YF17D‐based vaccines – standing on the shoulders of a giant

Live‐attenuated yellow fever vaccine (YF17D) was developed in the 1930s as the first ever empirically derived human vaccine. Ninety years later, it is still a benchmark for vaccines made today. YF17D triggers a particularly broad and polyfunctional response engaging multiple arms of innate, humoral and cellular immunity. This unique immunogenicity translates into an extraordinary vaccine efficacy and outstanding longevity of protection, possibly by single‐dose immunization. More recently, progress in molecular virology and synthetic biology allowed engineering of YF17D as a powerful vector and promising platform for the development of novel recombinant live vaccines, including two licensed vaccines against Japanese encephalitis and dengue, even in paediatric use. Likewise, numerous chimeric and transgenic preclinical candidates have been described. These include prophylactic vaccines against emerging viral infections (e.g. Lassa, Zika and SARS‐CoV‐2) and parasitic diseases (e.g. malaria), as well as therapeutic applications targeting persistent infections (e.g. HIV and chronic hepatitis), and cancer. Efforts to overcome historical safety concerns and manufacturing challenges are ongoing and pave the way for wider use of YF17D‐based vaccines. In this review, we summarize recent insights regarding YF17D as vaccine platform, and how YF17D‐based vaccines may complement as well as differentiate from other emerging modalities in response to unmet medical needs and for pandemic preparedness.


From YF17D history and legacy to vaccine platform
The yellow fever virus (YFV) is a highly pathogenic virus with a case fatality rate of more than 30%.It is endemic in large parts of Sub-Saharan Africa and Latin America.YFV is currently reemerging and causes, despite the availability of a very potent vaccine, several tens of thousands of deaths annually [1].The liveattenuated yellow fever vaccine (YF17D) strain YF17D was devel-provide long-term, if not life-long protection, global vaccination rates are insufficient, leaving populations vulnerable to YF outbreaks [6].Emergency mass vaccination is still the only means to manage YF outbreaks today.
The invention and history [7], clinical use and remaining gaps in knowledge of the YF17D vaccine have been summarized in several comprehensive reviews [7,8].Recent insights and open questions regarding efficacy and safety [9], epidemiology [10,11] and future developments in the field of yellow fever vaccination [12] are discussed elsewhere.Approaches on how to overcome issues of disruptions and shortages of YF17D supply by fractional dosing [3,13], dosing via the intradermal route [14], or modern state-ofthe-art bioprocess engineering [15], moving away from classical egg-based production [16][17][18], are out of the scope of this review.
Here, we focus on the more recent use of YF17D as a platform to generate recombinant live-attenuated vaccines to tackle unmet medical needs in the field of prophylactic and therapeutic immunization and for pandemic preparedness.Starting with a summary of the unique polyfunctional immune profile that is elicited by YF17D, and how this may translate into a favourable immunogenicity of YF17D-based vaccines derived thereof.In a second part, we describe the design and performance of currently licensed recombinant vaccines and vaccine candidates that are based on YF17D as a live viral vector.The use of subgenomic YF17D replicons [19,20] as a basis for self-amplifying RNA vaccines [21] or of related Flaviviruses as live viral vectors [22][23][24][25][26] is not considered.Technical aspects of recombinant vaccine virus construction, vaccine safety and unique requirements of preclinical animal testing linked to the biology of YF17D are described in separate text boxes.

Original immune responses to YF17Dleveraging for its use as vector
The YF17D legacy vaccine induces uniquely broad innate and multifunctional adaptive immune responses (Fig. 1) and is therefore considered a paradigm and benchmark in vaccinology [5].Yellow fever immunization is hallmarked by an extraordinarily quick (within 10-14 days), consistent (almost 100% seroconversion among healthy children and adults) [2][3][4] and long-lasting production of neutralizing antibodies (nAb), and balanced CD4 + and CD8 + T-cell responses (reviewed elsewhere [8,27]).In most vaccinees, lifelong or nearly lifelong protection is likely achieved after a single vaccine dose [28], with seroprotective levels of YFVspecific nAb detectable for up to 20 years in >90% [29,30] and 35 years in about 80% of vaccinated adults [31].In contrast to adults, nAb seems to wane more rapidly in children, particularly in those who received their primary vaccination at very young age (<2 years) [4,32,33], with a likely need to get re-vaccinated (boosted) within 5-10 years.YF17D effectiveness is remarkably high with only a handful cases of clinical yellow fever reported in persons with a history of 17D vaccination in almost 90 years of wide use.Actual rates of vaccine failure are difficult to estimate due to (i) changing epidemiology [34,35] and (ii) very low fre-quency.This however might be as low as 1 in 20-25 million [8].Such a high potency, efficacy and effectiveness appears unique; possibly even in comparison to other widely used human legacy vaccines such as measles-mumps-rubella [36] or smallpox vaccine [37].As specified in the following, different effector mechanisms complement each other in conferring yellow fever immunity.This multiplicity of mechanisms may also contribute, alone or in combination, to the protection YF17D-derived recombinant vaccines exert against a wide spectrum of other pathologies.
Despite its long-standing clinical use, little is known about the actual target cells and organs of live YF17D virus following its subcutaneous or intramuscular administration.Most insights have been gathered from ex vivo studies using human PBMCs [38][39][40][41] and preclinical biodistribution studies of YF17D-based recombinant vaccines [42][43][44] which, intriguingly, suggest a preference for lymphoid cells and organs.These data showing a tendency towards dendritic cells and other members of the monocyte/macrophage lineage are supported by virus tracking in humanized mice infected intravenously with YF17D [45].In this model, viral infection of, and replication in, antigen-presenting cells such as dendritic cells (DCs) and B cells leads to antigen persistence (up to 11 days after YF17D vaccination) within nascent germinal centres (GCs) [45].Thus, as a replication-competent live vaccine, YF17D mimics an acute viral infection.This supports the establishment of long-lived GC [46], whereby a sustained expression of YF17D-derived toll-like receptor (TLR) ligand and other pattern recognition receptor agonists causes a prolonged activation of B cells and follicular DCs, which in turn favours the establishment of (possibly affinity-matured) memory B cells and long-lived plasma cells.Besides boosting potent humoral responses, active replication of YF17D in DCs [38] further induces autophagy through the GCN2-eIF2α integrated stress response axis, enhancing antigen (cross-) presentation to both CD4 + and CD8 + T cells [41].The importance of active viral replication for the outstanding broad and long-lasting immunity conferred by YF17D becomes obvious in direct comparison to inactivated alternatives.Although deemed safer and highly immunogenic [47,48].β-propiolactone inactivated YF virus adjuvanted with alum required multiple doses to achieve complete protection and failed to trigger YF-specific CD8 + T cells [49].Similarly, a non-replicating modified vaccinia Ankara (MVA)-vectored YF vaccine candidate, expressing the YFV surface antigens, required multiple doses to match an original YF17D response [50].
Though clearly also other cell types, such as muscle and fibroblast cells, alike those playing a major role during vaccine production in embryonated chicken eggs [51,52], are highly permissive to YF17D infection and may significantly contribute to local amplification at the injection site.Leveraging on the abundance of lymphoid cells in the skin and mucosal tissues, both intradermal and intranasal immunization using YF17D has been trialled in humans [53] and non-human primates (NHPs) [54] with a dose-sparing effect and ease of administration, respectively, as possible benefits.
YFV-specific nAb are considered the major correlate of protection against YFV infection (for a critical discussion see [2]); whereas the expression of the receptor for the B-cell-activating factor BCMA/TNFRSF17 is seen as a key response-predictive signature [59].YF17D induces extraordinarily fast and sustained seroprotection.Protective levels of immunoglobulin M (IgM) are detectable within 4-7 days post-vaccination, peak around 2 weeks, and can persist up to 4 years [61,62].A more abundant and durable IgG antibody response develops and mounts more slowly, which peaks around 30 days post-vaccination and may remain detectable for more than 35 years [31].The durability of circulating YF-specific IgM that can persist for up to 3-4 years remains an intriguing feature [61].Originally, it was speculated that YF17D could cause non-pathogenic persistent infections [63], or that antigen storage in follicular DCs might account for the prolonged production of IgM nAbs [61].Alternatively, as a live-attenuated virus, YF17D might induce IgM long-lived plasma cells (LLPC), as shown for other viral infections such as influenza or lymphocytic choriomeningitis virus (LCMV).IgM LLPC accumulates in the spleen and, unlike IgG LLPC, can develop in the absence of GC, does not undergo further antigen selection (no viral persistence is required) and is, at least experimentally, sufficient to protect from lethal challenge (in an influenza mouse model) up to 1 year postimmunization [64].
Adoptive immune serum transfer experiments in mice, hamsters and monkeys have proven that YFV-specific nAb are sufficient to prevent lethal YFV infection [47,65,66].Antibodymediated protection might, nonetheless, be at least partially dispensable and supplementary to the strong T-cell response.By using two different chimeric vaccines (Chimerivax-JE and YF-ZIKprM/E) carrying heterologous glycoproteins prM/E that do not induce YFV-specific nAbs (see infra), we demonstrated that CD8 + alone [67], or in combination with an FcR-mediated effector mechanism involving non-nAb binding to non-structural (NS) protein 1 of YFV [68], may suffice to confer protection against YFV in mice, whereas CD4 + T cells and nAb appeared dispensable.Likewise, also in mice, immunization, using an AdV encoding only the NS3 protein of YF17D that does not induce any nAb, conferred protection against intracranial challenge, which relies on perforin-producing CD8 + T cells [69].
Following YF17D vaccination, antigen-driven rapid activation and proliferation of CD8 + T cells are featured by a transient upregulation of human leukocyte antigen-DR isotype, CD38 and Ki-67 and downregulation of the anti-apoptotic protein Bcell lymphoma 2. After peaking approximately 2-4 weeks postvaccination, peripheral YFV-specific CD8 + T cells divide extensively and then differentiate rapidly towards a polyfunctional and long-lived memory phenotype of CD45RA + CCR7 − [70][71][72].Upon re-exposure with YFV antigen, YFV-specific CD8 + T cells respond quickly by the production of IFN-γ (dominant cytokine), TNFα (high frequency), IL-2 and macrophage inflammatory protein-1beta.In addition, almost all active CD8 + T cells are positive for surface CD107a (also known as lysosomal-associated membrane protein 1, LAMP-1) and express substantial granzyme B and perforin, demonstrating activity of degranulation and cytotoxicity [73].Surprisingly, YFV-specific CD8 + T cells with a unique naivelike profile persisted in vaccinated individuals for more than 25 years with stem cell-like capacity of self-renewal ex vivo [74].YF17D-specific CD8 + T cells respond to YF17D by broadly targeting overlapping epitopes spanning from E, NS1, NS2b and NS3 proteins and beyond, which may confer overall a high-quality immunogenicity in humans [75].
A robust and polyfunctional CD4 + T-cell activation signature lies at the regulatory core of the integral immune response against YFV [38,76,77].Stimulated naive CD4 + T cells expand from a heterogenous repertoire, where the rarest precursor subsets become dominant in the effector response [78].This precedes the onset of CD8 + T cells, with CD4 + CD40 + IL-2/IFNγ-secreting T-cells circulating systemically within 24-48 h after vaccination [79].Likewise, circulating T follicular (cTfh) helper cells are detectable 72 h post-vaccination [80] and continue to increase up to 2 weeks post vaccination, following GC activity [81].These cells appear to be predominantly Th1-polarized, with a prominent expression of IFN-γ, IL-2 and IL-21.Unlike Th2-and Th17-polarized cTfh, Th1-polarized cTfh display epitope specificity against several viral proteins (i.e.C, E, NS1 and NS3) and correlate with high titres of nAb [80].Despite inducing a cytokine signature conventionally attributed to Th1 and Th2 responses [77,82], specific Th2 responses (e.g.induction of IgE; recruitment and activation of mast cells, basophils or eosinophils) have not been reported.Rarely observed allergic reactions after YF17D vaccination are related to egg allergy and residues originating from classical egg-based manufacture [83].Thus, IL-4-secreting CD4 + T cells might contribute to affinity maturation and memory Bcell generation during the earlier stages of the response [84], but YF17D-elicited antiviral memory responses appear predominantly Th1-polarized [85].
The CD4 + T-cell compartment remains nonetheless understudied, though of increasing importance considering the use of the platform to target more complex parasitic infections such as malaria [86][87][88] and Trypanosoma [89,90] or persistent viral infections such as HIV [87] or chronic hepatitis B [91].Nevertheless, besides inducing CD8 + and CD4 + T cells, concomitant stimulation of both proinflammatory Th17 cell responses [92] and antigen-specific regulatory T (Treg) [93] appears to be part of an overall balanced adaptive cellular immune response to YF17Dbased vaccines.Intriguingly, circulating Foxp3 + Tregs showed a rather delayed expansion and a relatively short-lived activation 10-14 days after YF17D vaccination [80,93,94]; followed by a rapid decrease in PD-1 expression (essential for their regulatory function) [94] and contraction to prevaccination levels.This is unexpected considering the usually high responsiveness of Tregs for cytokines such as IL-2 that are known to boost general T-cell development and are abundant early after YF17D administration (i.e.24-48 h; see above).For comparison, during acute infection with lymphocytic choriomeningitis virus or influenza A virus and resulting elevation of IFNγ levels in mice, Tregs acquire already within a week a Th1-like phenotype that limits CD8 + T-cell effector function, proliferation and memory formation [95].It is thus tempting to speculate how, in the case of YF17D, a uniquely postponed and abbreviated Treg-cell response may favour the fast activation and robust expansion of a vigorous antiviral cellular immunity.Hence, though highly functional, the generation of protective T-cell immunity to YFV, particularly preceding and during induction of effector T cells, remains largely unexplored.
In addition, though not understood in all complexity and functional consequences, YF17D also rapidly activates natural killer (NK) cells, monocytes, macrophages, granulocytes and gamma delta T cells (γδ + T cells) [96][97][98] and integrates the inflammasome [55] and the complement system to prime an optimal immunity [27].Analysis of the vigorous and balanced immunity seen in multiple applications of the vector platform (see infra) gives a reason to believe that the favourable immune profile triggered by genuine YF17D live-attenuated vaccine also translates to YF17Dderived vaccines.

YF17D as instructive example
Understanding of YF17D-induced immune signatures and cellular and molecular determinants contributing to protection has offered a measure to benchmark new vaccine candidates [5,27,59].Likewise, findings for YF17D might help rational selection of adjuvants (e.g.lipid nanoparticles and TLR agonists) or replication-competent recombinant viral vectors in order to guarantee broad activation of innate pathways to increase vaccine efficacies [59,99,100].Though appealing, direct translation might prove difficult since the unique properties of YF17D, whether used as genuine YF vaccine or as vaccine vector, cannot be explained by a particularly salient feature.Moreover, clinical studies comparing different vaccine platforms head-to-head are scarce and individual data hard to correlate due to confounding factors, such as pre-existing vector immunity, pre-vaccination immune signature and wide demographic and biological diversity of individuals.There are, nonetheless, several aspects for which YF17D seems to stand out.
(a) Replication-competent virus: The duration of immunity following YF17D vaccination could be attributed to a prolonged replication in lymphoid tissues.Similar feature has been described for measles virus vaccine, an equally successful live-attenuated vaccine with long-lasting humoral and cellular response [101].In contrast, current AdV, poxvirus and rabies virus-vectored vaccines are single-round replicating, replication-defective or inactivated constructs, showing suboptimal performance in stand-alone regimens [102,103].Likewise, an over-attenuation of parainfluenza virus-vectored vaccines leads to poor efficacy [104,105].
(b) Early adaptive transcriptional signature: In contrast to, for instance, live-attenuated varicella-zoster virus or rVSV-EBO-GP, YF17D triggers an early induction of B-and T-cell signatures [59].The latter is likely not antigen-specific and rather results from an early broad recruitment of adaptive immune cells into circulation.It remains to be elucidated how this unique feature of YF17D impacts the development of antigen-specific response thereafter and what can be learnt from this example for other vaccines.
(c) Broad and redundant activation of innate immune receptors: The activation of both TLR-dependent and independent virus sensing mechanism (i.e.TLR2, 3, 7, 8, 9, RIG-I and MDA-5) by YF17D contributes to the compound Th1/Th2 cytokine signature.This is mainly achieved by the redundant activation of transcription factors downstream MyD88 (i.e.NF-kB and IRFs) and TRIF (i.e.IRFs) (Figure 1).For instance, AdVes induce copious amounts of type I IFN mainly by the MyD88/TRIF-independent dsDNAmediated activation of IRF3, with modest production of NF-kBcytokines, IL-6 and TNFα [106].By this means the initial infection is efficiently controlled, whereas the humoral and cellular immunity against the AdV-vectored antigen might be hampered.In line, YF17D expressing hepatitis B virus (HBV) core antigen (HBc) induced a significantly higher frequency of polyfunctional HBc-specific CD8 + T cells producing IFNγ, TNFα and GmzB than AdV-vectored HBc assessed in parallel [91].This suggests that the absolute amounts of type I IFN are less relevant for the quality of the immune response than more subtle upstream pathways involved.Similarly, type I IFN responses to VSV and Newcastle disease virus are prominently dependent on RIG-I activation [107].By contrast, YF17D seems to temper RIG-I activation via modification of its nascent RNA by an intrinsic N1-2'O-methyltransferase activity of its NS5 protein [108].Thus, VSV-vectored Ebola vaccine (rVSV G-ZEBOV-GP; Ervebo) causes a more vigorous burst in type I IFN responses 24 h after vaccination, in contrast to a relatively late and moderate induction of IFN pathways by YF17D [59].While efficient in controlling viral replication, such an early exacerbated innate response by VSV might eliminate the vaccine virus too soon, with the consequently limited antigen persistence and compromised immune memory.Notably, rVSV G-ZEBOV-GP vaccination of macaques leads remarkably fast to full protection when challenged shortly after vaccination; immunity, however, waned rapidly, resulting in less than 50% survival when challenged after 3 months [109].
(d) Balanced polyfunctional immunity in stand-alone regime: Cellular immune responses induced by YF17D seem to be balanced to such an extent that vigorous long-lasting and polyfunctional (humoral and cellular) immunity can be achieved by YF17D alone.This may also apply to YF-based recombinant vaccines, as exemplified by the chimeric Japanese encephalitis vaccine Imojev (see infra).By contrast, vaccines based on other platforms, such as recently licensed Ebola vaccine (AdV + MVA [110]), as well as several vaccine candidates in different stages of development intended for prophylactic (e.g.AdV + MVA for malaria [111], DNA + MVA for Zika [112], AdV + DNA for Nipah [113]) or therapeutic use (e.g.AdV + MVA for cervical cancer [114] or chronic hepatitis B [115]), seem to clearly benefit, if not require, heterologous prime-boost schemes combing different modalities.A similar clinical benefit over single and homologous vaccination is observed as charged 'hybrid immunity' when COVID-19 vaccines based on different platforms are mixed [116].Though still speculative, potency of YF17D-based vaccines may not depend on such mixing.
In conclusion, the immunity elicited by YF17D and likely YF17D-based vaccines appears to leverage on (i) a robust replication capacity as a live-attenuated virus, (ii) an early adaptive transcriptional response and (iii) redundant activation of innate immune signalling that seems to feed into (iv) a vigorous, balanced and polyfunctional antigen-specific response.

Vaccines and vaccine candidates based on the YF17D vector
The unique immunological profile and outstanding potency of YF17D make it an attractive viral vector for the development of live-attenuated recombinant vaccines [146,147].Despite its compact size and somewhat limited coding capacity, the YF17D genome (Fig. 3a) can be manipulated to insert and express novel target antigens to generate (i) chimeric and/or (ii) transgenic vaccines.A comprehensive overview of the principal different designs is given by Bonaldo et al. [146].An updated list of YF17D-based vaccines and vaccine candidates is provided in Table 1.Technically, new target antigens are expressed as translational fusion within the genuine YF17D polyproteins, as summarized in Fig. 3. Regarding details on the molecular virology relevant for YF17D replication and polyprotein processing within the infected cell, we refer to Barrows et al. [148].An overview of the different reverse genetics systems used to generate recombinant YF17D derivatives is given in Box 2.

Chimeric Flavivirus vaccines
In Greek mythology, a chimera is a hybrid creature, composed of parts from different species.Similarly, chimeric YF17D-based vaccines are created by exchanging the coding sequence for the antigenic surface proteins (prM and E) of original YF17D for the corresponding genes of another Flavivirus (Figure 3b).In such a chimera, intracellular amplification of recombinant virus genomes is driven by the YF17D replication machinery.Infectious progeny virus is assembled containing the chimeric envelope, in this way triggering immunity against the new target antigen.Specific nAb responses against YF17D are not induced anymore.Two such chimeric vaccines have been granted market authorization, that is the Japanese encephalitis vaccine ChimeriVax-JE/Imojev [153,154] and the dengue vaccine CYD-TDV/Dengvaxia (both Sanofi-

BOX 1 -Animal models for preclinical assessment of YF17D-based vaccines
The in-depth study of YF17D and vaccines derived thereof is hampered by the lack of suitable animal models (Fig. 2).Immunocompetent (wild type, WT) mice, widely used as default model for mechanistic studies and preclinical vaccine testing, are naturally resistant to YFV infection.As such, WT mice do also not support active replication of live YF17D.Consequently, it can only elicit limited responses in mice, in sharp contrast to the vigorous immunity triggered in humans [117][118][119].The poor susceptibility to vaccination can readily be explained by the restriction innate immunity -in particular type I interferon (IFN-I) signalling -poses on the replication and dissemination of YF17D in mice [117,118,120].For successful infection, and hence vaccination, YF17D must overcome the antiviral state induced by IFN-I.In humans, this is accomplished via sequestration of STAT2 by the viral NS5 protein [121], whereas the necessary molecular interactions (phosphorylation of STAT1 and ubiquitinylation of NS5) are not compatible in mice [122].Nevertheless, YF17D shows a marked neurovirulence when inoculated directly in the mouse brain, which has historically been employed as potency test during vaccine manufacture.Mechanistically, neurovirulence is linked to a mutant allele of the antiviral Oas1b gene common in laboratory mice [123].On the other hand, this neurovirulence feature of YF17D in mice can be used experimentally.The high mortality following intracranial inoculation of YF17D in the mouse brain is frequently employed as stringent challenge model to assess the protection against YFV induced by immunization with novel vaccine candidates administered, for example, by the clinically relevant subcutaneous or intramuscular routes.Unfortunately, YF17D-induced protection is highly strain-dependent in mice and YFV-specific nAbs may not necessarily correlate with protection in WT mouse models [119].Alternatively, mice deficient in IFN-I signalling, such as ifnar −/− (mice lacking IFN-α/β receptors), have been proposed as valid surrogate to study the immunogenicity of YF17D and YF17D-based vaccines.YF17D replicates readily in ifnar −/− mice and elicits strong antibody and polyfunctional T-cell responses, almost independently from the inoculated dose [119,124].Obviously, in general, data obtained in mice, both WT strains as well as in strains with a compromised immune system, have to be critically judged regarding their relevance for human application [118,[125][126][127]. Syrian golden hamsters (Mesocricetus auratus) are naturally susceptible to YF17D, support robust antibody production after immunization, and represent a unique alternative small animal model for the study of YF17D and YF17D-derivatives.Several hamster-adapted virulent YFV strains, mimicking hallmarks of human yellow fever pathology, are available for challenge studies [128][129][130][131][132], and show a clear correlation between nAb levels and protection against infection in this model [47].Moreover, although YF17D infection in WT hamsters is asymptomatic and self-limited, as in humans, immunosuppressed hamsters [133] or hamsters with a STAT2 knock-out [134] can be used for vaccine safety studies related to YEL-AND and YEL-AVD (see Box 3) [42,92].Unfortunately, in-depth mechanistic characterization in hamsters remains challenging due to the lack of immunological reagents.As a natural host and reservoir for YFV, non-human primates (NHPs) represent likely the most relevant model for the mechanistic study of YF17D and derivatives thereof.YF17D immunization induces self-limited viraemia but considerate humoral and cellular responses in NHPs, alike YF17D immunization in humans [54,135,136].As a caveat, the use of NHP models is prohibitively expensive and raises ethical concerns.Nevertheless, WHO recommendations require that each vaccine virus seed lot is tested for viscerotropism, immunogenicity and neurotropism in NHP by intracranial inoculation before it is used for eggbased YF17D production [137].However, the consistency, reliability and hence relevance of results obtained by this historically established (in 1943) monkey safety test are under debate [138][139][140] and regulatory requirements may need to be adapted, in particular for YF17D-based vaccines employing other technologies.Innovations in vaccine quality control and safety testing may lead to a shift towards alternatives that are easier to standardize and interpret, for example recently validated small animal models for in vivo testing [42,47,140,141].Modern quality assessment will progressively rely on in vitro surrogate readouts generated by emerging molecular and cell-based methods such as deep sequencing [142,143] or use of organoids [144,145].Such innovations will be driven by growing insights in the biology of YF17D and how YF17D-based live-attenuated vaccines interact with the host immune system.
Pasteur) [155,156].Notably, leveraging on the excellent safety record (Box 3), genetic stability and efficacy of YF17D, Imojev and Dengvaxia were the first ever recombinant live vaccines licensed for human use, including for paediatric use.
The development of tetravalent Dengvaxia has been facing challenges that are linked to the complex immunological interplay between the four serotypes, common for every dengue vaccine [157], and hence not related to the YF17D backbone.Though poorly understood, the highest risk to suffer from severe dengue haemorrhagic fever (DHF) is observed in people infected

BOX 2 -Multiple approaches to construct recombinant live-attenuated YF17D-based viruses
Non-clonal reverse genetic systems: c) Fusion of 2-long subgenomic PCR amplicons 5 SP6, followed by invitro transcription and transfection [151].d) Infectious subgenomic amplicons (ISA) method: direct transfection of overlapping PCRs by 5' CMV promoter and 3' HDRz/SV40polyA [152].e) Circular polymerase extension reactions (CPER) or Gibson assembly (reviewed in [149,150]).by eukaryotic promoter such as SV40 promoter and 3' by a ribozyme i.e hepatitis delta virus ribozyme (HDRz).These plasmids can be transfected directly into permissive cells to produce recombinant vaccine virus.b) Generation of cDNA as in a) but the cDNA is 5' by a bacteriophage promoter such as SP6p.This plasmid is used as template for in vitro transcription, followed by transfection of the RNA into permissive cells to produce recombinant vaccine virus.Individual vaccine candidates and level of preclinical and clinical evaluation are summarized in Table 1.
for a second time by another antigenically distinct virus strain: originally described as antibody-dependent enhancement (ADE) of infection [158].Therefore, any dengue vaccine needs to induce a balanced response against all four dengue serotypes, in case of Dengvaxia by mixing of four live vectors.Dengvaxia has shown to provide good protection against dengue haemorrhagic fever in seropositive populations and is now recommended for use in naturally primed individuals as a kind of booster vaccine [159,160].
A vaccine candidate for the West Nile virus called ChimeriVax-WN02 has completed clinical phase II testing, though it was halted for further development [161].Despite promising safety and immunogenicity in humans, a changing epidemiology of the disease precluded trial set-up for assessment of clinical efficacy as required for the classical licensure pathway.Likewise, several chimeric Zika virus (ZIKV) vaccine candidates, designed following the same blueprint, have proven protective in mice [162][163][164], including from vertical transmission and congenital malformations in a stringent intra-placental challenge model [163].
The chimeric design seems to be feasible exclusively for the exchange of homologous surface antigens from members of the Flavivirus genus to which also YF17D belongs.Hereby multiple constraints (i.e. assembly of infectious progeny [165]; co-translational polyprotein processing [166]; RNA structures orchestrating genome replication [148]) pose a biological limit that already applies to the related hepatitis C virus proteins belonging to the greater Flaviviridae family [167].

Transgenic vaccines for prophylactic and therapeutic use
The YF17D genome also tolerates insertion of a wide range of foreign/heterologous sequences in different intragenic and intergenic sites depending on the size and the nature of the insert (Table 1 and Fig. 3c) [146,168].This design results in the expression of an add-on or transgene from the YF17D backbone without altering the expression of YFV proteins; meaning that No protection in marmosets [174,175] GP1 and/or GP2 E/NS1 CBA/J mice and guinea pigs
its immunogenicity as YF vaccine is maintained [146,168].This makes YF17D a platform to generate dual prophylactic vaccines possibly protecting against both, inserted target antigen plus YF [92,128,169] (Table 1) (see infra).
The transgene approach has been explored to target a wide range of viral infections, such as Lassa fever, HIV or most recently COVID-19 (Table 1 and references therein).Although small B-and T-cell epitopes can be accommodated in multiple insertion sites (E [170,171], NS1 [172], NS2A [173] and NS2B/3 [146]), most targeted diseases require the expression of complex and large antigens (i.e.Lassa virus [LASV]-glycoprotein complex [GPC], SARS-CoV-2 Spike, Ebola GP, simian immunodeficiency virus [SIV]-Gag or HIV-p24).In this regard, there are mainly three alternatives to insert and express these antigens (Fig. 3c): (i) insertion into the capsid gene, (ii) the E/NS1 intergenic region or (iii) creation of bicistronic RNA molecules [146].The latter option (iii) suffers from genetic instability and low antigen expression and has hence not been widely employed, whereas in design (i) and (ii) the foreign antigen is a translational fusion within the YF17D polyprotein and as such produced intracellularly at equally high (equimolar) levels.The intergenic E/NS1 insertion (ii) has been explored more extensively as it naturally represents a functional shift from structural to NS proteins (Fig. 3a), facilitating the accommodation of larger antigens without altering the replication and expression of the other YFV proteins [146].The first complex and large antigen that was inserted in this region was the LASV glycoprotein complex.This construct was immunogenic and protective in guinea pig models of lethal LASV infection [174,175].As these constructs faced stability issues, a second approach was tested, in which the LASV glycoprotein subunits 1 and 2 were inserted separately.These vaccine constructs were more stable, immunogenic and protective in guinea pigs.
Two other larger antigens that were inserted in this site were SIV gag protein (amino acid 45-269) [135] and HIV p24 protein [87].In both cases, although immunogenic (significant humoral and cellular responses), protection conferred was moderate or low, respectively, mainly due to instability of both constructs [87,135].In the case of p24, a second approach to insert into the YF capsid region (insertion site i) was also not successful (Table 1, [87]).Similarly, expression of only small parts of the SIV-gag protein elicited low levels of SIV-specific cellular responses [176].Besides shortcomings that might be overcome by an advanced construct design to sustain antigen expression, this may suggest that for more complicated viral infections such as HIV the expression of only one antigen may not be enough to protect.In such case, multiplexing of several YF17D vectors, each expressing distinct antigens, may be considered.Co-administration (simultaneously and consecutively) of multiple YF17D-based vaccines is feasible as showcased by the tetrameric dengue vaccine Dengvaxia (see above).
More recently, an even more complex and larger antigen, the Spike (S) protein of SARS-CoV-2, which has almost the double length (4 kb) of the previously reported maximum size [146], has been successfully inserted in the E/NS1 intergenic region [92].This vaccine candidate, called YF-S0, induced high neutral-izing antibody titres and balanced polyfunctional cellular immune responses against both SARS-CoV-2 and the YF-vector [92,119,128,169].Moreover, this vaccine candidate protected hamsters and NHPs against vigorous challenge with SARS-CoV-2 [92] and its adapted variant-proof version YF-S0* against emerging immune-escape variants dominating the ongoing COVID-19 pandemic [169].Furthermore, simultaneous protection against YFV was confirmed in a hamster model (using lethal YFV Asibi challenge) validating the potential of the platform to develop dual vaccines providing protective immunity for multiple infections or pathologies [128].Following the same vaccine design, a series of YF17D-based vaccine candidates are currently being developed against a range of Ebolavirus species, such as Sudan virus and Bundibugyo virus, that are not covered by current Ebola vaccines (specific for Zaire virus) [177].
Expression of classical bacterial antigens such as capsule structures present in conjugate vaccines has not been reported, which is obvious considering that YF17D vectors produce genuine eukaryotic proteins devoid of characteristic bacterial glycosylation patterns.
Use of live YF17D has also been explored as a platform to develop therapeutic vaccines for diseases, such as cancer, malaria, Chagas disease or chronic hepatitis B. The first YF17D-vectored transgenic vaccine constructs were generated by McAllister et al. [178] and expressed the cytotoxic T-lymphocyte epitope SIIN-FEKL derived from chicken ovalbumin (OVA).Insertion was tried at different sites of the genome, that is the amino terminus and the C/prM and NS2B/3 junctions (Table 1), which yielded viable recombinant virus progeny.The NS2B/3 insertion though proved to outcompete the other two constructs regarding replication fitness.Insertion at the NS2A/2B, NS3/4A and NS4A/4B junctions did not allow to recover viable viruses.The most promising candidate, called YF-pOva-8, was tested in a mouse tumour model.Herein, vaccination elicited SIINFEKL-specific CD8 + lymphocytes and induced protective immunity against lethal challenge with malignant melanoma B16 cells expressing ovalbumin.Furthermore, active immunotherapy with recombinant YF17D viruses induced regression of established solid tumours and pulmonary metastases [178].These data suggest that YF17D could be used as platform to develop therapeutic cancer vaccines.In this regard, the use of YF17D vaccine might be an asset as its oncolytic potential has been reported for intratumoral immunotherapy [179,180].Already genuine YF17D was able to replicate and kill mouse and human tumour cell lines as well as to delay tumour growth in mouse models after intratumoral injections in a CD8 + T-cell dependent manner.Such anti-tumoral activity warrants further investigation.It results, however, most likely from a bystander effect from the pro-inflammatory response induced by replication of the vaccine virus that seems to augment and modulate general T-cell activities [178,181].Similar as been observed for infections with related Zika virus increasing CD4 + and CD8 + Tcell intratumoral infiltration and activation [182], possibly resulting in clinical tumour remission [183].Pre-immunization of mice with YF17D improved successive intratumoral YF17D treatment in terms of local and distant antitumor immunity.Additive efficacy was observed upon co-administration of anti-CD137 and anti-PD-1 antibodies [179,180].
Besides this oncogenic application, YF17D was also considered a vector platform to express epitopes from parasitic diseases such as malaria and Chagas disease (Table 1).In either case, the insertion of both small as well as larger fragments from Plasmodium yoelii circumsporozoite and Trypanosoma cruzi ASP-2 in either the C protein, the E/NS1 intergenic region, or displayed at the surface of the viral particle as insertion in the viral E protein was explored [86, 88-90, 170, 171, 184].The malaria vaccine candidates induced cellular and/or humoral immune responses against plasmodia (depending on the inserted antigen, T-or B-cell epitope).Induction of antibodies was detected in rhesus macaques after vaccination with the construct expressing a B-cell epitope of Plasmodium falciparum [88,170,171,184].However, only few of these recombinant vaccine candidates induced some protection after a single dose.Increased protection was observed when administered in prime/boost combination, especially when using inactivated sporozoites as heterologous antigen [88,185].Similarly, the vaccine candidates tested for Chagas disease conferred only limited protection against challenge with T. cruzi, although vaccine efficacy was improved when two constructs expressing different antigens were combined [89,90].Combining several epitopes in strings or several recombinant YF17D vaccine constructs carrying different antigens could help tackling these more demanding pathologies [90,171].
Another therapeutic application recently explored by our group is chronic hepatitis B [91] (Table 1), where the exhaustion or anergy of the HBV-specific antiviral T-cell compartment is thought to facilitate ongoing viral replication, infection of new cells and the establishment of hepatocyte clones that harbour integrated copies of the HBsAg and HBx antigen genes, hampering virus elimination from the persistently infected liver for a functional cure [186].Two vaccine candidates were engineered carrying the codon sequence of HBV core antigen (HBc) inserted either in the YF17D C or E/NS1 junction.Following a protein prime, the YF17D/HBc-C candidate was found to induce particularly strong, polyfunctional and antigen-specific T-cell responses that were able to selectively kill HBV antigen-positive target cells in vivo; outperforming, in our experiments, the same HBc antigen expressed from an AdV vector used as benchmark [91].
In summary, especially data from different therapeutic applications (from cancer parasitic diseases to chronic hepatitis B) point towards unique properties conferred by the YF17D vector backbone.In general, these include (i) a vigorous cytotoxic T-cell activity as well as (ii) a balanced cellular and humoral immunity, (iii) Treg and (iv) proinflammatory Th17 cell functions (Fig. 1), on top of (v) classical nAb and (vi) non-nAbs that may act via multiple FcR-mediated mechanisms.In the protection against yellow fever, all activities seem to play a role and to complement each other.For the protection or treatment of other pathologies as indicated above by YF17D-derived recombinant vaccines the importance of these different arms of adaptive immunity remains to be determined and may play case-by-case an each more or less relevant role (Figure 4).YF17D co-expressing relevant antigens seems to be able to enhance/improve the immune response against a range of difficult to tackle target diseases.Hereby, YF17D might contribute to a more potent and durable immune response against these pathogens.Hence, YF17D may not simply serve as a delivery vehicle but provides at the same time a particularly strong and favourable immunological boost to enhance specific immune responses against the inserted antigen where other platforms depend on extra adjuvanting [99,100,147].
Dual activity of YF17D-based vaccines: One possible merit of YF17D over other viral vectors is the dual immunity and protection that is concomitantly provided by YF17D-based vaccines [67,92,128,135,177].Given that some pathogens such as malaria, HIV, HBV, Ebola and YFV circulate in overlapping geographic areas, such dual activity, killing two birds with one stone, may provide a relevant public health benefit; in particular in regions with underdeveloped health systems and immunization programs of limited capacity (for discussion see [128] and references therein).Intriguingly, chimeric vaccines may also display a dual activity, independently of the presence of nAbs by unconventional mechanisms, such as by cellular or Fc-mediated immunity [67,68].
Anti-vector immunity: Strong anti-vector immunity raises a general concern when considering a vector platform repeatedly for multiple indications [187,188].Such anti-vector immunity has indeed shown profound impact, particularly on nonreplicating viral vector vaccines, such as replication-defective herpes simplex virus 1 [189] and AdV-based recombinant vaccines [187,190].In contrast, pre-existing anti-vector immunity seems to have a limited impact on live replication-competent viral vectors, for example anti-measles immunity had only a minor impact on the efficacy of recombinant measles virus-based vaccines [191].Pre-existing Flavivirus immunity could affect the infection outcomes and immunogenicity of subsequent Flavivirus vector vaccination or viral infection [192,193].Likewise, responses to primary YF17D immunization may vary in magnitude between geographically different cohorts with distinct Flavivirus epidemiologies [194].Interestingly, in phase 3 human trials of YF17Dbased Dengvaxia, subjects with prior Flaviviruses infections demonstrated an improved immunogenicity compared to recipients immunologically naive to Flaviviruses [195,196].Concerning original YF17D, repeated dosing, as historically endorsed to warrant maintenance of protective YFV immunity, results in consistent boostability of YFV-specific nAb levels [197,198].Though a reduced magnitude of immune responses from secondary vaccination hints towards both longevity of protection as well as partial interference with (re-)vaccination [199,200].Interestingly, booster immunization led to a marked activation and expansion in particular of epitope-specific CD8 + T cells, including in people with considerably high nAb levels [199], hence favouring the use of YF17D-based vaccines for therapeutic indications requiring vigorous cell-mediated effector mechanisms.Nevertheless, results obtained with our YF-S0* COVID-19 vaccine candidate in mice and hamsters support the idea that full protection can be achieved even in fully YF17D-vaccinated individuals [128].

Conclusions and future prospects
The live-attenuated YF17D legacy vaccine can be employed as particularly potent live viral vector to develop vaccines against a range of targets.This both in a prophylactic or therapeutic set-

BOX 3 -General vaccine safety
The outstanding efficacy of YF17D and YF17D-based vaccines is combined with a favourable safety profile [9].While recognized as one of the safest vaccines, very rare severe adverse events of Yellow Fever Vaccine-Associated Viscerotropic (YEL-AVD) and Adverse Neurotropic Disease (YEL-AND) have been reported.These can occasionally be fatal, yet remain poorly understood [28]; also, in lack of full mechanistic insight on what actually confers attenuation of the live vaccine [40,142,201] compared to its highly pathogenic parent YFV Asibi strain.Besides particularly young age [28,202], inborn or functional deficiency in IFN-I signalling [203,204] may contribute to an enhanced individual risk for YEL-AND and YEL-AVD, respectively.Though very few cases have been studied in detail [205], there seems to be a particularly strong link to host genetics [206] and rare immune defects [204,205], similar to other live-attenuated vaccines, such as MMR, oral polio or Bacillus Calmette-Guerin (BCG) vaccine [207].WHO recommendations [137] (last updated in 2013) and, accordingly, the label of current WHO-prequalified YF vaccines, integrate these general safety concerns in a list of contraindications.Vaccine shortages during recent major YF outbreaks in Western Africa and Latin America [11] as well as disruptions in global vaccine supply [208,209] have fuelled major efforts to re-assess YF17D vaccine safety and alternative dosing regimens [3,13].A crisis is taken as opportunity to carefully revisit historical recommendations for the use of YF17D vaccine.Firstly, YF17D appears to be well tolerated in infants and children [210,211] with a reactogenicity and overall sideeffects comparable to other parenterally administered vaccines [212].Apparently, this does even not exclude more vulnerable individuals that are prone to complications due to (some degree of) immune suppression [213,214].Notably, a similarly promising safety has been observed in children for chimeric ChimeriVax-JE/Imojev [215], being superior to original live-attenuated JEV vaccine JE Sa14-14-2 [216,217].Further, recent meta-analysis compiling safety data from 25 cohorts, collected over the last 20 years since first cases of YEL-AVD were reported, could not affirm that immunocompromised individuals (such as HIV and cancer patients, organ transplant recipients or those under immunosuppressive therapy for other indications) have, regardless of aetiology, a higher risk of adverse events after receiving yellow fever vaccine [218].Recently it was also confirmed for the case of fractional dosing [219].Though very rare in absolute numbers [220] and hence difficult to address properly [221], there appears to be some elevated relative risk in elderly, correlated with increasing age [222], more recently linked to a high incidence of autoantibodies against Type I IFN in this population [203].With the development of the first YF17D-based chimeric vaccines, a solid regulatory framework was established.Extensive safety testing included intracerebral inoculation in suckling mice or, back then, even NHPs (see Box 1) to exclude the risk of AND, with original YF17D vaccine serving as a reference control ( [141] and references therein).In both mice and monkeys, all tested chimeric vaccines were significantly less neurovirulent than YF17D.Subsequently, preclinical data for YF17D-based vaccine candidates against Zika [67], SARS-CoV-2 [92], Ebola [177] and chronic hepatitis B [91], hence both chimeric and transgenic viruses (Fig. 3b and c), corroborate the generally improved safety, including in severely immunocompromised mouse (intracranial in suckling mice; ifnar −/− mice; IFN-α/β plus IFN-γ-receptor deficient AG129 mice) and hamster models (STAT2 −/− ) (Fig. 2).Accordingly, in hamsters, biodistribution of the transgenic COVID-19 vaccine candidate (YF-S0) and viraemia, that is detection of vaccine virus in the bloodstream as hallmark of viscerotropic spread, were markedly reduced [42].In conclusion, emerging evidence and cautious reevaluation of risks and benefits suggest the use of YF17D as a safe vector for a broad range of indications.This may include, of course, not precluding all necessary safety testing in relevant models prior to human assessment, (i) the development of paediatric vaccines, (ii) vaccines to be used in special patient populations such as elderly and people with underlying health conditions and (iii) cases with rather complex co-morbidities, for instance when considering therapeutic vaccines.ting, ranging from diverse pathologies, such as encephalitis (e.g.Japanese encephalitis), severe respiratory distress (e.g.COVID-19), congenital malformations (e.g.congenital Zika syndrome), parasites (e.g.malaria, Chagas disease), chronic virus infection (e.g.HIV and HBV) and possibly cancer.
Two rationally designed YF17D-based vaccines (Imojev and Dengvaxia) have been the first-in-class recombinant live virusvectored vaccines in human use, serving as important precedent for new synthetic vaccines and vector modalities (AdV-based, recombinant MVA-based, mRNA and self-amplifying RNA) seeing daylight essentially for the first time in the global emergency response to the COVID-19 pandemic ( [223,224] and references therein).
The YF17D platform has the flexibility to express different kinds of antigens, ranging from small peptides/epitopes to larger and more complex polypeptides, at different positions of its genome, while inducing a potent and protective immune response against both target disease and vector.Altogether, this makes the live-attenuated vector, YF17D, a viable platform with the potential to respond rapidly to new pathogen threats that could emerge, the so-called Disease X.
As a challenge ahead, the more complex transgenic YF17Dvectored vaccine candidates will have to prove their feasibility, safety and potency; end-to-end, from consistent manufacturing, clinical efficacy and safety testing to uptake and implementation in immunization programs; to demonstrate their real-world effectiveness.

Figure 1 .
Figure 1.Broad polyfunctional immunity triggered by the YF17D live-attenuated vaccine.(a) YF17D activates multiple surface and intracellular pattern-recognition receptors (PRRs) in antigen presenting cells (APCs), including toll-like receptors (TLRs) 2, 3 and 7-9 and retinoic acid-inducible gene I (RIG-I).This results in an integrated immune response comprising a robust production of type I interferon (IFN), inflammasome components (NLRP3, pro-caspase-1), and IL-1β-associated genes.In parallel, YF17D induces the formation of stress granules and autophagy, enhancing antigen presentation to both CD4 + and CD8 + T cells.Type I IFN signalling leads to the expression of IFN stimulated genes in APCs and somatic cells, through the activation and nuclear translocation of STAT1/STAT2 complexes.The interaction of YF17D nonstructural proteins NS4B and NS5 with STAT2 prevents its phosphorylation, thereby tempering type I IFN signal transduction and favouring prolonged vaccine virus replication and amplification.(b) Th1-polarised follicular helper T cells (Tfh1) and likely IL-4-secreting cells provide help to B cells at the germinal centres (GC), where active YF17D replication takes place.They support the proliferation of activated B cells, affinity maturation, antibody class-switching and generation of classical memory B cells and IgG-producing long-lived plasma cells (IgG LLPC).Long-lived antigen-induced IgM plasma cells (IgM LLPC) develop in the absence of CD40-CD40L interaction in the GC, the contribution of Th-derived soluble factors remains unknown.Th1 cells secreting IFN-γ and TNF-α promote the differentiation of perforin/granzyme B-producing cytotoxic CD8 + T cells.Three populations of YF-specific long-lasting CD8 +T cells are generated, that is effector memory (CD8 + T EM ), central memory (CD8 + T CM ) and stem cell-like memory CD8 + T cells (CD8 + Tscm).The contribution of specific Th17 to the overall adaptive immunity is yet to be determined.

Figure 2 .
Figure 2. Animal models to study YF17D-vectored vaccines.For details see Box 1.
of full length cDNA from viral RNA or, alternatively, synthetic DNA (one or several fragments) that is cloned into low copy plasmid or in bacterial or yeast chromosomes (BAC and YAC, respectively).The cDNA is 5'

Figure 3 .
Figure 3. Genome structure and design of YF17D-based recombinant vaccines and vaccine candidates.a) Schematic diagram of YF17D virus genome.b) Chimeric vaccines are engineered by swapping the YF prM/E genes with the equivalent ones from another Flavivirus.c) Transgenic vaccines are engineered by insertion of heterologous antigens in different places of YF17D genome, indicated with blue and purple arrowheads.Individual vaccine candidates and level of preclinical and clinical evaluation are summarized in Table1.
ifnar −/− mice Strong polyfunctional antigen-specific T-cell response; able to selectively kill HBc-positive target cells in vivo.Enhanced by heterologous protein prime Authors.European Journal of Immunology published by Wiley-VCH GmbH.

Figure 4 .
Figure 4. Relative importance of individual effector mechanisms conferred by YF17D-vectored vaccines contributing to protection against various target antigens (see Table1for details).Neutralizing antibodies (nAb) prevent cell entry, whereas non-neutralizing binding antibodies mediate cell death (i.e.antibody-mediated cell death, ADCC) by recruiting cytotoxic CD16 + NK cells to infected cells expressing pathogen-derived antigens on their surface.YF17D-specific cytotoxic CD8 + T cells (CD8 + CTL) target infected cells displaying viral epitopes on their MHC-I molecules.Activated CD4 + T cells provide support for antibody production and CD8 + T-cell activation.Size of circles to indicate relative contribution; ?, precise role under debate or unknown.EBOV, ebola virus; HBV, hepatitis B virus; JEV, Japanese encephalitis virus; LASV, lassa virus; ZIKV, Zika virus.

Table 1 .
List of YF17D-based recombinant vaccines and vaccine candidates.