Background

Gibbons, family Hylobatidae, are small arboreal apes, which inhabit tropical and subtropical rainforests of Southeast Asia and adjacent regions (Figure 1). Together with humans and great apes, they belong to the primate superfamily Hominoidea [1–4]. Among hominoids, gibbons were the first to branch off and they display a set of morphological and behavioural characteristics distinctly different from great apes and humans [1, 5, 6]. Most prominent in this respect is the predominantly monogamous life style, their territorial calls, and the typical brachiating locomotion [1, 4–7]. Due to their extensive karyotypic diversity [8–11], gibbons provide an excellent model organism to study chromosomal rearrangements and, hence, to better understand human diseases caused by such alterations.

Figure 1
figure 1

Geographical distribution of gibbons based on [2, 5, 23, 41]. Dotted and solid lines indicate country borders and major rivers, respectively. Historical distribution of N. hainanus and N. nasutus is hatched.

Full size image

Although in several aspects unique among primates and with up to 16 species the most diverse group of apes, gibbons are still in the shadow of great apes in respect of scientific studies, conservation efforts and public awareness. However, many gibbon species are on the brink of extinction and most of them are classified as "Endangered" or even "Critically Endangered" [12]. With approximately 20 individuals left in its native habitat, the Hainan gibbon (Nomascus hainanus) is the rarest primate in the world [6, 13, 3–5, 7, 17–35], neither a congruent phylogeny nor a consistent taxonomic classification was obtained. Moreover, incomplete taxon sampling as well as misidentified specimens resulted in only fragmentary or even false conclusions. Accordingly, the classification of gibbon taxa at various taxonomic levels as well as their phylogenetic relationships remain disputed and a consensus is far from being available.

For example, in early studies, small apes were divided into two genera, with one (Symphalangus) including the siamang, and the other (Hylobates) all the remaining species [17, 36]. Later on, the family was split into four major clades, which were recognized as subgenera [2, 5, 21] and eventually as genera [4, 16, 29, 37, 38]. This division is now widely accepted and takes into account the fact that species within each of the four major clades share a number of characteristics, most importantly a distinctive diploid chromosome number: Hoolock (2n = 38), Hylobates (2n = 44), Symphalangus (2n = 50) and Nomascus (2n = 52) [8]. Similarly, the number of species and subspecies is a matter of debate as well. While Symphalangus is consistently regarded as monotypic, the two Hoolock subspecies were recently elevated to species [38]. In Nomascus originally only one species was recognized [17, 18, 20, 39], but in current classifications four to six species were suggested [2, 4, 12, 16, 34]. In contrast, the genus Hylobates already comprised at least four species in early classifications [17, 39], but recent studies proposed six or seven species [2, 4, 16]. Due to this incongruence we follow the most recent classification of the IUCN Red List [12] with a total of 16 gibbon species (Table 1).

Table 1 Common names, IUCN classification and proposed classification of gibbons.
Full size table

In the present study, we analyse the complete mitochondrial cytochrome b (cytb) gene from 85 individuals, which represent all gibbon genera and species, and most subspecies. Based on our data, we are able to 1) provide the most complete phylogeny of gibbons on all taxonomic levels, 2) estimate divergence times between lineages, 3) establish a reliable classification, 4) elucidate gibbon phylogeography, and 5) provide a tool for the species identification of gibbon individuals.

Results

From all 85 gibbons, we successfully generated sequences of the complete mitochondrial cytb gene (1,140 bp). A contamination of our dataset with nuclear pseudogenes (numts) can be regarded as minimal, because no multiple amplifications of different copies were detected by direct sequencing. All sequences were correctly transcribed, and identical sequences were obtained for the same individual in cases where different material types were available. Moreover, no inconsistent positions were detected in alignments, which were assembled from overlap** sequences. Cross-contamination between individuals can be excluded as well, since all negative controls revealed no amplifications and randomly repeated PCRs for the same individual produced identical sequences.

Among the 85 individual gibbons studied, no identical haplotypes were detected. The cytb alignment comprising solely gibbons was characterized by 429 variable sites, of which 374 were parsimony-informative. In the complete alignment, which additionally contained great ape, human and hamadryas baboon representatives, we observed 565 variable sites, of which 462 were parsimony-informative.

Phylogenetic tree reconstructions based on maximum-parsimony (MP), neighbor-joining (NJ), maximum-likelihood (ML) and Bayesian algorithms revealed various strongly supported clades, which corresponded to genera, species and subspecies (Figure 2). All algorithms led to identical tree topologies, although several branching patterns gained only weak support. According to our reconstructions, hominoids diverged into a clade consisting of gibbons, and another with great apes and human. Among the latter, Pongo split off first, followed by Gorilla, before finally Pan and Homo diverged. Within gibbons, a basal position of Nomascus and a sister grou** of Hylobates and Hoolock was indicated, but support for this branching pattern was relatively low (Table 2). Similarly, with the exception of a strongly supported H. agilis + H. albibarbis clade, also the relationships among the species of Hylobates were not well resolved. However, at least species monophylies were clearly confirmed, though a common origin of H. agilis was only weakly supported. The relationships among the subspecies of H. muelleri and H. lar were less resolved. In Hoolock, the two species H. hoolock and H. leuconedys clearly segregated into two distinct clades. Within Nomascus, relationships among species were completely resolved, suggesting a N. hainanus + N. nasutus clade as sister lineage to the remaining species. Among them, N. concolor branched off first, followed by the divergence of N. gabriellae and N. leucogenys/N. siki. The monophyly of N. leucogenys was significantly supported, but evidence for a common origin of N. siki individuals was not obtained. Within N. concolor, specimens identified as N. concolor lu formed a distinct clade, while the remaining subspecies clustered together without further subdivision. However, support for a reciprocal monophyly of both clades was relatively low.

Table 2 Support values and Bayesian divergence date estimates (in mya)*.
Full size table
Figure 2
figure 2

Ultrametric tree showing phylogenetic relationships and estimated divergence ages among studied gibbon individuals based on complete mitochondrial cytb sequence data. For individual codes see Additional File 1. Circles indicate bootstrap or posterior probability values (filled circles: >90%, >0.95, open circles: <70%, <0.80). Nodes of interest are arbitrarily numbered (N1-N45). C2 and C3 refer to two of the three nodes used for calibration (C1 not shown). Light green bars indicate the four radiations. A geological time scale is given below. Full details of age estimates and node supports are presented in Table 2.

Full size image

Based on divergence age estimates, gibbons separated from great apes and humans 16.26 million years ago (mya) (for 95% credibility intervals see Table 2). Within hominids, Pongo branched off first (13.83 mya), followed by Gorilla (8.90 mya), before finally Homo and Pan diverged from each other (6.56 mya). The differentiation of Pongo and Pan into species occurred 4.12 and 2.74 mya, respectively. In an initial radiation, gibbons diverged within a relative short time period of only 1.65 million years (6.69-8.34 mya) into four genera. Within Hylobates, most species diverged from each other between 3.02 and 3.90 mya. The only exception was the separation of H. albibarbis from H. agilis 1.56 mya, which was in the time frame of subspecies splits within H. muelleri (1.42-1.78 mya). Differentiation of H. lar into subspecies occurred even later (0.52-1.05 mya). The two Hoolock species diverged 1.42 mya from each other. In Nomascus, differentiation into species took place over a longer time period, lasting from 4.24 until 0.55 mya. The most recent species divergence within Nomascus occurred between N. siki and N. leucogenys (0.55 mya), which was in a similar range as the separation of N. concolor lu from the other N. concolor subspecies (0.43 mya).

Discussion

By analysing all species and most subspecies, the present study provides the most complete view into the evolutionary history of the gibbon family. However, as in earlier molecular studies on gibbons [26–35], relationships on various taxonomic levels are less resolved and partially contradict earlier findings. While the herein depicted branching pattern among genera is identical with that found in earlier studies using also cytb [32] or D-loop [29] sequences, it differs from another cytb-based study [28] in placing Nomascus and not Symphalangus as most basal genus. Studies based on mitochondrial ND3-ND4 sequences [31] or chromosomal rearrangements [8] suggest Hoolock as most ancestral lineage, and Nomascus together with either Hylobates [31] or Symphalangus [8] as the most recently diverged genera. For Hylobates, our data indicate a basal position of H. klossii, and a further division into a clade consisting of H. lar, H. muelleri, H. agilis and H. albibarbis, and another one with H. moloch and H. pileatus. Various branching patterns among Hylobates species are proposed [27, 31, 32, 35], which all differ from our one, but respective support values are similarly low as in our study. In contrast, the relationships found among species of the genus Nomascus are well resolved and identical with that suggested by [30, 31, 33, 34].

According to our and earlier data, relationships among gibbon genera and Hylobates species remain disputed, which most likely can be explained by the separation of respective lineages within relative short time periods. This becomes even more obvious when considering estimated divergence ages, which fall into four temporal windows. In the first, between ~6.7 and ~8.3 mya, the four gibbon genera originated. In a second radiation, between ~3.0 and ~3.9 mya, Hylobates split into various species, and in a third burst, between ~1.3 and ~1.8 mya, H. muelleri, the H. agilis + H. albibarbis clade and Hoolock further differentiated. Finally, in a fourth radiation, between ~0.5 and ~1.1 mya, H. lar diverged into subspecies. In contrast, speciation in Nomascus was a continuous process, lasting from 4.24 until 0.55 mya.

Taxonomic implications

Our data show that mitochondrial DNA (mtDNA) provides a powerful tool for the identification and taxonomic classification of gibbons, because taxa form strongly supported monophyletic clades, or at least appear to form distinct lineages in those cases where only one individual per taxon was tested. Moreover, most differentiation events fall into four temporal periods, which allow a hierarchical ranking as proposed by Goodman et al. [40], though the threshold for the recognition of a certain taxonomic unit whether genus, species, or subspecies remains disputed. Hence, to provide a more reliable classification, we compare divergence ages among gibbon lineages with those among other Asian primates and hominids.

Accordingly and concordant with recent classifications [4, 12, 16, 29, 34, 37, 38, 41], the four major gibbon lineages are proposed as distinct genera (Table 1), since they split from each other in a similar time range as did colobine genera [[42, 43], Roos C, Zinner D, Schwarz C, Nash SD, ** PCR products were amplified with primers constructed on the basis of sequences from conspecifics (respective primers are available from the authors upon request). For all amplifications, wax-mediated hot-start PCRs were performed for 40 cycles, each with a denaturation step at 92°C for 1 min, annealing at 60°C for 1 min, and extension at 72°C for 0.5-1.5 min, followed by a final extension step at 72°C for 5 min. The results of the PCR amplifications were checked on 1% agarose gels. Subsequently, PCR products were cleaned with the Qiagen Gel Extraction kit and sequenced on an ABI 3130xl sequencer using the BigDye Cycle Sequencing kit. Sequences were assembled with Geneious v4.6.1 [78] and checked for their potential to be correctly transcribed. Gibbon haplotypes were deposited at GenBank and are available under the accession numbers GU321245-GU321329 (see also Additional file 1).

To prevent cross-species contaminations, laboratory procedures followed described standards [46]. To exclude contaminations of the dataset with numts, we mainly used material in which nuclear DNA is highly degraded (faeces, museum tissue) [79, 80]. Moreover, the applied primers are known to amplify solely the mitochondrial copy of the gene in hylobatids [34], and for cross-validation purposes, for some specimens, sequences were generated using different material types (blood, faeces).

Statistical Methods

For phylogenetic reconstructions, we expanded our dataset with orthologous sequences from various hominids (Homo, Pan, Gorilla, Pongo) and Papio hamadryas, which was used as outgroup. Phylogenetic trees were constructed with MP and NJ algorithms as implemented in PAUP v4.0b10 [81] as well as with ML and Bayesian algorithms, using the programs GARLI v0.951 [82] and MrBayes v3.1.2 [83, 84]. For MP analysis, all characters were treated as unordered and equally weighted throughout. A heuristic search was performed with the maximum number of trees set to 100. For NJ and ML reconstructions, the optimal nucleotide substitution model (GTR + Γ) was chosen using Akaike information criterion (AIC) as implemented in MODELTEST v3.7 [85]. Relative support of internal nodes was performed by bootstrap analyses with 10,000 (MP, NJ) or 500 replications (ML). In GARLI, only the model specification settings were adjusted according to the dataset, while all other settings were left at their default value. ML majority-rule consensus trees were calculated in PAUP. For Bayesian reconstructions, the dataset was partitioned into codon positions, each with its own substitution model. We used four Markov Chain Monte Carlo (MCMC) chains with the default temperature of 0.1. Four repetitions were run for 10,000,000 generations with tree and parameter sampling occurring every 100 generations. The first 25% of samples were discarded as burnin, leaving 75,001 trees per run. Posterior probabilities for each split and a phylogram with mean branch lengths were calculated from the posterior density of trees.

To estimate divergence times, a Bayesian MCMC method, which employs a relaxed molecular clock approach [86], as implemented in BEAST v1.4.8 [87], was used. A relaxed lognormal model of lineage variation and a Yule prior for branching rates was assumed. The alignment was partitioned into codon positions, and the substitution model, rate heterogeneity and base frequencies were unlinked across codon positions. Optimal nucleotide substitution models were chosen using AIC in MODELTEST.

For calibrations we used the fossil-based divergence between Homo and Pan, which was dated at 6 - 7 mya [88–90], the separation of Pongo from the Homo/Pan lineage ~14 mya [91], and the divergence of hominoids and cercopithecoids ~23 mya [92, 93]. Instead of hardbounded calibration points, we used the published dates as a normal distribution prior for the respective node. For the Homo - Pan divergence, this translates into a normal distribution with a mean of 6.5 mya and a standard deviation (SD) of 0.5 mya, for the separation of Pongo from the Homo/Pan clade into a mean of 14.0 mya and a SD of 1.0 mya, and for the hominoid - cercopithecoid divergence into a mean of 23 mya and a SD of 2 mya.

Since the estimation of phylogenetic relationships was not the main aim of this analysis, for the calculation an a-priori fixed tree topology as obtained from NJ reconstructions using the GTR + Γ model (Figure 2) was implemented. Four replicates were run for 10,000,000 generations with tree and parameter sampling occurring every 100 generations. The adequacy of a 10% burnin and convergence of all parameters were assessed by visual inspection of the trace of the parameters across generations using TRACER v1.4.1 [94]. Subsequently, the sampling distributions were combined (25% burnin) using the software LogCombiner v1.4.8, and a consensus chronogram with node height distribution was generated and visualized with TreeAnnotator v1.4.8 and FigTree v1.2.2 [95].