American Journal of Botany 88(9): 1527±1534. 2001.

THRIPS POLLINATION OF ANDRODIOECIOUS ELASTICA () IN A SEASONAL TROPICAL FOREST1

SHOKO SAKAI2

Smithsonian Tropical Research Institute, Apartado 2072, AncoÂn, Balboa, Republic of Panama

Androdioecy is a rare sexual system in nature, as predicted theoretically. Among the androecious species reported so far, (Moraceae) is unique in that ¯owers are unisexual and staminate and pistillate ¯owers on cosexual are produced on different in¯orescences. In addition, in¯orescence structure of staminate in¯orescences on males and staminate and pistillate in¯ores- cences on cosexes is markedly different. Staminate in¯orescences on males are bivalvate, while staminate in¯orescences on cosexes are ``®g-like'' and urceolate. Pistillate in¯orescences are discoidal. The difference may re¯ect different roles and requirements of the three in¯orescences in pollination and protection from herbivores. This study reports thrips pollination of C. elastica, demonstrated by a pollinator introduction experiment. Thrips pollination of the species may be an example of mutualism originating from ± herbivore interactions. In the Moraceae, shifts from simple herbivores on ¯owers to pollinators might have occurred many times, evolving into diverse pollination systems including the ®g±®g wasp mutualism. The family, of which little is known about pollination systems, provides interesting and unique opportunities to study evolution of pollination systems and roles of nonpollinating associates of in¯orescences.

Key words: androdioecy; in¯orescence structure; Moraceae; Panama; pollination; thrips.

The in¯orescences of the Moraceae are extremely diverse, Here, I report on the pollination system of Castilla elastica ranging from simple racemes, capitula, and spikes to urceolate Sesse (Moraceae). Although the species was formally impor- (urn-shaped) in¯orescences and syconia (Berg, 1989; tant for rubber production and was cultivated in the late 1800s Rohwer, 1993). The most important trend of evolution of in- and early 1900s (Pittier, 1910; Hammel, 1986), its reproduc- ¯orescences may be progressive condensation by fusion and tive biology is poorly studied. Three forms of in¯orescences, reduction of ¯oral parts (Rohwer, 1993), which can be asso- the most peculiar characteristic of the species, were described ciated with reduction of ¯owers and change from bisexual to by Cook (1903a, b) for the ®rst time. Cook (1903a, b) and unisexual condition (Berg, 1989). This tendency appears to be Pittier (1910) examined living and preserved materials as well most tightly linked with protection of ¯owers from destructive as dry specimens and mentioned that the species was ``partly organisms that consume or cause infection of ¯oral tissues. dioecious'' with two types of trees within a population. There The interactions that minimize loss to such ¯oral parasites may were trees that bore only male in¯orescences, while others in some cases lead to mutualism (Berg, 1989). In the case of produced pistillate in¯orescences ¯anked by a few staminate Ficus, seed predators have evolved into pollinators of the plant in¯orescences, which were smaller than primary ones. The sta- (Bronstein and McKey, 1989; Herre et al., 1996; Anstett, Hos- mens on both types of trees were fertile (Pittier, 1910). In other saert-McKey, and Kjellberg, 1997). In integer, words, the species was androdioecious. male in¯orescences infected by fungi provide mycelia as a As theoretical studies have predicted (Charlesworth and reward for gall-midge pollinators (Sakai, Kato, and Nagamasu, Charlesworth, 1978) androdioecy (cosexes and males in a sin- 2000). It is remarkable, however, that little is known about gle population) is an extremely rare sexual system in plants pollination systems in the Moraceae other than Ficus and some (Charlesworth, 1984). In most androdioecious species reported apparently wind-pollinated species (Berg, 1989; Endress, so far, a cosexual individual produces ¯owers 1994; Rohwer, 1993). Several available studies suggested that (Datisca spp. [Datiscaceae] [Liston, Rieseberg, and Elias, various pollination systems are involved (Bawa et al., 1985; 1989; Swensen, Luthi, and Rieseberg, 1998], Fraxinus spp. Momose et al., 1998; Sakai, Kato, and Nagamasu, 2000) and and Phillyrea angustifolia [Oleaceae] [Lepart and DommeÂe, that wind-pollination might have evolved more than once (Za- 1992; Yamazaki, 1993; Hiura and Ishida, 1994; Ishida and pata and Arroyo, 1978; Bawa and Crisp, 1980; Williams and Hiura, 1998; DommeÂe et al., 1999], Neobuxbaumia mezca- Adam, 1993; Bullock, 1994). laensis [Cactaceae] [Valiente-Banuet et al., 1997], Oxalis suks- dor® [Oxalidaceae] [Ornduff, 1964, 1972], Saxifraga cernua 1 Manuscript received 10 November 2000; revision accepted 15 February [Saxifragaceae] [Molau and Prentice, 1992], and Schizopepon 2001. bryoniaefolius [Cucurbitaceae] [Akimoto, Fukuhara, and Ki- The author thanks S. Joseph Wright and David W. Roubik for invaluable support throughout the study at STRI; Smithsonian Tropical Research Institute kuzawa, 1999]). A few plant species show ``monoecious'' cos- for the use of the canopy crane; Laurence A. Mound, Serguei V. Triapitsyn, exes with pistillate and staminate ¯owers on the same in¯o- and John M. Heraty for identi®cation; Vibeke Horlyck for arrangement rescences (Ricinocarpos pinifolius and Mercurialis annua [Eu- of my crane work; Amots Dafni for information about pollen staining; Allen phorbiaceae] [Thomson et al., 1989; Pannell, 1997], and Sag- Herre, Hidetoshi Nagamasu, Makoto Kato, and two anonymous reviewers for ittaria lancifolia [Alismataceae] [Muenchow, 1998]). constructive comments on the manuscript. This study was supported by JSPS Postdoctoral Fellowships for Research Abroad and Young Scientists. Androdioecy of C. elastica is unique in that staminate and 2 Current address: Graduate School of Human and Environmental Studies, Kyo- pistillate ¯owers occur on different in¯orescences of a co- to University, Sakyo, Kyoto, Japan 606±8501 (e-mail: [email protected]). sexual plant. In addition, staminate in¯orescences on male and 1527 1528 AMERICAN JOURNAL OF BOTANY [Vol. 88

Figs. 1±4. Castilla elastica. 1. Branches of a ¯owering male tree. 2. Primary staminate in¯orescence, from which many thrips ¯ew out when it was shaken by breeze (ϫ2). 3. A pistillate in¯orescence visited by thrips (Frankliniella diversa)(ϫ2.7). 4. Complemental staminate in¯orescence (ϫ2.7). cosexual plants and pistillate in¯orescences are strikingly dif- 1910; Berg, 1972). This deciduous tree can be 30 m in height. The natural ferent from one another in structure. Pistillate in¯orescences distribution of the species ranges from through Panama and the coast- are discoidal with a shallow central depression, in which pistils al region of western Colombia and western Ecuador, although the species was are aggregated (Fig. 3). Staminate in¯orescences on cosexes widely spread through the tropics as a result of introduction and cultivation are urceolate, ``®g-like'' (Cook, 1903b) with an apical hole for its during the late 1800s and early 1900s (Pittier, 1910; Berg, 1972; covered by imbricate bracts (Fig. 4), while staminate in¯ores- Hammel, 1986). Among the three subspecies recognized by Berg (1972), only cences on males are bivalvate (Fig. 2). Because of this varia- C. elastica subsp. costaricana (Liebmann) C. C. Berg is known from Panama. tion and the peculiarity of in¯orescence structure and the - I observed 13 reproductively mature trees within reach of the tower crane ual system, the pollination system of Castilla was considered (ϳ1 ha) in April±July 2000 (Table 1). of particular interest.

TABLE 1. Sex and diameter at breast height (dbh) of the trees observed MATERIALS AND METHODS in this study.

TerminologyÐIn this paper, I term staminate in¯orescences on male trees Plant ID Sex dbh (cm) ``primary staminate in¯orescences'' and those on cosexes ``complemental sta- minate in¯orescences'' following Cook (1903a, b), Pittier (1910), and Berg C1 Cosex 28.8 (1972). C2 Cosex 15.9 C3 Cosex 20.9 C4 Male 15.8 Study site and plantÐThe study was conducted in a seasonally dry forest C5 Male 49.0 in the Parque Natural Metropolitano near Panama City, Panama. Annual rain- C6 Cosex 33.1 fall at the site is 1740 mm on average, most of which occurs during the rainy C7 Cosex 15.9 season from May through December. The forest is ϳ80-yr-old second growth C8 Male 15.9 of 40 m in height. I used a 42 m tall tower crane with a 51-m jib at a C9 Cosex 24.0 permanent installation in the park (Parker, Smith, and Hogan, 1992) to reach C10 Male 14.6 tree canopies. C11 Male 21.6 Castilla elastica Sesse (Moraceae) is an abundant tree species in secondary C12 Cosex 18.2 C13 Cosex 18.0 forests and open areas such as clearings and forest edges of the region (Pittier, September 2001] SAKAIÐTHRIPS POLLINATION OF CASTILLA (MORACEAE) 1529

Observation and collection of ¯ower visitorsÐFlowers and ¯ower visitors test, P Ͼ 0.05, Table 1). On a single node of male tree branch- on staminate and pistillate in¯orescences of Castilla were observed from 10 es, about four primary staminate in¯orescences are borne to- April to 15 May for 20 h in total, including 4 h at night. Flower visitors on gether (Fig. 1). They are ¯abellate opening in a slit, or bi- staminate ¯owers were collected by sampling the whole in¯orescence. valvate, ϳ2.3 cm wide and 1.3 cm thick with a 0.5-cm pe- on pistillate ¯owers were collected with an aspirator. The insects were pre- duncle (Fig. 2). A primary staminate in¯orescence is fully served in 50% alcohol or AGA solution (60% alcohol, glacial acetic acid, and packed with ϳ400 ¯owers, each bearing a single stamen, on glycerin in the proportions 10:1:1) or in dried state and stored at the Smith- the internal surface. Anthers are ϳ1.2 mm in length and small- sonian Tropical Research Institute. Some were identi®ed to insect family and er than those in complemental in¯orescences (ϳ1.6 mm in to species in the case of adult thrips or used to examine pollen load under a length). On a branch of a cosexual tree, one or rarely two binocular microscope. To estimate density of the thrips, the numbers of thrips pistillate in¯orescences are borne on each node. The in¯ores- on pistillate in¯orescences were counted on the four trees, C1, C3, C6, and cences are almost sessile, shallow cup-shaped or discoidal with C7, and numbers of thrips on staminate in¯orescences on two trees, C5 and C7. a shallow central depression, 1.0 cm thick and 1.5 cm in di- ameter (Fig. 3). Approximately 16 pistillate ¯owers on an in- Sticky trapsÐI used sticky traps (TAT Fly Paper; Walco-Linck, Valley ¯orescence are fused with each other and with the receptacle Cottage, New York, USA) to compare population densities and activities of at the base. At the center of the in¯orescence, styles of the thrips among different sites. The traps catch insects on the sticky surfaces of ¯owers are aggregated. The ovary is 1-locular with a single a yellowish-brown paper (70 cm ϫ 4.7 cm). The trap has no odor and does ovule. Pistillate in¯orescences are usually accompanied by two not contain insecticide. To compare thrips density among sites, I placed traps or more complemental staminate in¯orescences. Complemen- for 24 h in tree canopies at the height of 10±25 m and in the understory 1.5 tal staminate in¯orescences are urceolate with a small apical m above the ground. The canopy traps included two traps on each of four opening, 1.2 cm long, and 1.0 cm in diameter with a peduncle Castilla trees (two cosexual trees, C1 and C6, and two male trees, C4 and of ϳ1.3 cm (Fig. 4). The opening is usually blocked by im- C5) and four traps at least 10 m away from any plants ¯owering intensively bricate bracts, but sometimes bracts are loosened during ¯ow- as a control. Understory traps were placed just below each of the canopy ering, and a small hole becomes visible. A staminate in¯ores- traps. Therefore, one trapping procedure involved 24 sticky traps, and trapping cence encloses ϳ50 staminate ¯owers, each with a single sta- was conducted twice on 27 and 28 April. men. Pollen grains are ϳ13 ␮m in a diameter with three or rarely four pores. Pollen grains of male and cosexual trees Pollination experimentÐI conducted two pollination experiments on cos- each stained red by methylene blue/phloxine-B solution, in- exual trees. In Experiment 1, I compared fruit set of in¯orescences enclosed dicating existence of cytoplasm in both, and there was no dif- in a bag of 0.2-mm mesh, which allowed the entry and exit of tiny insects ference between them under a compound microscope. I could such as thrips, with that of open-pollinated control on C3. Flowers on the not ®nd other rewards for pollinators besides pollen on sta- in¯orescences were receptive around 22 April. minate in¯orescences. Experiment 2 conducted on C1 involved four treatments: (1) bagged with ®ne cloth; (2) hand-pollinated; (3) thrips-introduced; and (4) open-pollinated The 13 C. elastica trees under observation ¯owered contin- control. In the former three treatments, in¯orescences were bagged with cloth uously or intermittently from April to May 2000. Most of their ®ne enough to exclude thrips before the in¯orescences became receptive. In leaves had been shed before the ¯owering period (Fig. 1), addition, in¯orescences were hand-pollinated with pollen of male tree C5 which was followed by an explosive leaf ¯ush. Primary in¯o- using a Chinese writing brush in treatment 2, and thrips collected from C5 rescences turned from green to yellow when they opened and were introduced into two bags enclosing ®ve in¯orescences each in treatment anther dehiscence began. Within 2 d all the anthers on an 3. Thrips for the treatment were collected from ®ve staminate in¯orescences, in¯orescence dehisced, and in¯orescences then abscised. In then released in a bag enclosing pistillate in¯orescences under the treatment. cosexes, pistillate and staminate in¯orescences were in ¯ower This procedure was estimated to introduce 247 Ϯ 82 thrips (98.9% of all simultaneously, and temporal segregation was not observed. introduced) and 2.8 Ϯ 2.2 other arthropods (1.1%) (N ϭ 10). Pistillate in¯orescences changed color from green to creamy Hand-pollination and thrips introduction were conducted twice on the same yellow as their stigma extended and probably became recep- in¯orescences on 9 and 11 May. tive (Fig. 3). Stigmas sometimes remained wet and fresh for Fruit set was examined on 10 June by dissecting infructescences to count Ͼ10 d. The receptive period of an in¯orescence may depend ®nished ¯owers and developing seeds in both experiments. At that time, in- on fertilization, since all the stigmas withered and turned fructescences (fruiting head) were still green but of almost mature size (4.3 brown Ͻ2 d after hand-pollination. Most pistillate in¯ores- Ϯ 0.4 cm in diameter, N ϭ 15). Differences in numbers of retained infruc- cences developed into infructescences, which reached maturity tescences and in fruit set of the in¯orescences between different treatments and turned salmon pink in July. The ¯owering period of com- were statistically examined using Fisher's exact test and a t test, respectively, plemental staminate in¯orescences seemed to be several days, using SAS (SAS, 1988). while it was dif®cult to examine since anther dehiscence could not be observed without damaging the in¯orescence. The in- Pollen stainingÐTo examine existence of cytoplasm in pollen grains, pol- ¯orescences had usually abscised 1 d after their color len grains from frozen ¯ower samples were dyed with methylene blue/phlox- ϳ ine-B solution. Phloxine-B stains cytoplasm of pollen red, and methylene blue changed from green to yellow. stains the exine blue (A. Dafni, University of Haifa, Israel, personal com- munication). Flower visitorsÐThe most abundant ¯ower visitors were two species of thrips, Frankliniella diversa and F. insularis, RESULTS both on staminate and on pistillate in¯orescences (Table 2). They were observed actively feeding on pollen grains on sta- Floral biologyÐTwo types of sex expression occurred with- minate in¯orescences. Many thrips emerged from primary sta- in a population of Castilla elastica: cosexual trees producing minate in¯orescences when the branch was shaken by a breeze both pistillate and staminate in¯orescences and male trees with (Fig. 2). On pistillate in¯orescences, thrips just stayed, some- staminate in¯orescences. Diameter at breast height (dbh) of times moving around (Fig. 3). Thrips accounted for 99.0% of the trees did not differ signi®cantly between the two (t ¯ower visitors (N ϭ 1320) collected on male trees and for 1530 AMERICAN JOURNAL OF BOTANY [Vol. 88

TABLE 2. Insects collected on in¯orescences of Castilla elastica.

Cosexual tree Male tree Order Family Pistillate in¯orescence Staminate in¯orescence Staminate in¯orescence Species C1 C2 C6 C7 C1 C7 C9 C12 C13 C5 C8 C10 C11 Total Thysanoptera Thripidae Frankliniella diversa Hood 2 8 10 39 12 47 81 56 384 19 271 170 1099 F. insularis Franklin 1 2 8 25 7 6 20 36 63 8 65 29 270 Sp. 1 2 2 Immature individuals 8 23 68 20 1 139 15 117 22 413 Subtotal of Thripidae 3 2 16 10 72 42 121 121 93 588 42 453 221 1784 Hymenoptera Apidae 1 1 Formicidae 1 1 Eulophidae 1 3 23 9 Vespidae 1 1 Coleoptera Curculionidae 2 1 3 Staphylinidae 3 3 Total 4 2 16 10 72 42 122 121 93 595 42 455 228 1802

99.8% (N ϭ 450) and 96.9% (N ϭ 32) on staminate and pis- ¯orescences were estimated to be 0.9 Ϯ 1.3 (N ϭ 59), 6 Ϯ tillate in¯orescence of cosexes, respectively. Their sex ratio 4.5 (N ϭ 7), and 202.8 Ϯ 16.0 (N ϭ 6), respectively. was biased in favor of female (80%, N ϭ 1079 in F. diversa; Other than thrips, wasps (Ceranisus americensis Girault, 91%, N ϭ 35 in F. insularis), some of which had a well- Eulophidae, Hymenoptera), which are parasitoids of thrips developed ovary. All thrips collected on male trees (N ϭ 55) (Triapitsyn and Headrick, 1995), and ants (Formicidae) were were observed to carry pollen grains. Seventy-one percent (N observed to visit both pistillate and staminate in¯orescences, ϭ 51) and 66% (N ϭ 44) of thrips collected on staminate and but only infrequently. Ants remained on the outside of primary pistillate in¯orescences of cosexual trees, respectively, had staminate in¯orescences and on the pistillate in¯orescences, pollen loads. The numbers of pollen grains on thrips collected catching the thrips as they entered or left the in¯orescence. from primary staminate in¯orescences were much higher (25.8 Beetles (Curculionidae and Staphylinidae, Coleoptera) and Ϯ 46.2 grains, N ϭ 25) than on those from complemental bees (Apidae and Halictidae, Hymenoptera) were occasionally in¯orescences (7.6 Ϯ 13.7, N ϭ 51). Thrips densities on pis- observed visiting ¯owers for pollen, only on primary stami- tillate in¯orescences, complemental, and primary staminate in- nate in¯orescences. High thrips activities around C. elastica were con®rmed by sticky traps (Fig. 5). In the canopy traps, the numbers of thrips trapped in male trees were 33 times as high as the control. Thrips trapped in cosexual-tree canopies were also more abun- dant than the control, while much less than those in male trees (Fig. 5). Thrips on understory traps did not show much dif- ference among sites. The numbers of arthropods other than thrips were almost constant at all sites both in the forest can- opy and in the understory. Pollination experimentsÐIn experiment 1, all the in¯ores- cences of both treatments, open-pollinated and bagged with mesh, were retained after ¯owering (Table 3). However, fruit set of the bagged in¯orescences was signi®cantly lower than and about half of that in the open-pollinated control (P ϭ

TABLE 3. Results of two pollination experiments on Castilla elastica.

In¯orescence Flowers Treatment N Retained (%) N Fruit set (%) Experiment 1 Control 8 100.0 120 90.8 Bagged with mesh 6 100.0 103 47.6 Experiment 2 Control 21 100.0 356 90.2 Bagged with ®ne cloth 19 0.0 Ð Ð Fig. 5. Mean Ϯ 1 SE of the numbers of thrips and other arthropods on Hand-pollinated 10 100.0 148 87.2 sticky traps in (canopy) and below (understory) the canopy of male and cos- Thrips-introduced 10 100.0 177 80.2 exual trees, and controls in the forest canopy and understory. September 2001] SAKAIÐTHRIPS POLLINATION OF CASTILLA (MORACEAE) 1531

0.005, t test, one-tailed). In experiment 2, in¯orescences in¯orescences changed color into yellow while ¯owering, al- bagged with ®ne cloth without additional manipulation pro- though it was green before and after ¯owering. duced no fruits, while all the in¯orescences of the other three Thrips, agricultural and horticultural pests causing damage treatments were retained (Table 3, P Ͻ 0.0001 for all three by feeding on plant tissues, are often found in ¯owers of many comparisons, Fisher's exact test, one-tailed). Fruit set of in¯o- wild plants (Kevan and Baker, 1983; Kirk, 1984). In a tropical rescences that were hand-pollinated and had thrips introduced forest of Panama, 85% of 285 species were inhabited by ¯oral were not signi®cantly different from those of the open-polli- thrips (F. Gattesco, the University of Milan, Italy and D. W. nated control (P Ͼ 0.05, t test, two-tailed). Roubik, Smithsonian Institution, personal communication). Thrips feed on ¯oral tissue, nectar and exposed liquids, and DISCUSSION pollen grains in ¯owers (Kirk, 1984). Although they are thought to rarely contribute to pollination, recent studies have Pollination systemÐAlthough some species of Moraceae reported thrips as primary pollinators in many plant species of are reported to be wind-pollinated, wind-pollination is unlikely several families including Annonaceae (Webber and Gottsber- in C. elastica. Staminate ¯owers were enclosed in bivalvate ger, 1995; Momose, Nagamitsu, and Inoue, 1998), Araceae or urceolate in¯orescences, and stamens were not exposed to (Rust, 1980), Lauraceae (Norton, 1984), Winteraceae (Thien, wind. Pistils were aggregated in the central depression of the 1980; Pellmyr et al., 1990), and Zamiaceae (Mound and Terry, pistillate in¯orescence, thus they may not catch airborne pollen 2001). Pollen grain size in plants pollinated by thrips is Ͻ34 effectively. ␮m, which corresponds well to grain size of C. elastica (F. Beetles and bees visited staminate in¯orescences on male Gattesco and D. W. Roubik, personal communication). trees for pollen, but they were not observed on pistillate or Host speci®city of thrips varies greatly among species, and complemental staminate in¯orescences. Pistillate in¯orescenc- true host relationships of thrips species have rarely been es- es or complemental staminate in¯orescences closed by imbri- tablished (Mound and Marullo, 1996). Frankliniella is a large cate bracts may not be attractive for them due to absence or thrips of ϳ180 species, 90% of which occur in the Neo- inaccessibility of pollen. In addition to thrips, ants and para- tropics. Its species diversity is especially high in Central sitoids of thrips visited both pistillate and staminate in¯ores- America. For most species, the number of individuals collect- cences. Although these insects might pollinate C. elastica ac- ed so far is quite small. For example, among 44 species of cidentally in rare cases, they are unlikely to visit C. elastica Frankliniella recorded from Costa Rica, 18 are known only ¯owers that lack their prey, thrips, and they could not be the from a single sample or individual, and only ten are known principal pollinators, considering their low visit frequencies. from six or more samples (Mound and Marullo, 1996). Frank- I conclude that Castilla elastica is pollinated by thrips for liniella diversa, the most abundant thrips on C. elastica, was the following reasons. First, thrips were observed and collect- described based on a single female specimen from Panama; a ed on all the three types of in¯orescences, primary and com- second female was collected in Costa Rica. These two females plemental staminate in¯orescences and pistillate in¯orescenc- are the only known specimens of this species prior to the pre- es, and accounted for 96.9±99.8% of the ¯ower visitors. Pres- sent study, despite extensive collections in Central America ence of thrips both in staminate and in pistillate in¯orescences (Mound and Palmer, 1992; F. Gattesco and D. W. Roubik, un- was also observed by Pittier (1910). Second, a high proportion published data). The rarity might re¯ect host speci®city of the of the thrips carried pollen on the body. All the thrips collected species. A large population of F. diversa was also found on on male trees, and two-thirds of thrips on pistillate in¯ores- cences had pollen loads. Third, the results of the pollination in¯orescences of (Moraceae) (S. Sakai, unpublished experiment 1 suggest that pollen vectors included very small data), thus they might be Moraceae specialists. On the other insects, because some ¯owers bagged with mesh developed hand, F. insularis has been frequently collected in Panama and into fruits. Finally, in experiment 2, thrips-introduced in¯ores- Costa Rica, especially from ¯owers of Bignoniaceae. A few cences enclosed in a bag of ®ne cloth showed fruit set as high individuals of this species were also collected from ¯owers of as hand-pollinated ones and the open-pollinated control. High Bignoniaceae and Apocynaceae at the study site, when Cas- fruit set (80.2%) cannot be explained if thrips did not pollinate, tilla was ¯owering (S. Sakai, unpublished data). although small numbers (1.1%) of insects other than thrips Extremely high fruit set in open-pollinated ¯owers (Table were accidentally introduced together. 3), which is no lower than hand-pollinated ones, suggests the Apparently, thrips visited staminate ¯owers for pollen. They focal population of C. elastica with rather high density of were observed actively feeding on pollen. They probably cop- ¯owering trees (13 trees in 1 ha) received enough pollination ulated and reproduced on staminate in¯orescences, because service. However, it is unknown how tree density affects pol- many thrips larvae as well as male and female adults were lination success and fruit set. Distance of thrips movement in found on staminate in¯orescences, and some females had de- forests has not been estimated yet. Momose, Nagamitsu, and veloped ovaries. On the other hand, it is unclear why thrips Inoue (1998) reported thrips pollination of an understory tree, visited pistillate ¯owers, which had no substantial reward. Pis- Popowia pisocarpa (Annonaceae), and suggested that limited tillate in¯orescences did not have soft tissue other than pistils, movement of thrips caused lower fruit set of isolated trees. On and there was no clear evidence that thrips fed upon and dam- the other hand, Appanah and Chan (1981) argued that thrips aged any part of the in¯orescences. Thrips might be attracted visiting ¯owers of emergent trees moved for long distances by odor and color of the pistillate in¯orescences that mimic between trees, carried by a wind, and effectively pollinated six those of staminate ones. Scent is thought to be a primary at- species of emergent trees (Shorea, Dipterocarpaceae). Obser- tractant for thrips (Kirk, 1985). Although I could not detect vations in isolated islands and on a boat far from the land odor of pistillate ¯owers, volatile released by ¯owering in¯o- suggest that some thrips could travel Ͼ30 km (Lewis, 1973). rescences remains to be examined. Color may also be an im- In C. elastica, low thrips densities on the forest ¯oor under- portant cue for the pollinators in C. elastica, because pistillate neath the tree canopies suggested that thrips rarely dropped 1532 AMERICAN JOURNAL OF BOTANY [Vol. 88 with abscised staminate in¯orescences. They might be dis- complemental staminate in¯orescences were present in much persed by wind from canopy branches. lower numbers than in primary in¯orescences. Pollen grains were also less abundant on thrips from complemental in¯o- Sexual systemÐCastilla elastica is androdioecious with rescences, probably due to imbricate bracts blocking the en- cosexes and male plants within a population. Pollen grains trance of the in¯orescences. Outstanding colonizing ability of were not different between male and cosexual plants, and ex- C. elastica is evidenced by trees growing in the wild out of istence of cytoplasm was con®rmed in pollen grains from both its original distribution dispersed by cultivated mother trees primary and complemental staminate in¯orescences. However, (Berg, 1972). Producing seeds through self-pollination must constancy of gender expression over more than one ¯owering be essential when they invade a new habitat without conspe- season is still unclear. Pittier (1910) suggested that younger ci®c trees. In addition, for C. elastica it may be important to trees blooming for the ®rst time might bear invariably male assure a high level of pollination, because selective abscission in¯orescences, but simple size-dependent gender switching of unfertilized ovaries (i.e., sterile fruits) is impossible due to does not seem to occur, considering that no clear size differ- fusion of all pistillate ¯owers on a discoid in¯orescence. All ence between male and cosexual trees was found. ¯owers on a developing infructescence were observed to grow Theoretical studies predict that for males to invade a cos- into fruits, some of which were without a seed. It may be exual population, pollen success of males must be at least advantageous to maintain high fruit set through self-pollination twice as high as cosexes; furthermore, high outcrossing rates rather than to produce many seedless fruits as a part of an and strong inbreeding depression are needed (Lloyd, 1975; infructescence. Charlesworth and Charlesworth, 1978). High outcrossing rates, inbreeding depression, and pollen production by male plants In¯orescence structureÐUrceolate in¯orescences occur in observed in another androdioecious species, Datisca glomer- at least three genera of three different tribes of the Moraceae: ata (Fritsch and Rieseberg, 1992; Rieseberg et al., 1993; Phil- Castilla in , Sparattosyce in Artocarpeae, and Ficus brick and Rieseberg, 1994) agree with the prediction. In ad- in Ficeae. They are considered to have evolved independently dition, Akimoto, Fukuhara, and Kikuzawa (1999) showed for protection of reproductive organs from insects destroying dominance of sel®ng in hermaphrodite populations and sig- reproductive organs (Corner, 1962; Berg, 1989). Although in- ni®cant correlation between frequency of males and inbreeding ¯orescences, or syconia, of Ficus are clearly different from coef®cients in androdioecious populations of Schizopepon those of Castilla in that monoecious Ficus have both staminate bryoniaefolius. In C. elastica, male trees may have higher pol- and pistillate ¯owers on a single in¯orescence, the pollination len success than cosexes, as indicated by higher thrips density system of C. elastica shares some common characters with on primary in¯orescences than complemental ones and much Ficus. Firstly, they are pollinated by small insects, which enter higher thrips activities in male canopies than those of cosexes, urceolate in¯orescences through small holes. Host speci®city as shown by data from sticky traps. Quantifying outcrossing of the pollinators may also be high. Next, the pollinators breed ratio and inbreeding depression, together with sex ratio and in the in¯orescences. Finally, both pollination systems are pos- sexual allocation, of C. elastica would be interesting problems sibly derived from herbivory at the in¯orescence level. for further studies. Interestingly, only complemental in¯orescences, rather than A distinctive characteristic of androdioecy in C. elastica is pistillate ones, have urceolate structure and, hence, the most that male and female ¯owers are produced on different in¯o- closed structure in C. elastica. In contrast, the primary sta- rescences. Thus male and female functions do not share any minate in¯orescences are opened by slits. This difference may structure. Additionally, the shape of staminate in¯orescences be due to different requirements for pollinator activities. For on male and cosexual plants is quite different. Androdioecy in male trees, it is essential to attract and release as many thrips C. elastica might have evolved from dioecy. Among three as possible, while smaller complemental staminate in¯ores- Castilla species, closely related C. elastica and C. ulei are cences may function primarily to ensure self-pollination if out- androdioecious, but C. tunu is strictly dioecious without com- cross pollen is unavailable. The closed structure of a comple- plemental in¯orescences (Berg, 1972). mental in¯orescence may protect pollinators as well as pollen. What is the advantage for cosexual plants of having sta- When there is no male tree nearby, density of pollintor thrips minate in¯orescences? Although functionally pistillate ¯owers may be low. Therefore, a cosexual tree should increase and of some species are known to produce inviable pollen to attract maintain a thrips population on its own staminate in¯ores- pollinators (reviewed by Mayer and Charlesworth, 1991), pol- cences throughout the ¯owering period to successfully produce len concealed in urceolate complemental staminate in¯ores- seeds through self-pollination mediated by thrips. Predators of cences of C. elastica is unlikely to serve as an attractant. In thrips such as ants and spiders were frequently observed out- C. elastica, production of staminate in¯orescences causes dis- side of staminate in¯orescences, although parasitoids were also advantages for cosexes in three different ways. In the ®rst found inside. In addition, the closed structure of staminate in- place, they should reduce production of pistillate ¯owers and ¯orescences promotes thrips visits to pistillate in¯orescences fruits as some resources are allocated to staminate in¯ores- prior to staminate ones. On the other hand, discoid pistillate cences. Second, there is the possibility that outcross pollen on in¯orescences are protected by the thick receptacle, rather than pollinator bodies is lost in a complemental staminate in¯ores- closed structure. This open structure may facilitate thrips vis- cence if thrips from other trees enter a staminate in¯orescence its. before a pistillate one. Furthermore, complemental staminate It is clear that both pollination and herbivory play important in¯orescences may increase inbreeding. roles in the evolution of in¯orescence structure of the Mora- I suggest that the most likely explanation for retention of ceae, although systematic relationships within the family and male function in cosexual C. elastica is reproductive conti- their pollination systems are poorly understood (Berg, 1989; nuity in colonists. Pollen of complemental staminate in¯ores- Humphries and Blackmore, 1989). In the course of evolution, cences must rarely contribute to outcrossing, because thrips in shifts from simple ``¯oral predators'' to pollinators might oc- September 2001] SAKAIÐTHRIPS POLLINATION OF CASTILLA (MORACEAE) 1533 cur several times. These pollinators are represented by thrips systematics, and fossil history of the Hammamelidae, vol. 2, 267±277. in this study, the ®g-wasps, that are both pollinators and seed Clarendon Press, Oxford, UK. predators of Ficus (Janzen, 1979; Wiebes, 1979), and gall ISHIDA, K., AND T. H IURA. 1998. Pollen fertility and ¯owering phenology in an androecious tree, Fraxinus lanuginosa (Oleaceae), in Hokkaido, Ja- midges pollinating Artocarpus integer, in which pollinator lar- pan. International Journal of Plant Sciences 159: 941±947. vae feed on fungi infecting a staminate in¯orescence (Sakai, JANZEN, D. H. 1979. How to be a ®g. Annual Review of Ecology and Sys- Kato, and Nagamasu, 2000). Some studies on different plant tematics 10: 13±51. groups suggest that preadaptations played essential roles in KEVAN,P.G.,AND H. G. BAKER. 1983. Insect as ¯ower visitors and polli- drastic changes of pollination systems, and the original func- nators. Annual Review of Entomology 28: 407±453. tion of these traits were often related to protection from de- KIRK, W. D. J. 1984. Pollen feeding in thrips (Insecta: Thysanoptera). Journal of Zoology 204: 107±117. structive ¯ower visitors (Pellmyr and Thompson, 1992; Pell- KIRK, W. D. J. 1985. Effects of some ¯oral scents on host ®nding by thrips myr et al., 1996; Armbruster, 1997). The Moraceae may pro- (Insecta: Thysanoptera). Journal of Chemical Ecology 11: 35±43. vide interesting and unique opportunities to study the roles of LEPART, J., AND B. DOMMEÂ E. 1992. Is Phillyrea angustifolia L. (Oleaceae) nonpollinators in shaping the evolution of pollination systems. an androdioecious species? Botanical Journal of the Linnean Society 108: 375±387. LEWIS, T. 1973. Thrips: their biology, ecology and economic importance. LITERATURE CITED Academic Press, New York, New York, USA. LISTON, A. L., L. H. RIESBERG, AND T. S. ELIAS. 1989. Datisca glomerata is functionally androdioecious. Nature 343: 641±642. AKIMOTO, J., T. FUKUHARA, AND K. KIKUZAWA. 1999. Sex ratios and genetic LLOYD, D. G. 1975. The maintenance of gynodioecy and androdioedy in variation in a functionally androdioecious species, Schizopepon bryon- angiosperms. Genetica 45: 325±339. iaefolius (Cucurbitaceae). American Journal of Botany 86: 880±886. MAYER,S.S.,AND D. CHARLESWORTH. 1991. Cryptic dioecy in ¯owering ANSTETT, M. C., M. HOSSAERT-MCKEY, AND F. K JELLBERG. 1997. Figs and plants. Trends in Evolution and Ecology 6: 320±325. ®g pollinators: evolutionary con¯icts in a coevolved mutualism. Trends MOLAU, U., AND H. PRENTICE. 1992. Reproductive system and population in Ecology and Evolution 12: 94±99. structure in three arctive Saxifraga species. Journal of Ecology 80: 149± APPANAH, S., AND H. T. CHAN. 1981. Thrips: the pollinators of some dip- 161. terocarps. Malaysian Forester 44: 234±252. MOMOSE, K., A. HATADA,R.YAMAOKA, AND T. I NOUE. 1998. Pollination ARMBRUSTER, W. S. 1997. Exaptations link evolution of plant±herbivore and biology of the genus Artocarpus, Moraceae. Tropics 7: 165±172. plant±pollinator interactions: a phylogenetic inquiry. Ecology 78: 1661± MOMOSE, K., T. NAGAMITSU, AND T. I NOUE. 1998. Thrips cross-pollination 1672. of Popowia pisocarpa (Annonaceae) in a lowland dipterocarp forest in BAWA, K. S., S. H. BULLOCK,D.R.PERRY,R.E.COVILLE, AND M. H. Sarawak. Biotropica 30: 444±448. GRAYUM. 1985. Reproductive biology of a rain forest in Costa Rica. American Journal of Botany 72: 346±356. MOUND,L.A.,AND R. MARULLO. 1996. The thrips of Central and : an introduction (Insecta: Thysanoptera). Associated Publishers, BAWA,K.S.,AND J. E. CRISP. 1980. Wind-pollination in the understory of a rainforest in Costa Rica. Journal of Ecology 68: 871±876. Gainesville, Florida, USA. MOUND,L.A.,AND J. M. PALMER. 1992. Thrips of Panama: a biological BERG, C. C. 1972. Olmedieae, Brosimeae (Moraceae). Flora Neotropica 7: 1±228. catalogue and bibliography (Thysanoptera). In D. Quintero and A. Aiello [eds.], Insects of Panama and , selected studies, 321±328. BERG, C. C. 1989. Systematics and phylogeny of the Urticales. In P. R. Crane and S. Blackmore [eds.], Evolution, systematics, and fossil history of the Oxford Univeristy Press, Oxford, UK. Hammamelidae, vol. 2, 193±220. Clarendon Press, Oxford, UK. MOUND,L.A.,AND I. TERRY. 2001. Thrips pollination of the central Aus- BRONSTEIN,J.L.,AND D. MCKEY. 1989. The ®g±pollinator mutualism: a tralian Macrozamia macdonnellii (Cycadales). International Journal of model system for comparative biology. Experientia 45: 601±604. Plant Sciences 162: 147±154. BULLOCK, S. H. 1994. Wind pollination of neotropical dioecious trees. Bio- MUENCHOW, G. E. 1998. Subandrodioecy and male ®tness in Sagittaria lan- tropica 26: 172±179. cifolia subsp. lancifolia (Alismataceae). American Journal of Botany 85: CHARLESWORTH, B., AND D. CHARLESWORTH. 1978. A model for the evo- 513±520. lution of dioecy and gynodioecy. American Naturalist 112: 975±997. NORTON, S. A. 1984. Thrips pollination in the lowland forest of New Zealand. CHARLESWORTH, D. 1984. Androdioecy and the evolution of dioecy. Biolog- New Zealand Journal of Ecology 7: 157±164. ical Journal of the Linnean Society 23: 333±348. ORNDUFF, R. 1964. The breeding system of Oxalis suksdor®i. American Jour- COOK, O. F. 1903a. Four new species of the Central American rubber tree. nal of Botany 51: 307±314. Science, New series 18: 436±439. ORNDUFF, R. 1972. The breakdown of trimorphic incompatibility in Oxalis COOK, O. F. 1903b. The culture of the Central American rubber tree. Gov- section Corniculatae. Evolution 26: 52±65. erment Printing Of®ce, Washington, D.C., USA. PANNELL, J. 1997. Variation in sex ratios and sex allocation in androdioecious CORNER, E. J. H. 1962. The classi®cation of Moraceae. Gardens' Bulletin Mercurialis annua. Journal of Ecology 85: 57±69. Singapore 19: 187±252. PARKER, G. G., A. P. SMITH, AND K. P. HOGAN. 1992. Access to the upper DOMMEÂ E, B., A. GESLOT,J.D.THOMPSON,M.REILLE, AND N. DENELLE. forest canopy with a large tower crane. BioScience 42: 664±670. 1999. Andromonoecy in the entomophilous tree Fraxinus ornus (Ole- PELLMYR, O., L. B. THIEN,G.BERGSTORM, AND I. GROTH. 1990. Pollination aceae). New Phytologist 143: 419±426. of New Caledonian Winteraceae: opportunistic shifts or parallel radiation ENDRESS, P. K. 1994. Diversity and evolutionary biology of tropical ¯owers. with their pollinators? Plant Systematics and Evolution 173: 143±157. Cambridge University Press, Cambridge, UK. PELLMYR, O., AND J. N. THOMPSON. 1992. Multiple occurrences of mutual- FRITSCH,P.,AND L. H. RIESEBERG. 1992. High outcrossing rates maintain ism in the yucca moth lineage. Proceedings of the National Academy of male and hermaphrodite individuals in populations of the ¯owering plant Sciences, USA 89: 2927±2929. Datisca glomerata. Nature 359: 633±636. PELLMYR, O., J. N. THOMPSON,J.BROWN, AND R. G. HARRISON. 1996. Evo- HAMMEL, B. E. 1986. The vascular ¯ora of La Selva Biological Station, Costa lution of pollination and mutualism in the yucca moth lineage. American Rica, Moraceae. Selbyana 9: 243±259. Naturalist 148: 827±847. HERRE, E. A., C. A. MACHADO,J.NASON,D.M.WINDSOR,S.S.MC- PHILBRICK,C.T.,AND L. H. RIESEBERG. 1994. Pollen production in the CAFFERTY,W.VAN HOUTTEN, AND K. BACHMANN. 1996. Molecular androecious Datisca glomerata (Datiscaceae): implications for breeding phylogenies of ®gs and their pollinator wasos. Journal of Biogeography system equilibrium. Plant Species Biology 9: 43±46. 23: 521±530. PITTIER, H. 1910. A preliminary treatment of the genus Castilla. Contribu- HIURA,T.,AND K. ISHIDA. 1994. Reproductive ecology in Fraxinus lanugi- tions from the United States National Herbarium 13: 247±249. nosa. I. breeding structure. Transactions of the Hokkaido Branch of the RIESEBERG, L. H., C. T. PHILBRICK,P.E.PACK,M.A.HASON, AND P. Japanese Forest Society 42: 58±60 (in Japanese). FIRTSCH. 1993. Inbreeding depression in androdioecious populations of HUMPHRIES,C.J.,AND S. BLACKMORE. 1989. A review of the classi®cation Datisca glomerata (Datiscaceae). American Journal of Botany 80: 757± of the Moraceae. In P. R. Crane and S. Blackmore [eds.], Evolution, 762. 1534 AMERICAN JOURNAL OF BOTANY [Vol. 88

ROHWER, J. G. 1993. Moraceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich species of the thrips-attacking genus Ceranisus Walker (Hymenoptera: [eds.], Flowering plants, dicotyledons: magnoliid, hamamelid and cary- Eulophidae). Transactions of the American Entomological Society Phil- ophyllid families, 438±453. Springer Verlag, Heidelberg, Germany. adelphia 121: 227±248. RUST, R. W. 1980. Pollen movement and reproduction in Arisaema triphyl- VALIENTE-BANUET, A., A. ROJAS-MARTIÂNEZ,M.C.ARIZMENDI, AND P. D AÂ VI- lum. Bulletin of the Torrey Botanical Club 107: 539±542. LA. 1997. Pollination biology of two columnar cacti (Neobuxbaumia SAKAI, S., M. KATO, AND H. NAGAMASU. 2000. Artocarpus (Moraceae)±gall mezcalaensis and Neobuxbaumia macrocephala) in the Tehuacan Valley, midge pollination mutualism mediated by a male-¯ower parasitic fungus. Central Mexico. American Journal of Botany 84: 452±455. American Journal of Botany 87: 440±445. WEBBER,A.C.,AND G. GOTTSBERGER. 1995. Floral biology and pollination SAS. 1988. SAS/STAT user's guide. Statistics, version 6. SAS Institute, Cary, of Bocageopsis multi¯ora and Oxandra euneura in Central Amazonia, North Carolina, USA. with remarks on the evolution of stamens in Annonaceae. Feddes Re- pertorium 106: 515±524. SWENSEN, S. M., J. N. LUTHI, AND L. H. RIESEBERG. 1998. Datiscaceae WIEBES, J. T. 1979. Coevolution of ®gs and their insect pollinators. Annual revised: monophyly and the sequence of breeding system evolution. Sys- Review of Ecology and Systematics 10: 1±12. tematic Botany 23: 157±169. WILLIAMS, G., AND P. A DAM. 1993. Ballistic pollen release in Australian THIEN, L. B. 1980. Patterns of pollination in the primitive angiosperms. Bio- members of Moraceae. Biotropica 25: 478±480. tropica 12: 1±13. YAMAZAKI, T. 1993. Tribe 4: Fraxineae. In K. Iwatsuki, T. Yamazaki, D. E. THOMSON, J. D., K. R. SHIVANNA,J.KENDRICK, AND R. B. KNOX. 1989. Boufford, and H. Ohba [eds.], Flora of Japan, vol. IIIa (Angiospermae, Sex expression, breeding system, and pollen biology of Ricinocarpos Dicotyledoneae, Sympetalae [a]). Kodansha, Tokyo, Japan (in Japanese). pinifolius: a case of androdioecy? American Journal of Botany 76: 1048± ZAPATA,T.R.,AND M. T. K. ARROYO. 1978. Plant reproductive ecology of 1059. a secondary deciduous tropical forest in Venezuela. Biotropica 10: 221± TRIAPITSYN,S.V.,AND D. H. HEADRICK. 1995. A review of the Nearctic 230.