Molecular and morphological evidences reveal a cryptic species in the Vinaceous Rosefinch Carpodacus vinaceus (Fringillidae; Aves)

Genetic analysis and phylogenetic reconstruction

We collected blood, liver or muscle tissues from 25 Carpodacus vinaceus vinaceus individuals (sixteen males, six females and three sex unidentified individuals) from Gansu Province (N = 8), Shaanxi Province (N = 8), Sichuan Province (N = 6) and Yunnan Province (N = 3) of China and 22 C. v. formosanus individuals (fifteen males, six females and one sex unidentified individuals) from Taiwan (Fig. 1). Owing to the lack of a phylogeny of genus Carpodacus, the sister taxon of C. vinaceus is unknown. Therefore, one sample each of C. nipalensis and C. edwardsii was used as outgroups in this study. Samples were immersed in 100% ethanol in the field and were transferred to a −80 °C freezer for long-term storage. Gross genomic DNA was extracted following a chloroform and LiCl precipitation protocol (modified from Gemmell & Akiyama 1996) and was then resuspended in double-distilled H₂O for later use.

Mitochondrial genes and Z-linked fragments were amplified via polymerase chain reactions (PCRs). Primer sequences for PCR amplifications are provided in Table 1. The PCRs were set up in 12-μL reaction volumes containing about 50 ng DNA, 10 mm Tris–HCl (pH 9.0), 50 mm KCl, 0.5 mm dNTP, 0.2 μm of each primer, 0.4 U of Taq DNA polymerase (GE Healthcare, Waukesha, WI, USA) and 1.5 mm MgCl₂. The PCR profile for both genes was one cycle of denaturation at 94 °C for 2 min, followed by 40 cycles at 94 °C for 30 s, 53 °C for 30 s and 72 °C for 1.5 min, and then a final extension at 72 °C for 2 min using iCycler Thermal Cycler (Bio-Rad, Hercules, CA, USA). Both strands of amplicons were sequenced using the Bigdye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Carlsbad, CA, USA) with the same PCR primers and electrophoresed on an ABI 3130xl Genetic Analyzer (Applied Biosystems, Carlsbad, CA, USA). Sequences were proofread and assembled with the aid of the program Sequencher 4.7 (Gene Codes Corporation, Ann Arbor, MI, USA). For two Z-linked fragments, all sequences were unphased and sites with ambiguous peaks were coded according to IUPAC standards.

We calculated the number of polymorphic sites, the number of haplotypes, haplotype diversity and nucleotide diversity (π; the average number of nucleotide differences per site between two sequences, Li 1997) using DnaSP 5.10 (Librado & Rozas 2009). The McDonald–Kreitman test (McDonald & Kreitman 1991) was used to examine the selective neutrality of two mitochondrial fragments by comparing the polymorphism within each C. vinaceus subspecies to that in C. punicwu. Two additional neutrality tests, Fu and Li’s D* (Fu & Li 1993) and Fu’s F_S (Fu 1997) tests were used to detect departures from the mutation–drift equilibrium that would indicate changes in historical demography and natural selection. All three tests were conducted using DnaSP. We used the uncorrected p-distance to calculate the average genetic distance within each C. vinaceus subspecies and the net distance (Nei & Li 1979) between the two subspecies with the software MEGA 4 (Tamura et al. 2007). For the unphased sequences of Z-linked loci, we only calculated the number of polymorphic sites and the number of genotypes.

Phylogenetic relationships among the concatenated mitochondrial DNA and zDNA sequences were reconstructed via both maximum likelihood (ML) and Bayesian analyses. For the zDNA genes, indels were excluded from the phylogenetic analysis. The ML analysis was performed using the program PHYML-aLRT (Guindon & Gascuel 2003; Anisimova & Gascuel 2006). Based on the Akaike information criterion (AIC), the model TPM2uf+G, where the gamma distribution shape parameter was 0.0110, was used for mitochondrial DNA (−lnL = 5873.5693, K = 96, AIC = 11939.1386) as suggested by the software jModelTest 0.1.1 (Posada 2008); the estimated base frequencies were A = 0.2792, C = 0.3226, G = 0.1523 and T = 0.2459; rates of nucleotide substitution relative to that between G-T were A-C = 35.6114, A-G = 272.0326, A-T = 35.6114, C-G = 1.0000 and C-T = 272.0326. For the Z-linked data set, the TrN+I substitution model was selected by the jModelTest; the estimated base frequencies were A = 0.2614, C = 0.2066, G = 0.2225 and T = 0.3095; rates of nucleotide substitution relative to that between G-T were A-C = 1.0000, A-G = 4.3530, A-T = 1.0000, C-G = 1.0000 and C-T = 2.1701. The topology of the tree was optimized, rather than the branch length. We used the approximate likelihood-ratio test (aLRT, Anisimova & Gascuel 2006) with the Shimodaira–Hasegawa-like procedure option to estimate the reliability of each node. For both mitochondrial and Z-linked data set, the Bayesian analysis was conducted using MrBayes 3.1.2 (Ronquist & Huelsenbeck 2003) under the GTR+I+G model, which permits a thorough exploitation of parameter sample space (Huelsenbeck & Rannala 2004). Two independent Markov chains, each of 10 000 000 steps, were run while sampling at every 1000 steps and discarding the first 2 500 000 steps as burn-in. All trees were rooted by the outgroup and visualized in the program FigTree v.1.3.1 (http://tree.bio.ed.ac.uk/software/figtree).

Morphometric analysis

We carried out a morphometric analysis of skin specimens of C. v. vinaceus and C. v. formosanus from China and Taiwan. The analysis of C. v. vinaceus utilized 12 males and seven females archived in the Institute of Zoology, Chinese Academy of Sciences, China, and nine males and 15 females from the Kunming Institute of Zoology, Chinese Academy of Sciences, China. The analysis of C. v. formosanus utilized eight males and nine females archived in the National Museum of Natural Science, Taiwan, and three males and five females from the Taiwan Endemic Species Research Institute, Taiwan. Voucher numbers of the skin specimens are listed in Table 2. We used a digital caliper (Digimatic Caliper, model number CD-8′CS; Mitutoyo Corp., Kanagawa, Japan) with a precision of ±0.02 mm to measure seven morphometric characteristics: beak depth (distance between commissure and point of beak), beak width (distance between the commissures on both sides) and lengths of culmen, gonys, tarsal, wing and tail. All measurements were taken from the right side of birds if possible, in units of mm (minimum: 0.1 mm). All measurements were taken by Shou-Hsien Li.

Two-way analysis of variance (anova) was conducted to detect the effects of subspecies and sex on the variance of different morphometric characteristics. Student’s t-test was applied to compare the means of morphometric traits between subspecies by sex. The α value for the Student’s t-test was adjusted by the Bonferroni correction. We also employed canonical discrimination analysis to classify both sexes of C. vinaceus individually with the morphometric data. We used the F-ratio test to select variables for model construction. The critical P value for a variable to be retained in a model was 0.05 and 0.10 for forward and backward selection procedures, respectively. All statistical analyses were performed using JMP 7.0 (SAS Institute Inc., Cary, NC, USA).

Plumage colour measurements

Eight study skins of each subspecies of C. vinaceus were used to measure male plumage coloration (C. v. vinaceus from the Kunming Institute of Zoology, Chinese Academy of Sciences, China and C. v. formosanus from the National Museum of Natural Science, Taiwan). Voucher numbers for the skin specimens examined are listed in Table 3. We used a USB2000 spectrometer (Ocean Optics, Dunedin, FL, USA) with illumination from an HL2000 halogen light source to measure reflectance from 300 to 700 nm of five parts of each individual’s body, namely the central positions of the breast and the rump, the cheek, the eyebrow and the wing spot on the third tertiary feather at the right side of the bird (T3 in Fig. 2). We pointed a R600-7-UV/125F probe (Ocean Optics, Dunedin, FL, USA) with a cover straight at the surface to measure the reflectance. All measurements were taken by Hsin-Yi Hung. A white standard (Labsphere, North Sutton, NH, USA) was used to calibrate the reflectance. Hue (wavelength at the mean between the maximum and minimum reflectance values from 550 to 700 nm), total brightness (average reflectance from 300 to 700 nm), red chroma (sum of reflectance from 550 to 700 nm divided by the sum of the reflectance from 300 to 700 nm) and UV chroma (sum of reflectance from 300 to 400 nm divided by the sum of the reflectance from 300 to 700 nm) were then calculated for each of the five body parts. All measurements were repeated three times. For each measurement, repeatability (Lessells & Boag 1987) was between 0.80 and 0.98. The significance of the colour difference between the two subspecies was examined by Student’s t-test with the α value adjusted by the Bonferroni correction.

Genetic polymorphism

In this study, we obtained 1121 bp of CYTB gene (N = 48) and 1505 bp of COXI gene (N = 47) from Carpodacus individuals (GenBank accession numbers: HQ735307-HQ735401), in which 149 CYTB and 235 COXI polymorphic sites were revealed. A majority of the substitutions occur at the third codon position (CYTB: 86%, COXI: 91%). Both gene fragments could be correctly transcribed, indicating that no pseudogene exists in our sequences. It is worth noting that 33 and 55 synapomorphic sites were found between C. v. vinaceus and C. v. formosanus from CYTB and COXI, respectively. Number of haplotypes, haplotype diversity and nucleotide diversity are higher in C. v. vinaceus than in C. v. formosanus (Table 4). The McDonald–Kreitman test showed no significant deviation from neutrality for either CYTB or COXI genes (P > 0.05 with Fisher’s exact test, Table 4). In contrast, both values of Fu and Li’s D* and Fu’s F_S were significantly negative in C. v. vinaceus. This may indicate a historical population expansion, rather than selection, when taking the results of McDonald–Kreitman test into account. The mean genetic distances for both CYTB and COXI genes are small within each subspecies: 0.002 for C. v. vinaceus and 0.001 for C. v. formosanus. However, the net genetic distances between the two subspecies of C. vinaceus are more substantial: 0.035 and 0.039 for CYTB and COXI genes, respectively.

For the two Z-linked loci, 614 or 617 bp of the PPWD1 gene (N = 44) and 606 bp of the SVEP1 gene (N = 42) were obtained (GenBank accession numbers: JF701189-JF701278). As with the mtDNA genes, the two zDNA genes show more genetic diversity in C. v. vinaceus than in C. v. formosanus (Table 4). For example, C. v. vinaceus has more polymorphic sites (eight and 12 for the PPWD1 and SVEP1 genes, respectively) than C. v. formosanus (one and two, respectively) and more genotypes (11 and 12 for the PPWD1 and SVEP1 genes, respectively, compared with three and four). We also found one and three synapomorphic sites between two C. vinaceus subspecies for the PPWD1 and SVEP1 genes, respectively.

Molecular phylogenetic analysis

Phylogenetic analyses were conducted based on a concatenated alignment (2626 bp) from 23 C. v. vinaceus, 21 C. v. formosanus and one each of C. nipalensis and C. edwardsii as both of their CYTB and COXI sequences were available. Phylogenetic trees reconstructed by ML and Bayesian analyses revealed identical topologies (Fig. 3A). Each of the two subspecies formed an exclusively monophyletic clade with high statistical support (aLRT > 0.95 and Bayesian posterior probability = 1.00 for both clades).

The molecular phylogeny inferred from the concatenated zDNA sequences (20 C. v. vinaceus and 21 C. v. formosanus; Fig. 3B) also indicate that C. v. formosanus is a monophyletic group with high statistical support (aLRT = 0.99 and Bayesian posterior probability = 1.00). However, C. v. vinaceus is a paraphyletic group: all but one of the C. v. vinaceus sequences formed a sister clade (aLRT = 0.83 and Bayesian posterior probability = 0.99) to the C. v. formosanus clade with high statistical support (aLRT = 0.75 and Bayesian posterior probability = 0.83). The exceptional one C. v. vinaceus sequence formed a weakly supported clade (both aLRT and the Bayesian posterior probability <0.5) with the two outgroup species.

Morphometric analysis

Mean and standard deviations of seven morphometric traits of the two C. vinaceus subspecies for both sexes are listed in Table 2. Results of two-way anova indicate that both subspecies and sex have significant effects (P < 0.05) on the variation of culmen (F_3,64 = 3.7146, P = 0.0158), wing (F_3,63 = 23.7627, P < 0.0001) and tail (F_3,55 =34.1711, P < 0.0001) lengths. Only subspecies, but not sex, has a significant effect on the variation of beak depth (F_3,64 = 10.3230, P < 0.0001) and tarsal length (F_{3, 63}<0.0001, P < 0.0001). Neither subspecies nor sex has a significant effect (P > 0.05) on the variation of beak width (F_3,64 = 1.6893, P = 0.178) or gonys length (F_3,64 = 0.8387, P = 0.4777). Results of the Student’s t-test indicate that male C. v. vinaceus tend to be larger than females and that both sexes of C. v. formosanus are larger than C. v. vinaceus (Table 2), as indicated by significantly longer beak depth, wing, tarsal and tail lengths.

Results of the canonical discriminant analysis suggest that the morphometry of the two C. vinaceus subspecies is highly distinct. For male rosefinches (N = 18 and 10 for C. v. vinaceus and C. v. formosanus, respectively), the stepwise selection procedure selected beak width, tarsal and tail lengths to construct the canonical discriminant function (P < 0.05). Because the first canonical variable explained 100% of the variance, we omitted the rest of the canonical variables constructed by the same model. The scoring coefficient of each morphometric variable of the canonical variable 1 for males is shown in the Table 5. Our results indicate that the subspecies of each male measured can be correctly predicted by our canonical discriminant functions (Table 4) with high posterior probabilities (P > 0.94). For females (N = 18 and 13 for C. v. vinaceus and C. v. formosanus, respectively), the stepwise selection procedure selected beak width, tarsal and tail lengths to construct the canonical discriminant function (P < 0.05). As for males, the first canonical discriminant variable explained 100% of the variance, so the other canonical variables inferred from the same model were not shown. The scoring coefficient of each morphometric variable in the canonical variable 1 for females is shown in Table 5. The canonical discriminant function can be used to predict all female C. v. vinaceus correctly (P > 0.780); however, it misassigns two C. v. formosanus females as C. v. vinaceus (P > 0.808).

Plumage colour measurements

Plumage reflectance patterns for males of the two C. vinaceus subspecies are shown in Fig. 4. Our measurements indicate that the plumage of male C. v. formosanus tends to be brighter and redder than that of male C. v. vinaceus (Table 3). Specifically, we found that C. v. formosanus males tend to have a brighter breast (t = 2.898, P = 0.012) and a higher red chroma of wing spot on the tertiary feather (t = 3.417, P = 0.004).

While measuring coloration, we also observed distinguishable patterns of the white patch on the tertiary feathers (Fig. 2). For all eight specimens of C. v. vinaceous examined (Fig. 2A), the patch was only present on the tip of external vane of the inner two tertiary feathers (denoted as T2 and T3 from the external to the most internal tertiary feathers), while in all eight specimens of C. v. formosanus examined (Fig. 2B), the patch was found on all three tertiary feathers (denoted as T1, T2 and T3 from the external to internal tertiary feathers) and the patches on T2 and T3 of C. v. formosanus extended to the internal vane.

Taxonomic recommendation

Because of their highly disjunctive distributions, the biological species concept, which emphasizes that a species is a naturally interbreeding unit (Mayr 1942, 1963), cannot be used to determine whether C. v. formosanus falls within C. vinaceus’ species boundary (Helbig et al. 2002). We can, however, adopt the phylogenetic species concept (e.g. Cracraft 1983; Donoghue 1985), by which a species is defined based on a diagnosable criterion and exclusive monophyly. Using this approach, the distinguishable sites of mtDNA and zDNA, the exclusive monophyly of both subspecies on the mtDNA tree (Fig. 3A), the significant differences in morphometric traits (Table 2) and the distinguishable plumage pattern and coloration (Fig. 2 and Table 3) all suggest that the two recognized C. vinaceus subspecies should be treated as two full species. Molecular dating further suggests that the two C. vinaceus subspecies may have evolved independently for over one million years, despite the frequent presence of land bridges connecting Taiwan and the Asian continent during the many ice ages in the Pleistocene epoch (e.g. Voris 2000). Their evolutionary integrity is therefore likely to persist in the future given no anthropogenic intervention. Consequently, it is also justifiable to treat C. vinaceus as two full species, when applying the evolutionary species concept that emphasizes the independent evolutionary history and fate of a species (Wiley 1978). Based on all lines of evidence, we conclude that the current taxonomic treatment includes a cryptic species under the name of C. vinaceus. We recommend revising the taxonomic status of C. vinaceus as follows: The taxonomic status of C. v. formosanus should be restored to a full species, namely C. formosanus, as first described by Ogilvie-Grant (1906, 1911); the name C. vinaceus can be reserved for its continental sibling species. Such a treatment would reflect the avian diversity of East Asia more appropriately. Given the existence of global warming, the fact that it has a restricted range in the alpine zone of Taiwan’s Central Mountain Range (Hachisuka & Udagawa 1951) means that C. formosanus deserves conservational attention as has been accepted for the Golden Bowerbird (Prionodura newtonia) in Australia (Hilbert et al. 2004). The birds’ range and density should be monitored from now on to evaluate whether their long-term persistence would be threatened by future climate change.

Biogeographic implications

In the past 3 million years, global climate has been characterized by periodic oscillations. During glacial periods, the climate was much cooler and drier than in the interglacial periods, driving species ranges to shift towards the warmer equator and lower altitudes (reviewed by Davis & Shaw 2001). The lower sea level in the glacial periods might also have allowed species to traverse the then dried-out Taiwan Strait, an effective geographic barrier between Taiwan and the Asian continent during the interglacial periods. These glacial periods might therefore have provided the most favourable conditions for connections between the montane biota of the eastern Tibet-Himalayas and Taiwan.

Assuming that the most recent common ancestor of C. formosanus colonized Taiwan before the split of C. vinaceus and C. formosanus, molecular dating of their split may provide estimate of when biota in montane Taiwan and the eastern Tibet-Himalayas region were connected. Based on mtDNA data, our molecular dating suggests that the two rosefinches split about 1.7 MYA. The range shift after the cool glacial period might then have isolated the Taiwan population from its mainland counterpart and allowed it to evolve independently. However, a recent study of Taiwan Hwamei (Leucodioptron taewanus), a Taiwan endemic, revealed that the speciation process of a continental island endemic can be very complicated (Li et al. 2010). Although molecular dating based on the divergence of mtDNA suggests that Taiwan Hwamei probably split from its sister species, Chinese Hwamei L. canorum, around 1.5 MYA (Li et al. 2006), analysis based on 18 nuclear loci suggests that the divergence of the two Hwamei species may have started 3.5 MYA. Moreover, a significant level of gene flow persisted between the two species until 0.5 MYA, i.e. one million years after the date of the split estimated by mtDNA divergence (Li et al. 2010). The incongruence between estimates based on mtDNA and nuclear loci implies that molecular dating based on mtDNA divergence alone may overlook the complexity of the speciation process. To depict the divergence history between C. formosanus and C. vinaceus properly, more genetic data and a test using more sophisticated speciation models are required in future investigation.