Molecular Ecology (2008) 17, 1469–1488 doi: 10.1111/j.1365-294X.2007.03681.x HowBlackwell Publishing Ltd many species are in the genus Batrachuperus? A phylogeographical analysis of the stream salamanders (family Hynobiidae) from southwestern China JINZHONG FU* and XIAOMAO ZENG† *Department of Integrative Biology, University of Guelph, Guelph, Ontario, Canada N1G 2W1, †Chengdu Institute of Biology, Chinese Academy of Sciences, Chengdu, Sichuan 610041, China Abstract Phylogeographical analysis of DNA sequence data has been routinely used to test species boundaries using the monophyly criterion; however, a complementary criterion, reproductive isolation, is often ignored. We used a combination of phylogenetic and population genetic approaches to determine species boundaries among stream salamanders in the genus Batrachuperus. First, cytochrome b sequence data from 174 Batrachuperus individuals, sampled from 78 populations, were used to reconstruct historical relationships within the genus. Second, allozyme data for 14 presumptive nuclear loci, from 463 individuals sampled from 60 populations, were collected and analysed to assess population similarity or disparity, as well as potential reproductive isolation. The DNA sequence data grouped all popula- tions into seven major monophyletic groups, and the allozyme data provided evidence for reproductive isolation among four of the seven groups, thereby supporting the species status of these groups. The allozyme data suggested that two of the other groups share the same gene pool, and therefore belong to a single species. Finally, the allozyme data revealed two reproductively isolated units within the seventh group, which we suggest represents a case of ‘budding speciation’ based on the DNA gene tree. In total, seven species of the genus Batrachuperus were defined, two of which were previously unknown. The phylogeographical analysis also revealed that vicariance events might have dominated the evolutionary history of this group, but the speciation events might precede the formation of the existing mountain topology. This study demonstrates the importance of including frequency data from multiple nuclear gene loci in determining species boundaries. Keywords: allozyme, DNA sequence, Hengduan Mountains, hynobiid salamander, phylogeography, species boundary Received 6 October 2007; revision received 6 December 2007; accepted 16 December 2007 the vast exploration of various nuclear genes, mtDNA Introduction sequence data continue to represent the marker of choice in Phylogeography reveals the history and formation of most recent phylogeographical studies, and it has also species (Avise 2000), and phylogeographical analysis of figured prominently in recent studies using molecular DNA sequence data has been routinely used to determine markers to identify species. Held (2003) argued that mtDNA species boundaries and patterns of speciation (e.g. Hewitt sequences alone are sufficient to identify species if the 2001). Mitochondrial DNA (mtDNA) has a small effective populations examined fulfil several criteria, such as bimodal population size and thus, a short coalescence time, which distribution of pairwise differences and persistence of high offers a major advantage over most nuclear genes for levels of genetic differentiation in sympatry, and Hebert phylogeographical analysis (Wiens & Penkrot 2002). Despite et al. (2004) proposed a standard sequence threshold for flagging provisional species based on a single mitochondrial Correspondence: Jinzhong Fu, Fax: (519) 767 1656; E-mail: gene (COI): 10 times the mean intraspecific variation for [email protected] the group under study. However, the usefulness of mtDNA © 2008 The Authors Journal compilation © 2008 Blackwell Publishing Ltd 1470 J. FU and X. ZENG sequence data in determining species boundaries is not reproducing organisms. Frequency data of nuclear loci are universally agreed upon. Many have argued against the useful for inferring gene flow between populations, and use of mtDNA haplotype data as mtDNA trees always the absence of gene flow between groups can be used as display a bifurcating tree fashion of ancestor–descendant evidence for reproductive isolation. relationships due to its maternal and nonrecombining Stream salamanders of the genus Batrachuperus (family mode of inheritance (e.g. Davis 1996). This pattern may Hynobiidae) provide an excellent system to test the power extend above the level of species as in cases of interspecific of DNA sequence data in determining species boundaries hybridization (e.g. de Queiroz & Donoghue 1990; Davis because their morphology is very conservative while their 1996), or below the species level if there is strong geogra- genetic divergence is high. Several studies have found few phical population structure. In the latter case, the popula- useful morphological characters in distinguishing species tions involved may appear to be multiple species while (e.g. Fei et al. 1983; Zhao et al. 1988), while Fu et al. (2001) they are in fact part of one and the same species with described up to 10.07% pairwise percentage difference of extensive male-mediated gene flow between them (e.g. mitochondrial cytochrome b gene between different spe- Roberts et al. 2004). In addition, all mtDNA genes belong to cies. Frost (2007) listed seven species in the genus Batrachu- a single linkage group and evolve as one locus, and using perus: B. cochranae, B. karlschmidti, B. londongensis, B. pinchonii a single locus to delimit species is always problematic, both (type species of the genus), B. taibaiensis, B. tibetanus, and B. practically and philosophically. Many other authors used a yenyuanensis. Nevertheless, the basic taxonomy of Batrachu- combination of mtDNA, morphological and distributional perus is not well resolved, despite its abundance in museum data in their analyses (Puorto et al. 2001; Wiens & Penkrot collections and extensive morphological studies. For exam- 2002). Sites & Marshall (2004) reviewed 12 commonly used ple, among the seven species listed by Frost (2007), Fei et al. methods for delimiting species boundaries. Among them, (2005) considered B. karlschmidti a junior synonym of B. 11 require molecular data, and all four tree-based methods tibetanus, and B. taibaiensis as invalid (without providing use DNA haplotype sequence data. specific reasons), while Zhao & Adler (1993) considered Methods of species delimitation are highly dependent both B. londongensis and B. cochranae synonyms of B. pinchonii. on the definition of ‘species’, of which Mayden (1997) These disagreements were largely due to the limited number summarized a total of 22 different concepts. Despite the of consistent morphological characters available and the apparent disagreement, de Queiroz (1998) convincingly various interpretations of homology and polarity. Since argued that there is an underlying consensus that a species studies employing molecular data have the potential to is a ‘segment of population level lineage’ (= ‘the general sidestep some of these pitfalls, the present study may resolve lineage species concept’). In addition, both Mayden (1997) these disputes. Furthermore, the data are capable of reveal- and de Queiroz (1998) highlighted the distinction between ing hidden species diversity that morphology-based studies the ‘species concept’, which provides a conceptual definition may overlook. Numerous studies have repeatedly con- of species, and the ‘operational (species) criteria’ which are firmed that species diversity is often much higher than practical guidelines for diagnosis. Furthermore, de Queiroz revealed by morphology, particularly in morphologically (1998) argued that many various versions of existing species conserved groups such as salamanders (e.g. Good & Wake concepts are actually operational criteria. Among them, the 1992). The hynobiid salamanders are no exception. Many monophyly criterion is the most pertinent to phylogeograph- species have been described due to the application of ical analysis of DNA sequence data, and following this molecular techniques (i.e. Matsui 1987; Zeng et al. 2006). So criterion, species are essentially monophyletic groups on a far, salamanders of the genus Batrachuperus have not been gene tree. However, there are two difficulties associated exposed to extensive molecular examination. with this criterion. The first is determining the level of Batrachuperus species are primarily distributed in the monophyly that corresponds to the level of species. Sev- Hengduan Mountains, southwestern China, which have eral rules have been proposed, such as diagnosablility perhaps the most complex topography on Earth. They (Cracraft 1983), exclusivity (Baum & Shaw 1995), and amount therefore provide an excellent system to study interactions of intraspecific variation (Hebert et al. 2004). The second is between geography, geology and biodiversity. Five major recognizing ancestral species that retain plesiomorphic rivers run north to south in parallel, separated by sharply characters (Funk & Omland 2003). This is likely to occur rising mountains that often climb to altitudes above 5000 m. when peripheral populations speciate from a widespread The differences in altitude from valley to mountaintop ancestor (Talbot & Shields 1996), resulting in paraphyly of often exceed 2000 m, creating dramatic variations in the ancestral species. These difficulties are equally applicable vertical ecological zones, resulting in geographical iso- to both mtDNA and nuclear gene sequences. A second lation. Batrachuperus