Ebola virus: from discovery to vaccine

Abstract

Ebola virus, being highly pathogenic for humans and non-human primates and the subject of former weapons programmes, is now one of the most feared pathogens worldwide. In addition, the lack of pre- and post-exposure interventions makes the development of rapid diagnostics, new antiviral agents and protective vaccines a priority for many nations. Further insight into the ecology, immunology and pathogenesis of Ebola virus will promote the delivery of these urgently required tools.

Main

Ebola virus (EBOV) was discovered in 1976, but remained largely unknown until 1995, when it re-emerged in Kikwit, Democratic Republic of the Congo (formerly Zaire). In this article, we describe the discovery of the virus and the events that have led to recent attempts at vaccine development.

Discovery of Ebola virus

First appearance. The story begins in 1976 in Central Africa. In June and July, the first cases of a haemorrhagic disease were reported from Nzara, a small town in the Western Equatorial Province of southern Sudan, bordering the African rain forest (Fig. 1). The outbreak started with the infection of individuals from a single cotton factory (index cases) and quickly spread to the close relatives of these individuals. The epidemic was augmented by the exportation of cases to neighbouring areas, and high levels of transmission occurred in the hospital of Maridi — a teaching centre for student nurses. The outbreak lasted until November, during which time approximately 15 generations of person-to-person transmission occurred. The total number of infected individuals was 284, with 151 deaths^1,2. By the end of August, a second epidemic started in the equatorial rain forest of northern Zaire, now Democratic Republic of the Congo (Fig. 1). The presumed index case came to the Yambuku Mission hospital for treatment of acute malaria. However, it remains unclear whether this individual was the source of the epidemic or became infected in the hospital. Most subsequent cases had contact with patients, but for more than 25% of cases the only known risk factor that was elucidated was receipt of injections at the hospital. Although transmission was focused in the outpatient clinics of the hospital, there was subsequent dissemination in the surrounding villages to people who were caring for sick relatives or attending childbirth, or through other forms of close contact. The case fatality rate in this outbreak was 88% (318 cases, of which 280 died), much higher than with the Sudan outbreak (53%)^2,3. Virus was isolated from patients of both outbreaks and named Ebola virus (EBOV) after a small river in northwestern Zaire. For several years, a direct link between the two EBOV epidemics was assumed. Later, it was discovered that the outbreaks were caused by two distinct species of filoviruses, today known as Sudan ebolavirus (SEBOV) and Zaire ebolavirus (ZEBOV).

Figure 1: Geographical locations of Ebola-virus outbreaks.

Outbreaks of Ebola virus (EBOV) were reported from one western and several central African countries. Case numbers and lethality are indicated. The stars indicate countries where EBOV-specific antibodies have been reported in parts of the population. The reports are mainly based on antibody detection by indirect immunofluorescence assay (IFA) (data are adapted from Ref. 91).

Notably, the first filovirus to be discovered was Marburg virus (MARV), which was isolated during an outbreak of haemorrhagic fever in Europe in 1967 (Ref. 4). Until the recent outbreak in the northern part of the Democratic Republic of the Congo, there have been only a few isolated cases of MARV haemorrhagic fever in Africa. This outbreak has been the largest so far with more than 100 cases, high mortality and several distinct introductions.

Emergence/re-emergence of Ebola viruses. EBOV surprised everyone when it emerged in the United States in 1989. The virus was detected in Cynomolgus monkeys (Macaca fascicularis), which were imported from the Philippines into a primate facility in Reston, Virginia⁵. Despite initial arguments, no clear link to Africa could be established, so the presumption prevails that this new virus, which is a distinct species known as Reston ebolavirus (REBOV), could be of Asian origin. The 1989 monkey outbreak spread through affected rooms by droplet contact with adjacent cages or to distant cages and to different rooms by larger droplets and/or small-particle aerosols. The potential airborne route of transmission and the presumptive lower human pathogenicity (no disease was found in infected workers) remain interesting features that are associated with the REBOV species.

Since the first description of EBOV, the virus has re-emerged several times in central Africa, which is considered to be endemic for EBOV (Fig. 1). In 1992 and 1994, two episodes of mortality were recorded among a troop of chimpanzees in the Tai National Park in the western Ivory Coast. Several of the dead animals showed signs of haemorrhages, and a necropsy was carried out on one of the animals in the field. A 34-year-old woman developed fever, headache and myalgia eight days after carrying out the necropsy. Five days later, she developed a syndrome that was similar to that described for surviving EBOV-infected patients. A distinct EBOV species, Ivory Coast ebolavirus (ICEBOV), was confirmed as the cause of this disease⁶ (Fig. 1).

EBOV became a real threat to public health when it re-emerged in Kikwit, Democratic Republic of the Congo, in 1995. This outbreak received much attention from the media and made EBOV one of the most feared infectious agents worldwide. Subsequent outbreaks in Gabon and Uganda, as well as the present outbreak in Congo-Brazaville, which has affected many great apes that have a role in transmission to humans⁷, keep this image alive and also emphasize the importance of EBOV infections for local and international public health. Since the tragic events of 11 September 2001, another dimension was added — bioterrorism. Active weapon-development programmes to produce large quantities of filoviruses (MARV and EBOV) were carried out in the past. Today, EBOV is not only feared as one of the most pathogenic human agents, but also as an agent that poses a potential bioterrorism threat⁸.Timeline

Key events

Molecular characterization

Originally MARV and EBOV were classified in the family Rhabdoviridae, but in 1982, the family Filoviridae was created on the basis of the unique morphological, morphogenic, physiochemical and biological features (Fig. 2 and Box 1). The molecular characterization of EBOV began at the end of the 1980s. Despite all of the attention on EBOV, surprisingly, the first complete genome sequence of a filovirus was published for MARV. Today, full-length sequences are available for two MARV isolates, the Mayinga strain of ZEBOV, as well as the Philippine and Pennsylvania strains of REBOV (Fig. 2).

Figure 2: Filovirus particles.

a | Surface structure of a particle. Purified Marburg virus (MARV) particles were spotted on poly-L lysine-coated glass coverslips, air-dried and analysed by atomic force microsopy. The image shows a filovirus particle of about 900 nm in length and 80 nm in diameter (Image courtesy of H. Schillers, Department of Physiology, University of Muenster, Germany with permission from Schattauer from Ref. 92 © (2003)). b | Schematic illustration of a particle. Four proteins are involved in the formation of the ribonucleoprotein complex (RNP): polymerase or large (L) protein, nucleoprotein (NP), virion structural protein 30 (VP30) and VP35. The glycoprotein (GP) is a type I transmembrane protein and is anchored with the carboxy-terminal part in the virion membrane. Homotrimers of GP form the spikes on the surface of the virion. VP40 and VP24 are membrane-associated proteins. c | A schematic illustration of the Ebola virus (EBOV) genome. The genome consists of a single, negative-stranded, linear RNA molecule. d | Generation of EBOV particles from cloned complementary DNA. The system is based on the co-transfection of six different plasmids: five expression plasmids under the control of the chicken β-actin promoter, encoding the four virus proteins that are required for transcription and replication (NP, VP30, VP35 and L) of the EBOV genome and the bacteriophage T7 polymerase, and a plasmid under the transcriptional control of the T7 promoter for transcribing the full-length EBOV genome (for details, see Ref. 10).

A scientific breakthrough in the field was made with the development of the infectious clone for EBOV^9,10 (Fig. 2). This success has to be attributed to the knowledge that was gained from work done with artificial minigenome systems^11,12. The infectious clone system was used to create an editing site mutant that led to the overexpression of transmembrane glycoprotein (GP) and the loss of expression of soluble glycoprotein (sGP)⁹. The recombinant virus had an increased cytopathogenic effect, but reduced virus production, indicating that RNA editing is required to reduce the cytotoxicity of transmembrane GP^9,13. A slightly different, and independently established, system was used to study the role of proteolytic cleavage for the activation of transmembrane GP^10,14. A recombinant virus that expresses non-cleaved GP molecules on the surface was isolated and shown to be infectious in tissue culture, indicating that proteolytic cleavage is not a requirement for infectivity¹⁰. The infectious clone system will be instrumental for future studies on protein function, genome transcription and replication, pathogenesis and the development of therapeutic and prophylactic interventions.

Concepts in pathogenesis

Although important scientific achievements have been made in the past, our knowledge of the pathogenic mechanisms of EBOV and the host immune response is still limited. EBOV is known to cause the most severe form of haemorrhagic disease in both humans and non-human primates, and shows the highest case-fatality rates for viral haemorrhagic fevers (reported up to 88%). The disease is caused by marked replication of virus together with immune and vascular dysregulation. In this way, it is as much an immune syndrome as a virus-induced vascular disease. As with some other viral haemorrhagic fevers, infection with EBOV is associated with fluid-distribution problems, hypotension, coagulation disorders and bleeding, which finally result in the sudden onset of severe shock. EBOV haemorrhagic fever can, therefore, be compared to a syndrome that is provoked by systemic treatment with cytokines or induced by an endotoxin¹⁵.

Impaired immune response. In humans and monkeys, there is marked disruption of the parafollicular regions in the spleen and lymph nodes, and replication of filoviruses in mononuclear phagocytic cells (cytolytic in vitro) has been shown^{16,17,18,19,20,21}. Widespread presence of bystander apoptosis of lymphocytes during infection with EBOV has been noted in non-human primates however, there is, as yet, little information on the role of apoptosis, in general, in the pathogenesis of EBOV, other than an observed concomitant loss of CD8⁺ T cells and plasma cells from the immune system¹⁸.

There is also evidence that filoviruses interfere with innate immunity, which normally has an important role in the control of virus replication, allowing the adaptive immune system to clear the virus. It has been shown in infected humans that there are clear differences in the expression of cytokines between fatal cases and patients who survive^22,23,24,25. In particular, the presence of interleukin-1β (IL-1β) and elevated concentrations of IL-6 in the plasma during the symptomatic phase have been indicated as markers of non-fatal infection, whereas the release of IL-10 and high levels of neopterin and IL-1 receptor A (IL-1RA) in the plasma, as soon as a few days after the onset of disease, is indicative of a fatal outcome (Fig. 3). Studies with ZEBOV have shown that the virus blocks double-stranded RNA (dsRNA)- and virus-mediated induction of interferon (IFN)-responsive promoters and the IFN-β promoter. In particular, viral protein 35 (VP35) was shown to function as an inhibitor of the type I IFN response and might, therefore, be an important feature in the pathogenicity of EBOV^26,27. The importance of the type I IFN response can also be shown in the mouse model²⁸. Although the mouse-adapted ZEBOV strain seems to have been selected during serial passage for its ability to block the production of type I IFNs, knockout mice that lack an effective type I IFN response are susceptible to lethal infection by several non-adapted EBOV and MARV strains²⁹. Recently, it was also shown that infected monocyte-derived dendritic cells were impaired in the secretion of pro-inflammatory cytokines, the production of co-stimulatory molecules and the stimulation of T cells³⁰.

Figure 3: Pathogenesis model.

The concept of the pathogenesis of filovirus haemorrhagic fever is illustrated. Viruses that enter the body through lymphatic and/or blood vessels get direct access to sessile monocytes/macrophages that become activated independently of virus replication. a | Infected cells might extravasate into tissues to infect other cells, such as hepatocytes, and induce focal necrosis. b | The production of cytokines by monocytes/macrophages can either promote or inhibit the immune response. Pro-inflammatory cytokines, such as interferon-γ (IFN-γ; additionally produced by cytotoxic T cells²²) and tumour-necrosis factor (TNF), can induce the activation of endothelial cells and increase vascular leakage. Direct infection of endothelial cells occurs, but probably has a minor role in the pathogenesis. The role of soluble glycoproteins is not based on experimental data (see also Refs 36,92).

For EBOV, it has been reported that sGP interacts with the host immune response by binding to neutrophils through CD16b — the neutrophil-specific form of the Fc-γ receptor III (FcγRIII) — and, subsequently, inhibiting the early activation of these cells^31,32. However, the direct binding of sGP to neutrophils has been challenged by others³³. Relatively high levels of sGP, the larger cleavage fragment GP1 and soluble GP1–GP2 (Box 1) are released into the medium of filovirus-infected cells, and sGP has been detected in patient serum. These findings led to the hypothesis that soluble glycoproteins, especially sGP, might effectively bind antibodies that might otherwise be protective^34,35. It has also been hypothesized that these soluble glycoproteins might function as mediators in the activation of mononuclear phagocytic cells and endothelial cells^21,36 (Fig. 3). In addition, filovirus transmembrane GP has a carboxy-terminal region that resembles a putative immunosuppressive domain that is also found in glycoproteins encoded by retroviruses^37,38, but it is not known whether this domain actually contributes to immune suppression.

Determinants of cytotoxicity. Infection with filovirus leads to a moderate cytopathic effect in susceptible tissue culture and host target cells. However, the mechanism of cell destruction is unknown. It is possible that either the marked production and accumulation of virus proteins or the maturation of virus particles at the plasma membrane are involved in this process. Alternatively, a virus protein might have specific cytotoxic potential. Several groups have shown cell alteration after the expression of EBOV-encoded GP^13,39,40. In one study, it was reported that a serine-threonine-rich mucin-like domain of GP1 mediated cytotoxicity in human embryonic kidney 293T cells and endothelial cells¹³. A second study showed the detachment of 293T cells after the expression of EBOV-encoded, but not MARV-encoded, GP. Cell detachment, in this case, occurred without cell death. This was largely attributed to a domain in the extracellular region of GP2 and seemed to involve a phosphorylation-dependent signal cascade³⁹. The ectodomain of GP and its anchorage to the cell membrane are required for GP-induced morphological changes⁴⁰. Using a reverse genetics system, it was also shown that the cytotoxicity of EBOV depends on the level of expression of GP. Overexpression of GP leads to the early detachment and cytotoxicity of infected cells⁹. These data show that the expression of GP, which is controlled by RNA editing — a mechanism that is not required for virus replication — is evolutionarily linked to the need to control cytotoxicity.

Impaired endothelial function. The disturbance of the blood–tissue barrier, which is controlled mainly by endothelial cells, is another important factor in pathogenesis. The endothelium seems to be affected in two ways: directly by infection with filoviruses, leading to activation and eventual cytopathogenic replication, and indirectly by a mediator-induced inflammatory response (Fig. 3). Such mediators seem to originate from filovirus-activated cells of the mononuclear phagocytic system, which are the main target cells of the virus^{15,19,20,21,23,41}. Increased levels of cytokines, such as tumour-necrosis factor (TNF) and IFN-γ, were detected in the serum of infected individuals²³. In particular, increased levels of IFN-γ were associated with fatal cases^22,23. In vitro, it was shown that the virus-induced cytokine release led to the activation of the endothelium, as shown by a breakdown of the endothelial barrier function¹⁹ (Fig. 3). Although the molecular mechanisms are largely unknown, experimental data have provided evidence for changes in the protein organization of the endothelial-cell junctions, in particular the cadherin–catenin complex of the vascular endothelium¹⁵. This might explain the imbalance of fluid between the intravascular and extravascular tissue spaces that is observed in patients. In vitro, the increased permeability could be blocked by antibodies that neutralize TNF¹⁹, indicating a crucial role for this pro-inflammatory cytokine in virus-induced shock. It has been shown by others that both TNF and IFN-γ, can alter the endothelial barrier function in tissue-culture models and in vivo^19,42. Therefore, infection with filovirus might lead to an uncontrolled release of cytokines, similar to that observed in lipopolysaccharide (LPS)-induced shock during infection with Gram-negative bacteria. Interestingly, recently reported preliminary data indicated that there are differences between filovirus- and LPS-induced cytokine release (U. Ströher et al., personal communication).

Present data indicate that the activation of endothelial and mononuclear phagocytotic cells could be triggered either by infection with virus or by the binding of soluble viral or cellular factors^19,21 (Fig. 3). Several soluble glycoproteins are expressed and secreted or released by alternative mechanisms during infection with filoviruses (Box 1). Comparison of the different strains of filovirus and virus variants raises questions that concern the relative roles of these proteins in pathogenesis. MARV, which causes comparable disease in primates and humans, releases GP1 in similar amounts to EBOV, but does not express sGP due to a different organization of its gene encoding GP³⁶. Similarly, a ZEBOV variant that expresses only small amounts of sGP is highly pathogenic in animal models⁴³, whereas the less pathogenic strain REBOV produces high levels of sGP³⁴. This argues against a general role of sGP in the pathogenesis of filoviruses and might point to a potential role of secreted GP1 and/or the soluble ectodomain of GP1–GP2. This might highlight the importance of cleavage of GP in pathogenesis, as the production of soluble GP1 depends on the efficacy of proteolytic cleavage^14,36. Interestingly, proteolytic cleavage is not required for infectivity of the virus¹⁰. However, all of this does not exclude a role for sGP as a biologically active protein in infection.

Clinical and laboratory data also indicate disturbances in haemostasis during infection with filoviruses. Although thrombocytopaenia is observed in severe infections of humans and non-human primates, studies on the role of disseminated intravascular coagulation (DIC), consumption coagulopathy, and platelet and endothelial dysfunctions are either missing, controversial and/or incomplete⁴⁴ (Fig. 3). During infection with filoviruses, DIC can be regularly observed in non-human primates and has been described in humans, but is less important in the mouse model of EBOV infection^28,45. In the non-human primate model, the endothelium remains relatively intact morphologically, even at the end stages of the disease, but, an increase in permeability was observed. Recent preliminary in vivo data (T. Geisbert et al., personal communication) indicated that the coagulation abnormalities might be triggered by the release of cytokines from infected mononuclear phagocytotic cells, as discussed earlier, rather than destruction of the endothelium itself.

Immunotherapy and vaccines

Immune response and immunotherapy. Despite being clearly immunosuppressive, there is good evidence for protective immunity during EBOV haemorrhagic fever. However, the mechanisms of recovery from infection are not well understood. In contrast to survivors and asymptomatic cases that show IgM and IgG antibody responses to virus antigens, fatal infections usually end with high viraemia and little evidence of a humoral immune response^22,24,46. If high-quality neutralizing antibodies could be given to patients at the time of disease onset, then this might limit the severity of disease, even if it fails to prevent it completely. Neutralizing antibodies specific for GP were shown to have protective and therapeutic properties in rodent models^47,48,49. An attempt to overcome the lack of serum available from convalescent patients was made by Russian investigators, who developed hyper-immune horse serum⁵⁰; this was shown to protect baboons from experimental infection with EBOV⁵¹. The treatment was also seen to be effective in guinea pigs, but failed to protect Cynomolgus monkeys from challenge with EBOV⁵². Furthermore, horse antibodies are not the best choice for a donor antisera, because horses produce a subclass of immunoglobulin (IgG_T) that is highly immunogenic in humans⁵³. Antiserum from other animals — for example, sheep and goats — is tolerated better by the human and primate immune systems because it does not contain IgG_T⁵⁴ and, therefore, might be a better source for therapeutic polyclonal antisera.

Protection in humans by convalescence sera has been anecdotally reported and convalescent plasma was transfused into patients during the EBOV outbreak in Kikwit⁵⁵. However, the efficacy of the treatment has been questioned because of the reduced severity of the disease at this point of the outbreak, even in non-treated patients. An evaluation of the efficacy of passive immunization in controlled clinical trials is missing at present. As with most RNA viruses, EBOV can rapidly mutate to produce antibody-escape mutants⁵⁶. This indicates that antibody therapy might require hyperimmune polyclonal serum or a panel of monoclonal antibodies of different epitope specificities to be successful. Despite the protective properties of antibodies, there has been some debate about the role of antibodies in enhancing EBOV infection and potentially exacerbating the disease^57,58,59. If this holds true for the infected host, strategies for vaccine development might have to be revisited.

Animal models. Non-human primates are the preferred animal model for human filovirus infection, as these animals are fatally infected with human virulent, non-adapted strains of EBOV and MARV. Numerous species of non-human primate have been used, such as baboons, African green monkeys, and Rhesus and Cynomolgus macaques, and the pathology of infected non-human primates is similar to that seen in humans^{60,61,62,63,64}. ZEBOV has been adapted for infection of both guinea pigs^65,66 and mice²⁸ by serial passage in each species. The pathogenesis of these adapted strains is similar to that in non-human primates and humans. Notably, in all cases, there is a conserved tropism for mononuclear phagocytic cells, pathological changes to the liver and spleen, high levels of viraemia and high mortality. The mouse-adapted strain seems to have reduced virulence in non-human primates, but is virulent in guinea pigs⁶⁷. Lymphocyte apoptosis was not reported to be a main feature of ZEBOV infection in mice or guinea pigs^28,66, but was a consistent feature of disease in humans²² and non-human primates¹⁸. Coagulopathy — in particular, haemorrhage and fibrin deposition — is not a main feature of disease caused by either of the rodent-adapted strains^66,67,68. This pathological feature is not consistently observed in primates and the occurrence in humans has not been well investigated because of the difficulty in obtaining samples⁶⁹. For vaccine research, the guinea pig is a poor immunological model with a restricted selection of specific reagents. However, mice provide an excellent model system with a wide range of reagents, as well as transgenic and mutant strains. The differences in pathology between model systems must be considered when planning vaccine studies, as should the observation that it is apparently harder to protect non-human primates than either species of rodent⁶⁹. Phase III clinical trials in humans will be difficult to conduct for any EBOV vaccine, because of the sporadic nature of the outbreaks and the potential difficulty in obtaining ethical approvals. In this case, licensing authorities will require that efficacy and safety have been shown in at least two animal species, and the most logical choices would be mice and non-human primates.

Vaccine development. There has been a long-standing discussion about the requirement for vaccines for EBOV and other rarely occurring viruses that induce haemorrhagic fever. However, in the past, the rare appearance of EBOV haemorrhagic fever and the remote locations of the outbreaks did not favour vaccine development. Development of effective vaccines requires industrial support and this did not seem to be feasible, knowing that there would not be a market for the vaccine. This view has changed with the existing threat from bioterrorism. Aside from better and more rapid diagnostics and antivirals, the development of a protective vaccine against 'List A agents', of which EBOV is one, today ranks at the top of the priority list for most countries. If a vaccine were available, present policies would probably include the vaccination of all at-risk medical personnel — such as first-responders, other public health workers, laboratory workers and military personnel.

Attempts to develop a vaccine for EBOV began shortly after the virus was discovered in 1976. Preparations of formalin-fixed or heat-inactivated virions were used in the early attempts to develop a vaccine against EBOV, using guinea pigs and non-human primates as the animal models. Unfortunately, the level of protection that was achieved was inconsistent. Guinea pigs were partially protected⁷⁰ and in one study, four out of five baboons survived challenge with EBOV after vaccination with an inactivated EBOV vaccine⁷¹. However, other studies indicated that inactivated EBOV did not induce sufficient immunity to protect hamadryl baboons reliably against a lethal challenge⁷².

Recently, investigators have mainly concentrated on the use of subunit vaccines that are based on a single or combinations of virus-encoded structural proteins to induce protective immunity against challenge with EBOV. GP, nucleoprotein (NP), and the virus structural proteins VP24, VP30, VP35 and VP40 have been tested as vaccine candidates using naked DNA, adenovirus, vaccinia virus or Venezuelan equine encephalitis virus (VEEV) replicons as delivery mechanisms^{73,74,75,76,77,78} (Table 1). Only EBOV GP is exposed on the cell surface of the virion and so seems to be the only target for neutralizing antibody. It is not surprising, therefore, that immunization with VEEV replicons that express the other virus proteins resulted in low endpoint titres (<1:20) in the 80% plaque reduction neutralization test (PRNT₈₀). Specific, non-neutralizing antibodies were produced in response to immunization with VP24, VP30, VP35, VP40 and NP, but it is not clear what role these antibodies have in protection. The EBOV GP–VEEV replicon resulted in antibody titres of between 10 and 320 using the PRNT₅₀, and protected 8 out of 10 guinea pigs and 18 out of 20 mice against challenge, and the EBOV NP–VEEV vector protected all 20 mice that were immunized⁷⁵.

Table 1 Comparison of Ebola virus vaccine candidates

Replicon immunization results in the production of virus antigen by host cells, leading to antigen presentation by dendritic cells in the context of both MHC class I and class II molecules. This should lead to the development of both cytotoxic and helper T-cell responses, in addition to the production of antibody, but unfortunately these T-cell responses were not measured. Consequently, when these VEEV replicon–EBOV GP/NP vaccine constructs failed to protect non-human primates from challenge with EBOV, it was impossible to determine whether a deficiency in antibody or cell-mediated immunity was responsible⁷⁸. Recombinant vaccinia viruses that express EBOV GP, NP, VP35, VP40 and VP24 (Refs 79,80) failed to protect guinea pigs, with the exception of the GP–vaccinia virus construct, which protected three out of five guinea pigs, but failed to protect Cynomolgus macaques⁷⁸. Once again, the antibody titres that were determined for the vaccinia virus constructs using enzyme-linked immunosorbent assay (ELISA) and PRNT₈₀ were low but, notably, T-cell proliferation and cytotoxic T lymphocyte (CTL) responses were not measured, resulting in the loss of relevant, and possibly crucial, immunological data.

There has been some success with DNA vaccination using genes encoding both GP and NP^73,74. EBOV GP DNA vaccination of guinea pigs resulted in variable protection, depending on the vaccine regimen, and premature sacrifice of the experimental animals makes interpretation of the challenge results problematic⁷⁴. Immune serum from the guinea pigs did not inhibit the growth of EBOV in vitro, and there was no passive protection conferred to naive animals after the transfer of serum. DNA immunization of mice with both EBOV GP and NP constructs resulted in some degree of protection from challenge and the development of CTL responses to both antigens. The most successful strategy so far involves using a DNA prime followed by an adenovirus boost⁷⁶. Cynomolgus monkeys were immunized twice with DNA encoding the GPs of ZEBOV, SEBOV and ICEBOV, and ZEBOV NP followed by a booster of adenovirus expressing ZEBOV GP five months after priming. The monkeys were challenged with six plaque-forming units of ZEBOV and all four animals survived after challenge. Antibody responses, T-cell proliferation and CTL responses indicated that antibody and T memory helper cells are essential for the protection, and that cell-mediated immunity, although possibly important, is not an absolute requirement. This is in agreement with observations that passive transfer of serum can be protective and might give some support to the concept of antibody therapy (discussed earlier).

The use of human adenoviruses as vaccine vectors is problematic because pre-existing immunity in the human population might eventually limit the efficacy of this approach. However, a study by Sullivan and colleagues⁷⁶ shows the first successful protection of non-human primates against challenge with EBOV, although this has been achieved previously for MARV using a GP-based alphavirus-replicon vaccine⁸¹. The study gives us insights into the mechanisms and indicates that appropriate live vaccine vectors or replication-deficient vectors might ultimately provide a successful human vaccine. If further enhancements in vaccine strategies are to be possible, the full spectrum of assays that are available to immunologists in the twenty-first century must be used in future vaccine studies. Particularly important would be the use of MHC class I tetramers to determine responses of CTLs to EBOV-encoded antigens, as well as the use of intracellular flow cytometry to determine cytokine responses to immunization and detailed study of the immunopathological responses to EBOV infection in unimmunized animals.

Finally, the recent development of EBOV-like particles⁸² might provide an interesting delivery system for a protective host immune response that seems to depend on both the cell-mediated effector mechanisms and the humoral immune response^69,78. Virus-like particles, which are generated by the expression of virus membrane proteins, might overcome the safety limitations that are associated with the use of attenuated filovirus particles or live vaccine vectors, as well as pre-existing immunity to vectors such as vaccinia or adenovirus.

Concluding remarks

The first cases of EBOV were reported from Sudan and Zaire in 1976, but the virus has only received real attention from scientists since 1995. Until recently, vaccine development was not considered to be a priority and despite several promising results, vaccine development is still in the experimental stages. The process of moving an experimental vaccine candidate into clinical trials is time-consuming and expensive. It has become clear that the rodent models (mouse and guinea pig), although important for the development of antivirals, therapeutics and vaccines, are not necessarily predictive for the efficacy of the same agents in non-human primates, which are the gold standard for predicting efficacy in humans. At present, there is no convincing evidence for any successful strategy in post-exposure prophylaxis. So, therapeutic antibodies might be the most promising short-term strategy. Despite some evidence for antibody-mediated enhancement of disease, therapeutic antibodies should be carried into the next stage involving the humanization of monoclonal antibodies, the production of human monoclonal antibodies and the evaluation of polyclonal and reconvalescent sera. A major drawback in the process of developing counter measures against EBOV, other viral haemorrhagic fevers and related diseases is the biocontainment that is required for animal work with these agents. Building new facilities is one way to respond and the political support is guaranteed in crisis situations. However, aside from maintaining facilities and long-term funding, the availability of well-trained personnel becomes a crucial issue.

Key events

Box 1 | Filoviruses

Filoviruses constitute a separate family in the order Mononegavirales. The family consists of the genera Marburgvirus (MARV) and Ebolavirus (EBOV); the genus Ebolavirus is further subdivided into four distinct species: Ivory Coast ebolavirus (ICEBOV), Reston ebolavirus (REBOV), Sudan ebolavirus (SEBOV) and Zaire ebolavirus (ZEBOV). Filoviruses consist of a single, negative-stranded, linear RNA genome that is non-infectious and does not contain a poly(A) tail. After entry into the cytoplasm of host cells, it is transcribed to generate polyadenylated subgenomic messenger RNA species. The genome has the following characteristic gene order: 3′ leader, nucleoprotein (NP), virion protein 35 (VP35), VP40, glycoprotein (GP), VP30, VP24, polymerase (L) protein and 5′ trailer (Fig. 2). Transcription and translation leads to the synthesis of seven structural polypeptides with presumed identical functions for all of the different filoviruses. Four proteins are associated with the virus genomic RNA in the ribonucleoprotein complex: NP, VP30, VP35 and the L protein. NP, VP35 and L are essential and sufficient for the transcription, as well as the replication, of MARV, whereas EBOV-specific transcription also depends on the presence of VP30. The three remaining structural proteins are associated with the membrane. GP is a type I transmembrane protein that forms the spikes on the virus particle. It is synthesized as a precursor (preGP) that is post-translationally cleaved by furin or a furin-like endoprotease into the disulphide-linked fragments GP1 and GP2. The homotrimeric GP1–GP2 functions in receptor binding and fusion and is the target for the neutralizing host immune response. Of the two membrane-associated non-glycosylated proteins (VP24 and VP40), VP40 functions as the matrix protein. The structure and function of VP24 has not yet been studied. A non-structural, soluble glycoprotein (sGP) is expressed as the primary product (non-edited mRNA) of the GP gene by EBOV, but not by MARV³⁶.

Ultrastructural studies indicate an association of virus particles with coated pits for the initiation of infection, indicating that filoviruses enter cells by receptor-mediated endocytosis. GP mediates receptor binding and subsequent fusion. The asialoglycoprotein receptor, folate receptor-α, integrins (especially the β1 group), DC-SIGN (dendritic-cell specific intercellular adhesion molecule-grabbing nonintegrin) and L-SIGN (liver/lymph node-SIGN) have all been described as potential attachment factors to promote infection with filoviruses^{40,83,84,85,86}. Uncoating is presumed to occur in a manner analogous to that of other negative sense RNA viruses. Transcription and genome replication take place in the cytoplasm and follow, in general, the models of Paramyxoviridae and Rhabdoviridae. Transcription starts at the conserved start site and polyadenylation occurs at a series of uridine residues in the stop site. The 5′-terminal non-coding sequences favour hairpin-like structures for all viral mRNA. Replication involves the synthesis of a full-length positive-stranded copy. During infection, nucleocapsids accumulate intracellularly and form intracytoplasmic inclusion bodies. Virions are released by budding through the plasma membrane³⁶.

For further reading on the history, epidemiology, molecular biology and pathogenesis of filoviruses, we refer the reader to several review articles and books (Refs 2,36,87–90).