Massively parallel sequencing for monitoring genetic consistency and quality control of live viral vaccines Alexander Neverov and Konstantin Chumakov1 Center for Biologics Evaluation and Research, Food and Drug Administration, Rockville, MD 20852 Edited* by Robert H. Purcell, National Institutes of Health, Bethesda, MD, and approved October 6, 2010 (received for review August 24, 2010) Intrinsic genetic instability of RNA viruses may lead to the accumu- uation. MAPREC is currently recommended by the World lation of revertants during manufacture of live viral vaccines, Health Organization (WHO) for screening of batches of OPV requiring rigorous quality control to ensure vaccine safety. Each before they can be released for use in humans (11, 18). lot of oral poliovirus vaccine (OPV) is tested for neurovirulence in Despite high sensitivity and accuracy of MAPREC, it can be animals and also for the presence of neurovirulent revertants. used to monitor only a limited number of genomic loci known to Mutant analysis by PCR and restriction enzyme cleavage (MAPREC) contain the markers of attenuation and misses mutations at other is used to measure the frequency of neurovirulent mutations at the sites, which can also contribute to loss of the attenuated phe- 5′ untranslated region (UTR) of the viral genome that correlate with notype. For most viruses, determinants of attenuation are un- the level of neurovirulence determined by the monkey neuroviru- known, which limits the utility of this approach for other live viral lence test. However, MAPREC can only monitor mutations at a few vaccines. Nevertheless, ensuring molecular consistency of vac- genomic loci and miss mutations at other sites that could adversely cine batches by requiring that no mutation accumulates beyond affect vaccine quality. Here we propose to use massively parallel the level present in past batches with good clinical record could sequencing (MPS) for sensitive detection and quantification of all help maintain vaccine safety (19). Therefore methods enabling mutations in the entire genome of attenuated viruses. Analysis of detection of unstable genomic loci in attenuated strains could be vaccine samples and reference preparations demonstrated a perfect used for the monitoring of genetic stability. Such studies are agreement with MAPREC results. Quantitative MPS analysis of usually conducted by serial passaging of attenuated strains in validated reference preparations tested by MAPREC produced iden- vivo and in vitro, followed by screening for the presence of small tical results, suggesting that the method could take advantage of numbers of mutants that may have accumulated in viral pop- MICROBIOLOGY the existing reference materials and be used as a replacement for ulations (16, 17, 20, 21). Traditional sequencing approaches can the MAPREC procedure in lot release of OPV. Patterns of mutations detect mutants only if their content exceeds 20–30% and are present at a low level in vaccine preparations were characteristic of therefore of limited utility. Other approaches based on analysis seed viruses used for their manufacture and could be used for iden- of electrophoretic mobility in gels (22, 23) do not always allow fi ti cation of individual batches. This approach may represent the ulti- precise location of mutations. Matrix-assisted laser desorption/ mate tool for monitoring genetic consistency of live viral vaccines. ioniation-time of flight (MALDI-TOF) mass spectrometry (24) and hybridization with microarrays of short oligonucleotides (25, fi mutant quanti cation | quasispecies | sequence heterogeneity | vaccine 26) are more sensitive, but are relatively laborious and may re- safety | neurovirulence quire follow-up by direct sequencing. In the past few years, a set of new technologies for massively igh mutation rates inherent to replication of RNA viruses parallel sequencing (MPS), which are also referred to as high- Hcreate a wide variety of mutants that are present in virus throughput, next generation, or ultradeep sequencing, have populations, which are often referred to as quasispecies (1). The emerged. They allow rapid generation of extremely large amounts diffuse, “cloud-like” nature of viral populations allows them to of sequence information (27) and are primarily used to sequence rapidly adapt to changing replicative environments by selecting genomes of higher organisms and for metagenomic studies of preexisting variants with better fitness (2–4). Many important virus populations of various bacteria and viruses (28, 29). These new properties cannot be explained by a mere consensus sequence, technologies are also very promising for analysis of viral quasis- but require knowledge about the microvariants present in viral pecies and have been used to study HIV (30, 31), severe acute stocks. One such property is virulence, which is critically relevant to respiratory syndrome (SARS) (32), hepatitis B (33), and other vaccine development and manufacture. RNA viruses (29). MPS was also used for detection of adventitious Attenuated Sabin strains used in manufacture of oral poliovirus agents in vaccines (34). Here we demonstrate that MPS can be vaccine (OPV) (5–7) can regain neurovirulence during growth used to identify very small amounts of mutant viruses in prepara- in vaccine recipients and propagation in cell cultures (8). It tions of live viral vaccines as well as for their accurate quantifica- occurs because of the accumulation of mutants with higher neu- tion. The ability to quantify potentially harmful mutations in rovirulence (9, 10), which include changes in the 5′ untranslated vaccine batches makes this method suitable for quality control region (UTR) (10). Because of this genetic instability, every batch and ensuring manufacturing consistency of live viral vaccines. of OPV must be tested for neurovirulence in monkeys (11) or transgenic mice susceptible to poliomyelitis (12, 13). Previously, we Results developed a highly sensitive molecular method, mutant analysis In the first experiment, pyrosequencing (technology developed by PCR and restriction enzyme cleavage (MAPREC), which en- by 454 Life Sciences, a Roche company, further referred to as abled us to quantify the 5′ UTR revertants in monovalent batches of OPV and demonstrate that their content in vaccine lots directly correlates with the results of the monkey neurovirulence test Author contributions: A.N. and K.C. designed research; A.N. and K.C. performed research; (MNVT) (14). The content of 472-C revertants in type 3 OPV is A.N. and K.C. analyzed data; and A.N. and K.C. wrote the paper. usually below 0.8% in batches of vaccine that passed the MNVT, The authors declare no conflict of interest. whereas batches that failed the test usually contain more than *This Direct Submission article had a prearranged editor. 1% of 472-C revertants (15). Similar methods were developed to Freely available online through the PNAS open access option. quantify mutants with 480-A in Sabin 1 and 481-G in Sabin 2 1To whom correspondence should be addressed. E-mail: [email protected]. strains (16, 17) that were also shown to be determinants of atten- gov. www.pnas.org/cgi/doi/10.1073/pnas.1012537107 PNAS Early Edition | 1of6 Downloaded by guest on September 24, 2021 454 Life Sciences) was used to analyze full-length PCR ampli- Table 1. Frequencies of mutations identified by pyrosequencing cons prepared from two samples of type 3 OPV: one that passed in plasmid DNA, PCR product, and viruses the monkey neurovirulence test and another that failed. Both Mutation frequencies were made from passage level 3 of the Sabin original seed stock (SO+3). The lot that passed the MNVT was manufactured in Total Insertions primary monkey kidney cells, whereas the failed lot was grown in nucleotides Substitutions, and WI-38 human diploid cells. Fig. 1A shows that the number of Sample read, 106 % deletions, % times each nucleotide was determined in positive and cDNA strands was about the same, but varied between 2,000 and 10,000 Plasmid 28.1 0.050 0.424 depending on the genomic location with the total for both PCR product 40.1 0.104 0.860 strands being between 6,000 and 15,000. Both samples contained Plaque 18.2 0.121 0.719 a number of mutants, the most prominent of which was the one HeLa p1 20.7 0.143 0.694 HeLa p7 36.6 0.155 0.858 with C2493U mutation, which changes threonine to isoleucine at amino acid 6 of the capsid protein VP1 (35, 36). The mutation Vero p5 19.0 0.197 1.232 does not affect neurovirulence, but the virus with 2493-U rapidly accumulates upon growth in cultured cells (15). Importantly, in agreement with our previous MAPREC results, there was a dif- 0.05%, whereas there were 0.104% mutations in PCR product ference in the content of 472-C mutants, which correlated with and 0.121% in the rederived virus. Although there was no ap- the MNVT results. The vaccine lot that passed the test contained parent bias for the type of mutations detected in plasmid 0.35% 472-C, whereas the lot that failed it contained 2.4%. DNA (Fig. 2), in all other samples prepared by PCR amplifica- There were a number of other mutations reaching several tion there were significantly more transitions (purine–purine and percent of the population in addition to a low level of less than pyrimidine–pyrimidine changes) than transversions (purine– 0.5% present at all nucleotide positions. This background level pyrimidine and pyrimidine–purine changes). After passaging the may represent true heterogeneity due to the quasispecies nature virus in HeLa cells