https://orcid.org/0000-0002-4816-1212
Abstract
The grass family (Poaceae) includes cereal crops that provide a key food source for the human population. The food industry uses the starch deposited in the cereal grain, which develops directly from the gynoecium. Morphological interpretation of the grass gynoecium remains controversial. We re-examine earlier hypotheses and studies of morphology and development in the context of more recent analyses of grass phylogenetics and developmental genetics. Taken in isolation, data on gynoecium development in bistigmatic grasses do not contradict its interpretation as a solitary ascidiate carpel. Nevertheless, in the context of other data, this interpretation is untenable. Broad comparative analysis in a modern phylogenetic context clearly demonstrates that the grass gynoecium is pseudomonomerous. A bistigmatic grass gynoecium has two sterile carpels, each producing a stigma, and a fertile carpel that lacks a stigma. To date, studies of grass developmental genetics and developmental morphology have failed to fully demonstrate the composite nature of the grass gynoecium because its complex evolutionary history is hidden by extreme organ integration. It is problematic to interpret the gynoecium of grasses in terms of normal angiosperm gynoecium typology. Even the concept of a carpel becomes misleading in grasses; instead, we recommend the term pistil for descriptive purposes.
Introduction
The grass family (Poaceae) is economically highly significant because it includes cereal crops that provide a key food source for the human population (e.g. McKevith, 2004; Laskowski et al., 2019). The food industry uses the starch deposited in the endosperm of the cereal grain, a one-seeded fruit that is accurately termed a caryopsis, which develops directly from the gynoecium. However, despite the economic importance of the grass gynoecium, and considerable available data on its structure, development, physiology, and topics related to genomics, its morphological interpretation remains controversial (Philipson, 1985; Linder and Rudall, 2005).
The primary goal of this review is to evaluate the evolution and morphological identity of the grass gynoecium. We re-examine earlier hypotheses and studies of morphology and development in the context of more recent analyses of grass phylogenetics and developmental genetics. We use the insights gained to highlight Kircher’s (1986) relatively underexplored hypothesis of the grass gynoecium, which proposed that stigmas of bistigmatic grasses belong to two sterile posterior-lateral carpels whereas the ovule belongs to the third, median carpel. Although Takhtajan (1987, 2009) cited Kircher’s work, his statement that gynoecia of most grasses are bicarpellate does not follow Kircher’s ideas. We show that the bistigmatic gynoecia of cereal crops and their wild relatives, although relatively simple in structure and development, underwent a remarkably complex evolutionary trajectory. We develop Philipson’s (1985) notion regarding the essential importance of a comparative approach for interpretation of plant ontogeny. The problem of the grass gynoecium demonstrates that even an in-depth knowledge of developmental processes and their underlying molecular machinery is insufficient without comparative information on outgroups. Taken together, developmental, comparative, and evolutionarily approaches provide a highly efficient tool. Evolutionarily successful groups such as grasses, with adaptively significant morphological novelties, provide inspiring opportunities for uncovering changes in developmental programs that have played key roles in macroevolution (Kellogg, 2000).
Molecular phylogenetic data provide an essential framework for studies of grass evolution (Fig. 1). There is a consensus in recognizing 12 subfamilies of Poaceae, of which three (Anomochlooideae, Pharoideae, and Puelioideae) form a species-poor basal grade (Grass Phylogeny Working Group II, 2012; Soreng et al., 2017; Saarela et al., 2018). Other subfamilies group into two species-rich clades, of which the PACMAD clade has a longer stem branch and higher support values than the BOP clade. Some pre-molecular classifications placed genera currently classified in Anomochlooideae, Pharoideae, and Puelioideae in a broadly defined Bambusoideae, a subfamily that was considered primitive relative to other grasses on the basis of a number of reproductive characters (e.g. Tzvelev, 1989). The more narrowly redefined subfamily Bambusoideae is currently placed in the BOP clade, though some early molecular studies suggested that the three subfamilies comprising the BOP clade instead form a grade, a hypothesis that may be congruent with the plesiomorphic nature of some bambusoid characters (see Nadot et al., 1994; Christin et al., 2009; Rudall et al., 2014).
Phylogenetic relationships among the subfamilies of grasses and the closest outgroups of Poaceae. The tree topology follows the plastid phylogenies of Givnish et al. (2018) and Saarela et al. (2018). Optimization of stigma number is based on our maximum parsimony analysis performed in Winclada (Nixon, 2002). Stigma numbers found in terminal groups are provided after their names. Instances of teratology are not considered here.
Gynoecium structure, diversity, and development in grasses
This section does not consider spontaneous abnormalities, mutants, hybrids, and cultivars. Here we focus on typical conditions found in wild-type material of certain species of grasses. Boxes 1 and 2 outline the basics of the complex descriptive terminology used in gynoecium morphology.
The gynoecium (pl. gynoecia) is an angiosperm-specific structure enclosing one or more ovules. The carpel is the structural unit of the gynoecium. Each carpel has a ventral side that is closest to the centre of the gynoecium; the opposite side is dorsal. Gynoecia can be apocarpous or syncarpous. The carpels of an apocarpous gynoecium are either free or united only through post-genital fusion, which is a process that can be observed during ontogeny, commencing with surface by surface contact of organs. In contrast, the carpels of a syncarpous gynoecium are congenitally united, either along their entire length or up to a certain level. In some species, the free parts of the carpels fuse post-genitally later in ontogeny. Congenital fusion does not represent a process in ontogeny, because the united parts are already conjoined at their initiation. When the distal parts of the carpels (e.g. the stigmas) remain unfused in a syncarpous gynoecium, two developmental patterns can be recognized. In early congenital fusion, the united parts of the different carpels develop before their free parts. In late congenital fusion, the free parts of the carpels can be observed early in ontogeny, while their united parts appear later by intercalary growth.
Each carpel of an apocarpous gynoecium bears and encloses one or more ovules. Although contentious, it is convenient in practice to consider the carpel as a sort of phyllome, a general term for all leaf homologues. Like other phyllomes, carpels are fundamentally bifacial, possessing a morphologically upper (adaxial) surface and a lower (abaxial) surface. Ovule(s) are attached to the adaxial surface, often close to its border with the abaxial surface. In many angiosperms, carpels have a sac-like proximal portion (the ascidiate zone) that is O shaped in transverse section, with an inner adaxial surface and outer abaxial surface, and a distal portion that is horseshoe shaped early in ontogeny (the plicate zone). The border between the adaxial and abaxial surfaces is located in the opening of the plicate zone. This border is more or less virtual, but in theory it could be visualized on the basis of expression patterns of genes governing the adaxial/abaxial polarity of the phyllomes. If more than one carpel is present in a gynoecium, the openings of their plicate zones are directed towards the centre of the flower. The opening is usually closed by post-genital fusion of the carpel margins to form a so-called ventral slit, which is absent from the ascidiate zone (note that closure of individual carpels and intercarpellary fusion represent different—though related—processes). The border between the ascidiate and plicate zone is called the cross-zone. If an angiosperm carpel has only one ovule, it is very often attached in the cross-zone. Some angiosperms have carpels without a plicate zone, while others lack an ascidiate zone.
A pistil is a physically continuous unit: either an individual carpel free from other carpels or an entire gynoecium composed of fused carpels.
An ovary is the region of a pistil with an internal cavity or cavities bearing the ovule(s). The distal part of the pistil bears the receptive stigmatic tissue and displays diverse morphology. A stigma can be sessile at the top of the ovary. Alternatively, there can be an attenuate or filiform style (more accurately termed a stylodium if formed by only one carpel) with terminal or decurrent stigmatic tissue. The style can be formed by the ascidiate zone (if the carpel is entirely ascidiate) or the plicate zone. A plicate style has a ventral slit along its length. In some taxa, the style is solid from inception and lacks a ventral slit. Some angiosperm carpels are distally bifid; in an extreme condition, a carpel has two stigmas (one at the left side and the other at the right side).
Further reading: Weberling (1989); Verbeke (1992); Endress (1994, 2011, 2015); Leins and Erbar (2010); Sokoloff et al. (2018); and Phillips et al. (2020).
Variation in structure and development of angiosperm gynoecia (ovules not shown). (A–D) Gynoecia early in ontogeny in vertical and transverse sections. (E–K) Mature gynoecia. asc, ascidiate zone; pli, plicate zone; cz, cross-zone. Yellow, areas of post-genital fusion. Arrows, ontogenies. A→E, A→I, B→F, apocarpy with no fusion between carpels. B→J, apocarpy with post-genital fusion between carpels. B→K, syncarpy with late congenital fusion between carpels. C→K, D→K, syncarpy with early congenital fusion between carpels. C→G, D→H, complete congenital fusion between carpels.
Many syncarpous gynoecia can be readily interpreted as consisting of carpels with recognizable ascidiate and plicate zones. Congenitally united ascidiate zones form a synascidiate zone, which from inception in ontogeny has as many locules as carpels. Congenitally united plicate zones form a symplicate zone, which is unilocular at initiation. A transverse section through a young symplicate zone shows that although each carpel is ventrally open, as all carpels are united by their margins, the entire gynoecium is closed. Later in development, the ventral slits of individual carpels may (or may not) close post-genitally in the symplicate zone to form as many locules as carpels. In syncarpous gynoecia, patterns of ovule attachment (termed placentation) differ among angiosperms. The ovules may be attached in the synascidiate or symplicate zone, in both zones, or only in the cross-zone. In the synascidiate zone and cross-zone, the ovule(s) are most commonly attached to the innermost side of a carpel (axile placentation). In the symplicate zone, they usually appear along the lines of congenital fusion of adjacent carpels (parietal placentation). As in other instances of congenital fusion, the precise boundaries between the carpels of syncarpous gynoecia cannot be traced. Adjacent carpels may share their lateral and/or ventral vascular bundles (then termed synlateral or synventral, respectively). Ovules of parietal placentae sometimes cannot be assigned to any particular carpel but are shared by the two carpels whose margins are united in this region. All types of style and stigma found in free carpels can also be found in carpels of syncarpous gynoecia. Styles of individual carpels (stylodia) are either free or united to various extents in a common style. When carpels are distally bifurcating and the lateral lobes of adjacent carpels are congenitally united with each other, so-called commissural stigmas are present. Each commissural stigma is shared by two adjacent carpels. If each carpel has its own stigma along its midline, the stigmas are carinal.
An orthodox view implies that angiosperm ovules always develop on carpels. Problematic apocarpous gynoecia include some uniovulate carpels with a basal ovule that is apparently initiated directly on the floral apex (e.g. in Illicium, Schisandraceae). One interpretation of this basal placentation is that the ovule develops in the axil of a phyllome that constitutes the carpel wall, in which case the carpel is a complex structure. A more widely held view is that the carpels are rooted in the flower axis (receptacle), and the region of ovule initiation belongs to the carpel. Among syncarpous gynoecia, it is almost impossible to assign particular ovules to particular carpels in cases where a radially symmetrical placenta is basally attached to the geometrical centre of a unilocular ovary composed of two or more carpels (free-central placentation). An extreme condition is realized in unilocular gynoecia with a single basal ovule. Even if the number of carpels can be counted using the number of free stigmas, the central ovule belongs to the entire gynoecium rather than to any particular carpel (a condition termed mixomery). An ‘unbiased’ view of the development of such gynoecia may lead to the conclusion that the ovule is initiated directly on the floral apex and the carpels merely surround it, but do not bear it. However, the ovule-on-carpel concept can be rescued by assuming that the ‘floor’ of the unilocular ovary is composed of tissue that belongs to united carpels rooted within the flower receptacle. Experimental testing of these contrasting interpretations is problematic.
Carpels can differ from each other within a single syncarpous gynoecium, a condition termed carpel dimorphism (or polymorphism). Pseudomonomerous gynoecia possess a fertile carpel united with one or more sterile carpels. Sterile carpels are often more or less reduced, but they can bear a functional stigma.
Further reading: Volgin and Tikhomirov (1980); Sattler and Lacroix (1988); Weberling (1989); Endress (1994, 2011, 2019); Leins and Erbar (2010); Odintsova (2012); and Sokoloff et al. (2017a).
Placentation in syncarpous gynoecia. Gynoecium zones are indicated. Ovules light brown, areas of post-genital fusion yellow. (A–D) Transverse sections. (E–G) Longitudinal sections. (A) Axile placentation. (B) Parietal placentation. (C) When ventral slits of carpels close post-genitally in the symplicate zone, placentation resembles the axile type. (D) Parietal placentation with ovules shared by adjacent carpels. (E) Relative positions of synascidiate and symplicate zones with axile and parietal placentation, respectively. (F) Axile placentation with each carpel possessing one ovule in the cross-zone. (G) An extreme condition of free-central placentation (mixomerous gynoecium).
Two stigma types in syncarpous gynoecia.
Structure of the grass gynoecium
The grass flower typically has a single pistil. The grass ovary is unilocular, possessing a single ovule (Figs 2–5). In all grasses with straightforward interpretation of spikelet morphology, the ovule is attached to the posterior side of the locule wall (the side closest to the spikelet axis and palea, Fig. 2A–F). The ovule is sessile and its attachment site (termed the placentochalaza) is considerably or strongly elongated along the vertical length of the pistil (Savchenko and Petrova, 1963; Batygina, 1987). In extreme cases, the placentochalaza covers almost the entire length of the ovary locule and the ovule (Fig. 2A, D). The micropyle of the ovule is oriented towards the ovary base (Fig. 2A–F).
Diagrams of longitudinal (A–F) and transverse (G) sections of grass gynoecia. Anterior side left, posterior side right. (A–C) Species with two stigmas. (A) Bromus inermis (Pooideae). (B) Poa annua (Pooideae). (C) Molinia coerulea (Arundinoideae). (D) Species with three stigmas and a common style, Pseudosasa japonica (Bambusoideae). (E) Species with one stigma, Nardus stricta (Pooideae). (F, G) Species with two congenitally united stigmas (silk), but no common style, Zea mays (Panicoideae). (F) Longitudinal section of the gynoecium (after Miller, 1919). (G) Transverse section of the silk with two bundles and associated strands of PTTT (after Modilevsky et al., 1958). Arrowheads indicate levels of transverse sections illustrated in subsequent figures. Yellow indicates the area of post-genital fusion mostly represented by PTTT; grey, strands of PTTT not derived from epidermis; light green, embryo sac; orange, non-stigmatic hairs; black and white, free stigmas out of the plane of the section. Vasculature not shown except in (G). cn, unclosed gynoecial canal; ld, lodicule; lm, lemma; pl, palea; st, stamen. Scale bars=200 µm (A, C, D, E), 50 µm (B), 1 mm (F), 100 µm (G).
Most grasses possess two lateral stigmas (Figs 6A–E, 7E–K), though some have a single anterior stigma (Fig. 8) or three stigmas (Figs 5A, B, 6F), of which one is anterior and two are lateral (the anterior side is farthest from the spikelet axis). Very few grasses have more than three stigmas. Phylogenetic data suggest complex evolutionary patterns of stigma number in grasses, including the three small subfamilies (Anomochlooideae, Pharoideae, and Puelioideae) that form a basal grade as successive sister lineages to all other grasses (Fig. 1). Among the two genera of Anomochlooideae, Anomochloa has a single stigma while Streptochaeta has three (Arber, 1934; Sajo et al., 2008, 2012), and three stigmas are present in Pharus, the sole genus of Pharoideae (Sajo et al., 2007; Kellogg, 2015). It is most parsimonious (Fig. 1) to consider the occurrence of two stigmas as a synapomorphy of the clade that is sister to Pharoideae, which was termed the Bistigmatic Clade by the Grass Phylogeny Working Group (2001). However, maximum parsimony does not capture the entire story, and stigma number remained somewhat labile in this clade (Kellogg, 2015). In Puelioideae, Guaduella has two stigmas, but the other genus Puelia has two or three (Kellogg, 2015). Stigma number varies in all subfamilies of the BOP clade and sometimes even exceeds three in subfamilies Bambusoideae and Pooideae (Fig. 1). Pistils with three stigmas are relatively common among the bamboos; for example, Ochlandra (Bambusoideae) has 3–9 stigmas. This genus is also unusual among grasses in having numerous stamens and a variable number of lodicules as well as fleshy fruits (Arber, 1929, 1934; Rudall and Dransfield, 1989). It is the species-rich and evolutionary derived PACMAD clade in which the bistigmatic condition became finally fixed (except for some teratological variation, described below).
Gynoecium morphology in grasses, SEM. (A–C) Molinia coerulea (Arundinoideae). (A) Spikelet at female stage of anthetis with exposed stigmas. One of two stigmas is visible in each of the two fertile flowers. The other stigma is exposed in the opposite side of the spikelet. Flowers of Molinia are protandrous (Taylor et al., 2001). (B) Pre-anthetic gynoecium. (C) Detail of (B) with cylindrical stigma stalks showing no evidence of ventral slits. (D, E) Sorghum halepense (Panicoideae). (D) Gynoecium. (E) Detail of plumose stigma. (F) Pseudosasa japonica (Bambusoideae). Gynoecium and androecium. Left, anterior stamen. Centre, lateral stamen with anther removed. Another lateral stamen not visible in this view. Right, sterile posterior inner whorl stamen. Androecium of this species has 3–4 stamens. (G, H) Festuca rubra s.l. (Pooideae), sporadic abnormal gynoecia with three stigmas. (G) Gynoecium and lodicules, view from the anterior side of the flower. There are two long posterior-lateral stigmas and a shorter anterior stigma. (H) Side view of gynoecium and androecium. Anterior side of the flower is on the right. The anterior stigma is much shorter than the two posterior-lateral stigmas. Other flowers of the same plant had bistigmatic gynoecia, which is a typical condition in Festuca. cs, common style; ld, lodicule; lg, lower glume; lm, lemma; pl, palea; sg, stigma; st, stamen; sst, sterile stamen; ug, upper glume. Scale bars=1 mm (A, F, D), 300 µm (B, G, H), 100 µm (C, E).
Gynoecium development in bistigmatic grasses, SEM. (A, B) Poa annua (Pooideae). Early developmental stage with gynoecial ridge and yet no evidence of stigmas. Note that stamens are short and located below the gynoecium. (C–L) Bromus inermis (Pooideae). (C, D) Early developmental stage with gynoecial ridge and yet no evidence of stigmas, views from the posterior side. In contrast to Poa annua, anthers greatly exceed the gynoecium at this stage. (E) Initiation of stigmas, view from the posterior side. (F, G) Two views of a flower before initiation of stigmatic branches. (F) Top view, posterior side down. (G) View from the posterior side. (H, I) Stigmas with developing branches, views from the posterior side. Note development of posterior lobes of gynoecium (asterisks). A narrow zone of joint growth of these lobes can be seen in (I). Further extensive growth of this zone produces the huge posterior outgrowth of the gynoecium. (J, K) Pre-anthetic gynoecia. Simple hairs belong to the ovary (mostly to its posterior outgrowth), multicellular branches belong to the two stigmas. (J) View from the anterior side. (K) View from the posterior side. (L) Pre-anthetic gynoecium with one stigma removed to show its attachment well below the hairy physical summit of the ovary (the summit is formed by the posterior outgrowth of the gynoecium). gr, gynoecial ridge; ld, lodicule; ov, ovule; ovr, ovary; pl, palea; po, posterior outgrowth of the gynoecium; sg, stigma; st, stamen. Scale bars=30 µm (A–C), 100 µm (D–I, L), 300 µm (J, K).
Flower development in monostigmatic grass Nardus stricta (Pooideae), SEM. (A–E) Stages with gynoecial ridge visible. (F–I) Development of stigma. (J) Gynoecium and androecium of pre-anthetic flower. (K) Detail of (J) with gynoecium orifice. gr, gynoecial ridge; lm, lemma; ov, ovule; ovr, ovary; pl, palea; sg, stigma; st, stamen. Scale bars=30 µm (A–E, G–I, K), 100 µm (F), 300 µm (J).
The grass stigma is either sessile (Fig. 7I–K) or has its own stalk (Fig. 6). Another structure that is present in some grasses is a common style (Figs 5C, 6F), which appears to be plesiomorphic in grasses, though the pattern of its evolution is complex (see Kellogg, 2015). When a common style is absent, the stigmas are normally attached at the top of the ovary. The two stigmas of Bromus (Pooideae) are attached on the anterior side (Figs 2A, 3A–C, 7I–L), whereas the posterior side of the gynoecium is prolonged above their insertion as a massive hairy appendage (Walker, 1906; Knobloch, 1942). Some other Pooideae have a similar—but less pronounced—condition, with stigmas attached slightly below the physical summit of the ovary on its anterior side (Walker, 1906; Arber, 1934; Barnard, 1957; Macfarlane, 1979). The ovary of Eremopyrum buonapartis (Pooideae) has a narrow, raised ridge on one (apparently anterior) side of the ovary, but this interesting feature has not yet been studied using fixed material (Macfarlane, 1979).
Flower anatomy of Bromus inermis (Pooideae). Descending series of transverse sections. Lemma not shown. The upper side of each image is the posterior side of the flower. (A) Level above the ovary (see Fig. 4I). There are two free plumose stigmas and three stamens. The thecae of the anthers are separate at this level. (B) Level of the posterior outgrowth of the ovary (see Fig. 4J). (C) Level of the gynoecium orifice. (D) Level of the canal of gynoecium. (E, F) Ovary locule. (G–I) Below the ovary locule. The lodicules are only vascularized in their basalmost portion (I). Dark grey, lodicules; light grey, palea; green, gynoecium; blue, stamens; black, vascular bundles. ab, anterior bundle of gynoecium; cn, canal of gynoecium; lb, lateral bundle of gynoecium; ov, ovule; pb, posterior (placentary) bundle of gynoecium; plb, bundle of palea; po, posterior outgrowth of the gynoecium; sb, bundle of stamen; sg, stigma; tt, pollen tube transmitting tissue (PTTT). Scale bar=0.1 mm.
Stigmas and their stalks are solid. They are often not radially symmetrical at anthesis as they enclose a strand of pollen tube transmitting tissue (PTTT) passing closer to the inner side (i.e. towards the other stigma or other stigmas) and a vascular bundle closer to the outer side (e.g. Kiew, 1973). Li and You (1991) reported the absence of PTTT in wheat, but their own data illustrated a functional transmitting tissue with thickened cell walls at intercellular corners (see also Batygina, 1987). Stigmas of most grasses are plumose, with multicellular branches that tend to develop on the inner side of the stigma (Aziz, 1972; Kiew, 1973). The stigmatic branches are usually four celled in transverse section, and pollen tubes grow between the four cell files (Li and You, 1991). Non-plumose stigmas with simple hairs or papillae occur in a few tristigmatic (Streptochaeta, Sajo et al., 2008; Pharus, Sajo et al., 2007), bistigmatic (Eremitis, Sajo et al., 2015), and monostigmatic grasses (e.g. Anomochloa, Sajo et al., 2012; Nardus, Fig. 8J). Even though some grasses with non-plumose stigmas can be pollinated by insects, this is clearly not the case in Nardus. A common feature of these grasses is their single-flowered spikelets.
The common style, when present, is hollow early in development. In grasses bearing a common style, the pistil opening is located between the bases of the stigmas or their stalks, often (perhaps invariably) at the posterior side of the style apex (Figs 2D, 5B). In anthetic flowers of most grasses, the style becomes solid by post-genital closure of the canal and formation of a strand of PTTT in its place (Fig. 5C). The transmitting tissue of the common style is derived from the epidermis of the canal. In Hordeum, for example, four cell layers of epidermal origin can be seen in the developing PTTT (Savchenko and Petrova, 1963). Cells of the outer integument and those of the ovary wall adjacent to the micropyle differentiate in the same way as those of PTTT, at least in barley (Savchenko and Petrova, 1963). The common style remains hollow at anthesis in Pharus (Sajo et al., 2007). Grasses that lack an externally visible common style do possess a canal in the ovary wall that opens at the base of the stigmas. PTTT derived from the epidermis of the canal is present. The canal, though narrow, remains open at anthesis in some grasses such as Zea (Fig. 2F).
Zea (maize) has a very long structure called the silk that should be interpreted as two congenitally united stigmas (Weatherwax, 1916; Kellogg, 2015). Only very short distal portions of the stigmas, often of unequal length, are free (Weatherwax, 1916; Miller, 1919; Bonnett, 1948). That the silk is not homologous with the common style of other grasses is clear from the fact that it is solid from inception, develops above the narrow opening of the ovary, and has two strands of PTTT associated with the two vascular bundles (Fig. 2F, G) (Miller, 1919; Modilevsky, 1958; Dresselhaus et al., 2011). Interestingly, Cenchrus americanus (=Pennisetum americanum, pearl millet) has a structure that is fully equivalent to the silk of maize in structure and development, though much shorter (Powers et al., 1980). Maize and pearl millet are relatively distantly related to each other within panicoid grasses (Kellogg, 2015). The closest relative of Zea—Tripsacum—normally has long free stigmas that should not be termed silks for consistency of terminology (Weatherwax, 1918). A stigma of Tripsacum has a vascular bundle and associated strand of PTTT like each half of the maize silk (Lausser et al., 2010). Hybrids between Tripsacum and Zea show a range of intermediate conditions with stigmas united to various degrees to form a silk (Newell and de Wet, 1974a). In addition, variation in stigma characters was found in native populations of Tripsacum dactyloides (Newell and de Wet, 1974b). These observations offer opportunities to investigate genetic bases of silk formation.
The morphological variation of the grass gynoecium presumably has functional implications. The bistigmatic condition is efficient in many-flowered spikelets, with the stigmas releasing on either side of the spikelet (Fig. 6A) in positions where airflows are expected to deliver pollen grains (Niklas, 1985). The monostigmatic condition correlates with single-flowered spikelets, non-plumose stigmas, and the absence of lodicules, a combination of characters that could well be of functional importance because it is found in members of three different subfamilies (see also Philipson, 1985). As pointed out by Hubbard (1968), the flowers of Nardus are protogynous (rather than protandrous, Fig. 6A), the single minutely hairy stigma protruding from the apex of the floret before the anthers, as in other grasses lacking lodicules. However, seed production in Nardus is mainly apomictic and its protogyny could also be a relictual feature without any function today (Kissling et al., 2006).
Vasculature of the grass gynoecium
Gynoecium vasculature in grasses was reviewed by Arber (1934), Butzin (1965), Philipson (1985), and Pizzolato (1990, 1991). The ovary wall has a posterior bundle supplying the ovule (Figs 3–5) that either terminates or sometimes continues above the level of ovule attachment (e.g. Pizzolato, 1990). Reports of two branches of the ovular bundle that fuse with each of the two stigmatic bundles (Walker, 1906; Knobloch, 1942) are incorrect and based on misinterpretation of the PTTT as procambium. In Zea, Sorghum, and some other grasses, there is a posterior arc of numerous small bundles that supply the massive ovule (Schuster, 1910; Arber, 1934; Pizzolato, 1991). A complex structure of the ovular bundle is found in Nardus (Fig. 5G). The ovary wall also has unbranched bundles that directly correspond to the stigmas (Figs 3–5). When three stigmas are present, they are supplied by two lateral bundles and an anterior bundle in the ovary wall. Some bamboos with large pistils possess additional small bundles between those that correspond to the stigmas (Arber, 1934). When there is no anterior stigma, the anterior bundle is either absent (Fig. 4K–P) or present (Fig. 3). Grasses with a single (anterior) stigma lack lateral bundles. The stigmas are unvascularized in Agrostis canina and its small-sized ovary possesses only a posterior ovular bundle (Philipson, 1935). In Poa annua, tiny lateral bundles are present in the posterior side of the ovary, not extending into the stigmas; they join the ovular bundle in the lower portion of the ovary (Fig. 4E–G). In grasses of the PACMAD clade, the lateral bundles supplying the two stigmas are located in the radii between the median and lateral stamens (Fig. 4P), whereas in pooid and other grasses the lateral bundles of the gynoecium are in the radii of the lateral stamens (see Butzin, 1965, p. 139). We note that the two pooid grasses illustrated here differ in the physical position of the lateral stamens; this feature correlates with the position of the lateral bundles in the ovary wall (Figs 3, 4E, F). In Poa annua, both lateral stamens and lateral ovary bundles are shifted towards the posterior side of the flower (Fig. 4E, F). This feature is related to the question of whether the lateral stamens of the most common tristaminate type of grass androecium always belong to the outer whorl of the androecium (Cocucci and Anton, 1988; Rudall and Bateman, 2004).
Flower anatomy in bistigmatic grasses, descending series of transverse anatomical sections, light microscopy (LM). The upper side of each image is the posterior side of the flower. (A–H) Poa annua (Pooideae). (A) Free stigmas. (B) Gynoecium orifice (arrowheads). (C) Open canal of gynoecium (arrowheads). (D) Canal of gynoecium post-genitally closed. (E–G) Ovary. (H) Below the ovary. (I–J) Bromus inermis (Pooideae). (I) Free stigmas. (J) Free stigmas and posterior outgrowth of the ovary. (K–P) Molinia coerulea (Arundinoideae). (K) Free stigmas. (L) Non-receptive stalks of stigmas. (M) Gynoecium orifice (arrowheads). (N) Canal of gynoecium (arrowheads). (O) Ovary locule. (P) Below the ovary locule. es, embryo sac; f, stamen filament; lb, lateral bundle of gynoecium; ld, lodicule; ov, ovule; pb, posterior (placentary) bundle of gynoecium; pl, placenta; po, posterior outgrowth of the gynoecium; pttt, pollen tube transmitting tissue; ps, pollen sac; sb, stamen bundle; sg, stigma; sn, synergid; ss, stigma stalk; vb, vascular bundle supplying the gynoecium. Scale bars=50 µm (A–H, K–P), 100 µm (I–J).
Ontogeny
In most grasses studied so far, the earliest stages of gynoecium development show a posterior dome-shaped or almost flat bulge that is circular in outline, and an anterior crescent-shaped projection termed a gynoecial ridge (Aziz, 1972; Kiew, 1973; Sattler, 1973; Batygina, 1987; Yamaguchi et al., 2004; Bess et al., 2005; Reinheimer et al., 2005; Sajo et al., 2012; Pilatti et al., 2019). The arms of the gynoecial ridge approach the circular bulge. The circular bulge produces the ovule. The gynoecial ridge ultimately becomes circular by growth of the posterior side of the developing gynoecium, involving the posterior part of the circular bulge and the base of the developing ovule (Figs 7A–D, 8A–E). As a result, the ovary develops as a tubular structure with the ovule on its posterior wall. When a common style is present, it develops by growth of the distal part of this tubular structure. When a common style is absent, the gynoecial opening that occurs initially near the bases of the stigmas or stigma stalks closes post-genitally to a greater or lesser degree.
Stigmas and their stalks (when present) develop as a result of more extensive growth of certain areas at the edge of the gynoecial ridge (Figs 7, 8). Kellogg (2015) noted that in grasses with a single stigma such as Nardus, it is unclear whether the stigma is borne on an unbranched common style or an individual stalk. This question can be resolved using developmental data: the portion below the gynoecium orifice corresponds to the common style of other grasses. Grass stigmas and their stalks lack a ventral slit and are more or less cylindrical or triangular in transverse section (Li and You, 1991) early in development. The only exception known to us is the monostigmatic grass Anomochloa, in which the gynoecium closes by a conspicuous longitudinal slit located at the base of the stigma (Baillon, 1892; Sajo et al., 2012). Since Anomochloa represents one of only two genera of the basal subfamily Anomochlooideae, this unusual condition could either represent a plesiomorphic condition or a slit formed by rudiments of the lateral stigmas that have become integrated with the functional stigma.
The ovule curves by extensive growth in the region that produces the placentochalaza. This process takes place during the course of initiation and development of the integuments (Savchenko and Petrova, 1963; Kiew, 1973; Batygina, 1987). The very young ovule has a nucellus facing towards the upper anterior side of the gynoecium. The mature ovule has a micropyle facing towards its lower anterior side. The grass embryo has an asymmetrical position near the bottom of the anterior part of the caryopsis, its root pole (coleorhiza) usually facing the lower posterior side of the caryopsis (Batygina, 1987). As the root pole is directed towards the micropyle, its lower posterior direction indicates that the grass ovule continues its rotation after fertilization. Post-fertilization growth in the posterior side of the gynoecium that continues the process of ovule rotation often results in the appearance of a posterior groove in the caryopsis (the crease: Kellogg, 2015). The groove is long and conspicuous in Triticum and Secale, short in panicoid grasses, and absent in Oryza (Petrova et al., 1985).
Developmental proportions vary among species of grasses that have been examined, especially the relative timing and degree of development of the anterior and posterior sides. In Poa labillardieri, the two sides develop almost synchronously and the youngest documented gynoecium is ring shaped in outline (Ahmad et al., 2009). Bromus and Zea provide two extremes of variation. In Bromus (Fig. 7), the posterior part of the gynoecial ridge shows extensive growth. It first develops as two posterior-lateral extensions of the crescent-shaped gynoecial ridge, which increase in height before they meet on the posterior side to make the gynoecial ridge completely circular. Growth of the posterior part of the ridge takes place at very late stages, after initiation of the stigmatic papillae, but this growth is enormously extensive. It forms a huge solid posterior outgrowth overtopping the morphological apex of the ovary. The outgrowth is distally bilobed (Macfarlane, 1979; Kellogg, 2015), with the lobes corresponding to the two posterior-lateral extensions of the gynoecial ridge that occur early in development. The two stigmas are inserted on the anterior side of the gynoecium well below its physical summit, but the place of their insertion is a morphological apex of the ovary. The narrow gynoecium orifice (which remains open in pre-anthetic flowers) is located between the bases of the stigmas. In Zea, the gynoecial ridge becomes completely circular at an early stage (Cheng et al., 1983; Mena et al., 1996; Sundberg and Orr, 1996; Orr and Sundberg, 2004, 2007). The anterior side of the gynoecial ridge strongly elongates and produces the long and solid silk bearing distally two short free stigmatic lobes. The gynoecium orifice is located below the base of the silk, at the posterior side of the gynoecium. The early appearance of a complete unlobed ridge surrounding the ovule appears to be an uncommon condition in grasses other than Zea and some other panicoids (Reinheimer et al., 2010), though is also well documented in Anomochloa (Sajo et al., 2012).
Taxon sampling remains uneven in studies of grass flower development. More developmental data are required on the morphologically diverse subfamily Bambusoideae. Current knowledge suggests that early stages of gynoecium development in bamboos with a single stigma (Wang et al., 2014), two stigmas (Sajo et al., 2015), and three stigmas (Philipson, 1985) resemble each other and those observed in pooid grasses. Apart from Bambusoideae, development of gynoecia with three stigmas was studied in members of the basal grade of grasses, Streptochaeta and Pharus (Sajo et al., 2007, 2008). These taxa show very early appearance of the three lobes that ultimately produce the stigmas. In the earliest documented stages, these lobes are already united at the base to form a complete ring around the future ovule (Streptochaeta) or a ring that is almost open on the posterior side (Pharus). Unfortunately, even younger stages were not available in the material studied by Sajo et al. (2007, 2008), and it is unclear whether a crescent-shaped stage is present during early gynoecium development of these early-divergent grasses.
Early morphological interpretations of the grass gynoecium
Early ideas on the morphological identity and evolution of the grass gynoecium were reviewed by Kaden (1958a, b, 1959) and Butzin (1965). To early workers, it was always clear that grasses are monocots, but their high morphological and biological specificity raised the question of whether they are derived from ancestors with typical trimerous-pentacyclic monocot flowers or represent an ancient, isolated lineage. The first view stimulated interpretation of the organs of the grass flower as homologues of those in trimerous-pentacyclic flowers. Accordingly, it was proposed that the grass gynoecium is ancestrally tricarpellate and that the three stigmas found in some grasses belong to three different carpels. This idea was developed in detail by Schuster (1910). The grass ovule was interpreted as belonging to a parietal placenta shared by the left posterior and right posterior carpels (Čelakovský, 1894; Schuster, 1910; Arber, 1934) or belonging to only one of the two posterior carpels (Puri, 1952), though the latter view is based on rare observations of abnormal flowers (Arber, 1929) and is not supported by the bulk of other data (Butzin, 1965). The third, median anterior carpel was interpreted as sterile. Two supernumerary bundles occasionally found in the gynoecia of bamboos (Fig. 9B) were interpreted as synventral bundles of two other, sterile parietal placentas (Arber, 1934; Chandra, 1963). Progressive reduction of the median carpel resulted in loss of its stigma and ultimately in the loss of the anterior bundle of the grass ovary. In their seminal work on grass flowers, Cocucci and Anton (1988) did not use any outgroup comparisons and followed Schuster’s (1910) interpretation of gynoecium morphology. In modern terminology (see Box 2), this interpretation implies that the ancestors of grasses possessed a gynoecium with a fertile symplicate zone and carinal stigmas. Therefore, we here designate it the ‘Symplicate Hypothesis’ (Fig. 9A–C).
Contrasting morphological interpretations of the origin of the grass gynoecium (A, B, D, E, after Kircher, 1986). All images show diagrams of transverse sections of ovaries. (A–C) Evolutionary series according to the Symplicate Hypothesis. (D–F) Evolutionary series according to Kircher’s hypothesis (Kircher, 1986). (A, D) Hypothetical ancestor of grasses. (A) Proposed ancestor with fertile symplicate zone of gynoecium and ovules on parietal placentas. (D) Proposed ancestor with fertile synascidiate zone and ovules on axile placentas as in closest extant relatives of grasses such as Anarthria (Fig. 11). (B, E) Tristigmatic grasses. (C, F) Bistigmatic grasses. Red, yellow, and brown, parts belonging to three different carpels. The ovules are shared by adjacent carpels in (A–C) as indicated by colours. Black, vascular bundles. The two bundles in (B) shown as open circles were found only in a few tristigmatic bamboos. These are interpreted by the Symplicate Hypothesis as vestigial synventral bundles of the two sterile placentas, while the third synventral bundle supplies the remaining ovule of the fertile placenta.
In at least some PACMAD grasses, the location of the lateral bundles of the gynoecium in the radii between those of the stamens could suggest that the two stigmas supplied by these bundles are commissural rather than carinal (see Box 2 for terminology). Indeed, if the three stamens of most grasses belong to the outer whorl of the androecium and the inner whorl is lost (but see Rudall and Bateman, 2004), then the three carpels hypothetically comprising the grass gynoecium should be located in the radii of stamens. If the grass stigmas are carinal, they should be in the stamen radii. If the lateral stigmas are located between the radii of the median and lateral stamens, then each stigma must be commissural, namely shared by two adjacent carpels (see Butzin, 1965). We agree with Butzin (1965) that this interpretation is not convincing, because the position of the stigmas between the radii of the stamens can be equally explained by assuming that they are carinal and belong to posterior-lateral carpels shifted into almost lateral positions during the progressive reduction (and complete loss?) of the anterior carpel that is postulated for PACMAD grasses in the framework of the Symplicate Hypothesis.
The interpretation of the grasses as a taxonomically isolated, probably ancient group is congruent with the view that the grass gynoecium consists of a single carpel (e.g. Eichler, 1875; Engler, 1892). This interpretation was supported by observations of the initiation of the gynoecial ridge as a single crescent-shaped structure; these clear and convincing early observations were made using light microscopy (Wigand, 1854; Payer, 1857; Schmalhausen, 1870). At first sight, the fact that the free parts of the pistil (stigmas) are initiated after the common wall of the ovary is a strong argument in favour of the monomerous interpretation. However, these data can also be interpreted as early congenital fusion (see Box 1) between carpels, which is well documented in some monocots with undoubtedly tricarpellate gynoecia when their individual free tips appear in ontogeny after the appearance of the common triangular-tubular base of the gynoecium (e.g. Dioscorea, Dioscoreales: Remizowa et al., 2010a; Tricyrtis, Liliales: Remizowa et al., 2010b). In Costus (Zingiberales), the primordia of the two posterior-lateral carpels are initiated as congenitally united with each other, but free from the primordium of the anterior carpel (Kirchoff, 1988). Early congenital fusion of petals is common among members of the campanulid clade of eudicots (early sympetaly, Erbar and Leins, 2011), but the presence of multiple petals is not questioned, at least since the work of De Candolle (1827), who challenged earlier interpretations of a ‘monopetalous’ corolla (e.g. Rivinus, 1690).
The unicarpellate theory provides an explanation for the presence of more than one stigma in the pistils of most grasses. There are examples of bifid stigmas or even double stigmas in individual carpels of various angiosperms, but finding examples with three or more stigmas is problematic. It was therefore proposed that grass stigmas are not homologous to those of other angiosperms because they are ab initio solid and lack a ventral slit (Kaden, 1959). Such stigmas were viewed as merely late-developing outgrowths of the carpel margin. For example, they were compared with multiple awns found in the lemmas of some grasses (Schmalhausen, 1870; Eichler, 1875; Engler, 1892; Smirnov, 1953; Kaden, 1958a). The specificity of the grass stigma can be explained in terms of adaptation to wind pollination (Endress, 1995). The occurrence of ab initio solid styles (stigmas) corresponds to wind pollination (Endress, 2015). Among monocots, Endress (2015) listed only grasses as possessing this type, but noted that the character merits further studies. Contrary to the view of Kaden (1959), the stigmas of most Cyperaceae, another wind-pollinated family of Poaceae, develop in exactly the same way as those of grasses. They are ab initio circular in transverse section and are usually initiated after the formation of a ring-like common gynoecium primordium (Vrijdaghs et al., 2005a, b, 2009, 2011; Reynders et al., 2012). Ab initio solid and cylindrical styles/stigmas are also found in the wind-pollinated monocot genus Tetroncium (Juncaginaceae: Alismatales, Sokoloff et al., 2017b).
Ab initio solid and cylindrical stigmas could be interpreted as unifacial (with a morphologically abaxial surface only), by analogy with the unifacial foliage leaves of some angiosperms (Endress, 2015). Alternatively, by analogy with subunifacial foliage leaves (Ozerova and Timonin, 2009), they could be interpreted as bearing a narrow strip of an indistinctly bordered adaxial surface on the ventral side. Finally, the indistinctly bordered morphologically adaxial surface may be not especially narrow (a ‘cryptic bifacial’ condition). To our knowledge, no detailed research has been conducted attempting to distinguish between unifacial, cryptic bifacial, and subunifacial styles (Fig. 10). Data on expression patterns of genes of abaxial/adaxial polarity may help to resolve the problem.
Types of carinal stigma in angiosperm carpels: diagrams of transverse sections at anthesis; stigmatic papillae not shown. Light green, morphologically adaxial surface; yellow, site of post-genital fusion. (A) Ascidiate. (B, C) Plicate. (D–G) Ab initio solid. (D) Flat bifacial. (E) ‘Cryptic bifacial’. (F) Subunifacial. (G) Unifacial. In practice, distinguishing between types (E), (F), and (G) is potentially problematic.
A question debated within the unicarpellate concept is whether the grass ovule is purely terminal and merely becomes enveloped by the carpel during development, or whether it develops on the ventral side of the carpel (e.g. Klaus, 1966; Maze et al., 1971; Sattler, 1973). However, it is clear that the grass ovule belongs to the carpel(s) from its attachment to the ovary wall in anthetic gynoecia (Kaden, 1959; Klaus, 1966). The notion of a ‘cauline ovule’ probably appeared by simplified interpretation of developmental data (Endress, 2019).
Walker (1906), Saunders (1925), and Barnard (1957) proposed hypotheses that the grass gynoecium is a composite structure produced by a fertile carpel and varying numbers of sterile carpels. Their theories somewhat resemble the interpretation of Kircher (1986) outlined below, but differ in details. Importantly, these three earlier authors could not use detailed comparisons with phylogenetically related taxa, but paid special attention to data on floral vasculature whose significance was apparently overestimated. According to Saunders (1925), the grass gynoecium is ancestrally two whorled and trimerous, with all outer-whorl carpels sterile and solid, producing stigmas in 1-, 2- and 3-stigmatic grasses. The inner posterior inner-whorl carpel is fertile (with dorsal placentation). However, this theory fails in a modern phylogenetic context, because other commelinid monocots lack examples of two-whorled gynoecia or carpels with a single dorsally attached ovule. One could expect that the two-whorled concept of Saunders (1925) might help in understanding those rare gynoecia of bamboos that have more than three stigmas, but bundles that supply all stigmas of such gynoecia form a single series rather than two series, as would be expected in a two-whorled gynoecium (Arber, 1927, 1934).
The grass gynoecium in a phylogenetic context
The precise relationships of grasses remained controversial until the acquisition of molecular phylogenetic data. Nevertheless, by the 1980s, all key theories on the grass relatives were already focused on families currently classified in Poales. Cronquist (1981, 1988) considered Poaceae as sister to Cyperaceae. Dahlgren et al. (1985) and Takhtajan (1987, see also Takhtajan, 1966) recognized Flagellariaceae, Joinvilleaceae, Ecdeiocoleaceae, Restionaceae, Anarthriaceae, and Centrolepidaceae as the closest phylogenetic relatives of grasses. Subsequently, the latter view was fully supported by molecular phylogenetic data (reviewed by Givnish et al., 2018). Anarthriaceae, Restionaceae, and Centrolepidaceae form a well-supported restiid clade (Briggs et al., 2014) recognized as Restionaceae in a broad sense (APG IV, 2016). Poaceae plus three species-poor families, Flagellariaceae, Joinvilleaceae, and Ecdeiocoleaceae, form a graminid clade that is sister to the restiid clade (Fig. 1). Whether Ecdeiocoleaceae alone or Joinvilleaceae plus Ecdeiocoleaceae form a sister group of grasses remains under discussion, but the overall phylogenetic pattern is rather stable (Givnish et al., 2018; Wu et al., 2022).
Flowers of Restionaceae were used in the 1980s to update views on the origin and identity of the grass gynoecium (Philipson, 1985; Kircher, 1986). Like grasses, Restionaceae are wind pollinated. More derived groups of Restionaceae possess uniovulate ovaries and one-seeded fruits, thus resembling other derived lineages of wind-pollinated angiosperms, including grasses and sedges. Those one-seeded Restionaceae that belong to the family in its narrow sense have pseudomonomerous gynoecia (Kircher, 1986; Linder 1992b). There is a consistent view that the ancestral conditions of the restiid clade include a tricarpellate gynoecium with three locules, each having a pendent ovule, and a dehiscent three-seeded fruit (Kircher, 1986; Ronse De Craene et al., 2001, 2002; Fomichev et al., 2019). An example of such a gynoecium with three fertile ovules and three equal stigmas is illustrated here for a species of the Western Australian genus Anarthria (Fig. 11). This tricarpellate gynoecium can be interpreted as possessing a long synascidiate zone and a short symplicate zone. At anthesis, the inner space of the symplicate zone is sealed by post-genital fusion of the inner surfaces of the gynoecium with cells derived from the inner epidermis differentiating as PTTT. The ovules are inserted at the uppermost part of the synascidiate zone.
Gynoecium anatomy of Anarthria scabra (Restionaceae–Anarthrioideae). (A) Longitudinal section. (B–G) Transverse sections. db, dorsal bundle; it, inner tepal; ot, outer tepal; ov, ovule; pl, placenta; sm, staminode; svb, synventral bundle; tt, pollen tube transmitting tissue (grey); yellow dashed line, area of post-genital fusion. Scale bar=1 mm.
Philipson (1985) used comparisons with Restionaceae to defend the view that the gynoecium of grasses is pseudomonomerous. When a grass gynoecium has three stigmas, it should be interpreted as possessing three carpels. Kircher (1986) greatly refined the interpretation of grass gynoecia; he criticized the view of Schuster (1910) who implied that the grass ovule belongs to a parietal placenta shared by the two posterior-lateral carpels (the Symplicate Hypothesis; Fig. 9B, C). Instead, Kircher (1986) proposed that the ancestors of grasses possessed gynoecia with a pronounced trilocular synascidiate zone and axile placentation, like many extant Restionaceae such as Anarthria (Fig. 9D–F). Therefore, the grass ovule is borne on an axile placenta of the anterior-median carpel; the posterior-lateral carpels are sterile and their individual locules (in the synascidiate zone) were lost during the course of evolution. Only the anterior carpel retained a recognizable ascidiate zone. This conclusion has interesting implications for interpretation of bistigmatic gynoecia in grasses (Fig. 9F). The stigmas belong to the two sterile carpels. When a median stigma is absent, as in most grasses, the fertile carpel lacks a stigma. This represents the greatest possible degree of functional and structural dimorphism of the carpels in angiosperms; it is also documented in a few scattered eudicots (Lagoecia, Apiaceae, Apiales: Magin, 1980; Buchanania, Anacardiaceae, Sapindales: Bachelier and Endress, 2009). The eudicot Pennantia corymbosa (Pennantiaceae, Apiales) has a trimerous gynoecium in which the stigma of the fertile carpel is smaller than in the two sterile carpels (Karpunina et al., 2022). In the trimerous gynoecium of Tricomaria usillo (Malpighiaceae, Malpighiales), the stigma of the median anterior carpel is generally smaller than in the posterior carpels or even sometimes missing, but fruits typically contain a single seed that develops in the anterior locule (Aliscioni et al., 2019). Thus, this trend of gynoecium evolution proposed for grasses is not unique among angiosperms.
Kircher’s (1986) theory provides a straightforward interpretation of grass gynoecia with two and three stigmas, but it is problematic with respect to those possessing a single median stigma. For Nardus and Lygeum, two sister genera that belong to the Bistigmatic Clade (Fig. 1), complete suppression of the posterior-lateral carpels plus regain of the stigma in the fertile carpel would be required. The anterior vascular bundle of Nardus is sometimes displaced from the median position (Fig. 5G, H), suggesting that the only stigma supplied by this bundle is derived from one of the two lateral stigmas of bistigmatic grasses. However, early stages of gynoecium development in Nardus show no evidence of asymmetry (Fig. 8A–G; see also Schmalhausen 1870; Baillon, 1893a; Philipson, 1985). The slightly displaced position of the stigma in some flowers of Nardus can be readily explained by its tight packing with the three anthers in the flower bud. Data on gynoecium vasculature in Lygeum are controversial and not documented by photographs (Schuster, 1910; Arber, 1928). Re-investigation of this key genus may shed new light on the evolution of monostigmatic grasses.
Flower anatomy in tristigmatic (A–D) and monostigmatic (E–I) grasses, LM. Descending series of transverse anatomical sections. The upper side of each image is the posterior side of the flower. (A–D) Pseudosasa japonica (Bambusoideae). (A) Free stigmas. (B) Gynoecium orifice (arrowheads). (C) Common style with canal of gynoecium post-genitally closed. (D) Ovary. (E–I) Nardus stricta (Pooideae). (E–H) Serial sections of flower with non-median position of the anterior bundle of the gynoecium. (E) Style. (F) Upper part of the ovary. (G) Lower part of the ovary; note that the posterior bundle is composed of four distinct parts (asterisks). (H) Below the ovary. (I) Section similar to (H) from a flower with median position of the anterior bundle of the gynoecium. ab, anterior bundle of the gynoecium; es, embryo sac; f, stamen filament; lb, lateral bundle of gynoecium; ld, lodicule; ov, ovule; pb, posterior (placentary) bundle of gynoecium; pttt, pollen tube transmitting tissue; sb, stamen bundle; sg, stigma. Scale bars=50 µm (A, B, D, F–I), 100 µm (C, E).
Broad morphological studies fully support Kircher’s (1986) view that grasses are derived from an ancestor with a trilocular, tricarpellate gynoecium and a single pendent ovule per carpel with axile placentation. Gynoecia with a single ventrally attached pendent ovule and a single median stigma per carpel characterize (without exception) all the families that form a phylogenetic grade leading to the grass family: Restionaceae, Flagellariaceae, Joinvilleaceae, and Ecdeiocoleaceae (Linder, 1992a, b; Rudall and Linder, 1988; Linder and Rudall, 1993; Rudall et al., 2005; Sajo and Rudall, 2012). Gynoecia of Flagellariaceae and Joinvilleaceae are always tricarpellate, while those of Ecdeiocoleaceae are tricarpellate (Georgeantha) or bicarpellate (Ecdeiocolea) (Rudall et al., 2005). Gynoecia of Restionaceae are almost certainly ancestrally trimerous. Among the three genera of Restionaceae–Anarthrioideae, two have three carpels, but Hopkinsia has a single carpel (Fomichev et al., 2019). In Restionaceae–Centrolepidoideae, the gynoecium has from one to >40 carpels, but individual carpel morphology strictly follows the common groundplan of the family (Sokoloff et al., 2009, 2015). In core Restionaceae (Restionaceae sensu stricto), gynoecia are trimerous, dimerous, or pseudomonomerous (Linder, 1992a, b; Ronse De Craene et al., 2001, 2002).
Data on gynoecium structure and development in the basal grade of grasses, Pharus and Streptochaeta (Sajo et al., 2007, 2008), support earlier views that their three stigmas are homologous with stigmas of other monocots and represent carpel tips. In particular, the arrangement of the three stigmas in the earliest available stages of flower development in Streptochaeta is almost symmetrical, a condition that would be expected in a normal tricarpellate gynoecium. To summarize, comparative morphology in a broad phylogenetic context leaves no doubt about the plausibility of Kircher’s (1986) interpretation of grass gynoecia, at least in bistigmatic and tristigmatic grasses.
Heterochrony as a clue for understanding the evolution of the grass gynoecium
Grasses differ in ovule morphology from their closest phylogenetic relatives. Ovules of Restionaceae, Flagellariaceae, Joinvilleaceae, and Ecdeiocoleaceae are orthotropous (Hamann, 1962; Kircher, 1986; Rudall, 1997; Rudall and Linder, 1988; Linder and Rudall, 1993; Rudall et al., 2005; Sajo and Rudall, 2012; Fomichev et al., 2019). Like those of other graminids, grass ovules possess a micropyle oriented towards the base of the ovary, but they are never orthotropous (Savchenko and Petrova, 1963; Aulbach-Smith and Herr, 1984; Petrova et al., 1985; Batygina, 1987). Grass ovules vary in degree and mode of curvature, and thus are variously termed in the literature, but curvature is always present, sometimes approaching the common angiosperm condition of an anatropous ovule. In addition, they often have a vertically elongated zone of attachment (placentochalaza), a feature that is unusual among angiosperms (also found in Posidonia, Alismatales, Remizowa et al., 2012). These aspects of the grass ovule should be interpreted together with the developmental specificity of their gynoecium. In our view, major evolutionary events related to the origin of grasses were the loss of ovules in the posterior-lateral carpels coupled with a shift in the initiation of the remaining ovule of the anterior carpel at progressively earlier developmental stages. Such a heterochronic shift resulted in ovule initiation at a developmental stage when the locule of the fertile carpel is still virtually absent. The ancestral ovule was orthotropous and pendent. The orientation of the micropyle towards the base of the gynoecium is apparently functionally important for grasses and their relatives (note that it determines the position of the root pole of the embryo). Young grass ovules cannot be directed to the ovary base, because the locule is delayed in development. Thus, a process of ovule rotation takes place during development. When the process of rotation involves combined growth of the ovule base and the posterior wall of the ovary, an elongate attachment zone is formed (see Savchenko and Petrova, 1963).
Early initiation of the grass ovule together with progressively delayed stigma initiation resulted in developmental ‘decoupling’ of the grass ovule from the stigma of its carpel. When grass stigmas are initiated on the margin of the gynoecial ridge, this represents a developmentally simple process. The wider the ridge and the narrower the stigmas at initiation, the greater the number of stigmas that can be initiated in the gynoecium. Such variation in proportions is apparently responsible for the extensive variation in stigma number in bamboos. Developmentally, these stigmas already belong to the entire gynoecium rather than to individual carpels. This interpretation is complementary to ideas on carpel dimorphism outlined above. Although of very different evolutionary origin, the grass gynoecium and the behaviour of its stigmas approach the condition of mixomery found in Cyperaceae (Reynders et al., 2012; Sokoloff et al., 2018).
Developmental genetics and teratology of the grass gynoecium
Developmental genetics has provided considerable data on grass flowers since their establishment as model taxa. Regulation of grass flower development follows broadly the same general principles as in other angiosperms, including the model eudicot Arabidopsis (Whipple and Schmidt, 2006; Li and Liu, 2017), though a member of the YABBY family, DROOPING LEAF (DL), plays a key role in carpel specification in rice, and its function is not the same as that of its Arabidopsis orthologue CRABS CLAW (Yamaguchi et al., 2004; Whipple and Schmidt, 2006). Hirano et al. (2014) stated that DL transcripts appear to accumulate in three roughly separate domains in a transverse section of the rice pistil, implying that it is formed from three congenitally fused carpels. However, the only transverse section illustrated by Yamaguchi et al. (2004) is at a late developmental stage, which reduces the significance of these observations.
Of special interest are morphologically characterized mutants of various grasses with sporadic or constantly increased numbers of pistils. These can be discussed together with developmental abnormalities discovered in the pre-molecular era (reviewed by Kaden, 1958b; Clifford, 1986). In the vast majority of cases, multiplication of the female organs in grass flowers results in several pistils, each usually having more than one stigma and often two stigmas (Kaden, 1958b; Semenov and Ryabinina, 1978; Wang et al., 1989, 1990; Yamaguchi et al., 2004; Zhang et al., 2004; Ohmori et al., 2009; Li et al., 2011; Sugiyama et al., 2019; Saha et al., 2020, Preprint). Detailed developmental studies of zag1-mum1 mutants in maize showed a loss of flower determinacy and successive development of more than one (typically sterile) open pistil enclosed in each other (Mena et al., 1996). Each pistil behaved like a phyllome early in development, but produced an apically bilobed silk in exactly the same way as the wild-type maize gynoecium (Mena et al., 1996). Note that these bilobed silks must be interpreted as stigmas of two congenitally fused carpels (see above). Thus, developmental plasticity of teratological flowers provides no evidence of the separate behaviour of individual carpels that would be expected in the framework of either the Symplicate Hypothesis or Kircher’s (1986) hypothesis. The literature on grass genetics tends to use the terms carpel and pistil as synonyms, which is highly misleading (see Box 1). The grass gynoecium is such a developmentally integrated unit that it behaves like a single phyllome when studied using experimental methods.
We illustrate an abnormal flower of Bromus inermis found in a plant with otherwise normal flowers (Fig. 12E–H). It shows abnormalities that are also recorded in other grasses, including wheat, where similar morphotypes are of potential agricultural importance (Wang et al., 1989, 1990; Li et al., 2020). This type of flower can be interpreted as incompletely peloric; that is, its abnormalities reduce the degree of monosymmetry (Rudall and Bateman, 2003). There are two additional pistils, each in front of the two anterior lodicules. A lodicule is present in the posterior position, in the radius of the ‘main’ pistil (the posterior lodicule is normally absent in most pooid grasses including Bromus). Our flower is at a young stage, but each of the two additional pistils of the three-grain wheat forms two stigmas in the same way as the ‘main’ pistil (Fig. 12E). Such an abnormality can by no means be interpreted as an atavistic indication of ancestral apocarpy. If each pistil is interpreted as a carpel, then the carpels have inverted polarity, with their putative ventral sides bearing ovules facing outwards (Fig. 12E). This and many other abnormalities demonstrate the considerable integrity of grass pistil development.
Normal and abnormal flowers of pooid grasses. (A–E) Floral diagrams, stigmas red. (A) Diagram of typical trimerous pentacyclic monocot flower (after Remizowa et al., 2013). (B) Typical bistigmatic condition of pooid grasses; note organ homologies with (A). (C, D) Sporadic abnormal flowers with three stigmas in contrasting positions. (E) Abnormal flower with three pistils of the type shown in (F–H). (F–H) Sporadic abnormal flower with three pistils in Bromus inermis (Pooideae), SEM. The same individual plant also had normal flowers, some of which are illustrated in Fig. 7. (F) View from the anterior side. (G) As in (F), but anthers removed. (H) Top view of the flower with anthers removed, posterior side down. ld, lodicule; mp, main pistil; ov, ovule; pl, palea; snld, supernumerary lodicule; snp, supernumerary pistil; st, stamen; Scale bars=0.1 mm.
There are many reports of grass species that normally have two lateral stigmas but occasionally develop a third stigma (Fig. 6G, H) or a small non-receptive appendage. Interestingly, the third stigma or appendage does not always develop in an anterior position, where we would expect to find it on theoretical grounds (compare Fig. 12A and B). For example, Schmalhausen (1870) observed the sporadically occurring third stigma in either the anterior (Fig. 12C) or posterior (Fig. 12D) position in different flowers in Hordeum. He also documented a gynoecium with five stigmas of which the odd one was in a posterior position (Schmalhausen, 1870). In the framework of the Symplicate Hypothesis, Butzin (1965, p. 139) interpreted the occasional median posterior stigma as a ‘genuine commissural stigma’ shared by the two posterior-lateral carpels. Kircher’s (1986) hypothesis allows a simpler explanation, such as the appearance of yet another sterile carpel with its carinal stigma (see also Čelakovský, 1889). Most reports of the median posterior stigma are not supported by illustrations (e.g. Schenck, 1867; Baillon, 1893b; but see Schmalhausen, 1870), and SEM-based developmental data are urgently needed. Some reports of infraspecific variation of stigma number in grasses do not consider the anterior versus posterior position of the third stigma (e.g. Singh et al., 2012). The variation in the position of the occasional third stigma in derived lineages of grasses resembles the variation in orientation of bistigmatic gynoecia found in Cyperaceae (Reynders et al., 2012).
A need for more comparative developmental data on grass flowers
Well-documented anatomical and developmental data on grass gynoecia still cover a tiny fraction of the phylogenetic diversity of the family. Use of SEM is important for proper documentation. For example, the pioneering study of Payer (1857) illustrated the stigmas in Panicum as derived from the abaxial surface of the gynoecium wall, well below the margin of the gynoecium orifice. It is tempting to suggest that these observations are incorrect and that the stigmas of Panicum are initiated at the boundary of the abaxial and adaxial surface of the gynoecium in the same way as in other grasses, but the SEM-based study of Reinheimer et al. (2005) did not cover all stages of stigma development and possibly investigated a different species of Panicum, a genus that is polyphyletic in its traditional concept (Aliscioni et al., 2003). Payer (1857) stated that he studied P. aduncum, but this name is not used in any taxonomic accounts of the genus. Intriguingly, some images in an incomplete developmental series of another panicoid grass, Moorochloa eruciformis, give an impression of stigmas attached somewhat below the gynoecium orifice (Reinheimer et al., 2010). Thus, variation of a character of fundamental importance remains to be understood in panicoid grasses.
The closest outgroup of grasses with well-investigated early stages of gynoecium development is Anarthria in Restionaceae (Fomichev et al., 2019). It remains unknown whether the carpels of the related families Flagellariaceae, Joinvilleaceae, and Ecdeiocoleaceae are initiated as distinct primordia in the same way as in Anarthria, or as a common ridge, as in grasses. It is possible that stigmas of Flagellariaceae, Joinvilleaceae, and Ecdeiocoleaceae are plicate, as in Anarthria, rather than solid from inception, as in grasses, but more data on early development are required to test this hypothesis.
Conclusions and perspective
Broad comparative analysis in a modern phylogenetic context clearly demonstrates that the grass gynoecium is pseudomonomerous. A bistigmatic grass gynoecium has three carpels: two sterile carpels that each produce a stigma, and a fertile carpel that lacks a stigma (Kircher, 1986). To date, studies of grass developmental genetics and developmental morphology have failed to fully demonstrate the composite nature of the grass gynoecium because its complex evolutionary history is hidden by extreme organ integration. As nicely demonstrated by Klaus (1966), taken in isolation, data on gynoecium development in bistigmatic grasses such as barley have nothing that contradicts its interpretation as a solitary ascidiate carpel with the ovule initiated in the cross-zone. Nevertheless, in the context of other data, this interpretation is clearly untenable.
There are apparently optimal modes of development for an angiosperm pistil possessing a single ovule. The most reduced types of grass gynoecium (e.g. in Nardus) show close similarity to the free ascidiate carpels found in various angiosperms (Philipson, 1985). It is highly problematic to interpret the evolutionary history of the grass gynoecium in terms of normal angiosperm gynoecium typology. Such gynoecia are too integrated and too reduced to recognize (sym)plicate and (syn)ascidiate zones (see also Bachelier and Endress, 2007; Karpunina et al., 2022). Even the concept of carpel becomes misleading in grasses; instead, we recommend the term pistil for descriptive purposes.
The grass gynoecium provides a case study that allows us to discuss the origin of the angiosperm carpel and angiosperm gynoecium. Assuming an unlikely hypothetical scenario where grasses were the only monocots available for comparative studies, even the most in-depth possible developmental, genetic, genomic, and other experimental methods would be unable to uncover the evolutionary origin and morphological identity of the grass gynoecium. In the same way, outgroup data are essential in uncovering the nature of the angiosperm carpel. Even if developmental data on extant species suggest that the angiosperm carpel represents a simple phyllome, the real evolutionary history of the angiosperm carpel could well be far more complex (e.g. Doyle, 2008).
Acknowledgements
We are grateful to Rainer Melzer for his invitation to write this review and to two anonymous reviewers for helpful suggestions. The use of a scanning electron microscope was greatly facilitated by the assistance of the staff of the Laboratory of Electron microscopy at the Faculty of Biology, Lomonosov Moscow State University.
Author contributions
DDS, CIF, PJR, TDM, and MVR: evaluation of hypotheses on grass gynoecium; DDS, CIF, and MVR: generation and interpretation of illustrative material. DDS: writing—original draft. All authors contributed to its final form.
Conflict of interest
The authors have no conflict of interest to declare.
Funding
The work is supported by the Russian Science Foundation (project 19-14-00055, flowers of Poaceae) and the Russian Foundation for Basic Research (project 20-34-90162, flowers of Anarthria).
Data availability
All data are available within the paper.
Comments