Fortunately, vaccine developers did not wait for the conclusions of the prospective epidemiology studies before initiating work on prophylactic HPV vaccines. Perhaps because these investigators were primarily laboratory-based researchers, and even, as in our case, involved in the studies of the HPV oncogenes, they were not unduly inhibited by the weak correlation between HPV infection and cervical disease seen in the early epidemiologic studies. In any event, considerable activity aimed toward developing HPV vaccines was underway by 1990. A vaccine based on a live attenuated HPV strain was not considered a viable option for two reasons. First, HPVs could not be propagated in replicating cells in culture, so there was no reasonable means of virus production. Second, the virus was known to contain at least three oncogenes, E5, E6, and E7, and a vaccine that delivered oncogenes might not be considered safe for general use as a prophylactic vaccine. Therefore, most efforts involved subunit vaccine strategies. Since it was generally appreciated that most effective viral vaccines functioned primarily through the induction of neutralizing antibodies , attention was mainly focused on subunit vaccine strategies that might induce neutralizing antibodies to the papillomavirus virion proteins.
Almost equally important were the studies to develop in vitro neutralization assays that could critically evaluate the antibody responses to the vaccine candidates. In addition to protection from experimental challenge with wart-derived virions in rabbit, bovine, and canine models, there were only two assays for measuring neutralizing antibodies induced by papillomavirus vaccine candidates as of 1990. First, infectious events by bovine papillomavirus (BPV) types 1 and 2, but not HPVs, could be monitored in vitro by the induction of transformed foci in NIH3T3 and C127 mouse cell lines . Neutralizing antibody titers were determined from the reduction in the number of foci induced by a standard stock of wart-derived virions (Fig. 1). We had developed this assay in the late 1970s, and it formed the basis for studying morphologic transformation by BPV. Second,
no Ab no virus ariti-wtAcMNPV
Fig. 1 In vitro neutralization of BPV1 virions by BPV1 L1 VLP antisera conducted in 1992. Foci of transformed cells induced by BPV1 infection of mouse C127 cells were visualized by staining with methylene blue/carbol fuchin. "Anti-AcBPV-L1" was sera from a rabbit vaccinated with BPV1 VLPs derived from L1 recombinant baculovirus infected insect cells. Anti-wt AcMNPV was serum from a rabbit vaccinated with an extract of wild type baculovirus infected insect cells. Numbers refer to the dilution of the sera used in the assay. "No Ab" demonstrates the number of focal transformation events induced by the BPV1 inoculum without added serum and "preimmune" demonstrates the number of foci induced in the presence of the rabbit's serum prior to VLP vaccination no Ab no virus ariti-wtAcMNPV
Fig. 1 In vitro neutralization of BPV1 virions by BPV1 L1 VLP antisera conducted in 1992. Foci of transformed cells induced by BPV1 infection of mouse C127 cells were visualized by staining with methylene blue/carbol fuchin. "Anti-AcBPV-L1" was sera from a rabbit vaccinated with BPV1 VLPs derived from L1 recombinant baculovirus infected insect cells. Anti-wt AcMNPV was serum from a rabbit vaccinated with an extract of wild type baculovirus infected insect cells. Numbers refer to the dilution of the sera used in the assay. "No Ab" demonstrates the number of focal transformation events induced by the BPV1 inoculum without added serum and "preimmune" demonstrates the number of foci induced in the presence of the rabbit's serum prior to VLP vaccination wart-derived virions could be used to infect mucosal or cutaneous epithelial chips that were then placed under the renal capsule of athymic mice . Infection was monitored by hyperproliferative changes in the transplanted chips and antibodies could be semiquantitatively evaluated for inhibition of in vitro infection prior to transplantation. Infection by BPV1, CRPV, HPV1, and HPV11 (but not other HPVs), and infection inhibition by type-specific antibodies, was evaluated by this relatively cumbersome xenograft assay .
Although papillomavirus virions were relatively poorly characterized at the time, it was known that the nonenveloped icosahedral capsid was composed of 360 copies of L1, the major capsid protein, and 12-72 copies of L2, the minor capsid protein . Early studies employing animal papillomaviruses had established that interperitoneal or intramuscular injection of wart-derived virions, which does not induce an active infection, was effective at inducing neutralizing antibodies  and protecting from experimental infection [23, 24]. In contrast, IM injection of denatured virions or disordered L1 polypeptides derived from Escherichia coli inclusion bodies was ineffective at inducing neutralizing antibodies or protecting from experimental challenge [25, 26]. From these results, it was concluded that L1 needed to be in a "native" conformation to induce high titers of neutralizing antibodies. To this day, no relatively small peptide fragment of L1 has been shown to induce neutralizing antibodies. However, low levels of neutralizing antibodies, and partial protection from experimental challenge with homologous virus was induced using a bacterial fusion protein of either CRPV or BPV L1, suggesting that at least a minority of L1 molecules could display conformation-dependent neutralizing epitopes after bacterial expression [25, 27, 28].
At this point in time, it was appreciated that the major capsid protein of at least some naked icosahedral viruses had the intrinsic capacity to self-assemble into VLPs. For instance, E. coli-derived VP1 of several polyomaviruses, which are structurally similar to papillomaviruses, had been shown to self-assemble into morphologically correct VLPs from capsomeric subunits in in vitro reactions . In addition, the commercial hepatitis B vaccine since the 1980s was based upon self-assembly of the viral S protein into lipid bilayer-containing VLPs . Thus, by 1990 it was reasonable to consider a VLP-based vaccine displaying conforma-tionally correct L1, or L1 and L2, as a potential candidate for an HPV vaccine.
In 1991, Zhou and Frazer reported that coexpression of HPV16 L1 and L2 in monkey CV-1 cells via a vaccinia virus vector resulted in the generation of "viruslike" particles that could be concentrated by sucrose gradient centrifugation . These irregular particles had a mean diameter of 35-40 nm, compared to the 50-55 nm symmetrical particles reported for authentic virions, and were described as "incorrectly assembled arrays of HPV capsomeres." Particles were not detected when L1 was expressed separately. The vaccine potential of these L1/L2 capsomere arrays could not be critically evaluated because no HPV16 neutralizing assay was available at this time. Whether this study was the foundation for the subsequent development of the HPV prophylactic vaccines or taught against their development became the subject of much discussion, particularly among patent lawyers.
Shortly thereafter, Richard Schelgel and colleagues published the production of the L1 of HPV1 (a cutaneous wart type) in COS cells using an SV40 replicon vector . The protein reportedly reacted with monoclonal antibodies that recognized native but not denatured HPV1 virions. However, no VLPs could be detected in cell extracts, and the L1 was neither purified nor evaluated for the induction of neutralizing antibodies. Thus it was unclear from these reports whether L1 alone, or L1 and L2 together, could assemble into conformationally correct VLPs or, if they could, whether they would efficiently induce neutralizing antibodies, and thus be attractive vaccine candidates. Because papillomavirus virions are only produced in terminally differentiated squamous epithelial cells, there was a concern that molecular chaperones unique to these cells might be required for correct folding and assembly of the virion proteins into capsids, and so it might be impossible for generation of morphologically correct VLPs in normal cell culture.
Our involvement in papillomavirus vaccine development began in early 1991, soon after Reinhard Kirnbauer, a dermatologist from the University of Vienna, arrived to begin a postdoctoral fellowship. As an initial project, we asked Reinhard to attempt to generate and characterize papillomavirus VLPs, both as vaccine candidates and as reagents for basic studies of virion/cell interactions, since there was no ready source of authentic virions. This project was a clear departure from the core activities of the laboratory, which for more than 10 years had centered on the molecular biology of the viral transforming genes and the regulation of viral gene expression. We have often reflected how fortunate we were to be in the intramural program of the National Cancer Institute, because our review was primarily retrospective. It is very doubtful that, given our lack of experience in virion structural proteins or vaccines, we would have convinced an extramural grant review committee to fund the project. After a review of potential production systems, we decide to express L1 in insect cells via recombinant baculovirus vectors, primarily for two reasons. First, they were known to produce exceptionally high levels of recombinant protein, and a critical capsid protein concentration might be needed to drive the VLP assembly reaction . Second, the FDA had already approved clinical trials of other proteins produced in this system. Thus there was reason to expect that, should preclinical vaccine studies produce encouraging results, there would be a reasonable path to GMP production and human vaccination trials.
We focused initially on expressing BPV1 L1, rather than an HPV L1, because we had a stock of cow wart-derived infectious BPV virions in the laboratory and expertise in the in vitro focal transformation assay that could be used to evaluate neutralizing antibodies elicited by any vaccine candidate that would be generated in the study. In relatively short order, particularly considering his limited expertise in molecular biology, Reinhard was able to generate the L1 recombinant baculovirus, infect insect cells, and demonstrate robust L1 expression in the infected cells. We were thrilled when the first set of electron photomicrographs of thin sections from the infected insect cells revealed approximately 50 nm particles, the size expected for authentic papillomavirus virions. In addition, particles purified by CsCl gradient centrifugation displayed a regular array of capsomeric structures in transmission electron photomicrographs (Fig. 2). Most importantly, rabbits injected with partially
purified VLPs, or even crude extracts of the infected cells, generated very high titers of BPV neutralizing antibodies (Fig. 1). The titers were so unexpectedly high that Reinhard had to conduct three consecutive neutralizing antibody titration experiments, each involving successively higher sera dilutions, until he finally reaching an endpoint titer when the sera was diluted more than 100,000-fold. In contrast to the titers of 105 induced by our BPV1 VLPs, the highest titer reported for bacterially derived BPV1 L1 had been 36 . Consistent with the concept that neutralizing antibodies recognize conformationally dependent L1 epitopes, no neutralizing activity was detected in the sera of rabbits inoculated with denatured VLPs. The above results established that BPV L1 without any other viral proteins had the intrinsic capacity to assemble into VLPs that were able to induce high titers of neutralizing antibodies . It is worth noting that assembly of a major capsid protein in VLPs did not necessary predict the ability of the VLPs to efficiently induce neutralizing antibodies. Neil Young's laboratory had recently published the production of B19 parvovirus VLPs . That study demonstrated that the major capsid protein was sufficient to generate VLPs, but significant neutralizing antibodies were induced only when the minor capsid was coassembled into the particles.
Since the prevention of cow warts was not our ultimate goal, the obvious question became whether these findings could be translated into HPV vaccines. There were several possible explanations for why L1 VLPs were not detected in the HPV16 and HPV1 studies described earlier. One possibility was that the levels of L1 expression were insufficient to drive the assembly reaction. Another possibility was that L1s of different types had different intrinsic self-assembly capabilities. The latter possibility was raised since BPV1 belongs to a group of animal papillomavirus that uniquely induces cutaneous fibropapillomas and is distantly related to HPV16 and HPV1. To address these possibilities, Reinhard expressed HPV16 via an analogous recombinant baculovirus and was able to demonstrate abundant L1 protein expression in the infected insect cells. However, to our surprise and consternation, it was exceedingly difficult to find VLPs in the extracts. We estimated that the efficiency of HPV16 VLP formation was 1,000-fold lower than for BPV L1 . Fortunately, Reinhard had also begun work on generating L1 VLPs of rhesus papillomavirus (RhPV1), since we were considering the possibility of future vaccine studies in nonhuman primates. RhPV1 L1 produced excellent yields of VLPs in our production system, much like BPV1 L1. This was an important observation because RhPV1 is closely related to HPV16, in fact more closely related to HPV16 than HPV16 is to some other high-risk HPVs, such as HPV18. We therefore felt that it was unlikely that the inefficient assembly of HPV16 was attributable to its phylogeny. It seemed more likely that the L1 of the prototype HPV16 clone used in our study was an assembly defective mutant. This widely distributed clone was the initial HPV16 DNA isolated by Harald zur Hausen's group . The fact that it was isolated from a cervical cancer, and cancer cells are genetically unstable, supported the mutant hypothesis. To test this possibility, we obtained from the DKFZ's Mathias Durst and Lutz Gissmann two HPV16 clones that they had isolated from low-grade virus-producing lesions. We were much relieved to find that the L1 of these clones efficiently produced VLPs in our baculovirus expression system  (Fig. 2). Sequencing of the clones revealed that a single aspartate to histidine change in the prototype strain was responsible for the assembly defect of the prototype L1. Whether the inability to demonstrate assembly of HPV16 L1 into morphologically correct VLPs in the 1991 study  was due to the use of the prototype gene, the production system used, or some other technical difference was never firmly established.
In the same year as we published the efficient assembly of HPV16 L1 VLPs, the laboratory of Denise Galloway published the production of HPV1 L1 VLPs using a vaccinia virus vector, indicating that recombinant vaccinia virus infected cells could be permissive for L1 VLP self-assembly . In addition, Bob Rose et al. published the production of HPV11 L1 VLPs using a baculovirus vector . Thus, by the end of 1993, there was considerable evidence that the ability to self-assemble into VLPs is a general property of wild type papillomavirus L1 proteins. These findings were substantiated in many subsequent studies involving the L1s of other human and animal papillomavirus types.
Armed with the data demonstrating induction of high titer neutralizing antibodies by L1 VLPs, and a provisional patent application, we began a series of visits to commercial vaccine manufactures. We were received with interest, but also some reservations, by most of the companies. No doubt this was in part because we lacked credentials in vaccine development. However, there was also considerable skepticism in general about the prospects of developing an effective vaccine against a sexually transmitted mucosal pathogen. A cautious view at that time was understandable, given the conspicuous failure, despite extensive efforts in both academic and commercial sectors, to develop effective vaccines against HIV, HSV, and other STIs. The exceptional response came from Maurice Hilleman, one of the godfathers of modern vaccinology, who was then an emeritus employee at Merck. After a short private presentation of our data in his office, he unequivocally stated that the vaccine was going to work and Merck was going to make it. He turned out be correct on both accounts. Shortly after our discussion with Hilleman, we were approached by MedImmune, a biotechnology company headquartered just a few miles north of our laboratory in Bethesda, with an expression of interest in our vaccine concept. We were enthusiastic about the prospects of more than one company undertaking the commercial development of the vaccine. In our view, public health interest would most likely be served by competition during the development phase and hopefully, eventually in the marketplace. In keeping with its general policy, the NIH ultimately granted nonexclusive licenses to both Merck and Medimmune. Merck also exclusively licensed competing patent applications from Zhou and Frazer, and MedImmune exclusively licensed competing patent applications from the Schlegel and Rose groups. This led to a long series of patent disputes that for practical purposes was settled in 2005 when Merck and GlaxoSmithKline (which by this time had subli-censed Medimmune's HPV vaccine patent portfolio) agreed to a financial settlement. This agreement gave the two companies unrestricted and coexclusive access to the papillomavirus VLP vaccine patent claims of all four parties. This exclusivity solidified the sustained commercial investment in the vaccines.
Further insights into papillomavirus VLP vaccines were provided by the publication in 1995-1996 of several proof-of-concept trials in animal models. Intramuscular injection of low microgram amounts of L1 VLPs of COPV, CRPV, or BPV4, even without adjuvant, was shown to induce strong protection from high dose experimental challenge in dogs, rabbits, and calves, respectively [39-42]. Protection was type specific and could be passively transferred in immune sera or purified IgG, indicating that neutralizing antibodies were sufficient to confer protection. However, VLP vaccination did not induce regression of established lesions, suggesting that HPV VLP vaccines would not be therapeutic. These overall encouraging results strengthened commercial and academic interest in the vaccines.
The animal challenge studies noted earlier assessed cross-type protection against distantly related animal papillomavirus types. It also seemed important to assess the potential for cross-protection of HPV VLP vaccines against genital HPV types, which form relatively closely related clusters around HPV16, HPV18, and HPV6, respectively. Such an assessment would help in making decisions concerning the valency of an HPV VLP vaccine aimed at preventing cervical cancer and/or genital warts. Richard Roden, then a postdoctoral fellow in the laboratory, initially addressed the question by investigating the ability of homologous and heterologous L1 VLP rabbit sera to inhibit VLP agglutination of mouse red blood cells. Hemagglutination inhibition (HAI) has been used for a number of viruses as a surrogate for a true virus neutralization assay. Homologous HAI titers of several thousand or more were obtained with the VLP sera. However, only low HAI titers were seen across types, and then only for closely related pairs such as HPV6 and HPV11 or HPV18 and HPV45, arguing that VLPs would induce type-restricted protection against HPV infection . However, we discovered that only a subset of the monoclonal antibodies that neutralized BPV1 in our focal transformation assay exhibited HAI activity. Therefore, we considered HAI to be a somewhat imperfect surrogate assay for assessing HPV infection inhibition.
The limitations of the HAI assay led Richard to develop an in vitro neutralization assay based on HPV pseudoviruses . The pseudovirions were generated by coexpression of L1 and L2 via Semliki Forest Virus vectors in a hamster cell line that contained a relatively high copy number of autonomous replicating BPV1 genomes. Expression of L1 and L2 in these cells led to assembly of capsids that had incorporated the BPV genome. Infectious events could be scored by counting transformed foci on C127 cells, as in the case of the authentic BPV1 (Fig. 1). Using this rather laborious neutralization assay, or easier later versions in which a marker gene-expressing plasmid rather than the BPV genome was encapsidated , we were able to demonstrate that wild type HPV16 L1 VLPs, but not the prototype HPV16 L1 protein, generated high titers of HPV16 neutralizing antibodies. In agreement with the HAI data, VLP-induced neutralizing antibodies were clearly type restricted, with low levels of cross-neutralization detected only for closely related types. However, antibodies raised to VLPs of one HPV16 L1 variant were equally effective at neutralizing pseudovirions of other HPV16 variants . From these studies, we concluded that HPV genotypes represent distinct serotypes, but that it is unlikely that distinct serotypes exist within a given genotype. The results supported the prediction that immunoprophylaxis by HPV VLP vaccines would be type restricted, which implied that multivalent vaccines would be needed for broad-spectrum protection against HPV-induced diseases.
Was this article helpful?
Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.