The seroprevalence of SFV was evaluated in 84 mandrills (mean age, 8 years; range, 1-29), comprising 38 males (mean age, 7 years; range, 1-20) and 46 females (mean age, 8.6 years; range, 2-29). Of these, 28 were juveniles (< 4 years); 36 were sub-adults (5-10 years); 6 were adults (11-15 years), and 14 were old adults (> 16 years) (Table 1). We found by Western blot analysis that 70 of the 84 mandrills had gag doublet reactivity, and they were thus considered SFV seropositive (Figure 1), for an overall seroprevalence of 83%. Four were of indeterminate seropositivity, and the 10 others were considered seronegative. As seen in Table 1 the seroprevalence increased significantly with age (p < 0.001), from 57% in juvenile monkeys to 94% in adults and 100% in older mandrills. No significant difference was found between males (84%) and females (82%).

The DNA samples obtained from peripheral blood mononuclear cells (PBMCs) from the 84 mandrills were examined by nested PCR targeting a 425-bp fragment of integrase, a region in the polymerase gene. The 14 seronegative and indeterminate samples were PCR negative. SFV DNA was detected in 61 of 70 seropositive samples (87%); although the other nine mandrills were serologically positive, no SFV DNA could be detected. The sequence of the integrase fragment was obtained for 53 PCR-positive samples (Table 1). Nucleotide sequence comparison showed that 52/53 sequences were closely related, with 94-100% sequence similarity, and they were also closely related to the five SFV sequences previously obtained by Calattini et al. 3 . The one divergent sample, Mnd31CDP, from a wild-born mandrill introduced into the colony at the age of 2 years, showed greater nucleotide divergence (8-9%) than all the other mandrill SFV sequences.

The phylogenetic analysis confirmed these findings, as shown in Figure 2. This tree represents the 11 main SFV strains circulating in the colony and, in the insert, all 53 sequences, including the 11 main strains (in colour). These 53 newly obtained SFV strains belong to a large clade comprising all the available sequences from mandrills and drills, with a high bootstrap value (100%). This clade contains two main clusters. The first comprises most of the new sequences and others previously obtained from mandrills, including the five sequences of Calattini et al. 3 , from the same breeding centre. The second consists of the unique Mnd31CDP strain, which is localized between the large clade of mandrills and that of drills (Figure 2).

To determine the origin and distribution of the different clades in the mandrill colony, a 267-bp portion of the cytochrome b sequence was amplified and sequenced from 21 SFV-infected monkeys in the colony and from eight mandrills caught in the wild (Figure 3) in various regions of Gabon (Figure 4).

As seen in the phylogenetic tree, two distinct clusters could be distinguished, with perfect correlation between cytochrome b sequences and the origin of the wild mandrills. One cluster consisted of mandrills from regions north of the Ogooué River and the second of animals from regions south of the Ogooué. The Mnd31CDP cytochrome b sequence clustered with sequences obtained from mandrills originating in southern Gabon, as did 14 of 21 analysed sequences of cytochrome b from our colony. Only six sequences from other mandrills in our colony clustered with sequences from mandrills from northern Gabon (above the Ogooué River, see Figure 4).

To confirm the hypothesis that mandrills are infected naturally with two different SFV strains, we amplified and sequenced SFV from DNA in blood or tissue samples collected from eight mandrills (pets or 'bush meat') from northern Gabon and seven from the southern part (Figure 4). Eight SFV sequences were obtained and compared with the SFV in our colony. Phylogenetic analysis confirmed that mandrills are infected with two SFV strains (Figure 5). Mnd31CDP clustered with the SFV obtained from wild monkeys from the south, whereas the other strain clustered with newly obtained viruses from wild northern animals. Cytochrome b phylogenetic analysis also confirmed the geographical separation of the wild mandrills (Figure 3).

We evaluated the possible transmission of mandrill SFV to humans by examining 20 people (15 men and 5 women; mean age, 39 years; range, 20-54) occupationally exposed to mandrills as animal caretakers or veterinarians at the Primatology Centre. The mean duration of exposure to nonhuman primates was 12 years (range, 5 months to 27 years). Two of these people (10%) were found to be SFV-seropositive by Western blotting (Figure 1). The SFV integrase sequence was detected by nested PCR in PBMCs from the two seropositive persons, who were found to be the only ones who had been bitten by nonhuman primates during their work at the Centre. The first person (H1CIRMF) was bitten by a chimpanzee on a finger in 1996 and by a mandrill (Mnd2ACDP) on a shoulder during the same year. The second person (H2CIRMF) recalled a bite on a finger by an unknown monkey in 1985. SFV sequences were obtained from amplified 425-bp integrase fragments in PBMC DNA from the two SFV seropositive persons as well as from the chimpanzee and the mandrill Mnd2ACDP. Phylogenetic analysis (Figure 6) showed that the viruses from H1CIRMF and from mandrill Mnd2ACDP were almost identical, with only one base difference (99.7% nucleotide identity). This sequence was not related to the sequence obtained from the chimpanzee. Phylogenetic analysis of the SFV obtained from the second person showed that the virus was located in the clade of Asian SFVs (bootstrap, 96%) and clustered with Macaca fascicularis (Figure 6). The two SFV-infected humans are healthy and show no clinical signs related to a retroviral infection, 15 years after the bites.

To evaluate the genetic variability of SFV in vivo, we investigated the virus population in one mandrill at an interval of 10 years, and we also studied the genetic variation of the virus after transmission to a human through a severe bite. We studied several clones obtained in a single PCR: 18 clones from mandrill Mnd2ACDP in 1996 on the day H1CIRMF was bitten, 13 clones from the same animal 10 years later, and 11 clones from the bitten person 10 years after the bite. Comparative sequence analysis showed strong nucleotide sequence similarity (data not shown), with a major identical strain (12/18 and 9/13 clones identical) among the sequences obtained in the mandrill on day 0 and 10 years later. The major strain (4/11 clones) in the infected person differed by one base from the major mandrill strain. The other clones in the two mandrill and the human samples differed only slightly, with a divergence of one or two bases. Also, as seen in Additional file 1 clones sequenced from H1CIRMF clustered mainly at the top of the tree, while sequences of the clones from Mnd2ACDP clustered in the middle, close to some published sequences. Only one clone sequence from H1CIRMF, CIRMF1C9, was closely related to a clone sequence from Mnd2ACDP (Mnd2AC10Y10).

We studied 84 mandrills in the semi-free-ranging colony housed at the Primatology Centre of the International Centre for Medical Research in Franceville, Gabon. Wild-born, wild-caught animals and animals born at the Centre were maintained in accordance with the guidelines of the United States National Institutes of Health. The six male and eight female mandrills that founded the colony were brought from various parts of Gabon and released into the enclosure in 1983 43 46 . A small colony of macaques (M. fascicularis) was also founded in 1983. Blood samples from monkeys in this colony are collected every year, stored at -80°C and tested for different retroviruses, including SFV. Thus, between November 2006 and January 2007, 7 ml of blood were collected from mandrills in EDTA-K2 tubes under ketamine HCl anaesthesia (10 mg/kg body weight). Plasma and PBMCs obtained after Ficoll separation were kept frozen.

Wild mandrills were collected in cities and villages throughout the country and in the Lopé Reserve. We collected blood samples from locally captured live animals (pets) and from wild mandrills as previously described 43 . Small amounts of tissue (donated by hunters) were also collected from fresh cadavers in villages or on markets 54 . No money or favours were exchanged for these samples in order to prevent any increase in demand for 'bush meat'. All samples were collected with the approval of the Gabonese Government and in accordance with national laws. Tissue samples were immediately preserved and then stored at -20°C until tested.

To evaluate possible transmission of SFV from mandrills to humans, blood was collected from caretakers or veterinarians working at the Primatology Centre. The participants were volunteers, and fully informed consent was obtained from each person before testing. The samples were anonymous, but age and information about the contact, such as a bite, scratches or other wounds, were retained (for 12 years of mean length of potential exposure to animals). The study obtained ethical clearance from the public health authorities.

Plasma from mandrills was screened for the presence of foamy virus antibodies as described previously 4 25 60 . Briefly, a Western blot assay was performed with an SFV-infected BHK-21 cell line as the source of foamy viral antigens 27 . Plasma was tested at 1:100 dilution. Western blot seropositivity was defined as the presence of reactivity to the Gag doublet of 70 kDa and 74 kDa, as previously described 4 . Samples without reactivity to either Gag protein were considered seronegative, and those with reactivity to a single band in the 70- to 74-kDa molecular mass range were considered indeterminate. The Western blot positive control was serum from an SFV-positive chimpanzee, used by Calattini et al. 16 . The negative serum was obtained from a human who had never been in contact with a nonhuman primate.

High relative molecular mass genomic DNA was extracted from PBMCs from the tested animals and tested against several positive and negative controls with the Qiagen kit (QIAmp blood Mini Kit, Courtaboeuf, France). The first round of PCR involved a described set of primers 61 (primer 1: GCC ACC CAA GGG AGT TAT GTG G, and primer 2: GCT GCA CCC TGA TCA GAG TG) for amplifying an integrase fragment of 590 bp (a region in the polymerase gene), under the following conditions: 40 cycles of 30 s of denaturation at 94°C, 30 s of annealing at 55°C and 1 min of extension at 72°C. A 425-bp fragment corresponding to another portion of the integrase was amplified under the same conditions with nested primers (primer 3: CCT GGA TGC AGA GTT GGA TC and primer 4: GAA GGA GCC TTA GTG GGG TA), as reported previously 25 60 61 .

The presence and quality of the extracted DNA were verified by amplifying an albumin gene fragment. Amplification and detection of albumin were carried out as described for SFV pol sequences, but with specific primers (forward: AlbF: GCT GTC ATC TCT TGT GGG CTG T and reverse: AlbR: ACT CAT GGG AGC TGC TGG TTC) 62 . Molecular amplification was also performed, with the same program, to study the 267-bp cytochrome b region, which was sufficiently variable to differentiate the northern and southern populations of mandrills 54 with these specific primers: L14725: CGA AGC TTG ATATGA AAA ACC ATC GTT G and H15149: AAA CTG CAG CCCCTC AGA ATG ATA TTT GTC CTC A 63 . Positive PCR products were directly sequenced. In order to evaluate genetic drift in vivo, purified PCR products were cloned with the pCR2.1 TOPO plasmid (Invitrogen, Carlsbad, California, USA), and various positive clones were selected, extracted, purified and sequenced with an automatic sequencing system (GATC, Germany).

All the SFV and cytochrome b sequences from mandrills and humans obtained in this study have been submitted to GenBank as cytochrome b (accession numbers GU169713 to GU169741) and SFV (accession numbers GU169742 to GU169847).

For the phylogenetic analysis, the new SFV sequences were aligned with the ClustalW (1.81) program 64 and then analysed and edited with Bioedit http://www.mbio.ncsu.edu/BioEdit/bioedit.html. The final alignment was submitted to the the Bayesian method implemented in MrBayes version 3.1 software (2005) 65 with the Jones, Taylor and Thornton model 66 and the rtREV model 67 of evolution and gamma distributed rates at sites, with one million generations and burn-in of 2.5%. Bayesian parameters were examined with the Tracer program http://evolve.zoo.ox.ac.uk/Evolve/Software.html to determine convergence to a stable log likelihood value. Likelihood traces between replicate runs were compared for convergence to similar log likelihood values. All estimated sample sizes were greater than 545 68 . If replicate runs converged, all trees after burn-in were combined to create a single consensus tree. BMCMC posterior probability values represent the proportion of MCMC samples that contain a particular node. The final phylogenetic tree was obtained by majority rule consensus and after editing with the graphic resources contained in the FigTree v1.2 software http://tree.bio.ed.ac.uk/software/figtree.

SFV serological status in relation to sex and age group was analysed statistically by the chi-squared test with Yates correction, and prevalence and odds ratios were calculated. The corresponding 95% confidence intervals were reported as measures of statistical significance. The Mann-Whitney U test was also used for statistical analysis. Significance was assumed at p < 0.05. All analyses were performed with Statistica software v7.1 (StatSoft France, http://www.statsoft.fr).