Histological Study of Post‐pollination Events in Spathodea campanulata Beauv. (Bignoniaceae), a Species with Late‐acting Self‐incompatibility

Abstract

The reproductive biology of Spathodea campanulata was investigated by means of hand‐pollination experiments, observations of pollen tube growth using fluorescence microscopy, and serial sections of ovules in selfed and crossed pistils. Only cross‐pollinated flowers developed fruits, and all selfed flowers abscised within 3–4 d. However, self pollen tubes grew successfully to the ovary, penetrating and fertilizing the majority of ovules by 48 h, indicating that S. campanulata is a species with late‐acting self‐incompatibility. The incidences of ovule penetration, fertilization and endosperm initiation were all significantly slower in selfed vs. crossed pistils, although no other signs of malfunctioning were detected. The possible role of such slow self pollen tube effectiveness as a recognition event is discussed within the context of the slow but not entirely suppressed self pollen tube growth reported for some species with conventional homomorphic self‐incompatibility.

INTRODUCTION

The predominantly Neotropical family Bignoniaceae is composed of around 800 species distributed in 112 genera (Spangler and Olmstead, 1999). Breeding systems have been investigated in 27 species of this family (Bawa, 1974; Stephenson and Thomas, 1977; Petersen et al., 1982; Bullock, 1985; Bertin and Sullivan, 1988; Amaral, 1992; Gobatto‐Rodrigues and Stort, 1992; Vieira et al., 1992; Gibbs and Bianchi, 1993, 1999; James and Knox, 1993). Except for Pyrostegia venusta (Gobatto‐Rodrigues and Stort, 1992), Astianthus viminalis (Bullock, 1985), Tecoma stans (Singh and Chauhan, 1996; Dutra and Machado, 2001) and Tabebuia chrysotricha (N. S. Bittencourt Jr and J. Semir, unpubl. res.), all species were self‐incompatible. In 17 of these species investigated for the site of pollen tube incompatibility reaction, all were shown to present ovarian or late‐acting self‐incompatibility (LSI) sensu Seavey and Bawa (1986) and Sage et al. (1994).

In late‐acting self‐incompatibility, selfed pistils are uniformly rejected despite the fact that self pollen tubes grow to the ovary and penetrate ovules. Seavey and Bawa (1986) recognized that it may be difficult to distinguish between self pistil rejection due to active self rejection (i.e. some kind of novel late‐acting SI) and the effects of early‐acting inbreeding depression, and they suggested various criteria which might help distinguish these two phenomena. Klekowski (1988) has challenged most of these criteria and argued for an inbreeding depression explanation for most LSI‐type phenomena. Nic Lughadha (1998) presented additional evidence favouring embryonic lethals as an explanation for selfed pistil rejection. Certainly in some species, e.g. Epilobium obcordatum (Seavey and Carter, 1994, 1996), Dalbergia miscolobium (Gibbs and Sassaki, 1998), Gomidesia fenzliana, G. cerqueiria (Nic Lughadha, 1998) and Calluna vulgaris (Mahy and Jacquemart, 1999), the occurrence of embryonic malformations and/or delayed abscission in selfed pistils/young fruits favours an inbreeding depression explanation. But there is evidence that in some taxa ovarian incompatibility may be due to late‐acting self‐incompatibility mechanisms with major gene control (Cope, 1962; Jacob, 1980; Lipow and Wyatt, 2000), or that selfed pistil abscission may be triggered by events that occur during self pollen tube growth in the style, prior to their arrival at the ovules (Gibbs and Bianchi, 1999; Lipow and Wyatt, 1999; Sage et al., 1999).

However, most studies involving taxa with LSI‐type phenomena lack detailed histological analyses of the consequences of self vs. cross pollen tube penetration of ovules (but see Gibbs and Bianchi, 1993; Seavey and Carter, 1994; Sage et al., 1999). In this study, the reproductive biology of Spathodea campanulata, a member of the Bignoniaceae—a family known to have taxa with ovarian sterility or LSI—was investigated, and the histology of post‐pollination events in pistils of selfed and crossed flowers was compared to determine whether inbreeding depression events or major gene control are implicated for the self‐sterility in this species.

MATERIALS AND METHODS

The species

Spathodea campanulata, in a monospecific genus, is native to tropical forests in West Africa. This species is widely cultivated as a pan‐tropical ornamental tree in urban landscapes, and is also reported as an invasive colonizer in various areas, e.g. Micronesia (Smith, 1985; McConnell and Muniappan, 1991), Puerto Rico (Rivera and Aide, 1998) and Brazil (pers. obs.). The inflorescence of S. campanulata is a corymb of around 40–50 flowers with acropetal maturation. The showy orange, ornithophilous flowers have a fairly typical bignoniaceous zygomorphic corolla, with four didynamous stamens, and a bicarpellate pistil with a 7·0–7·5 cm long style which terminates in an exserted, bilamellate, touch‐sensitive stigma. The ovary contains around 1000 ovules.

Hand pollinations

Controlled pollinations were carried out with trees cultivated on the campus of the Universidade Estadual de Campinas (UNICAMP), São Paulo state, Brazil. Initial hand‐pollination experiments were made with five trees during February–March of 2000, and subsequently with eight trees at the same site between December 2001 and February 2002. In the first year of study, whole inflorescences with mature basal buds were enclosed in waterproof paper bags, and subsequently either selfed or crossed during the morning of the first day of opening, before being rebagged. Mature fruits were collected after 4 months.

At least five cross‐ and self‐pollinated pistils were fixed in FAA50 (Johansen, 1940) at 24‐, 48‐ and 72‐h intervals. These pistils were subsequently partially dissected by removing the ovary wall and stained with aniline blue. Pollen tube growth in the style and ovary was studied using fluorescence microscopy (Martin, 1959). However, attempts to estimate the incidence of ovule penetration by scoring pollen tube ‘tails’ using ovules dissected from the ovaries and mounted in a droplet of aniline blue (Gibbs and Bianchi, 1999) were unsuccessful, and sectioned pistils (see below) obtained from pollinations in the following season were used instead.

In the 2002 season, self‐ and cross‐pollinations were made as before, and pistils were either collected at 24–72‐h intervals and fixed in FAA50, or collected at 24–96‐h intervals for selfs and 1–14‐d intervals for crosses and fixed in 2 % glutaraldehyde in 0·1 m sodium phosphate buffer at pH 7·0 (Gabriel, 1982). Pistils were partially dissected to remove the ovary wall prior to fixation. For FAA50‐fixed pistils, dissected ovaries were infiltrated with tertiary butanol, embedded in paraffin (Johansen, 1940) and transverse serial sections were cut at 8 µm using a Micron HM 240 E rotary microtome. Sections were stained with astra blue and basic fuchsin (Roeser, 1972), so as to contrast cellulose walls from the protoplasmic content of the embryo sac/early endosperm and surrounding cells. Ovules were sectioned throughout the entire length of each ovary.

For histological analyses, all sections derived from three selfed and three crossed ovaries selected at random for the 24‐, 48‐ and 72‐h fixation intervals were examined. Results were pooled for each treatment interval, giving a total of 13 090 ovules from 18 ovaries. The χ² statistic was used to compare numbers of penetrated ovules and ovules showing various stages of post‐penetration events in these ovaries.

Dissected ovaries from glutaraldehyde‐fixed pistils were embedded in glycol methacrylate, and semi‐serially sectioned at 6 µm. Sections were stained with toluidine blue O, with acid fuchsin and toluidine blue O, or with PAS and aniline blue black (O’Brien and McCully, 1981, modified). These sections were used to confirm histological details of events in the embryo sac, but the incomplete sets of sections from the resin‐embedded ovaries were not analysed statistically as for the paraffin‐embedded sections. Both paraffin‐ and methacrylate‐embedded sections were examined using an Olympus BX40 light microscope and photographic records were made with an Olympus BX50 photomicroscope, using Fujichrome Provia 100 F.

RESULTS

Hand pollinations

Results of the pollination experiments are presented in Table 1. In all self‐pollinated flowers, abscission of the corolla and style, or the whole flower, occurred 3–4 d after pollination. In a few self‐pollinated flowers the pistil persisted after corolla abscission, but fell off with the dry calyx in the following 1 or 2 d. No swelling occurred in selfed ovaries. Fruit set was 55 % in hand cross‐pollinated flowers. In successfully cross‐pollinated flowers, corolla and style abscission occurred 3–4 d after anthesis, and initial swelling of the pistil was visible 1 week after pollination.

Table 1.

Experimental pollinations and fruit‐set in Spathodea campanulata.

	Self‐pollination		Cross‐pollination
Tree	Number of flowers	Fruit‐set	Number of flowers	Fruit‐set
1	20	0	25	12
2	24	0	26	15
3	22	0	28	14
4	17	0	12	8
5	18	0	9	6
Total	101	0 (0 %)	100	55 (55 %)

Ovule penetration

In all selfed and crossed pistils examined 24 h after pollination, many pollen tubes had traversed the base of the style into the ovarian transmitting tissue and some ovules had been penetrated (Table 2). However, the incidence of penetrated ovules was significantly higher for crossed pistils at 24 h, and also at 48 and 72 h post pollination (χ² = 253·7, d.f. = 1, P < 0·001; χ² = 140·1, d.f. = 1, P < 0·001; and χ² = 92·8, d.f. = 1, P < 0·001, respectively), although the majority of ovules in both selfed and crossed pistils were penetrated and fertilized by 48 h (Table 2). No variation was observed in the number of penetrated ovules at different levels of the ovary.

Table 2.

Spathodea campanulata: post‐pollination events at 24, 48 and 72 h in ovules of pistils with self‐ vs. cross pollination

	Time after pollination
	24 h		48 h		72 h
	Self	Cross	Self	Cross	Self	Cross
Not penetrated	1694 (84·9 %)	1372 (65·6 %)	477 (18·7 %)	168 (7·5 %)	102 (4·5 %)	0
Penetrated	221 (11·1 %)	668 (31·9 %)	1961 (76·9 %)	2058 (91·5 %)	2026 (90·0 %)	1924 (98·4 %)
Fertilized	0	0	1871 (73·3 %)	1931 (85·8 %)	1999 (88·8 %)	1924 (98·4 %)
Endosperm two‐celled	0	0	3 (0·12 %)	266 (11·8 %)	835 (37·1 %)	1180 (60·3 %)
Endosperm three‐celled	0	0	0	2 (0·08 %)	0	636 (32·5 %)
Endosperm three‐/five‐celled	0	0	0	0	0	62 (3·2 %)
Without embryo sac	52 (2·6 %)	39 (1·9 %)	91 (3·6 %)	15 (0·67 %)	97 (4·3 %)	27 (1·4 %)
Abnormal embryo sac	26 (1·3 %)	11 (0·52 %)	22 (0·86 %)	9 (0·4 %)	26 (1·6 %)	4 (0·20 %)
Number of ovules observed	1993	2090	2551	2250	2251	1955

Each column pools observations from three pistils sampled at random for each treatment.

Histology of the embryo sac

Although ovules were not observed on the first day of anthesis, almost all ovules in ovaries fixed on the second day of anthesis contained a mature, normally developed embryo sac (Fig. 1A; Table 2). The mature Polygonum‐ type embryo sac of S. campanulata is elongated, with a swollen micropylar region surrounded by a compact pecto‐ cellulosic, amyloplast‐containing envelope, derived from the degeneration of the micropylar portion of the nucellar epidermis and part of the endothelium. The two prominent hooked synergids present a conspicuous PAS‐positive filiform apparatus at the micropylar end, and are conspicuously vacuolated at the chalazal end (Fig. 1B). The egg cell is positioned chalazally to the synergids, and its narrow micropylar end is not in contact with the micropylar channel. The egg cell nucleus is positioned at the chalazal end, while the middle and micropylar portions of this cell are conspicuously vacuolated (Fig. 1A). The cytoplasm that surrounds the egg cell nucleus contains numerous amyloplasts. There is no cell wall between the chalazal end of the egg apparatus and the central cell, and a very narrow space filled by the protoplast of the central cell occurs between the lateral walls of the egg apparatus and the embryo sac envelope. The two polar nuclei in the central cell usually remain unfused until the second day of anthesis (Fig. 1A) and, although fusion may occur before ovule penetration creating a conspicuous secondary nucleus, unfused or incompletely fused polar nuclei were frequently observed in recently penetrated embryo sacs. Amyloplasts are abundant in the cytoplasm of the mature central cell, and they are subsequently inherited by the cells of the early developing endosperm (Fig. 1F). In the final stage of embryo sac maturation, i.e. shortly before pollen tube penetration, the chalazal end of the egg cell was very prominent, adjacent to the fused or unfused polar nuclei. The antipodals are densely cytoplasmic but vacuolated cells (Fig. 1C) that persisted until the early stages of endosperm formation.

Fig. 1. Longitudinal sections of ovules at different stages of development, oriented with chalazal end uppermost. A, Ovule containing a mature, second day anthesis embryo sac. Arrow indicates polar nuclei. B, Egg cell apparatus before pollen tube penetration, showing the two synergids. Arrow indicates the filiform apparatus; asterisk indicates the chalazal end of the egg cell in tangential section; white arrowheads indicate the embryo sac envelope. C, Chalazal end of the mature embryo sac showing the three antipodals (arrows). D, Primary endosperm nucleus in early prophase (arrow). E, Egg cell apparatus in a penetrated embryo sac just prior to fertilization. Arrow indicates the degenerate synergid protoplasmic loop between egg and central cell. F, Two‐celled endosperm. Arrowheads indicate amyloplasts. G, Three‐celled endosperm immediately after the second cycle of mitosis. H, Three‐celled endosperm, after expansion of the micropylar transverse tier of cells. White arrows indicate the two nuclei at the chalazal chamber in different plane of focus; black arrows indicate the cell wall around the chalazal end of the zygote. I, Initiated seed with a five‐celled endosperm. Arrow indicates the emerging wing; arrowheads indicate remnants of entangled pollen tubes above the external surface of the placental epidermis. All sections from glutaraldehyde‐fixed and resin‐embedded material, stained with toluidine blue O and acid fuchsin, except D (FAA50‐fixed and paraffin‐embedded material, stained with astra blue and basic fuchsin). a, Antipodal; cc, central cell; ds, degenerate synergid; e, egg cell; ee, embryo sac envelope; h, hypostase; mc, micropylar channel; ne, nucelar epidermis; ps, persistent synergid; s, synergid; z, zygote.

Penetration events in selfs and crosses

The first sign of degeneration in one of the synergids, i.e. a more darkly stained cytoplasm, was observed only after pollen tube penetration into the micropyle, but before any contact with the filiform apparatus. After full penetration, the pollen tube discharged its dense and dark‐staining contents into the degenerating synergid, so that it was difficult to distinguish the contents of this cell. Subse quently, a protoplasmic loop of the degenerating synergid was always observed to develop intrusively between the chalazal end of the egg cell and the central cell (Fig. 1E and G). We were unable to verify unequivocally the transference of the sperm cell nuclei from the degenerating synergid to the egg and central cell, but the dense cytoplasmic loop observed between egg and central cell seems to be related to these transferences.

Likewise, we were unable to visualize directly and unmistakably the processes of syngamy and triple fusion. However, triple fusion may occur prior to syngamy since, in the majority of penetrated embryo sacs with the protoplasmic loop of the degenerate synergid around the egg cell, a prominent primary endosperm nucleus in early prophase was clearly observed (Fig. 1D), whilst no sign of nuclear change was visible at this stage in the egg cell. The conspicuous presence of the cytoplasmic loop linking the synergid to the egg cell was used to score ovules as fertilized. Significantly more fertilized ovules were observed in crossed vs. selfed pistils at 48 and 72 h [χ² = 76·6, d.f. = 1, P < 0·001; χ² = 118·7, d.f. = 1, P < 0·001, respectively].

Endosperm development

In the Bignoniaceae the initial stages of endosperm development are very characteristic (Govindu, 1950; Johri et al., 1992; Shivaramiah, 1998) and so can be used to monitor events in penetrated ovules. Endosperm initiation is cellular. The primary endosperm nucleus divides transversely to give a small chalazal cell and a large micropylar cell (two‐celled stage). The latter then divides longitudinally to give two cells (three‐celled stage). The third mitotic cycle is signalled by a longitudinal division of the nucleus of the chalazal cell which may or may not be followed by cell wall formation (four‐celled or three‐celled but four‐nucleate endosperm, respectively). Subsequently, in most taxa, the nuclei of the micropylar cells each divide transversely to give two intermediate or central cells (six‐celled, or five‐celled, six‐nucleate endosperm). Thereafter, endosperm development proceeds conventionally by waves of synchronized divisions of nuclei derived from the central cells only.

At 48 h after cross‐pollination, a two‐celled endosperm (Fig. 1F) was observed in approx. 12 % of the crossed ovules, and in a very few selfed ovules (Table 2). Although the second mitotic division of the endosperm in the micropylar chamber (Fig. 1G) was usually longitudinal, occasional transverse divisions of the micropylar chamber were also observed. A few ovules with a three‐celled endosperm were seen in crossed pistils at 48 h after pollination, but no ovules with three‐celled endosperm were seen at this stage in selfed pistils (Table 2).

At 72 h after pollination, the majority of ovules in crossed pistils were in the two‐celled or later stages of endosperm formation, whilst in selfed ovaries only around 37 % of the ovules had undergone the first endosperm division. Again this difference was significant (χ² = 2277, d.f. = 1, P < 0·001). While at this time interval 32 % of the crossed ovules presented an endosperm at or beyond the three‐celled stage (Table 2), in selfed ovules a three‐celled endosperm was seen in only one of the few self‐pollinated pistils that survived for 96 h after pollination. Consequently, the difference between selfed and crossed pistils for ovules with a three‐celled endosperm was highly significant. The difference between self‐ and cross‐pollinated pistils in the number of ovules fertilized, but not yet with a developing endosperm (not shown), was also statistically significant.

The presence of a PAS‐positive wall around the chalazal end of the egg cell indicated that it must be interpreted as a zygote (Schulz and Jensen, 1968; Olson and Cass, 1981). Before the third cycle of divisions in the endosperm, the two micropylar cells underwent a period of expansion and vacuolation (Fig. 1H) which caused the lateral walls of the egg apparatus to separate from the embryo sac envelope. Some crossed ovules with such vacuolated micropylar cells were observed at 72 h after pollination. The third mitotic cycle of the endosperm was usually observed at the chalazal cell (Fig. 1H). This division was not followed by wall formation, and gave rise to the bi‐nucleate single‐celled chalazal haustorium of the endosperm (Fig. 2F).

Fig. 2. Longitudinal sections of ovules and recently initiated seeds at different stages of development, oriented with chalazal end uppermost. A, Five‐celled endosperm. B, Recently formed proembryonal tube, 5 d after pollination. White arrow indicates the micropylar nucleus; black arrow indicates the chalazal nucleus. C, Endosperm and proembryonal tube, 7 d after pollination. Proembryonal tube micropylar nucleus is not shown. Arrow indicates proembryonal tube chalazal nucleus. D, Proembryonal tube chalazal tip laterally entering the chalazal end of the endosperm body (8 d after pollination). Arrow indicates proembryonal tube chalazal nucleus. E, Proembryonal tube micropylar nucleus (arrow). F, Chalazal end of the endosperm, showing the two‐nucleated chalazal haustorium (arrow). G, Young seed 13 d after pollination. Arrow indicates a remnant of pollen tube at the micropyle opening. H, Non‐penetrated non‐functional ovule (without embryo sac), 96 h after pollination. Arrow indicates the micropylar channel. I, Penetrated non‐functional ovule, with a pollen tube coiled into the embryo sac cavity. All sections from glutaraldehyde‐fixed and resin‐embedded material, stained with toluidine blue O and acid fuchsin. ds, Degenerate synergid; pt, proembryonal tube; z, zygote.

The simultaneous transverse division of the two elongated micropylar cells of a three‐celled endosperm may occur after or at the same time as karyokinesis takes place in the chalazal chamber. These fourth cycle divisions are unequal, resulting in two smaller, middle cells and two larger, micropylar cells (Fig. 2A). Therefore, at this stage the five‐celled endosperm comprised the two larger micropylar cells, the two middle cells and the one‐celled, bi‐nucleate future chalazal haustorium. The persistent synergid was still intact at this stage, but degenerated soon afterwards. Although some five‐celled endosperms were observed in crossed ovules at 72 h after pollination (Table 2), they were most common after 96 h. In just one self‐pollinated pistil that survived to 96 h, only four ovules with a five‐celled endosperm were observed, and this was the most advanced stage of endosperm development found in any selfed ovules.

A tangential elongation of cells at the margins of the single integument, signalling the initiation of the seed wing, was clearly observed in ovules at the three‐ to five‐celled endosperm and later stages (Fig. 1I), but was also observed in non‐viable ovules (Fig. 2H and I). At 14 d post pollination, the last stage monitored, the endosperm of crossed ovules usually comprised 11 cells, which derived from the middle tier of cells, with the two micropylar cells simply undergoing expansion and vacuolation (Fig. 2C and G). These subsequent divisions were usually perpendicular to the longitudinal axis of the endosperm, so that the body of the multicellular endosperm is composed of two longitudinal tiers of cells which are progressively smaller from the micropylar to the chalazal end of the endosperm (Fig. 2C, F and G). The two micropylar cells never showed dense cytoplasm in any observed stage of development. Although some endosperms with a cell arrangement somewhat different from that described above were occasionally seen, no signs of endosperm malformation were observed in selfed or in crossed ovules. Effectively, up to 96 h post pollination in the selfed ovules, and up to 14 d post pollination in crossed ones, no signs of endosperm malfunction (other than slower development in the former) were observed.

Formation of the proembryonal tube

After fertilization the zygote began to acquire a cylindrical appearance. Although karyokinesis was not observed directly in the zygote nucleus, a two‐nucleate proembryonal tube was clearly evident from the fifth or sixth day after (cross) pollination (Fig. 2B). The micropylar nucleus is smaller than the chalazal one, and an expanding vacuole began to develop between them. In subsequent stages, the proembryonal tube continued to grow towards the chalazal end of the endosperm, usually between the two longitudinal tiers of cells of the endosperm body (Fig. 2C). Initially, this elongation of the proembryonal tube seemed simply to follow the longitudinal expansion of the endosperm. However, an intrusive growth of the proembryonal tube chalazal tip was evident starting from the seventh day after pollination, when it began to penetrate into the chalazal end of the endosperm (Fig. 2D), where the smaller endosperm cells were located. The proembyonal tube chalazal nucleus was prominent, containing a conspicuous nucleolus, and was surrounded by a dense cytoplasm, whereas the micropylar nucleus remained inconspicuous, with a small nucleolus (Fig. 2E). The two nuclei were separated by a very large vacuole. No wall formation was seen between them, and the proembryonal tube remained binuclear until the last examined stage of the young (crossed) seed at 14 d after pollination.

Remnants of the synergids remained at the micropylar end of the endosperm, adjacent to the proembryonal tube base, and the pollen tube was also visible at the micropylar opening until the last monitored stage (Fig. 2G). As with endosperm development, no proembryonal tube malformations were seen up to 14 d, although some variation in developmental progress was observed, with some proembryos and endosperms presenting earlier stages of development than the majority of young seeds in the same fruit.

Non‐functional ovules

Less than 4 % of the 13 090 ovules observed in sectioned ovaries of selfed and crossed pistils (Table 2) did not present an embryo sac (Fig. 2H), or contained an embryo sac with an anomalous structure, i.e. in an early one to four‐nucleated stage, with a degenerate embryo sac, or with an apparently mature embryo sac, but lacking some cells. In some cases, ovules lacking a normal embryo sac were penetrated and the pollen tube coiled inside the embryo sac cavity (Fig. 2I). As with functional ovules that presented different stages of pre‐ and post‐fertilization development in the same pistil, non‐functional ovules were distributed throughout the length of the ovary.

DISCUSSION

Spathodea campanulata is clearly a species with marked self‐incompatibility (no fruit set from 101 selfings on five trees), but this is the result of some kind of LSI since the majority of ovules in selfed pistils were penetrated and fertilized within 48 h. This study brings the number of bignoniaceous species with SI to 28, and the number reported with LSI to 18. Given that the remaining species were not studied for post‐pollination events, it is likely that LSI is also present in some, if not all, of the ‘self‐incompatible’ species reported in this family.

Post‐penetration development in selfed vs. crossed ovules

As is commonly the case in embryological studies, it was difficult to establish the precise timing of karyogamy to form the zygote and of the triple fusion to form the primary endosperm nucleus. However, endosperm initiation in the Bignoniaceae, with its consistent pattern of cell divisions, is sufficiently distinctive to permit this process to be followed in detail in self‐ and cross‐pollination treatments.

Gibbs and Bianchi (1999) noted an initially slower rate of ovule penetrations in selfed pistils compared with crosses in the LSI bignoniaceous species Tabebuia nodosa and Dolichandra cynanchoides, although this difference levelled off at 96 h. The histology of penetrated ovules was not examined in that study. In S. campanulata there is a similar significantly slower rate of ovule penetrations in selfed vs. crossed pistils, at 24, 48 and 72 h, i.e. over the life‐span of selfed flowers in this species. Moreover, our histological observations revealed that at 48 h post pollination, there was nearly a 100‐fold difference between selfs and crosses in the presence of ovules at the two‐celled stage of endosperm development. By 72 h, effectively the last day before abscission for most selfed pistils, a significantly smaller number of ovules with a two‐celled endosperm was again observed in selfed (37 %) vs. crossed (60 %) pistils, and no selfed ovules had reached the three‐celled endosperm stage, compared with 32 % ovules in crossed pistils.

Overall, selfed pistils in S. campanulata present an LSI situation where progress at important phases of post‐pollination, from incidence of ovule penetration and fertilization to initial stages of endosperm development, are characteristically slower than in crossed pistils, but development is otherwise normal. At no stage up to abscission of selfed flowers were any indications of malformation observed within ovules penetrated by self pollen tubes. Thus, there is no evidence to support the view that selfed pistils are abscised following embryo or endosperm malfunction owing to the action of deleterious recessive alleles. Lack of obvious embryo malformations also characterized the crossed ovules up to the 14 d limit of this survey.

Early reports of major gene control of LSI in the Sterculiaceae (Cope, 1962; Jacob, 1980) and, in particular, the more recent finding based on diallel crosses that LSI in Asclepias exaltata is controlled by a single gene with multiple alleles operating in a sporophytic or possibly gametophytic manner (Lipow and Wyatt, 2000), open new possibilities for LSI studies. Likewise, the study by Sage et al. (1999) with the LSI species Narcissus triandrus, in which it was demonstrated that growth of self vs. cross pollen tubes in the style has different effects on ovule maturation in the ovary, raises the possibility that ‘negative’ pollen tube–stylar interactions may be triggered in LSI species even though self pollen tubes continue to grow to the ovary and ovule penetrations/fertilizations occur.

Given the possibility of major gene control of LSI in some taxa, as demonstrated in Asclepias, and thus a similarity with the genetic control in homomorphic, gametophytic self‐incompatibility (GSI), the observations of Lush and Clarke (1997) on pollen tube growth following SI pollinations in Nicotiana alata are particularly intriguing. These authors reported that in this GSI species, growth of incompatible pollen tubes is not completely suppressed by the incompatible response, but that tubes continue to grow in the style, albeit at a much slower rate than crossed tubes, until the flowers senesce. These observations are similar to those by Herrero and Dickinson (1980) who reported that incompatible pollen tubes in Petunia hybrida, another GSI species, continued to grow slowly for 5 d after pollination.

A striking feature of LSI in S. campanulata is that selfed flowers have a longevity that is similar to that of unpollinated flowers and they abscise within 3 d, despite the fact that at this time the majority of ovules in the ovary had been fertilized. The only difference that we observed between crossed and selfed pistils in this species was a slower incidence of ovule penetrations in selfed pistils and, consequently, a slower rate of development within the embryo sac. The difference between this slower action of self pollen tubes in the LSI species S. campanulata and the ‘fatally’ slow growth of self tubes described in various GSI species, may be a quantitative rather than qualitative one. In the latter, self pollen tube growth is depressed so severely that the tubes never reach the ovary, whereas in S. campanulata (and possibly other LSI species) self pollen tube growth in the style is more active, with tubes reaching the ovary within 24 h, but, nevertheless, the relative delay in effecting self fertilizations may mean that such selfed pistils are already programmed to abscise.

ACKNOWLEDGEMENTS

N.S.B. Jr acknowledges financial support from the Fundo de Apoio ao Ensino e à Pesquisa (UNICAMP), and the CNPq. P.E.G. likewise acknowledges financial support from the CNPq with the award of a Pesquisador Visitante grant.

Received: 20 September 2002; Returned for revision: 9 December 2002; Accepted: 20 February 2003    Published electronically: 27 March 2003