records are derived from the north-eastern highlands of Rio Grande do Sul. Throughout this paper, we follow the orchid classification of Chase et al. (2003).

Floral features and nectar properties

Ten fresh, intact flowers from each species were used to record and measure flower features (column and pollinarium structure, spur length, nectar-column height, as well as nectar volume and concentration; see Table 1). All flowers came from five bagged specimens of each species, and nectar parameters were always measured for each of these taxa during the period when their pollinators had been recorded in activity. Nectar volume was measured by means of a microsyringe, and nectar concentration (total sugar) was measured using a manual pocket (0–32) refractometer. Because of the small quantity of nectar produced by H. montevidensis flowers, nectar volume in this species was measured by means of a P2 micropipette, and nectar concentration measurements were based on the total accumulated nectar of all the flowers analysed. Data on spur length, nectar volume, height of nectar column and nectar concentration (total sugars) for all studied species were tested for normality by means of a Shapiro–Wilk test and, later, statistically compared by means of a Kruskall–Wallis test (because the data did not display a normal distribution) followed by a Dunn post-hoc test, using the BioEstat software (Ayres et al., 2007) (Table 1).

Flower morphology was studied using fresh and alcohol-preserved flowers [70% (v/v) ethanol]. Plant vouchers were deposited at the ICN Herbarium of the Universidade Federal do Rio Grande do Sul (UFRGS) under the following accession numbers: H. johannensis (M. Pedron 5), H. macronectar (M. Pedron 1); H. megapotamensis (M. Pedron 10); H. montevidensis (M. Pedron 9). Throughout this paper, we follow the orchid morphology concepts of Dressler (1993).

Pollination

The pollination biology of H. megapotamensis was studied in the field during the flowering seasons of 2010 and 2011 (Table 2). The other three species were studied in 2011 (Table 2). Generally, the observation period for both years began in late January or early February and was completed by mid March, totalling 181 observation hours (Table 2). Both, crepuscular–nocturnal and diurnal observations were made on each species. Here, we define ‘diurnal observations’ as those made between 06:00 and 18:00 h and ‘crepuscular–nocturnal’ observations as those made from 18:00 h onwards. The specific observation period for each species and the specific number of diurnal and crepuscular–nocturnal hours spent observing each species are detailed in Table 2. When our early observations clearly indicated the prevalence of some kind of pollinator activity (e.g. diurnal vs. crepuscular–nocturnal), our fieldwork schedule was adapted accordingly. Therefore, most observation hours for H. johannensis, H. macronectar and H. megapotamensis were performed during the nocturnal period (Table 2). Conversely, most observation hours for H. montevidensis were made during the day. However, as nocturnal, non-pollinating visitors were also recorded for this species, intact flowers (36 flowers from 21 inflorescences) were marked at the end of the observations for the day and checked at the beginning of the next day, before observations had begun, in order to verify whether they had been pollinated and/or had their pollinaria removed during the night. Pollinator behaviour was documented for all species studied using field notes, photography and, when possible, video. The video record made it possible to gain a better understanding of the pollination process for most of the species studied, especially those visited by crepuscular–nocturnal pollinators. Individuals of both, pollinating and non-pollinating insects were collected and sacrificed for taxonomic identification. These insect vouchers were deposited at the entomological didactic collection of the Insects Ecology Laboratory, Zoology Department, UFRGS.

Breeding system, fruiting success and pollination efficiency

Breeding system experiments were performed in situ, by bagging inflorescences in order to exclude natural pollinators. Bags were supported with the help of wooden stakes. Four treatments were applied to these inflorescences: intact flowers (control); emasculation; manual self-pollination; and cross pollination (Table 3). Treatments that set fruit were compared using a χ²-test (α = 0.05). The number of plants per species and flowers used per treatment are shown in Table 3. Intact flowers of these plants were also used to record flower lifespan for each species.

In order to assess the efficiency of pollination, the fruiting success (number of fruit divided by the number of flowers produced) was calculated for each species, at the end of the observation periods. During 2010, 30 inflorescences of H. johannensis (totalling 269 flowers), 49 inflorescences of H. macronectar (totalling 526 flowers) and 32 inflorescences of H. megapotamensis (totalling 1568 flowers) were available. It was not possible to study the fruiting success for H. montevidensis in 2010, as fruiting inflorescences were destroyed, probably eaten by unidentified animals. During 2011, 45 inflorescences of H. johannensis (totalling 387 flowers), 16 inflorescences of H. macronectar (totalling 154 flowers), 17 inflorescences of H. megapotamensis (totalling 561 flowers) and 35 inflorescences of H. montevidensis (totalling 429 flowers) were available. Fruiting success (mean fruit set per inflorescence) in both years for all species except H. montevidensis (see above) was statistically compared using an independent, two-sample t-test (unequal sample sizes, equal variance) by means of PAST software (Hammer, Harper & Ryan, 2001) (Table 4).

In 2011, Nilsson's male efficiency factor (percentage of pollinated flowers divided by the percentage of flowers acting as pollen donors; Nilsson et al., 1992) was also used to calculate pollination effectiveness for all species studied. The total number of flowers used in these calculations and the percentages of flowers that were pollinated and that acted as pollen donors are shown in Table 4.

Floral features

Flowers are resupinate and clustered in long, terminal, racemose inflorescences bearing a maximum of 19 to 87 flowers, depending on the species (Table 1; and see also Supporting Information, Fig. S1). The flowers of H. johannensis, H. macronectar and H. montevidensis are greenish–white and those of H. megapotamensis are light green (Fig. 1A; see also Supporting Information, Fig. S1). The lifespan of intact flowers ranged from 12 to 21 days (Table 1). In all studied species, the dorsal sepal and lateral petals form a hood-like structure that partially hides the column. In H. montevidensis, the lateral petals are entire, and in the other three species they are bifid. The labellum is trilobed, and its posterior part is prolonged to form a spur that is partially filled with nectar (Fig. 1A). Mean spur length ranged from 1.62 to 13.09 cm, and statistically significant differences were found between the studied species (H = 36.19, P < 0.0001, d.f. = 3; see Table 1), except when comparing H. johannensis with H. megapotamenis, H. macronectar with H. megapotamensis and H. macronectar with H. montevidensis (Table 1). The rostellum is trilobed, and the anther canals are adnate to the lateral lobes (rostellar arms) so that a pad-like viscidium is placed at the end of each rostellar arm (Fig. 1B–E). In H. johannensis, the median rostellar lobe is keel-like and anteriorly projecting (Fig. 1B). During floral ontogenesis, the anther undergoes a division that ultimately results in the formation of two separate pollinaria concealed in their respective anther sacs. Each pollinarium is placed alongside a lateral rostellar lobe (Fig. 1B–E). In all studied species, the pollinarium consists of massulate pollinia, an arm-like, hyaline caudicle and a pad-like, terminal viscidium. Pollinarium length ranged from 3 to 15 mm (Table 1). In H. johannensis and H. macronectar, the rostellar arms are long and arched; with both viscidia almost adjacent and upward pointing (Fig. 1B–C). In H. megapotamensis, the rostellar arms are short and curved inwards, so that the spur entrance is closely flanked by the viscidium of each pollinarium (Fig. 1D). In H. montevidensis, the rostellar arms are straight and slightly divergent (Fig. 1E) and the distance between the apices of the rostellar arms is c. 3 mm. In all studied species, the two stigmatic surfaces are placed below the pollen sacs, surrounding the spur entrance (Fig. 1C–E). In H. johannensis and H. macronectar, the stigmatic surfaces are long-stalked and involute (Fig. 1B–C), whereas those of H. megapotamensis and H. montevidensis are short and slightly convex (Fig. 1D–E). In H. megapotamensis, a tooth-like process partially occludes the spur entrance (Fig. 1D). This process is formed by the protruding, erect projection of the apex of the inner margin of the stigmatic lobes, and this is located in front of the spur entrance, thereby dividing the entrance of the spur into two (Fig. 1D). The flowers of H. johannensis, H. macronectar and H. megapotamensis emit a sweet fragrance, the secretion of which is perceptible after 18:00–19:00 h. The flowers of H. montevidensis produce a faint, sweet fragrance throughout the whole day. This fragrance is best perceived by enclosing the flowers in a vial.

Nectar volume in intact flowers of the same species is subject to remarkable variation and mean values ranged from 1.27 to 35 μL, depending on the species (Table 1). Mean nectar volumes showed significant differences between the studied species (H = 34.88, P < 0.0001, d.f. = 3; see Table 1), except when comparing H. johannensis with H. macronectar and H. macronectar with H. megapotamensis. In intact flowers of all species, the nectar forms a conspicuous column. Except for H. montevidensis, the height of this nectar column shows remarkable variation in intact flowers of the same species (Table 1). Mean values of nectar column height ranged from 0.52 to 3.32 cm, depending on the species (Table 1). Nectar column heights showed statistically significant differences (H = 26.36, P < 0.0001, d.f. = 3; see Table 1), except when comparing H. macronectar with H. megapotamensis. Mean values of nectar concentration ranged from 18.12 to 26.20% (Table 1). Significant differences were found in nectar concentration (H = 15.46, P < 0.0004, d.f. = 2; see Table 1), except when comparing H. macronectar with H. megapotamensis.

Pollination mechanism

Pollinarium withdrawal takes place when insects insert the proboscis into the spur and press the surface of the viscidia with a scaleless, smooth body part (e.g. proboscis, ventral region between the palpi or the eyes, depending on the species) (Fig. 3), thus dislodging the pollinaria on leaving the flower. Pollination takes place when a pollinarium-laden insect visits a flower; the pollinia contact the stigmatic surface and massulae are left behind (Fig. 3B, E, G). In H. johannensis, H. macronectar and H. montevidensis, the insects are able to insert the proboscis along a straight course. In H. johannensis, pollinator movements are restricted by the projected, keel-like median rostellar lobe (Fig. 1B), and the pollinators are mechanically guided against the upward facing viscidia (Fig. 3A). In H. megapotamensis, two tooth-like stigmatic processes partially block the spur entrance (Fig. 1D). Therefore, the insect has to insert its proboscis laterally, thus pressing it against the viscidia that are orientated toward the spur entrance (Fig. 3C).

Pollinators, pollinator features and behaviour

All species investigated are pollinated by Lepidoptera (Figs 2-4), a fact which is consistent with overall floral morphology. Most insects recorded bearing five or more pollinaria had proboscides that were shorter (see Table 4) than the mean spur length of the orchids that they pollinated (see Table 1). However, a larger pollinator sample is necessary to address this phenomenon adequately.

During our observations, H. johannensis, H. macronectar and H. megapotamensis were pollinated solely by crepuscular–nocturnal hawkmoths (Sphingidae) that were recorded at the flowers from 19:00 to 21:35 h (Figs 2, 4A–B, Table 4). Diurnal visitors or pollinators were never recorded at these three species. Our fieldwork observations and our photographic and film record showed that all Sphingidae displayed similar behavioural patterns. On approaching, the hawkmoth hovers in front the flower and inserts its proboscis into the spur (Fig. 3A, D). In H. johannensis and H. macronectar, the moths use their forelegs to grasp the perianth (Fig. 2A, C), but this behaviour was not observed for H. megapotamensis. During our observations, H. johannensis was pollinated by Manduca rustica (proboscis length: 13.5 cm, n = 1) and M. sexta (proboscis length: 10.3 cm, n = 1) (Fig. 2A). These moths carried pollinaria of H. johannensis (15 and 13, respectively, Table 4) attached to the underside of the base of the proboscis (Fig. 2A, B). Because of weather conditions, the behaviour of the first moth species could not be followed in detail. However, the photographic record suggests that it behaves much like M. sexta. This latter pollinator was seen moving between two neighbouring inflorescences, spending around 25 s at the flowers. Four additional hawkmoth visits were recorded during the observation period, but environmental conditions and the long distance that separated us from the visited inflorescences precluded the unequivocal taxonomic identification of these insects. Also during our observations, Eumorpha labruscae (proboscis length: 5.9 cm, n = 1) (Fig. 2D) and M. cf. lucetius (proboscis length: 6.4–7.4 cm, n = 3) (Fig. 2C) were recorded as the pollinators of H. macronectar. Both moth species carried pollinaria (up to 12 and 3, respectively, see Table 4) attached between the palpi. The first moth species was sighted only once, whereas the second was sighted 13 times. On five of these occasions, the moths carried pollinaria of H. macronectar, and on two occasions were also bearing pollinaria of H. megapotamensis (that was flowering nearby) attached to the base and sides of the proboscis. It was possible to film some individuals of M. cf. lucetius visiting the flowers of H. macronectar (see also Supporting Information, Video S1), and our video record indicates that moths spent between 2 and 17 s probing individual flowers and between 12 and 140 s at each inflorescence. Remarkably, videos revealed that the same moth inserted its proboscis very differently into flowers on the same inflorescence (inserting or not the full length of the proboscis, see also Supporting Information, Video S1).The sphingid Agrius cingulatus (proboscis length: 9–12.8 cm, according to Singer & Cocucci, 1997) was photographed at the flowers. This insect never removed pollinaria, possibly because its proboscis is considerably longer than the spur (see also Supporting Information, Fig. S2). The only pollinators recorded for H. megapotamensis were moths of M. cf. lucetius (Fig. 4A, B, see also Supplorting Information, Video S2). Seven visitation events were recorded and insects were observed carrying pollinaria (up to 15, see Table 4) fixed to the sides and underside of the proboscis base. As stated above, this moth was also observed pollinating flowers of nearby H. macronectar. They visit several flowers per inflorescence, usually starting at the base and proceeding towards the top of the inflorescence. During our observations, the pollinators spent 2–9 s at each flower and 30–125 s at each inflorescence. Also, during our observations, H. montevidensis was pollinated solely by butterflies of the family Hesperiidae (Fig. 4C, D, Table 4). Generally, pollinator activity was observed from 06:00 to 18:00 h. All pollinator species behaved in a similar manner. Insects landed at the base of the inflorescence and proceeded towards the apex, systematically visiting the flowers. Column structure allows pollinators to place their heads between the rostellar arms (Fig. 3F) and all recorded pollinators bore the pollinaria on their eyes (Figs 3G, 4C, D). Cumbre sp. and Vehilius clavicula (proboscis length of both: 13 mm, n = 1) were sighted only once, bearing two and three pollinaria, respectively. Butterflies of Urbanus teleus (proboscis length: 11 mm, n = 1) and U. zagorus (proboscis length: 15 mm, n = 1) were more frequent, bore more pollinaria (up to 6 and 14, respectively, see Table 4) and are considered to be the main pollinators of H. montevidensis in the study area (Fig. 4C, D, see also Supporting Information, Video S3). In total, Urbanus spp. were seen c. 167 times visiting inflorescences. These insects visited one to four inflorescences per visit to the population, spending between 1 and 11 s at each flower and between 14 and 80 s at each inflorescence. These skippers were occasionally recorded actively attempting to remove the pollinaria attached to their eyes with their forelegs. Butterflies of Hesperocharis erota (Pieriidae) and Phocides pialia (Hesperiidae) were photographed and filmed visiting inflorescences, but without dislodging pollinaria (see also Supporting Information, Fig. S2). Although measurements could not be made in these two species, the photographic and film record clearly indicates that the eyes of both insect species do not come into contact with the orchid viscidia. Four visits of the nocturnal hawkmoth Xylophanes fosteri (Sphingidae, proboscis length: 22 mm, n = 1) (see also Supporting Information, Fig. S2 and Video S3) were recorded. This moth consistently visited flowers without removing pollinaria, as would be expected from an insect with a proboscis significantly longer than the nectariferous spur. Furthermore, intact, marked flowers used to ascertain the possibility of nocturnal pollination (see Material and Methods), were still intact the following morning.

Breeding systems

All species studied (Table 3) are self-compatible, but pollinator dependent. No fruit were formed by emasculated or intact flowers, strongly indicating that all species studied require the agency of a pollinator in order to set fruit. All species showed high fruiting success, either by self-pollination or cross-pollination, and there was no significant difference between these two treatments (Table 3).

Fruiting success and pollination efficiency

All species investigated displayed high fruiting successes in both sampling years, ranging from 69.48 to 97.40% (Table 4) (fruit set in H. montevidensis could only be determined for 2011, see Material and Methods). Mean fruit set per inflorescence was also high in both years, ranging from 68.62 to 97.38%. When comparing fruit set per inflorescence for 2010 and 2011, significant differences were found in H. johannensis (F = 41.464, P = 0.0113, d.f. = 71, Table 4) and H. macronectar (F = 15.436, P = 0.0027, d.f. = 64, Table 4). Habenaria megapotamensis was the only species that did not show a significant difference (F = 10.377, P = 0.1175, d.f. = 48, Table 4).

All studied species displayed male efficiency factors slightly higher than 1 (Table 4); i.e. at least one flower was pollinated for each flower acting as pollen donor. Percentages of pollinated flowers ranged from 72.38 to 86.06% and the percentages of flowers acting as pollen donors were also high, ranging from 63.81 to 83.55% (Table 4). However, several flowers used to calculate male efficiency factors contributed only one pollinarium. In H. johannensis, 23.97% of the flowers acting as pollen donors contributed a single pollinarium. In H. macronectar, H. megapotamensis and H. montevidensis, this value reached 38.8, 21.76 and 38.17% of the flowers acting as pollen donors, respectively.

Concluding remarks

The present study shows that four sympatric southern Brazilian Habenaria spp., the flowering periods of which overlap, are isolated (in terms of pollination) by a set of factors that includes differences in floral morphology (especially spur length and column morphology) and different pollinators and/or different pollinarium placement on the pollinator. Species with similar flower features (e.g. H. macronectar and H. megapotamensis) can share pollinators, but these carry the pollinaria of the different orchid species on different body parts, thereby physically preventing hybridization.

Long-spurred, hawkmoth-pollinated orchids have captivated the attention of scientists ever since Darwin (1862) made his famous comments on the putative pollinators of Angraecum sesquipedale Thouars and proposed a co-evolutionary race between plant and pollinator. Several Neotropical Habenaria spp. rank among the longest spurred orchids in the Americas, with (for instance) all the species of section Macroceratitae having long spurs (≥ 6 cm in length), and that of H. longicauda reaching 25 cm (Renz, 1992; Batista et al., 2006). Therefore, these orchids parallel the mainly African–Madagascan Angraecinae, at least in terms of floral morphology (see Micheneau, Johnson & Fay, 2010). Thus, it is tempting to ask whether the long-spurred Habenaria spp. and their sphingid pollinators co-evolved. Although the present contribution did not set out to address that question, nevertheless some of the results presented here suggest that a co-evolutionary relationship (in its strictest sense, see Ridley, 2004) is unlikely as: (1) pollination by Sphingidae occurs in more than one Habenaria section/species group; and (2) one studied species (H. macronectar) is pollinated by phylogenetically unrelated moths. Well-supported phylogenies are already available for Sphingidae, and these studies indicate that the observed pollinators of H. macronectar are distantly related and belong to different clades (see cladograms in Kawahara et al., 2009). A comprehensive molecular phylogeny of Neotropical Habenaria is currently being prepared (J. A. N. Batista, pers. comm.), and this will be the basis for a more robust and reliable infrageneric classification, prompting a re-evaluation of the characters that are normally used to separate species groups. Furthermore, it could be used as a framework to elucidate the evolution of pollination-related features (e.g. long spurs). Only then will it be possible to ascertain whether the long-spurred Habenaria spp. have a co-evolutionary relationship with their sphingid pollinators.