Breeding biology and distyly in Palicourea rigida H. B. & K. (Rubiaceae) in the Cerrados of Central Brazil

(Reproductive biology and distyly in Palicourea rigida H. B. & K. (Rubiaceae) in the Cerrados of Central Brazil). Distyly is a fl oral polymorphism more common among the Rubiaceae than in any other angiosperm group. Palicourea rigida is a typically distylous species of the Rubiaceae widely distributed in the Brazilian Cerrados. This work aimed to study the fl oral biology and breeding system of P. rigida in order to verify if there was asymmetry between fl oral morphs. The work was carried out at Fazenda Água Limpa, Brasília-DF, from 1993 to 1995; and at Serra Caldas Novas State Park-Goias and in Clube Caça e Pesca Itororó de Uberlândia-Minas Gerais in 2005 and 2006. Density, height and pin/thrum ratio were assessed for fl owering individuals in all areas. Plants were investigated for differences in fl oral morphology, nectar production, reproductive success and site of self incompatibility reactions. Blooming period was long and concentrated during the rains. Flowers were clearly distylous and with reciprocal herkogamy. They produced nectar and lasted for a single day. In spite of differences in density and height, populations were mostly isoplethic. Nectar production varied in volume and concentration but the differences could not be associated with fl oral morphs. The species is self-incompatible but reproductive success was always high and independent of fl oral morphs. There were differences in the site of incompatibility barriers between fl oral morphs, which were similar to those observed for other Rubiaceae. The main fl oral visitors and pollinators were the hummingbirds Colibri serrirostris and Eupetomena macroura. High fruit-set indicates that the pollinators transported enough compatible pollen grains between fl oral morphs, despite their territorial behavior.


Introduction
Distyly is found in ca.27 angiosperm families and it is understood as a mechanism to promote pollination and cross fertilization (Darwin 1877; Barrett et al. 2000b;Li & Johnston 2001).Distylous plants present individuals with fl owers with long styles and short stamens, the pin morph, and individuals with fl owers with short styles and long stamens, the thrum morph (Ganders 1979).The ratio between morphs in populations is commonly 1:1, an equilibrium denominated isoplethy (Ganders 1979;Sobrevila et al. 1983).Together with the fl oral dimorphism, distylous species usually have a diallelic sporophytic self-incompatibility system, in such a way that fruit set arises only from pollination between fl oral morphs (Barret & Richards 1990).Genetic control of the self-incompatibility mechanism and fl oral dimorphism seems to be determined by a heteroallelic "supergene", comprising a tightly linked series of loci (Ganders 1979;Gibbs 1990).
Besides the position of reproductive structure, the fl oral morphs may also exhibit differences in stigma morphology, pollen grains, corolla size and nectar production (Johnston 1952;Levin 1968, Ganders 1976;Ornduff 1976;Ganders 1979;Barrett 1992;Dulberger 1992;Castro et al. 2004;Castro & Oliveira 2002;Ornelas et al. 2004;Teixeira & Machado 2004a).Little is yet known of the origin and selective pressures that lead to these ancillary differences between fl oral morphs, but it is believed that they also affect the pollen fl ow between morphs (Feinsinger & Busby 1987;Stone 1995;Lau & Bosque 2003).
The family Rubiaceae has the largest number of distylous genera and species among the families for which the polymorphism has been reported (Ganders 1979;Barrett & Richards 1990).It is the fourth largest family of the angiosperms, with ca.650 genera and 12,000 species of Pantropical distribution, and the largest family in the Neotropics, with some 200 genera and 1500 species (Delprete 2004).It is also one of the main families of the Brazilian fl ora, with ca.130 genera and 1500 species (Souza & Lorenzi 2005).In the Cerrado Biome, the Neotropical Savannas in Central Brazil, it is considered the seventh most species-rich family, with at least 376 species (Mendonça et al. 2008), among which ca. 41% are distylous species (Consolaro 2008).Palicourea has ca.200 species and distyly is a very common, basal feature of the genus (Vuilleumier 1967;Taylor 1997).Its odorless fl owers have a bright colored tubular corolla and are mainly pollinated by hummingbirds (Taylor 1997).Palicourea rigida H. B. & K. is a widely distributed species, occurring in more than 37% of the cerrado areas studied so far (Ribeiro & Walter 2008;Ratter et al. 2003), possibly the most common Palicourea species of the Brazilian cerrados.
Cerrado is the second largest biome in Brazil, a complex of plant formations with ecological and physiognomic similarities with other savanna areas in the Neotropics, Africa and Australia, with forest, open savanna and grassland areas (Ribeiro & Walter 2008).Its flora is characteristic and differentiated, although it shares some genera with surrounding biomes, and is directly or indirectly affected by climate, soil and anthropic disturbance, especially fi re frequency (Ribeiro & Walter 2008).The effect of these factors appears in the structure, spatial distribution and fl oristic composition (Oliveira 2008;Ribeiro & Walter 2008).
Recent works on the Cerrado biome have shown variation and anomalies in the features of distylous plant populations (Coelho & Barbosa 2003;Consolaro et al. 2005;Toledo et al. 2007;Consolaro et al. 2009).In this sense, our work aimed to characterize the fl oral biology of P. rigida and to study different populations comparing morph distribution and ancillary features associated with distyly.

Materials and methods
Period and study areas -The data were collected during the fl owering season between 1993and 1995(PI), and between 2005and 2007 (PII).The study during PI was carried out at the Água Limpa Farm (FAL) (15°57'S; 47°55'W), Brasília, DF, a research area of 4,500 ha belonging to the Universidade de Brasília.The second study period (PII) took place at Serra de Caldas Novas State Park (PESCAN) and in the Clube de Caça e Pesca Itororó de Uberlândia (CCPIU).PESCAN (17º43'S; 48º40'W) comprises 12,500 ha in Caldas Novas, Goias and Rio Quente, Goias and CCPIU (18º55'S; 48º17'W) is a private 640-hectare reserve southwest of Uberlândia, Minas Gerais.The areas are dominated by cerrado plant formations.Species studied -Palicourea rigida is a shrub or treelet with characteristic thick bark deeply cleft into rectangular blocks.It is commonly known as "batecaixa" and "gritadeira" due to its rigid, coriaceous leaves (Silva-Júnior 2005).It has terminal thyrsoid infl orescences with zygomorphic fl owers (Robbrecht 1988).It is widely distributed in the savannas of South America (Taylor 1997), and frequent in the open cerrado and cerradões (denser woodlands) in the states of Bahia, Goiás, Minas Gerais, Mato Grosso, Mato Grosso do Sul, Paraná, São Paulo, Tocantins and Distrito Federal (Silva-Júnior 2005).Floral biology -Anthesis, fl ower life-span, stigmatic receptivity and pollen release were observed in the fi eld (a minimum of 20 fl owers each study period) and in branches brought to the laboratory and kept in vases with water.Stigma receptivity was tested with 3% hydrogen peroxide (Kearns & Inouye 1993).
Phenology -Quantitative phenodynamic data was collected only for fl owering during PI.The beginning, peak (more than fi fty percent of open fl owers) and end of the fl owering period were recorded for 50 individuals.More general phenological observations were done during PII in order to determine if there were marked differences between populations and study periods (69 individuals at PESCAN and 100 at CCPIU).Population features and isoplethy -In order to describe basic population structure and to estimate isoplethy, blooming plants were sampled and marked (50 plants at FAL, 69 at PESCAN and 100 at CCPIU, the same plants used for phenological observations).The number of infl orescences per plant was counted for the FAL and CCPIU populations, while height and density were estimated for the PESCAN and CCPIU populations only.Density was estimated using the nearest neighbor method (Brower & Zar 1984), with the distance between individuals obtained by GPS coordinates.Density was later estimated for the FAL population using the individuals mapped in a 0.3328 ha area.The fl oral morph of each individual was determined and the morph ratio was calculated for each population.Isoplethy (1:1 morph ratio equilibrium) was tested in each case.Floral morphometry and nectar -Some fl owers were collected at random and fi xed in 70% ethanol for measurements in the laboratory (20 fl owers for each fl oral morph, from at least 5 individuals of each morph) and other fl owers were measured directly in the fi eld (28 pin fl owers and 14 thrum fl owers in PI).All fl ower measurements were carried out using a digital pachymeter (error 0.01 mm), under stereomicroscope whenever necessary.The following structures were measured: corolla length and diameter, stamen and anther length, and style and stigma length.
Pollen size and morphology were investigated for each morph using both light and scanning electron microscope (SEM).The latter was used only in PI (20 pollen grains from thrum and 24 from pin fl owers) to detail pollen morphology.The material for SEM observations was prepared using anthers from fresh fl owers directly fi xed in Osmium tetroxide vapor during 24 hours (Smith & Tiedt 1991).SEM analyses were used only for pollen measurements.The presence of lipids on the pollen grains was tested using Sudan IV (Johansen 1940).
The amount of pollen was estimated by counting the total number of pollen grains in one anther per fl ower.The anthers were extracted from fl oral buds the day before anthesis and fi xed in 70% alcohol during the PI only (10 fl oral buds for each morph).The pollen/ovule ratio (sensu Cruden 1977) was calculated from these estimates.Pollen viability was estimated from fl owers of different periods after anthesis by using Alexander's stain (modifi ed as in Dafni 1992).
Nectar collection was done at the end of the morning in previously bagged fresh fl owers.The volume was measured using glass micropipettes or microseringes and concentration was determined using a hand refractometer.Five to seven fl owers from each morph were used during PI, and 50 thrum fl owers and 43 pin fl owers were used during PII, three to fi ve fl owers from 10 individuals of each morph.Breeding system -Controlled hand pollination tests were carried out in order to determine the breeding system using plants at FAL during PI.The index of self-incompatibility (ISI) was calculated by using fruit-set results from those tests, as the fruit-set ratio between self-and between morph crosspollination tests (sensu Bullock 1985).Flowers on the day before anthesis or freshly open were marked with threads of different colors to identify treatments, and were then isolated using nylon mesh bags.The tests followed the same protocol adopted for breeding system studies in other distylous species of Rubiaceae (Bawa & Beach 1983): hand self-pollination (fl ower pollinated with its own pollen) or cross-pollination (fl ower pollinated with pollen from other plants, either from the same morph -intramorph or illegitimate pollination -or from the other morph -between morph or legitimate pollination).Some fl owers were left untouched in order to test spontaneous self-pollination (fl owers bagged before anthesis).
At CCPIU (during PII), legitimate and illegitimate cross pollination was done using previously bagged fl owers in order to observe pollen-tube growth.Hand pollinated fl owers were collected six hours after pollination and fi xed in 70% alcohol.Five pistils of each treatment and morph were analyzed (from five different plants) using fluorescence microscopy technique (Martin 1959).
In order to estimate the efficiency of natural pollination, 67 inflorescences of thrum individuals (12,646 flowers) and 91 of pin individuals (13,658 fl owers) were marked in the fi rst fl owering season in PI (1993) and 61 infl orescences of thrum individuals (10,353 fl owers) and 86 of pin individuals (10,243 fl owers) during the second fl owering season in PI (1994).During PII, 120 fl oral buds from each morph were marked at PESCAN and 60 at CCPIU (fi ve fl owers per infl orescence and two to four infl orescences per individual).Also during PII, 12 infl orescences for each morph were analyzed in order to count fruits developed after natural pollination (a total of 2,485 fl owers for thrum and 3,252 for pin at PESCAN, and 2,247 fl owers for thrum and 2,319 for pin at CCPIU).This analysis could be used for P. rigida because the ovary and nectary disc of non-fertilized fl owers persist up to infl orescence abscission or leave scars that can be used to count fl owers.Developed fruits and non-fertilized fl owers were counted 30 or 40 days after fl owering (fruit development can be assessed ca. 10 days after anthesis).
In PII, pollen tube growth under fl uorescence microscopy (Martin 1959) was also used to estimate natural pollination success.Flowers from both morphs were collected and fi xed in 70% alcohol at the end of anthesis.The amount of pollen grains germinating on the stigma and pollen tubes effectively growing along the style were quantifi ed (30 fl owers for each morph in each population, three to fi ve fl owers per individual).Flower visitors -The pollinators were determined according to fl ower architecture, morphology and behavior of the visitors for each study period and population.The main pollinators were the most frequent pollinators during the study period, appearing at most of the observation sessions and throughout the fl owering season.Additional pollinators were less frequent, appearing in less than half of the observation sessions and sometimes only during part of the fl owering season.As much as possible, visitors were photographed and/or captured for identifi cation by specialists.Although fl owers were visited also by insects of various groups, observations were restricted to hummingbirds and bees.Statistical analysis -Morphometric comparisons between fl owers were done using one-factor ANOVA (morph).They were done using ranked values whenever the data did not reach normality.Student t-test was used to compare pollen size and nectar measurements between fl oral morphs.Fisher's exact test for chi-square approximation was applied to test isoplethy between fl oral morphs in each population, verify differences in reproductive success between fl ower morphs, and natural fruit set in marked infl orescences between populations.All analyses were done using the program Statistica 6.0 (StatSoft Inc. 2001).

Results
Floral biology -Flowers of Palicourea rigida are presented in dense terminal infl orescences, pedunculate cymose thyrses arising between a pair of leaves.The development from bud primordia to anthesis of the fi rst fl owers took ca.60 days.The fl owers are zygomorphous, dichlamydeous and hermaphrodite.The ovary is inferior, mostly bilocular with a basally fi xed ovule at each locus.Above the ovary there is a conspicuous nectary disc which surrounds the base of the style.The stigma is of the dry papillate type (sensu Heslop-Harrison & Shivanna 1977), usually bifi d, but more rarely trifi d.The form and si ze of the stigma differed markedly between the morphs.The androecium is formed by fi ve stamens adnate to the corolla by two thirds of their length.Anthers are long, dorsifi xed, bitecae, introrse with longitudinal dehiscence.The hairy hypanthium has fi ve discrete lacinea.The corolla is tubular with a gibbous base and an internal ring of longer hairs which somewhat protects access to the nectar.The color of the corolla varies from yellow to reddish.
Anthesis began before dawn, between 3 and 5 o'clock.Flower duration was around 24 hours, with most fl owers wilting, darkening and falling off the morning of the second day.In some thrum fl owers anther dehiscence occurs before fl ower opening, leaving the stigma, still unreceptive, covered with pollen.In pin fl owers the stigma is receptive before anthesis and the pollen is released a bit later.The test with H 2 O 2 indicated that receptivity continues throughout the day.Sometimes it was possible to observe a slight contraction of the stigmatic lobes after pollen deposition.Phenology -During the study, the populations presented a single annual fl owering peak in the rainy season (Figure 1), but it was common to observe extemporaneous fl owering in other periods of the year (Silva 1995).The fl owering period in Brasília (PI) in 1993 began at the end of September, peaked in mid-November and fi nished at the end of January 1994.The next blooming season began in mid-September 1994, peaked again in November and continued until March 1995.There were no marked differences between morphs in phenodynamics but there were differences between years in PI (Figure 1).In the study areas during PII, both in 2005 and 2006, fl owering at PESCAN began in September while at CCPIU it began only in November.In all cases fl owering ended in March of the following year.Population characteristics and isoplethy -The density of reproductive individuals of Palicourea rigida during PI was 150.24 individuals/ha and in PII was 119 individuals.ha -1 for PESCAN and 182.61 individuals.ha - for CCPIU.The height of the individuals, estimated without distinction between morphs, was 2.12 ± 0.52 m for PESCAN and 1.14 ± 0.27 m for CCPIU.Reproductive effort, estimated as the number of infl orescences per individual, did not vary between morphs.In the census of the FAL population (PI), thrum individuals produced an average of 8.00 ± 7.00 infl orescences and signifi cantly more than the pin individuals which produced 6.00 ± 4.00 infl orescences (p<0.05).At CCPIU (PII), thrum individuals produced an average of 4.82 ± 3.28 infl orescences while the pin ones produced 5.77 ± 5.75 infl orescences (U= 1181.0;P= 0.65).
Both in the PI and PII, the height of the stigma was statistically larger in the pin flower and the height of the anthers and stigmatic lobe length were larger in the thrum fl owers (Figures 2, 3 and 4), characterizing distyly.Measurements indicated reciprocal herkogamy (Table 1 and Figure 2) but we did not test it directly.The length of the anthers did not differ between morphs in the FAL (PI), but it was larger in thrum fl owers both at PESCAN (U= 74.00; P< 0.001) and CCPIU (U= 76.00; P< 0.001) (Table 1).
Nectar production began at fl ower opening and continued during the day, accumulating in the gibbous base of the corolla but could reach most of the length of the corolla tube.
The effi ciency of fruit-set after natural pollination varied widely during the PI.In 1993, thrum plants of the FAL population presented fruit-set of 22.20% and the pin plants 14.70%.In 1994, thrum plants had a fruit-set of 30.04% while pin plants had 39.90%.During PII, the analysis of fruit set from marked fl oral buds did not reveal signifi cant differences between morphs.At PESCAN, thrum plants had 41.70% fruit-set and pin plants 31.80%(t= 1.56; P> 0.05).At CCPIU, fruit-set was 52.33% for thrum plants and 56.66% for pin plants (t= 0.37; P> 0.05).
Pollen tube growth analysis after natural pollination in PII showed large pollen load in most pistils and in all of them we observed germinating pollen grains and pollen tubes growing along the style.Thrum flowers presented larger amounts of pollen tubes than pin flowers in both populations and only some pin flowers at CCPIU presented pistils with less than 10 pollen grains germinating.Thrum flowers at PESCAN presented 86.70% of the analyzed pistils with more than 50 germinating pollen grains while pin flowers presented more than 40% of the pistils with that amount of pollen grains.At CCPIU, 96.70% of thrum flowers had more than 50 pollen grains germinating on the stigma while pin flowers presented more than 30% of the stigmas with that amount of pollen grains.
Pistils of thrum flowers after self-pollination and illegitimate cross-pollination presented pollen tubes which appeared fragile (enlarged and curved) and grew relatively short into the stigma (Figure 5 and 6).In self-pollinated pistils of pin fl owers, pollen tubes presented bright fl uorescence, many callose plugs, enlarged portions and apparent branching (Figure 7), and larger diameter than the tubes observed in legitimate cross-pollinated pistils.These self pollen tubes grew along the style and seemed to be blocked at the base, before reaching the ovary (Figure 8), forming callose plugs (Figure 9).The pollen tubes formed in illegitimate crosspollination pistils were also observed along the style but did not reach the ovary.After legitimate pollination in both morphs, pollen tubes grew without problems and were observed in the ovary (Figures 10 and 11).Floral visitors -P.rigida flowers were visited by 11 hummingbird species and 21 species of insects of different orders (table 3).In FAL, Colibri serrirostris (Figure 12) was the most frequent fl ower visitor and was considered the main pollinator.It showed clear territorial behavior, although it visited individuals of different morphs in the defended territory.C. serrirostris collected nectar sequentially in every open fl ower throughout the day (6:00 -18:00 h) at intervals that varied between 10 and 30 minutes (longer intervals during the warmer parts of the day).A similar situation was observed at PESCAN and CCPIU where Eupetomena macroura (Figure 13) was the most frequent visitor and, despite its territorial behavior, visited plants of different morphs and was considered the main pollinator.
Another nine species of hummingbirds visited P. rigida fl owers and appeared, in some cases, to have well defi ned foraging routes.However, it was not possible to characterize clearly this foraging pattern.These hummingbirds were less frequent and visited quickly, though legitimately, some fl owers per bout, being considered additional pollinators.
Bees were observed constantly on P. rigida fl owers (Table 3), collecting nectar or pollen but were not always legitimate pollinators.They were considered additional pollinators only when their size and behavior made possible the contact with anthers and stigma.In many cases, fl oral resources, especially nectar, were collected with the bee inserting only part of its body or the proboscis inside the corolla tube, and eventually touching the stigma or stamens.However, some species pillaged nectar by perforating the corolla base or using apertures made by other visitors.Analysis of the collected specimens showed very small amounts of pollen on the body, except on Bombus morio workers, which had a lot of pollen adhered to the hairs on the thorax and accumulated on the corbiculae.

Discussion
The P. rigida populations studied differed in some structural aspects, such as density and height of individuals, but the values are similar to the values described for populations in other Neotropical savannas, as the 187 individuals/ha found in a population in Venezuela (Wütherich et al. 2001).They presented density much higher than the 34 individuals/ha cited as an average density for P. rigida (Silva-Júnior 2005), probably because this average is based on phytosociological surveys limited to individuals over 10cm dbh.Despite the differences in density and size, the populations studied did not present any distyly anomaly that could be associated with these structural differences.Neither were these structural differences related clearly to conservation status or history of the study areas.
The fl owering pattern for the populations studied is of the annual extended type, since there was a single main fl owering event each year with over fi ve months duration (Newstrom et al. 1994).Actually, in 1993 the fl owering period was less than fi ve months and the pattern in this case would be classifi ed as annual intermediate.However, since the difference between fl owering pattern types is less than a month and during the other years the period was longer, species fl owering phenology may be better described as annual extended.Observed differences between fl owering peaks in PI and small differences in the beginning and duration of fl owering periods are common in tropical environment and especially in a widely distributed species as P. rigida.
The occurrence of extemporaneous fl owering, as in PI, is characteristic of subannual fl owering (Newstrom et al. 1994) and would explain the presence of herbarium sheets with fl owering specimens most of the year (Silva 1995).This type of fl owering strategy may benefi t the plants, since fl owers and fruits will be available most of the year.This has been cited for other Cerrado plants (Oliveira & Gibbs 2000).However, these phenological asynchronies should be treated with caution since they do not occur every year and are not well understood yet (Newstrom et al. 1994).
The isoplethy found in the populations studied during PII contrasted with the anisoplethy recorded for the FAL population.But it should be noted that the number of individuals in that area was much smaller than those surveyed at PESCAN and CCPIU, and the anisoplethy may be result of this limited sample.In another survey carried out in the Distrito Federal, in an area near to the FAL, isoplethic populations were recorded for different distylous Rubiaceae, including P. rigida (H.N.Consolaro pers.obs.).Smaller populations of Hottonia palustris, a distylous European species, tended to present greater deviation from the 1:1 morph ratio than larger populations (Brys et al. 2007).
Studies with Palicourea padifolia (Ree 1997;Ornelas et al. 2004) and P. fendleri (Sobrevila et al. 1983) showed isoplethic populations.In Brazil, specifically in an Atlantic forest remnant, a P. longepedunculata population was isoplethic while P. marcgravii was anisoplethic (Pereira et al. 2006).In the Cerrado, P. coriacea and P. macrobotrys presented deviation from the equilibrium morph ratio, while P. offi cinalis and P. marcgravii were isoplethic (Consolaro et al. 2009).Apparently, the primary mechanism maintaining equilibrium morph ratio is density dependent, favored in populations where there are higher numbers of legitimate cross-pollinations than illegitimate ones (Barrett et al. 2000a).
Anisoplethy may be also caused by vegetative multiplication and loss of intramorph self-incompatibility (Barrett 1992).Breakdown of self-incompatibility and rising and dominance of homostylous individuals have been reported for Rubiaceae species in forest habitats inside the Cerrado biome and may be associated with disturbance and habitat fragmentation (Coelho & Barbosa 2003;Consolaro et al. 2009).Initial stages of functional distyly loss seem to be associated with changes in reciprocal herkogamy (Vallejo-Marín & Barrett 2009), especially in the genus Palicourea (Consolaro et al. 2009).But the reciprocal herkogamy seems to be present in the studied populations of P. rigida, depite the ancillary differences between fl ower morphs.Larger structures in the thrum fl owers, as stigmatic lobes and anthers, are common in distylous species, especially in the Rubiaceae (Baker 1956;Ornduff 1980, Sobrevila et al. 1983;Murray 1990;Dulberger 1992;Richards & Koptur 1993;Castro & Araújo 2004;Castro et al. 2004;Coelho & Barbosa 2004;Rossi et al. 2005;Pereira et al. 2006).However, distylous species in other families present larger stigmatic lobes in pin fl owers (Ornduff 1971;Dulberger 1992).These ancillary differences have been associated with the effi ciency of legitimate pollinations.Pollen grains and stigmatic surface dimorphism would reduce or avoid illegitimate cross pollination and its reproductive consequences (Vuilleumier 1967;Ganders 1979;Barrett 1992).However, these trends on ancillary fl oral features cannot be generalized since, even for the Rubiaceae, there are distylous species without either pollen or stigmatic lobe dimorphism (Castro et al. 2004;Pereira et al. 2006).
P. rigida did not present marked pollen dimorphism and its pollen is described as monomorphic in palynological studies for the region (Salgado-Laboriau 1971).Despite these clear differences, morphometric analysis indicated signifi cant differences in size between thrum and pin pollen.Pollen dimorphism is one of the most common features associated with heterostyly (Dulberger 1992), together with differences on the amount of pollen produced (McKenna 1992;Barros 1998).But distylous Rubiaceae present an enormous variation in amplitude of pollen dimorphism and in pollen grain features (Dessein et al. 2005), with examples of very similar pollen, as in Palicourea lasiorrachis (Feinsinger & Busby 1987).
The amount of pollen in each P. rigida fl ower is directed to cross pollination since the fl ower produces only two ovules.In order to secure fruit formation, some two to six pollen grains would be needed germinating on the stigma (Cruden 1977).So, the amount of pollen grains produced in a P. rigida, fl ower would be enough to pollinate 500 other fl owers.
There was no clear relation either between nectar production and fl oral morphs, contrasting with the observed for other species of Palicourea (Consolaro et al. 2009).Pin fl owers of Psychotria barbifl ora presented a greater volume of nectar and lower concentrations than thrum fl owers (Teixeira & Machado 2004a).Larger amounts of nectar in pin fl owers were also observed for a population of Palicourea padifolia (Ornelas et al. 2004).P. rigida nectar sugar content varied between 19 and 22%, concentrations similar to those found in bat-pollinated fl owers (18.9%) and hawkmoth fl owers (22.1%) (Pyke & Waser 1981).For hummingbirds, average nectar sugar concentration is commonly a bit higher (25.4%) but much lower than the concentration preferred by bees (41.6%).The volume of nectar in P. rigida fl owers was smaller than the observed in other typical hummingbird fl owers (Baker 1975;Buzato et al. 2000).Smaller amounts of nectar may keep pollinators moving, obliging more visits to satisfy their energy requirements, and increasing pollen fl ow between plants (Navarro 1999;Coelho & Barbosa 2003).
Reproductive success of P. rigida was relatively high and similar between fl oral morphs.Natural fruit-set results at PESCAN and CCPIU during PII was higher than the observed for the Brasília population (PI), but was similar to the observed in other Palicourea, such as P. padifolia (Ree 1997), P. petiolaris and P. fendleri (Sobrevila et al. 1983), and P. macrobotrys (Coelho & Barbosa 2003;Consolaro et al. 2009).It is important to note that this latter species is homostylous and self-compatible.Other Rubiaceae, as Psychotria suterella (Lopes & Buzato 2005), P. jasminoides (Castro et al. 2004), P. nuda (Castro & Araújo 2004) and Sabicea cinerea (Teixeira & Machado 2004b) presented similar fruit set results and are also markedly self-incompatible.
The reproductive effi ciency of P. rigida is also relatively high when compared to other self-incompatible woody cerrado species which presented fruit set around 7% (Oliveira & Gibbs 2000).The fruit set is explained by the high frequency of pollinated pistils observed in the studied populations.The higher amount of pollen grains observed on the stigma of thrum flowers of P. rigida indicates a greater potential for fruit set in these plants, which contradicts the male function sometimes attributed to thrum fl owers in distylous species (Monteiro et al. 1991).The higher amount of pollen grains seems to be associated with the greater stigmatic surface in these thrum fl owers, as described for Palicourea fendleri (Lau & Bosque 2003).Since incompatibility reaction in thrum fl owers occurs early on the stigma, the pollen tubes growing down the style in P. rigida thrum-fl ower pistils indicate that the pollinators are effi cient enough to transport pollen between fl oral morphs and perform legitimate cross pollination.
Palicourea rigida presented typical heteromorphic SI, with fruit set only from between morph pollinations (legitimate).This self-incompatibility mechanism is associated with characteristic pollen arresting mechanisms (Bawa & Beach 1983).The site of incompatibility reaction differed between morphs: in thrum fl owers they occurred on the stigma while in pin fl owers they occurred further down along the style.This pattern is similar to the observed in other distylous Rubiaceae studied (Bawa & Beach 1983;Coelho & Barbosa 2004;Lopes & Buzato 2005;Pereira et al. 2006;Consolaro et al. 2009), and also for heterostylous species in other plant families (Barros 1998).
In the Serra de Caldas Novas State Park (PESCAN), no Clube Caça e Pesca (CCCPIU) and in the Fazenda Água Limpa (FAL-Brasília), P. rigida seems to be an important resource for hummingbirds during the fl owering period, presenting high density of flowering individuals and many flowers.The large offer of fl owers and nectar often attract territorial hummingbirds, which drive away visitors with other foraging strategies.Since P. rigida is self-incompatible, territorial behavior could limit between-morph pollen fl ow.However, the patches defended by hummingbirds usually included plants of both morphs.Moreover, territories and territorial behavior varied throughout the day and blooming period, resulting in heterogeneous pollen fl ow and explaining reproductive success.
Based on the populations studied, it is possible to consider P. rigida a typically distylous species, presenting fl oral dimorphism, isoplethic distribution and heteromorphic self-incompatibility.The populations differed in density and height of the individuals, but these differences did not seem to interfere with the function of distyly.Recent studies with other Cerrado species of Palicourea showed both typically distylous and anomalous species occurring in the same area, indicating that each species may respond in a different way to habitat conditions and conservation history (Consolaro et al. 2009).Despite its vast distribution in the Cerrado biome, which probably exposes its populations to different environmental pressures and habitat fragmentation, it seems that P. rigida did not present the distylous anomalies that have been observed in other Cerrado Rubiaceae.

Figure 1 .
Figure 1.Flowering phenology of of thrum ( ) and pin ( ) plants of Palicourea rigida H. B. & K. (Rubiaceae) marked and followed in the Fazenda Água Limpa (FAL) in 1993 (dark symbols) and 1994 (open symbols).Arrows indicate the beginning of fl owering and peaks (more than 50% of the plants fl owering).

Table 1 .
Floral morphometrics of thrum and pin fl owers of Palicourea rigida H.B. & K. (Rubiaceae) in the Fazenda Água Limpa (FAL).Parque Estadual da Serra de Caldas Novas (PESCAN) and Clube de Caça e Pesca Itororó de Uberlândia (CCPIU) (cm, mean ± standard deviation).Means followed by different letters indicate statistical differences at P< 0,05 between morphs in each population, n = number of fl owers, F and P values of ANOVA analysis for comparison between morphs.

Table 2 .
Data obtained from the experimental pollination tests of Palicourea rigida H.B. & K. (Rubiaceae) carried out in the Fazenda Água Limpa (FAL) in 1993 and 1994.T: thrum, P: pin, n= number of plants, fl = number of pollinated fl owers, fr= number of fruits produced, % of fruits).

Figures 3 -
Figures 3-11.Flower morphology and pollen tube growth in Palicourea rigida H. B. & K. (Rubiaceae) from the Fazenda Água Limpa (FAL), Parque Estadual da Serra de Caldas Novas (PESCAN) and Clube Caça e Pesca Itororó de Uberlândia (CCPIU).3. Pin fl ower (above) and thrum fl ower (bellow).4. Longitudinal sections of the same fl owers showing the differences between pin and thrum morphs.5. Epifl uorescence photomicrography showing pollen germination and pollen tube growth on the stigma of a self-pollinated pistil of a thrum fl ower.6.Similar to fi gure 5, but an illegitimate pollinated pistil of a thrum fl ower.Note in fi gures 5 and 6 the callose rich vesicles that seem to be associated with incompatibility reaction (arrows).7. Epifl uorescence photomicrography showing pollen tube growth of the stigma of a self-pollinated pistil of a pin fl ower.8. Epifl uorescence photomicrography mosaic of pin fl ower self-pollinated pistil showing pollen tube arresting at the second third of the style (arrow).9. Details of anomalous callose plugs formed at the tip of pollen tubes after illegitimate pollination in a pin fl ower.10.Pollen tube germination and growth in a legitimate pollinated pistil of a pin fl ower.11.Pollen tube growing normally down to the base of the style of a legitimate pollinated pin fl ower (arrow).All pistils fi xed six hours after hand pollinations.Figures 5, 6 and 10 scale bar=100μm; 7 and 9 scale bar=200μm; 8 and 11 scale bar=1000μm.