The resurrection of Neohattoria Kamim. (Jubulaceae, Marchantiophyta): a six decade systematic conflict resolved through a molecular perspective.

The systematic placement of Frullaniaherzogii has been contentious since its description six decades ago. Over the years it has been interpreted as either a member of the genus Frullania or segregated into its own genus, Neohattoria, due to morphological similarities with both Frullania and Jubula. Here we provide molecular evidence that supports the recognition of the genus Neohattoria and its inclusion within the Jubulaceae, together with Jubula and Nipponolejeunea. Jubulaceae are placed sister to Lejeuneaceae rather than to the monogeneric Frullaniaceae.

Introduction

The liverwort S.Hatt. was originally described by Hattori (1955) from a poor, sterile specimen collected on Mt. Hayachine in Iwate Prefecture, northern Honshu, Japan. Since that time the generic and even familial placement of the species has remained controversial. The species also has remained poorly known partially because of its seemingly limited distribution in the subalpine coniferous forest zones of Honshu and Hokkaido, Japan, and the Kuril Islands (Inoue et al. 1981, Stotler and Crandall-Stotler 1987). Hattori (1955) remarked that the leaf morphology, with acute teeth along the margin, differed from all the other Japanese Raddi species known by him. A few years later, in his monograph of Japanese , Kamimura (1961) erected the new genus Kamim. to separate this taxon from other species. He stated that although his new genus superficially resembled species of (Spruce) Schiffn. or , there was an important similarity between the branching patterns of (S.Hatt) Kamim. and species in the genus Dumort. In both and , the branches replace the lobule of the leaf at the point of insertion, and the leaf lobes are attached to both the main stem and to the branch. Although Kamimura (1961) noted the similarity of cell shape between and , he considered the combination of branching architecture and leaf denticulation sufficient to recognize as a distinct genus. A year later he had to give a new name, Kamim., to his recently described genus (Kamimura 1962), because of the almost simultaneous although earlier description of by Schuster for a liverwort in the (Schuster 1961).

Later Schuster (1963), in a key for the Southern Hemisphere genera of liverworts, expanded the circumscription of PageBreak to include two more species, Pearson from New Caledonia, and (R.M.Schust.) R.M.Schust. ex von Konrat, L.Söderstr. & A.Hagborg from Fiji. He based his taxonomic decision on the morphology of the reduced leaves on branch bases, the subfloral innovations, and the sharply delimited bracts and bracteoles of , and on the toothed leaf lobes of this species. Schuster (1963) did not provide any argument for placement of the Fijian in , other than the hyaline margins of the leaves that can be seen in this species and in (as inferred from the key). However, his key is restricted to the Southern Hemisphere and did not include the type of the genus, which completely lacks a hyaline border in leaf lobes. Schuster (1970) later expanded this generic concept even further, including the Australasian (Hook.f. & Taylor) Hook.f. & Taylor ex Gottsche, Lindenb. & Nees (as R.M.Schust.) and von Konrat, Braggins, Hentschel & Heinrichs (as R.M.Schust.), the SE Asian (Sande Lac.) Grolle & S.Hatt. [as (Steph.) R.M.Schust.], the New Caledonian (R.M.Schust.) R.M.Schust. and J.J.Engel (as R.M.Schust.). Of these, is now considered a member of von Konrat, Hentschel & Heinrichs (von Konrat et al. 2010), while the rest of the taxa are currently included in (R.M.Schust.) R.M.Schust. The current taxonomic placement of these taxa is based on both morphological (Hattori and Mizutani 1982, Schuster 1992) and molecular evidence (Hentschel et al. 2009, von Konrat et al. 2012).

Asakawa et al. (1979) demonstrated, based on chemical compound differences, that sensu lato should be divided into three families, i.e. , and . This view has been confirmed by most molecular phylogenies published to date (e.g., Forrest et al. 2006, Heinrichs et al. 2005, 2007). Asakawa et al. (1979) listed 11 morphological characters that support the separation of and , and placed together with in the . Hattori (1982, 1984, 1986) and Hattori and Mizutani (1982) also accepted the separation between and and argued that R.M.Schust., a genus formerly considered by Schuster (1970, 1980) as intermediate between and , should be placed within . This view was first held by Engel (1978), who had earlier reduced to a synonym of .

In 1987, Stotler and Crandall-Stotler published a thorough treatise of the taxonomic history of (S.Hatt.) Kamim. in the context of a detailed re-evaluation of its morphology, including the discovery of immature female inflorescences. In that contribution they came to the conclusion that this taxon should be considered within the circumscription of , although in its own subgenus, Stotler & Crand.-Stot. Their conclusion was based on both vegetative and reproductive characters, including the morphology of the bracts surrounding the female gametangia, lobule anatomy, leaf cell pattern, and the morphology of regenerants. Although they recognized that leaf-lobe insertion, branch morphology, and morphology of stylus are more similar to than to , they concluded that on the basis of the -like inflorescences and regenerants, should be synonomized with . This synonomy was adopted by Grolle and Meister (2004) who described a morphologically similar plant from Oligocene amber from Bitterfeld (Germany) as Grolle. However, this fossil species appears morphologically closer to than to , and this issue will be explored in detail in a forthcoming monograph of the latter subgenus.

Lack of useable specimens has previously precluded inclusion of in molecular phylogenetic studies. As a result of recent collecting activities, fresh material became available that allowed for successful DNA extraction and amplification. In the present study, we use molecular sequence data to investigate the phylogenetic position of . We investigate whether the genus should be placed in the or the and evaluate whether molecular evidence supports the recognition of as a distinct genus.

Methods

Microscopy

For the production of microscopic images an Olympus BX51 microscope was used, equipped with both a QICAM Fast1394 camera from QIMAGING (Surrey, Canada), and a slide scanner (moving platform stage attached between the objectives and the condenser) from Objective Imaging Ltd. (Cambridge, UK). The software “Surveyor” from the latter company was used for the digitally rendered images.

DNA extraction, PCR amplification and sequencing

We worked with two independent datasets to address two different questions, (1) what is the position of relative to the , and , and once we obtained results from these analyses, we asked (2) what is the position of within the . For dataset 1 sequences were generated for two mitochondrial (nad1, rps3), and two chloroplast loci (psbA, rbcL), following DNA extraction, amplification and sequencing methods described by Shaw et al. (2003), and using primer sequences provided in Cooper et al. (2011). For dataset 2 we used the aforementioned plastid regions (psbA and rbcL) together with the nuclear ITS region following the methods described by Shaw et al. (2003), and the chloroplast trnL-trnF region, amplified and sequenced as described in von Konrat et al. (2012). All sequences were edited and manually aligned in PhyDE v0.9971 (www.phyde.de) following the alignment rules and hotspot definitions presented in Kelchner (2000), Olsson et al. (2009), and Borsch and Quandt (2009).

Taxon sampling and outgroup selection

For dataset 1 seven species of were selected as outgroup taxa following the results already published in recent liverwort phylogenies (Davis 2004, Forrest et al. 2006, Feldberg et al. 2014, Heinrichs et al. 2005, 2007). The same criteria were undertaken for dataset 2, including all taxa with sequences available in GenBank for and S.Hatt. (Ahonen 2006, Ahonen et al. 2003, Konstantinova and Vilnet 2011, Pätsch et al. 2010, Wilson et al. 2004, 2007), using selected taxa of the and species of as outgroup based on results from dataset 1. GenBank accession numbers for both newly generated sequences and for already published sequences are provided in Appendices 1 and 2 for datasets 1 and 2 respectively.

Phylogenetic inferences

Both datasets were analysed with PartitionFinder v1.1.0 (Lanfear et al. 2012, 2014) to develop best-fit partitioning schemes and models of molecular evolution. Dataset 1 PageBreakwas partitioned setting one separate data block for each of the four genes used, each of them divided in three according to each codon position; introns and/or spacers were coded as extra partitions. Dataset 2 was partitioned in four parts, corresponding to the regions included only, without inner codon partition for the coding regions analysed. For dataset 1, phylogenetic reconstructions under maximum likelihood (ML) were performed in GARLI v2.01 (Zwickl 2006), setting up seven different models for the eleven partitions determined by PartitionFinder. Two independent searches each with 100 bootstrap replicates were made, and the 50% majority-rule consensus tree from all obtained trees was obtained with SumTrees v3.3.1 included in the package DendroPy v3.12.2 (Sukumaran and Holder 2010). Bayesian Posterior Probabilities analyses (PP) were executed in MrBayes v3.2.2 (Huelsenbeck and Ronquist 2001, Ronquist and Huelsenbeck 2003) also with the partitioned data set as given by PartitionFinder, and setting a different model for the individual partitions from the available options in MrBayes, with all characters given equal weight and gaps treated as missing data. The default settings of the program for a priori probabilities were used. Four runs, each with four MCMC chains (one million generations each) were run simultaneously, with the temperature of the heated chain set to 0.2 (default setting). Chains were sampled every 100 generations. Calculation of the consensus tree and posterior probabilities of clades was based on the set of trees sampled after the chains had converged, as observed graphically using Tracer v1.5 (Rambaut and Drummond 2007). For dataset 2, phylogenetic reconstructions under ML were performed in GARLI v2.01 and Bayesian analyses were executed with MrBayes v3.2.2 following the protocols as described above. For this dataset only three different partitions were suggested by PartitionFinder, and the models given by this software for each partition were incorporated into the settings of both the ML and the Bayesian analysis. Trees were edited and support values added using TreeGraph v2.0.54-364 beta (Stöver and Müller 2010).

Results

The complete alignment for dataset 1 including all four regions mentioned above, with flanking areas pruned to avoid ambiguous readings, comprised 4818 characters for 54 accessions, of which 694 were parsimony informative. A total of 101 new sequences were generated for this study (Appendix 1). In the analysis of the sequences with accessions of the , and (dataset 1), is strongly supported (as defined by Pedersen et al. 2007) as one of three clades belonging to the in both ML and Bayesian analyses, with accessions of , resolved in a second clade and those of , in a third clade (Fig. 1), although the latter with low support (ML = 52, PP = 0.6). The is resolved as sister to the with strong support in both types of analysis. The position of the as sister to this latter clade ( + ) was strongly supported by the Bayesian analyses (PP = 1.0), but it was not recovered by the ML analyses. The Bayesian analyses also resolved as sister to the rest of the ( + ) with strong support (PP = 1.0).

Figure 1.

Maximum likelihood (ML) tree showing the systematic position of relative to the , and . Wide black branches indicate ML bootstrap support > 90 % and PP > 0.95.

The complete alignment for dataset 2 including all four regions included, and after pruning the flanking areas to avoid ambiguous readings and deleting unalignable areas of the ITS region, comprised 3737 characters for 55 accessions, of which 548 were PageBreakparsimony informative. The four different regions were not equally represented in the matrix, as shown in Appendix 2. The results of the analyses (Fig. 2) confirm with strong support the placement of within the (ML = 100, PP = 1.0), and forming a sister clade to , although recovered with strong support only by the Bayesian analysis (ML = 64, PP = 0.97). was resolved as the sister clade to the - clade, although with low support (ML = 65, PP = 0.5).

Figure 2.

Maximum likelihood (ML) tree showing the systematic position of within the . Only 1/2 of the length of the branch between the and the / clade is depicted. Wide black branches indicate ML bootstrap support > 90 % and PP > 0.95.

The voucher of used for DNA extraction is illustrated in Figure 3.

Figure 3.

. A Habit, dorsal view B Habit, ventral view with distal lobules detached C Regenerant shoot originating from a detached lobule D Lobule E Underleaf F–K Leaves. All from Furuki 22673 (F). Scale bar: 350 µm (A, B), 200 µm (C), 180 µm (D), 300 µm (E), 150 µm (F–K).

Discussion

Our molecular analyses support recognition of the genus as distinct from the genus , as first proposed by Kamimura (1961) almost 55 years ago. Moreover, our molecular analysis strongly supports its inclusion within the , together with and . A close relationship with , based on similarities in branch morphologies, was first suggested by Kamimura (1961, p. 94), and also accepted by Hattori et al. (1972). Inoue et al. (1981) provided new karyological, chemical and ecological data on and concluded that the biosystematic evidence collected suggested distance between and , but, nonetheless, retained in the . While morphologically closer to than to which it is sister, it is clearly not nested in the clade. This combination of molecular and morphological evidence, in fact, supports its recognition as a distinct genus in the .

Circumscription and relationships of the

Our results strongly support the position of the (containing , and ) sister to the , and the as sister of the latter clade, although without significant support (Fig. 1). These results agree with several molecular phylogenies (e.g. Ahonen 2004, Forrest et al. 2006, Heinrichs et al. 2005, 2007). Thus the traditional view of a widely circumscribed including is further rejected in this study.

These three families (, and ) share several morphological characters, including the leaves divided into two (or three) parts [lobe, lobule (and stylus)], the beaked perianths, the sporophyte enclosed in a stalked true calyptra, the bistratose capsule wall, and the vertically aligned elaters that are attached to the valve apices (Crandall-Stotler et al. 2009, Gradstein et al. 2001, Schuster 1992). However, these characters need to be carefully evaluated to understand their evolution and their role in demonstrating the history of these lineages. In the past, PageBreak, and members of the were placed in a single taxonomic group (the subtribe Jubuleae), based largely on the similarities among their sporophytes (e.g. Müller 1915). Verdoorn (1930) argued that based on most characters (e.g., number of archegonia, seta form, and lobule ontogeny) belongs nearest to , which later lead Schuster (1992, p. 6) to describe as a “bona-fide genus of [= ]”. Mizutani (1961) was the first to propose that, except for the lobule structure, had no alignment with , and subsequently placed into the . However, Asakawa et al. (1979) concluded that chemically, both and are quite different from Lib. species. Interestingly, the phylogenetic analysis by Crandall-Stotler and Stotler (2000) of 40 gametophyte and 21 sporophyte characters distributed among 34 liverwort families, resolved Mont. as sister to a clade containing (Steph.) Verd. and (Ehrh.) Lindb. However, in the systematic treatment of the same work (Crandall-Stotler and Stotler 2000) is presented as including both and , whereas the is presented as a separate family, following accepted classifications of the time. The revised version of that classification, incorporating some recent molecular data, presents the , and as three separate families within the suborder (Crandall-Stotler et al. 2008, 2009), which is accepted here but with the transfer of from the to the .

Assessing the importance of different morphological characters in circumscribing , and has been a difficult problem, but there are several characters that are consistent with the molecular phylogenetic results presented here. In most a true stylus does not develop, but instead a single, unstalked slime papilla is formed at the junction of the lobule base and the stem, while in and there is a one- or two-celled filament terminated by a slime papilla in this position (Crandall-Stotler and Guerke 1980, Stotler and Crandall-Stotler 1987). Both types of structures are clearly different from those of the , where the stylus is always formed by more than two cells and is usually very conspicuous. The and can be clearly differentiated from the by the lobule, which is almost free from the larger dorsal lobe, and typically modified into an inflated, balloon-like to helmet-shaped sac whose aperture is directed either toward the shoot base or toward the stem, with the exception of which has -like lobules. Guerke (1978) hypothesised that was more advanced than on the basis that has many specialized characteristics e.g., a highly reduced stylus, seta, and foot, and features associated with the sporeling. In contrast, Schuster (1992, p. 9) stated that taxa such as (Gola) R.M.Schust. (≡ Gola), with a 16 + 4 seriate seta and monogynous gynoecia, diminish the distinctions between the two groups such that he prefers not to attempt a “subfamilial separation” at all. However, revision of the chemical, morphological, and ecological data provided support for the recognition of two subfamilies in the (Guerke 1978, von Konrat 2004). Alternatively, Asakawa et al. (1979), on the basis of biochemical and morphological evidence, proposed two families: (, ) and (, S.Hatt., , and S.Hatt.). Hattori (1982, 1984, 1986) and Hattori and Mizutani (1982) also accepted two families. This approach has been adopted in most recent hepatic floras and classifications (Paton 1999, Damsholt 2002, Casas et al. 2009, Crandall-Stotler et al. 2009, Frey and Stech 2009).

Schuster (1980, 1992) questioned the division into two families and argued that only the single family should be recognized, but commented that this area of classification remains replete with ambiguities and contradictions. Interestingly, he also suggested that there was a possibility that might share a closer affinity to (= ) than to (Schuster 1996), a view first expressed when Grolle (1966) transferred E.A.Hodgs. & S.W.Arnell, which is the generitype of R.M.Schust., to . However, recent molecular analyses (e.g., Heinrichs et al. 2005, Forrest et al. 2006) have demonstrated that (= ) is far removed from the .

Morphologically, the monogeneric can be differentiated from the by: (1) plants usually with conspicuous secondary pigmentation, often reddish; (2) initial leaves of branches either trifid or bifid; and (3) spores with rosette-like protrusions. Conversely, in the the plants are: (1) soft and without secondary pigmentation (thus usually dull green to pale brown); (2) the initial leaves of branches are small, subtriangular, and never tri- or bifid; and (3) the spores without rosette-like protrusions. The first two of these characters support the placement of within rather than (spores remain unknown in ).

Chemically, species in general, produce significant amounts of sesquiterpene lactones, diterpenoids, and bibenzyl derivatives, which are considered important chemosystematic markers of the group (Asakawa et al. 1981, 1983, 1987, Kraut et al. 1994). On the other hand, cyclocolorenone and maalioxide have been isolated as major components of (Steph.) Horik. & Ando (Asakawa et al. 1979); interestingly cyclocolorenone is also widely distributed in the . In contrast, no members of or produce paraffinic hydrocarbons which are characteristic for (Inoue et al. 1981).

Interestingly, Schuster (1996) suggested that there was a possibility that might share a closer affinity to (= ) than to . This view was first expressed when Grolle (1966) transferred E.A.Hodgs. & S.W.Arnell, which is the type species of R.M.Schust., to . However, preliminary unrooted trees made for this contribution including C.Massal., Trevis., Steph., R.M.Schust., Dumort. (= ) and L. together with representatives outside the , showed far away from but within (results not depicted). These results are basically the same as the ones observed in recent molecular phylogenies (e.g. Heinrichs et al. 2005, Forrest et al. 2006), demonstrating that these groups are only distantly related to either the or the .

Circumscription and relationships of

Our results place PageBreak in the with strong support, together with and . Within the , is resolved as sister to , and this latter clade sister to , although this relationship is sensitive to taxon sampling (cf. Figs 1 and 2), and not strongly supported in the analyses. When describing the genus (later renamed ), Kamimura (1961) conceived it as a monotypic genus containing only the Japanese endemic . The singularity of this taxon was well described and illustrated, highlighting its closer affinities to instead of , mostly because of its branching pattern and leaf insertion: “[…] the branch replaces the lobule of leaf in origin and the lobe is inserted partly to the stem and partly to the branch. The first leaf and underleaf of branches are much deformed, being the “Vorblätter” of Verdoorn (1930).” (Kamimura 1961, p. 94). The characteristic combination of traits that led Kamimura to describe this new genus vanished when Schuster (1963, 1970) added more species in the circumscription of as explained above. Schuster (1970) still recognized the taxonomic singularity of when placing it in its own subgenus within , but failed to see the relationships of this taxon with other species, precisely because of his wide concept of that includes members of and .

Oil-bodies in PageBreak are homogenous, usually more than ten per cell, and similar in size to chloroplasts (Hattori et al. 1972, Inoue et al. 1981). Hattori et al. (1972) reported 10–20 oil-bodies per leaf lobe median cell for and later Inoue et al. (1981, p. 25) reported a similar number “usually 7–15 per leaf-lobe cell (rarely up to 22)”. Hattori et al. (1972) stated that oil-bodies of are hyaline and homogenous, and Inoue et al. (1981) recorded in their specimen of that the oil-bodies were completely colourless and homogenous. However, they noted that sometimes they were faintly papillose with a few distinct granules; Inoue et al. (1981) were uncertain if this was due to degeneration of the oil-bodies. Reports of oil-body numbers for are ambiguous: although Guerke (1979) and Paton (1999) suggested they range between 3–7 in all taxa, Schuster (1992) stated that the oil-bodies are numerous in the North American material of (≡ ), ranging from 6–16 per cell, and Mizutani (1961) reported 2–10 for Japanese . All authors agree that the oil-bodies in are faintly granular or homogeneous. In , on the other hand, the oil-bodies range between 3–5(7) per cell, are hyaline to somewhat grayish, and are formed by 15–20 internal oil-globules (Mizutani 1961). In the oil-bodies are usually larger, finely to coarsely papillose rather than smooth, and few per cell, with their number generally increasing from the leaf-lobe marginal cells to the basal cells, except in the species that have basal ocelli; however, this number rarely reaches the number of oil-bodies seen in or . The average number of oil-bodies from the 22 species studied by von Konrat (2004) is 4.3 per median lobe cell. One remarkable exception is the North American species (R.M.Schust.) R.M.Schust., which has up to 16 oil-bodies per median cell (von Konrat 2004). A survey of over sixty species (including literature data) suggests that this is a rare condition in the genus (von Konrat 2004). Schuster (1992) described the oil-bodies of as formed of numerous oil-globules and usually appearing coarsely to finely papillose, the only exception being the oil-bodies of , which are smooth and frequently appear as almost homogeneous oil-droplets (von Konrat 2004). The oil-bodies of then appear closer to the other genera in appearence (although smooth, homogeneous oil-bodies are also seen in ) and number, notwithstanding the number reported for and some reports of taxa with fewer oil-body numbers.

Nomenclatural novelties

Kamim., Journal of Japanese Botany 37: 218. 1962.

≡ Stotler & Crand.-Stotl., Memoirs of The New York Botanical Garden 45: 542. 1987 (“”). syn.nov. – Type: S.Hatt.