Comparing mitogenomic timetrees for two African savannah primate genera ( Chlorocebus and Papio )

Complete mitochondrial (mtDNA) genomes have proved to be useful in reconstructing primate phylogenies with higher resolution and confidence compared to reconstructions based on partial mtDNA sequences. Here, we analyse complete mtDNA genomes of African green monkeys (genus Chlorocebus ), a widely distributed primate genus in Africa representing an interesting phylogeographical model for the evolution of savannah species. Previous studies on partial mtDNA sequences revealed nine major clades, suggesting several cases of para- and polyphyly among Chlorocebus species. However, in these studies, phylogenetic relationships among several clades were not resolved, and divergence times were not estimated. We analysed complete mtDNA genomes for ten Chlorocebus samples representing major mtDNA clades to find stronger statistical support in the phylogenetic reconstruction than in the previous studies and to estimate divergence times. Our results confirmed para- and polyphyletic relationships of most Chlorocebus species, while the support for the phylogenetic relationships between the mtDNA clades increased compared to the previous studies. Our results indicate an initial west–east division in the northern part of the Chlorocebus range with subsequent divergence into north-eastern and southern clades. This phylogeographic scenario contrasts with that for another widespread African savannah primate genus, the baboons ( Papio ), for which a dispersal from southern Africa into East and West Africa was suggested.


INTRODUCTION
The availability and analyses of genetic data have a tremendous impact on the understanding of phylogenetic relationships and evolutionary history of organisms, at which different genetic markers (mitochondrial, autosomal or gonosomal) or respective whole genomes can provide insights into different aspects of the evolutionary history of taxa (Moore, 1995;Hoelzer, 1997;Maddison, 1997;Nichols, 2001;Funk & Omland, 2003;Avise, 2004). It has been shown that phylogenies based on mitochondrial and different nuclear markers are often incongruent, due to sex-biased dispersal, nuclear swamping or mitochondrial capture after introgression (Moore, 1995;Hoelzer, 1997;Maddison, 1997;Avise, 2004;Roos et al., 2011;Zinner, Arnold & Roos, 2011a;Liedigk et al., 2012;Wang et al., 2012;Haus, Roos & Zinner, 2013a). Moreover, in taxa with female-biased dispersal, as in the majority of mammal species including primates (Greenwood, 1980), the mitogenomic population structure often reveals *Corresponding author. E-mail: s.dolotovskaya@gmail.com; croos@dpz.eu Figure 1. Distribution of the six Chlorocebus species and the nine major mtDNA haplogroups (1-9; Haus et al., 2013b). Coloured species distributions are modified from Lernould (1988) and Kingdon (1997). Dashed lines indicate the distribution of the nine major mtDNA haplogroups, and circles indicate the geographic provenance of samples used in this study (see Table S1). Green monkey drawings by Stephen Nash.
C. tantalus and C. pygerythrus) were included in their study, and the geographical provenance of only two samples was known. Another mitogenomic study did not include C. djamdjamensis (Guschanski et al., 2013). Recently, the first Chlorocebus phylogeny based on nuclear genomic data was presented by Warren et al. (2015), but C. djamdjamensis was also lacking in the study.
The most geographically broad sampling including all taxa was conducted by Haus et al. (2013b), who analysed mitochondrial cytochrome b (cytb) gene sequences, including samples covering almost the entire range of the genus. This study revealed nine major mtDNA clades that reflect geographic distributions rather than morphological taxa (Haus et al., 2013b). However, monophyly of several clades was not well supported, and phylogenetic relationships among them could not be resolved, suggesting that cytb sequence information alone might not be sufficient for resolving the apparently rapid radiation of African green monkeys' mtDNA lineages.
However, applying complete mtDNA genomes has improved confidence in reconstructed primate phylogenies compared to shorter mtDNA fragments Liedigk et al., 2012;Finstermeier et al., 2013;Guschanski et al., 2013;Zinner et al., 2013;Pozzi et al., 2014). In our study, we therefore provide new complete mtDNA genomes of Chlorocebus and base our phylogeographic analysis on these data.
The observed discordances between mtDNA haplogroups and morphological phenotypes of Chlorocebus are not unique; rather, they resemble those reported from another widespread African savannah primate genus, Papio baboons (Zinner et al., 2009(Zinner et al., , 2011b(Zinner et al., , 2013Keller et al., 2010;Fig. 2). Both genera are primarily found in savannah woodland and have largely overlapping geographical distributions (Figs 1, 2). The similarity in habitat preference and distribution of Chlorocebus and Papio makes it likely that the evolutionary histories of both genera were influenced by similar climatic and geological events, while potential differences in divergence times and population histories of the genera might indicate taxon-specific responses to environmental changes.
In this study, we analyse complete mtDNA genomes of ten Chlorocebus samples representing the nine mtDNA clades revealed by Haus et al. (2013b) and compare the resulting phylogenetic relationships with Figure 2. Distribution of the six Papio species and the seven major mtDNA haplogroups (A-G; Zinner et al., 2009Zinner et al., , 2013. Species distributions are modified from Zinner et al. (2013). Dashed lines indicate the distribution of the seven major mtDNA haplogroups, and circles indicate the geographic provenance of samples used in this study (see Table S1). Baboon drawings by Stephen Nash. a morphology-derived phylogeny (Groves, 2001(Groves, , 2005 and divergence times to those based on partial mtDNA sequence data (Haus et al., 2013b), Y-chromosomal DNA (Haus et al., 2013a) and nuclear genome data (Warren et al., 2015). We expect that the analysis of complete mtDNA genomes leads to a better resolution of the phylogenetic relationships among lineages than in the previous studies based on partial mtDNA sequence data. We also include complete mtDNA genomes of Papio (Zinner et al., 2013) in our phylogenetic analysis and compare branching patterns and divergence times of Chlorocebus taxa with those of Papio taxa in order to evaluate the potential role of climatic events in their radiations and to investigate general phylogeographic trends for African savannah mammals.

Sampling
We used faecal samples of wild African green monkeys originating from ten sites in Ethiopia, Tanzania, Zambia, Ghana, Nigeria and the Republic of South Africa (RSA). These samples represent the nine mtDNA clades revealed by Haus et al. (2013b) (Fig. 1, Supporting Information Table S1). The samples were collected between 2005 and 2010 by Haus et al. (2013b). Faecal samples were preserved in > 90% ethanol for ≥ 24 h, dried and transferred into tubes with silica beads for further storage (Nsubuga et al., 2004). Sample collection was conducted in compliance with the animal care regulations and the principles of the American Society of Primatologists for the ethical treatment of nonhuman primates. All samples were collected from wild, non-habituated groups without threatening or harming the subjects. The research complied with protocols approved by the German Primate Center and adhered to the legal requirements of the countries in which samples were collected.

DNA extraction, amplification anD Sequencing
We extracted total genomic DNA from faecal samples with the First-DNA all-tissue kit (Gen-Ial, Troisdorf, Germany), following the standard protocol as provided by the company with minor changes. The volume of lysis buffers 1, 2 and 3 was increased to 1000, 100 and 600 µL, respectively, the volume of proteinase K was increased from 10 to 20 µL and an additional chloroform separation step was used. Extracted DNA was dissolved in 50 µL molecular-grade water. Until further processing, DNA extracts were stored at −20 °C.
Since DNA extracted from faecal samples is usually degraded, we amplified complete mtDNA genomes via 23 fragments with sizes of 1-1.2 kb and an overlap of 100-300 bp (primer information is available upon request). We used 1 U BiothermTaq 5000 (Genecraft, Cologne, Germany) in a 30 μL polymerase chain reaction (PCR) mix (1× reaction buffer, 0.16 mM of each dNTP, 0.33 μM of each primer) with a wax-mediated hot-start technique and with the following thermocycler conditions: 94 °C for 2 min, followed by 40 cycles of 94 °C for 1 min, 50-52 °C for 1 min, 72 °C for 1.5 min and 72 °C for 5 min. As template, c. 100 ng total genomic DNA was added. We conducted all PCRs with at least one PCR blank (molecular-grade water). We ran and checked PCR products on 1% agarose gels, excised DNA fragments of relevant lengths and purified PCR products with the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany). Sequencing was performed on an ABI 3130xl sequencer using the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and the respective forward and reverse primers.
Since in some species the hypervariable region II of the control region contains two or even three homopolymers, direct sequencing of PCR products from both ends revealed incomplete sequences. Thus, respective PCR products were cloned into the pGEM-T Easy vector following the manufacturer's instructions (Promega, Madison, WI, USA) and transferred into One Shot TOP10 Electrocomp Escherichia coli cells (Invitrogen, Carlsbad, CA, USA). Plasmid DNA was extracted with the QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany) and sequenced using M13 primers.
For phylogenetic reconstructions, we expanded our data set with 53 additional mtDNA genomes of Catarrhini taxa derived from GenBank: five additional African green monkeys, five other Cercopithecini, 23 Papionini, ten colobines, three gibbons and seven hominids (Table S1). We included only mtDNA genomes from GenBank that were complete and had fewer than ten ambiguous sites. Sequences were aligned with Muscle 3.8.31 (Edgar, 2010) as implemented in AliView 1.18 (Larsson, 2014). We removed indels and poorly aligned positions with Gblocks 0.91b (Castresana, 2000) using standard settings. Maximum-likelihood (ML) and Bayesian approaches were applied to reconstruct phylogenetic trees using IQ-TREE 1.3.13 (Nguyen et al., 2015) and MrBayes 3.2.6 (Ronquist et al., 2012), respectively. The TIM3 + I + G model was selected as the appropriate model of nucleotide substitution with IQ-TREE and jModelTest 2.1.7 (Dariba et al., 2012) using the Akaike information criterion and the Bayesian information criterion. For Bayesian tree reconstructions, we conducted four independent Markov chain Monte Carlo runs with default temperature of 0.2. We ran all repetitions for 1 million generations with tree and parameter sampling in every 100 generations. We discarded the first 25% of samples as burn-in, resulting in 7501 trees per run. To check the adequacy of the burn-in and convergence of all parameters, we used the uncorrected potential scale reduction factor (Gelman & Rubin, 1992) as calculated by MrBayes and visually inspected the trace of the parameters across generations using the software Tracer 1.6.0 (http://beast.bio. ed.ac.uk/Tracer). We applied AWTY (Nylander et al., 2008) to check whether posterior clade probabilities were also converging. We calculated posterior probabilities (PP) for each split and a phylogram with mean branch lengths from the posterior density of trees. The ML analysis was run with 1000 ultrafast bootstrap (BS) replications (Minh, Nguyen & von Haeseler, 2013).
We estimated divergence times using a Bayesian approach implemented in the BEAST 2.3.2 package (Bouckaert et al., 2014), assuming a relaxed lognormal clock model of lineage variation (Drummond et al., 2006). Eight independent analyses were conducted with varying combinations of substitution models (TrN + I + G or GTR + I + G), tree prior models (Yule or Birth-Death) and two slightly differing fossil-based calibration sets (Table 1). Calibration points in Set 1 were taken from Pozzi et al. (2014), while those applied in Set 2 were derived from Liedigk et al. (2014Liedigk et al. ( , 2015. The rationales to use these calibration points and a discussion of fossils supporting respective nodes are provided in Liedigk et al. (2014Liedigk et al. ( , 2015 and Pozzi et al. (2014). Each analysis was run for 25 million generations, with tree and parameter sampling occurring every 1000 generations. To assess the adequacy of the burn-in and convergence of all parameters, we visually inspected the trace of the parameters across generations using Tracer. We combined sampling distributions of multiple independent replicates with LogCombiner 2.3.2 and summarized and visualized  (2007) and Leakey (1993)

RESULTS
We generated complete mtDNA genomes from ten Chlorocebus individuals representing all six species of the genus and the nine mtDNA haplogroups detected by Haus et al. (2013b). Newly generated mtDNA genomes had lengths of 16 259 to 16 635 bp and consisted of two rRNA genes, 13 protein-coding genes, 22 tRNA genes and the control region. Contamination of our data set with nuclear mitochondrial pseudogenes (numts) can be excluded, because faecal material contains generally highly degraded nuclear DNA (Thalmann et al., 2004), direct sequencing of PCR products revealed no multiple amplifications of different copies (as indicated by double peaks in the electropherograms), and no inconsistent nucleotides were detected in overlapping sequences. Furthermore, all protein-coding genes were correctly translated without premature stop codons.
For phylogenetic analyses, we generated an alignment including the ten newly generated mtDNA genomes and 53 additional mtDNA genomes of Catarrhini taxa obtained from GenBank. After removing indels and poorly aligned positions, the original alignment of 17 216 bp was reduced to 15 454 bp. ML and Bayesian tree reconstructions resulted in nearly identical tree topologies (Fig. S1)  Within Chlorocebus, tree reconstructions revealed nine major clades/lineages (Fig. 3) comprising the following species as delineated by phenotypes and geographic regions: C1: C. tantalus from Nigeria and Central African Republic and C. aethiops from Ethiopia; C2: C. aethiops from Ethiopia, but with unknown exact geographic provenance and intermediate phenotype C. aethiops × C. pygerythrus; C3: C. cynosuros from Zambia; C4: C. cynosuros with unknown geographic provenance and C. pygerythrus from RSA; C5: C. djamdjamensis from Ethiopia; C6: C. pygerythrus from Ethiopia; C7: C. sabaeus; C8: C. pygerythrus from Tanzania and C9: C. tantalus from Ghana. With exception of C. sabaeus and C. djamdjamensis, which were both represented by only one sample, phenotypes of all species are found in more than one clade (C. tantalus: C1 and C9; C. aethiops: C1 and C2; C. cynosuros: C3 and C4 and C. pygerythrus: C4, C6 and C8), revealing para-and polyphylies within the genus Chlorocebus. The branching pattern among lineages and clades is strongly supported in both ML and Bayesian reconstructions, and only the common origin of lineages C3 and C4 (t16) gained lower statistical support (63%, 0.84).

DISCUSSION
We reconstructed a well-supported mtDNA phylogeny for Chlorocebus and estimated divergence times of the different lineages using complete mtDNA genome data. Papio (below) as obtained from the BEAST reconstruction using the GTR + I + G substitution model, a Yule tree prior and Calibration Set 1 (the complete tree is shown in Fig. S1). Newly generated Chlorocebus mtDNA genome sequences are marked with an asterisk (see Table S1). C1-C9 indicate the nine major mtDNA clades within Chlorocebus according to Haus et al. (2013b); x = putative hybrid due to intermediate phenotype. A-G indicate the seven major mtDNA clades within Papio (Zinner et al., 2009(Zinner et al., , 2013. All nodes are numbered (Chlorocebus: t6-t18; Papio: t26-t34). BS and PP values lower than 100% and 1.0 are given at respective nodes. Blue bars represent 95% confidence intervals of divergence times. The timescale is calibrated in million years. The estimated divergence times are given in Table S2. Primate illustrations by Stephen Nash.
Further, based on our phylogenetic reconstruction, we propose a phylogeographic scenario for Chlorocebus and compared it with a scenario for Papio.
As in Haus et al. (2013b), the mtDNA clades revealed from complete mtDNA genomes reflect geographic distributions rather than nominal species. All clades demonstrate good correspondence to geographic regions, except for one C. aethiops sample from Ethiopia, which clusters with C. tantalus from Nigeria and Central African Republic into C1. This exception, however, is in concordance with findings by Shimada (2000) and Haus et al. (2013b). Paraand polyphyletic relationships for all Chlorocebus species were confirmed, except for C. sabaeus and C. djamdjamensis. However, these two taxa were represented by only one sample each, so that possible nonmonophyletic relationships could not be detected. Likewise, also the nuclear genomic phylogeny presented by Warren et al. (2015) is based on a single specimen per species and, thus, allows no inferences about possible para-or polyphylies. Since, in our phylogenetic reconstruction, geographically close lineages cluster together, we follow the suggestion by Haus et al. (2013b) that introgression is the most possible cause for the presence of para-and polyphylies, although other factors such as incomplete sorting cannot be excluded.
Most splits in our phylogenetic reconstruction have stronger support than in Haus et al. (2013b), and the branching pattern among Chlorocebus lineages/clades is almost completely resolved; only node t16 gains lower statistical support. Although clades C1-C9 correspond to the nine clades distinguished by Haus et al. (2013b), the order of branching in our tree reconstruction is different. The first divergence within Chlorocebus is between the westernmost C. sabaeus clade (C7) and all other lineages. This is in agreement with other mtDNA and nuclear DNA studies (Wertheim & Worobey, 2007;Chatterjee et al., 2009;Guschanski et al., 2013;Ayouba et al., 2015;Warren et al., 2015). The most recently emerged lineage within Chlorocebus is represented by the southern lineage consisting of C. cynosuros and C. pygerythrus. This also is in agreement with previous studies on mtDNA (Guschanski et al., 2013;Ayouba et al., 2015) and nuclear DNA (Warren et al., 2015), but in contrast to Haus et al. (2013b), whose analysis could not resolve the relationships of the southern lineages with other clades. Within the southern lineage, as in Haus et al. (2013b), C. cynosuros is found in two clades: C3 and C4, although neither Haus et al. (2013b) nor our analysis provided significant statistical support for this splitting. As suggested by Haus et al. (2013b), this pattern can simply be due to wrong taxonomic determination of the respective specimens. Unfortunately, since the geographic provenance of the C. cynosuros sample from C4 is unknown, this question cannot be solved here.
The overall catarrhine divergence date estimates are on a similar timescale as estimates from earlier molecular studies (Chan et al., 2010;Perelman et al., 2011;Roos et al., 2011;Liedigk et al., 2012). Within Chlorocebus, our dates appear older than the estimates obtained by Guschanski et al. (2013), who dated the radiation within Chlorocebus to 2.2-1 Myr. Interestingly, all mtDNA-based estimates for the Chlorocebus radiation are in stark contrast to those based on nuclear genomic data. According to Warren et al. (2015), the radiation within Chlorocebus is much younger and occurred within the last 550 000 years. This discrepancy could be due to different population histories reflected by different parts of the genome or more likely due to differences in applied methods for estimating divergence times [fossil-based calibration in this study, substitution rate in Warren et al. (2015)]. Within Papio, our mtDNA-based estimates are generally consistent with dates from Zinner et al. (2013), who dated the radiation within the genus to 2.2-1.0 Myr.

phylogeographic implicationS
The first split sets the westernmost clade representing C. sabaeus as sister taxon to all other Chlorocebus taxa. This suggests the origin of Chlorocebus to be north of the Central African rainforest. Our phylogenetic reconstruction further suggests a complicated biogeographic history with at least two temporally independent range expansions into eastern Africa [C5 and C6 (Ethiopia) and C1, C2 and C8 (East Africa)] and one into southern Africa (C3 and C4). Interestingly, similar mitochondrial haplotypes are found in C. tantalus from Nigeria/Central African Republic and C. aethiops from Ethiopia (C1a and C1b). Both taxa are parapatric, but a contact zone might have existed or still exists in southern Sudan.
Although the divergence of Papio appears to be slightly more recent than the divergence of Chlorocebus, it occurred on a similar timescale (2.5-1.4 Myr and 3.5-1.6 Myr, respectively). According to the phylogenetic scenario suggested by Zinner et al. (2009Zinner et al. ( , 2011bZinner et al. ( , 2013, Papio originated in southern Africa and was first prevented from dispersal to the north by the equatorial forest belt. Due to climate changes during glacial periods, the savannah biome expanded, enabling the dispersal to the north. After reaching the northern part of the range, baboons further dispersed to the west and to the east, possibly in two subsequent waves. The same two-wave pattern was suggested for giraffes Giraffa camelopardalis, as populations from Niger seem to be more closely related to East African than to neighbouring Central African populations (Hassanin et al., 2007). A scenario of several temporally independent expansions is also conceivable for Chlorocebus.
The para-and polyphyletic relationships within Chlorocebus and Papio suggest a complicated biogeographic history of these genera. Although it is difficult to make inferences on temporal concordance of evolutionary and climatic events, the last 10 Myr were characterized by repeated climate fluctuations in Africa, leading to dynamic changes in forest and savannah cover (Bonnefille, 2010). The dispersing lineages within both Chlorocebus and Papio probably experienced multiple phases of isolation and reconnection, triggered by recurrent expansion of unsuitable forest or desert habitats during Pleistocene (2.588-0.012 Myr) glacial and interglacial periods, creating various isolated savannah refuges at certain periods which were subsequently reconnected (Nichol, 1999;deMenocal, 2004). The speciation within both Chlorocebus and Papio also appears to be temporally in accordance with general aridification in Africa around 2.9-2.4 Myr that led to changes in savannah habitats: savannah expansions in West and East Africa and loss of savannah habitat in North Africa (deMenocal, 2004;Bonnefille, 2010).
The phylogeographic scenarios for Chlorocebus and Papio represent two main phylogeographic patterns of sub-Saharan African savannah mammals. Chlorocebus represents a scenario with an initial split of western and eastern clades with subsequent dispersal to the south accompanied by divergence into north-eastern and southern clades. A similar phylogeographic pattern of an initial west-east division has been suggested for other African savannah animals, such as common warthogs Phacochoerus africanus (Muwanika et al., 2003), African elephants Loxodonta africana (Nyakaana, Arctander & Siegismund, 2002), African buffalo Syncerus caffer (Van Hooft, Groen & Prins, 2002) and roan antelopes Hippotragus equinus (Alpers et al., 2004). Papio, on the other hand, represents the second phylogeographic pattern with an initial separation of southern and northern lineages with subsequent splits into eastern and western clades in the North. Apart from baboons, this scenario was shown for antelopes of the genus Alcelaphus (Arctander, Johansen & Coutellec-Vreto, 1999;Flagstad et al., 2001), giraffes G. camelopardalis (Hassanin et al., 2007) and lions Panthera leo (Bertola et al., 2011).

CONCLUSION
We generated a phylogeny for the genus Chlorocebus based on complete mtDNA genomes of ten samples representing all species and all major mtDNA clades. We obtained stronger statistical support in the phylogenetic reconstruction than in previous studies based on partial mtDNA sequence data. In accordance with Haus et al. (2013b), the nine major mtDNA clades indicate several cases of paraphyly within the genus and reflect geographic distributions rather than taxonomy. Our results suggest an initial west-east division in the northern part of the genus' range with subsequent divergence into a north-eastern and a southern clade. The general phylogeographic scenario for the dispersal of Chlorocebus thus contrasts with that for another widespread African savannah primate genus, baboons of the genus Papio, which diverged on a similar timescale. The origin of baboons is believed to have been in southern Africa with an initial south-north division with subsequent separation of eastern and western clades in the northern part of the range. The opposing dispersal scenarios for Chlorocebus and Papio thus represent primate cases of the two main phylogeographic dispersal patterns found in sub-Saharan African savannah mammals.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's website: Table S1. Samples, their origin and GenBank accession numbers. Table S2. Support values and estimated divergence ages. Figure S1. Ultrametric tree showing phylogenetic relationships and divergence ages among 63 mtDNA genome sequences as obtained from a BEAST reconstruction using the GTR + I + G substitution model, a Yule prior and calibration set 1. Blue bars indicate 95% credibility intervals of divergence times and the timescale below shows million years before present. Nodes are numbered and refer those listed in Figure 3 and Table S2. Details about ML bootstrap support and Bayesian posterior probabilities as well as divergence times with their 95% credibility intervals are provided in Table S2.