| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Anatomy and Morphology |
2Departamento de Ecología Funcional, Instituto de Ecología, A.P. 70-275, Universidad Nacional Autónoma de México (UNAM), México D.F. 04510, México; 3Institute of Molecular Biosciences, Massey University, Private Bag 11-222, Palmerston North, New Zealand; 4Departamento de Biología, Facultad de Ciencias, UNAM, A.P. 70-356, México D.F. 04510 México; 5Department of Botany, The Natural History Museum, Cromwell Road, London SW7 5BD, UK and Micromorphology Group, Jodrell Lab., Royal Botanical Gardens, Kew, Richmond, Surrey TW93A13, UK; 6Departamento de Botánica, Instituto de Biología, UNAM, México D.F. 04510, México
Received for publication May 16, 2005. Accepted for publication August 25, 2005.
ABSTRACT
The enigmatic monocotyledon family Triuridaceae is composed of inconspicuous mycoheterotrophs, that have been variously interpreted. We present here the first report of a thorough floral developmental series for any member of the Triuridaceae. The two known Mexican Triuridaceae species were studied with anatomical sections and scanning electron microscopy. While both species have ephemeral and reduced radially symmetric flowers arising in a counterclockwise spiral on racemose inflorescences, Lacandonia schismatica is hermaphroditic with a central androecium and Triuris brevistylis is dioecious. Tepals are connately fused at their bases, and during development the subapical caudal tips continue to elongate, while at maturity the tepals are reflexed. Carpel primordia develop centrifugally from compound primordia in both species, with contrasting androecium development. In Lacandonia schismatica stamen and carpel primordia arise from a common precursor. The two species differ in tepal and carpel number and timing of organ development. This paper provides a developmental framework to understand floral characters in the Triuridaceae. Notably, we addressed if L. schismatica and T. brevistylis bear true flowers or pseudanthia, and our data support the former. The role of particular genes in determining the floral developmental patterns studied and the evolutionary significance of these patterns are discussed.
Key Words: flower development • homeosis • Lacandon Forest (Chiapas, Mexico) • Lacandonia • Lacandoniaceae • mycoheterotrophs • pseudanthia • Triuridaceae
The basic ground plan for all bisexual flowers in the angiosperms consists of sterile perianth organs (sepals and petals or tepals) on the outside followed by male reproductive organs (stamens) and then female reproductive organs (carpels) in the center. The flowers of the mycoheterotrophic monocot species, Lacandonia schismatica (Martínez and Ramos, Triuridaceae (Lacandoniaceae): Pandanales; APG, 2003
) have been interpreted as an anomaly to this fixed floral Bauplan of the angiosperms. The taxonomic description of this taxon implied the presence of homeotic flowers, in which a group of central stamens (mostly three) is surrounded by a multicarpellate gynoecium (Martínez and Ramos, 1989
). This reproductive morphology is mostly fixed among stable and healthy populations of the species that have been monitored since its discovery 20 years ago in the Lacandon rainforest of southeastern Mexico.
Lacandonia schismatica is endemic to Mexico, but it grows in proximity to a second triurid, the species Triuris brevistylis, that grows at higher altitudes than L. schismatica and extends to other neotropical areas. The areas of distribution for both species seem to be associated with an ancient lake (Vergara-Silva et al., 2003b
). Interestingly, populations of normally dioecious Triuris brevistylis also include a few individuals with heterotopic hermaphroditic flowers (Vergara-Silva et al., 2003b
) suggesting that a common ancestor of the two Mexican triurids might have had flowers with inverted reproductive whorls.
The members of Triuridaceae (Gardner) are all mycoheterotrophic achlorophyllous perennial herbs mostly found in the New and Old World tropics (Giesen, 1938
; Maas and Rübsamen, 1986
; Rübsamen-Weustenfeld, 1991
). They range in stature from 5 cm to 1 m, though most are not longer than 15 cm and similar to many mycoheterotrophs with the vegetative and reproductive parts extremely reduced (Leake, 1994
) and thus hard to interpret.
After the original interpretation of the L. schismatica flower as a homeotic structure, alternative hypotheses regarding the position of the reproductive whorls have been considered, referring either to a possible secondary torsion of the stamens toward the center or to the flower actually representing a reduced inflorescence (pseudanthia) (Stevens, 1991
). Along with the original taxonomic description of the species, a series of studies (Márquez-Guzmán et al., 1989
, 1993
; Vázquez-Santana et al., 1998
) has described the reproductive anatomy, ovule and seed development, and fertilization of L. schismatica. The continuity between the epidermal layer of the stamens and the central part of the receptacle found in these studies refuted the hypothesis of secondary torsion of the stamens in L. schismatica (Márquez-Guzmán, 1989
), but it did not provide evidence against a pseudanthial interpretation.
Recently, it has been proposed that the floral structures of L. schismatica and of triurids in general may not be flowers, but pseudanthia, in which clusters of highly reduced male and female flowers together appear as a single flower (Rudall, 2003
). In such a case, the central position of L. schismatica stamens could then be accounted for through reduction in inflorescences that originally bore male flowers in the distal part and female ones more proximal. Results of the phylogenetic analyses performed by Chase et al. (2000)
based on molecular matrices provided some support to this alternative interpretation. These analyses placed the triurid genus Sciaphila as the sister group of Freycinetia, one of the three extant genera of Pandanaceae (the screw pines). An important morphological diagnostic feature of Pandanaceae is the presence of pseudanthia (Stone, 1972
; Claßen-Bockhoff, 1990
). According to Chase et al. (2000)
, Triuridaceae and Pandanaceae compose a clade together with the families Cyclanthaceae, Stemonaceae, and Velloziaceae, which corresponds to a newly circumscribed order Pandanales (APG, 2003
).
The pseudanthial interpretation of the reproductive axes of the Triuridaceae is mainly based on structural grounds (Rudall, 2003
), particularly anatomical and morphological analyses of adult reproductive units of triurids compared to pseudanthial inflorescences of selected members of Pandanaceae and Cyclanthaceae. Rudall (2003)
has stated as the main tenet of her argument that the structures of Triuridaceae have, in general, elongated receptacles and vasculature patterns similar to those found in pseudanthial axes. In this paper, we directly address the hypothesis of triurid pseudanthia, also on structural grounds. We interpret our results using the framework and terminology of classical morphology.
For our purpose, we present extensively documented developmental series for the reproductive axes of L. schismatica and its close relative, T. brevistylis. We have chosen to describe these axes as bearers of true flowers, but to avoid circularility, we have looked in our data for features that would support viewing "the flowers" as pseudanthia. After this critical analysis, we have concluded that the reproductive structures of both taxa have, in accordance with the original interpretation, a evanthial character. Conceived in this way, the "inside-out," putatively true flowers of L. schismatica could have originated from a homeotic mutation (Stevens, 1991
; Mabberley, 1997
; Vergara-Silva et al., 2003b
). The developmental series provided in this paper complement previous accounts (Rübsamen-Weustenfeld, 1991
), are also useful to clarify floral morphological character states in other genera of Triuridaceae, and could constitute the basis for schematic models of the expression patterns of developmental genes in reproductive axes of Pandanales.
MATERIALS AND METHODS
Inflorescences of Lacandonia schismatica and Triuris brevistylis were collected over 10 years. Inflorescences were collected at Frontera Corozal for L. schismatica and at Nahá and El Censo for T. brevistylis. All locations are found in the Lacandon rain forest (see Vergara-Silva et al., 2003b
for precise locations and description of sites). Samples were fixed in the field at 4°C in FAA (4% formaldehyde, 2% acetic acid, 50% ethanol). Samples were left at 4°C for 36 h, then dehydrated through a graded ethanol series. In preparation for SEM, samples were critically point dried in CO2, mounted on aluminum stubs, dissected, and coated with gold-palladium. Some samples were analyzed on a Cambridge 360 SEM (Cambridge Instruments, Cambridge, UK) at 10 kV accelerating voltage. Others were observed on a JEOL 5310LV (JEOL Ltd., Tokyo, Japan) or Hitachi S3500N SEM (Hitachi High-Technologies Corp., Tokyo, Japan) at 10–15 kV.
For stained sections, some FAA fixed samples were placed in a graded ethanol–Histoclear (National Diagnostics, Atlanta, Georgia, USA) series and then a graded Histoclear–paraplast series. Samples were embedded in Paraplast (TYCO Healthcare Group, Mansfield, Massachusetts, USA) and 8–10 µm sections were cut on a rotary microtome. The paraplast sections were cleared in Histoclear, brought gradually into an aqueous solution, and then stained with toluidine blue (pH 5.2). The sections were rinsed in water, gradually brought into Histoclear, and mounted in Permount (Fisher Scientific, Pittsburgh, Pennsylvania, USA). Additional flowers were fixed in 5% glutaraldehyde–4% paraformaldehyde in 0.1 M s-collidine buffer at pH 7.2, dehydrated in graded ethanol series, and embedded in JB-4 (Polysciences Inc., Niles, Illinois, USA) or Epon 812 (Shell Chemicals, Houston, Texas, USA) resins. Sections were cut at 1–2 µm and stained with toluidine blue. Histological sections were observed with an Olympus Provis AX70 microscope (Olympus Corp., Tokyo, Japan).
RESULTS
General growth conditions and morphology of the Mexican Triuridaceae
The herbaceous L. schismatica and T. brevistylis are found on decaying logs or in wet leaf litter where the forest floor is generally free of green seedlings. These plants tend to grow in clumps. As long as the forest floor remains moist, L. schismatica flowers from August to January and T. brevistylis flowers from August to November. The roots and rhizomatous mat of this latter species is creeping, and evidence of vegetative propagation has been found. The aboveground parts of both species (10–20 cm tall) are erect and white to amber in color (Figs. 1, 38, 39). These achlorophyllous plants appear to have an inconspicuous stem with reduced leaves (acaulescent), and the aboveground parts consist mostly of the inflorescence (Figs. 1, 3, 38, 39).
|
|
Floral organogenesis in L. schismatica
Perianth organogenesis
The floral meristem is dome shaped and surrounded by a bract (Fig. 7). At this stage, the bilayered tunica and the multilayered corpus are conspicuous (Fig. 7). Upon initiation of the six tepal primordia, the floral meristem becomes broadly convex (Fig. 8). The first tepal primordium develops opposite the bract (Figs. 8, 9). The tepal primordia develop asynchronously on the flanks of the floral meristem in a single whorl, where the tepals opposite the stamens develop slightly faster compared to the ones in alternate positions (Figs. 10, 11). The tepals are slightly triangular in shape, connately fused at their bases, and grow to completely cover the bud when the stamen primordia begin to develop (Figs. 10–12).
|
|
|
|
In each gynoecial fascicle that radiates towards the periphery of the flower meristem, the carpel primordia develop ovaries in a synchronous manner (Figs. 22–24). Each single carpel primordium develops a nucellus with basal placentation to form a unicarpellar ovary (Figs. 23, 24). The ovule curvature proceeds in the opposite direction to the carpel curvature (Fig. 25). The carpel primordia acquire an obovoid shape as the style develops subapically (Figs. 25, 31). Stigmas are not observed.
|
|
). Double fertilization occurs at or close to this stage (Márquez-Guzmán et al., 1993
). As the tepals open, ovules have been fertilized (Figs. 31–35). The style is persistent and the anthers remain unopened (Figs. 32–35). Flowers at anthesis have ca. 40–80 unicarpellar fruits (Fig. 33). Each fruit has a seed with copious endosperm and reduced embryo (Fig. 35; Márquez-Guzmán et al., 1993
; Vázquez-Santana et al., 1998
).
When the indehiscent dehydrated fruits are mature, they fall from the dome-shaped floral receptacle (Figs. 36, 37), and the style and epidermis of the pericarp collapse (Vázquez-Santana et al., 1998
). In the oldest flowers, the stamens are persistent (Figs. 36, 37). After fruit detachment, collapsed pollen exine walls can be seen inside the unopened anthers (data not shown; Vázquez-Santana et al., 1998
).
Inflorescence development of Triuris brevistylis
The aboveground parts of female T. brevistylis plants (8–15 cm) tend to be taller than the male plants (10 cm at the most; data not shown). The unisexual flowers of T. brevistylis are radially symmetric, and female (Fig. 38) and male flowers (Fig. 39) each have three tepals that are basally connate with long subapical caudal tips. However, the caudal tips of the male flower (6.5–9 mm) tend to be longer than those of the female flower (1–4 mm). The first female flower tends to be larger than subsequent flowers on the inflorescence, while all male flowers on the same inflorescence tend to be more equal in size (Figs. 38, 39).
Female and male racemose inflorescences of T. brevistylis arise terminally (Figs. 38–41). This triurid species has 1–4 flowers spirally arranged around the inflorescence (Fig. 41). The pedicels are well developed and range in length from 1– 5 mm (Figs. 38, 39, 41, and data not shown). New floral meristems arise in a counterclockwise spiral, each enclosed within its own bract (Figs. 40–42, 45). The bract completely covers the floral meristem and then sheathes the pedicel. The inflorescence axis and pedicel of T. brevistylis is ridged or striated, and the diameters of both axes are similar to each other, ranging from 0.5 to 1 mm (Figs. 40, 41).
|
|
|
|
|
Androecial organogenesis in T. brevistylis
In the staminate flowers, the three stamens alternate with the tepals (Fig. 79; earlier stages were unavailable). The anther primordia are basifixed on short filaments during microspore mother cell development and when the anther wall is four-layered (Figs. 68– 70). The anther wall cells rapidly differentiate towards epidermis and endothecium, and thickenings of the latter cell layer become apparent (Fig. 69). At this stage, the three distinct short filaments of the stamens arise from the center of the flower, where a small, still unelongated receptacle can be found (Figs. 70, 71). The three bithecate anthers begin to diverge from each other through massive meristematic activity (Figs. 72–75) mainly along axial and radial axes, giving rise to an androphore, which appears as a greatly elongated, central structure with the shape of a winged pyramid with three faces alternate to the anthers. The androphore cells are mainly parenchymatous (Figs. 72 and 76).
The filaments appear to become assimilated into the androphore, suggesting that the meristematic activity comes from the base of the filaments and connective tissue (Figs. 70–74). Therefore, the bases of the filaments seem to begin to expand upward and outward to develop the androphore, and the filaments postgenitally fuse to each other at their edges and in the center and expand upward. The center of the flower where the filaments meet expands upward more quickly than their fused edges (Figs. 72–74, 77). The area where the three filaments fused will appear as ridges on the androphore until late in development. In this manner, the anthers, which hang directly on the sides of the growing androphore, are oriented outwards, but through continued upward growth of the androphore, the anthers eventually hang down (Figs. 72–75). The functional extrorse dehiscence of the anthers is through longitudinal slits (Figs. 77–80). In the abaxial face of the anthers, cavities are formed by epidermal cells of the filament base and the receptacle epidermis (Figs. 75, 79, 80). Papillate and secretory epidermis cells line these three cavities (Fig. 75 and data not shown).
DISCUSSION
Developmental and structural anatomical studies on the Mexican triurids: the nature of reproductive axes in Triuridaceae
Systematic background
The reproductive axes of Triuridaceae have been traditionally regarded as flowers, organized in terminal bracteate racemes (Giesen, 1938
). The original description of the Mexican species L. schismatica depicted a mycoheterotrophic plant of triurid affinity, whose putatively true flowers had an apparent homeotic interconversion involving stamens and carpels (Martínez and Ramos, 1989
). On the basis of this character state, a monotypic family Lacandoniaceae was erected. Subsequent publications on the structural anatomy and population-level morphological variation in L. schismatica (Márquez-Guzmán et al., 1989
, 1993
; Vázquez-Santana et al., 1998
; Vergara-Silva et al., 2003b
), as well as posterior rejections of the validity of Lacandoniaceae by other taxonomists (Rübsamen-Weustenfeld, 1991
; Maas-van de Kamer, 1995
; Maas-van de Kamer and Weustenfeld, 1998
), have followed a euanthial interpretation of its reproductive structures. In contrast, other authors have considered the plausibility of a different morphological description of the reproductive structures of triurids. Stevens (1991)
expressed the possibility of viewing the putative flowers of L. schismatica as pseudanthia to explain their organ arrangement. However, this author argued that at the time there was no evidence to support this interpretation. On the basis of new structural analyses, Rudall (2003)
has argued that the floral units of Triuridaceae may be "hidden pseudanthia," where sets of very reduced individual flowers appear as the floral organs. At the time of its publication, this hypothesis had indirect support from molecular systematics research, which had overturned decades-old taxonomic assumptions for Triuridaceae within the monocots (Chase et al., 2000
).
The family Triuridaceae has been of problematic phylogenetic affinity since its first description (reviewed in Maas and Rübsamen, 1986
). The taxonomic group is even more intriguing now that the oldest monocot fossil flowers have been ascribed to it (Gandolfo et al., 1998
; Gandolfo et al., 2002
). Currently, Triuridaceae is included in the order Pandanales, along with the families Stemonaceae, Velloziaceae, Cyclanthaceae, and Pandanaceae (APG, 2003
). Reproductive axes of the latter two families have been unequivocally interpreted as pseudanthia (Harling, 1958
; Stone, 1972
; Claßen-Bockhoff, 1990
), while those of Stemonaceae and Velloziaceae have never been proposed to have undergone processes of reduction (Kubitzki, 1998a
, b
). The family-level taxonomic circumscription for Pandanales was based predominantly on the broad phylogenetic analyses for the monocots by Chase et al. (2000)
. Reasonably, the retrieval of a sister group relationship between Triuridaceae and Pandanaceae in those analyses has inspired the reconsideration of a pseudanthial hypothesis for triurids. We must acknowledge, though, that subsequent studies (Vergara-Silva et al., 2003a
; Davis et al., 2004) did not corroborate the previous hypothesis of relationships for Triuridaceae and Pandanaceae.
Pseudanthia could still occur in triurids, even if the group is not sister to Pandanaceae. This distribution of character states would imply two at least partially independent evolutionary events, converging on a putatively homologous construction of reproductive structures. Hypotheses about the relationships among the families of Pandanales are not based on the congruence of morphological character states; they are based on analyses of molecular matrices. The proposal of triurid pseudanthia put forward by Rudall (2003)
is primarily based on structural evidence of mostly adult structures. Therefore, attempts to refute this proposal call for a specific consideration of detailed developmental information.
Evidential support for the pseudanthial and euanthial hypotheses
The structural anatomy argument of Rudall (2003)
stems from two separate, although mutually reinforcing, groups of observations. In the first group, the morphology of adult or almost adult flowers from several triurid taxa was analyzed exclusively in order to suggest a pseudanthial nature of the structures without reference to nontriurid taxa. The second set of observations was comparative and included an explicit reference to female pseudanthia from selected members of Pandanaceae and Cyclanthaceae. For the sake of her hypothesis, Rudall underscored that in monoecious monocot species, the basal part of the inflorescence is female and the distal one is male (Tomlinson, 1982
). Under the assumption of organ and internode reduction, this spatial location of flowers in an inflorescence would be consistent with the reproductive structure described for L. schismatica. If the floral units of this species were reduced inflorescences, the male structures would be located in the center (the apical part) and the female ones in the periphery (the basal part). However, in T. brevistylis we have found individuals with both male and female flowers, and the female are sometimes terminal rather than basal (data not shown). Rudall (2003)
also noted that Andruris has axially elongated and branched reproductive axes with a vasculature system reminiscent of reduced inflorescence axes with shortened internodes. In both Mexican triurids, however, floral vasculature is extremely reduced (Fig. 5 and Martínez and Ramos, 1989
); this is typical of true flowers. The female floral units of both species have broad receptacles, and the vascular traces depart into each tepal, but not into each carpel or stamen (Figs. 1, 5, 68, 69, 76, and Martínez and Ramos, 1989
).
Male flowers of Triuris could also be interpreted as pseudanthia with either a single tristaminate flower or more highly reduced unistaminate flowers subtended by three or six bracts (Rudall, 2003
). Particularly, the androphore characteristic of T. brevistylis and other species of Triuris could be interpreted as a residue of meristematic apical tissue, perhaps suggesting an indeterminate meristem typical of inflorescences but not of flowers. However, the developmental evidence shown here clearly shows that the filaments of the three stamens are apparent at early stages of male flower development, and evidence of these filaments should remain if the androphore originated only from meristematic activity from the receptacle. Finally, the androphore is a determinate rather than indeterminate structure.
Rudall (2003)
also suggested that the variable number of carpels in the triurids reflected an indeterminate inflorescence meristem. However, our developmental series shows that the carpels arise from common and compound primordia and not in whorls, and this may explain the varying numbers of carpels formed. Rudall (2003)
also compared the female floral unit of Triuridaceae to pistillate flowers of Pandanus and Sararanga (Pandanaceae) and to Dicranopygium and Carludovica (Cyclanthaceae). The homologization of reproductive axis organization inherent to the pseudanthial hypothesis requires that the structures in Triuridaceae be "subtended by a petaloid bract (three bracts forming a bracteate pseudocorolla)" (Rudall, 2003
, p. S315), but it is hard to think that six tepals or four tepals (as some members of Triuridaceae have) also form a bracteate pseudocorolla together. According to our interpretation of the electron micrographs, both triurids studied here have trimerous patterns in the perianth and androecium whorls, suggesting that these are part of true monocot flowers rather than pseudanthia.
The morphological differences between meristems that give rise to what would be floral units, with respect to the putative apical meristems of the inflorescence in both species studied here, also suggest a euanthial hypothesis for triurids. The inflorescence meristems appear indeterminate, naked, and central with respect to the flanking floral meristems. In contrast, each floral meristem is determinate and enclosed by a clearly distinguishable bract. When fully developed, this bract is clearly distinguished from the tepals once these are formed later in flower development. Developmental data in species of Pandanaceae (and other members of Pandanales) with true pseudanthia are very scarce. However, the available information for Freycinetia indicates that bracts and either stamens or carpels arrest at early stages of development in each of the multiple flowers that compose the adult pseudanthial spikes (Stone, 1972
; Cox, 1990
; Huynh, 1992
).
The issue of organ arrest, in reference to its participation in the formation of pseudanthia in other angiosperm families, deserves special attention in our attempt to assess the hypothesis of reduced inflorescences in the triurids. In further support of an euanthial nature for the floral units of the two triurid species studied here, we did not detect, at any stage of development analyzed, arrested floral organ primordia or perigon structures subtending the individual reproductive units that, according to the former hypothesis, would be very reduced flowers instead of floral organs. In certain species with pseudanthia, individual flowers can be so reduced that they completely lack a perianth and can be reduced to just a single reproductive organ when observed at maturity. Plant species in genera outside Pandanales with extreme floral organ reduction in inflorescences, such as Euphorbia, have stamens with obvious constrictions between its glabrous filament and the hairy peduncle, distinguishing this single stamen as an individual male flower of a reduced inflorescence (Weberling, 1989
). In axes of both L. schismatica and T. brevistylis, we were unable to find any constrictions at the base of carpels or stamens that would indicate a floral nature for these organs.
Pseudanthia are often showy and have been postulated to be adaptations to attract pollinators (Claßen-Bockhoff, 1990
). Regardless of the accepted interpretation of their identity, the colorless perianth of the Mexican triurids does not function in attraction because the species with bisexual flowers is cleistogamous, and its ovules are fertilized preanthesis (Márquez-Guzmán et al., 1993
). In both Mexican triurid species, the spatiotemporal pattern of tepal emergence is reminiscent of flower perianth organs rather than bracts, which would emerge in a spiral arrangement. Furthermore, both species have prophyllate aestivation (i.e., the first tepal arises nearly opposite of the single bract), a characteristic pattern of floral organ development with respect to prophylls in true flowers (Weberling, 1989
). Seemingly, if the tepals and bract have any function, it is probably related only to protection for the developing bud.
The presence of common and compound organ primordia also supports that L. schismatica and T. brevistylis reproductive structures are true flowers. It is noteworthy that common primordia of other species generally give rise to petal and stamen primordia and compound primordia usually give rise to many stamen primordia (e.g., Weberling, 1989
; Tucker and Bernhardt, 2003
), rather than to stamens and carpels or to carpels only, as in L. schismatica and pistillate T. brevistylis flowers, respectively.
In many bisexual flowers, floral organs usually arise in acropetal succession with the sepals forming first, followed by the petals, then the stamens, and finally the carpels, located in the center of the receptacle (Weberling, 1989
). But there are deviations from this spatiotemporal pattern in several flowering species (Tucker, 2003
). Lacandonia schismatica follows a typical angiosperm temporal progression of organ development according to organ identity (perianth organs first, then stamens, and finally carpels), but the spatial distribution of the floral organs is different in a way that suggests homeosis with respect to reproductive axes in which female structures are central. It is difficult to reconcile a pseudanthial interpretation of the reproductive axes in L. schismatica with the observed features of its organ ontogeny. The developmental series for L. schismatica described in this paper shows that tepals develop first in the outermost whorl, then the stamen primordia arise in the central part of the flower meristem, and finally carpel primordia appear between the stamen and tepal primordia. The pseudanthial hypothesis entails that triurid flowers—in particular, those of L. schismatica—were instead derived from inflorescences that underwent extensive processes of internode and organ reduction, yet still retaining an inflorescence character. Under such an interpretation, the flowers at the bottom of the inflorescence would develop first (tepals), followed by the male flowers at the apex (stamens in the center), and finally, the female flowers (carpels) would emerge in the middle region of the inflorescence. To our knowledge, such a pattern has not been found among inflorescences.
The results in L. schismatica led us to address if in female T. brevistylis the central carpel primordia also appeared before the more peripheral ones. Interestingly, female T. brevistylis has a temporal pattern of organ primordia initiation very similar to that found in L. schismatica. In both species, three primordia arise in the center of the meristem, and divisions along axes radiating from the central primordia give rise to more primordia. Perhaps, this pattern of reproductive organ formation favored the origin of the inversion of reproductive organ identity because, although in very few cases, reproductive axes bearing stamens that are central with respect to carpels have also been found in T. brevistylis (Vergara-Silva et al., 2003b
). The study of reproductive development in hermaphroditic axes of Triuridaceae that do not have an inverted position of stamens and carpels would be helpful to corroborate our findings on the sequence and pattern of organ formation in the Mexican triurids. Such reproductive units exist in species of the genus Sciaphila (Schlechter, 1913
; Giesen, 1938
).
Finally, our results indicate that the stamen primordia arise in the center of the floral receptacle and continue their development to maturity in that position of the flower, as suggested by previous anatomical studies (Márquez-Guzmán et al., 1993
). This evidence argues against secondary physical inversion. Although stamens and carpels arise from a common precursor and much of the subsequent growth occurs on the abaxial face of the common primorida, the identities of the reproductive organs are likely established before any primorida emerge. In summary, the floral developmental series shown here strongly argue against the pseudanthial and secondary physical inversion hypotheses to explain the unique spatial organization of the reproductive whorls of L. schismatica. From the classic morphological standpoint, this implies that the reproductive axes of the Mexican triurids are true flowers; we thus conclude that designating the arrangement of floral organs in these flowers as homeotic is valid.
Structural and developmental differences in the reproductive axes of the Mexican triurids: utility for further studies inside Triuridaceae
The number of genera in Triuridaceae varies from five to nine, while the number of recognized species varies from 30 to >80, depending upon the author (Giesen, 1938
; Maas and Rubsamen, 1986
; Rübsamen-Weustenfeld, 1991
; Maas-van de Kamer and Weustenfeld, 1998
; Gandolfo et al., 2002
). The many anatomical and developmental similarities between the Mexican triurids are expected in view of their close taxonomic alliance and further support it. However, there are some contrasting features in their reproductive morphology and development. An account of these differences, in the light of general information on extant members of the family and recent publications on fossil Triuridaceae (Gandolfo et al., 2002
), indicates their possible utility in subsequent systematic and taxonomic treatments of the family as a whole.
Inflorescence and floral morphology in the Mexican triurids
Most members of the family Triuridaceae have been described as having racemose inflorescences. However, some authors have interpreted the inflorescences of Triuris and L. schismatica as being sympodial (Giesen, 1938
; Jonker, 1943
; Standley and Steyermark, 1958
; Martínez and Ramos, 1989
). Our results agree with the interpretation of the inflorescence of both T. brevistylis and L. schismatica as being monopodial racemes (Maas and Rübsamen, 1986
).
The existence of compound or common primordia has not been reported for Triuridaceae, and floral developmental series of other triurid taxa would reveal whether the formation of compound or common primordia is a common developmental feature of this family. An important difference in reproductive organ development between the two triurids studied here concerns the fate of the compound primordia present in both species. In the female flowers of T. brevistylis, only carpels arise from all six compound primordia, while in L. schismatica, the three central primordia become stamens and the rest carpels. In both species, organs initiate from the compound primordia in a centrifugal manner as the floral receptacle expands and as space allows. There are heterochronic differences in the subsequent differentiation of the carpels. In L. schismatica, after the inception of all carpel primordia, subsequent differentiation is simultaneous. In contrast, the carpels of T. brevistylis differentiate sequentially as they arise centrifugally. After differentiation, the carpels grow rapidly in size. Therefore, the first arising central carpels are slightly more developed and much larger than the later arising peripheral carpels in T. brevistylis.
The stamens of L. schismatica and Triuris species have been reported as lacking filaments (Jonker, 1943
; Maas and Rübsamen, 1986
; Maas-van de Kamer and Maas, 1994
). In virtue of this character state coding, the latter authors described a new species of Triurideae, Triuridopsis peruviana, in which stamens are filamented (Maas-van de Kamer and Maas, 1994
). Our results here show that both L. schismatica (see also Martínez and Ramos, 1989
) and T. brevistylis have filaments.
The stamens of L. schismatica have anthers with one theca and two pollen sacs, while those of T. brevistylis bear anthers with two thecae and two pollen sacs per theca. The number of stamens for T. brevistylis has been reported as three or six. Our results show that there are three stamens, each having two divergent theca. The anthers of L. schismatica dehisce longitudinally, while those of T. brevistylis dehisce either longitudinally or transversely, and the former has introrse stomia, while the latter has extrorse stomia. Gandolfo et al. (2002)
lists all extant members of Triurideae (Peltophyllum, Triuridopsis, Triuris, and L. schismatica), as well as the two fossil genera as having extrorse anthers.
Sterile appendages in extant and fossil Triuridaceae
Gandolfo et al. (2002)
scored 20 morphological characters for the Triuridaceae and 13 morphological characters for the fossil genera Mabelia and Nuhliantha. Two well-defined clades were obtained in a cladistic analysis: one includes all members of tribe Sciaphileae (Seychellaria, Hyalisma, Sciaphila, and Soridium), and the other includes both fossil genera and the tribe Triurideae. This analysis not only places the fossil genera as sister to the Triurideae, but also includes Triuris, Triuridopsis, Peltophyllum, and L. schismatica in the tribe. Therefore, it also supports the idea that L. schismatica belongs in the tribe Triurideae in the family Triuridaceae, and not in a separate family. The results shown here suggest that there are discrepancies for some of the scored morphological characters. It is worth investigating if correcting these for Triuris and L. schismatica would result in the same relationships after new analyses.
The thecal divergence in T. brevistylis could be due to residual meristematic activity at the base of the connective tissue that may also contribute to androphore growth. Meristematic activity at the base of the connective has been noted for other species where thecae have diverged (Weberling, 1989
). This interpretation coincides with those regarding the origin of other sterile appendages in other triurids as the result of extended meristematic activity of connective tissue. The development of the androphore in T. brevistylis is not comparable to what occurs in L. schismatica, but other members of Triuridaceae develop sterile projections. It is not clear if there is any biological role for the androphore or sterile appendages found in Triuridaceae; however, studies on the pollination biology of these species may provide clues.
Both extant and extinct members of Triurideae have some sort of sterile appendages. The fossil flowers have filaments that are extended and fused. Triuridopsis peruviana has a sterile projection in the center of the male flower and Peltophyllum has a small androphore, while Triuris has a large androphore (Maas and Rübsamen, 1986
; Maas-van de Kamer and Maas, 1994
). Some members of the tribe Sciaphileae also have sterile appendages including androphores (Maas and Rübsamen, 1986
), whereas Sciaphila arfakiana has filiform connective appendages in the male flower (Schlechter, 1913
). Andruris and Seychellaria both have connective appendages (Maas and Rübsamen, 1986
). Other members of Sciaphila have basally fused filaments similar to T. brevistylis before the growth of the androphore obscures them. Our observations indicate that L. schismatica lacks sterile appendages. Given the discrepancies in coding states for these appendages—for example, Maas van de Kamer and Maas (1994)
list Peltophyllum as having no sterile appendages, while Maas and Rubsamen (1986)
note that Peltophyllum has a small androphore— and given the presence of different types of sterile appendages in both the tribes of Triuridaceae and the fossil genera, we consider that these structures warrant further developmental studies, to establish them as important diagnostic characters for this family.
Spatiotemporal patterns of gene expression and the evolution of floral structures in Triuridaceae: hypotheses and perspectives
The molecular genetic basis of floral development has been thoroughly studied in a few model systems, most of them belonging to the eudicot clade (see Friedman et al., 2004
; Pruitt et al., 2003
; and Ferrario et al., 2004
for recent reviews). The ABC model (Coen and Meyerowitz, 1991
) has been instrumental in establishing a link between developmental genetics and evolutionary studies by providing a straightforward framework for comparison of expression patterns of homologous sequences that potentially could have been involved in determining peculiar floral or inflorescence features in specific taxa. This model establishes that the activity of A genes that are expressed in the two outer whorls (sepals and petals) alone determines sepal identity; the combined activity of A and B genes, that are expressed in the second and third whorl, determine petal identity; the activity of B and a single C gene, that is expressed in the third and fourth whorls, determine stamen identity; and the activity of C alone determines carpel identity.
We have hypothesized (Vergara-Silva et al., 2003b
) that the putative floral organ homeotic inversion present in the flowers of L. schismatica could be due to a spatial displacement of the expression patterns of genes homologous to the B-function genes in A. thaliana (the MADS-box genes APETALA3; AP3; Jack et al., 1992
) and PISTILLATA (PI; Goto and Meyerowitz, 1994
). Simultaneous transcription of B-function homologous sequences—whose products usually function as heterodimers—in the center of L. schismatica floral meristems could specify the development of stamens if homologues of the C function gene (corresponding to the MADS-box gene AGAMOUS; AG; Yanofsky et al., 1990
) from A. thaliana are also expressed in the corresponding cells. According to our original hypothesis, the sole presence of transcripts of C-function homologous genes in the periphery of the developing stamens could also explain mechanistically the formation of carpels surrounding the central androecium (Vergara-Silva et al., 2000
; Vergara-Silva et al., 2003b
).
The variants that were uncovered in T. brevistylis caused us to speculate that more complex interactions of the floral organ identity genes might be necessary to explain the phenotype of the heterotopic hermaphroditic flowers described (Vergara-Silva et al., 2003b
). The developmental series presented here for both Mexican triurids allows us to interpret the normal phenotype of L. schismatica and variants of T. brevistylis by the simplest molecular explanation—a shift in B function activity.
Lacandonia schismatica has three stamen–carpel common primordia and three carpel compound primordia, while T. brevistylis has six carpel compound primordia. Other species have been reported to have common and compound primordia (Ferrándiz et al., 1999
; Tucker, 2003
), but the development of stamens and carpels from a common precursor as shown here has not been reported before. These angiosperm species, whose morphogenetic patterns are divergent with respect to those of model systems should become important for discovering the molecular mechanisms that underlie the overall spatiotemporal patterns of floral organ initiation and determination, as well as the timing and control of identity of organs emerging within whorls. In the case of the Mexican triurids studied here, though, it would still be necessary for B and C activity to occur in the center of the flower and C activity in the periphery before any growth is seen on the compound primordia, as floral organ identity genes are active before primordia are even evident.
It is likely that additional genetic factors would be involved in the shift of B function toward the flower center. A network model of regulatory gene interactions that includes the ABC genes has been proposed to comprise a module for the combinatorial selection of gene activity (Espinosa-Soto et al., 2004
) and puts forward a mechanistic explanation for how the concerted action of ABC and non-ABC genes might underlie the combinatorial ABC model. Spatiotemporal analyses of this network model will be a useful framework for postulating more elaborate alternative hypotheses on the genetic alterations that can cause spatiotemporal patterns of floral organ determination. However, it is likely that additional mechanisms underlying primordia location and growth (for example auxin signaling) are coupled to this network and jointly specify the spatiotemporal patterns of floral organ primordia determination and growth in different parts of flowers. These unknown mechanisms should be conserved between L. schismatica and T. brevistylis because in both species the central primordia emerge before those in the next whorl that in both cases differentiate into carpels.
For example, we may speculate that distinct roles for homologues of the A. thaliana zinc-finger gene SUPERMAN (SUP) in triurids could be involved in these processes. The function of SUP was originally associated with the definition of the spatial boundary between the third and fourth whorls (Bowman et al., 1992
; Sakai et al., 1995
). Homologues of SUP, as well as conserved elements in their mechanisms of action, have already been found in monocotyledon taxa (e.g., rice; Nandi et al., 2000
). This and other hypotheses of the molecular mechanisms underlying the unique floral arrangement of L. schismatica can now be tested in the developmental framework of this study and the formal dynamical models proposed for model systems.
The fundamental features of the ABC model also seem to be conserved between dicots and monocots (Ambrose et al., 2000
; Whipple et al., 2004
) in a manner that allows the postulation of organ homologies between the specific floral structures of the grasses and the cruciferous flower. Comparative studies of this sort are not abundant in other monocot taxa. Concerning the phylogenetic vicinity of Triuridaceae inside Pandanales, our results, coupled to the observations of researchers working in Pandanaceae, suggest that homologous patterns of expression for the ABC MADS-box genes can be postulated between early developmental stages of euanthial units in the Mexican triurids and similar stages for each of the multiple floral primordia that comprise the inflorescences (spikes) in Freycinetia species (Stone, 1972
; Cox, 1990
