Molecular Phylogenetics and Evolution 107 (2017) 538–550
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
Improved sampling at the subspecies level solves a taxonomic dilemma – A case study of two enigmatic Chinese tit species (Aves, Passeriformes, Paridae, Poecile)
Christian Tritsch a, Jochen Martens b, Yue-Hua Sun c, Wieland Heim d,e, Patrick Strutzenberger a, ⇑ Martin Päckert a, a Senckenberg Naturhistorische Sammlungen, Königsbrücker Landstrabe 159, D-01109 Dresden, Germany b Institut für Zoologie, Johannes Gutenberg-Universität, 55099 Mainz, Germany c Key Laboratory of Animal Ecology and Conservation, Institute of Zoology, Chinese Academy of Science, 100101 Beijing, China d Institut für Biochemie und Biologie, AG Tierökologie, Universität Potsdam, Maulbeerallee 1, 14469 Potsdam, Germany e Amur Bird Project, Roseggerstrabe 14, 14471 Potsdam, Germany article info abstract
Article history: A recent full species-level phylogeny of tits, titmice and chickadees (Paridae) has placed the Chinese Received 1 July 2016 endemic black-bibbed tit (Poecile hypermelaenus) as the sister to the Palearctic willow tit (P. montanus). Revised 25 November 2016 Because this sister-group relationship is in striking disagreement with the traditional affiliation of P. Accepted 9 December 2016 hypermelaenus close to the marsh tit (P. palustris) we tested this phylogenetic hypothesis in a multi- Available online 11 December 2016 locus analysis with an extended taxon sampling including sixteen subspecies of willow tits and marsh tits. As a taxonomic reference we included type specimens in our analysis. The molecular genetic study Keywords: was complemented with an analysis of biometric data obtained from museum specimens. Our phyloge- Poecile hypermelaenus netic reconstructions, including a comparison of all GenBank data available for our target species, clearly Poecile weigoldicus Multi-locus phylogeny show that the genetic lineage previously identified as P. hypermelaenus actually refers to P. weigoldicus Phylogeography because sequences were identical to that of a syntype of this taxon. The close relationship of P. weigoldi- DNA barcoding cus and P. montanus – despite large genetic distances between the two taxa – is in accordance with cur- rent taxonomy and systematics. In disagreement with the previous phylogenetic hypothesis but in accordance with most taxonomic authorities, all our P. hypermelaenus specimens fell in the sister clade of all western and eastern Palearctic P. palustris. Though shared haplotypes among the Chinese popula- tions of the two latter species might indicate mitochondrial introgression in this part of the breeding range, further research is needed here due to the limitations of our own sampling. Ó 2016 Elsevier Inc. All rights reserved.
1. Introduction pleteness of taxon sampling and the choice of target taxa at the species level also largely depend on the species concept applied Density of taxon sampling has been repeatedly evaluated as one and the taxonomic authorities consulted. Taking these basic con- of the crucial factors affecting the accuracy of phylogenetic analy- cerns into consideration, our study refers to a problematic sister- ses and the resulting topologies (Omland et al., 1999; Johnson, group relationship in a near-complete multi-locus phylogeny of 2001; Braun and Kimball, 2002; Zwickl and Hillis, 2002; Funk tits, Paridae (Johansson et al., 2013). and Omland, 2003; Heath et al., 2008; Albert et al., 2009; McKay Tits, titmice and chickadees of the speciose passerine family and Zink, 2010; Päckert et al., 2010; Nabhan and Sarkar, 2011). Paridae are widely distributed across Eurasia, Africa and North Adding missing taxa to an incomplete sampling might already America. Two diversity hotspots are situated in the western improve phylogenetic accuracy even with 50–90% missing Palearctic and along the eastern margin of the Qinghai-Tibet Pla- sequence information (e.g. only one out of n loci analyzed was teau (Fig. 1). Ancestral range reconstructions by Tietze and available for single taxa; Wiens and Tiu, 2012). However, com- Bothakur (2012) identified China as a cradle of diversity from where ancestors of Eurasian Paridae repeatedly colonized adjacent regions, such as the northern Palearctic, the Himalayas, Indonesia ⇑ Corresponding author. and the Philippines. High levels of species richness in the Heng- E-mail address: [email protected] (M. Päckert). duan Shan and the adjacent northern Chinese mountain systems http://dx.doi.org/10.1016/j.ympev.2016.12.014 1055-7903/Ó 2016 Elsevier Inc. All rights reserved. C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 539
Fig. 1. Worldwide distribution of tits, titmice and chickadees (Paridae); colours indicate local species richness from 1 (dark blue) to 11 (red) species; shapefiles downloaded from BirdLife International (2016). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) have been documented for several bird families and the complex cluster with all Chinese subspecies of the marsh tit (P. palustris) topography and landscape mosaic is regarded as the most impor- with this entire clade being the sister group of all willow tit (P. tant factor that made this region a center of avian endemism (Lei montanus) specimens (Fig. 2B). However, a multilocus phylogeny et al., 2006, 2014). This is reflected by pronounced intraspecific for all Paridae based on a near-complete sampling at the species phylogeographic structure in many bird species from China and level by Johansson et al. (2013) did not confirm this classification the adjacent eastern Himalayas (in Paridae, Parus minor: Zhao but recovered P. hypermelaenus as sister to the willow tit (P. mon- et al., 2012; Periparus ater: Martens et al., 2006; Tietze et al., tanus) with strong support (Fig. 2A). Apparently, the conflict 2011; Pentzold et al., 2013; in Aegithalidae, Aegithalos concinnus: between these two phylogenetic studies has remained unrecog- Dai et al., 2011). nized so far and cannot be resolved easily, because Dai et al. The first comprehensive phylogenetic family tree of Paridae was (2010) did not include western Palearctic marsh tit populations, exclusively based on the mitochondrial cytochrome-b gene (Gill whereas Johansson et al. (2013) did not include eastern Palearctic et al., 2005). As a consequence of this study several taxonomic marsh tit populations in their sampling. Clarification can only be authorities followed the lead of the American Ornithologists’ Union achieved with an improved taxon sampling including the Sichuan (AOU) who had already elevated the Nearctic Poecile and Baeolo- tit (P. weigoldicus) as a key taxon which was missing from most phus to genus rank (AOU, 1998) and divided the former Parus sensu previous phylogenetic studies (except Salzburger et al., 2002). lato into up to seven genera (former subgenera; see Gosler and This species is endemic to the south-western Chinese provinces Clement, 2007). Eight years later Johansson et al. (2013) published Sichuan, NW Yunnan, SE Qinghai and eastern Tibet (Harrap and the first near-complete multilocus species-level phylogeny includ- Quinn, 1996), where its breeding distribution widely overlaps with ing 56 parid species recognized by Gosler and Clement (2007) plus the scattered range of the black-bibbed tit, P. hypermelaenus. thirteen selected subspecies that raised their sampling to 69 parid Johansson et al. (2013) tentatively included P. weigoldicus in the taxa (covering all 67 Paridae species distinguished by Clements willow tit, P. montanus, following earlier classifications (Snow, et al., 2015 except one, see below). Apart from a confirmation of 1967; Quaisser and Eck, 2002; Gosler and Clement, 2007). Based the generic subdivision of Paridae, the new multi-locus phylogeny on its rather high genetic divergence from other P. montanus sub- included one rather surprising grouping in the Old World clade of species (4.6–5.9% K2P distance for cytochrome-b) documented by genus Poecile concerning the position of the black-bibbed tit (Poe- Salzburger et al. (2002) some authors had separated P. weigoldicus cile hypermelaenus). This species is a Chinese endemic with a from the willow tit at the species level before (see Table 1). Other restricted and presumably scattered distribution range in south- authors united all brownish-headed Asian willow tits and allies western China (Provinces Shaanxi, Sichuan, Yunnan, Guizhou – under the species name P. songarus (the ‘‘songar tit” including Cen- and probably also outliers to Hubei, Gansu and south-eastern tral Asian P. s. songarus and Chinese P. s. affinis, P. s. stoetzneri and P. Tibet) and isolated populations in the Chin Hills (south-western s. weigoldicus; see Table 1), but so far this arrangement has not Myanmar; Harrap and Quinn, 1996). been corroborated by molecular phylogenetics (Kvist et al., 2001; Classifications based on comparative morphology and biomet- Salzburger et al., 2002). A recent phylogeographic study by Song rics assigned P. hypermelaenus to close kinship of P. palustris et al. (2016) compared Chinese marsh tit and willow tit popula- (Vaurie, 1957; Harrap and Quinn, 1996; Eck, 2006) and most taxo- tions only without considering their trans-Palearctic counterparts nomic authorities agreed on this relationship (compare Table 1). and thus did not provide a finite solution of the systematic- This taxonomic arrangement received first support from a phy- taxonomic dilemma either. logeny of Asian parid species based on two mitochondrial markers With this study we aim at clarifying the phylogenetic relation- (Dai et al., 2010). Though their overall taxon sampling for Paridae ships of the two Chinese endemic Poecile species, P. weigoldicus was less than complete, sequences of P. hypermelaenus formed a and P. hypermelaenus with particular respect to the contradictory 540 C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550
Table 1 Taxonomy and Systematics of Old World Poecile tits according to the classification by different authorities (only taxa considered in our study listed); taxonomic classifications into: two species = Gosler and Clement, 2007; three species = Johansson et al., 2013 (only taxa discussed in their paper are listed); four species A = Clements Checklist, 2013; Dickinson, 2003; Harrap and Quinn, 1996; four species B = Clements Checklist, 2015; Gill and Donsker, 2014; Dickinson and Christidis, 2014; Eck and Martens, 2006; ⁄ = ssp. jeholicus synonymized with hellmayri by some of the authors; distribution according to Dickinson and Christidis (2014).
Three species Four species A Four species B Two species Willow tits and allies Distribution Willow Tit (P. montanus) montanus montanus montanus montanus Alps and Carpathian Mts to N Greece rhenanus rhenanus rhenanus W Europe S to France salicarius salicarius salicarius C Europe borealis borealis borealis Scandinavia, EC Europe uralensis uralensis uralensis W Siberia, SE Russia, N Kazakhstan baicalensis baicalensis baicalensis C Siberia, Altai, Russian Far East to Korea and N China kamtschatkensis kamtschatkensis kamtschatkensis Kamchatka, N Kuril Isls anadyrensis anadyrensis anadyrensis NE Siberia sachalinensis sachalinensis sachalinensis Sakhalin, Japan Hokkaido restrictus restrictus restrictus Japan Honshu Songar Tit (P. songarus) songarus songarus songarus songarus E Kazakhstan, Tien Shan, NW China affinis affinis affinis affinis C China (E Qinghai to SW Shaanxi) stoetzneri stoetzneri stoetzneri stoetzneri NE China (Shanxi, Nei Mongol to Hebei) Sichuan Tit (P. weigoldicus) weigoldicus weigoldicus P. weigoldicus weigoldicus WC China (E Xizang, Sichuan, N Yunnan) Marsh tits and allies Marsh Tit (P. palustris) palustris palustris palustris palustris W Europe to Scandinavia, SE to Balkans brevirostris brevirostris brevirostris SE Siberia, Russian Far East to N Korea, N Mongolia and NE China ernsti ernsti ernsti Sakhalin hellmayri hellmayri hellmayri S Korea jeholicus⁄ jeholicus⁄ jeholicus NE China (S of brevirostris) hensoni hensoni hensoni S Kuril Isls, N Japan Black-bibbed Tit (P. hypermelaenus) P. hypermelaenus P. hypermelaenus P. hypermelaenus hypermelaenus C and SW China, W Myanmar
A P. superciliosus P. lugubris lugubris P. davidi P. palustris marsh tit 1 P. lugubris hyrcanus 0.99 IOZ1239 Palearctic/ Chinese 1 P. palustris hypermelaenus P. montanus (W Palearctic) willow tits 0.95 + 1 P. montanus affinis black-bibbed tit Nearctic Poecile
B P. palustris hypermelaenus IOZ1239 Gansu 1 Chinese P. palustris hypermelaenus IOZ7991 Hubei marsh tits 1 Shanxi P. palustris hellmayri IOZ8680 + P. palustris hellmayri IOZ2458 Shaanxi P. palustris brevirostris IOZ5675 Jilin black-bibbed tit 1 P. montanus baicalensis IOZ5944 Heilongjiang 1 P. montanus baicalensis IOZ5954 Heilongjiang P. montanus salicarius IOZ2237 Slovakia Palearctic/ Chinese 0.99 P. montanus affinis IOZ2178 Shaanxi willow tits
Fig. 2. Competing phylogenetic hypotheses on the relationships of the black-bibbed tit, P. hypermelaenus by (A) Johansson et al. (2013; schematically redrawn) and (B) Dai et al. (2010; re-analyzed from the original data set and the original settings with MrBayes; Markov chain 2,000,000 generations, burnin = 2000; tree including branch lengths and posterior support values not shown in Dai et al. 2010); IOZ catalogue numbers of the relevant specimens indicated at the tip clades. placement of the latter species in the phylogenies by Dai et al. sequence data derived from type specimens in our analyses. (2010) and Johansson et al. (2013) (compare Fig. 2). Our sampling Our genetic analyses are complemented by comparison of bio- is the first to include populations from all across Eurasia from metric traits taken from museum specimens. In the following both willow tits and marsh tits and from their Chinese relatives we use the taxonomy by Clements et al. (2015) who treated the in the same data set covering a total of eighteen taxa. In order two Chinese endemics P. hypermelaenus and P. weigoldicus as sep- to enable an authoritative taxonomic classification we included arate species. C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 541
2. Material and methods and ODC followed the protocols in Johansson et al. (2013). For amplification of the COI gene, the primers BirdF1 and BirdR1 2.1. Molecular genetics (Hebert et al., 2004), each one modified with a 50 M13 tail, were used (forward-primer: M13F(-21) - BirdF1, reverse-primer: 2.1.1. Material M13R-pUC(-26) - BirdR1. The PCR profile was: 94 °C for 1 min, Our genetic analysis relied on 46 fresh samples of Poecile spe- 35 cycles of 94 °C for 1 min, 48.5 °C for 1 min, 72 °C for 1:30 min, cies including the Sichuan tit (P. weigoldicus), the black-bibbed tit final elongation of 5 min at 72 °C). Sequencing was performed (P. hypermelaenus), the sombre tit (P. lugubris), the marsh tit (P. using the non-modified primer pair BirdF1 and BirdR1 (Hebert palustris) with five subspecies and the willow tit (P. montanus) with et al., 2004). ten subspecies (see Supplementary Table S1). Our sampling was For museum specimens, we amplified the two mitochondrial completed by nine samples taken from historical specimens markers in multiple overlapping fragments of about 200 bp in (Figs. S1 and S2) and most of these represented type specimens length: ND2: 1091 bp in seven fragments using primer pairs pro- of Chinese target taxa (listed in Eck and Quaisser, 2004). vided in Johansson et al. (2013) with slight modifications; COI: 447 bp in three fragments using specifically designed primer pairs (i) one syntype of P. Salicarius weigoldi Kleinschmidt, 1921 (by for Poecile tits (for primer combinations and detailed PCR profiles reasons of homonymy amended to P. weigoldicus; refer to Supplementary Table S2). PCRs using DNA extracts from Kleinschmidt, 1921: Falco 17 (2), 4); MTD C23803, a male toepads were performed in two multiplex reactions, each of five Ò collected by H. Weigold at Mauntschi (Barongschiba), north non-overlapping fragments using the Type-it microsatellite PCR of Batang (Sichuan province) on 16th October 1915 (see kit (Qiagen) according to the manufacturers’ instructions (without comments on this type series in Eck and Martens, 2006 adding the optional Q-Solution). The first amplification step was and Lecroy, 2010); another male from the Weigold series, followed by a specific re-amplification of each fragment (from a MTD C23808, collected at Janeti ‘‘3 days west of Batang” 1:10 dilution of the original PCR product) in a second PCR. To on 29th September 1915; exclude possible contamination, negative controls were always (ii) the holotype and three paratypes of P. Salicarius stoetzneri performed for each multiplex PCR. PCR products were purified Kleinschmidt, 1921; all collected by H. Weigold, 30 m km using ExoSAP-IT (GE Healthcare) according to the manufacturer’s north of Balihandien, NE of Jehol [presently Chengde]; advice. MTD C23845, holotype, male, 30th April 1916; MTD For sanger-sequencing a standard protocol of 25 cycles with C25168, paratype, female, 29th April 1916; MTD C25169, 10 s at 96 °C, 5 s at 50 °C and 4 min at 60 °C was used for all genes paratype, unsexed, 29th April 1916; MTD C25170, paratype, and samples and both strands were sequenced. The products were female, 30th April 1916; purified and sequenced on an ABI 3130xl capillary sequencer (iii) two syntypes of Parus communis jeholicus Kleinschmidt and (Applied Biosystems). The concatenated sequences were aligned Weigold, 1922; both collected by H. Weigold ‘‘30 m km manually using Mega version 5.10 (Tamura et al., 2011) and exam- north of Balihandien, NE of Jehol”; MTD C25165, male, ined by eye and controlled for possible stop codons or frame shifts 27th April 1916; MTD C40803, male, 25th April 1916. in coding sequences. (iv) One additional specimen of Poecile montanus songarus (Sev- ertzov, 1873): MTD C48361, male, 15th Octobert 1962, col- 2.1.3. Sequence analysis lected by Krylow, at lake Issyk Kul, Kyrgyzstan; this For comparison with previous studies we added 24 ND2 specimen is not a type. sequences (most from Johansson et al., 2013) and 65 COI sequences (including those from Dai et al., 2010) from GenBank to our data set (see Supplementary Table S1). We reconstructed all phyloge- 2.1.2. DNA extraction, PCR and sequencing netic trees using Bayesian inference of phylogeny (BI) with For DNA extraction from fresh tissue we used the innuPREP MrBAYES vers. 3.2.3 (Ronquist and Huelsenbeck, 2003) or BEAST DNA Mini Kit (for muscle tissue) or the innuPREP BloodDNA Mini vers 1.8.1 (Drummond et al., 2012) and Maximum Likelihood Kit (for blood) respectively (both Analytik Jena AG, Germany). For (ML) using RAxML (Stamatakis, 2006, 2014). DNA extraction from dry toe pad tissue of museum specimens Testing for the appropriate substitution models was done using we used the LGC sbeadex Forensic Kit (LGC genomics, Berlin, Ger- PAUP vers. 4.0 (Swofford, 2003) and MrModeltest vers. 2.2 many). All procedures were carried out according to the manufac- (Nylander, 2004). The best fitting partition scheme for the multi- turer’s instructions except for overnight incubation of toepad gene phylogenies was identified with the rapid hill climbing tissue with proteinase K (instead of one hour) and only 60 ll elu- method in RaxML vers. 8.1 (Stamatakis, 2014). tion volume (instead of 100 ll) in order to yield a sufficiently high We reconstructed a single-locus mtDNA phylogeny with the concentration of DNA extracts. All toe pad samples were analyzed ND2 data set in two different analyses with MrBayes and raxML in a separate clean lab. There, each step of analysis (sampling, (i) in a single partition under the GTR + Gamma + Invariant Sites extraction and PCR) was done on separate working benches. In (GTR + C + I) substitution model, (ii) partitioned by codon position order to avoid cross-contamination working benches were cleaned under the GTR + Gamma + Invariant Sites (GTR + C + I) substitution with DNA-away (Molecular Bio Products, Inc.), after each step all model. In the latter analysis we allowed the overall rate to vary benches and lab rooms were decontaminated with UV-light for between partitions by setting the priors hratepr = variablei and at least four hours. model parameters such as gamma shape, proportion of invariable For phylogenetic comparison, we chose the same markers as sites and the rate matrix were unlinked across partitions. We con- used in the multi-locus analyses by Johansson et al. (2013): the ducted two runs per analysis with one cold and three heated protein coding mitochondrial gene NADH-dehydrogenase subunit2 MCMC chains of 1,000,000 generations with trees sampled every (ND2) was sequenced for all samples and in addition sequences of 100 generations and a burn-in of 3000 trees. The ND2 ML tree two nuclear introns myoglobin (myo) and ornithine decarboxylase was generated with RAxML v. 7.3.0 in raxmlGUI1.1 (Silvestro and (ODC) were generated for at least one specimen of each target Michalak, 2012) in 1000 replicates of an ML + thorough Bootstrap taxon. As a fourth additional gene, we sequenced the DNA barcod- search under the GTR + C + I model. ing marker cytochrome-c oxidase subunit I (COI) for all samples. A multigene phylogeny was reconstructed with BEAST from a Amplification and sequencing procedures for ND2, myoglobin- combined dataset of ND2 (1041 bp, three partitions by codon 542 C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 position, GTR + C + I model) and the nuclear introns myo (730 bp, with P. hypermelaenus. Means were compared in a one-way ANOVA one partition, HKY + C model) and ODC in (643 bp, one partitions, including Levene’s test to assess the equality of variances for the GTR + C model). We ran BEAST for 30,000,000 generations (param- taxon groups compared, because our sample sizes differed greatly eters were logged and trees sampled every 1000 generations) among taxon groups. Levene’s test revealed non-equality of vari- under the uncorrelated lognormal clock model for all loci with ances for four out of eight variables at a significance level, the ‘‘auto-optimize” option activated and a birth-death process p < 0.05. Therefore, to test for significance of multiple comparisons prior applied to the tree. The log files were examined with Tracer among groups we used Dunnett’s T3 test as a post-hoc test for 1.6 (Rambaut et al., 2014) in order to ensure effective sample sizes unequal variances and small sample sizes (for P. hypermelaenus (ESS; which yielded reasonable values for all parameters after and P. weigoldicus). For within-species comparison among sub- 30,000,000 generations) and trees were summarized with species groups (including the enigmatic forms weigoldicus and TreeAnnotator v1.4.8 (posterior probability limit = 0.5) using a hypermelaenus) we performed a principal component analysis burn-in value of 3000 trees and the median height annotated to (PCA) and further statistical tests with SPSS 14.0. each node. The multigene ML analysis was performed using raxML- In the following, we limit the presentation and discussion of our GUI1.1 as described above with five partitions, but using GTR + C results to the intraspecific and interspecific relationships of willow + I model for all partitions. tits and marsh tits without going into those details of the entire All obtained phylograms were edited in FigTree vers. 1.4.2 Poecile clade that have already been comprehensively discussed (Rambaut, 2012). by Johansson et al. (2013). In addition, mitochondrial haplotype networks of the mtDNA markers were produced separately for the Willow Tits and for 3. Results the Marsh Tits (COI: 694 bp, aDNA sequences only 427 bp) using TCS vers 1.2.1 (Clement et al., 2000). 3.1. Molecular genetics We used a General Mixed Yule Coalescent (GMYC) (Fujisawa and Barraclough, 2013) approach to delimit genetic clusters in The topology of the Palearctic Poecile crown clade was largely our dataset in order to further corroborate our taxonomic hypothe- congruent among single-locus and multi-locus reconstructions ses. We used BEAST v1.8.1 to estimate trees based on a 144 and was divided into two major clades (Figs. 3 and 4). The strongly sequences COI alignment as BEAST was found to produce best- supported marsh tit clade was formed by one sister species pair P. performing trees for subsequent GMYC analyses (Talavera et al., palustris and P. hypermelaenus. The willow tit clade included three 2013; Kekkonen et al., 2015). BEAST runs were performed with a species: (i) the willow tit, P. montanus that was divided into three chain length of 3.3 10E7 states, with trees being sampled every genetic lineages (see below), (ii) the Sichuan tit, P. weigoldicus and 3000th state. We applied a GTR + I + G model with empirical base (iii) the Caspian tit, P. hyrcanus. Monophyly of these three was frequencies to each codon position. A single clock model set to log- strongly supported in all reconstructions. However, phylogenetic normal relaxed clock with an estimated rate was used. One tree relationships among these three were not unambiguously resolved estimation each was performed using a Yule tree prior and alterna- (in the ND2 tree P. montanus was sister to P. weigoldicus, whereas in tively a Coalescent: Constant Size prior. All other priors were left at the multi-locus tree P. montanus was sister to P. hyrcanus; neither the default values except for ucld.mean which was set to an expo- of these topologies received reliable node support; Figs. 3 and 4). nential prior with mean = 0.01. Tracer v1.6 was used for examina- Delimitation of evolutionary significant clusters using GMYC tion of log files for convergence and sufficient effective sample identified 22 entities. Trees reconstructed with constant size and sizes. Trees were summarized with TreeAnnotator v1.8.1 where Yule tree prior resulted in the same clustering scheme. Comparison the first 1000 trees were removed as burnin and median heights to the null model with LR tests resulted in p < 0.001 for both trees were annotated to the maximum clade credibility tree. GMYC clus- demonstrating that the obtained classification is significantly dif- ter delimitation was performed with the R package splits (Fujisawa ferent from the null model. and Barraclough, 2013) using the single threshold model. Resulting classifications were compared to the null model using the LR test implemented in ‘splits’. 3.1.1. The marsh tit clade In the un-partitioned ND2 tree (Fig. 4) marsh tits, P. palustris, 2.2. Morphometry were subdivided in a well supported western subclade (subspecies palustris) and an eastern subclade (subspecies brevirostris, hell- For comparison of body and feather dimensions we measured mayri, hensoni, jeholicus) that received only poor support. The seven morphological traits for a total of 264 individual whole skins east-west split was reflected in the Bayesian multi-locus tree and from two bird collections: Senckenberg Natural History Collections corresponded to two separate COI haplotype clusters separated Dresden (SNSD) and Museum für Naturkunde Berlin (MFN). Mea- by a minimum of three substitutions only (Fig. 5). surements were taken according to the standards in Eck et al. The most common haplotype of the western group was found in (2011) for: bill length (bill-to-skull, bsk), bill height (bp) and bill eleven out of thirteen individuals. Ten haplotypes of the eastern width (bwp), wing length (maximum chord, wmax), Kipp’s dis- cluster differed by a maximum of five substitutions (n = 25). None tance (kipp), tail length (t1) and tarsus length (t2; compare Eck of the tip haplotypes differed by more than two steps from either et al., 2011). As derived traits we calculated a relative tail index of the two central haplotypes (both n = 7, Fig. 5b). In most phyloge- (=t1/wmax) and roundedness index (kipp/wmax) for comparison netic reconstructions, the eastern clade collapsed into a basal poly- with mean values for this frequently used trait from the literature. tomy, e.g. in the partitioned ND2 analysis and in raxML All measurements were taken by W.H. using a digital caliper (digi- reconstructions. GMYC analysis of COI barcode sequences did not Max; Ecotone, Poland) with an accuracy of 0.01 mm and a magni- identify two separate marsh tit clusters either but separated P. fier lamp (Waldmann RLL 122T, magnification (120 mm £) 2.5, 4 hypermelaenus from P. palustris (Fig. S3). diopter) to achieve maximum precision. We calculated parameter In the ND2 tree and the COI tree (Figs. 4, S1), the sister clade of means for subspecies groups according to genetic clusters identi- P. palustris included our P. hypermelaenus specimens and a third fied by molecular genetic analysis. According to the classification sequence inferred from one of the two P. p. jeholicus type speci- by Eck and Martens (2006; also Cheng, 1987) we grouped speci- mens (MTD C40803). In contrast, the second P. p. jeholicus type mens of P. palustris dejeani [loc. typicus Tsékou, N Yunnan, China] (MTD C25165) was firmly nested in P. palustris (Figs. 4, S1). In C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 543 Marsh Tits 1 / 100 P. p. palustris P. p. palustris* West Palearctic 1 / 99 P. p. jeholicus P. p. brevirostris East Palearctic 1 / 100 0.99 / - P. p. hellmayri P. p. hensoni P. hypermelaenus Black-bibbed Tit P. m. sachalinensis P. m. rhenanus P. m. baicalensis 1 / 93 P. m. montanus* Willow Tits 1 / 71 P. m. restrictus North Palearctic 0.86 / 47 P. m. kamtschatkensis P. m. salicarius P. m. montanus 1 / 82 P. m. songarus Central Asia 1 / 76 P. m. affinis* 0.65 / 59 1 / 99 P. m. affinis SW China 1 / 95 P. m. stoetzneri P. hyrcanus* CaspianTit 1 / 100 P. weigoldicus „P. p. hypermelaenus“ * IOZ1239 SichuanTit 0.99 / 64 P. davidi* P. carolinensis* 0.98 / 43 1 / 100 P. atricapillus 0.45 / 25 1 / 74 P. atricapillus* Poecile P. gambeli* 1 / 91 1 / 95 P. sclateri* 1 / 100 P. cinctus* 1 / 89 P. hudsonicus* P. rufescens* 1 / 70 1 / 100 P. superciliosus 0.82 / - P. superciliosus* 1 / 100 P. lugubris anatoliae 1 / 90 P. l. lugubris* 1 / 100 Sittiparus semilarvatus* Sittiparus varius* 1 / 100 Lophophanes cristatus* Lophophanes dichrous*
0.07
Fig. 3. Multilocus phylogeny of Poecile tits based on three markers ND2 (1041 bp), myoglobin (643 bp) and ODC (730 bp); Bayesian inference of phylogeny with BEAST, node support from Bayesian posterior probabilities followed by ML bootstrap support indicated at nodes, missing values (–) indicate that a node was not represented in the respective analysis; asterisk denotes sequences inferred taken from GenBank (all from Johansson et al., 2013; the position of the crucial ‘‘P. p. hypermelaenus” specimen IOZ1239 is indicated at the corresponding clade). the COI tree the P. hypermelaenus subclade comprised the two ‘‘P. p. relationships received strong support. The three willow tit lineages hypermelaenus” specimens from Dai et al. (2010; specimens corresponded to the same COI haplotype clusters in the minimum IOZ1239 and IOZ7791) plus two further GenBank sequences of P. spanning network (Fig. 5). The northern Palearctic haplotypes p. hellmayri from the same study (Fig. S3). In contrast, the P. p. bre- formed a central cluster with the most common central haplotype virostris specimen IOZ5675 from Dai et al. (2010) clustered with P. found in 22 (out of 46) individuals. That most frequent haplotype palustris (Fig. S3). The truly deep divergence between P. hyperme- was separated from the Central Asian P. m. songarus haplotype by laenus and P. palustris as inferred from our sequence data (Figs. 3 10 substitutions and from the most common Chinese haplotype and 4) could not have been inferred from the tree with unscaled (affinis-stoetzneri group; found in 10 individuals) by 8 substitu- branches published by the latter authors (compare Fig. 2B). tions. The same three genetic clusters of willow tit subspecies- groups were also identified by GMYC analysis of COI barcode 3.1.2. The willow tit clade sequences (Fig. S3). In all phylogenetic reconstructions willow tits were divided into All our five P. weigoldicus ND2 sequences, including the Dresden three separate genetic lineages (Figs. 3 and 4, S3): (i) northern syntype MTD C23803, formed a strongly supported clade (Fig. 5). Palearctic populations of the montanus-group (subspecies mon- Surprisingly, the ‘‘P. p. hypermelaenus” sequence from specimen tanus, salicarius, rhenanus, baicalensis, uralensis [COI], IOZ1239 by Johansson et al. (2013) was firmly nested in this clade kamtschatkensis, anadyrensis [COI], sachalinensis, restrictus), (ii) and was identical with one of our sequences (MAR1865). All P. wei- Central Asian P. m. songarus, (iii) Chinese P. m. affinis and P. m. goldicus samples shared the same COI barcode haplotype (includ- stoetzneri. The northern Palearctic clade and the Chinese clade ing the type specimen MTD C23803) that was separated from the received strong support in all reconstructions (Central Asian P. m. central haplotype of the northern Palearctic montanus cluster by songarus was represented by a single specimen only). Phylogenetic 20 substitutions (Fig. 5). One further COI sequence from GenBank relationships among these three differed among single-locus and (from a specimen identified as P. m. affinis) was identical to the multi-locus reconstructions and none of the sister-group P. weigoldicus COI haplotype (Fig. S3). 544 C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550
P. p. palustris (1) P. p. palustris (2) 1 / 97 P. p. palustris (3) P. p. palustris (4) P. palustris (West Palearctic)
P. p. palustris (5) Marsh Tits P. palustris (6) 1 / 95 P. p. hellmayri (1) P. p. hellmayri (2) P. p. brevirostris (1) P. p. brevirostris (2) P. p. jeholicus (1) P. palustris (East Palearctic) 1 / 100 0.47 / 64 P. p. jeholicus (2) syntype P. p. jeholicus (3) P. p. hensoni (1) P. p. hensoni (2) 1 / 99 P. hypermelaenus (1) P. hypermelaenus 1 / 96 P. hypermelaenus (2) „P. p. jeholicus“ (4) syntype Black-bibbed Tit
P. m. sachalinensis (1) P. m. salicarius (1) P. m. salicarius (2) P. m. salicarius (3) P. m. salicarius (4) P. m. kamtschatkensis (1) P. m. kamtschatkensis (2) P. m. kamtschatkensis (3) P. m. baicalensis 0.97 / 70 P. m. salicarius (6) P. montanus P. m. montanus (1) P. m. restrictus (1) (North Palearctic) 0.99 / 88 P. m. rhenanus (1)
P. m. salicarius (5) Willow Tits P. m. montanus (2) P. m. rhenanus (2) 1 / 93 P. m. montanus (3) P. m. rhenanus (3) P. m. rhenanus (4) P. m. songarus P. m. songarus P. m. stoetzneri (1) P. m. stoetzneri (2) P. m. affinis (1) 0.44 / 38 P. m. affinis (2) 0.76 / 68 P. m. affinis (3) 0.69 / 42 1 / 100 P. m. affinis (4) P. m. affinis - P. m. stoetzneri (3) holotype P. m. stoetzneri (4) paratype P. m. stoetzneri P. m. stoetzneri (5) paratype P. m. stoetzneri (6) paratype P. m. stoetzneri (7) 1 / 91 P. m. stoetzneri (8) P. weigoldicus (1) 0.25 / - P. weigoldicus (2) 1 / 100 P. weigoldicus (3) P. weigoldicus „P. p. hypermelaenus“ (3)* IOZ1239 Sichuan Tit P. weigoldicus (4) syntype P. weigoldicus (5) P. hyrcanus P. hyrcanus P. davidi Caspian Tit 0.73 / 56 1 / 100 P. lugubris anatoliae 1 /100 P. lugubris (1) P. lugubris lugubris (2) 1 / 98 P. atricapillus (1) 0.9 / 51 P. atricapillus (2) 1 / 100 P. gambeli (1) P. gambeli (2) 1 / 79 0.96 / 58 1 / 100 P. carolinensis (1) 0.7 / 72 P. carolinensis (2) P. sclateri 0,49 / 54 1 / 100 P. cinctus 1 / 59 P. hudsonicus 1 / 100 0.49 / 72 P. rufescens (1) P. rufescens (2) 1 / 95 Sittiparus semilarvatus 0.4 / - Sittiparus varius 1 / 100 P. superciliosus (1) P. superciliosus (2) 0.98 / 93 Lophophanes cristatus Lophophanes dichrous
0.05
Fig. 4. Single-locus tree based on 1041 bp of the mitochondrial ND2; Bayesian inference of phylogeny with BEAST, node support from Bayesian posterior probabilities followed by ML bootstrap support indicated at nodes, missing values (–) indicate that a node was not represented in the respective analysis; type specimens indicated in bold; the position of the crucial ‘‘P. p. hypermelaenus” specimen IOZ1239 is indicated at the corresponding clade; numbers behind taxa indicate consecutive specimen number as listed in the material Table S1.
3.2. Morphometry explained a further 11.8% of the total variance – thus together the first two components explained a cumulative 96.3% of the total Principal component analysis (PCA) included a total of 119 variance. PC2 was most heavily loaded by wing length and Kipp’s marsh tit and black-bibbed tit specimens. The first principal com- distance and negatively loaded by tail length. The scatterplot of ponent had an eigenvalue of 24.7 and explained 84.6% of the total PC1 versus PC2 reflected the wide central Palearctic distribution variance. It was most heavily positively loaded by wing length and gap of that species, because eastern and western marsh tits largely tail length. The second component had an eigenvalue of 3.5 and represented two clusters with only marginal overlap (Fig. 6A). C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 545
Fig. 5. Distribution of DNA barcode (COI) haplotypes of (A) willow tits and Sichuan tit (P. montanus, P. weigoldicus; 585 bp), (B) marsh tits and black-bibbed tit (P. palustris, P. hypermelaenus; 580 bp); minimum spanning networks shown in boxes at the lower left side, colours indicate distinct haplotype clusters; open circles = COI sequences from Dai et al. (2010), site coordinates were roughly estimated according to the Chinese provinces listed in this study; distribution areas of target species according to shape files downloaded from BirdLife International (2016); the map in (A) includes the range of P. hyrcanus, the Caspian vicariant species of P. montanus and P. weigoldicus (pink range only at the southern coast of the Caspian Sea; note that BirdLife International (2016) did not separate this form from Western Palearctic P. lugubris). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Morphometric variation was characterized by greater feather length. The scatterplot of PCA1 versus PCA2 did not show distinct dimensions (longer wings and tails) in eastern birds. Strikingly, clusters. Generally, northern Palearctic willow tits (montanus the Chinese black-bibbed tit, P. hypermelaenus, did not resemble group) ranged at smaller feather dimensions compared to their Chinese eastern marsh tits, but specimens ranged at the opposite Chinese counterparts along the PC1 axis (shorter wings and tails; extreme of the scatterplot with smallest feather dimensions Fig. 6B). Along the PC2 axis Chinese P. weigoldicus and Central Asian (Fig. 6A). P. m. songarus ranged at greatest feather dimensions (longer and In comparison, we analyzed a total of 145 willow tit and more pointed wings [due to greater Kipp] but relatively shorter Sichuan tit specimens. The first principal component had an eigen- tails; Fig. 6B). value of 19.1 and explained 78.2% of the total variance. Like in Among all species analyzed P. hypermelaenus stood out as the marsh tits PC1 was most heavily positively loaded by wing length smallest and extremely short-tailed species with lowest Kipp’s dis- and tail length. The second component had an eigenvalue of 3.8 tance (Fig. 7; significant differences in kipp only between P. hyper- and explained a further 15.7% of the total variance – thus together melaenus and four other taxa, Dunnet T3 test, p < 0.05). Within the the first two components explained a cumulative 93.9% of the total large variation of kipp in the northern Palearctic montanus group variance. Again like in marsh tits, PC2 was most heavily loaded by characteristic outliers corresponded to subspecific variation: low- wing length and Kipp’s distance and negatively loaded by tail est values were observed for P. m. rhenanus and P. m. salicarius from 546 C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550
3 4 A B 3 2
2 1 1 0 PC2 PC2 0
-1 -1
-2 -2
-3 -3 -3 -2 -1 0 1 2 3 -4 -3 -2 -1 0 1 2 3 PC1 PC1 palustris West palustris East montanus weigoldicus hypermelaenus songarus affinis/ stoetzneri
Fig. 6. Intra- and interspecific morphometric variation of (A) marsh tit and black-bibbed tit (P. palustris, P. hypermelaenus) and (B) willow tit and Sichuan tit (P. montanus, P. weigoldicus); scatterplot of principal components 1 and 2 from principal component analysis.
80 15 A B kipp
14 52 52 245 122 70 13 50 212 122 38 54 228 56 57 59 12
60 11
234 162 232 10
50 9 249 234 230 233 * wing 8 71 232 * 248 * 40 tail 7 * N = 82 7 42 14 53 49 6 N = 83 7 41 13 53 49 6 E W W E affinis affinis montanussongarus montanussongarus palustris weigoldicuspalustris weigoldicuspalustrispalustris hypermelaenus hypermelaenus
Fig. 7. Interspecific morphometric variation among Old World Poecile tits of (A) wing length and tail length, (B) Kipp’s distance; boxplots with means (black bars), quartiles (boxes) and standard deviation (lines); outliers indicated by open circles and stars; significant differences of Kipp’s distance between P. hypermelaenus and other species indicated by asterisk (Duncall T3 test, p < 0.05). western and central Europe, whereas highest values were observed Sarkar, 2011). Further drawbacks of phylogeny reconstruction for P. m. baicalensis and P. m. anadyrensis from Mongolia, the Rus- might result from imperfect taxonomy in cases where cryptic spe- sian Altai, Nei Mongol and one of our northernmost records from cies are neglected due to lumping of allegedly intraspecific taxa Russian Autonomous Region of Tchukotka (Fig. 7B; outliers mon- (Funk and Omland, 2003; McKay and Zink, 2010). It must be tanus group). emphasized that the study by Johansson et al. (2013) was indeed Contrary to the genetic findings, all P. palustris jeholicus type based on an exhaustive sampling at the species level and provided specimens differed greatly from P. hypermelaenus specimens in lar- the first robust phylogenetic hypothesis for all parid genera. In ger body size proportions (Fig. S2). their conclusive taxonomic recommendations Johansson et al. (2013) suggested upgrading only two out of thirteen selected parid subspecies to species status. Among these P. weigoldicus was miss- 4. Discussion ing, though its genetic distinctiveness had been documented before (Salzburger et al., 2002) and it was accordingly separated 4.1. Incomplete taxon sampling and misidentifications as a source of from P. montanus at the species level by several authors (Eck and phylogenetic error Martens, 2006; Dickinson and Christidis, 2014; Gill and Donsker, 2014; Clements et al., 2015). In that respect taxon completeness Poor density of taxon sampling has been frequently criticized as is a matter of the investigator’s perspective, i.e. the species concept a major pitfall in phylogenetic studies (review in Nabhan and applied and the taxonomic authority consulted. Our results show C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 547 that phylogenetic relationships of P. hypermelaenus could not be C23808). Measurements of the syntype matched those of other P. reliably evaluated without inclusion of further missing taxa at weigoldicus specimens analyzed and size dimensions given in the the subspecies level. In a global analysis Phillimore and Owens literature with respect to longer wings and relatively shorter tails (2006) documented that about 36% of all avian subspecies ana- (compared to P. hypermelaenus and Asian P. montanus except P. lyzed were phylogenetically distinct and thus represent relevant m. songarus; Table 2). Therefore, the large-sized populations of units to be considered in evolutionary and conservation biology. the southern Chinese genetic lineage nested in the P. montanus/P. Inclusion of single subspecies or missing species in a rather dense hyrcanus clade must be named P. weigoldicus and do not corre- taxon sampling can result in unexpected tree topologies that might spond to P. p. hypermelaenus as understood by Johansson et al. contradict previous systematic classifications, like in case of (2013). Cyanistes teneriffae cyrenaicae from Libya (Päckert et al., 2013; Gohli et al., 2015; Stervander et al., 2015) or in some Tibetan spe- 4.3. Phylogenetic relationships and systematics cies of Carpodacus rosefinches (C. waltoni, C verreauxi: Tietze et al., 2013; C. sillemi: Sangster et al., 2016). The genetic distinctiveness of P. weigoldicus against all major Apart from the density of taxon sampling, incorrect taxonomy genetic lineages of P. montanus was already inferred from single- was suggested to be one major cause for species-level paraphyly locus analyses by Salzburger et al. (2002). Like the latter study in avian phylogenetic studies, particularly in cases of taxon pairs our multi-locus analysis did not support a monophyletic ‘songar that were ‘‘notoriously difficult to identify and may not be separate tit’ clade (including the taxa songarus, affinis, stoetzneri and weigol- species” (Funk and Omland, 2003; McKay and Zink, 2010). In fact, dicus). These Asian populations had been previously united as a distinction of P. palustris against P. montanus by morphological distinct subspecies group (songarus-group: Eck, 1980, 2006; but means is challenging (Broughton, 2009; Beaman and Madge, Eck and Martens, 2006 excluded P. weigoldicus) or separated from 1998), and difficulties rather increase when statistical analyses P. montanus as a species of its own (Clements et al., 2013; encompass large intraspecific variation. PCA of biometric traits Dickinson, 2003; Harrap and Quinn, 1996). However, a species- could not identify clusters that would have matched any of the level taxon named ‘songar tit’ (P. songarus according to classifica- phylogroups of our target taxa – except eastern and western P. tion ‘‘four species A”, Table 1) does not meet the main criteria of palustris that could be separated by biometric means, too. Even the phylogenetic species concept. First, it does not represent the interspecific comparison showed that marsh tits and willow tits smallest exclusive monophyletic group of common ancestry (de are very similar in biometric parameters of body size and beak Queiroz and Donoghue, 1988). Furthermore, P. montanus and P. shape (Shao et al., 2016). Misidentification of species and mislabel- songarus are not even reciprocally monophyletic and due to the ing of samples and specimens can thus be another potential source large biometric and phenotypical variation within and among of error in phylogenetic studies that is being more and more criti- these Asian taxa the ‘songar tit’ would hardly even meet the crite- cally addressed for phylogenetic analyses (Philippe et al., 2011; rion of diagnosability (Cracraft, 1983; Sangster, 2014). Kozlov et al., 2016) or for species distribution modelling (Costa In disagreement with Johansson et al. (2013), all our specimens et al., 2015). classified as P. hypermelaenus belonged to the sister clade of all northern Palearctic and Chinese P. palustris, and had significantly 4.2. Authoritative assignment of taxon names to genetic lineages based smaller body size dimensions compared to eastern and western on type specimens marsh tit populations and to P. weigoldicus, respectively. Despite having treated P. hypermelaenus as a distinct species of its own, In the case of Chinese Poecile the same genetic lineage has been Eck and Martens (2006) already stated that ‘‘a close relationship assigned to two different scientific taxon names by two different of hypermelaenus with the Marsh Tit is evident and has never been author teams. This conflict could be resolved only by examination questioned”. Accordingly, several authors had included these and genotyping of type specimens. In all our phylogenetic recon- southwestern Chinese populations and those from Mount Victoria structions, the genetic Poecile lineage from south-western China in P. palustris before (P. p. hypermelaenus: Vaurie, 1957; Snow, closely related to P. montanus and P. hyrcanus also included one 1967; Eck, 1988; Cheng, 1987; Sibley and Monroe, 1990; Cramp of the name-bearing syntypes of P. weigoldicus (MTD C23803) and Perrins, 1993; Gosler and Clement, 2007), and those authors and a further historical specimen from the same series (MTD who separated P. hypermelaenus and P. palustris at the species level
Table 2 Body size dimensions of Willow Tits, Marsh Tits and their allies; means of eight morphometrical parameters (all in mm, except relative tail in% = t1/wmax).
Clade/specimen/reference Bill (bsk) Bill (bp) Bill (bwp) Wing Kipp Tail Tarsus Relative tail montanus mean ± sd 10.2 ± 0.5 4.7 ± 0.3 4.9 ± 0.3 63.3 ± 2.6 10.5 ± 1.2 57.9 ± 4.0 15.9 ± 0.6 91.4 ± 4.3 n8484848383828283 songarus mean ± sd 11.5 ± 1.2 4.7 ± 0.2 5.0 ± 0.2 69.8 ± 1.9 11.1 ± 0.7 60.5 ± 3.5 17.1 ± 0.6 86.7 ± 3.9 n7 77777 77 affinis/stoetzneri mean ± sd 10.1 ± 0.7 4.7 ± 0.3 4.9 ± 0.3 64.4 ± 2.1 10.4 ± 0.9 60.6 ± 2.7 16.5 ± 0.5 94.4 ± 3.3 n4545454241434542 weigoldicus mean ± sd 9.8 ± 0.4 4.8 ± 0.2 4.9 ± 0.3 66.9 ± 2.6 10.4 ± 1.2 58.9 ± 2.64 16.5 ± 0.6 87.9 ± 3.1 n1414141513141515 weigoldicus syntype C23803 10.2 5.0 5.0 69.5 12.0 64.4 17.2 92.7 weigoldicus Harrap and Quinn, 1996 [10.5–12.6] – – [66–69] – [54–57] – – weigoldicus Eck, 2006 (n = 32) – – – 67.4 ± 1.6 – 58.2 ± 1.4 – 0.86 ± 1.7 palustris West mean ± sd 9.1 ± 0.6 4.7 ± 0.3 4.9 ± 0.3 63.7 ± 2.1 10.6 ± 0.8 54.2 ± 2.9 15.7 ± 0.5 85.0 ± 3.0 n5454545353535353 palustris East mean ± sd 9.0 ± 0.5 4.7 ± 0.3 4.8 ± 0.3 64.0 ± 2.2 10.9 ± 0.8 60.6 ± 3.7 15.9 ± 0.6 95.0 ± 4.1 n5454545050535350 hypermeleanus mean ± sd 8.7 ± 0.5 4.3 ± 0.3 4.7 ± 0.2 62.0 ± 3.9 9.0 ± 0.8 50.7 ± 2.6 15.3 ± 0.5 81.8 ± 1.9 n7 77777 77 hypermelaenus Harrap and Quinn, 1996 [9–11] – – [55–71] – [42.5–60] [14–16.5] – Eck, 2006 (n = 5) – – – – – – – 0.80 ± 1.6 548 C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 emphasized the close relationships of these two species 2006; Fedorov et al., 2009; Periparus: Martens et al., 2006) and of (Stresemann and Heinrich, 1939; Harrap and Quinn, 1996). Justifi- the Nearctic (Poecile: Reudink et al., 2007; Kingston et al., 2012; cation for species rank of P. hypermelaenus was inferred from its Davidson et al., 2013). However, mitochondrial introgression is morphological and genetic distinctiveness against all P. palustris not necessarily limited to narrow hybrid belts but might also (Eck and Martens, 2006; who gave a cyt-b distance value of 3.3% extend across wide zones of intergradation, e.g. in western Europe between the two species) and this reasoning was later followed between northern and southern genetic lineages of the coal tit, by Dickinson and Christidis (2014) and Clements et al. (2015). Periparus ater (Pentzold et al., 2013). However, according to our data set, genetic distinctiveness of P. hypermelaenus is not unambiguous because of shared mitochon- 5. Conclusions drial haplotypes between this species and its adjacent counterparts in the North, P. palustris hellmayri and P. p. jeholicus. These findings 5.1. Taxonomy can only be elucidated with respect to spatial distribution patterns of phenotypes and haplotypes. The black-bibbed tit was originally described as Poecile hyper- melaena Berezowski and Bianchi, 1891; synonyms (according to 4.4. Distribution ranges and biogeography Snow, 1967 and Dickinson et al., 2006): Parus Dejeani Oustalet 1897 (recognized as a valid subspecies by Cheng, 1987); Lopho- Eurasian breeding ranges of marsh tits and willow tits largely phanes poecilopsis Sharpe 1902 (synonymized by Rothschild overlap though in most regions the two species are ecologically (1926) in his avifauna of Yunnan). After having been transferred separated and only rarely occur in local syntopy (Cramp and to genus Parus the epitheton’s gender was correctly changed to Perrins, 1993; Harrap and Quinn, 1996; Gosler and Clement, Parus hypermelaenus by all authorities. In the course of a recent dis- 2007) but local co-occurrence has been observed in parts of cussion on the correct gender of Poecile, Harrop (2011) made a south-eastern Siberia (J.M. pers. obs.). In Asia, the largely overlap- strong case for treating Poecile as feminine and consequently ping Chinese breeding ranges of marsh tits, willow tits and allies changing species names to Poecile montana and Poecile hyperme- are poorly known and in some cases insufficiently documented. laena. This was opposed by several authors who treated the word Remarkably, all our genotyped records of the P. hypermelaenus lin- as masculine (P. montanus, P. hypermelaenus) and this debate eage fall in the range of P. p. hypermelaenus according to the map in seems to be currently unresolved (David and Gosselin, 2008, Cheng (1987) but are located north of the species’ breeding range 2012; Sangster et al., 2015). For the time being and for the sake according to BirdLife International (2016, because they did not of taxonomic stability, it might be wise maintaining the most fre- include the scattered populations from Shaanxi, Gansu and Hubei; quently used masculine gender. Today the deviant spelling P. compare Harrap and Quinn, 1996; fig. 47.1). In fact, our P. hyperme- hypermelas is predominantly being used in the Chinese literature laenus lineage comprises all samples and sequences from Shaanxi (Cheng, 1987; Dai et al., 2010), but it might date back to Vaurie and Shanxi (including specimens identified as P. m. hellmayri by (1957, 1959) who had transferred the female Poecile hypermelaena Dai et al. 2010) and Song et al. (2016) identified further four pop- from the original description to the masculine Parus as Parus hyper- ulations from Shanxi, Shaanxi and Hubei that belonged to the melas (see note by Brooke, 1975 in his list of references). mitochondrial P. hypermelaenus lineage. Considering the spatial distribution of mtDNA haplotypes the northern range limits of P. hypermelanus should at least extend across Shaanxi to adjacent 5.2. Species limits and superspecific classification southern Shanxi. The type locality of P. hypermelaenus ‘Lan-shya- kou’ at the border of southern Shaanxi to Gansu (Berezowski and Delimitation of biological species-level taxa (according to the Bianchi, 1891; cf. Eck and Martens, 2006) falls into that range. Biological Species Concept, BSC; Mayr, 1942) remains problematic Harrap and Quinn (1996) listed these populations from Shanxi, in our target groups, because the taxonomic operational units Shaanxi and Sichuan under P. p. hellmayri but stressed that they identified by genetic analyses have a largely allopatric Asian distri- differed slightly from the populations of NE China and Korea in a bution and only limited spatial information is available where ‘‘more heavily washed buff on the underparts and in having a breeding ranges of northern and southern Chinese vicariant forms slightly more extensive black bib”. Possibly, these populations might overlap. In case of P. palustris and P. hypermelaenus, there is from Central China represent a zone of phenotypical intergradation limited evidence of mitochondrial introgression (our study) and between northern Chinese P. palustris and the black-bibbed tit, P. nuclear gene flow in central China (Song et al., 2016), however hypermelaenus (though Snow (1957) stressed that no intermedi- assessment of reproductive barriers in the sense of the BSC should ates of the latter Chinese form and P. p. hellmayri had ever been col- rely on more information from field surveys including DNA barcod- lected). Putative admixture of the two latter forms in Central China ing data and phylogeographic analyses. is corroborated by recent evidence of asymmetric nuclear gene According to the Phylogenetic Species Concept, at least P. wei- flow from northern into southern Chinese populations of marsh tits goldicus is reciprocally monophyletic against its closest relatives (Song et al., 2016; with southern populations corresponding to P. (Donoghue, 1985; de Queiroz and Donoghue, 1988), however, hypermelaenus, respectively). In our study, we found local co- diagnosability (according to Cracraft, 1983, 1987; Sangster, 2014) occurrence of P. hypermelaenus haplotypes and that of the eastern by morphometric means is problematic and must rely rather on P. palustris lineage at one of our sampling sites, i.e. the type locality plumage color patterns (Fig. S1; Kleinschmidt, 1921; Harrap and of P. palustris jeholicus ‘‘30 km N of Balihandien, Jehol”. Because the Quinn, 1996). At least P. hypermelaenus is diagnosable against other phenotypes of both type specimens clearly matched that of eastern Palearctic Poecile species by body size, tail length and wing shape. P. palustris we tend to assume an accidental record of mitochon- The close relationship of the willow tit P. montanus and the Caspian drial introgression of the southern P. hypermelaenus lineage into tit P. hyrcanus were the main argument for splitting the latter spe- northern P. palustris populations. Mitochondrial introgression has cies from P. lugubris at the species level (Johansson et al., 2013) and been documented for other sister species pairs in narrow contact this split received further support from biometric analyses (Loskot, zones (including the presence of phenotypical hybrids) at the east- 2014). Both P. hyrcanus and P. weigoldicus are genetically strongly ern Tibetan Plateau margin (long-tailed tits, Aegithalos: Dai et al., divergent from P. montanus and separation of these three at the 2011; Wang et al., 2014) and among closely related tit taxa of species level seems justified. Therefore, we see our results in good the Palearctic (Parus: Päckert et al., 2005; Kvist and Rytkönen, accordance with the taxonomy and systematics by Clements et al. C. Tritsch et al. / Molecular Phylogenetics and Evolution 107 (2017) 538–550 549
(2015) and we would refrain from recommending any changes Brooke, R.K., 1975. Application for a declaration that species-group names formed until further population genetic data from Chinese species are from or ending in unlatinized Greek adjectives are indeclinable. Bull. Zool. Nom. 32, 188–191. available. The superspecies concept by Amadon (1966) might Broughton, R.K., 2009. Separation of Willow Tit and Marsh Tit in Britain: a review. accommodate the close relationships among species of the willow Brit. Birds 102, 604–616. tit clade (three allospecies) and among those of the marsh tit clade Clement, M., Posada, D., Crandall, K.A., 2000. TCS: a computer program to estimate gene genealogies. Mol. Ecol. 9, 1657–1660. (two allospecies). We suggest the following classification. Clements, J.F., Schulenberg, T.S., Iliff, M.J., Roberson, D., Fredericks, T.A., Sullivan, B. L., Wood, C.L., 2013. The eBird/Clements Checklist of Birds of the World: v2013. 1. Willow tits and allies, Poecile [montanus] (Conrad, 1827) Downloaded from
Gosler, A.G., Clement, P., 2007. Family Paridae. In: del Hoyo, J., Elliott, A., Christie, D. Rambaut, A., 2012. FigTree v1. 4. University of Edinburgh, Edinburgh, UK. Available A. (Eds.), Handbook of the Birds of the World, Picathartes to Tits and Chickadees, at: