Isolates of Zaire ebolavirus from wild apes reveal genetic lineage and recombinants

Results

Between 2001 and 2006, we discovered 47 dead animals, of which 23 were great apes (18 gorillas, 5 chimpanzees), in the Gabon/RC region [supporting information (SI) Table 2]. Using a combination of diagnostic tests, we confirmed infection with ZEBOV for 13 gorillas, 3 chimpanzees, and 1 duiker. Because of the advanced decomposition of several carcasses, we only succeeded in amplifying and sequencing full-length GP gene sequences [2,031 nucleotides (nt)] from three gorillas and one chimpanzee, but shorter GP sequences (538 nt) could be obtained from another three gorillas. Phylogenetic analysis of the full GP revealed that these ape-derived sequences represent a new, distinct ZEBOV lineage (group B) that is 2–3% divergent from the lineage containing all GP sequences described to date (groups A and R) (Fig. 1A). Analysis of the partial GP data set containing the additional three sequences from gorillas confirmed this result (SI Fig. 4). In addition to the virus from ape carcasses, GP sequences generated from the two most recent human outbreaks in Mbandza (2003) and Etoumbi (2005) in RC also fell into the new group B (Fig. 1A).

Three full GP sequences that originated from ape carcasses located in the Lossi sanctuary in RC (GOR1_Lossi_dec02, GOR2_Lossi_dec02, and CH_Lossi_feb03) grouped together in the phylogeny, with the two sequences from gorillas clustering most closely (Fig. 1A). In December 2002, two gorilla carcasses were found together, a few meters from each other and in the same state of decomposition, indicating that they were infected and died around the same time. They probably belonged to the same group because a single group of gorillas was known to inhabit this part of the sanctuary. Despite this finding, GP sequences from these carcasses differed substantially at both the nucleotide and amino acid levels (11 nt and 5 aa differences) (SI Table 3), suggesting that they had not been infected by the same source.

We also obtained partial sequence data (1,448 nt) from the nucleoprotein (NP) gene from all human outbreaks since 2001, including the most recent ones in Mbandza and Etoumbi, as well as from two gorilla carcasses discovered in the Lossi and Odzala sanctuaries. Consistent with results for GP, sequences from apes and the two recent human outbreaks clustered together (group B) and were genetically distinct from viruses found in the 1976–1997 outbreaks (group A) (Fig. 2B). Similar to GP, both groups were ≈2% divergent. However, the phylogenetic position of viruses detected in October 2001 to January 2003 differed markedly for the two tree topologies: Whereas for GP, these viruses fell into group A and represented direct descendents of viruses seen during previous outbreaks (Fig. 1A), their NP sequences clustered in a more basal position in the tree together with group B viruses, from which they were not genetically distinct (Fig. 1B). This result was highly suggestive of a recombination event between group A and group B viruses, which both occupied stable positions in the GP and NP trees. To further examine the evidence for recombination, we statistically evaluated the observed level of incongruence between the two tree topologies. Shimodaira–Hasegawa tests identified the GP and NP topologies as significantly incongruent regardless of how the comparison was made (in all cases, P < 0.001) (see Materials and Methods).

A molecular clock-based analysis estimated that the root of all ZEBOV sequences just barely predated the first observed outbreak in 1976. Estimates based on GP sequences placed this most recent common ancestor at 1975 (highest posterior density interval: 1972–1976), whereas NP data produced a similar point estimate of 1971, but with a wider interval (1955–1976). Ancestors of groups B and R, as well as those of the viruses from two Lossi gorillas, were estimated to be of an even younger age (Table 1).

Plotting the month of emergence for all ZEBOV isolates since 2001 included in the genetic analysis against their longitudinal distribution revealed a positive relationship, consistent with outbreaks spreading east at a fairly regular rate, regardless of the genetic type of virus involved (R² = 0.859, P < 0.001) (Fig. 2).

Discussion

Up to now, our understanding of ZEBOV epidemiology and evolution has been exclusively based on molecular data from human infections and largely involved only a single viral gene (GP). By presenting new genetic data from infected ape carcasses and recent human outbreaks for the GP and NP genes, we are able to provide a more complete, but also more complex, picture of the history and conditions of ZEBOV emergence in central Africa.

The existence of additional ZEBOV lineages in the African rainforests, other than those known from previous human outbreaks, has been suspected for some time. Although IgG prevalence rates in some great ape (5) and human populations with no history of hemorrhagic fever (2, 13) suggested the presence of apathogenic lineages, our data demonstrate the existence of another ZEBOV lineage capable of causing significant mortality in human and wildlife populations. Spatially, this new genetic group has so far only been found in a small area that shows almost no overlap with other groups, but that is situated immediately to the east of the area affected by the group R viruses a few years earlier (Fig. 3). In fact, ignoring the molecular data and considering solely the temporal and spatial sequence of outbreaks between 2001 and 2005, the pattern appears highly consistent with ZEBOV continuously spreading eastward (Fig. 2). However, our new genetic data reveal that the more recent outbreaks are not caused by genetic descendents of previously emerging viruses, as seen in previous data, but instead involve a genetically different type of ZEBOV. Thus, because outbreaks do not appear to be epidemiologically linked in a consistent fashion, these findings challenge the notion of ZEBOV emergence in recent years being driven by a single viral lineage spreading through the affected area of Gabon/RC (9).

A key aspect of ZEBOV epidemiology that is still poorly understood is the role and extent of animal–animal transmission among susceptible species like gorillas and chimpanzees (8). Recent data from habituated gorillas in the Lossi sanctuary that experienced mass mortality due to ZEBOV strongly suggest efficient transmission within family groups, as well as repeated spread to neighboring groups (6). If these gorillas indeed acquired infection predominantly from contact with infected conspecifics, diseased individuals from one group or locality would be expected to harbor highly related viruses. For example, several studies have found no genetic changes during chains of transmission involving human–human passage of ebolavirus for several weeks or months (14, 15). In contrast, we found that the two gorillas from Lossi, which were found next to each other in December 2002 and were likely infected and died around the same time, yielded viruses for which GP gene sequences differed at 11 sites (SI Table 3). Our molecular clock-based analysis indicates that this level of divergence implies independent evolution for a minimum of 8 months (Table 1). Thus, it seems most likely that the two gorillas became infected from different sources. Both sequences tightly grouped with each other and with a third Lossi sequence from a chimpanzee, still suggesting that all three animals were infected by the same local ZEBOV variant (Fig. 1A). Whether divergence among sequences reflects variability of a local virus in its reservoir or whether these changes accumulated during prolonged animal–animal transmission remains unresolved.

Results discussed so far point to a multiple independent emergence of ZEBOV in the Gabon/RC border area over the last decade. However, in line with previous studies (9, 16), we also continued to find low divergence among viral sequences, implying a recent common evolutionary origin for all ZEBOV lineages discovered so far. Despite the new group B viruses being clearly distinct from the longer known group A viruses, the two lineages are estimated to have split only in the early 1970s and thus just before the first recognized outbreak of ZEBOV in 1976, which was in the Democratic Republic of the Congo, 2,000 km from the Gabon/RC region. Similarly, the common ancestor of all group B viruses is estimated to date back no further than 1998 (Table 1). These dates are clearly at odds with the virus being a long-term resident in the affected area while being maintained by a widespread reservoir, mostly likely some fruit bat species: Hypsignathus monstrosus, Epomops franquetti, and Myonycteris torquata (1, 10).

Furthermore, it remains suspicious that groups A and B viruses emerged in such direct temporal and spatial succession and that the distribution of outbreaks from either type exhibited the same pattern consistent with eastward spread (Fig. 2). This finding suggests the existence of some underlying ecological factor that connects all outbreaks regardless of the genetic type of virus involved. Movement and contact processes among animals involving either reservoir or susceptible species, or both, may be potential mechanisms for this (8, 9).

Our comparative phylogenetic analysis of two viral genes produced a strong indication that a recombination event has occurred in the recent evolutionary history of ZEBOV (Fig. 1). All GP sequences from the putative recombinants are estimated to have shared a common ancestor in 1999 (Table 1), suggesting that a single incident gave rise to the recombinant lineage around this time. Where this event took place, in terms of both the geographic context and the host species, and how recombination affected viral fitness cannot be discerned at this point. However, it is interesting to note that all of the recombinant GP sequences contained a stop codon in amino acid position 481, which would be predicted to result in a 29% shorter translated protein product. Despite a truncated GP, the virus obviously remained viable, because members of this group were involved in a series of outbreaks. In future research, we hope to address the potential fitness benefits (or costs) of recombination in ZEBOV, as well as to describe the molecular mechanisms underlying it. Our finding represents the first evidence of a recombination event within the Filoviridae family (Ebola and Marburg viruses) and one of only a few examples of such an event in negative-strand RNA viruses (17). This result has important implications for vaccine development (18, 19) because it raises the possibility of recombination between a live attenuated vaccine strain and wild-type strains, a scenario that could have grave public health repercussions.

In conclusion, our results demonstrate that ZEBOV was likely introduced into susceptible wildlife and human populations repeatedly in the last 30 years, as evidenced by the emergence of different genetic lineages. What remains unclear is how often such introductions have occurred and to what extent transmission among susceptible animal hosts, such as great apes, may have subsequently contributed to the spatial propagation of outbreaks. Similarly, it remains to be resolved whether the temporal-spatial patterns of ZEBOV emergence, which are in many ways suggestive of a spreading process, could be the result of transmission processes in the reservoir species or whether other factors could have generated such patterns. In addition, our study reveals an unanticipated role of recombination in ZEBOV evolution. Besides raising a number of important questions about the biology of the virus, recombination is likely to further complicate efforts to understand the history of ZEBOV emergence by using molecular epidemiological approaches.

Materials and Methods

Supplementary Material