Morphological diversity and evolution of Centrolepidaceae (Poales), a species‐poor clade with diverse body plans and developmental patterns

Abstract

• Premise of the study: The small primarily Australian commelinid monocot family Centrolepidaceae displays remarkably high structural diversity that has been hitherto relatively poorly explored. Data on Centrolepidaceae are important for comparison with other Poales, including grasses and sedges.

• Methods: We examined vegetative and reproductive morphology in a global survey of Centrolepidaceae based on light and scanning electron microscopy of 18 species, representing all three genera. We used these data to perform a cladistic analysis to assess character evolution.

• Key results: Each of the three genera is monophyletic; Centrolepis is sister to Aphelia. Some Centrolepidaceae show a change from spiral to distichous phyllotaxy on inflorescence transition. In Aphelia and most species of Centrolepis, several morphologically distinct leaf types develop along the primary shoot axis and flowers are confined to dorsiventral lateral spikelets. Centrolepis racemosa displays secondary unification of programs of leaf development, absence of the leaf hyperphyll and loss of shoot dimorphism. Presence or absence of a leaf ligule and features of inflorescence and flower morphology are useful as phylogenetic characters in Centrolepidaceae.

• Conclusions: Ontogenetic changes in phyllotaxy differ fundamentally between some Centrolepidaceae and many grasses. Inferred evolutionary transformations of phyllotaxy in Centrolepidaceae inflorescences also differ from those in grasses. In contrast with grasses, some Centrolepidaceae possess ligulate leaves where the ligule represents the boundary between the bifacial hypophyll and unifacial hyperphyll. All the highly unusual features of the morphological‐misfit species Centrolepis racemosa could result from the same saltational event. Centrolepidaceae offer good perspectives for studies of evolutionary developmental biology.

When considering the diverse range of body plans that occur in angiosperms, there is highly uneven distribution of structural diversity along the phylogenetic tree. Several relatively species‐poor lineages appear to exhibit disproportionally high structural diversity. Among monocots, well‐known examples are the order Pandanales and a core group of the order Alismatales (e.g., Posluszny and Charlton, 1993; Rudall and Bateman, 2006). Although quantitative data on species richness are at least partially available (e.g., Salamin and Davies, 2004; Davies and Barraclough, 2007), insufficient correlative data exist on structural and developmental diversity in the morphologically most complex lineages, both at species and genus levels. Inferring organ homologies is typically problematic in groups with high structural diversity.

In this paper, we examine structural diversity and organ homologies in a species‐poor, but morphologically highly complex commelinid monocot family, Centrolepidaceae (order Poales), which includes three genera, Aphelia R.Br. (six species), Centrolepis Labill. (ca. 30 species), and Gaimardia Gaudich. (four species), with principal species diversity occurring in mainland Australia and Tasmania. We provide a comprehensive review of reproductive morphology and related vegetative characters (shoot and leaf morphology, including the transition from vegetative to reproductive growth) in all three genera of Centrolepidaceae. We focus on patterns of flower arrangement, as well as structure, development, and position of phyllomes associated with flowers and phyllomes of the primary inflorescence axis in Centrolepidaceae. Flower‐subtending bracts are essential in understanding inflorescence architecture (e.g., Prenner et al., 2009; Whipple et al., 2010), and knowledge of flower position in an inflorescence allows adequate analysis of the spatial arrangement of phyllomes associated with flowers (e.g., recognizing their adaxial, abaxial, or transversal position on floral pedicel/axis). We use the resulting morphological data for a cladistic analysis of Centrolepidaceae. The analysis allows discussion of the taxonomic and phylogenetic significance of morphological characters within the family, as well as character evolution within the order Poales.

Data on reproductive morphology in Centrolepidaceae are important for comparison with other Poales, including grasses and sedges. The evolutionary origins of the grass flower and inflorescence are among the most controversial topics in comparative and evolutionary plant morphology (e.g., Philipson, 1985; Clifford, 1986; Cocucci and Anton, 1988; Kellogg, 2001; Skvortsov, 2003; Rudall et al., 2005, Sajo et al., 2008, 2012, 2015,). This debate is fuelled in part by the high species diversity of the grass family (Poaceae) and their prominence both in worldwide ecosystems and agriculture, but also because grass flowers and spikelets are considered to be particularly complex and specialized compared with the reproductive structures of many other monocots. Some sedges (Cyperaceae) are also problematic with respect to interpretation of flower‐inflorescence boundaries as well as interpretation of partial inflorescences as spikelets or monochasia (e.g., Vrijdaghs et al., 2007, 2010; Prychid and Bruhl, 2013). Centrolepidaceae are relatively species‐poor, economically insignificant and geographically restricted, yet the origin and the morphological identity of their reproductive structures are enigmatic, as in grasses and sedges.

Background: Systematics and morphology

Centrolepidaceae are most closely related to Restionaceae; indeed, some recent molecular analyses (as well as an earlier morphological study: Linder et al., 2000) have indicated that Centrolepidaceae are nested phylogenetically within Restionaceae and could be classified as a subfamily (Briggs and Linder, 2009; Briggs et al., 2010, 2014). This close relationship is supported by several morphological and anatomical characters, including the shared occurrence of bisporangiate, monothecal anthers (e.g., Hamann, 1962, 1975; Linder and Rudall, 1993).

Restionaceae (and the closely related Anarthriaceae) are perennial and often robust plants, but Centrolepidaceae are mostly miniaturized annuals (Fig. 1A–C), or cushion‐forming perennials. In Centrolepidaceae, leaf blades are typically longer than the sheath. In contrast, leaves in Restionaceae are typically reduced and leaf blades, when present, are short. However, the occurrence of well‐developed leaf blades in Restionaceae seedlings is consistent with the hypothesis that Centrolepidaceae evolved from Restionaceae via neoteny (Linder and Caddick, 2001). With very few exceptions (Linder et al., 1998; Briggs et al., 2014), Restionaceae are dioecious, bearing unisexual or functionally unisexual flowers. In contrast, dioecy is not recorded in Centrolepidaceae and reproductive units are more commonly bisexual.

Figure 1.

Photographs of (A–C) general habit of some Centrolepidaceae and (D–E, SEM) an example of bizarre flower morphology in Centrolepis. (A) Aphelia cyperoides, Western Australia. (B) Centrolepis strigosa, South Australia. (C–E) Centrolepis banksii. (C) General habit, Northern Territory, Australia. (D) Young flower. (E) Gynoecium and androecium of preanthetic flower. a = anther; abp = abaxial phyllome associated with a flower; adp = adaxial phyllome associated with a flower; c = carpel; cs = common stalk of gynoecium and stamen (possibly receptacular); fc = floral center (possibly receptacular); fsb = short phyllome associated with a flower, presumed flower‐subtending bract; gs = stalk of gynoecium (possibly receptacular); sf = stamen filament; sg = developing stigmas. Scale bars: D = 40 µm; E = 100 µm.

Whereas flower identity is clear and uncontroversial in Restionaceae, morphological interpretation of the basic reproductive units is contentious in Centrolepidaceae (reviewed by Sokoloff et al., 2009a). These units are interpreted either as flowers (euanthia) or as compact groups of reduced and often fused, unisexual, perianthless flowers (pseudanthia or synanthia). Proponents of the pseudanthial concept have viewed the unusual pattern of carpel arrangement in multicarpellate species of Centrolepis (in two rows along a stalk‐like structure, see Fig. 1D, E) as the principal supporting evidence (e.g., Eichler, 1875; Hamann, 1962). However, a recent study of gynoecia in all genera of Centrolepidaceae (Sokoloff et al., 2009a) found that the pattern of carpel arrangement does not contradict interpretation of the carpels as belonging to the same flower (see also Hieronymus, 1873). In flowers of Centrolepidaceae, carpels always form a single whorl, as in most other monocots (Remizowa et al., 2010), but the whorl becomes strongly asymmetric in Centrolepis due to one‐sided elongation of the receptacle (Sokoloff et al., 2009a). Another possible interpretation is that the floral center is formed by united ventral parts of carpels rather than by the receptacle, but this alternative interpretation does not affect any aspects considered here. If all the carpels of a reproductive unit are viewed as belonging to a single flower, there is no argument against interpreting the bisexual reproductive units of Centrolepidaceae as bisexual flowers. This interpretation is followed in the present paper. Multiplication of carpels within the single whorl and their placement at different levels due to one‐sided growth coupled with the occurrence of only one monothecal stamen (Fig. 1D, E) make flowers of some Centrolepis one of most unusual and peculiar among angiosperm flowers.

The fact that Restionaceae (and the closely related family Anarthriaceae) are typically dioecious does not contradict interpretation of the bisexual units of Centrolepidaceae as flowers, because flowers of Restionaceae and Anarthriaceae often retain staminodes and pistillodes. Indeed, there are a few Restionaceae species with bisexual flowers (Linder et al., 1998; Briggs et al., 2014).

In Restionaceae, as in many other monocots, the basic condition of the perianth is six tepals in two whorls of three, though reduction to four tepals occurs occasionally and the tepals are lost in female flowers of two African species (Linder et al., 1998). In Centrolepidaceae, flowers are commonly viewed as lacking a perianth (e.g., Hieronymus, 1873, 1886; Hou, 1957; Wu and Larsen, 2000), but this interpretation needs clarification.

Flowers of Restionaceae are normally organized in spikelets with spirally arranged flower‐subtending bracts, though this pattern can be obscured by a reduced number of flowers per spikelet and aggregation of small spikelets into compact groups (Kircher, 1986). Floral prophylls (bracteoles) are absent from Restionaceae and their records for some members of the family (e.g., Überfeld, 1925) are apparently based on incorrect interpretation of the inflorescence architecture (Kircher, 1986).

Inflorescences of Centrolepidaceae are diverse and controversially interpreted by various authors. Most species of Centrolepis possess two large bracts on the primary inflorescence axis, each subtending a lateral group of flowers arranged in a zig‐zag pattern. The morphological nature of these lateral structures was problematic due to the apparent absence of flower‐subtending bracts. Hieronymus (1873) viewed the zig‐zag pattern as evidence supporting the arrangement of the flowers into a cincinnus (a kind of monochasium). Following this interpretation, the entire inflorescence could be described as a thyrse (see Endress, 2010). In contrast, Bentham (1877, 1878) interpreted the lateral units in inflorescences of Centrolepis as dorsiventral distichous spikelets, so that the entire inflorescence should be viewed as a double spike. This interpretation implies that spikelet dorsiventrality in Centrolepis involves oblique insertion of flowers. Dorsiventral spikelets with pronounced oblique insertion of flowers are superficially remarkably different from nondorsiventral ones (Fig. 2A–F). On the other hand, in angiosperms that lack flower‐subtending bracts entirely, dorsiventral distichous spikelets and cincinni can appear superficially very similar (Fig. 2G–J). However, considerable fundamental differences exist between these two patterns of flower arrangement, allowing for divergent evolutionary and taxonomic interpretations (Fig. 2G–J).

Figure 2.

Differences between (A–C) nondorsiventral and (D–F) dorsiventral distichous spikelets and superficial similarity between (G, H) lateral dorsiventral distichous spikelets and (I, J) cincinni in angiosperms that lack flower‐subtending bracts. (A–F) Schematic views of angiosperm spikelets showing normal orientation of flowers in the (A–C) nondorsiventral type and (D–F) their oblique insertion in dorsiventral spikelets. (A, D) Side views. (C, F) Frontal views. (B, E) Details of A and D, respectively, with different sides of flowers labeled. Median and transversal planes of flower are indicated. Flower outlines are elliptical rather than rounded, to reflect a situation when a flower is monosymmetric in A–C, becoming asymmetric in D–F. Flower‐subtending bract gray; it can be reduced in taxa studied here. (G–J) Two contrasting interpretations of lateral partial inflorescences with zig‐zag flower arrangement and no visible trace of reduced flower‐subtending bracts. (G, H) Dorsiventral spikelet. (I, J) Cincinnus. Hypothetical positions of aborted bracts are indicated. f1, f2, f3, f4, etc. = subsequent flowers; fsb1, fsb2, fsb3, fsb4, etc. = their respective flower‐subtending bracts; ssb = spikelet‐subtending bract. Note that the mode of presentation of the spikelet diagram is unorthodox in H, as it implies that the spikelet axis is just below the plane of the diagram and parallel to it (consider the scheme in F and imagine a section cut just above its plane). The same mode of presentation of the lateral spikelets is used in the diagrams of inflorescences of Aphelia and Centrolepis below.

An SEM‐based developmental study of inflorescences in Centrolepis exserta (R.Br.) Roem. & Schult., which possesses inflorescences typical of Centrolepis, revealed sporadic reduced phyllomes in positions of flower‐subtending bracts predicted by the dorsiventral spikelet hypothesis (Sokoloff et al., 2010). Sokoloff et al. (2009b) described a new species, Centrolepis racemosa D.D. Sokoloff & Remizowa from northern Australia, with highly unusual inflorescences represented by a simple, nondorsiventral spike with flowers located in the axils of well‐developed foliage leaves. Comparison of patterns of orientation of monosymmetric flowers in C. racemosa and C. exserta further supports Bentham's (1877, 1878) interpretation of inflorescences in Centrolepis as fundamentally racemose, with lateral units being dorsiventral spikelets (Sokoloff et al., 2010). This interpretation is followed in the present paper.

The inflorescence of Gaimardia is a simple spike with a single functional flower (Hou, 1957), whereas Aphelia has a zig‐zag rachis with two rows of numerous large bracts. As highlighted by Cooke (1995), the rachis can be interpreted either as a single axis or as a sympodial system. However, this question remains unresolved, and we disagree with Cooke (1995) that it is one of terminology rather than substance. Inflorescences of the aberrant species C. racemosa share some important characters with those of the two other genera of the family, namely, the presence of well‐developed flower‐subtending bracts (shared with Gaimardia) and the occurrence of numerous bracts on the primary inflorescence axis (shared with Aphelia, if its axis is a monopodium). Sokoloff et al. (2010) suggested that the unusual inflorescences of C. racemosa are more likely derived rather than primitive, but this hypothesis has not been tested in any phylogenetic analysis.

Flowers of Centrolepidaceae can be associated with one, two, or three scale‐like phyllomes, termed secondary bracts (Cooke, 1992, 1995), floral bracteoles or floral prophylls (Hieronymus, 1873, 1886; Wu and Larsen, 2000), or glumes (Hou, 1957). According to Bentham (1877: p. 505), the scales “may possibly represent a reduced perianth, but are perhaps more likely to be bracts or bracteoles”. Interpretation of flowers in Centrolepidaceae as perianthless but bracteolate contrasts with the presence of a perianth and absence of bracteoles in the closely related Restionaceae; this apparent contradiction has not been addressed specifically in the literature. In the present paper, we simply use a descriptive term “phyllomes associated with flowers” for scale‐like phyllomes that are associated directly with flowers.

The presence or absence (Cooke, 1995) of phyllomes associated with flowers, as well as their number and spatial arrangement (Hou, 1957) are of great taxonomic importance in Centrolepis (see also Bentham, 1878). However, due to the small overall size of inflorescences with densely crowded flowers and the tiny size of some of these phyllomes, it is difficult to analyze them using herbarium material and a dissecting microscope (i.e., the common taxonomic approach). As a result, current knowledge of phyllomes associated with flowers of Centrolepidaceae is inadequate. For example, Cooke (1995) in his detailed revision of Australian Centrolepis did not describe the spatial arrangement of the phyllomes; a feature regarded as taxonomically important by Hou (1957), although the latter considered only the three Malesian species. In this present paper, using SEM investigations, we document the occurrence in some species of additional phyllomes that are not mentioned in the taxonomic literature.

MATERIALS AND METHODS

A list of specimens examined is given in Appendix 1. Our species sampling covers major aspects of morphological diversity within each genus of Centrolepidaceae. In Centrolepis, apart from the previously recognized taxa, we analyze a new species, C. milleri (its formal description is provided by Barrett and Sokoloff [2015]). This species represents one of several recently discovered, but so far undescribed species of Centrolepis (M. D. Barrett et al., unpublished manuscript). Centrolepis milleri is of interest because it possesses a combination of characters related to phyllomes associated with flowers that was previously unknown in Centrolepis but is similar to that of Aphelia.

Plant material was fixed in formalin‐acetic acid‐alcohol (FAA) or 70% ethanol and stored in 70% ethanol. For light microscopy (LM), material was sectioned using standard methods of Paraplast embedding and serial sectioning at 10–15 µm thickness. Sections were stained in safranin O and Alcian blue (Tolivia and Tolivia, 1987) and mounted in DPX (Agar Scientific, Stansted, UK) mounting medium. Alternatively, the material was embedded in Technovit (Heraeus‐Kulzer, Wehrheim, Germany) 7100, sectioned at 5 µm thickness, stained in toluidine blue O (0.5% w/v in distilled water) and mounted in DPX. Digital photomicrographs were taken using a Zeiss Axioplan photomicroscope. For scanning electron microscopy (SEM), material was dissected in 70% ethanol. Material examined at RBG Kew was dehydrated through absolute ethanol and critical‐point dried using an Autosamdri‐815B CPD (Tousimis Research, Rockville, Maryland, USA), then coated with platinum using an Emitech (Kent, UK) K550 sputter coater and examined using a Hitachi (Wokingham, UK) cold‐field emission SEM S‐4700‐II at 1 kV. Material examined at Moscow University was dehydrated through absolute acetone and critical‐point dried using a Hitachi HCP‐2 critical point dryer, then coated with gold and palladium using a Giko (Tokyo, Japan) IB‐3 ion‐coater and observed using a JSM‐6380LA SEM (JEOL, Tokyo, Japan) at 20 kV. Some images were merged and some colored digitally using Adobe Photoshop (San Jose, California, USA). Information on the morphology of Centrolepis racemosa in the descriptions below is based on Sokoloff et al. (2009b).

A list of characters used in the morphological cladistic analysis is given in Appendix 2. The matrix used in the analysis is available in Appendix 3. It contains 47 characters and 19 species of Centrolepidaceae, plus four outgroup species representing all three currently recognized subfamilies of Restionaceae s.s. (Briggs et al., 2014): Ceratocaryum argenteum (Restionoideae), Apodasmia similis (Leptocarpoideae), Sporadanthus gracilis, and Lepyrodia heleocharioides (Sporadanthoideae). Data on morphology of Restionaceae are taken from literature sources (Cutler, 1969; Kircher, 1986; Linder et al., 1998; Meney and Pate, 1999; Ronse De Craene et al., 2001; Briggs and Johnson, 2012; Hargreaves, 2013). Taxon sampling in Centrolepidaceae was dictated mainly by availability of fixed material, but we believe that it covers all major taxonomic groups of all three genera. Maximum parsimony analyses of the morphological data set were performed using the program WinClada (Nixon, 2002), with the Ratchet algorithm (10000 iterations, 20 trees to hold per iteration, 6 characters to sample). Unsupported nodes were collapsed in all trees. A bootstrap analysis was performed with 100 replications.

RESULTS

Morphological diversity of Centrolepidaceae

Vegetative morphology

Gaimardia and some species of Centrolepis are perennials. In our taxon sampling, the group of perennials in Centrolepis is represented by C. fascicularis Labill. The majority of Centrolepis species, as well as all species of Aphelia, are annuals. Inflorescences are terminal in all Centrolepidaceae, so that shoot systems are sympodial in perennials (e.g., Fig. 3A). Shoot branching is also extensive in annuals, and each branch is terminated in an inflorescence. Usually, branches of several orders are formed. In annuals, internodes are short (Fig. 3D, H) in vegetative parts of all shoots (except sometimes the epicotyls and very rarely the first internode after the epicotyl). In perennials, internodes can be more or less elongated (but this is not the case in C. fascicularis). In Gaimardia, Aphelia, and most species of Centrolepis, the last internode before the inflorescence (termed the basal internode: Weberling and Müller‐Doblies, 1989) is elongated and often conspicuous, raising the inflorescence above the rosette of vegetative leaves (Fig. 3A, D). However, in Centrolepis curta D.A.Cooke, C. racemosa and some other species of Centrolepis, the last internode before the inflorescence is short like other internodes.

Figure 3.

Shoots and phyllomes of Centrolepidaceae. (A–C) Gaimardia setacea. (A) Fragment of a shoot system with a terminal inflorescence and a lateral shoot. Leaves of the lateral shoot are numbered sequentially. (B) Vegetative foliage leaf viewed from adaxial side. (C) Lower phyllome of the primary inflorescence axis viewed from adaxial side. (D–G) Aphelia cyperoides (H.P. Linder 6077). (D) Lateral shoot viewed from side of its mother axis with leaves numbered sequentially. (E) Cotyledon, adaxial view. (F) Vegetative foliage leaf, adaxial view. (G) Lower phyllome of the primary inflorescence axis, adaxial view. (H–L) Centrolepis milleri. (H) Second order shoot with leaves numbered sequentially. 1 and 2 are foliage leaves; 3, 4, 5 are cataphylls; cataphyll 5 encloses a young inflorescence. There is a third‐order shoot in axil of leaf 1. Its first leaf (black star) is in a transversal position. (I) Cotyledon, adaxial view (arrowhead, boundary between bifacial and unifacial part). Inset, blade to lamina transition magnified. (J) Vegetative foliage leaf, adaxial view. (K, L) The first phyllome of the primary inflorescence axis. (K) Adaxial view. (L) Phyllome artificially unrolled to show sheath to blade transition. (M) Centrolepis strigosa subsp. pulvinata (R.Br.) D.A.Cooke, the first phyllome of the primary inflorescence axis dissected along its midline. The figure shows one half of the phyllome viewed from the adaxial side. Arrowhead, sheath to blade transition. (N, O) Centrolepis aristata. (N) Vegetative foliage leaf, sheath to blade transition. (O) The first phyllome of the primary inflorescence axis, sheath to blade transition. (P) Centrolepis racemosa, vegetative leaf, adaxial view (line drawing modified from Sokoloff et al. [2009b]). a = auricle; bd = blade of phyllome; is = inflorescence stalk (=basal internode); li = ligule of phyllome; ppa1, ppa2 = the first and the second phyllomes of the primary inflorescence axis; sc = seed coat; sh = sheath of phyllome; sl = subtending leaf of lateral shoot. Scale bars: A–K, M–P = 500 µm; L = 100 µm.

Phyllotaxis is distichous (or spirodistichous) to spiral with a Fibonacci pattern. Distichy is well recognizable on shoots with conspicuous internodes and relatively sparse branching, as in our material of Gaimardia setacea Hook.f. (Fig. 3A) and G. fitzgeraldii Rodway. As pointed out by Cooke (1992), the basically distichous phyllotaxis is obscured in most species of Centrolepis by the crowding of leaves from the numerous growing points forming each tuft, sometimes producing a false spiral phyllotaxis in species with numerous linear leaves such as C. strigosa (R.Br.) Roem. & Schult. Our data confirm the occurrence of plants with basically distichous (to spirodistichous) phyllotaxis obscured by extensive branching (e.g., in C. milleri). However, in many species with densely crowded leaves (e.g., Centrolepis fascicularis, C. strigosa, C. racemosa, Gaimardia australis Gaudich., and Aphelia nutans Hook.f. ex Benth.), we found true spiral phyllotaxis. When vegetative leaves are distichous, the pattern of phyllotaxis does not change on transition to an inflorescence. In species with spiral phyllotaxis, except C. racemosa, the pattern changes to distichy on transition to an inflorescence. In C. racemosa, the Fibonacci spiral continues directly to the inflorescence.

Cotyledon morphology was investigated only in annuals due to availability of plant material. The cotyledon (Fig. 3E, I) has an open sheathing base with membranous margins and a filiform photosynthetic lamina with a haustorial tip that remains associated with the endosperm and seed coat up to the time of anthesis and fruit set; there is no ligule. The main shoot produces foliage leaves immediately after the cotyledon in all species of Aphelia and Centrolepis for which we have observations.

Lateral shoots of Gaimardia possess a bladeless adaxial prophyll forming an open membranous sheath. In G. setacea and G. fitzgeraldii (Fig. 3A), the prophyll is followed by foliage leaves. The first foliage leaf is abaxial and the plane of distichy of lateral shoots is almost the same as that of the main shoot. In G. australis, the leaf following the prophyll is also abaxial, but has a reduced lamina; other leaves have normal laminas and follow a spiral pattern. In Aphelia (Fig. 3D) and Centrolepis (Fig. 3H), lateral shoots lack an adaxial prophyll, and the leaf series starts with two leaves in transversal or obliquely transversal positions (these can be interpreted as prophylls analogous to those of most eudicots). If the shoots are distichous, the plane of distichy of a lateral shoot is perpendicular to that of the main shoot. In Centrolepis, the first leaves of the lateral shoots are normally foliage leaves (Fig. 3H). An exception was found in plants of C. polygyna (R.Br.) Hieron., which bear branches of several orders. Branches of higher orders did not form foliage leaves at all and the two first phyllomes were bladeless cataphylls. In Aphelia, one (Fig. 3D) or two first leaves of lateral shoots are membranous bladeless cataphylls. These can be followed by foliage leaves (or a single foliage leaf; rarely all leaves are foliage leaves). However, one or two cataphylls may be the only leaves produced by a lateral shoot before its transition to an inflorescence. In small plants of Aphelia, the only vegetative foliage leaves are on the main shoot, but in vigorous plants, there also are foliage leaves on at least some lateral shoots. Both patterns are recorded in the same species (A. cyperoides R.Br.). We did not record shoot branching in the axil of the prophyll of Gaimardia, but branching often takes place in the axils of the first leaf (as well as subsequent leaves) in Aphelia and Centrolepis (e.g., in Fig. 3H there is branching in the axil of the first leaf of a second order shoot, labeled 1).

Vegetative foliage leaves of Gaimardia (Fig. 3B) possess an open sheath with membranous margins distally joining on the adaxial side to form a ligule. The lamina is elliptical in cross section and possesses one to three vascular bundles. In Aphelia and Centrolepis, the vegetative foliage leaves have an open sheath with membranous margins (Fig. 3F, J, N) and a filiform lamina with single vascular bundle. According to Cooke (1995, 1998), a ligule is absent from both Aphelia and Centrolepis, though a minute ligule is reportedly present in C. monogyna (Hook.f.) Benth. (Cooke, 1992). In our taxon sampling, a ligule was present in Aphelia (Fig. 3D, F) and absent in Centrolepis (Fig. 3J, N). Ligule length increases sequentially in leaves along the main shoot in Aphelia. The leaf blade is likely unifacial in Aphelia and Gaimardia, though developmental data are needed to test this hypothesis. In Centrolepis aristata (R.Br.) Roem. & Schult., the membranous margins of the sheathing base are continuous with longitudinal ribs in the proximal part of the lamina and there is no ligule (Fig. 3N). These ribs can be interpreted as borders of adaxial and abaxial surfaces and thus, blades of the vegetative leaves of Centrolepis are bifacial. However, the ribs are not clearly recognizable in many species of the genus (e.g., Fig. 3J), but again, there is no ligule.

In C. racemosa, the sheath to lamina transition is gradual rather than abrupt (Fig. 3P). The lamina of C. racemosa (a distal part of a leaf that is bent backward) is as long as—or slightly longer than—the sheathing base. In other species of Centrolepis and in Aphelia, the lamina is normally much longer than the sheath (Fig. 3D, F, J).

In most species of Centrolepis, one, two, or several (up to five in C. milleri) leaves below the inflorescence and above the vegetative foliage leaves are bladeless (Fig. 3H) and subtend no axillary buds or lateral branches. They have only open membranous sheathing bases, unvascularized in all species studied here (with a single bundle in some specimens of C. humillima Benth.: Cooke, 1992). These leaves were termed cataphylls by Cooke (1980, 1992, 1995). Normally, the term cataphyll is used for reduced bladeless leaves located at the beginning of a leaf series on a shoot, preceding foliage leaves and commonly functioning as a protective organ of a bud (Weberling and Müller‐Doblies, 1989). At first glance, this is not the case in Centrolepis. However, as the phyllomes on the main inflorescence axis often possess laminas, one could speculate that the reduced leaves below the inflorescence start a new quantum of growth of the same axis that first produced the vegetative foliage leaves (another elementary shoot, see Gatsuk, 1974; Mikhalevskaya, 2008). For this reason, we retain the term cataphyll for Centrolepis in the same sense as introduced by Cooke, but highlight a need for developmental studies to test the hypothesis regarding the occurrence of two elementary shoots.

Cataphyll(s) cover and protect young inflorescences until elongation of the basal internode. Elongation of this internode, when present, takes place due to intercalary growth at the time when the early stages of floral development are complete, at least for some flowers of an inflorescence. The intercalary growth is located in the proximal part of the basal internode, which thus remains under the protection of the cataphyll(s). Cataphylls below the inflorescence are absent from Gaimardia (Fig. 3A) and Centrolepis racemosa. Presence or absence of a cataphyll is variable in C. glabra (F.Muell. ex Sond.) Hieron. The lamina of the last leaf before the inflorescence is much shortened in C. aristata or occasionally entirely absent (Cooke, 1992; this study).

In Aphelia, cataphylls located above the foliage leaves and below the inflorescence are normally absent (Fig. 3D). However, when a leaf series on a lateral shoot is short and lacks foliage leaves, bladeless cataphylls are present just below the basal internode in Aphelia. The situation is then superficially similar to that of Centrolepis, but in reality the differences are significant. In most species of Centrolepis, reduced leaves are present after the vegetative foliage leaves, whereas in lateral shoots of Aphelia, they are located before the vegetative foliage leaves, if the latter are present.

Primary inflorescence axis and its phyllomes

Internodes of the primary inflorescence axis are either completely suppressed, or quite obvious, but even in the latter case, they do not exceed the lengths of the sheathing bases of phyllomes situated on it. The inflorescence axis is most commonly terete, except in the Australasian species of Gaimardia where it is strongly flattened perpendicular to the plane of distichous phyllotaxis (Figs. 4A, B, 5D).

Figure 4.

Diagrams of inflorescences of (A–D) Gaimardia and (E–I) Aphelia species. (A) G. fitzgeraldii, inflorescence with typical flower. (B) G. fitzgeraldii, inflorescence with atypical flower bearing a sterile carpel. (C) G. australis, interpreted as having a terminal flower. (D) G. australis, interpreted as having a lateral flower situated in the axil of the second phyllome of the primary inflorescence axis. The internal structure of the gynoecium has not been investigated in G. australis. (E) A. cyperoides (P.J. Rudall 908). (F) A. cyperoides (H.P. Linder 6077). (G) A. brizula. (H, I) A. drummondii. (E–H) Relative arrangement of organs as seen in serial sections perpendicular to the primary inflorescence axis; however, all the structures appearing at all consecutive nodes of the primary inflorescence axis have been placed in the plane of a single diagram. We interpret the structures in the axils of the phyllomes of the primary inflorescence axis of Aphelia as lateral dorsiventral spikelets whose axes are parallel to the plane of the diagram (and located just below this plane, as explained for the spikelet diagram in Fig. 2). In E, all lateral spikelets one‐flowered. In F and H, all but the first lateral spikelets are one‐flowered. In G, all but the two lower spikelets are one‐flowered. I is the same diagram as H redrawn in an interpretative form to show the axes of the lateral spikelets (black arrows). Green = vascularized phyllomes of the primary inflorescence axis; blue = long unvascularized phyllomes composed of two cell layers only (the outer and inner epidermises); magenta [RGB code 255,0,255] = short unvascularized phyllomes with more than two cell layers; yellow = stamens; red = carpels; black circle or ellipse = primary inflorescence axis; black star = sterile flower.

Figure 5.

Reproductive structures of Gaimardia (SEM). (A–F) G. setacea. (A–C) Postanthetic inflorescences in different views; note different blade length of the first phyllome in A and C. (D) Postanthetic inflorescence with the first phyllome of the inflorescence axis removed to show a fertile flower in its axil. (E) Postanthetic inflorescence with its continuation above the first phyllome removed. (F) Fruit (dissection of inflorescence and view angle as in D). (G–I) G. australis. (G) Inflorescence with fruit. (H) The same specimen as in G with the first phyllome removed. (I) Inflorescence with fruit releasing seeds. bd = blade of phyllome; f = stamen filament; fl = fruit locule; fs = fruit stalk; fv = fruit valve; g = gynoecium (of two united carpels); is = inflorescence stalk; pa = primary inflorescence axis (the inflorescence is apparently uniaxial); ppa1, ppa2 = the first and the second phyllomes of the primary inflorescence axis; s = style; sd = seed; sh = sheath of phyllome. Scale bars: A–E, G–I = 400 µm; F = 200 µm.

In anthetic and especially postanthetic inflorescences of Aphelia, the primary axis is zig‐zag (Fig. 6C). The zig‐zag curvature can be also seen in the lower node of the inflorescence of some Centrolepis species (e.g., C. exserta). In our opinion, the zig‐zag curvature does not indicate that the inflorescence axis is sympodial, as proposed earlier by Cooke (1995) with respect to Aphelia. Its monopodial nature is supported by two arguments. (1) Phyllomes of the primary inflorescence axis bear axillary structures (usually spikelets, see below), easily interpretable in the monopodial model. If a sympodial model is followed, these spikelets could be viewed as terminal, but in this case, the continuation of the primary inflorescence axis appears to be a lateral branch without a subtending phyllome, which is a relatively complex interpretation. (2) The zig‐zag curvature appears late in inflorescence development. The young inflorescence axis is straight (Fig. 6G).

Figure 6.

Inflorescences and flowers of Aphelia cyperoides (P.J. Rudall 908), SEM. (A) Lateral view of an inflorescence with six phyllomes on its primary axis. (B) Young phyllome of the primary inflorescence axis removed with its axillary spikelet and viewed from the adaxial side. (C) Lateral view of preanthetic inflorescence with four of six phyllomes of the primary axis (the blade of the fifth phyllome is also removed); for diagram of this particular inflorescence, see Fig. 4E. (D, E) Young single‐flowered spikelets of different orientation (left‐ or right‐handed); phyllomes of the primary inflorescence axis that subtend these spikelets are removed. (F) Top view of inflorescence tip with young one‐flowered spikelets in axils of phyllomes of the primary axis. (G) Lateral view of distal part of inflorescence with single‐flowered spikelets at different developmental stages in the axils of phyllomes of the primary inflorescence axis. a = anther; bd = blade of phyllome; g = gynoecium (single carpel); ia = apex of the primary inflorescence axis; p1, p2 = phyllome 1 and phyllome 2 associated with flowers (see text for details); pa = primary inflorescence axis; ppa = phyllomes of the primary inflorescence axis or their scars (numbered sequentially in A, C, G); sh = sheath of phyllome; arrowheads = ligule. Digital coloring (as in Fig. 4): green = phyllomes of the primary inflorescence axis or their scars, when removed; magenta = phyllome 1 associated with flower; blue = phyllome 2 associated with flower; yellow = stamens; red = carpels. Scale bars: A, C = 1 mm; B, D, E, G = 100 μm; F = 50 μm.

The primary axis of each inflorescence bears only two phyllomes (primary bracts in the terminology of Cooke, 1992) in all species of Centrolepis (Figs. 7A–C, 8A, 8B–D, 8H, 8I, 9A) except for C. racemosa, where it bears up to 18 phyllomes (only the upper part of an inflorescence is shown in Fig. 7D). Very rarely, in species of Centrolepis other than C. racemosa (e.g., C. exserta), a reduced third phyllome is observed. In species of Aphelia, even less vigorous inflorescences possess four or more phyllomes on the primary axis (Figs. 4E–I, 6). In our material, the highest number of phyllomes on the primary axis (14) was observed in some inflorescences of A. nutans (Fig. A in Appendix S1, see Supplemental Data with the online version of this article) and the lowest number (four) in some inflorescences of A. cyperoides, specimen H.P. Linder 6077 (Fig. A in Appendix S2, see online Supplemental Data). According to Cooke (1995), up to 18 phyllomes can be observed in the latter species. In Gaimardia, two (Fig. 5A–C, G) or three phyllomes are present. Phyllomes of the primary inflorescence axis are always vascularized in Centrolepidaceae.

Figure 7.

Diagrams of inflorescences of Centrolepis species. (A) C. banksii. (B) C. strigosa. (C) C. milleri. (D) C. racemosa. The diagrams are not interpretative and show relative arrangement of organs as can be seen on serial sections perpendicular to the primary inflorescence axis. The only difference with respect to actual sections is the placement in the plane of a single diagram, all the structures appearing on all the consecutive nodes of the primary inflorescence axis. In A–C, structures in the axils of the phyllomes of the primary inflorescence axis are (in our interpretation) lateral dorsiventral spikelets whose axes are parallel to the plane of the diagram (and located just below this plane, as explained for the spikelet diagram in Fig. 2). In D, flowers develop directly in the axils of the phyllomes of the primary inflorescence axis. Green = vascularized phyllomes of primary inflorescence axis; blue = unvascularized phyllomes composed mostly of two cell layers only; magenta = short unvascularized phyllomes with more than two cell layers (their presence is sporadic in A); yellow = stamens; red = carpels; black = primary inflorescence axis.

Figure 8.

Phyllomes of the primary inflorescence axis in Centrolepis. (A, B) C. strigosa. (A) Side view of postanthetic inflorescence. (B) Very early stage of inflorescence development with two phyllomes of its primary axis initiated (arrowheads, morphological tips of the phyllomes). (C) C. aristata, side view of preanthetic inflorescence. (D–G) C. polygyna. (D) Young preanthetic inflorescence (asterisk, second phyllome of the primary inflorescence axis). (E) Anthetic inflorescence. (F) Adaxial view of the first phyllome of the primary inflorescence axis. (G) Sheath to blade transition of the same phyllome with margins of the sheathing base removed (arrowhead, a ridge separating adaxial and abaxial surfaces). (H) C. glabra, anthetic inflorescence (arrowhead, sheath to blade transition). (I) C. milleri, side view of anthetic inflorescence. a = anther; ab = abaxial surface; ad = adaxial surface; bd = blade of phyllome; g = gynoecium; is = inflorescence stalk; ppa1, ppa2 = the first and second phyllomes of the primary inflorescence axis; s1, s2 = primordia of the first and second spikelet (located in axils of ppa1 and ppa2, respectively); sh = sheath of phyllome. Scale bars: A, C, E = 1 mm; B, 30 µm; D, G = 200 µm; F, H, I = 500 µm.

Figure 9.

Flower arrangement in Centrolepis. (A) C. pilosa, young inflorescence with the first phyllome of the primary axis removed. This phyllome subtends a three‐flowered spikelet with zig‐zag flower arrangement. The second phyllome has a sheathing base and a blade; it subtends a spikelet, in which two first flowers are visible. (B–E) C. banksii. (B, C) Different views of a young lower spikelet in inflorescence (its subtending phylome and continuation of the inflorescence axis are removed). The spikelet is distichous, with zig‐zag flower arrangement. White arrowhead, presumed flower‐subtending bract; black arrowhead, just initiated membranous phyllomes associated with a flower. (D, E) Tristichous spikelets. (D) Lower spikelet of an inflorescence removed and viewed from side of its subtending phyllome, i.e., from the abaxial side. (E) Lower spikelet of an inflorescence seen from side of the primary inflorescence axis, i.e., from the adaxial side (continuation of the primary axis is removed). a1, a2 = anthers of flowers of the upper spikelet of inflorescence (numbered after numbers of flowers); A1–A4 = anthers of flowers of the lower spikelet of inflorescence (numbered after numbers of flowers); abp = abaxial, membranous, unvascularized phyllome associated with a flower; adp = adaxial, membranous, unvascularized phyllome associated with a flower; bd = blade of phyllome; F1–F12 = flowers numbered according to their positions in spikelet; g1 = gynoecium of the first flower of the upper spikelet; G1–G5 = gynoecia of flowers of the lower spikelet (numbered after number of flowers); ia = inflorescence axis; ip = inner phyllome associated with a flower; ppa1, ppa2 = the first and the second phyllomes of the primary inflorescence axis; sh = sheath of phyllome. Green = vascularized phyllomes of primary inflorescence axis; blue (in A–C) = long thin membranous unvascularized phyllomes associated with flowers; magenta (in C) = short thick unvascularized phyllomes (presumed flower‐subtending bracts); yellow (in A–C) = stamens; red (in A–C) = carpels; light blue (in D, E) = lateral rows of flowers; orange (in D, E) = median row of flowers. Scale bars = 100 μm.

Phyllomes of the primary inflorescence axis are arranged distichously in all Centrolepidaceae (Figs. 4, 5A–C, G, 6A, C, G, 7A–C; 8, 9A; Figs. A, D in Appendix S1) except C. racemosa, where they are spirally arranged (Fig. 7D).

In Gaimardia, the lowermost phyllome of the primary inflorescence axis has a sheathing base with free membranous margins and a terete lamina. The phyllome base completely encircles the inflorescence axis (Fig. 5B, E). As in the vegetative leaves, a ligule is present. In contrast with vegetative leaves, the lamina is shorter than the sheathing leaf base (Figs. 3C, 5C). The lamina is further reduced and can be inconspicuous in the second (Fig. 5B) and third (if present) phyllomes in the inflorescence of Gaimardia.

In Aphelia cyperoides, all phyllomes on a primary inflorescence axis are morphologically similar (Figs. 1A, 6A; Fig. A in Appendix S2). They have a sheathing base and a lamina, which varies in length between accessions, being longer than the sheath (Cooke, 1995), as long as the sheath (see P.J. Rudall 908 – Fig. 6A) or shorter than the sheath (see H.P. Linder 6077 – Fig. 3G). In P.J. Rudall 908, the lamina bears an attenuated tip (Fig. 6A) composed of elongate cells (also illustrated in Cooke, 1995), though such a tip was absent in H.P. Linder 6077 (Fig. 3G). The phyllome base about one half encircles the inflorescence axis (Fig. 6C). The margins of the sheathing base are membranous and proximally extended into incurved auricles (Figs. 3G, 6A; Fig. A in Appendix S2). At the distal end of the sheathing base, the membranous margins of both sides join to form a short ligule (Figs. 3G, 6B). A row of densely inserted short and narrow typically two cellular hairs with spirally twisted tips is present along the proximal portion of the membranous margins, including auricles. A hair is attached to every marginal cell. In addition, thick hairs 2–4 cells long with spirally twisted tips are inserted along the abaxial surface of the sheathing base (Fig. 6A; Figs. A, B in Appendix S2). According to Cooke (1995), these thick hairs are sometimes missing from the two lowermost phyllomes in the inflorescence.

In three species of Aphelia (Cooke, 1995), the two proximal phyllomes of the primary inflorescence axis differ in morphology from the distal phyllomes. Of these species, we here investigated A. nutans and A. brizula F. Muell. All phyllomes normally have a base that encircles approximately one half of the circumference of the inflorescence axis (Fig. D in Appendix S1; Figs. A, E, F in Appendix S2), but the width of the base of the first phyllome is variable (Figs. C, E in Appendix S1). Distal phyllomes have sheathing bases only (a very short lamina is rarely present in A. brizula); the proximal phyllomes vary in presence or absence and length of the lamina (in A. nutans, the lamina is only rarely present and always short). Ligules are present only in phyllomes with well‐developed laminas. In both species, the distal phyllomes differ from the two proximal ones in the presence of incurved basal auricles and conspicuous hairs along the membranous margins (at least in their basal portion, extending into the auricles). In A. brizula (Figs. D–F in online Appendix S3), all phyllomes are strongly folded along their midline (the folding is not shown in the diagram in Fig. 4G), resulting in the entire inflorescence being distinctively flattened (Fig. F in Appendix S3). The distal phyllomes of A. brizula possess a row of thick (1)2(3)‐celled hairs in the basal portion of the keeled midrib (Fig. F in Appendix S3). The second cell is longer and thinner than the lower cell and its tip is either straight or slightly twisted. In A. nutans, all phyllomes are U‐shaped in cross section; proximal ones lack a keel, distal ones have an inconspicuous and glabrous keel (Fig. A in Appendix S1).

In two species of Aphelia (Cooke, 1995), the lowermost phyllome of the primary inflorescence axis differs in morphology from the other phyllomes (here termed the distal phyllomes). We examined one of these two species, A. drummondii (Hieron.) Benth. The lowermost phyllome base encircles about two thirds of the inflorescence axis (Fig. F in online Appendix S4). The phyllome is broader above the base and folded and the young inflorescence is enclosed completely by its overlapping membranous margins (Fig. A in Appendix S4). In our material, the phyllome consists almost entirely of a sheathing base; the lamina is extremely short and cap‐like. In some other specimens, a short lamina is present (Cooke, 1995). The distal phyllomes (e.g., ppa2 in Fig. D, Appendix S4) have sheathing bases only; a lamina is either absent or very short. All phyllomes are glabrous and lack basal auricles and ligules (Figs. A, D in Appendix S4). The distal phyllomes differ from the lowermost one in that the base is equal to about one half of the circumference of the primary axis and especially by the presence of a conspicuous dorsal keel bearing multicellular emergences.

In most species of Centrolepis, the two phyllomes on the primary inflorescence axis at most differ quantitatively, e.g., in relative lamina to sheath length (Fig. 8A). Both phyllomes have broad bases, almost completely encircling the inflorescence axis), with membranous margins and a terete or at least basally dorsiventrally flattened lamina (Figs. 3M, 3O, 8A, 8C, 9A, 9E). The lamina is longer than the sheath (C. aristata, Fig. 8C, C. fascicularis), about as long as the sheath (C. pilosa Heiron., Fig. 9A), much shorter than the sheath (C. exserta, Fig. 8A; C. strigosa, Fig. 3M), or almost completely reduced, being only occasionally recognizable as a very short mucro [C. banksii (R.Br.) Roem. & Schult.]. The phyllomes are either glabrous (Fig. 8C), or covered with multicellular hairs (Fig. 8A), but thick or incurved hairs similar to those found in some species of Aphelia are not recorded in Centrolepis. Basal auricles are always absent. Membranous margins of the sheath are often extended above the transition to the lamina as two distal auricles. The bases of these auricles can extend onto the ventral side of the phyllome and usually come close to each other. When the two auricles unite, a bifid (C. aristata, Fig. 3O; C. fascicularis), or entire (C. drummondiana, C. pilosa) ligule is present. These observations suggest that the lamina of the phyllomes on the primary inflorescence axis is unifacial, in contrast with the bifacial lamina in vegetative foliage leaves. However, in some species with monomorphic phyllomes (C. strigosa, Fig. 3M; C. exserta), the membranous leaf margins do not extend distally onto the ventral side and the inner surface of the lamina continues directly into the adaxial surface of the sheathing base. Here, the lamina (which is very short in these species) appears to be bifacial.

In one group of Centrolepis species, the two phyllomes on the primary inflorescence axis clearly differ in morphology (Fig. 8D, H, I). Of these, our sampling covers three species. In all three species, the base of the upper phyllome about one half surrounds the inflorescence axis, but that of the lower phyllome almost completely encircles the axis (Figs. 3K, 8D, 8E, 8H, 8I). In C. polygyna (Fig. 8D–G) and C. glabra (Fig. 8H), the lower phyllome has a sheathing base with membranous margins and a long or short filiform lamina, while the upper phyllome lacks a lamina (Fig. 8D, H); both phyllomes lack any hairs or papillae. In C. polygyna (Fig. 8F, G), the membranous margins of the sheathing base extend distally into the ventral side of the lower phyllome, but their bases are not adjacent to each other, as described above. In the space between their distal ends, a transverse ridge is present (Fig. 8G). Anatomical investigation clearly shows that this ridge is the boundary between the adaxial and abaxial surfaces, whose histological patterns differ dramatically in the bifacial sheathing base. However, the lamina is unifacial, rounded in cross section, with a uniform epidermis and subepidermal chlorenchyma across the section. In C. glabra, the membranous margins do not extend distally onto the ventral side and the lamina is slightly flattened (Fig. 8H). This leaf could be bifacial, but more comparative and developmental data are required to test this hypothesis. In C. milleri (Figs. 3K, 3L, 8I), the lower phyllome lacks hairs or papillae. It has a laterally compressed sheath (Figs. 3K, 8I) that extends distally into very short auricles and a lamina that is longer than the auricles (Fig. 3L). The upper phyllome has a sheath with a smooth central region and papillate membranous margins distally prolonged into very short auricles and a lamina that is about as short as the auricles (Fig. 8I). In both phyllomes, the margins of the sheathing base extend distally only slightly onto the ventral surface, so that there is a clear gap between them (Fig. 3L), as in the lower phyllome of C. glabra. On the other hand, the lamina is rounded or almost rounded in cross section, as in the lower phyllome of C. polygyna.

In C. racemosa, the lowermost (of up to 18) phyllomes on the primary inflorescence axis are similar to the vegetative leaves (in particular, the lamina is unquestionably bifacial); their length gradually decreases along the inflorescence axis. In other examined members of the family, phyllomes on the primary inflorescence axis differ morphologically from the vegetative foliage leaves as well as from the cataphylls (if the latter are present).

For early development of the phyllomes on the primary inflorescence axis, we have observations on all studied species of Aphelia (see Fig. 6F, G; Figs. D and E in Appendix S1, Figs. A and B in Appendix S3, Fig. F in Appendix S4) and some species of Centrolepis (C. exserta, Fig. 8B; C. strigosa, C. pilosa). In Aphelia, the primordium is crescent‐shaped and has as broad a base as a mature phyllome. If a lamina is present, it differentiates due to more rapid growth of the median part of the young phyllome. Simultaneously with lamina development, the membranous margins of the sheath base differentiate. At the next stage, a ligule appears. The attenuated tip of the lamina and the basal auricles (when present) are formed late in development. The long abaxial hairs are initiated much earlier than the short marginal hairs in A. cyperoides and the distal phyllomes of A. brizula. The phyllomes of Centrolepis also initiate as broad crescent‐shaped primordia. The lamina results from more rapid growth of the median part of the young phyllome. The sheath margins initially do not extend distally onto the ventral surface in all examined species. Later in development, such a continuation of the sheath margins onto the ventral surface occurs in C. pilosa.

Patterns of flower arrangement

In our interpretation, all Centrolepidaceae (except possibly G. australis, see below) lack a terminal flower or flower‐like structure and thus, the inflorescences are always polytelic. We interpret the inflorescences of C. racemosa (Fig. 7D) and Australasian species of Gaimardia (Fig. 4A, B) as simple spikes. In these taxa, each phyllome on the primary inflorescence axis directly subtends a flower (see Fig. 5D, E). We see no evidence for any more complex interpretation (e.g., for interpreting each flower as a reduced one‐flowered spikelet). In the reduced one‐flowered spikelets of C. polygyna (see below), the orientation of monosymmetric flowers with respect to the phyllome of the primary inflorescence axis is almost opposite that of C. racemosa. In C. polygyna, the stamen of a unistaminate flower is closer to the subtending phyllome of the one‐flowered spikelet (see Fig. 8D: phyllome is labeled by an asterisk), while in C. racemosa, the stamen is closer to the primary inflorescence axis (Fig. 7D).

In C. racemosa, all the phyllomes of the primary inflorescence axis subtend bisexual flowers or sometimes one of the phyllomes is empty. In all species of Gaimardia, the inflorescence normally develops a single fertile bisexual flower. In Australasian species of the genus (Figs. 4A, 4B, 5D, 5E), the lowermost phyllome of the primary inflorescence axis subtends a fertile flower, whereas the second phyllome may subtend a sterile flower. In the South American species G. australis, the only flower is borne after the second phyllome of the primary inflorescence axis (Fig. 5G–I). In the absence of a continuation of the primary axis, the flower can be viewed as either terminal (Fig. 4C) or located in the axil of the second phyllome (Fig. 4D). The absence of morphologically terminal flowers has been documented in Restionaceae (Kircher, 1986). On the other hand, the orientation of floral parts relative to the final phyllome in G. australis differs from the orientation of floral parts relative to the subtending phyllome in the Australasian species, which could indicate that the flower of G. australis is terminal.

The inflorescences of Aphelia and Centrolepis (except C. racemosa) are interpreted here as homothetic double spikes (i.e., with all simple spikes are lateral, see Weberling, 1989; Endress, 2010). They differ from those of Gaimardia and C. racemosa in a parameter called axiality that indicates the minimal branch order ending in a flower (Notov and Kusnetzova, 2004). Gaimardia (at least the Australasian species) and C. racemosa are diaxial plants (the minimal branch order ending in a flower is 2), whereas all other Centrolepidaceae are triaxial plants (i.e., the minimal branch order ending in a flower is 3).

In most species of Centrolepis studied here (Figs. 7A, 7B, 8A, 9A), each of the two phyllomes of the primary inflorescence axis subtends a spikelet with several flowers. In some species (C. pilosa, Fig. 9A; and especially C. aristata, Fig. A in Appendix S5), the upper lateral spikelet was more advanced developmentally than the lower lateral spikelet of the same inflorescence. In a group of species of Centrolepis with dimorphic phyllomes on the primary inflorescence axis (here represented by C. milleri, Fig. 7C; C. polygyna, C. glabra), only the upper phyllome subtends a spikelet (which is 4–6‐flowered in C. glabra and C. milleri and 1–3‐flowered in C. polygyna).

The lateral spikelets in inflorescences of Centrolepis in both species groups discussed are dorsiventral (Figs. 7A–C, 9D–C). Dorsiventrality is manifested by the fact that the flowers are oriented toward the adaxial side of the inflorescence axis. This dorsiventrality clearly reflects the spatial constraints created by the massive phyllomes of the primary inflorescence axis that subtend spikelets (the phyllomes are massive in the context of small‐sized plants of Centrolepis). Flowers of Centrolepis are either all bisexual, or sometimes proximal flowers in spikelets are bisexual, while the distal flowers (or flower) are female (e.g., C. drummondiana, C. glabra).

The lateral spikelets of Centrolepis are distichous (Figs. 7A–C, 9A–C), or sometimes also tristichous (see below). The plane of distichy is transversal with respect to the axil of the subtending phyllome. As the spikelets are dorsiventral, the two rows of flowers are shifted toward the adaxial side of the inflorescence axis. When it was possible to compare the two spikelets on the same inflorescence (e.g., Fig. 9A), they had the same orientation (either right‐ or left‐handed). In left‐handed spikelets (Figs. 7A, 7B, 9A–C), the first flower is located on the left transversal side with respect to the subtending phyllome of the primary inflorescence axis; in right‐handed spikelets (Fig. 11A, B), the first flower is located on the right side.

Figure 11.

(A–I) Flowers and (J) fruit of (A–D) Centrolepis curta and (E–J) Centrolepis strigosa. (A, B) Two views of a lower spikelet with three flowers initiated. The first flower has all organs initiated, the second has only phyllomes initiated, and the third is an undifferentiated primordium. (C) Two‐flowered spikelet with the second flower at a stage of an undifferentiated primordium. Note the presence of a presumed flower‐subtending bract as well as an unusual stamen position in the first flower. (D) Fully formed preanthetic flower. (E) Young flower viewed from a transversal basiscopic side. (F) Young flower seen from a transversal acroscopic side. (G) Top view of an isolated young flower. (H) Top view of an upper spikelet of inflorescence with several flowers at different stages of development. Median of a subtending phyllome of the spikelet was located on top left side of the figure. (I) Fully formed preanthetic flower. (J) Fruit after dehiscence. a = anther; abp = abaxial or abaxial/basiscopic transversal phyllome associated with a flower; adp = adaxial or adaxial/acroscopic transversal phyllome associated with a flower; c = carpel; F = flower; fsb = short phyllome associated with a flower, presumed flower‐subtending bract; ia = inflorescence axis; ip = inner phyllome assocated with a flower; sd = seed; sf = stamen filament; sg = stigmas; abbreviations with numbers indicate flower number in a spikelet (e.g., a2, anther of the second flower). Scale bars: A–C = 50 µm; D = 500 µm; E–H = 30 µm; I, J = 300 µm.

In some inflorescences of C. fascicularis, C. banksii, and C. exserta, the lateral spikelets had three rows of flowers, the third row being on the adaxial side with respect to the axil of the subtending phyllome of the primary inflorescence axis (Fig. 9D, E). Inflorescences with tristichous and distichous spikelets can occur on the same plant.

In Aphelia nutans (Figs. A, C, E in Appendix S1) and A. brizula (Fig. 4G), each of the two lower phyllomes of the primary inflorescence axis subtends a lateral spikelet with 2–4 male flowers. Each of the subsequent phyllomes subtends a one‐flowered female spikelet. In A. drummondii (Fig. 4H, I; Figs. B–E in Appendix S4), the lowermost phyllome of the primary inflorescence axis subtends a spikelet with up to eight male flowers. Each of the subsequent phyllomes subtends a one‐flowered female spikelet. The two examined accessions of A. cyperoides differ with respect to inflorescence morphology (Fig. 4E, F). In one specimen (P.J. Rudall 908), each phyllome of the primary axis subtends a one‐flowered lateral spikelet and all flowers are bisexual. In another specimen of this species (H.P. Linder 6077), the lowermost phyllome of the primary inflorescence axis usually subtends a two‐flowered lateral spikelet with a proximal bisexual and a distal male (Fig. B in Appendix S2) or sterile flower. The second phyllome subtends a single‐flowered spikelet with a bisexual flower (as in P.J. Rudall 908). The remaining phyllomes subtend single‐flowered spikelets with female flowers (Fig. E in Appendix S2).

Spikelets of Aphelia, when they contain more than one flower (e.g., Figs. B–E in Appendix S4), are distichous and dorsiventral in the same way as the spikelets of Centrolepis. Given the sister‐group relationship between the two genera (Briggs et al., 2010; this study), once the spikelet interpretation is adopted for Centrolepis, it would be most parsimonious if Aphelia could be interpreted in the same way. As in Centrolepis, the plane of distichy is transverse with respect to the median plane of the subtending phyllome of the primary inflorescence axis. Thus, both left‐ and right‐handed single‐flowered spikelets can be recognized by the position of the first flower in a spikelet. One‐flowered spikelets of Aphelia also can be left‐ or right‐handed, which can be seen because the insertion of the flower is clearly transverse. There is no clear pattern in the arrangement of left‐ and right‐handed spikelets along the primary inflorescence axis (e.g., Fig. 4E).

We emphasize that structures interpreted here as one‐flowered spikelets in Aphelia cannot be viewed merely as single flowers located in the axils of the phyllomes of the primary inflorescence axis. Our interpretation (Fig. 4I) is supported by data on the orientation of flowers relative to the phyllomes of the primary inflorescence axis. As outlined later, flowers of Aphelia are associated with two phyllomes of different morphology, which are spaced in a precise way with respect to a stamen and/or a carpel. The positions of the phyllomes allow us to see that the orientation of the first flower in two‐ or many‐flowered lateral spikelets is the same (or mirrored) as the orientation of the flower in one‐flowered spikelets. The similarity is especially obvious in H.P. Linder 6077 (A. cyperoides) because both the lowermost and the second spikelet have bisexual flowers of identical morphologies.

Phyllomes associated with flowers

Flowers of Gaimardia (Figs. 4A–D, 5D, 5E) are naked, i.e., they are not associated with any phyllomes, apart from conspicuous phyllomes that clearly belong to the primary inflorescence axis. In Aphelia (Fig. 10A–C) and Centrolepis (Fig. 10D–J), each flower is always associated with at least one phyllome, sometimes up to three. These phyllomes always lack vasculature. In the following description, we use “adaxial” and “abaxial” with respect to the corresponding sides of individual flowers as inferred from our interpretations of inflorescence morphology provided in the previous section. In distichous dorsiventral spikelets, the left and the right transversal sides of a flower are often not equal (Fig. 10A–C, E–I), because the flowers are asymmetric and obliquely inserted (such oblique insertion of lateral structures is recorded in many other plants with dorsiventral shoots; Goebel, 1913). Therefore, it is useful to distinguish between basiscopic and acroscopic transversal sides. The former is closer to the base of the spikelet and to the primary inflorescence axis; the latter is closer to the apex of the spikelet and to its subtending phyllome of the primary axis (Fig. 2D–F).

Figure 10.

Floral diagrams of species of Aphelia and Centrolepis. Orientation of the acroscopic and basiscopic sides of the flower is shown in the diagram at top left (see also Figs. 4 and 7). (A) Bisexual flower of A. cyperoides. (B) Male flower of A. cyperoides (similar in male flowers of other Aphelia spp.). (B^/) The first flower in male spikelet of A. nutans. (C) Female flower of A. cyperoides (similar to female flowers of other Aphelia spp.). (D) C. banksii (similar to some flowers of C. exserta). (E) Some flowers of C. exserta. (F) Some flowers of C. curta. (G) C. strigosa. (H) Centrolepis drummondiana (similar to C. aristata, C. pilosa, C. fascicularis). (I) C. milleri. (J) Bisexual flower of C. glabra (female flower of this species differs in the absence of a stamen). Thick blue arcs = long unvascularized phyllomes composed mostly of two cell layers; thin blue arcs = short unvascularized phyllomes composed of two cell layers; magenta = short unvascularized phyllomes with more than two cell layers; yellow = stamens; red = carpels; asterisk = position of spikelet axis apex. Dotted lines indicate outlines of organs or their parts that can be either present or absent in a species. abp = abaxial or abaxial/basiscopic transversal phyllome associated with a flower; adp = adaxial or adaxial/acroscopic transversal phyllome associated with a flower, ip = inner phyllome associated with a flower; p1 = phyllome 1 associated with a flower (presumed flower‐subtending bract); p2 = phyllome 2 associated with a flower.

In Aphelia (Figs. 4E–H, 6B–G; Fig. B in Appendix S1; Figs. C–E in Appendix S2; Figs. B, C, E in Appendix S3; Fig. C in Appendix S4), each flower is associated with two (rarely one: Fig. C in Appendix S1) phyllomes of different morphology and precisely fixed positions. Phyllome 1 is in the inferred abaxial position. It is short (0.1–0.2 mm), but relatively thick, triangular or elliptical in cross section. The epidermal cells are large and radially elongated with dark‐brown contents (Fig. D in Appendix S2). Subepidermal cells are parenchymatous and appear to be unspecialized (at least with a light microscope). Although phyllome 1 resembles a gland, no secretion was found on its surface in our fixed material, but no observations have been made in the field. Phyllome 1 was found in all examined flowers of all species examined here. It was not mentioned by Cooke (1995), but was observed by Hieronymus (1873) in A. cyperoides and female flowers of A. brizula. Phyllome 2 (the secondary bract in the terminology of Cooke, 1995) was found in all flowers of all species examined here, except in the first flower of the male spikelets of A. nutans (Fig. 10B^/ ; Figs. C, E in Appendix S1). Its position is intermediate between adaxial and basiscopic transversal. Phyllome 2 is much longer than phyllome 1 in mature flowers and its base is much broader (Figs. 6C; Fig. B in Appendix S1, Figs. C–E in Appendix S2, Fig. E in Appendix S3, Fig. C in Appendix S4). It is membranous and has two cell layers (the outer and inner epidermises) throughout its length (Figs. C, D in Appendix S2), sometimes with a discontinuous internal cell layer. The margins are covered by hairs or glabrous. Phyllome 2 is typically folded inward along an eccentric midline to enclose a stamen and/or a carpel in preanthetic flowers. In female flowers of A. brizula, the midline of the phyllome is keeled, with a row of thin hairs along the keel (Fig. D in Appendix S3). Phyllome 2 of Aphelia species is usually broadest just above the base and its margin closest to phyllome 1 is folded either outward (i.e., toward the primary inflorescence axis) or inward (both conditions can be found in the same species, A. cyperoides, Figs. 6C, Fig. B in Appendix S2). When the margin is folded outward (Fig. 6C), phyllome 2 superficially appears to be outermost relative to phyllome 1; when the margin is folded inward (Fig. B in Appendix S2), the phyllome 1 appears to be the outermost one. Since the bases of phyllomes 1 and 2 usually do not overlap, it is not possible to determine whether one of them is the lowermost or outermost. However, in cases where we observed real overlapping of the bases (lines of attachment), phyllome 1 was always outermost (Fig. C in Appendix S3); hence the numbering used here. Developmentally, phyllomes 1 and 2 appear in a very rapid sequence and phyllome 2 soon becomes crescent‐shaped (Figs. 6F, G; Figs. D, E in Appendix S1; Figs. A, B in Appendix S3; Fig. F in Appendix S4). At least in some cases, we were able to document that phyllome 1 is initiated slightly before phyllome 2 and that the primordium of the latter is relatively narrow at initiation.

In tropical Australian species of Centrolepis (here represented by C. banksii, C. exserta, C. curta, C. racemosa), each flower is associated with two membranous phyllomes that consist of two (epidermal) cell layers only. The two phyllomes are located in inferred abaxial and adaxial positions, respectively, though the adaxial phyllome often extends more onto the acroscopic side and the abaxial onto the basiscopic side (Figs. 9B, 9C, 10D–F, 11A–D). Each phyllome has a horseshoe‐shaped base half encircling the flower, so that the two phyllomes completely surround a preanthetic flower. The phyllome tip is obtuse, emarginate, or irregularly serrate. Most marginal cells are extended into hair‐like papillae (except at the phyllome base).

Some flowers of C. exserta (see also Sokoloff et al., 2010), C. curta (Fig. 11C), and C. banksii (Figs. 1D, 9C), in addition to the two membranous phyllomes, possess a third one in the abaxial position (magenta in Fig. 10D–F). The third phyllome is always much shorter than the two membranous phyllomes and its width is variable; when it is narrow, it resembles phyllome 1 in Aphelia. Like phyllome 1 of Aphelia, the third phyllome consists of more than two cell layers, but its epidermal cells lack conspicuous brown content. It is inserted at a conspicuously lower level than the two long membranous phyllomes, so that a kind of short pedicel is present between them. The presence of this third phyllome is variable; it can be present in some flowers of a spikelet and absent in the others.

Developmental data are available for C. exserta (Sokoloff et al., 2010; this study), C. curta, and C. banksii (this study). Sokoloff et al. (2010) concluded that the adaxial membranous phyllome is initiated before the abaxial membranous phyllome (they did not examine young stages of flowers associated with three phyllomes), but this conclusion was based on the size difference of the phyllomes associated with very young flowers. In the current study, having examined two more species and more material of C. exserta, we were unable to document young stages with only a single phyllome present. Therefore, the two or three phyllomes are initiated either simultaneously or in a very rapid sequence. Figures 9B and 9C (C. banksii) and 11A and 11B (C. curta) show spikelets where the youngest flowers already possess all phyllomes. In younger stages available, none of the phyllomes was initiated (e.g., a flower labeled F2 in Fig. 11C).

In C. strigosa, C. fascicularis, C. pilosa, C. drummondiana, and C. aristata, each flower is surrounded by three membranous phyllomes (Figs. 10G, 10H, 11E–I, 12A–D; Fig. B in Appendix S6; Figs. C, E, F in Appendix S5). Two, with broad bases, are located outside the third phyllome, which has a relatively narrow base. One of the outer phyllomes is inserted on the inferred adaxial and acroscopic transversal side of a flower (labeled adp in Figs. 11E–I and 12A–D; Figs. A, B in Appendix S6, Figs. C–F in Appendix S5), while another occupies the abaxial and basiscopic transversal side (labeled abp in Figs. 11E–I, 12A–D; Figs. A, B in Appendix S6, Figs. C–F in Appendix S5). The bases of these phyllomes do not overlap. The third (inner) phyllome (labeled ip in Figs. 11E, G–I, 12A, B, D, Figs. B, C in Appendix S6, Figs. B–F in Appendix S5) is adaxial to the basicopic one.

Figure 12.

Flowers of Centrolepis pilosa (A–D), Centrolepis milleri (E–I), and Centrolepis glabra (J, K). (A) Young lower spikelet of inflorescence viewed from side of primary inflorescence axis. The first and the second flowers of the spikelet are visible. (B) Flower isolated from a spikelet, abaxial view. (C) Spikelet in the same view as in A on a later stage of development (asterisks = abp). (D) Adaxial view of fully formed preanthetic flower (note that the flower is mirror‐shaped relative to B). (E, F) Different views of young spikelet with three flowers at successive stages of development (stars = adp2). (G) Preanthetic spikelet with ca. six flowers, side view. (H) Adaxial view of fully formed preanthetic flower. (I) Detail of H magnified to show a very short abaxial/basiscopic transversal phyllome. (J) Spikelet with gynoecium and stamen of the first flower removed to show the absence of any phyllomes on adaxial side. (K) Flower viewed from abaxial side. a = anther; abp = abaxial or abaxial/basiscopic transversal phyllome associated with a flower; adp = adaxial/acroscopic transversal phyllome associated with a flower; c = carpel; g = scar of removed gynoecium; ia = inflorescence axis; ip = inner phyllome associated with a flower; ppa2 = the second phyllome of the primary inflorescence axis; s = scar of removed stamen; sg = stigmas; st = style; abbreviations with numbers indicate flower number in a spikelet (e.g., a2, anther of the second flower). Scale bars: A–D, G, I, J = 100 µm; E, F = 50 µm; H, K = 300 µm.

Initiation of the phyllomes was studied in C. strigosa. The two outer phyllomes are initiated simultaneously; the inner phyllome appears slightly later. Early in development, the outer adaxial/acroscopic phyllome grows faster than other phyllomes (Fig. 11G). However, this is the shortest phyllome at the time of anthesis: it is 2–4 times shorter than the inner phyllome (Fig. 11I). The abaxial/basiscopic phyllome is either as long as the inner phyllome or intermediate between it and the adaxial/acroscopic phyllome. All three phyllomes surrounding the anthetic flowers of C. strigosa are thin (two cell layers), membranous and bear some hair‐like papillae distally.

In C. fascicularis, C. pilosa (Fig. 12A–D), C. drummondiana, and C. aristata (Figs. C, D in Appendix S5) the shortest phyllome at anthesis is the abaxial/basiscopic one, which is much shorter than the shortest phyllome of C. strigosa and in most cases is not obviously membranous. In C. pilosa and C. fascicularis, the abaxial basiscopic phyllome does not exceed the level of insertion of the lowest ovary locule and the margin is slightly lobed; in C. drummondiana and C. aristata, it bears several irregular protuberances, the longest of which extends to the upper end of this locule (Fig. D in Appendix S5). The outer adaxial/acroscopic and the inner abaxial/basiscopic phyllomes are conspicuous membranous structures composed of two cell layers (the outer and inner epidermises), sometimes with traces of mesophyll at the very base of the inner phyllome. Only these two phyllomes were described by Cooke (1992) as secondary bracts. The relative lengths of the two conspicuous phyllomes depend on species and stages of development (though we lack data on actual initiation of the phyllomes). In C. aristata, the inner phyllome is much longer than the outer adaxial/acroscopic phyllome at mid stages of development, but later the size difference is inconspicuous. In C. pilosa, the inner phyllome is shorter than the outer adaxial/acroscopic one at mid stages of development (Fig. 12A, B), but this is the longest phyllome in the mature flower bud and anthetic flower (Fig. 12D). In C. drummondiana, the inner phyllome is the longest at mid stages as well as at anthesis (Fig. C in Appendix S6). It is also the longest in anthetic flowers of C. fascicularis.

In C. milleri (Figs. 10I, 12E–I), each flower is associated with an adaxial phyllome (extending to the transversal acroscopic side) and an abaxial phyllome (extending to the transversal basiscopic side). The abaxial/basiscopic phyllome is very small (0.1–0.2 mm wide and 0.04–0.08 mm long), not exceeding the level of insertion of the lowest ovarian locule (Fig. 12G, I), entire or lobed. The adaxial/acroscopic phyllome (Fig. 12H) is longer than all other floral parts in the mature floral bud, completely membranous and distally lobed or serrate. Because this phyllome has a wide base and extends onto the acroscopic (rather than basiscopic) side of a flower (Figs. 10I, 12G), we believe that it is homologous with the outer adaxial/acroscopic phyllome rather than the inner phyllome of the species described above. The two phyllomes of C. milleri are initiated simultaneously and are similar in shape and size during early developmental stages (Fig. 12E, F).

In C. glabra (Figs. 10J, 11J, 11K) and C. polygyna, each flower is associated with only one phyllome (not recorded by Cooke, 1992), which is similar in length and position to the short phyllome of C. milleri, though it is thicker.

Androecium, gynoecium, and fruits

No species of Centrolepidaceae is dioecious. The fertile flowers of Gaimardia are always bisexual (Figs. 4A–D, 5D, 5H). In Centrolepis, the flowers are bisexual (Fig. 10D–J), but there are sometimes also female flowers that are distal in the spikelet. In most species of Aphelia (Fig. 4G, H), flowers and spikelets are unisexual, with female spikelets located above male spikelets. In A. cyperoides (Fig. 4E, F), either the flowers are all bisexual, or bisexual with male and female flowers co‐occuring on the same plant, in which case the female flowers are in separate distal spikelets and the male flowers are located in the same spikelets as bisexual flowers.

Flowers of Gaimardia have two stamens and two (rarely three) carpels. Flowers of Aphelia have a stamen and/or a carpel. Bisexual flowers of Centrolepis have a stamen and one to about 45 carpels. In Australasian species of Gaimardia (Figs. 4A, 5D), one stamen is in a position intermediate between adaxial and transverse and the other one is on almost maximum possible distance from it. The two carpels almost alternate with the stamens and also have oblique positions. When a third (sterile in our material) carpel was present, it was in an obliquely adaxial position (Fig. 4B). In the South American G. australis, it is equivocal whether flowers are terminal, or axillary (see above). The two carpels of G. australis are median and the two stamens are transverse (Figs. 4C, 4D, 5H)—the median plane is that which passes though the middle of the final phyllome of the inflorescence axis.

In bisexual flowers of Aphelia cyperoides, stamen position can be interpreted as intermediate between adaxial and transverse basiscopic (Fig. 10A). The stamen lies on the same radius as phyllome 2 (discussed earlier). The margins of the phyllome enclose the stamen before anthesis. The anther dehisces toward the acroscopic side of the flower. The carpel is inserted on the acroscopic side or on the side intermediate between acroscopic and adaxial (we infer carpel position by the orientation of its ventral slit). Male and female flowers of Aphelia spp. differ from the bisexual flower described earlier merely in the absence of a carpel, or a stamen (Fig. 10B, C). Rudiments of the opposite gender were not recorded in anthetic flowers or in early developmental stages.

In Centrolepis, the stamen is in the adaxial position or in a position intermediate between adaxial and transversal‐basiscopic (Figs. 9A–C, 10D, 10E, 10G–J, 11A, 11B, 11H, 12A, 12F). In some flowers of C. curta, stamen position can be interpreted as transversal‐basiscopic (Figs. 10F, 11C). The anther dehisces toward the acroscopic or abaxial side of a flower (or intermediate). Carpels in gynoecia of Centrolepis develop sequentially. The first‐formed carpel is abaxial and the last‐formed carpel is adaxial. To summarize, stamen position is similar in Aphelia and Centrolepis, while the position of the first‐formed carpel in Centrolepis differs from that of the solitary carpel of Aphelia (Fig. 10). There are no data on carpel orientation in unicarpellate species of Centrolepis.

Stamens of all Centrolepidaceae are bisporangiate and monothecal (Figs. 6D, 6E, 9A–C; Fig. C in Appendix S6), with a distinct filament (Figs. 11D; Figs. A, C in Appendix S1; Figs. B, E in Appendix S2; Figs. E, F in Appendix S3; Fig. B in Appendix S5). Carpels of all Centrolepidaceae possess solitary pendent orthotropous ovules, the ovary is (syn)ascidiate, and the stigma is plicate.

The gynoecium of Aphelia consists of a single carpel. In Australasian species of Gaimardia, the gynoecium is syncarpous with long synascidiate, short symplicate, and long asymplicate zones; the two carpels are inserted at the same level (Sokoloff et al., 2009a). There are no data on the anatomy of anthetic flowers in G. australis. In Centrolepis, the carpels form a single whorl, which is strongly monosymmetric in anthetic flowers. Initiation of the carpels is somewhat staggered and due to the one‐sided growth of the receptacle, the last‐formed carpels are inserted at higher levels than the first‐formed carpels; the result is that the carpels appear to be arranged in two rows along a stalk‐like extended receptacle. With the extended taxon sampling here, we are able to recognize two types of structure and development of pluricarpellate gynoecia in Centrolepis.

In anthetic flowers of some species of Centrolepis (C. aristata, C. drummondiana, C. fascicularis, C. glabra, C. milleri, C. pilosa, C. polygyna, and some others: Cooke, 1992), there is a common style bearing individual stigmas (Fig. 12D, G, H, K). Gynoecium development has been examined in C. fascicularis (Prakash, 1970), C. drummondiana, C. pilosa, C. polygyna (Sokoloff et al., 2009a), C. aristata, and C. milleri (this study). In these species, the gynoecium is circular in outline at the time of carpel initiation (Figs. 9A, 12A, 12B). The strong one‐sided elongation of the receptacle takes place during later developmental stages (see Fig. 12D). Early stages of gynoecium development can be interpreted readily as early formation of a synascidiate zone compatible with that of many other angiosperms.

Gynoecia of anthetic flowers of some other species of Centrolepis (C. banksii, C. curta, C. exserta, C. racemosa, C. strigosa) lack a common style; each carpel has its own, individual style (Figs. 1E, 11J). We did not record the basal connation of individual styles that was reported for C. strigosa (see Cooke, 1992). In all species examined (see also Sokoloff et al., 2009a), the one‐sided elongation of the receptacle is already conspicuous at the time of carpel initiation (Figs. 1D, 9B, 9C, 11A, 11B, 11E–H). In contrast with the previous group of species, the young carpels are connected with each other via the floral center only. Their lateral sides are free from each other (Figs. 1D, 9B, 9C, 11A–C, 11E, 11F).

In both types of gynoecium, the stamen filament base is sometimes united basally with the gynoecium (C. banksii, Fig. 1E; C. polygyna; see also Cooke, 1992). Our data show that fusion is congenital and occurs when the part of the receptacle below the stamen is partly involved in the one‐sided growth of the receptacle that causes the torsion of the gynoecium. Alternatively, the stamen filament can be united basally (for a very short distance) with the inner adaxial to basicopic phyllome associated with a flower (in C. drummondiana).

The fruits of Aphelia are indehiscent and dispersed together with persistent phyllomes subtending the spikelets. In Australasian species of Gaimardia (Fig. 5F) and Centrolepis (Fig. 11J), each fruit locule (derived from basal part of a carpel) dehisces dorsally to release a seed and the locules remain isolated from each other. The fruit of G. australis (Fig. 5I) is distally unilocular, dehiscing along dorsal lines of the two carpels to reveal two seeds. However, in the absence of developmental data, we do not know whether the unilocular part is formed by a long symplicate zone or by rupture of a septum in synascidiate zone. Another peculiar feature of G. australis is the long stalk‐like base of the fruit, which is longer than the dehiscing region that contains the seeds (Fig. 5H).

Morphological cladistic analysis of Centrolepidaceae (Fig. 13)

Figure 13.

Trees inferred from maximum parsimony analysis of Centrolepidaceae using the morphological character set. (A) One of the shortest trees (the first tree generated in the analysis) with bootstrap support indices indicated above branches. Arrowheads = nodes that collapse in strict consensus of all shortest trees. (B) The same tree as in A, with maximum parsimony reconstruction of character state changes. (C) Detail of tree with alternative placement of Centrolepis racemosa. In B and C, figures above branches indicate character numbers; those below branches are acquired character states (Appendix 2); black squares = unique synapomorphies, open squares = homoplasies; equivocal events (i.e., whose exact position depends on using Acctran vs. Deltran optimization) not shown.

Maximum parsimony analysis yielded 287 shortest trees (length = 98 steps, consistency index = 61, retention index = 83). Each of the three genera is monophyletic. Gaimardia is sister to a clade consisting of Aphelia + Centrolepis that received bootstrap support of 86%. Monophyly of Gaimardia is supported by two synapomorphies in our analysis: (1) inflorescences without lateral spikelets and (2) adaxial or obliquely adaxial tepals or floral prophylls absent. Formally speaking, none of these synapomorphies is uniquely derived in the context of our analysis. However, morphologies similar to Gaimardia occur in distantly related parts of our phylogenetic tree and we believe that we have enough evidence for monophyly of the genus. The Aphelia + Centrolepis clade has five synapomorphies, all uniquely derived in Centrolepidaceae: (1) annual life cycle (with a reversal in a group of species of Centrolepis here represented by C. fascicularis); (2) two first leaves of a lateral shoot (prophylls) transversal; (3) leaf epidermal cells overlapping (see Cutler, 1969); (4) terminal spikelet absent; (5) male or bisexual flowers unistaminate. Aphelia has bootstrap support of 91% and three synapomorphies: (1) female or bisexual flowers unicarpellate; (2) fruits indehiscent; (3) phyllomes subtending spikelets dispersed with fruits (the first and the third features are uniquely derived in the family). Synapomorphies of Centrolepis (bootstrap support of 74%) are: (1) the first leaf of a lateral shoot is a foliage leaf; (2) vegetative leaves not ligulate; (3) a cataphyll is present below inflorescence and above foliage leaves; (4) inflorescence with less than three lateral spikelets; (5) all spikelets with more than one flower; (6) gynoecium with more than four carpels (with a reversal in C. fascicularis and some other taxa not sampled here); (7) first formed carpel abaxial.

Three major groups of Centrolepis species can be recognized based on our phylogenetic analysis (see Fig. 13A). Groups 1 and 2 are clades with bootstrap support of 70% and 89%, respectively. Species of Group 3 form a clade in the tree first generated in the analysis (Fig. 13A, B), but not in all shortest trees.

Group 1 includes C. strigosa, C. curta, C. banksii, C. exserta, and C. racemosa. This clade is defined by, among other characters, a large membranous abaxial phyllome associated with a flower, the gynoecium elliptical in outline at the time of carpel initiation and with carpels united via floral center only. Within Group 1, the mostly temperate C. strigosa is sister to a subclade comprising tropical annual species (C. curta, C. banksii, C. exserta, C. racemosa).

Group 2 includes C. polygyna, C. glabra, and C. milleri. The most important shared features of these species are inflorescences with only one lateral spikelet (a uniquely derived synapomorphy in Centrolepidaceae, at least in our taxon sampling) and dimorphic phyllomes on the primary inflorescence axis.

Group 3 includes C. fascicularis, C. aristata, C. pilosa, and C. drummondiana. These four species have similar inflorescence morphology, an identical set of phyllomes surrounding the flowers and similar gynoecia. In our analyses, most shared features of the four species of Group 3 are interpreted as symplesiomorphies (and possibly ancestral character states in Centrolepis). In the trees where species of the Group 3 form a clade (Fig. 13), it is defined by synapomorphies in two characters that are homoplastic in Centrolepis. These are (1) long lamina of the lowermost phyllome of the primary inflorescence axis and (2) occurrence of two adaxial or obliquely adaxial phyllomes associated with flowers.

The shortest trees found in our analysis provide two contrasting scenarios for relationships between members of Groups 1, 2, and 3 in Centrolepis. In one group of trees (e.g., Fig. 13A, B), species of Groups 2 and 3 collectively form a clade that is sister to a clade formed by species of Group 1. The presence of a conspicuous common style is inferred as a synapomorphy of the Group 2+3 clade (Fig. 13B). In other shortest trees (e.g., Fig. 13C), species of Group 3 form a clade that is sister to a clade formed by species of Groups 1 and 2. The absence of a ligule of the lowermost phyllome of the primary inflorescence axis is inferred as a synapomorphy of the Group 1+2 clade (Fig. 13C).

DISCUSSION

Ontogenetic and phylogenetic transformations of phyllotaxis in centrolepids and grasses

Both Fibonacci spiral phyllotaxis and distichous phyllotaxis occur in Centrolepidaceae. When vegetative leaves are distichous, the distichy is continued into the primary axis of a terminal inflorescence. Centrolepis racemosa is the only member of the family with spiral phyllotaxis that is continuous from the vegetative part of a shoot to the primary axis of a terminal inflorescence. In all other Centrolepidaceae with spirally arranged vegetative leaves, a change to distichy takes place on transition to a terminal inflorescence. This type of ontogenetic transformation of phyllotaxis is fundamentally different from transformations that are characteristic of many grasses (Kellogg et al., 2013) and some other Poales (e.g., some Eriocaulaceae: Stützel and Trovó, 2013).

Whereas leaves are produced in a distichous pattern in many grasses, with the primordia separated from each other by an angle of 180°, the inflorescence branches are produced in a spiral. However, in several grasses from subfamily Pooideae, a change in phyllotaxis does not occur, and primary inflorescence branches are produced distichously (Kellogg et al., 2013). The absence of a change in phyllotaxis and the distichous arrangement of primary inflorescence branches are apomorphic features in grasses, based on character optimization onto a molecular phylogeny (Kellogg et al., 2013).

In contrast, our data on Centrolepidaceae show that the spiral phyllotaxy on the primary inflorescence axis is an autapomorphy of C. racemosa. In a more general context, this feature of C. racemosa can be viewed as a reversal, because outgroups of Centrolepidaceae such as Restionaceae have spiral phyllotaxy in their inflorescences (e.g., Kircher, 1986). Kellogg et al. (2013) reconstructed the spiral phyllotaxy in inflorescences of Centrolepis as a feature inherited from a common ancestor with Restionaceae. However, this conclusion was biased by insufficient taxon sampling in Centrolepidaceae. Their interpretation of flower arrangement in C. aristata as spiral (Kellogg et al., 2013; their fig. 1D and E illustrates a dorsiventral lateral spikelet) is contradicted by our results, which demonstrate that the phyllomes on the primary inflorescence axis are distichous in all Centrolepidaceae except C. racemosa.

Kellogg et al. (2013) highlighted the phylogenetic significance of differences in phyllotaxis within the two‐ranked pattern in grass inflorescences. They used the term “distichous” for cases in which the two ranks of primary branches are initiated at angles of 180° and “two‐ranked” if the ranks are less than180° apart on one side of the inflorescence. Thus, “distichous” is a subset of “two‐ranked.” The two terms are identical etymologically (Greek stikhos means a row or a line), so we prefer to use them as synonyms. The second condition, in which the two ranks are less than 180° apart on one side of the axis, is an example of pendulum symmetry, which is viewed here as the regular alternation of mirror‐shaped metamers on distichous shoots, where it is correlated with dorsiventrality (Charlton, 1997). Pendulum symmetry is characteristic of distichous shoots of various taxa across the angiosperms (Charlton, 1997, see also Goebel, 1882, 1913; Sokoloff et al., 2007; Prenner, 2013).

A kind of pendulum symmetry is present in the vegetative shoots of grasses, where it is manifested in regular alternation of left‐ and right‐handed leaves (identified using characters such as patterns of overlapping of open sheath margins, e.g., Macloskie, 1895). Grass rhizomes are also often dorsiventral, because axillary buds are displaced around the axis away from the midline (or the midline itself is displaced?); left‐hand and right‐hand displacement regularly alternate at nodes along a rhizome (Bell and Bryan, 2008). A potential correlation between pendulum symmetry in the vegetative shoot and the inflorescence axis would merit further exploration. Goebel (1913) observed that in some legumes and grasses, dorsiventrality characterizes the entire plant. As noted by Kellogg et al. (2013), the developmental basis for the regulation of dorsiventrality remains unknown in grasses. Pendulum symmetry is such a common feature in distichous shoots across angiosperms that we predict common patterns of developmental regulation, based on the regulation of distichy. Since pendulum symmetry occurs in model systems such as cereals and legumes [Lotus japonicas (Regel) K.Larsen, see Sokoloff et al., 2007], a comparative analysis of its genetic control is potentially achievable.

In Centrolepidaceae, distichous phyllotaxis of the primary inflorescence axis is not accompanied by pendulum symmetry. In contrast, the distichous lateral spikelets of Centrolepis and Aphelia demonstrate pronounced pendulum symmetry. In grasses, pendulum symmetry can be found in the primary axes of the inflorescence, but the ultimate units (spikelets) are not pronouncedly dorsiventral. Thus, patterns of pendulum symmetry differ between grasses and centrolepids.

Patterns of initiation of phyllotaxis on lateral shoots are of phylogenetic significance. In Gaimardia, the first leaf of a lateral shoot is adaxial. This feature is common among monocots, including grasses and Restionaceae, and therefore we interpret it as plesiomorphic in Centrolepidaceae. When shoots are distichous in Gaimardia, the plane of distichy is the same in both mother and daughter shoots. In contrast, in grasses, the plane of distichy is transversal with respect to the axil of a subtending leaf (both in vegetative parts and inflorescences, e.g., Smirnov, 1953). The arrangement of the first two leaves in the transversal positions is a synapomorphy of Aphelia+Centrolepis and when shoots are distichous, the plane of distichy of a lateral shoot is perpendicular to that of the primary shoot, as in grasses. These different patterns of initiation could indicate that distichy has appeared more than once during the evolution of the graminid clade.

Diversity and evolution of vegetative leaves and phyllomes of the primary inflorescence axis

Linder and Caddick (2001) suggested that the ligulate foliage leaves of Restionaceae and Centrolepidaceae have unifacial laminas. In their interpretation, the ligule margin represents a morphological boundary between the abaxial and adaxial morphological surfaces, so that both surfaces are present in the leaf sheath but only the abaxial one in the lamina. This interpretation is supported by the fact that the laminas are often elliptical to almost round in cross section in both families. In terete laminas, epidermal characters do not show a clear boundary between the abaxial and abaxial surfaces (Cutler, 1969). In Centrolepidaceae (Hieronymus, 1873; Cooke, 1992; Tillich, 1995, 2007; Kellogg et al., 2013; this study) and those Restionaceae that have epigeal germination (Linder and Caddick, 2001; Tillich, 2007), the terete photosynthetic part of the cotyledon (phaneromer sensu Tillich, 2007) resembles the lamina of foliage leaves. Morphologically, the phaneromer is the proximal region of the cotyledonary hyperphyll (Oberblatt). The hyperphyll of Centrolepidaceae and Restionaceae is interpreted as unifacial (Tillich, 2007), and it is tempting to adopt the same interpretation for the (broadly similar) laminas of epicotyledonary leaves.

On the other hand, there are arguments against interpreting all laminas in Restionaceae and Centrolepidaceae as unifacial. When the lamina has three vascular bundles (e.g., in Gaimardia spp.), the bundles are arranged in the same plane, as would be expected in a bifacial leaf (Cutler, 1969). A ligule is absent in many Restionaceae, and its presence is not universal in Centrolepidaceae. In vegetative nonligulate leaves of Restionaceae (Briggs et al., 2014) and Centrolepis (this study), the margins of the sheathing base are continued into the lamina, though the ribs indicating the presumed boundary between the abaxial and adaxial surfaces soon disappear along the length of a lamina, at least in our material of Centrolepis. Briggs et al. (2014) highlighted a narrow groove, interpreted as the adaxial surface, on the upper side of a lamina in Anarthriaceae and those Restionaceae that possess more developed laminae. Although Briggs et al. (2014) termed such laminae unifacial, they are, in their interpretation, fundamentally bifacial, with very narrow adaxial surface (subunifacial: see Ozerova and Timonin, 2009, for a review of terminology). Briggs et al. (2014) also recognized a narrow adaxial surface in the leaf lamina of Centrolepis and Gaimardia, but this observation seems less convincing based on their illustrations.

Arber (1922; see also Goebel, 1913; Cutler, 1969) described the lamina of phyllomes on the primary inflorescence axis of C. aristata as belonging to an extremely reduced ensiform type, with a second bundle above the median bundle, resulting from the fusion of two laterals. However, only one bundle is present in the vegetative leaves (Arber, 1922). This difference fits well with our data where only phyllomes on the inflorescence axis are ligulate.

We interpret the sheath and lamina of Centrolepidaceae as formed by the hypophyll (Unterblatt) and hyperphyll (Oberblatt), respectively. At least for phyllomes on the primary inflorescence axis, this interpretation is supported by our developmental data. Even if the concept of Unterblatt and Oberblatt is problematic for some monocots (Rudall and Buzgo, 2002), we consider it useful in Centrolepidaceae. Note that typical grass leaves (including the lamina) are interpreted as formed mainly by a hypophyll; the hyperphyll is restricted to an inconspicuous precursor tip (Vorläuferspitze), where the margins of the blade become confluent at the leaf apex (Kaplan, 2001). Therefore, ligules are not strictly homologous between Centrolepidaceae and grasses.

Centrolepis racemosa differs from other Centrolepidaceae in that leaves on the inflorescence axis (as well as vegetative leaves) lack clear differentiation into a sheath and lamina. In our interpretation, the leaves of C. racemosa have a reduced hyperphyll. Their proximal sheath‐like and distal lamina‐like regions (which gradually merge into one another) are both formed by the hypophyll. In this respect, the leaves of C. racemosa approach the concept of the grass leaf (Kaplan, 2001). In our view, almost the entire leaf of C. racemosa is homologous with a leaf sheath in other Centrolepidaceae. Leaves of C. racemosa are pronouncedly bifacial, with a row of hairs along the margins. Similar marginal hairs are characteristic of the sheathing leaf bases of various Centrolepidaceae.

Distichous and tristichous partial inflorescences in Centrolepis and their interepretation as spikelets

As mentioned in the introduction, partial inflorescences of Centrolepis described here as distichous spikelets are sometimes interpreted as monochasia (cincinni, Fig. 2). Normally, monochasial vs. racemose types of flower arrangement can be distinguished using the relative positions of flowers, flower‐subtending bracts and inflorescence axes. However, this criterion is equivocal in Centrolepis, because of problems in distinguishing the flower‐subtending bracts from tepals (see below). We believe that the sporadic occurrence of partial inflorescences with three rows of flowers (documented here for the first time for Centrolepis) further supports their interpretation as spikelets, rather than cincinni. Indeed, both tristichous and distichous partial inflorescences are accommodated readily by the spikelet model, differing merely in the phyllotaxis along the spikelet axis (1/2 and 1/3). Similarly, partial inflorescences in Leguminosae, tribe Fabeae (=Vicieae) are interpreted as dorsiventral racemes with two, three (e.g., Sinjushin, 2013), or even more (Goebel, 1913) rows of flowers.

In contrast, the units with three rows of flowers cannot be explained readily as cincinni, or any other type of monochasial flower arrangement (e.g., Müller‐Doblies, 1977; Kondorskaya, 1990). At first glance, the variation in number of flower rows found in Centrolepis suggests a comparison with pair‐flowered cincinni. Pair‐flowered cymes, which are probably restricted to certain members of the eudicot order Lamiales (Weber, 2013) have axes of all orders ended in a pair of flowers rather than a flower. Compact cincinni of Epithema Blume (Gesneriaceae, Weber, 1976: fig. 7) indeed superficially resemble the partial inflorescences of Centrolepis. However, they possess four, rather than three rows of flowers; four rows is precisely what should be expected theoretically in such a pattern of flower arrangement. Four rows of flowers should be also expected in a geminiflorous cincinnus. A geminiflorous cyme is a special type of cyme with one of the two dichasial branches reduced to a single flower (Weber, 2013). In theory, a pattern with three rows of flowers can be found when geminiflorous (or pair‐flowered) cymes are alternating regularly with normal cymes along a cincinnus, but this would be an extremely complex explanation. To our knowledge, nothing similar is found in any other monocots. We accept the much simpler spikelet model.

A tendency to developmental retardation or loss of the lower spikelet in inflorescences of Centrolepis

In our taxon sampling, the loss of a spikelet in the axil of the lower phyllome of the primary inflorescence axis is a uniquely derived synapomorphy of a clade comprising species of Group 2. The presence of only one spikelet correlates with dimorphic morphology of the phyllomes of the primary inflorescence axis. The upper phyllome (that subtends the only remaining spikelet) is narrower and shorter than the lower phyllome. Clearly, the lower phyllome of species of Group 2 plays an important protective role for the spikelet. Among the species not sampled here and not studied developmentally, lower spikelets can be either present or absent in C. mutica (R.Br.) Hieron. (Barrett and Sokoloff, 2015). In some species with two spikelets (C. aristata, C. pilosa of Group 3, see Fig. 13), the upper lateral spikelet was more advanced developmentally than the lower lateral spikelet of the same inflorescence. This can be regarded as a step toward a complete loss of the lower spikelet. Thus, a tendency to suppression and loss of the lower spikelet can be seen in Centrolepis, at least in Groups 2 and 3. We did not record pronounced developmental retardation of the lower spikelet in examined members of the Group 1. Is this an argument in favor of a potentially close phylogenetic relationship between species of Groups 2 and 3 (as in Fig. 13A, B)? Published molecular phylogenetic data (Briggs et al., 2014) are insufficient to resolve this issue, due to relatively low taxon sampling in Centrolepis.

For a morphologist, a question arises: Why is the upper spikelet of Centrolepis lateral (as interpreted here) and not terminal on the inflorescence axis? Indeed, the developmental retardation of the lower spikelet could be viewed as an indication of the terminal nature of the upper spikelet. The following lines of evidence support our interpretation of the upper spikelet as lateral. (1) Continuation of the primary inflorescence axis after insertion of the upper spikelet occurs in at least one species (C. exserta). It bears a third, reduced phyllome that subtends no lateral structure. (2) Developmental retardation of the lower spikelet can be seen as an evolutionary step toward its complete reduction (see above). (3) The plane of distichy of the upper spikelet does not continue the plane of distichy of the primary inflorescence axis (Fig. 7A, B). (4) The developmental retardation of the lower spikelet is not a universal phenomenon in Centrolepis.

Phyllomes associated with flowers and the interpretation of bracts and tepals

Our study highlights the taxonomic and phylogenetic significance of the number, position, and morphology of phyllomes associated with the flowers of Aphelia and Centrolepis. We found high structural similarity of inflorescences across Aphelia and Centrolepis, so that the position of each phyllome can be described precisely and comparisons between species can be made.

Many species of Centrolepis possess two thin, membranous, but conspicuous phyllomes associated with the flowers (“secondary bracts”: Cooke, 1992). These two phyllomes are readily detectable using routine taxonomic herbarium‐based practices. Our data show that species with two conspicuous phyllomes belong to two different groups and that members of these groups differ in position of the phyllomes (the differences are scored in characters 30 and 31 of our data matrix). In species of Group 3 (see Fig. 13A: C. aristata, C. fascicularis, C. pilosa, C. drummondiana), both long phyllomes are located on the adaxial side of the flower, but on different radii (Fig. 10H, long phyllomes are thick blue arcs). In the group of tropical annuals of Group 1 (C. banksii, C. exserta, C. curta, C. racemosa), one long membranous phyllome is adaxial, and the other is abaxial (Fig. 10D–F, long phyllomes are thick blue lines). Centrolepis strigosa (a basal member of a clade comprising species of Group 1) has three long, conspicuous phyllomes (Fig. 10G) and could be considered intermediate between species of the Group 3 and tropical annuals because the two phyllomes on the adaxial side of a flower are in the same positions as in Group 3, but there is also a conspicuous phyllome on the abaxial side, as in tropical annuals. The position of C. strigosa inferred from the morphological cladistic analysis is perfectly compatible with its intermediacy between members of Group 3 and tropical annuals.

SEM and LM investigations have revealed the sporadic occurrence of an additional, short and inconspicuous phyllome on the abaxial side of flowers of some tropical annuals (C. banksii, C. exserta, C. curta). In our view (see also Sokoloff et al., 2010), these small phyllomes could be interpreted as vestigial flower‐subtending bracts. The spatial arrangement of the phyllomes is what might be expected from flower‐subtending bracts in our model of a dorsiventral spikelet. Their position relative to the floral parts is the same as the large subtending phyllomes of C. racemosa (Sokoloff et al., 2010). The phyllomes that we interpret as flower‐subtending bracts in C. banksii, C. curta, and C. exserta are inserted below the two conspicuous phyllomes, so that a short stalk is present between their levels of attachment, which could be interpreted as a pedicel. The insertion of the proposed flower‐subtending bracts is such that they can be readily interpreted as belonging to the spikelet axis.

If our interpretation of the flower‐subtending bract in C. banksii, C. curta, and C. exserta is correct, the bract is suppressed in most flowers of these species. When the number, position, relative size, and timing of development of the floral organs do not differ between flowers with and without visible flower‐subtending bracts (as in the case of C. banksii, C. curta, C. exserta), we postulate the occurrence of a cryptic bract whose positional information is used for the correct patterning of the floral organs (e.g., Remizowa et al., 2013). For example, we predict cryptic flower‐subtending bracts in the superficially bractless flowers of C. banksii, C. exserta, and C. curta. Based on our morphological interpretations, we also predict the absence of cryptic bracts associated with flowers of C. racemosa because these flowers develop directly in the axils of the large phyllomes of the primary inflorescence axis. These predictions should be viewed as testable hypotheses because cryptic bracts can be visualized using gene expression patterns, as demonstrated in the eudicot family Brassicaceae and the monocot family Poaceae (e.g., Long and Barton, 2000; Whipple et al., 2010).

Our SEM and LM investigations have revealed the regular occurrence of a short and inconspicuous abaxial (more precisely, abaxial/basiscopic) phyllome associated with the flowers of all species of Group 3. By analogy with C. exserta, C. curta, and C. banksii, can this phyllome be interpreted as a vestigial flower‐subtending bract? This hypothesis can be falsified by visualization, using gene expression patterns, of cryptic flower‐subtending bracts distinct from the phyllome in species of Group 3. In the absence of such information, we highlight three observations from comparative morphology suggesting that the short abaxial phyllome occurring in species of Group 3 does not represent a flower‐subtending bract. (1) In contrast with C. exserta, C. curta, and C. banksii, this phyllome is not inserted at a conspicuously lower level than two other (long) phyllomes. (2) In contrast with C. exserta, C. curta, and C. banksii, this phyllome has only two cell layers (like the long membranous phyllomes of all species of the genus). (3) This phyllome has the same location and shape (broad base extending into the basiscopic side) as one of the large and conspicuous phyllomes of species of C. strigosa. Plotted onto diagrams, the phyllomes of members of Group 3 (Fig. 10H) and C. strigosa (Fig. 10G) are so similar that they are most likely homologous. Furthermore, the drastic size difference between the abaxial/basiscopic phyllomes of species of Group 3 and C. strigosa appears late in development; the phyllomes are very similar at mid stages. On the other hand, the large abaxial (or abaxial/basiscopic) phyllomes of tropical annuals and C. strigosa are positionally and structurally similar. Thus, the large abaxial/basiscopic phyllome of C. strigosa is homologous with a similar large phyllome of tropical annuals and not with the presumed flower‐subtending bracts of C. exserta, C. curta, and C. banksii.

If we accept the postulated homologies between the three phyllomes observed in Group 3 (two long, one short) and C. strigosa (all three long), it is possible that this morphology is plesiomorphic in Centrolepis.

Members of Group 2 differ considerably from the taxa discussed in previous paragraphs. Centrolepis milleri (Fig. 10I) has one phyllome that is long, conspicuous, and adaxial/acroscopic and one that is very short, inconspicuous, and abaxial/basiscopic. On the basis of their similar locations, definitive morphology and developmental similarity, we conclude that these phyllomes are homologous with the adaxial/acroscopic and abaxial/basiscopic phyllomes of species of Group 3. It is likely that the absence of an adaxial/basicopic phyllome (which is present in Group 3 and C. strigosa) is a derived condition in C. milleri. Centrolepis glabra (Fig. 10J) and C. polygyna form a clade (Group 2) with C. milleri, but differ from it in the presence of a single phyllome associated with a flower. The phyllome is short and abaxial to basiscopic. Therefore, we believe that this single phyllome is homologous with the short phyllome of C. milleri and Group 3.

Species of Aphelia have two phyllomes associated with the flowers, a short one on the inferred abaxial (or nearly abaxial) side and a long and membranous one in a nearly adaxial position (Fig. 10A–C, long phyllomes are thick blue arcs). At first sight, these phyllomes appear comparable with those of Centrolepis milleri. However, in terms of phylogeny, C. milleri is not closely related to Aphelia. Also, the long membranous phyllome of Aphelia is adaxial/basiscopic rather than merely adaxial, while the long membranous phyllome of C. milleri is adaxial/acroscopic. Therefore, the two long phyllomes are unlikely to be homologous. The short phyllome of C. milleri has a broad base and consists of two cell layers, as in Group 3 of Centrolepis, whereas the short phyllome of Aphelia has a narrow base and consists of more than two cell layers. In these characters, the short phyllome of Aphelia resembles the vestigial flower‐subtending bracts of C. banksii, C. curta, and C. exserta, but the hypothesis on their homologies requires further testing.

To summarize, we believe that the short vestigial phyllomes of C. banksii, C. curta, and C. exserta and possibly the short phyllomes of Aphelia are flower‐subtending bracts. All other phyllomes of Aphelia and Centrolepis belong to the floral axis. We speculate that these phyllomes can be interpreted as tepals rather than floral prophylls (bracteoles) because floral prophylls are not recorded in the closely related family Restionaceae (Kircher, 1986). In species of Centrolepis with three tepals (Group 3 and C. strigosa), the adaxial/basiscopic tepal occupies in an inner position with respect to the other two, so we interpret it as an inner‐whorl tepal, strongly associated (or even basally fused) with the stamen. Strong association between stamens and tepals occurs in a broad range of monocots (reviewed by Endress, 1995; Remizowa et al., 2010), indeed, it is more common between inner‐whorl stamens and tepals (petals) than between outer‐whorl stamens and tepals (sepals). In Restionaceae, the family most closely related to Centrolepidaceae, only inner‐whorl stamens are present, either as fertile stamens or as staminodes in functionally male flowers (Ronse De Craene et al., 2001, 2002); these inner‐whorl stamens are basally connected with the inner‐whorl tepals (H. P. Linder, University of Zurich, personal communication, 2011).

Centrolepis racemosa as a morphological misfit and perspectives of evo‐devo in Centrolepidaceae

Our phylogenetic data clearly show that C. racemosa is only distantly related to Aphelia and Gaimardia. Its similarities with Aphelia (numerous phyllomes of the primary inflorescence axis) and Australasian Gaimardia (plants diaxial: flowers borne directly in the axils of phyllomes of the primary axis) should probably be interpreted as homoplasies. Centrolepis racemosa clearly belongs to Centrolepis and, more precisely, falls into the group of tropical species that is defined by a suite of floral features. Although its phylogenetic position in the tropical group (predicted by Sokoloff et al., 2009b) is supported by the current study, C. racemosa differs from other tropical species and in fact from all other species of Centrolepis in an impressive list of characters (online Appendix S7). Our phylogenetic analysis provided two alternative hypotheses for a potential sister‐group relationship of C. racemosa (both within the group of tropical annuals), one with C. curta (Fig. 13C) and the other in an unresolved position among other tropical annuals (Fig. 13B). These two topologies imply that the branch leading to C. racemosa is eight or nine steps long, respectively. In all shortest trees found, the branch leading to C. racemosa is the longest branch within Centrolepidaceae. In contrast, the branch length for its potential sister groups is either zero (C. curta, Fig. 13C), or 1–2 steps (see Fig. 13B).

We believe that C. racemosa can be classified as a morphological misfit (e.g., Minelli, 2015). The distinctive features of C. racemosa could be interpreted as losses of developmental boundaries or as unification of developmental programs that are diversified in other species of the genus. Shoot apical meristems of phylogenetically related species of Centrolepis produce three distinct types of phyllomes in a precise sequence, i.e., the foliage leaves, the cataphylls and the phyllomes of the primary inflorescence axis (Appendix S7). In C. racemosa, all the phyllomes produced by the shoot apical meristem are basically similar to each other. In other species of Centrolepis, the leaf lamina is formed by the hyperphyll (Oberblatt), at least in phyllomes of the primary inflorescence axis. In C. racemosa, the leaves are apparently formed by the hypophyll only (its leaves lack clear demarcation between hypophyll and hyperphyll). In Centrolepis species other than C. racemosa, the inflorescence is clearly demarcated from the vegetative part of a shoot. The last internode before the inflorescence (the final internode) is the only long internode in many species (though not in C. curta). In many species, phyllotaxy changes at the transition to the inflorescence, but in C. racemosa, the boundary of the inflorescence is not sharp (an empty phyllome is sometimes inserted between the two flower‐subtending phyllomes), the basal internode is not differentiated, and there is no phyllotactic change at the transition to the inflorescence. In other species of Centrolepis, spiral phyllotaxy, when present, is confined to the vegetative parts of the shoot.

All species of Centrolepis except C. racemosa (also all species of Aphelia) have two shoot types (Appendix S7): (1) vegetative shoots continued into the primary inflorescence axis and (2) specialized dorsiventral short‐shoots (spikelets) bearing suppressed or vestigial flower‐subtending bracts. As mentioned already, type 1 shoots produce three distinct types of phyllomes. The suppressed or vestigial bracts represent the fourth phyllome type of more typical Centrolepis species. In contrast with these species, in C. racemosa, differentiation into two shoot types is lost and the flowers appear in the axils of ordinary leaves in the distal part of every shoot (Appendix S7).

An important question is whether the impressive differences of C. racemosa appeared during the course of evolution by successive and possibly gradual acquisition of different features or simultaneously and rapidly, in a saltationary manner. To test saltation, Bateman and DiMichele (2002) recommended describing as many features as possible and treating separately the features that are potentially developmentally correlated. This is the approach that we used in our cladistic analysis. Long branches like the one leading to C. racemosa in our analysis are thought to be indicators of saltational events, so that all character transformations in the branch could result from a single mutation (Bateman and DiMichele, 2002; see also Nuraliev et al., 2014).

If such a macromutation occurred in an ancestor of C. racemosa, we speculate that a key factor could have been the unification of four developmental programs for different leaf types characteristic of related species of Centrolepis (vegetative leaf, cataphyll, phyllome of the primary inflorescence axis, cryptic/suppressed bract). It seems that none of these programs (including that for vegetative leaves) has remained unchanged.

One of the most extensively studied cases of partial loss of differentiation between different developmental programs involves the resuppression of cryptic bracts in various mutants of model plants that normally bear bractless inflorescences. Inflorescences with suppressed bracts are characteristic of model organisms belonging to the eudicot family Brassicaceae and the monocot family Poaceae (grasses). Comparative analyses have demonstated strong differences in the genetic mechanisms of bract suppression in these two families (Chuck et al., 2010; Whipple et al., 2010). This difference is in accordance with the high phylogenetic distance between the grasses and Brassicaceae. On the other hand, Centrolepidaceae are the closest grass relatives that show an independent event of bract suppression, so it will be interesting to investigate whether the genes involved in bract suppression are more similar between grasses and centrolepids than between grasses and Arabidopsis.

Phenotypes of tassel sheath (tsh) mutants of maize with resuppressed bracts show several parallels with C. racemosa. These mutants showed some defects in branching, so that spikelet pairs were often initiated instead of long branches in male inflorescences and single spikelets sometimes developed instead of spikelet pairs. These features can be interpreted as a shift toward a loss of differentiation into several hierarchical types of shoots, just as in C. racemosa (Chuck et al., 2010; Whipple et al., 2010). Resuppression of bracts in maize can alter patterns of phyllotaxis (Chuck et al., 2010). However, an important issue is that suppressed bracts in grasses subtend inflorescence branches, whereas the suppressed bracts of Brassicaceae and Centrolepis subtend individual flowers. Flower‐subtending bracts are not suppressed in grasses.

The small family Centrolepidaceae represents an attractive target for evolutionary developmental genetics for several reasons, both practical (annual life cycle in many species and small size of individuals), phylogenetic (relatively close phylogenetic relationship to grasses), and also because its immense morphological diversity provides several potentially testable hypotheses. Further investigations of centrolepids and the entire restiid clade could shed light into the early evolution and origin of grasses.

ACKNOWLEDGEMENTS

We are grateful to Richard Bateman and Barbara Briggs for discussion of many aspects of the paper, Peter Linder for providing material of several species and helpful comments, Terry Macfarlane for discussion and essential input in collecting material of Centrolepis from the Northern Territory, Evgeny Mavrodiev for discussion of grass phyllotaxy, and several anonymous reviewers for helpful comments. We thank the staff of the Department of Electron Microscopy at the Biological Faculty of Moscow University for their assistance and for providing SEM facilities. Studies of phyllome morphology, shoot branching and architecture, phylogenetic analysis and evolutionary interpretation were performed by D.D.S. and M.V.R. with funding from Russian Scientific Fund (project 14‐14‐00250).

Supporting information

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

List of examined specimens. Herbaria abbreviations are specified as in Index Herbariorum (http://sciweb.nybg.org/science2/IndexHerbariorum.asp).

Taxon, Collection area, Voucher specimen, (Herbarium code)

Aphelia brizula F.Muell., Western Australia, H.P. Linder 5547 (MW); Aphelia cyperoides R.Br., Western Australia, P.J. Rudall 908 (MW); Aphelia cyperoides R.Br., Western Australia, H.P. Linder 6077 (MW); Aphelia drummondii (Hieron.) Benth., Western Australia, P.J. Rudall 910 (MW); Aphelia nutans J.D.Hook. ex Benth., Western Australia, M.D. Barrett 2607 (PERTH); Centrolepis aristata (R.Br.) Roem. & Schult., Western Australia, J.G. Conran JGC3076 (MW); Centrolepis curta D.Cooke, Northern Territory, T.D. Macfarlane et al. 4321 (MW); Centrolepis banksii (R.Br.) Roem. & Schult., Northern Territory, T.D. Macfarlane et al. 4256 (MW); Centrolepis drummondiana (Nees) Walp., Western Australia, H.P. Linder 5550 (MW); Centrolepis exserta (R.Br.) Roem. & Schult., Northern Territory, T.D. Macfarlane et al. 4222, 4289, 4330 (MW); Centrolepis fascicularis Labill., Australia, J.J. Bruhl 2819 (UNE); Centrolepis glabra (F.Muell. ex Sonder) Hieron., Western Australia, R.L. Barrett 5037 (PERTH); Centrolepis milleri M.D.Barrett & D.D.Sokoloff, Western Australia, B. Miller s.n. (PERTH, MW); Centrolepis pilosa Hieron., Western Australia, H.P. Linder 5551 (MW); Centrolepis pilosa Hieron., Western Australia, J.G. Conran JGC3077 (MW); Centrolepis polygyna (R.Br.) Hieron., Western Australia, P.J. Rudall 25 (MW); Centrolepis polygyna (R.Br.) Hieron., Western Australia, H.P. Linder 6076 (MW); Centrolepis polygyna (R.Br.) Hieron., Western Australia, R.L. Barrett 5065 (PERTH); Centrolepis strigosa (R.Br.) Roem. & Schult., native to Australia, cultivated in Bonn Botanic Garden, coll. P.J. Rudall s.n., 28 Sept. 2006 (no voucher); Centrolepis strigosa subsp. pulvinata (R.Br.) D.A.Cooke, Australia, Kent Group, Bass Strait, J. Whinray s.n., 29 Nov. 1970 (Kew spirit collection no. 33017); Gaimardia australis Gaudich., Falkland Islands, S.W. Greene 84/1 (K); Gaimardia fiztgeraldii F.Muell. & Rodw., Tasmania, W.M. Curtis s.n., Jan 1948 (Kew spirit collection no. 22474); Gaimardia setacea Hook.f., Tasmania, W.M. Curtis s.n., Jan 1948 (Kew spirit collection no. 22476).

List of morphological characters used in cladistic analysis.

0. Lifespan: 0, perennial; 1, annual.

1. Plant size: 0, robust plants (more than 15 cm tall); 1, diminuitive plants (less than 10 cm tall).

2. Stems (culms): 0, solid; 1, hollow.

3. Leaf and stem epidermal cells: 0, non‐overlapping; 1, overlapping (Cutler, 1969; Meney and Pate, 1999).

4. Protective cells surrounding substomatal cavities: 0, present; 1, absent (Cutler, 1969; Meney and Pate, 1999).

5. Phyllotaxis of vegetative leaves: 0, spiral; 1, distichous.

6. Position of the first leaf of a lateral shoot: 0, adaxial; 1, transversal.

7. Typical foliage leaves (with conspicuous laminas): 0, present; 1, absent.

8. Morphology of the first leaf of a lateral shoot: 0, cataphyll; 1, foliage leaf.

9. Ligule on vegetative foliage leaves: 0, present; 1, absent.

10. Sheath to blade transition in foliage leaves: 0, distinct; 1, indistinct.

11. Number of vascular bundles per foliage leaf: 0, one; 1, three.

12. Simple (unbranched) hairs on leaf blades (except the very tips): 0, present; 1, absent.

13. Cataphyll(s) located below inflorescence and above foliage leaves on a shoot: 0, present; 1, absent.

14. Basal internode: 0, elongate (=inflorescences on scapes); 1, short (=inflorescences sessile).

15. Primary inflorescence axis: 0, not compressed; 1, compressed.

16. Number of lateral spikelets (or branches, for outgroups) per inflorescence: 0, one; 1, two (three); 2, more than three; 3, absent (unordered character).

17. Lateral spikelets: 0, not dorsiventral; 1, dorsiventral.

18. Plant sexuality: 0, dioecy; 1, monoecy.

19. Number of flowers in bisexual or female spikelets: 0, one (without additional sterile flowers); 1, two or more flowers (sometimes only one flower is fertile).

20. Terminal spikelet: 0, absent; 1, present.

21. Phyllotaxis on the primary inflorescence axis: 0, spiral; 1, distichous.

22. Number of phyllomes on the primary inflorescence axis: 0, several; 1, two (three).

23. Phyllomes on the primary inflorescence axis: 0, monomorphic; 1, the lowermost phyllome differs from the others; 2, the two lowermost phyllomes differ from the distal ones (unordered character).

24. Structure in the axil of the lowermost phyllome of the primary inflorescence axis: 0, flower; 1, male spikelet; 2, bisexual spikelet; 3, no axillary structure; 4, complex branch; 5, female spikelet (unordered character).

25. Structure in the axil of the second phyllome of the primary inflorescence axis: 0, flower or sterile flower; 1, male spikelet; 2, bisexual spikelet; 3, female spikelet; 4, complex branch (unordered character).

26. Hairs on the phyllomes of the primary inflorescence axis (except their margins): 0, absent; 1, present.

27. Lamina of the lowermost phyllome of the primary inflorescence axis: 0, absent or less than 1/2 length of the sheath; 1, longer than 1/2 length of the sheath.

28. Ligule (or auricles with ventrally approaching margins) of the lowermost phyllome of the primary inflorescence axis: 0, present; 1, absent.

29. Phyllomes of the primary inflorescence axis at anthesis: 0, upright; 1, widely gaping.

30. Number of adaxial or obliquely adaxial phyllomes associated with a flower (tepals or prophylls): 0, absent; 1, one; 2, two; 3, three (ordered character).

31. Long (longer than the ovary) abaxial membranous phyllome associated with a flower: 0, absent; 1, present.

32. Short (shorter than one half of the ovary height) abaxial phyllome associated with a flower: 0, absent; 1, present.

33. Stamens per male or bisexual flower: 0, two or three; 1, one.

34. Gender allocation in inflorescence: 0, all stamens in male flowers (i.e., bisexual flowers absent); 1, bisexual flowers present (sometimes together with male and/or female flowers in the same inflorescence).

35. Stamen to gynoecium connation: 0, stamen(s) free; 1, filament basally united with gynoecium.

36. Typical carpel number per gynoecium: 0, one; 1, two to four; 2, five to eight; 3, usually more than eight.

37. Intercarpellary fusion: 0, typical syncarpy; 1, fusion via floral center.

38. Gynoecium outline at carpel initiation (in gynoecia with more than two carpels): 0, rounded; 1, elongate.

39. Position of ovary locules in anthetic flowers: 0, all ovary locules at the same level; 1, ovary locules on different levels.

40. Position of the first‐formed or the only carpel: 0, abaxial; 1, oblique or transversal.

41. Conspicuous common style in pluricarpellate gynoecia: 0, abset; 1, present.

42. Styles of individual carpels (=stylodia): 0, of equal length and attached to common style or ovary at the same level; 1, of unequal length and attached at different levels.

43. Fruits: 0, dehiscent; 1, indehiscent.

44. Wide (about as wide as the fruit) and long carpophore: 0, absent; 1, present.

45. Fruit locules: 0, fruits plurilocular; 1, fruits unilocular.

46. Phyllomes subtending spikelets: 0, not dispersed with fruits; 1, dispersed with fruits.

Morphological data matrix (“–” denotes missing character; “?” denotes unknown character; A = 0, 1; B = 2, 3; C = 3,4; D = 4,5).

Taxon	Character no.
	00000000001111111111222222222233333333334444444
	01234567890123456789012345678901234567890123456
Ceratocaryum argenteum	00A00??10???0‐0020011?00DC00?03–000100010010‐0
Apodasmia similis	001010010???1‐00200010004400?03–00010?0?1010‐0
Sporadanthus gracilis	00100??10?0?1‐0020001?004400?03–00010?0?000000
Lepyrodia heleocharioides	00100??1000?1‐0020001?004400003–00010?0?000000
Aphelia brizula	11011?100000110021100102111A011011000‐‐‐1‐‐10‐1
Aphelia cyperoides	1101101000001100211A0100221A01101110‐‐‐1‐‐10‐1
Aphelia drummondii	11011??0??001?00211001011300A?1011000‐‐‐1‐‐10‐1
Aphelia nutans	110??01000001100211001021100‐11011000‐‐‐1‐‐10‐1
Centrolepis aristata	110111?0A1001A001111011022010020111020010100000
Centrolepis banksii	110????011001000111101102200‐111A11131110010000
Centrolepis curta	110????0110000101111011022101111A11031110010000
Centrolepis drummondiana	11011??01100100011110110220A0020111020010100000
Centrolepis exserta	11011010110000001111011022101111A11AB1110010000
Centrolepis fascicularis	010110101100A0001111011022110120111010010100000
Centrolepis glabra	110??11011001A000111011132001000111020?10100000
Centrolepis milleri	110??110110010000111011132001010111120010100000
Centrolepis pilosa	110??0101100A0001111011022110120111020010100000
Centrolepis polygyna	11001110A10010000111011132011000111130010100000
Centrolepis racemosa	110??0101110‐1103‐111000000‐1?11?1103??10010000
Gaimardia australis	01001000000111003‐10111030000000001010?0?000110
Gaimardia fitzgeraldii	01001100000111013‐11111000000000001010?01000000
Gaimardia setacea	01001100000011013‐11111A00000000001010?01000000