Our reinvestigation of the original specimens of Almargemia incrassata and A. dentata revealed remarkable differences between the two species. These differences necessitated the transfer of A. incrassata to a different genus, viz. Eobowenia gen. nov.
Comparison of Almargemia dentata and Eobowenia incrassata
The leaves from the Lower Cretaceous of Portugal later referred to by Florin  as Almargemia dentata were first described by Heer  as Ctenidium dentatum Heer and C. integerrimum Heer. In the generic diagnosis, Heer  distinguished Ctenidium from Ptilophyllum and Ptilozamites by the decurrent leaf bases and from Ctenis by the absence of vein anastomoses. Florin  investigated the epidermal anatomy of the specimens described by Heer  in detail and, as a consequence, transferred both species in the new combination Almargemia dentata, correctly recognizing that the genus name selected by Heer  was pre-occupied by a genus of extant mosses. The main diagnostic epidermal characters of Almargemia according to Florin  were the predominantly incompletely amphicyclic haplocheilic stomata, arranged irregularly in stomatal bands running between the veins on the abaxial surface of the leaflets, the sunken guard cells and the presence of both weakly and strongly cutinised pavement cells. Macromorphologically, the diagnostic characters included slightly contracted leaflet bases, parallel (rarely dichotomizing) venation and the presence of lobe-like teeth (Fig. 1 a, b, e).
When Archangelsky  described Eobowenia incrassata (as Almargemia incrassata), he decided to assign such specimens to Almargemia on the base of the serrate margin of the leaves (erroneously identified as ‘dentate’ by Archangelsky ) and the differently thickened cutinization of the epidermal cells. However, most of the other diagnostic characters of A. dentata are absent in E. incrassata (Table 1). The stomatal characters are strikingly different (Fig. 4), with E. incrassata having guard cells at the same level of the epidermal cells, monocyclic stomatal complexes and longitudinally oriented guard cells, while Almargemia dentata has stomata sunken below the epidermal level, as in most extant Zamiaceae, incompletely amphicyclic stomatal complexes and randomly oriented guard cells. To use only the differentially thickened cutinisation of the epidermal cells is, in our opinion, too weak a character to assign the fossils in question (viz. Eobowenia incrassata) to Almargemia, because of their common presence in most members of extant Zamiaceae [34, 41, 54]. Moreover, the dentation of the margin in the two species is quite different, with E. incrassata having relatively small, acute teeth and A. dentata having larger, lobe-like teeth. For these reasons, we reconsider the allocation made by Archangelsky  by erecting a new genus because a new generic definition is needed for this fossil taxon.
Comparison of Eobowenia incrassata with other fossil cycadophytes
The leaves of Eobowenia incrassata are easily distinguishable from all other cycadalean leaf taxa described from the Baquero Group (i.e. Ticoa, Mesosingeria, Mesodescolea, Sueria; [40, 43]) by their leaf shape and epidermal anatomy (see [40, 43]). Among other Mesozoic cycadophyte leaves with parallel venation, E. incrassata differs from Pseudoctenis  by its basally converging veins, the serrate margin and by epidermal characters (i.e. guard cells at the same level as the epidermal cells, darker-staining pavement cells, longitudinally elongated pavement cells), and from Ctenis  by the absence of vein anastomoses as well as the very different cuticle. It differs from segmented Nilssonia leaves  by the lateral attachment of the leaflets and the anatomy of the cuticle and from Encephalartites by the leaf base that is contracted only on the acroscopic side, and by the oblong leaflets. Eobowenia incrassata is distinguished from any segmented bennettitalean leaf by the haplocheilic architecture of the stomata in contrast to the syndetocheilic architecture characterising bennettitalean leaves [19, 14].
A similar combination of differentially thickened epidermal cells, monocyclic stomatal complexes and guard cells at the same level with the aperture is present in some species assigned to the tentative pteridosperm genus Stenopteris. Monocyclic stomatal complexes with differentiated subsidiaries are present in S. nana T.M.Harris from the Bajocian of Yorkshire , but the overall morphology of the leaf easily distinguishes this species from Eobowenia incrassata. Another interesting species is S. cyclostoma K.Saiki, T.Kimura et J.Horiuchi, from the Lower Cretaceous Choshi Group of Japan . The cuticle of this species presents many similarities with E. incrassata including the rows of dark staining cells , but presents a very dissimilar morphology of the leaf. However, there are differences even at the cuticular level, with S. cyclostoma being clearly amphistomatic and having an external vestibulum. Moreover, we were not able to identify the peculiar perforations of the substomatal complexes in the illustrations of Saiki et al. . The cycad-like characters of S. cyclostoma are definitely interesting, but a more thorough discussion would include a revision of the morphology of the entire genus, and falls outside the scope of the present investigation.
Comparison of Eobowenia and Bowenia
Our reinvestigation pinpoints numerous similarities between Eobowenia incrassata and the extant cycad genus Bowenia (Table 1). Among the most interesting characters are the flush guard cells, which clearly separate Eobowenia from Almargemia dentata as well as from all Zamiaceae and Cycadaceae sensu Stevenson  (Fig 4, Additional file 3: Figure S2; Additional file 4: Figure S3; Additional file 5: Figure S4; Additional file 6: Figure S5; Additional file 7: Figure S6). The cuticle of the guard cells also presents cuticular thickenings both on the dorsal and ventral surfaces, and single cuticular ridges running parallel to the dorsal wall of the guard cells (Fig. 3 c, d; Fig. 4 a, b), the monocyclic stomatal complexes (Fig. 3 a, c), and the presence of substomatal cell complexes with secondarily thickened walls (Fig. 3 e). The first set of characters is present among extant cycads in Bowenia and Stangeria, with some differences between the two genera . Monocyclic stomatal complexes with stomata at the same level with the epidermis are restricted in extant Zamiaceae to Bowenia [41, 57, 58] (Additional file 6: Fig. S4). The perforations associated with some of the stomatal complexes in Eobowenia incrassata presents some striking similarities to the substomatal complex in Bowenia, which present secondarily thickened cell walls. This structure was interpreted by Greguss  as a perforation of the subsidiary cells, not dissimilar to the condition present in Cycas , where all epidermal cells present perforations of the inner periclinal wall. The structures in Eobowenia incrassata more closely resemble the structures in Bowenia (which also occur but are less developed in some species of Encephalartos and Macrozamia; see Additional file 7: Figure S6) in being mostly restricted to the substomatal complexes (Fig. 3 e, f). The main difference between the epidermis/cuticles of Eobowenia incrassata and Bowenia is the presence of files of short cells with thickened cuticle in the former. This character has been compared to the state present in Ceratozamia  by Kvaček , where files of short, dark-staining cells are present on both surfaces of the leaflets. However, the slightly concave and sometimes wavy anticlinal cell walls of the dark-staining cells is closer to the cuticle of the cell files present in Dioon (Additional file 3: Figure S2 C, D) . Darker staining cells are present in Bowenia, but they are organised as single or small groups of cells, commonly of the same length as the other epidermal cells (Fig. 3 b, d).
Eobowenia and Bowenia not only share significant and interesting characters in epidermal and cuticular anatomy, but also share commonalities at the macromorphological level, one being the serrate leaflet margin, which occurs in many extant cycads, such as some species of Zamia and Stangeria (Fig. 6). Marginal teeth are also present in a few species of Encephalartos (Fig. 6 b). In Bowenia, a serrate margin is present in both B. serrulata (Fig. 6 a) and individuals of B. spectabilis Hook. ex Hook.f. growing in more open environments , as well as in the fossil B. eocenica R.S.Hill  and other fossil members of the genus that have not yet been formally described . The thickened, almost glandular-like aspect of the teeth in Eobowenia is compatible with the situation present in extant (Fig. 6) as well as fossil cycads (i.e. Restrepophyllum, ). In extant cycads, the thickened aspect of the tooth is given by a concentration of marginal fibres. The teeth in Almargemia dentata, on the other hand (Fig. 1) remind more closely of the lobe-like teeth present in some species of Encephalartos (Fig. 6 e). The basally converging veins in the leaflets are another character shared between Eobowenia incrassata and Bowenia. This character is also present in members of Zamia, but Zamia has articulated leaflets in contrast to the decurrent insertion of the leaflets in Bowenia and Eobowenia. The striations on the leaflets of Eobowenia remind of similar striations present in fossil representatives of Bowenia (described by  as “veinlets”), which correspond to interspersed fibres in the leaflets of extant Bowenia.
However, Bowenia and Eobowenia also differ in details that are mainly restricted to the morphology of their leaflets. All extant and extinct species of Bowenia are characterised by dichotomous venation, with veins ending at the margin. In the species with serrate leaflet margin, the veins commonly end in the teeth. In Eobowenia, the details of the venation are not clear from the material available, even if some dichotomies are potentially present on the specimen (Fig. 2 a, b).
Another striking difference between Eobowenia and Bowenia lies in the size of the leaflets. The two extant species of Bowenia have leaflets with length varying from 9 to 14 cm , which are markedly larger than the 0.6–1.0 cm long leaflets of Eobowenia. However, fossil leaves assigned to Bowenia commonly have rather short leaflets (e.g. B. eocenica and B. papillosa R.S.Hill: 3–4 cm; ). Extant Bowenia is characterised by bipinnate leaves, which are an autapomorphy of the genus, while the fragmentary nature of the leaflets of E. incrassata does not allow us to evaluate the character in this taxon.
Despite the striking similarities presented between Eobowenia incrassata and Bowenia, we refrain from assigning the specimens to Bowenia, mainly in the light of the differences outlined above, and considering the institution of the new genus Eobowenia to represent the best solution for the accommodation of this fossil taxon.
On the other hand, the differences and uncertainties in macromorphological characters do not preclude a relationship between the two genera. Regarding, for example, leaflet size, size variation is not uncommon among extant and fossil Cycadales. For example, in extant Zamia, leaflet length can vary from 1 to 8 cm in Z. pygmaea Sims  to 30–60 cm in Z. wallisii A.Braun. In the fossil genus Ctenis, leaflet length can vary from 1.5–3.5 cm in C. nathorstii Moeller  to 15–20 cm in C. kaneharai Yokoyama .
Phylogenetic evidence for the placement of Eobowenia and Almargemia
Our investigation is not the first to hypothesise a link between Eobowenia (Almargemia) incrassata and Bowenia. In their phylogenetic analysis of extant and fossil cycads, Martinez et al.  retrieved a maximum parsimony tree with Almargemia (predominantly coded after A. incrassata) as sister to Bowenia plus Stangeria and Mesodescolea. However, such relationship does not receive any support from the bootstrap analysis, and it is not retrieved in other analyses of morphology, which consider Almargemia predominantly coded for A. incrassata [61, 62].
Using the topology from Salas-Leiva et al.  and the modified matrix from Martinez et al.  as a backbone, the best placement for Eobowenia is as sister to Bowenia (Fig. 5 a). However, alternative placements are possible at the cost of only one or two steps more. This could be due to the low number of characters coded for Eobowenia (21 characters out of 89), and the few informative epidermal characters linking the different clades of the Zamiaceae. If we consider the placements which are only one step longer, these placements imply that the unique characters of the stomatal complex of Eobowenia (guard cells at level with epidermis and monocyclic stomatal complexes) either evolved independently in this taxon and in Bowenia (if Eobowenia is placed as sister to the Ceratozaminae or the Encephalartinae), or represent a potentially plesiomorphic status of all Zamiaceae except Dioon. This would imply that all the similarities of the stomatal complexes of the Encephalartinae and Dioon could represent parallel evolution of sunken, protected guard cells.
Our phylogenetic analyses based on the Martinez et al.  matrix retrieve a relationship between Eobowenia and Bowenia in both the MP and BI analyses of the morphological data, with Eobowenia being sister to the Stangeriaceae sensu Stevenson  but such relationships only receive low support in the BI analysis. In the MP analysis this is partially due to the uncertainties surrounding the relationships between Bowenia and Stangeria and many other fossil taxa with peculiar character combinations, such as Kurtziana, Mesosingeria, Sueria and Pseudoctenis. When information from the molecular analysis of Salas-Leiva et al.  is added, resulting in the breakup of the Stangeriaceae, Eobowenia is preferentially retrieved as sister to Bowenia instead of Stangeria. The characters linking Eobowenia and Bowenia in these topologies regard the unique structure of the stomatal apparatus, which combines the flush guard cells with the lack of encircling cells. The combined analysis using a Bayesian framework retrieves the strongest support for the sister relationship of the two genera. This is in our knowledge the first attempt of integrating morphology and molecular data in a matrix that includes fossil taxa in the Cycadales, and shows the potential of this practice to resolve some of the uncertainties in the relationships between extant and fossil cycads.
The placement of Almargemia, on the other hand, is much more uncertain, with no clear placement in any of the analyses. However, a sister relationship between Eobowenia and Almargemia is never retrieved.
Our phylogenetic analyses show that the link between Eobowenia and Bowenia is the best hypothesis to explain the relationship between the fossil taxon and the diversity of the Cycadales, even when adopting a conservative approach to its macromorphological character coding. Such phylogenetic evidence, which is lacking for many fossil cycads that have been linked with extant groups, such as Restrepophyllum  and Austrozamia , as well as for the many fossil leaves assigned to extant genera [21–24, 26–28], make Eobowenia a reliably placed cycad fossil foliage.
Such a placement is also compatible with at least some of the inferred age for the divergence of Bowenia based on molecular dating. The age of the deposition of the Anfiteatro de Ticó Formation, where Eobowenia is found, is very well constrained to 118.23 ± 0.09 Ma  or 116.85 ± 0.26 Ma  representing an Aptian (Lower Cretaceous) age, which is compatible with the ages inferred for the stem of Bowenia by Nagalingum et al.  using a relaxed log-normal clock and by Condamine et al.  using the favoured birth-death prior with both the calibration implemented, but is older than the dates retrieved by Salas-Leiva et al.  (Table 2). This probable early divergence of the genus Bowenia is, however, compatible with the phylogenetic placement retrieved by the multilocus analysis of Salas-Leiva et al. , which sees Bowenia as sister to all the other Zamiaceae apart from Dioon. A Cretaceous stem history of Bowenia/Eobowenia is also compatible with the presence of cuticle indistinguishable from modern Bowenia in the Upper Cretaceous of Central Australia .
Eobowenia and the biogeography of Bowenia
The occurrence of a potential sister of Bowenia in the Early Cretaceous of Patagonia helps to strengthen some of the hypotheses around the biogeography of Bowenia. Until now, the phylogenetic isolation of Bowenia, as well as the presence of fossil records limited to Australia, had complicated the resolution of the biogeography of the genus. Indeed, Salas-Leiva et al.  retrieved two different results in their analysis: using S-DIVA, they retrieved an ancestral area including Australia, Africa and Mexico for the stem of Bowenia, while their DEC analysis hypothesises a model of stasis in Australia. The presence of Eobowenia in Patagonia during a period of connectivity between southern America and Australasia supports the hypothesis of a Gondwanan distribution for the stem of the group, with subsequent extinction shaping the current Australian endemic distribution. Bowenia would indeed represent yet another case of eastern survival . Even if we know that some cycads persisted in southern South America until the Palaeocene , the identification of the precise timing of the extinction of Eobowenia in South America is hindered by the potential rarity of this fossil leaf type in the record. However, it is clear that this fossil represents another important clue to the biogeography of Gondwana coming from Patagonia .