In this study, we analyzed the morphological affinities of the 24 species of Amphitecna based on detailed morphological studies and multivariate cluster analyses. Our results suggest that the genus Amphitecna includes six morphological groups that can be easily distinguished based on floral and fruits characteristics: A. donnell-smithii group, A. macrophylla group, A. megalophylla group, A. molinae group, A. spathicalyx group, and A. steyermarkii group. A new species from Mexico, Amphitecna fonceti, is described. This new species is clearly differentiated by the predominantly ramiflorous inflorescences bearing multiple flowers per shoot, buds rounded at the apex, large flowers with a transverse fold in the corolla throat, calyx surface pubescent and strongly costate, and fruits elliptic, apiculate at the apex. We discuss the characteristics of each morphological group and their geographical distribution, provide a detailed description of the new species including ethnobotany notes, and propose the re-establishment of the giant-leaved species A. megalophylla.

Chiapas, conservation, Crescentieae, Neotropics, Mexico

Amphitecna Miers is a Neotropical genus of small to medium-size trees (Gentry 1980). The genus includes around 25 species, most restricted to tropical rainforests. Molecular phylogenetic and morphological data indicate that Amphitecna is closely related to the Neotropical genera Crescentia L. and Parmentiera DC., from which it differs by the combination of simple and alternate leaves, greenish corollas with petal lobes fused, and pepo-type fruits (Gentry 1980, Grose and Olmstead 2007).

The distribution range of Amphitecna encompasses two regions with high species diversity (Table 1). The first region includes 14 species and encompasses the rainforests from Mexico to Honduras. The second region contains eight endemic species that range from Northern Nicaragua to the North of Colombia (Table 1). The only widely distributed species is A. latifolia (Mill.) A.H. Gentry, that occurs in the coastal areas of Florida (USA), the Caribbean, the Pacific slope of Mexico, and coastal areas of Central America to Colombia, Ecuador, and Venezuela (Gentry 1980). The wide distribution of A. latifolia may be associated with its water-dispersed fruits, while most of the narrowly distributed species are mammal-dispersed (Gentry 1980).

Reproductive characteristics of species of Amphitecna are quite variable, with most species exhibiting clear differences in flower and fruit morphology (Gentry 1980; Gentry 1982a; Ortiz-Rodriguez et al. 2016). However, most species are similar vegetatively, often leading to taxonomic confusion and misidentifications. For example, the four giant-leaved species, with leaves 50–100 cm long × 10–15 cm wide, i.e., Amphitecna costata, A. megalophylla, A. macrophylla, and A. regalis, are frequently misidentified in herbaria. Despite the similarity in leaf traits, these four species are easily differentiated reproductively (Gentry 1980, Table 2). Among the most variable reproductive features of Amphitecna are the corolla shape, inflorescence position, number of flowers per shoot, pedicel length, and the number of calyx components (Gentry 1980; Gentry 1982a; Burger and Gentry 2000). Flower bud shape, fruit morphology, and calyx surface also represent important features for species identification (Ortiz-Rodriguez et al. 2016).

Here, we carried out a multivariate cluster analyses of all 24 species of Amphitecna currently recognized to infer the morphological affinities among species and establish the position of a newly described species of Amphitecna from the Sierra Madre de Chiapas, Mexico.

To infer the morphological similarities among the 24 species of Amphitecna, we performed a hierarchical clustering analysis on a matrix that included 15 flower traits. The data were analyzed using the unweighted pair group method with arithmetic mean (UPGMA, Sokal and Michener 1958) and the Gower index (Gower 1971), allowing a simultaneous use of binary and continuous characters (Dunn and Everitt 1982; Yang et al. 2007; Zanella et al. 2011; Tuler et al. 2017; Svoboda and Ballard 2018; Wahlsteen and Tyler 2019). All morphological characters were obtained from original species descriptions (Gentry 1980; Gentry 1982b; Burger and Gentry 2000; Ortiz-Rodriguez et al. 2016), herbarium specimens deposited at MEXU (www.ibdata.ib.unam.mx), and type specimens available online (https://www.gbif.org/ and https://plants.jstor.org/).

The UPGMA results were contrasted with those derived from other clustering algorithms, specifically Ward, single linkage, complete linkage, WPGMA, WPGMC, and UPGMC, implemented in the R-package stats, using the function hclust (R Core Team 2020: https://www.r-project.org/). We then determined the similarities and differences among the various clustering dendrograms by calculating the cophenetic correlation (a Pearson’s measure) between each clustering result using the cor.dendlist and the corrplot functions from the corrplot R-package (Wei and Simko 2017). For each dendrogram, the agglomerative coefficient was calculated using the agnes function from the cluster R-package (Maechler et al. 2019). The agglomerative coefficient measures the amount of clustering structure, with values closer to 1 suggesting stronger clustering structure. Also, the Fowlkes-Mallows Index (from the dendextend R-package, Galili 2015) was used to compare the species composition within clusters (k = 3–8) obtained from the UPGMA analysis and other algorithms. The optimal number of morphological clusters in Amphitecna was determined based on the greater similarity between clustering algorithms (values closer to 1). We further performed an internal clustering validation (a cluster stability test) by calculating the average silhouette width (Si) for each cluster (k = 3–8) resulting from each of the algorithms used. While Si values greater than 0.71 suggest strong structure and cluster stability, values between 0.51 and 0.70 are interpreted as reasonable, values between 0.26 and 0.50 indicate weak structure, and values lower than or equal to 0.25 are not worth further discussion (Kaufman and Rousseeuw 2005). The graphical representation of the UPGMA dendrogram was carried out in the R software, using the function hclust implemented in the R-packages ape, and ggtree (Yu et al. 2017; Paradis and Schliep 2018).

The new species described was recognized by a unique combination of features (Donoghue 1985) identified through comparisons with morphologically similar taxa and literature review (Gentry 1982a, b, Burger and Gentry 2000; Ortiz-Rodriguez et al. 2016). 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 2019).

The UPGMA dendrogram is shown in Figure 1. The results of this analysis are very similar to those obtained using other clustering algorithms (correlation values between 0.78 and 0.98, Suppl. material 1: Figure S1). The agglomerative coefficient value for the UPGMA dendrogram was 0.67 (between 0.47 and 0.83 in analyses conducted with other approaches), suggesting a moderate to strong structure among species of Amphitecna. The Fowlkes-Mallows Index showed that six groups (k = 6) show significantly similar clusters when the UPGMA is compared to the other clustering algorithms (FM values between 0.71 and 1). Silhouette width values consistently showed the highest values (Si value 0.35 for all algorithms) when each dendrogram was divided into six clusters.

Figure 1. 

Morphological similarities among the 24 species of Amphitecna currently recognized A dendrogram based on the results from an UPGMA analysis B main flower types found in Amphitecna, flowers with a transverse fold in the corolla throat (Amphitecna costata, top right) and flowers radially symmetric, without a transverse fold in the throat (Amphitecna tuxtlensis, bottom right). Photographs by Hector Gómez Domínguez (A. costata) and Pablo Carrillo Reyes (A. tuxtlensis). NM = Northern Mesoamerica; SM = Southern Mesoamerica.

The UPGMA results indicate that the genus Amphitecna can be classified into six morphological groups (Figure 1A) and three main clusters (A–C). The A. molinae, A. megalophylla, and A. spathicalyx groups are characterized by corollas with a transverse fold in the throat (Figure 1B). In the UPGMA dendrogram, these species are nested within cluster “A” (Figure 1A). The Amphitecna molinae group consists of 10 species with sessile leaves, 15–40 cm long, flower buds rounded at apex, and smooth fruit surface, rarely warty. The A. megalophylla group includes two species with short-petiolate leaves, long leaf blades (50–100 cm long), flower buds rounded at the apex, and fruit surface costate. The new species, Amphitecna fonceti, is part of the A. molinae group and is morphologically most similar to the A. latifolia and A. sessilifolia groups (Table 3). Amphitecna spathicalyx is also placed within cluster “A” (i.e., the A. spathicalyx group) and is distinguished from other species by the pointed flower buds and spathaceous calyces.

Cluster “B” is composed of the A. macrophylla and A. donnell-smithii groups (Figure 1A). The species from these groups are best recognized by their short-petiolate leaves, inflorescences born along the main trunk or leafless branches, mostly composed of 1–2 flowers per shoot, flowers without a transverse fold in the corolla throat, and calyces bilabiate or trilabiate (Figure 1B). The A. macrophylla group contains five species with cauliflorous inflorescences and long and funnelform corollas (up to 70 mm long). The A. donnell-smithii group consists of two ramiflorous species with small flowers (less than 28 mm long) and broadly campanulate corollas. The only two species of Amphitecna with broadly campanulate flowers are included in the A. donnell-smithii group.

Cluster “C” consists of the Amphitecna steyermarkii group and is composed of three species that are characterized by their terminal inflorescences with several flowers per shoot, flower buds with a sharp acumen, flowers without a transverse fold in the corolla throat, and spathaceous calyces.

We are grateful to Yuyini Licona Vera, Ivett Reynoso, Juan Francisco Ornelas, José Luis Villaseñor, Lulu Rico Arce, Alan Paton, and an anonymous reviewer for providing many useful comments on the manuscript. We also thank Christopher Davidson and Sharon Davidson for their support during field work.