Interpretation of the fossil and its systematic position
The vegetative and reproductive features of the new fossil exclude a relationship to any other group than the Gnetales. Ovules surrounded by one or two seed envelopes and with the integument extended into a long exposed micropylar tube are known only for extant and extinct Gnetales, for at least some members of the Bennettitales [14, 18, 31] and for a number of extinct plants thought to be closely related to Gnetales: i.e., the Erdtmanithecales [33] and a number of unassigned taxa based on dispersed seeds, such as Buarcospermum, Lignierispermum, Lobospermum, Rugonella [14]Raunsgaardispermum [18].
A close relationship with the Bennettitales can be ruled out based on vegetative as well as reproductive characters. The Bennettitales have ovules borne on long stalks, like Siphonospermum simplex, but the Bennettitales are otherwise distinct from Siphonospermum in having ovules borne in densely crowded heads and interspersed with fleshy interseminal scales. The Bennettitales are further distinguished by their typically larger and compound leaves. The Erdtmanithecales are known only from dispersed reproductive structures (seeds, male organs, pollen) [see summary e.g., in [13]]. The seeds of Erdtmanithecales are distinguished by their three-valved seed envelope, but further comparison with Erdtmanithecales is not possible.
Higher level relationship of Siphonospermum simplex (its systematic position within the Gnetales) is more difficult to establish due to the restricted amount of preserved information. Its morphology differs substantially from that of other gnetalean fossils from the Yixian Formation, which have compound cones comprising cone bracts and seeds [7, 11, 12]. They are generally interpreted as close relatives of Ephedra. In contrast, the reproductive units of Siphonospermum consist only of single ovules (with surrounding envelope[s]); there are no supporting bracts.
The long and conspicuous micropylar tubes of Siphonospermum are well preserved and have not been bent or curled during fossilization. Their walls are prominent (Figure 2b) and one tube (Figure 2a, left ovule) is broken at the apical level of the seed envelope. In Ephedra, Gnetum and Welwitschia, the inner epidermis of the micropylar tube is strongly thickened, probably lignified [[34], and pers. obs.], and the same appears to hold for Siphonospermum. Conceivably, the thickening and support for the micropylar tube have been important in the Gnetales in order to withstand visits from pollinators. Nothing is known about the pollination biology of Siphonospermum, but in extant Gnetales, pollination drops secreted via the micropylar tube serve as reward for pollinators (mainly Dipterans and Hymenopterans) [35–38]. Abiotic pollination is probably also important, at least in Ephedra [38, 39].
Vegetative characters of Siphonospermum simplex are similar to extant Ephedra, with striate, erect stems and linear leaves opposite at nodes and with joined bases. However, these characters are not unique to Ephedra, but are general (ancestral) characters of the Gnetales, at least partly present also in Gnetum, Welwitschia and other gnetalean fossils [e.g., [4, 12]]. The venation of the leaves of Siphonospermum is difficult to assess with certainty due to poor preservation and slightly folded leaves. There are three primary veins and perhaps also second order venation. The latter feature is interesting because it is a derived feature within the Gnetales, which characterizes the Gnetum-Welwitschia clade. Extant Ephedra species have two primary veins in leaves and cone bracts and so have most Early Cretaceous ephedroids such as Liaoxia [12, 40]. One ephedroid fossil, Liaoxia robusta [12], has three or four primary veins but second order venation is unknown in fossil ephedroids as well as in extant Ephedra.
Siphonospermum has several features in common with Gnetum. The "reticulate" venation of Gnetum leaves develops from successive dichotomies in 5-10 parallel veins located in the centre of the leaf [41] and basically, the venation pattern of Siphonospermum is not different from that seen in Gnetum. The reproductive units of Siphonospermum have a similar shape as those of Gnetum and are (like in Gnetum) completely exposed with no remnants of cone bracts. Weak impressions of a structure at the lower left of the reproductive unit with the broken micropylar tube (Figure 2a, arrow 4) are not remains of a bract but reflect the original position of the ovule. This reproductive unit appears to have been detached and slightly displaced.
In Gnetum the nucellus is surrounded by three structures, an integument that forms the extending micropylar tube, a sclerenchymatous inner seed envelope and a second seed envelope, composed of parenchymatous cells and sclereids [42, 43]. In Siphonospermum, the integument is surrounded by a persistent coalified tissue and its preservation mode indicates that it is sclerenchymatous. In the acuminate apex region, there are weak impressions of an additional tissue. There are no remains of sclerenchymatous tissue in this zone; it has obviously only left an imprint without cellular remains, which suggests soft tissue.
Further, the numerous supportive strands in the sclerenchymatous zone form a pattern with dichotomies and anastomoses, which is very similar to that formed by vascular bundles in the "endotesta" of Gnetum [44]. In the Gnetales, vascular bundles are absent in the integument but present in the seed envelopes [24], which supports our interpretation that the supportive strands of Siphonospermum are originally derived from the seed envelope.
Conflicting information and uncertain evolutionary relationships
An evolutionary origin of the spikes of Gnetum, with sessile reproductive units arranged in whorls, from the pedunculate and solitary units of Siphonospermum, could be hypothesized by introducing hypothetical steps. However, there are several noteworthy differences between Siphonospermum and Gnetum, which make evolutionary relationships uncertain. First, the linear leaves and arrangements of the reproductive units are different from those of Gnetum (but likely plesiomorphic features in the Gnetales). Second, in Gnetum the micropylar tube is closed by radially expanding cells of the integument [45], whereas the micropylar tubes of Ephedra and Welwitschia are hollow [28, 42] (and pers. obs.). In Siphonospermum, the micropylar tube is isodiametric and there are no indications of a closure tissue. Third, it is not possible to say if the outermost tissue of the reproductive units of Siphonospermum represents an additional seed envelope like in Gnetum [24], or an outer parenchymatous zone of a single seed envelope like in most species of Ephedra [34].
Thus, several seemingly "ephedroid" features of Siphonospermum may instead be plesiomorphic in the Gnetales, and retained in Siphonospermum. The general structure of the reproductive units indicates a sister relationship with Gnetum. If so, the closure tissue in the micropylar tube is a unique feature for Gnetum [and at least some species of Bennettitales [45]], not shared by Siphonospermum (or is missing in Siphonospermum because of an early stage of development). Alternatively, Siphonospermum could be sister to the Gnetum-Welwitschia clade; further studies of the diversity in the Gnetales and related taxa, and a better understanding of character evolution, are needed to elucidate higher level relationships among living and fossil species and the exact phylogenetic position of Siphonospermum simplex.
