A fossil species of the enigmatic early polypod fern genus Cystodium (Cystodiaceae) in Cretaceous amber from Myanmar

The monospecific fern genus Cystodium (Cystodiaceae; Polypodiales) occurs exclusively in the tropical forests of the Malay Archipelago, the Admiralty Islands, the Louisiade Archipelago, and the Solomon Islands. Divergence time estimates suggest that the genus originated in the Mesozoic; however, fossil evidence to validate this suggestion has been lacking. Amber from Myanmar (Burmese amber) is an important source of new information on the diversity of vascular cryptogams in the Cretaceous. This paper describes the fossil taxon Cystodium sorbifolioides nov. sp. based on a fragment of a fertile leaf preserved in Burmese amber that represents the first fossil evidence of the family Cystodiaceae. Cystodium sorbifolioides is used to obtain a minimum age estimate for the Cystodiaceae and the closely related, monogeneric Lonchitidaceae and Lindsaeaceae. The fossil strengthens the hypothesis that the forest ecosystems of Malesia and Melanesia represent refugia for many tropical plant lineages that originated in the Cretaceous.

clearly argued against affinities of C. sorbifolium to the Dicksoniaceae, including the grooved upper surface of the petiole, cylindrical sorus receptacles, and the near-vertical position of the sporangium annulus 1,14,15 . Based on these structural features, the genus was excluded from the Dicksoniaceae by Croft 15 and transferred to a new family, Cystodiaceae J. R. Croft. Moreover, this author 15 noted similarities of Cystodium to the Dennstaedtiaceae s.l. with regard to soral structure. However, this notion was not accepted 2 . Application of DNA sequence variation eventually provided evidence for close affinities of Cystodium to basal polypod lineages such as the Lindsaeaceae C. Presl ex M. R. Schomb. [16][17][18] . The most recent classification of ferns places Cystodium in a separate family, Cystodiaceae, nested in the suborder Lindsaeineae together with the families Lindsaeaceae and Lonchitidaceae Doweld 19 .
Divergence time estimates have variously been used to suggest an origin of Cystodium in the Early Cretaceous (113.0-153.0 Ma) 12 , Jurassic (167.1 Ma) 11 or even Triassic (180-220.5 Ma) 13 . However, the Cystodiaceae represent a ghost lineage that lacks a documented fossil record, which could be used to corroborate its presence in the Mesozoic. Over the last years, fossils of fern foliage [20][21][22][23] exquisitely preserved in mid-Cretaceous amber from Myanmar [24][25][26][27] have become increasingly important as a source of new information on Cretaceous fern diversity and have contributed greatly to the resolution of several striking discrepancies between the age estimates obtained by molecular dating and the minimum ages of lineages provided by the fossil record, for instance with regard to the crown group Dennstaedtiaceae Lotsy 21 . Moreover, the recent discovery of a fossil assignable to the Lindsaeaceae in Burmese amber 22 provided evidence of the presence of early diverging polypod ferns of the suborder Lindsaeineae in the Cretaceous. Although this find is in agreement with divergence time estimates in molecular phylogenetic frameworks [10][11][12][13]21 , the fossil has not yet been used to calibrate a dated phylogenetic hypothesis for its putative extant relatives.
In this study, we describe the fossil taxon Cystodium sorbifolioides nov. sp. based on a fragment of a fertile leaf preserved in Burmese amber. This discovery is important because it represents the first piece of macrofossil evidence of the existence of the Cystodiaceae in the mid-Cretaceous. Moreover, it suggests the coexistence in the Cretaceous forests of Myanmar of fossil members of the families Cystodiaceae and Lindsaeaceae, both within the suborder Lindsaeineae. The newly described fossil is used to obtain a minimum age estimate for the fern family Cystodiaceae, as well as for the closely related Lonchitidaceae and Lindsaeaceae. This estimate is compared with results obtained from estimates in which either only the unnamed Lindsaeaceae fossil has been used 22 or both fossils are used together to calibrate a phylogenetic hypothesis for the extant members of the suborder Lindsaeineae and the closely related Saccolomatineae. Finally, the new fossil validates the hypothesis that the forest ecosystems of Malesia and Melanesia represent refugia for many tropical plant lineages that originated in the Cretaceous.

Results
Systematic Paleontology. Classification Etymology. The specific epithet refers to the striking morphological resemblance of the fossil to foliage of the extant Cystodium sorbifolium.
Description. Three portions of blades of this fern occur in a single block of Burmese amber (4.2 × 1.9 cm) that has subsequently been cut into three pieces to allow better photographic documentation of the individual inclusions. The blade of this fern was at least pinnate-pinnatifid (probably 2-pinnate). Pinnae (or pinnules) are up to 3.8 cm long and 7 mm wide. Axes are adaxially grooved. Individual pinnule segments are up to 5 mm long and 7 mm wide, with the margins mostly entire or 2-3 blunt-dentate at apex. The venation is free and catadromous, with the secondary veins forking once (rarely twice) in their course to the margin. Pinnae (or pinnules) are hypostomatic. The epidermis of the adaxial and abaxial sides is composed of elongate cells, 80-200 × 15-60 μm in size, that are regularly arranged parallel to the main vein and characterized by weakly undulate anticlinal walls. Pinnae (or pinnules) and axes appear to be glabrous (based on the areas where the epidermis is preserved). Stomata are 40-55 long, 20-30 μm wide, and anomocytic. Sori are apical, marginal, Divergence time estimates. The divergence time estimates obtained by applying three different approaches are suggestive of a diversification of the Lindsaeineae crown group some time between the Early Jurassic and Late Cretaceous, and Saccolomatineae between the Late Cretaceous and Paleogene-Miocene (Table 1, Fig. 3). Narrower intervals were acquired when the Cystodium fossil described in this study, as well as the unnamed fossils of Lindsaeaceae mentioned in the Introduction section of this paper, were included in the analysis (Fig. 3), while the youngest ages were obtained by including only the new Cystodium fossil.

Discussion
Systematic position. Members in the order Polypodiales, which comprises >80% of extant fern diversity, are characterized by sporangia with a vertical or near-vertical ring of annulus cells 19,28,29 . The occurrence of sporangia with a (near-) vertical annulus in the fossil Cystodium sorbifolioides is therefore indicative of systematic affinities of this fern to the Polypodiales. Precisely as in the extant Cystodium sorbifolium, however, annulus orientation in the fossil C. sorbifolioides does not entirely coincide with that seen in typical representatives of Polypodiales. Tetrahedral trilete spores are common in all the six families of non-eupolypod Polypodiales (i.e. Saccolomataceae Doweld, Cystodiaceae, Lonchitidaceae, Lindsaeaceae, Pteridaceae E. D. M. Kirchn. and Dennstaedtiaceae [28][29][30]. Within the Saccolomataceae, only a single genus, Saccoloma Kaulf., is currently recognized 19 . This taxon differs from the fossil described in this study in that the blade is anadromously dissected 31 , the acroscopic sides of all divisions are larger than the basiscopic sides, and the ultimate segments are asymmetrical at base 32 . Members of Lindsaeaceae producing trilete spores (Lindsaea Dryand. p.p., Nesolindsaea Lehtonen & Christenh., Odontosoria Fée and Sphenomeris Maxon) also are distinct from the fossil in that they possess an anadromously dissected lamina 33 . Moreover, the sori typically occur on the uniting ends of several veins (Lindsaea p.p. and Nesolindsaea). Although Odontosoria and Sphenomeris generally share with the fossil the formation of sori on single veins at segment apices (one per segment), they differ from the latter in having strongly anadromous blades with the ultimate divisions usually connected to each other by a narrow wing of tissue 33 . Ferns in the families Lonchitidaceae and Pteridaceae with marginal or intramarginal sori are distinguished from the fossil by having elongate sori that are protected by the reflexed pinnule margin, rather than a true indusium 28,31,34 .
The only remaining repository for the fossil within Polypodiales is the Cystodiaceae, a monogeneric taxon comprised of the monotypic genus Cystodium 15,16,19 . The sole extant species in this genus, C. sorbifolium, shares with the fossil several key morphological features (Fig. 4), including (1) pinna axes with a prominent adaxial groove; (2) catadromous lamina division; (3) fertile pinnules with pronounced teeth protecting the sori (Figs 1d, 4a,b);  Table 1. Comparison of divergence time estimates of the polypod fern suborders Saccolomatineae and Lindsaeineae obtained from the following approaches: (1) assignment of the fossil Cystodium sorbifolioides to the split between Cystodiaceae and Lonchitidaceae-Lindsaeaceae, (2) assignment of the unnamed Lindsaeaceae fossil 22 as minimum age of the split between Lonchitidaceae and Lindsaeaceae and (3) assignment of both Cystodium sorbifolioides and Lindsaeaceae fossils. Published estimates [11][12][13] . In square brackets values of 95% highest posterior density intervals. All values given in mybp. Approach (3) shown in Fig. 3. (4) pinnule veins that are pinnately arranged, free, simple or forked; (5) submarginal sori positioned terminal on the veins; (6) a short receptacle that is slightly raised, round in cross-section, and protected by a reflexed lobe of the lamina (outer or false indusium); (7) a small and delicate, inner indusium that is attached at the base of the receptacle (Figs 1e,f, 4b,d); (8) sporangia with an annulus that is slightly oblique to almost longitudinal and interrupted by the pedicel (Fig. 1g); (9) numerous simple, multiseptate, non-glandular paraphyses interspersed with the sporangia (Figs 1f, 4b,d); and (10) trilete spores (Fig. 1h) that are globose and possess a low-sculptured perine 1,2,15,30,38 . The epidermis composed of polygonal cells seen in the outer indusium of Cystodium sorbifolium is yet another feature that the extant taxon shares with the fossil (Figs 1d,e, 4e). Taken together, the unique complement of morphological correspondences between the extant C. sorbifolium and the fossil renders Cystodium the only reasonable repository for the fossil. The fossil is largely identical morphologically to the extant C. sorbifolium, but there are also several minor but decisive structural differences that prompted us to introduce a new species, Cystodium sorbifolioides, for the fossil, rather than simply assigning the specimen to the extant C. sorbifolium. Cystodium sorbifolium is characterized by hypostomatic leaves, and an adaxial epidermis composed of elongate cells with weakly undulate anticlinal walls that are regularly arranged parallel to the main vein, precisely as seen in the fossil (Figs 2a; Fig. 4 in 14 ). On the other hand, the abaxial epidermis of C. sorbifolium differs markedly from that of the fossil because it is constructed of cells that usually are as long as wide and have prominently undulating anticlinal walls ( Fig. 4f; Fig. 5 in 14 ). Moreover, polocytic stomata, which are a typical feature of C. sorbifolium (Fig. 4f), have not been observed in the fossil (Fig. 2b,d). However, the preserved fragments of abaxial epidermis of C. sorbifolioides are small, and we therefore cannot rule out the possibility that polocytic stomata occurred somewhere in the fossil. Unfortunately, comparison of the fossil spore morphology and ornamentation with the granular-striate pattern seen in the spores of C. sorbifolium 30,38 is impossible because all fossil spores are ill preserved and probably have been altered during embedding and fossilization in the amber matrix.
Discovery of a fossil representative of the genus Cystodium enables us to obtain a minimum age estimate for the Cystodiaceae and their relatives, the monogeneric Lonchitidaceae and the Lindsaeaceae. Lindsaeaceae fossils that have previously been described in the Burmese amber 22 and provided support to the hypothesis that this lineage of ferns diversified before the Late Cretaceous. This result concurs with the age estimates obtained in previous studies for these lineages [11][12][13] . However, our analyses are not designed to obtain a maximum age estimate for the diversification of these lineages. In this context, it is important to note that the position of the Saccolomataceae remains incompletely resolved; it might be sister to all extant polypodialean lineages or sister to the Lindsaeineae 12,17,18 .
Together with the other ferns described to date in Burmese amber, the fern fossil detailed in this study suggests that, during the mid-to late Cretaceous, a fern flora existed in this region of the world that was composed mainly of representatives of families which still today occur in the lowland forests of SE Asia and Malesia, including Cystodiaceae, Dennstaedtiaceae 21 , and Lindsaeaceae 22 23 . Given the restricted present-day distribution of Cystodium in tropical lowland forests in Malesia and part of Melanesia, the fossil C. sorbifolioides supports the hypothesis that the tropical forests in this region represent refugia for many tropical plant lineages that originated in the Cretaceous.

Methods
Burmese amber comes from several localities near the village of Tanai in Kachin State, Myanmar 24,27 . Biostratigraphic studies 25 and U-Pb dating of zircons 26 suggest a late Albian to earliest Cenomanian age of Burmese amber, with a minimum age of 98 Ma.
Amber block (4.2 cm long, 1.9 cm wide) GZG.BST.21964 from the amber collection of the Geoscientific Collections of the University of Göttingen was studied using a Zeiss Stemi 2000 dissection microscope and a Zeiss AxioScope A1 compound microscope. After initial inspection, the block containing the fossil was cut into three portions (GZG.BST.21964 a to c) to come closer to the individual inclusions. Each portion was then ground and polished close to the inclusions using a series of wet silicon carbide papers (grit from FEPA P 600-4000, Struers) to minimize light scattering during subsequent detailed analysis and photographic documentation. Images were captured with a Canon EOS 5D digital camera attached to a compound microscope; incident and transmitted light were used simultaneously. To enhance the depiction of three-dimensional structures figures have been prepared as photomicrographic composites, digitally stacked from up to 60 focal planes using the software package HeliconFocus 6.7. Images of the extant Cystodium sorbifolium (B 200138084 (B), collected by C. E. Carr in 1935, in Papua New Guinea, forest at 1599 feet near Koitaki) were produced in Panorama 3D using a Keyence VHX-5000 Microscope. In order to prevent degradation of the amber specimens over time, the three amber specimens were embedded in a high-grade epoxy (EPO-TEK 301-2, Epoxy Technology, ratio of 100 [resin]: 37 [hardener]) under vacuum 40 .
A possible classification of the fossil was assessed using morphological evidence and comparisons to descriptions of extant fern families and genera in the literature 41 . Taxonomical updates were considered 19 .
The impact of the newly described fossil, as well as that of an unnamed Lindsaeaceae fossil 22 on minimum age estimates was tested using a dated phylogenetic hypothesis for the putative extant relatives of the suborders Saccolomatineae and Lindsaeineae 19 . This hypothesis was derived from published DNA sequences of five plastid markers (atpA, atpB, rbcL, trnH-psbA and trnL-trnF) downloaded from Genbank (see Supplementary Material Table S1). The sequences were aligned manually using BioEdit 7.0.5.3 42 and the chronogram was implemented in BEAST 1.8.2 43 . The GTR + I + G model was chosen as the best-fitting nucleotide substitution model in PARTITIONFINDER 44 employing the Akaike Information Criterion. We tested three approaches generated using a lognormal relaxed clock: (1) the amber fossil here described placed as minimum age of the split of the Cystodiaceae and Lonchitidaceae-Lindsaeaceae lineages; (2) the unnamed Lindsaeaceae fossil 22 conservatively considered a stem member of the family, thus assigned as minimum age of the split between Lonchitidaceae and Lindsaeaceae; and (3) an approach using both fossils. Calibrations were incorporated adopting a lognormal distribution with a minimum age set to 98 Ma 26 and a truncate range of 98-400 Ma, with a birth-death prior accounting for incomplete sampling 45 , and a single partition for all five markers. Analyses were run for 80 million generations sampling every 8,000 generations. The results were inspected in TRACER 1.5 46 and ESS values > 200 indicated ample mixing of the MCMC and a sufficient number of generations. These hypotheses were compared with published dating [11][12][13] for the clades of Saccolomatineae and Lindsaeineae. Data availability. All data analysed during this study are included in this published article and its Supplementary Information: Supplementary Table S1.