Bredia hispida (Sonerileae, Melastomataceae), a species occurring in southeastern Sichuan, China, is newly described based on morphological and molecular data. The generic placement of B. hispida is well supported by phylogenetic analysis and morphological characters, including basally cordate, hairy leaf blade, cymose inflorescence, basally gibbous anthers and enlarged ovary crown enclosing an inverted frustum-shaped depression. Both molecular and morphological divergence showed that B. hispida is well separated from its close relatives, justifying its recognition as a distinct species. The new species resembles B. repens, B. changii and B. guidongensis in the prostrate habit and isomorphic stamens but differs markedly in the unequal opposed leaves, the 2–4 mm long, stout bristles on the adaxial surface of leaf blade and acuminate leaf apex. Bredia hispida co-occurs with B. esquirolii in the wild. No morphologically putative hybrids between them were observed despite their overlap in flowering season. The isolating mechanism remains unclear, pending further investigation.

Bredia, Melastomataceae, phylogeny, taxonomy

Bredia Blume was originally established based on B. hirsuta Blume, a species endemic to Taiwan and the Ryukyu islands (Blume 1849). Circumscription of this genus had long been controversial. The dispute mainly concerned whether to include Tashiroea Matsum. ex T. Itô & Matsum. and certain species of Phyllagathis Blume in Bredia (Diels 1924, 1932; Merrill and Chun 1940; Li 1944; Chen 1979, 1984; Hansen 1992; Chen and Renner 2007). By combining molecular phylogenetic and morphological data, recent studies have provided strong evidence for a new generic limit of Bredia (Zhou et al. 2019a, 2019b, 2019c). Bredia was redefined as excluding Tashiroea while incorporating seven species previously treated in Phyllagathis (Zhou et al. 2019b). Together with two recently published species, Bredia as currently circumscribed includes 23 species distributed from central and southern mainland China, Taiwan, to the Ryukyu islands and northern Vietnam (Zhou et al. 2019b; Wen et al. 2019; He et al. in press). Species of Bredia are characterized by the leaf blade papery, usually hairy, inflorescences cymose, umbellate, or a cymose panicle, anther basally gibbose or tuberculate, and ovary crown persistent and enlarged enclosing an inverted frustum-shaped depression.

During a survey of specimens in Chinese herbaria for a project on species delimitation of Bredia, several collections from Xuyong County, southeastern Sichuan Province, caught our attention. These collections (e.g. Fig. 1A–C) were identified as Phyllagathis deltoidea C. Chen (Fig. 1D–F). Upon closer examination, however, the plants from Xuyong are morphologically quite different from P. deltoidea in having basally cordate (vs. cuneate) lamina adaxially hispid with stout long bristles (vs. puberulous and sparsely setose) and linear-lanceolate calyx lobes (vs. broadly triangular lobes) (Fig. 1). In addition, P. deltoidea is only recorded from its type locality in Ningming County, southwestern Guangxi, which is about 800 kilometers away from Xuyong County, southeastern Sichuan. Both morphology and distribution suggest that the current identification is erroneous.

Figure 1. 

Herbarium specimen of Bredia hispida (A–C) and Phyllagathis deltoidea (D–F). A–C W. B. Ju and H. N. Deng, HGX12702 (CDBI) collected from Shui-wei town, Xuyong County, Sichuan, China, showing the stout long bristles on adaxial surface of the leaf blade and linear-lanceolate calyx lobes, images from National Plant Specimen Resource Center D–F H. H. Su 68119 (IBK, IBSC) collected from Aidian, Ningming County, Guangxi, China, holotype (IBSC) (D, F) and isotype (IBK) (E) of Phyllagathis deltoidea.

In September 2019, we made a field expedition to Xuyong County and collected flowering and fruiting specimens of the plant in question (Figs 2, 3). This plant possesses all the synapomorphies of Bredia aforementioned. It most closely resembles B. changii W. Y. Zhao, X. H. Zhan & W. B. Liao, B. guidongensis (K. M. Liu & J. Tian) R. Zhou & Ying Liu and B. repens R. Zhou, Q. J. Zhou & Ying Liu in habit and isomorphic stamens (Fig. 4), but differs markedly from the latter species in leaf morphology (Fig. 2). Judging from morphological aspects, this plant may represent an undescribed species in Bredia.

Figure 2. 

Bredia hispida. A Habitat B a flowering branch, showing the prostrate habit C adaxial leaf surface D abaxial leaf surface E lateral view of leaf, showing the stout long bristles F terminal inflorescence G axillary inflorescences on old branchlets. All from Y. Liu 764 (A, PE, SYS).

To test the generic affiliation of the unknown plant and its closest relative in the genus, we performed phylogenetic analyses based on DNA sequence data of nuclear ribosomal internal transcribed spacer (nrITS), sampling all species so far recorded in Bredia. We also calculated pairwise genetic distances among this plant and 23 species of Bredia to evaluate its distinctness. The results confirmed our suspicion that this plant represented a species new to science, which we described as B. hispida below.

Figure 3. 

Floral details and young fruit of Bredia hispida. A Side view of a flower B top view of a flower C longitudinal section of a flower showing isomorphic stamens and ovary crown D top view of a young fruit E side view of a young fruit showing enlarged ovary crown exserted from hypanthium. Scale bar: 5 mm (C). All from Y. Liu 764 (A, PE, SYS).

To test the generic affiliation of B. hispida and its position in the genus, we selected ingroup taxa based on previous studies (Zhou et al. 2019a, 2019b, 2019c). The final nrITS dataset contained 35 accessions representing Blastus Lour., Dissochaeta Blume, Fordiophyton Stapf, Tashiroea, Phyllagathis, Scorpiothyrsus H. L. Li, Blakea P. Browne, and 23 species so far recorded in Bredia, with Blakea (Blakeeae) chosen as an outgroup according to Goldenberg et al. (2012). The sequences of B. hispida, B. violacea H. L. Li and B. reniformis C. M. He, Y. H. Tong & S. J. Zeng were newly sequenced, while the rest were downloaded from GenBank. The source of the materials and GenBank accession numbers are given in Supplementary material 1.

Total DNA was extracted from fresh leaves using the modified CTAB procedure (Doyle and Doyle 1987). The nrITS region of B. hispida, B. violacea and B. reniformis were amplified and sequenced using universal primers ITS4 and ITS5 (White et al. 1990), following the procedure described in Zou et al. (2017).

Sequences were aligned using SeqMan v.7.1.0 (DNASTAR Inc., Madison, WI). The Akaike information criterion in Modeltest version 3.7 (Posada and Crandall 1998) was used to select the best-fitting nucleotide substitution model (GTR+G) prior to phylogenetic analyses. Bayesian inference (BI), maximum likelihood (ML) and maximum parsimony (MP) analyses were performed using MrBayes 3.2.6 (Huelsenbeck and Ronquist 2001), RAxML version 8.2.10 (Stamatakis 2014) and PAUP version 4a165 (Swofford 2002) respectively. For BI analysis, two independent Markov chain Monte Carlo analyses (MCMC) were performed with four simultaneous chains of 2,000,000 generations sampling one tree every 100 generations. The first 25% of trees were discarded as burn-in and the remaining were used to construct a majority-rule consensus tree with Bayesian posterior probabilities (PP). We verified that the average deviation of split frequencies had reached a value below 0.01 at the end of MCMC analyses. We also assessed the effective sample sizes (ESS) for all parameters and statistics using Tracer version 1.7.1 (Rambaut et al. 2018). ML analyses were performed under GTR+G model as recommended by the author. Node support was estimated with 1,000 bootstrap replicates using a fast bootstrapping algorithm (Stamatakis et al. 2008). For MP analyses, a heuristic search strategy was conducted of 1000 random addition replicates, with the tree-bisection-reconnection (TBR) branch swapping algorithm and MultTrees on. Maxtree was set to 500. Node support was evaluated by 1000 bootstrap replicates of 1000 random additions. Pairwise genetic distances among B. hispida and species of Bredia were calculated using the Kimura 2-parameter method (Kimura 1980).

The aligned sequence matrix contained 665 characters. Statistics of sequences sampled were summarized in Supplementary material 2. The tree resulting from ML analysis is shown in Fig. 5, with PP, ML bootstrap support values (BSML), and MP bootstrap support values (BSMP) labeled at nodes. Bredia hispida was nested within the well supported Bredia clade (PP = 1.0, BSML = 100%, BSMP = 96%), forming a subclade with B. repens, B. tuberculata (Guillaumin) Diels and B. yunnanensis (H. Lév.) Diels (PP = 1.0, BSML = 87%, BSMP = 81%). Pairwise genetic distances among B. hispida and 23 species of Bredia are provided in Supplementary material 3.

We thank Herbarium of Chengdu Institute of Biology (CDBI) for their kind assistance during herbarium survey. We are grateful to Dr. David E. Boufford and the anonymous reviewer for their valuable comments on the manuscript. This work was supported by the National Natural Science Foundation of China (31770214), Science and Technology Program of Sichuan Province (2018JY0487) and Science and Technology Project of Yuexiu District, Guangzhou City (2018-GX-001).