Frullania knightbridgei, a new liverwort (Frullaniaceae, Marchantiophyta) species from the deep south of Aotearoa-New Zealand based on an integrated evidence-based approach

Abstract Frullania is a large and taxonomically complex genus. A new liverwort species, Frullania knightbridgei sp. nov. from southern New Zealand, is described and illustrated. The new species, and its placement in Frullania subg. Microfrullania, is based on an integrated evidence-based approach derived from morphology, ecology, experimental growth studies of plasticity, as well as sequence data. Diagnostic characters associated with the leaf and lobule cell-wall anatomy, oil bodies, and spore ultra-structure distinguish it from all other New Zealand species of Frullania. A critical comparison is also made between Frullania knightbridgei and morphologically allied species of botanical regions outside the New Zealand region and an artificial key is provided. The new species is similar to some forms of the widespread Australasian species, Frullania rostrata, but has unique characters associated with the lobule and oil bodies. Frullania knightbridgei is remarkably interesting in comparison with the majority of Frullania species, and indeed liverworts in general, in that it is at least partially halotolerant. Maximum parsimony and maximum likelihood analyses of nuclear ribosomal ITS2 and plastidic trnL-trnF sequences from purported related speciesconfirms its independent taxonomic status and corroborates its placement within Frullania subg. Microfrullania.


Introduction
Frullania Raddi (Frullaniaceae) is a large and complex liverwort (Marchantiophyta) genus with a worldwide distribution (Yuzawa 1991, von Konrat andBraggins 2001a). Th e number of published Frullania names has been reported to be over 2000 . Published estimates of the number of accepted species for the genus that have become widely recognized range from 300-375 accepted species (e.g., Schuster 1992, Gradstein et al. 2001. However, there is no evidence whatsoever to support these suppositions as no worldwide monographic treatment of Frullania has ever been attempted (von Konrat et al. 2006a. On the contrary, growing data and evidence may suggest the number of 300-375 species is a minimum estimate at best ; in some cases the underestimation of the Frullania species diversity has been attributed to conservative morphology within species complexes . In New Zealand, the current number of accepted taxa is 33, including 31 species and two varieties of which 10 are apparently endemic (Hattori 1979a, b;1983;Glenny 1998;von Konrat and Braggins 2005;von Konrat et al. 2006b;von Konrat et al. 2010von Konrat et al. , 2011. Here we present a study of a newly discovered species that is morphologically close to Frullania rostrata (Hook. f. et Taylor) Hook. f. et Taylor, which is considered a widespread, polymorphic and common Australasian species (von Konrat et al. 2006b). Th e new species would be lumped under F. rostrata based on overall gross morphology.
Our paper is part of a broader, on-going, regional study of Frullania species by us that includes the botanical regions of New Zealand, Australia, the Pacifi c, South East Asia, and South America. Our new species occurs on Stewart Island/Rakiura and the Auckland Islands group of the New Zealand botanical region (as defi ned by de Lange and Rolfe 2010). Both areas are regarded as extremely biologically and biogeographically signifi cant (Given and Hnatiuk 1995;McGlone and Wilson 1996;Wilson 1987;Wagstaff et al. 2011). Stewart Island/Rakiura, is the southernmost and third largest island of the New Zealand archipelago, with about 85% of the island comprising Rakiura National Park (Heenan et al. 2009). Although the fauna and fl ora has been partially modifi ed as a consequence of historical Maori and European settlement, the island's vegetation patterns are little altered from their pre-human state (Wilson 1987). On the other hand, the Auckland Island group is one of fi ve New Zealand sub-Antarctic island groups (including Snares, Bounty, Antipodes, Auckland, and Campbell islands) and are all World Heritage Areas (Chown et al. 2008). Th e distinctive fl ora of the subantarctic islands includes some of the last remnants of a once-diverse Antarctic fl ora, with examples of many plants possibly still retaining distinctive features of their ancestors (Wagstaff et al. 2011).
Th e new species of Frullania described below is assigned to Frullania subg. Microfrullania (R.M.Schust.) R.M.Schust., which is confi rmed by molecular evidence. Detailed microscopic and SEM micrographs as well as a brief comparison with morphologically similar species are provided. Th is new species is remarkable in comparison with the majority of Frullania species in New Zealand for its least partial tolerance and exposure to salt spray. For consistency and clarity through this article, the results and discussion that follows refers to the newly recognised species as Frullania knightbridgei.

Living material
In order to monitor the stability of character-states and assess whether some characters varied independently of the genotype, material was grown under uniform environmental conditions described by von Konrat and Braggins (2001a). Th e same individual colonies were grown under three diff erent light and water regimes and characters subsequently monitored for stability or variation. Voucher specimens of cultivated material are deposited at AK and F.

Morphological study
Where necessary, plant material was cleared to remove pigmentation using the method outlined by von Konrat and Braggins (2001a) and the cell layers of the capsule wall were separated as described by von Konrat et al. (1999). Microscopy techniques, measurements, the use of descriptors to indicate abundance and frequency, terminology of spore ornamentation, preparation of material (including spores for the SEM studies) are outlined in detail by von Braggins (2001b) andvon Konrat et al. (2006a, b).

DNA extraction, PCR amplification and sequencing
Dried tissue was disrupted with the aid of a sterile steel bead in a Qiagen tissuelyser (Qiagen Inc. Hilden, Germany) set at 30 Hz for 45 s. Genomic DNA was extracted and purifi ed using an Invisorb Spin Plant Mini Kit (Invitek, Berlin, Germany) according to the manufacturer's specifi cations. Two molecular markers, the internal transcribed spacer 2 of nuclear ribosomal DNA, and the plastidic trnL-F region were sequenced using the primer sets presented by Hartmann et al. (2006) and Gradstein et al. (2006). Approximately 525 base pairs (bp) of the 5.8S-nrITS2 region were sequenced, along with 500 bp of the trnL-trnF region per isolate. PCR for each sample was performed in a total of 25 μl and contained 2.5 μl dntp mix, 2.5 μl MgCl 2 , 5 μl of Bovine Serum Albumin (New England Biolabs, Ipswich, Massachusetts, USA) 1 μl of the forward primer, 1 μl of the reverse primer, 0.5 μl Taq (Roche diagnostics, Indianapolis, Indiana, USA), 10.5 μl of dH 2 0, and 2 ul of sample DNA. PCRs were run for 37 cycles in a Dyad DNA engine (Bio-Rad Laboratories, Inc., Hercules, California, USA) set to the following parameters: initial denaturation at 94°C for 2 min, denaturation at 94°C for 1 min, annealing at 55°C for 1 min, extension at 72°C for 1 min, then fi nal extension at 74°C for 7 min followed by a cool down stage at 4°C. Th e amplicon was purifi ed using a Nucleofast 96 well PCR plate (Macherey-Nagel, Evanton, Pennsylvania, USA). Cycle sequencing was performed using the same primer sets as for the PCRs. Sequencing reactions were done using the BigDye terminator cycle sequencing kit (Applied Biosystems, Foster City, California, USA) and amplicons were run on an ABI 3730 (Amersham Pharmacia Biotech Inc., Piscataway, New Jersey, USA). A consensus sequence was constructed and edited using Sequencher version 4.10 (Gene Codes Corp., Ann Arbor, Michigan, USA).

Taxon sampling and outgroup selection
Initially, the new sequences were compared with GenBank sequences using the BLASTN program (Altschul et al. 1990). Th e BLAST searches confi rmed the position within Frullania subg. Microfrullania. Ingroup taxa representing representatives of this subgenus were selected to test taxonomic hypotheses based on morphology. Several accessions of F. rostrata were included because F. knightbridgei shares several morphological characters with this morphologically rather variable taxon. Based on the analyses of Hentschel et al. (2009), three representatives of F. subg. Th yopsiella [F. asagrayana Mont., F. microphylla (Gottsche) Pearson, F. tamarisci (L.) Dumort.] were designated as outgroup taxa for phylogenetic reconstruction. Taxa studied, including GenBank accession numbers and voucher details, are listed in Table 1.

Phylogenetic analyses
All sequences were aligned manually in Bioedit version 7.0.5.2 (Hall 1999). Ambiguous positions were excluded from the alignment and lacking parts of sequences were coded as missing. Maximum parsimony (MP) and maximum likelihood (ML) analyses were carried out with PAUP* version 4.0b10 (Swoff ord 2000).
MP heuristic searches were conducted with the following options: heuristic search mode, 1.000 random-addition-sequence replicates, tree bisection-reconnection (TBR) branch swapping, MULTrees option on, and collapse zero-length branches off . All characters were treated as equally weighted and unordered. Non-parametric bootstrapping values (Felsenstein 1985) were generated as heuristic searches with 1.000 replicates, each with ten random-addition replicates. Th e number of rearrangements was restricted to ten million per replicate. Bootstrap percentage values (BP) above 70 were regarded as good support (Hillis and Bull, 1993). Where more than one most parsimonious tree was found, trees were summarised in a strict consensus tree. Th e two genomic regions were fi rst analysed separately to check for incongruence. Th e strict consensus trees of the non-parametric bootstrap analyses were compared by eye to identify confl icting nodes supported by at least 70% (Mason-Gamer and Kellog 1996). Th e trees gave no evidence of incongruence. Hence the datasets were combined. jModeltest 0.1 (Posada 2008) was used to select the TIM2 + G model of evolution for the ML analysis of the combined dataset. Th e analysis was performed as heuristic search using ten random-sequence addition replicates, MULTrees option on, collapse zero length branches off , and TBR branch swapping. Th e confi dence of branching was assessed with PAUP* using 200 non-parametric bootstrap resamplings generated as heuristic searches.

Species concept
Although the determination of species is regarded as one of the most important activities of the taxonomist, the majority of systematists undertaking monographs and revisions of vascular plants do not discuss the concepts or the criteria to delimit species (McDade 1995). A similar statement can undoubtedly be applied to liverwort systematics (von Konrat et al. 2006a. Here, we adopt a hierarchical model as promoted by Mayden (1997). Th is model considers the Evolutionary Species Concept as a theoretically robust primary species concept, as well as a bridging, secondary or operational species concept. Th is is discussed in the context of Frullania by von Konrat et al. (2006a).

Data resources
Th e occurrence data underpinning the analysis has been uploaded as a Darwin Core Archive (DwC-A), to the Global Biodiversity Information Facility (GBIF) via the Pensoft Data Hosting Center at the GBIF's Integrated Publishing Toolkit (IPT) (http://ipt.pensoft.net/ipt/manage/resource.do?r=deep_south_frullania_species). Th e genomic sequences are deposited at GenBank and their hyperlinked accession numbers are listed in Table 1.
In addition to the current paper semantically tagged and enhanced using the Pensoft Mark Up Tool (PMT), repository data and images, including images with zoom capability can also be accessed at www.discoverlife.org and www.symbiota.org for selected species that are closely allied to the newly described species. Th e Consortium of North American Bryophyte Herbaria (CNABH) was created to serve as a gateway to distribute data resources of interest to the taxonomic and environmental research community, off ering a common web interface, including tools to locate, access and work with a variety of data (see http://symbiota.org/bryophytes/index.php).

Results and discussion
In the present study, hypotheses of species diff erences are based on support from multiple lines of evidence, including morphology, experimental growth studies, and nucleotide sequences. Th is is discussed below.

Phylogeny
Of a total of 964 molecular characters, 127 were parsimony informative, 62 autapomorphic, and 775 constant (Table 2). Th e MP analysis resulted in two trees of 304 steps with a consistency index of 0.78 and a retention index of 0.80 (not depicted). A single most likely tree was found in the ML analysis ( Fig. 1 Frullania rostrata is split in two robust subclades. Hence, the ML phylogeny indicates that F. rostrata -despite exclusion of F. knightbridgei -is part of a species complex, possibly with some geographical structure. Th is may be supported by the signifi cant number of synonyms and herbarium specimens summarised under F. rostrata (von Konrat et al. 2006a). Frullania rostrata might well be regarded as a Southern Hemisphere equivalent of the Holarctic Frullania tamarisci. Heinrichs et al. (2010) investigated F. tamarisci, which is typically regarded as a single polymorphic species. Using sequences from the nrITS region and plastid trnL-trnF and atpB-rbcL, their analyses resolved eight partly sympatric monophyletic groups representing distinct species rather than subspecies or varieties.
Th e number of molecular studies at the population level in liverworts is still limited. Th is hampers our eff orts to quantify the contribution of cryptic species to the global biodiversity of liverworts . Existing studies suggest a signifi cant part of bryophyte biodiversity is undetected with traditional morphological concepts alone ). It is clear, we urgently need more species-level phylogenies with extensive population sampling to approximate the actual diversity of Frullania, and to elucidate speciation processes and distribution range formation , Ramaiya et al. 2010.

Growth studies
In Frullania, as well as liverworts generally, there remains a large gap between characters used for delimitation and our understanding and knowledge of their plasticity in nature (von Konrat et al. 2006b). Frullania rostrata and the new species, F. knightbridgei, responded well to growing in controlled environmental conditions in a glasshouse unit. Oil bodies in particular were monitored. Th e stability of oil body characters indicates that the diff erences have some underlying genetic basis; thus it is likely that the salient characters of this species are genetically dependent rather than infl uenced by the environment.

Morphology
Many critical morphological features have often been neglected in liverwort systematics (Schuster 1992;von Konrat et al. 1999von Konrat et al. , 2001avon Konrat et al. , 2006b, and scores of studies have been restricted to herbarium material where ephemeral structures; e.g., sporophytes and oil bodies, have not been available (von Konrat et al. 2006a,b;Heinrichs et al. 2009). Th e new species is morphologically aligned to a group of species representing F. subg. Microfrullania, which has been resolved as a monophyletic group ). Inclusion of F. knightbridgei in F. subg. Microfrullania is also supported by molecular evidence as discussed above. Frullania subg. Microfrullania represents a clade with the most historical confusion out of all Frullania subgenera with taxa occurring in southern South America, Australasia and islands of the South Pacifi c, New Guinea, and Indonesia (von Konrat et al. 2006a. Th e new species appears almost to lie intermediate between F. rostrata, of New Zealand and Australia, and F. pseudomeyeniana S. Hatt. of New Caledonia. Th e latter is only known from the type material (New Caledonia, Mont Mou, N of Paita, 1100 m., Kitagawa 21422, NICH), which was examined by the senior author. Frullania knightbridgei also has some similarity with F. magellanica (Spreng.) F. Weber et Nees of Chile. Frullania knightbridgei superfi cially strongly resembles some forms of F. rostrata in plant size, the large styli and lobules, and the entire underleaves. However, with fresh material, F. knightbridgei is immediately discernable from F. rostrata by the presence of large oil bodies (usually only 2 per cell, occasionally 1 or 3) that almost occupy the entire cell lumen of basal and median cells of the leaf lobe ( Fig. 2a-b). In the absence of oil body data, careful consideration has to be given to the anatomy of the leaf-lobe and -lobule to help diff erentiate between these species. In F. knightbridgei, cells towards the lobule apex progressively develop a more regular shape (quadrate to rectangular) and the cell walls become semi-straight (Fig. 2e). Conversely, the cell walls of both F. rostrata and F. pseudomeyeniana are fl exuose with indistinct trigones, and with small, nodulose intermediate thickenings throughout the lobule, from the base to the apex (Fig. 2f ).
Th e unique cell anatomy of the leaf lobe in F. knightbridgei further places it into an isolated position within subg. Microfrullania; this species is seemingly unique in having a group of conspicuously enlarged cells, originating from the base of the lobe and extending 10-14 cells out toward the apex, forming a partial band or pseudo vitta up to 4-6 cells wide (Fig. 3). Th e cells are enlarged to accommodate the typically 2 large oil bodies. Th e features of the oil bodies are unique within F. subg. Microfrullania. In those species examined thus far, the oil bodies of the leaf lobe median cells number from 2-4(5) per cell, are of small size and lack any signifi cant ornamentation, almost appearing as homogeneous oil droplets (von Konrat et al. 2006a (Fig. 2c-d). Th e position of the lobules in relation to the stem as well as styli shape and form are often used to help distinguish between taxonomic units of varying levels in Frullania taxonomy. Lobule position varies in F. knightbridgei, ranging from parallel to subparallel with the stem or with the lobule spreading at a strong angle, so that the lobuli are tilted inwards. Frullania pseudomeyeniana and some phenotypes of F. rostrata also have lobuli that lie almost parallel or subparallel to the stem for both species. Frullania magellanica also has at least some phases with lobules more or less parallel to the stem (Engel 1978). Interestingly, the parallel lobule position is typically a feature associated with species of subg. Th yopsiella. Th us lobule position must be used secondary to and in collaboration with more salient characters in circumscribing F. subg. Microfrullania.
Historically, characters associated with the capsule wall and spore surface ultrastructure have rarely been utilized in Frullania systematics (von Konrat et al. 2006b). Yet, it is clear that characters associated with these structures have great utility at various taxonomic levels (e.g., von Konrat et al. 1999Konrat et al. , 2006b. Th e spores of F. knightbridgei have a "rosette" with conspicuous protuberances lacking secondary branches and deposits -a feature used to help characterize F. subg. Microfrullania ). Th e spores can also be used to distinguish Frullania knightbridgei and F. rostrata (Table 3, Fig. 5). Diff erences are also refl ected in the epidermal wall of the capsule. In F. knightbridgei, the walls are nodular, where the lobes extend irregularly for a short distance over the tangential face toward the centre of the cell and have intermediate thickenings near the middle of the longer walls (Fig. 5a). In F. rostrata, the lobes extend toward the centre of the tangential face for a short distance, as short rounded or obtuse lobes and the juxtaposed corner thickenings form 3-4 lobed fi gures; intermediate thickenings are also lacking (Fig. 5b).
Tables 3-5 summarizes the characters diff erentiating F. knightbridgei from two morphologically similar species that it might be confused within the New Zealand bo-
Leaf-lobe: to 20 cells long, from base to apex, by 35 cells at widest region; with a band of conspicuously enlarged cells originating from the lobe base and extending out towards the lobe apex 10-12 cells, and up to 6 cells wide at the widest region. Lobe marginal cells ± rectangular to subquadrate, small to 8 μm long × 6 μm wide, hyaline walls subequally thickened, cell cavities brownish red. Cells of the middle region of the lobe are ± dimorphic in size; Type One [see below]: 4-6 rows of median cells, cells to 30 μm long × 22.5 μm wide (usually 2-2.75 × long as wide), thus similar in size to basal cells; Type Two [see below]: cells gradually becoming reduced in size (median cells to 15 μm long × 10 μm wide, usually 1.25-2 × long as wide, between central band of enlarged cells and lobe margin). Both cell types usually pentagonal or hexagonal, hyaline walls subequally thickened, intermediate thickening rare to absent, wall thickness to 2.75 μm wide, cell cavities of median cells brownish red. Cells becoming gradually larger basally, cavities of the basal cells to 40 μm long × 25 μm wide; walls of basal cells with small indistinct trigones and semi-straight walls without any intermediate thickenings, walls and cavities brownish red. Median cells of underleaves vary in shape and size, cells with heavily equally-thickened walls so that the hyaline trigones and intermediate thickenings become indistinct. Median cells of lobule as long as wide or slightly longer than wide, cell cavities to 14 μm long × 9 μm wide; cells near lobule mouth, irregular in shape with fl exuose walls, indistinct trigones and occasional small nodulose intermediate thickenings; towards the lobule apex, cells gradually becoming more regular in shape, quadrate to rectangular and the cell walls becoming semi-straight.
Walls of epidermal layer of capsule wall nodular where the lobes extend irregularly for a short distance over the tangential face toward the centre of the cell and have intermediate thickenings near the middle of the longer walls. Spores globose, 35-45 μm at widest axis, spore wall papillae densely distributed, equatorial face interspersed with 8-10 rosettes, 2.5-3 μm wide in the equatorial diameter, bearing a ring of 7-10 conspicuous primary projections, 0.75-1.5 μm long × 0.5-1.0 μm wide at base (often with a 1.5-2:1 length to width ratio), tapering gradually to a rounded apex, never papillate or bearing secondary short branches.
Etymology. Th e species epithet knightbridgei is named in honour and memory of an esteemed New Zealand conservation botanist and ecologist, Phil (Philip) Ian Knightbridge (1969Knightbridge ( -2011 who passed away in April 2011. Th is southern species of Aotearoa-New Zealand is a small tribute to Phil by the senior authors who knew him as a colleague and friend.
Distribution and ecology. Th is species is currently known from only four collections; two from the shore margin of Paterson Inlet, Stewart Island and two from the Auckland Islands. Frullania knightbridgei probably has a more extensive distribution than is currently known and it is quite likely that it resides unrecognized in New Zealand herbaria fi led as a form of F. rostrata. Nevertheless, it would appear that F. knightbridgei is a species of southern distribution and it would be interesting to see if the distribution extends south to the Campbell Islands of the New Zealand botanical region; further fi eld work is required to establish if the species grows on the other two main islands of the New Zealand archipelago, North and South Islands, it should for example be looked for along the Fiordland and Foveaux Strait coastline of the southern South Island. Note the type of F. pseudomeyeniana is a high elevation taxon at 1,100 m whereas the New Zealand taxon is seemingly restricted to or near the shoreline or low elevation.
F. knightbridgei is noteworthy in comparison with the majority of F. species in New Zealand for it appears to be a salt tolerant species. Th is is clearly evident in Stewart Island/Rakiura where F. knightbridgei was growing on twigs of Dracophyllum immediately adjacent the shoreline, at a height of about 50 cm above the sea. It was evident that for at least some periods of the year, this represented a harsh coastal environment where signifi cant exposure to salt spray from violent wave action would be common. Elsewhere, F. ericoides reportedly develops in rock crevices exposed to sea in the Madeiran archipelago (Sim-Sim and Sergio 1992), and Schuster (1992) reported a variety of F. kunzei growing in mature mangrove swamp forest where salt spray and even rare submersion was possible. Th e high rainfall of the region, providing fresh water, is possibly a critical factor as Engel and Schuster (1973) described for tidal zone liverworts in southern Chile. Interestingly however, Engel and Schuster (1973) noted for Stewart Island a notable "lack of any Hepaticae in the intertidal zone where sea spray is a factor". At least in the New Zealand botanical region, it is clear this is a habitat area that has traditionally been underexplored for liverworts. It would also be interesting to perform glasshouse experiments to investigate test the extent of salt tolerance in these organisms.