Amer. J. Bot. 72(10): 1537-1543. 1985.

FUNCTIONAL DIOECISM IN THE : THE BREEDING SYSTEM OF SPACHEA MEMBRANACEA CUATR. KIME. STEINER Botany Department, University of California, Davis, California 95616

ABSTRACT The breeding system of Spachea membranacea (Malpighiaceae), an understory tropical rain forest tree, was studied on Barro Colorado Island in central . Two types of trees were found, those with hermaphroditic flowers and those with female flowers. Crossing experiments indicated that morphological hermaphrodites normally function as males but may be induced to set fruit if emasculated prior to cross-pollination. Thus, although morphologically gynodi- oecious, S. membranacea is functionally dioecious. This is the first experimental evidence for functional dioecism in the Malpighiaceae. The proportions of hermaphrodites (functional males) in the three populations examined were 44.7% (N = 47), 60.0% (N = lo), and 87.5% (N = 17). The distribution of hermaphrodites and females in size classes was quite similar, suggesting that sexual expression is not influenced by the age of the individual. Flowers are visited and presumably pollinated primarily by bees of the genera Paratetrapedia (Anthophoridae) and Trigona (Apidae).

RECENTbreeding system surveys in tropical W. R. Anderson (1977, 1981) was the first communities have shown that dioecism to note the existence of unisexuality in the Mal- is common among tropical trees (Bawa and pighiaceae. This large, predominantly Neo- Opler, 1975; Ruiz and Arroyo, 1978; Bawa, tropical family shows a wide range in growth 1980; Sobrevila and Arroyo, 1982). This re- forms and fruit types but a relatively constant alization has prompted a closer examination floral morphology ("floral conservatism" of W. of plant breeding systems in general and con- R. Anderson, 1979). Anderson's (1 98 1) study sideration of the selective factors involved in of herbarium collections led him to hypoth- the evolution of dioecism in particular (Bawa esize that Spachea elegans (G. F. W. Meyer) and Beach, 198 1; Baker, 1984; Charlesworth, Adr. Jussieu and S. tricarpa Adr. Jussieu, pre- 1984). Unfortunately, a plant's functional dominantly South American , are gyno- breeding system is not always accurately re- dioecious or functionally dioecious. Collec- flected by its floral morphology, a situation that tions with mature fruits had flat unopened is particularly evident in some species whose anthers, and with pollen-bearing anthers hermaphroditic flowers are functionally male never had enlarging fruits on old inflorescences (Arroyo and Raven, 1975; G. Anderson, 1979; (W. R. Anderson, 198 1). Both Croat (1978) Atsatt and Rundel, 1982). Thus, it is important and Cuatrecasas and Croat (1980) failed to to evaluate sexual expression on a functional mention the possibility of gynodioecism or rather than morphological basis as has been functional dioecism in the related Spachea emphasized by Lloyd (1980), Ross (1982), and membranacea Cuatr., which occurs in tropical Charlesworth (1984). moist forests of Panama and (Cua- trecasas and Croat, 1980). The purpose of this ' Received for publication 15 October 1984; revision study was to document whether populations accepted 30 April 1985. of S. rnembranacea on Barro Colorado Island I thank William R. Anderson for suggesting this study (BCI) in central Panama exhibited a floral di- and for comments on the manuscript. Steve Buchmann and an anonymous reviewer also made many helpful sug- morphism analogous to that observed by W. gestions on the manuscript. I am particularly grateful to R. Anderson (1981) and if so, to determine Robin Foster for providing unpublished data on the lo- whether they were functionally gynodioecious cation of S. membranacea in the 50 ha mapped plot on or dioecious. BCI and to Lynn Loveless for assistance in data collection. Roy Snelling and David Roubik kindly identified the bees, Spachea rnembranacea is a locally common and K. Douthit illustrated the flowers. Special thanks are understory tree (up to 10 m) on BCI that bears due to the Smithsonian Tropical Research Institute for small (1-cm diam) pink flowers on pendent permission to work on BCI and to the residents who made racemes during the early wet season (May- this the most enjoyable of all my visits. Present address: National Botanic Gardens of South Af- June). The flowers consist of five sepals, four rica, Kirstenbosch, Private Bag X7, Claremont 7735, Rep. of which bear paired oil-secreting glands (cf. of South Africa. elaiophores of Vogel, 1974; W. R. Anderson, AMERICAN JOURNAL OF BOTANY ' [Vol. 72

Fig. 1. Spachea membranacea. a-c. Female flower, gynoecium, nonfertile stamen; scale lines: a) 1 mm; b) 0.5 mm; c) 0.5 mm. d-f. Stamens, gynoecium, hermaphrodite flower; scale lines: d) 0.5 mm; e) 0.5 mm; f) 1 mm.

1979), five separate petals, 10 stamens, and an son 8, Shannon 13, and Van Tyne 12, were ovary that usually (but see Results) contains studied during May and June 1983. Plant two ovules (Fig. 1). voucher specimens of the individuals studied are at DAV, while bees have been deposited METHODSAND STUDY SITE-Barro Colorado at the Los Angeles County Museum. Island, the site for this study, is a 12 km2nature reserve situated in Lake Gatun, a part of the Floral morphology-Two types of trees were Panama Canal. Spachea membranacea is lo- found on BCI, those with morphologically her- cally common in the young forest (ca. 70 yr) maphroditic flowers with normal-looking on BCI (Steiner, pers. obs.) but is only occa- styles, ovules, and anthers bearing pollen, and sionally found in old forest (100f yr). One those with flowers that looked similar but lacked small population, referred to as the Plateau pollen-bearing anthers (females). Ovary and population, has been recorded from a mapped ovule size, style length, and stamen length were 50 ha plot in the old forest (Hubbell and Foster, measured on five flowers from four hermaph- unpubl.). This population, and portions of rodite and four female trees in the Pearson populations at BCI trail locations called Pear- population and from seven hermaphrodite and October, 19851 STEINER- DIOECISM IN MALPIGHIACEAE 1539 four females from other populations. Ratios of TABLE1. Spachea membranacea flower size (mm) com- hermaphrodite to female trees were deter- parisons within the Pearson population on BCZ.Five mined for the Plateau, Pearson, and Shannon flowers from each of four hermaphrodites and four populations, while diameters at breast height females were measured (DBH's) were recorded for the Plateau and Hermaphrodites Female F-ratio P Shannon populations. I Ovary length Experimental pollinations -Five pollina- 1.38 i 0.15 1.50 + 0.14 6.6 <.01 tion treatments were performed on hermaph- Ovule length roditic flowers. Flowers were left either un- 0.90 + 0.09 1.16 k 0.08 87.9 <.001 manipulated and unbagged (=open pollination) Style length or unmanipulated but enclosed in a fine mesh 1.68 i 0.14 1.55 +. 0.23 4.72 <.05 nylon bag, or self-pollinated and bagged, or Stamen length cross-pollinated after emasculation and bagged, 3.08 +. 0.21 1.83 i 0.17 425.2 <.001 or cross-pollinated without emasculation and bagged. Female flowers were either unmanipu- lated and bagged or cross-pollinated and bagged. one-loculate and ovulate flowers were found in low frequency on BCI. Five of 103 (4.9%) Floral visitors-Floral visitors were ob- female flowers examined had three carpels and served and collected primarily in conjunction ovules (3 of 13 trees), while only 1 of 3 10 (0.3%) with the pollination experiments. These ob- hermaphrodite flowers (1 of 20 trees) showed servations spanned the entire flowering period this character. Cases where the carpel and ovule and totaled about 20 hr but involved mainly number were reduced to one were found in 3 hermaphrodites at the Plateau, Pearson, and of 310 (1%) hermaphrodite flowers (3 of 20 Shannon sites. trees), but none were found in female flowers (0/103, 13 trees). RESULTS-Floral characteristics-Spachea The proportion of hermaphrodite and fe- membranacea flowers open between 0530 and male trees varied between populations. The 0630 local time. If left unpollinated, the entire Plateau population was strongly biased in favor hermaphrodite flower abscises after 1 day, but of hermaphrodites (87.5%, N = 16), while the only the petals ofthe female flower abscise after Pearson and Shannon populations had nearly the first day. The remainder of the female flow- equal proportions with slightly more her- er abscises after 2-3 days. An average her- maphrodites at Pearson (60°/o, N = 10) and maphrodite inflorescence has 47 flowers (N = more females at Shannon (55.39'0, N = 47). 30, range: 26-75) with a variable number of The distribution of hermaphrodites and fe- flowers opening each day over a 5-10 day pe- males in DBH size classes was quite similar riod. Individual trees flower for 3-4 wk, while (Fig. 2), suggesting that sexual expression is a population will flower for 4-5 wk. genetically fixed within an individual and is Hermaphrodite flowers have smaller ovar- not related to size or age. Many of the females ies and ovules, longer styles and stamens than examined had retained some fruits from the female flowers in the Pearson population (Ta- previous year, providing additional evidence ble 1). Except for style length (in the Plateau for sexual constancy. population females had longer styles than her- maphrodites), these differences were charac- Experimentalpollinations -Hermaphrodite teristic of other populations as well. The an- flowers normally last a single day before they thers and stigmas ofhermaphrodite flowers are abscise, but after emasculation and cross-pol- at the same position in the flower, but the stig- lination some flowers remained for 3-4 days. mas extend beyond the shorter sterile stamens Fruiting success was, therefore, not scored until in female flowers (Fig. 1). Although the ovules at least 1 wk following pollination. of hermaphrodite and female flowers look es- Results of the experimental pollinations show sentially the same, ovules from hermaphrodite that only emasculated hermaphrodites and fe- flowers (with one exception) were always male flowers that were cross-pollinated pro- smaller. duced fruit (Table 2). Cross-pollinated female In the generic description, Cuatrecasas and flowers resulted in significantly more fruit than Croat (1980) note that the ovary can be two- all other treatments, while emasculated, cross- or three-loculate and ovulate, but variability pollinated hermaphrodite flowers produced of this sort is not mentioned under the de- significantly more fruit than all treatments ex- scription of S. membranacea. Both three- and cept 4 and 7 (Table 2). This was probably due AMERICAN JOURNAL OF BOTANY [Vol. 72

hermaphrodite trees .female trees 80 - -

h 8 60- V 0" C - Q) u3 g 40 - I.L r -

20 -

niln 1. n rn I I I I I I 4.1- 6.1- 8.1- 10.1- 121- 14.1- 16.1- 18.1- 201- 221- 241- 26A- 281- 301- 321- 341- 361- 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 22.0 24.0 26.0 28.0 30.0 32.0 34.0 36.0 38.0

DBH Size Class (cm) Fig. 2. Size class distribution of femaleand hermaphroditetrees of Spachea membranacea in the Shannon population on BCI (N = 47).

to the small sample sizes used in Treatments havior of these two bees differs. P. calcarata 4 and 7. Even though the sample size for non- collects oil by scraping the calyx glands with emasculated, cross-pollinated hermaphrodites its special foreleg basitarsal scraper setae (Neff (Treatment 4) was small, consisting of flowers and Simpson, 198I), while T. pallens, which from a single tree, the same tree when emas- lacks the modified setae, scrapes the glands culated prior to crossing had the highest fruit with its mandibles (cf. Vogel, 1974).Both bees set (70°/o, N = 30) of the trees sampled. Vari- then transfer the oil to their hind scopae,where ability in the success of emasculated cross-pol- it is carried either alone or admixed with pollen linationsamonghermaphrodite trees was quite to the nest. high. Fruit setper tree ranged from 0-70% (X = Most pollinator observations took place in 15.79'0,SD = 26.8%, N = 7) with only three of the Plateau population where there were only the seven trees producing fruits. two poorly flowering female trees. Since no bees were observed to visit female flowers, it Floral visitors-Visitation rates to Spachea is not possible to compare foraging on the two membranacea were very low considering the floral morphs. The presence of oil in female largenumber ofavailable flowers. The two most flowers, however, suggests a reward sufficient commonly observed visitors to Spachea mem- to attract both P. calcarata and T. pallens. branacea flowers were Paratetrapedia calcar- The much largerbee, Centrislongimani (body ata Cresson (body length = 8.5 mm) and Tri- length = 21 mm) was observed systematically gona pallens Fabricius (bodylength = 6.0 mm), visiting hermaphrodite flowers once at the Pla- but Paratetrapedia xanthaspis Cockerrell, P. teau site and twice at the Shannon site. At the gigantea Schr., T. fuscipennis Friese, T. ful- Plateau site, a single bee visited flowers on the viventris Guerin, and Centris longimani Fabr. tallest and most heavily flowering tree over a also visited flowers. Both P. calcarata and T. 4-hr period, but it did not move to surrounding pallens collect pollen and oil from hermaph- trees, which were smaller and had fewer flow- rodite flowers; however, the oil collection be- ers.At the Shannon site, two Centris longimani October, 19851 STEINER-DIOECISM IN MALPIGHIACEAE 154 1

TABLE2. Crossing results for Spachea membranacea on Burro Colorado Island

Treatment Fru~tslflowers O/o Success No. of trees 1. Unmanipulated hermaphrodite 0/522 0.0 9 2. Self-poll. hermaphrodites 0/132 0.0 4 3. Open-poll. hermaphrodites 0/3054 0.0 6 4. Cross-poll. hem., no emasculation 0/10 0.0 1 5. Cross-poll. hem. + emasculation 33/184 17.9 7 6. Cross-poll. females 24/4 1 57.7 1 7. Unmanipulated females 0/8 0.0 1 Significance levels from Chi-square two-way contingency tables: Treatment 5 vs. Treatments 1, 2, 3, 6 P < .001 Treatment 5 vs. Treatment 4 P > .I0 Treatment 5 vs. Treatment 7 P > .I0 Treatment 6 vs. Treatments 1, 2, 3 P < .001 Treatment 6 vs. Treatment 4 P < .025 Treatment 6 vs. Treatment 7 P < .05 bees were observed collecting either pollen were observed developing on any nonmanip- (short visits) or oil (longer visits) from two ulated hermaphrodites, but until further com- hermaphrodite trees. parisons between emasculated and nonemas- culated cross-pollinations can be made, the DISCUSSION-The evolution of gynodioe- relative importance of emasculation and cross- cism is presumed to occur when a male sterile pollination cannot be firmly established. The mutant becomes favored in a population of lack of fruit development on open pollinated hermaphrodites (Ross, 1982). For this to oc- hermaphrodites argues against cross-pollina- cur, the seed fitness of the mutants (females) tion as the factor governing induced fruit de- must be twice that of the hermaphrodites to velopment, since fruits developed on open- compensate for the lack ofreproduction through pollinated females in the Pearson population pollen (Lewis, 194 1). The ability of male sterile where pollinators, even on hermaphrodites, mutants to spread through a population of her- were not commonly observed. maphrodites may be explained by the selective The mechanism by which emasculation in- advantage of increased outcrossing (Lloyd, fluences seed development remains unknown, 1982). Although selection for strict femaleness but the role ofhormones in the control of sexual has been intense in the evolution of gynodioe- expression is well established even though the cism (Lloyd, 1982), this has not been the case details vary from plant to plant. Removal of with female sterility in the evolution of sub- the stamens in hermaphrodite flowers of Spa- dioecism and dioecism. Selection for female chea membranacea must upset the hormonal sterility appears to be much less intense in her- balance responsible for the maintenance of fe- maphrodites of gynodioecious species, as many male sterility and thereby allow fruit devel- show substantial variation in their degree of opment to take place. female sterility (Arroyo and Raven, 1975; Lloyd, 1975; Webb, 1979; Atsatt and Rundel, Pollination -The low visitation rates by pol- 1982). The result is a spectrum of functional linators to Spachea membranacea may have or physiological types ranging from full female been the result of competition for pollinators fertility (functional gynodioecy) to total female from other flowering species and/or reduced sterility (functional dioecy), all within the mor- bee populations due to a long and unusually phological guise of gynodioecism. The prob- severe dry season. Bee populations on other lems of reconciling physiological and morpho- flowering Malpighiaceae species also appeared logical breeding system indicators within a rigid low compared to visitation rates in 1980 (Stei- system of classification is discussed by Bawa ner, unpubl.). and Beach (198 1). Most dioecious species of tropical forests are Spachea membranacea is at the functionally pollinated by small, generalist, social and sol- dioecious end of the gynodioecious spectrum itary bees in the families Halictidae, Megachili- as morphological hermaphrodites fail to set dae, and/or Apidae (Bawa, 1980). The flowers seed, functioning solely as males. Only through of the plants visited by these bees are corre- emasculation can one induce fruit formation sondingly small and lack morphological spe- from some "hermaphrodite" flowers. No fruits cializations that might suggest pollinator spec- 1542 AMERICAN JOURNAL OF BOTANY [VO~.72

ificity. Although Spachea flowers are also small, prising that the functional males have retained they possess the specialized oil-secreting calyx viable ovules and the ability to set fruit oc- glands typical of most Neotropical Malpighi- casionally (albeit with physiological modifi- aceae (Vogel, 1974). cation); however, this supports the notion that There is a close association between Neo- selection for female sterility has been much tropical Malpighiaceae and oil-collecting a?- less intense than selection for male sterility in thophorid bees in the genera Centris, Epz- the evolution of dioecism from gynodioecism charis, and Paratetrapedia (Vogel, 1974; Neff (Lloyd, 1982). and Simpson, 198 1; Simpson and Neff, 198 1, Two additional Spachea species have been 1983).Centris and Epicharis preferentially vis- examined and shown to be morphologically it canopy trees and lianas (Frankie et al., 1983), gynodioecious and probably functionally dioe- the predominant habit among Panamanian cious (W. R. Anderson, 1981), and in both Malpighiaceae (Cuatrecasas and Croat, 1980). cases, ovules are present in the "hermaphro- These medium to large bees (> 12 mm) are dite" flowers. Based on the drawings in An- rarely observed foraging in the forest under- derson (198 l), the size difference between the story (Steiner, unpubl.), which may explain why ovaries in hermaphrodite and female flowers they were not commonly observed on Spachea. in Spachea elegans is even greater than it is in Paratetrapedia species, which are much small- S. membranacea, yet it remains to be seen er bees (5-10 mm), visit a wide variety ofplants whether the physiological capability to set fruit with small flowers in forest understory as well has been retained in this and other species of as more open forest edge habitats (Steiner, pers. Spachea. Although one might expect different obs.). They also tend to spend longer on a par- Spachea species to occupy different positions ticular plant than more mobile larger bees. This on the morphological and/or functional spec- tendency to restricted foraging is typical of the trum between gynodioecy and dioecy, there is pollinators of many dioecious tropical trees no evidence to suggest that differentiation along (Bawa, 1980) and may explain why the only this axis has occurred. Apparently evolution- malpigs with unisexual flowers are understory ary divergence, which has resulted in the rec- trees that are presumably pollinated primarily ognition of distinct morphological species by these and Trigona bees. Malpighia rome- within the , has occurred without a con- roana Cuatr. (Malpighiaceae), an understory comitant divergence in breeding system. Clear- shrub on BCI with small pink flowers similar ly, a more detailed look at the functional breed- to those of Spachea, is also pollinated by Par- ing systems of the remaining Spachea species atetrapedia and Trigona spp.; however, it shows is in order. no evidence of unisexuality (Steiner, unpubl.). LITERATURE CITED Evolution of dioecism -One would predict, over evolutionary time, a divergence in the ANDERSON,G. J. 1979. Dioecious Solanum species of hermaphroditic origin is an example of broad con- morphologies of the sexual parts in female and vergence. Nature (London) 282: 836-838. hermaphrodite (functionally male) Spachea ANDERSON,W. R. 1977. Byrsonimoidae, anew subfamily flowers based on sexual selection and resource of the Malpighiaceae. Leandra 7: 5-18. allocation arguments (cf. Ross, 1982). With . 1979. Floral conservatism in Neotropical Mal- elimination of the male function in female pighiaceae. Biotropica l l : 2 19-223. flowers, ovules should become larger and/or . 198 1. The botany of the Guayana Highland- Part XI-Malpighiaceae. Mem. N.Y. Bot. Gard. 32: more numerous, and stamens should become 21-305. reduced. Functional male flowers should de- ARROYO,M. T. K., AND P. H. RAVEN. 1975. The evo- velop larger stamens with more pollen and lution of subdioecy in morphologically gynodioecious smaller and/or fewer ovules. Such is in fact the species of Fuchsia sect. Encliandra (Onagraceae). Evo- case for S. membranacea where female flowers lution 29: 500-5 1 1. have larger ovules and smaller stamens while ATSATT,P. R., AND P. RUNDEL.1982. Pollinator main- tenance vs. fruit production: partitioned reproductive functional male flowers have smaller ovules effort in subdioecious Fuchsia lycioides. Ann. Mo. Bot. and larger stamens. Gard. 69: 199-208. The difference between the two flower types BAKER,H. G. 1984. Some functions of dioecy in seed in Spachea membranacea is also evident from plants. Amer. Nat. 124: 149-158. the proportions of one- and three-carpelate BAWA,K. S. L980. Evolution of dioecy in flowering plants. flowers observed. The frequency of female Annu. Rev. Ecol. Syst. 11: 15-39. -, AND J. H. BEACH. 1981. Evolution of sexual flowers that had three carpels (three ovules) systems in flowering plants. Ann. Mo. Bot. Gard. 68: was much larger than that for functional males, 254-274. while only in functional males were cases of a -, AND P. A. OPLER. 1975. Dioecism in tropical single carpel (ovule) found. It is somewhat sur- forest trees. Evolution 29: 167-1 79. October, 19851 STEINER -DIOECISM IN MALPIGHIACEAE 1543

CHARLESWORTH,D. 1984. Androdioecy and the evolu- structures in the Anthophoridae (Hymenoptera): mor- tion of dioecy. Biol. J. Linn. Soc. 22: 333-348. phology, function and use in systematics. J. Kans. CROAT,T. B. 1978. Flora of Barro Colorado Island. Stan- Entomol. Soc. 54: 95-123. ford University Press, Stanford, Calif. Ross, M. D. 1982. Five evolutionary pathways to sub- CUATRECASAS,J., AND T. B. CROAT.1980. Malpighiaceae. dioecy. Amer. Nat. 1 19: 297-3 18. In R. E. Woodson, Jr., and R. W. Schery [eds.], Flora Ru~z,T. ZAPATA,AND M. T. K. ARROYO.1978. Plant of Panama. Ann. Mo. Bot. Gard. 67: 851-945. I reproductive ecology of a secondary deciduous trop- FRANKIE,G. W., W. A. HABER,P. A. OPLER,AND K. S. ical forest in Venezuela. Biotropica 10: 221-230. BAWA.1983. Characteristics and organization of t,he SIMPSON,B. B. 1983. Evolution and diversity of floral large bee pollination system in the Costa Rican dry rewards. In C. E. Jones and R. J. Little [eds.], Hand- forest. InC. E. Jones and R. J. Little [eds.], Handbook book of pollination biology, pp. 360-372. Van Nos- ofexperimental pollination biology, pp. 4 1 1-497. Van trand-Rheinhold, New York. Nostrand-Reinhold, New York. -, AND J. L. NEW. 198 1. Floral rewards: alternatives LEWIS,D. 194 1. Male sterility in natural populations of to pollen and nectar. Ann. Mo. Bot. Gard. 68: 301- hermaphrodite plants. New Phytol. 40: 56-63. 322. LLOYD,D. G. 1975. The maintenance of gynodioecy and SOBREVILA,C., AND M. T. K. ARROYO.1982. Breeding androdioecy in angiosperms. Genetica 45: 325-329. systems in a montane tropical cloud forest in Vene- . 1980. Sexual strategies in plants. 111. A quanti- zuela. Plant Syst. Evol. 140: 19-37. tative method for describing the gender of plants. N. VOGEL,S. 1974. Olblumen und Olsammelnde Bienen. Z. J. Bot. 18: 103-108. Trop. Subtrop. Pflanzenwelt 7: 1-267. -. 1982. Selection of combined versus separate sexes WEBB,C. J. 1979. Breeding systems and the evolution in seed plants. Amer. Nat. 120: 571-585. of dioecy in New Zealand apioid Umbelliferae. Evo- NEW, J. L., AND B. B. SIMPSON.1981. Oil-collecting lution 33: 662-672.