The filoviruses (ebolaviruses and marburgviruses) have caused lethal human outbreaks since 1967 
, devastating primate epizootics in Africa 
, as well as several non-human primate outbreaks originating in the Philippines 
; they have also been implicated in a livestock epizootic in the Philippines 
. Filovirus disease in primates is generally a severe hemorrhagic fever syndrome that has no proven specific treatments 
. The total number of human cases of filoviral disease is near 2,500, spread over roughly 30 outbreaks and laboratory accidents 
. Filovirus disease is often lethal; the human fatality rate averaged over all known cases is in the vicinity of 70% 
. Because of the recurrently emergent and extremely serious nature of this disease, significant effort has been made to develop vaccines. However, a pan-filoviral vaccine, or even a clear characterization of what would be required to make one, remains elusive. Here we report theoretical vaccine designs using mosaic techniques first applied to the hyper-variable human immunodeficiency virus type 1 (HIV-1) 
and preliminary experimental results. While the techniques used here are very similar to those used for HIV-1 mosaic vaccine design, a pattern of repeated introductions of the filoviruses into humans (and primates generally) gives a crucial difference from HIV-1. HIV-1 shows great diversity within the pandemic, but that diversity has developed continuously, leaving intermediate isolates in its wake. In contrast, known filovirus diversity has episodically increased as new outbreaks are found to result from novel viruses, lacking intermediates. This crucial difference is reflected in the phylogeny of the viruses, discussed below. Using the mosaic technique we give the first characterization of the likely requirements for a broadly-protective filoviral vaccine; we also discuss potential limitations on pan-filoviral vaccines and report in vivo
murine antigenicity and protection.
In the family Filoviridae
there are currently six species recognized as causing disease in primates: Marburg marburgvirus, Zaire ebolavirus, Sudan ebolavirus, Reston ebolavirus, Taï Forest ebolavirus
(formerly Cote d'Ivoire ebolavirus
), and Bundibugyo ebolavirus
. The genetic diversity of these viruses is large. Protein sequence phylogeny provides a measure of the diversity of protein sequences that is relevant to designing recombinant mosaic proteins. Phylogenetic trees based on protein sequence alignments of filoviral glycoprotein (GP) and nucleoprotein (NP) and the equivalent HIV-1 M-group proteins Env and Gag (current mosaic vaccine targets) show that the diversity of proteins within the Ebolavirus genus is comparable to the protein diversity within the HIV-1 M-group (Figure S1
). Including Marburgvirus
sequences increases the protein phylogenetic diversity beyond what is observed in HIV-1 M-group proteins. The relatively sparse and long branches in the filovirus trees as compared with the HIV-1 trees are a marked difference in their phylogeny. While one can choose a vaccine target population with diversity comparable to HIV-1 (e.g. Ebolavirus
), there may be additional challenges arising from the notable differences in their phylogeny.
The human fatality rate after infection with the members of the five non-Reston filoviral species appears to be outbreak dependent, and in large outbreaks (>10 cases) ranges from 20–90%. Reston virus has not caused recognized human clinical cases, although there is serological evidence for human infection (i.e. subclinical infection). The human fatality rate from Taï Forest virus disease is unclear, as only a single human case is known. The remaining viruses (Zaire, Sudan, and Bundibugyo viruses, and Marburg virus) give human fatality rates ranging from 34% to 82%, averaged over all known cases 
. Fatality rates in wild non-human primates are not well documented, but epizootics are thought to be responsible for precipitate declines in primate populations in Africa 
. Even though fatality rates in large outbreaks of this serious disease are high by any reasonable measure, the strong variation in fatality rates between viral species and between outbreaks, indicates the importance of genetic variation for the course of the disease.
In the context of providing broad vaccine coverage, it is worth emphasizing the apparently idiosyncratic nature of the filoviruses. In recent decades, previously-known filoviruses have continued to re-emerge, sometimes in unexpected ways, as with “pig Ebola” 
in the Philippines, and new strains have also emerged 
, including a virus in a new genus and species, Lloviu cuevavirus
. This new virus, identified in European bats, is as phylogenetically distant from marburgviruses and ebolaviruses as they are from each other. The first filovirus discovered, Marburg virus, is still a phylogenetic outlier, even after the discovery of half a dozen other filovirus varieties. While a major reservoir in Africa is likely to be closely connected to African bats 
, known bat strains have not reproduced the full diversity of known filovirus strains, much less anticipated new ones (e.g. Bundibugyo virus). On the other hand, for the most geographically distant filovirus (Reston virus, traced to the Philippines), no natural reservoir has been found, and no plausible explanation has been given to tie Reston virus to a hypothetical origin in Africa, even though it is a member of the genus Ebolavirus
, an otherwise African taxonomic grouping. Clearly, the filovirus reservoir and the sequence diversity of the viruses are not fully understood. The implication is that a vaccine against the filoviruses should strive for good coverage of common epitopes from the maximum number of types and strains currently available, in the hope that future outbreaks will retain these elements, so the vaccine will still be effective when challenged by a novel strain in a new outbreak. However, the natural history of repeated filovirus introductions to primate populations leads to marked differences in the phylogeny in the form of long sparse branches (see Figure S1
), and is likely to have numerous implications for vaccines.
The large genetic variation between known filovirus species, the recurrent emergence of viruses from different species, and the failure to date to produce a filovirus vaccine that protects across all six species 
indicates that new strategies for vaccine development should be considered. One such strategy was developed by our group for HIV-1 
. By maximizing the vaccine's coverage of common potential T cell epitopes found throughout the viral population, the mosaic vaccine strategy is intended to optimize T-cell immunity. In recent years T-cell immunity has been found to be important and effective in combating many viral infections, including filovirus infections. Immune responses to vaccines can confer protection from the lethal effects of ebolaviruses in animal models 
, and cytotoxic T lymphocyte (CTL) responses have been specifically shown to confer protection 
. Furthermore, inter-strain CTL-mediated protection in non-human primates has also been demonstrated using nucleoprotein (NP) and glycoprotein (GP) as antigens 
. Promising cross-protection results for a recombinant adenovirus-based GP-rAd5 vaccine 
depend on CD8 CTL responses 
. Despite the difficulties associated with defining immune correlates of protection in humans for a disease that rapidly kills such a high fraction of those infected 
, there are indications that survival of symptomatic filovirus disease in a human Sudan virus outbreak was correlated with CTL-mediated immunity 
The mosaic vaccine design strategy distills the essential antigenic variation of a collection of sequences by computational recombination of potential CTL epitopes into a small set of chimeric intact protein sequences. The mosaic design algorithm favors common over rare peptide 9-mers and rejects all recombinants that include non-natural peptide 9-mers. The intent of mosaic design is to generate proteins that resemble natural proteins, and are expressed, folded, and processed naturally while maximizing coverage of natural epitope variation 
. In this way the mosaic proteins can be delivered in vaccines using the same strategies as natural proteins, and should be expressed and processed comparably. Because CTL epitopes are typically 9 amino acids long, 9-mer coverage is optimized, though other epitope lengths are also well covered 
So far all HIV mosaic proteins tested have been highly immunogenic, and, as anticipated, elicited responses that are more cross-reactive than responses elicited by natural-strain vaccines 
. Both CD8 and CD4 T cell responses were enhanced using mosaic vaccine inserts 
. Although the mosaic proteins were designed to optimize T cell epitope coverage, when used as immunogens they also elicited B cell responses that were comparable to or better than natural strain responses 
. HIV-1 epitopes from mosaic proteins are processed appropriately – T-cells specific for 13 out of 13 distinct commonly-targeted HIV epitopes isolated from 22 infected individuals were able to recognize human target cells expressing HIV mosaic proteins 
. The mosaic design approach was subsequently applied to Hepatitis C Virus (HCV) 
, where cocktails of mosaic proteins resulted in improved coverage of potential CTL epitopes from the highly diverse sequence population.
Here we make several comparisons between our mosaic designs and several sets of data. First, we computationally compare coverage of natural epitopes by cocktails of filovirus mosaic proteins with cocktails comprised of an equal number of best-natural sequences. The best-natural sequences are the combination of naturally occurring proteins that provides the best (9-mer) coverage. Second, we compute the coverage of known epitope-containing regions. This gives an indication whether the mosaic strategy retains known good epitopes. Third, we compare our results with the theoretical coverage numbers for a promising ebola GP protein rAd5 vaccine construct 
. Finally, we show that a mosaic-based ebolavirus vaccine immunogen conferred protective immunity in a mouse model against challenge by a mouse-adapted Zaire ebolavirus strain, protection comparable to protection conferred by a natural Zaire strain vaccine.