Rift Valley fever virus (RVFV) is a devastating mosquito-borne viral zoonotic disease that causes serious morbidity and mortality in both humans and livestock. In ruminants, RVF is characterized by substantial mortality of young animals (especially of lambs), fetal deformities and abortion (Flick and Bouloy, 2005, Gerdes, 2004, Swanepoel and Coetzer, 2003). In humans the disease is often associated with benign fever but can lead to more complicated cases such as retinal vasculitis, encephalitis, neurologic deficits, hepatic necrosis, or fatal hemorrhagic fever (Flick and Bouloy, 2005, Geisbert and Jahrling, 2004, Meegan, 1979). Interestingly, human case fatality rates increased significantly during the last several years. While historically less than 2% of infected individuals developed a fatal hemorrhagic fever, analysis from recent outbreaks (2007/2008) reveal a 20–30% fatality rate in humans (LaBeaud et al., 2008). However, differences in case definition, accuracy in disease surveillance methods and data gathering methodology likely impact these numbers.

RVFV is a member of the Bunyaviridae family, which includes more than 300 viruses grouped into five genera (Orthobunyavirus, Hantavirus, Nairovirus, Phlebovirus, and Tospovirus). Bunyaviruses are enveloped viruses with a tripartite, single-stranded RNA genome of negative and sometimes ambisense polarity (Elliott, 1996, Elliott et al., 1991, Schmaljohn and Hooper, 2001). The large (L) genomic RNA segment encodes the RNA-dependent RNA polymerase (L), the medium (M) segment the glycoprotein precursor, which is post-translationally processed into the two mature spike proteins (G1 and G2, or by the new convention: G_N and G_C), and in some viruses a non-structural protein (NS_M), while the small (S) segment encodes the nucleoprotein (N), and in some viruses a non-structural protein NS_S (Elliott, 1996, Schmaljohn and Hooper, 2001).

RVFV has traditionally caused recurrent outbreaks affecting humans and ruminants predominantly in Sub-Saharan Africa, but spread to Egypt in 1977 and to the Arabian Peninsula in 2000 (Al-Hazmi et al., 2003, Anonymous, 2000, Balkhy and Memish, 2003, Madani et al., 2003, Shoemaker et al., 2002). More recently, RVFV circulated in East Africa causing serious epidemics in Kenya, Tanzania, Somalia, Sudan and was reported in the Comoros Islands (LaBeaud et al., 2008, WHO, 2007) in 2007, and subsequently expanded to Madagascar and South Africa in 2008 [http://www.fao.org/docs/eims/upload//242253/EW_rvf_apr08.pdf].

RVFV is a prototype of emerging/re-emerging pathogens and is classified as a Category A High Priority Pathogen by the National Institute for Allergy and Infectious Diseases (NIAID) (http://www3.niaid.nih.gov/topics/BiodefenseRelated/Biodefense/research/CatA.htm), is on the Center for Disease Control (CDC) Bioterrorism Agents (http://www.bt.cdc.gov/agent/agentlist-category.asp#a) and is also classified as a Department of Health and Human Services (HHS), United States Department of Agriculture (USDA) overlap select agent (USDA, 2005).

The lack of prophylactic and therapeutic measures, the potential for human-to-human transmission, and the significant threat to livestock associated with RVFV make infection with this pathogen a serious public health concern not only in endemic, developing countries, but also in many non-endemic developed countries due to recent bioterror threats, and clearly illustrates the need for more RVFV vaccine research and development.

MP12, a highly attenuated (by 5-fluorouracil treatment in cell culture) human virus isolate of RVFV (Caplen et al., 1985, Vialat et al., 1997), has recently been tested in a phase II safety/efficacy clinical trial (ClinicalTrials.gov identifier: NCT00415051) to determine if it is safe to give to humans (results not yet published). MP-12 also has potential veterinary applications (Hunter et al., 2002).

A formalin-inactivated RVFV vaccine, TSI-GSD-200, has been developed; however, it is not licensed and not commercially available (Pittman et al., 1999). TSI-GSD-200 is only provided to veterinarians working in endemic areas, high containment laboratory workers and others at high risk for contracting RVFV (Pittman et al., 1999). Unfortunately, this vaccine is (i) expensive, (ii) difficult to produce, (iii) in short supply, (iv) requires larger dose relative to an attenuated vaccine and three initial inoculations followed by a 6-month booster (v) and requires continued annual boosters to maintain protective immunity (Frank-Peterside, 2000, Kark et al., 1982, Kark et al., 1985, Niklasson et al., 1985).

The use of virus-like particles (VLPs) is a promising approach for the development of a safe and efficient RVFV vaccine. Expression of structural proteins of many non-enveloped and enveloped viruses leads to the formation of VLPs (Garcea and Gissmann, 2004, Grgacic and Anderson, 2006a, Grgacic and Anderson, 2006b, Noad and Roy, 2003). Such VLPs frequently exhibit a morphology very similar to that of wild-type (wt) viruses (Johnson and Chiu, 2000). Since VLPs have a tropism similar to that of the wt virus and show comparable cellular uptake and intracellular trafficking, the formation of VLPs can be used to study virus assembly and morphogenesis, budding processes, genome packaging, receptor binding, and virus entry (Bos et al., 1997, Johnson and Chiu, 2000, Li et al., 2003, Licata et al., 2004, Overby et al., 2006, Schmitt et al., 2002, Ye et al., 2006). Especially for viruses classified as high containment agents, e.g., RVFV and Ebolavirus, the development of VLP systems are of practical use because subsequent work can be performed under lower biosafety conditions (Habjan et al., 2009, Naslund et al., 2009, Warfield et al., 2004, Watanabe et al., 2004). VLPs present viral antigens in a native conformation and are effectively recognized by the immune system (Grgacic, 2006; Grgacic and Anderson, 2006; Noad and Roy, 2003).

Many promising vaccine candidates based on VLPs are at various stages of development, including vaccine candidates for hepatitis B virus (HBV), human papillomavirus (HPV), Norwalk virus, human polyomavirus, Bluetongue virus, rotavirus, retroviruses, bunyaviruses and filoviruses (Garcea and Gissmann, 2004; Grgacic, 2006; Habjan et al., 2009, Naslund et al., 2009, Noad and Roy, 2003).

These promising attempts to generate VLP-based vaccines against many different animal and human pathogens encouraged us to evaluate RVF VLPs as vaccine candidates against RVFV. Here we describe the generation of chimeric RVF VLPs, a novel concept for bunyaviruses, the optimization of VLP production and their successful use as vaccine candidates. Vaccine efficacy was analyzed through immunological studies of vaccinated mice and in lethal challenge studies in two different rodent models. High protection rates and robust and long-lasting immune response of vaccinated animals demonstrate that chimeric RVF VLPs are a promising approach to generate safe and efficient RVFV vaccines.

The structural proteins of many viruses can assemble into VLPs (Grgacic and Anderson, 2006). VLPs are often described as being more efficacious in the activation of immune responses than conventional protein immunogens/subunit vaccines because their immunogenic protein components are displayed in a high density, more authentic conformation, often with intact biochemical functionality that is effectively immunogenic (Garcea and Gissmann, 2004; Grgacic, 2006; Noad and Roy, 2003). This is further enhanced by the particulate nature of VLPs that appears to be optimal for uptake by dendritic cells (Fifis et al., 2004). In addition, they are safer than inactivated and attenuated virus vaccines because they are usually free of viral genetic material and therefore are not encumbered by the possible safety-related drawbacks including reversion, recombination and re-assortment. At this stage, the ability to develop VLPs does not appear to be limited to any one type of virus or virus family, or by the complexity of the virus particle.

Mimicking the structure of virus particles allows the use of lower doses of antigen to elicit a similar protective response when compared to subunit vaccines (Noad and Roy, 2003). In addition to their ability to stimulate B-cell-mediated immune responses, VLPs have been shown to be highly effective at stimulating CD4 proliferative and cytotoxic T lymphocyte responses (Murata et al., 2003, Paliard et al., 2000, Schirmbeck et al., 1996). This feature of VLP-based vaccines likely plays a major role in their effectiveness as vaccines against viral diseases. The well-documented immunogenicity of VLPs is likely facilitated by their interaction with dendritic cells (Warfield et al., 2003).

Several VLP-based vaccines are in human clinical trials or are FDA-approved, including those for hepatitis B virus (Andre and Safary, 1987, McAleer et al., 1984, Sitrin et al., 1993), trivalent influenza H1N1, H3N2 and B vaccine (“FluBlok”) (Cox et al., 2008, Treanor et al., 2007), H5N1 “bird flu” (Perrone et al., 2009), human papillomavirus (2007, Giannini et al., 2006, Harper et al., 2004, Harro et al., 2001, Joura et al., 2007), human immunodeficiency virus (Young et al., 2006) and Norwalk virus (Tacket et al., 2003). Other VLP-based vaccines with very promising pre-clinical results include VLPs for the severe acute respiratory syndrome (SARS) coronavirus (Lokugamage et al., 2008), human polyomavirus (Goldmann et al., 1999), rotavirus (Ciarlet et al., 1998, El-Attar et al., 2009, Jiang et al., 1999), and Ebola and Marburg viruses (Swenson et al., 2008, Swenson et al., 2005, Warfield et al., 2003, Warfield et al., 2005b, Warfield et al., 2007). Additionally, VLP vaccines also have important agricultural applications, including promising vaccines for livestock diseases including bluetongue (Noad and Roy, 2003, Roy, 2000, Roy et al., 2009, Roy and Noad, 2008) and foot-and-mouth (Li et al., 2008, Remond et al., 2009).

Few studies have been undertaken to develop VLPs for bunyaviruses. Bunyamwera VLPs were generated by co-expression of G_N, G_C and NSm, in addition to a minigenome system (bunyamwera L, N and minigenome) in mammalian cells and were used to identify viral protein components required for virus assembly (Shi et al., 2009, Shi et al., 2006, Shi et al., 2007). Similarly, it has been shown that the expression of recombinant G_N and G_C glycoproteins of Uukuniemi (UUK) virus, a phlebovirus closely related to RVFV, leads to the assembly and budding of VLPs from transfected mammalian cells (Overby et al., 2007a, Overby et al., 2006, Overby et al., 2007a, Overby et al., 2007b). These VLPs are similar in structure to wt virus and are neutralized by UUK-specific antibodies. However, no immunological in vivo studies were performed to determine immunogenicity.

Generation of RVFV VLPs in insect cells has been demonstrated by Liu et al. (2008) using a single recombinant baculovirus that expresses the RVFV glycoproteins (G_N/G_C) and the N protein (Liu et al., 2008). Efficient generation of RVF VLPs in mammalian cells has been recently demonstrated by Habjan et al. (2009) using transfected DNA encoding the complete RVFV M segment as well as the RNA polymerase L, the nucleoprotein N and a GFP-expressing minigenome (Habjan et al., 2009, Naslund et al., 2009). Näslund et al. (2009) showed that these RVF VLPs can be used for vaccine studies. Three intraperitoneal injections of 1 × 10⁶ RVF VLPs in mice induces antibody titers from 1:300 to 1:900 against G_N and G_C proteins but does not result in the development of detectable N-specific antibodies. Importantly, these VLPs protect 11 of 12 vaccinated mice from lethal virus challenge (2.4 × 10⁴ pfu), whereas only 1 of 12 survived in the unvaccinated control group (Naslund et al., 2009).

The generation of chimeric RVF VLPs and its successful use as a vaccine candidate is unique to the field of bunyaviruses. We have established an efficient system to generate RVFV chimVLPs and VLPs from 293-gag and 293 cells (Fig. 1, Fig. 2). These VLPs are immunogenic as indicated by the generation of neutralizing antibodies in vivo by immunized mice (Fig. 3) and antigen-specific secretion of immune-related cytokines by splenocytes from vaccinated mice (Fig. 4). Furthermore, these VLP-based vaccine candidates are partially protective in mice and 100% protective in rats against lethal challenge (Fig. 5, Fig. 6). Interestingly, RVF VLP production requires only the expression of the two glycoproteins G_N and G_C, and therefore RVFV N is not required. While consistent with findings by Overby et al. (2006) who were able to generate UUK and bunyamwera VLPs without N, this contradicts recent findings that suggest RVF VLPs could only be generated through expression of RVFV glycoproteins together with RVFV N (as part of the minireplicon system) (Habjan et al., 2009). A possible explanation for this discrepancy is that the use of different expression systems (e.g., chicken β-actin vs. immediate-early cytomegalovirus promoter) leads to substantially different amount of RVFV G being produced, and RVF VLPs derived only from the expression of RVFV G requires high levels of expression.

The generation of RVF VLPs lacking the N protein facilitates the rational vaccine design for the generation of a safe and highly efficient RVFV vaccine following the DIVA (Differentiating Infected from Vaccinated Animals) concept to differentiate vaccinated from infected individuals (see Bird et al., 2008, Capua et al., 2004). Because RVFV-specific antibodies against the N proteins are easily detected in infected individuals, a vaccine candidate lacking the N antigens facilitates DIVA. However, further studies have to be performed (e.g., increased vaccine dose, different adjuvants) to increase N-lacking VLP vaccine efficiency (see Fig. 5).

chimVLPs containing a retroviral gag protein (either MoMLV or simian immunodeficiency virus (SIV) gag) and the antigen of interest (e.g., influenza hemagglutinin and neuraminidase) have been recently described (Guo et al., 2003, Haynes et al., 2009). Unfortunately, the generation of RVF chimVLPs is more complicated because RVFV G and MoMLV gag localize to the Golgi (Gerrard and Nichol, 2002, Schmaljohn and Hooper, 2001, Wasmoen et al., 1988) and plasma membranes (Soneoka et al., 1997), respectively, in mammalian cells. However, over-expression of RVFV G leads to some G_N/G_C localization at the cell surface (Filone et al., 2006, Gerrard and Nichol, 2002, Gerrard and Nichol, 2007, Liu et al., 2008), which allows the generation of RVF chimVLPs. Further attempts to increase RVFV G surface localization by generating chimeric RVFV G proteins containing the ectodomain of RVFV G and the transmembrane domain and cytoplasmic tail of the MoMLV Env (C-terminal 56aa of the envelope polyprotein, accession number GI:331936), which removes a putative Golgi retention signal of RVFV G_N, did not significantly increase RVFV G content on cell surfaces as demonstrated by immunofluorescence studies and did not result in increased chimVLP yields (data not shown).

Optimization of VLP production is important for the ability to scale-up for the generation of material required for non-human primate and livestock studies and ultimately clinical grade vaccine production. First, as seen in Fig. 2A, the ratio of transfected expression plasmid for the RVFV G and N proteins influences VLP yields. Second, optimal generation of VLPs is clearly observed when the N plasmid is included, and increasing amounts of N expression plasmid enhances the generation of RVFV VLPs. Similar findings were also reported for the generation of UUK VLPs (Overby et al., 2006). The addition of a RVFV-specific minigenome did not significantly increase the N content of generated VLPs despite the fact that VLPs were able to package minigenome and transfer reporter gene activity into VLP-infected cells (M. Bouloy and R. Flick, unpublished data). This contradicts previous findings by Overby et al. (2006) who showed that omission of a minigenome leads to almost no nucleoprotein incorporation into UUK VLPs. Generation of VLPs is optimal between 24 and 60 h post-transfection (Fig. 2B) as determined via time course experiments. Furthermore, multiple harvests during a  120-h time period post-transfection results in higher VLP yields compared to less frequent harvests. This is consistent with our previous studies which show that multiple harvests of Ebolavirus and Lassa virus chimVLPs yields substantially more VLPs compared to a single harvest or collections at intervals longer than 12 h (data not shown). This might reflect VLP stability or binding to the producing cells.

While inactivated viral vaccines are often ineffective at eliciting neutralizing antibodies (Green et al., 2001), robust vaccines can elicit the development of neutralizing antibodies that are maintained for prolonged times (Kan et al., 2007). A PRNT₈₀ of 1:40 is generally accepted as protective against RVF disease in mice (Peters et al., 1986), rats (Anderson et al., 1991a, Anderson et al., 1987), hamsters (Niklasson et al., 1984) and Rhesus macaques (Peters et al., 1988). Pittman et al. (1999) examined the neutralizing antibody responses in 598 human subjects vaccinated with the TSI-GSD-200 inactivated RVFV vaccine. 540 (90.3%) had serum neutralizing antibody titers > 1:40 after their primary series of three injections while 58 individuals (9.7%) had titers of < 1:40. PRNT₈₀ > 1:40 was maintained for 183 days in 85% of recipients but decreased to only 35% at 1 year indicating that durability might be an issue with this particular vaccine.

Correlation between the development of neutralizing antibodies and protective efficacy for RVFV vaccines has also been shown more recently in several systems. Naslund et al. (2009) showed that five of six mice vaccinated with 1 × 10⁶ RVF VLPs developed neutralizing antibody titers (PRNT₈₀) from 250 to 1250. Eleven of 12 vaccinated mice were protected from lethal challenge. Bird et al. (2008) showed in a rat model that neutralizing antibody titers (PRNT₅₀) of 1:640–1:7040 were obtained with a highly attenuated RVFV strain lacking the NSs and NSm genes that is 100% protective from lethal challenge.

Here we demonstrate that RVF VLPs induce neutralizing antibody titers of  ≥ 1:640 detectable 6 months post-immunization (Fig. 3), indicating the robustness and durability of the VLP-based RVFV vaccine candidates. Interestingly, as seen in Fig. 3, while similar neutralizing antibody titers are generated by RVF VLPs with or without N, these results do not correlate with vaccine efficacy in the mouse model, as seen in Fig. 5, where N-containing RVF VLPs are substantially more efficacious compared to VLPs lacking RVFV N. As demonstrated with DNA vaccines expressing only the RVFV M ORF, while it is likely that immunity to RVFV is determined by the response to the RVFV G (Lagerqvist et al., 2009, Spik et al., 2006, Wallace et al., 2006), the results described above suggest that N might represent an important component of an efficacious RVFV vaccine. It has been demonstrated that both RVFV N (Lagerqvist et al., 2009, Wallace et al., 2006) and Toscana virus (a related bunyavirus) N (Gori Savellini et al., 2008) are partly protective against lethal challenge in mice. Interestingly, while previous passive transfer studies would suggest that the generation of neutralizing antibodies to RVFV G_N and G_C is predictive of vaccine efficacy (Besselaar and Blackburn, 1991, Schmaljohn et al., 1989), our RVF VLP without N, while sufficient for the generation of neutralizing antibodies (see Fig. 3), is not efficacious in the mouse challenge model (see Fig. 5).

Furthermore, we were able to demonstrate that RVF VLPs clearly induce antigen-specific cytokine secretion by isolated splenocytes from vaccinated animals (Fig. 4). Consistent with these results, Ebolavirus VLPs produced in mammalian and insect cells have been shown to stimulate secretion of cytokines such as IL-6, IL-10, IL-12, and TNF-α from dendritic cells (Bosio et al., 2004, Ye et al., 2006). Taken together, measurement of immune correlates clearly demonstrate that VLPs are immunogenic; however, only vaccine efficacy studies can clearly demonstrates the potency of vaccine candidates.

We therefore employed two different rodent models to determine RVF VLP-based vaccine efficacy. The results shown in Fig. 5, Fig. 6 clearly demonstrate that our RVF VLP vaccine candidates are partially protective in mice and fully protective in rats. Importantly, RVFV ZH501 challenge in rats was performed with a 2 log₁₀ higher dose than is often reported in the literature (see Bird et al., 2008) to ensure 100% lethality in the model and to demonstrate vaccine potency.

Overall, this novel approach of using chimeric RVF VLPs as vaccine candidates yielded promising immunological and efficacy data in two different rodent models and sets a strong precedent for the generation of an efficacious vaccine against RVFV that is urgently needed for the high containment laboratory worker, indigenous people in endemic areas, and for the Strategic National Stockpile and National Veterinary Stockpile as protection against the emergence of this disease and for a potential bioterror event.