In this report we describe the laboratory confirmation of the first known marburgvirus outbreak in West Africa, along with an extensive comparative genomic analysis of marburgviruses. Virus associated with previous Marburg HF outbreaks had all originated in East Africa, and up to 21% virus genetic diversity had been found based on analysis of partial genome sequences. Consequently, the discovery of Marburg HF in Angola was a surprise, as was the finding that the Angolan marburgvirus was not more distinct, differing from the main group of East African marburgviruses by only ~7%. Presumably the absence of a greater genetic difference reflects virus natural host reservoir similarities between the East and West African locations. However, due to the resource-poor nature of the affected area, an index case was never identified, and thus the importation of the virus from East Africa, while unlikely, cannot be ruled out. By the same token, the large virus genetic difference seen between the Ravn/09DRC lineage and the other marburgviruses may represent reservoir host differences, such that viruses within this lineage may be associated with a different host species or subspecies than the other marburgviruses. Interestingly, earlier ecological niche modeling approaches had predicted that parts of northern and eastern Angola were within the range for potential filovirus reservoirs (39
The natural reservoir for marburgvirus remains unknown, but marburgvirus emergence in Angola will likely extend the scope of the reservoir search beyond East Africa. As recently reported (5
), the marburgvirus outbreak in Durba, DRC, was closely associated with illegal mining activity. Epidemiological data linked over 70% of the cases with mines or caves, suggesting that the natural reservoir could well be associated with such environments. With the Angola outbreak, difficulties in surveillance and contact tracing, combined with the delay in the identification of the outbreak, led to poor epidemiological linkage of marburgvirus cases and ultimately to a lack of success in identifying a point source or mounting any ecological study. Filovirus outbreaks in general are relatively rare events. A recent report has suggested that bat species with rather wide geographic distributions may be a potential reservoir for ebolavirus (24
). If the natural reservoir of marburgvirus is similar to that of ebolavirus, the emergence of marburgvirus in western Africa should not be surprising, as the sites of multiple large ebolavirus outbreaks are less than 500 miles away, including areas which have experienced almost yearly activity over the last decade (25
The initial diagnosis was based on 9 of 12 clinical specimens testing positive by virus isolation, antigen, and/or PCR-based methods, including a newly designed Q-RT-PCR assay designed to detect the VP40 genome region of all known strains of marburgvirus. This Q-RT-PCR assay outperformed all other assays designed to detect acute marburgvirus infection, including the “gold standard” virus isolation assay. The assay was also sufficiently robust to allow deployment into a field setting in Angola. The extensive virus genomic analysis presented here also confirms that the VP40 gene target was an excellent choice for a broadly reactive marburgvirus detection assay, as it is the most conserved virus gene. These features suggest this should be a highly useful assay for detection of potential naturally occurring or deliberate Marburg HF outbreaks in the future.
Following the initial diagnosis, we determined the full-length genome sequence directly from clinical material from 11 patients representing the temporal and geographic distribution of the outbreak. Our purpose was to fully characterize the level of sequence variation generated during the course of a large marburgvirus outbreak with a high case fatality rate and to use this information to determine whether multiple marburgvirus introductions into the human population had occurred. To generate a context with which to compare the Angola marburgvirus sequence data, we determined the complete genome sequence of the Ravn marburgvirus, first isolated in 1987 in Kenya, as well as three additional, uncharacterized marburgviruses from Durba, DRC, the location of the largest outbreak on record prior to the one in Angola. Our data increased the number of the complete genome reference sequences from three to eight. Therefore, we used this expanded database to review the known genetic elements throughout the genome in an effort to assess their importance based upon the degree to which the features are maintained across the eight marburgvirus strains. One of the most striking features is the high degree to which nearly all the genetic elements previously identified were maintained. To some extent this was expected, but we predicted that perhaps the Ravn and 09DRC lineages, differing by greater than 21% from all other marburgviruses at the nucleotide level, would show more versatility as to how the viral proteins could potentially carry out their functions. Diversity within marburgviruses is still, even with this more complete data set, considerably less than that seen with the ebolaviruses, to the point that the Ravn/09DRC lineage would not be considered a different species. In fact, the three most conserved genes, VP40, VP24, and L, among the major species of ebolavirus (Zaire, Sudan, and Reston) show 20 to 25% protein diversity (44
). In contrast, those same genes within the marburgvirus lineages shown here have, respectively, 1.6, 4.3, and 12.3% maximum diversity.
Given the general error-prone nature of RNA virus polymerases, the fact that there were comparatively few changes observed among the 11 Angolan marburgvirus isolates was somewhat surprising. In fact, four of the complete genome sequences (19,114 nucleotides in length) from specimens collected over a month and a half and representing at least two to three human to human transmission cycles were 100% identical. In the absence of precise knowledge of transmission events from the natural reservoir, possibly bats (24
), this analysis attempts to answer the question of what level of sequence divergence constitutes a distinct lineage versus the degree of sequence variation to be expected in a large outbreak of single origination. To answer this question, we felt it important to compare the diversity seen in the Angola outbreak with that in the Durba, DRC, outbreak. Among the sequences from the Angola outbreak, the representative from the Songo municipality (specimen 0754) had 11 nucleotide differences from the reference isolate (0.07% variation). However, this maximum level of variation is 1/10 of the minimum diversity seen between the two closest lineages within the Durba outbreak, 05DRC and 07DRC, respectively (0.8% variation). Based on this comparison, we do not consider the Songo lineage to be indicative of a second introduction, although such a possibility cannot be ruled out. For the Angola marburgvirus outbreak, it would be interesting to measure the rate of accumulation of nucleotide changes over time, but unfortunately, the epidemiological links between the 11 patients could not be established and there is no Angola marburgvirus isolate from the presumed beginning of the outbreak in October 2004.
Genome plasticity and rapid evolution in response to positive selection are general features of RNA viruses which possess error-prone polymerases and lack proofreading mechanisms (12
). Such features have been well documented in recent outbreaks of emergent RNA viruses, such as severe acute respiratory syndrome coronavirus and human immunodeficiency virus (HIV), in which genetic differences have been shown to accumulate during human-to-human transmission events and, in some instances, within the same tissues of a single host (27
). Severe acute respiratory syndrome coronavirus has recently been calculated to accumulate approximately two mutations per genome per human transmission event (50
). Why, then, was so little variation found within this marburgvirus outbreak? The answer likely involves at least two related components, namely, (i) the time of progression within the host from initial infection to disease outcome and (ii) the host immune response to marburgvirus infection. With HIV, highly diverse quasispecies have been shown to develop from homogeneous populations in response to the development of host neutralizing antibodies and cytotoxic T-lymphocyte responses (6
). Yet, in the later stages of HIV infection towards the development of AIDS, the immune system is no longer able to exert strong selective pressure on the replicating quasispecies and, as a result, HIV evolution dramatically slows (11
). For marburgvirus, like Zaire ebolavirus, disease progression is so rapid that most individuals die before an effective immune response can be mounted (1
). For ebolavirus the median time to death after onset of symptoms is only 8 to 9 days (22
). In addition to the rapid course of infection, ebolavirus actively suppresses the immune system by (i) directly antagonizing the host interferon system through an uncharacterized mechanism involving the virus VP35 protein (4
) and (ii) preferentially infecting dendritic cells and macrophages as sites of primary infection, thus diminishing host cell populations that are critical for the establishment of innate and adaptive immune responses (16
). These features together allow for explosive virus growth within the host in a manner virtually unchecked by the immune system. The lack of IgM and IgG responses shown in Table is consistent with this view. The lack of genetic diversity within the Angola marburgvirus outbreak is therefore not surprising and is in fact consistent with the absence of genetic differences seen among genome fragments of viruses analyzed in previous Zaire and Sudan ebolavirus outbreaks (42
A trial sampling procedure that emerged during the outbreak was the use of oral swabs for sampling suspected Marburg VHF corpses. RNA extracted from oral swabs sometimes gave low CT
values in the Q-RT-PCR assay (Table and data not shown), indicating the potential for high viral loads in these secretions. Unfortunately, definitive interpretations of CT
values from oral swabs are difficult, since the swabbing technique and the volumes recovered are inherently variable. In general, very high viral loads are seen in sera of individuals infected with filoviruses. For instance, during human infections with Sudan ebolavirus, viral loads in patient serum can vary from 105
RNA copies/ml at the time of fever onset to greater than 1010
RNA copies/ml at the time of death (48
). However, the timing and extent to which this range of viral load is reflected in human oral/nasal secretions are currently unknown. Of particular concern is the use of this sampling technique as a means of establishing a particular etiology of a VHF outbreak when multiple agents must be considered and tested, many of which may not ever produce viral loads that can be detected by oral swabs. Blood-based sampling as a reliable source of virus for serology-, antigen-, and PCR-based detection assays is well documented and should remain the method of choice for sampling suspected VHF patients.