Unique safety issues associated with virus-vectored vaccines: Potential for and theoretical consequences of recombination with wild type virus strains

2.2. Virus recombination

Recombination describes a process by which nucleic acid sequences from two different parental viruses are exchanged so that the progeny contain sequences derived from both parents. Both RNA and DNA viruses may undergo recombination when two related genomic variants of a virus co-infect a cell. In viral systems there are three different mechanisms of recombination, dictated by the structures of the viral genomes. For DNA viruses, recombination occurs by the physical breakage and rejoining of parental DNA molecules through regions of sequence homology, in a fashion similar or identical to the same process in bacteria or higher organisms. For RNA viruses containing segmented genomes, gene exchange occurs primarily through reassortment of individual parental genome segments into progeny viruses, however intragenic recombination has also been reported for the segmented orthomyxoviruses, reoviruses and bunyaviruses [12], [13], [14], [15], [16]. Recombination has been observed in several single-stranded RNA (ssRNA) virus families representing both positive and negative sense genomes both in the laboratory and in the wild; picornaviruses, coronaviruses, togaviruses and retroviruses, all with positive sense ssRNA genomes, display relatively efficient recombination [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31]. The frequency of recombination among negative sense RNA viruses (excluding reassortment of segmented genomes) seems to be relatively low [31]. Recombination in RNA viruses, including retroviruses, is thought to occur during replication via “copy choice”, namely switching RNA templates during replication with the result that the newly synthesized genome contains sequences from two different parental molecules [32], [33].

While recombination clearly requires coinfection of a cell with two different viruses, the circumstances leading to such a coinfection in vivo are not clearly understood. Coinfection could theoretically result from infection with a heterogeneous population of viruses, by simultaneous or overlapping serial infections with different viruses, or by infection of an individual harboring a persistent, latent or reactivated infection with a different virus. Nevertheless, recombination among viruses in the human population clearly occurs as exemplified by a recent study describing interclade recombinants of varicella zoster virus [34].

2.3. Vaccine viruses in the wild

Although experience with virus-vectored vaccines in humans is limited, perspective on the issue of recombination between virus-vectored vaccines and wild type viruses can be informed by experience with traditional live, attenuated human virus vaccines and virus-vectored veterinary vaccines. Vaccine viruses may establish a long-term reservoir in the wild, and recombination between attenuated vaccine strains and circulating wild type viruses or even between two different live attenuated vaccine strains has been documented. Specifically, evidence exists that vaccinia virus used as vaccine during the smallpox eradication campaign in Brazil has established a durable reservoir in the wild and is the cause of numerous cowpox-like infections in cattle and humans [35]. Likewise, the bovine herpesvirus vaccine may establish a latent reservoir in vaccinated animals which, through reactivation, may spread to other animals [36]. Numerous examples document probable recombination between live attenuated vaccine viruses and wild viruses. Phylogenetic analysis revealed recombination between wild, circulating strains of Newcastle disease virus (NDV), an avian paramyxovirus, and attenuated NDV vaccine strains [37]. Attenuated viruses contained in the oral poliovirus vaccine frequently recombine with related indigenous human enterovirus strains to produce circulating vaccine-derived polioviruses (cVDPV), which can cause paralytic disease [38]. A reassortant Rift Valley Fever virus strain containing both wild type virus-derived and vaccine virus-derived genomic segments was isolated from a patient who received a needle stick injury while vaccinating sheep [39]. Analysis of two independently isolated disease strains of the pestivirus bovine viral diarrhea virus (BVDV) demonstrated that these strains arose via both homologous and non-homologous recombination between a persistent BVDV strain and a vaccine strain, resulting in evolution of strains with enhanced pathogenicity relative to the parental strains [40]. Analysis of a strain of the poxvirus myxoma, originally isolated from a wild rabbit, suggests that it resulted from a recombination between a wild myxoma strain and a vaccine strain [41]. Genome analysis of a disease strain of porcine reproductive and respiratory syndrome virus (PRRSV), an arterivirus, demonstrated that it is a recombinant between an attenuated PRRSV vaccine strain and a field strain [42]. Additionally, recombination between independently derived attenuated avian herpesvirus vaccine strains can give rise to circulating pathogenic recombinant viruses [43]. Lastly and most extraordinarily, evidence exists that the retroviral disease reticuloendotheliosis was introduced into avian populations via contamination during development of fowlpox (a poxvirus) and Marek's disease (a herpesvirus) vaccines, and now circulates in the wild as an integrated provirus in some fowlpoxvirus genomes [44]. All of these examples attest to the potential for genetic interaction between vaccine viruses and viruses in the wild. By contrast, it is noteworthy that more than 100 million doses of poxvirus-vectored recombinant rabies virus vaccine has been distributed in the wild in the United States, Eurasia and Western Europe resulting in reproducible reduction or elimination of wild rabies where applied and with no reports of recombination with wild type virus strains, attesting to the utility of recombinant vectored vaccines [45]. Furthermore, a wide variety of viral-vectored recombinant vaccines have been routinely used in veterinary practice for well over a decade with no recorded evidence of problems resulting from recombination with wild viruses [11]. Thus, while the occurrence of a recombination event is likely to be rare, the unanticipated consequences of such an event must be considered by developers and regulators alike.

3.1. Intrinsic recombination properties of the parental viruses

As noted above, different virus families are associated with different intrinsic frequencies of recombination, and these intrinsic properties will affect the probability that recombination will take place. Generally, DNA viruses are subject to relatively high frequencies of recombination. Although RNA viruses generally display lower recombination frequencies compared to DNA viruses, retroviruses and some positive stranded RNA viruses (picornaviruses, coronaviruses, togaviruses, noroviruses), readily recombine. Other positive stranded RNA viruses recombine only inefficiently, and while recombination among negative stranded RNA viruses can be demonstrated on an evolutionary scale, the frequencies are sufficiently low as to make recombination under laboratory conditions difficult to detect. Segmented viruses display extremely high rates of reassortment, and concern has been expressed that live attenuated influenza vaccines might readily reassort with circulating wild type strains [51]. One must assume that vectors based on virus families with segmented genomes or families prone to high recombination rates would also be most prone to recombination with related wild type viruses should the opportunity for recombination arise. For viruses of any given family, the intrinsic rates of recombination both in cell culture and in laboratory animals can often be determined experimentally, and recombination among related viruses in the wild can sometimes be deduced based on phylogenetic analyses. The intrinsic recombination properties of any given virus should be taken into account during vaccine vector development.

3.2. Sequence relatedness of vector and wild virus

For both DNA and RNA viruses, the majority of observed recombination events occur through regions of nucleic acid sequence homology between parental genomes. Thus, any sequence alterations to a vector that reduce homology with the wild type virus should theoretically reduce the probability of recombination between vector and wild type virus. Viruses can be attenuated by changing the genome sequence to substitute less common codons [52]. Sequences can be changed in silent positions such that the original amino acid sequence of the gene is preserved, but multiple substitutions of less frequently used codons, “codon deoptimization”, results in loss of virulence. Attenuation can also be achieved by the introduction of many uncommon combinations of pairs of codons, thus changing the “codon pair bias” [53]. Vectors engineered in this fashion should theoretically have a reduced frequency of recombination with wild type homologs. Alternatively, vectors intended for human use which are based on naturally occurring non-mammalian viruses, for example the use of canarypox virus as an HIV vaccine vector [54], are sufficiently divergent in sequence homology from mammalian viruses such that the probability of recombination between the vector and a mammalian virus in the vaccinated host should be lowered. Lastly, gross alterations in gene arrangement of a vector could suppress productive recombination with a wild type virus homolog [55], [56], [57]. Notably, non-homologous recombination among viruses does occur, albeit at a relatively low frequency, and manipulations to influence homologous recombination will theoretically not affect non-homologous recombination events. Furthermore, sequence homology is irrelevant in reassortment of genome segments in segmented viruses.

3.3. Host range, pathogenesis and transmission

The probability of recombination between two viruses should be directly proportional to the probability that a cell will be co-infected by the two parental viruses under consideration. The probability of co-infection, in turn, may theoretically be influenced by virus host range, pathogenesis (i.e. whether infections are normally acute, persistent, chronic or latent) and transmission. Considering the influence of host range, use of an avipoxvirus vector in humans or non-avian vertebrates should limit the potential for recombination of the vector with wild avipoxviruses [54]. Likewise, use of the avian Newcastle disease virus (NDV) as a vector in humans would present minimal opportunity for co-infection with the wild type avian virus, thus limiting the opportunity for recombination [49]. By contrast, use of a human adenovirus as a vaccine vector carries the clear potential for genetic interaction between the vector and circulating human adenoviruses. Considering the influence of pathogenesis, one would expect that a virus that in the wild causes a latent infection in the target population would provide a greater opportunity for recombination compared with a virus that results in an acute infection. As a specific example, herpes viruses have been proposed as vaccine vectors [58]. Because a large fraction of the human population carries several different herpes viruses in a latent state, these latent wild type genomes could theoretically provide stable, persistent populations of virus genomes in a target population which could recombine with vaccine vectors. By comparison, if the vector parent usually undergoes an acute infection such as in the flavivirus yellow fever, the opportunities for recombination would be relatively rare. Transmission mechanism could also theoretically affect the probability of recombination between a viral vaccine vector and a circulating virus. For example, different species of arthropod-borne flaviviruses exhibit differences in observable recombination frequencies in nature attributable to differences in mechanism of vectoring by ticks and mosquitoes and by differences in both host and vector ecology among different flaviviruses [30].

3.4. Replication competency of vector in target host

The probability of recombination between a virus-vectored vaccine and a wild virus should be proportional to the virus load in a vaccinated individual, and thus limited by the replication competency of the vectored vaccine. Virtually all vaccine vectors are either defective for virus replication or naturally or artificially attenuated for pathogenicity in their target populations, and attenuation sometimes equates to reduced virus replication, hence reduced virus load and reduced opportunity for recombination. “Naturally” attenuated vectors comprise those in which a virus specific for one animal species is used as a vector for vaccination of another normally non-permissive species. Examples include canarypox virus and Newcastle disease virus engineered for use in humans as described above [49], [59]. Canarypox in particular undergoes an abortive infection in non-avian cells, virtually nullifying the opportunity for recombination [60]. Vectors may be artificially attenuated using classic methods of serial passage in vitro or more modern methods involving engineered gene deletion or rearrangement. Modified vaccinia Ankara provides an example of a poxvirus vector attenuated by serial passage [61], the MRKAd5 HIV-1 clade B gag/pol/nef vaccine provides an example of an adenovirus vector artificially attenuated by deletion of the essential viral E1 regulatory region [62], and vesicular stomatitis virus vectors have been attenuated by rearrangement of gene order in the viral genome [57].

3.5. Mechanism of attenuation

The specific mechanism of attenuation impacts the consequences of recombination between a vaccine vector and a wild type virus. Specifically, if attenuation is genetically linked to transgene expression, reversion to virulence seems unlikely. For example, as noted above, in an adenovirus vector the transgene encoding an immunizing antigen may be inserted into a deleted E1A region [62] so that any recombination event that restores virulence to the vector also deletes the transgene and conversely transfer of the transgene to a wild type virus also transfers attenuation. Thus, theoretically, no recombinant should be more virulent than the vaccine vector itself. By contrast, in an attenuated vaccinia vector (MVA) multiple genes are mutated or deleted to confer attenuation, and the transgene is expressed only from one site [61]. Recombination with a wild orthopox virus (cowpox virus, vaccinia virus or monkeypox virus) could in theory either “repair” some virulence mutations or transfer the transgene into a wild type virus background, generating a recombinant with improved replication properties relative to the original vaccine vector and also leaving the transgene expression intact.

3.6. Additional factors potentially affecting virulence

Vector designs have been proposed that incorporate expression of immune modulatory molecules, for example interleukins, to stimulate or otherwise regulate the immunogenicity of the recombinant vaccine [63]. However, in some cases these genes may act as virulence factors. For example, in a model poxvirus system, expression of interleukin-4 (IL-4) from ectromelia virus (a natural pathogen of mice; mousepox virus) enhanced the pathogenicity of the virus and conferred resistance of the recombinant virus to pre-existing immunity in infected animals [63], [64]. Theoretically, recombination of a vector expressing such a virulence factor with a wild type virus could result in production of a wild virus with increased virulence.

We thank the following for helpful advice: (1) Mark Pallansch, David Bloom, Kathy Spindler, Pennathur Sridhar, Ann Arvin, Harry Greenberg, Nancy Davis, Valery Grdzelishvili; and (2) other members of the V3SWG during the preparation of this document: Baevin Carbery, Louisa Chapman, David Garber, Najwa Khuri-Bulos, Bettina Klug, Sonali Kochhar, and Janet McNicholl.