Systematics of the Cyclanthaceae, especially Sphaeradenia and Chorigyne
The Cyclanthaceae have an exclusively Neotropical distribution, and consist of perennial herbs, vines, and epiphytes. The leaves are distichous or spirally arranged, with a lamina that is entire (Ludovia), divided into four segments (Carludovica), or bifid (remaining genera). The plants are monoecious, and the inflorescence is a terminal or lateral, peduncled spadix, with spathes inserted on the peduncle. Pistillate and staminate flowers are situated in alternate cycles where individual flowers are not discernible (Cyclanthus), or in spirally arranged groups where the pistillate ones are tetramerous with a unilocular ovary (remaining genera). Most species prefer humid habitats at low and medium high altitudes, and the family forms a conspicuous part of the flora in such areas.
Among the genera treated by Harling (1958), the existence of the monotypic genus Pseudoludovia Harl. has been questioned in recent studies of the Cyclanthaceae (Wilder 1978, 1981; Wilder & Hammel 1989; Eriksson 1993). I examined the type of Pseudoludovia andreana, and concluded that this taxon is based on a mixed collection. This specimen includes a leaf of the recently described Ludovia bierhorstii G. Wilder and a young infructescence of Sphaeradenia laucheana (Mast.) Harl., mounted to simulate connection. The infructescence was selected as lectotype (Eriksson 1993), with the consequence that P. andreana is reduced to a synonym of S. laucheana, and Pseudoludovia is placed in synonymy under the earlier described genus Sphaeradenia.
The number of presently recognized species in the Cyclanthaceae totals
222 in 12 genera, including the recent taxonomic alterations in Sphaeradenia
(Eriksson 1995). The genera are generally well circumscribed and widely
accepted, while the number of species certainly will increase with future
The family Cyclanthaceae is usually placed in an order or superorder of its own, to express its isolated position (e. g. Harling 1958; Cronquist 1981; Thorne 1983, 1992; Dahlgren & al. 1985), and suggested relatives have been Araceae, Palmae, or Pandanaceae. Morphological, embryological, anatomical, and chemical data accumulated during the last decades indicate that the Cyclanthaceae are most closely related to Pandanaceae, and only distantly so to the other families (Harling 1958, Wilder 1976, Harris & Hartley 1980, French & al. 1983, Tomlinson & Wilder 1984, Dahlgren & al. 1985). The monophyly of the Cyclanthaceae, where Cyclanthus is the sister-group to the other genera, is well established (e. g. Harling 1958, French & al. 1983, Tomlinson & Wilder 1984, Hammel & Wilder 1989, Eriksson 1994b).
The evolution in the Cyclanthaceae has clearly taken place along two lineages, which Harling (1958) emphasized by a subdivision into two subfamilies, viz. Carludovicoideae and the monotypic Cyclanthoideae. Harling also constructed a genealogical tree of the family, and informally divided Carludovicoideae into the Asplundia group (Asplundia, Carludovica, Evodianthus, Dicranopygium, Schultesiophytum, and Thoracocarpus) and the Sphaeradenia group (Ludovia, Sphaeradenia, Stelestylis, and the now synonymized Pseudoludovia). In order to place their new genus Dianthoveus within the family, Hammel & Wilder (1989) used a cladistic approach, taking into account all genera in the Carludovicoideae except Chorigyne. Their analysis suggested that Dianthoveus is the sister-group to Evodianthus, and supported the monophyly of Harling`s informal groups, although with a different evolutionary scenario. It was evidently necessary to do a more elaborate phylogenetic analysis, based on all genera and a larger character set.
My phylogenetic analysis of the Carludovicoideae (Eriksson 1994b), using Cyclanthus as outgroup, is based on 63 characters, of which 38 are informative for determining relationships, and the rest for demonstrating monophyly of either the ingroup or the terminal taxa. As terminal taxa of the ingroup all genera were selected, except in Asplundia where the subgenera where used. A cladistic search using the Branch and Bound algorithm of PAUP 3.0 (Swofford 1989) yielded 18 equally parsimonious trees. In order to test the strength of individual clades, I used the Bootstrap procedure with 100 replications, and constructed consensus trees of all cladograms up to one or two steps longer than the minimum-length ones. These methods generated similar results.
The Sphaeradenia group (including the new genus Chorigyne) is monophyletic, and characterized by the following synapomorphies: predominantly epiphytes, distichous phyllotaxy, strongly birefringent walls of the parenchyma-like dead cells in the lamina, predominantly coriaceous lamina, subapical or apical placentation, and endosperm with thick cell walls. Furthermore, Ludovia is the sister-group to the other genera in the Sphaeradenia group, and Sphaeradenia is the sister-group to Stelestylis and Chorigyne. The Sphaeradenia group is thus acceptably resolved, with the outlined relationships supported by the strict consensus tree and the Bootstrap analysis, where these clades appear in almost all of the replications. This hypothesis is in accordance with the results by Harling (1958) and Hammel & Wilder (1989), with the addition of Chorigyne.
The Asplundia group, as presently circumscribed, is not supported
in my analyses, contrary to ones by Harling (1958) and Hammel & Wilder
(1989), and may be paraphyletic. Schultesiophytum is possibly the
sister-group to the other genera in the Carludovicoideae, as indicated by
the strict consensus tree and the Bootstrap procedure. A formal recognition
of the groups proposed by Harling (1958) is not justified under such circumstances.
Outside the Sphaeradenia group, the relationships are largely uncertain,
and only the Evodianthus-Dianthoveus clade is well supported.
Key to the genera
Because the generic delimitations in the Cyclanthaceae have partly changed since the last publication of a complete key (Harling 1958), this new one is given.
1. Leaf blades not plicate, lateral costae ending in apex; staminate
and pistillate flowers in alternate cycles (Cyclanthoideae) ... Cyclanthus
The Cyclanthaceae are one of the few exclusively cantharophilous families, and recent studies (e. g. Beach 1982; Gottsberger 1990, 1991; Eriksson 1994a) have indicated a specialized relationship regarding these plants and their pollinators.
Cyclanthus bipartitus (Cyclanthoideae) is pollinated by Cyclocephala spp. (Scarabaeidae), which arrive at the protogynous inflorescence when it is in the pistillate phase, and leave at the end of the staminate phase c. 24 hours later, covered with pollen (Beach 1982). While visiting the inflorescence, the pollinators consume specialized food tissue produced by the spathes, which also serve as shelter against predators and mating site. Pollinator exclusion tests have suggested that Cyclanthus is obligately xenogamous (Beach 1982).
The genera of the Carludovicoideae all have inflorescences with rather uniform functional morphology in terms of pollination biology. The inflorescences are protogynous in most studied species (e. g. Gottsberger 1990, 1991; Eriksson 1994a), and, during anthesis, the staminate flowers cover the pistillate ones, such that a pollination chamber is formed above the stigmas. Scent emitting staminodes, the main pollinator attractant (Gottsberger 1991; Eriksson 1994a), protrude between the staminate flowers, where also the small entrances to the pollination chambers are situated. The staminate flowers are usually shed soon after anthesis.
Previous observations (Harling 1958; Hammel 1984; Gottsberger 1990, 1991; Schatz 1990) have shown that the Carludovicoideae are pollinated by weevils (Curculionidae). They arrive during the pistillate phase, effect pollination while using the inflorescence for feeding, shelter, and mating, and leave it covered with pollen during the staminate phase. The occasional occurrence of weevil larvae in the ovaries of fruiting spadices has indicated that the visitors also use the inflorescence as oviposition site (Harling 1958).
My investigations of Sphaeradenia hamata and its pollinators, and scattered observations in other carludovicoid genera, confirm this generalized pollination pattern (Eriksson 1994a). Previous studies have, however, failed to recognize an important fact. The weevils use the inflorescence for oviposition, but the eggs are laid in the pedicel of the staminate flowers, and usually not in the ovaries. The inflorescence morphology of the Carludovicoideae restricts the spectrum of effective pollinators mainly to weevils, and the shape of the pollination chamber is such that oviposition is promoted in the staminate flowers. The last-mentioned adaptation is a very effective way of minimizing loss of invested energy for a plant that relies on being pollinated by ovipositing visitors, since the staminate organs are dispensable and not nourished after anthesis. Although cantharophily is sometimes viewed as a one-sided adaptation of flowers for the needs of visiting beetles (Gottsberger 1990), a co-evolved plant-pollinator relationship must be taken into consideration. Adaptations for being attracted to the carludovicoid inflorescence increase the probability of reproductive success for the weevils, since the inflorescence offers a location with high probability of finding mates, a safe oviposition site, and secured larvae development for many individuals at the same time.
Many species of the Carludovicoideae are probably xenogamous, this being
ensured by strict protogyny. Geitonogamy is with certainty known only from
cultivated individuals (Eriksson 1989, 1994a). It may occur in species where
the pistillate and staminate phases are overlapping, as observed by Gottsberger
(1991), and then probably promoted by weevils as pollen vectors.
Maintained by Roger Eriksson and last updated August 17, 2006.