Open Access

Research Article

• Download: XML | PDF | Citation
• E-mail this Article
• Order Reprints
• Print this Article
• Bookmark this page:
  

Introduction

Foamy viruses (also termed spumaviruses) are complex retroviruses that naturally infect numerous mammal species, including primates, felines, bovines and equines, but not humans [1]–[4]. Simian foamy viruses (SFVs) have been identified in a wide variety of primates, including prosimians, New World and Old World monkeys as well as apes, and each species has been shown to harbor a unique (species-specific) strain of SFV [5]–[13]. Moreover, closely related SFVs have been isolated from closely related primate species: a comparison of phylogenies derived from SFV integrase and primate mitochondrial DNA sequences revealed highly congruent relationships, indicating virus-host co-evolution for at least 30–40 million years [13]. This ancient relationship may be responsible for the non-pathogenic phenotype of SFV: Although highly cytopathic in tissue culture, the various SFVs do not seem to cause any recognizable disease in their natural hosts [2],[3],[14]. SFVs are highly prevalent in captive primate populations, with infection rates ranging from 70% to 100% in adult animals [2], [3], [5], [15]–[19]. Transmission is believed to occur through saliva because large quantities of viral RNA, indicative of SFV gene expression and replication, are present in cells of the oral mucosa [3], [20]–[22]. However, little is known about the prevalence and transmission patterns of SFV in wild-living primate populations.

Although there is no human counterpart of SFV, humans are susceptible to cross-species infection by foamy viruses from various primate species. Indeed, the first “human foamy virus” [23] isolated from a Kenyan patient with nasopharyngeal carcinoma more than three decades ago was subsequently identified to be of chimpanzee origin [7],[8]. Since then, SFV strains from African green monkeys, baboons, macaques and chimpanzees have been identified in zookeepers and animal caretakers who acquired these infections through occupational exposure to primates in captivity [19], [24]–[27]. More recently, about 1% of Cameroonian villagers who were exposed to primates through hunting, butchering and the keeping of pet monkeys were found to be SFV antibody positive, and genetic analysis of three such cases documented infection with SFV strains from DeBrazza's monkeys, mandrills and gorillas [10]. Finally, a large proportion of individuals (36%) who were severely bitten and injured while hunting wild chimpanzees and gorillas had detectable SFVcpz or SFVgor sequences in their blood [28]. Thus, humans are susceptible to a wide variety of SFVs and seem to acquire these viruses more readily than other retroviruses of primate origin, such as simian immunodeficiency viruses (SIVs) or simian T-lymphotropic viruses (STLVs). Interestingly, these infections appear to be non-pathogenic and thus far exhibit no evidence of onward transmission by human-to-human contact; however, additional studies will need to be conducted to fully characterize the natural history of SFV infections in humans [10], [24], [28]–[30].

Among wild primates, chimpanzees (Pan troglodytes) are of particular public health interest since they harbor SIVcpz, the precursor of the human immunodeficiency virus type 1 (HIV-1) [31]–[34]. There are four proposed chimpanzee subspecies which have been defined on the basis of geography and differences in mitochondrial DNA (mtDNA) sequences [35],[36]. These include P. t. verus in west Africa, P. t. vellerosus in Nigeria and northern Cameroon, P. t. troglodytes in southern Cameroon, Gabon, Equatorial Guinea and the Republic of Congo, and P. t. schweinfurthii in the Democratic Republic of Congo and countries to the east (Figure 1). Two of these, P. t. troglodytes and P. t. schweinfurthii, are naturally infected with SIVcpz, but only P. t. troglodytes apes have served as a reservoir of human infection [31]–[34]. It is now well established that SIVcpzPtt has been transmitted to humans on at least three occasions, generating HIV-1 groups M, N and O. Moreover, two of these cross-species infections (groups M and N) have been traced to distinct P. t. troglodytes communities in southeastern and southcentral Cameroon, respectively [33]. The reason for the emergence of SIVcpzPtt strains, but not SIVcpzPts strains, in humans is unknown, but could reflect regional differences in the types and frequencies of human/chimpanzee encounters. Thus, examining humans for SFVcpz infection might be informative as to the location(s) where human/chimpanzee contacts are most common; however, no information exists regarding the prevalence, geographic distribution and genetic diversity of SFVcpz in chimpanzees in the wild.

Figure 1. Location of wild chimpanzee study sites.

Field sites are shown in relation to the ranges of the four proposed chimpanzee subspecies. White circles indicate forest areas where fecal samples were collected for prevalence studies (Table 3). These were located in Cote d'Ivoire (TA), Cameroon (MF, WE, MP, MT, DG, DP, BQ, CP, EK, BB, MB, LB), the Central African Republic (ME), Gabon (LP), Republic of Congo (GT), Democratic Republic of Congo (BD, WL, WK), Uganda (KB), Rwanda (NY), and Tanzania (GM-MT, GM-KK, GM-KL, MH). Black circles indicate forest sites where eight ancillary samples (BA432, BF1167, EP479, EP486, KS310, UB446, WA466, WA543) were collected. International borders and major rivers are shown.

doi:10.1371/journal.ppat.1000097.g001

In this study, we sought to develop an experimental strategy that would allow us to identify and molecularly characterize SFVcpz infection in wild-living chimpanzees by entirely non-invasive means. The rationale for this approach was two-fold. First, we wished to explore whether large scale screening of endangered primates for infectious agents other than primate lentiviruses was feasible. Second, we wished to examine whether SFVcpz could serve as a test case in efforts to develop suitable early warning systems for pathogens that might infect humans exposed to wild animals. To this end, we tested whether fecal based methods previously developed for SIVcpz could be adapted to the non-invasive detection and molecular characterization of SFVcpz. Our results show that this was indeed possible. Using these newly developed methods, we determined the prevalence of SFVcpz infection in wild chimpanzee communities throughout equatorial Africa, molecularly characterized 120 new SFVcpz strains, examined the subspecies association and phylogeography of SFVcpz, documented numerous instances of SFVcpz co-infection and recombination, investigated the routes of SFVcpz transmission in the wild, and examined the frequency of SFV cross-species transmissions from prey species. Our results reveal important new insights into the molecular ecology and natural history of SFVcpz infection that could not have been gained from studies of captive chimpanzees, and show more generally how endangered primates can be studied by non-invasive molecular approaches to elucidate the circumstances and mechanisms of pathogen transmission.

Results

Fecal-based methods for SFVcpz antibody and nucleic detection

SFV infection of primates and humans is generally diagnosed by documenting virus specific anti-Gag antibodies in serum or plasma using ELISA and/or Western blot approaches [8],[10],[18],[19]. The infecting SFV strain is then molecularly characterized by amplifying viral DNA from peripheral blood mononuclear cell (PBMC) or other tissue DNA [8], [10], [11], [13], [16]–[18],[24],[26],[28]. Since collecting blood from wild chimpanzees is not feasible, we sought to develop methods of SFVcpz detection that are entirely fecal-based. To accomplish this, we examined whether existing methods of SIVcpz fecal antibody and nucleic acid detection [33],[37],[38] could be adapted to the non-invasive identification and molecular characterization of SFVcpz.

Western blot strips were prepared from sucrose purified SFVcpz virions and used to test 40 fecal extracts from 23 SFVcpz infected chimpanzees from the Yerkes Primate Research Center (Table 1). Reactivity with the two SFVcpz Gag proteins p74 and p71 was scored positive, following interpretive guidelines established for serum antibody positivity [8],[10],[18]. The absence of viral bands was scored negative, and samples that did not meet either criterion were classified as indeterminant. Using this approach, SFVcpz specific IgG antibodies were detected in 29 of 40 fecal extracts from infected chimpanzees (Table 1). All samples reacted with the Gag doublet and a subset also recognized the accessory Bet (p60) protein (Figure 2A). In contrast, none of 21 fecal extracts from uninfected human volunteers exhibited false-positive or indeterminant Western blot reactivities (Figure 2A; Table 1).

Figure 2. Detection of SFVcpz antibodies in chimpanzee fecal samples.

Enhanced chemiluminescent (ECL) Western blot analysis of fecal extracts from (A) human volunteers and captive chimpanzees, and (B) SFVcpz infected wild chimpanzees representing four different chimpanzee subspecies. Strips were prepared using an infectious molecular clone (pMod-1) of SFVcpzPts (see Methods). Samples are numbered, with letters indicating the species (panel A) or collection site (panel B) of origin. Molecular weights of SFVcpz specific Gag and Bet proteins are shown. The banding pattern of plasma from an SFVcpz infected chimpanzee (used at a 1:100,000 dilution) and an uninfected human are shown as positive (Pos) and negative (Neg) controls, respectively.

doi:10.1371/journal.ppat.1000097.g002

Table 1. Validation of Fecal-Based Antibody and Nucleic Acid Detection Assays Using Samples from SFVcpz Infected Captive Chimpanzees and Uninfected Human Volunteers.

doi:10.1371/journal.ppat.1000097.t001

We also investigated whether SFVcpz nucleic acids could be detected in fecal samples using primers designed to amplify a conserved 425 bp fragment (pol-IN) in the viral pol gene (Figure 3) [11]–[13],[18]. In vitro studies have shown that foamy viruses, in contrast to other retroviruses, reverse transcribe their RNA genome before they assemble into virus particles and bud from infected cells [39],[40]. Thus, infectious foamy virus particles have been reported to contain mostly viral DNA, while productively infected cells contain mostly viral RNA [1],[40],[41]. Using nested PCR to analyze fecal samples from the 21 infected chimpanzees, we found SFVcpz DNA in only 2 of 40 samples (Table 1). However, RT-PCR of fecal RNA from these same specimens yielded amplification products for 30 samples. Sequence analysis confirmed the authenticity of the amplification products and identified 11 distinct SFVcpz strains (Table 1). Omission of the cDNA synthesis step during the RT-PCR procedure failed to yield detectable amplification products. These results thus indicate that SFVcpz is present in chimpanzee fecal samples mostly as viral RNA, the source of which (cell associated, cell free, or both) remains to be determined.

Figure 3. Location of RT-PCR derived amplicons in the SFVcpz genome.

Amplification products are shown in relation to the corresponding regions in the SFVcpz genome, with the length of the amplified fragments indicated. The genomic organization of SFVcpz is shown on the top (structural and accessory genes are drawn to scale) [79].

doi:10.1371/journal.ppat.1000097.g003

The sensitivities of SFVcpz antibody and viral nucleic acid detection in fecal samples from captive chimpanzees were determined to be 73% and 75%, respectively (Table 2). Assay specificities were 100% (Table 1). Interestingly, not all fecal vRNA positive chimpanzees were also fecal Western blot positive (and vice versa). Two SFVcpz infected apes (CPZ6, CPZ23), each of whom had detectable RNA in at least two independent stool samples, were repeatedly fecal antibody negative (Table 1). Since both individuals had high titer antibodies in their blood, this was not due to a recently acquired SFVcpz infection. Two other apes (CPZ7, CPZ20) were fecal antibody positive, but virion RNA negative (Table 1). Thus, antibody or virion RNA screening alone would have missed SFVcpz infection in these individuals. Nonetheless, Western blot together with RT-PCR correctly diagnosed SFVcpz infection in 21 of 23 captive chimpanzees, suggesting that the newly developed assays were of sufficient sensitivity and specificity for field surveys, especially when used in combination.

Table 2. Sensitivities of Antibody and Viral RNA Detection in Fecal Samples From Captive and Wild Chimpanzees.

doi:10.1371/journal.ppat.1000097.t002

Geographic distribution, subspecies association and prevalence of SFVcpz in wild-living chimpanzees

To determine to what extent chimpanzees are infected with SFVcpz in the wild, we tested 724 fecal samples from 25 different field sites across equatorial Africa for virus specific antibodies and/or viral RNA (Table 3). Samples were selected from existing specimen banks based on their geographic representation, available host genetic information (mtDNA, microsatellite and sex markers), and remaining quantities of material. Figure 1 depicts the geographic location of the sites with respect to the ranges of the four proposed chimpanzee subspecies. Except for P. t. verus, all other subspecies were sampled at multiple sites. Specimens from the Taï Forest (TA) as well as from Gombe (GM-MT, GM-KK) and Mahale Mountains (MH) National Parks were collected from individually known (habituated) chimpanzees under direct observation. Samples from the Goualougo Triangle (GT), several field sites in Cameroon (DP, EK, BB, MB, LB), and the Kalande community (GM-KL) in Gombe National Park were obtained from non-habituated chimpanzees, but were subsequently genotyped using mtDNA, microsatellite and sex markers and thus also represent known numbers of individuals [33] (B. Keele and B. H. Hahn, unpublished). Samples from the remaining field sites in Cameroon (MF, MP, WE, MT, BQ, DG, CP), Gabon (LP), the Central African Republic (ME), the Democratic Republic of Congo (BD, WL, WK), Rwanda (NY) and Uganda (KB) were derived from an unknown number of chimpanzees. All samples were previously screened for SIVcpz antibodies and/or vRNA [33],[37],[42] (F. van Heuverswyn and M. Peeters, unpublished) and their integrity was confirmed by mtDNA analysis (Table S1).

Table 3. Prevalence Rates of SFVcpz Infection in Wild Chimpanzees throughout Equatorial Africa.

doi:10.1371/journal.ppat.1000097.t003

Of 724 fecal samples included in the analysis (Table 3), 706 were tested by Western blot analysis and 211 were found to be SFVcpz antibody positive (18 samples were of insufficient quantity for immunoblot analysis but were used for RT-PCR amplification). All of these reacted with the Gag p74/p71 doublet and a small number also recognized the p60 Bet protein (Figure 2B). Interestingly, two samples from the DP site reacted only with the Bet protein and were thus classified as indeterminant (not shown). The remaining 493 fecal extracts exhibited no detectable bands and were thus classified as antibody (SFVcpz IgG) negative. A subset of samples (n = 392) was also examined for SFVcpz nucleic acids (Table 3). RT-PCR of fecal RNA yielded pol-IN (425 bp) amplification products for 175 samples, all of which were shown to contain SFVcpz sequences (two samples were RT-PCR positive using LTR and pol-RT primers, respectively). In contrast, amplification of fecal DNA from these same samples failed to yield viral sequences (not shown), providing further evidence for the presence of SFVcpz RNA, and not DNA, in fecal material. A breakdown of antibody and RNA positive samples for each field site is shown in Table 3. The results revealed SFVcpz infected chimpanzees at all field sites.

We next sought to determine the prevalence of SFVcpz infection at each of the 25 field sites. To accomplish this, we examined whether fecal antibody and vRNA detection tests yielded similar data for captive as well as wild communities. Inspection of Table 3 indicated that this was not the case. For example, at the TA field site all of 16 fecal samples were SFVcpz antibody positive (100%), but only 7 contained vRNA (44%). In contrast, at the ME field site none of 21 fecal samples contained antibodies (0%), while 16 were vRNA positive (76%). Importantly, the latter was not due to a lack of antibody cross-reactivity since other P. t. troglodytes samples (e.g., 11 of 20 GT samples) were Western blot positive using the same antigens (Table 3, Figure 2B). To examine this further, we re-calculated test sensitivities using only samples from SFVcpz infected wild chimpanzees (Table 2). This yielded surprising results: not only did test results vary extensively between different field sites, the sensitivities of antibody and vRNA detection were also inversely correlated (Figure 4). To account for this in prevalence estimations, we decided to calculate a “field sensitivity” for each test by averaging values across all collection sites. The rationale for this was that the strong negative correlation between the two assay sensitivities would predict that if the sensitivity of one test was underestimated, the sensitivity of the other test would be overestimated to a roughly similar degree. Thus, if samples were subject to an equal number of both tests, these effects would tend to even out. While many samples were not subject to equal numbers of the two tests, this nonetheless seemed to represent the most reasonable approach. For both Western blot and RT-PCR assays, the average sensitivities across all sites were around 56%. Therefore we pooled results from all tests to obtain a general sensitivity value (56.3%) that was then used to calculate the prevalence rates.

Figure 4. Inverse correlation of fecal antibody and viral RNA detection at different field sites.

Fecal viral RNA (x-axis) and antibody (y-axis) detection sensitivities are plotted for field sites with known numbers of infected chimpanzees (Table 2). The size of the circle is directly proportional to the number of samples tested (results from the three Gombe communities were combined). Color coding and corresponding two letter codes are as in Figure 1. Test sensitivities are significantly inversely correlated (P<0.001).

doi:10.1371/journal.ppat.1000097.g004

Table 3 summarizes the prevalence of SFVcpz infection at 25 different field sites. For 11 sites, these values were calculated based on the proportion of infected individuals. For the remaining sites, prevalence rates were estimated based on the proportion of antibody and/or vRNA positive fecal samples, but correcting for repeat sampling and sample degradation (see Methods). The results revealed uniformly high infection rates for all sites, similar to previously reported values for captive primate populations [15]–[19]. The highest prevalences (>90%) were found at a P. t. verus field site in Cote d'Ivoire (TA); two P. t. vellerosus field sites in central Cameroon (MF, MP); four P. t. troglodytes field sites in Cameroon (DG, CP), Gabon (LP) and the Central African Republic (ME); and three P. t. schweinfurthii field sites in the DRC (BD), Uganda (KB) and Tanzania (GM-MT). The lowest infection rates (<60%) were identified at three P. t. troglodytes field sites in southern Cameroon (BQ, MB, LB). Given that the confidence intervals for the prevalence estimates across the various sites showed extensive overlap (Table 3), these differences are unlikely to be significant. These data thus indicate that SFVcpz is widely distributed and infects chimpanzees at very high prevalence rates throughout their natural habitat.

Documentation of SFVcpz/SIVcpz co-infections

The fact that all 724 fecal samples had independently been tested for SIVcpz antibodies and/or viral nucleic acids provided an opportunity to compare the two viruses with respect to their relative infection frequencies and geographic distribution. As reported previously, natural SIVcpz infection has been identified only in two of the four chimpanzee subspecies (P. t. troglodytes and P. t. schweinfurthii), and then only in a fraction of sampled communities [33],[37],[42]. Thus, SIVcpz is clearly much less common and widespread among wild chimpanzees than is SFVcpz. Nonetheless, the current survey included field sites where both SFVcpz and SIVcpz infections were endemic. To examine whether the two infections were epidemiologically linked, we selected seven sites with known numbers of SFVcpz and/or SIVcpz infected chimpanzees. Four of these were located in Cameroon (MB, LB, EK, DP), while the other three were located in Gombe National Park in Tanzania (GM-KK, GM-KL, GM-MT). Table 4 summarizes the results: Of a total of 130 chimpanzees tested, 55 were infected only with SFVcpz, 7 were infected only with SIVcpz, and 15 harbored both viruses. Thus, SFVcpz/SIVcpz co-infections are not uncommon at locations where both viruses are endemic; however, examination of the relative frequencies of single and dual infections at individual sites, or sites in combination, revealed no association between SFVcpz and SIVcpz (Fisher exact test; P>0.2). Thus, there was no evidence that infection with one of these viruses increased or decreased the likelihood of infection by the other.

Table 4. Number of SFVcpz and SIVcpz Infections in Chimpanzee Communities Harboring Both Viruses.

doi:10.1371/journal.ppat.1000097.t004

Patterns of SFVcpz transmission in the wild

To determine under what circumstances chimpanzees acquire SFVcpz in the wild, we screened members of two habituated communities for evidence of infection. The Kasekela and Mitumba communities are located in Gombe National Park and have been under human observation since the 1960s and 1980s, respectively [43]. Chimpanzees from both communities are followed daily (with particular individuals selected for all-day observation) and their reproductive states and social interactions are recorded. Thus, for many Mitumba and Kasekela apes, especially the more recent offspring, the date of birth is known. This provided an opportunity to compare the frequency of SFVcpz infection among individuals representing different age groups. Testing the most recent fecal sample available, we found no evidence of SFVcpz infection in four infants age 2 years or younger. In addition, only three of ten chimpanzees ages 2.1 to 9 years were found to be SFVcpz antibody and/or viral RNA positive. In contrast, all of 13 adult chimpanzees ages 14 to 45 years were SFVcpz infected (Figure 5). Thus, there was a significant increase of SFVcpz infection with age, suggesting horizontal rather than vertical (perinatal) transmission as the predominant route of infection in these communities.

Figure 5. Increase of SFVcpz infection rates with age.

Members of the habituated Mitumba and Kasekela communities in Gombe National Park were non-invasively tested for SFVcpz infection and their infection rate (y-axis) plotted by age group (x-axis). Group 1 comprises 4 infants age 2 or younger; group 2 comprises 10 chimpanzees age 2.1 to 9 years; and group 3 comprises 13 adult chimpanzees age 14 to 45.

doi:10.1371/journal.ppat.1000097.g005

To investigate whether perinatal transmission was responsible for at least some of the newly acquired infections, we tested longitudinal samples from the three SFVcpz positive offspring and their infected mothers. As shown in Table 5, Fansi (born in November 2001) was fecal Western blot positive in June 2004 (2.6 years of age), but both fecal antibody and viral RNA negative two years earlier in August of 2002. Similarly, Flirt (born in July 1998) was fecal Western blot positive in October 2001 (3.2 years of age), but antibody and viral RNA negative one year earlier in November 2000. Although false negative results at the earlier timepoints cannot be excluded, these data suggest that the two infants acquired SFVcpz after their first and third year of life, respectively. Analysis of the third mother/offspring pair also failed to provide evidence for perinatal transmission. Although Tarzan (born in October 1999) was SFVcpz fecal antibody positive at the earliest timepoint (2.6 years of age) and harbored a virus that was identical in its pol-IN sequence to that of his mother's, the same pol-IN sequences were also recovered from three other chimpanzees, including Flirt and one unknown individual from the neighboring Kalande community. Thus, it is unclear whether Tarzan acquired his SFVcpz infection from his mother during or shortly after birth, or whether he became infected later by another route. Taken together, none of these three mother/offspring pairs provided conclusive evidence for vertical transmission of SFVcpz in the wild.

Table 5. SFVcpz Infection in Three Mother-Offspring Pairs in Gombe National Park.

doi:10.1371/journal.ppat.1000097.t005

SFVcpz evolution at the subspecies level

To determine the evolutionary relationships of SFVcpz strains infecting wild chimpanzees in different parts of equatorial Africa, we selected 392 fecal samples for RT-PCR analysis. Using primers designed to amplify a conserved 425 bp pol-IN fragment [11]–[13],[18], we recovered SFVcpz sequences from 175 samples (one sample yielded only LTR and another only pol-RT sequences; not shown). Pol-IN sequences were also amplified from two P. t. vellerosus apes housed in a Cameroonian sanctuary (SA161 and SA163) as well as eight wild-living P. t. schweinfurthii apes who were sampled at different locations within the Democratic Republic of Congo (BA432, BF1167, EP479, EP486, KS310, UB446, WA466, WA543; Figure 1). The phylogenetic relationships of these SFVcpz sequences to each other and to subspecies specific SFVcpz reference sequences from the database are shown in Figure 6. The analysis revealed three well-defined SFVcpz clades for viruses from P. t. verus, P. t. vellerosus, and P. t. schweinfurthii apes, respectively, each supported with very high posterior probabilities. In contrast, SFVcpz strains from P. t. troglodytes formed two distinct (well-supported) groups in the maximum clade credibility (MCC) tree: (i) one major group which comprised the great majority of the newly identified P. t. troglodytes strains, and (ii) one minor group which included only four strains from the Lope Reserve in Gabon and which formed a sister clade to SFVcpz from P. t. schweinfurthii (Figure 6). Since the placement of the Lope group apart from the other P. t. troglodytes strains was not supported by a high posterior probability, we wondered whether its unexpected position in the MCC tree might be due to the short length of the pol-IN (425 bp) fragment. To clarify these relationships, we amplified additional gag (616 bp) and pol-RT (717 bp) fragments from a subset of samples. Indeed, phylogenetic analysis of these larger fragments placed a representative of the “Lope variant” (LP29) together with the other SFVcpzPtt strains within a single cluster. In the gag region, this clade was supported with a highly significant posterior probability (Figure 7). In the pol-RT region, where the posterior probability was not high, the MCC tree nevertheless placed all P. t. troglodytes sequences in a monophyletic clade (Figure 8). Moreover, an analysis of combined pol-IN and pol-RT data (not shown) yielded a monophyletic P. t. troglodytes SFVcpz clade, with 100% posterior probability. Thus, SFVcpz strains from wild chimpanzees grouped into four major lineages according to their subspecies of origin.

Figure 6. Evolutionary relationships of newly derived SFVcpz strains in the pol-IN region.

Pol-IN (425 bp) sequences were analyzed using the Bayesian Markov chain Monte Carlo (BMCMC) method implemented in BEAST. Sequence LM183 (from a wild bonobo) was included as an outgroup. The maximum clade credibility (MCC) tree topology inferred using TreeAnnotator v1.4.7 is shown, with branch lengths depicting the mean value for that branch in the upper half of the MCMC sample. Posterior probabilities (expressed as percentages) are indicated on well-supported nodes, either as asterisks (100%) or filled circles (90%–99%). Newly identified SFVcpz strains are color coded according to their subspecies of origin (as shown in Figure 1). Representative strains from the database are shown in black. Plus signs (+) denote sequences that represent placeholders of multiple viruses with identical sequences (a complete list is provided in Table S2). Sample WE464 (boxed) was collected in the P. t. vellerosus range, but has a P. t. troglodytes mtDNA haplotype (Figure S1). Arrows identify distinct SFVcpz strains (termed A or B) that were found in the same sample. The scale bar represents 0.02 substitutions per site.

doi:10.1371/journal.ppat.1000097.g006

Figure 7. Evolutionary relationships of newly derived SFVcpz strains in the gag region.

Gag (616 bp) sequences were analyzed as described in Figure 6. The gag tree was rooted using a relaxed clock. Posterior probabilities are indicated on well-supported nodes, either as asterisks (100%) or filled circles (90%–99%). Newly identified SFVcpz strains are color coded according to their subspecies of origin (Figure 1). Representative strains from the database are shown in black. Plus signs (+) denote sequences that represent placeholders of multiple viruses with identical sequences (Table S2). Sample WE464 (boxed) was collected in the P. t. vellerosus range, but has a P. t. troglodytes mtDNA haplotype (Figure S1). The scale bar represents 0.02 substitutions per site.

doi:10.1371/journal.ppat.1000097.g007

Figure 8. Evolutionary relationships of newly derived SFVcpz strains in the pol-RT region.

Pol-RT (718 bp) sequences were analyzed as described in Figure 6. The tree was rooted using LM183 as an outgroup. Posterior probabilities are indicated on well-supported nodes, either as asterisks (100%) or filled circles (90%–99%). Newly identified SFVcpz strains are color coded according to their subspecies of origin (Figure 1). One representative strain from the database (HFV) is shown in black. Plus signs (+) denote sequences that represent placeholders of multiple viruses with identical sequences (Table S2). Sample WE464 (boxed) was collected in the P. t. vellerosus range, but has a P. t. troglodytes mtDNA haplotype (Figure S1). Arrows identify distinct SFVcpz strains (termed A or B) that were found in the same sample. The scale bar represents 0.02 substitutions per site.

doi:10.1371/journal.ppat.1000097.g008

To examine further the evolution of SFVcpz at the subspecies level, we obtained mitochondrial DNA sequences (hypervariable D loop region) from all SFVcpz vRNA positive fecal samples and performed a Bayesian Markov chain Monte Carlo (BMCMC) phylogenetic analysis (Figure S1). The topology of this tree was similar to previous mtDNA phylogenies in several key features [33]: (i) P. t. verus and P. t. vellerosus as well as P. t. troglodytes and P. t. schweinfurthii clustered together, forming two highly divergent lineages; (ii) P. t. verus and P. t. vellerosus formed two well separated sister clades; and (iii) P. t. schweinfurthii fell within the P. t. troglodytes radiation. Comparison of this mtDNA phylogeny with those of SFVcpz pol-IN, pol-RT and gag regions (Figures 6, 7, 8) revealed a number of differences. Most notably, SFVcpz strains from P. t. vellerosus were much more distant from SFVcpz strains infecting P. t. verus than would have been predicted based on mtDNA phylogenies of their respective hosts. In both gag and pol-RT trees, P. t. vellerosus viruses shared a most recent common ancestor with strains from P. t. troglodytes rather than with strains from P. t. verus (as with the placement of the Lope strains, the gag pattern was mirrored in the MCC tree from the pol-RT analysis, albeit without significant support). In addition, SFVcpz from P. t. troglodytes apes formed a single clade (Figures 7 and 8), while their corresponding mtDNA sequences were paraphyletic, being separated by the P. t. schweinfurthii clade (Figure S1). In many cases, chimpanzees with highly divergent mtDNA haplotypes harbored closely related SFVcpz strains, and vice versa. Finally, one fecal sample (WE464) collected north of the Sanaga River contained SFVcpz sequences from P. t. vellerosus, but mtDNA sequences from P. t. troglodytes (boxed in Figures 6, 7, 8, S1). While the latter finding is most simply explained by the migration of a P. t. troglodytes ape across the Sanaga River some time in the past, followed by infection of her progeny with the local variety of SFVcpz, the other discordances are more difficult to interpret. It is clear that SFVcpz is not strictly maternally inherited, since its evolutionary history shows differences with the mtDNA tree. Moreover, the mtDNA phylogeny (Figure S1) offers only a limited perspective on the ancestral relationships of chimpanzee populations, even setting aside any possible inaccuracies due to the short fragment analyzed. Thus, deciphering chimpanzee evolution in the more recent past will require additional study. However, the fact that 120 naturally occurring SFVcpz strains clustered in strict accordance with their mtDNA-defined subspecies of origin provides compelling evidence for virus-host co-evolution.

Phylogeography of SFVcpz

As shown in Figure 1, three of the four chimpanzee subspecies were sampled at multiple locations. This provided an opportunity to examine whether viruses from P. t. vellerosus, P. t. troglodytes and P. t. schweinfurthii apes clustered according to their collection sites of origin, as previously reported for SIVcpz [33],[42]. Inspection of Figures 6–8 revealed that this was generally not the case. Although each of the major SFVcpz clades exhibited considerable structure, the great majority of sublineages were comprised of viruses from multiple field sites. Moreover, geographic distance did not predict viral diversity. For example, viruses from the single DG field site in southern Cameroon exhibited as much pol-IN inter-strain diversity (0% to 5.8%) as did viruses collected hundreds of kilometers apart at the CP and LB/MB field sites (0% to 4.1%). Nonetheless, there were some notable exceptions. Significant geographic clustering was observed for (i) P. t. troglodytes viruses from the ME and GT field sites in the Central African Republic and the Republic of Congo (Figures 6, 7, 8); (ii) P. t. troglodytes viruses from the LP field site in Gabon (Figure 6); and (iii) P. t. schweinfurthii viruses from the BD field site in the Democratic Republic of Congo (Figure 6). Interestingly, all of these were associated with potential barriers to chimpanzee movement. GT and ME were the only P. t. troglodytes field sites east of the Sangha River; LP was separated from all other P. t. troglodytes sites by the Ogooue River; and BD was the only P. t. schweinfurthii collection site north of the Uele River (Figure 1). Thus, in addition to delineating the subspecies ranges, major rivers and other biogeographical barriers appear to also have influenced the dispersal of SFVcpz within existing subspecies ranges.

SFVcpz co-infection and recombination

GENECONV analyses and inspection of phylogenetic trees inferred for each independently amplified gene fragment (Figures 6, 7, 8) identified several SFVcpz strains with a strong signal of distinct evolutionary histories in different parts of their genome. For example, MF1269 was most closely related to other MF strains in gag and pol-RT regions (Figures 7 and 8), but clustered with MP and WE viruses in the pol-IN region (Figure 6). Such discordant branching patterns can be indicative of viral recombination but also of co-infection with divergent viruses [44]–[47]. Similarly, DG534, DP157, and CP470 were all found by GENECONV to be members of sequence pairs with globally significant (P<0.05) evidence of mosaic evolution. In the case of DG534 and DP157, highly significant putative recombination breakpoints were detected at or near the junction of independently amplified sequence fragments, strongly suggesting that some were due to the amplification of two or more variants from the same sample rather than intramolecular recombination per se.

To differentiate between these possibilities, we selected five such samples (DG534, DP157, MF1279, MF1269, CP470) for additional RNA extraction, RT-PCR and sequence analyses. Comparison of independently amplified gag, pol-IN, and pol-RT sequences yielded unequivocal evidence of SFVcpz co-infection in three of the five fecal samples. As shown in Figures 6 and 8, DP157 harbored two clearly distinct variants (DP157A and DP157B). The observation that these sequences fall into different, highly supported clades (P = 1.0) within the P. t. troglodytes SFVcpz radiation leaves little doubt that this chimpanzee was co-infected with more than one strain. The alternative, i.e., that the divergent sequences trace back to a single infection that diversified extensively within a single chimpanzee, is inconsistent with the fact that sequences from other apes are interspersed between the DP157 variants. Similarly, DG534 and MF1279 each exhibited (at least) two distinct SFVcpz strains as determined by phylogenetic analysis of independently amplified (and directly sequenced) pol-RT sequences (Figure 8). To follow up on these observations, we subjected two of these samples (DP157 and MF1279) to single genome amplification (SGA). This approach amplifies single viral templates, precludes Taq polymerase errors and in vitro recombination, and provides an accurate representation of the viral population in vivo [48]–[50]. Targeting both pol-IN (Figure 9A) and pol-RT regions (not shown), we generated SGA derived sequences for MF1279 and DP157. Phylogenetic analysis of these sequences confirmed co-infection of MF1279 with two SFVcpz strains, and revealed the presence of at least four distinct SFVcpz strains in DP157 (Figure 9A).