Inside-out flowers of Lacandonia brasiliana (Triuridaceae) provide new insights into fundamental aspects of floral patterning

Material examined

Material was obtained from the following sources: the spirit collection at the Royal Botanic Gardens, Kew (K), the spirit collection at the National Herbarium of the Netherlands, Leiden (NHNL), and wild-source collection (UFP) from Brazil, from a lowland Atlantic Forest fragment with a white sandy soil that is known locally as “tabuleiros.”

Material imaged for this study (for authorities of species names, see Table 1):

Lacandonia brasiliana: collected in alcohol in Brazil by one of the authors (MA): UFP: A. Melo et al., 1195, 25 August 2013, Brazil, Paraíba, Reserva Biológica Guaribas, 100–150 msm, 06°42′24″S–35°10′35″W. SisBio collecting permit no. 19769-4 (June 2012–June 2014).

Lacandonia schismatica: flowers donated by Dr. Francisco Vergara-Silva, UNAM, Mexico.

Triuris hexophthalma: pre-prepared microscope slides made by Traudel Rübsamen-Weustenfeld and housed at the Department of Botany, University of Bochum, Germany). These slides were examined by kind permission of Prof. Thomas Stützel.

Triuridopsis peruviana H. Maas & Maas (K: T. Laessoe 265, Peru).

Peltophyllum luteum Gardn. (K: Philcox 4245, Brazil).

Microscopic examination

For light microscopy (LM), flowers were embedded in LR-White resin and sectioned using a Leica rotary microtome. Sections were stained in toluidine blue, dehydrated through an alcohol series to 100% ethanol and then placed in Histoclear. Sections were mounted in DPX mounting medium (distrene, with dibutyl phthalate and xylene). Slides were examined using a Leica DMLB photomicroscope fitted with a Zeiss Axiocam digital camera.

For whole mounts, flowers and buds were dissected and dehydrated in an ethanol series. Dehydrated material was critical-point dried using an Autosamdri 815B critical-point drier, dissected where necessary, and mounted onto scanning electron microscope (SEM), stubs using double-sided adhesive tape. Some specimens were imaged using a Nikon Shuttlepix P-MFSC optical system, where necessary subsequently using EDX image stacking to achieve an average focus from multiple primary optical frames (Fig. 2). For scanning electron microscopy, mounted specimens were coated with platinum using an Emtech K550X sputter-coater, and examined under a Hitachi cold-field emission SEM S-4700-II at 2 kV. The resulting images were recorded digitally for subsequent manipulation in Adobe Photoshop. Earlier versions of some SEM images (Figs. 10, 11 and 12D) were published previously by one of us (Rudall, 2003; Rudall, 2008), reproduced here with permission from the International Journal of Plant Sciences (University of Chicago Press).

Flowers of Lacandonia brasiliana

Flowers of L. brasiliana are bisexual with six tepals arranged in a single whorl. Numerous carpels surround the stamens, which themselves form a single whorl around a central depression (Figs. 2–4). The six tepals are fused at their margins in closed buds (Figs. 5A and 5B) and remain basally connate in open flowers (Fig. 3B). Tepals of open flowers have long filamentous tendril-like structures close to their apices. These subapical filamentous appendages begin to develop in closed buds and grow downwards between the central stamens (Fig. 6). When extended, the tepal appendages consist of a thick epidermal layer with persistent cytoplasmic contents; the epidermis encloses a small amount of thin-walled parenchyma and a central region of vascular tissue (Fig. 4E).

Three (rarely two or four) stamens are present (Figs. 2 and 3), though they are sometimes barely visible below surrounding carpels (Fig. 2B). Development of the stamens is precocious with respect to the carpels. Each stamen consists of an anther borne on a very short filament (i.e., the anther is almost sessile: Fig. 4D). Each anther possesses a slit-like aperture, or sometimes two or three apertures (Figs. 5E and 5F). The anther apertures apparently remain firmly closed, even in open flowers. Occasional anthers appeared partially fused together (Fig. 5G). Early stages possess sporogenous tissue and three to four anther wall layers (Fig. 7A) that subsequently differentiate into an endothecium and tapetum (Fig. 7B). At later stages, the tapetum degenerates and the endothecial cells possess thickenings on their inner walls (Fig. 7D). Mature (three-celled) pollen grains are present in well-developed anthers in unopened buds before the carpels are fully developed (Figs. 4D and 7A–7C). Germinating pollen grains were visible in many anther locules, typically with several pollen tubes growing along the middle lamellae of parenchyma cells in the filament (pollen tubes arrowed in Figs. 7D–7F).

The stamens are surrounded by numerous carpels with papillate surfaces that make them readily distinguishable from the stamens (Figs. 2 and 3). At early stages the carpels appear as undifferentiated primordia, even as the stamens have begun to differentiate (Fig. 4A). Each carpel possesses a subapical style-like appendage that develops at around the same time as the filamentous tepal appendages (Fig. 5). Each carpel is uniovulate (Fig. 8). The subapical style-like carpel appendages remain unvascularized and lack an obvious transmitting tract (Fig. 8B). The ovules are anatropous and each have two integuments (Figs. 8A and 8D). We observed megagametophytes at the megaspore-mother cell (Fig. 8A) and four-celled stages (Figs. 8C and 8D). In older flowers, many ovules are apparently fertilized and the developing seeds contain endosperm and a proembryo at the micropylar end (Figs. 8E and 8F).

Lacandonia brasiliana closely resembles L. schismatica in flower structure

Our study found remarkably few morphological differences between flowers of Lacandonia brasiliana (Figs. 2–8) and L. schismatica (Fig. 9), despite the large geographical separation of these apparently isolated species (Fig. 1). Both species possess inside-out flowers characterized by three central stamens that develop precociously with respect to the carpels (Márquez-Guzmán et al., 1989; this paper). Some morphological variation is evident in both species. For example, occasional flowers have reduced or absent stamens (Vergara-Silva et al., 2003; this paper). In both L. brasiliana and L. schismatica, the tepals have relatively smooth surfaces and are basally connate with late-developing subapical filamentous appendages. Both species have almost sessile anthers that remain closed. Remarkably, in both species fertilization is apparently achieved by pollen tube growth from germinating pollen grains produced in the same cleistogamous flower (Márquez-Guzmán et al., 1993; this paper). The phenomenon of precocious pollen germination within intact anthers, followed by extragynoecial fertilization without pollination, is rare in angiosperms. It has been reported in cleistogamous flowers of some eudicots: a few species of Malpighiaceae (Anderson, 1980) and several species of Callitriche (Philbrick, 1984). Among other free-carpellate monocots, pollen-tube growth via the receptacle has also been documented in Sagittaria potamogetifolia (Wang, Tao & Lu, 2002). In contrast, pollen grains have not been observed germinating within the same anther locule of other Triuridaceae, such as Triuris hexophthalma.

One possible difference is that the subapical tepal appendages appear shorter in L. schismatica than in L. brasiliana, but this apparent distinction requires quantification. Tepal appendages are also relatively short in Triuridopsis peruviana (Maas-van de Kamer & Maas, 1994). Melo & Alves (2012) observed that L. brasiliana differs from L. schismatica in having solitary flowers or inflorescences with up to six flowers (compared with up to three in L. schismatica), but this feature requires improved population analysis and statistical tests. There is also a potential phenology difference, although this could simply reflect contrasting latitudes: L. brasiliana typically flowers in August and September, compared with November and December in L. schismatica. These characteristics require more detailed comparative data, both morphological and molecular, to determine whether they adequately delimit bona fide species boundaries.

At least some of the features shared between L. brasiliana and L. schismatica are also common to other species of the neotropical tribe Triurideae (Figs. 10–12). However, the close similarity between the two Lacandonia species makes it appear unlikely that they arose independently from two separate homeotic transformation events. Thus, they could either represent sister species that border on the definition of cryptic species, or two populations of a single disjunct species. In either case, the genus Lacandonia is more widespread than was previously believed. Such disjunct distributions are not unknown, especially in mycoheterophs, which are inconspicuous and ephemeral and hence are undoubtedly seriously under-recorded (Merckx, Smets & Specht, 2013). For example, the widespread species Triuris hyalina is known from disjunct populations in southeast Brazil, the Brazilian Amazonas region and Central America (Fig. 1), encompassing the previously named taxa T. brevistylis, T. major and T. mycenoides (Maas & Rübsamen, 1986; Maas, Maas & Melo, 2015). Interestingly, Vergara-Silva et al. (2003) postulated that L. schismatica could have originated from mutations in populations of “T. brevistylis” within the Mexican Lacandon rainforest (see also Mabberley, 2008; Espinosa-Matías et al., 2012). “Triuris brevistylis” itself represents populations from the Yucatán Peninsula (Guatemala and Mexico) of the widespread species T. hyalina. Although “T. brevistylis” is typically dioecious or monoecious, Vergara-Silva et al. (2003) observed occasional bisexual flowers with varying positions of stamens and carpels, including homeotic variants possessing inside-out bisexual flowers. It is also noteworthy that most species of Triurideae are narrow endemics known from a single locality or a few localized populations (Fig. 1 and Table 1).

Taxonomic boundaries within Triuridaceae–Triurideae are ambiguous, though the tribe has two clear morphological synapomorphies

The neotropical tribe Triurideae, which is monophyletic in both molecular and morphological studies (Rudall & Bateman, 2006; Mennes et al., 2013), consist of four genera: Lacandonia, Triuridopsis, Triuris and Peltophyllum (Table 1). The tribe is apparently well-defined by two morphological synapomorphies: (1) the long filamentous appendages that grow from the tepal tips at relatively late stages of development, and (2) carpel orientation and development (i.e., carpel fascicles—structures that are discussed in more detail in the next section). Both of these remarkable features occur in all species of Triurideae, including Lacandonia brasiliana. The developmental timing and anatomy of the subapical tepal appendages, which occur in both male and female flowers of unisexual-flowered Triurideae, suggest that they are osmophores (Rudall, 2003; Rudall, 2008). Filamentous appendages are common in sapromyiophilous mycoheterophs that produce otherwise inconspicuous flowers (Vogel, 1990). Although filamentous structures are common in general in the family Triuridaceae, in species of tribes Sciaphileae and Kupeaeae they grow from different parts of the flower, such as the bases of the tepals in male flowers of Seychellaria (Maas-van de Kamer & Weustenfeld, 1998; Rudall, 2008). The epidermal hairs at the tepal tips of some Sciaphila species (e.g., S. picta: Maas & Rübsamen, 1986) are non-homologous with the relatively large, vascularized and organ-like appendages of Triurideae.

Our study underlines the problematic generic and species boundaries within Triurideae. Existing molecular data do not resolve this issue. The molecular analysis of Mennes et al. (2013) placed Lacandonia schismatica (using the single available plastid sequence, atpA) as sister to Triuris (T. hexophthalma plus T. hyalina), with relatively long branch lengths between taxa. Population-level studies are needed to resolve the systematics of Lacandonia, including species of Triuridopsis and Peltophyllum and improved sampling within Triuris, perhaps focusing on nuclear gene regions.

In terms of flower structure, Peltophyllum (Figs. 2F and 10) is arguably the least complex genus of the tribe Triurideae because its male flowers lacks the central appendage that characterizes male flowers of Triuridopsis and Triuris (Figs. 11 and 12). The remarkable ovoidal or conical central structure in Triuris was termed an androphore by Maas & Rübsamen (1986). The term “androphore” normally indicates a central column formed by united filaments in a male flower. We use this term here for convenience, though there is no evidence that this unusual structure is formed from united filaments in Triuris. In some other Triuridaceae the filaments are united at their bases. The androphore is late-developing, at least in “Triuris brevistylis” (Vergara-Silva et al., 2003; Espinosa-Matías et al., 2012). It contains enigmatic regions of cells with dense cytoplasmic contents (Fig. 12B) and is most likely a scent-producing structure (osmophore: see also Rübsamen-Weustenfeld, 1991; Rudall, 2003). Such retarded development is typical of structures that are additional to the normal floral organ complement, such as many osmophores and some types of nectary (Rudall, 2010).

The Peruvian species Triuridopsis peruviana was segregated from Triuris based primarily on two characters. Firstly, its anthers possess filaments, compared with sessile anthers in Triuris. Secondly, it has a sterile projection in the centre of the male flower instead of an androphore (Maas-van de Kamer & Maas, 1994); this appendage appears to emerge directly from the receptacle (Figs. 11B and 11D). Subsequently, Franke, Beenken & Hahn (2000) added a second species of Triuridopsis, T. intermedia from Bolivia, which differs from T. peruviana in possessing a relatively short central appendage. However, the androphore itself appears highly plastic. In “Triuris brevistylis,” Vergara-Silva et al. (2003) documented male-like flowers with a malformed or arrested androphore and female-like flowers lacking an androphore entirely and possessing intermingled anthers and carpels, or even inside-out flowers. On the other hand, they found no floral variants of Lacandonia that possessed an androphore, and we also found none in our material. Thus, existing comparative morphological data are conflicting, but the discovery by Melo & Alves (2012) of Lacandonia in northern Brazil opens exciting prospects, even making more plausible the possibility of a close relationship with Peltophyllum, which occurs in Paraguay, Argentina, and south-western Brazil (Fig. 1 and Table 1).

New insights on character evolution in Triuridaceae

The debate on the evolutionary origin of the inside-out Lacandonia flower has become entangled in multiple issues. To clarify the problem, we present an evolutionary scenario (Fig. 13) that illustrates the different phases that led to this bizarre phenotype. Contrasting hypotheses are applicable to at least two levels in this process.

To work backwards from the Lacandonia phenotype, the arguments for homeotic transformation of organ types appear highly plausible, especially given that studies of the B-function (Ls-AP3 and Ls-PI) and C-function (Ls-AG) MADS-box floral-organ identity genes show that B-function occurs in the flower centre in Lacandonia (Álvarez-Buylla et al., 2010). Homeotic transformations are certainly common within Triuridaceae, and could have caused a saltational origin of Lacandonia from a “hopeful monster” (cf. Bateman & DiMichele, 2002). Hemsley (1907) noted in Sciaphila (Triuridaceae) that even in species bearing typically unisexual flowers, many are “quasi-hermaphrodite” and often irregular as to the number and location of the floral parts. He commented on Triuridaceae in general that “it would appear that these small flowers are peculiarly subject to disturbances in their development” (page 75). However, as shown in Fig. 13, it remains ambiguous whether the homeotic transformation in Lacandonia resulted from a replacement of carpels with stamens in a female-like flower (homeoheterotopy) or de novo origin of carpels in a new location in a male-like flower (neoheterotopy: Bateman & DiMichele, 2002; Baum & Donoghue, 2002). Indeed, since both male and female flowers occurred in the putative ancestor, it is difficult to see how developmental-genetic studies that inevitably investigate a snapshot in evolutionary time could conclusively distinguish between these processes.

As discussed in the previous section, two clear synapomorphies place Lacandonia firmly within the neotropical tribe Triurideae: filamentous subapical tepal appendages and the highly unusual (possibly unique) carpel fascicles, which have also been subject to contrasting interpretations. In all species of Triurideae, including both species of Lacandonia, the carpels are arranged on ridges in radial double rows (Vergara-Silva et al., 2003; Rudall, 2008; this paper). This feature is sometimes more clearly discernible at early developmental stages, as the ridges are more disorganized in older buds (Figs. 5D, 9C and 10B). Organ primordia develop along the ridges, maturing from the inside outwards. In other Triurideae, all of these organ primordia become carpels, but in Lacandonia, the apices of the three primary ridges develop into stamens. Rudall (2008) compared the ridges (termed carpel fascicles) with stamen fascicles, which occur widely in angiosperms, including some species of the related family Velloziaceae (Sajo, Mello-Silva & Rudall, 2010). Sokoloff et al. (2007) suggested that the carpel fascicles or ridges result from folding of an enlarged (perhaps fasciated) floral apex (see also Rudall, 2008; Endress, 2014), resulting in a single sinuous (star-shaped) whorl of carpels, with the coils compressed together into zig-zag radial lines. In this case, the three stamens in Lacandonia are located at the tips of three upward loops of a sinuous whorl. Regardless of interpretation, obscurity still surrounds the evolutionary derivation of a female flower with carpel fascicles from one without them. We speculate that understanding the basis for this structure could clarify the processes that led to the inside-out phenotype in Lacandonia.

The final conundrum in this series is the derivation of the unisexual flowers of Triuridaceae from a “typical” monocot flower, which is trimerous and pentacyclic, with three organs per whorl arranged in five whorls: two tepal whorls, two stamen whorls, and a single central carpel whorl with three fused carpels and septal nectaries (Remizowa, Sokoloff & Rudall, 2010). Most lilioid monocots are relatively conservative in their floral groundplan. In Velloziaceae, the sister family to all other Pandanales (e.g., Mennes et al., 2013), the flowers are at least superficially closer to the typical monocot structure, though some species possess a corona of six petaloid appendages, and others show increased stamen number (Sajo, Mello-Silva & Rudall, 2010).

Conversely, in some Pandanales, loss of the typical monocot flower groundplan is so extensive that precise delimitation of the inflorescence–flower boundary is problematic (Rudall, 2003; Rudall & Bateman, 2006; Rudall, 2008). Like Triuridaceae, flowers are also unisexual in most Cyclanthaceae and Pandanaceae, with highly atypical organ numbers. In Cyclanthus, female and male flowers are united in rings, with a loss of flower individuality (Sajo et al., 2014). The female reproductive units of Sararanga (Pandanaceae) possess up to 80 carpels united to form a unilocular ovary, and the carpels follow a zigzag arrangement along a folded axis, reminiscent of the asterid eudicot Tupidanthus (Araliaceae: Sokoloff et al., 2007). The pseudanthial theory proposed by Rudall (2003) for Triuridaceae was partly dependent on a postulated close relationship with Pandanaceae and Cyclanthaceae. The results of subsequent molecular (Mennes et al., 2013) and morphological (Rudall & Bateman, 2006) cladistic analyses suggesting a closer relationship with Stemonaceae, which have typical bisexual flowers, appeared to contradict this hypothesis. Flowers of Stemonaceae are bisexual but resemble those of Triuridaceae in many respects (Rudall et al., 2005; Rudall & Bateman, 2006). In some Stemonaceae (e.g., Croomia), the carpel whorl is reduced to a single carpel, leading Rudall & Bateman (2006) to postulate derivation of the Triuridaceae flower by carpel multiplication. However, a recent whole-plastid genome analysis (albeit with only Sciaphila sampled for Triuridaceae) found strong evidence that Triuridaceae is the sister group of a clade comprising Cyclanthaceae and Pandanaceae (Lam, Gomez & Graham, 2015), a topology that would support a loss of developmental homeostasis in these taxa (cf. Rudall & Bateman, 2006).

In conclusion, a homeosis theory for Lacandonia origin is not in conflict with any of the various hypotheses for the origin of the Triuridaceae flower, which occurred much earlier in the evolutionary history of this unusual family and coincided with a shift to unisexuality. Indeed, the unisexual yet highly plastic flowers of Triuridaceae could have pre-adapted the origin of the bizarre Lacandonia morphology.