HPV-mediated cervical carcinogenesis: concepts and clinical implications

Authors


Abstract

Persistent infection with a high-risk human papillomavirus (hrHPV) is generally accepted as a necessary cause of cervical cancer. However, cervical cancer is a rare complication of an hrHPV infection since most such infections are transient, not even giving rise to cervical lesions. On average, it takes 12–15 years before a persistent hrHPV infection may ultimately, via consecutive premalignant stages (ie CIN lesions), lead to an overt cervical carcinoma. This argues that HPV-induced cervical carcinogenesis is multi-step in nature. In this review, the data from hrHPV-mediated in vitro transformation studies and those obtained from analysis of clinical specimens have been merged into a cervical cancer progression model. According to this model, a crucial decision maker in the early stages following infection involves individual susceptibility for certain HPV types depending on the genetic make-up of immune surveillance determinants. Once a CIN lesion has developed, altered transcriptional regulation of the viral E6/E7 oncogenes, resulting in genomic instability and distinguishing the process of cell transformation from a productive viral infection, probably provides the subsequent important step towards malignancy. The additional (epi)genetic alterations that subsequently accumulate in high-grade CIN lesions may result in overt malignancy via immortality and growth conditions that gradually become less sensitive to growth-modulating influences mediated by cytokines and cell–cell and cell–matrix adhesions. The potential implications of hrHPV testing and some other biomarkers deduced from this model for cervical screening and the clinical management of CIN disease are also discussed. Copyright © 2006 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Infections with high-risk HPV are a necessary cause of cervical cancer

Cervical cancer is the second most common cancer amongst women world-wide, with a mean age standardized incidence rate of up to 18.7 per 100 000 women, as found in less developed countries 1. Based on strong epidemiological evidence, supported by basic experimental findings, there is no doubt that persistent infections with high-risk types of human papillomavirus (hrHPV) represent a necessary cause of cervical cancer 1–4.

Together, the viral genes E6 and E7 are responsible for the induction as well as maintenance of the transformed phenotype of cervical cancer cells, particularly by abrogating cell-cycle control and apoptosis mechanisms. Both E6 and E7 proteins can bind to multiple cellular targets (reviewed in ref 5). The interactions that are thought to be most relevant for their transforming functions are E6 binding, via the cellular protein E6-AP, to the tumour suppressor gene product p53, and E7 binding to the retinoblastoma tumour suppressor gene product pRb and its related pocket proteins, p107 and p130 6, 7. The first interaction results in rapid ubiquitin-dependent proteolytic degradation of p53, which prevents cells from undergoing p53-mediated apoptosis 8. A consequence of E7–pRB interaction is interference with cell-cycle control. In combination, the E6–p53 and E7–pRB interactions seem to compromise the accuracy of mitosis. In addition, hrHPV E6 can activate the telomere-lengthening enzyme telomerase independent of p53 binding, and E7 can induce abnormal centrosome duplication through a mechanism independent of inactivation of pRb and its family members (reviewed in ref 9). It is likely that these latter properties also contribute to the transforming characteristics of these viral oncoproteins.

Although HPV 16 and 18 are, in descending order, the most common HPV types in cervical cancer, a recent pooled analysis of world-wide case–control studies led to the classification of 15 different HPV types as hrHPV types, whereas another three types were considered probably high risk 4. hrHPV DNA can be detected in up to 99.7% of cervical squamous cell carcinomas (SCCs) 2, 4 and in 94–100% of cervical adenocarcinomas and adenosquamous carcinomas 10, 11.

Cervical SCCs, representing the most common histological type of cervical cancer, develop from pre-existing non-invasive squamous precursor lesions, referred to as cervical intraepithelial neoplasias (CINs) or squamous intraepithelial lesions (SILs). These lesions are classified histologically on the basis of atypia of epithelial cells that progressively extends from the lower parabasal layers of the squamous epithelium up to the whole thickness of the epithelium, depending on the grade (Figure 1). CIN 1 and low-grade SIL (LSIL) correspond to mild dysplasia, CIN 2 to moderate dysplasia, and CIN 3 to both severe dysplasia and carcinoma in situ, which in increasing order reflect atypia in the lower third to more than two-thirds of the epithelium. HSIL represents the combination of CIN 2 and CIN 3.

Figure 1.

Schematic presentation of the morphological alterations seen in consecutive premalignant cervical lesions and correlation of the cervical intraepithelial neoplasia (CIN) classification with squamous intraepithelial lesions (SILs)

From normality to CIN

It is well known that hrHPV infections, although necessary, are far from sufficient for the pathogenesis of cervical cancer. In fact, cervical cancer is only a rare complication of an hrHPV infection and requires several additive conditions and events to accumulate once an infection has become evident. This is particularly exemplified by the fact that HPV infections are very common in young women but frequently resolve spontaneously. The life-time risk to ever contract HPV is estimated to be 80% 12, and at least 80% of the hrHPV infections are likely to be transient, not even giving rise to CIN lesions 13. Why certain HPV infections tend to persist and give rise to lesions, whereas many others do not, is still questionable, but inter-individual variations in the capacity to clear the infection by an effective immune response would be one explanation. In this context, it is noteworthy that certain alleles of the polymorphic human histocompatibility leukocyte antigen (HLA) class I (such as HLA-Cgrp1 or HLA-Cgrp2) and II (such as HLA-DRB1*1301 or HLA-DRB1*1501), as well as killer immunoglobulin-like receptor (KIR) genes (such as KIR2DL1 or KIR3Ds1), alone or in combination, seem to confer protection or susceptibility in terms of high-grade CIN development 14, 15. Interestingly, carriers of commonly reported protective HLA II alleles displayed lower viral loads, short-term HPV infections, and a decreased risk of cervical carcinoma in situ15, 16. In earlier studies, low viral loads were found to confer a decreased risk of viral persistence and CIN development and therefore mainly represent clinically irrelevant HPV infections (reviewed in ref 17). On the other hand, it should also be realized that, according to the current concept, viral persistence originates from primary infections of the proliferating basal cells of the squamous epithelium. Primary HPV infections targeting differentiated, more superficial cells are a priori transient, since the viral DNA will be lost as the infected cells are shed during the terminal differentiation process. In conclusion, both the genetic background of the host with regard to immune surveillance mechanisms and the nature of the infected target cells are apparently decisive for the development of CIN lesions.

Heterogeneity of CIN lesions: productive infections versus cellular transformation

It is assumed that it takes on average 12–15 years before a persistent hrHPV infection may ultimately lead to overt cervical carcinoma 18, 19. Together with the often spontaneous regression of CIN lesions 20, a phenomenon associated with viral clearance 20, 21, this argues for a multi-step nature of HPV-induced cervical carcinogenesis.

Originally, it was thought that cervical SCC would always evolve from infected normal cervical epithelium via a continuum of long-lasting, consecutive CIN 1, CIN 2, and CIN 3 lesions. However, an alternative concept that finds increasing support is that many of the clinically relevant CIN 2/3 lesions may be rapidly induced within 2–3 years following infection 22, whereas it can be deduced from Wallin et al18 and Zielinski et al19 that it takes another 10–12 years to develop invasive cervical cancer. As a consequence, most CIN 1 lesions and some CIN 2 lesions should not be considered as true precursor stages of cervical cancer, but rather the cytopathological effect of a productive viral infection. Support for the latter comes from studies showing that low-grade CIN lesions more than incidentally harbour low-risk HPV types that confer a negligible risk of progression 23. Moreover, CIN 1 and some CIN 2 lesions that harbour hrHPV types display viral expression patterns suggestive of productive viral infections 24, 25. In these infections, active viral replication and virion production are strongly coupled to the differentiation programme of the infected epithelium. Low levels of viral activity in the infected basal cells characterize this process, with monomeric episomes being co-replicated with the genome of the host cell. Upon differentiation, viral transcription, including that of the viral E6 and E7 oncogenes, is markedly increased and vegetative DNA amplification and assembly of new virions occur only in squamous epithelia undergoing terminal differentiation (Figure 2A). Ultimately, newly assembled viruses will be released from the shed terminally differentiated cells. Except for providing the E1 and E2 proteins necessary for viral DNA replication, HPVs rely entirely on the host cell DNA replication machinery for viral DNA synthesis (reviewed in ref 26). As studied in organotypic raft cultures of primary human keratinocytes, the viral E7 protein appears competent in reactivating the cellular DNA replication machinery in a differentiation-dependent manner, thereby creating an S-phase environment supporting vegetative viral DNA replication in differentiating spinous cells 27. A likely function of E6 would be to prevent p53-mediated apoptosis that is normally induced in pRb-deficient cells 28. The fact that differentiated cells have already lost the ability to divide and are destined to be shed explains the low risk of progression of these low-grade CIN lesions, at least as long as E6/E7 expression is differentiation-dependent.

Figure 2.

Schematic presentation of a productive viral infection showing strong E6/E7 expression in differentiated epithelial cells (A). Nuclei expressing E6/E7 are indicated in blue. Viral episomes are indicated as black circles and viral capsid proteins are indicated in red. Virions are depicted schematically as black circles in red, icosahedral structures. (B) Schematic presentation of deregulated E6/E7 expression in (para)basal cells associated with cell transformation and failure of the viral life cycle

By contrast, some CIN 2 lesions, and CIN 3 lesions, exhibit a dramatic topographical change in viral gene expression, which includes an increase in E6/E7 expression in proliferating dysplastic cells 24, 25 (Figures 2B and 3). Although the mechanism underlying deregulated E6/E7 expression in proliferating cells is not yet understood, in vitro studies using epithelial raft cultures have shown that altered histone deacetylation may be a contributing factor 29. In addition, high-grade CIN lesions and cervical carcinomas often show integration of the viral genome into that of the host cell 30, a phenomenon that is accompanied by DNA aneuploidy 31. In fact, integration may also interfere with the normal regulation of E6/E7 expression, either by interruption of transcriptional control mediated by the viral E2 protein, or by increased stability of chimeric E6/E7–host cell transcripts, or a combination thereof 3, 32. Whatever the exact mechanism, deregulated E6/E7 expression in proliferating cells can induce chromosomal instability at both the numerical and the structural level and provide the driving force for further progression (reviewed in ref 33), similar to somatic alterations compromising the p53 and pRb pathways in many tumours that are unrelated to HPV. Although chromosomal instability is tightly linked to HPV 16 integration in cervical keratinocytes, it is still unclear whether integration represents the cause or simply the consequence of genomic instability 34.

Figure 3.

Deregulated HPV 16 E6/E7 expression in lower, dysplastic cell layers of a CIN 2 lesion at the transition to normal cervical epithelium. Left panel: Haematoxylin–eosin (HE)-stained section. Right panel: E6/E7 RNA in situ hybridization (RISH) on a consecutive section using a probe and methodologies described previously 118. Visualization was by confocal laser scan microscopy. Transcripts are indicated by red signal. Arrows indicate identical positions in both images

In summary, it is likely that uncontrolled E6/E7 expression in proliferating basal and parabasal epithelial cells is a phenomenon that distinguishes the process of cell transformation from a productive viral infection.

Additive events in cervical carcinogenesis: the value of in vitro model systems

In order to resolve the additive steps that are necessary for progression of CIN lesions towards cervical cancer, in vitro models have proven to be of great value. Originally, the transforming properties of hrHPV types in epithelial cells were established by studies showing their capacity to induce immortalization of primary human keratinocytes 35–38. Upon prolonged culturing, it was found that fully tumourigenic clones can emerge from HPV-immortalized cells 39, 40. This is in line with the assumption that hrHPV alone can induce malignant growth, as long as genomic instability triggered by its E6/E7 functions is allowed to proceed uninterrupted, leading to the accumulation of relevant additive (epi)genetic changes. Analysis of the transformation of epithelial cells in vitro has revealed at least four consecutive stages characterized by different phenotypes, ie extended lifespan, immortalization, anchorage-independent growth, and tumourigenicity in nude mice 39 (see the lower part of Figure 4). An extended lifespan is a direct consequence of hrHPV E6/E7 functions 41, and shutting off E6/E7 expression by either anti-sense RNA or repression of the promoter by ectopic E2 results in growth arrest with features resembling the senescence state where normal cells reach the end of their lifespan 5, 42. Somatic cell fusion experiments involving multiple hrHPV-containing cell lines demonstrated that each of the other three phenotypes mentioned above is under the control of a recessive regulatory process 39, 40. This argues that the inactivation of tumour suppressor genes is pivotal for these processes. Moreover, four complementation groups could be identified for immortalization, versus only one for anchorage-independent growth and two for tumourigenicity. This suggests that particularly an immortal phenotype can be obtained by interference with different regulatory events that can complement each other.

Figure 4.

Alignment of the different steps during cervical carcinogenesis (upper part) and in vitro transformation of epithelial cells mediated by hrHPV (lower part). See text for further description. Potential relevant genetic alterations are indicated in red. Chromosomes are indicated by ‘#’. Compl. = complementation

Most interesting is that for all of these phenotypes, HPV-transformed cell lines generated in vitro shared complementation groups with cell lines derived from cervical cancers, suggesting that common gene defects are involved in both in vitro progression and cervical carcinogenesis 39, 40. This indicates that both the functional data obtained from HPV-transfected keratinocytes and the data from somatic cell hybrids of cervical cancer cell lines can be combined to compose a progression model of cervical cancer. That in vitro progression data can be mapped to different stages of cervical carcinogenesis, as depicted in Figure 4, finds further support from observations following transfection of full-length HPV 16 or HPV 18 genomes into primary keratinocytes 41, 43, 44 (Steenbergen et al, unpublished data); all stable transfectants showed integrated viral DNA, histological features in organotypic raft cultures that progressively increased during passage from mild to severe dysplasia, E6/E7 expression in proliferating dysplastic cells in raft cultures, and cytogenetic instability at the numerical and structural level throughout all stages. In the next paragraphs we will mainly focus on functional findings from in vitro model systems that can be translated into a cervical carcinogenesis model, as depicted in the alignment scheme of Figure 4. The multiple recurrent genetic and (epi)genetic alterations with an as yet unknown functional consequence that are shared by in vitro transformed cells and clinical specimens have been reviewed elsewhere 45.

Immortalization

Whereas normal human keratinocytes have a finite proliferative lifespan and enter senescence within 50–100 cell divisions, cells with deregulated E6/E7 expression invariably show an extended lifespan 41. Recent data suggest that interference with p53 and pRb functions is sufficient to overcome this senescence barrier 28, 46. Eventually, E6/E7-expressing cells enter a period of crisis, after which most cells die and from which immortal subclones may arise, but only at low frequency. The bypass of this second barrier has been associated with activation of the telomere-lengthening enzyme telomerase and concomitant arrest of telomere shortening 41, 47. Telomere erosion is considered to provide an intrinsic division count system that determines the lifespan of normal somatic cells 48. Telomerase is a reverse transcriptase enzyme that can add six base-pair repeats to telomere ends, thereby compensating for telomere shortening during cell division. Of the subunits of the telomerase complex, the catalytic subunit (hTERT) is considered the rate-limiting factor. Its expression correlates strongly with enzyme activity, which is generally restricted to only subsets of normal cells including germ cells and stem cells, but is strongly induced in most cancer cells 49. That increased hTERT expression is sufficient to produce an immortal phenotype in hrHPV-containing epithelial cells can be concluded from studies in which ectopic hTERT was introduced into pre-crisis HPV 16- and HPV 18-containing keratinocytes 50 (Steenbergen et al, unpublished data). Ectopic hTERT counteracted telomere shortening and allowed cells to bypass crisis. Despite the fact that hrHPV E6 is capable of inducing hTERT and telomerase activity 51, several studies have shown that in the context of the whole HPV genome, E6 by itself does not necessarily cause telomerase activation in cervical keratinocytes 41, 50, 52. It is therefore more likely that E6 facilitates telomerase activation in conjunction with cellular changes. Indeed, microcell-mediated transfers of several human chromosomes, or part thereof, in HPV-immortalized cells or cervical cancer cells resulted in growth arrest and crisis-like features without having a known effect on E6 expression 50, 53–56. These chromosomes include chromosomes 2, 3, 4, 6, and 10, and further fine mapping analysis assigned the responsible region of chromosomes 4 and 10 to 4q35.1-qter and 10p14-p15, respectively 54, 55. For three chromosomes (ie chromosomes 3, 4, and 6), the immortalization suppressive effect correlated with suppression of telomerase 50, 53. Ectopic hTERT could prevent growth arrest mediated by chromosome 6 in both HPV 16-immortalized keratinocytes and the HPV 16-containing cervical cancer cell line SiHa, which provides functional evidence that reversal towards a mortal phenotype by this chromosome involves interference with hTERT activity.

Translation of the experimental data to clinical specimens representing the whole CIN to carcinoma sequence reveals intriguing results in favour of the alignment depicted in Figure 4. First, increased telomerase activity and concomitant elevated hTERT expression have been shown in almost all cervical SCCs and approximately 40% of CIN 3 lesions, whereas normal cervix and CIN 1 and CIN 2 lesions were devoid of any detectable telomerase activity 57. Although speculative, an attractive hypothesis would be that telomerase-positive CIN lesions have gained an immortal phenotype and have reached a point of no return in terms of malignant potential. In addition, allelic imbalances at 6q14–q22 were significantly increased in CIN 3 lesions and cervical carcinomas with telomerase activity and elevated hTERT expression 58. Combined with the in vitro data, this argues for the presence of a telomerase repressor on 6q. Finally, deletions at 3p, 4q35.1-qter, and 10p14–p15 are also relatively common events in high-grade CIN lesions and cervical carcinomas, although their relationship with telomerase activity is either not evident or not tested for 54, 55, 58. A potential candidate gene at the 10p locus may be GATA-3, since it showed markedly reduced expression in HPV-immortalized keratinocytes, cervical cancer cell lines, a small subset of CIN 3 lesions, and the majority of cervical carcinomas 59. Although further studies are necessary to establish the responsible genes at these various chromosomal locations, the data collected provide sufficient ingredients to support their role in cervical carcinogenesis.

Anchorage-independent growth and tumourigenicity

Loss of tumour suppressor gene(s) at chromosome 11 has been implicated in the progression from an immortal to a tumourigenic phenotype, since introduction of chromosome 11 into SiHa cells abrogated their capacity to form tumours in nude mice, without affecting immortality 60. The significance of this finding was supported by allelotyping studies, showing frequent deletions at 11q loci, particularly 11q22-23, in cervical carcinomas 61. Recently, we collected functional evidence that the TSLC1 gene (also named IGSF4 or NECL-2) may be the candidate suppressor of tumourigenicity on chromosome 11 62. In earlier studies, this gene had already been shown to suppress tumourigenicity of lung cancer cell lines 63. The TSLC1 gene, which resides at 11q23, encodes an immunoglobulin (Ig)-like cell surface protein that belongs to the Nectin and Nectin-like (Necl) family of molecules 64. TSLC1 is involved in cell-cell adhesion via homotypic contacts and heterotypic interactions with other Nectins and Necls. Absence of TSLC1 in epithelial cells contributes to loss of cell polarity and cell-cell adhesion, leading to neoplastic transformation and metastasis. In addition, TSLC1 has been demonstrated to interact with CRTAM (class I-restricted T-cell-associated molecule), a receptor only expressed on activated NK cells and CD8+T-cells, thereby promoting immune responses. Thus, silencing of TSLC1 also results in ineffective immune surveillance against (pre)neoplastic cells 65.

TSLC1 was found to be silenced in 91% (10/11) of cervical cancer cell lines, mostly as a result of promoter hypermethylation alone or combined with allelic loss 62. TSLC1 promoter hypermethylation was also detected in 58% of cervical carcinomas and 35% of high-grade CIN lesions, but not in low-grade CIN lesions and normal cervix. Moreover, ectopic expression of TSLC1 suppressed both tumour formation in nude mice and anchorage-independent growth of SiHa cells. The latter suggests that, albeit apparently necessary, inactivation of TSLC1 is not sufficient for tumourigenicity but rather predisposes cells with anchorage-independent growth properties to become tumourigenic as a result of an additional hit.

From that perspective, it is interesting that from a model system of non-tumourigenic hybrids of the HPV 18-containing cervical cancer cell line HeLa with fibroblasts and their tumourigenic segregants, a change in the composition of the AP-1 complex emerged as being relevant for tumourigenicity 66–68. AP-1 is a transcription factor consisting of different proteins (eg c-Jun, c-Fos or Fra-1) in homo- or hetero-dimer complexes. It is involved in multiple regulatory pathways such as proliferation and differentiation, and can influence the sensitivity against growth-inhibitory cytokines and chemokine production of HPV-containing cells in nude mice 66, 68. Whereas in non-tumourigenic cells AP-1 mainly consists of Jun family members complexed with Fra-1, tumourigenic cells revealed a reduction in Fra-1 but increased amounts of c-Fos, changing the Jun/Fra-1 ratio in favour of Jun/c-Fos 66, 67. The constitutive expression of c-fos in cervical cancer cells has recently been shown to result from the loss of Net, a protein negatively regulating the SRE motif of the c-fos promoter 69. Interestingly, ectopic expression of c-fos in non-tumourigenic HeLa–fibroblast hybrids resulted in a tumourigenic phenotype accompanied by a change in the AP-1 complex 66. Together with the frequent overexpression of c-Fos in cervical carcinomas 70, this argues for a prominent role of an altered AP-1 composition in cervical cancer development. To what extent alterations within the Notch-1 signalling pathway may be involved in the modification of AP-1 is still unclear 71, 72.

Merging in vitro with clinical data: concept of cervical carcinogenesis

After merging all the in vitro data with the data obtained from clinical samples, a model of cervical carcinogenesis can be composed as depicted in Figure 5. A likely crucial decision-maker in the early stages following infection involves individual susceptibility for certain HPV types depending on the genetic make-up of immune surveillance determinants (see above). Once a CIN lesion has developed, altered transcriptional regulation of the viral E6/E7 oncogenes probably provides the subsequent important step towards malignancy.

Figure 5.

Progression model of cervical cancer based on in vitro transformation steps and data from clinical samples. See text for further description. Potential relevant genetic alterations are indicated in red. TSGs = tumour suppressor genes. ↑ indicates increased activity resulting from (epi)genetic alteration(s). ↓ indicates decreased activity resulting from (epi)genetic alteration(s), such as deletion or promoter hypermethylation

Although not discussed so far, down-regulation of MHC-1 expression is common in cervical carcinomas and increases in proportion to increasing severity of CIN lesions 73, 74. There is evidence that the frequently observed allelic loss at the 6p21.3 locus, harbouring the HLA class I antigen genes, alone or in combination with somatic mutations, underlies MHC-1 down-regulation in most cases 75–77. It is tempting to speculate that the change in localization of viral E6/E7 activity from ‘immune-privileged’ differentiated compartments in the squamous epithelium to the (para)basal layers provides an extra trigger to the cellular immune system, which needs to be counteracted by MHC-1 down-regulation to ensure further viral persistence.

Additional (epi)genetic alterations that subsequently accumulate in the CIN 3 stage would in turn lead to overt malignancy via immortality and growth conditions that gradually become less sensitive to growth-modulating influences mediated by cytokines and cell–cell and cell–matrix adhesions.

Clinical implications

Given the prerequisite of an hrHPV infection for the pathogenesis of cervical cancer, it is not surprising that many studies have focused on the value of hrHPV DNA testing in cervical screening and clinical policies to overcome the drawbacks of current cytological examination of cervical smears in terms of false negativity and false positivity. Indeed, data are accumulating indicating that addition of hrHPV DNA testing to cytology can improve the efficacy of cervical screening and triage programmes for the following reasons:

  • (1)Adding hrHPV testing to cytology improves the negative predictive value from 96–97% to more than 99% 78. Of the women with normal cytology, those with a positive hrHPV test are at risk of ≥CIN 3 and therefore need additional surveillance, eg yearly follow-up. On the contrary, the vast majority of women in a screened population, ie those with normal cytology and a negative hrHPV test, may be screened at much longer intervals (ie 8–10 years), because their risk of ≥CIN 3 is negligible 13, 79. Preliminary cost calculations show that this approach is cost-effective, decreasing the costs by about one-third (Berkhof et al, unpublished data).
  • (2)Women with equivocal or mildly abnormal Pap smears can be referred directly on the basis of the presence of hrHPV because only those who are hrHPV-positive are at risk of ≥CIN3 80–82. This would markedly reduce the number of redundant repeat smears and referrals to the gynaecologist and is cost-effective 83.
  • (3)HPV testing is more effective than cytology in detecting cervical adenocarcinomas and their precursors. Many studies have shown that adenocarcinoma in situ (ACIS) and adenocarcinoma (AdCa) of the cervix are frequently missed by conventional cervical cytology. Reasons for missing AdCa and its precursors might be their location higher in the endocervical canal than squamous lesions, with consequently less accessibility to the brush 84, or failure of cytologists to recognize these lesions 85. The failure of cervical cytology to detect cervical AdCa and its precursors efficiently is reflected by changes in the ratio of incidences of AdCa compared with squamous cell carcinoma in screened populations. Whereas the incidence of squamous cell cancer has declined, that of AdCa has either remained the same or has even increased in some countries despite well-organized screening programmes 86, 87. Adding hrHPV testing to cervical cancer screening might thus improve the detection rate of ACIS and AdCa 88, 89.

The value of testing for hrHPV types, as a pool, alone or as an adjunct to cervical cytology is currently being tested in several large follow-up studies, including trials in the UK (TOMBOLA study 90 and HART study 91) and the Netherlands (POBASCAM trial) 92.

Finally, hrHPV testing can be used to monitor women for post-treatment CIN 3. Recent studies have shown that hrHPV testing in conjunction with cytology is more effective in detecting recurrent/residual CIN 3 after ablative therapy, the consequence being less redundant visits to the gynaecologist compared with monitoring by cytology solely 93–95.

Additional opportunities for hrHPV DNA testing

A problem with current population-based cervical screening programmes is that the compliance rate is subject to improvement. Annually, in The Netherlands, 28% of the invited women do not participate in the cervical screening programme and, as is the case in the UK and USA 96, 97, half of the cervical cancers are diagnosed in this group of women 98. A user-friendly self-sampling method for collecting representative cervical cell material at home may lower the threshold for women to participate in the screening, as we found in a study on more than 2500 women who even after a recall did not respond to the invitation for screening (Bais et al, unpublished observations). More than one-third of these women responded actively by sending vaginal material, collected via a self-sampler to the test laboratory. Upon hrHPV testing, 24% of the hrHPV-positive women appeared to have prevalent lesions ≥CIN 2. Self-sampled vaginal material is not representative of the cytological status of the cervix but is highly representative of cervical hrHPV DNA status 99, 100 (Brink et al, unpublished observations). Thus, in combination with hrHPV DNA testing, self-sampling may increase the coverage of the population-based cervical screening programme since a substantial number of women who do not participate in the regular screening programme might participate if an hrHPV self-sampling kit were to be offered. Indeed, if one-third of non-responder women were to re-enter the screening programme by offering self-sampling, this might yield a reduction of 1/6 in the incidence of cervical cancer (ie one-third of the 50% of cervical cancers that are diagnosed in non-responder women).

Further challenges

Despite the promising performance of the hrHPV DNA test, the positive predictive value of this test for ≥CIN 3, however, calls for improvement. As an example, the population-based cervical screening programme in The Netherlands includes approximately 1% of women with ≥CIN 3, and 6–7.5% of women who are hrHPV DNA-positive by the consensus primer GP5+/6+ PCR assay. Besides the 1–1.3% women with severe cervical lesions, the latter group mainly includes women with equivocal/mildly abnormal smears (1–1.5% of screened women) and normal cytology (about 4–5% of the screening population). Of the hrHPV-positive women with normal cytology, only 8% will have or acquire >CIN 3 101. Thus, it is impossible to give an individual assessment of the risk of having or developing ≥CIN 3 for an hrHPV-positive woman with unknown cytology. This emphasizes the need for additive markers to reduce the number of redundant follow-up smears and visits to the gynaecologist of hrHPV-positive women who will not have or develop ≥CIN 3. Our current insight into cervical carcinogenesis has now yielded several novel markers that potentially may have additive value for risk assessment of cervical cancer. Some of these are listed below.

Viral load assessment

Given the fact that an increased viral load is associated with an increased risk of a clinically relevant cervical lesion, the positive predictive value of HPV testing for ≥CIN 3 could be improved by viral load assessment in cervical scrapings. To that end, we have initiated viral load analysis for six hrHPV types by type-specific real-time PCR in a cohort of 44 102 women participating in the POBASCAM study 92. In a first analysis, we used baseline cervical scrapings of women with normal cytology who were positive for HPV 16, 18, 31 or 33 DNA by consensus primer GP5+/6+ PCR as a reference to define various viral load cut-off points (25th, 33rd, and 50th percentiles). These cut-off points were subsequently used to assess the discriminative power of viral load assessment for CIN 3 using a large series of abnormal cervical scrapes of women with underlying CIN lesions containing one of these four hrHPV types (Hogewoning et al, unpublished observations). Although women with CIN 3 generally displayed increased viral loads, there exists a substantial overlap between viral DNA load levels in women with and without CIN 3. However, preliminary data indicate that all women with CIN 3 had viral load levels above the 25th percentile threshold of women with normal cytology containing the respective hrHPV type (sensitivity 100%). Therefore, from these data, we infer that viral load analysis will allow us to distinguish at most 25% of women without clinically relevant hrHPV infections that are detectable by GP5+/6+ PCR (Hogewoning et al, unpublished observations).

Viral typing

There is now evidence that the different hrHPV types pose different risks for cervical cancer. As a consequence, it has been proposed that not all HPV types classified as high risk should be included in hrHPV detection systems, since the beneficial effect of only a slightly increased sensitivity for high-grade cervical premalignant lesions would be out-weighed by the adverse effect of a decreased specificity 102, 103. Moreover, one may consider scenarios in which hrHPV typing may aid in the stratification of hrHPV-positive women who, depending on the HPV type present, are at greatest risk of cervical cancer. A recent meta-analysis performed by the IARC on cross-sectional data comparing HSIL with cervical SCC indicates that the risk of cervical carcinoma is, in decreasing order of magnitude, increased in the presence of an infection with HPV 16, 18, and 45 104. Interestingly, cross-sectional population-based data from the POBASCAM study revealed that the risk of CIN 3 is increased for HPV 16 and 33, but not for HPV 18 and 45 105. Combined, these data imply that during different morphological phases of the oncogenic process, different HPV types have different oncogenic potentials. Thus, HPV 33 might have a lower oncogenic potential in terms of progression from CIN 3 to cervical carcinoma, compared with HPV 18 and 45. On the contrary, the risk posed by HPV 33 to induce CIN 3 seems higher than that of HPV 18 and 45. HPV 16 exerts the highest risk for both CIN 3 and cervical carcinoma. Studying a prospective cohort of 20 810 women for up to 10 years, Kahn et al106 showed a particularly excessive risk posed by HPV 16 and 18 for cervical (pre)cancers compared with other oncogenic HPV types. In addition, Castle et al107 showed for HPV 16 an increased 2-year absolute risk for cervical pre-cancer in women with equivocal or mild cervical abnormalities compared with women with other hrHPV types. These data indicate that HPV typing at least for HPV 16 and 18 may identify women who would benefit from more aggressive management than women with other HPV types. Long-term ongoing follow-up studies are presently underway that may further strengthen the conclusions based on these cross-sectional and follow-up data, and allow HPV testing algorithms to be adapted accordingly.

Factors related to E6/E7 expression and viral integration

Given the importance of hrHPV E6/E7 expression, its assessment may also provide a valuable marker for clinically relevant HPV infections. The recent introduction of preservation media for the collection of cervical scrapes for liquid-based cytology purposes has opened possibilities for HPV transcript analysis, since these media sufficiently preserve RNA for amplification purposes 108. Interestingly, data from some small studies indeed suggest that hrHPV mRNA testing may distinguish clinically relevant from irrelevant hrHPV infections, and there are indications that lack of E6/E7 mRNA in HPV DNA-positive cytologically normal cervical scrapes predicts future loss of HPV DNA 109–111. However, these data are based on small studies and larger studies with long follow-up need to be performed to give an impression about the possible value of E6/E7 mRNA analysis to stratify hrHPV DNA-positive women for risk of ≥CIN 3.

Since p16INK4A expression is regulated by an Rb-dependent negative feed-back loop, continuous inactivation of Rb by hrHPV E7 results in increased p16INK4A levels, as can be detected immunohistochemically in hrHPV-containing CIN lesions and cervical carcinomas 112. Hence, the detection of increased p16INK4A expression in liquid-based cytology specimens may provide a promising marker for the detection of HPV-induced dysplasia with deregulated E7 expression 113. However, also in this case, large follow-up studies involving p16INK4A immunostaining on liquid-based cytology samples should be awaited before firm conclusions are drawn about the value of this method.

Finally, the detection of viral integration via specific amplification of chimeric virus–host cell-derived transcripts may emerge as a potentially valuable marker for risk assessment 114.

Markers involving host cell factors

Application of host cell markers that reflect later stages in the carcinogenic process might be particularly valuable for identifying hrHPV-positive women who need immediate referral for colposcopy. Although increased activity of telomerase seems relevant for most cervical carcinomas, the detection of telomerase activity in fresh cervical scrapes as well as the detection of hTERT by immunohistochemistry in thin-layer cervical cytology preparations did not reflect the status of these markers in the underlying lesions 115–117. Admixed mononuclear cells that have some regulated telomerase activity in the cervical swab can in part explain these results. In conclusion, when applied to cervical scrapings, these markers are currently not sensitive or specific enough to predict cervical cancer.

Interestingly, TSLC1 promoter hypermethylation could be detected in archival cervical smears from women with cervical cancer taken up to 7 years before cancer diagnosis 62. This indicates that testing for TSLC1 silencing in cervical smears may become a powerful tool for identifying women who have CIN lesions with invasive potential.

Altogether, the data collected during recent decades have resulted in a comprehensive cervical cancer progression model from which hrHPV DNA tests and other potential markers for cervical screening can be deduced. HrHPV DNA testing has so far been analysed most extensively for cervical screening and owing to its high sensitivity and very high negative predictive value for ≥CIN 3, it is currently the most promising candidate to be implemented in cervical screening, triage of women with equivocal smears, or monitoring women for post-treatment CIN 3 either alone or in conjunction with cytology. The somewhat lower specificity of hrHPV DNA testing for ≥CIN 3 can probably be compensated for by performing viral load and/or E6/E7 mRNA testing and typing for at least HPV 16 and HPV 18. In addition, a series of other candidate markers, for example p16INK4a immunostaining and TSLC1 methylation analysis, may, alone or in combination, further identify high-grade CIN lesions that have a high chance of progression to invasive cervical cancer. In addition, the progression model can form the basis for the discovery of novel targets for novel (immune) therapeutic intervention modalities. Together with prophylactic vaccination efforts, these should result in a marked reduction in the incidence of cervical cancer.

Acknowledgements

RDMS was supported by a fellowship of The Netherlands Academy of Arts and Sciences.

Ancillary