Main

Globally, cervical cancer is one of most common cancers in women, killing about 0.25 million women per year. Cervical cancer (Box 1) is attributed to persistent infection with a 'high-risk' subset of human papillomaviruses (HPVs) (Table 1), and is the first cancer recognized by the World Health Organization (WHO) to be 100% attributable to an infection2. At present, optimal treatment of cervical cancer combines surgery or radiotherapy with adjuvant chemotherapy, and delivers cure rates of almost 100% for locally invasive Federation of International Gynaecologists and Obstetricians (FIGO) stage 1 disease. Disease that has spread beyond the pelvis — FIGO stage 4 — is not generally regarded as curable, and many patients present with stage 4 disease in the developing world. Prevention and early detection of cervical cancer relies on cytology screening programmes, which have markedly reduced cervical cancer death rates where they are available. However, more than 80% of cervical cancer occurs in the developing world, where neither population-based screening nor optimal treatment is available. Further, present treatments and to a large extent present screening strategies do not acknowledge the viral aetiology of this common cancer.

Table 1 Classification of papillomaviruses
Full size table

The link between cervical cancer and HPV

Several cancers are now attributed, in part, to the consequences of chronic viral or bacterial infection. For example, the most successful available cancer vaccine prevents hepatocellular cancer that results from chronic infection with hepatitis B virus (HBV)3. The crucial observation by Zur Hausen4 that infection with papillomavirus might be responsible for cervical cancer in humans was subsequently confirmed through long-term cross-sectional and case-controlled epidemiological studies. Their conclusions are supported by in vitro studies, showing the probable molecular mechanisms by which infection with HPV promotes epithelial cell immortalization and initiates the development of cervical cancer5.

Only a minority of HPV GENOTYPES infect the anogenital epithelium, and most of these infections are self limiting. However, persisting infection of the anogenital epithelium — over five years to a lifetime — with one of the more than ten high-risk genotypes of HPV is causally linked to cervical cancer, and other anogenital malignancies are also, to some degree, caused by HPV infection. Infection with HPV16 — the highest risk genotype — is responsible for more than 50% of cervical cancers worldwide6. Nevertheless, more than 95% of incident HPV infections of the anogenital tract resolve over three to five years7, and cancer can be estimated to develop in less than 5% of HPV16-infected individuals over their lifetime8.

The biology of HPV infection

Papillomaviruses are double-stranded DNA viruses (Fig. 1), which replicate exclusively in stratified squamous epithelia, using the differentiation of the epithelium to regulate their replication (Fig. 2). Virions penetrate the epithelium through microabrasions and infect epithelial stem cells that are located in the basal epithelial cell layer. In these cells and their progeny, known as transit amplifying cells, they replicate their DNA episomally, using two virus-encoded non-structural proteins, early 1 (E1) and E2, and the cellular DNA-replication machinery. Transit amplifying cells populate the basal layer of the epithelium. Expression of the viral E6 and E7 non-structural proteins delays cell-cycle arrest and differentiation, which is normally observed as epitheial cells move up from the basement membrane to become mature keratinocytes. This delay of cell-cycle arrest allows further viral episome replication using the host DNA-replication machinery in suprabasal epithelial cells, and produces the thickening of the skin (or wart) characteristic of some papillomavirus infections. When the differentiation of replicating epithelial cells to non-replicating mature keratinocytes eventually occurs, virus-encoded structural proteins, late 1 (L1) and L2, are assembled in the cell nucleus. Mature virions then assemble and are released from the epithelium within the superficial epithelial cells.

Figure 1: Schematic of the human papillomavirus 16 (HPV16) genome showing the arrangement of the major non-structural and capsid genes.
figure 1

The three circles correspond to the three reading frames in which the sense strand can be translated. There are no known gene products produced from the antisense strand. Viral protein early 4 (E4) is encoded by a messenger RNA transcript that includes the initial amino acids of the E1 gene. The region between late 1 (L1) and E6 is an important transcriptional regulatory region — the mRNAs encoding most non-structural (E6, E7, E1, E2, E4 and E5) and capsid (L1 and L2) genes originate in this region. Most papillomavirus genomes resemble HPV16 in general organization.

Full size image
Figure 2: The location in squamous epithelium of the main stages of the papillomavirus life cycle.
figure 2

Cervical stratified squamous epithelial cell architecture and the expression of human papillomavirus (HPV) proteins after infection. Daughter cells of epithelial stem cells divide along the basement membrane and then mature vertically through the epithelium without further division (right side). After introduction of HPV into stem cells in the basal layer of the epithelium, expression of viral non-structural proteins occurs. Under the regulation of these proteins, the dividing-cell population expands vertically and epithelial cell differentiation is delayed and is less complete. Viral proteins are expressed sequentially with differentiation as shown, and mature virions are produced only in the most superficial layers of the epithelium. Intraepithelial antigen-presenting cells (APCs) are depleted in the HPV-infected epithelium.

Full size image

Infection with high-risk HPV sometimes results in integration of the viral episome into host DNA. If integration interrupts the viral E2 gene, overexpression of E6 and E7 proteins occurs owing to loss of E2-mediated repression of transcription of these genes. As a result, HPV-infected cells with integrated HPV DNA acquire extended lifespans, retain the capacity to proliferate, and tend to develop and perpetuate mutations in the germline DNA attributable to the actions of E6 and E7 proteins. Such cells are dysplastic, and marked dysplasia of the cervical epithelium is the precursor lesion of cervical malignancy, although dysplasia and infection might resolve spontaneously (see later).

The natural immune response to HPV

Resolution of HPV infection involves specific immune responses, as immunosuppression in transplant recipients9 or patients infected with HIV10 prevents the resolution of anogenital HPV infection and increases the risk of progression of HPV infection to malignancy. The natural immune response to HPV after infection is however weak, when compared with that of most other viral infections11 (Box 2). There is little tissue destruction associated with HPV infection, and no virus-associated double-stranded RNA to invoke innate immune responses and initiate specific adaptive immunity. Papillomaviruses would, therefore, seem able to evade the immune responses of their mammalian hosts by avoiding the main triggers that initiate an immune response to viral infection — the host remains immunologically ignorant of the infection. In confirmation of this, natural immune responses to primary HPV infection are slow to appear, even though the viral proteins, if delivered optimally with adjuvants, are immunogenic in animals, as well as in uninfected and infected humans (see later). Papillomavirus functions provide additional potential mechanisms for evading the induction of immune responses and immune effector mechanisms12. The E6 protein inhibits epithelial-cell–dendritic-cell interactions13, perhaps accounting for the depletion of dendritic cells observed in the HPV-infected cervical epithelium14. The E6 and E7 proteins block the production of and responsiveness of infected cells to type 1 interferons (IFNs)15,16,17,18. The E5 protein inhibits pH-dependent processing of antigenic peptides19. Evaluation of the significance of these mechanisms for viral immune evasion awaits the production of function-deficient infectious virions.

Antibodies specific for the HPV16 L1 major capsid protein (the most readily detectable virus-specific immune response) develop four months to five years after the first infection, although one-third to one-half of women develop no measurable capsid-specific antibody response at all20. Antibodies specific for the E7 non-structural protein appear only with the onset of invasive cervical cancer21, even though the protein is expressed in acute infection and during the 10–20 year progression of chronic infection to invasive malignancy. Whether the immune response, which eventually results in natural infection in some people, protects the host from further viral challenge is hard to establish. However, cows22,23, rabbits24 and dogs25 infected or immunized with their species-specific papillomaviruses develop neutralizing antibodies, and seem resistant to further viral challenge, indicating that protective immunity can develop after successful resolution of primary infection. Cell-mediated immune responses to papillomavirus proteins following natural infection in humans have been harder to define, mainly for technical reasons that relate to the probable low frequency of circulating effector cells. However, regression of papillomavirus-associated lesions in humans, as in animals, is associated with a cellular infiltrate of T cells26,27; there is presumptive evidence of specific cellular immune responses to at least some viral non-structural proteins28,29, and one study indicates that delayed-type hypersensitivity to the viral E7 protein correlates with resolution of clinical disease30.

Vaccines for HPV infection

Viral vaccines that are presently licensed for use in humans are prophylactic against future challenge with infectious virus. It is believed that all prophylactic vaccines work through the induction of virus-neutralizing antibodies, and markedly reduce the number of cells that are infected after challenge with virus, and so prevent the clinical disease associated with infection. Successful vaccines immunologically mimic the infections they prevent, priming the adaptive immune system to recall specific effector functions ('memory') to future encounters with the infectious agent, and this restimulation on challenge boosts immunity and, therefore, protection against future viral challenge.

Vaccines might also be developed to eradicate cells that are already infected with virus. Such vaccines, known as therapeutic vaccines, would be designed to prime the antigen-specific T-cell-mediated effector mechanisms that are used by the host immune system to control and eradicate viral infections. These include CD8+ T-cell-mediated cytotoxicity targeted at infected cells, and secretion by T cells of cytokines with direct (IFN-α and IFN-γ) or indirect (interleukin-1 (IL-1) or IL-12) effects on virus-infected cells. If a virus-specific effector T-cell response were present during initial viral challenge, as a consequence either of recent immunization or of stimulation by viral antigens of memory cells induced by previous immunization, then early termination of viral replication might prevent disease. If the effector T-cell responses were stimulated when viral pathology was already present, therapeutic elimination of infected cells and, therefore, of disease would be the goal.

HPV prophylactic vaccines

As discussed earlier, the main effector mechanism for vaccine-induced prevention of viral infection is neutralizing antibody. Pioneering studies by Jarrett and colleagues31 used formalized bovine papillomavirus virions as a vaccine, and these studies showed that vaccination could be used to induce papillomavirus-specific antibodies and prevent infection with virus. However, replication of papillomavirus requires epithelial cell differentiation to allow the production of viral capsid proteins32,33, as described earlier, and this prevents conventional approaches to the production of virus for vaccines in continuous cell culture in vitro. Although systems have recently been developed that allow recapitulation of the entire papillomavirus lifecycle in epithelial cell raft culture34 and in yeast35, the yield from these culture systems is insufficient to generate a conventional killed vaccine.

Virus-like particles. Recombinant DNA technology allows expression of the L1 major capsid protein of papillomavirus (Table 2) by various cell types, and expression of L1 protein by eukaryotic cells is followed by self assembly of the recombinant L1 into VIRUS-LIKE PARTICLES (VLPs)36,37. L1 VLPs mimic the natural virus structurally and, unlike denatured L1 protein, elicit high titres of virus-neutralizing antibodies in animals and humans. Although the amino-acid sequence of L1 is conserved between HPV genotypes, most genotypes seem to be serotypically distinct. An explanation for this observation was provided by definition of the crystal structure of the L1 molecule as assembled in VLP pentamers38. Variability between papillomavirus genotypes is concentrated over a few amino acids in the solvent-facing apices of three loops of the ∼500 amino-acid L1 peptide chain; these loops contain the sites to which neutralizing antibodies are directed. In keeping with this observation, vaccines based on papillomavirus VLPs seem to offer genotype-specific protection against infection with papillomavirus in animals and humans. Passive-transfer studies in animals have shown that antibody is sufficient39 to provide host protection against papillomavirus infection after VLP immunization.

Table 2 Papillomavirus proteins
Full size table

Two important randomized placebo-controlled studies1,40 on young sexually active women have shown that vaccination with HPV16 VLPs induces absolute protection over 12–18 months against persisting infection with HPV16, in association with high and persisting titres of neutralizing antibody specific for the virus, and similar protection against virus-associated premalignant lesions of the cervix. Interestingly, both studies have shown a lower incidence, but not prevention, of transient cervical infection in immunized individuals than in controls. As increased risk of cervical cancer is associated with persistent infection, transient infection in immunized individuals is unlikely to be of clinical consequence. Immunologically, however, the observation that immunization reduces the incidence of transient infection, but absolutely prevents persistent infection is puzzling. As protection is antibody dependent and immunization with VLPs with conventional alum adjuvants mainly induces an antibody response, a possible explanation is that transient infection results from infection of more superficial cells in the cervical epithelium, which are less protected by systemic antibody, and which have a defined and short lifespan. Epithelial stem cells are self renewing, and therefore infection of these cells is more likely to produce persistent infection. As these cells lie deeper in the epithelium they would be more exposed to systemic virus-neutralizing antibody. Although VLP-based vaccines induce both systemic and mucosal immunity41, the relative contribution of each to protection is undefined, yet these data indicate that antibody might have a particular role in protecting cells that are vulnerable to persisting infection.

Protection against HPV-associated diseases after VLP vaccination is genotype specific in animals, and antibody induced by VLP-based vaccines efficiently neutralizes in vitro only the genotype to which the antibody was raised42. Prophylactic vaccines that are envisaged at present incorporate two high-risk genotypes of HPV (HPV16 and HPV18), which together account for about 70% of cervical cancers6. If vaccine-induced protection is genotype specific in humans, vaccines presently under trial will, at best, prevent only two- thirds of cervical cancer in successfully immunized women. Further clinical trials will define the duration of immunity to HPV infection after immunization of healthy and immune compromised men and women, and children prior to sexual activity, and so will better define the role and optimal use of these vaccines as part of the various strategies available for the prevention of cervical cancer. Follow-up studies on immunized individuals might also define the pathogenic significance of HPV infection for cancers, in which the causal role of oncogenic HPV infections remains controversial, including oesophageal and oral cancers.

HPV therapeutic vaccines

Animal models have recently confirmed a role for specific immunity in cancer surveillance43. Cervical cancer is the only human cancer that is 100% attributable to infection with a virus — human papillomavirus — and the cancer and its precursor lesion, CERVICAL INTRAEPITHELIAL NEOPLASIA (CIN), continue to express defined viral proteins (Table 1). HPV infection should, therefore, be an optimal 'test case' for the probable effects of immunological intervention to prevent or treat a human cancer, as it meets many of the requirements for successful specific immunotherapy (Box 3).

VLP-based prophylactic vaccines might be able to prevent infection with high-risk HPV, if delivered to women before exposure to the virus. But even if prophylactic vaccination were to begin today, there is nevertheless a compelling case for developing immunotherapy for existing HPV infection, as this has the potential to prevent the estimated 5 million cervical cancer deaths that will otherwise occur over the next 20 years as a consequence of existing HPV infections. Immunotherapy could have an immediate impact on the incidence of cervical cancer, whereas prophylactic vaccines will take many years to reduce deaths from this disease. Cervical cancer cells and cells of the premalignant anogenital epithelial lesions that arise from chronic HPV infection express up to eight papillomavirus-encoded proteins (L1, L2 and E1–E7) (Table 2). E2, E6 and E7 proteins are of particular interest for vaccine development at present. The E7 protein seems to induce protective cellular immunity in human premalignancy30. The E7 protein also induces genotype-specific antibody after the onset of invasive cervical cancer21, although this immune response is clearly not host protective.

Animal models of therapy for HPV infection

Papillomavirus infections have been studied in several animals and have shed light on HPV immunology and immunotherapy. Animal infections, including bovine papillomavirus (BPV), cottontail rabbit papillomavirus (CRPV), canine oral papillomavirus and rabbit oral papillomavirus, have allowed detailed study of the natural immune response to infection with papillomavirus, and have been crucial in defining the use of prophylactic vaccination. Most animal papillomavirus infections are, in contrast to high-risk HPV infections, rapidly self limiting, induce strong immune responses that protect against further infection and do not promote malignancy. The natural regression of these lesions is thought to reflect expression, early in epithelial cell differentiation, of the viral late proteins and a consequent better natural immune response to infection44. Some animal papillomavirus infections (BPV1, BPV4 and CRPV) more closely resemble high-risk HPV infections, with general resolution but occasional persistence and risk of malignancy; the risk of progression being, in part, genetically determined and MHC linked45, as is also observed in humans46. Immunotherapy, targeted at virus-encoded structural and non-structural proteins, can modulate the course of disease in each of these animal infections, if given early after viral challenge and before the onset of lesions. For self-limiting infections, the E2 protein seems an effective target. For persisting infections, the best result, perhaps unsurprisingly, occurs after administration of a vaccine based on all of the viral non-structural proteins47.

A range of mouse transplantable tumours have been engineered that express papillomavirus proteins. These tumours are generally susceptible to a wide range of specific and non-specific immunotherapeutic approaches48. These include polynucleotide vaccines47,49,50,51; viral52,53 and bacterial54 vectors that express various viral non-structural proteins; and protein-55 or peptide-based56 immunogens, either incorporating innate immune stimulants (such as heat-shock protein57) or delivered with adjuvants that are designed to promote cell-mediated immune responses58. Cell-based therapy with antigen-primed dendritic cells has also proven to be effective in these animal models59. Effective immunotherapy administered before tumour challenge includes an antigen-specific component, whereas effective immunotherapy after tumour challenge can be achieved with enhancement of either innate or adaptive immunity, and seems to be optimal with both60. Rejection depends on antigen-specific CD8+ T cells61 and the induction of memory, and recall responses are determined by the generation of concomitant CD4+ T-cell responses61,62. Immunotherapeutic approaches used, with varying success, in human clinical trials (Table 3) have invariably proven to be effective in the transplantable tumour models, indicating that these models might be rather non-discriminatory for useful clinical products.

Table 3 Clinical trials of HPV-specific immunotherapy
Full size table

To explore more critically the role of immunotherapy for specific tumour-associated viral antigens, we established an animal model, in which skin expressing various papillomavirus antigens from a promoter (keratin 14, K14) that is active only in epithelial cells is transplanted into naive immunocompetent recipients63. This model shows that papillomavirus-encoded nuclear antigens, such as E6 and E7, invoke weak natural immune responses, which do not lead to graft rejection, in contrast to secreted antigens, such as ovalbumin (OVA)64, which prime strong immune responses with ensuing rapid graft rejection. Enhancing innate immunity, through concomitant infection with Listeria monocytogenes65 or a range of viruses, promotes the rejection of newly placed, but not well-established, K14–E7-transgenic grafts. Rejection is antigen specific and induces long-lasting memory, preventing the establishment of further grafts. Rejection of K14–E7 grafts can also be achieved by passive transfer of specific CD8+ T cells, activated in vivo by specific immunization, although not by immunization alone. Combined active/passive specific immunotherapy is, however, most effective for recently placed grafts. No therapy has yet been identified that reliably induces rejection of well-established grafts, even though these continue to express the E7 tumour antigen, which eventually promotes malignancy in the graft.

Taken together, these animal models indicate that high-risk HPV-encoded early proteins, such as nuclear antigens, are poorly cross-presented by epithelial cells to produce effective therapeutic immunity, and are poorly presented by epithelial cells to effector T cells induced through vaccination. Whether poor presentation reflects poor trafficking of effector cells to a skin target that lacks pro-inflammatory signals, or poor direct presentation by epithelial cells of relevant peptides in the context of MHC to effectors that reach the skin site of infection, or both, remains to be resolved. As E7-specific effector T cells cannot kill E7-expressing keratinocytes in vitro, whereas the same E7-specific effector T cells can kill most E7-expressing cell lines, and can also kill keratinocytes that are exposed to the relevant E7 peptide, processing of endogenous antigen for direct presentation by keratinocytes to T cells might be faulty66. As large numbers of primed antigen-specific T cells achieve skin-graft rejection in vivo, whereas small numbers can achieve transplantable tumour rejection, directing effector cells to traffic to skin targets might also be required for the development of an effective therapeutic vaccine for papillomavirus infection.

Human trials of immunotherapy for HPV infection

Several possible immunotherapeutic interventions have been tested for cervical cancer and the preceding premalignant lesion (Table 3). Cervical cancers, unlike their preceding premalignant HPV-induced lesions, express only two viral genes (E6 and E7) consistently as a result of selective integration of the papillomavirus genome into transformed cells. In general, trials for cervical cancer interventions have been carried out in patients with late-stage disease, who are already partially immune compromised by radiotherapy and/or chemotherapy. In addition, mutations and deletions in genes that are involved in antigen processing and presentation are commonly observed in cervical cancer cells67,68, indicating that these will be poor targets for specific immunotherapy (Box 3). Despite these problems, specific cellular and humoral immunity has been induced in some immunized individuals, albeit with limited evidence of therapeutic benefit (Table 3). One practical problem is that there is little information available from previous successful immunotherapeutic interventions in humans about optimal dose and frequency of delivery of antigen, or about the choice of adjuvant. Repeated immunizations and presently licensed alum-based adjuvants tend to bias the resulting immune responses towards the production of antibody and T helper 2 (TH2)-type cytokines (IL-4 and IL-5), and away from cytotoxic T cells and TH1-type cytokines (IFN-γ and tumour-necrosis factor)69,70 and consequently, are unlikely to be successful for eradicating tumour cells. Further, previous exposure to antigen as a result of infection with HPV might have already primed for ineffective TH2-biased immune responses69.

Trials of immunotherapy in patients with HPV-associated premalignancy are more likely to be informative, as the cancer-associated immunosuppressive mechanisms, of which the most relevant is impaired antigen presentation by cervical cancer cells as a consequence of mutations in MHC and TAP genes72,73, are unlikely to be of relevance, although there are potential immune-evasive mechanisms that are attributable to HPV infection itself71. Induction of specific immunity to papillomavirus early proteins can be achieved by several antigen-delivery systems, including saponin and enhanced oil in water emulsion adjuvants, recombinant viral vectors, heat-shock protein fusion proteins and polynucleotide antigen-expression systems. Clinical benefits are harder to define in these early phase clinical trials. Where these are indicated, they might, in part, reflect the induction of innate immunity, as the responses seen are not specific to disease caused by the papillomavirus genotype to which the vaccine is targeted. These results nevertheless justify pursuit of further randomized placebo-controlled trials, which might extend to trials of combined prophylactic and therapeutic vaccination to prevent and accelerate the elimination of early HPV infection.

Conclusions and the future

Vaccines to prevent infection with high-risk HPVs are conventional vaccines, inducing high titres of neutralizing antibodies, which provide host protection against infection and the potential malignant consequences of infection. Their use should follow ongoing Phase III studies of efficacy worldwide. Further studies with HPV VLP-based vaccines will be undertaken, which might broaden the range of HPV genotypes that are protected against and better define the duration of protection. These data will enable the rational introduction of these vaccines to appropriate target groups of young women before the onset of sexual activity. Educating healthcare professionals and public-health authorities about the benefits of the use of papillomavirus vaccines in the developed and developing world will require better definition of the local prevalence and natural history of HPV infections. As cervical cancer can largely be prevented by well-deployed cytology screening programmes, the interaction between a vaccine programme, which does not cover all papillomavirus genotypes and can protect against only 70% of cervical cancers, and a screening programme, which will probably find far fewer abnormalities for further management after vaccine introduction, will need careful planning.

Vaccines to treat persistent infections with papillomavirus are experimental. There is a need to define effective methods to overcome the challenges imposed by, on the one hand, poor presentation of viral nucleoprotein antigens that are expressed at low levels and, on the other hand, poor trafficking of effector T-cell populations to non-inflamed skin sites. However, papillomavirus-specific immune responses can be induced in humans, enhancement of innate immunity seems to assist the natural resolution of persistent infection with papillomavirus, and vaccines that promote trafficking of effectors preferentially to the skin are under development, allowing cautious optimism that effective immunotherapy for papillomavirus infection can be achieved.