Heteromorphic Incompatibility and Efficiency of Pollination in Two Distylous Pentanisia Species (Rubiaceae)

Abstract

• Background and Aims Distyly has been hypothesized to promote cross-pollination by reducing intrafloral and geitonogamous self-pollination, and enhancing intermorph pollination. Distylous plants typically display both reciprocal herkogamy and a heteromorphic incompatibility system, which allows mating only between morphs. Distyly and its pollination consequences were examined in two Pentanisia species with long-tubed flowers which are pollinated almost exclusively by butterflies.

• Methods Anther and stigma heights were measured to quantify reciprocal herkogamy. The type of incompatibility system was determined by observing pollen tubes and seed production following controlled hand pollination. Pollen loads on pollinators and stigmas were also examined to assess the efficiency of intermorph pollen flow.

• Key Results Pentanisia prunelloides and P. angustifolia exhibit reciprocal herkogamy and a host of ancillary dimorphisms, including pollen colour, exine sculpturing, stigmatic papilla shape and floral-tube pubescence. Controlled hand-pollinations revealed the presence of a strong heteromorphic incompatibility system in both species. The site of incompatibility differed between the morphs; intramorph pollen tubes were blocked in the style of the short-styled morph and on the stigmatic surface of the long-styled morph. Butterflies carried pollen from the short- and long-styled morphs primarily on their head and proboscis, respectively. Natural pollination resulted in a higher proportion of pollen transfer from long- to short-styled plants than vice versa. Nevertheless, fruit set did not differ between morphs.

• Conclusions Both Pentanisia species are fully distylous. Reciprocal herkogamy results in pollen from the two morphs being carried on different locations on pollinators' bodies, which in turn promotes intermorph pollination. Intramorph pollination does not result in fertilization, because of an effective heteromorphic incompatibility system.

INTRODUCTION

Hermaphroditism in flowers allows for simultaneous pollen removal and deposition by pollen vectors, but also exposes plants to pollinator-facilitated self-pollination, which may reduce pollen export and lead to inbreeding depression if the ovules are self-fertilized (Barrett and Harder, 1996). The separation of floral sexual parts either in time (dichogamy), or space (herkogamy) reduces the likelihood of self-pollination (Thomson and Stratton, 1985; Lloyd and Webb, 1986; Webb and Lloyd, 1986; Harder et al., 2000). In some species, herkogamy is incorporated in a polymorphism for, either stigma and anther heights (reciprocal herkogamy), or stigma height only (stylar polymorphism) (Barrett et al., 2000).

Distyly is a dimorphism that involves reciprocal herkogamy, a heteromorphic incompatibility system and a suite of associated morphological and physiological characters (Ganders, 1979). A distylous population is comprised of individuals with either long-styled flowers (hereafter L-morph), with the stigmas placed above the anthers (approach herkogamy), or short-styled flowers (hereafter S-morph), with the stigmas placed below the anthers. This reciprocal herkogamy should effectively reduce both self- and intramorph pollination, while promoting intermorph pollen flow (Ganders, 1974), because pollinators tend to carry pollen from the two morphs on different locations on their bodies (Darwin, 1877; Lloyd and Webb, 1992a). However, the efficiency of this reciprocity has been questioned (Faivre and McDade, 2001), because some species classified as distylous do not show precise reciprocity of the height of their sexual organs. In addition to reciprocal herkogamy, distyly is usually characterized by a heteromorphic incompatibility system, which precludes both self- and intramorph cross-fertilization and allows only intermorph cross-fertilization (Barrett, 1990). Distylous species also exhibit a variety of ancillary morphological polymorphisms involving primarily the pollen and stigmas of the two floral morphs (Dulberger, 1992). The range of ancillary polymorphisms associated with distylous species varies between species (Dulberger, 1992) and no specific set of polymorphisms has been defined as the standard for distylous species.

Field observations in South Africa revealed that populations of Pentanisia prunelloides Eckl. and Zeyh. and P. angustifolia Hochst. (Rubiaceae) consist of both L-morph and S-morph plants. Pentanisia appears to exhibit dimorphisms typically associated with heterostylous plants. Although the genus has been included in lists of known heterostylous species (Ganders, 1979), neither the breeding or pollination system of the genus has been studied previously (Ornduff, 1974). Differences in pollen colour and exine sculpturing between the L- and S-morphs of these Pentanisia species provided us with a unique opportunity to examine pollen loads from L- and S-morphs on both pollinators and stigmas, and thus test Darwin's hypothesis that reciprocal herkogamy promotes mechanical efficiency of pollen transfer in heterostylous plants (Darwin, 1877).

The objectives of this study were to report the presence of distyly in Pentanisia prunelloides and P. angustifolia, and to study their pollination system using the pollen colour dimorphism by (a) determining whether reciprocal herkogamy occurs in P. prunelloides and P. angustifolia; (b) testing for the existence of morph-specific incompatibility; (c) describing ancillary polymorphisms; and (d) quantifying the distribution of pollen loads on pollinators and stigmas using exine sculpturing and pollen-colour dimorphism to distinguish between the pollen grains of the two morphs.

MATERIALS AND METHODS

Study species and sites

Pentanisia (Rubiaceae) comprises 15 herbaceous species occurring in Africa and Madagascar (Launert, 1989). Pentanisia prunelloides and P. angustifolia are clonal perennials, which inhabit grasslands in South Africa and southern Mozambique (Launert, 1989) and flower prolifically after burning. Pentanisia prunelloides can be distinguished from P. angustifolia by its flatter inflorescence, broader, shorter leaves, and denser clonal growth. Individual ramets of both species are easily identified in the field and are hereafter referred to as ‘plants’. The actinomorphic, tubular flowers of both species are pale blue to mauve, with epipetalous anthers. Anthesis lasts 3–4 d for individual flowers. The bi-locular ovary produces one ovule per locule and fruit maturation takes approx. 4 weeks. The two floral morphs are easily distinguished by their pollen colour (yellow for the L-morph and blue for the S-morph), and form of herkogamy (approach versus reverse herkogamy for the L- and S-morph, respectively).

Both species were studied in KwaZulu Natal, South Africa. Pentanisia prunelloides was studied at the Vernon Crookes Nature Reserve (30°16′06″S, 30°36′29″E) on the coast, south of Durban. Flowering starts at the end of August and lasts approx. 6 weeks. The experiments on P. angustifolia were conducted at the Umgeni Valley Nature Reserve, located north of Howick (29°26′20″S, 30°14′06″E). Pentanisia angustifolia flowers from early January to late February. Morph frequencies of both populations were determined from censuses.

Floral morphology and ancillary polymorphisms

Flower production per inflorescence and floral morphology were compared between the two morphs of both species. Average flower production was counted for randomly collected inflorescences (59 L-morph and 60 S-morph for P. prunelloides; 42 L-morph and 41 S-morph for P. angustifolia). Floral morphology was compared using individual flowers from 26 L-morph and 24 S-morph plants of P. prunelloides and 34 L-morph and 42 S-morph plants of P. angustifolia. For each flower, corolla diameter, and the heights of the stigma (stigma height), anthers (anther height) and mouth of the flower tube (tube height) above the ovary were measured with digital callipers. The length of a stigmatic lobe was measured with the aid of a dissecting microscope.

To compare pollen characteristics between morphs, pollen was collected from one flower from each of 11 L-morph and ten S-morph P. prunelloides plants, and five L-morph and five S-morph P. angustifolia plants. The anthers of each flower were brushed against previously prepared electron microscope stubs. The pollen grains were examined using a Philips XL 30 environmental scanning electron microscope to identify differences in exine sculpture between the morphs, and to measure the lengths of the polar and equatorial axes (×3500 for P. prunelloides and ×2500 for P. angustifolia). Pollen grains were also tested for starch by immersing them in a 5 % solution of potassium iodide and iodine for 1 h (Olesen and Warncke, 1989).

Pollen production was counted for one flower from each of 16 L-morph and 20 S-morph P. prunelloides plants, and 21 L-morph and 19 S-morph P. angustifolia plants. Undehisced anthers were placed in microcentrifuge tubes containing 100 μL of ethanol with one drop of ‘Tween 20’ (polyoxyethylene), and were crushed using a glass rod. To digest all the tissues, except the exine, 300 μL of HCl were added to the tubes which were then placed in an oven at 40 °C for 2 h. The tubes were then centrifuged at 5000 rpm (198 g) for 10 min, after which the supernatant was removed and the pellet was resuspended in 400 μL of a 3 : 1 solution of lactic acid and glycerol. After the solution had been vortexed for 30 s, a 1-μL subsample was taken and the number of pollen grains counted under a dissecting microscope at ×50. The total number of pollen grains on the flower was calculated by multiplying the subsampled pollen count by 400.

Stigma morphology and the pubescence of the floral tube were examined for five flowers of each morph per species. To fix the tissue, the flowers were placed in vials containing 3 % buffered glutaraldehyde, which were then refrigerated for 24 h. The glutaraldehyde was replaced with two 30-min washes of 0·05 m cacodylate buffer. The stigmas were then dehydrated through an ethanol series (30 %, 50 %, 70 %, 80 %, 90 % and 3× 100 %), with 10 min per stage. The stigmatic papillae and inner surface of the floral tube (split longitudinally) were examined using a Philips XL30 environmental electron microscope.

Pollination, fecundity and compatibility

To identify which visitors to Pentanisia flowers carried pollen and to see if pollinators carry L-morph and S-morph pollen on different parts of their bodies, visitors were captured in the field and placed in individual envelopes. Quantitative data could be obtained only for visitors of P. prunelloides, as too few butterflies were captured on P. angustifolia flowers. Pollen grains were collected from the middle of the proboscis (hereafter called proboscis), and the labial palps and adjacent head regions (hereafter called head) with small pieces of clear gelatine. The gelatine was then placed on a microscope slide, melted, and the number and morph of pollen grains were determined at ×40 under a compound microscope. Pollen grains could be attributed to a particular floral morph based on exine sculpture (gelatine preparation eliminated pollen-colour differences between morphs).

Pollen loads on stigmas were examined by collecting slightly wilted flowers in the field and placing them in empty individual microcentrifuge tubes. The stigmas were then examined in the laboratory at ×10 using a compound microscope with reflected illumination. Pollen grains on the stigma were counted separately for each morph and could be distinguished by their colour (pollen grains of the two morphs retain their distinctive colour after pollen tube germination). S-morph flowers were dissected longitudinally to expose their stigmas.

Fruit set of open-pollinated plants for the two morphs was compared based on the samples collected to estimate flower production per inflorescence (see Floral morphology and ancillary polymorphisms). The samples were collected about 4 weeks after flowering, so that fruit set could be determined by the presence of a swollen ovary. The number of seeds per ovary was also counted for some fruits.

To assess the mating potential of both morphs, flowers on previously bagged inflorescences (n = 43 for P. prunelloides, n = 21 for P. angustifolia) were hand-pollinated. Each pollination bag was supported with a wire structure, which prevented the bag from touching the anthers or stigmas (depending on morph) during windy conditions. Flowers were assigned randomly to one of the following treatments: intermorph pollination, intramorph cross-pollination or self-pollination. To obtain pollen from anthers of the L-morph and to pollinate flowers of the S-morph, the corolla tube was split along half of its length. This manipulation should not affect differences between pollination treatments, because all S-morph flowers were split, regardless of treatment, and split flowers remained turgid. After pollination, the inflorescences were rebagged. Sixty-three crosses were performed per treatment on P. prunelloides (28 on L-morph flowers and 35 on S-morph flowers) and 66 on P. angustifolia (31 on L-morph flowers and 35 on S-morph flowers). Fruit set was recorded 4 weeks later.

To identify the site of incompatibility, the extent of pollen-tube growth after 48 h was examined using a modification of the staining technique described by Martin (1959). Cuttings with several flower buds were kept in the laboratory with their stems immersed in water. Two to three days later, newly opened flowers received either intra- or intermorph cross-pollination. After 48 h, the flowers were collected and placed in vials containing an FAA (formalin, acetic acid and alcohol) solution. Forty-eight hours later each flower's tube was split, the style at the top of the ovary was excised, washed with tap water and placed in a 4 n NaOH solution for 36 h. The NaOH was subsequently removed and replaced with tap water for 24 h. Styles were then placed in a 1 % solution of aniline blue for 8 h before being mounted on microscope slides. Pollen tubes at were examined at ×10 using an Olympus Provin AX70 fluorescent microscope (wavelength = 365 nm).

Data analysis

The reciprocity between the anther and stigma heights of the two morphs was determined using a reciprocity index (R) derived by J. H. Richards, D. G. Lloyd and S. C. H. Barrett (unpublished result cited in Richards and Koptur, 1993):

where A is the height of a morph's anthers and S is the stigma height of the alternate morph. With perfect reciprocity, R = 0.

Most of our statistical analyses involved simple comparisons between morphs for each species, although some comparisons of pollination involved two-factor analyses. Whether the two morphs were equally frequent within a population was assessed with goodness-of-fit likelihood-ratio (G) tests (Sokal and Rohlf, 1995). All statistical comparisons of floral morph characteristics involved two-tailed t-tests (Sokal and Rohlf, 1995). Whether the occurrence of pollen receipt varied independently of morph was assessed with Fisher's exact tests (Sokal and Rohlf, 1995). To compare the total amount of pollen received between both species and morphs, a two-factor analysis of variance was used (Neter et al., 1996; SAS Institute, 2001). Because several flowers had been sampled from individual P. angustifolia plants, this analysis considered the average pollen receipt per plant. A two-factor ANOVA, which accounted for repeated measures, was used to compare the receipt of pollen from either morph as pollen donors by both morphs as pollen recipients (SAS Institute, 2001). This analysis used restricted maximum likelihood (Jennrich and Schluchter, 1986) to estimate the variance–covariance structure of the repeated measurement of individual stigmas (i.e. pollen from two sources). For both species, compound symmetry with heterogeneous variances was the most appropriate variance-covariance model. Denominator degrees of freedom for F-tests were calculated by Kenward and Roger's approximation (Kenward and Roger, 1997). When using contrasts to interpret significant interactions between pollen source and recipient morph, the Dunn–Šidák procedure was applied to control the experiment-wise Type I error rate to α = 0·05 (Kirk, 1995). All analyses of pollen receipt considered square-root transformed data to rectify non-normality. For these analyses, back-transformed results, which resulted in asymmetric standard errors, were presented.

RESULTS

Reciprocal herkogamy, morph ratios and compatibility

Populations of both species included two distinct floral morphs. Although stigma and anther positions varied continuously within both populations, the stigmas in one group of plants are positioned above the floral tube and well beyond the anthers (L-morph), whereas the stigmas in the other group are situated within the floral tube, well below the anthers (S-morph: Fig. 1). Stigma and anther heights differ significantly between the two floral morphs for both species (Table 1). The reciprocity indices for both morphs for the two species are close to 0 (P. prunelloides: long level R = 0·02, short level R = 0·03; P. angustifolia: long level R = 0·04, short level R = 0·04), indicating strong reciprocity for sex-organ position between the two morphs. The observed reciprocity indices are comparable with those for distylous species (Richards and Koptur, 1993). The height of the epipetalous anthers correlates strongly with floral-tube length in both morphs of both species (P. angustifolia: L-morph, r = 0·991, P < 0·001; S-morph, r = 0·963, P < 0·001; P. prunelloides: L-morph, r = 0·786, P < 0·001; S-morph, r = 0·941, P < 0·001).

Fig. 1.

Variation of anther and stigma height in flowers of (A) Pentanisia prunelloides, and (B) P. angustifolia. Flowers have been ranked according to stigma height, to illustrate the reciprocal placement of anthers and stigma heights between the L- and S-morphs.

Table 1.

Mean (±s.e.) and range (in parentheses) of flower production per inflorescence and flower dimensions (mm) of the two morphs of Pentanisia prunelloides and P. angustifolia

	P. prunelloides			P. angustifolia
Character	L-morph (n = 26)	S-morph (n = 24)	L-morph (n = 34)	S-morph (n = 42)
Flower number	45·9 ± 1·89 (20–82)	43·0 ± 1·75 (16–68)	72·9 ± 4·42 (11–123)	64·2 ± 3·70 (12–105)
Tube height	16·3 ± 0·28 (13·5–19·5)	15·4 ± 0·27 (12·7–17·9)	18·2 ± 0·45 (14·0–23·0)	18·1 ± 0·39 (12·7–23·7)
Anther height	14·6 ± 0·29* (11·6–17·4)	17·9 ± 0·32 (15·6–20·9)	16·7 ± 0·43* (12·4 –20·8)	20·0 ± 0·45 (14·1–26·0)
Stigma height	19·3 ± 0·37* (16·4–22·2)	13·8 ± 0·26 (11·6–17·1)	21·2 ± 0·48* (17·1–27·0)	15·4 ± 0·31 (11·9–19·2)
Stigma lobe length	1·5 ± 0·10 (0·3–3·2)	1·4 ± 0·05 (1·0–1·9)	1·4 ± 0·05 (0·5–2·1)	1·3 ± 0·05 (0·8–2·2)
Corolla diameter	11·9 ± 0·33 (9·6–14·6)	11·4 ± 0·28 (9·0–13·6)	9·2 ± 0·20 (6·8–12·1)	9·4 ± 0·18 (6·9–11·4)

^*Morphs differ significantly (P < 0·05) based on Student's t-test.

The L- and S-morphs were equally abundant in the populations of both species. The P. prunelloides population had an L : S morph ratio of 1·10 : 1 (G_adj = 0·47, P > 0·1, n = 224). Similarly, in the P. angustifolia population the ratio was 1·08 : 1 (G_adj = 0·28, P > 0·5, n = 426).

Hand-pollinations revealed a clear pattern of incompatibility, which is characteristic of a strong heteromorphic incompatibility system (Table 2). No self-pollinated flowers produced fruit, indicating strong self-incompatibility. Intramorph cross-pollination also resulted in incompatibility reactions; however, pollen tubes were arrested at different locations within the style, depending on the floral morph. No pollen tubes were found in the styles of L-morph plants, indicating that they did not grow beyond the stigmatic surface (Fig. 2A). In contrast, intramorph pollen tubes were blocked in the stylar canal of S-morph plants (Fig. 2C). As expected from the pollen-tube results, intramorph pollination resulted in very low fruit set (Table 2). In contrast to intramorph cross-pollination, intermorph pollination resulted in pollen tubes that extended from the stigmatic surface through the stylar canal to the ovary (Fig. 2B and D). Correspondingly, intermorph pollination resulted in high fruit set, especially for P. angustifolia (Table 2).

Table 2.

Percentage fruit set in response to pollination treatment for the two floral morphs of P. prunelloides and P. angustifolia

	P. prunelloides			P. angustifolia
Treatment	L-morph	S-morph	L-morph	S-morph
Autogamy	0 (28)	0 (35)	0 (31)	0 (35)
Intramorph	0 (28)	5·7 (35)	3·2 (31)	3·4 (35)
Intermorph	71·4 (28)	45·7 (35)	77·4 (31)	84·4 (35)
Open	61·3 (60)	56·9 (59)	28·5 (42)	33·1 (41)

Sample sizes (number of flowers for the controlled hand-pollination treatments and number of plants for the open pollination treatment) are given in parentheses.

Fig. 2.

Relative extent of pollen-tube growth in stigma lobes of Pentanisia prunelloides following intramorph and intermorph pollination. (A) L-morph stigma lobes after intramorph pollination. Fluorescent patches on the stigma surface are sites where pollen grains attached prior to the washing process, but there is no evidence of pollen-tube growth. Abbreviations: S, stigma surface; C, stylar canal. (B) L-morph stigma lobes after intermorph pollination with well-developed pollen tube. (C) S-morph stigma lobes after intramorph pollination with pollen tubes arrested in the stylar canal. (D) S-morph stigma lobes after intermorph pollination with well-developed pollen tubes. Scale bars: A = 73 μm; B = 69 μm; C = 66 μm; D = 64 μm.

Ancillary polymorphisms

In addition to reciprocal herkogamy, flowers of the two morphs differ with respect to several other aspects of floral morphology. The two morphs of both species produce equivalent numbers of flowers per inflorescence. The floral morphs do not differ significantly in corolla width, floral-tube length, or length of a stigmatic lobe (Table 1), although they do exhibit some qualitative differences (Table 3). For both species, the inner surface of the corolla tube of the L-morph (Figs 3G and 4G) is less pubescent than that of the S-morph (Figs 3C and 4C). In addition, the stigma of the L-morph has larger papillae (Figs 3E and 4E) than that of the S-morph (Figs 3B and 4B).

Table 3.

Qualitative and quantitative [mean ± s.e. (n)] ancillary polymorphisms of P. prunelloides and P. angustifolia

	P. prunelloides			P. angustifolia
Character	L-morph	S-morph	L-morph	S-morph
Pollen production	8950 ± 1134·5 (16)	10050 ± 780·9 (20)	2743 ± 228·3 (21)	2674 ± 178·5 (19)
Polar axis† (µm)	36·7 ± 0·25* (55)	45·2 ± 0·29 (50)	36·1 ± 0·32* (25)	40·5 ± 0·44 (25)*
Equatorial axis† (µm)	24·8 ± 0·19* (55)	25·9 ± 0·21 (50)	24·0 ± 0·23* (25)	24·8 ± 0·27 (25)
Floral tube pubescence	Less hairy	More hairy	Less hairy	More hairy
Pollen colour	Yellow	Blue	Yellow	Blue
Exine morphology	Reticulate	Verrucate	Reticulate	Verrucate
Stigma papillae shape	Conical	Spatulate	Dome shaped	Nipple shaped

Sample sizes for quantitative measures are given in parentheses.

^*Morphs differ significantly (P < 0·05) based on Student's t-test.

^†Pollen dimensions measured using a scanning electron microscope.

Fig. 3.

Morphology of Pentanisia prunelloides flowers: (A) exine sculpturing of an S-morph pollen grain; (B) papillae of an S-morph stigma; (C) floral-tube pubescence of the S-morph; (D) cross-section of flowers; SS, S-morph; LS, L-morph. (E) papillae of an L-morph stigma; (F) exine sculpturing of an L-morph pollen grain; (G) floral-tube pubescence of the L-morph. Scale bars: A, F = 20 μm; B, E = 50 μm; C, G = 200 μm; D = 3·37 mm.

Fig. 4.

Morphology of Pentanisia angustifolia flowers: (A) exine sculpturing of an S-morph pollen grain; (B) papillae of an S-morph stigma; (C) floral-tube pubescence of the S-morph; (D) cross-section of flowers; SS, S-morph; LS, L-morph. (E) papillae of an L-morph stigma; (F) exine sculpturing of an L-morph pollen grain; (G) floral-tube pubescence of the L-morph. Scale bars: A, F = 10 μm; B, E = 20 μm; C, G = 200 μm; D = 3·35 mm.

The two morphs of both species produce equivalent numbers of pollen grains, but the grains differ with respect to exine sculpturing and size (Table 3 and Figs 3A and F and 4A and F). For both species, pollen grains of the L-morph are shorter (polar axis) and narrower (equatorial axis) than those of the S-morph. The L-morph produces yellow pollen with a reticular exine, whereas the S-morph produces blue pollen with a verrucate exine. Starch was not detected in the pollen of any morph.

Pollination and fecundity

Butterflies were the primary visitors to flowers of both species (Fig. 5), with occasional visits by bombyliid flies and bees, which mainly removed pollen from S-morph flowers. Pollinators of the P. prunelloides population at the Vernon Crooks Nature Reserve included seven butterfly species (Table 4), which were active mainly from morning to early afternoon (0800–1400 h). In the P. angustifolia population at the Umgeni Valley Nature Reserve, where observations were more limited, three butterfly species (Papilio demodocus, Junonia archesia and Danaus chrysippus aegiptus) and a bombyliid (Systoechus sp.) were observed. In this population, pollinators were active throughout the day, with peak activity during morning (0800–1000 h) and late afternoon (1600–1800 h).

Fig. 5.

Butterflies (A, Catacroptera cloanthe; B, Vanessa cardui) visiting S-morph flowers of Pentanisia prunelloides. Scale bars: A = 5 mm; B = 4 mm.

Table 4.

Average number of L- and S-morph pollen grains on the proboscides and heads of different butterfly species caught visiting P. prunelloides flowers, in order of increasing mean proboscis length

	Proboscis length (mm)	Proboscis			Head
Species (n)		L-morph	S-morph	L-morph	S-morph
Eurema brigitta (3)	8·1 ± 0·59	0	0	0	0·33
Danaus chrysippus (1)	11·1	0	0	0	0
Junonia oenone (1)	12·0	0	0	4	0
Catacroptera cloanthe (1)	13·2	13	0	3	7
Precis octaviai (4)	13·4 ± 0·19	11·25	1·0	15·5	12·5
Vanessa cardui (5)	13·4 ± 0·61	9·8	0·2	5·4	16·0
Catopsilia florella (1)	17·1	1	0	0	0

Not all butterflies caught on P. prunelloides flowers carried pollen and those that did tended to carry pollen of the two morphs on different locations of their bodies (Table 4). The species with the largest pollen loads (Catacroptera cloanthe, Precis ocataviai and Vanessa cardui) had proboscides about 13 mm long, whereas those that carried little pollen had either shorter or longer proboscides (Table 4). Pollen loads on the proboscis consisted almost entirely of L-morph pollen, whereas pollen on butterfly heads comprised an equal mixture of L- and S-morph pollen (Fig. 6; t₁₄ = 2·91, P < 0·05; t₁₄ = 1·75, P > 0·05).

Fig. 6.

Mean (±s.e.) pollen-load segregation on the bodies of butterfly pollinators of Pentanisia prunelloides.

Overall, stigmas of wilting P. prunelloides flowers received more pollen grains than those of P. angustifolia. Only 4·7 % of the 128 P. prunelloides stigmas examined did not receive any pollen. For this species, the S-morph had slightly but significantly higher pollination failure (8·1 %) than the L-morph (0 %: P = 0·039). In comparison, 43·6 % of the 156 P. angustifolia stigmas examined had not received any pollen, with L-morph flowers having significantly higher pollination failure (52·3 %) than S-morph flowers (32·4 %: Fisher's exact test, P = 0·015). On average, P. prunelloides stigmas received about seven times more pollen (mean = 27·5 grains, lower s.e. = 2·3, upper s.e. = 2·4; based on square-root transformed data) than P. angustifolia stigmas (mean = 3·8 grains, LSE = 1·0, USE = 1·2; F_1,104 = 86·52, P < 0·001), even though P. prunelloides produced only about three times more pollen than P. angustifolia flowers (Table 3). This difference was not consistent among morphs (species × morph interaction, F_1,104 = 8·72, P < 0·005), as pollen receipt did not differ among morphs for P. prunelloides (t₁₀₄ = 1·68, P > 0·1), whereas S-morph stigmas of P. angustifolia received, on average, significantly more pollen than those of L-morph plants (t₁₀₄ = 2·45, P < 0·05).

Pollen was not dispersed equally within and between morphs for either species (Fig. 7; morph × pollen source interaction: P. angustifolia, F_1,105 = 4·52, P < 0·05; P. prunelloides, F_1,61 = 30·23, P < 0·001). For both species, L-morph plants experienced equivalent intra- and intermorph pollination as pollen recipients (P > 0·5 in both cases), whereas S-morph plants received significantly more pollen via intermorph pollination (P < 0·01 in both cases). As a consequence, more pollen dispersed from L-morph donors to S-morph recipients than between any other combination of morphs. The main difference in pollination patterns between the two species arose because equivalent numbers of S-morph pollen grains were found on L- and S-morph stigmas for P. angustifolia (t₄₁ = 2·10, P > 0·1), whereas for P. prunelloides significantly less S-morph pollen was found on S-morph stigmas than on L-morph stigmas (t₆₁ = 6·28, P < 0·001). For both species, about three-quarters of the pollen on stigmas originated from L-morph plants (P. angustifolia, 71·0 %; P. prunelloides, 75·1 %).

Fig. 7.

Mean (±s.e.) receipt of pollen from L- and S-morph donors by L- and S-morph plants of (A) Pentanisia prunelloides and (B) Pentanisia angustifolia. Based on square-root transformed data.

Fruit set of open-pollinated plants ranged from 29–61 %, being much higher for P. prunelloides than for P. angustifolia (Table 2). Fruit set did not differ significantly between morphs for either species. All examined fruits produced two seeds.

DISCUSSION

Populations of Pentanisia prunelloides and P. angustifolia consist of two flower morphs characterized by the reciprocal placement of stigmas and anthers (reciprocal herkogamy) (Fig. 1), a strong heteromorphic incompatibility system (Table 2 and Fig. 2) and a wide range of ancillary polymorphisms (Table 3). Reciprocal herkogamy promotes efficient pollen transfer between morphs in these species, as indicated by pollen segregation on pollinators' bodies and disassortative pollen flow between the two morphs (Table 4 and Fig. 7). Therefore, both species are morphologically, genetically and functionally distylous.

Flower size and the positions of anthers and stigmas within flowers vary continuously in both species, with no discontinuity between the two morphs (Fig. 1). Nevertheless, the two morphs could be distinguished based on the relative positions of anthers and stigmas within flowers. An absence of distinct style levels has been reported in several other distylous species (Richards and Koptur, 1993; Arroyo and Barrett, 2000). Unlike some other distylous species, such as Guettarda scabra, Gaertnera vaginata and Psychotria poeppigiana (Richards and Koptur, 1993; Pailler and Thompson, 1997; Faivre and McDade, 2001), floral-tube length did not differ significantly between the two morphs even though the positions of the epipetalous anthers differed significantly. Therefore, differences in filament length seem to be the primary cause of dissimilar anther height between the morphs (cf. Scribailo and Barrett, 1991).

The presence of a heteromorphic incompatibility system was confirmed by low fruit set and arrested pollen-tube growth following intramorph pollination (Fig. 2 and Table 2). The dissimilar sites of incompatibility between the floral morphs that were observed has been reported in some other distylous species, particularly in the Rubiaceae, although, in most of these studies, pollen tube arrest is on the stigma of the S-morph and in the style of the L-morph (Bawa and Beach, 1983; Dulberger, 1992; Faivre, 2002). Differences in incompatibility sites may not have functional significance per se, but may simply reflect different origins of the incompatibility systems for the L- and S-morphs (Lloyd and Webb, 1992b). Although not examined quantitatively, the stigmatic papillae of the L-morph appear larger than those of the S-morph, as has been reported for other distylous species (Ganders, 1979; Scribailo and Barrett, 1991; Dulberger, 1992).

Pollination biology

The location of pollen on butterflies' bodies is partially consistent with Darwin's proposal (Darwin, 1877) that the contrasting anther positions of L- and S-morph flowers of distylous species results in pollen being carried on different body locations, thereby facilitating intermorph transfer. Butterfly proboscides carried P. prunelloides pollen predominantly from the inserted anthers of the L-morph (Fig. 6), which would have been positioned suitably for transfer to the stigmas of S-morph plants. In addition, most of the pollen from the exserted anthers of the S-morph was carried on butterflies' heads (Fig. 6), and so would have been positioned suitably for transfer to L-morph stigmas. Both of these outcomes should promote intermorph pollination. However, butterfly heads also carried as many L-morph pollen grains as S-morph grains (Fig. 6). The presence of L-morph pollen on the heads and palps of butterflies, even though it must be extracted from the flower by the proboscis, may arise through two mechanisms. Pollen could be transferred from a butterfly's proboscis to its head when the proboscis is coiled during flights between plants. In addition, the hairs in the upper floral tube may collect some pollen as butterflies proboscides move past, which may then be transferred to the heads of butterflies that visit subsequently. This latter mechanism may be particularly common for butterflies with proboscis lengths shorter than the floral tube, as was the case for the species that carried most Pentanisia pollen (cf. Tables 1 and 4). Regardless of mechanism, the presence of abundant L-morph pollen on butterflies' heads should cause considerable intramorph pollination of the L-morph.

The composition of stigmatic pollen loads is largely consistent with the distribution of pollen from the two morphs on butterflies' bodies. In particular, pollen flow was asymmetric, with more intermorph pollen grains present on S-morph stigmas than on L-morph stigmas (Fig. 7), despite comparable pollen production between the morphs (Table 3). Similar results have been reported for other distylous species (Ganders, 1974; Lloyd and Webb, 1992a). It was not possible from the present data to determine whether the higher incidence of intramorph pollination for the L-morph resulted from intramorph cross-pollination, geitonogamous self-pollination or within-flower self-pollination. Of these alternatives, cross-pollination may dominate, because most L-morph pollen on stigmas was involved in intermorph transfer (Fig. 7). By contrast, Ganders (1979) and Piper and Charlesworth (1986) suggested that most intramorph pollination of other distylous species resulted from self-pollination (also see Lloyd and Webb, 1992a).

Most studies of pollen dispersal for distylous species that have found support for Darwin's hypothesis have considered bee- and/or hummingbird-pollinated species (e.g. Ganders, 1974; Piper and Charlesworth, 1986; Feinsinger and Busby, 1987; Stone, 1996; Lau and Bosque, 2003), whereas Stone (1996) reported that butterflies did not transfer pollen effectively between morphs of Psychotria suerrensis. The present results for P. prunelloides illustrate that butterflies can disperse pollen effectively between the morphs of distylous species (Fig. 7A). By contrast, for P. angustifolia, butterflies deposited more S-morph pollen on S-morph stigmas than on L-morph stigmas (Fig. 7B). Together, the present results and those of Stone (1996) suggest that the effectiveness of butterflies as pollinators of distylous species depends strongly on specific conditions. Given that the only pollinators carrying significant amounts of P. prunelloides pollen had proboscides 13–14 mm long (Table 4), an important condition may be the abundance of butterfly species whose proboscis length is well-suited to exchange pollen with flowers, given their floral-tube length.

Despite differences in the amount and composition of pollen loads between the L- and S-morphs, fruit set did not differ between morphs (Table 2). Similar results have been reported for other distylous species (Price and Barrett, 1984; Barrett and Glover, 1985). This result is understandable for P. prunelloides, as most stigmas received many more legitimate pollen grains than would be required to fertilize the two ovules per ovary (Fig. 7A). By contrast, fruit set in the P. angustifolia population studied seems to have been highly pollen limited, so that differences in the receipt of legitimate pollen between the morphs should have affected their fruit set. In this population, 44 % of flowers received no pollen and only about 30 % of flowers set fruit (Table 2). The average L-morph flower received less than one legitimate pollen grain, whereas the average S-morph flower received four legitimate grains (Fig. 7B). Nevertheless, the morphs had equivalent fruit set (Table 2).

Overall, the reciprocal herkogamy, heteromorphic incompatibility system and ancillary polymorphisms in Pentanisia prunelloides and P. angustifolia, reported here for the first time, indicate that these species warrant the classification of distylous. The pollen colour dimorphism in these Pentanisia species will facilitate further investigations of the functional significance of distyly, selection on floral traits that might influence pollen receipt, pollen fates in distylous plants and the potential for negative frequency-dependent selection to maintain sexual dimorphisms (cf. Thompson et al., 2003).