A Prototype-Pathogen Approach for the Development of Flavivirus Countermeasures

Abstract

Flaviviruses are a genus within the Flaviviridae family of positive-strand RNA viruses and are transmitted principally through mosquito and tick vectors. These viruses are responsible for hundreds of millions of human infections worldwide per year that result in a range of illnesses from self-limiting febrile syndromes to severe neurotropic and viscerotropic diseases and, in some cases, death. A vaccine against the prototype flavivirus, yellow fever virus, has been deployed for 85 years and is highly effective. While vaccines against some medically important flaviviruses are available, others have proven challenging to develop. The emergence and spread of flaviviruses, including dengue virus and Zika virus, demonstrate their pandemic potential. This review highlights the gaps in knowledge that need to be addressed to allow for the rapid development of vaccines against emerging flaviviruses in the future.

WHAT ARE FLAVIVIRUSES?

The Flaviviridae family comprises 4 genera of positive-strand RNA viruses: Flavivirus, Pestivirus, Hepacivirus, and Pegivirus [1]. More than 70 members of the genus Flavivirus have been described, some of which constitute ongoing and expanding threats to human health [2]. Most flaviviruses are transmitted by mosquitoes or ticks to mammalian and/or avian species (Figure 1). Humans are incidental dead-end hosts for many flaviviruses, including West Nile virus (WNV) and tick-borne encephalitis virus (TBEV). However, an urban cycle supports sustained transmission between mosquitoes and humans for some flaviviruses, such as dengue (DENV), Zika (ZIKV), and yellow fever (YFV) viruses. Additional modes of transmission have been documented, such as sexual and vertical routes for ZIKV [3, 4]. Typically, less than half of flavivirus-infected individuals show any clinical symptoms, although this number varies among flavivirus species. The mildest symptomatic infections cause rash, fever, headache, and myalgia, whereas severe disease syndromes can include encephalitis, flaccid paralysis, hemorrhagic disease, hypovolemic shock, jaundice, and congenital complications that result in significant morbidity and mortality [5]. Flaviviruses have a broad capacity to infect different host tissues and can be viscerotropic and/or neurotropic [6]. More than 100 million clinically apparent flavivirus infections are estimated to occur annually, and numerous outbreaks over the last few decades highlight their pandemic potential [2].

Figure 1.

Transmission cycle of 3 prototype flavivirus pathogens. Dengue virus (DENV) and West Nile virus (WNV) are transmitted via mosquito vectors. For WNV, birds serve as a reservoir for virus and can transmit virus during a mosquito blood meal. Mosquitoes replicate virus and transmit to other vertebrate species, with humans (and horses, not shown) serving as an incidental dead-end host. In contrast, DENV in its urban cycle can be transmitted from a human, because of its high viremia, to a mosquito, and following an intrinsic incubation period in the female mosquito can be transmitted to a naive human host. Rodents serve as a primary reservoir for tick-borne encephalitis virus (TBEV) and, as its name implies, it is transmitted via ticks. Humans are incidental hosts for this virus.

Pestiviruses infect a variety of mammals, including economically important ruminants and swine [7]. Notable pestiviruses include classical swine fever virus and bovine viral diarrhea viruses, although new viruses in this group are continually being discovered [8]. Transmission occurs by contact with respiratory droplets, feces, or urine [7]. It is unclear if pestiviruses productively infect humans, although some human sera have pestivirus-reactive antibodies (Abs) [9]. The potential for pestiviruses to cross the species barrier to infect humans warrants further study.

Hepatitis C virus (HCV) is a hepatotropic hepacivirus with an estimated 58 million people currently infected worldwide (World Health Organization), contributing significantly to the incidence of liver disease and hepatocellular cancer. HCV biology, pathogenesis, and immunity have been reviewed extensively [10]. Transmission occurs predominantly by exposure to blood and has accelerated in recent years, in part due to the opioid crisis and injection drug usage. The development of multiple classes of highly effective antiviral drugs has had a dramatic impact on the prevention and treatment of HCV in the industrialized world [11]. Additional hepaciviruses that infect primates and other mammals have been described [12].

The pegiviruses are a group of viruses related to hepaciviruses but are not yet linked to human diseases, despite evidence that a substantial number of human infections have occurred [1]. Viruses of this genera have been detected in humans, primates, bats, and rodents [12].

The National Institute of Allergy and Infectious Diseases (NIAID) pandemic preparedness plan prioritizes basic and translational research on viral “prototypes” to enable the design of diagnostic antigen and vaccine platforms that could rapidly be applied to multiple viral species within a given family (reviewed in [13]). Flaviviruses are among 10 families prioritized by this initiative. The NIAID plan and this review focus on viruses of the genus Flavivirus due to their ability to cause severe disease in humans, established pandemic potential, and general lack of countermeasures.

FLAVIVIRUS VIROLOGY

Flaviviruses encode an approximately 10.8 kilobase RNA genome that is modified at the 5′ end with a type 1 7-methyl guanosine and 2′-O methyl cap structure but is lacking a polyadenosine tract at the 3′ end [14]. The genomic RNA is translated as a polyprotein that is co- and posttranslationally cleaved by host and viral proteases into 10 proteins (Figure 2A). The 3 amino-terminal (N-terminal) proteins function as structural proteins comprising the virion, whereas the other 7 proteins have roles in RNA synthesis, host and viral protein cleavage, and antagonism of the host immune response and pathogenesis. Each structural and nonstructural (NS) protein has membrane affinity and/or contains 1 or more transmembrane domains, with the exception of NS5, which probably associates with membranes by its association with other viral proteins. Among the NS proteins, flavivirus NS1 has a unique biogenesis. The first step in biogenesis of NS1 is the signal sequence-directed translocation of the polypeptide into the endoplasmic reticulum (ER) where it forms a membrane-associated dimer and regulates early steps of RNA synthesis. It likely functions as a scaffold, as viral replication occurs on the opposite side of the ER membrane in the cytoplasm associated with vesicle packets [15]. NS1 also transits into the trans-Golgi network, where it associates with lipids, forms oligomers, and is secreted or trafficked to the plasma membrane. Secreted NS1 acts as virulence factor that interferes with complement function and can influence endothelial cell and vascular permeability [16]. NS1 is also expressed on the surface of infected cells where it can inhibit complement protein deposition and may drive tissue-specific infectivity [17]. Abs against NS1 can protect against flavivirus infection in mouse and nonhuman primate models, through multiple mechanisms including Fc effector functions and mitigating endothelial cell dysfunction [18–21]. Nonetheless, there are many unanswered questions about the biological functions of the different cellular forms of NS1, which warrant further study to fully realize its potential as a possible antigen in future flavivirus vaccines designed to prevent severe disease.

Figure 2.

Flavivirus genome organization, polyprotein translation, and virus particle morphologies. A, The polyprotein genome organization and topology with proteolytic cleavages shown by black arrows (signal peptidase), red arrowhead (furin), and green arrowheads (viral 2B/3C protease). B, A side view of the prM-E trimer colored by domains: E domain I, red; domain II, yellow; domain III, blue; transmembrane domain, cyan; and prM, pink. On the right is a surface-shaded view of the immature virion. C, The E dimer found in the mature virion colored as in (B). The surface shaded view of the mature virion is shown to the right. D, Surface shaded view of subviral particles. E, Surface shaded images of flavivirus particles that have been observed or inferred from cryoelectron microscopy studies. Abbreviations: C, cytoplasmic capsid protein; E, envelope protein; NS, nonstructural protein; prM, precursor membrane protein.

The 3 structural proteins contain a signal sequence that directs each of the downstream proteins across the ER membrane (Figure 2A). This results in a cytoplasmic capsid protein (C) and 2 membrane-anchored glycoproteins (precursor membrane protein [prM] and envelope [E]) in the lumen of the ER. The C recruits newly synthesized genomic RNA to the site of engagement with prM and E, and envelopment and budding occurs into the lumen of the ER [22]. The resulting immature virion contains 60 trimers of prM:E heterodimers arranged as spikes (Figure 2B) [23]. The immature particle advances through the secretory pathway where the low pH environment of the trans-Golgi network triggers a dramatic rearrangement of the envelope glycoproteins characterized by a fold-down of the heterotrimeric spike and the formation of E protein homodimers [24]. This pH-induced transformation results in a particle with a relatively smooth surface and the exposure of a cleavage site recognized by host cell furin proteases, allowing for proteolysis of prM into pr and M [25]. The N-terminal pr peptide remains associated with the virus particle, preventing premature fusion until exposure to neutral pH upon release from the cell, resulting in an infectious virion (Figure 2C and 2D). However, prM cleavage is inefficient for many flaviviruses grown on laboratory cell lines, particularly DENV, resulting in partially mature virions [26] (Figure 2E). Partially mature virions are more sensitive to the binding of antibodies targeting prM and prM-sensitive E protein epitopes (the presence of prM alters presentation of specific E epitopes and renders some inaccessible) and neutralization compared to fully mature virions. An important gap in knowledge is the maturation state of flavivirus within the human host and the functional relevance of any noncleaved prM. A potential impact of prM or prM-sensitive E protein epitopes needs to be considered in future studies to design optimal vaccine antigens (see below).

The mature virions have E proteins covering most of the particle surface; the residues of M not embedded in the viral membrane are found beneath the E protein and are predicted to be solvent inaccessible. The virion has quasi-icosahedral symmetry, with the E protein dimer situated on the 2-fold icosahedral axis and 3 E protein dimers arranged in parallel as a 6-membered raft [27] (Figure 2). The E protein consists of 3 well-defined ectodomains, with domain I serving as a bridge between domains II and III. Domain II, referred to as the dimerization domain, is an extended, mostly β-stranded structure with a highly conserved fusion loop sequence located at its distal end [28]. As the E protein is dimeric and related by a 2-fold axis of rotation, the fusion loop of one E is nestled underneath a glycan-containing loop from domain I from the adjacent E protein. Studies have shown that the glycan loop is flexible, allowing exposure of the fusion loop [29]. Domain III has an immunoglobulin-like fold and connects with the stem and the anchor domains that embed the E protein to the membrane. Particles lacking C and genome RNA, but containing the 2 membrane glycoproteins, can be formed and released from the cell and are referred to as subviral particles (SVPs) (Figure 2D). Two SVP sizes have been reported, with 1 approximately the same size as the mature particle and the other having a size consistent with a T = 1 icosahedral E protein arrangement [30]. SVPs have been evaluated as immunogens for flavivirus vaccines in preclinical and clinical studies, although the organization of E and M and its implications for immunogenicity require further study [31, 32].

The structural diversity of flaviviruses is not limited to different maturation states or SVPs (Figure 2D and 2E). DENV serotype 2 (DENV-2) virus exists in at least 2 states governed by temperature: the low temperature “native” mature state (Figure 2D) and a “bumpy” mature state observed when particles are incubated at >33.5°C (Figure 2E) [33]. Incubation of DENV-3 at elevated temperatures has also revealed a particle morphology resembling a dumbbell or club-shaped particle [34]. These studies highlight gaps in our understanding of the possible ensembles of structures that flaviviruses can adopt under different physiological temperatures and conditions. Flaviviruses are not static well-ordered arrangements of E proteins and may sample multiple structural states at equilibrium, reminiscent of the “viral breathing” described for enteroviruses [31], which are also prioritized in the prototype pathogen component of the NIAID pandemic preparedness plan [13, 35]. An understanding of the ensemble of structures sampled by flaviviruses at equilibrium and the role dynamic virions play in entry and the replication cycle is incomplete. Viral “breathing” of surface E proteins may be a critical element of attachment, internalization, and fusion processes [36]. E protein sequences that determine the ensemble of states created by conformational dynamics are unknown and merit additional study [37]. Viral breathing alters the surfaces of the virion that may be recognized by antibodies and therefore may contribute to immune evasion and complicate our understanding of the protective antibody response to flavivirus virions or vaccine antigens [38, 39]. Studies are also needed to compare the structural heterogeneity of contemporary clinical strains and historical laboratory-adapted reference virus strains (widely used for research) to understand the impact of laboratory adaptation on viral structure and function.

Despite substantial effort, no consensus has emerged for a dominant proteinaceous entry receptor for any flavivirus [40]. Several reports have identified members of the TIM and TAM families of phosphatidylserine receptors as mediators of flavivirus entry into cells [41]. Other molecules, such as DC-SIGN and CLEC5A, which bind glycans attached to prM and E proteins, were shown to enhance entry by promoting attachment of the virion to the cell surface or in the case of CLEC5A to act as a pattern recognition receptor influencing the cellular response to infection [42]. In many instances, the removal of signals required for endocytosis does not reduce the ability of attachment factors to enhance infectious entry. The identification of cellular molecules that are utilized for particle endocytosis represents an important gap in knowledge. Furthermore, in vitro studies have shown that a broad diversity of cells support infection; however, the restrictions within hosts that govern tropism are not understood and remain a key knowledge gap. The process of low-pH fusion mediated by the class II E fusion proteins is broadly understood, but the precise molecular and structural details of fusion remain elusive [43]. Cryoelectron microscopy has provided many structures of flavivirus mature particles, which appear frequently as a relatively smooth, tightly packed shell of E proteins [44]. The dynamic behavior that has been observed suggests that this shell has flexibility and may therefore facilitate protein movement for receptor engagement and fusion.

FLAVIVIRUS IMMUNITY

Flavivirus infection is controlled by the concerted action of cell-intrinsic, innate, and adaptive immune mechanisms (reviewed in [45]). The innate immune system is composed of a series of pattern recognition systems that detect conserved features of microbes. These pathogen-associated molecular patterns (PAMPs) include nucleic acids, lipoproteins, glycans, and bacterial cell wall components. The recognition of PAMPs by innate immune receptors triggers signaling cascades that stimulate antiviral responses, including the production of type I (eg, α and β) and III (eg, λ) interferons (IFNs). These IFN proteins signal in an autocrine and paracrine manner to induce the expression of many IFN-stimulated genes (ISGs) with antiviral function. The critical importance of the innate response in controlling flavivirus infection is reflected by the large number of antagonists encoded by their genome and their diverse modes of action [46]. However, viral innate immune antagonists often are species specific. Many flaviviruses cannot replicate efficiently in mice due to an inability to blunt the innate immune response to infection [47]. As such, mouse strains lacking innate signaling cascades or ISGs have advanced our understanding of innate immunity to flaviviruses and serve as key animal models to study pathogenesis and countermeasure development.

Humoral Immunity

The role of B cells and humoral immunity to flaviviruses is a critical component of a protective response and has been studied extensively [45]. The antigenic relatedness among flaviviruses is a complex facet of understanding flavivirus pathogenesis and immunity, developing specific diagnostics, and developing vaccines. Flavivirus antibodies may be specific for only a single flavivirus species (flavivirus type specific) or capable of binding to 1 or more heterologous species (cross-reactive). Sequential infections with heterologous flavivirus species occur despite a cross-reactive polyclonal antibody response, highlighting the importance of defining type-specific immunity.

The prM and E structural proteins are dominant targets of the antibody response [45]. Antibodies that bind prM may be cross-reactive, have a limited capacity to neutralize infection in vitro, and enhance disease, at least in some mouse models of DENV [48]. Antibodies that bind the E protein vary considerably in specificity and function. The characterization of mouse and human monoclonal antibodies (mAbs) revealed that virtually every exposed surface of the E protein can be recognized. While the amino acids that comprise an antibody epitope may be contained entirely within a single E protein monomer, biochemical and structural studies have identified multiple modes of quaternary recognition that involve antibody binding across the antiparallel E protein dimer or among E dimers [49–51] (Figure 3). Studies of multiple flaviviruses identify a key contribution of antibodies that bind quaternary determinants in the neutralizing or protective activity of human polyclonal immune sera [52–54]. Many antibodies, including those that potently neutralize infection, bind epitopes that are not predicted to be accessible for recognition on the structurally mature virion [55] (Figure 2). The accessibility of poorly accessible or “cryptic” antibody epitopes is modulated in some instances by the retention of prM on infectious partially mature virions [56, 57]. A deeper understanding of the biology of virions in these structural states and the features that govern transitions among these ensembles are needed to understand antibody recognition.

Figure 3.

Antibodies produced following native flavivirus infection or vaccination. Epitopes elicited by flaviviruses differ based on the immunogen presented. In native infections, mature, immature, and mixed virions (partially mature) present complex epitopes that are conformationally dependent on the arrangement of not only E protein dimers, but E protein rafts and the presence of prM, resulting in a diversity of antibody specificities. A gap in knowledge is which types of antibodies provide the best protection and which vaccine strategy can elicit optimal responses. Abbreviations: E, envelope protein; LAV, live-attenuated vaccine; prM, precursor membrane protein.

Many surfaces of the highly immunogenic NS1 protein are recognized by antibodies that can protect against infection [58, 59]. NS1 has shown promise as a vaccine antigen in animal studies of multiple flaviviruses [60–62]. A favorable characteristic of NS1 vaccine antigens is they elicit antibodies that do not support antibody-dependent enhancement (ADE), discussed below.

The detection of antibodies to nonstructural protein targets can be a useful approach in clinical trials for identifying new infections in the context of vaccine-elicited immunity elicited by vaccine platforms that exclusively express or incorporate structural proteins. Type-specific NS1 antibodies have been shown to have considerable diagnostic utility [63]. However, incorporating NS1 epitopes or antigens in candidate vaccines could complicate use of NS1 antibodies to distinguish vaccine-elicited responses from infection within flavivirus experienced populations. Antibodies specific for NS2B and/or NS5 may also be useful diagnostically [64, 65].

Antibody Mechanisms of Protection

Antibodies contribute to protection and clearance of flavivirus infections via multiple mechanisms. Antibody-mediated neutralization is defined herein as an antibody-mediated reduction in the capacity of a virion to infect target cells. Multiple neutralization mechanisms have been defined, including the inhibition of virion attachment to cells and interference with conformational changes in the structural proteins required for internalization and endosomal membrane fusion [66]. Multiple factors influence neutralization potency, including antibody subclass, complement binding, attachment factor utilization, Fc-receptor (FcR) expression, maturation state, and temperature [38, 67, 68]. Assay format, including cellular substrate, viral passage history, specific infectivity, and propagation technique may also impact apparent neutralization potency. The development of standardized quantitative approaches of reference strains and international standards to measure virus neutralization activity is critical [69]. A correlate of protection (CoP) is a marker of immune function that statistically correlates with protection after vaccination or infection, but may not be the mechanism of protection [70]. For flaviviruses, neutralizing antibody activity is a frequently assumed CoP during vaccine development and has been established experimentally for Japanese encephalitis virus (JEV), TBEV, and YFV [71–73]. However, neutralization alone appears to be an imperfect CoP in other circumstances, as some children enrolled in DENV vaccine trials who developed neutralizing antibodies were not protected from disease [74].

Antibody effector functions mediated by interactions between constant regions of antibody molecules and complement or FcRs on immune cells also contribute to flavivirus immunity. In vivo protection by some flavivirus E protein-reactive antibodies have been shown to depend on an ability to engage FcRs of complement [75, 76], as observed in studies of other medically important viruses. These mechanisms are important for understanding how NS1-reactive antibodies that do not directly bind virions contribute to protection from disease [18, 77]. Antibody effector functions correlate with or predict infection outcomes in humans [21, 78]. In DENV infection, afucosylated IgG1 (the glycan on the Fc region is missing core fucose sugar units) can be associated with more severe disease outcomes, although the mechanism is not clear and therefore this represents a gap in knowledge.

Antibody-Dependent Enhancement of Infection

ADE describes an antibody concentration-dependent increase in the infection of cells [79]. The principal mechanism of ADE involves an increase in the efficiency of attachment of virion immune complexes to cells that express FcRs [80, 81]. Concentrations of antibodies that support ADE in vitro are defined at the upper limit by the stoichiometric requirements for antibody neutralization and at lower concentrations by the minimal number of antibodies that support immune-complex attachment to cells [79]. Most flavivirus-reactive antibodies support ADE in vitro at some antibody concentration. Curiously, antibodies may bind the virion with an angle of approach that does not support efficient interaction with FcRs, greatly limiting ADE activity [82]. Autoreactive antibodies that bind epitopes shared by the virus and cell surface molecules may also increase the efficiency of infection by tethering virions to the cell surface [83]. Antibody binding also may stabilize virions in a conformation that enhances viral fusion, providing a mechanism to increase the specific infectivity of virions at a postbinding step in the replication cycle [84, 85]. While a growing body of work points to some flavivirus cross reactive antibodies enhancing dengue infections in humans, a major gap that stifles vaccine research is the absence of laboratory models and assays for studying and predicting ADE in vivo.

Secondary DENV infections by viral strains grouped within a heterologous DENV serotype can be associated with severe disease outcomes [86, 87]. Prior ZIKV infection has also been demonstrated to exacerbate secondary DENV2 infections in Nicaragua and Brazil [88]. While investigators have debated the roles of different adaptive immune mechanisms in severe dengue disease, the results from a recent dengue vaccine clinical trial strongly implicate dengue-specific antibodies as drivers of severe disease. When a live attenuated chimeric YFV-DENV vaccine containing the structural proteins (main targets of host antibodies) from DENVs and the nonstructural proteins from YFV (main targets of T cell responses) was administered to dengue-seronegative children, the vaccine induced DENV-specific, weakly neutralizing antibodies and increased the risk of severe disease compared to unvaccinated children [89, 90]. Further experimental support for this concept comes from the passive transfer of flavivirus–cross-reactive antibodies into animal models that increases viremia or disease in animal models of DENV infection [91]. The exact conditions that predict immune-enhanced disease in vivo are not well-established or linked to any laboratory test. Although ADE can readily be observed in cell culture with many flavivirus antisera, we currently do not understand the specific immune molecules and mechanisms responsible for severe dengue disease. Immune-enhanced disease is an important risk for flavivirus vaccines expected to be deployed in regions with endemic flaviviruses and administered to travelers in these endemic regions. Additional mechanistic and epidemiological insights into heterotypic immunity are essential. Furthermore, vaccine and therapeutic strategies designed to limit the potential for ADE should be prioritized [75, 92].

T Cell Immunity

T cells have important and diverse functions in the immune response to flaviviruses, where they generally contribute to protection via multiple mechanisms [91]. Activated CD8⁺ T cells are recruited into infected tissues, where they locate and eliminate virus-infected cells via several effector molecules, including the production of granzymes and perforin [93]. In addition to direct lysis, T cells shape innate and adaptive responses through the production of proinflammatory cytokines (such as IFN-γ and tumor necrosis factor-α [TNF-α]) and chemokines that amplify immunity through effector cell recruitment and activation [94]. Likewise, CD4⁺ T cells are activated by viral infection; these cells provide crucial help for humoral responses and instruct the transition to long-term memory in both the B- and T cell compartments in lymphoid tissues. For some viral infections, CD4⁺ T cells can also be cytolytic and kill virus-infected targets, though the extent of this in flavivirus infection is currently unclear.

Animal models and ex vivo analyses of human T cells isolated from infected individuals point to a protective role for antiviral T cells during most flavivirus infections. During murine DENV and ZIKV infection, depletion or genetic ablation of CD8⁺ T cells results in increased viral burden and mortality [95]. Similarly, CD4⁺ T cells can protect against DENV infection, even independent of their role providing T and B cell help [96]. Adoptive transfer of activated DENV-reactive T cells has established that cross-protective immunity is possible in some circumstances for antigenically related viruses [97]. ZIKV/DENV cross-reactive T cells expanded in ZIKV-infected mice and vaccinations with cross-reactive epitopes reduced ZIKV burden in a CD8⁺ T cell dependent manner [98]. Thus, protective cross-reactive CD8⁺ T cells could be elicited by vaccination and thus deployed against emerging flaviviruses.

Human flavivirus-specific T cells are expanded during acute infection, and much work has been dedicated to deciphering peptides presented by different HLA molecules for T cell recognition [99, 100]. During acute human DENV infection, T cells are preferentially generated against peptides from nonstructural viral proteins, suggesting T cells could complement neutralizing antibody responses that are directed against less conserved structural viral proteins [101]. Some studies also indicate a role for T cells in both protection and immunopathology in the JE serogroup viruses [102]. Cross-reactive T cells have also been identified after human flavivirus infection, although their contribution to protection and pathology remains uncertain [103, 104].

While antiviral T cells can have a protective role during viral infection, overexuberant or inappropriate responses can result in immunopathology, which has been observed during some flavivirus infections. For DENV, CD8⁺ T cells may contribute to pathogenesis via a process called original antigenic sin, whereby cross-reactive T cells elicited by a first infection expand and predominate following a secondary heterologous DENV challenge [105]. These anamnestic cross-reactive responses are thought to be of lower affinity and exhibit altered cytokine production to nonhomologous DENV. While ex vivo human studies have confirmed the skewing of T cell responses predicted by this model, this pattern has not been associated with poor outcomes [99]. In fact, detailed transcriptomic analysis of both CD4 and CD8 DENV-specific T cells did not reveal significant differences in responses associated with severe disease [106, 107].

Approaches to optimize T cell contributions to vaccine-elicited immune protection are an important consideration for future flavivirus antigen design and vaccine platform selection. Identification of the most protective T cell peptide epitopes is an important gap in our understanding of antiflavivirus immunity and could enable vaccine strategies that maximize the activation and long-term retention of protective and potentially broadly reactive T cells. In this context, identification of epitopes/antigen regions that are broadly conserved within different flavivirus species might be of particular interest [108].

FLAVIVIRUS MODELS OF DISEASE

Animal models have been developed for flavivirus infections that result in encephalitis, viscerotropism, hemorrhagic disease, or teratogenicity [109–112]. While much has been learned from their use, many of these models have limitations because they do not fully recapitulate important features of human disease. Many flavivirus mouse models utilize immunocompromised, gene knock-in or gene knock-out strains to generate clinical signs of disease. Differences in placenta structure and biology between mice and humans have limited our understanding of vertical transmission, which is critical for understanding the most severe ZIKV outcomes [113]. Although many flaviviruses infect primates, the nonhuman primate models have limited utility and often require large viral inocula to generate any clinical disease. Discoveries enabling the development of additional more humanized animal models could accelerate medical counter measure development. Determining why flaviviruses do not replicate efficiently in many animals is an important knowledge gap. This understanding could be informed by greater knowledge of the determinants of tropism, including specific flavivirus cellular receptors, host factors required for flavivirus replication, and species-dependent immune evasion mechanisms. Some flavivirus infections can result in Guillain-Barré syndrome [114], although this occurs at low frequency making it challenging to study. Models to explore the mechanisms of Guillain-Barré syndrome would advance the field. Recent advances in development of in vitro, organoid, 3-dimensional culture models may offer opportunities to develop models of flavivirus infection and injury in relevant cells of the central nervous system [115]. The ability to replicate and manipulate the cellular environment of a specific organoid culture model and to track virus infection may prove to be an important addition to understanding molecular pathogenesis.

In recent years, human cohort studies and candidate vaccine clinical trials have generated considerable information about mechanisms of flavivirus pathogenesis and immunity, particularly for DENV and ZIKV. Controlled human infection models (CHIMS) of attenuated strains of DENV and ZIKV can expedite the clinical development of vaccine candidates and may become a powerful approach for identifying immune correlates of protection [116, 117]. A CHIM may prove critical for the evaluation of vaccines for which the incidence of infection is low, such as many of the flavivirus that have humans as dead-end hosts. Rapidly developing a CHIM for an emerging flavivirus has challenges that relate to requirements for understanding viral pathogenesis and disease, mechanisms of transmission, and the characterization and manufacture of appropriately attenuated challenge strains. Furthermore, the criteria for limiting risk for study participants and ethical considerations concerning the potential value to society are complex and critical issues that must be discussed before embarking on these human studies. The availability of rescue therapy for study participants (small molecule inhibitors or therapeutic mAbs) might simplify the establishment of these models.

FLAVIVIRUS VACCINES

How long will it take to develop a vaccine for a newly emerged flavivirus? Historically, vaccine development from discovery to licensure has taken over 10 years. The COVID-19 pandemic revealed that greatly accelerated vaccine development timelines are possible; the first COVID-19 vaccine took only 326 days from discovery to emergency use authorization [118]. This remarkable achievement was made possible in part by a sophisticated understanding of the structure of protective coronavirus spike protein antigens and prior experience with the modified messenger RNA (mRNA) platform. Despite the rapid development timeline, this may not be quick enough to control an outbreak or provide sufficient opportunity to establish the efficacy of new countermeasures. For example, while the ZIKV outbreak in 2015–2016 resulted in the development of multiple vaccine candidates in a short time frame, greatly reduced levels of virus transmission by 2017 limited the ability to conduct efficacy trials. Timelines from discovery to the deployment of 100 days have been proposed as a target by some [119], which is a laudable goal and a great challenge.

Flavivirus E proteins are class II fusion proteins that share considerable amino acid homology among members of the genus and a similar overall structure and arrangement on the virion (Figure 2). It is likely that the immunogenic and protective features of one flavivirus antigen could guide the development of vaccines against a poorly characterized emerging flavivirus. It remains unclear if immune responses control all flaviviruses in the same manner. For example, immunity that prevents entry into the brain and neurological disease may differ from that preventing virus entry into the placenta and congenital disease. Detailed comparative studies of neutralizing and nonneutralizing antibodies and T cell-mediated immunity [120] between natural virus infection and new vaccine candidates will be instructive. Analysis of 2 related ZIKV vaccines identified qualitative features of the antibody response rather than neutralizing activity as predictive of vaccine-elicited immune protection [74]. The identification of CoPs and reproducible, quantitative methods to measure relative vaccine-elicited responses will be critical to guide vaccine candidate downselection and evaluation during rapid vaccine development campaigns.

Multiple licensed flavivirus vaccines have proven successful at controlling encephalitic, viscerotropic, and hemorrhagic disease, suggesting vaccines can prevent most disease syndromes caused by flavivirus infections. Can insights from these successes be leveraged for rapid development of vaccines to protect against new flavivirus diseases?

Formalin-inactivated vaccines have been developed for JEV, TBEV, and Kyasanur Forest disease virus [121]. Each of these vaccines requires 2 doses to induce protective immunity and booster doses to maintain protective immunity. In contrast, empirically derived live-attenuated vaccines have been developed for YFV (strain 17D) and JEV (strain SA14–14-2) that require just 1 dose to induce long-term protective immunity [122]. The CoP for the JEV vaccine has been shown to be neutralizing antibodies with PRNT₅₀ (50% plaque reduction neutralization test) titers of 10 being protective [123]. Despite the availability of YFV-17D and SA14-14-2 vaccines for 85 and 48 years, respectively, an understanding of the molecular mechanism of attenuation and protective immunity is still rudimentary [124–127]. Additional investment to understand these mechanisms will advance development of vaccines against other flaviviruses.

In the last decade, rationally derived live-attenuated vaccines have been developed for JEV and DENV. Both are recombinant chimeric viruses engineered using YFV-17D live-attenuated vaccine backbone where the prM and E genes are replaced with those of JEV SA14-14-2 or wild-type DENV genes representing the 4 DENV serotypes. Live-attenuated recombinant chimeric DENV vaccines also have been developed using similar approaches and the backbone of DENV-2 or DENV-4 viruses [128]. The tetravalent NIAID DENV vaccine candidate includes 1 chimeric component; all viruses in this vaccine encode deletions in the 3′ untranslated region, which attenuate the virus by sensitizing it to type I IFN responses [129].

Rapidly Deployable Platforms

Multiple antigen designs and vaccine platforms have been evaluated for flaviviruses. The rapid vaccine development following the emergence of ZIKV in the Americas was the most recent application of the state-of-the-art for flavivirus vaccine design, with more than a dozen candidates rapidly advanced into human clinical trials. Vaccines examined in clinical or preclinical studies included whole inactivated vaccines [130–132] and rationally designed live-attenuated vaccines [133, 134]. Strategies to reduce barriers for manufacturing, regulatory approval, and deployment of vaccines approaches derived from infectious virus merits continued consideration. Recombinant soluble E proteins, SVPs, and virus-like capsid-containing particles have been studied for ZIKV and other flaviviruses [135, 136]. More recently, investigators have used structure guided computational methods to design stable E dimers displaying quaternary epitopes that induced more potent neutralizing and protective antibody responses compared to traditional E monomer subunit vaccinees [92, 137]. NS1 has also shown promise as a vaccine antigen [60, 138]. Flavivirus antigens can be delivered using multiple vaccine platforms including DNA, mRNA, and viral vectors (eg, adenovirus, measles virus, modified vaccinia virus) [139–141]. An adenoviral construct expressing ZIKV M and E proteins evaluated in phase 1 clinical trials [142] using mRNA, DNA, or viral-vectored ZIKV vaccines were all shown to be immunogenic, protective in animal models [139, 143–148], and some were advanced into clinical trials [149, 150]. The speed, diversity of approaches, and success/failures of the vaccine development response to ZIKV holds lessons for how to respond to the next emerging flavivirus.

Despite rapid progress, the optimal composition of a flavivirus vaccine still is unclear. While the E protein is likely the most important target of a protective immune response, the success of NS1 vaccines in preclinical studies, and the importance of this protein in disease pathogenesis, raises questions regarding its inclusion in future vaccines [151, 152]. Vaccines that deliver SVP antigens are straightforward to develop, but studies with ZIKV raised the possibility that heterogeneous structures described previously may elicit antibody responses that are inferior compared to natural infection [74]. A major goal will be to design E protein-based antigens that display most protective epitopes but limit recognition by antibodies with the greatest potential for ADE [153]. One approach to achieving this goal has been the introduction of mutations into cross-reactive epitopes hypothesized to be undesirable, including the highly conserved fusion loop [148, 154]. Another strategy is the production of disulfide-linked E protein dimers [92, 137, 155], which would reduce exposure of the fusion loop and other cryptic epitopes yet remain recognized by highly neutralizing E protein-dimer reactive antibodies.

Strategies to optimize CD4 and CD8 T cell immunogenicity to complement existing vaccine concepts requires further study [156, 157]. While more broadly cross-reactive T cells that recognize different flavivirus species are rare and associated with lower reactivity [158], it was shown that the majority of T cell responses following immunization with the tetravalent NIAID DENV vaccines are focused on epitopes conserved across different serotypes [159]. This suggest that a vaccine component focusing responses on highly conserved regions might be synergistic with vaccine designed to optimally inducing neutralizing antibodies responses, and of particular interest in the context of pandemic preparedness [108]. A comprehensive approach to vaccine development for flavivirus prototypes is required to establish key relationships between antigen structure, the immune repertoire, protection mechanisms, CoP, and protection.

ROLES FOR MONOCLONAL ANTIBODIES

Neutralizing antibodies have been established as the CoP for some flavivirus vaccines, leading to studies investigating antibodies for prophylaxis and as therapeutics. Polyclonal antisera and mAbs can be used prophylactically to mediate protection in animal models for a variety of flavivirus infections [160–162]. Identifying antibodies with the greatest therapeutic potential is an important goal with translational implications [162]. Recent technological advances have enabled detailed studies of the human antibody repertoire to identify mAbs with promising characteristics. Flavivirus-reactive mAbs can be discovered using viral antigen-agnostic approaches or by antigen-dependent sorting of B cells expressing antibodies with desirable features. While antibody discovery studies undertaken in mice have yielded many potent mAbs that have advanced the field from a mechanistic and structural perspective, the antibody repertoire in humans and mice may differ. Factors that combine to define the potential clinical utility of antibodies have been reviewed [163] and merit additional study. For flaviviruses, passive administration of potently neutralizing antibodies engineered to prolong half-life in vivo or to be incapable of FcR-receptor interactions, and thereby unable to mediate ADE, holds promise to prevent infection. Therapeutic applications are possible, but more challenging. Protection against flaviviruses can occur if mAbs are given shortly after virus challenge. Additional study is required to define the window of opportunity to treat flavivirus infections with viremias of short duration, or before viruses seed target tissue(s) with potentially limited access to antibody. In particular, there may be significant differences in response depending upon a viscerotropic versus an encephalitic flavivirus.

FLAVIVIRUS PROTOTYPES

Prototype recommendations focused on members of the flavivirus genus rather than the other genera found within Flaviviridae, based on the assessment of pandemic potential by a working group supporting the NIAID. Identifying these prototypes, and the gaps in current knowledge, may advance the rapid development of vaccines against emerging flaviviruses in the future. Two major factors were considered: ecology and clinical disease syndrome. Approximately 50% of flaviviruses are transmitted by mosquitoes and 25% by ticks; the remaining 25% have either no known vector or are insect-specific flaviviruses. The study of prototypes transmitted by both vector types is prudent and the inclusion of prototypes with varying disease tropism was also deemed important. Given the large number of flaviviruses and complexity of transmission and disease, 3 prototype pathogens were recommended: WNV, DENV serotype 2, and TBEV. ZIKV was also discussed as a potential prototype because it is transmitted by mosquito, vertical, and sexual routes. However, it mutates frequently with cell passage, which has the potential to complicate results obtained. Box 1 highlights the gaps in knowledge as identified by the flavivirus working group that can be applied to these prototype as well as other flaviviruses.

Box 1. Gaps in Knowledge.

• Evaluation of the potential of pestiviruses to cross the species barrier into humans.

• Partially mature virions are defined as virus particles with structural features of mature and immature virus particles. What is the impact of the noncleaved prM present on some virions on virus attachment and entry, immunity, pathogenesis?

• The precise molecular details of low-pH fusion mediated by class II E fusion proteins are poorly understood.

• What is the function(s) of the dynamic breathing behavior of flaviviruses?

• What are the features of virions that govern transitions among the structural ensembles? Can a detailed understanding of these interactions improve our understanding of antibody recognition and inform antigen design?

• A greater understanding of NS1 functions in disease pathogenesis and protective immunity is required. What are the mechanisms of protective NS1 antibodies?

• There is no clear consensus for any flavivirus on entry receptors and the cellular molecules that are utilized for particle endocytosis and fusion.

• The contributions and mechanisms of non-neutralizing antibodies to immune protection should be explored.

• Deeper insight is needed into how and under what circumstances the modifications to antibody Fc domains occur.

• Antibody-dependent enhancement of infection (ADE) is a potential risk for flavivirus vaccines expected to be deployed in regions with endemic flaviviruses. Mechanistic and epidemiological insights into the heterotypic immunity is essential.

• Research to enable the development of vaccine and antibody therapeutic strategies that limit the potential for ADE should be prioritized.

• Most animal models for flavivirus infections do not fully recapitulate disease seen in humans. Improved animal models are needed for many flavivirus infections to enhance our understanding of mechanisms of pathogenesis.

• Flavivirus clinical syndromes need additional studies, including Guillain-Barré syndrome and how flaviviruses cross the placenta and infect the fetus.

• There is limited understanding of the mechanism of attenuation and protective immunity of the licensed live-attenuated YFV and JEV vaccines. A deeper understanding of these vaccines will aid development of vaccines against other flaviviruses.

• What are the most protective flavivirus antibody epitopes? What is the optimal antigen design to elicit these?

• Are prM- and E protein-reactive antibodies targeting partially immature features of virions desirable or undesirable components of the antibody response and therefore need to be considered for vaccine antigen design?

• What are the best ways to best elicit protective T cell responses? Are cross-reactive T cells protective?

• Should NS1 be included in vaccines?

West Nile Virus

WNV is a neurotropic flavivirus transmitted by mosquito vectors. WNV is the most important arboviral disease in the United States for which there are no medical countermeasures. It grows to high titers in cell culture, mutates little when passaged in vitro, and infects multiple nonimmunocompromised animal species. There are well-characterized rodent models available and nonhuman primate models are described. Large numbers of reagents are available to study the virus, which has recently been reclassified to allow handing using biosafety level 2 (BSL-2) conditions.

Dengue Virus Serotype 2

DENV is the most widespread mosquito-borne viral infection of humans, with as many as 100 million symptomatic cases reported each year [164]. The global infections caused the 4 DENV serotypes comprise an ongoing pandemic and, therefore, must be prioritized. DENV-2 is considered an optimal model for all DENV serotypes because there are many assays and reagents already available, as well as a large body of biological and structural data. A CHIM has been developed for DENV and multiple dengue cohort studies are ongoing, creating opportunities to incorporate information from clinical and epidemiological studies into research. The 4 DENVs are classified as BSL-2 agents in the United States.

Tick-Borne Encephalitis Viruses

Tick-borne and mosquito-borne flaviviruses differ in their transmission route and degree of genetic diversity in the field. Infection by several tick-borne flaviviruses is associated with severe disease and a deeper understanding of these viruses is an important preparedness goal. Powassan virus (the deer tick genotype) currently causes outbreaks in the United States. While Powassan virus transmission has been limited and outbreaks have been small, there is potential for large outbreaks. Furthermore, cofeeding could amplify transmission and may have the potential to increase virus spread.

TBE viruses are the major tick-borne flaviviruses and grow to high titers in cell culture and are also relatively stable when passaged in cell culture. They are currently researched using BSL-4 containment and practices, and all but the Central European TBE subtype (TBEV-Ce) are classified as select agents in the United States, complicating the study of these viruses. These obstacles were recently reduced by the recommendation that TBEV-Ce can be studied using BSL-3 containment and practices, provided all laboratory staff are vaccinated. The Food and Drug Administration has approved a TBEV-Ce vaccine for use in the United States, and this has been endorsed by the Centers for Disease Control and Prevention Advisory Committee on Immunization Practices. The availability of the vaccine should facilitate the study of TBEV-Ce as the prototype tick-borne flavivirus.

Notes

Acknowledgments. We thank members of the Flaviviridae community for their input on gaps in scientific knowledge and recommendations of the prototype pathogens for this viral family.

Financial support. T. C. P. is supported by the Intramural Program of the National Institute of Allergy and Infectious Diseases.

Supplement sponsorship . This article appears as part of the supplement “Pandemic Preparedness at NIAID: Prototype Pathogen Approach to Accelerate Medical Countermeasures—Vaccines and Monoclonal Antibodies,” sponsored by the National Institute of Allergy and Infectious Diseases, NIH, Bethesda, MD.