Immunogenicity Testing in Human Papillomavirus Virus-Like-Particle Vaccine Trials

The recently introduced human papillomavirus (HPV) prophylactic vaccines are the first widely approved vaccines specifically designed to prevent a sexually transmitted infection and are only the second (after the hepatitis B virus vaccine) designed to prevent human cancer, in this case cervical cancer and several other anogenital and head and neck cancers. The antigens in the HPV vaccines are naked icosohedral virus-like particles (VLPs) composed of the major HPV virion protein L1. These subunit vaccines were independently developed and tested by 2 companies, GlaxoSmithKline (GSK) and Merck [1]. The GSK vaccine, Cervarix, is bivalent—it contains the VLPs of HPV-16 and -18, the types that cause ~70% of cervical cancers worldwide—and is produced in insect cells. Cervarix also contains the proprietary adjuvant AS04, which is composed of monophyosphoryl lipid A and an aluminum salt. The Merck vaccine, Gardasil, is tetravalent—it contains the HPV-16 and -18 VLPs and the VLPs of HPV-6 and -11, which cause 80%–90% of genital warts—and is produced in yeast. Gardasil is adjuvanted with a simple aluminum salt. The vaccines are delivered by intramuscular injection in 3 doses over 6 months. Both have been remarkably effective in phase 3 trials conducted in young women, providing nearly complete protection against persistent genital tract infection and premalignant neoplastic disease end points caused by the HPV types targeted by the respective vaccines [2]. Both vaccines have been licensed in >50 countries, starting with Merck’s in 2006 and GSK’s in 2007, and millions of doses have been sold. However, despite the rapid and successful introduction of these 2 vaccines, they can reasonably be viewed as introductory products that will likely be followed by second-generation vaccines that will target more types and/or be less expensive to produce and deliver. This article will focus on one aspect of HPV VLP vaccines, the assessment of immunogenicity.

WHY IMMUNOGENICITY TESTING IS IMPORTANT

Immunogenicity testing contributes substantially to 5 aspects of HPV vaccine development and deployment. The first area is quality control of the vaccine-manufacturing process and the stability of the vaccine during storage and distribution. Vaccine manufacturers tend to prefer physical characterization for routine quality-control purposes, because reproducible quantitative results are more easily achieved than with biological assays. However, direct evaluation of the immune response to a vaccine in an animal model or human subjects remains the most relevant test of vaccine quality. Second, immunogenicity bridging studies are being used to extend vaccine approval for populations that were not evaluated in pivotal phase 3 studies, which were limited to females aged 15–26 years [3, 4]. The widespread regulatory approval of the vaccines for younger adolescent girls is based on the noninferior immunogencity observed in this age group. Regulatory bodies in some countries have also approved the vaccines for older women and young men, on the basis of the noninferiority of the immune responses to the vaccines. Third, immunogenicity studies may help to predict the duration of protection. Typical for a newly introduced vaccine, the duration of protection of the HPV vaccines is unknown at present. However, the durability of the presumably relevant immune response (discussed below) supports an optimistic projection. Fourth, immunogenicity studies can lead to the establishment of an immune correlate of protection. Such a correlation can be of immense help in the development of second-generation vaccines by providing an early indication of potential efficacy and, in some cases of closely related vaccines, might even permit regulatory approval without costly and time-consuming efficacy trials. Fifth, immunogenicity testing can be used to assess the relative merits of competing vaccines in the absence of a head-to-head efficacy trial. As discussed below, caution must be taken in comparing vaccines that have been evaluated by different manufacturers using different assays. However, in an interesting development GSK has announced that it is undertaking a randomized clinical trial in young women that will directly compare the immunogenicity of Cervarix and Gardasil [5].

WHY THE FOCUS IS ON ANTIBODY RESPONSES

HPV VLPs induce potent T and B cell responses in animal models and human subjects [6]. However, immunogenicity testing in clinical trials has focused primarily on antibody responses for several reasons. One is that the protection induced by established prophylactic viral vaccines is largely, if not entirely, mediated by antibodies that prevent virus infection [7]. Another is that the protection from experimental challenge in animal papillomavirus models can be passively transferred in serum or purified immunoglobulin (Ig) G from VLP-vaccinated animals, which indicates that VLP-induced antibodies alone are sufficient to confer protection [6]. In addition, the characteristics of the antibody response induced by VLP vaccination are consistent with the protection in humans being antibody mediated in that there is a correlation between the type-restricted protection that has been seen in clinical trials and a similar spectrum of type restriction in in vitro antibody- neutralization assays [6]. The high rate of protection seen in clinical trials also correlates with the >99.5% rate of sero-conversion detected in VLP vaccinees.

Antibodies induced after intramuscular VLP injection may protect from cervi-covaginal HPV infection by a combination of 2 mechanisms [6]. The first is transudation of serum antibodies into genital tract mucus, which is quite pronounced at the cervix [8]. The second is that infection is thought to require trauma that exposes the epithelial basement membrane to the virus, and this wounding is expected to result in direct exudation of systemic antibodies at the site of virus infection [9].

The expression pattern of L1, the viral protein that comprises the VLPs, makes it unlikely that T cell responses to the VLPs contribute to protection. Rather than being expressed in the proliferating basal cell keratinocytes, where infection is maintained, L1 is detectably expressed only in terminally differentiating keratinocytes of an infected stratified squamous epithelium [10]. It therefore seems unlikely that T cell effector responses would substantially influence infection or lesion development. Consistent with this conjecture, VLP vaccination was not effective against prevalent infection [11]. It is likely the case that T helper responses contribute to the uniformly robust B cell responses seen in VLP vaccinees. However, even in this regard there is evidence that VLPs, like other well ordered multivalent antigens, can induce T cell–independent antibody responses [12]. In summary, the focus on antibody responses in immunogenicity studies of VLP vaccines seems well justified.

ASSAYS TO MEASURE VLP ANTIBODY RESPONSES

Three assays to measure serum antibodies predominate in the evaluation of VLP antibody responses in clinical trials. Each measures an overlapping but distinct subset of the antibody responses to the VLPs (figure 1), and each has a unique set of strengths and weaknesses.

Figure 1.

Relationships among virus-like-particle (VLP) antibodies detected by enzyme-linked immunosorbent assay (ELISA), competitive Luminex immunoassay (cLIA), and secreted alkaline phosphatase neutralization assay (SEAP-NA). The large background circle depicts the entire spectrum of antibodies induced by VLP vaccination, and the interior circles depict the subsets of antibodies detected by ELISA, SEAP-NA, and cLIA. Note that the size of the circles and the extent of their overlap will vary depending on the specific performance characteristics of the assays for a given human papillomavirus type. Abs, antibodies.

Enzyme-linked immunosorbent assay (ELISA)

GSK has used a VLP-based ELISA as the principal assay of immunogenicity in its trials [13]. This assay measures antibodies that bind to a VLP antigen fixed to a solid surface. The bound antibodies are detected by the addition of a secondary antibody that recognizes the constant region of a human antibody of a specific class (e.g., IgG) or subclass (e.g., IgG1). The secondary antibody has an enzyme (e.g., alkaline phosphatase) conjugated to it, and the enzyme’s activity is monitored by a change in the substrate (e.g., a color change that can be measured in a spectraphotometer). The major advantages of the ELISA are that it is sensitive, rapid, and reproducible and can be automated. Correlations across laboratories can be increased by reporting results relative to the responses to type-specific standard antisera [14].

Although the ELISA can measure neutralizing antibodies, which for HPV L1 exclusively recognize conformation-dependent virion surface epitopes, it has the disadvantage of also potentially measuring nonneutralizing antibodies that are elicited by VLPs and that may be recognized by VLP antigens in the ELISA. The fraction of reactivity that is attributable to nonneutralizing antibodies (which would not be protective) would depend on the degree to which the VLPs used in the vaccine and/or as ELISA antigen are properly folded and assembled. Some ELISAs can produce misleading results. For instance, VLP ELISAs can exhibit extensive cross-reactivity among genital HPV types, whereas protection in the trials is more type restricted [15]. Such cross-reactivity between types may be a significant issue in evaluating the immunogenicity of multivalent vaccines, because it precludes categorical assignment of reactivity to a particular VLP type in the vaccine.

A unique feature of the ELISA is that it detects reactivity to only 1 class or subclass of antibodies. Most studies focus on detecting only IgG, the predominant immunoglobulin class induced in the serum of most vaccinees. Contributions of other classes, such as IgA, to protection are not routinely evaluated, potentially diminishing the association between antibody response and protection.

Competitive Luminex immunoassay (cLIA)

Merck has mainly used a cLIA to evaluate VLP antibody responses in its clinical trials [16]. The assay involves fixing each of the 4 VLP types in the vaccine to Luminex microspheres (Invirtogen) with distinct fluorescent properties. Serum samples from the vaccinees are evaluated for their ability to prevent VLP binding by a type-specific neutralizing monoclonal antibody that has a phycoerythrin fluorescent tag. Thus, the strength of the antibody response is inversely proportional to the detection of the monoclonal antibody binding signal. This assay has 2 major advantages. First, it is highly type specific, because it is based on monoclonal antibodies that were specifically chosen to not cross-react with the VLPs of even very closely related HPV types. Second, individual reactivities to multiple VLPs can be simultaneously evaluated in a single reaction, and the assay is amenable to high-throughput processing. Unlike the VLP ELISA, the cLIA simultaneously measures VLP-binding antibodies of all immunoglobulin classes.

The major disadvantage of the assay is that it measures only the subset of neutralizing antibodies that compete with the specific monoclonal antibody for VLP surface binding (figure 1). The responses of some vaccinees could be dominated to varying degrees by VLP-binding (and potentially neutralizing) antibodies that do not compete with the monoclonal antibody. Thus, the assay can underrepresent the potentially protective antibody response induced by the vaccine. It is also unclear to what degree the epitopes recognized by the different type-specific monoclonal antibodies overlap the immunodominant epitopes in humans. Therefore, titers of antibody to the different VLP types in a given vaccine cannot be directly compared in a cLIA.

Neutralization assay

The third assay is an in vitro neutralization assay. The first reproducible quantitative neutralization assay that is not limited by availability of infectious capsids was recently developed and is becoming widely adopted by both companies and academic laboratories [15]. It involves the cell culture production of high-titer infectious L1/L2 pseudovirions that have encapsidated a gene whose activity can easily be measured as a marker of infection. In the most widely employed variation, the pseudovirus carries the gene for secreted alkaline phosphatase (SEAP), which enables infection to be quantitated by measuring the ability of culture supernatants to cleave a colorigenic substrate. The sensitivity of the SEAP neutralization assay (SEAP-NA), in terms of titers measured in response to natural infection or VLP vaccination, is similar to that of standard VLP-based IgG ELISAs [15, 17, 18]. The assay has been adapted to a high-throughput 96- well-plate format. Nevertheless, a major disadvantage of this assay is that it is considerably more laborious than either ELISA or cLIA. However, the results of this assay are the most likely to correlate with protection, because it presumably measures all neutralizing antibodies (regardless of immunoglobulin class) and only potentially protective antibodies.

The in vitro neutralization assay was found to be considerably more type specific than VLP ELISAs using research grade VLPs [15]. However, a strong correlation between individual ELISA and neutralizing-antibody titers was observed in trials of the GSK vaccine [17, 18]. In some situations, the neutralization assay may be less type specific than the cLIA. For instance, serum samples from HPV-18 VLP vaccinees had low but readily detected neutralizing-antibody titers in an HPV-45 pseudovirus neutralization assay, whereas there was no cross-reactivity in the cLIA using the type-specific HPV-18 monoclonal antibody [19]. However, HPV-18 and -45 are very closely related, and significant protection from HPV-45 infection was demonstrated in the GSK clinical trials [20]. Thus, cross-reactivity detected in the neutralization assay appears to correlate with biologically significant cross-protective responses. Titers measured in neutralization assays of different HPV types cannot always be directly compared. This is because the pseudovirus preparations of the different types can contain different percentages of noninfectious capsids, which can compete for neutralizing- antibody binding. An assay based on a pseudovirus preparation with a high particle-to-infectivity ratio will result in a lower apparent neutralizing-antibody titer than will an assay based on a preparation with a low particle-to-infectivity ratio [15].

QUESTIONS AND CONTROVERSIES

Although there are a number of interesting issues involving immunogenicity assessment of the HPV VLP vaccine clinical trials, 3 predominate at present. First, is there is one assay that should be considered the reference standard? National regulatory agencies have endorsed in a de facto fashion both the cLIA and the ELISA in that immunogenicity bridging data generated with them have served as the basis for extending authorization of the Merck and GSK vaccines to groups not evaluated in the efficacy trials [2]. Both assays have performance characteristics that make them appropriate for high-throughput analysis of the large numbers of samples that are typically assessed in immunogenicity bridging studies. However, it is important to note that neither assay has been validated as an immune correlate of protection. In large measure, this lack of validation may result from the exceptional efficacy of the vaccines. There have been very few cases of breakthrough infection or disease in the VLP vaccinees, perhaps too few to reasonably assess the assays as immune correlates. World Health Organization guidelines for HPV vaccines suggest that neutralizing assays should be considered the reference standard for assessing potentially protective antibodies induced by the vaccines [21]. Therefore, as breakthrough cases accumulate it may be preferable to consider using SEAP-NA as the primary assay for evaluating immune correlates of protection against infection and disease. Although the GSK ELISA results correlated very well with SEAP-NA results for both serum and cervical mucus specimens collected from a random sample of young women vaccinated with Cervarix [18, 22], the ELISA may somewhat overestimate the protective antibody response (i.e., generate false-positive results) if some individuals predominantly generate a response to nonneutralizing epitopes. Conversely, the cLIA may underestimate the protective antibody response in some individuals (i.e., generate false-negative results) if the immunodominant epitopes in certain individuals do not substantially overlap the epitopes recognized by the competing monoclonal antibody used for that HPV type in the assay. Determination of an immune correlate of protection might be accomplished in a nested case-control study evaluating the infrequent vaccinees with breakthrough infection or disease and a randomized subset of the vaccinees who remain free of infection and disease. Such a study would not overly tax the performance capacity of the neutralization assay. Ideally, it would be preferable to evaluate both serum and local cervical antibody levels, because it is possible that some of the rare breakthrough infections may result from unusually poor transudation of serum IgG into the cervical mucus.

Merck has initiated clinical trials of a nonavalent vaccine containing VLPs of an additional 5 high-risk types (31, 33, 45, 52, and 58). The cLIA will be useful in evaluating nonoverlapping type-specific antibody responses. This analysis will address the question of whether an increase in valency diminishes the antibody responses to the 4 original VLP types. However, it would also seem important to evaluate the potential added value of each additional VLP type (or the combination of VLP types) by assessing the ability of the various VLP types (or the combination of VLP types) to induce cross-neutralizing antibodies against related types by means of the SEAP-NA. For instance, would a vaccine that includes types 16, 31, and 52 induce a sufficiently high titer of HPV-58 cross-neutralizing antibodies that this type could be omitted from the second-generation vaccine? Similarly, could a second-generation GSK vaccine including HPV-6 VLPs potentially protect against both HPV-6–induced and HPV-11–induced genital warts, analogous to the protection against HPV-45 observed with the GSK HPV-16/18 vaccine? Insight into these possibilities cannot be obtained from a cLIA or ELISA.

The second major issue is whether loss of a detectable antibody response to a specific VLP type indicates loss of protection against that type. This question was brought to the forefront by the report that, by 5 years after vaccination, 35% of Gardasil- vaccinated women had lost detectable antibodies to HPV-18 VLPs, as measured in Merck’s cLIA [23]. In contrast, only 1% of vaccinees had undetectable HPV-16–specific cLIA responses at this time point. Also, 5-fold higher peak titers were measured for HPV-16 than for HPV- 18. There are 2 reasonable explanations for these findings. One is that Merck’s HPV- 18 VLPs are intrinsically less immunogenic than its HPV-16 VLPs. The HPV-18 VLPs do not disassemble under the low salt-reducing conditions used to disassemble HPV-6, -11, and -16 VLPs during the manufacturing process [24]. This finding suggests that HPV-18 VLPs may have some structural properties that differ from the other Merck VLPs and from the GSK HPV-18 VLPs produced in insect cells, which disassemble under similar conditions. The second possibility is that the HPV-18 cLIA is simply less sensitive and/ or does not detect a substantial fraction of the protective antibody responses in a subset of women. Consistent with the latter interpretation is the observation that protection against HPV-18 infection does not appear to be waning in years 4 and 5 after Gardasil vaccination [25]. It is possible that, in selecting an HPV-18 monoclonal antibody that does not cross-react with the closely related type HPV-45, an antibody was chosen that is less representative of the polyclonal antibody response to HPV-18 VLP vaccination, compared with the degree to which the HPV-16 monoclonal antibody is representative of the polyclonal antibody response to HPV- 16 VLPs. It may be possible to distinguish between these alternatives by conducting HPV-16–specific and HPV-18–specific SEAP-NAs, which measure neutralizing antibodies regardless of where they bind on the capsid. However, the caveat that SEAP-NA titers are influenced by the presence of noninfectious interfering particles would have to be taken into account. Interestingly, Cervarix induced mean serum titers of antibody to HPV- 18 and -16 that differed by only ?2-fold, as accessed by GSK’s ELISA and the SEAP-NA, and titers remained in the detectable range for both assays over the 4 years women were examined after vaccination [22, 26]. A competitive ELISA based on J4, the same monoclonal antibody used in Merck’s cLIA, was less sensitive than GSK’s direct ELISA or the SEAP-NA [18].

The third major question is whether a vaccine that is more immunogenic is necessarily a better vaccine. This issue will likely receive increased attention once the results of a GSK clinical trial directly comparing the immunogenicity of Cervarix and Gardasil are announced [27]. If other characteristics of 2 vaccines are comparable, in general one would favor the vaccine that is more immunogenic. However, because no immune correlate of protection has been established for these vaccines, we do not know the minimal levels of antibodies needed for solid protection and therefore do not know whether higher levels would induce longer protection. For both vaccines, the mean VLP antibody titers reached a plateau by 2 years after vaccination, and there has been no sign of waning protection 4–5 years after vaccination [20, 28, 29]. Therefore, both vaccines might be sufficiently immunogenic to provide long-term or even lifetime type-specific protection. Also, virologists generally consider antibody titers on a logarithmic scale, and differences in titers of a few fold might be considered of questionable biological significance.

Because cross-neutralizing titers are lower than vaccine type–specific titers, there is a greater likelihood that they may wane to nonprotective levels over time. One could imagine that subtle differences in the folding of the conformation-dependent neutralizing L1 epitopes under the influence of the chaperones in yeast and insect cells might influence the display of the immunosubdominant cross-neutralization epitopes by the 2 vaccines. These differences could influence the titers of antibodies to nonvaccine types induced. Therefore, it will be important to also evaluate the titers of neutralizing antibodies to nonvaccine types. There are a number of HPV-16–related types and a number of HPV-18–related types that are each associated with 1%–5% of cervical cancers, and some of these types might be preferentially neutralized by the antibodies induced by 1 of the 2 vaccines. However, HPV-16 and -18 are only distantly related to HPV-6 and -11, which makes it unlikely that Cervarix, as currently constituted, would induce cross neutralizing antibodies to these genital wart–associated types.

CONCLUSION

There is a strong rationale for focusing on the antibody responses to the HPV VLP vaccines. Each of the 3 types of antibody assays discussed here will be useful in the evaluation of current and future prophylactic HPV vaccines. However, it will be important to remain cognizant of the unique strengths and limitations of each assay when interpreting the immunogenicity results of clinical trials.