Recognition of the genus Thaumatophyllum Schott − formerly Philodendron subg. Meconostigma (Araceae) − based on molecular and morphological evidence

Abstract Philodendron subgenus Meconostigma has been a well-circumscribed group since 1829. Members of this group are easily distinguished by diagnostic morphological characters as well as by a distinct ecology and geographical distribution. Based on molecular, morphological and cytological evidence, we propose the recognition of P. subg. Meconostigma as a distinct genus, Thaumatophyllum Schott. We also present the necessary new combinations, an emended key and some nomenclatural and taxonomic corrections regarding 21 names of Thaumatophyllum.


Introduction
Philodendron Schott is the second most species-rich and diverse genus in the family Araceae and also in the "Homalomena clade" (sensu Cusimano et al. 2011), comprising 487 formally recognised species (Boyce and Croat 2018). The genus ranges from northern Mexico to southern Uruguay (Mayo et al. 1997), most commonly in tropical humid forests as epiphytes and hemi-epiphytes. Most rarely, it also occurs as terrestrial plants in open habitats (e.g. seasonal dry forests of South America). Cabrera et al. (2008) published a family-wide molecular phylogeny that included species from 102 genera. Cusimano et al. (2011) re-analysed and augmented a molecular data set with a more complete genus sampling and compared the resulting phylogeny with morphological and anatomical data, proposing informal names for the suprageneric clades. The "Homalomena clade" (composed of the genera Adelonema Schott, Cercestis Schott, Culcasia P.Beauv., Furtadoa M.Hotta, Homalomena Schott and Philodendron Schott) was recovered in both molecular and morphological analyses and was supported by the occurrence of sclerotic hypodermis and resin canals in the roots and absence of endothecial thickenings in the anthers (present in Homalomena). The clade is composed of two sister groups: "Culcasieae clade" (Cercestis, Culcasia) and "Philodendron clade" (Furtadoa, Homalomena, Philodendron). Mayo et al. (2013) gave an alphabetical table of the clades that is a useful complement to the listing in Cusimano et al. (2011).
The evolutionary history of the "Philodendron clade" has been discussed in several recent papers (Tam et al. 2004, Gauthier et al. 2008, Mayo et al. 2013, Wong et al. 2013, Loss-Oliveira et al. 2014, Wong et al. 2016, Loss-Oliveira et al. 2016, as well as the relationship amongst the three subgenera of Philodendron as independent lineages (Gauthier et al. 2008, Loss-Oliveira et al. 2014, Loss-Oliveira et al. 2016. A question recently answered was how the Asian-Neotropical distribution of the genus Homalomena originates (sensu Mayo et al. 1997). Based on molecular evidence (Gauthier et al. 2008, Wong et al. 2013, Wong et al. 2016, the American species of Homalomena were recognised as a separate lineage and consequently Schott's old genus Adelonema was recognised once more (Wong et al. 2016). The "Philodendron clade", still needs better phylogenetic resolution for two other lineages: Homalomena + Furtadoa and Philodendron subgenera Philodendron and Pteromischum. Several research articles (Wong et al. 2016, Loss-Oliveira et al. 2016 have proposed different hypotheses for the relationship amongst these lineages as summarised in Fig. 1. The recent recognition of the genus Adelonema for the American species of Homalomena (Wong et al. 2016) makes the genus Philodendron paraphyletic in some of the current proposed phylogenetic hypothesis (Figs 1B,3A). The most recent studies (Loss-Oliveira et al. 2016, Vasconcelos 2015 recovered two major lineages: P. subg. Meconostigma (= Thaumatophyllum) and Philodendron subg. Philodendron plus subg. Pteromischum. Vasconcelos (2015) recovered P. subgenus Pteromischum as monophyletic and sister clade to P. subg. Philodendron.  Gauthier et al. (2008), maximum parsimony. B Gauthier et al. (2008), Bayesian analysis; Wong et al. (2013), Wong et al. (2016).
Philodendron subgenus Meconostigma (= Thaumatophyllum) has been a well-circumscribed group since the genus Philodendron was first recognised taxonomically by Schott (1829). It is now comparatively well-studied taxonomically; the last taxonomic revision included 15 species (Mayo 1991, with updates by Gonçalves andSalviani 2002, who recognised 19 species). Members of this subgenus are easily distinguished by diagnostic morphological characters as well as by a distinct ecology and geographical distribution that ranges from the Amazonian and Atlantic forests to the savannahlike landscapes of the Cerrado biome (Mayo 1991).
Based on the evidence now accumulated (most recently, by Calazans et al. 2014, Loss-Oliveira et al. 2014, we propose the recognition of P. sugb. Meconostigma as a distinct genus, Thaumatophyllum Schott (1859), a taxon that was accepted by experts as recently as Bunting (1980). Barroso (1962) was the first botanist to formally assign the name Thaumatophyllum to the synonymy of Philodendron and Mayo and Barroso (1979) gave a detailed account of the confusion that had misled previous authors regarding the floral morphology of T. spruceanum. The aim of this paper is, therefore, to formally propose this change of status and validly publish the necessary new combinations. We also provide an emended key and some nomenclatural and taxonomic corrections concerning six names in this genus.

Taxon and gene sampling
We have sampled data for 110 extant species of Philodendron, 21 species of Thaumatophyllum and six species of Homalomena and five of Adelonema of the nuclear 18S and external transcribed spacer (ETS) and the chloroplast trnK intron, maturase K (matK) genes, trnL intron, trnL-trnF intergenic spacer. Species from the genera Cercestis, Culcasia, Colocasia, Dieffenbachia, Heteropsis, Montrichardia, Nephthytis, Furtadoa and Urospatha were included as the outgroup. The species list, the voucher and GenBank accession numbers are listed in Suppl. material 1: table 1. The majority of the used sequences were generated by a previous study of our group (Loss-Oliveira et al. 2016).
Additionally, we generated a subsampled dataset comprised of species from our original data with available ETS and 18S sequences and at least two available chloroplast sequences. This strategy aimed to reduce the impact of missing data in the concatenated analysis. This taxon sampling is described in Suppl. material 1: table 1.

Alignment and phylogenetic analysis
The methodological approach of Loss-Oliveira et al. (2016) was followed in order to estimate individual gene trees and a supertree. We have used MAFFT 7 (Katoh and Standley 2013) to individually align the molecular markers and SeaView 4 (Gouy et al. 2010) to manually adjust them. Bayesian analysis was conducted in MrBayes 3.2.2 Ronquist 2001, Ronquist andHuelsenbeck 2003) for individual gene trees (Fig. 1, Suppl. material 1) using the GTR + G substitution model. The Markov chain Monte Carlo (MCMC) algorithm was run twice for 10,000,000 generations with four chains, which were sampled every 100 th cycle. We have applied a burnin of 20%. Individual gene trees were used to estimate a supertree with PhySIC_IST algorithm (http://www.atgc-montpellier.fr/physic_ist/) in order to avoid the impact of missing data in the estimation (Scornavacca et al. 2008).
We have also performed phylogenetic analysis for concatenated chloroplast markers separated from nuclear markers from the subsample consisting of species with available ETS and 18S sequences and at least two chloroplast markers in order to compare the estimated trees. Both chloroplast and nuclear datasets were used to estimate trees from Maximum Likelihood and Bayesian analysis approaches.
A maximum likelihood approach was performed in PhyML, implemented in Seaview (Gouy et al. 2010). The GTR+G model of sequence evolution was used for both chloroplast and nuclear concatenated sequences.
The Bayesian analysis were performed in MrBayes 3.2.2 Ronquist 2001, Ronquist andHuelsenbeck 2003) using the GTR + G substitution model for both chloroplast and nuclear concatenated sequences. The MCMC algorithm was run twice for 10,000,000 generations, using four chains. Chains were sampled every 100th cycle and a burn-in of 20% was applied.

Phylogenetic analysis
As observed in Figure 2, Philodendron subg. Meconostigma was recovered as monophyletic and as a sister group of P. subg. Philodendron and P. subg. Pteromischum.
The subsampled chloroplast analyses (Figure 2, Suppl. material 1) were inconclusive. They presented very low posterior probabilities for Bayesian analysis (Figure 2A, Suppl. material 1), as well as very low aLRT values for Maximum Likelihood estimates  Figure  3B, Suppl. material 1), with high posterior probabilities and aLRT support, respectively. The estimated phylogenetic relationships are also very similar to those found in the estimated supertree.

Discussion
The genus Thaumatophyllum Schott Molecular evidence. Barabé et al. (2002) used the trnL intron and trnL-trnF intergenic region to estimate the phylogenetic relationships of 33 genera of Araceae; they included six species of Philodendron and found that three species of subg. Philodendron formed a sister group to a clade composed of species of Thaumatophyllum, subg. Pteromischum and Homalomena; P. insigne (subg. Philodendron sect. Baursia) was sister group to all these. Tam et al. (2004) analysed the trnL-trnF with the same six species of Philodendron within a larger analysis focused on subfam. Monsteroideae, but this part of their tree was largely unresolved. Gauthier et al. (2008) carried out a more complete analysis of Philodendron based on over 80 species using ETS and ITS markers. Their ETS tree recovered P. subg. Meconostigma as monophyletic and sister group to P. subg. Philodendron, with P. subg. Pteromischum as the basal component of Philodendron (Fig. 3A). In their ITS tree, the three subgenera formed a trichotomy. Similarly, Loss-Oliveira et al. (2014) recovered P. subg. Meconostigma (= Thaumatophyllum) as monophyletic with 100% aLRT support and 100% posterior probability through the analysis of nuclear ETS and 18S markers and chloroplast matK, trnK, trnL intron and trnL-trnF intergenic region (Fig. 3C). Thaumatophyllum was recovered as sister group of P. subg. Philodendron and P. subg. Pteromischum.
The analysis conducted by Vasconcelos (2015), using the chloroplast markers rpl32-trnL, trnV-ndhC and trnQ-5'-rps16 and the nuclear ITS, corroborate the monophyly of Thaumatophyllum and its position as sister group of P. subg. Philodendron and P. subg. Pteromischum (Fig. 3D). Wong et al. (2016) used nuclear ITS and plastid matK markers in an analysis which included Asian Homalomena, Adelonema (previously American Homalomena) and Philodendron (sensu lato) and also found that Thaumatophyllum and P. subg. Philodendron were sister groups. In this study, P. subg. Pteromischum emerged as sister to Adelonema.
These results are consistent with our findings that Thaumatophyllum is a monophyletic and isolated lineage in Philodendron, the sister group of P. subg. Pteromischum and P. subg. Philodendron.
Morphological evidence. As here defined, Thaumatophyllum is a Neotropical genus composed of 21 species. It is defined by an arborescent habit, very much thickened spathe, well developed sterile intermediate zone in the inflorescence equal or longer than the staminate zone, the gynoecium always having stylar lobes and an axial vascular system independent of the funicle supply (Mayo 1991. Other distinctive features of Thaumatophyllum are (Mayo 1991): 1. sympodial articles diphyllous, internode between prophyll and preceding foliage leaf never developed, internode between prophyll and succeeding foliage leaf sometimes developed but usually very short; 2. leaf blade cordiform, sagittate or hastate, never unlobed at the base; 3. adaxial spathe resin canals J-shaped in longitudinal section, extending to the epidermal surface and secreting resin at anthesis; 4. abaxial spathe resin canals distributed throughout the abaxial parenchyma zone; 5. stamens normally long, slender, more than 3 times longer than wide (except T. leal-costae); 6. stamen vascular trace unbranched (French 1986); 7. style lobes always present; 8. central style dome often present; 9. separate stylar canals occasionally absent; 10. vascular plexus normally present in style; 11. basal vascular complex of gynoecium multi-stranded; 12. lobed central vascular cylinder in the roots (V.T. Rosa, personal comm.); 13. lack of cell wall thickening in the inner root endodermis and neighbouring cortical cells; and 14. collenchyma rather than sclerenchyma sheaths around root resin canals.
Shoot morphology and arborescent habit. Stem architecture in Thaumatophyllum is similar to Philodendron, since the mature stems of both genera are sympodia composed of diphyllous articles (terminology after Ray 1987). However, in those species of Thaumatophyllum which have appreciably elongated internodes, the pattern of elongation is different from that of Philodendron. The position of the 'intravaginal squamules' (Dahlgren andClifford 1982, Mayo 1991) is also different in the two genera and is evidence of the two contrasting patterns of internode elongation. The squamules are always found immediately above the prophyll scar in mature internodes of Philo-dendron. However, in Thaumatophyllum, the squamules occur immediately below the prophyll scar and often surround the foliage leaf scar as well. Also in Thaumatophyllum the squamules frequently persist on the adult stem and are normally spinose or aculeate projections; their number, size, shape and persistence are taxonomically useful.
Inflorescence. Thaumatophyllum is characterised by normally solitary inflorescences in each floral sympodium and very thick, weakly constricted or unconstricted spathes with a uniformly white inner surface. In the spadix, the long staminodial zone that equals or exceeds the fertile male zone is the most useful diagnostic character and distinguishes it from the genus Philodendron. This long staminodial zone plays an important role in the floral biology, serving as a food resource and as the main source of the very large temperature elevations observed during flowering (Gibernau et al. 1999, Gibernau and Barabé 2000, Barabé et al. 2002, Gibernau et al. 2005. Pistillate flowers and the Gynoecium. Unlike Philodendron, the style lobes are conspicuous in Thaumatophyllum and, together, constitute the style crown (Mayo 1991); they resemble stigma lobes as they are frequently covered by stigmatic tissue but are distinct from other kinds of lobed stigma because the lobing is caused by the style apex tissues rather than differential growth of the stigma trichomes. In many species the central region of the style apex is elongated into a more-or-less cylindrical axial portion, the central dome. The central dome may be excavated itself into a pit or even a long canal and may itself have lobed margins. The gynoecial type, typical of Thaumatophyllum, was designated by Mayo (1986Mayo ( , 1989Mayo ( , 1991 as type A, based on a sample of only four species. Calazans et al. (2014) studied 19 out of 21 species and recognised a further three subtypes within Mayo's type A: subtype A1: stylar body absent and stylar canals short, central stylar dome absent and compitum deep (T. adamantinum, T. dardanianum, T. speciosum and T. williamsii); subtype A2: undeveloped stylar body present with long stylar canals, central stylar dome absent and compitum shallow (T. corcovadense, T. lundii, T. paludicola, T. saxicola, T. stenolobum, T. tweedieanum and T. uliginosum); subtype A3: well developed stylar body present with stylar canals long, central stylar dome present and compitum shallow (T. bipinnatifidum, T. brasiliense, T. mello-barretoanum, T. petraeum, T. spruceanum, T. solimoesense, T. undulatum and T. venezuelense).
Based on molecular evidence, Loss-Oliveira et al. (2014) suggested that the common ancestor of Thaumatophyllum probably possessed short stylar lobes, long stylar canals, a stylar body, a vascular plexus in the gynoecium and druses in the stylar parenchyma. These authors also proposed that the morphological diversity observed in the gynoecium of Thaumatophyllum species is the result of an ongoing process of fusion of its floral structures and that the resulting reduction of energy wastage and increase in stigmatic surface are likely to be evolving under positive selection. Evolutionary history. Mayo (1988) hypothesised that Thaumatophyllum was the first lineage to emerge as a distinct clade from ancestral Philodendron and the Eastern and Southern South America species would present a higher number of plesiomorphic gynoecial characters (low number of locules and simple style) and the Amazonian species would have more apomorphic characters (high number of locules and elaborated style). Results from the morphology-based phylogenetic reconstruction of Calazans et al. (2014) partly agreed with Mayo's (1988)  Etymology. from Ancient Greek "θαυματ-" ("thaumato-", wonder, miracle) + "φύλλον" ("phyllum", leaf ); wonderful leaf, referring to the beautiful and peculiar leaves of the type species.   Remarks. The species was previously synonymised under P. tweedieanum (= T. tweedieanum) by Croat and Mount (1988). We propose its reinstatement as an accepted species based on the following morphological differences: herbs erect and rupiculous (x herbs decumbent or rhizomatous subterranean acaulous in T. tweedieanum), prophyll deciduous when still herbaceous (x marcescent and persistent-membranous in T. tweedieanum), denudation of posterior division absent (x present in T. tweedieanum), presence of stylar central dome in pistillate flowers (x absence of stylar central dome in T. tweedieanum). Besides having a different gynoecium type , the majority-rule consensus tree based on morphological characters support these species as different lineages. Furthermore, the phylogeny based on molecular characters supports the two species as separate taxa (Loss-Oliveira et al. 2014). Thaumatophyllum petraeum was first described for Paraguay with four varieties and are still recorded only from this country. We have no evidence to recognise the varieties as distinct taxa, except for. P. petraeum var. tobatiense Chodat & Vischer, which is a synonymous of P. undulatum (= T. undulatum).