Molecular Genetic Insights Into Varicella Zoster Virus (VZV), the vOka Vaccine Strain, and the Pathogenesis of Latency and Reactivation

Abstract

Genetic tools for molecular typing of varicella zoster virus (VZV) have been used to understand the spread of virus, to differentiate wild-type and vaccine strains, and to understand the natural history of VZV infection in its cognate host. Molecular genetics has identified 7 clades of VZV (1–6 and 9), with 2 more mooted. Differences between the vOka vaccine strain and wild-type VZVs have been used to distinguish the cause of postimmunization events and to provide insight into the natural history of VZV infections. Importantly molecular genetics has shown that reinfection with establishment of latency by the reinfecting strain is common, that dual infections with different viruses can occur, and that reactivation of the superinfecting genotype can both occur. Whole-genome sequencing of the vOka vaccine has been used to show that vesicles form from a single virion, that latency is established within a few days of inoculation, and that all vaccine strains are capable of establishing latency and reactivating. Novel molecular tools have characterized the transcripts expressed during latent infection in vitro.

Varicella zoster virus (VZV), with a genome of approximately 125000 base pairs and 71 open reading frames (ORFs), is the smallest of the human herpesviruses and the first to be fully sequenced [1]. The virus is known to encode at least 7 clades, designated 1–5 at an international meeting in 2010 [2] with a further 2 confirmed [3–5] and 2 more proposed but awaiting confirmation [5]. Early investigations using restriction enzyme analysis noted that variation between unrelated VZV strains was less than between herpes simplex virus strains [6]. In 1995, Takayama et al used long polymerase chain reaction (PCR) analysis to amplify fragments of DNA from 6.8 to 11.4 kb from 40 Japanese viruses [7]. Restriction digests of the fragments, which covered ORFs 12–16, 38–43, and 54–60, were generated with 10 enzymes, yielding 12 variable restriction sites. Comparison with the only full-length sequence then available, the Dumas strain, showed 28 substitutions in 65000 nucleotide residues analyzed, giving an estimated interstrain variation of 0.043% (1 in 2300 bases) [8]. Our data from heteroduplex mobility assays of 37 fragments across 10 VZV genomes found 92 nucleotide substitutions, compared with the Dumas strain sequence, in 15059 nucleotides, giving a variation of 0.061% (1 in 1637 nucleotides) [8]. Both estimates were considerably lower than those for herpes simplex virus (0.32%–0.81%), cytomegalovirus (2.5%), human herpesvirus type 8 (1.5%–2%), and pseudorabies virus (2%–3%) [8].

Several groups working in this area used single-nucleotide polymorphisms (SNPs) and whole-genome sequencing to identify distinct VZV clades [8–10], and in 2010 an international meeting agreed on a common terminology for these [2]. Clades 1–5 were defined from the start, while single representatives of 2 putative new clades were also identified and were designated by the Roman numerals VI and VII until further examples of these clades could be found. Since then, whole-genome sequencing has confirmed clade 6 [3]; uncovered 2 further clades, clade VIII and IX [5]; and, most recently, confirmed clade 9 [4]. These clades can be distinguished by a new SNP typing schema [4], which replaces the original, first proposed in 2010 [2]. However, the recent development of high-throughput next-generation sequencing methods that can generate whole VZV genomes from the low amounts of DNA found in clinical samples are proving as cheap and tractable for VZV genotyping, especially if the variable regions are not required [11]. With whole-genome sequencing, a maximum variation of 0.2% (1 in 1000 bases), excluding variable regions, between clades 5 and 2 [12] has been observed in one study, while in another study clades 1 and 2 were the most divergent [12, 13]. In addition, intraclade diversity has been calculated to range from 0.03% (1 in 3333 bases) to 0.07% (1 in 1429 bases), using sequences from 8 clade 1 viruses [13]. Whole-genome sequencing has also confirmed early experiments showing recombination between different strains [3, 14].

Molecular Epidemiology of VZV

Before the advent of routine whole-genome sequencing, SNP markers that distinguished between clades fueled extensive studies of VZV molecular epidemiology. Much of this work was driven by the need to distinguish between the vOka vaccine strain and circulating wild-type viruses. Two restriction sites, a Bgl1 site in ORF 54 and a Pst1 site in ORF 38, can be used to distinguish Japanese-derived vaccine and wild-type strains from US wild-type strains [15, 16], although these clade-based markers will not distinguish wild-type clade 2 from vaccine strains that are also clade 2. We used the SNPs delineating these sites to survey >240 VZV strains from varicella and herpes zoster cases collected from our local population [17]. In common with the findings by La Russa et al that 3 of 20 wild-type US strains were Bgl1 positive [16], we found 20 of 244 viruses (8%) circulating in London to be Bgl1 positive [16, 17]. All were Pst1 positive. This confirmed that United Kingdom wild-type strains could be distinguished in the same way as US viruses from the Japanese vOka vaccine, a finding of public health utility should the vaccine be introduced into the United Kingdom [18].

The development of more-extensive SNP based assays enabled us and others to demonstrate the geographical segregation of clades. Clades 1 and 3 predominated in Europe and in countries that had experienced significant European immigration, clade 5 predominated in Africa and in countries with had African migration, clade 2 predominated in Japan and other countries in East Asia, and clade 4 was present in Asia. Clade 6 has since been found in Europe [3]. Making use of the finding that clades 1 and 3 were negative for the Bgl1 restriction site in ORF 54, whereas clades 2, 4, and 5 were Bgl1 positive, we performed several studies of VZV epidemiology. On the premise that the virus isolated from a patient with zoster is identical to the original infecting strain that caused varicella, we showed that, among the cases of zoster, Bgl1-positive viruses were significantly more common (P < .05) in subjects who had immigrated from Africa, India, Asia, and East Asia than in subjects who had grown up in the United Kingdom [17]. In contrast, we found no strong association of Bgl1-positive VZV with immigrants. Instead, there was a significant rise (from 5% to 40%; P < .001) over 25 years in the prevalence of Bgl1-positive varicella cases, suggesting that Bgl1-positive viruses had been introduced with immigration and were now spreading in the United Kingdom [17]. Although limited in its scope, this study provided the first indication that Bgl1 and Pst1 genotyping was useful for investigating questions about VZV strain distribution, spread, and possibly even pathogenesis, especially in view of the simplicity and low cost of this analysis.

Further analysis of viruses from around the world confirmed that the Bgl1-positive genotype was present in 100% of 100 viruses collected from different countries in Africa, East Asia, and the Indian subcontinent but accounted for <20% of viruses collected in Europe, the United States, and other countries settled by Europeans [19]. A prospective study of >400 patients presenting with clinical zoster allowed us to type VZV from 200 white British-born subjects aged 5–98 years [19, 20]. Again, assuming that the VZV detected in these patients was identical to the virus that caused varicella and that, having grown up in the United Kingdom, their episode of varicella occurred at ≤10 years of age, we showed that Bgl1-negative viruses (clades 1 and 3) accounted for 80%–90% of viruses circulating in London over the past 100 years [20]. From this, we concluded that the data obtained from opportunistic sampling of viruses from cases of varicella and zoster, which is the basis of most molecular epidemiological studies performed to date, provided a reasonably accurate picture of VZV strain prevalence across the world [20]. Moreover, our data confirmed that Bgl1 genotyping can be used to distinguish strains of European (Bgl1-negative) and non-European (Bgl1-positive) origin and, thus, is a useful molecular epidemiological tool [20, 21]. These simple tools, now used by many, support the geographical distribution of VZV clades and provide further confirmation that Bgl1-positive African and Asian clade 4 and 5 strains are spreading in countries with European populations [17, 20, 22–24].

In contrast, very few European clade 1 and 3 viruses and no Japanese clade 2 viruses have been found in surveys of >100 strains from African countries (ie, Guinea Bissau, Zambia, Sudan, Chad, and the Democratic Republic of the Congo) [10, 19, 25]. In 2 cases, clade 1 viruses, which were genotyped as part of an outbreak in Guinea Bissau, were in fact imported by children returning from Europe [26]. Despite the presence of many household members who later acquired a local African strain of VZV, both imported European strains spread only to a single household contact. A few European clade 3 viruses causing atypical palmar and plantar lesions, which were confused clinically with monkeypox, have also been found in the Democratic Republic of the Congo [27], and one clade 3 virus has been opportunistically identified in Sudan, again as part of genotyping of a monkeypox outbreak [28]. Nonetheless, the impression is that non–clade 5 viruses have not spread in Africa. Molecular genotyping of viruses in an outbreak occurring in Guinea Bissau has confirmed that VZV transmission generally is impaired, compared with transmission during outbreaks in temperate climates, but can be overcome to some extent by high-density living and larger family size [26]. While environmental factors are thought to be of most importance, the possibility that strain differences are responsible cannot be excluded. Further studies would be informative, such as in areas of large-scale European settlement, like South Africa.

Molecular Genetics and VZV Pathogenesis

Molecular epidemiological tools have also provided insights into VZV pathogenesis. Restriction enzyme analyses to genotype VZV causing varicella and zoster in the same patient proved Hope Simpson’s hypothesis that zoster is due to reactivation of latent VZV originally acquired from chickenpox [29]. This finding was confirmed in 3 additional patients [30]. All of these patients were immunosuppressed, with primary varicella and zoster occurring close together. In 2006, we made use of SNP genotyping to investigate a case of recurrent zoster in an immunocompetent man. This young man, who had a self-reported history of a single varicella episode at 5 years of age, presented as an adult with 2 episodes of zoster 3 years apart, with the first case in the ocular division of the left trigeminal nerve and the second in the left thoracic region [31]. Genotyping showed the first strain to be a clade 5 strain and the second to be a clade 4 strain. Microsatellite typing of human DNA associated with the strains proved that the host was identical in both cases, thus ruling out laboratory contamination. This suggested that infection with 2 different viruses had occurred and that both had established latency and reactivated. The result overturned previous dogma that zoster is always caused by the strain of virus that caused primary varicella. Furthermore, the data corroborated findings from earlier studies, which, using Bgl1 genotyping, had shown that zoster due to Bgl1-negative viruses (clades 1 and 3) occurred in 30% of adults who had immigrated to the United Kingdom from areas where Bgl1-positive (clade 2/4/5) viruses circulate [19]. None of these adults had a history of varicella while in the United Kingdom. Together with studies of healthcare workers who received vOka vaccine and developed proven infection due to wild-type virus after having seroconverted to the vaccine strain [32], these findings confirm that asymptomatic reinfection of latently infected individuals is part of the natural history of VZV, with establishment of latency plus reactivation, at least in some cases, by the second strain. These findings have been further corroborated by studies showing that 86% of immunized children have wild-type VZV within their enteric neurons despite lacking a history of varicella [33]. Importantly, most reinfections are asymptomatic with no evidence of skin rash, supporting the likelihood that hematogenous transfer of virus to ganglia is possible [34].

The evidence that VZV is highly recombinant provides conclusive evidence that coinfections must be occurring, although where in the body the process of viral recombination takes place remains a matter for conjecture [3, 12, 35]. Intriguingly, genotyping of viruses from varicella outbreaks in the United Kingdom and Guinea Bissau showed that multiple strains can cocirculate within a single outbreak. In the former outbreak, there were strains from different clades, and in the latter there were 4 distinct strains from clade 5 [36, 37]. We have also demonstrated coinfection with 2 different viruses within the same individual [38]. In that study, limiting dilution PCR analysis allowed genotyping of single molecules, providing evidence that a swab of vesicle fluid from a child with chickenpox contained clade 1 and clade 3 viruses in a ratio of 3:1.

Molecular Epidemiology of the Vaccine Strain

Where molecular genetic tools have been most extensively applied is in distinguishing the live attenuated vOka vaccine strain from circulating wild-type strains. This is particularly important as 2%–5% of vaccinees can develop rashes or other complications following receipt of vOka vaccine. The vOka vaccine was developed in Japan in 1974 to reduce severe or fatal complications of varicella among immunocompromised children [39]. The vaccine, named after the child from whom the original virus (known as the parental Oka strain [pOka]) had been obtained, was attenuated through serial passage of the isolate in guinea pig embryo cultures, with the attenuated virus then propagated in human diploid cells (WI-38) [40, 41]. The vOka strain is attenuated for replication in skin but less so in other target tissues, such as T cells and trigeminal ganglia [42, 43]. The live attenuated vOka strain (GenBank accession number AB097932) comprises a mixture of VZV genotypes [44] and, by deep sequencing of multiple batches, has been shown to differ from wild-type Oka strains by at least 224 SNPs [45]. Which of these is primarily responsible for attenuating the virulence of VZV in live vaccines remains unclear, but vaccine mutations at positions 106262, 107252, and 108111 in ORF 62 are believed to play key roles [46]. Although highly prevalent in vOka vaccine and in rashes caused by the vaccine, nucleotides at these positions have been shown not to be fixed, with up to 6% of clones from the vaccine testing positive for the wild-type nucleotide at these positions [47]. Nucleotides at a further 11 positions have been shown to be significantly more likely to be wild type in vaccine viruses recovered after immunization from individuals with rash, encephalitis, and retinitis, suggesting a role for the vaccine allele in attenuating replication in these tissues [45]. Of these, a stop codon at position 560 in ORF 0 has been shown to reduce VZV replication in epithelial xenografts in SCID-hu mice [48], while a substitution of lysine for proline at position 446 in a transactivating region of ORF 62 is always wild type in rashes (P = 7.25 × 10^–11), suggesting a critical role in the recovery of replicative ability in skin [45, 49]. Both pOka and vOka strains are, like many Japanese viruses, of clade 2 origin and are positive for the Bgl1 and Pst1 restriction sites. Genotyping of viruses recovered following clinical events in patients who have been vaccinated with the vOka strain, using the Bgl1, Pst1, and the vaccine SNPs 106262, 107252, and 108111, which are at near fixation, has been performed in postsurveillance studies of varicella and zoster vaccines. Using these data, it has been shown that vaccine-associated rashes occurring ≤14 days after vaccination are largely due to wild-type virus, while those occurring >14 days after vaccination are more likely to be caused by the vOKa strain [50, 51]. An important finding from studies in leukemic children has been that herpes zoster due to the vOka strain occurs less frequently (approximately 7 times less often) than herpes zoster due to wild-type virus, confirming that reactivation of the vOka strain is impaired [52]. While vOka associated events, albeit mostly benign, occur in 2%–5% of children immunized against varicella, very few vOka strain–associated events have been documented following use of the more concentrated vOka vaccine against herpes zoster. Three cases of herpes zoster due to the vOka strain have been identified. One led to disseminated infection and death in an immunocompromised man. The third case resulted in retinitis; virus recovered from the vitreal fluid was sequenced and found to carry some of the 11 mutations associated with an increased risk of vOka complications [53]. At least 3 commercial vOka preparations are available for prevention of varicella, and 1 is available for prevention of zoster.

Molecular Genetics of the Vaccine and VZV Natural History

The unique opportunity posed by having a live vaccine that results in varicella-like and zoster-like rashes prompted us to compare the genome sequences of the infecting virus (ie, the vaccine virus) with that of the virus recovered from rashes. This provided a model for the natural history of VZV, a human-restricted virus for which there are no tractable animal models that accurately reproduce its natural history. Using statistical genetic analyses, we established that the live vaccine is highly stable from batch to batch [45, 54]; that, following inoculation, all viruses within the vaccine are capable of establishing latency and reactivating [45]; and that there were no neurotropic variants within the vaccine and no evidence of selection acting on any of the vOka variants [55].

Mathematical modeling of the results yielded 3 important findings. First, the vaccine virus evolves rapidly, at a rate of approximately 10^–3 substitutions/site/year. This is about 10 times faster than has previously been calculated from dated-tips calculations for short-term evolution of double-stranded DNA viruses [55]. A rate of 10^–6–10^–5 substitutions/site/year was calculated from circulating VZV transmitted in varicella outbreaks [36]. Conversely, a rate of 10^–9 substitutions/site/year was based on a small subset of highly conserved genes and calculated over long periods [56]. The new estimates of VZV mutation rates date the origins of the clades to between 20 and 50000 years ago, which is earlier than hitherto supposed but fits with other estimates [35] and with evidence for considerable rates of recombination [3] and of global spread of African and Asian strains associated with population migrations [3]. Second, we were able to show that most individual VZV vesicles arise from a single virion, with a few apparently originating from up to 3 virions [45]. Individual vesicles within the same zoster rash harbored different varicella strains. Third, we were able to show that mutation rates for viruses recovered from varicella rashes were several logs higher than for viruses recovered from zoster rashes. In the former, the mutation rate remained constant with time, while in the latter, the mutation rate declined over time [55].

The best explanation for these findings is that viruses reactivating to cause herpes zoster are not replicating during the time that they are latent in neurons and therefore do not acquire mutations. When the mutation rates derived from replicating viruses that had caused varicella were applied to the zoster-associated viruses, we calculated that the latter viruses had replicated for a mean of 13 days during the period of establishing latency and reactivating to cause the zoster rash [55]. These findings provide a unique insight into VZV natural history in its cognate host.

New Applications of Molecular Genetic Technologies to Transcriptomes

The studies of VZV vOka natural history were made possible by the development by our group of new technologies that allow recovery and enrichment for very-low-copy-number viral genetic material from clinical samples for whole-genome sequencing [11]. In studies of VZV pathogenesis, these methods were used to recover viral genomes. By applying these methods to transcriptomes recovered from VZV-infected cells, it has been possible to examine the viral transcripts elaborated by wild-type– and vOka–infected cells, even when these are at low levels [57]. Analysis of the transcriptome of induced pluripotent stem cell neurons infected with VZV provided data on the behavior of wild-type parental strain pOKa and vOka VZV in a new in vitro model of VZV latency and reactivation. The results suggest that low-level transcription is occurring across the genome during both pOka and vOka latency in this model but that the level is lower for vOKa. Both pOka and vOka were able to establish latency, but pOKa reactivated nearly 5 times more frequently [57].

Most recently, the same method has been used to demonstrate the expression of a unique antisense VZV latency-associated transcript during viral latency in trigeminal ganglia neurons [58]. In summary, molecular tools have provided unique insights into the molecular epidemiology, transmission, and pathogenesis of VZV and the live attenuated vOka vaccine strain.

Notes

Supplement sponsorship. This work is part of a supplement sponsored by the Royal Society of Medicine (Royal Charter number RC000525) funded through unrestricted educational grants from Merck, Sanofi Pasteur MSD, The Research Foundation for Microbial Diseases of Osaka University, Seqirus and GlaxoSmithKline.

Potential conflicts of interest. Author certifies no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References