New insights into the phylogeny and relationships within the worldwide genus Riccardia ( Aneuraceae , Marchantiophytina )

1,2,3,6 Université Pierre et Marie Curie – Sorbonne Universités – Institut de Systématique, Évolution, Biodiversité, ISYEB, UMR 7205, CNRS-MNHN-UPMC-EPHE, Muséum national d’Histoire naturelle, 57 rue Cuvier, CP 39, 75005 Paris, France. 4 Staatliches Museum für Naturkunde Stuttgart, Rosenstein 1, 70191 Stuttgart, Germany. 5 Nees-Institut für Biodiversität der Pflanzen, Rheinische Friedrich-Wilhelms-Universität Bonn, Meckenheimer Allee 170, 53115 Bonn, Germany.

Riccardia species exhibit a simple, pinnate to multi-pinnate thallus, generally without internal differentiation.The development of the gametangia in two rows on short, specialized branches is a synapomorphy of the genus (vs gametangia in several rows in Aneura).The species present great phenotypic plasticity and the lack of constant diagnosis characters has led to difficulties in defining and identifying them.It is widely accepted that Riccardia is one of the most challenging genera of the Marchantiophytina (Hewson 1970;Meenks & Pócs 1985;Schuster 1989;Furuki 1991).
Riccardia has a cosmopolitan distribution but is predominant in the southern hemisphere; only seven species are encountered in the temperate regions of Europe and North America (Schuster 1992;Paton 1999;Grolle & Long 2000).The species colonize moist to rather wet substrates (soil, rock, rotten log, bark) in a wide range of habitats, usually under high atmospheric humidity, and do not tolerate desiccation (Schuster 1992).
The monophyly of Aneuraceae has been well supported in large-scale liverwort phylogenies (e.g., Qiu & Palmer 1999;Forrest et al. 2006;He-Nygrén et al. 2006).The first assumptions on relationships among Aneuraceae were proposed in an investigation of symbiosis between Aneura and fungi (Kottke et al. 2003;Wickett & Goffinet 2008;Bidartondo & Duckett 2010;Krause et al. 2011).The first phylogeny of the family focused on the genus Lobatiriccardia, including a few Riccardia sequences from Ecuador and Europe (Preuβing et al. 2010).However, a phylogeny focusing on the relationships within the large genus Riccardia has not been conducted.
The aim of the present work is to explore the relationships among Riccardia species using a large geographical sampling.This paper is part of a larger study addressing species delimitation, phylogenetic relationships and character evolution within the genus Riccardia, with special reference to Africa.
In this study, we provide support for the monophyly of the genus and highlight several well-supported clades.We also describe a new monospecific genus, Afroriccardia, based on morphological and molecular evidence.

Sampling
We studied 98 samples from Europe, southern South America, Tropical America, Africa, Asia and Oceania (Appendix).Specimens were kindly provided by collectors from all around the world and by several herbaria (AK, BORH, CONN, E, HSNU, KLU, PC; see Acknowledgments).Dates of collections spread from 1973 to 2013.Due to low amplification success, specimens older than about 30 years were not selected for molecular work.

Identification
Each sample was studied in the light microscope after treatment of the thallus with bleach (20%), degrading the cellular content, and followed by coloration with methylene blue (Rico 2011;Reeb & Bardat 2014).This greatly enhanced observations of the anatomical structure of the thallus and facilitated identification.Traceability of observations has been insured by numerous photographs of the plants in toto and in transverse section, taken with a Nikon CoolPix P5000 camera.

Choice of markers
The plastid markers rps4, trnL-F and psbA-trnH, classically used in phylogenetic reconstructions among bryophytes (Quandt & Stech 2004;Preuβing et al. 2010;Carter 2012), were selected because of their small size (< 1000 bp) allowing good amplification success and their potential informative variability at the infra-generic level (Liu et al. 2010;Stech & Quandt 2010).In total, 291 new sequences generated from 98 samples were used in this study: 95 rps4 sequences, 98 trnL-F sequences and 98 psbA-trnH sequences.GenBank accession numbers are given in the Appendix, together with voucher details.

DNA isolation, amplification and sequencing
Prior to extraction, samples were cleaned under the binocular microscope (dry or humidified with distilled water) using tweezers to remove micro-epiphytic leafy liverwort and debris.A few thalli, preferably green, chlorophyll-rich terminal thallus branches, were selected and placed in a 2 ml Eppendorf tube.Two tungsten beads (2 mm) and one volume of pure silica sand (sable de Fontainebleau) were added to the disrupted tissues, which were crushed in 2-3 iterations at 30 Hz for 1 min using Quiagen TissueLyser.DNA was extracted from the resulting powder.For older herbarium specimens (> 10 years), a supplementary CTAB procedure was applied beforehand: 400 µl AP1 lyse buffer + 30 µl CTAB buffer + 30 µl proteinase K were added to each specimen and tubes were placed for 20-24 hours in a thermocycler at 42°C.460 µl of CIA (96 : 4 chloroform : isoamyl alcohol) was added to purify and solubilize remaining impurities.Tubes were gently mixed by inversion and centrifuged for 15 min at 13000 rpm and 4°C.DNA was then extracted following an adapted protocol of Dneasy® Plant Mini Kit Quiagen.Elution was performed in 50 µl of AE buffer and deposited a second time on the membrane of the spin column.

Sequences alignments
Sequence assembly and elimination of primer annealing sites were conducted using PhyDE v. 0.9971 and Geneious v. 6 (Kearse et al. 2012).The whole data set was aligned manually in PhyDE using the data set of Preußing et al. (2010) as a scaffold and applying the criteria laid out in Kelchner (2000).We identified two hairpin associated inversions in the psbA-trnH intergenic spacer (inversion 1: 14 nt stem, 22 nt loop; inversion 2: 23 nt stem, 5 nt loop) both of which were positionally separated in the alignment.Both inversions were included as reverse complemented in the phylogenetic analyses, as discussed in Quandt et al. (2003) and Borsch & Quandt (2009).Variable and parsimony-informative sites were estimated using MEGA v. 5.2 (Tamura et al. 2011).

Molecular species delimitation
Identification of Riccardia species is a challenging exercise and misidentifications are very common among herbarium materials (Reeb & Bardat 2014).Our sampling contained multiple accessions of several morphological species, e.g., R. chamedryfolia (With.)Grolle (13 specimens), R. longispica (Steph.)Pearson (8) and R. fucoidea (Sw.)Schiffn.(6); others were represented by singletons only (R. diminuta Schiffn., R. crenulata Schiffn., etc.; see Appendix).We used molecular species delimitation tools to check the congruence of morphological identifications with genetic signals and to clarify the initial dataset for the phylogenetic analyses while keeping the largest sampling of potential species.
We first used a non-tree based method, ABGD (Automatic Barcode Gap Discovery; Puillandre et al. 2012), not requiring monophyly to propose species delineation, and a tree-based method (Fontaneto et al. 2015), here the Poisson Tree Processes model (PTP; Zhang et al. 2013).We analysed the initial dataset with ABGD in order to test morphological species delimitation, especially for samples identified as the same taxon.The following parameters were selected: distance Kimura-Nei, P = 0.0057 (psbA-trnH and trnL-F) and P = 0.0037 (rps4).Each gene was analysed independently.
We also ran PTP on the bPTP server (http://species.h-its.org/ptp/),with 500 000 MCMC generations, thinning set to 100, burn-in 0.25 and the "remove out-groups" option selected.Input trees were RAxML trees calculated on CIPRES Science Gateway (Miller et al. 2010) under default parameters.Two analyses are provided: PTP_ML (maximum likelihood solution) which gives the most likely solution among the dataset, and PTP_sh (Bayesian solution) which considers the frequency of the nodes across the sampling (Lang et al. 2015).We only retained molecular species with posterior delimitation probabilities higher than 0.91 (Zhang et al. 2013).
All PTP retained species were congruent with ABGD results.Three strategies were selected to keep the largest number of species hypotheses: (1) if two samples assigned to the same morpho-species were considered as separate species with ABGD, we kept the two accessions; (2) if several samples assigned to the same morpho-species were considered as one species with ABGD, we selected the most informative accession (length and sequence quality); (3) if several samples assigned to different morpho-species were considered as the same species with ABGD, identifications were checked; when the initial species hypotheses were confirmed, the accessions of the different morpho-species were retained.Finally, we built a concatenated alignment with the reduced dataset (Appendix) based on ABGD / PTP analyses.

Phylogenetic analyses
File commands for the parsimony ratchet analysis (Nixon 1999) were generated by PRAP2 (Müller 2007) and run in PAUP.4.0 (Swofford 2002) with the following parameters: 10 cycles of 200 iterations each, with 25% positions chosen randomly and overweight to 2. Gaps were coded as missing data.
Branches of minimum size of 0 were automatically collapsed.
The dataset was partitioned a priori on the basis of gene identity, i.e., rps4, psbA-trnH and trnL-F.
For each alignment, the best partitioning scheme and the best nucleotide substitution model were defined using Partition Finder v. 1.1.1 (Lanfear et al. 2012) based on the Akaike Information Criterion.
The GTR + Γ + I model of sequence evolution and the restriction site model (F81) for binary data were selected.Bayesian analysis was performed using MrBayes v. 3.2.6 (Huelsenbeck et al. 2001) on CIPRES Science Gateway (Miller et al. 2010) and 10 Markov Chain Monte Carlo (MCMC) runs with 4 chains (1.5 × 10 6 generations each) were run simultaneously.Chains were sampled every 1000 generations with the respective trees written to a tree file.We visualized the results with Tracer v. 1.6 (Rambaut et al. 2014) to verify the convergence of the runs.Calculation of the consensus tree and the posterior probabilities of clades were performed after removing the burn-in samples (25%).Finally, a 50% majority-rule consensus tree was built in MrBayes.
Consensus topologies and support values were compiled using Inkscape v. 0.91 (The Inkscape Team: https:// inkscape.org).Each name on the tree is formed by the name of the taxon followed by its voucher number.

Sequences and alignments
The final concatenated plastid matrix contained 1882 positions, including 706 conserved sites, 892 variable sites, and 728 informative sites (Table 3).Length variation for each marker is also given in this table.As we did not get full rps4 sequences for all specimens, we excluded the terminal part in order to avoid a high level of missing data.A homo-polynucleotide stretch (position 350 to 356) within P8 of the trnL group I intron (compare with Quandt & Stech 2005) wa been excluded from the phylogenetic analysis due to its ambiguous homology assessment (compare with Kelchner 2000).

Phylogenetic analyses
Five main lineages were detected by both Bayesian inference (Fig. 1) and parsimony ratchet (Fig. 2) within the Aneuraceae: Lobatiriccardia, Verdoornia, Aneura, Riccardia and a fifth lineage proposed as a new genus, Afroriccardia gen.nov.(see below).Verdoornia is sister to all other genera.The clade (2) species hypotheses were scattered through the tree, not forming a monophyletic group (R. aeruginosa Furuki, R. sp8), although this might also due to identification problems.
The origin of our identified samples was congruent with the published distribution of the taxa except for the New Caledonian sample of Riccardia cf.nagasakiensis (Steph.)S.Hatt.The latter species is considered endemic to Japan (Fig. 2).

Description of the new genus Afroriccardia
In the consensus tree Afroriccardia comosa is a strongly supported lineage sister to the genus Riccardia (Figs 1-2).The four samples of this taxon in the initial dataset (Appendix) were confirmed by all species delimitation analyses as a single molecular species.It resembles Lobaticcardia in the broad, pinnately branched thallus and the wide expansion of rhizoids on the ventral face (Furuki 1991;Preußing et al. 2010), and was therefore initially assigned to the latter genus by Reeb & Bardat (2014).Hence Aneura comosa Steph. is not cited in the world checklist (Nebel 2016).However, the species clearly differs from Lobatiriccardia by having long female branches (to 1 cm long) and two regular rows of gametangia.The latter two characters are shared with Riccardia but the dense clusters of rhizoids covering the archegonia clearly separate Afroriccardia comosa from Riccardia.Since the morphological differences with Riccardia are subtle, Afroriccardia comosa could be considered a separate subgenus of the latter.

Diagnosis
Thallus (bi)pinnate.Main axis of thallus 2.5-4.0 mm wide.Rhizoids present over the whole width of the ventral thallus surface.Female branches to 1 mm long, archegonia in pairs, covered by a dense cluster of rhizoids with strongly thick-walled tips originating from beneath the apex of the female branch.

Description
Dioicous.Thallus green, to 7 cm long, main axes 2.5-4.0 mm wide, creeping, ± regularly (bi-)pinnate, with 1-2 reiterations, branches alternate to subopposite, stolons not observed.Rhizoids developing over the whole width of the ventral surface of the thallus.Main axes plano-convex to biconvex, 6-8(-10) cells thick, margin entire, acute to rounded, un-winged, epidermal cells in cross section 1.5-2.0× smaller than medullary cells, all cells thin-walled.Terminal branches to 8 mm long, 0.8-2.0mm wide, 4-5 cells thick, with a conspicuous, 3-4(-6) cells wide wing, branch margins parallel, crenulate, thallus surface cells becoming smaller towards the margin, not or slightly bulging; branch apex rounded to truncate and usually narrowly incised (to 130 µm deep).Mucilage papillae on branches ca 20, present below the apex and in four rows on the ventral branch surface.
Female branches solitary or grouped on main axes and primary branches, 0.5-1.0mm long, archegonia (unfertilized ones seen only) in pairs, covered by a dense cluster of rhizoids originating from beneath the apex of the female branches, rhizoids up to 0.7 mm long, with strongly thick-walled tips.Multicellular paraphyses lacking.Male branches, calyptra and sporophyte not seen.Vegetative reproduction not observed.

Distribution
Afroriccardia comosa is a rare species that was known only from a few old, 19 th century collections from La Réunion and Mauritius; the species is newly reported here from Madagascar and Uganda.The species occurs in evergreen humid forest at mid-montane elevations in Uganda and Madagascar (1100- European Journal of Taxonomy 273: 1-26 (2017) 14 1600 m), and at lower elevations in La Réunion (175-300 m).Where habitat information is available, it was always collected on damp rock surfaces, in shaded places, close to water beds (shaded rivers, entrance of caves with water).

Discussion
Differences in success of amplification for the three markers We observed a significant difference of the PCR success between the three plastid markers (Table 5).
The matching of primers on rps4 sequences allowed us to detect variability near the 3' end of the two primers (F and R).This variability was only found in Riccardia and might affect primer annealing and thus reduce the amplification success.New primers were designed and tested in order to enhance rps4 PCR success (Tables 1-2).Only a part of rps4 could be amplified, for 40% of our final dataset.Improving this marker amplification for classical PCR still remains a challenge.
Most of the samples were herbarium specimens of various ages, dried under unknown conditions.Since these poikilohydrous plants are quite sensitive to humidity variation in their storage area, the DNA of these herbarium specimens can easily degrade.Internal primers generate shorter sequences, allowing amplification of deteriorated DNA.However, it implies an increase of experimentation time and budget, especially in the case of a large dataset.

Species delimitation
Molecular species delimitation using ABGD and PTP is sensitive to balanced sampling, especially in number of replicates per taxon (Puillandre et al. 2012;Zhang et al. 2013;Lang et al. 2015).The initial dataset (Appendix) contained singletons and some taxa with more than ten samples (Riccardia chamedryfolia).Some specimens of the latter species, from the Atlantic islands, Africa and Guadeloupe, were not initially identified as R. chamedryfolia but their proximity was revealed by the molecular analyses.Even if psbA-trnH, trnL-F and rps4 are located in the same area of the plastid genome (Wicke et al. 2011), they have not necessarily evolved in the same way or at the same speed (Preuβing et al. 2010).ABGD was initially a unilocus tool based on calculated barcode gap outside the confidence interval (at 95%) of the population mutation rate θ, given a prior P of maximum intraspecific divergence (Puillandre et al. 2012;Fontaneto et al. 2015).We therefore analysed each gene independently.Although the results of the analysis of each of the three markers were rather similar, the dataset was most finely split with psbA-trnH.On the other hand, analysis with the concatenated markers did not separate all species.However, the latter has to be considered carefully.If the inversions that might occur at the population level (compare with Quandt et al. 2003), as, e.g., observed here in the psbA-trnH intergenic spacer, are not detected, the molecular species delimitation using ABGD and PTP will return a wrong concept (data not shown).
With PTP, even though the convergence of the runs was moderate to good, only singleton species were highly supported (> 0.91).Changing of parameters did not improve the results.It is possible that the PTP results were affected by unbalanced sampling and missing data.
Our phylogenetic results show that at least two species, Riccardia aeruginosa and R. sp8, are not monophyletic (Figs 1-2).These results may indicate that (1) samples were misidentified, (2) samples may represent undescribed species, (3) PCR contamination occurred, or (4) the species is paraphyletic.The latter case may be verified with tools such as Haplowebs (Fontaneto et al. 2015).Non-monophyly of species is frequently detected in bryophytes and molecular analyses are an important tool to reveal the existence of morphologically distinct species that would otherwise have remained undetected (e.g., Sukkharak et al. 2011;Hutsemékers et al. 2012;Aranda et al. 2014;Hedenäs et al. 2014;Heinrichs et al. 2015).
Some samples morphologically identified as the same species (Riccardia alcicornis, R. elata, R. sp14, R. chamedryfolia, R. longispica, R. conimitra, R. stipatiflora) were separated by at least one ABGD analysis but appeared to form a monophyletic group in all phylogenetic analyses.This could be due to (1) genetic variations among species and/or (2) high sensibility of the marker psbA-trnH, on which these delimitations were based.
The results indicate that the holarctic Riccardia incurvata Lindb., R. multifida (L.) Gray and R. palmata (Hedw.)Carruth.form a clade together with several African, Asian and Australasian species, but that the largely holarctic R. chamedryfolia is not a member of this clade (see also Preußing et al. 2010).Riccardia chamedryfolia is more widespread in the tropics and it seems to often be overlooked or misidentified The clade containing Riccardia amazonica (Spruce) Schiffn.ex Gradst.& Hekking, R. longispica, R. sp8 and R. cataractarum (Spruce) Schiffn.includes species from warm, low elevation areas of tropical Africa and tropical America.It is an interesting and somewhat puzzling group because of the strong polymorphy of some of the species in this group, contrasting with close genetic distances (Reeb unpubl. res.).All authors agree that R. amazonica is an Afro-American species (e.g., Meenks & Pócs 1985;Wigginton 2004;Perold 2003; but see Gradstein 2013).However, this is not supported by the experiment, showing that the spores of R. amazonica lose their capacity of germination after a few hours (Van Zanten & Gradstein 1988;Gradstein 2013), making successful long-distance-dispersal by spores unlikely.Some authors suggest polyploïdy as a possible explanation of the large range of morphological variability in R. amazonica (Berrie 1966).A closer look at this species is needed to improve our understanding of the delimitation and biogeography of this widely distributed and highly variable taxon.

Infrageneric placement
Subgeneric and sectional attribution of the species, following Nebel ( 2016) is shown on the consensus tree (Fig. 1).It appears that only two subgenera, subg.Arceoneura (R. prehensilis (Hook.f.& Taylor) C.Massal.) and subg.Riccardia (11 spp.) and three sections of subg.Riccardia (sects Alcicornia, Crassantia and Riccardia, each with 2 spp.) are represented in this study.The subgeneric or sectional placement of the great majority of Riccardia species analysed in this study is uncertain ("incertae sedis").Therefore, only limited conclusions can be drawn here on the infrageneric placement of Riccardia species.The data indicate that the southern temperate sect.Alcicornia, represented here by R. alcicornis and R. conomitra, and the circumpacific sect.Crassantia (R. crassa (Schwägr.)Carrington & Pearson and R. graeffii (Steph.)Hewson) are polyphyletic because the species of these sections are placed in different lineages in the phylogeny.The two species in sect.Riccardia, R. multifida (type of the genus Riccardia) and R. filicina (Colenso) E.A.Hodgs., are nested in a clade together with three unclassified members of subg.Riccardia (R. aeruginosa, R. nagasakiensis and R. palmata) and four members of the incertae sedis group (R. crenulata, R. diminuta, R. elata and R. incurvata).This suggests that the latter four species belong in subg.Riccardia.The placement of R. eriocaula (Hook.)Besch.& C.Massal. and R. chamedryfolia in subg.Riccardia (Nebel 2016), is not supported by our phylogeny.Riccardia eriocaula is a morphologically highly unusual species that was placed in subg.Arceoneura by Brown & Braggins (1989).
Although our sampling has been insufficient to evaluate the infrageneric classification of Riccardia, these first results are suggestive of the very incomplete state of knowledge of the relationships of species within this large genus.A broader sampling of the genus, including representatives of the subgenera not analysed here, is needed to arrive at a better understanding of its phylogeny.In addition, revisions of species at continental and worldwide scales should be carried out, using an integrative taxonomy approach.In future, we plan to extend our sampling and use additional markers, including nuclear ones, in order to produce a more complete phylogeny, including reconstruction of character evolution in the worldwide genus Riccardia.Appendix.List of specimens, with country of origin (norm ISO 3166-1 alpha-3), family, ABGD partition, bPTP partition, species based on morphological identification, voucher number, herbarium acronym (see Thiers continuously

Fig. 1 .
Fig.1.The 50% majority-rule consensus of the trees generated by Bayesian Inference (BI) of the combined psbA-trnH, rps4 and trnL-F dataset.Posterior probability percentages above 80% are indicated under each node.Species names appear in italics, followed by the corresponding voucher number.Outgroup labels appear in light grey.Infra-generic divisions (sub-genera and sections) of Riccardia according to Nebel(2016) are reported in front of each name.The corresponding colour scheme for each sub-generic division is shown in the bottom left proportion of the figure.

Fig. 2 .
Fig. 2.This topology shows the result of the maximum parsimony ratchet.Bootstrap proportions above 80% are indicated under each node.Species names appear in italics, followed by the corresponding voucher number.Outgroup labels appear in light grey.The geographical origin of each Riccardia or Afroriccardia sample is marked by bold rectangles.The continental/geographical repartition of each species according to the literature is indicated with rectangles coloured by continents: Europe (purple), North America (red), southern South America (orange), tropical America (brown), Africa (yellow), Asia (green) and Australasia (blue).A grey rectangle indicates new continental records for the species.

Table 2 .
List of PCR protocols used for each primer couple.

Table 3 .
Number of sequences obtained and number of informative sites for each marker.All the numbers refer to the aligned matrix, except "Range of amplicon size", which refers to unaligned sequences.rps4F3 refers to the part of rps4 obtained with the newly designed primers.trnL-F information is detailed for the two internal portions often sequenced and concatenated for older samples.Percentage values are indicated in brackets next to absolute values for conserved, variable, parsim-info and singleton sites.
including Aneura and Lobatiriccardia is sister to the clade formed by Afroriccardia and Riccardia.The monophyly of each genus was strongly supported in all analyses.Within the genus Riccardia, clades were usually very well supported, with few exceptions.In the case of conserved species hypotheses with the same name, two cases were observed: (1) all hypotheses formed a clade (R. alcicornis (Hook.f.& Taylor) Trevis, R. conimitra (Steph.)A.Evans, R. elata (Steph.)Schiffn., R. pallida (Spruce) Meenks & C.De Jong, R. stipatiflora (Steph.)Pagan);

Table 4 .
Comparison of results given by three different methods of species delimitation used for our study in terms of number of species hypotheses: ABGD, PTP and morphological delimitation.
Monospecific, contains only A. comosa from Madagascar, the Mascarene Islands and Uganda.

Table 5 .
Overview of PCR success depending on primer couples employed.