Cyphostemma in

Dang, Viet Cuong, Nguyen, Van Hieu, Dao, Baohui, Yang, Wanliu, Li, Faliang, Yang, Xinkai, Wen, Jun, Chen, Zhiduan & Lu, Limin, 2017, A New Species and New Records of Cyphostemma (Vitaceae) from China and Vietnam Based on Morphological and Molecular Evidence, Systematic Botany (Basel, Switzerland) 42 (3), pp. 449-457 : 451

Morphological Observations —Specimens of Cyphostemma in several major herbaria in China (PE) and elsewhere (E, HN, K, MO, P, US, and VNM) were examined.Seed morphology was observed and photographed with a stereomicroscope (a Nikon SMZ1000 with a Nikon DXM 1200F digital camera) in the State Key Laboratory of Systematic and Evolutionary Botany, Institute of Botany, Chinese Academy of Sciences. Terminology for describing leaves and seeds follows Jiménez-Saa (2011) and Chen and Manchester (2007), respectively.

Phylogenetic Analyses —The phylogenetic position of both the new species and the new record was inferred in a global sampling scheme from Lu et al. (2013) using the nuclear ribosomal ITS and five chloroplast DNA regions (atpB-rbcL, rps16, trnC-petN, trnH-psbA, and trnL-F). Thirty-six new sequences from six newly collected samples of Cyphostemma were generated for this study with the other sequences from Lu et al. (2013, 2017) and Liu et al. (2016) downloaded from GenBank (Appendix 1). Sequence assembly and alignment followed the protocols described in Lu et al. (2013, 2017). Single gene trees were constructed to check whether wellsupported incongruences (bootstrap. 70%) exist between the chloroplast and ITS topologies ( Hillis and Bull 1993) prior to combining all individual data sets.

The combined dataset was analyzed with the maximum likelihood (ML) and Bayesian inference (BI) methods. The ML analyses were conducted in RAxML 8.1.11 ( Stamatakis 2006), applying 1,000 bootstrap replicates. A partitioned Bayesian analysis was conducted in MrBayes 3.2.6 ( Ronquist et al. 2012) as implemented on the CIPRES Science Gateway Portal ( Miller et al. 2010). Parameter configuration and convergence estimation followed Lu et al. (2013).Best-fitting model for each of the single partition was calculated using MrModeltest 2.3 ( Nylander 2004) under the Akaike information criterion (AIC). GTR 1 G was found to be the most appropriate nucleotide substitution model for atpB-rbcL, rps16, trnC-petN, trnH-psbA, and trnL-F; GTR 1 I 1 G for ITS.