The most prolific source of exquisitely preserved bryophyte fossils is amber, which often contains inclusions in a three-dimensional and life-like state of preservation. In recent years, many fossil species have been described from 16 worldwide deposits ranging in age from the Miocene to the Cretaceous. One of the oldest is mid-Cretaceous Kachin amber from Myanmar. It includes the moss genera Vetiplanaxis and Calymperites as well as the leafy liverwort genera Frullania, Gackstroemia, Protofrullania, and Radula. All liverwort fossils belong to the mainly epiphytic Porellales, a group which was probably strongly influenced by the rise of angiosperms and underwent significant lineage turnover in the Cretaceous. Hence, Kachin amber provides important information on the evolution of leafy liverworts during a period characterized by fundamental changes in terrestrial ecosystems. Here, we describe two new species of the mainly epiphytic leafy liverwort genus Radula (Radulaceae), R. patrickmuelleri sp. nov. and R. tanaiensis sp. nov., and emend the description of Frullania kachinensis (Frullaniaceae). Radula is now represented by four species and one fossil only described to genus level, which indicates a high diversity already in the mid-Cretaceous. Furthermore, we describe additional fossil specimens of Frullaniaceae, including the third fertile specimen of Frullania baerlocheri, one sterile fossil of F. cretacea, and 12 sterile fossils of Protofrullania cornigera. The new fossil evidence necessitates an emendation of a recently published determination key for Cretaceous Jungermanniidae.

Amber fossils, Cretaceous, epiphytes, leafy liverworts

The description of bryophyte fossils from worldwide amber deposits underwent much progress in recent years with the number of species as well as the number of deposits increasing steadily (e.g., Bouju et al. 2021; Feldberg et al. 2021a, 2021b; Wang et al. 2021). Most of these fossils derive from Cenozoic deposits and have been found in middle Miocene Zhangpu amber from China (Wang et al. 2021), early Miocene Chiapas amber from Mexico (e.g., Estrada-Ruiz and Riquelme 2017; Feldberg et al. 2021a), early Miocene Dominican amber (e.g., Frahm and Newton 2005; Feldberg et al. 2021a), early Miocene Shewa amber from Ethiopia (Bouju et al. 2021), Miocene Cape York amber from Australia (Hand et al. 2010), Paleogene Bitterfeld amber from Germany (e.g., Frahm 2010; Feldberg et al. 2021a), late Eocene Baltic amber (e.g., Frahm 2010; Feldberg et al. 2021a), late Eocene Rovno amber from Ukraine (e.g., Ignatov et al. 2019; Feldberg et al. 2021a), late middle Eocene Anglesea amber from Australia (Stilwell et al. 2020), middle Eocene Sakhalin amber from Russia (Ignatov and Perkovsky 2013), early Eocene Cambay amber from India (e.g., Heinrichs et al. 2016a), and Paleocene-Eocene amber from the Chickaloon Formation in Alaska (Grimaldi et al. 2018). The oldest sources are Late Cretaceous Taimyr amber from Russia (e.g., Ignatov et al. 2016), Late Cretaceous Kuji amber from Japan (e.g., Katagiri et al. 2013; Katagiri and Shinden 2020), mid-Cretaceous Alaskan amber (Heinrichs et al. 2011), and mid-Cretaceous Kachin amber from northern Myanmar (e.g., Hedenäs et al. 2014; Heinrichs et al. 2014; Feldberg et al. 2021a, 2021b; Li et al. 2021b, 2022b). The latter includes the moss genera Vetiplanaxis with four species and Calymperites with one (Bell and York 2007; Hedenäs et al. 2014; Heinrichs et al. 2014) as well as a very diverse flora of porellalean leafy liverworts which has recently been revised by Feldberg et al. (2021b) and regularly yields new and interesting fossil species (e.g., Li et al. 2020, 2021b, 2022b).

Amber fossils provide important information on the evolution of leafy liverworts (Jungermanniidae) which are otherwise comparatively rare in the fossil record (e.g., Heinrichs et al. 2018; Tomescu et al. 2018). In contrast, thalloid liverworts (Pelliidae, Metzgeriidae, Marchantiopsida) and mosses (Bryophyta) are found more frequently in sedimentary rocks since the Middle Devonian and the Carboniferous, respectively (Tomescu et al. 2018; Ignatov and Maslova 2021). The often excellent and three-dimensional preservation state in amber allows utilization of many amber inclusions as minimum age constraints in divergence time estimates based on molecular phylogenies of extant species (e.g., Feldberg et al. 2014, 2021a; Laenen et al. 2014; Bechteler et al. 2017). Subsequent diversification analyses of Jungermanniidae indicate that the order Porellales, which mainly comprises epiphytes and epiphylls, underwent a lineage turnover beginning in the Cretaceous when flowering plants were on the rise (Feldberg et al. 2014). The Angiosperm Terrestrial Revolution (e.g., Magallón et al. 2019; Benton et al. 2022) caused fundamental changes in terrestrial ecosystems and has likely influenced the evolution of epiphytic lineages. Furthermore, Cretaceous Porellales often show distinct morphological differences to modern taxa which is another indicator of significant changes within this taxon during this period (Feldberg et al. 2021a, 2021b).

The most striking differences can be found in Frullaniaceae, by far the most diverse taxon included in amber. The family includes one extant genus, the mainly epiphytic Frullania, which is morphologically well defined but has a very complex subgeneric taxonomy with 576 accepted species in several subgenera whose circumscription is often unclear (e.g., Hentschel et al. 2009b, 2015; Heinrichs et al. 2010; von Konrat et al. 2013; Söderström et al. 2016; Carter et al. 2017). Extant Frullania has a subcosmopolitan distribution with a center of diversity in humid tropical regions, but also occurs in temperate as well as arctic and alpine areas (Schuster 1992; Hentschel et al. 2009b). The family also includes the extinct Protofrullania cornigera Heinrichs from Kachin amber (Heinrichs et al. 2017a; Feldberg et al. 2021b) and probably the extinct Kaolakia borealis Heinrichs et al. from Alaskan amber (Heinrichs et al. 2011) as well as Pseudofrullania hamatosetacea (Grolle) Heinrichs et al. from Bitterfeld amber (Heinrichs et al. 2018).

Despite having an extant center of diversity in the tropics, Cenozoic Frullania fossils are rarely reported from tropical regions. To date, only one inclusion of Frullania has been found in Dominican amber, which otherwise contains a very diverse leafy liverwort flora (Heinrichs and Schmidt 2010; Feldberg et al. 2021a), and two species have recently been described from Ethiopian Shewa amber, namely F. shewanensis Feldberg et al. and F. palaeoafricana Feldberg et al. (Bouju et al. 2021). In contrast, the genus is highly diverse in Cenozoic European ambers (Feldberg et al. 2021a). Fifteen species of Frullania have been described, with F. grabenhorstii Heinrichs et al. and F. mammilligera Grolle occurring only in Bitterfeld amber (Grolle 2003; Feldberg et al. 2018), F. acutata Casp. only in Baltic amber (Caspary 1887), and F. ekaterinae Mamontov et al., F. riclefgrollei Mamontov et al., F. rovnoi Mamontov et al., F. schmalhausenii Mamontov et al., F. vanae Mamontov et al., and F. zerovii Mamontov et al. only in Rovno amber (Mamontov et al. 2015, 2017, 2018, 2019, 2020). Frullania baltica Grolle, F. casparyi Grolle, F. schumannii (Casp.) Grolle, and F. truncata Casp. occur in Bitterfeld as well as Baltic amber (Caspary 1887; Grolle 1981, 1985; Grolle and Meister 2004), F. pycnoclada Grolle occurs in Baltic and Rovno amber (Grolle and Meister 2004; Mamontov et al. 2020), and F. varians Casp. occurs in Bitterfeld, Baltic, and possibly also Rovno amber (Caspary 1887; Grolle and Meister 2004; Konstantinova et al. 2012; Mamontov et al. 2020). While some of these species show some unique characters, which make an assignment to extant subgenera or sections difficult, many are very similar to the extant diversity and could be employed in divergence times estimates, given extended DNA datasets of extant species which can solve the remaining taxonomic problems (Mamontov et al. 2020; Feldberg et al. 2021a).

The oldest unequivocal Frullaniaceae fossils have been described from Kachin amber and these tend to differ strongly from the extant diversity (Feldberg et al. 2021a, 2021b). The most common leafy liverwort inclusion found in this deposit is Protofrullania cornigera (Heinrichs et al. 2017a; Feldberg et al. 2021b), which differs from Frullania by its elongated ovate to strap-shaped underleaves which carry rhizoid bundles at their apical regions, a character unknown in Frullania where the undivided or bifurcated underleaves carry rhizoid bundles near their base or near the middle. But Protofrullania is not the only Frullaniaceae with morphological characters unknown or rare in extant representatives, also the mid-Cretaceous species assignable to Frullania show unique or rare characters and character combinations. One example is F. cretacea Hentschel et al. (Hentschel et al. 2009a; Heinrichs et al. 2012) with its long ciliate underleaves, and another is F. kachinensis Y.Li et al. (Li et al. 2021b), which has bifurcated underleaves with serrate lobes and rhizoid bundles inserted just below the sinus. The remaining fossil species F. baerlocheri Heinrichs et al. and F. partita Y.Li et al. might look more similar to the extant diversity, but do not fit any extant lineage within the genus due to the unusual combination of characters (Feldberg et al. 2021b; Li et al. 2021a).

Another diverse genus found in Kachin amber is Radula, which comprises ca. 200 extant species in seven subgenera (e.g., Devos et al. 2011; Söderström et al. 2016), most of which are epiphytes or epiphylls. Radula has a subcosmopolitan distribution with a center of diversity in tropical to warm temperate regions. In recent years several molecular phylogenetic studies greatly extended the knowledge of this taxonomically challenging genus (Devos et al. 2011; Renner et al. 2013a, 2013b; Renner 2014, 2015; Patiño et al. 2017). However, species delimitation based on morphological characters alone can be difficult because of the high degree of morphological homoplasy and sometimes subtle differences among relatively unrelated taxa (e.g., Renner et al. 2013a, 2014, 2015). This especially applies to fossil material which is often fragmentary and sterile.

Radula has a comparatively extensive fossil record with nine species known from different amber deposits (Grolle 1987; Heinrichs et al. 2016b, 2018; Kaasalainen et al. 2017; Stilwell et al. 2020; Feldberg et al. 2021a, 2021b). Most appear very similar to the extant diversity, though this might be misleading. The youngest fossils are R. intecta M.A.M.Renner et al. (Kaasalainen et al. 2017) and R. steerei Grolle (Grolle 1987) from Miocene Dominican amber, which are sterile and cannot be assigned to a subgenus with confidence. Radula intecta may belong to subg. Volutoradula and R. steerei has been compared to the extant Japanese endemic R. fauriana Steph. (Grolle 1987), which is listed as incertae sedis in Söderström et al. (2016) but was included in subgen. Radula sect. Faurianae by Yamada (1979). Three fossil species have been found in Cenozoic ambers of Europe. Radula baltica Heinrichs et al. is known only from the holotype in Eocene Baltic amber, while R. sphaerocarpoides Grolle and R. oblongifolia Casp. are common fossils in Baltic as well as Bitterfeld amber (Caspary 1887; Grolle 1980; Grolle and Meister 2004; Heinrichs et al. 2016b). The latter species is autoicous and is often preserved with androecia, gynoecia, and sporophyte, while the former two are only known in sterile condition and might represent several biological entities given their morphological heterogeneity (Heinrichs et al. 2016b). Furthermore, two fossil specimens have been reported from late middle Eocene amber of Anglesea, Victoria, Australia (Stilwell et al. 2020) but have not been formally described. The oldest previously described species are R. cretacea Bechteler et al., which is preserved with apparently unfertilized gynoecia, and the likely distantly related and sterile R. heinrichsii Feldberg et al. from Kachin amber (Bechteler et al. 2017; Feldberg et al. 2021b; Wang et al. in press). Radula cretacea is probably a member of subg. Odontoradula and may even represent one of the several crown group lineages, given the shared possession of characters displayed by extant species of Radula subg. Odontoradula sect. Acutifoliae (Bechteler et al. 2017). The relationship of R. heinrichsii is more obscure, because its microphyllous (amentulose) branches have no exact equivalent among extant species, though their presence along with the longitudinal lobule stem insertion suggests the species belongs to subg. Amentuloradula (Feldberg et al. 2021b).

Here we describe two further fossil species of this diverse genus from Kachin amber as R. patrickmuelleri sp. nov. and R. tanaiensis sp. nov. and provide an updated determination key for Cretaceous Jungermanniidae. Furthermore, we obtained several new specimens of previously described fossil species which we list and evaluate. These include several sterile specimens of Protofrullania cornigera, one sterile specimen each of Frullania cretacea and F. kachinensis, the third fertile specimen of F. baerlocheri with gynoecium, as well as a fossil showing strong similarities to Radula cretacea. Furthermore, two additional specimens of the moss Calymperites burmensis Heinrichs et al. are illustrated. Most specimens are in good accordance with the type material, others require diagnoses and descriptions to be emended.

All fossils derive from Kachin amber of northern Myanmar which was mined from outcrops about 20 km southwest of the village of Tanai located on the Ledo Road ca. 105 km north of Myitkyina in Kachin State, northern Myanmar (Cruickshank and Ko 2003; Grimaldi and Ross 2017). Several lines of evidence point to a mid-Cretaceous age of Kachin amber (Grimaldi et al. 2002; Grimaldi and Ross 2017). Biostratigraphic studies by Cruickshank and Ko (2003) suggested an age of ca. 100 Ma (late Albian) of the amber-bearing deposit, and Shi et al. (2012) assigned a minimum age of 98 Ma (earliest Cenomanian) based on recent U-Pb dating of zircons. These age estimates have recently been confirmed by the discovery of an ammonite shell of the genus Puzosia, a late Albian–late Cenomanian index fossil, enclosed in a piece of Kachin amber (Yu et al. 2019).

The specimens of Protofrullania cornigera (BuB 3811, 3812, 4333–4338, 4394, 4408, 4409, 4413), Frullania cretacea (BuB4411), and Calymperites burmensis (BuB 4339, 4398) are housed in the Patrick Müller Amber Collection (Amber Research Group, c/o Geological-Palaeontological Institute and Museum (CeNak) of the University of Hamburg). Frullania baerlocheri GZG.BST.22039 (BuB4220), F. kachinensis GZG.BST.22040 (BuB4431), Radula patrickmuelleri GZG.BST.22041 (BuB4395), R. tanaiensis GZG.BST.22042 (BuB4329), and Radula sp. GZG.BST.22043 (BuB4420) are housed in the Geoscience Centre (GZG) at the University of Göttingen, Germany (all formerly Patrick Müller Amber Collection). Frullania kachinensis PB22711a (holotype) is housed in the Nanjing Institute of Geology and Palaeontology, Chinese Academy of Sciences.

Relating to the recent conflicts in Myanmar (e.g., Sokol 2019), we declare that all Kachin amber pieces mentioned in this study have been collected before the year 2017 and that we follow the recommendations by Haug et al. (2020). The specimens from the American Museum of Natural History were provided to the museum prior to 2010 (refer to Hentschel et al. 2009a and Heinrichs et al. 2012). The fossils deposited in the Nanjing Institute of Geology and Palaeontology and the Zhejiang Museum of Natural History were collected in compliance with the laws of Myanmar and China, including Myanmar’s import and export regulations of jewelry, and China’s fossil law. Specimens deposited in the remaining collections were legally purchased from an authorized trader from Myanmar in 2016.

Frullania kachinensis (PB22711a, holotype) was examined under a ZEISS Axio ZoomV16 microscope equipped with a high-resolution digital camera (Axiocam 512 colour). All images were digitally stacked photomicrographic composites from ca. 50 individual focal planes using the software package ZEN 2.3 pro. All other specimens were prepared for study according to procedures described by Sadowski et al. (2021). Photographic documentation of the specimens was carried out under Zeiss AxioScope A1 compound microscopes equipped with Cannon Eos 70D and Cannon Eos 5D digital cameras. All images are digitally stacked photomicrographic composites from up to 76 individual focal planes using the software package HeliconFocus version 7.6.1. Pro.

The new fossil material described in this publication includes some well-preserved sterile specimens of Frullania cretacea (Suppl. material 1: Fig. S2F, G), which is represented by 15 fossils in total, and Protofrullania cornigera (Suppl. material 1: Figs S3–S5), which is now represented by altogether 30 fossils. The latter is by far the most common species found in Kachin amber and the only species for which androecia as well as gynoecia are known (Feldberg et al. 2021b). Frullania baerlocheri (Suppl. material 1: Fig. S2A–E) is now represented by 12 fossils, including three fertile specimens with gynoecia, while F. kachinensis (Figs 1, 2) is known from eight sterile fossils, and F. partita from only two specimens (Li et al. 2020, 2022a). Several of these species occur together in individual amber pieces, namely F. baerlocheri and P. cornigera (SNSB-BSPG 2021 XII 1; described in Feldberg et al. 2021b), F. kachinensis and P. cornigera (PB22711; described in Li et al. 2021b), F. kachinensis and F. baerlocheri (ZMNH155047; described in Li et al. 2021a), and F. baerlocheri and Radula heinrichsii (ZMNH155044; described in Wang et al. in press). This indicates that these species grew very close to each other, obviously even on the same tree.

The new material allowed us to emend the diagnosis and description of F. kachinensis. Like many fossil species from Cretaceous ambers, F. kachinensis shows some characters that are unusual or even unknown in extant Frullaniaceae, e.g., rhizoid bundles inserted directly below the underleaf sinus (Fig. 1G, H). Extant Frullania has rhizoid bundles inserted at the base of underleaves or towards the middle (e.g., Hattori 1975, 1983; Schuster 1992), and the unusual insertion point aligns F. kachinensis with Proto­frullania cornigera. The form of the underleaves, however, is completely different. While P. cornigera mostly has strap-shaped underleaves with a wider apical area armed with two to four cilia and carrying the rhizoid bundle (Suppl. material 1: Figs S3–S5), F. kachinensis has elliptic to elongate ovate underleaves which are bifid to ca. 0.3–0.8 × their length into two lanceolate lobes and are more consistent with typical underleaves of extant Frullania in this aspect (Figs 1F–H, 2C, D). However, another difference to extant species is the strong armature of the underleaf lobes which can be weakly to strongly toothed or ciliate and resemble the bracteoles of extant species more than vegetative underleaves (Figs 1G, H, 2C). Some extant Frullania species from different subgenera, e.g., F. monocera (Hook.f. & Taylor) Gottsche, Lindenb. & Nees of subgen. Acutilobae, F. blepharozia Spruce of subgen. Chonanthelia, F. curvistipula Steph. of subgen. Diastaloba I, and F. amamiensis Kamim. of subgen. Frullania, also have toothed vegetative underleaf lobes, but these teeth are mostly shorter and situated on the outer side of the triangular lobes (Yuzawa 1991, 2001). Another example would be F. papulosa Steph., which has lanceolate, pilose underleaf lobes with two spinose teeth at the lower margins.

Radula tanaiensis, R. patrickmuelleri, and the additional fossil GZG.BST.22043 can be confidently assigned to the genus. The presence of complicate bilobed lateral leaves with a large dorsal lobe and a small Lejeunea-type lobule, the absence of underleaves, as well as the presence of rhizoids on the lobules in the case of R. patrickmuelleri allow a definite assignment to Radula. The new species are morphologically well differentiated from the other two fossil species from Kachin amber, while the classification of R. sp. is ambiguous.

Radula tanaiensis can be clearly differentiated from R. cretacea, R. heinrichsii, and R. patrickmuelleri by the very large lobules on what are probably primary shoots. Lobules are up to 0.4 × as big as the lobes and their free antical margin often reaches the antical margin of the lobe (Fig. 5A, B). Lobules in R. cretacea are less than 0.2 × as big as the lobes and also the acute triangular-ovate lobes separate it from the new species (Fig. 4F). Radula patrickmuelleri has similarly small lobules and the lobes are less wide and have nearly parallel postical and antical margins in their middle parts (Fig. 3). Radula heinrichsii is the most similar species, but the lobules are only 0.2–0.3 × as big as the lobes, obovate to nearly orbicular instead of ovate to rounded trapezoid, and have more broadly rounded apices (Fig. 5A, B, F). The lobes of R. heinrichsii are more elongated, and the free exterior cell walls of the marginal lobe cells are not as distinctly thickened as in R. tanaiensis (Fig. 5E, G). Contrary to R. heinrichsii, the cell walls of R. tanaiensis also appear smooth for the largest part (Fig. 5D, E), but the cells are often indistinct, and some might have rounded mamillae in their center. Furthermore, microphyllous axes (Fig. 5F) have not been seen in the new species. It has three intact branches with a few small and somewhat reduced basal leaves which soon become main shoot-like and some broken remnants of branches whose character is not discernible (Fig. 5A, C).

Radula patrickmuelleri has roughly the same shoot size as R. heinrichsii and R. tanaiensis but can be clearly separated by the size and shape of the small lobules which are less than 0.2 × as big as the lobes, the absence of microphyllous branches, and the gemmae formation (Figs 3, 4). Radula cretacea resembles R. patrickmuelleri at first glance, because it also has small lobules as well as irregular gemmae on the lobe margins (Fig. 4F–H). The form of the lobules and the longitudinal insertion might be similar, but those of R. cretacea are quadrate to trapeziform, have slightly curved exterior and antical margins which are somewhat irregular due to bulging marginal cells, and an obtuse to slightly attenuate apex with a papilla in a shallow notch (Fig. 4F). In contrast, the lobules of R. patrickmuelleri are rounded rectangular to ovate, have more strongly curved exterior and antical margins which do not show any bulging marginal cells, and subacute to broadly rounded apices (Fig. 3D–F). Notches with papillae have not been seen. The most conspicuous differences are the size of the plants and the form of the dorsal lobes. Radula cretacea is much smaller and has triangular-ovate leaf lobes (Fig. 4F) while R. patrickmuelleri is larger and has oblong-elliptic lobes (Fig. 3A–D). The antical and postical margins of the lobes also show a different curvature. The antical lobe margin of R. cretacea is more strongly arched than the postical margin, whereas R. patrickmuelleri has nearly parallel medial postical and antical margins. Furthermore, R. cretacea has acute to slightly acuminate lobe apices, whereas R. patrickmuelleri has broadly rounded lobe apices. Another differentiating character are the gemmae. While the first cell division in the gemmae of R. cretacea seems to be mainly anticlinal (Fig. 4G, H), all clearly visible first divisions in R. patrickmuelleri are periclinal (Fig. 4B–E).

The presence of four species in Kachin amber and recent divergence time estimates indicate a high diversity of Radula already in the Cretaceous (Bechteler et al. 2017; Patiño et al. 2017; Feldberg et al. 2021a, 2021b). Radula cretacea and R. heinrichsii show close morphological affinities to the extant subgenera Odontoradula and Amentuloradula, respectively (Bechteler et al. 2017; Feldberg et al. 2021b). The assignment of the two new species, however, is more difficult due to the scarcity of the fossil material. Based on the available evidence, it is not possible to assign R. patrickmuelleri and R. tanaiensis to any subgenus with confidence, and because of the high degree of morphological homoplasy in extant Radula, a comparison with extant lineages is premature. The gemmae formation in R. patrickmuelleri is perhaps consistent with an epiphyllous lineage in extant subg. Metaradula, where the first cell division is always periclinal. However, the species also shares some similarities with the extant R. pugioniformis M.A.M.Renner of subg. Odontoradula but differs in having larger leaf lobe cells with thin walls and distinct trigones, as well as gemmae (Renner et al. 2013b). A more accurate assignment of the new species might be possible when more fossil material is discovered. Both the age of each Radula subgenus, and the relatively long stem lineages leave open the possibility that fossils belong to extinct lineages that diverged from the various stem lineages. Bechteler et al. (2017) evaluated how misleading an assignment solely based on morphological similarity may be, and assigned R. cretacea to different nodes in the molecular phylogeny of Radula. An assignment of R. cretacea to the stem of subgenus Odontoradula proved to be most likely and resulted in an estimated mean age of 176.3 Ma for the genus.

DNA-based divergence time estimates provide evidence for a species turnover within Jungermanniidae during the Angiosperm Terrestrial Revolution, and fossils in Kachin amber offer unique insights into the evolution and diversification of leafy liverworts during this time span (Feldberg et al. 2014; Laenen et al. 2014; Benton et al. 2022). They are essential as age constraints in divergence time estimations, because only very few Mesozoic fossils from the rock record show enough details to allow a reliable assignment to extant lineages (Heinrichs et al. 2018; Tomescu et al. 2018; Feldberg et al. 2021a, 2021b).

Unfortunately, neither Radula patrickmuelleri nor R. tanaiensis can contribute to age constraints beyond what is already gleaned from other fossils, which is an absolute minimum age for the lineage stem node of Radula, though they provide more evidence of a high species diversity of Radula already in the mid-Cretaceous. The newly detected fossil of Frullania kachinensis allowed emendation of the species diagnosis and description for some important characters but can likewise not provide new age constraints, because some unusual characters make it impossible to assign it to any group within the crown group of Frullania. The rhizoid insertion resembles Protofrullania cornigera more than any extant Frullania and it might thus be more closely related to this genus.

Our results suggest that new fossil liverwort material should preferably be described based on as many specimens as possible, as every specimen is incomplete. Not only does fossil material often consist of sparse and damaged fragments which do not show all relevant characters, leafy liverworts often show significant morphological homoplasy between lineages as well as some morphological plasticity within a species, with characters varying in expression between individuals or even between different parts on the same plant, e.g., the main shoot and branches. This high variability may obscure species boundaries in extant leafy liverworts and is even more obstructive in the description of extinct species.

K.F. designed research. K.F. and Y.L. prepared and documented the specimens. K.F., A.S.-V., Y. L., and M.A.M.R. analyzed the data and wrote the paper. All authors commented on the manuscript.

The authors have no competing interests to declare.

We would like to thank Patrick Müller (Zweibrücken) and Jörg Wunderlich (Hirschberg an der Bergstraße) for providing specimens for study. We are grateful to Alexander R. Schmidt (Göttingen) for commenting on an earlier version of the manuscript as well as to Grzegorz J. Wolski, an anonymous reviewer, and the editor Florian Witzmann for constructive comments that improved the manuscript. Financial support from the German Research Foundation (DFG, project 428174246 to K.F.) and the National Natural Science Foundation of China (grant number 41790454 to Y.L.) is gratefully acknowledged.