Yellow fever vaccine.

Yellow fever, a mosquito-borne viral hemorrhagic fever, is one of the most lethal diseases of humankind. The etiologic agent is the prototype member of the genus Flavivirus, family Flaviviridae, a group of small, enveloped, positive-sense, single-strand RNA viruses. Approximately one in seven people who become infected develop a rapidly progressive illness, with hepatitis, renal failure, hemorrhage and cardiovascular shock, with a case fatality rate of 20-50%. Yellow fever occurs in sub-Saharan Africa and tropical South America, where it remains a continuing public health problem of varying magnitude, depending on the level of vaccination coverage in the human population and cyclical, ecologic and climatic factors that influence virus transmission.

In the decade 1986-1995, there was a dramatic upsurge in the incidence of yellow fever (YF) due, in part, to a true increase, as well as to improved reporting. During this interval, there were a total of 24,753 cases (2159 in South America and 22,594 in Africa) and 6594 deaths (1398 in South America and 5196 in Africa) officially notified to the World Health Organization (WHO). Outbreaks continue to occur on a regular basis, and between 2000 and 2004, 16 countries reported one or more epidemics [1]. The true incidence, which has been estimated to be approximately 200,000 cases per year, exceeds the reported incidence by a factor of 10-500 [1,2]. However, this estimate, based on historic epidemics, serosurveys and calculations of endemic disease burden [3], is dated and needs revising by on-the-ground surveillance, which almost does not exist, particularly in Africa where the disease incidence is significantly higher than in tropical America. In addition to the impact of YF in endemic regions, there is a long-recognized risk of an emergence of urban epidemics of YF in densely populated coastal regions of Africa and South America, and in distant parts of the world that are infected with the vector Stegomyia aegypti (formerly Aedes aegypti), which is capable of interhuman transmission of the virus. The occurrence of urban YF in Asia, southern USA or Central America would increase demands for vaccines far beyond current supply. Recent introduction of YF vaccine into the Expanded Program of Immunization (EPI) in South America and a number of African countries [1], as well as the expansion of travel from USA and Europe to endemic countries, have increased demands for vaccines. Almost every year, there are temporary shortages of 17D vaccine for travelers.
YF virus is transmitted between wild monkeys and mosquitoes and cannot be eradicated. Vaccination represents the most efficient means of preventing and controlling the disease. In 1936, Max Theiler, Wray Lloyd and Hugh Smith at the Rockefeller Foundation developed the 17D live attenuated vaccine [4], and field trials of the vaccine in Brazil commenced the following year. Manufacturing of 17D vaccine began during World War II in the USA and the UK, and later in at least ten other countries [1], when it became widely used for protection of travelers and military populations. In 1951, Theiler was awarded the Nobel Prize in Medicine for the development of 17D vaccine. Mass vaccination campaigns in countries of tropical South America, beginning in the late 1930s in Brazil and later in other countries, led to a reduction in the incidence of YF. In Africa, 17D vaccine was used primarily for emergency control of epidemics until 1988, when the WHO recommended introduction of 17D into the EPI. This program has met with highly variable success, and coverage in many countries is low [1]. However, wherever high vaccine coverage rates have been achieved, the incidence of YF has diminished. Previous reviews have discussed the characteristics and usage of YF vaccines [5,6]. The present paper will be limited to a review of 17D vaccines. A different live attenuated YF vaccine, the French neurotropic vaccine (FNV), is no longer in use and will not be covered. FNV has an interesting history [7], and has been the subject of several molecular studies.

Development of 17D vaccine
In 1927, YF virus was first isolated from blood of a human patient in west Africa (the Asibi strain). The 17D vaccine was derived by empirical methods of sequential passage of the Asibi virus in tissue cultures that were restrictive for growth of the virus [4]. Initial in vitro passages were performed in cultures of minced murine embryo tissue. After 18 passages, the virus was cultured in minced whole chicken embryo. After 58 passages, the virus, now designated as subculture series 17D, was grown in minced chick embryo from which the brain and spinal cord had been removed. Intermediate passage levels were checked for two critical biologic characteristics: neurovirulence for adult mice and monkeys after intracerebral inoculation and viscerotropism, that is, the ability of the virus to cause fever, illness, hepatitis and viremia in monkeys. These evaluations revealed that the virus had become attenuated with respect to neurotropism and viscerotropism for monkeys between the passages 89 and 114. Monkeys inoculated by the subcutaneous route remained well, developed neutralizing antibodies and were protected against challenge with virulent Asibi virus. To ensure safety, continued passages were made in minced chick embryo cell culture before preparation of vaccine batches in embryonated hens' eggs. The first clinical trials in 1936 employed virus at passage 227 (produced in New York, NY, USA) and passage 229 (produced in Brazil) [4,8]. A simplified passage history during the development of YF vaccine is shown in FIGURE 1, and further details have been published elsewhere [6,9]. Two different substrains (17D-204 and 17DD) derived from a common ancestor are used today for manufacturing YF vaccines.

17D-204 substrain
This subculture series was used for development of vaccines manufactured in all countries except Brazil (and, for a period of time in the past, in Senegal). The original 17D subculture series in minced chick embryo (without nervous tissue) was continued to passage 204, at which point a series of 17 passages (the 17D-204 series) was made using normal human serum in the culture medium. Beginning at passage 221, the virus was propagated in embryonated hens' eggs to prepare vaccine batches at the Rockefeller Foundation in NY, USA, and the YF Laboratory, Bogota, Colombia. A batch prepared in Colombia at passage 228 (Colombia-88) was returned to the Rockefeller Foundation in 1940 and was the common ancestor of seed stocks used for vaccine manufacture as shown in FIGURE 1. In 1977, the Robert Koch Institute (RKI) in Germany prepared for the WHO new seed viruses from the RKI primary seed stock (passage 236). A new primary seed at passage 237 was designated 17D-213-77 [9]. In the literature, this material is often referred to as 17D-213 but, in fact, it represents the 17D-204 substrain. All current 17D-204 series vaccines are currently manufactured in embryonated eggs at passage levels between 235 and 238.

17DD substrain
Virus at passage 195 in minced chick embryo (without nervous tissue) was used to initiate a separate subculture series at the Rockefeller Foundation, designated 17DD. At passage 229, the virus was sent to the Instituto Oswaldo Cruz, Rio de Janeiro, Brazil. The virus was passed 14 times in chick embryo tissue cultures and then, beginning at passage 243, in whole embryonated hens' eggs. Primary and secondary seeds were prepared in Brazil at passages 284 and 285, respectively, and the current vaccine is at passage 286. Recently, a new secondary seed was produced that raised the vaccine passage level to 287.

Lessons learned during the early days of vaccine development Passage level & dose
Between 1938 and 1941, 17D vaccine batches were deployed belonging to several independent subculture series, including 17DD, 17D-204 and others (17D2, 17D3, 17D low and 17D-NY104), without control of passage level within these series. Experience with these batches revealed the importance of control of substrain and passage level. Some batches were overattenuated and poorly immunogenic, whereas others were associated with the appearance of serious adverse events postvaccinal encephalitis [10]. The clinical and epidemiologic observations were extended to experimental studies in monkeys, which demonstrated that batches associated with encephalitis had increased neurotropism [11]. These observations led to the institution of a 'seed-lot' system in 1941, in which vaccine passage level was stabilized. Primary seed and manufacturer's working (secondary) seeds were prepared and characterized for safety and immunogenicity in monkeys. The seed-lot system was formally established as a biologic standard in 1945 [12].
There was considerable batch-to-batch variation in virus concentration in some early lots of vaccine. Following the problems noted above, Fox and Penna investigated the relationship between dose, viremia and antibody responses in rhesus monkeys [8]. These studies concluded that low doses were associated with higher, later and more long-lasting viremia, and a delayed immune response, and suggested that low dose may be associated with severe adverse events. This phenomenon will be discussed later with respect to new YF-vectored vaccines.
Based on historic experience and early dose-ranging field studies [13], the minimum dose specification for YF vaccine was set at 1000 mouse intracerebral 50% lethal doses (LD 50 ), or its equivalent in cell culture plaque-forming units (pfu), and this has been carried forward into modern requirements for YF vaccines [14]. Current vaccines are formulated by the manufacturer to contain the appropriate dose in the 0.5 ml inoculum.

Adventitious viruses
Early manufacturing methods employed normal human sera. The infected chick embryos were harvested, triturated in a blender, suspended in human serum and centrifuged to remove solids. The supernate was then sterilized by filtration. The human serum was necessary to reduce losses across the bacteriologic filter. The use of human sera resulted in large outbreaks of hepatitis B in Brazil in 1939 [15] and in US servicemen in 1942 [16,17]. Human serum was removed from the vaccine in the early 1940s, but this required elimination of sterile filtration by substituting aseptic methods for harvesting and processing the embryos.
In 1966, YF 17D vaccine was discovered to be contaminated with avian leukosis virus (ALV) [18]. New seeds free from ALV were developed in the 1970s, and most manufacturers now employ ALV-free seeds, as stipulated by WHO standards. An epidemiologic study revealed no increased cancer risk in people who had been vaccinated with ALVcontaminated vaccine [19]. Subsequently, with the introduction of sensitive retrovirus tests, such as the product-enhanced reverse transcriptase (PERT) assay, it was found that eggs and egg-based vaccines are positive, reflecting the presence of endogenous retrovirus sequences. For a new YF vaccine to be approved, it would be necessary to show the absence of infectious or inducible replication-competent endogenous retroviruses in the seed stocks and, possibly, each vaccine lot [20,21].

Current vaccine manufacture & control
YF 17D-204 vaccines approved by the WHO are manufactured in six countries (TABLE 1). The passage history of these vaccines is shown in FIGURE 1. Vaccines for local consumption are produced in Colombia, Russia and China. In the USA and Europe, vaccines are produced by aseptic inoculation of secondary seed into viable embryonated hens' eggs from closed, special pathogen-free flocks. Embryos must be less than 12 days of age at the time of harvest to reduce the risk of allergic reactions. A general schema for vaccine production is shown in FIGURE 2, but it may vary between manufacturers. The yield per egg is typically in the range of 100-300 doses.
Biologic standards for YF vaccines have been established by the WHO [14], but individual national control authorities govern approval for use and, thus, vaccine standards differ from country to country, particularly with respect to quality control test requirements. An overview of the manufacturing steps and  [6,9] quality control test requirements established by the WHO and/or found in the USA and European pharmacopeias are shown in FIGURE 2. In addition to the tests shown, YF vaccines are tested for thermostability. The lyophilized vaccines are able to withstand 14 days at 37°C with retention of minimum potency specification and loss of less than 1 log in the infectivity titer. Primary (master) and secondary (manufacturer's working) seeds must pass a safety test in rhesus or cynomolgus monkeys [14,22]. A minimum of ten monkeys are inoculated intracerebrally with 0.25 ml of virus containing 5000-50,000 mouse LD 50 . Animals are monitored for clinical signs, viremia and antibody responses. Neurovirulence is determined by measuring semiquantitative histopathologic lesion scores in brain and spinal cord. For comparison, a reference control is inoculated into ten other animals. The viremia profile (a test for viscerotropism) and histopathologic scores must not exceed those of the reference control.

Molecular characterization & the basis for attenuation
The molecular structure of YF 17D virus was elucidated 50 years after the vaccine was developed. This has led to a basic understanding of functional components of the virus and a partial understanding of the molecular coordinates of virulence, and also provided tools for rational construction of new vaccines using 17D as a vector.
The entire genome of YF 17D-204 virus was first sequenced by Rice and colleagues in 1985 [23]. Subsequently, full or partial sequences of other 17D-204 strains (including 17D-213) and 17DD vaccine have also been partially or fully sequenced [24][25][26][27]. The sequence of the parental Asibi virus [28] allowed for the initial comparisons that partially elucidated the molecular basis for attenuation of 17D vaccine. It should be emphasized that, as expected given the passage history, high mutational rates of flaviviruses and lack of plaque purification of 17D vaccines, the vaccine contains a heterogeneous mixture of virion subpopulations (i.e., a 'genetic swarm'). This has been shown by characterizing viruses isolated by plaquing for their virulence in mice [29], reactivity with monoclonal antibodies [30], and by consensus sequencing [31]. Heterogeneity in nucleotide sequence was shown for ARILVAX ® (17D-204 manufactured in the UK) at residues that differed in previously published sequences of 17D-204 and 17DD [31], many of which were performed with plaque-purified or molecularly cloned virus. This suggests that some differences across 17D strains may reflect heterogeneities (genotype mixtures) at individual residues, rather than fixed differences in amino acid sequence evolved during independent passages.
The YF 17D genome contains 10,862 nucleotides, and is composed of a short 5´-noncoding region (NCR), a single openreading frame (ORF) of 10,233 nucleotides and a 3´-NCR [23]. The ORF contains genes for three structural proteins at the 5é nd (capsid [C], premembrane [preM], and envelope [E] proteins), followed sequentially by seven nonstructural (NS) proteins, NS1, NS2A, NS2B, NS3, NS4A, NS4B, NS5 (FIGURE 3). The structural proteins, together with the viral RNA, comprise the mature virion, whereas the NS proteins are responsible for virus replication and post-translational processing of the viral proteins. Virus assembly occurs in the rough endoplasmic reticulum (ER) and virions are released by exocytosis or following cell lysis. The E-protein of YF virus contains the antigenic determinants that elicit immune responses, as well as ligands for cell receptors and multiple virulence determinants [32]. The crystal structure of the YF E-protein is predicted to be similar to that revealed for tick-borne encephalitis (TBE) virus [33] and dengue viruses [34]. The predicted 3D structure of the E-protein of 17D virus is shown in FIGURE 4, together with the mutations associated with attenuation of the 17D virus.
Comparisons of wild-type, virulent Asibi and various attenuated 17D-204 and 17DD vaccines derived from Asibi have revealed 20 amino acid differences (0.59%) and four nucleotide differences in the 3´-NCR (FIGURE 3). It is likely that attenuation of the vaccine viruses depends on multiple changes in different genes or the 3´-NCR. A higher rate of mutation occurred within the E-and NS2A-proteins than in others. As the E-protein is involved in virus entry into cells and in virion assembly and maturation, one or more of the eight amino acid differences in the Eprotein between Asibi and vaccine strains are suspected to play a role in attenuation.
Two biologic properties of 17D have been attenuated during the empirical derivation in tissue culture: neurovirulence (the capacity to replicate in brain tissue and cause inflammation, neuronal damage and encephalitis) and viscerotropism (the capacity to replicate in extraneural tissue, cause viremia and damage vital organs, including liver, kidney and heart). It has not been possible to pinpoint precisely which of the mutations, shown in FIGURES 3 & 4, determine the difference in neurotropism or viscerotropism between parental and vaccine strains. Studies of YF and other flaviviruses have identified three areas of the E-glycoprotein in which mutational changes alter virulence properties. These include the tip of the fusion domain (domain II), the hinge region between domains I and II, and the upper lateral surface of domain III containing the receptorbinding site [35]. Future studies may result in fine mapping of the 17D-specific mutations by reverting them singly or in clusters. A study of neurovirulence determinants using a chimeric virus constructed by inserting the prM-E genes of an attenuated (SA14-14-2) strain of Japanese encephalitis (JE) virus into the 17D infectious clone showed that full attenuation of neurovirulence depended on multiple genetic determinants in the E-protein [36].

Putative receptor binding region
The mutations at E305, E325 and E380 are in domain III in a region that is involved in receptor binding. E325, which is spatially close to E305 in the 3D structure of the virus, maps to a 17D-204 substrain-specific epitope. Substitutions at E305 and E325 have been shown to be neurovirulence determinants in mice [37]. A mutation at an adjacent amino acid to E325 (at E326) was associated with increased neuroinvasiveness of a 17D strain that had been adapted by serial brain passages [38]. Similarly, a mutation near E305 (at E303) occurred in a virus isolated from a fatal human case of 17D encephalitis and was shown to have increased neurovirulence in animals [39]. E380 is in the RGD motif implicated in receptor binding and mutations have been shown to alter neuroinvasiveness and neurovirulence of 17D virus [38,40,41]. 3'

Molecular hinge region
The E52 GlyArg mutation is in the hinge region between domains I and II, and thus could alter the acid-dependent conformational change of the E-protein in the endosome and reduce the efficiency of virus internalization. Mutations at this site have been shown to affect neuroinvasiveness and neurovirulence of 17D virus [38,41]. E52 is also implicated in neurovirulence of JE virus [42].

Interface of prM & E-proteins
The E173 determinant defines a wild-type epitope mapped by monoclonal antibody [43]. A ThrIle mutation occurred during the attenuation of 17D virus. This amino acid residue has been shown to affect neurovirulence and neuroinvasiveness for mice [38,43]. This mutation is in a region of the E-protein that interfaces with prM following the rearrangement from dimer to trimer after exposure to low pH in the endosome. Mutations at this location have also been shown to modify neurovirulence of other flaviviruses, including a YF chimeric virus [36] and TBE virus [44].

Stem-anchor region
A mutation at E407 is in the stem-anchor loop of the E-protein and is functionally involved in conformational changes of the E-protein during acid-induced fusion events.

Nonstructural proteins
The NS proteins are critical to the replication of flaviviruses, and subserve a range of functions including proteolytic processing of the polyprotein translated from the ORF and interactions with the 3´ stem-loop structure of the viral RNA to form the replicase complex. A total of 11 amino acid changes in the NS proteins are associated with attenuation of 17D (FIGURE 3), with the largest number of substitutions in the NS2A gene. NS2A is involved, together with other NS proteins, in formation of the replicase complex [45], and is also implicated in assembly or release of virions [46]. A complex of the N-terminal region of NS2B and NS3 constitutes the serine protease involved in post-translational processing of the flavivirus polyprotein [47]. The NS2B IleLeu mutation at NS2B109 is located in the NS2B-NS3 protease. The AspAsn change in NS3 at residue 485 is in a region of the C-terminal part of the NS3 protein that has RNA helicase and triphosphatase activities involved in unwinding RNA. The function of NS4B is not clear, but the protein, based on its localization at the subcellular level, is presumed to be involved in RNA replication and it is clearly implicated in attenuation/virulence (discussed later). NS5 contains the RNA-dependent RNA polymerase and methyltransferase enzymes involved in RNA replication. The two mutations in the NS5 protein may affect replication efficiency and contribute to attenuation of 17D [48].

3´ noncoding region
The 3´-NCR contains in order from the 3´ end: a 3´-extreme stem-and-loop structure conserved across all flaviviruses that cyclizes with the 5´ terminus and is presumed to function as a promoter for minus-strand RNA synthesis; two conserved sequence (CS) elements CS1 and CS2, which share a high degree of nucleotide sequence homology across all mosquitoborne flaviviruses; and (in the case of 17D and West African wild-type YF strains) three copies of a repeat sequence (RS) element located in the upstream portion of the NCR. Mutations may thus interfere with formation of the replicase complex or the initiation of RNA synthesis. Deletions or mutations in the 3´-NCR of dengue and TBE viruses have been shown to attenuate these viruses [49,50]. There are four nucleotide substitutions in the 3´-NCR common to 17D/17DD vaccine strains that could play a role in attenuation. Nucleotide 10,367 and 10,418 are unlikely to be involved in attenuation. Deletion of the three RS elements that included nucleotide 10,418 did not affect 17D replication in vitro [51]. The former deletion mutant also had a HindIII site engineered immediately upstream from the first RS, which changed nucleotide 10,367 and surrounding nucleotides without affecting replication. In contrast, the changes at nucleotides 10,800 and 10,847 could contribute to attenuation. These mutations are in the predicted 3´ stem-loop structure proposed to function as a promoter for minus-strand synthesis. A profound effect of the nucleotide 10,847 change on the appearance of the structure in 17D was predicted [52].
Less is known about the molecular determinants of viscerotropism than neurotropism, in part because until recently, monkeys were the only animal model available in which to test changes in virulence. A wild-type YF virus was adapted by seven serial liver-liver passages in hamsters and caused fatal hepatitis resembling human disease [53,54]. The adapted, viscerotropic virus had accumulated seven amino acid mutations [55], of which five were in the E-protein (E27, E28, E155, E323 and E331), but none corresponded to residues associated with virulence of Asibi with respect to comparisons with 17D. E323 and E331 are in the part of domain III involved in cell receptor interactions. In addition, there were two substitutions in NS proteins, NS2A and NS4B. Eight amino acid residues differentiate wild-type Asibi from 17D vaccine virus [34]. Seven amino acid residue types in parental Asibi virus are shown first, followed by the residue type of the yellow fever 17D virus. Residue A407V is not within the known crystal structure, and therefore is not shown here. Amino acids are shown by CPK representation (displays spheres sized to the van der Waals radii).
When Asibi virus was passed in HeLa cells, neurotropism and viscerotropism were rapidly attenuated after a few passages [56]. The HeLa virus had ten mutations, five in the Eprotein, one in NS2A and three in NS2B. Interestingly, an NS4B mutation at residue 95 seen in 17D and associated with hamster viscerotropism was again implicated, indicating that this determinant may play an important role. Also intriguing is that three E-protein mutations (at E27, E155, E331) were also associated with adaptation in the hamsters [5], suggesting that they may be viscerotropism determinants.

Productive infection: the early events after inoculation of YF 17D vaccine
The restricted replication and attenuation of 17D vaccine are reflected by the low levels of viremia compared with those caused by parental wild-type virus. Whereas wild-type YF virus causes high viremias in monkeys and humans, 17D virus levels detected by infectivity assays are either absent or barely detectable and do not exceed 2.3 log 10 pfu/ml. Published studies on the level and time course of viremia, measured by infectivity assays in humans inoculated with 17D virus, paint a consistent picture of low, transient viremia in a subset of subjects [57][58][59][60][61]. Reverse transcriptase polymerase chain reaction (RT-PCR) may provide a more sensitive measure of circulating virus [59]. Subjects who had no detectable viremia by plaque assay were positive by RT-PCR, and sera contained detectable viral RNA for at least 1 day longer than infectious virus. As expected, no viremia was detectable in immune subjects who were revaccinated with 17D. Viremia occurs between 4 and 6 days after inoculation and, thus, the incubation period and duration of viremia are not dissimilar to natural infection with wild-type YF virus. Titers of virus in blood are far below the infection threshold for blood-feeding mosquitoes. The very low viremia following 17D vaccine reflects markedly restricted replication in extraneural tissues, the apparent low risk of congenital infection and the low risk of neuroinvasion and postvaccinal encephalitis. Importantly, the low viremia levels in normal subjects provides a benchmark for comparison with viremia levels in subjects with yellow fever vaccine-associated viscerotropic adverse events (YEL-AVD), in which high viremia levels or prolonged viremia may be of diagnostic value.
In the authors' studies of 17D-vectored chimeric vaccines, humans and two species of macaque (rhesus and cynomolgus) were given commercial 17D vaccine (YF-VAX ® ) as an active control. Of 36 human subjects included in several clinical studies, 50% developed detectable viremia, with a mean peak level of 1.7 log 10 vero cell pfu/ml and a typical time course (FIGURE 5). In contrast, virus appeared earlier and reached higher levels in monkeys, with rhesus sustaining a more active infection than cynomolgus macaques. This pattern precisely mirrors the infection of monkeys and humans with wild-type YF virus, in which rhesus are the most susceptible species and develop a fulminating infection (FIGURE 6) [62]. Monkeys are more efficient viremic hosts than humans in transmission cycles of wild-type YF virus. The higher susceptibility of monkeys than humans to YF viruses is also important in considering how testing might be performed for safety and immunogenicity of new live YF vaccines or vaccines employing 17D as a vector, as the animals would be expected to reveal a more active infection than humans.

Sites of replication in vivo
The in vivo sites of replication of 17D virus in humans have not been determined. In vitro, 17D virus replicates in myeloid dendritic cells [63] and, as demonstrated for dengue [64], dendritic cells expressing DC-SIGN in the skin may be early sites of replication after subcutaneous inoculation. Indeed, 17D vaccine was successfully administered by percutaneous administration (scarification) during field trials in Africa [65]. Not unexpectedly, a study of recombinant 17D vaccine administered to monkeys by the epidermal route resulted in viremia and immunity [66]. Undoubtedly, subcutaneous vaccination with 17D exposes these cells in the epidermis to the viral inoculum. The predilection of virus for Langerhans cells is logical, given the adaptation of flaviviruses to infect via saliva of hematophagous insects and ticks deposited in the epidermis during probing and feeding. It is likely that draining lymph nodes, to which the activated Langerhans cells home under the control of interleukin (IL)-1β, are the next sites of replication [67]. Virus reaches other organs via the lymphatic drainage and bloodstream. In cynomolgus macaques that were perfused to remove blood from the tissues, small amounts of 17D virus were found by RT-PCR in the skin at the site of inoculation, draining axillary lymph nodes and mesenteric lymph nodes at the time of peak viremia (day 3), and in a wider assortment of tissues on day 7 including spleen, liver, thymus, adrenal gland and bone marrow. By day 14, virus was still present in lymph nodes and spleen. No virus was detectable at the last sampling point (day 46) [68]. Virus was present in tissues, despite the appearance of neutralizing antibodies in 60% of the animals on day 7 and 90% of the animals on day 14. These data indicate that 17D virus, although restricted in its capacity to replicate in visceral organs, has a tissue tropism not unlike that of the parental wild-type virus [69]. This is an important observation as, under circumstances in which replication is unrestrained (as in YEL-AVD), the virus appears to target the same array of tissues as wild-type virus.

Stimulation of innate immunity
YF 17D vaccine is believed to stimulate a strong innate immune response. This may result in a transient state of resistance to superinfection with other viral agents administered within a short interval, both homologous (e.g., in monkeys sequentially infected with 17D and wild-type YF [70]) or heterologous flaviviruses (e.g., in experimental sequential infection with 17D and wild-type dengue virus [71]), or a completely unrelated virus (e.g., a member of the Reoviridae [72]). Monkeys treated with poly(I)-poly(C), which induces interferon (IFN)-α, were protected against lethal challenge with YF virus [73]. An important component of the innate immune responses to many virus infections, including flaviviruses, is the activation of IFN regulatory factor (IRF)-3, leading to the induction of IFNβ. Activation of IRF-3 occurs via two signaling pathways, the most important being Toll-like receptor (TLR)3, which recognizes double-stranded RNA (an intermediate step in flavivirus replication). IRF-3 is phosphorylated and translocated to the nucleus and transcribed, leading to the induction of IFN and IFN receptor-mediated signal transduction and expression of multiple genes involved in antiviral activities. Activation of IRF-3 has been demonstrated in West Nile-infected cells [74], and induction of Type 1 IFN [75] and the IFN-stimulated gene product, 2´5´-oligoA synthetase [76], have been demonstrated in humans following administration of 17D vaccine. Activation of natural killer cells also mediated by TLR recognition of viral RNAs results in the induction of IFN-γ and proinflammatory cytokines. IFN-γ was shown to mediate an antiviral state in monkeys challenged with YF [77].
Based on the expanding interest in TLR triggering of innate immunity via transcriptional induction of multiple genes, it is predictable that the rudimentary state of knowledge about flaviviruses will rapidly improve and that 17D vaccine will play a central role in the exploration of these responses in humans. The recent demonstration that the serine protease of hepatitis C (equivalent of NS2B-NS3 in 17D) inhibits TLR3mediated TRF-3 activation indicates that flavivirus proteins may downregulate innate immunity [78]. The review by King and colleagues provides an additional insight into how early events in the infection cycle may downregulate or evade the host response to flavivirus infection [79].
In addition to limiting replication of 17D virus in the host, crossregulation between the innate and adaptive immune response plays an important role in shaping the adaptive immune response. The potent innate response to live 17D vaccine may be responsible for the rapid and long-lasting antibody and T-cell responses that characterize this vaccine.

Adaptive immune response
Neutralizing antibodies elicited by 17D vaccine are considered the principal mediator of protection against future infection and are exceptionally durable. Neutralizing antibodies encounter virus at the earliest stage of infection, deposited in the skin by a blood-feeding mosquito. The role of neutralizing antibodies in  protection was demonstrated by transferring serum from immune monkeys before or shortly after challenge with wildtype YF virus [80,81] and in passively immunized mice challenged intracerebrally with 17D virus [82]. Epitopes responsible for inducing neutralizing antibodies reside in the E-glycoprotein [80][81][82][83]. Antibodies against 17D strain-specific, virus-specific and flavivirus group-reactive epitopes neutralize virus and, when administered to mice, have been demonstrated to protect against intracerebral challenge.
Studies with another flavivirus using naked DNA to elicit immunity only to the E-protein in B-and T-cell knockout mice suggested that protection is mediated by antibodies and not by CD8 + T-cells to the E-protein [84]. Immune responses to NS proteins, including complement-dependent cytolytic antibodies to NS1 [85,86] and cellular responses to NS3 and other proteins, may also play a role in protection against future YF infection, but these mechanisms are not well defined, particularly in humans. Further reference to cellular immunity is made later in this review.
Vaccination with 17D virus is followed by the rapid appearance of neutralizing antibodies. In recent large vaccine trials, 99% of 1440 adults [87] and 93% of 1107 children [88] developed neutralizing antibody responses, with mean log 10 neutralization indices (LNIs) of approximately 2.0 in adults and 1.3 in children. The reasons for the lower response in children are unclear. A recent bridging study in 288 subjects compared vaccines produced in France (Stamaril ® , sanofi pasteur), Germany (Robert Koch Institute [2]), and Switzerland (Flavimune ® , Berna) [89]. Neutralizing antibody response rates were 100%, with geometric mean plaque-reduction neutralization titers 1 month after inoculation varying between 612 and 1184.
Antibodies appear several days earlier in monkeys (days 6-7) than in humans (days 10-14), as would be expected based on the kinetics of the viremic responses (FIGURE 5). In one study, 86-88% of adult human subjects developed neutralizing antibodies by day 14, whereas 99-100% were seropositive on day 28 after vaccination [90]. The International Health Regulations state that the vaccination certificate for YF is valid 10 days after administration of 17D vaccine.
The primary response to 17D vaccine in humans is characterized by an early response involving immunoglobulin (Ig)M neutralizing antibodies, which appear between days 4 and 7, several days before detection of IgG antibodies, and may provide an element of early protection [91]. IgM neutralizing antibodies were 16-to 256-fold higher than IgG antibodies during the first month after immunization and persisted for many months and possibly years. IgM antibodies (against West Nile virus) were shown to confer protection against encephalitis in mice [92].
The mechanism by which neutralizing antibodies protect cells from flavivirus infection is not completely understood. In vitro studies with West Nile and JE viruses indicate that neutralizing antibodies can interfere with fusion and virion internalization from the endosome [93,94].
The redundancy of neutralizing epitopes in the E-protein ensures that antibodies protect broadly against wild YF virus strains, which may differ slightly in their amino acid sequences. Indeed, while YF virus strains found in both continents represent a single serotype defined by neutralization, it has long been known that African and South American strains can be distinguished antigenically [95], and that antibodies induced by 17D vaccine have higher homologous titers than those against heterologous YF strains, including parental Asibi virus. However, the natural evolution of wild-type YF viruses, which are classified into seven distinct genotypes (five in Africa and two in South America) based on nucleotide sequencing [96], is not accompanied by antigenic changes sufficient to escape immunity induced by 17D virus.
Various methods may be applied to the measurement of neutralizing antibodies. The constant serum-varying virus neutralization test has been standardized and the cut-off for a positive test (LNI 0.7) established by showing that it correlates to protection against challenge in monkeys [97]. For measurement of the immune response in an individual, the difference in titer between pre-and postvaccination sera expressed as the LNI represents the neutralizing capacity of the serum. This quantity of virus neutralized in undiluted or minimally diluted serum may be more biologically relevant as a serum dilution endpoint titer. However, the test is slightly less sensitive than the serum dilution-constant virus plaque-reduction method. The LNI has been validated and used to support pivotal trials of 17D vaccines [87].
The hemagglutination inhibition (HI) test is less sensitive than neutralization and has little value for measuring vaccineinduced responses. The complement-fixation (CF) test is rarely used today, but has some value as individuals without prior flavivirus exposure do not develop CF antibodies to 17D vaccination [98]. IgG ELISA is insensitive for the detection of primary response to 17D vaccine, but may be somewhat better for detecting responses after revaccination. An interesting approach that has been rarely studied is the measurement of antibodies to NS proteins. This could be of particular value in distinguishing artificial 17D immunization from wild-type YF infection (as for CF antibody), or in the case of chimeric vaccines in which the backbone vector is 17D virus, when one wishes to distinguish vaccine immunity from natural infection with the wild-type virus corresponding to the prM-E genes being carried by the vector. NS1 antibodies do not appear after primary 17D vaccination (YOUNG P, GUIRA-KHOO F, MONATH T, UNPUBLISHED DATA), but there are no data on wild-type YF. Similarly, although no data are currently available on antibodies to NS5 following YF vaccination or infection, NS5 antibody has been useful in the diagnosis of West Nile infection [99].
The measurement of antibodies has been helpful in our understanding of the serious viscerotropic adverse events that occasionally follow 17D vaccine (see Safety & tolerability section). Some survivors of these events have shown very high titers of YF neutralizing antibodies [100], suggesting that the antigenic mass following these severe infections surpassed the usual more limited infection in typical vaccination. Although there are no data as yet, it is likely that people with severe viscerotropic reactions will mount a more active antibody and T-cell response to NS antigens, and that anti-NS1 (for example) might be a useful diagnostic test in these patients. There is a real need for noninvasive tests distinguishing viscerotropic accidents from other adverse reactions, and a systematic accumulation of comparative virologic and serologic data on normal and YEL-AVD patients.

Revaccination
The YF vaccination certificate for international travel is valid for 10 years, after which revaccination is necessary. However, studies of individuals vaccinated up to 35 years before testing have shown that neutralizing antibodies at protective levels are probably lifelong in the vast majority of people [101]. It is of interest that revaccination was not associated with an IgM response when the interval between primary vaccination and revaccination was short (2 years) [102], whereas subjects revaccinated after more than 10 years developed IgM antibodies [59], suggesting that some memory B-cells may senesce over time.
Since most individuals that are revaccinated have preformed neutralizing antibodies, it is not surprising that the response to revaccination with a live vaccine is blunted compared with primary immunization, and that a booster response is more likely in people with a baseline low neutralizing antibody titer [103,104].

Flavivirus interactions
Shared, crossreactive antigenic determinants between YF and other flaviviruses invariably result in a dramatic broadening of the immune response to 17D in subjects who are later infected with a heterologous flavivirus or in people who were previously infected with another flavivirus and are subsequently vaccinated with 17D [105]. The prior infection elicits memory responses of IgG-producing B-cells and restimulates CD4 + helper cells and follicular dendritic cells, so that an accelerated and crossreactive response is observed. Thus, in people previously immunized with 17D and subsequently given TBE vaccine, TBE antibodies appeared earlier, at higher titers and demonstrated more crossreactivity with heterologous flavivirus antigens than in nonimmune subjects [106]. Indeed, there was considerable interest several years ago in using a sequential immunization scheme with a limited array of flavivirus antigens to stimulate immunity more broadly to multiple flaviviruses.
The possibility also exists that prior infection with a heterologous flavivirus could crossprotect and thus interfere with 17D vaccination. Experimental studies in monkeys showed that certain flaviviruses, including dengue, Wesselsbron and Zika viruses, crossprotected animals against infection with wild-type YF virus [107,108]. One early study suggested that prior dengue, but not St Louis encephalitis immunity, interfered with 17D vaccine, but more extensive vaccine trials did not demonstrate interference by dengue antibody [88]. In people with antibodies to JE or broadly crossreactive flavivirus antibodies, there was no interference with 17D immunization and no alteration of viremia [103,109].

Dose-response
Dose requirements for YF 17D vaccine were established in the 1930s and 40s, based on limited clinical trials and experience with lots that varied in titer. Fox and colleagues, in 1943, found that 70% of subjects responded to a dose of 14 mouse median lethal doses (MLD 50 ) and all responded at 140 MLD 50 [13]. The minimal potency requirements were eventually set at 1000 MLD 50 or the number of pfu in cell culture shown by the manufacturer to be equivalent to that dose. An excess (typically 0.7 log) is added to ensure that the minimum potency specification is met at the end of shelf life. Manufacturers have sometimes experienced difficulty in validating the ratio of MLD 50 to pfu, and the WHO is currently in the process of establishing reference standards based on cell culture assays. In general, there are approximately 5-8 Vero cell pfu/MLD 50 when weanling outbred mice are used in the assay. Recent dose-response studies indicate that doses of as little as 100-200 pfu result in seroconversion of more than 90% of persons vaccinated [110,111]. Thus, the minimum potency requirements (3.0 log 10 MLD 50 = 3.7 log 10 pfu) set by the WHO exceeds the 90% immunizing dose by 50-fold or more.
A consistent, if subtle, finding has been an inverse relationship between dose and viremia and antibody levels. This has been observed in humans and nonhuman primates with 17D vaccines [11,111], and the author has reported similar findings with chimeric viruses utilizing YF as a vector for the envelope genes of heterologous flaviviruses [104]. Historically, there has been some concern that a low dose might delay the onset of immunity and enhance viremia and the risk of neuroinvasion, but limited experience in settings in which low doses have been intentionally administered (e.g., for delivery of 17D by scarification or intradermal inoculation during skin testing of persons with egg allergy) have not revealed a safety issue. The reason for the inverse dose-response is uncertain, but may be due to induction of a robust innate immune response by high doses compared with low doses, or possibly due to a higher concentration of defective interfering virus. The mechanism involved is certainly deserving of study.
YF 17D-specific T-cell responses have been measured by cytotoxic T-cell assay, IFN-γ ELISPOT and lymphoproliferation [113]. In recent studies conducted at Emory University (Atlanta, GA, USA), CD8 + T-cell responses to 17D vaccine were studied in healthy adults (J MILLER & R AHMED, PERS. COMM., 2005). At 2 weeks following inoculation, maximum expansion of a well-defined pool of effector CD8 + T-cells was observed, characterized by the expression of the T-cell activation markers, CD38 and HLA-DR. At the peak of the CD8 + T-cell response, between 4 and 13% of total CD8 + T-cells coexpressed CD38 and HLA-DR. Additional phenotypic analysis at the peak of the response showed that activated CD8 + T-cells also expressed Ki-67, perforin and granzyme B, and downregulated Bcl-2 molecules, implying that responding T-cells proliferate, acquire cytolytic potential and are highly susceptible to apoptosis. From 2-4 weeks post inoculation, virus-specific effector populations contracted five-to tenfold and downregulated expression of activation markers. By stimulating PBMC with 17D virus-infected cell lines, functional 17D-specific CD8 T-cells were detected that exhibited similar response kinetics to CD8 + T-cells coexpressing CD38 and HLA-DR. In additional experiments monitoring the activation status of cytomegalovirus, Epstein-Barr and influenza-specific CD8 + T-cells in subjects inoculated with YF vaccine showed that bystander activation does not contribute to the peak of the anti-17D CD8 + T-cell response. This observation suggested that the observed T-cell responses were primarily 17D specific.

Durability of the immune response
One of the most remarkable characteristics of 17D vaccine is the long persistence of neutralizing antibodies following a single inoculation. In various studies evaluating neutralizing antibodies 10-35 years after vaccination, 75-97% of subjects without a likelihood of natural exposure to YF were seropositive [101]. Long-term studies of cellular responses have not been performed. However, in one recent study, functional 17D-specific CD8 + T-cells were found 1 year after vaccination, showing that the vaccine induced long-term CD8 + T-cell memory (J MILLER & R AHMED, PERS. COMM., 2005) or, possibly, that there is persistent antigenic stimulation.
The mechanisms underlying the long-term durability of immunity are uncertain, but it is likely that the very strong innate immune response to 17D shapes the durable adaptive response. Chronic persistent infection or storage of antigen in vivo, such as in follicular dendritic cells, with recurring antigenic stimulation is also a possible explanation for the durability of YF immunity. This was suggested by the prolonged synthesis of IgM antibodies in persons immunized with 17D virus [91]. Reports of chronic infection with YF virus in vivo include the recovery of virus from brains of mice [114] and rhesus monkeys several months after inoculation [115]. These observations are consistent with other reports in the literature of chronic, persistent infection of monkeys and humans with various other flaviviruses, including TBE, West Nile and JE.

Protective efficacy
Studies in nonhuman primates have demonstrated protection afforded by 17D vaccine against lethal challenge with wildtype YF virus [97,116]. Although controlled clinical trials have never been performed, the efficacy of 17D vaccine has been illustrated by accumulated experience over the past 60 years. Wherever vaccine coverage rates are high, the incidence of YF has declined. Very few instances of primary vaccine failure have been documented. If new YF vaccine(s) are developed, licensure would need to be based on immunologic surrogates (neutralizing antibodies) in noninferiority trials or, depending on the nature of the vaccine, by use of the exclusionary rule showing protection in two animal models.
Nevertheless, effectiveness and safety studies of existing YF vaccines would provide important information in populations with a high background of HIV infection, in which vaccine efficacy may be diminished and vaccine safety may be in question. As vaccine manufacturers are unlikely to undertake such studies, they will be left to government agencies using the vaccine, particularly in the setting of outbreak control.

Safety & tolerability Commonly reported, expected adverse reactions to YF 17D
Early studies performed in the 1930s in Brazil showed that 17D vaccine was well tolerated. Smith and colleagues followedup 2457 vaccinees, finding 14.6% with headache for 1-2 days and 10.2% with pains in the body; 1.4% missed time from work (usually only for 1 day) [8]. These reactions, which were generally mild, occurred 5-7 days after immunization and thus fit the expected temporal evolution of the vaccine infection. Many other studies have been performed over the years, with varying methodology and degrees of rigor with respect to capturing adverse events, and have supported the conclusion that 17D vaccines are associated with a relatively low incidence of mild local and systemic side effects during the first week after inoculation [87][88][89][90]110]. However, none of these studies have included a placebo control, rendering interpretation difficult.
Recently, two large studies using diary cards and frequent follow-up clinic visits to assess safety and tolerability in adults [87] and children [88] have been reported. These trials were designed to compare the licensed vaccine in the USA (YF-VAX ® ) with the vaccine manufactured in the UK (ARILVAX ® ). Both vaccines belong to the same substrain (17D-204), and differ by two amino acid changes associated with the 14 passages that separate the vaccines from a common ancestor [31]. In adults, common side effects were asthenia (fatigue), malaise, headache, myalgia and injection-site reactions, which were generally mild and transient (FIGURE 7). YF-VAX differed statistically from ARILVAX only in local reactogenicity, which may reflect either the level of impurities in the vaccine or higher dose. In children, who were less able to report certain systemic effects, principal treatment-emergent adverse events were fever, gastrointestinal and upper respiratory symptoms. Interestingly, local reactions were few, indicating perhaps that adults are more likely to be sensitized to egg protein or the gelatin excipient in the vaccine, and that local reactions in adults are due to an allergic response rather than the virus itself. Since there were no placebo controls, the adverse events in children appear likely due to intercurrent infections, and as many of the events reported by adults (e.g., headaches and fatigue) are commonly reported in clinical trials of any investigational drug, it is difficult to attribute causality in these trials.
One very small placebo-controlled study was conducted as part of a clinical trial involving other treatment groups receiving chimeric YF-JE vaccine [61]. Although the sample size is too limited for definitive analysis, the results do shed some light on the problem of interpreting data in the uncontrolled studies. Systemic symptoms, such as headache, myalgia and fatigue, were commonly reported by placebo recipients, and the only adverse experience that was clearly related to 17D vaccine was injection site erythema.

Significant & serious adverse events YF vaccine-associated neurotropic adverse events
The predominant type of neurologic accident associated with 17D vaccines is acute meningoencephalitis caused by invasion of the CNS and direct viral injury, analogous to the encephalitis produced by intracerebral injection of mice and monkeys. The median onset of clinical signs is 11 days after vaccination (range 2-23 days). The diagnosis of VF vaccineassociated viscerotropic disease (YEL-AND) is made on the basis of temporal association to 17D vaccination, the clinical picture and, most importantly, the finding of IgM antibody in the cerebrospinal fluid. With one exception, all such cases have been in subjects undergoing primary immunization. The acute disease is usually self-limited, although long-term sequelae have occasionally been reported (e.g., a 29-year-old with mild ataxia 11 months after onset [117]), but careful follow-up has not been described in most cases. YEL-AND caused by acute brain infection occurs at a higher incidence in infants under 12 months of age, estimated to be 0.5-4 per 1000 infants [118,119], whereas the incidence in older people is estimated to be in the order of one in 2.5 to 16 per million. The majority of published reports of YEL-AND occurred during the 1950s, before restrictions were imposed on vaccination of young infants. Of 23 cases reported between 1945 and 2001, 13 (57%) were in infants less than 4 months of age and 16 (70%) in those less than 9 months of age. In the 1960s, recommendations against immunization of infants under 6 months led to a reduction in such reports. The increased risk of encephalitis in human infants mirrors the increased susceptibility of suckling mice to neuroinvasion and neurovirulence of 17D virus and other flaviviruses, and may be due to immaturity of the blood-brain barrier, prolonged or higher viremia and/or immunologic immaturity and delayed viral clearance. There are no data on 17D viremia levels or on the kinetics of the immune response in infants compared with adults, but it is interesting that the mean neutralizing antibody titers in children are significantly lower than in adults [88], implying that a delayed or lower immune response and viral clearance may be important factors in age-related susceptibility.

Figure 7. Treatment-emergent adverse events in a pivotal trial in healthy adults and children 9 months to 10 years comparing two yellow fever 17D vaccines (YF-VAX® and ARILVAX™).
Data from [87] for adults and [88]  The only YEL-AND case report in the USA from 1945-2001, and one of only two fatal cases in the literature, was a 3-year-old child who died in 1965 [120]. This case is interesting because it represents the only instance in which the virus was recovered from brain tissue. The virus isolate contained three point mutations determined by consensus sequencing and had increased neurovirulence for mice and monkeys [39]. One mutation was in domain III at E303 Glu Lys, and was near a known virulence determinant at E305. Two other mutations (at E155 and NS4B72) were also found. The mutation at E155 was probably not responsible, because some attenuated vaccine strains have a wild-type residue at this locus and also because a neutralization escape mutant of 17D at E155 did not show any change in neurovirulence [39,121]. The child had no known risk factors for potentiated infection. It is uncertain whether the mutations in the brain isolate were the result of selection of a neurovirulent subpopulation present in the 17D vaccine, in vivo mutation in extraneural tissues that increased neuroinvasiveness or mutation of the virus during replication in the brain.
Retrospective reviews of VAERS data that have been conducted back to 1990 In addition, beginning in 2001, the Centers for Disease Control (CDC) and US Food and Drug Administration (FDA) have formed a working group to scrutinize Vaccine Adverse Events Reporting System (VAERS) data for 17D vaccine-associated adverse events and to obtain additional clinical and diagnostic information [122]. As a result, a total of 11 cases of YEL-AND with clinical signs and a temporal relationship to 17D meeting an agreed case definition have been identified among US citizens during the 1990-2004 interval. Between June 2001 and August 2002, four cases of acute encephalitis were found, all in adults [123]. As the number of doses sold to civilian travelers was approximately 250,000 during this interval, the incidence of YEL-AND may be estimated as 1:62,500 (16 per million), about tenfold higher than vaccine-associated poliomyelitis. Of the 11 cases found between 1990 and 2004, four had typical postvaccinal encephalitis, four had Guillain-Barré syndrome and three had acute demyelinating syndromes. A confounding factor in the evaluation of some cases was the administration of other vaccines, such as tetanus, typhoid and hepatitis A; although a role for these vaccines in the neurologic accidents cannot be ruled out, they are regarded as far less likely causative agents based on accumulated safety data. In Europe, the incidence of YEL-AND was estimated from spontaneous safety reports and the number of ARILVAX doses sold in Europe during the period from 1991 to 2003. The incidence from 1991 to 1996 was 1.3 per million, and between 1996 and 2003, 2.5 per million [124].
Other neurologic syndromes have been associated with 17D vaccine, both in the literature and postmarketing surveillance reports. These include Guillain-Barré syndrome [125], bulbar palsy, Bell's palsy, optic neuritis [126] and other mononeuritides. Although not well characterized, these adverse events, the reports of Guillain-Barré syndrome and acute demyelinating disease suggest that 17D vaccine may elicit autoimmune neurologic reactions. Whether these autoimmune events require replication of the virus in nervous tissue in order to trigger an adverse immune response is uncertain. Data are needed on IgM antibodies in the cerebrospinal fluid of cases of YEL-AND of suspected immunopathologic origin.
All reported cases are associated with use of vaccines produced with the 17D-204 substrain. Interestingly, the 17DD vaccine produced in Brazil is characterized by a slightly higher neurovirulence profile than 17D-204 when tested in mice or monkeys [127]. The absence of reported cases of YEL-AND in Brazil may be the result of insensitive adverse event reporting, the higher background of dengue and other crossprotective flavivirus immunity, or the primary immunization of young infants having some residual (protective) maternal antibody.

YF vaccine-associated viscerotropic adverse events
This rare complication of 17D vaccines has emerged since 1996. The incidence of YEL-AVD has been estimated to be 2.5 per million [124]. YEL-AVD is a severe acute illness with a short (2-5 day) incubation period characterized by hepatitis, multi-organ failure and a high lethality. The clinical syndrome resembled wild-type YF in many respects, with viral antigen in multiple tissues, hepatitis, midzonal necrosis of the liver, Councilman bodies, renal failure and shock. Cases have been associated with vaccines manufactured in Brazil (17DD substrain), France (17D-204), the USA (17D-204) and the local Chinese vaccine. A total of 23 cases of YEL-AVD have been reported to date, the majority in international travelers. The clinical and diagnostic criteria applied to these cases have not been uniform, and the CDC-FDA Working Group are developing case definitions that (when published) should improve reporting. As noted below, these rare events appear to be related to genetic susceptibility of a small proportion of individuals to unrestrained 17D virus infections.
The reason for the emergence of YEL-AVD after 1996 has been a matter of speculation. One theory is that prior to 1996, immune serum globulin (ISG) was concurrently administered with 17D vaccine to travelers for prevention of hepatitis A, whereas after 1996, hepatitis A vaccines have been used. ISG contains high titers of YF-neutralizing antibodies [128], as 5-10% of plasma donors who contribute to pooled plasma used for immunoglobulin production have been vaccinated during prior military service: current lots of intravenous immunoglobulin (IGIV) also contain high titers of antibody. Thus, travelers before 1996 were receiving active immunization while being passively immunized with preformed antibody. The latter would be expected to protect against severe infection. Active-passive immunization was performed prior to development of 17D vaccine, when the available vaccine candidates were incompletely attenuated [129]. Vaccine-associated viscerotropic adverse events can occur only in the setting of primary vaccination of people without pre-existing YF immunity.
From 1996 through to 2003, nine cases (six fatal) of YEL-AVD have been reported following administration of 17D-204 vaccine in the USA [100,122,123,130,131]. Most US cases were elderly (median age 67 years; range 22-79 years), and had onset 2-5 days after vaccination, multi-organ failure, shock, acute respiratory failure, hepatitis, lymphopenia and thrombocytopenia. Two patients also had signs of encephalitis. 17D virus was isolated from the blood of two cases, but most revealing, YF antigen was found by immunohistochemistry (IHC) in the liver in one person and in multiple tissues (lung, lymph node, spleen, heart, liver and muscle) in two fatal cases, indicating that 17D virus was responsible for the disease syndrome. Although no data on the biodistribution of 17D in normal subjects are available, the data from monkeys (referred to previously) suggests that lymphoid tissue, but not critical viscera, are involved in replication. The subjects affected by YEL-AVD had very high neutralizing antibody titers compared with typical vaccinees, eliminating immunosuppression as a susceptibility factor and suggesting that an overwhelming infection (high antigenic mass) led to an exaggerated antibody response.
A total of 14 additional cases have been reported outside the USA [131][132][133][134]. The first three cases were recognized in Brazil and Australia between 2000 and 2001. A consistent pattern emerged of histopathologic changes in the liver, characterized by midzonal necrosis and viral antigen detected by IHC. 17D virus strains isolated from tissues were characterized phenotypically and by consensus sequencing. These studies essentially ruled out mutational virus-specified factors as being responsible for the adverse events [135]. A strain recovered from a fatal case of 17DD YEL-AVD showed no alteration of its attenuated phenotype in monkeys (R MARCHEVSKY, UNPUBLISHED DATA) or hamsters [136] inoculated by the intrahepatic route.
A genetic basis for increased susceptibility is suspected to underlie the pathogenesis in many cases. Two patients (one confirmed) in a single family have been recorded in Brazil. Genetic resistance of mice to 17D virus was first demonstrated in 1931 [137]. This phenomenon has attracted many investigations over the years [138], and the flavivirus susceptibility gene (Flv) in mice was ultimately shown to be a member of the 2´5´-oligoA synthetase family. Given this, it is likely that the susceptibility of humans to severe 17D infection may also be controlled by genes determining early (innate-immune) antiviral defenses. However, in the absence of any immunologic evaluation of cases with YEL-AVD to assess transcriptional activation of cellular antiviral genes, IFN and other cytokine levels and T-and B-cell function, the underlying pathogenesis remains unclear.
Two acquired host factors for YEL-AVD have been identified: advanced age and thymic disease. The role of advanced age as a risk factor for serious adverse events has been the subject of concern in the USA [100,123]. A retrospective analysis of VAERS data revealed a higher incidence of serious adverse events (neurologic or multisystem involvement) to 17D vaccine in elderly people, with people over 65 years of age having a risk 12-32-times higher than adults 25-44 years of age, suggesting the possibility that senescence of immune responsiveness may play a role [100].
A similar conclusion was reached after analysis of postmarketing surveillance data for ARILVAX in the UK (M CETRON, TP MONATH, UNPUBLISHED DATA) and Stamaril ® in Australia [139]. Although the cases of vaccine-associated viscerotropic events in Brazil were in children and young adults [132], a predilection of severe reactions in the elderly would not be apparent in Brazil, since older people have been previously vaccinated.
A history of thymectomy, thymoma or related conditions (e.g., myasthenia gravis and DiGeorge syndrome) has emerged as an important risk factor for YEL-AVD [140]. The YF-VAX label has now been amended to include thymic disease as a contraindication to vaccination. Of 23 patients with YEL-AVD reported through 2003, four (17%) had undergone thymectomy. As the incidence of thymomas is only 1.5 per million person-years in the USA [141] and only about ten adult thymectomized people would be vaccinated annually, the relative risk of acquiring YEL-AVD if thymectomized approximates 1.5 × 10 5 . Thymoma is associated with a variety of immunologic deficits in adults and susceptibility to viral (and other) infections, including Good's syndrome [142] and severe T-and B-cell deficiency [143]. Interestingly, a patient with myasthenia gravis had high titers of neutralizing antibodies to type-1 IFN [144]. In adult mice, administration of antithymocyte serum increased lethality of 17D virus infections [145].

Hypersensitivity reactions
Allergic reactions to 17D vaccine were first reported during large-scale immunization of military personnel in World War II. These reactions were characterized as either immediate (anaphylaxis), delayed serum sickness syndrome occurring 3 and 7 days after vaccination, or erythema multiforme. Most of these cases were in people with known egg allergies. A history of allergy (oral intolerance) to eggs has for many years been a contraindication to administration of 17D vaccines. People with a general history of allergies/atopy may be at higher risk of allergic reactions to 17D [146]. Egg-sensitive patients should undergo allergy testing with diluted vaccine and, if necessary, may undergo a desensitization procedure (described in the YF-VAX ® package insert). Persons without a history of egg allergy may occasionally develop anaphylactic reactions to 17D vaccine, and hydrolyzed porcine gelatin (used as a stabilizer in 17D-204 vaccines) may be responsible. In one interesting report, a history of intolerance to foods containing raw but not cooked eggs was elicited, and the patient had a negative skin test to cooked eggs, but positive tests to raw egg antigen and 17D vaccine [147]. As few foods contain raw eggs and these are mixed with other ingredients, people sensitized to raw eggs might not give an allergic history when queried prior to vaccination.
VAERS data collected between 1990 and 1997 revealed 45 cases of nonfatal hypersensitivity-type reactions (urticaria, angioedema, bronchospasm and anaphylaxis) temporally associated with YF vaccine. Based on the number of vaccine doses sold during the interval, the incidence of allergic reactions was estimated to be one in 131,000 [148].

Pregnancy & lactation
There is no evidence that 17D vaccine (or for that matter, the wild-type virus) causes severe infection of the fetus or stillbirth or birth defects. Manufacturers' postmarketing safety reports do contain cases of congenital anomalies, but it is not possible to assign causality, as such anomalies occur in the general population.
Recent studies in which 200-300 women were inadvertently vaccinated during pregnancy, and cord bloods examined in about 80 cases for IgM antibody, found no evidence for untoward effects and a low frequency of inapparent congenital infection (1-2%) [149,150]. The WHO and Advisory Committee on Immunization Practices (ACIP) recommend that pregnant women be vaccinated only if there is a high risk of exposure to wild-type virus. Inadvertent immunization is definitely not an indication for therapeutic abortion. Immunization during pregnancy appears to be relatively ineffective, a conclusion based on a single study showing a significantly lower seroconversion rate (39%), compared with nonpregnant subjects in an age-matched control group [149]. This suggests that women who are intentionally vaccinated during pregnancy should either be tested to ensure serologic response or warned that they may not be effectively protected and revaccinated after parturition.
With regard to the vaccination of lactating mothers, there are no data suggesting that 17D virus is transmitted to infants via breast milk. Nevertheless, a hypothetical concern is founded on analogies to other flaviviruses, such as West Nile (rare mother-infant transmission) TBE and Alkurma viruses, which are transmissible to humans orally via milk of infected goats and sheep. The ACIP recommends caution against use of 17D vaccine in breast-feeding mothers, while the product label in the UK states that vaccination is contraindicated [151]. Again, the judgment must be made based on an assessment of risk and benefit.

Vaccine interactions
A number of studies have been performed on simultaneous or sequential administration of 17D vaccine with other vaccines, including vaccinia, diphtheria-pertussis-tetanus, bacillus Calmette-Guerin (BCG), measles, live and subunit typhoid, live and inactivated cholera, hepatitis A, hepatitis B and meningococcus A/C plus typhoid vaccines. In the case of live viral vaccines, such as measles, it is recommended that they be administered concurrently at different sites or spaced 30 days apart to avoid any interference effects induced by innate immunity. A complete analysis of vaccine interactions is beyond the scope of this review, and the interested reader is referred elsewhere [6]. The most important area for future research will be on the interactions between 17D vaccine and new live attenuated vaccinates being developed against other flaviviruses: dengue, JE and West Nile. As discussed below, 17D virus is used as a live vector for delivery of E-genes of these target flaviviruses. Consequently, pre-existing immunity to NS proteins of YF could interfere with immunization by the chimeric vaccines or, more importantly, with the intentional vaccination against YF in individuals who were first given the chimeras. To date, however, the available data, albeit limited, suggest that such interactions are not problematic and that enhancement of immunity rather than interference may result from sequential administration.

Use of 17D virus as a live vector
This subject has been reviewed in several recent publications, including a recent issue of Expert Review of Vaccines [152][153][154]. Briefly, a full-length cDNA clone of 17D was used to construct chimeric flaviviruses in which the prM-E structural protein genes of 17D were replaced with the corresponding genes of the target virus. The full-length chimeric cDNA is transcribed to RNA, which is transfected into an appropriate cell culture for vaccine production. This strategy has led to the rapid development of 17D-based live attenuated vaccines against dengue, JE and West Nile viruses. The advantage of the approach, which distinguishes it from other live vector systems, is that replacement of the E-gene (which contains neutralization determinants) eliminates the problem of antivector immunity. The 17D backbone and the chimerization process confer a strong attenuating effect on the vaccine candidates as well as an important biologic feature -the inability to infect mosquito vectors. However, for the JE and West Nile chimeric vaccines, which contain E-genes from neurotropic viruses, it has been necessary to ensure a fully attenuated phenotype by using either donor genes from a vaccine strain (in the case of the JE chimera [155]) or to introduce attenuating mutations in the E-gene (e.g., for West Nile chimeras [156]). Encouraging results have been reported from a number of preclinical and clinical studies with vaccines against dengue, JE and West Nile. The product profile of the chimeras appears to mirror that of the vector (i.e., minimal side effects, more than 95% seroconversion and long-lasting antibody responses). The chimeric vaccine against JE will enter Phase III trials in 2005.

Expert commentary
Until about 5 years ago, YF 17D was often cited as the safest and most effective live attenuated viral vaccine ever developed. A more conservative approach has been mandated by the recognition of YEL-AVD as a rare, but life-threatening serious adverse reaction to the vaccine. These reactions occur in individuals who appear to have host factors rendering them more susceptible to unrestrained replication of 17D virus, which then causes direct viral injury to the liver and vital organs, with a pathogenesis essentially identical to that of wild-type virus. The host factors appear to be both genetic and acquired; the former possibly related to defects in innate immunity and the latter related to immune dysfunction of advanced age and thymic disease. The incidence of YEL-AVD appears to be higher than severe or life-threatening adverse events associated with other licensed viral vaccines. This observation has triggered recommendations from the ACIP to limit vaccination to people truly in need due to possible exposure to wild-type virus [152]. Assessment of the risks and benefits of 17D vaccination now fall to the practitioner of travel medicine and to those responsible for public health policy in endemic countries. In situations in which the vaccine is used for mass immunization in epidemic control, active surveillance for adverse events should be incorporated into the control plan, as witnessed in Kenya (1992) and the Ivory Coast (2001) [157]. No cases of YEL-AND or YEL-AVD were detected in the Ivory Coast. The defects in host response associated with YEL-AVD deserve further dissection, principally because they will reveal general mechanisms underlying human susceptibility to flavivirus infections, the ratio of inapparent to apparent infections and the severity of disease expression.
Although the adverse events associated with 17D are of concern, it is important to emphasize that 17D is a highly effective vaccine that induces long-lasting protection against a fatal and devastating disease. Overall, the benefits far outweigh the risks of adverse events in individuals traveling or residing in YFendemic areas. These risks have been reviewed in detail elsewhere [158]. The benefit of 17D vaccination in areas of Africa and South America, where periodic upsurges in wild-type virus activity have spawned intense epidemics or increased risk of jungle YF, is undisputed and, with assistance from the Gates Foundation (WA, USA), the vaccine is increasingly used in routine childhood immunization programs. The cost-benefit ratio is favorable in endemic/epidemic areas [159].
A question that has been considered by industry and the WHO is whether or not a new, safer vaccine against YF is required or whether other approaches, such as active-passive immunization, should be considered. To date, there is no effort toward a safer vaccine, although sanofi pasteur and the CDC are collaborating on a clinical trial of active-passive immunization using commercial lots of ISG. It is expected that the coadministration of YF antibody will not interfere with the vaccine [128,129], while reducing viremia and extraneural replication. As coadministration of ISG (for hepatitis A) and 17D vaccine was common prior to the advent of hepatitis A vaccines, this strategy has practical merit at least for travelers or those with risk factors (thymic disease and advanced age). With respect to the development of a safer vaccine, the obstacle is commercial, not technical. Both safer live vaccines (based on chimeric technology), inactivated whole virion, subviral particle or recombinant subunit vaccines could be developed without difficulty, and data from other flavivirus vaccines provides evidence that such approaches would be feasible and effective in the case of YF. Similarly, although the molecular tools now exist to effect a straightforward shift from egg-based to cell-culture systems for manufacture of standard 17D vaccine, there is no commercial incentive to do so.
It cannot be determined yet whether the rare serious adverse events will complicate the future development of chimeric vaccines using 17D as a vector. With respect to neurotropism, the chimeric vaccines have been engineered to be significantly less neurovirulent than parental 17D [150][151][152][153][154] and, based on animal toxicity studies, one would expect that the chimeras will have significantly lower rates of neurotropic accidents than parental vaccine. The concern is more on viscerotropism, and although it is likely that changing the E-gene (containing ligands for cell receptors) will alter replication sites in vivo, large-scale clinical trials and postmarketing surveillance will ultimately determine safety of the chimeric vaccines.

Five-year view
Scientifically, there remain many important questions that will be addressed in the next 5 years. The 17D vaccine will prove to be a useful tool for conducting clinical research, for example, in the exploration of signal transduction events underlying innate immune responses. It is likely that these studies will provide an explanation for the remarkable durability of the adaptive immune response to 17D vaccine and will elucidate general mechanisms underlying intrinsic resistance to viral infections. Recently, a full-sponsored research effort has been planned to utilize 17D as a model infection to explore the pathogenesis of cutaneous immune dysfunction associated with Th2-oriented responses in atopic individuals. The interesting observations cited in this review on dendritic cells as targets for virus replication in the skin, and the recent observations on skin immunization of monkeys with 17D vectored vaccines, suggest the possibility that this and other flavivirus vaccines are ideally suited for needleless delivery methods.
At a virologic level, new data will be available on the molecular basis for attenuation and virulence of 17D and its parental (Asibi) virus. Clinical research on the interactions of 17D and chimeric vaccines using 17D as a vector will lead to critical new data on the role of NS genes in immunity, vaccine interactions, and immunologic memory and crossreactions between flaviviruses. Although 17D vaccine, developed nearly 70 years ago, is entering old age, it will continue to provide a wealth of new information about immunologic and virologic first principles.
Finally, any futuristic view must take into account the possibility that YF will emerge in Asia, receptive areas of Central America or the USA, or in densely populated coastal regions of South America infested with Stegomyia aegypti. The likelihood of such an occurrence is enhanced by increased air travel. The disease is so virulent and easily diagnosed that little delay is expected in recognizing it and mounting control measures, which would predominantly involve the use of vaccine. Shortages of approved vaccine in the USA and other developed countries would be greatly amplified in such an emergency, as only a few hundred thousand doses are distributed annually.

Key issues
• Yellow fever (YF) is a zoonotic disease requiring prevention/control by means of human vaccination.
• The molecular basis for attenuation of 17D vaccine is reviewed.
• The mechanism of action depends on neutralizing antibody responses.
• The remarkable immunogenic properties of 17D (rapid and durable adaptive immunity) depends on generating strong innate responses. • Rare YF vaccine-associated neurotropic and viscerotropic adverse events are caused by direct viral injury.
• Host factors associated with serious adverse events include advanced age, thymic disease and other forms of immune suppression.
• YF 17D vaccine is now being explored as a live vector for foreign genes.