Regulatory Approaches for Control of Viral Contamination of Vaccines

Vaccines are among the most effective ways to control infectious diseases. Through vaccines, previously fatal or disabling diseases such as poliomyelitis and smallpox have become a distant memory, and morbidity associated with other vaccine-preventable illnesses is also significantly reduced. Vaccines work both by preventing disease in vaccinated individuals and interrupting transmission of the disease agent. For infections transmitted from person to person, high vaccination rates promote the herd immunity required for the additional benefit of substantially reducing disease transmission, even to the non-immunized. In turn, high vaccination rates are dependent upon high levels of public confidence in vaccines, especially in their safety. Thus, measures to increase confidence that vaccines are safe will help to maintain the amply demonstrated public health benefit of vaccines.

Because vaccines are produced in biological systems from biological materials, there is the theoretical potential for adventitious agents to be present. Adventitious agents in a vaccine, if they lead to adverse events or to a perception of reduced safety, could undermine public confidence in vaccines and reduce the expected benefit. Thus, testing strategies and vaccine production methods should assure that vaccines are free of adventitious agents, and if adventitious agents are nonetheless detected, further analysis to assess the potential impact on vaccine benefit/risk ratios is needed.

Because it is impossible to “prove a negative,” a test sample is considered to be “free” of an adventitious agent if assays demonstrate that the agent is undetectable in a defined quantity of vaccine to a defined level of sensitivity. The level of assay sensitivity can be determined experimentally using reagents and controls that have been standardized or qualified using methods appropriate for validation of these assays. In addition to testing for adventitious agents, adherence to current good manufacturing processes (cGMPs) ensures high-quality vaccines that have reduced probability of containing adventitious agents. Also, in product manufacture, clearance steps in a manufacturing process that are validated to remove a potential adventitious agent, should it be present, to a defined level may support a conclusion that the product is free of that agent.

Specific FDA recommendations for testing vaccines for adventitious agents have been promulgated (1). These methods include broad, overlapping schemes to detect as wide an array of viruses as possible. The evolution of this approach over time recapitulates the history of virus discovery—adapting assays that were used to discover and identify an increasing number of new viruses.

Infectivity assays in animals and cell culture permit determination of the presence of an infectious virus. If there is a method for detecting an infection (for example, by cytopathic effect or hemagglutination in cell culture, or animal illness in in vivo assays), a virus can be detected even without specific knowledge of the actual virus or its sequence. Other assays also can be non-specific—including assays for reverse transcriptase (RT), an essential component of retroviruses, and electron microscopy, which detects viral particles (albeit relatively insensitively). Other adventitious agent detection methods focus on specific agents, for example, those that use antibody tests to identify viruses in cell culture or the polymerase chain reaction (PCR) to identify potential adventitious agent nucleic acid sequences.

These strategies to detect potential adventitious agents in vaccines, while powerful, have not completely prevented their presence in distributed vaccines. Simian Virus 40 (SV40) did not show cytopathic growth in the rhesus monkey kidney cell cultures initially used to produce and test poliovirus vaccines. The virus was discovered as an adventitious agent in these poliovirus vaccines in 1960 when cytopathic effect was observed in African green monkey kidney cell cultures (2). Although the poliovirus in these vaccines was formalin-inactivated, the non-enveloped SV40 was not fully inactivated by this manufacturing step, and thus many individuals who received poliovirus vaccine in the 1950s were likely exposed to infectious SV40. An SV40-free inactivated poliovirus vaccine (IPV) was produced as quickly as possible, and licensure of the live-attenuated oral poliovirus vaccine (OPV) was delayed until an SV40-free vaccine could be made. Although there were concerns that marketed OPV might contain SV40, testing performed at that time by the manufacturer on every lot of oral polio vaccine would have detected SV40 if it were present, and subsequent studies using PCR also found no evidence of SV40 in U.S. licensed oral polio vaccines (3). Although some early recipients of IPV in the 1950s sero-converted to SV40, long-term follow-up studies did not reveal adverse consequences (4). While several investigators reported the presence of SV40 DNA sequences using highly sensitive PCR assays in samples extracted from human cancer tissues, a multi-center investigation did not confirm the presence of SV40 in human cancers (5).

In 1996, an improved PCR-based RT assay (product-enhanced RT, or PERT, assay) was used to show that previously undetectable quantities of RT were present in some avian cell–produced vaccines (6). Additional studies showed this RT activity to be due to a defective, non-infectious endogenous avian retrovirus (EAV) in measles and mumps vaccines (7, 8) and a defective version of avian leukosis virus (ALV-E) in some yellow fever vaccines (9). Both viruses have always been present in eggs consumed by humans, and neither virus is infectious in humans. The long safety record of hens' egg-produced vaccines and the absence of evidence of potential harm to humans (10) were also important considerations in determining that these egg-produced vaccines could safely contain defective endogenous avian retroviruses (11). This experience led to the inclusion of improved RT assays in product testing (1) to further assure that other retroviruses are not present in vaccines.

In 2010, investigators used a combination of nuclease digestion, non-specific PCR, and high-throughput sequencing to further characterize live virus vaccines. They found sequences from porcine circovirus (PCV) 1 in Rotarix® live-attenuated rotavirus vaccine (12). Porcine circoviruses are small, single-stranded, circular DNA-containing viruses and are ubiquitous in pigs, but are not known to cause disease in humans.

To further characterize the PCV sequences identified in rotavirus vaccine, the FDA undertook its own investigation. The primary goal of the investigation was to determine whether vaccines contained infectious PCV. This investigation focused on Rotarix® and IPV (produced in a cell bank related to that used to produce Rotarix®) (GSK Biologicals, Rixenxart, Belgium), and Rotateq® (Merck and Co., West Point, PA), for which studies by the manufacturer suggested the presence of PCV-2 DNA fragments.

The FDA's studies first focused on characterization of the nucleic acids in the vaccines, identifying particle-associated full-length PCV-1 genomes in Rotarix® but not in the other vaccines. Infectivity assays in cell culture revealed that Rotarix® but not the other vaccines contained infectious PCV-1. None of the vaccines contained infectious PCV-2 (13).

After discussion of these issues at the May 7, 2010 Vaccines and Related Biological Products Advisory Committee meeting, committee members concluded that even with the presence of infectious, non-pathogenic PCV-1 in Rotarix® and reports of PCV-1 and PCV-2 DNA in RotaTeq®, the benefits of these vaccines so clearly outweighed the theoretical risks associated with PCV that these vaccines should continue to be used (14). Committee members also recommended that the manufacturers develop PCV-free preparations of vaccines (15).

The success of these new molecular techniques in identifying an adventitious agent in a vaccine demonstrates the potential utility in using such techniques to provide greater assurance of vaccine safety. Routine use of such techniques will likely be possible once several scientific issues are addressed.

Most of these new molecular virus detection techniques follow a similar general framework (Table I). There is selection of a sample, a purification method that extracts nucleic acids from that sample, an amplification scheme that increases the quantity of sequence from the sample, a detection scheme used to identify the detected sequence, and an analysis method that is used to put the results into context. For many of these methods, a confirmatory study may also be needed.

In implementing these new assays, consideration must be given to the sample sources that are most appropriate for testing. One could consider testing raw materials and reagents, the cell banks used to produce a vaccine, vaccine intermediates, or the final product. In some cases, it may be reasonable to consider employing these new techniques in conjunction with previously used methods, such as cell culture. Implementation of the new molecular approaches in conjunction with cell culture assays performed in the production cell substrate may be particularly useful for some samples because uninfected cells and material from early time-points during the culture can potentially be used as controls to confirm (or rule out) detectable replication of the agent (or its nucleic acid) in the production cell substrate.

Sample preparation is also an important consideration. When they detected PCV-1 in Rotarix, Victoria et al. enriched their vaccine samples for virus capsids (12), which increased the quantity of adventitious agent relative to cellular nucleic acids, potentially increasing assay sensitivity and also increasing the likelihood that any identified viral sequence would represent infectious virus. However, sensitivity could be reduced if the purification conditions destroyed particles of labile viruses.

Other investigators have analyzed RNA transcriptomes from cells using massively parallel sequencing techniques (16). Such an approach can be used to characterize transcription from potential adventitious agents (and may also directly identify RNA viruses) in a cell bank, but may be less useful when examining products or reagents directly because extracellular RNA is usually rapidly degraded when it is not part of a virus. Transcriptomes are also cell cycle–dependent, so these results may be more difficult to standardize, and analysis can yield endogenous retroviral sequences that require careful analysis for interpretation of the results. Cellular genomes (DNA) can also be analyzed by massively parallel sequencing or other techniques, with the advantage of representing all DNA sequences in a sample, but this strategy cannot detect non-integrating RNA viruses and the dilution of viral sequences in the background of the entire cellular genome may lead to reduced relative sensitivity. Thus, sample preparation methods should be matched to the type of viral signature that is being sought and to the amplification and detection methods that are being used.

Different methods can also be used to amplify purified nucleic acids. Typically, non-specific amplification is used to increase the likelihood of finding mutant or as-yet undiscovered viruses. However, there are several different non-specific PCR methods that have varying sensitivity to detect small quantities of nucleic acid. While for many studies the actual PCR sensitivity may not be critical because the massively parallel sequencing readout generates a sufficient number of sequences to compensate for low PCR sensitivity, for other methods, in which viral particles are first purified, improved PCR sensitivity may be more important. In our laboratory's experience, inclusion of a 3′ anchor sequence in a non-specific degenerate oligonucleotide primer (17, 18) can improve assay sensitivity for small quantities of viral nucleic acid.

Because many of these techniques rely on non-specific PCR that is capable of amplifying all nucleic acids, such tests are unusually susceptible to contamination. In our experiments, we observed that reagents used for the tests, including the polymerases and even the water, may contain small DNA fragments from bacteria or algae that could yield false signals depending on how they are interpreted and what controls are run.

These next-generation methods can be paired with various detection methods, including mass spectrometry, as is used in the PLEX-ID system (19); this system relies on virus family–specific primers and would not work with completely general primers, microarrays that can yield very rapid results and are suitable for automation (20) but may have reduced ability to detect mutated or completely unknown viruses), and massively parallel sequencing (also known as deep or high-throughput sequencing). Different massively parallel sequencing platforms may have varying utility because of differences in number, accuracy and length of reads and in sources of nucleic acids that each platform sequences.

Use of massively parallel sequencing-based platforms for adventitious agent detection will also lead to the need for more standardized and validated bioinformatic tools. Sequence databases will need to be curated in order to assure that potential adventitious agent sequences are correct and that no important sequences are excluded. While the algorithms for searching these databases have been utilized in research settings, they have not all been validated for use in a regulatory setting. For example, each new massively parallel sequencing run may yield sequences that do not currently match anything in the existing database; a plan would also need to be devised to determine how to address those sequences, potentially with follow-up analyses as the sequence databases become more complete.

Introduction of massively parallel sequencing-based methods for adventitious agent detection could be facilitated by including appropriate controls directly within the experiments. The use of barcoded primers in massively parallel sequencing reactions will permit the inclusion of internal negative and positive controls to facilitate the interpretation of confusing results in both the sample and the controls.

It will be important to select appropriate standards or positive controls for these new molecular virus detection assays. Ideally, such controls would mimic the type of sample being tested, and would contain a variety of viruses chosen to test the expected weaknesses of each specific purification, amplification, and detection scheme. For some samples, specific nucleic acids spiked into samples or non-adventitious-agent sequences already known to be present in the sample may suffice as controls, while other samples may require more complex controls.

If these novel methods were to detect viral sequences in a product, reagent, or a cell bank, follow-up studies will be necessary. Regulators would have to consider each situation on a case-by-case basis, but at a minimum, confirmation of results, and perhaps performance of studies similar to those used to evaluate the PCV sequences identified in rotavirus vaccines, could be considered. However, evaluation of these situations may be complex in cases where infectivity assays do not exist and human experience to evaluate the potential pathogenicity of a potential adventitious agent is lacking.

The new generation of molecular assays has significant promise for improving vaccine safety, and thus confidence in vaccines. A concerted effort to identify the best uses for these assays and the best approaches for implementing them will have a positive public health impact.

• © PDA, Inc. 2011