Molecular phylogeny of the Drosophila obscura species group, with emphasis on the Old World species
© Gao et al. 2007
Received: 03 December 2006
Accepted: 07 June 2007
Published: 07 June 2007
Skip to main content
© Gao et al. 2007
Received: 03 December 2006
Accepted: 07 June 2007
Published: 07 June 2007
Species of the Drosophila obscura species group (e.g., D. pseudoobscura, D. subobscura) have served as favorable models in evolutionary studies since the 1930's. Despite numbers of studies conducted with varied types of data, the basal phylogeny in this group is still controversial, presumably owing to not only the hypothetical 'rapid radiation' history of this group, but also limited taxon sampling from the Old World (esp. the Oriental and Afrotropical regions). Here we reconstruct the phylogeny of this group by using sequence data from 6 loci of 21 species (including 16 Old World ones) covering all the 6 subgroups of this group, estimate the divergence times among lineages, and statistically test the 'rapid radiation' hypothesis.
Phylogenetic analyses indicate that each of the subobscura, sinobscura, affinis, and pseudoobscura subgroups is monophyletic. The subobscura and microlabis subgroups form the basal clade in the obscura group. Partial species of the obscura subgroup (the D. ambigua/D. obscura/D. tristis triad plus the D. subsilvestris/D. dianensis pair) forms a monophyletic group which appears to be most closely related to the sinobscura subgroup. The remaining basal relationships in the obscura group are not resolved by the present study. Divergence times on a ML tree based on mtDNA data are estimated with a calibration of 30–35 Mya for the divergence between the obscura and melanogaster groups. The result suggests that at least half of the current major lineages of the obscura group originated by the mid-Miocene time (~15 Mya), a time of the last developing and fragmentation of the temperate forest in North Hemisphere.
The obscura group began to diversify rapidly before invading into the New World. The subobscura and microlabis subgroups form the basal clade in this group. The obscura subgroup is paraphyletic. Partial members of this subgroup (D. ambigua, D. obscura, D. tristis, D. subsilvestris, and D. dianensis) form a monophyletic group which appears to be most closely related to the sinobscura subgroup.
Species of the Drosophila obscura group (41 species assigned to six subgroups) are mostly inhabitants of temperate forest throughout the Holarctic region, with some can adapted into high-elevation temperate-like habitats in the Afrotropical, Neotropical and Oriental regions. Some of these species (e.g., D. pseudoobscura and its close relatives) have served as favorable models for evolutionary biology since the influential works of Dobzhansky and his colleagues in the 1930's [1, 2]. The whole-genome sequence of D. pseudoobscura was determined following D. melanogaster. Comparisons between the two species have shed new light on Drosophila genome evolution . In addition, in the past two decades, increasing number of evolutionary studies have been conducted in a historical background of the obscura species group on varied subjects, e.g., evolution of genome size , evolution of karyotype and P elements , origin and evolution of Drosophila Y chromosome  and genetics of morphological evolution .
Since the 1950's, a number of studies have been conducted to reconstruct phylogeny of the obscura group via a variety of approaches [2, 8]. Recent molecular phylogenetic studies [9–13] clearly support the monophyly of the obscura species group and recover several well-supported lineages, for example, the affinis, pseudoobscura, and subobscura subgroups, the D. ambigua triad (D. obscura, D. ambigua and D. tristis), give essential support to the monophyletic origin of the New World species, i.e., those of the affinis and pseudoobscura subgroups. In spite of this, the relationship among the Old World obscura, subobscura, microlabis, and sinobscura subgroups, and their relationship to the New World clade are still poorly resolved. This phylogenetic predicament was partially ascribed to the "rapid radiation" history of the obscura group [10, 12]. An alternative hypothesis to explain the lack of resolution at the base of this phylogeny is a bias in taxon sampling. For example, none of the previous phylogenetic studies has dealt with the obscura group as a whole: different studies employed different set of taxa, with species from the Afrotropical region (5 species) and Oriental region (8 described + 2 undescribed species) have rarely been investigated [14, 15], probably due to the difficulty in collecting and/or culturing these poorly known taxa.
Gene loci sampled in the present study. Numbers show aligned lengths and numbers of parsimony informative sites (PI, given in parentheses) for nucleotide or translated amino acid sequences of each locus.
Translated amino acid sequences
NADH dehydrogenase subunit 2 (ND2)
Cytochrome oxidase subunit I (COI)
Cytochrome oxidase subunit II (COII)
Cytochrome b (Cyt b)
Alcohol dehydrogenase (Adh)
28S ribosomal RNA (28S)
Species sampled in the present study and collection data of samples used for DNA sequencing.
Unknown site, Switzerland
Unknown site, USA*
Unknown site, USA*
Kunming, Yunnan, China
Yakutsuk, East Siberia, Russia
Kunming, Yunnan, China
Koganei, Tokyo, Japan
Chitou, Taiwan, China
Lugu Lake Nature Reserve, Yunnan, China
Shennongjia Nature Reserve, Hubei, China
Kunming, Yunnan, China
Mt. Elgon, Kenya
All sequences from GenBank
All sequences from GenBank
Accession numbers for sequences. Sequences with underlined accession numbers are used for the statistical test of temporal pattern only.
D. hubeiensis (HB)
D. hubeiensis (KM)
Results of pairwise partition homogeneity test (PHT). Numbers above and below diagonal show P -values resulted of the un-weighted and the six-parameter weighting methods, respectively.
Incongruence between data partitions indicates that the two partitions compared have had different histories or that one of them violate the assumptions of the phylogenetic method . The PHT is currently implemented with only parsimony, which assuming small number of actual sequence changes per site. Higher P values obtained by six-parameter weighting may indicates that, the six-parameter parsimony method fits the NT data better by accounting for the effect of multiple hits (as suggested in the saturation plot in Figure 1), thus reduces the incongruence in several pairwise comparisons.
The MP tree deduced with the un-weighted method (Figures 2A; henceforth referred to as uwMP tree) clusters the D. ambigua triad with the sinobscura subgroup, while puts the D. dianensis/D. subsilvestris pair outside this cluster. However, the MP tree deduced with the six-parameter weighting method (Figure 2B; henceforth referred to as 6pMP tree), the ML tree (Figure 2C) and the Bayesian tree (Figure 2D) congruously suggest a cluster of the D. ambigua triad and the D. dianensis/D. subsilvestris pair. This cluster (henceforth referred to as obscura cluster) forms a larger cluster with the sinobscura subgroup (henceforth referred to as obscura-sinobscura cluster). However, the supports for these relationships are also low.
D. microlabis, as the single representative of the microlabis subgroup, is placed at the basal position in the uwMP and 6pMP trees. However, this species is clustered with the subobscura subgroup in the ML tree (Figure 2C, BP = 93) and Bayesian tree (Figure 2D; PP = 1.00). The uwMP, 6pMP and Bayesian trees suggest weakly (BP = 6–39; PP = 0.81) a close relationship between the obscura-sinobscura cluster and the New World clade. However, the ML tree clusters all the Old World species into a large group with weak support (BP = 28).
The Bayesian trees of AA data recover the same relationships within each of the major lineages as those of the NT data, except for that they suggest a branching order of (ambigua, (obscura, tristis)) in the D. ambigua triad, and that the tree inferred with the Poisson model clusters the Kunming (KM) strain of D. hubeiensis with D. sinobscura, instead of its conspecific Hubei (HB) strain. However, all these relationships are very weakly supported.
Our estimate for the origin of the microlabis - subobscura clade (~19.5/16.7 Mya) falls close to previous estimation based on mutation distance of 11 genes (17.7 ± 4.4 Mya; D. pseudoobscura vs. D. subobscura) ; the estimate for the splitting between the affinis - pseudoobscura clade and the obscura - sinobscura cluster (~17.9/15.4 Mya) is clearly older than the previous one based on Adh sequences (13.1 ± 1.74 Mya; obscura subgroup vs. pseudoobscura subgroup) , and the estimate for the D. pseudoobscura - D. miranda divergence (4.93/3.76 Mya) differs greatly from the estimate based on mutation distances (2.00 ± 0.6 Mya) . This is mainly due to that 1) our estimates is not directly based on pairwise distances, but on a given tree with branch lengths; 2) we use different calibration point from those studies [31, 32].
Phylogeny of the obscura species group is investigated with dense taxon-sampling from the Old World, especially the Oriental region, using both nucleotide and translated amino acid sequences of multiple loci. The results corroborate some previously well-recognized relationships, and shed some new light on the evolutionary history of the obscura group, especially the relationship among major lineages.
The MP trees of NT data suggest with low confidence that D. microlabis alone, as a long-branch taxon, represents the first branch in the obscura group. However, it was strongly suggested in the ML tree (Figure 2C) and Bayesian trees (Figures 2D, 3B, and 3C) that D. microlabis forms a monophyletic group with the subobscura subgroup. A suspicion arises whether the basal position of D. microlabis is true, or an artifact due to long-branch attraction (LBA) by those outgroups? As demonstrated by Anderson and Swofford , if this relationship is true, MP method is prone to positively recover it, thus seems to perform as good as, or even better than ML method. Otherwise, ML will outperform MP by recovering the true relationship. As shown by some studies with empirical data and/or computer simulation [35–37], model-based methods (ML and Bayesian methods) can be relatively robust against branch-length differences, even against model violation. Therefore, it is very likely that the basal relationship of D. microlabis alone in the MP trees is an artifact due to LBA, while the ML and Bayesian trees suggest the true relationship between the microlabis and subobscura subgroups.
The Bayesian trees with AA data (Figures 3B, 3C) support with high confidences a basal position of the microlabis - subobscura clade in the obscura group. This relationship is also suggested by previous cladistic analyses . In some previous studies lacking any representative of the microlabis subgroup [10, 12, 38], the subobscura subgroup alone was placed as basal to the rest in the obscura group. Moreover, comparison of more than 48 morphological characters among obscura group species (except for the microlabis and sinobscura subgroups) suggests that D. subobscura (as the only representative of the subobscura subgroup) differs more from the other Eurasian species than the latter differ from each other .
The monophyly of the obscura cluster is also suggested by morphological data: in the obscura group, D. dianensis, D. subsilvestris and D. obscura are the only species characterized by pale spots on several abdominal tergites in female [30, 40], and large, somewhat quadrate 10th sternite in male [30, 41]. In the present study, the obscura cluster appears to be most closely related to the sinobscura subgroup, while the remainder species of the obscura subgroup, i.e., the D. limingi/D. tsukubaensis pair and D. bifasciata/D. imaii pair, appear to have diverged earlier. Based on these evidences, we propose here a revised notion of the obscura subgroup, i.e., the cluster of the five species D. ambigua, D. obscura, D. tristis, D. dianensis and D. subsilvestris.
Consistent with some previous study , the present study clusters the Palearctic D. helvetica with the Nearctic D. affinis with strong support, clearly indicating its adscription to the affinis subgroup. Morphologically, D. helvetica possesses some diagnostic characters pertained to the affinis subgroup, e.g., very small distal sex-comb, 6 rows of acrostichal setulae. Some morphological similarities between D. helvetica and D. tolteca, a member of the affinis subgroup, are also found . Given the Old World Origin of the obscura group  and the monophyletic nature of the affinis - pseudoobscura clade, D. helvetica undoubtedly represents a refluence of the New World element back into the Old World.
It was demonstrated that Bayesian posterior probability can overestimate the true probability of node confidences if substitution model used for phylogenetic analysis is oversimplified , and/or if concatenated sequences data are used . In the 6pMP and ML trees of NT data, the BP supports for the obscura cluster are relatively low, and so are the BP supports for the obscura - sinobscura cluster. However, the corresponding PP values in the Bayesian tree of NT data seem to be excessively high. Also in the analyses with AA data, a remarkable discrepancy between BP and PP is found for the large clade consisting of the affinis, pseudoobscura, sinobscura, and obscura subgroups. In our Bayesian analysis with NT data, partition-specific models are used. The Bayesian analyses of AA data with simple (Poisson) and comprehensive (GTR) models yield comparable PP supports for the above relationships. Therefore, the great discrepancy between BP and PP may be partially due to our using of concatenated sequences. On the other hand, it was shown that BP in ML analyses is generally a conservative estimate of statistical confidence , and that compared to the BP in ML analyses, BP in MP analyses shows lower correlation with Bayesian PP .
The effect of taxon sampling on phylogenetic accuracy has been addressed by a number of studies [46–49], most of which favor addition of taxa, especially for breaking up long branches to improve information about state of internal nodes and rate at individual sites . In the present study, adding of a number of Old World taxa enables us to trace some additional, ancient branching events, resulted in some largely congruent basal relationship in the obscura group, e.g., the close relationship between the D. ambigua triad and the D. dianensis/D. subsilvestris pair, that between the obscura cluster and the sinobscura subgroup, and the sister relationship between the microlabis and subobscura subgroups. The obscura group is presently known for 41 described and at least 2 undescribed species. Future studies with denser taxon sampling and larger number of characters (especially for nuclear gene sequence characters) are desirable to fully resolve the basal relationship in this group.
Throckmorton's  study with data of palegeography and fossil record has proposed that the founder of the obscura group arose and existed for short time before its expanding with the temperate forest. The temperate forest was proposed to began to spread in Northern Hemisphere with decreasing of temperature by about 10~15°C in Oligocene . However, according to our time estimation (Figure 4), the Old World diversification of the obscura group began well after the origin of this group. By the mid-Miocene time, at least half of the current major lineages had come into being, indicating a more or less rapid major radiation of the obscura group. This is also suggested by the results of statistical test of temporal pattern (Figure 5), and reflected by the short internal branches in the phylogenetic trees (e.g., Figures 2C, 2D, 3B and 3C). On the other hand, obvious nucleotide substitution saturation (Figure 1) and base composition bias have been observed in the mitochondrial loci. All these results lend supports to the previous proposal  that either the rapid radiation, or the special evolutionary dynamics of the mtDNA in the obscura group may account for the phylogenetic predicament concerning the obscura group.
Two major patterns during the evolution of the family Drosophilidae have been proposed by Throckmorton  based on morphological and biogeographical data: 1) the primary tropical disjunction involving species groups, subgenera and genera; and 2) the temperate-forest disjunction involving species subgroups and species, represented typically by the obscura species group. Due to lacking of data about drosophilid faunas from either the Afrotropical or the Oriental region, the disjunction pattern of the obscura group within the Old World was thought to be not clear . However, the pattern is now much more clearly seen: there are about 14 species (3 of the subobscura subgroup, 4 of the microlabis subgroup, 1 of the affinis subgroup, 5 of the obscura subgroup, and 1 ungrouped) are restricted to or mainly distributed in Europe/the Afrotropical region; and at least 11 species (3 of the sinobscura subgroup, 6 of the obscura subgroup, and 2 undescribed) restricted to East/Southeast Asia. Among the Eurasian species of the obscura group, at least 9 are restricted or mainly distributed in the Oriental region, with the southmost records from the Mt. Kinabalu of Malaysia. This clearly indicates a thorough adaptation of the group into high-elevation temperate-like habitats in the Oriental region, a pattern parallel to those in the Afrotropical and Neotropical regions [8, 51, 52].
Our time estimation for the major radiation of the obscura group is overlapped largely to the hypothetical time span of the developing of temperate forest in Northern Hemisphere (mid-Oligocene to mid-Miocene) [50, 53]. It was proposed based on biogeographical data that, from the mid-Tertiary times, the temperate drosophilid faunas developed and spread with the temperate forest, until the time of the temperate-forest disjunction in mid Miocene age . It is very likely that the fragmentation of the temperate forest had enforced the Old World diversification, and that the gradual desertification of the Asia interior onset from the early Miocene epoch  played important role in enforcing the disjunction of the temperate forest and thus the east-west disjunction of the obscura group within the Old World.
The cooling of the climate in the Qinghai-Tibet area of South Asia resulted from the uplift of the Qinghai-Tibet Plateau in the Tertiary period was thought to provide favorable conditions for the Palearctic insect fauna to invade southwards . It is reasonable to presume that changing of climate might have also facilitated the adaptation of the founders of the Oriental elements of the obscura group into South Asia. Probably the intensified uplift of the plateau in late Pliocene  has accelerated these elements (e.g., the sinobscura subgroup, initiated to diversify ~2.6–3.0 Mya) to spread around, giving rise to the current species in south China, India and Malaysia.
In conclusion, our phylogenetic study suggests that, the obscura group began to diversify rapidly in the Old World before invaded into the New World. Among the Old World lineages, the microlabis and subobscura subgroups form a monophyletic group basal to the rest of the obscura group. Our results corroborate the finding by the previous studies that the traditional obscura group is paraphyletic, with some of its members (the D. ambigua triad plus the D. dianensis/D. subsilvestris pair) forming a monophyletic cluster, which appears to be most closely related to the sinobscura subgroup.
Samples of twenty-one species of the D. obscura group and one species of the D. melanogaster group (Table 2) were used for DNA sequencing. DNA was extracted from single fly by standard phenol-chloroform method. The PCR cycle program comprised an initial 2 min of predenaturation at 94°C, 35 cycles of amplification (50 s of denaturation at 94°C; 1 min of annealing at 55°C for ND2 and COII, 51.5°C for Cyt b, 52°C for COI and Adh, 60°C for 28S; 1 min of extension at 72°C), and 5 min of sequence postextension at 72°C. The primers (all given left to right from 5' to 3' ends) for the PCR and sequencing of the ND2, COI, COII, Cyt b and 28S genes were: nd2-1 ATATT TACAG CTTTG AAGG, and nd2-2 AAGCT ACTGG GTTCA TACC for the ND2 gene ; UEA5 AGTTC TAGCA GGAGC TATTA CTAT, and UEA8 AAAAA TGTTG AGGGA AAAAT GTTA for the COI gene ; coii-1 ATGGC AGATT AGTGC AATGG and coii-2 GTTTA AGAGA CCAGT ACTTG  for the COII gene; Cyt b -F TTATG GTTGA TTATT ACGAA, and Cyt b -R CAAAA CATAT GCTTA TTCAA for the Cyt b gene; 28S-H CCCGA AGTAT CCTGA ATCTT TCGCA TTG (designed by T. Katoh in Hokkaido University), and 28S-T TCTTA GTAGC GGCGA GCG  for the 28S gene. PCR products were separated on 2.0% agarose gels, then excised from the gels and purified using Watson™ gel extraction mini kit (Watson Biotechnologies).
The Adh fragments of D. hubeiensis (KM), D. luguensis, D. dianensis, and D. limingi were amplified using the primers adh-e2+ CTGGAC TTCTG GGACA AGCG, and adh-e3- TAGAT GCCCG AGTCC CAGTG , and the PCR product was cloned into the PMD18-T Vector (TaKaRa), then transformed into Escherichia coli as host. Thereafter, the recombinant DNA was extracted then, and the Adh fragment was sequenced with the M13 universal primers AAGCT TGCAT GCCTG CAGGT CGACG and CGGTA CCCGG GGATC CTCTA GAGAT. After purifying of the product of sequence reaction, the sequences were determined using ABI 377 or ABI 3700 sequencer according to the protocol by manufacturer.
The newly collected sequences were edited using the Editseq module of the DNAStar package . For each of the ND2, COI, COII, Cyt b, Adh and 28S gene fragment, homologous GenBank sequences were downloaded and aligned with newly determined sequences by the ClustalW method . The intron region of the Adh gene was excluded from all analyses. The alignment was then adjusted by eye to make it conform to the codon assignments. Then the ends of a few COI and 28S sequences were trimmed slightly, so as to make the majority of homologous sequences well overlapped. We use MEGA3  to calculate base composition and ti/tv ratios in each data partitions. Detection of substitution saturation in mitochondrial and nuclear data partitions was performed by plotting the ratio ti/tv for sequence pairs versus corresponding number of whole substitutions with respect to codon positions (1st+2nd or 3rd), with the pairwise ratios and numbers of substitutions calculated in MEGA3 , and plots worked out with the Microsoft Excel program. Only ingroup species are included for saturation analysis.
Before the phylogenetic analyses, the NT data was subjected to pairwise PHT  between data partitions of different loci under either un-weighted or six-parameter weighting parsimony scheme [63, 64] with PAUP* 4.0b10 , with heuristic search for 1000 replicates. Modeltest 3.6  was used to estimate parameters of DNA substitution model for the six-parameter MP, ML and BI analyses.
Phylogenetic analyses with NT data set were performed using MP, ML and Bayesian inferring methods. The MP tree was constructed with either un-weighted or six-parameter weighting parsimony methods with heuristic search (initial trees obtained by 100 replicates random addition; branch swapping with TBR algorithm). For the six-parameter method, models specific to each locus were implemented in PAUP*4.0b10 , with each substitution classes was weighted based on its substitution rate (Rij, i.e., rate of transformation between nucleotide i and j) estimated with Modeltest3.6 : wij = -ln (Rij/∑Ri). The weighting parameter stepmatrix for each locus was adjusted for satisfaction of triangle inequality in PAUP*4.0b10 . To access the support level for each node on the MP trees, bootstrap (BP)  analyses were performed with 1000 replicates and heuristic search. The MP analysis with AA data was performed with similar strategy as that of the NT data set.
The ML analysis of NT data was performed using PAUP*4.0b10 , with parameters assigned as follows: base frequencies of A (respectively C, G and T) = 0.3089 (respectively 0.1385, 0.1319 and 0.4207); substitution rates of A-C (respectively A-G, A-T, C-G, C-T and G-T) = 2.0353 (respectively 13.2831, 6.7964, 5.9419, 33.5370 and 1.0000); proportion of invariable sites (I) = 0.5765; and gamma distribution shape parameter (α) = 0.8598.
Bayesian inferring was implemented in MrBayes3.1 . The starting tree was randomly selected and four chains were run. For the analysis with NT data set, parameters are set as follows: "nst = 6" + "invgamma" applied to the character partition of mitochondrial genes, "nst = 6" + "gamma" to that of the Adh gene, and "nst = 2" + "gamma" to that of the 28S gene. Bayesian analyses of AA data were performed with either the Poisson or the GTR models, with gamma-distributed rate variation across sites and a proportion of invariable sites. For all the Bayesian analyses, two independent runs were implemented in parallel, with the Markov chains been sampled every 100 cycles. The runs were stopped after 2,000,000 cycles of MCMC (Markov Chain Monte Carlo) for NT data, but 1,000,000 cycles for AA data, till the average deviation of split frequencies fall well below 0.01. For all the runs, 1,000 trees sampled at early phase of the chain (well before the end of this phase, the likelihood values stop to increase, and start to fluctuate within a stable range) were discarded, and the remainders were summarized to obtain a majority rule tree which showing all the compatible partitions.
Before estimating divergence times, relative-rate test using the program Phyltest2.0  is conducted to examine constancy of sequence evolution between lineages in the obscura group. Since no time calibration point of fossil record or by geological dating is available for our estimation, we cite that used by Beckenbach et al. : an interval of 30–35 Mya for the divergence between the obscura and melanogaster groups. The program r8s1.71 , which enables estimating divergence time in the absence of a molecular clock, is used to estimate the divergence times in the obscura species group. A ML tree constructed with mtDNA sequence data was used as the input tree file for time estimating. The model for the ML search is selected by Modeltest3.6 : base frequencies are 0.3246, 0.1028, 0.1116 and 0.4610 for A, C, G and T, respectively; rates = 3.4175, 25.8349, 10.1514, 4.7640, 86.1667 and 1.0000 for A-C, A-G, A-T, C-G, C-T and G-T, respectively; I = 0.5629; α = 1.0481. A penalized likelihood (PL)  method is used for divergence time reconstructing, with a truncated Newton (TN) algorithm for finding optima of the objective functions. Cross-validation are checked over a range of smoothing values by set the parameters cvstart = 0, cvinc = 0.5 and cvnum = 10. Divergence time for all the nodes except for the root is estimated by rerunning of the input data with a selected smoothing parameter (= 316, which has the lowest cross-validation score) by checking of the cross-validation.
We perform a statistic test  of temporal pattern in the obscura group. To reduce the effect of incomplete sampling of extant taxa, GenBank sequences of some additional obscura group species (Table 3) are also used. Based on the aligned sequence of 27 obscura group species, a F84 distance data matrix is created, and a so-called KITSCH tree is constructed, using the program DNADIST and KITSCH, respectively, both packed in Phylip3.6 . For distance estimating using DNADIST, shape parameters of the Gamma distribution (α = 1.0353) and base frequencies (A = 0.3132; C = 0.1401; G = 0.1320; T = 0.4147) are estimated in Modeltest3.6 , and the ratio of ti/tv (= 1.1) is calculated in MEGA3 . The branching times for the resulted KITSCH tree were normalized between zero (the time of the first branching event) and one (the present), cumulative frequency distribution (CFD)  of the scaled branching times for all the nodes (n = 25) in the tree is plotted. The dissimilarity between the resulted empirical CFD and expected (i.e., average) CFD specific for same number of extant taxa was quantified using a Kolmogorov-Smirnov (K-S) goodness-of-fit statistic D . The average CFD for specific number of extant species has been generated by Wallenberg et al.  by computer simulations under null model of stochastic lineage bifurcation and extinction. Therefore, we get the average CFD for 27 extant taxa by interpolated those for 25 and 30 taxa .
We thank Drs. W. Pinsker, E. Haring, L. Serra, and H. Takamori for providing us with samples. We are grateful to Drs. M.J. Toda, G. Baechli, E. Haring, and two anonymous reviewers for their invaluable suggestions and comments. The present study was supported by grants of the State Key Basic Research and Development Plan, NSFC (30460026, 30621092), Bureau of Science and Technology of Yunnan Province, and JSPS (No. 12375002, 16370040, 19570077).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.