Early Cretaceous Umkomasia from Mongolia: implications for homology of corystosperm cupules

Introduction

The so-called ‘Mesozoic seed ferns’, a loose assemblage of extinct gymnosperms, are generally taken to comprise Caytonia, corystosperms and peltasperms (Taylor et al., 2006; Taylor & Taylor, 2009). These three groups, together with the Permian glossopterids, have been considered to be of crucial importance for understanding the phylogeny of seed plants and the evolution of their key structural features, including, especially, the carpel and ovule of angiosperms (Doyle, 2006; Hilton & Bateman, 2006). However, phylogenetic relationships among these groups, and their relationships to other lineages of living and fossil seed plants, remain uncertain, in large part because of incomplete knowledge and uncertain homologies among their reproductive organs (Doyle, 2006; Hilton & Bateman, 2006).

The corystosperms (=Umkomasiaceae Petriella, 1981) were first recognized by Thomas (1933) based on compression fossils from the Late Triassic Molteno Formation of Upper Umkomaas Valley, Natal, South Africa. The group initially included the seed-bearing organs Umkomasia Thomas, Pilophorosperma Thomas and Spermatocodon Thomas, the pollen organ Pteruchus Thomas, and pinnately compound leaves assigned to Dicroidium Gothan. These organs were thought to have been produced by the same group of plants because of their close association in the same bed, similarities in the structure of their cuticles and the occurrence of Pteruchus pollen grains in the micropyles of Umkomasia seeds (Thomas, 1933).

Since the initial description of the corystosperms, much new information has accumulated. The earliest occurrences are from the Late Permian of Jordan (Kerp et al., 2006) and India (Chandra et al., 2008), and the group appears to have attained its maximum diversity and widest distribution in Gondwana during the Middle and Late Triassic (Kerp et al., 2006), with some relictual occurrences as late as the Eocene, based on leaves (McLoughlin et al., 2008) as well as putative dispersed pollen (Harris, 1965) from southeastern Australia. There are also scattered reports of probable corystosperms from the Triassic (Zan et al., 2008), Jurassic (Kirchner & Müller, 1992) and Cretaceous (Stockey & Rothwell, 2009) of the Northern Hemisphere.

Seed-bearing organs of corystosperms are generally assigned to Umkomasia. They consist of a branch system, generally of several orders. Usually, there is a central axis with helically arranged, variously forking, lateral fertile branches that bear terminal cupules (Playford et al., 1982; Anderson & Anderson, 2003; Holmes & Anderson, 2005). Cupules are reflexed and generally enclose a single seed. The extent to which the seed is enclosed by the cupule varies among the species of Umkomasia (Axsmith et al., 2000; Klavins et al., 2002). Based on the original material from South Africa, the cupules were reconstructed as lobed or unlobed structures, within which the seed was loosely enclosed (Thomas, 1933). However, studies of other compression material from the Triassic of Antarctica (Axsmith et al., 2000), and permineralized specimens from the Early Cretaceous of Canada (Stockey & Rothwell, 2009), have shown that, in presumed immature cupules, the seed is completely enclosed, except for a small distal opening, and that, at maturity, the seeds were shed when each cupule split into three valves.

In this article, we describe a new species of Umkomasia based on abundant, three-dimensional, lignified mesofossils from the Early Cretaceous of Mongolia. These exceptionally well-preserved specimens provide more detailed information on the morphology of corystosperm cupules than has been available previously. The material preserves major features of anatomy and also includes cupules of different sizes that were presumably preserved at different stages of development. Combined with information from compression fossils (Axsmith et al., 2000) and anatomical information from permineralized material (Klavins et al., 2002; Stockey & Rothwell, 2009), the new specimens from Mongolia highlight persistent questions about how the seed-bearing structures of corystosperms should be compared with those of other seed plants. Different interpretations of the homology of corystosperm cupules have significantly different implications for understanding seed plant phylogeny, including the origin of angiosperms.

Materials and Methods

Bulk samples of poorly consolidated lignite containing the mesofossils described here were collected during two field seasons (2011, 2013) from the Tevshiin Govi Formation at Tevshiin Govi, an open-cast lignite mine (45°58′54″N, 106°07′12″E) in central Mongolia, c. 220 km south of Ulaanbaatar. The Tevshiin Govi Formation is one of several non-marine Lower Cretaceous coal/lignite-bearing beds in eastern and central Mongolia that formed in ancient peat swamps (Ichinnorov, 2003). The formation is thought to be Aptian–Albian in age (125–100.5 million yr ago (Ma)) based on palynological assemblages and stratigraphic correlations (Ichinnorov, 2003). The assemblage of plant fossils at Tevshiin Govi is dominated by wood, leaves, seed cones and seeds of probable stem group and crown group Pinaceae and Cupressaceae, as well as other conifers (Leslie et al., 2013; Shi et al., 2014; Herrera et al., 2015). Other seed plants are also present, including the seed-bearing organs of corystosperms described here.

Mesofossils were extracted from bulk lignite samples by disaggregation and sieving in soap and water. Fossils were sorted under a stereomicroscope. Measurements are based on c. 100 specimens. Selected specimens were cleaned with hydrofluoric and hydrochloric acid to remove adhering mineral matrix, and thoroughly rinsed in distilled water. Adhering organic matter was removed manually using a small brush and forceps. Photographs and measurements were made using a Leica MZ16 stereomicroscope, with a Leica DFC420 digital camera system, and an FEI XL-30 ESEM-FEG scanning electron microscope (SEM) in the Department of Geology and Geophysics at Yale University, New Haven, CT, USA. Synchrotron-radiation X-ray tomographic microscopy (SRXTM) was carried out at the Advanced Photon Source in the US Department of Energy Argonne National Laboratory using the techniques described previously (Schönenberger et al., 2012).

Mesofossils for anatomy were first soaked in 10% hydrochloric acid, followed by Aerosol OT (10% solution of sodium dioctyl sulfosuccinate in alcohol). Specimens were taken through a series of ethanol concentrations (70% to absolute ethanol) and embedded in Technovit 7100 following the prescribed mounting protocol. Embedded mesofossils were sectioned into transverse and longitudinal sections c. 4–7 μm thick using a Leica 2030 microtome.

Results

Systematic palaeontology

Umkomasiales Doweld

Umkomasiaceae Petriella

Umkomasia Thomas

Umkomasia mongolica Shi, Leslie, Herendeen, Herrera, Ichinnorov, Takahashi, Knopf et Crane sp. nov.

Etymology: The specific epithet derives from Mongolia, where the new species was found.

Holotype: PP55648 (Fig. 1).

Other illustrated material: PP55781 (Fig. 2a), PP55773 (Fig. 2b), PP55771 (Fig. 2c), PP55779 (Fig. 2d), PP55774 (Fig. 2e), PP55817 (Fig. 2f), PP55840 (Fig. 2g), PP55803 (Fig. 2h), PP55760 (Fig. 2i), PP55810 (Fig. 2j), PP55846 (Fig. 2k), PP55753 (Fig. 3a–e), PP55767 (Fig. 3f), PP55793 (Fig. 3g), PP55757 (Fig. 3h,j), PP55783 (Fig. 3i), PP55754 (Fig. 3k), PP55890 (Fig. 4a,b,d), PP55891 (Fig. 4e,f). For addition material, see Supporting Information Notes S1. All specimens are deposited in the Paleobotanical Collections of the Department of Geology, Field Museum, Chicago, IL, USA.

Locality and horizon: Tevshiin Govi coal mine (45°58′54″N, 106°07′12″E), Central Mongolia. Tevshiin Govi Formation, Aptian–Albian stage (125–100.5 Ma), Early Cretaceous.

Diagnosis: Seed-bearing unit consisting of a narrow, elongate, bract subtending, and partially fused to, an axis that bifurcates into two cupule stalks, each of which is strongly reflexed distally and bears a cupule containing a single erect seed. Each cupule consists of the reflexed cupule stalk, which bears a single seed at, or near, its tip, and two lateral flaps that are fused to form a bilobed structure. The reflexed portion of the cupule stalk results in the micropyle of the seed being oriented back towards the base of the cupule stalk. Together, the reflexed portion of the cupule stalk and the two lateral flaps enclose the seed. Seeds ovate in lateral outline, three-angled, with three large lateral faces that have weakly developed wings along the margins, and a flat basal triangular attachment scar. Integument with a thin outer cuticle and a well-developed sclerotesta composed of two to three layers of thick-walled sclerenchyma.

Description and comments on the species

The Mongolian Umkomasia material includes seed-bearing units, isolated cupules and isolated seeds. Each seed-bearing unit consists of an elongate bract subtending and partially fused to a bifurcating axis (Fig. 1). Each fork of the bifurcating axis (the cupule stalk) bears a strongly reflexed cupule at, or near, the tip (Fig. 2a,b). Each cupule is formed by the reflexed distal-most portion of the cupule stalk and the two lateral flaps of the bilobed structure. The cupule stalk and the two lateral flaps correspond to the three lateral faces of the single seed borne in each cupule (Figs 3b–e, 4b,c). At the base of each seed-bearing unit, there is an elongated abscission zone (Fig. 1a,c) which suggests attachment to a main axis. As all the dispersed units are identical in their organization (Figs S1, S2), they are probably the deciduous lateral units of a cone-like structure. From the shape and orientation of the abscission zone on the surface away from the bract (Figs 1, S2i,g), the cupules of seed-bearing units were reflexed upwards (adaxially) towards the putative main cone axis (Fig. 5).

The bract extends from the base of each seed-bearing unit approximately to the point at which the axis bifurcates (Figs 1b,d, 2b). It is narrowly elongate, 4.0–5.5 mm long, 0.8–1.5 mm wide, and broadens gradually towards the rounded apex. The bract has an entire margin, is c. 300 μm thick and is crescent shaped in transverse section. It is fused with the axis in the midrib region, but free at the apex and along its lateral margins (Fig. 2c).

The axis of each seed-bearing unit is 3.2–5.3 mm long, 0.5–0.9 mm wide, and broadens gradually from the base to the point of bifurcation. In most specimens, both forks of the axis are similar in length and both cupules of a single seed-bearing unit are equally well developed (Fig. 2a,b). In a few seed-bearing units, one cupule is aborted (Fig. S2o,p). Isolated cupules occur frequently in our material, but none shows a distinct abscission layer near the base of the cupule or along the cupule stalk, suggesting that the cupules persisted on the bifurcating axes at maturity and that the isolated cupules recovered were broken before deposition or, in some cases, during preparation. Isolated cupules are often laterally compressed (Fig. 2d–f).

Larger (presumed mature) cupules are obovoid (Fig. 2a,k), 2.3–3.5 mm long and 1.5–2.7 mm wide. The cupule stalk is 0.8–1.1 mm wide and 3.3–4.4 mm long from the point of bifurcation to the point of attachment of the bilobed structure. The cupule stalk is commonly flattened, especially adjacent to the seed, where it forms one of the three sides of the cupule (Figs 2b, 3c–e, 4b,c).

In many cases, there is a protrusion on the back of the cupule, which may be very short or more pronounced. This results in some cupules being pointed, rather than rounded at the end (Fig. 2b,d,k). In some specimens, the protrusion is quite conspicuous (Fig. 2b,d) and, in these cases, like the cupule stalk, it appears to be composed of woody tissue (Fig. S3c,d). The protrusion is not thin textured and leaf like.

Most of the large cupules appear empty (Fig. 2a,b,d,k), suggesting that the seeds were only loosely enclosed and were shed at maturity. However, where seeds are still present, the flaps loosely enclose the seeds. The two lateral faces of the three-angled seed are covered by the lateral flaps of the cupule; the other face is covered by the cupule stalk (Figs 3b–e, 4b,c). This organization is consistent in all of our material, and in none of the specimens is the cupule unlobed, as in some other corystosperm ovulate structures. The lateral flaps of the cupule are generally fused along their outer margins into a symmetrical or asymmetrical bilobed structure (Fig. 2a,k). The bilobed structure is attached close to the point of attachment of the seed. It is free from the cupule stalk, except at the base (Figs 2d,e, S4). The outer surface of the lateral flaps is slightly wrinkled (Fig. 2a,d,e), perhaps indicating that they were originally fleshy.

Small, presumed aborted, cupules in otherwise well-developed seed-bearing units probably reflect preservation at an early stage of development (Figs 2g–i, S5). In these smaller cupules, the cupule stalk is less flattened than in the larger cupules (Fig. 2g,h) and the seed is commonly visible between the lateral flaps and the cupule stalk (Fig. 2i). In many of these cupules, there is a distinct median ventral cleft on the cupule stalk immediately distal to the point of seed attachment. This ventral cleft is flanked on each side by a distinct bulge, and each bulge extends into one of the two lobes of the bilobed structure (Figs 2i, S5). This may suggest that the bilobed structure was formed by two lateral flaps, which were initiated separately. The inner margins of the lateral flaps are appressed to, but not fused with, the edge of the cupule stalk (Fig. 2g–i).

The cuticles of the seed-bearing units are very delicate. We were unable to isolate them by maceration, but some aspects of the cuticular structure are visible with scanning electron microscopy (Fig. 3f). Trichomes or papillae are absent. Below the point of bifurcation, on the surface of the axis that faces away from the bract (adaxial), epidermal cells are rectangular and usually arranged in longitudinal files with straight anticlinal and smooth periclinal walls; no stomata were observed. Epidermal cells on the abaxial surface of the bract are generally similar to those of the axes in form and arrangement. In some cases, stomata are present, although their structure is not clear.

The cuticle of the lateral flaps forming the bilobed structure contains a few scattered stomata that are irregular in their orientation and distribution. Stomata are haplocheilic (anomocytic) and monocyclic (Fig. 3f). Stomatal pits are elliptical, c. 16 μm long and 8 μm wide. Guard cells are sunken and surrounded by 7–10 subsidiary cells that are irregular in shape and slightly smaller than ordinary epidermal cells. Near the stomata, the epidermal cells are irregular in shape and orientation. Epidermal cells in non-stomatal regions are regularly arranged in longitudinal files. They are rectangular and longitudinally elongate, 40–65 μm long, 11–18 μm wide.

The axis, cupule stalk and cupular flaps show distinct inner and outer cortical zones (Fig. 4a,b,d). The inner cortical zone is especially well developed in the axis and cupule stalk (Fig. 4a,b). It is less prominent in the flaps and absent in the most distal part of the cupule (Figs 4d, S3). At the point of seed attachment, the inner cortical zone is thin and horseshoe shaped (Figs 4d, S3). Cells of the outer cortical zone are thin walled, anisodiametric (rounded to polygonal) and typically 20–25 μm in diameter. In longitudinal section, they are up to 200 μm in length. Thick-walled sclerified cells are common and are scattered through the outer cortical zone (Fig. 4a,b). Cells of the inner cortical zone are more consistently thick walled compared with cells in the outer cortical zone. They are variable in shape (rounded to elliptical), typically 10–15 μm in diameter and generally filled with dark contents. In longitudinal sections, these cells are elongated, and up to 1 mm long (Fig. S3).

Seeds borne in the cupules are erect, sessile and have three lateral faces. The seed is borne at, or near, the apex of the strongly reflexed portion of the cupule stalk (Figs 2f, S4). The micropyle of the seed is oriented back towards the base of the cupule stalk (Figs 2e,f, S4).

Isolated seeds, similar in form and size to those occurring in cupules, are common in the same samples as the seed-bearing units. They are ovate in lateral view, three-angled, 2.3–2.9 mm long, 1.6–2.4 mm wide with three, weakly developed, narrow lateral wings (Fig. 3g,h). Each seed has a basal triangular attachment scar, c. 0.7–0.9 mm across (Fig. 3h). The seeds have an elongated micropyle (Fig. 3g), which is commonly broken or abraded in our material (Fig. 3h).

The outer cuticle of the seed is delicate and 5–8 μm thick. It lacks stomata and is composed of elongated rectangular cells that are arranged in more or less regular longitudinal files (Fig. 3i). Beneath the cuticle, the seed coat is formed by a sclerotesta, typically c. 20–40 μm thick, which appears light in our sections, and an endotesta, typically c. 12–30 μm thick, which is dark in our sections (Fig. 4e,f). The sclerotesta consists of two to three layers of isodiametric, thick-walled cells, each of which is c. 15–20 μm in diameter.

In most dispersed seeds, the outer cuticle is missing, exposing the outer surface of the sclerotesta, which is composed of more or less isodiametric cells, 20–35 μm in diameter, that are slightly elongated transversely (Fig. 3j). Each cell on the outer surface of the sclerotesta has a very thin outer periclinal wall and the cell lumen is an angular pit, 10–14 μm in diameter (Fig. 4e,f). This pit probably reflects the former presence of a cuboidal crystal (Figs 3j, 4e,f). The endotesta is composed of c. 10 layers of thick-walled cells, each with a very small lumen.

Several bisaccate pollen grains of the Alisporites-type were found on the tip of the micropyle of one dispersed seed. Similar pollen grains are commonly found adhering to the cupules and seeds (Fig. 3k). They are elliptical, with sacci that do not protrude from the outline of the grain. Pollen grains are 60–83 μm long (including sacci) with a corpus 48–64 μm in width. Sacci are crescentic, and attached laterally to the body of the grain.

Discussion

Umkomasia mongolica provides important clarification of the structure of corystosperm cupules: the flattened cupule stalk covers one of the three lateral faces of the single three-angled seed, whereas the two cupule lobes cover the other two faces (Figs 3b–e, 4b,c). This organization is identical to that of Doylea and similar to that in U. resinosa. In Doylea, the ‘three valves’ of the mature cupule comprise the flattened cupule stalk with a single vascular bundle, and two lobes that lack vascular bundles (Stockey & Rothwell, 2009). These two lobes correspond to the two cupule flaps in U. mongolica. In U. resinosa, two collateral vascular bundles enter the base of the cupule stalk and fuse distally into a single bundle. In bilobed cupules, this single vascular strand divides to supply each of the two lobes (Klavins et al., 2002).

Small cupules of U. mongolica (Figs 2g–i, S5) show clearly that the cupule stalk is both morphologically and functionally part of the seed-enclosing structure. It is responsible for the inversion of the seed and also contributes to its covering. A potentially similar situation also occurs among extant angiosperms, where structural evidence and indications from molecular developmental genetics suggest that the development of the outer integument and ovule curvature are initimately functionally connected (Endress, 2011).

Several interpretations of the homologies of the seed-bearing organs of corystosperms have been offered (Thomas, 1933; Axsmith et al., 2000; Klavins et al., 2002; Stockey & Rothwell, 2009). Thomas (1933) interpreted the entire corystosperm ovulate structure as a fertile branch system with two orders of branching, based on bracts borne near the base of the ultimate and penultimate fertile axes, especially towards the tips of the fertile branch system. The bracts were taken to indicate that the axes arising in their axils were shoots. Thomas (1933) suggested that, in his material, the seed was borne singly and terminally on a shoot, rather than on a structure derived from a leaf (sporophyll). Under this interpretation, the cupule stalks are fertile branches with an erect seed at their tip.

Studies of permineralized corystosperm seed-bearing organs (Klavins et al., 2002) clearly show that they are fertile branch systems, as suggested by Thomas, because the main axis bearing the fertile lateral units has typical cauline anatomy with a radially symmetrical vascular cylinder surrounding a pith (Fig. S6a–g). However, most interpretations regard the ultimate units of the seed-bearing organs, equivalent to the cupule stalk in our material, as foliar (Axsmith et al., 2000; Klavins et al., 2002; Stockey & Rothwell, 2009), and this has figured prominently in discussions of angiosperm origins (Doyle, 2006).

The evidence for this interpretation comes mainly from permineralized material of U. resinosa, in which each cupule stalk is supplied by two flattened, leaf-like, collateral vascular bundles, and curvature of the cupule is towards the phloem side of the vasculature within the stalk (Klavins et al., 2002; see Fig. S6a–g). Similarly, in compression fossils of U. uniramia (Fig. S6u), the cupule was inferred to be a modified leaf that bears the seed on the lower (abaxial) surface and that is reflexed downwards relative to the axis on which the seed-bearing unit is borne (Axsmith et al., 2000).

Umkomasia mongolica is very similar to both U. resinosa and U. uniramia, but the abundant material and three-dimensional preservation provide more complete information on cupule morphology. In particular, the position of the bifurcating cupule-bearing stalk in the axil of a bract raises the question as to whether the ‘axial’ interpretation of Thomas (1933) or the standard ‘foliar’ interpretation has the strongest support.

These two alternatives (Fig. 5) have significantly different implications for the homology of ovulate structures among seed plants (Table 1). For example, under the ‘foliar’ interpretation, the supposed abaxial position of the seed suggests similarities with Callistophyton and perhaps peltasperms, whereas the ‘axial’ interpretation points instead to similarities with Ginkgo (Table 1).

In our view, the possibility of a close relationship between corystosperms and Ginkgo deserves closer scrutiny (see also Meyen, 1984; Anderson & Anderson, 2003). The vasculature of teratological forms of Ginkgo, in which seeds are borne on individual stalks (Figs S6h–t, S7), is very similar to that of U. resinosa, and Thomas (1933) compared the cupule of corystosperms to the collar at the base of the seed in Ginkgo. In many young ovules of living Ginkgo, a flap of tissue develops in the groove between collar and seed (Douglas et al., 2007), which could potentially be homologous to the bilobed flap that forms part of the U. mongolica cupule. Karkenia Archangelsky, an extinct Ginkgo relative known from the Lower Permian to Lower Cretaceous (Archangelsky, 1965; Zhou, 2009), also has cone-like reproductive structures in which the micropyles of the seeds point towards the cone axis, raising the possibility that reflexed ovules may be basic in Ginkgoales (Zhou, 2009).

In addition, leaves of the early corystosperm-like plant Kannaskoppia (Kannaskoppifolia) are similar in form and vein organization to those of Ginkgo (Anderson & Anderson, 2003; Bomfleur et al., 2014), U. uniramia has a Ginkgo-like habit with distinct short shoots (Axsmith et al., 2000), and basal frond elements of Dicroidium, the predominant form of corystosperm foliage, resemble ginkgophyte leaves in their shape and venation (Bomfleur et al., 2012). Leaves of Pseudotorellia Florin, a presumed ginkgophyte, are also common in the low-diversity Tevshiin Govi samples that yield U. mongolica (G. Shi et al., pers. obs.), although it remains to be determined whether they are part of the same plant as U. mongolica cupules.

The ‘foliar’ interpretation of the cupule hinges mainly on the observation that the cupule stalk in Doylea and in U. resinosa is supplied by collateral, leaf-like, vascular bundles. However, collateral bundles are not an infallible indicator of foliar structure (Kaplan, 2001). For example, in Ginkgo, the stalks that bear each pair of ovules usually have two pairs of collateral vascular bundles, but the most recent interpretation, based on a thorough study of morphology, anatomy and development, concludes that ‘the ovulate stalk in Ginkgo is a simple axial and axillary [dichotomizing] structure’ (Douglas et al., 2007). Similarly, collateral vascular bundles occur in the ovuliferous scales of many conifers, which are conventionally interpreted as modified shoots (Florin, 1951). An extreme case occurs in the ovulate structures of Podocarpaceae (Fig. S8), in which the ovules are borne on a reflexed stalk supplied by a pair of collateral vascular bundles (Sinnott, 1913).

Important for the resolution of whether the structure bearing the seed in corystosperms is ‘foliar’ or ‘axial’ is the precise position of seed attachment. Numerous specimens of U. mongolica provide the clearest insight so far into the attachment of the seed in the corystosperm cupules. Small cupules that were presumably preserved at an early stage in development (Figs 2g–k, S5) show that the seed is borne at, or very close to, the apex of the reflexed portion of the cupule stalk, rather than on its abaxial surface. The point of attachment is far out along the cupule stalk (Figs 2f, S4) and it is the curvature of the cupule stalk that results in the inversion of the seed.

That the cupule stalk may be a shoot is also supported by the structure of the two lateral flaps on either side of the seed. These are distinct from the cupule stalk. They are thinner in texture with a less prominent inner cortical zone and scattered stomata on the outer cuticle. These lateral flaps may be foliar structures borne on the reflexed portion of the cupule stalk near the point of attachment of the seed. Thomas (1933) compared them to the envelope that encloses the seed in Gnetales and Bennettitales (see also Friis et al., 2007).

We conclude that the structure of the vascular bundle alone is insufficient to determine whether the cupule stalk of corystosperms is ‘foliar’ or ‘axial’ in nature, and that the evidence from small cupules of U. mongolica is more consistent with the ‘axial’ than the ‘foliar’ interpretation. The cupule stalk of corystosperms, not just the branch on which the cupule is borne, is as likely to be a fertile shoot bearing a terminal seed, as a fertile leaf.

Irrespective of whether the ‘foliar’ vs ‘axial’ interpretation is correct, new information from U. mongolica is important for comparisons with the cupules of Caytonia and Petriellaea, which have figured prominently in discussions of the origin of the angiosperm second integument (Doyle, 2006). Doyle (2006) regards the cupules of corystosperms, and those of Caytonia and Petriellaea, as fundamentally different (Table 1), based on the interpretation that the seed in the corystosperm cupule is borne on the abaxial surface of a foliar structure, whereas, in Caytonia and Petriellaea, the seeds are borne adaxially.

Under this interpretation, the cupule of corystosperms arises by ‘downwards’ folding of a foliar structure away from the subtending axis (towards the phloem side of its collateral vascular bundle), whereas the cupule of Caytonia and Petriellaea arises by ‘upwards’ folding of a foliar structure towards the subtending axis (towards the xylem side of its collateral vascular bundle). The new material of U. mongolica complicates this interpretation because curvature of the cupules is ‘upwards’ (adaxial) towards the inferred main axis of the reproductive structure (Fig. 5), as is assumed for Caytonia and as is inferred for Petriellaea based on its vasculature (cupule curved towards the xylem side of its vascular bundle). However, at the same time, the cupule of U. mongolica could be interpreted as folded downwards (abaxial) with respect to the unbifurcated portion of the cupule stalk (cupule curved towards the phloem side of its vascular bundle). Similar considerations apply to the seeds of conifers (Fig. S8), which are generally interpreted as attached to the adaxial surface of the ovuliferous scale, but that are ‘abaxial’ (using the criterion of Doyle, 2006) with respect to the orientation of the phloem in the scale vasculature.

We conclude that referring to the cupules of corystosperms, Caytonia and Petriellaea as abaxially vs adaxially folded is potentially confusing, and that the cupules of corystosperms, Caytonia and Petriellaea are less distinct than has previously been supposed (Doyle, 2006). The direction of folding may have less significance for phylogenetics than for pollination biology. A downward orientation of the micropyle ensures effective flotation of saccate pollen (Leslie, 2008). This is important because, if the cupules of corystosperms are as relevant as those of Caytonia for understanding the origin of the angiosperm second integument (Doyle, 2006), then a bract, like that subtending the seed-bearing units in U. mongolica, may have played a role in the enclosure of the bitegmic ovule and the origin of the carpel.

Current hypotheses of seed plant relationships place great weight on whether the seeds in different groups are borne ‘terminally on a shoot’, or adaxially or abaxially on modified leaves (Table 1; Doyle, 2006). For corystosperms, the axial interpretation suggested by U. mongolica points to a possible relationship to Ginkgo, whereas the reflexed cupule may suggest a relationship to angiosperms (Table 1). These two ideas may not be incompatible. If the fundamental phylogenetic division between angiosperms and all extant gymnosperms is correct, as suggested by analyses of molecular data (Wickett et al., 2014), corystosperms, Caytonia and perhaps other extinct groups may be close to the point at which the two extant clades diverged.

Author contributions

P.S.H., M.T. and P.R.C. designed the research. P.S.H., P.R.C., M.T., A.B.L. and N.I. collected the palaeobotanical samples. G.S., A.B.L. and F.H. prepared, described and illustrated the fossil material. G.S. and P.R.C. wrote the manuscript, in discussion with A.B.L., P.S.H. and P.K.