Phylogenetic analyses and morphological characteristics support the description of a second species of Tridimeris (Annonaceae)

Abstract Based on phylogenetic and morphological evidence, Tridimeris chiapensis Escobar-Castellanos & Ortiz-Rodr., sp. n. (Annonaceae), a new species from the karst forest of southern Mexico, is described and illustrated. The new species differs from Tridimeris hahniana, the only described species in the genus, in that the latter has flowers with sepals densely tomentose outside, one (rarely two) carpel(s) per flower and fruits densely covered with golden-brown hairs, while Tridimeris chiapensis has flowers with glabrous sepals outside, two to five carpels per flower and glabrous fruits. Furthermore, a shallow triangular white patch at the base of the inner petals is found in Tridimeris chiapensis, a morphological character shared with the sister genus Sapranthus but absent in Tridimeris hahniana. Geographically, both species occur allopatrically. With just one known locality and seven individuals of Tridimeris chiapensis recorded in one sampling hectare, and based on application of the criteria established by the IUCN, we conclude tentatively that the species is critically endangered.


Introduction
Annonaceae is a plant family composed of about 110 genera and 2,500 species of trees and lianas . Most genera of Annonaceae (except Asimina, endemic to the USA) are exclusively tropical and many are important fl oristic elements in several lowland forests. Within continents, the endemism at generic level is very high and only the genus Xylopia is pantropical (Doyle and Le Th omas 1997). Recent phylogenetic analyses (Richardson et al. 2004 show that Annonaceae is composed of four major lineages, and on that basis, the family is now classifi ed into four subfamilies: Anaxagoreoideae, Ambavioideae, Annonoideae and Malmeoideae. Of these, Annonoideae (50 genera and 1600 species) and Malmeoideae (50 genera and 700 species) are the most species-rich lineages in Annonaceae (Pirie and Doyle 2012). In Mexico the family is represented by eleven genera, Anaxagorea, Annona, Cymbopetalum, Desmopsis, Guatteria, Mosannona, Sapranthus, Stenanona, Tridimeris, Unonopsis and Xylopia, of which Tridimeris is endemic.
Tridimeris is a monotypic and poorly studied genus. Baillon (1869) described its only species, Tridimeris hahniana Baill., based on exemplars from Veracruz, Mexico (Turner 2013). Th e species, restricted to eastern Mexico in the states of San Luis Potosí, Puebla and Veracruz (Figure 1), is easily recognizable by its greenish and dimerous fl owers (two sepals and four petals) and its large and fl eshy fruits densely covered with golden-brown hairs (Schatz 1987). Dimery is uncharacteristic in Neotropical Annonaceae and it has been recorded only in Anaxagorea silvatica R. E. Fr., Ephedranthus dimerus J. C. Lopes, Chatrou & Mello-Silva and, Malmea dimera Chatrou (Lopes et al. 2014). Phylogenetic analyses based on molecular characters , Chaowasku et al. 2012, Ortiz-Rodriguez et al. 2016 show that Tridimeris hahniana belongs to the Malmeoideae tribe Miliuseae, where along with Desmopsis, Sapranthus and Stenanona it forms the subtribe Sapranthinae (Ortiz-Rodriguez et al. 2016). Sapranthus and Tridimeris are closely related and together form the sister group of the remaining members of Sapranthinae. Although Sapranthus and Tridimeris have contrasting fl oral characteristics, with Tridimeris showing axillary, dimerous and greenish fl owers whilst Sapranthus shows leaf-opposed, trimerous and usually brown to purple colored fl owers, both genera are characterized by large and fl eshy fruits, these last characteristics being their most obvious synapomorphy (Schatz 1987).
During a fl oristic study in southern Mexico, several individuals of an unusual species of Annonaceae were collected. Th e general characteristics of its fl owers, notably dimery, suggested that it probably was a species related to the genus Tridimeris; however, its fruit characteristics did not fi t with those of Tridimeris hahniana. To elucidate this, we performed molecular phylogenetic analyses including one sample of the putative new species and studied its morphological characteristics in detail to corroborate its identity at the genus level and to determine whether the collections from Chiapas represent a second species of Tridimeris for the Mexican fl ora.

Phylogenetic analyses
Phylogenetic relationships among taxa were estimated using Bayesian inference (BI), maximum likelihood (ML), and parsimony methods. Analyses of six cpDNA regions were conducted separately (for regions with information available for the new species) and in combination.
For BI, three partitioning strategies were used: (1) data matrix divided into six partitions based on DNA region identity, (2) six plastid markers concatenated and analyzed without partitioning, and (3) 2-partitioned, distinguishing coding (matK, ndhF, rbcL and ycf1) and non-coding (psbA-trnH and trnL-F) regions. jModelTest ver. 3.06 (Posada 2008) was used to determine the appropriate DNA substitution model and gamma rate heterogeneity for each partition using the Akaike Information Criterion (AIC). Th e BI analysis was performed using MrBayes ver. 3.2.1 (Ronquist and Huelsenbeck 2003) with two independent runs of four chains and ten million generations with trees sampled every 1000 generations. Convergence was assessed using the standard deviation of split frequencies as a convergence index, with values <0.01 interpreted as indicating good convergence. Tracer ver. 1.5 (Rambaut and Drummond 2007) was used to determine whether the parameter samples were drawn from stationary, unimodal distribution, and whether adequate eff ective sample sizes (ESS) for each parameter (ESS>200) were reached. Th e initial 25% of samples of each MCMC run were discarded as burnin, and the remaining trees were summarized as posterior probabilities; PP values ≥ 0.95 were considered to represent strong support. Analysis performance of each partitioning strategy was assessed using Bayes factor. Bayes factors were calculated from the estimated harmonic means of likelihood using the sump command in MrBayes. Decisions were taken based on the 2ln Bayes factor criterion (Kass and Raftery, 1995), for which the Bayes factor scale of strength of evidence in favor of one hypothesis is: 0-2 (not worthy of mention), >2-6 (positive), >6-10 (strong), >10 (very strong).
Th e most parsimonious trees were obtained using the ratchet strategy (Nixon 1999) in Winclada ver. 1.0000 (Nixon, K. C. 1999(Nixon, K. C. -2002, running NONA ver. 2.0 (Goloboff 1993) on a combined dataset of six plastid regions, with nucleotide characters treated as unordered and equally weighted, 1000 iterations, holding 10 trees per iteration with 10% of nodes constrained, and all other parameters set to default. Branch support was assessed using bootstrap resampling, 1000 bootstrap-resampled pseudoreplicate matrices were each analyzed using 100 random addition sequences (multi*100). Ten trees were retained during TBR swapping after each search initiation (hold/10) using NONA ver. 2.0 and performed in WinClada, with the same interpretations of support level as in the ML analyses.

Morphological differentiation
We examined the specimens of Tridimeris hahniana deposited at XAL herbarium (Th iers 2016). Also, we consulted the digitized type specimens available at JSTOR Global Plants (http://plants.jstor.org/). Th e putative new species was recognized using the unique combination of features criteria (Donoghue 1985) through comparisons with morphologically similar species and literature review (Schatz 1987). Finally, we elaborated the species description following terminology presented in Hickey (1973).

Conservation status
We assessed the conservation status by calculating the extent of occurrence (EOO) and the area of occupancy (AOO) using the GeoCAT tool (Bachman et al. 2011) and applying the IUCN Red List Categories and criteria (IUCN 2001).

Analysis of individual cpDNA regions
Each individual cpDNA region provided a relatively good resolution within Sapranthinae clade, with most branches resolved in the four separate trees (Suppl. material 1-4). Analyses of the matK, rbcL and ycf1 coding regions showed that Sapranthinae is composed of two main subclades, the Desmopsis-Stenanona clade and the Sapranthus-Tridimeris clade, while the analysis of trnL-F spacer showed very low resolution recovering only the Sapranthus-Tridimeris clade (Suppl. material 1-4). Each phylogenetic hypothesis unequivocally placed the new species within the Sapranthus-Tridimeris clade.

Analysis of combined data
Th e concatenated 32-accession dataset contained 6419 aligned positions, of which 746 were variable and 208 were parsimony informative. For the Bayesian analyses, the substitution model was GTR+G for matK, trnL-F, psbA-trnH and unpartitioned datasets, GTR+G+I for rbcL, ndhF and coding datasets and HKY+G for non-coding dataset.
Th e six partitioned strategy considerably improved the mean −lnL values in the Bayesian analyses (mean −lnL non-partitioned = -15754.57; mean -lnL 2-parti-tioned= -15725.73; mean -lnL 6-partitioned = -15722.11). Bayes factor comparison indicated that the analyses using six partitions provided a better explanation of the data than unpartitioned and 2-partitioned analyses. For the ML analyses the likelihood score of the optimal ML tree, was ln L = -15572.87. Th e parsimony analysis of the combined regions resulted in 20 most parsimonious trees of 1030 steps with a Consistence Index of 0.79 and a Retention Index of 0.60. Th e subsequent presentation of the results is restricted to the 50% majority rule consensus tree derived from Bayesian analyses using six partitions.

Morphological differentiation
Morphologically, Tridimeris chiapensis has a set of morphological characters that clearly distinguish it from T. hahniana (Fig. 3, Table 1). In addition, both species occur in disjunct locations and therefore distributed allopatrically (Fig. 1).

Discussion
Th e phylogenetic analyses showed that Tridimeris chiapensis and T. hahniana form a strongly supported monophyletic group (Fig. 2). Th e two species of Tridimeris share axillary infl orescences, dimery fl owers (two sepals and four petals), greenish petals, and large and fl eshy fruits. Furthermore, both species have pocket domatia in the axils of secondary veins. As in previous studies, Tridimeris and Sapranthus appear to be closely related , Chaowasku et al. 2012, Ortiz-Rodriguez et al. 2016) and together form the Sapranthus-Tridimeris clade. Tridimeris chiapensis clearly diff ers from T. hahniana by its number of carpels per fl owers, fruit surface, glabrous pedicels and sepals, and by the presence of a triangular white patch near the base of inner petals (Fig. 3). A similar white patch is found in the inner petals of Sapranthus viridifl orus G.E. Schatz, which have been considered by Schatz (1998) as a vestigial food body since food bodies are morphological modifi cations of a specifi c area of inner petals as food reward for fl oral visitors (Schatz 1987). However, a more detailed study of these structures is needed, as well as the compounds present in this structure and its anatomical characteristics in order to determine its function.
Ecologically T. chiapensis inhabits wet forests on karstic topography around 1000 m elevation, while T. hahniana occurs in lowland wet forests (200-900 m) or even in cloud forests in the northern portion of its distribution (Schatz 1987). Th e type locality of T. chiapensis, the protected natural area La Pera in Chiapas, is a karstic zone covered mostly by tropical rain forest. Th is area among other similar regions of southern Mexico, are considered centers of plant endemism (Wendt 1987).
Tree 3-9 m tall and 3-14 cm DBH; young branches slightly pubescent, trichomes appressed and golden-brown in color, glabrescent with age. Leaves membranaceous to chartaceous, alternate, phyllotaxy distichous, 11-20 cm long to 3.5-8 wide, narrowly elliptic to obovate, the apex acute to acuminate, the base acute to obtuse, sometimes asymmetrical; upper surface glabrous, the lower side glabrescent; venation brochidodromus, 6-9 veins per side, pocket domatia in the axils of the main veins; the midrid impressed above and slightly canaliculate toward the base (sometimes with erect to appressed light-brown hairs), lateral veins barely elevated above; the midrib and lateral veins prominently elevated below and with sparsely light-brown hairs, lateral veins decurrent at midrid insertion ; petiole swollen, 0.5-1 cm long, canaliculate, with sparsely light-brown hairs. Inflorescences always oneflowered, axillary, sometimes arising on leafl ess part of branches (ramifl ory), the pedicel glabrous, 1-1.7 cm long, bearing 2-3 minute, densely golden tomentose and broadly ovate basal bracts. Sepals 2, connate, to 2 mm long × 4-5 mm wide, decurrent along the pedicel, broadly ovate, rounded at apex, glabrous inside and outside, the margins ciliate. Petals 4, in two subequal whorls, 8-14 mm long × 3-5 mm wide, lanceolate to triangular, green to yellowish green, glabrous inside and outside, the margins ciliate, acute at apex, the base truncate and cusped around the stamens; the outer petals, more or less thin, with faint venation, refl exed at anthesis; the inner petals thicker and fl eshier and not refl exed with a shallow, more or less triangular white patch near the base. Stamens, c.a. 40, 1-1.5 mm long, extrorse, fi lament very short, apical part of connective expanded over the thecae, shield-shaped, ellipsoid to angulate, glabrous. Carpels, 2-5 per fl ower, to 2.5 mm long; the stigma more or less globose and essentially glabrous; style absent; the ovaries ellipsoid and more or less curved, like a small banana with sparsely light-brown hairs; the ovules, 12-18, lateral, in two rows. Monocarps, 1-4 per fruit, large and fl eshy, 8-11 cm long × 3-5 cm wide, ellipsoid, the apex and base rounded, glabrous, shortly stipitate, stipes to 7 mm long; young monocarps green, yellow to light brown when ripe with a peach-like sweet odor; seeds lunate to wedge-shaped, 1.3-2.2 cm long with lamellate ruminations.
Habitat and ecology. Th e type locality of Tridimeris chiapensis is locally named as "La Pera" and "Pozo Turipache" or "El Pozo" and it lies within the ecological state reserve La Pera, which is mostly covered by tropical rainforest. Th in soils, rough limestone outcrops, caves, crevices, sinkholes and almost no surface water that form a typical karst landscape can be observed around El Pozo (Wake and Johnson 1989). Also, fogs forming cloudbanks are common throughout the year, though absent during the dry season (Wake and Johnson 1989). Phenology. Th e species was found in full bloom in August and bearing fruits in March and May.
Etymology. Th e specifi c epithet is in honor of the Mexican state of Chiapas where the species was found.

Conservation status.
Tridimeris chiapensis is known only from the type locality at the ecological state reserve La Pera. According to the criteria established by the IUCN, it is possible to tentatively determine that the species is Critically Endangered [CR B1ab (iii)]. Th e Area of occupancy (AOO) of T. chiapensis is 0.314 km² and the Extent of occurrence (EOO) is 1.519 km², suggesting a very restricted overall distribution. Although the only known population of the species is located within a protected natural area, only 7 individuals of Tridimeris chiapensis in one hectare of sampling were recorded (Escobar-Castellanos 2016). Th e 3000 ha of La Pera's rainforest estimated by Espinosa (2014) and its species are threatened by non-sustainable activities (logging, fi res, illegal settlements) and forests in this region are fragmented and only some remnants persist which are surrounded by roads, croplands and cattle pastures (Medina et al. 2006, Luna-Reyes et al. 2015.

Acknowledgements
We extend our sincere gratitude to Antonio Arreola, owner of Campamento Trepatroncos where the type species was collected, and Christopher Davidson and Sharon Christoph for partially fi nancing visits to the type locality through the project "Floristic diversity and evolutionary ecology of endangered species in protected natural areas from Chiapas, México". We thank Yuyini Licona, Juan Francisco Ornelas, Eduardo Ruiz and Francisco Lorea for helpful comments on previous versions of the manuscript. We especially thank Andrew Vovides (Laboratorio de Biología Evolutiva de Cycadales) for his careful revision that considerably improved the writing of this paper; and Th omas Couvreur, George Schatz, Lars Chatrou, Roy Erkens and one anonymous reviewer for their valuable criticisms and suggestions. Th e fi rst author thanks CONACyT (Consejo Nacional de Ciencia y Tecnología, México) for the support of his doctoral scholarship number 262563. Th is work constitutes partial fulfi llment of AEOR's doctorate in Biodiversity and Systematics at the Instituto de Ecología, AC. Th is research was partially supported by a competitive grant (155686) from the Consejo Nacional de Ciencia y Tecnología, México (CONACyT) and research funds from the Instituto de Ecología . Th e Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited.