Flower structure and development in Vietnamese Sciaphila (Triuridaceae: Pandanales): refined knowledge of the morphological misfit family and implications for taxonomy

Groundplan of male flowers: number of whorls and their merism

Developmental data obtained for male flowers of Sciaphila suggest that the perianth is two-whorled, because the two alternating sets of tepals strongly differ in timing and manner of their initiation. Assuming that in nearly all monocots the floral whorls initiate in an acropetal sequence (Rudall, 2010; Remizowa, 2019; but see Stützel, 1984; Vrijdaghs et al., 2009), we interpret the early-arising tepals as an outer whorl, and the tepals with late-arising lobes as an inner whorl (Fig. 31). Similarly, in S. densiflora, the single species studied here with an androecium of six stamens, the androecium is also two-whorled, with the outer stamens in the radii of the outer tepals, and the inner stamens in the radii of the inner tepals. The two-whorled interpretation of the androecium in S. densiflora is supported by delayed initiation of the inner whorl of stamens. Therefore, in our interpretation, the male flowers of S. densiflora show a perfect alternation of all floral whorls, which is a basic pattern for angiosperms (Endress, 2011). Additionally, the abnormal flower of S. nana with four tepals and four stamens can also be assumed as having a two-whorled androecium, in contrast to the common single-whorled condition in this species.

Based on the idea of whorl alternation, we assume the three stamens in male flowers of S. arfakiana, S. nana and S. stellata to be homologous to the outer whorl of the androecium in S. densiflora. Thus, we describe the male flowers of Sciaphila with three stamens as tricyclic and those with six stamens as tetracyclic, differing in the absence or presence of the inner stamen whorl, with each floral whorl being typically trimerous (Fig. 31). This is consistent with the speculations by Nuraliev, Cheek & Beer (2016) who supposed two-whorled perianth and floral trimery in male flowers of Seychellaria, a genus of Triuridaceae closely related to Sciaphila.

The floral organs in the male flowers of Seychellaria which were treated as “filamentous structures” by Rudall (2008) and are commonly termed staminodes (see Nuraliev, Cheek & Beer, 2016) alternate with the stamens, and therefore their position in a flower fully corresponds to that of the inner stamens in six-stamened species of Sciaphila (such as S. densiflora). In our opinion, the criterion of position corroborates the androecial nature of these structures in Seychellaria. The male flower of Seychellaria is thus tetracyclic, with a remarkable stamen dimorphism among the androecial whorls.

Our results are in good agreement with illustrations of male flower development of S. arfakiana presented by Rübsamen-Weustenfeld (1991: Taf. 34–35). Rübsamen-Weustenfeld (1991) interpreted the tepals of each whorl as developing sequentially, which is, however, not evident from her SEM images, which show all three primordia of a whorl being of nearly equal size at each stage. This interpretation also contradicts our findings that the tepals initiate simultaneously and develop with a delay at the abaxial side of a flower (see below). Although early perianth development was thoroughly studied and illustrated only for S. arfakiana (a single appropriate illustration provided for Soridium spruceanum Miers is quite inconclusive), Rübsamen-Weustenfeld (1991) extrapolated her data to assume the perianth of some other Triuridaceae as two-whorled, with the stamens occupying the radii of the outer whorl (see also Maas-van de Kamer, 1995; Maas-van de Kamer & Weustenfeld, 1998). These representatives include the neotropical S. rubra Maas characterized by a perianth of four tepals and androecium of four stamens. Our observations on two more species, S. nana and S. densiflora, support the idea that the two-whorled nature of the perianth is stable in Sciaphila.

Our observations do not support the hypothesis that tepals of Triuridaceae are arranged in pseudowhorls rather than true whorls (Rübsamen-Weustenfeld, 1991). The term “pseudowhorl” (Charlton, 1973; Sattler & Singh, 1978; Posluszny & Charlton, 1993) was introduced to describe patterns of organ arrangement in which phyllomes follow a spiral, but are closely spaced in groups superficially resembling whorls (i.e., each phyllome is arranged in its own node). In our view, the perianth of Sciaphila follows the typical monocot groundplan with two trimerous whorls of tepals. We disagree with the view of Rübsamen-Weustenfeld (1991) that the early-arising tepals are initiated sequentially as 1, 2, 3 on the floral axis. In addition, sequential initiation of the phyllomes does not preclude their arrangement in a whorl, and this condition is well-documented in some monocots with typical flower groundplan (e.g. Greller & Matzke, 1970).

In both male and female flowers of Sciaphila, the tepals are known to be either equal or unequal, and in the latter case, the larger and smaller tepals alternate with each other (Van de Meerendonk, 1984). Since the perianth of Sciaphila (at least in the male flowers) is here shown to be two-whorled, a question arises for the species with dimorphic tepals regarding the correspondence between the tepal size (larger and smaller) and their attribution to the perianth whorls (outer and inner) in anthetic flowers. Perianth orientation with respect to the flower-subtending bract cannot be used for determination of the outer and inner tepals at anthesis, because the orientation becomes obscured due to pedicel elongation and torsion. The other way to distinguish tepal whorls in anthetic male flowers is to assess the positions of the tepals with respect to the stamens. In male flowers of Sciaphila which have unequal tepals and the stamens half as numerous as the tepals (i.e., two or three stamens), the stamens are known to occupy the radii of the larger tepals (Van de Meerendonk, 1984). From this uniform pattern it follows that the larger tepals are the outer ones and the smaller tepals are the inner ones in the flowers of this groundplan. However, in flowers with an equal number of tepals and stamens, like those of S. densiflora, the position of the stamens cannot be used for determination of perianth whorls, because each tepal is arranged in a stamen radius, and the stamen whorls are also indistinguishable. For such flowers, one can only suppose the same regularity as for the flowers with single-whorled androecium, but a possibility cannot be excluded that the inner tepals become larger than the outer tepals in the course of late developmental stages. Thus, for S. densiflora this question is still to be resolved by observation of subsequent stages of tepal growth till their maturity or possibly by comparative vascular anatomy.

Patterns of perianth initiation in male flowers

We found that in the male flowers of Sciaphila the outer tepals are in a median abaxial and transversal-adaxial positions, and the two transversal-adaxial tepals grow considerably faster than the abaxial tepal. This type of floral orientation and early development is known to be common for lateral flowers in monocots which lack floral prophylls (bracteoles) (Endress, 1995; Remizowa et al., 2013). Accordingly, all species studied here were proved to be ebracteolate. The unidirectional nature of floral development is also evident from young inner tepals. At early stages, the perianth and the entire flower is thus prominently monosymmetric with a median plane of symmetry, and a shift to floral polysymmetry takes place when the tepals of each whorl become uniform in size. Floral polysymmetry with an early monosymmetric phase (also termed transient monosymmetry) has been reported to frequently occur in plants with spikes or racemes and to be often expressed by delayed early development of the abaxial half of the flower (Endress, 1999, 2012). It was supposed that the delay is caused by the influence of the flower-subtending bract, and several hypothetical mechanisms of this phenomenon have been suggested (Endress, 1999; Remizowa et al., 2013). In Sciaphila, the flower-subtending bract is much larger than the flower and covers it at the time of tepal initiation, which makes it possible that the developmental delay is caused by the physical pressure of the bract (Ronse De Craene, 2018). On the other hand, the influence of the bract through an inhibitory positional signal cannot be excluded. Notably, the developmental delay discussed here is not pronounced in the SEM images of S. arfakiana published by Rübsamen-Weustenfeld (1991: Taf. 34a–c), but her work illustrates only three relevant flowers.

Ambrose et al. (2006) reported that “the first tepal primordium develops opposite the bract” in Lacandonia schismatica E. Martínez & Ramos. The illustration provided by Ambrose et al. (2006: fig. 9) shows an abaxial delay of perianth development similar to that in Sciaphila. In contrast to Sciaphila, the published image of flower of L. schismatica at the stage before gynoecium initiation (Ambrose et al., 2006: fig. 9) provides no evidence of the arrangement of the tepals in two whorls. Late developmental stages and anthetic flowers of both species of Lacandonia E. Martínez & Ramos show six equal tepals basally united in a tube (Martínez & Ramos, 1989; Vergara-Silva et al., 2003; Ambrose et al., 2006; Rudall, Alves & Sajo, 2016). Thus, the perianth of Lacandonia can be interpreted as either single-whorled or two-whorled. Notably, the perianth of Triuris Miers, the closest phylogenetic relative of Lacandonia (Mennes et al., 2013), consists of only three tepals (Maas & Rübsamen, 1986; Vergara-Silva et al., 2003) and is therefore clearly single-whorled.

Sciaphila reveals a highly unusual sequence of development of perianth parts: the outer whorl of tepals initiates as separate primordia, followed by formation of the perianth tube and finally by the initiation of free lobes of the inner tepals on the tube surface. Thus, the outer tepals demonstrate late congenital fusion, whereas the inner tepals exhibit early congenital fusion with adjacent outer tepals. The terms early and late congenital fusion (Sokoloff et al., 2018) are derived from the concept of early and late sympetaly (Erbar, 1991; Leins & Erbar, 1997, 2010) and can be applied to describe a wide array of developmental processes in plants. The perianth tube itself appears to be of an intermediate (or perhaps dual) nature with respect to this feature in Sciaphila, as it is characterized by a combination of early and late syntepaly. A perianth tube of this type is apparently rare in angiosperms. A similar example is that of the calyx tube of Coronilla L. (Fabaceae), implying early synsepaly between two of the sepals and late synsepaly between the other three ones (Sokoloff et al., 2007, 2018).

Nature of the stamen appendages

The stamen appendages reported here for some species and specimens of Sciaphila are to be compared with so-called filamentous structures described in detail by Rudall (2008), that is, the floral organs of Triuridaceae with an elongate shape, which are supposedly not homologous to any floral element or its part. Although Rudall (2008) did not mention the stamen appendages of Sciaphila among the filamentous structures, their homology is not really obvious. As we described above, the appendage is attached close to the base of the stamen filament, and thus can be interpreted as either a part of stamen or an organ independently inserted on the receptacle. In our opinion, it is important here that we did not observe any variation in the number and arrangement of the appendages with respect to the stamens: there was constantly a single appendage associated with each stamen, and it occupied a strictly adaxial position. This contrasts with the relations of the stamens and tepals, which were found to infrequently vary in number and arrangement rather independently (see also Rübsamen-Weustenfeld, 1991; Maas-van de Kamer, 1995). For this reason, we treat the appendages as parts of the stamens.

Within this interpretation, the stamen appendages of Sciaphila and a closely related Seychellaria have been homologized with various parts of the stamen, including a distal portion of the filament (Van de Meerendonk, 1984) and an extension/appendage of anther connective (Hemsley, 1907; Maas & Rübsamen, 1986; Rübsamen-Weustenfeld, 1991; Maas-van de Kamer, 1995; Maas-van de Kamer & Weustenfeld, 1998; Ambrose et al., 2006; Rudall & Bateman, 2006; Rudall, 2008; Merckx et al., 2013). In fact, the appendage represents a basal adaxial outgrowth of the stamen filament, whereas the filament itself continues beyond the appendage attachment and is terminated by the anther.

It is noteworthy that the three fossil species described by Gandolfo, Nixon & Crepet (2002) in the genera Mabelia Gandolfo, Nixon & Crepet and Nuhliantha Gandolfo, Nixon & Crepet are characterized by true connective extensions (i.e., supraconnectives). The authors assumed connective extensions to be a common feature of these fossils and the appendage-bearing species of Sciaphila and Seychellaria, and used this similarity to support the attribution of the fossil taxa to Triuridaceae. In fact, the stamens of Sciaphila and Seychellaria show a different morphology and lack supraconnectives, as shown above. The supraconnectives are known to be completely absent from the extant Triuridaceae (Maas-van de Kamer & Weustenfeld, 1998), which is an argument against placement of Mabelia and Nuhliantha in this family. The affinities of the fossil flowers reported by Gandolfo, Nixon & Crepet (2002) with Triuridaceae have already been doubted on the basis of pollen characters (Furness, Rudall & Eastman, 2002; Merckx, Bidartondo & Hynson, 2009).

Structure and development of carpels

In carpels of Sciaphila (Rübsamen-Weustenfeld, 1991; Rudall, 2008; this study), Seychellaria (Rudall, 2008) and other Triuridaceae where carpel development has been documented (Ambrose et al., 2006; Rudall, 2008; Rudall, Alves & Sajo, 2016), the ovules become discernible well before carpel closure. Late carpel closure (and early ovule development) was considered by Endress (2015) as a rare and apparently derived condition in angiosperms, found for example in derived families of Alismatales and in certain members of Poales, and also scattered among the eudicots, but unknown in any basal angiosperms. Notably, Endress (2015) has not listed any Pandanales in a review of cases of this condition. One can argue that the roundish structure in young carpels of Triuridaceae that we interpret as an ovule could be a short and thick adaxial wall of ascidiate zone of the carpel, whereas the ovule is formed after carpel closure (as demonstrated for Hopkinsia Fitzg., Restionaceae, by Fomichev et al. (2019). Special anatomical investigations of closed carpels at various developmental stages are needed to ultimately resolve this question. Currently we believe that the latter option is less probable, and this view is supported by illustrations of longitudinal sections of carpels of Sciaphila in Rübsamen-Weustenfeld (1991: Taf. 12d, e, 13f).

Endress (2015) highlighted the difficulty in interpretation of free carpels characterized by a single median ovule and precocious ovule development. This assumption is fully applicable to Sciaphila. Indeed, the determination of carpel zones in Sciaphila is somewhat controversial, which is a result of peculiar carpel structure. At first glance, the ovary seems to be entirely formed by the folded carpel blade, with the ventral slit extending directly from the ovary base. This pattern of ovary development fits the idea of a plicate carpel with basal (but not dorsal or ventral) ovule attachment, lacking an ascidiate zone. However, at early stages of carpel development we observed a short ventral part of the carpel, and the ventral (but not basal) part is ovule-bearing. Initiation of the ovule closer to the ventral side cannot be interpreted in any other way than as an evidence in favor of ascidiate zone of the carpel (although very short). To summarize, we assume the carpel of Sciaphila as consisting of both ascidiate and plicate zones, the former being extremely short and undetectable in late development and in definitive carpels. The level of ovule attachment corresponds to the cross-zone within our interpretation. The ascidiate zone is hardly evident below the cross-zone. The plicate zone forms the rest of the carpel, that is, the major part of the ovary wall and the style.

The ascidiate zone is visible during an apparently short period of carpel development, and the features of carpel shape that indicate the presence of the ascidiate zone (the central depression and the flanges connecting the two carpel parts) are small and comparable to the carpel cells in their size. This leaves a possibility for argumentations contra our views. It is remarkable that the presence of an ascidiate zone is accepted in the earlier works which discussed the question of carpel structure in Triuridaceae (Rübsamen-Weustenfeld, 1991; Igersheim, Buzgo & Endress, 2001). Very similar gynoecium and carpel development (though without a gynobasic position of the style) have been reported for Sagittaria L. (Alismataceae) (Huang, Wang & Wang, 2014). Notably, the carpels of Sagittaria were shown to be capable of sharing the pollen by means of the growth of the pollen tubes through the receptacle, the latter thus functioning as an extragynoecial compitum (Wang, Tao & Lu, 2002). The arrangement of the ovules of Triuridaceae close to the receptacle tissues can possibly be related with the same mechanism. Pollen tube growth from anthers to the ovules through the receptacle has been reported to perform the pollination in the bisexual flowers of Lacandonia (Márquez-Guzmán et al., 1993; Rudall, Alves & Sajo, 2016).

Ovule position in Sciaphila can be compared to that in Triuris. Triuris brevistylis Donn.Sm. is known to have a highly similar carpel structure at early developmental stages, differing in the apparent attachment of the ovule to the receptacle (Ambrose et al., 2006: Fig. 54, 55). This phenomenon requires additional explanation, because angiosperm ovules are believed to be parts of carpels, and therefore are borne exclusively within their tissues (Endress, 2019). This inconsistency can be resolved by employment of the notion (supported by histological data) that young carpels are “embedded” within the receptacle (Van Heel, 1983, 1984; Endress, 2019). It thus can be accepted that the ovule of T. brevistylis is surrounded by the tissues of the ovary wall instead of being directly attached to the receptacle. These “embedded” carpel tissues surrounding the ovule possess an annular shape (in apical view), and can be treated as the basal part of the ascidiate zone.

The style in Sciaphila is entirely cylindrical, without a ventral furrow (see also Igersheim, Buzgo & Endress, 2001), and thus it seems to be unifacial, with the entire surface of the style formed by abaxial side of the carpel. However, in styles of Sciaphila, we observed no transitional area between bifacial and unifacial parts, which is usually present at bases of unifacial phyllomes (or their parts) (Franck, 1976). For this reason, there is a possibility that a narrow adaxial side is present throughout most of the style length. In the latter case, the style is to be described as a subunifacial (Ozerova & Timonin, 2009) structure. Apparently, Endress (2015) used the term “unifacial” to describe only the appearance of the style, but not its morphological nature, and implied a meaning of the cylindrical shape, opposed to the plicate condition. Both unifacial and hypothetical subunifacial types of the style differ from the plicate style in the absence of a ventral slit, and therefore of a postgenital closure.

Gynoecium phyllotaxis

Our data on S. arfakiana show a high variation in patterns of arrangement of the outermost carpels with respect to the tepals, with all possible options detected: a carpel in a tepal radius, a carpel in an intertepal radius, two carpels in a tepal sector. Nevertheless, we found a tendency to the paired arrangement of carpels in tepal sectors. In both available flowers clearly showing six tepals and twelve carpels, such a pattern was manifested all round. In cases with this feature investigated only in a floral sector, the pairwise arrangement was found together with the other types of arrangement.

In S. densiflora, the staminodes occupy the tepal radii and thus are most likely arranged in two whorls by analogy of the fertile stamens of the male flowers of the same species, at least in the flower with six staminodes (Fig. 31). The outermost carpels of S. densiflora alternate with the staminodes (though this is based on study of a single flower). One can suppose that the presence of the staminodes plays an important role in the transference of the positional information from perianth to gynoecium in S. densiflora, and this positional signal is significantly weaker in flowers of Sciaphila lacking the staminodes.

The three flowers of S. nana used for the investigation of the phyllotaxis of the entire gynoecium show a remarkable diversity of carpel arrangement. One of the flowers possessed regular parastichies and whorled structure of the gynoecium, with all whorls being 9-merous except for the two distal ones. The other flower had moderately smooth parastichies and the numbers of parastichies of opposite directions were inferred as 10 and 11. This can be seen as indicative of a special kind of spiral phyllotaxis with divergence angle of c. 34.3° (rather than c. 137.5° of the Fibonacci spiral). Alternatively, this can be viewed as a whorled system with non-integer merism, namely, a merism of 10.5 in several basal whorls. The choice between these two interpretations is rather conventional, as discussed by El, Remizowa & Sokoloff (2020) using examples from androecia of Nuphar (Nymphaeaceae). The third flower of S. nana examined here in detail is characterized by rather a chaotic carpel arrangement that cannot be assigned to accurate sets of parastichies. Notably, none of these flowers possesses a gynoecium with right and left sets of parastichies differing in more than one parastichy, that would unambiguously point to the spiral phyllotaxis. However, the studied examples demonstrate that it is still difficult to provide an unequivocal characteristic of carpel phyllotaxis for S. nana and the genus Sciaphila in general. Rather, we suggest to interpret this feature in the studied plants as a result of a balance between a particular kind of order and an irregular arrangement.

In a given flower, the balance seems to be shifted to one or the other side, probably depending of the conditions of the development of the flower (such as exact size of the floral apex and the tepals at the time of carpel initiation or prepatterning). The degree of irregularity of floral phyllotaxis is known to be positively correlated with the number of involved organs in angiosperms (Endress & Armstrong, 2011; Rutishauser, 2016; El, Remizowa & Sokoloff, 2020 and references therein), and thus the irregular patterns in carpel arrangement in Sciaphila are quite expected. The opposite tendency, that is, an ordered phyllotaxis, can possibly be inherited by a gynoecium from the preceding floral whorl through the positional signal produced by the perianth and/or androecium. This idea is consistent with the merism of the outermost whorl of carpels which is often divisible by three (Fig. 31), as assessed from diagramming of parastichies (S. nana: 9, 10 and possibly 10.5 carpels) and from direct counting of primordia at early stages of gynoecium development (S. arfakiana: usually 12 carpels, also 9 and 11 carpels; S. nana: 9 carpels). In contrast to the staminodes which are supposed, as discussed above, to govern rather strongly the position of the basalmost carpels in S. densiflora, the tepals seem to possess a relaxed control over the number and position of carpels. The precise patterns of transference of the positional information from the tepals to the carpels are either highly variable or species-specific: we observed a tendency to the presence of twelve outermost carpels in S. arfakiana and nine in S. nana. Both these figures are likely to be somehow related to the trimerous nature of the two perianth whorls, but imply different carpel arrangement with respect to the tepals. We interpret the carpel whorls with merism that is not divisible by three in Sciaphila as formed by a significant contribution of irregularity.

Our interpretation of gynoecium phyllotaxis in Sciaphila generally agrees with those of Rudall (2008) for Sciaphila and Seychellaria. Rudall (2008) described the carpel arrangement in Sciaphila major Becc. and S. tenella Blume as “somewhat chaotic”. She characterized the carpel arrangement in Seychellaria madagascariensis C.H. Wright and S. thomassetii Hemsl. as “spiral or chaotic”, and noted that the numbers of right and left parastichies in a gynoecium is apparently the same, which indicates, according to Rudall, “a whorled arrangement, although this becomes chaotic in places”. Thus, Rudall (2008) did not propose a conclusion for this question, which is readily understandable in the light of the broad variation uncovered in our investigation. It is remarkable that a gynoecium of Seychellaria thomassetii is coloured as having twelve parastichies (Rudall, 2008: Fig. 10B). This gynoecium could be 12-merous, a condition we also found in S. arfakiana, but this hypothesis should be tested by drawing a set of parastichies of the opposite direction.

Patterns of carpel arrangement in Sciaphila and Seychellaria outlined above allow questioning the interpretation of gynoecium phyllotaxis as spiral in Kupea martinetugei Cheek & S.A. Williams, a member of the tribe Kupeaeae (Rudall et al., 2007). It will be essential to clarify the occurrence of whorled or spiral carpel arrangement in Kupea Cheek & S.A. Williams by mapping positions of individual carpels as performed here for S. arfakiana and S. nana. It is possible that the gynoecium of Kupea is whorled or irregular as in Sciaphila.

Structure of inflorescence in Sciaphila and Seychellaria

Rübsamen-Weustenfeld (1991; see also Maas-van de Kamer & Weustenfeld, 1998) has summarized that the entire family Triurudaceae is characterized by racemose inflorescences, with the exception of most species of Seychellaria. In Seychellaria thomassetii, the inflorescence is also a raceme, whereas other species of this genus show diverse deviations in inflorescence structure, as outlined by Rübsamen-Weustenfeld (1991) and described in detail by Nuraliev, Cheek & Beer (2016). Seychellaria madagascariensis possesses a thyrse, with a cyme of up to four flowers in an axil of each phyllome of the main axis. The flowers of S. madagascariensis are known to be bracteolate, with usually two bracteoles per cyme, and thus the presence of the bracteoles in the flowers of higher orders of the cyme is still questionable. Seychellaria africana Vollesen has racemes with bracteolate flowers; this condition was regarded as intermediate between a raceme and a thyrse by Rübsamen-Weustenfeld (1991), see also Remizowa et al. (2013). A similar inflorescence structure was found in the recently described species Seychellaria barbata Nuraliev & Cheek, with the difference in the presence of underdeveloped flowers in the axils of bracteoles, which makes the inflorescence more close to a thyrse than to a raceme (Nuraliev, Cheek & Beer, 2016). Additionally, the bracteoles of some species of Seychellaria are laterally united with flower-subtending bracts (Nuraliev, Cheek & Beer, 2016), which complicates clarification of the morphological nature of bracteole and the inflorescence structure in this genus.

Our investigations in Sciaphila confirm the idea of complete absence of bracteoles and the truly racemose nature of inflorescence in the genus. Given that Sciaphila is likely to be paraphyletic with respect to Seychellaria (Mennes et al., 2013), we suggest that the cymose partial inflorescences of Seychellaria have appeared in the course of evolution from the single axillary flowers of Sciaphila, and the appearance of the bracteoles acted as a key innovation that allowed inflorescence branching.

Maas-van de Kamer (1995) indicated, apparently based on the study by Rübsamen-Weustenfeld (1991), that in the entire family Triuridaceae the median tepal of the outer whorl is always arranged abaxially (facing the flower-subtending bract). Indeed, it is likely to be so for all the representatives of the family with racemose inflorescences and lacking bracteoles, and it was observed in the species of Sciaphila studied here. However, we suppose that all the species of Seychellaria with bracteolate flowers possess variable floral orientation, as it has already been reported for Seychellaria barbata (Nuraliev, Cheek & Beer, 2016) and found to be common in other monocots which possess a single floral prophyll (Remizowa et al., 2013).

Euanthial interpretation of reproductive units of Triuridaceae in the light of new data

The reproductive units of Triuridaceae, traditionally termed flowers, were stressed to show certain features of inflorescences by Rudall (2003, 2008) and Rudall & Bateman (2006). Their suggestion of a pseudanthial interpretation stimulated comprehensive studies of these structures. As a result, the reproductive units of Lacandonia and Triuris were concluded to be homologous to flowers of the other angiosperms (Ambrose et al., 2006).

Our results allow consideration of the reproductive units of Sciaphila, the largest genus of Triuridaceae. In Sciaphila, structure, orientation and developmental patterns of the reproductive units appear to closely meet the groundplan of a typical lateral monocot flower lacking a bracteole (Endress, 1995; Remizowa, Sokoloff & Rudall, 2010; Remizowa et al., 2013). The male flowers of Sciaphila are commonly trimerous, with a two-whorled perianth, one- or two-whorled androecium and a precise alternation of all the floral whorls. Perianth development is abaxially delayed in the same way as in other monocots with lateral flowers lacking a bracteole. The female flowers are also six-tepaled. The morphologically bisexual flowers (including functionally bisexual and staminode-bearing) show sequentially arranged perianth, androecium and gynoecium, each corresponding to the same floral zone of a unisexual flower in number and arrangement of the elements. Thus, the available data on Sciaphila supports the idea of the euanthial nature of the reproductive units in the entire family Triuridaceae.

Taxonomic implications

Problematic assessment of anther structure in the genus Sciaphila

Anther morphology is widely used for taxonomic segregation and identification of species of Sciaphila. The anthers are believed to be uniformly dithecal in Sciaphila (Maas & Rübsamen, 1986; Merckx et al., 2013). Endress & Stumpf (1990) characterized the anthers of Triuridaceae (exemplified by Lacandonia, Sciaphila and Soridium Miers) as dithecal but at the same time synthecal and di- or trisporangiate, which is partly in contradiction with our findings and earlier investigations. Particularly, the anthers of Sciaphila are here proved not to be synthecal, but there is still a possibility that synthecal anthers are present in Soridium (Maas & Rübsamen, 1986: Figs. 15B and 15D). Clearly, the number of thecae in anthers of most Triuridaceae remains an open question, because they lack a pronounced connective, which makes them similar to monothecal anthers irrespective of the number of microsporangia.

Two variable characters are mainly used in taxonomy of Sciaphila, viz. the number of locules (= microsporangia, cells) and the number of lobes. Apparently, the number of locules and number of lobes are not necessarily equal in a given anther: for example, the anthers are stated to vary in the genus within 1–4-locular and 2–4-lobed by Van de Meerendonk (1984), who, however, has not indicated number of anther locules for most of the species in his account. In contrast to Van de Meerendonk (1984), other authors (Maas-van de Kamer, 1995; Maas-van de Kamer & Weustenfeld, 1998; Merckx et al., 2013) described the anthers of Sciaphila as 3–4-locular and mentioned 2-locular anthers only for some other genera of Triuridaceae.

Whereas the 3-locular condition seems to be easily recognizable (and apparently coincides with 3-lobed shape), evaluation of the 2- vs. 4-locular (and -lobed) condition is not always straightforward. In our investigation, it is demonstrated by the example of S. stellata. All the studied male flowers of this species possess 2-lobed (more precisely, 2-valved) anthers that strongly resemble 2-locular anthers, including the closed and dehisced ones. Most likely, it is the presence of two valves around a seemingly single cavity that led Averyanov (2007) to indicate the anther structure of S. stellata as unilocular. However, our investigation of the internal structure of an anther uncovered the presence of four locules. Thus, in 2-lobed anthers of Sciaphila the number of locules cannot be established with confidence by external observations alone. It means that the complete absence of 2-locular anthers in Sciaphila is highly probable, and the reports of such anthers in the genus (e.g. Xu, Li & Chen, 2011; Suetsugu et al., 2017; Suetsugu, Kinoshita & Hsu, 2019; Suetsugu & Kinoshita, 2020) are possibly misinterpretations and require careful verification. The case of S. secundiflora Thwaites ex Benth., a species morphologically very similar to S. stellata (see below), is particularly illustrative: the anthers of this species were described as 4-locular and 4-lobed by Van de Meerendonk (1984) but stated to be 2-locular by Suetsugu, Kinoshita & Hsu (2019).

Although 2-locular anthers are probably absent from Sciaphila, the remarkable diversity in anther shape could still have taxonomic significance in this genus. In order to make this character clear and unequivocally applicable, it is essential to precisely define the difference between the 2-lobed and 4-lobed conditions, that is, what is termed by the anther lobe for both closed and dehisced anthers.

Taxonomic status of Sciaphila stellata

Sciaphila stellata was described on the basis of two specimens collected in northern Vietnam (Averyanov, 2007). Since description, no more findings of this species have been reported, except for a specimen from Guangxi (China) (Jiang et al., 2011), which is very close geographically to the type location of the species, but the illustrations of this plant (along with the general taxonomic uncertainties regarding S. stellata) do not allow us to verify the identification.

Sciaphila stellata belongs to a taxonomically difficult complex of species, in which the earliest described species is S. secundiflora. The protologue of S. stellata notes the similarity of this species with S. secundiflora, but does not indicate explicitly the differences between them (Averyanov, 2007). The comparison is currently complicated by the lack of a commonly accepted species concept for this group. Within the treatment by Van de Meerendonk (1984), S. secundiflora is an extremely variable and widespread species (with a large number of heterotypic synonyms), and the original material of S. stellata (the type and the paratype) along with the specimens studied here would certainly fit this circumscription. Some other authors, in contrast, tend to accept the taxa of S. secundiflora complex as distinct species. Particularly, Suetsugu, Kinoshita & Hsu (2019) have reinstated two species from the synonyms of S. secundiflora using differences in size of some floral parts. The latter approach is difficult to follow, because some of the measurements are unknown for type specimens of some species from this complex (including S. stellata), and because the distinctness of the segregate species has never been tested by a quantitative analysis of a relevant amount of material.

Our identification of the specimens treated here under S. stellata was based on their remarkable similarity with the description and the images from the protologue of this species and on the relative geographical proximity of our specimens to its type location. The known populations of S. secundiflora complex geographically closest to our specimens are those of Sciaphila stellata, although they are found in another part of Vietnam. We argue that comprehensive morphological investigation is crucial for further clarification of taxonomic boundaries within this species complex, and the detailed descriptions verified by SEM images would allow accurate documentation of plant structure and avoid such uncertainties as the discrepancy in anther locules of S. secundiflora pointed above.

Striking morphological diversity of Sciaphila nana

Among the studied specimens of S. nana, the specimen Nuraliev et al. 972 differs remarkably from the others in several aspects of the male flowers. First, it showed a nearly globose knob at the apex of the inner tepals, which is sharply delimited from the narrow portion of tepal (vs. cylindrical, smoothly delimited knob). Second, the stamens of this specimen completely lack the appendages (vs. prominent appendages, usually much longer than the anthers). Additionally, the specimen Nuraliev et al. 972 possesses shorter male pedicel (± as long as flower vs. much longer than the flower) and somewhat larger male flowers at anthesis (ca. 3 mm vs. ca. 1.5–2 mm in diameter when perianth lobes spread; this is to be verified by more numerous measurements). The specimen in question was found within several kilometers of another specimen of S. nana, Nuraliev 498, that has “typical” morphology (as accepted here). However, the species of Sciaphila frequently occur in mixed populations, with sometimes just a few meters between the individuals of different species (as is evident from the other specimens studied here).

Both morphotypes outlined above generally fall within the current morphological concept of S. nana (Van de Meerendonk, 1984; Chantanaorrapint & Thaithong, 2004; Averyanov, 2007). Sciaphila nana is known to be characterized by significant morphological variation; moreover, it was suggested to unite this species with S. arfakiana (Guo & Cheek, 2010), which would make it even more variable, but this view has never been adopted and is not supported by our study. Particularly, stamen appendages are stated to be often present in S. nana in Flora Malesiana (Van de Meerendonk, 1984, “filaments often exceeding the anthers”) and are apparently absent from the Thai specimen assigned to this species (Chantanaorrapint & Thaithong, 2004). Ohashi et al. (2008) described S. nana as having stamens with “a minute cylindrical appendage” and also its male flowers as having three reduced carpels, a feature never mentioned elsewhere for this species. Short but prominent appendages are reported for the Korean populations of S. nana (Yim, Kim & Song, 2011). The stamen appendages are absent from the Myanmarese specimen of the species (Jin, Zhu & Mint, 2018). Averyanov (2007) stated “filaments hardly exceeding anthers” for S. nana; notably, his specimens listed under this species were collected in the same location as our “unusual” specimen Nuraliev et al. 972 (Chu Yang Sin National Park), and their photographs (Averyanov, 2007: Fig. 2C and 2D) are highly similar to it.

The remarkable diversity of Asian specimens of Triuridaceae with respect to the absence or presence of the stamen appendages, that substantially contributes to the appearance of the entire flower, has been acknowledged long time ago. The plants with appendaged stamens were formerly assigned to the genus Andruris Schltr., as opposed to Sciaphila s.s. which was stated to lack the appendages (Schlechter, 1912; Giesen, 1938). As shown later, Andruris and Sciaphila s.s. are indistinguishable in other respects; consequently, Andruris is currently being synonymized with Sciaphila (Gandolfo, Nixon & Crepet, 2002; Hsieh, Wu & Yang, 2003; Ohashi et al., 2008; Govaerts, Maas-van de Kamer & Maas, 2020; but see Maas-van de Kamer & Weustenfeld, 1998; Merckx et al., 2013), and furthermore, some of its species are treated as heterotypic synonyms of certain species of Sciaphila, including a great number of them under S. arfakiana (Van de Meerendonk, 1984). Thus, within the current views, this feature frequently varies at intraspecific level in Sciaphila.

Our preliminary conclusion is that two distinct morphotypes are indeed present in Vietnam, but the differences are insufficient (at least, at the current state of the knowledge) to treat them as separate taxa. The alternative option can become more plausible when a special comprehensive investigation is carried out that should include consideration of the corresponding characters of the type of S. nana. The occurrence of evolutionary lineages within this species and their congruence with the morphotypes and geographical distribution is to be tested by molecular phylogenetic data.

Sciaphila densiflora: nuances of identification and a new record for Vietnam

The specimen Nuraliev 1670 treated here as S. densiflora shows a high similarity with S. arcuata Aver. Sciaphila arcuata is a Vietnamese endemic which is known only from the type gathering collected 287 kilometers from the location of the specimen Nuraliev 1670 (Averyanov, 2007). The protologue of this species states that the male flowers are unknown, and the identity of S. arcuata is quite obscure for the reason that male flowers contain most of the characters necessary for comparison with the other species (Van de Meerendonk, 1984). During investigation of the isotype of S. arcuata (LE: LE01041991) we have found that there are several flower buds in the terminal portion of the inflorescence, which are almost definitely male according to the inflorescence structure common in the genus. However, destructive sampling of this collection is problematic due to its unique nature.

Within the treatment in Flora Malesiana (Van de Meerendonk, 1984), which also covers most of the species of Sciaphila known in the mainland Southeast Asia, S. arcuata is most similar, in our opinion, to the widespread S. densiflora. The protologue (Averyanov, 2007) contains a brief comparison of S. arcuata with S. micranthera Giesen, but does not mention S. densiflora. Sciaphila arcuata shares with both S. densiflora and S. micranthera a non-secund inflorescence (i.e., with flowers arranged all around and not turned to one side) and bearded tepals of female flowers. However, S. arcuata is closer to S. densiflora in the number of carpels in a gynoecium: there are 45 carpels illustrated in the drawing of S. arcuata (Averyanov, 2007: Fig. 1D), ca. 15–40 carpels in S. densiflora and ca. 15–25 carpels in S. micranthera (Van de Meerendonk, 1984). As is evident from the protologue and the isotype of S. arcuata, it differs from S. densiflora in subequal (vs. conspicuously unequal) and triangular-ovate (vs. long-triangular) free parts of female tepals and in the attachment of the style subapically on the carpel (vs. at the base or middle of the carpel). Therefore, we consider S. arcuata as a distinct, although poorly known, species until more evidence on its morphology and phylogenetic placement become available.

The specimen studied here (Nuraliev 1670) nearly perfectly fits the description and illustration of S. densiflora provided by Van de Meerendonk (1984) with respect to the inflorescence, male and female flowers. At the same time, it fails to fit S. arcuata by its pronouncedly unequal, long-triangular lobes of female tepals, the longer ones with a caudate apex and all of them lacking an apical knob, and by distinctly basal attachment of the style. Thus, we assign our specimen to S. densiflora and report a significant range extension for this species, which has earlier been known to inhabit Sri Lanka, Borneo, the Philippines, the Lesser Sunda Islands, the Maluku Islands, New Guinea and New Caledonia (Van de Meerendonk, 1984; Tsukaya & Suetsugu, 2014) and is presented here as a new record for Vietnam and for the entire mainland Asia.

Staminodes in Sciaphila densiflora and the uncovered similarity of S. densiflora with S. tenella

The flowers of S. densiflora (including S. flexuosa Giesen, S. longipes Schltr., S. nutans Giesen, S. reflexa Schltr., S. trichopoda Schltr.) are known to be entirely unisexual, that is, staminate and carpellate flowers without any morphological structures of the opposite sex (Schlechter, 1912; Giesen, 1938; Van de Meerendonk, 1984). Here we describe for the first time the female flowers of this species as morphologically bisexual (functionally unisexual), according to our observation that they possess the staminodes. Although these structures show very simple shape and lack any specific androecial features, we consider them to be homologous to stamens on the basis of their number and arrangement which correspond with the stamens of male flowers of S. densiflora. Furthermore, if the presence of staminodes in S. densiflora is considered, the groundplan of the female flowers of this species appears to be identical to that of bisexual flowers of another widespread Asian species, S. tenella (Giesen, 1938; Van de Meerendonk, 1984; Ohashi et al., 2008; Guo & Cheek, 2010; Chantanaorrapint & Chantanaorrapint, 2012). This can be viewed as additional support for a staminodial interpretation of the sterile organs of S. densiflora in question.

Apart from the bisexual flowers, S. tenella possesses male flowers which are of the same structure as those of S. densiflora, including the peculiar 3-locular anthers. In total, it appears that the presence of (functionally) female vs. bisexual flowers is the only character that clearly delimits S. densiflora from S. tenella. The other possible difference is the filaments that are sometimes described as basally connate for S. tenella (Van de Meerendonk, 1984), but this feature is most likely indistinct and variable, especially if S. maculata Miers is treated as the synonym of the latter species, as suggested by Ohashi et al. (2008) and accepted by Govaerts, Maas-van de Kamer & Maas (2020). Our finding of morphologically bisexual flowers in S. densiflora leads to the understanding of its closest similarity to S. tenella, and brings us to a question, if they indeed represent two distinct species or should be merged into one species. The latter option would be in agreement with the known morphological variability already documented for both species as well as with the geographical evidence. The question is to be resolved by investigation of morphology of greater amount of specimens and molecular phylogenetic studies.

A more general speculation can be suggested based on our results on S. densiflora, that is, that the morphologically bisexual flowers in Sciaphila are more widespread than is currently documented. In S. densiflora, the staminodes are very small (much smaller than the carpels), completely hidden between the perianth and the gynoecium and are undetectable without special examination. Therefore, similar staminodes may be overlooked in some other species of Sciaphila with female flowers (though we have confirmed their absence in three other species studied here). In support of this hypothesis, there is an example provided by Rübsamen-Weustenfeld (1991) who has observed bisexual flowers in the neotropical S. albescens Benth. which was previously known to be strictly monoecious.