|Home | About | Journals | Submit | Contact Us | Français|
The live attenuated Oka varicella vaccine (vOka), derived from clade 2 wild-type (wt) virus pOka, is used for routine childhood immunization in several countries, including the United States, which has caused dramatic declines in the incidence of varicella. vOka can cause varicella, establish latency, and reactivate to cause herpes zoster (HZ). Three loci in varicella-zoster virus (VZV) open reading frame 62 (ORF62) (106262, 107252, and 108111) are used to distinguish vOka from wt VZV. A fourth position (105705) is also fixed for the vOka allele in nearly all vaccine batches. These 4 positions and two vOka mutations (106710 and 107599) reportedly absent from Varivax were analyzed on Varivax-derived ORF62 TOPO TA clones. The wt allele was detected at positions 105705 and 107252 on 3% and 2% of clones, respectively, but was absent at positions 106262 and 108111. Position 106710 was fixed for the wt allele, whereas the vOka allele was present on 18.4% of clones at position 107599. We also evaluated the 4 vOka markers in an isolate obtained from a case of vaccine-caused HZ. The isolate carried the vOka allele at positions 105705, 106262, and 108111. However, at position 107252, the wt allele was present. Thus, all of the ORF62 vOka markers previously regarded as fixed occur as the wt allele in a small percentage of vOka strains. Characterization of all four vOka markers in ORF62 and of the clade 2 subtype marker in ORF38 is now necessary to confirm vOka adverse events.
Varicella-zoster virus (VZV) is the first human herpesvirus for which a vaccine has been licensed. In 1995, the United States became the first country to implement routine varicella vaccination for healthy children aged 12 to 18 months (12), resulting in a dramatic decline in varicella morbidity and mortality (9, 33). More recently, a higher-dose formulation of the same vaccine seed virus was licensed and recommended for the prevention of herpes zoster (HZ) in persons ≥60 years of age (4).
Differentiation of the live attenuated Oka varicella vaccine (vOka) from wild-type (wt) virus has become important for at least two reasons. Testing can be used to assess vOka effectiveness by identifying cases of breakthrough varicella caused by wt virus. Breakthrough disease occurred in 3 to 25% of vOka recipients in outbreak settings (5, 23, 50) and is severalfold more likely to occur among vOka recipients with low VZV IgG levels at 6 weeks postvaccination (<5 glycoprotein enzyme-linked immunosorbent assay [gpELISA] U/ml) (23, 33)). Strain discrimination testing is also used to document adverse events associated with vOka. Serious adverse events due to vOka are rare, with only 8 laboratory-confirmed reports of meningitis or encephalitis (6, 7, 12, 17, 21, 27, 28, 37) and 7 cases of secondary transmission (13, 17, 18, 25). One of the most common complications postimmunization is a varicella-like rash that occurs within 1 to 6 weeks postimmunization (6, 17, 43). The incidence of rash is approximately 5% in healthy children (14, 49). vOka can also establish latency and reactivate to cause HZ. While the incidence of wt HZ after vaccination has declined (8, 19, 26, 50, 51), the incidence of HZ caused by vOka is less well defined, as most recipients of the vaccine are children, in whom HZ (caused by vOka or wt virus) is uncommon. In addition, very few HZ cases occurring postvaccination are identified as attributable to vOka versus wt virus by laboratory testing.
vOka is a live attenuated virus produced by serial passage of pOka in tissue culture (46). Three preparations of vOka are in commercial production, vOka/Biken (Biken Institute, Japan), Varivax (Merck & Co.), and Varilrix (GlaxoSmithKline [GSK]). The complete DNA sequence of the vOka/Biken genome revealed base substitutions at 42 loci compared with pOka, over a third of which clustered in the major viral gene transactivator protein encoded by open reading frame 62 (ORF62) (16). Most of the vOka-specific loci were determined to be mixtures of the wt and vOka nucleotides, revealing that the vOka is a heterogeneous population of viral strains. This has since also been confirmed for the Varivax and Varilrix preparations, which are derived from the same seed stock (48). Genetic variation has been reported between all 3 vaccine preparations and among different lots from the same manufacturer (22, 42, 47, 48). Differences in the production of these vaccine preparations probably account for some of this variation. The original vOka/Biken was produced through passage of pOka 11 times in human embryo fibroblast cells at 34°C, 12 times in guinea pig embryo fibroblast cells at 37°C, and a further three times in human WI-38 and MRC-5 cells (46). Varivax was produced by a further seven passages of vOka/Biken-infected cells in MRC-5 cells (44), whereas Varilrix was produced through a series of five limiting-dilution clonings using cell-free virus, resulting in a further 6 passages in W1-38 cells and 8 passages in MRC-5 cells (10).
Five confirmed and 2 provisional wt VZV clades have been described (clades 1 to 5, VI, and VII) (3). The clades are differentiated by evaluating a panel of single nucleotide polymorphisms (SNPs) in ORF1, -21, -22, -50, and -54 (1, 24, 30). However, these SNPs do not differentiate vOka and similar strains from other clade 2 viruses (24). A PstI restriction site in ORF38 across SNP 69349, which is present in clades 1, 3, 4, 5, VI, and VII and in 70% of clade 2 strains but absent in vOka and 30% of clade 2 wt viruses (20, 30, 45), is required for that purpose. This PstI site and a BglI restriction site across SNP 95241 in ORF54 were initially used to distinguish vOka from other wt strains in countries such as the United States, where no vOka-like (clade 2 PstI−) wt strains had been identified. Recently, however, clade 2 PstI− wt isolates have been identified in the United States and in Australia, rendering that approach problematic (30, 37).
Limits in the sensitivity of conventional sequencing methods for detecting low-level variant strains (typically <10%) have left unanswered the possibility that the observed differences in the SNP profiles of various vaccine preparations may in fact simply reflect a failure to detect variant strains that are present in very low numbers. Whole-genome sequencing of the 3 vaccine preparations and early analysis of vOka-associated SNPs were performed using direct Sanger sequencing (16, 48). This method has a detection limit of approximately 10% (22), and thus an allele present in a mixture but on <10% of genomes would have appeared to be fixed. Recent studies have used more sensitive techniques, including pyrosequencing, with sensitivity limits ranging on average from 2% to 10% (22, 47) and analysis of vaccine-derived clones (47). The latter method can theoretically detect an allele present on a single genome within a mixture of viruses.
Currently, three loci in ORF62, 106262, 107252, and 108111, are used to distinguish vOka from wt virus (i) because they were reported as fixed for the novel vOka allele in all 3 preparations (16, 22, 31, 48) and (ii) because they have been detected in every isolate from cases of lab-documented vOka-caused varicella or HZ. However, a recent study by Thiele et al. (47) reported that 106262 and 108111 were mixed in Varivax and that 107252 was fixed for the novel vOka allele. A fourth position (105705) was reported to be fixed in all but one batch of Varivax (22). Two loci in ORF62 of Varivax, 106710 and 107599, were reported to lack the vOka allele present in Varilrix and/or vOka/Biken (40, 47).
The aim of our study was to analyze full-length ORF62 clones derived from Varivax to further investigate the presence of alleles reported to be absent in this vaccine preparation and, if detected, to determine whether any of them cooccur in some strains. Related to this, we also describe a case of HZ caused by a unique vOka variant.
In March 2011, approximately 23 months after receiving 1 dose of vOka (Varivax; Merck & Co.) (vaccine lot number not indicated), an otherwise healthy 3-year-old female child was hospitalized in New York City, NY. She presented with a rash consisting of 50 to 249 (exact number not specified) skin lesions on her left arm, affecting dermatomes C5, C6, and T1. The patient's guardian reported no history of varicella or HZ. The patient had been on acyclovir for 2 days prior to specimen collection. A swab was taken from a vesicle 2 days after HZ onset. On-site testing identified the presence of VZV DNA and absence of herpes simplex virus DNA by PCR. The patient was found to be VZV IgM negative and IgG positive.
For the HZ case, total nucleic acid was extracted from 200 μl of vesicular fluid using the MagNA Pure LC automated NA purification system (Roche) according to the manufacturer's instructions and was eluted in 100 μl of elution buffer. For evaluation of the ORF62 SNPs on Varivax-derived clones, DNA was extracted from a batch of Varivax (lot LL46F, released in 2008) using a QIAamp DNA Minikit (Qiagen Ltd., United Kingdom) according to the manufacturer's instructions. Prior to extraction, the lyophilized vOka preparation was resuspended in 500 μl of the provided hydration buffer. DNA was extracted from 2 200-μl aliquots and 1 100-μl (mixed with 100 μl of PBS) aliquot. The 3 200-μl aliquots of eluted DNA were merged to form a 600-μl DNA pool which was used for subsequent experiments.
Detection of VZV and discrimination between wt and vOka were performed at position 106262 using a Förster resonance energy transfer (FRET)-based real-time PCR assay as previously described (29) and at position 107252 using an in-house FRET assay. All FRET PCRs were performed using the LightCycler (Roche). A TaqMan-based real-time allelic discrimination assay, performed using the MX3000p platform (Stratagene), was used to analyze a third SNP (108111) in ORF62 that was reported to distinguish between vOka and wt strains (40, 48). Position 105705 was characterized by sequencing, which was also used to confirm results of the real-time PCR assays. Primers were designed to sequence SNPs 105705 (5′-CAAAGCGTGTTCTCTGTCGTCTG-3′), 106262 (5′-CTATGTGCCGCCTCGTCCA-3′), 106710 (5′-ATGATCAGAAGCCTCACATCCTCCG-3′), 107252 and 107599 (5′-GGTGTCTCCCTAATCTTGTCG-3′), and 108111 (5′-TGCTGCCTGTAGTTTCACTTCCC-3′). Sequencing reactions were performed using the BigDye Terminator cycle sequencing kit version 1.1 and analyzed on an ABI prism 3100 Genetic Analyzer (Applied Biosystems) in accordance with the manufacturer's instructions. SNPs in ORF1, -21, -22, -38, -50, and -54 that differentiate the 7 wt clades of VZV were analyzed as follows. The BglI and PstI sites in ORF54 and -38, respectively, were characterized using an in-house FRET-based real-time PCR assay. SNPs in ORF1, -21, -22, and -50 were analyzed by PCR and sequencing as previously described (1, 30). PCR amplification of the entire ORF62 (4,087 bp) was performed using a forward (5′-CCCGCACAGACAGACAGACACT-3′) and reverse (5′-CTGCGAGAGCGTTTGGAAAACT-3) primer set. The PCR master mix included 12.5 μl Extensor Hi-Fidelity PCR master mix including buffer I (AB-0794; Thermo Scientific), 4 μl of forward and reverse primers (each at 6 pmol/μl), and 4.5 μl of template DNA. PCRs were performed in a GeneAmp 9700 PCR system (Applied Biosystems) with initiation for 2 min at 95°C followed by 10 cycles (94°C for 10 s, 55°C for 30 s, and 68°C for 4 min) and 20 cycles (94°C for 10 s, 55°C for 30 s, and 68°C for 4 min, increased by 10 s per cycle) of amplification, ending with a step at 68°C for 7 min. The PCR product was visualized by electrophoresis on a 1% gel at 85 V for 1.5 h.
Gel purification and cloning of the Varivax-derived ORF62 PCR product were performed using a TOPO XLPCR cloning kit with One Shot TOP10 chemically competent Escherichia coli (KA4750-10; Invitrogen) and suicide gene selection according to the manufacturer's instructions. Over three hundred colonies were picked, and colony PCR using M13 vector primers was performed to amplify a 4,335-bp fragment (ORF62 insert plus M13 primer and flanking vector regions 132 bp from reverse primer site to ORF62 insert and 118 bp from forward priming site to ORF62 insert) to confirm the presence of the ORF62 insert. Positions 105705, 106262, 106710, 107599, and 108111 were characterized on each clone by sequencing as described above. Position 107252 was characterized by restriction fragment length polymorphism (RFLP) analysis using NaeI restriction endonuclease as described below. ORF62 from pOka was also cloned as described above except that only 2 colonies were screened by PCR.
The NEBcutter tool (New England BioLabs) was used to verify previously described restriction endonuclease cleavage sites that could be used to determine the presence of the wt allele at positions 106262 and 107252 in the ORF62 clones. NEBcutter also generates a virtual gel image to illustrate how the restriction fragment length polymorphisms (RFLPs) should look depending on which allele is present.
The vOka alleles at position 106262 and 107252 have been reported previously to create SmaI and NaeI restriction endonuclease sites, respectively (15, 29), creating unique vOka and wt RFLPs. In our ORF62 clone, the presence of an additional 8 SmaI sites produced RFLPs that could not be interpreted by gel electrophoresis. Three additional NaeI sites were identified in ORF62 at nucleotides (nt) 105511 to 105516, nt 105768 to 105773, and nt 108780 to 108785. We replaced the vOka and wt alleles at all 16 vOka loci in ORF 62 and found that these additional sites were not affected, nor were any new sites created, by any of the vOka mutations. Using NaeI, RFLPs that clearly differentiated between the presence of the vOka allele and the wt allele were produced (Fig. 1). NaeI digestion in the presence of the wt allele resulted in a total of 4 fragments due to the absence of the restriction site at position 107252. The sizes of the minor fragments varied slightly depending on the orientation of the ORF62 fragment inserted into the vector (3012 bp, 535 to 552 bp or 514 to 531 bp, and 257 bp). The presence of the vOka allele at position 107252 caused cleavage of the NaeI site, resulting in 5 fragments (1530 bp, 1482 bp, 535 to 552 bp or 531 to 514 bp, and 257 bp). Restriction digestion of the PCR product was performed in a 10-μl reaction volume consisting of 2 μl of PCR product, 5 U of NaeI restriction endonuclease (New England BioLabs), 1 μl of 10× NEBuffer 4, and 6.5 μl molecular-grade water. Reaction mixtures were incubated at 37°C for 3 h, and fragments were separated by gel electrophoresis on 1% agarose at 80 V for 3 h. The 3 additional NaeI sites served as an internal control for the enzymatic reaction. Position 107252 was also sequenced in all clones to confirm the RFLP results.
We estimated the probability of occurrence of the wt allele as the number of variant (i.e., wt) alleles that we found divided by the total number of sequences that we analyzed. We assumed a binomial distribution to obtain a 95% confidence interval for the probability, using the calculator at the website http://statpages.org/confint.html#Binomial.
All nucleotide positions referred to are those in strain Dumas (GenBank accession no. X04370; RefSeq NC_001348).
The SNP profile for the VZV isolate obtained from the HZ case is shown in Fig. 2. DNA was analyzed at positions 105705, 106262, 107252, and 108111 in ORF 62 in order to determine if the virus was vOka derived or wt. At positions 105705, 106262, and 108111 the virus was found to have the vOka allele, confirming it as vOka derived. However, at position 107252 the virus carried the wt allele. Analysis of SNPs in ORF1, -21, -22, -38, -50, and -54 collectively identified the virus as a clade 2 strain with the ORF38/ORF54 SNP profile of the vOka.
The frequencies of the wt allele at four loci in ORF62 (105705, 106262, 107252, and 108111) in 304 ORF62 clones derived from a batch of Varivax were determined (Fig. 3). At positions 106262, and 108111, no clones with the wt allele were detected. At positions 105705 and 107252, the wt allele was found on 3% (9/304) and 2% (6/304) of clones, respectively. Among the 6 clones with the wt allele at position 107252, 3 also carried the wt allele at position 105705. The clones were also analyzed at positions 106710 and 107599, both of which are reported to lack the novel vOka allele present in Varilrix and/or vOka/Biken. At position 107599, the vOka allele was found on 18.4% (56/304) of clones, but only the wt allele was detected at position 106710.
Effective monitoring for vOka adverse events requires genomic markers that reliably distinguish vOka from wt viruses. The VZV genome is highly conserved, however, and even the most distantly related wt clades have only about 0.1% sequence variation at the DNA level (34, 36). Complete sequencing of vOka and pOka revealed only 42 SNP that differ between the two viruses (16). vOka has been shown to contain a mixture of strains, and only 4 of 42 vOka-associated SNPs were reported as fixed for the vOka allele in all vOka strains (16, 22, 31, 48). The remaining SNPs occur as mixtures of vOka and wt alleles in the vaccine preparations, and the SNP profiles vary both between manufacturers and between batches from the same source (22, 42, 47, 48).
The vaccine is known to be attenuated for replication in skin (35), but the genetic basis for attenuation remains elusive. It may involve multiple mutations within both ORF62 and several other genes (reviewed in reference 39). The presence of the wt alleles is also likely to contribute to the immunogenicity and pathogenicity of the vaccine.
The approach to discriminating vOka from wt strains has evolved with improved understanding of genomic variation in vOka and with apparent shifts in the biogeography of wt VZV clades (16, 17, 22, 34, 40, 47, 48). At first, vOka-derived viruses were distinguished using two SNPs located in ORF38 and ORF54 that distinguished the wt clade 2 variant of vOka (clade 2 PstI−) from all other wt strains (24). This was an unacceptable approach in Asia, where strains of this subclade were in common circulation, but it was initially useful in the United States since none of these clades appeared to be present. After pOka and vOka had been completely sequenced, SNPs that could distinguish vOka from all wt strains were identified (16). Based on conventional DNA sequencing methods, four SNPs appeared to be fixed in vOka, all of which were located in ORF62 (positions 105705, 106262, 107252, and 108111) (16, 31). This finding was supported by the observed SNP profiles of viruses isolated from laboratory-confirmed cases of vOka varicella and HZ, although four isolates with the wt SNP at position 105705 were detected (41). Assays targeting any of the SNPs at position 106262, 107252, or 108111were therefore regarded as the most reliable approach for identifying vOka, particularly if a combination of methods targeting more than one SNP was used. Complicating matters, however, was the observation of a wt clade 5 strain containing the vOka allele at position 107252 isolated from varicella cases that had been transmitted from an 86-year-old HZ patient in a long-term care facility (32). Since the index case patient would likely have experienced her primary infection in childhood, this finding suggested that wt strains bearing one of the vOka markers from a clade known to be circulating in North America were present in the early 20th century and are likely to still be circulating today. Thus, it has now become necessary to characterize both wt genotyping markers (1, 30) and vOka-specific markers to maximize the reliability of vOka-wt discrimination.
Recognizing the inherent limit in the sensitivity of DNA sequencing methods for the detection of mixed alleles, we sequenced a large number of TA clones derived from one lot of vOka to determine whether a small fraction of strains in vOka carried the wt allele at any of the fixed vOka loci. In our study, 2% of vOka clones carried the wt allele at position 107252 in ORF62, and 3% of clones carried the wt allele at position 105705. None of the 304 ORF62 clones evaluated in this study carried the wt alleles at either position 106262 or 108111. A recent study of shorter clones (47), each containing one of the fixed vOka SNPs, found precisely the opposite: 6% of clones had the wt marker at positions 106262 and 108111, and none of the clones had the wt allele at either position 105705 or 107252. While it is unclear whether the discrepant findings reflect lot-to-lot variation, differences between the experimental approaches or the numbers and types of clones evaluated, or other factors, it is now apparent that vOka contains small numbers of strains that carry at least one wt allele at any of the four vaccine markers previously regarded as fixed. Moreover, we found that 3 of the 6 clones that carried the wt allele at position 107252 also carried the wt marker at position 105705. On the basis of the findings from our study and that by Thiele and colleagues (47), reliance on a single “fixed” vOka marker would result in a 1-in-17 to 1-in-50 chance of mischaracterizing a vOka adverse event as a wt infection. A testing algorithm that evaluates all 4 of these SNPs would reduce the risk of mistaken identification by multiple orders of magnitude. In addition, since one of the vOka markers (position 107252) has been observed in a North American clade 5 virus, additional testing to evaluate the PstI marker at position 69349 in ORF38 will also be needed to determine whether an isolate has the SNP profile of vOka-like clade 2 strains.
We initially screened position 107252 using RFLPs generated by the NaeI restriction endonuclease. We demonstrated that this approach is a robust, rapid, and cost-effective screening tool for characterization of this locus through RFLP patterns generated from the entire OFR62 fragment. The consistency of the additional NaeI sites in all vOka genomes makes this method applicable to the analysis of all batches of vOka.
Our observation of a case of vOka HZ illustrates the need for a more complex approach to discriminating vOka from wt virus. If a method targeting only the 107252 locus had been employed, this virus would have been identified as wt. Evaluation of SNPs at multiple loci in ORF1, -21, -22, -38, -50, -54, and -62 unambiguously identified the strain as a clade 2 virus of the vOka/pOka subtype, (clade 2 PstI−), with 3 of the 4 vOka markers that were previously considered fixed.
The Varivax vaccine preparation has also been reported to contain two fixed wt alleles in ORF62 (positions 106710 and 107599) that are present as a mixture of wt and vOka alleles in vOka/Biken and/or Varilrix (16, 40, 48). Since the clones used in our study included the entire ORF62 reading frame, we were also able to evaluate the frequency of these markers in Varivax and to detect any linkage between one or more vOka markers. We detected the vOka allele at position 107599 on 18% of clones, which was previously reported to be fixed for the wt in all batches (31, 41, 47, 48). This percentage is well within the detection limits of all currently employed methodologies and may therefore be indicative of interbatch variation. None of the 304 clones carried the vOka marker at position 106710, consistent with previous observations using DNA prepared directly from the vOka (31, 48). The apparent loss of this vaccine marker may reflect differences in the manufacturing processes used by the three companies that produce this vaccine. The Varivax and Varilrix vaccines underwent additional passages in human cells, which may have introduced selective pressure against the vaccine marker at position 106710.
Another concern for diagnostic approaches to discriminating between vOka and wt strains is the possibility of recombination between vOka and wt virus in coinfected persons. Recombination between wt VZV strains has been reported in vitro and in vivo (1, 2, 11), and phylogenetic analyses of the published complete genomic sequences for VZV have led to the conclusion that recombination has played an important role in the establishment of currently circulating VZV clades (34, 36, 38). Our characterization of the HZ case virus was limited to the analysis of widely scattered SNPs across a large portion of the genome. While the data are not sufficient to make definitive conclusions about recombination, the SNP profile of this virus provided no suggestion that recombination between vOka and a wt virus had occurred.
It is now clear that all 42 of the vOka-associated SNPs are present as a mixture of vaccine and wt alleles, albeit in widely variable proportions. The basis for vOka attenuation is therefore likely to be more difficult to ascertain than might have been hoped. A number of the genetic differences between vOka and pOka therefore seem likely to contribute collectively to the reduced pathogenicity of the vaccine. We are currently working to isolate a series of viable clones from the vOka preparation, each with different vOka-associated SNP profiles, for characterization using in vitro and in vivo pathogenicity assays.
In view of our observations, it now seems necessary to evaluate all four of the high-frequency vOka markers in ORF62 (positions 105705, 106262, 107252, and 108111) in addition to the clade 2 subtype marker in ORF38 in order to confidently verify vOka adverse events. Additional genotypic analysis may prove necessary in the future, particularly if the occurrence of vOka-wt recombination is documented.
We thank Michael J. Cannon for technical assistance and intellectual input during the preparation of the manuscript.
The views included in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
The authors report no conflicts of interest.
Published ahead of print 29 February 2012