The anatomical and developmental evidence support the morphological floral variation in all studied species. For Rhynchospora sparsiflora, a distinct ontogenetic sequence for the perianth formation was reported, highlighting a different developmental pattern for the perianth in Cyperoideae (Vrijdaghs et al. 2009). For Cyperus surinamensis, we reported the presence of bisexual and unisexual flowers within the same spikelet, differing from previous taxonomic descriptions (i.e. Araujo and Longhi-Wagner 1996; Chen et al. 2009; Ribeiro et al. 2015). The variation in the number of stamens of Carex male flowers was notable and caused by the fusion of primordia, corroborated by the staminal vasculature. Floral variation is supported by the anatomical and developmental evidence in each floral whorl, which is detailed discussed below.
Perianth – The presence/absence and variation on the number of perianth parts is commonly a stable character state for Cyperoideae tribes. For Cypereae (Bauters et al. 2014; Reutemann et al. 2014; Vrijdaghs et al. 2009, 2011), Cariceae (Gehrke et al. 2012; Smith 1966; Smith and Faulkner 1976) and Abildgaardieae (Reutemann et al. 2015) the absence of a perianth is a stable character state, whilst in Eleocharideae, a perianth is always present (San Martin 2014; Vrijdaghs et al. 2009). However, the stability on the presence/absence of perianth within tribe Rhynchosporeae is not a rule where this character state is remarkably labile (Lucero et al. 2014; Monteiro et al. 2017; Vrijdaghs et al. 2009). Moreover, the tribe also exhibits a notable lability on the number of perianth parts. This is the case of Rhynchospora sparsiflora, previously placed in Pleurostachys and described as exhibiting five perianth parts (Alves and Thomas 2015; Thomas and Alves 2008; Thomas et al. 2013). However, in the present study we observed the development of six perianth bristles, although the outer abaxial bristle is weakly developed. The absence or non-development of the outer abaxial perianth part may be related to the pressure applied by the glume subtending the developing meristem, as previously reported by Lucero et al. (2014). This observation highlights the relevance of developmental studies for taxonomic description, mainly in families with reduced structures such as Cyperaceae.
Whilst the presence and number of perianth parts can be quite labile in Cyperoideae flowers, the sequence of development is reported as a stable pattern, where the formation of perianth parts seems to start after the formation of the stamens and simultaneously with the appearance of the ovary primordium (Vrijdaghs et al. 2009, 2010, 2011). Based on our results, the late formation of a perianth was not observed in Rhynchospora sparsiflora, on the contrary, the outer perianth parts starts slightly before all stamens are formed, and before any sign of the ovary primordium. The same pattern was reported for flowers in Oreobolus R. Br. (Mora-Osejo 1967), which corroborates a distinct ontogenetic sequence in the perianth from the general pattern described to Cyperoideae (Vrijdaghs et al. 2009, 2010). These previous observations raise the question about the variation in the developmental patterns of the perianth in Cyperoideae, which may be explained by an initial development of perianth primordia, followed by an interruption of such development, and then a regain in growth after the appearance of the stamens. Further investigations are needed in Cyperoideae on this matter.
Androecium – The variation in the number of stamens in Cyperoideae flowers ranges commonly from zero to six stamens per flower (Vrijdaghs et al. 2009, 2010, 2011). Such observation is reinforced here by the variation among studied species, from Rhynchospora sparsiflora maintaining the trimerous condition in the androecium, with occasionally bisexual flowers in Cyperus surinamensis with a single stamen, and the extreme variation on Carex brasiliensis ranging from four to six stamens in the male flowers. The instability of the number of stamens within species is remarkable, which in terms, may hamper the taxonomic description of some species. This is the case of Cyperus surinamensis, which is described in taxonomic studies as having bisexual flowers with one stamen (i.e. Araujo and Longhi-Wagner 1996; Chen et al. 2009; Ribeiro et al. 2015). However, most of the flowers of C. surinamensis analysed are unisexual, consisting of a single gynoecium, with no vestige of a stamen primordium, confirming that they are strictly female. This is a pattern previously reported for Cyperus eragrostis Lam., with no correlation between sex and position of flowers in the spikelet (Barnard 1957). Furthermore, in Carex brasiliensis, male flowers exhibiting four to six stamens differ from the typical flowers of Carex which usually exhibit three stamens (Gehrke et al. 2012; Smith 1966; Smith and Faulkner 1976) and seems to be a derived condition, since Cyperaceae flowers ancestral condition is reported as exhibiting a single whorl of three stamens, opposite to the outer perianth whorl (Remizowa et al. 2010; Vrijdaghs et al. 2009).
The reduction of floral parts is a common tendency among Cyperoideae flowers. However, the increase on the number of stamens in male flowers of Carex brasiliensis is notable. Considering the development of such flowers, although two or more anthers share the same filament, we did not observe separate primordia fusing later in development. Moreover, we did not report a single primordium splitting into two equal daughter primordia. Therefore, we believe this increase of the number of stamens is not a case of an equal division of a primordium, known as dédoublement (Ronse De Craene and Smets 1993). However, the anatomical analyses shed light on such increase. Regarding the stamen vascularization, two patterns are found: 1) For flowers with four stamens, a single vascular bundle supplies each stamen; 2) For flowers with more than four stamens, two or three vascular bundles run into a single filament and supply separate anthers. This latter observation suggests that two or more stamen primordia are congenitally fused, since stamens are usually vascularized by a single vascular bundle (Puri 1951). The same pattern of vasculature was reported previously for male flowers of other species of Carex (Carex flacca Schreb., Carex nigra All. and Carex panicea L.) (Smith 1966; Smith and Faulkner 1976).
Although the anatomy clarifies the fused nature of stamens in the male flowers of Carex brasiliensis, the reasons concerning such random and extreme variation remain unclear. Here we highlight two distinct interpretations for the increase on the number of stamens in male flowers of Carex brasiliensis, 1) the higher number of stamens is due to a secondary splitting of initial primordia; 2) the higher number of stamens is explained by the nature of the male flower of Carex brasiliensis as an extremely reduced and condensed spikelet, bearing several male flowers. The first interpretation is based on previous report in other angiosperms exhibiting an increase in the number of stamens, which are commonly basally fused in the mature flower, due to a secondary division of the initial primordia, known as secondary polyandry (Endress 1994; Ronse De Craene 2022). However, in this current study we did not observe any evidence of a secondary splitting of the initial primordia in male flowers primordium. For the second interpretation, three main features here observed is noteworthy: 1) the flowers exhibit a dome-shape primordium, commonly observed in spikelets and differing from the elliptical floral primordium in Cyperoideae (Vrijdaghs et al. 2009, 2011); 2) a plexus with numerous vascular bundles is present in the receptacle, resembling the typical vascular pattern in racemose inflorescence in monocots with spiral phyllotaxis (Remizowa et al. 2013). This pattern differs from the consistent feature observed in Cyperoideae flowers, which is the presence of three main bundles in the receptacle that fuse to form a central vascular plexus before the divergence of floral parts (Blaser 1941a, b; Monteiro et al. 2017; Reynders et al. 2012). Therefore, we hypothesize that such features suggest the nature of the male flower of Carex brasiliensis as an extremely reduced and condensed spikelet, bearing four male flowers, each one composed of only one or more (two or three) stamens.
Gynoecium – The gynoecium in Cyperoideae flowers develops from a ring primordium surrounding a central meristematic zone (corresponding to the floral apex) as a result of the congenital fusion of carpels (Reynders et al. 2012). The annular ovary wall primordium was interpreted as causing a break in the strict positional pattern of the stigma lobes and allowing shifts in the number and position of stigmas responsible for the variation in pistil types in Cyperoideae (trimerous, dorsiventrally or laterally flattened dimerous) (Reynders et al. 2012). Considering that the position of styles and stigmas is strongly fixed and predictable in the floral meristem (Ronse De Craene et al. 2002; Ronse De Craene 2022), we believe that new gynoecium morphs are more likely to be a matter of mechanical constraint of the glume, which is in direct contact with the abaxial side of the primordium. In Cyperus surinamensis we observed that a ring is formed from which three lobes appear and the abaxial one is slightly less developed than the adaxial ones, probably due to the pressure applied on the gynoecium by the glume. This is also observed in Rhynchospora sparsiflora, where spatial constraints seem to be restricting the floral development on the abaxial side (in contact with the glume). Evidence for this are: 1) the undeveloped abaxial perianth part; 2) the delayed development of the abaxial stamen; 3) the presence of only two latero-adaxial stigmas. Such observations reinforce the hypothesis that one of the factors modulating the dimerous dorsiventrally flattened pistil in Rhynchospora sparsiflora is the mechanical constraint of the glume.