(a) Mucosal HPV

A subset of mucosotropic HPVs that belong to the alpha genus, including the high-risk HPV types 16 and 18, are associated with more than 99% of cervical carcinomas (Walboomers et al., 1999). In these cancers, the papillomaviral DNA genome is often found integrated into the host chromosome (Boshart et al., 1984; Schwarz et al., 1985; Yee et al., 1985). Cervical epithelial cells that harbour integrated HPV 16 DNA have a selective growth advantage over cells that harbour normal extrachromosomal viral genomes; this growth advantage correlates with the increased expression of two viral genes in particular, E6 and E7 (Jeon et al., 1995). The early proteins, E6 and E7, bind and inactivate the tumour-suppressor gene product, p53, and the retinoblastoma tumour-suppressor protein (pRb), respectively (Dyson et al., 1989; Münger et al., 1989a; Werness et al., 1990). In cell lines derived from HPV-positive cervical cancers, these genes are not inactivated mutationally, whereas they are mutated in cell lines derived from HPV-negative cervical cancers (Scheffner et al., 1991). The expression of the E6 and E7 viral genes is required for the continued growth of cell lines derived from cervical cancers (Hwang et al., 1993; Francis et al., 2000; Goodwin & DiMaio, 2000; Goodwin et al., 2000; Nishimura, A. et al., 2000; Wells et al., 2000). These facts support the hypothesis that E6 and E7 are causally related to the onset and maintenance of human cervical cancers. In addition, continuous expression of these early proteins can lead to the accumulation of mutations in the cellular genome that are required for malignant conversion (reviewed in zur Hausen, 1999). Both E6 and E7 co-operate to induce transformation of epithelial cells (Münger et al., 1989b); however, a fully malignant phenotype is only observed after prolonged cultivation of the transformed cells (Hurlin et al., 1991; Dürst et al., 1995), which supports the multistep nature of HPV-induced transformation. In a transgenic mouse model, the expression of HPV 16 E6 alone has been shown to be sufficient to induce carcinomas (Song et al., 1999). In contrast, E6 and E7 of the low-risk mucosal types have very low transforming activities in vitro (Farr et al., 1991; Sang & Barbosa, 1992). Exceptions include immortalization by the low-risk type HPV 6 of human mammary epithelial cells which are not the natural host cells of these viruses (Band et al., 1993). A recent report demonstrated the presence of integrated genomes of the low-risk type HPV 11, but not those of HPV 6, in cancers that developed in patients with early-onset recurrent respiratory papillomatosis (Reidy et al., 2004). This points to the relevance of integration of the viral genome in malignant transformation, as discussed in Section 4.1.4.

(b) Cutaneous HPVs

Epidemiological studies have clearly demonstrated that a subset of cutaneous HPV types classified into the beta genus (approximately 25 types, also called epidermodysplasia verruciformis (EV)-HPV types, have been sequenced so far) are commonly and consistently found in non-melanoma skin cancers. These skin tumours arise predominantly at sites exposed to the sun and, contrary to mucosal types, the EV-HPV DNA copy number appears to be much lower than one copy/cell (de Villiers, 1998; Iftner et al., 2003). Cutaneous HPVs have been studied less extensively than mucosal types and their capacity for cell transformation and molecular mechanisms are still largely unknown. Most cutaneous HPV types express E6 and E7 gene products that are structurally similar to those of the mucosal types but their genome does not harbour an identifiable E5 open-reading frame (ORF) (Pfister, 2003).

Schmitt et al. (1994) performed a comparative analysis of various properties of the E6 and E7 proteins of EV-associated type 8 and non-EV-associated type 1 cutaneous HPVs by transfecting the genes into different cell lines. HPV 8 E6, HPV 16 E6 and E7 and HPV 1 E7 but not cottontail rabbit papillomavirus (CRPV) long E6 or HPV 8 E7 were able to transform immortalized mouse fibroblasts (C127 cell line) while cells that expressed HPV 1 E6 or CRPV short E6 exhibited a weak transformed phenotype. The in-vitro retinoblastoma protein (Rb)-binding affinity (relative to that of HPV 16 E7) was 66% for HPV 1 E7, 34% for HPV 8 E7 and 11% for CRPV E7. None of the E6 or E7 proteins of the cutaneous HPV types 1 or 8 or CRPV revealed true immortalizing activities in primary human keratinocytes. In these cells, only a weak induction of proliferation was observed with HPV 8 E7, and only HPV 8 E7 transformed primary rodent cells co-transfected with the EJ-Ras oncogene.

HPV 5 and 8 E7s were shown to form complexes with the Rb protein, but with lower affinities than that of HPV 16 E7 (Yamashita et al., 1993). Ciccolini et al. (1994) found that HPV 1 E7 binds to pRb with an affinity similar to that of high-risk E7 proteins but has no transforming activity in primary rodent cells. HPV 8 E6 protein expressed in vitro was shown not to bind murine p53 (Steger & Pfister, 1992). Similarly, HPV 1 and 8 E6 proteins bound to neither human p53 nor E6-associated protein (E6-AP) (Elbel et al., 1997). Furthermore, HPV 8 E2, E4 and E6 were shown to interact with the TATA box-binding protein (TBP) and a number of TBP-associated factors (Enzenauer et al., 1998).

Caldeira et al. (2003) analysed the in-vitro properties of E7 proteins of cutaneous EV-(HPV 20 and 38) and non-EV- (HPV 10) HPV types that are frequently detected in the skin. It was shown that HPV 38 E7 binds to and inactivates the tumour suppressor pRb and induces loss of G1/S transition control. In contrast, HPV 10 and HPV 20 E7 proteins do not display in-vitro transforming activities. Moreover, E6 and E7 of HPV 38 were shown to immortalize primary human keratinocytes, which suggests a role of HPV 38 infection in skin carcinogenesis.

(a) E5

The study of the HPV E5 protein and recognition of its tumorigenic potential arose from the analysis of the transforming potential of BPV 1 in mouse C127 cells (see Section 3.3) (Yang et al., 1985; DiMaio et al., 1986; Schiller et al., 1986; Schlegel et al., 1986; Settleman et al., 1989). The E5 ORF and the hydrophobic nature of its gene product are conserved in many papillomaviruses, although the degree of conservation of the primary amino acid sequence is variable (Bubb et al., 1988). High-risk HPV E5 is considered to be tumorigenic because it transforms murine fibroblasts and keratinocytes in tissue culture (Leechanachai et al., 1992; Pim et al., 1992; Straight et al., 1993), enhances the immortalization potential of E6 and E7 (Stöppler et al., 1996) and, in cooperation with E7, stimulates the proliferation of human and mouse primary cells (Bouvard et al., 1994b; Valle & Banks, 1995). Two major biochemical activities have been attributed to HPV 16 E5 in vitro: the ability to enhance the activity of the epidermal growth factor receptor (EGFR) in the presence of ligand (Leechanachai et al., 1992; Pim et al., 1992; Straight et al., 1993) and the ability to bind and inactivate the 16-kDa pore-forming membrane component of the vacuolar H⁺ adenosine triphosphatase (v-ATPase) (Conrad et al., 1993; Adam et al., 2000; Briggs et al., 2001). Multiple studies have suggested that EGFR, a 170-kDa tyrosine kinase receptor, mediates the biological activities of HPV 16 E5 protein, and purport that E5 activates EGFR and induces mitogenic signalling and transformation of cells via this receptor (Straight et al., 1993; Crusius et al., 1998; Tomakidi et al., 2000). Co-immunoprecipitation studies have indicated that HPV 16 E5 can form a complex with growth factor receptors when both proteins are overexpressed (Hwang et al., 1995), but this binding has not always been observed (Conrad et al., 1994). The binding of HPV 16 E5 to the 16-kDa subunit of the v-ATPase (Conrad et al., 1993) is thought to delay endosomal acidification in human keratinocytes (Straight et al., 1995), which has been implicated in the enhancement of EGFR phosphorylation in keratinocytes, since failure to acidify endosomes may result in decreased receptor degradation and increased receptor recycling to the cell surface (Straight et al., 1993, 1995).

(b) E6

Mucosal high-risk E6 proteins are best known for their ability to associate with the cellular tumour suppressor p53 (Werness et al., 1990). Association of E6 with p53 leads to degradation of p53 via recruitment of an ubiquitin ligase, E6-AP ( Scheffner et al., 1990; Huibregtse et al., 1991; Scheffner et al., 1993), and results in the inhibition of the transcriptional regulatory activities of the p53 protein in tissue culture cells (Lechner et al., 1992; Mietz et al., 1992). E6 proteins from multiple human and animal papillomaviruses bind to cellular proteins other than p53 and E6-AP. These include (a) transcription factors such as p300 (Patel et al., 1999; Zimmermann et al., 1999), myc (Gross-Mesilaty et al., 1998), interferon regulatory factor 3 (IRF3) (Ronco et al., 1998) and autocrine motility factor 1 (AMF-1/Gps2) (Degenhardt & Silverstein, 2001); (b) factors that determine adhesion, cytoskeleton and polarity, such as paxillin (Tong & Howley, 1997; Tong et al., 1997; Vande Pol et al., 1998), the mammalian homologue of Drosophila disk–large tumour-suppressor gene product (DLG) (Kiyono et al., 1997; Lee et al., 1997), Scribble (Nakagawa & Huibregtse, 2000), membrane-associated guanylate inverted-1 (MAGI-1) (Glaunsinger et al., 2000) and multiple PDZ protein 1 (MUPP1) (Lee et al., 2000); (c) apoptosis factors such as the pro-apoptic Bcl2 protein, Bak (Thomas & Banks, 1998, 1999); (d) replication factors and DNA repair factors such as mcm7 (Kühne & Banks, 1998; Kukimoto et al., 1998) and XRCC1 (Iftner et al., 2002); and (e) other proteins such as E6 target protein 1 (E6TP1) (Gao et al., 1999), E6 binding protein 1 (E6BP1) (Chen et al., 1995) and protein kinase PKN (Gao et al., 2000). In addition, E6 can induce telomerase activity by inducing the expression of human telomerase reverse transcriptase (hTERT) (Klingelhutz et al., 1996; Gewin & Galloway, 2001; Oh et al., 2001; Veldman et al., 2001). A more complete compendium of factors that interact with E6 is available (Mantovani & Banks, 2001).

Several E6 targets (paxillin, IRF3, E6BP1, E6TP1 and E6-AP) share a common α-helical structural motif that is known or suspected to mediate their binding to E6 (Tong et al., 1997; Chen et al., 1998; Vande Pol et al., 1998; Be et al., 2001). Mutations in E6 that disrupt its ability to bind α-helical partners also cause defects in multiple biological properties (Liu et al., 1999). One of these partners, E6-AP, has attracted considerable attention because it is believed to be responsible for mediating high-risk HPV E6-dependent destabilization of multiple targets including p53 (Scheffner et al., 1993), myc (Gross-Mesilaty et al., 1998) and the PDZ domain protein, Scribble (Nakagawa & Huibregtse, 2000). This ubiquitin ligase, however, probably does not account completely for the ability of E6 to target cellular factors for proteasome-mediated degradation. Among the cellular partners of high-risk HPV E6 that are destabilized independently of E6-AP are human DLG and members of the MAGI family (Grm & Banks, 2004). It is also important to note that cutaneous HPV E6s, including the EV-associated HPV 8 E6, do not bind E6AP (Elbel et al., 1997). Another putative α-helical partner, E6TP1, a Rap guanosine triphosphatase-activating protein, has been implicated in the capacity of E6 to immortalize mammary epithelial cells (Gao et al., 2001). It has been demonstrated that, for bovine papillomavirus (BPV) 1 E6, the selective binding of E6, but not of E6-AP, to paxillin mediates its transforming potential (Vande Pol et al., 1998); however, it remains unclear whether an interaction with paxillin is of biological importance for the transforming potential of HPV E6 protein.

The interaction of high-risk E6 proteins with proteins that contain the PDZ domain, such as DLG, Scribble, MAGI1–3 and MUPP-1, has drawn a lot of interest. E6 can target PDZ-domain proteins for degradation (Gardiol et al., 1999; Nakagawa & Huibregtse, 2000; Massimi et al., 2004). The intracellular location and levels of DLG are altered in cervical cancers and their precursor lesions (Watson et al., 2002; Cavatorta et al., 2004; Lin et al., 2004). Disruption of the C-terminal domain in E6 that mediates its interaction with PDZ-domain proteins leads to defects in the transforming potential of E6 in certain tissue culture-based assays (Kiyono et al., 1997) and its tumorigenic potential in mice (Nguyen et al., 2003), but not its capacity for immortalization.

(c) E7

High-risk HPV E7 proteins are best known for their ability to associate with the cellular tumour suppressor, pRb (Dyson et al., 1989; Münger et al., 1989a; Gage et al., 1990). Association of high-risk E7 with pRb also promotes the degradation of pRb (Boyer et al., 1996; Jones et al., 1997a) through a proteasome-mediated pathway (Berezutskaya & Bagchi, 1997; Gonzalez et al., 2001) and disrupts the capacity of pRb to bind and inactivate functionally cellular E2F transcription factors (Phelps et al., 1991; Chellappan et al., 1992). In addition to binding pRb, high-risk E7 proteins can bind to other pocket proteins (p107 and p130) that are related to pRb (Dyson et al., 1992; Davies et al., 1993) and also interact with different members of the E2F family of transcription factors (Dyson et al., 1993; Classon et al., 2000). The inactivation of pocket proteins by E7 is necessary but not sufficient to elicit the transforming potential of E7 (Heck et al., 1992; Phelps et al., 1992; Kiyono et al., 1998). High-risk E7 is also purported to complex with cyclins (Dyson et al., 1992; Arroyo et al., 1993; Tommasino et al., 1993; McIntyre et al., 1996) and to inactivate the cyclinassociated kinase inhibitors p21 and p27 (Funk et al., 1997; Jones et al., 1997b). Thus, E7 can associate with and/or alter the activities of multiple cellular factors that normally contribute to the regulation of the cell cycle. Other interactions have been identified between high-risk E7 and cellular factors including the S4 subunit of the 26 S proteasome (Berezutskaya & Bagchi, 1997), Mi2beta, a component of the nuclease remodelling and deacetylase (NURD) histone complex (Brehm et al., 1999), the fork head domain transcription factor MPP2 (Lüscher-Firzlaff et al., 1999), the transcription factor activator protein-1 (AP-1) (Antinore et al., 1996), insulin-like growth factor binding protein 3 (Mannhardt et al., 2000), TBP (Massimi et al., 1996, 1997; Phillips & Vousden, 1997), TBP-associated factor110 (Mazzarelli et al., 1995) and a novel human DnaJ protein, hTid-1 (Schilling et al., 1998).

(a) Immortalization

The E6 and E7 proteins of mucosal high-risk HPVs have transforming activity in tissue culture. They act independently or synergistically to immortalize multiple cell types including human foreskin keratinocytes, cervical epithelial or mammary epithelial cells (Dürst et al., 1987a; Pirisi et al., 1987, 1988; Hawley-Nelson et al., 1989; Kaur et al., 1989; Band et al., 1990; Hudson et al., 1990; Halbert et al., 1991; Wazer et al., 1995). To date, the E6 and E7 proteins of only one cutaneous EV-associated HPV type, HPV 38, have shown to immortalize human primary keratinocytes (Caldeira et al., 2003). Mucosal high-risk (HPV 16, 18, 31) E7s but not low-risk (HPV 6, 11) E7s cooperate with an activated ras to transform neonatal rat kidney or human cervical epithelial cells (Matlashewski et al., 1987; Crook et al., 1988; Phelps et al., 1988; Storey et al., 1988). The contribution of HPV 16 E7 to immortalization correlates with its disruption of the p16/pRb pathway (Kiyono et al., 1998; Jarrard et al., 1999). While E7 appears to be critical for efficient immortalization of multiple human epithelial cell types (Hawley-Nelson et al., 1989; Halbert et al., 1991), it may be dispensible depending on the method by which the cells are cultured (Ramirez et al., 2001). Low-risk mucosal HPV 6 E7 and cutaneous EV-HPV 8 E7 demonstrate weak immortalizing activity compared with that of high-risk HPV 16 E7 (Halbert et al., 1992; Schmitt et al., 1994). This difference potentially correlates with the relative capacity of these E7 proteins to induce the degradation of pRb rather than their affinity for pRb (Giarrè et al., 2001). The E5 protein can enhance the immortalization of keratinocytes effected by the combination of E6 and E7 (Stöppler et al., 1996). The mechanism by which E6 contributes to immortalization is controversial. Some studies have correlated its potential for immortalization with its ability to induce the expression of telomerase (Kiyono et al., 1998), since telomerase activity is clearly induced by E6 (Klingelhutz et al., 1996). Whereas E6 modulates the transcription of hTERT through a direct stimulation of mycmediated transactivation of the hTERT promoter (Veldman et al., 2003), another study argued that induction of myc was not found (Gewin & Galloway, 2001). An alternative hypothesis that was recently put forward is that E6 relieves the repression of the telomerase promoter by inducing degradation of the transcriptional repressor, NFX1-91 (Gewin et al., 2004). The level of telomerase activity in E6-positive cells increases further when they become immortalized although levels of E6 expression do not change (Fu et al., 2003), which indicates that other events contribute to telomerase activation. Other studies have linked the immortalization potential of E6 in mammary epithelial cells and keratinocytes to its inactivation of p53 (Dalal et al., 1996; McMurray & McCance, 2004).

(b) Genomic instability

Another hallmark of cells that express E6 and E7 is genomic instability, which has been observed in multiple epithelial cell types (Smith et al., 1989; Hashida & Yasumoto, 1991; Reznikoff et al., 1994; White et al., 1994; Coursen et al., 1997; Steenbergen et al., 1998; Duensing & Münger, 2002; Shen et al., 2002b). Abnormalities included monosomies and trisomies, chromatid gaps and breaks, double minutes and aberrant chromosomes. Structural changes are more commonly detected in chromosomes 1, 3 and 5 and less frequently in chromosomes 7, 8, 10, 12, 13, 16 and 22. Some of these allelic losses have been associated with particular genes that could be involved in malignant conversion and/or progression. Among these, losses in 3p and 10p have been associated with telomerase activation (Steenbergen et al., 1998), which is a crucial step for cell immortalization mediated by high-risk HPVs (Klingelhutz et al., 1996; Coursen et al., 1997). Mitotic abnormalities can be induced by high-risk HPV 16 (but not low-risk HPV 6) E6 and E7 proteins by direct subversion of the mitotic spindle checkpoint (Thomas & Laimins, 1998; Duensing et al., 2000).

The ability of E6 to induce genomic instability probably reflects its ability to inhibit the function of p53 (Havre et al., 1995), which leads to the disruption of normal DNA repair processes and a consequent accumulation of genetic change. The genomic instability induced by E7 may reflect its effect on centrosome biogenesis and the consequent defects in segregation of daughter chromosomes during cell division (Duensing et al., 2000; Duensing & Münger, 2001; Duensing et al., 2001a,b). However, the manner in which E7 induces genomic instability remains unclear. While studies in mice have indicated that inactivation of pRb is sufficient to induce centrosome abnormalities (Balsitis et al., 2003), other studies have demonstrated that E7 can induce centrosome abnormalities through a pRb-independent mechanism (Duensing & Münger, 2003).

Using an HPV 16-positive cell line (W12) derived from a low-grade squamous intraepithelial lesion (LSIL), Pett et al. (2004) recently suggested that acquisition of chromosomal instability is correlated with integration of the viral genome. The contrary has been shown in raft cultures of human keratinocytes that contain episomal HPV 16 in which genomic instability was observed in the absence of viral integration (Duensing et al., 2001b). Progress in this area has been hampered by the lack of experimental models and methods to determine HPV integration that are suitable for use on large series in human biological specimens.

(c) DNA damage responses

Both E6 and E7 can abrogate normal DNA damage responses (Kessis et al., 1993; Slebos et al., 1994; Demers et al., 1996). This is thought to reflect the activity of both E6 and E7 in inhibiting p53-mediated cell-cycle arrest. This correlates at least in part with the ability of E6 to bind and inactivate p53 (Song et al., 1998); for E7, this correlates not only with its ability to disrupt the function of the cell-cycle regulator pRb, but also with its ability to inactivate p21 (Funk et al., 1997; Jones et al., 1997b), the cyclin-dependent kinase inhibitor that is induced when p53 is activated in response to DNA damage (Helt et al., 2002). Abrogation of the DNA-damage responses is hypothesized to contribute to the accumulation of genetic alterations in HPV-positive cells, which include those that might contribute to tumorigenicity.

(d) Cell proliferation and differentiation

Whereas suprabasal cells are withdrawn from the cell cycle in normal stratified epithelia, HPV can re-programme suprabasal cells to sustain DNA synthesis which may contribute to the production of progeny virus through amplification of the viral genome. E7 is necessary (Flores et al., 2000) and sufficient (Cheng et al., 1995; Herber et al., 1996) for this phenotype. It remains uncertain whether the ability of E7 to induce DNA synthesis in differentiated cells reflects the failure of cells that express E7 to withdraw from the cell cycle (Sacco et al., 2003) or an ability of E7 to re-programme the differentiated cell to reenter the cell cycle (Cheng et al., 1995; Chien et al., 2002). Inactivation of pRb alone can cause suprabasal cells to sustain DNA synthesis (Balsitis et al., 2003); however, other properties of E7 may also contribute to this phenotype, notably its recognition by casein kinase as a substrate for phosphorylation (Chien et al., 2000). E6 has also been shown to induce suprabasal DNA synthesis (Song et al., 1999), a p53-independent activity that correlates with the ability of E6 to bind PDZ-domain proteins (Nguyen et al., 2003). A potentially related activity of E6 and E7 is their ability to inhibit keratinocyte differentiation (Schlegel et al., 1988; Barbosa & Schlegel, 1989; Pan & Griep, 1994; Herber et al., 1996; Gulliver et al., 1997; Sherman et al., 1997; Pei et al., 1998; Song et al., 1999). For E7, this probably reflects its inactivation of pRb (Gulliver et al., 1997; Balsitis et al., 2003), but inactivation of p53 by E6 is not sufficient for the inhibition of keratinocyte differentiation (Sherman et al., 1997). It remains unclear whether the re-programming of supra-basal cells to sustain DNA synthesis by E6 and E7 or their inhibition of differentiation contribute to the tumorigenic potential of HPV; however, recent studies indicate that the ability of E6 to bind PDZ-domain proteins, which is required for the induction of both of these acute effects on stratified squamous epithelia, correlates with its contribution to the promotion phase of carcinogenesis (Simonson et al., 2005).

High-risk E6 and E7 proteins modulate apoptosis. E7 can induce apoptosis in mouse lens and retina (Howes et al., 1994; Pan & Griep, 1994; Nakamura, T. et al., 1997). E7 can also sensitize human keratinocytes, mammary epithelial cells and uroepithelial cells to agents that induce apoptosis such as tumour necrosis factor (TNF) (Stöppler et al., 1998; Basile et al., 2001), ionizing radiation (Puthenveettil et al., 1996) and ultraviolet (UV) radiation (Carlson & Ethier, 2000) but it has the opposite effect on TNF-induced apoptosis in human fibroblasts (Thompson et al., 2001) or hydrogen peroxide-induced apoptosis in astrocytes (Lee, W.T. et al., 2001). In mouse lens fibre cells and mouse fibroblasts, apoptosis is mediated through the inactivation of pRb by E7 (Alunni-Fabbroni et al., 2000) and the consequent up-regulation of E2F activity (McCaffrey et al., 1999). In hepatocytes, however, it is hypothesized that E7 mediates apoptosis through inactivation of pRb and induction of p21 (Park et al., 2000a). The apoptosis induced by E7 occurs through p53-dependent as well as p53-independent pathways (Pan & Griep, 1995). E7 induces expression of p53 (Song et al., 1998; Seavey et al., 1999), but this p53 is not fully transcriptionally active (Eichten et al., 2002). Whereas E7 can inhibit p53-induced cell-cycle arrest, it does not inhibit p53-induced apoptosis (Wang, Y. et al., 1996) which indicates that these two pathways are separable, and that E7-induced p53 is capable of triggering apoptosis.

(a) Requirement of HPV gene expression for cell growth and invasion

Expression of HPV E6 and E7 proteins is essential for cellular immortalization, but other factors are required for the acquisition of a fully transformed phenotype. HPV-immortalized cells have been shown to become tumorigenic either spontaneously (Pecoraro et al., 1991) or after treatment with chemical carcinogens (Garrett et al., 1993) or exposure to γirradiation (Dürst et al., 1995). The necessity of expression of viral proteins for transformation to the malignant phenotype has been demonstrated using various cell lines in these experimental models. Using an inducible promoter, it has been shown that continued expression of HPV 16 E7 in the presence of activated Ras is necessary for the maintenance of a transformed phenotype in primary rodent cells (Crook et al., 1989). Nonmalignant revertants were obtained in a model in which transcription of E6 and E7 was impaired (von Knebel Doeberitz et al., 1992, 1994). In another study, stimulation of nontumorigenic HeLa (HPV 18)-fibroblast hybrids to invasive growth was shown to involve loss of TNFα-mediated repression of viral transcription and participation of AP-1 (Soto et al., 1999). Moreover, a key role for the constitutive expression of c-fos in the transformation of cervical cancer cells has been demonstrated (van Riggelen et al., 2005).

Interference with the expression of HPV 16 E6 and E7 was studied in the HPV 16-positive cervical cancer cell line SiHa using E6 short-interfering RNA (siRNA). Yoshinouchi et al. (2003) showed that E6siRNA decreased the levels of mRNA that encode E6 and E7 and induce nuclear accumulation of p53. Moreover, E6siRNA suppressed monolayer and anchorage-independent growth of SiHa cells, which was associated with induction of p21 and hypophosphorylation of pRb.

The contribution of HPV genes to the development of malignancy has also been studied in vivo through the generation and characterization of HPV transgenic mice. These studies are described in Section 4.1.6.

(b) Integration of HPV sequences

In most invasive cancers, high-risk HPV genomes are integrated into the host genome. Integration of HPV can also be found in premalignant lesions, particularly in grade 2/3 cervical intraepithelial neoplasia (CIN2/3). In contrast, HPV DNA is commonly found extrachromosomally in benign and low-grade lesions. Low-risk HPV types are very rarely found integrated in tumours. However, Reidy et al. (2004) studied tissue specimens from patients with a history of benign early-onset recurrent respiratory papillomatosis who developed laryngeal cancer. Integrated HPV 11 was found in these specimens as judged by the absence of full-length E2 transcripts measured by real time-polymerase chain reaction (PCR) in a manner similar to that of high-risk HPVs in cervical cancers. An increased ability of high-risk HPV types to integrate into host DNA compared with low-risk types also has been suggested in vitro (Kessis et al., 1996).

Integration is considered to be an important molecular event in HPV-induced carcinogenesis. Integrated sequences of DNA have been consistently identified in cervical cancers by southern blot hybridization (Dürst et al., 1985; Cullen et al., 1991). A systematic analysis of large series of HPV-infected cells and tissues has been hampered by a lack of less time-consuming and labour-intensive methods to determine integration of HPV. Since integration of HPV often disrupts the E2 gene (Schwarz et al., 1985; Romanczuk & Howley, 1992), determination of a lack of amplification of E2 sequences by PCR-based protocols has been considered; however, results obtained with these methods are ambiguous because of the concomitant presence of non-integrated and integrated molecules in the biological specimens and to technical failures due to presence of inhibitors or other factors. However, detection of early gene transcripts by reverse-transcription PCR is more sensitive both in cancers (Park et al., 1997) and in benign or dysplastic cervical swabs, in which the presence of integrated genomes has been shown to correlate with severity of disease, particularly for HPV 18 (Hudelist et al., 2004). Recently, alternative methods for the accurate determination of the physical status of HPV genomes have been proposed (Klaes et al., 1999; Luft et al., 2001; see Section 1.3.3). In the study by Klaes et al. (1999), transcripts derived from integrated HPV were more frequently detected in high-grade lesions and cervical cancer than in normal or low-grade dysplastic tissues. Integration of HPV 16 and 18 in high-grade lesions is often accompanied by chromosomal abnormalities (Hopman et al., 2004). This supports the potential use of measurements of HPV integration as markers of progression in cervical cancer. In tonsillar cancers, the presence of extra-chromosomal HPV 16 genomes and high viral loads were correlated with better prognosis (Mellin et al., 2002).

Integration of HPV genomes affects both viral and host gene expression. HPV gene expression is regulated by viral and cellular transcriptional activators and repressors (see Section 1.1.6). Normal regulation is altered by viral integration and leads to the continuous expression of E6 and E7 proteins and, consequently, selective growth advantage, as shown by Jeon et al. (1995), who used clonal populations of the W12 cell line that harbour non-integrated or integrated HPV 16 DNA. Integration correlated with increased E7 protein synthesis; cells with integrated viral DNA had growth advantages and phenotypic changes compatible with those of high-grade neoplasia compared with cells that harboured extrachromosomal viral DNA.

This consistent pattern of disruption seen in the viral genome does not seem to occur in the host genome. The occurrence of integrated HPV DNA sequences at preferential sites of human chromosomes has been reported and suggests a non-random pattern of integration. In cervical carcinomas, Ferber et al. (2003a) observed HPV integration into and around the hTERT gene, which resulted in an increase in hTERT expression. Furthermore, HPV 18 DNA was found integrated in the proximity of c-myc in several cervical cancers (Ferber et al., 2003b) but no up-regulation of endogenous proto-oncogene expression was observed. Cytogenetic and molecular studies have shown that HPV 16 and 18 DNA sequences can be found integrated in particular chromosomal loci known as common fragile sites in cervical cancers (Ferber et al., 2003b; Thorland et al., 2003), in HPV 16immortalized keratinocytes (Popescu & diPaolo, 1990) and in an HPV 16-positive cell line derived from a cancer of the tongue (Ragin et al., 2004). One of these regions, the FRA3B common fragile site that encompasses the fragile histidine tetrads (FHIT) tumoursuppressor gene is mapped on chromosome 3p. Butler et al. (2002) showed a clear association between the loss of FHIT expression and progression of HPV 16-positive CIN. Invasive cervical cancers that express high-risk HPV E6 and E7 transcripts were shown to contain normal FHIT transcription, while fewer viral transcripts were detected when FHIT was abnormally expressed, which suggests that E6 and E7 could be repressed in the presence of FHIT aberrations (Segawa et al., 1999).

The studies described above favour the concept that HPV genomes may interfere with critical cellular functions by insertional mutagenesis. However, this has not been confirmed in recent studies that used HPV transcript and genome-based amplification techniques (see Section 1.3.3). A comprehensive analysis of integration sites of HPV 16 and 18 in 21 anogenital cancerous and pre-cancerous lesions revealed only single integration events in which E6 and E7 transcripts could be detected (Ziegert et al., 2003). This could be an indication that the major function of HPV integration is the conservation and stabilization of HPV gene expression. A thorough review of integration sites of HPV in cervical dysplasia and cancer (Wentzensen et al., 2004) concluded that these are randomly distributed over the whole genome with a clear predilection for genomic fragile sites. The relative impact of physical and functional disturbance of viral and cellular genes in HPV-mediated carcinogenesis needs further study. Many observations have demonstrated that the malignant phenotype cannot be attributed exclusively to the expression of HPV genes. It has been hypothesized that modification of host cell genes that interfere with the expression or function of viral genes will eventually contribute to immune evasion, and tumour progression and invasion (zur Hausen, 1999).

(c) Chromosomal abnormalities in HPV-associated cancers

Numerical chromosomal changes have been described in several HPV-associated cancers including cervical (see below), vulvar (Pinto et al., 1999; Rosenthal et al., 2001) and head and neck tumours (Braakhuis et al., 2004).

Aneuploidy has been observed in cervical cancers and their precursor lesions (reviewed in Lazo, 1999). Loss of heterozygosity (LOH), which most frequently involves chromosomes 1, 3, 6, 11, 17 and 18 has been reported. Losses in the short (p) arm of chromosome 3 and gain on the long (q) arm are among the most frequent events associated with progression from high-grade lesions to cervical cancer (Heselmeyer et al., 1996; Larson et al., 1997; Wistuba et al., 1997; Lin, W.M. et al., 2000; Nishimura, M. et al., 2000). An association between the severity of anal intraepithelial neoplasia (AIN) and chromosomal changes detected by comparative genomic hybridization has been described (Haga et al., 2001). The most common alteration involved 3q, similar to the commonest alteration seen in cervical cancer, which suggests that a common molecular pathway for these HPV-associated malignancies exists. Moreover, LOH at 3p was more frequent in HPV 16- and 18-positive cervical tumours, whereas LOH at the 5p regions was more frequent in HPV-negative tumours (Mitra, 1999). Kersemaekers et al. (1999) analysed the CIN component, invasive carcinoma and lymph node metastases from 10 patients with primary squamous-cell carcinoma of the cervix for LOH. In CIN lesions, LOH was frequently found at 3p, 6p and 11q. During progression to an invasive tumour, losses of genes were observed on 6q, 17p and 18q. It was suggested that loss of an additional locus on the X chromosome and activation of the erbB2 oncogene are important in progression to metastases. Evidence that integration of high-risk HPV is associated with genomic alterations measured by comparative genomic hybridization was provided in a study that examined different degrees of cervical squamous intraepithelial lesions (SIL): more numerical chromosomal aberrations were found in high-grade lesions with integrated HPV DNA than in low-grade lesions (Alazawi et al., 2004). ELhamidi et al. (2004) analysed 164 CIN for LOH at 12 micro-satellite loci and found that LOH at D3S1300, D3S1260, D11S35 and D11S528 was associated with CIN, which showed a tendency to persist and/or progress. An indication of geographical distribution of genetic alterations in oesophageal carcinomas has been reported by Si et al. (2004): HPV-positive tumours from Hong Kong, but not from Sichuan, had a higher frequency of LOH at D5S82, D6S397 and D13S260 than HPV-negative tumours.

Alterations in the pRb and p53 genes have also been studied. Kim et al. (1997) reported infrequent LOH at pRb (14%) and p53 (5.5%) loci in 55 primary cervical carcinomas. The genes p53 and pRb are less frequently mutated in HPV-positive than in HPV-negative cervical cancer cell lines and tumours (reviewed in Tommasino et al., 2003). Similarly, expression of HPV 16 E6 in head and neck cancers correlated with the absence of mutations within the p53 gene (Braakhuis et al., 2004; Dai et al., 2004).

Continuous expression of papillomavirus early genes may directly promote genetic abnormalities that often result in impaired function of genes that are critical for cell homeostasis. Furthermore, expression of high-risk HPV E6 and E7 genes in the basal cell layer induces chromosomal instability and aneuploidy. It has recently been suggested that this event precedes integration of the viral genomes which in turn triggers the continuous expression of early genes (Melsheimer et al., 2004). It could therefore be speculated that the physical status of the HPV genome and testing for aneuploidy could be used as prognostic tools (Kashyap & Das, 1988).

(d) Alterations of specific proto-oncogenes

Several investigations have addressed the structural or functional alteration of different proto-oncogenes. Most of these were descriptive in nature and few were designed to correlate the observed alteration with progression of disease. Therefore, a direct role for these genetic alterations in HPV-associated carcinogenesis is difficult to establish.

(a) Effects of other infectious agents

The proposed mechanisms through which infectious agents might act as co-factors in HPV-associated tumorigenesis include direct biological interactions, such as modification of HPV replication and transcription, and indirect effects, such as inflammation and damage to the epithelial barrier that protects against HPV infection. Herpes simplex virus-2 (HSV-2) is one of the infectious agents that has been most frequently studied as a potential co-factor for cervical cancer. However, epidemiological studies have provided conflicting results (see Section 2.7.2) for an association of HSV-2 with cervical cancer.

Because of the lytic behaviour of many herpesviruses, including HSV-2, abortive infections would need to be involved for HSV-2 to have a direct effect on HPV-associated tumorigenesis. Accordingly, UV-inactivated HSV-2 can transform rodent cells in vitro (Duff & Rapp, 1971a,b, 1973). Several studies have demonstrated an interaction between HSV-2 and HPV in transformation in vitro (Dhanwada et al., 1993), whereas others have found that HSV-2 can suppress HPV gene expression (Fang et al., 2003). Thus, laboratory data similarly to the epidemiological data are not consistent with regard to a possible interaction of HSV-2 in HPV-associated tumorigenesis.

Laboratory studies conducted during the 1970s demonstrated the ability of UV-inactivated HSV-2 and HSV-1 to transform hamster cells, and showed that these transformed cell lines caused tumours in newborn rodents (Duff & Rapp, 1971a,b, 1973). The continued presence of the HSV genome was observed in some tumorigenic cell lines, but the transformed phenotype was also found to persist in the absence of detectable HSV viral sequences (Davis & Kingsbury, 1976). These findings and the inconsistent detection of HSV DNA in specimens of human cervical cancer gave rise to hypotheses of a possible ‘hit and run’ mechanism (Davis & Kingsbury, 1976; Skinner, 1976; Galloway & McDougall, 1983), i.e. the concept that a virus may be involved in the initiation or promotion of cancer without being required for the maintenance of the transformed phenotype.

At least two separate genomic regions of HSV have been shown to transform rodent cells in vitro: the morphological transforming region II in the BglII N fragment and the morphological transforming region III in the BglII C fragment (Jones, 1995). In contrast to rodent cells, the BglII N fragment in human keratinocytes was found to induce tumorigenic clones in cells that had been immortalized by HPV but not in normal cells (DiPaolo et al., 1990), a finding that is consistent with the hypothesis that HSV is a co-factor in HPV-associated cervical tumorigenesis, but is not itself an important etiological agent. Moreover, although BglII N sequences were not detected in tumour-derived cell lines (DiPaolo et al., 1990, 1998), when only the Xho2 segment of the BglII N fragment was used, HPV-immortalized cells could still be transformed and the Xho2 segment was found to be maintained stably in an integrated form in the host cell genome (DiPaolo et al., 1998). These data were interpreted as evidence that the Xho2 fragment contains the transforming sequences of the BglII N fragment, and that the remaining sequences have an inhibitory effect on stable integration. However, a study of 200 specimens of human cervical cancer failed to detect any HSV-2 sequences using sensitive PCR methods (Tran-Thanh et al., 2003). In-vivo studies of the detection of HSV and other infectious agents in cervical specimens are mainly reviewed in Section 2.7.2.

Other herpesviruses that are reported to infect the cervix have also been shown to transform epithelial cells in tissue culture, including cytomegalovirus (CMV) (Galloway et al., 1984; Doniger et al., 1999), human herpesvirus (HHV) 6 (Razzaque, 1990; Kashanchi et al., 1997) and Epstein-Barr virus (EBV) (Lopes et al., 2003; Busson et al., 2004; Thompson & Kurzrock, 2004), but there is no strong evidence that these viruses are involved in cervical cancer.

Chlamydia trachomatis is a microbial agent that has most consistently been shown to be associated with cervical cancer in epidemiological studies that controlled statistically for HPV infection (see Section 2.7.2). Although it is not believed to have a direct effect on host DNA or on the transcription of HPV genes, several biological mechanisms by which C. trachomatis may increase the risk for cervical cancer have been described. First, it may have anti-apoptotic effects (Fan et al., 1998): resistance of infected cells to apoptosis ensures the persistence of C. trachomatis infection, while cell death at the end of the infection cycle triggers release and initiates a new infection cycle (Fan et al., 1998, Perfettini et al., 2003a,b). These anti-apoptotic effects could result in increased persistence of epithelial cells that are co-infected with HPV and/or reduce the probability of cell death following the development of chromosomal abnormalities, which increase in frequency with increasing grade of cervical neoplasia (Lorenzato et al., 2001, Melsheimer et al., 2001). Second, infection by C. trachomatis is associated with squamous metaplasia and hypertrophic ectopy, which have been shown to be a risk factor for cervical neoplasia (Moscicki et al., 1999). Third, C. trachomatis may cause human cervical epithelial cells to separate from each other due to the breakdown of the cadherin–catenin junctions in the epithelium (the N-cadherin/β-catenin complex) and thereby increase the exposure of basal cells to HPV (Prozialeck et al., 2002). Fourth, it may increase the risk for HPV infection and its persistence through modulation of immune factors: C. trachomatis is reported to inhibit the expression of interferon (IFN) γ-inducible major histocompatibility complex (MHC) class II (Zhong et al., 1999), as well as the expression of MHC class I (Zhong et al., 2000; Hook et al., 2004). If this is true, it could impair the adaptive immune response to HPV. Quantitative or qualitative alterations in expression of MHC class I (e.g. due to the presence of viral antigens) can result in stimulation of natural killer (NK) cells (core effector cells of the innate immune system that can kill a broad range of intracellular microbially infected cells without prior sensitization). C. trachomatis may, however, inhibit NK cell function and result in a decrease in the lytic capability of NK cells, reduced NK cell production of TNFα and IFNγ and a decrease in antibody-dependent cellular cytotoxicity (Mavoungou et al., 1999). Last, chronic infection with C. trachomatis is associated with a predominantly T-helper (Th)2 (humoral immune) cytokine pattern, whereas Th1 (cellular immune) cytokines are important in the control of intracellular microbes such as C. trachomatis and HPV (Stephens, 2003).

Overall, the immune response to microbial infection (i.e. cervical inflammation) may play a role in HPV-associated tumorigenesis and help explain the possible associations of cervical cancer with a range of pathogens, including herpesviruses, C. trachomatis, Trichomonas vaginalis, Neisseria gonorrhoeae, Candida albicans and others (Castle & Giuliano, 2003). The mechanisms by which inflammation might cause an increased risk for cervical cancer have been best described for C. trachomatis. Many of the cytokines that are secreted during C. trachomatis infection, including TNFα and IFNγ, could cause tissue damage by inducing apoptosis of uninfected cells (Perfettini et al., 2000), and infiltrating macrophages may cause further tissue damage through release of reactive oxygen species (Castle & Giuliano, 2003). Together, these effects probably result in partial disruption of the tissue barrier and exposure of basal cells to HPV infection. Furthermore, it has been hypothesized that the reactive oxygen species released by infiltrating macrophages could cause host cell DNA damage that leads to increased risk for cervical cancer in cells that are protected against apoptosis by HPV (Gravitt & Castle, 2001; Smith et al., 2004). Support for this proposed mechanism has come from laboratory studies that showed an association between inflammatory host responses and oxidative DNA damage (Zhuang et al., 2002; Touati et al., 2003).

The incidence of cervical cancer is significantly increased in women who have human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS) (Mbulaiteye et al., 2003), and biomarkers of host immune status in HIV-positive women, including HIV RNA level and CD4⁺ T-cell count, are associated with risk for HPV infection and cervical neoplasia (see Section 2.8.3). It is uncertain, however, whether there might also be a direct biological interaction between HIV and HPV. In-vitro studies have shown that HIV TAT protein can co-activate HPV (Tornesello et al., 1993; Vernon et al., 1993; Buonaguro et al., 1994). Whereas some studies have reported that epithelial cells can be infected by HIV in vitro (Moore et al., 2003; Yeaman et al., 2003), there is little evidence of this in vivo (Spira et al., 1996; Greenhead et al., 2000; Miller & Shattock, 2003; Wu, Z. et al., 2003). Taken together, it seems improbable that HPV-infected cervical epithelial cells could be coinfected with HIV, which limits the opportunity for the two viruses to interact directly at the molecular level.

Adeno-associated virus (AAV) may have a protective effect against HPV-associated cervical tumorigenesis. AAV is a helper-dependent parvovirus that requires co-infection with other DNA viruses, such as adenovirus, for its replication (Leonard & Berns, 1994). In tissue culture, AAV inhibits the tumorigenic effects of HPV (Hermonat, 1994a) and BPV (Hermonat, 1989) and furthermore, HPV can support replication of AAV (Walz et al., 1997; Meyers et al., 2001), a finding that is consistent with possible HPV/AAV coinfection in nature. Initial laboratory studies found that AAV suppressed papillomavirus replication (Hermonat, 1992) and attributed this effect to Rep 78, the major non-structural regulatory protein of AAV. Rep 78 can interfere with transcription factors and HPV promoter activity (Zhan et al., 1999; Su et al., 2000; Prasad et al., 2003). However, recent studies have found that the effects of AAV on HPV replication are complex (Agrawal et al., 2002). High levels of AAV decreased but low levels increased HPV replication (Meyers et al., 2001) and, under certain culture conditions, AAV actually increased the tumorigenicity of papillomavirus (Hermonat et al., 1998). The significance of these recent findings is not yet clear, but AAV also has anti-neoplastic effects that are independent of its proposed biological interaction with HPV. Through the direct interaction of AAV proteins with cellular genes, AAV has been shown to induce differentiation of tumour cell lines (Bantel-Schaal, 1995), down-regulate c-FOS and c-MYC (Hermonat, 1994b), inhibit cell proliferation (Walz & Schlehofer, 1992) and reduce carcinogen-induced mutagenicity (Schlehofer & Heilbronn, 1990).

(b) Hormones and anti-estrogens

Epidemiological data suggest an association between hormonal status, use of hormonal contraception and parity and the risk for preneoplastic lesions of the cervix and cervical cancer (see Section 2.7). The experimental evidence for such an association derives primarily from the fact that the endogenous level of the steroidal hormone, progesterone — the major ingredient of oral contraceptives and injectable hormonal contraceptives — increases during pregnancy (Pater et al., 1994), and experimental studies indicate the presence of hormonal recognition elements in the LCR of high-risk mucosal HPV and increased production of the E6 protein in response to exogenous hormonal stimulation in vitro (reviewed by Pater et al., 1994; Moodley et al., 2003; de Villiers, 2003). There is also experimental evidence that hormones may mediate changes in the immune status of the cervical mucosa (Roche & Crum, 1991).

In a review of the literature, Pater et al. (1994) summarized the evidence that supports an association between hormones and HPV-mediated tumorigenesis. Several reports before 1994 observed hormone-enhanced transformation of primary rodent cells, immortalization of genital keratinocytes and enhanced expression of HPV in cervical cells. In addition, inhibition of transformation and expression by anti-hormones (e.g. RU-486) had been observed in several reports. Since that time, the evidence to support a role of hormones, including estrogen and progesterone, in the increase in HPV expression has accumulated (Chen et al., 1996; Khare et al., 1997; Webster et al., 2001).

In addition to the above-mentioned effects of hormones on the expression of HPV and the ensuing carcinogenesis, there is evidence for other mechanisms of action of hormones in the development of HPV-related cervical cancer. Auborn et al. (1991) observed enhanced 16α-hydroxylation of estradiol activity in both cervical and foreskin cells immortalized with HPV 16. As 16α-hydroxyestron is known to be a risk factor for other estrogen-sensitive cancers, increased concentrations of this metabolite in target cells in the cervix may enhance cervical carcinogenesis by increasing cell proliferation in the presence of HPV 16. Monsonego et al. (1991) examined estrogen and progesterone receptor profiles in CIN and invasive cervical cancers and found high levels of expression of progesterone receptors in the underlying stromal cells of preneoplastic lesions of the cervix. These data suggest that, in vivo, sex steroid hormones, particularly progesterone, may act indirectly on HPV-infected epithelial cells and be implicated as co-factors in HPV-related cervical neoplasia. These results could also explain the relative predisposition to malignant transformation of the cervical mucosa compared with vulvar and penile mucosa.

More recently, evidence of a co-carcinogenic role for hormones in cervical cancer has accumulated from studies conducted in HPV-infected transgenic mice. Arbeit et al. (1996) demonstrated that exposure of K14-HPV 16 transgenic mice to 17β-estradiol increased the incidence of proliferating cells in the cervical and vaginal squamous epithelium and resulted in a concomitant up-regulation of E6/E7 gene expression through all stages of carcinogenesis. In addition, exposure of these K14-HPV 16 transgenic mice to estrogen induced hyperplasia in the lower uterine gland, and continuous exposure to estrogen resulted in the development of squamous metaplasia and neoplastic progression (Elson et al., 2000). In transgenic mice that express HPV 16 E6 or E7 alone, Riley et al. (2003) showed that E7 and estrogen combined are sufficient to induce cervical cancer, and that E6 contributes to increased tumour growth. In a more recent study, estrogen was found to contribute not only to the genesis but also to the maintenance and malignant progression of cervical cancers in HPV 16 transgenic mice (Brake & Lambert, 2005). Michelin et al. (1997) observed significant activation of the viral upstream regulatory region in response to exogenous estrogen and progesterone and pregnancy in HPV 18 transgenic mice.

(c) Nutrients

A large number of different food constituents and nutrients have been associated with a reduction in the persistence of HPV and the development of preneoplastic lesions and invasive cancer of the cervix. The mechanisms by which these food-derived compounds confer protection are not entirely clear. Three main types of nutrient and/or nutrient metabolite have been identified: those that are involved in oxidation reactions (e.g. carotenoids, vitamins C and E), those that are involved in methylation or one-carbon transfer reactions (e.g. folic acid, vitamin B12, vitamin B6, cysteine and the biological marker, serum homocysteine) and nutrient metabolites that have hormone-like activity (e.g. retinoic acid and its isomers).

Epidemiological research over the past few decades has indicated that anti-oxidant nutrients such as carotenoids that are found in fruit and vegetables as well as vitamins C and E may confer protection against the persistence of HPV and the development of pre-neoplastic lesions and invasive cervical cancer (see Section 2.7.1). The mechanism by which these nutrients might prevent cervical cancer remains unclear. Reactive oxygen species appear to play a central role in cell signalling by activating transcription factors, AP-1 and nuclear factor (NF)-κB, cell proliferation and apoptosis (Palmer & Paulson, 1997). In animal and in-vitro models, reactive oxygen species increased viral titres (Peterhans, 1997) and the infectivity of influenza virus (Hennet et al., 1992). As anti-oxidants, carotenoids and vitamins C and E have a multitude of effects that may be chemopreventive. These compounds have been shown to quench reactive oxygen species that can lead to cellular damage and dysregulation of cell signalling (Palmer & Paulson, 1997). Carotenoids and vitamins C and E may also potentiate host cellular and humoral immunity (Meydani et al., 1995). Studies of HIV and influenza virus also indicate a role for anti-oxidants (in particular nutrient anti-oxidants) in the down-regulation of viral replication and expression.

Administration of anti-oxidants to animals infected with influenza virus protected against the lethal effects of influenza (Oda et al., 1989). In vitro, increases in the cellular oxidant load have been shown to increase the replication of HIV (Pace & Leaf, 1995; Peterhans, 1997). This effect is thought to be due to the fact that reactive oxygen species activate NF-κB, a nuclear transcriptional factor that is obligatory for HIV replication (Pace & Leaf, 1995), and in-vitro studies have consistently demonstrated inhibition of NF-κB activation by anti-oxidants (see review by Epinat & Gilmore, 1999). Further molecular epidemiological studies are required to delineate to what extent anti-oxidant nutrients have a protective effect in vivo and the mechanisms by which they act.

Evidence has accumulated to suggest that reactive oxygen species and their down-regulation by anti-oxidants may have a similar effect on HPV infection. Activation of AP-1, a central transcription factor for the expression of E6 and E7 proteins of high-risk HPV types (Cripe et al., 1990; Offord & Beard, 1990), has been shown in vitro to be inhibited by antioxidants. Rösl et al. (1997) demonstrated that the anti-oxidant pyrrolidine-dithiocarbamate selectively suppressed AP-1-induced HPV 16 gene expression in HPV 16-immortalized human keratinocytes, and suggested that manipulation of the redox potential may be a novel therapeutic approach to interfere with the expression of high-risk HPVs.

In addition to its effects on target cells and cell signalling, the oxidant–anti-oxidant balance is an important determinant of immune cell function, and affects the maintenance of immune cell membrane lipids, control of signal transduction and gene expression of immune cells (Meydani et al., 1995; Anderson & Theron, 1990), events that are important for the loss of HPV infection and regression of CIN.

Whitehead et al. (1973, 1989) proposed a correlation between folic acid status and the use of oral contraceptives and increased risk for cytological abnormalities among women in the USA. This increased risk was thought to be due to a deficiency of folic acid in local tissues, which could not be detected by measuring folate in serum or in the diet. In support of this hypothesis, several laboratory studies have delineated a role for folic acid in the prevention of cancer at several sites. Folic acid is essential for the synthesis of purine nucleotides and thymidilate, which are essential for the synthesis of DNA during cell replication and repair. In addition, folic acid is necessary for the synthesis of S-adenosylmethionine, the main donor of methyl groups in various methylation reactions, such as methylation of the DNA base cytosine (Poirier, 2002). Low tissue levels of folate increase the frequency of fragile sites on DNA (Ames & Wakimoto, 2002), enhance the risk of attack on DNA by carcinogens and viruses (Hsieh et al., 1989), and decrease DNA repair (Ames & Wakimoto, 2002) and DNA methylation (Wainfan et al., 1988). Kim et al. (1994) observed significant increases in global DNA methylation in cervical tissue with increasing grade of cervical lesion, which suggested that the change in methylation status may be an early event in cervical carcinogenesis. Extending the work to genital HPV types, DNA methylation within the upstream regulatory region has been shown to regulate expression of high-risk HPV in vitro (Rösl et al., 1993; Thain et al., 1996). One in-vitro study demonstrated that sequence-specific methylation of CpG sites in the constitutive enhancer region of the HPV 18 upstream regulatory region resulted in a down-regulation of transcriptional activity (Rösl et al., 1993). Methylation of a novel transcription factor-binding site decreased the activity of the HPV 16 enhancer and suppressed viral transcription (List et al., 1994). Additional in-vitro studies demonstrated that methylation of specific CpG sites within the HPV 18 E2-binding site abolishes binding (Thain et al., 1996) and leads to a direct effect on E6 and E7 transcription. In summary, changes in the methylation of host and viral DNA may result in an increase in the production of viral proteins and hence in the risk for carcinogenesis. However, it remains unclear whether nutritional status of methyl donors such as folic acid and vitamin B12 influence the methylation patterns of host and viral DNA.

As described in Section 2.7.1, topical all-trans-retinoic acid was found to effect regression of CIN2 lesions in one placebo-controlled clinical trial (Meyskens et al., 1994). Chemoprevention studies that used retinoic acid to prevent preneoplastic lesions or make them regress at other epithelial sites have been successful. A growing body of basic experimental research indicates that retinoic acid and related compounds also have chemopreventive activity in the cervix. Retinoic acid is essential for terminal differentiation of cervical epithelial cells because it decreases cellular proliferation and DNA replication. It differentially inhibits the growth and differentiation of HPV 16-immortalized cervical epithelial cells (Agarwal et al., 1991; Eckert et al., 1995) and low-passage human foreskin keratinocytes (Pirisi et al., 1992; Khan et al., 1993; Creek et al., 1994) compared with normal human keratinocytes in the absence of an HPV infection. In addition to decreasing cellular proliferation in HPV 16-immortalized low-passage human keratinocytes, physiological concentrations of retinoic acid inhibit the expression of HPV 16 E6 and E7 (Pirisi et al., 1992; Khan et al., 1993; Creek et al., 1994). Retinoic acid may indirectly reduce levels of HPV mRNA by influencing the activity of AP-1 (Schüle et al., 1991) or the expression of transforming growth factor β (TGFβ) (Batova et al., 1992). In addition, retinoic acid suppresses cell growth. However, this suppression appears to be lost in late stages of HPV 16-induced transformation of human keratinocytes (Borger et al., 2000) and cervical carcinoma cell lines. In several in-vitro model systems, cells in the late stages of HPV 16-induced transformation acquire resistance to retinoic acid-induced differentiation through several different mechanisms, including loss of growth inhibition, loss of sensitivity to TGF β (Borger et al., 2000), continued growth stimulation (Higo et al., 1997; Sizemore et al., 1998) and loss of expression of the retinoid receptor (Bartsch et al., 1992). This resistance to retinoic acid is consistent with observations that therapy with retinoic acid does not reduce recurrence rates of invasive cervical cancer (Wadler et al., 1997; Look et al., 1998; Weiss et al., 1998) nor does it increase regression of CIN3 (Meyskens et al., 1994). Taken together, these data suggest that retinoic acid may only be effective in the early stages of cervical carcinogenesis by modulating the clearance and persistence of HPV, viral load and the regression of moderate CIN.

(d) Tobacco smoke

The evidence from epidemiological studies is sufficiently strong to conclude that, in the presence of HPV infection, tobacco smoking is a co-factor in the development of pre-neoplastic lesions of the cervix and invasive cervical cancer (see Section 2.7.1). Currently, there appear to be two different mechanisms by which tobacco smoking can increase the risk for cervical diseases. Cigarette smoke contains mutagens, carcinogens and other components that may act as initiators and/or promoters of uterine cervix carcinogenesis. These components can either affect immune function and allow HPV infection to persist and progress or act directly as co-carcinogens in cervical tissue or both.

Products of cigarette smoke have been found in body fluids outside the lung such as in the breast fluid of lactating smokers, the amniotic fluid of smokers, the urine and saliva of infants of mothers who smoke and secretions of the cervical mucus of smokers (Holly et al., 1986). Both nicotine and cotinine are measurable in the cervical tissue of smokers (Sasson et al., 1985; Schiffman et al., 1987; Hellberg et al., 1988; McCann et al., 1992) and nonsmokers exposed to secondhand tobacco smoke (Jones et al., 1991). On average, concentrations of nicotine were more than 45 times higher in cervical tissue than in serum among women who smoked (Sasson et al., 1985).

Nicotine is metabolized by oxidative N-nitrosation to 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone that is in turn metabolized by α-hydroxylation to nitrosamine, which is considered to be one of the most potent carcinogens in cigarette smoke (Hellberg et al., 1988). Farin et al. (1995) demonstrated that both HPV 16-immortalized oral and cervical cell lines express cytochrome P450 enzymes that are necessary for the activation of nitrosamines and polycyclic aromatic hydrocarbons. Significantly higher concentrations of the tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (Prokopczyk et al., 1997) and benzo[a]pyrene metabolites (Melikian et al., 1999a) have been observed in the cervical mucus of smokers compared with that of nonsmokers.

As determined by the Ames Salmonella mutagenicity test, smokers are at an increased risk for having mutagenic cervical fluid (Holly et al., 1986). Cervical tissues from smokers have significantly elevated levels of DNA damage as measured by DNA adducts compared with nonsmokers (Simons et al., 1993, 1995; Melikian et al., 1999b). In-vitro experiments (Nakao et al., 1996; Yang et al., 1996) suggest that cigarette-smoke condensate induces faster growth in serum, higher saturation density, anchorage-independent growth and tumorigenicity in cells immortalized by HPV 16 and 18. Melikian et al. (1999b) also suggested that HPV 16-infected cells in vitro are more susceptible to DNA damage by benzo[a]pyrene than non-HPV-infected cells. Finally, tobacco smoke has been associated with aberrant hypermethylation of the tumour-suppressor gene p16 which significantly correlated with the grade of cervical disease (Lea et al., 2004). Collectively, these data provide biochemical evidence that components of cigarette smoke have a carcinogenic potential on cervical tissue which, in an HPV-infected cervix, would increase the risk for progression to cervical carcinoma.

In addition to its direct role in carcinogenesis, tobacco smoking has been associated with a generalized suppression of the immune system, including a significant decrease in NK cells and NK cell activity, in circulating levels of immunoglobulin (Ig)G and IgA (Ferson et al., 1979) and in Langerhans cells (Barton et al., 1988; Poppe et al., 1996). Langerhans cells are dendritic cells that are localized in the epithelium and present antigen to T lymphocytes. A reduction in the number of Langerhans cells available to detect and present viral antigens may facilitate the establishment and persistence of local viral infection. Giuliano et al. (2002c) demonstrated that tobacco smoking was associated with an increased risk for persistence and duration of high-risk HPV infection. The resultant viral persistence may increase the probability of the development of virally induced neoplastic transformation.

(e) Radiation

(a) Tumorigenic properties of HPV genes in mouse skin

The most common type of cancer caused by HPVs in humans is squamous carcinoma of the anogenital tracts and oral cavity. The first evidence in HPV transgenic mice that HPV 16 genes could induce squamous carcinoma derived from the analysis of αAcryE6/E7 mice (in which the viral genes were placed under transcriptional control of the αA crystallin promoter) that expressed these genes ectopically in the skin (Lambert et al., 1993). The mice developed squamous carcinomas as adults (Lambert et al., 1993), as well as tumours of the lens (Griep et al., 1993) and retina (Griep et al., 1998). Transgenic mice in which the expression of HPV 16 E6 (Song et al., 1999), HPV 16 E7 (Herber et al., 1996) and all early HPV 16 genes (Coussens et al., 1996) was directed to the basal compartment of stratified squamous epithelia by the keratin 14 (K14) promoter developed cancers. The severity of the cancers that arose in mice that expressed E6 or E7 differed individually. In K14E7 mice, the tumours were primarily benign or low-grade carcinomas, whereas in the K14E6 mice, the majority of tumours were malignant carcinomas of higher grade (Song et al., 2000). These differences were also seen in the synergy between HPV genes and chemical carcinogens in the induction of skin cancer. E6 was found to contribute to two stages in carcinogenesis — promotion, which is a required step in the formation of benign papillomas, and progression, which is the process that converts a benign tumour to a malignant cancer — whereas E7 only contributed to promotion (Song et al., 2000). Exclusively benign skin tumours were observed in other strains of mice in which the expression of HPV 16 or 18 E6 and E7 was directed by the tyrosinase or human keratin (hK) 1 promoters (Greenhalgh et al., 1994; Kang et al., 2000). Activated forms of the Ras proto-oncogene acted synergetically with HPV 16 E6 and E7 to produce tumours (Schreiber et al., 2004).

The carcinogenic properties of a cutaneous HPV in mice have now been described (Schaper et al., 2005). The early region of the EV-HPV 8 genome was placed under transcriptional control of the hK14 promoter. While multiple independent lines of HPV 8 transgenic mice did not display acute phenotypes of epithelial hyperplasia as seen in mice that expressed HPV 16 E6 and E7 genes separately or together (see above), they developed both benign and malignant tumours of the skin at frequencies of 91 and 6%, respectively. The frequency of tumours was higher and tumours occurred at an earlier age than those in HPV 16 transgenic mice. This provides the first evidence that a cutaneous HPV can cause tumours in vivo in an animal model.

(b) Tumorigenic properties of HPV genes in the reproductive tracts of mice

The first mouse model for HPV-induced cervical cancer used a recombinant retrovirus to transduce HPV 16 E6 and E7 genes into the mouse cervix (Sasagawa et al., 1992). Cervical cancer developed when the mice exposed to the E6/E7 recombinant retrovirus were treated with the tumour promoter 12-O-tetradecanoylphorbol-13-acetate or the mutagen N-methyl-N′-nitro-N-nitrosoguanidine. A more tractable model for HPV-induced carcinogenesis was developed in which K14HPV16 transgenic mice were treated with exogenous estrogen. These mice developed a progressive disease that led to the formation of squamous carcinoma of the cervix over a 6-month period and closely reflected the histopathological characteristics of the progressive disease that leads to cervical cancer in humans (Arbeit et al., 1996; Elson et al., 2000). The individual role of E6 and E7 in cervical cancer was found to differ from that observed in the skin; E7 played a more dominant role in the cervix and E6 in the skin (Riley et al., 2003).

The validity of these HPV 16 transgenic mice as models for human cervical cancer has been demonstrated at several levels. First, the histopathological progression of disease in mice closely parallels that in human cervical cancer (Riley et al., 2003). Second, the role of estrogen as a co-factor in the development and progression of cervical cancers in mice (Brake & Lambert, 2005) parallels the epidemiological evidence for a role of estrogen in human cancers (see Section 2.7.1). Third, there is a close parallel in the expression pattern of biomarkers for human and murine cervical cancer (Brake et al., 2003). Last, the demonstration that the anti-estrogenic drug, indole-3-carbinol, inhibits cervical cancer in the HPV-transgenic mouse model (Auborn et al., 1991) has led to its successful use in the clinical treatment of CIN2/3 (Bell et al., 2000).

(c) Use of HPV transgenic mice for immunological studies

Many investigators have used HPV transgenic mice to investigate host immune responses to viral antigens and to develop protocols to induce antigen-specific therapeutic immune responses against viral antigens. Although viral genes that are expressed as trans-genes should be considered as self-antigens, many studies have demonstrated that HPV transgenic mice can mount immune responses to E7 when stimulated appropriately. When immunized with E7 protein, HPV transgenic mice (line 19 αAcryE6E7) that develop squamous-cell carcinomas (Griep et al., 1993) produced antibody and showed Th responses to E7 that were indistinguishable from those seen in immunized syngeneic, non-transgenic mice (Frazer et al., 1995; Herd et al., 1997). Non-immunized line 19 mice, when allowed to age and develop skin lesions, showed spontaneous E7-specific immune responses (Frazer et al., 1995). While these antigen-specific responses neither contributed to nor modulated the skin diseases that occurred in these mice, non-specific local inflammatory responses did contribute positively to skin disease (Hilditch-Maguire et al., 1999). K14E7 mice that express E7 from the K14 promoter display split tolerance to E7, which is characterized by the ability to mount normal Th and B-cell responses to E7, but the inability to mount E7-specific CTL responses on genetic backgrounds (i.e. H-2b) in which E7-specific CTL-restricted epitopes exist (Doan et al., 1998; Frazer et al., 1998; Doan et al., 2000).

The absence of CTL responses in K14E7 mice was associated with a non-specific down-regulation of CD8-positive T cells (Tindle et al., 2001) that was probably attributable to transgene-associated disturbance of the K14E7 thymic architecture (Malcolm et al., 2003). Similar split tolerance was observed in HPV transgenic mice in which the expression of E7 was directed from the K10 promoter (Borchers et al., 1999). The underlying reason for this split tolerance is not understood, but could reflect the fact that keratin promoters are active in thymic epithelial cells (Frazer et al., 1998).

The generation of HPV transgenic mouse models that express viral antigens in the skin offered the opportunity to use graft studies to investigate immune recognition of these antigens, when expressed in clinically relevant amounts in keratinocytes by naïve, immunologically intact animals. E7 did not function as a classical minor transplantation antigen with regard to its expression in mouse epidermis. When grafted onto non-transgenic mice, skin from K14E7 mice was not rejected even when these mice were immunized against E7 and developed E7-specific CTL responses (Dunn et al., 1997; Frazer et al., 2001). Similarly, grafts in which E6 antigen was expressed from the K14 promoter were also accepted (Matsumoto et al., 2004). However, recipient animals were induced to reject the skin grafts in an E7-specific manner after stimulation of systemic pro-inflammatory responses (Frazer et al., 2001) or through passive transfer of E7-specific CTLs in combination with E7specific immunization (Matsumoto et al., 2004). The reason why the immune system has difficulty in recognizing E7 antigen in the skin remains unclear, but E7 may possibly inhibit host immune responses. In one study in transgenic mice, it was suggested — on the basis of a comparison of expression levels of various IFN-responsive genes using real-time PCR in E7 transgenic versus non-transgenic mice — that E7 can suppress innate immune responses through its interaction with IRF1 (Um et al., 2002). In another study, however, no effect of E7 was noted on the levels of MHC class I protein, a major IFN-response gene product, following induction by IFN when keratinocytes from K14E7 transgenic mice were compared with those from non-transgenic mice (Leggatt et al., 2002). An alternative explanation for the poor recognition of E7 by the host immune response is that, as a non-secreted antigen, it is poorly cross-presented by antigen-presenting cells and/or that the local immune environment of the skin is not conducive to recognition by the host immune system of keratinocyte-expressed antigens. Regardless of the mechanism, the insights gained from these graft studies have clear implications with regard to the design of effective therapeutic vaccines to treat patients with HPV-associated disease.

(a) T-Helper cell responses

Most priming of CD4⁺ T cells occurs in the lymph nodes following the transit of Langerhans cells from infected tissues. Most infiltrating T lymphocytes carry HLA-DR antigens — a late activation marker — but only a few express CD25 — an early activation marker (Coleman et al., 1994), which indicates that they are probably activated at distant sites. T-Helper cells largely function through the up-regulation of CD40 ligand which interacts with CD40 on dendritic cells and leads to their maturation (Melief et al., 2002). CD40L on CD4⁺ T cells can also directly signal to CD8⁺ CTL. In addition, CD4⁺ cells play a crucial role by interacting with B cells to induce antibody production. The interaction of T-helper cells with dendritic cells determines whether they secrete Th1 or Th2 cytokines.

Activation of CD4⁺ occurs through recognition of MHC class II/peptide complexes present on antigen-presenting cells. Much effort has been made to identify which viral peptides can be recognized by CD4⁺ T cells and what are the consequences of having viral-specific CD4⁺ T-cell clones. In theory, all viral antigens can serve as targets that will be presented by class II antigens. Numerous studies have shown that CD4⁺ T cells proliferate in response to both early and late antigens (Tindle et al., 1991; Kadish et al., 1994; Luxton et al., 1997), and the responses tended to be type-specific. Which T-cell epitopes are presented is governed by the ability of a particular MHC molecule to bind a given peptide.

In addition to their critical function in priming other cells in the cell-mediated immune response, CD4⁺ T cells deliver cytokines to the infected tissue, thereby influencing the outcome of the infection. The production of Th1-type cytokines is benefical for the reduction of lesions, whereas the elaboration of Th2 cytokines has a deleterious effect. Under the influence of IL-12, a Th1 response yields TNF and IFNg, whereas under the influence of IL-10, the Th2 cytokines IL-4 and IL-5 are generated, which favour B-cell development. The extent to which CD4⁺ T cells kill keratinocytes that have up-regulated HLA-DR is unclear.

Assays to detect CD4⁺ T-cell responses have relied on the use of peptides to stimulate T-cell proliferation, as measured by incorporation of tritium. More sensitive assays now measure the release of IL-2 from stimulated cells. Numerous studies have detected CD4⁺ T-cell responses in sera of patients. In some studies, Th1 responses to either E7 or E2 correlated with the clearance of lesions (Kadish et al., 1997). However, in other studies, the frequency of responses was similar in cases with low-grade lesions and in controls (de Gruijl et al., 1996), even in studies that used virgin women as controls (Nakagawa et al., 1996).

(b) Cytotoxic T-cell responses

The role of CD8⁺ T cells in killing HPV-infected cells is extremely important as most therapeutic vaccine strategies are aimed at induction of this type of T cell. It is now recognized that the ability of CTL to kill is not based solely on the interaction of the CD8⁺ T-cell receptor with antigenic peptides presented by MHC class I. These interactions, in the absence of other signalling cascades, frequently lead to tolerance. In contrast, productive CTL memory and killing requires signalling from highly activated dendritic cells and CD4⁺ T cells (Melief et al., 2002).

Numerous mouse models have shown that tumours that express HPV 16 or 18 E6 or E7 antigens are killed by generating a CTL response that is antigen- and MHC-restricted (Chen et al., 1991; Feltkamp et al., 1993; Tindle & Frazer, 1994). In these cases, the epitopes that served as recognition sequences were specified by the mouse MHC class I loci.

CTLs to E7 have been difficult to detect in the peripheral blood of patients known to have been exposed to HPV 16 or 18 (Borysiewicz et al., 1996); however, they can be detected at higher frequency in lesions that contain infiltrates from tumours (Evans et al., 1997). Oligopeptide epitopes or HLA-matched cervical cancer cell lines rarely stimulated peripheral blood CTLs, whereas viral vectors with recombinant E6 and/or E7 stimulated CTLs from 30-60% of HPV 16-positive CIN patients in vitro (Nakagawa et al., 1997; Nimako et al., 1997).

In a longitudinal study, more CTL responses to HPV 16 E6 were detected in women who had HPV 16 infection without SIL than in women who had SIL (Nakagawa et al., 2000). Fifty-one women who had HPV 16 infection and three HPV 16-negative control women were enrolled; 22 (55%) of 40 women who cleared HPV 16 infection had an E6 CTL response at least once compared with none of nine women who had persistent HPV16 (p = 0.003). No such difference was demonstrated for E7; 25 (63%) of 40 women who cleared HPV 16 infection responded versus five (56%) of nine women with persistent HPV 16 (p = 0.720). It appears that the lack of response to E6 is important in the persistence of HPV 16 infection. More studies on the CTL response to HPV antigens are needed to understand fully the role of CTLs in mediating regression.

(a) MHC class I

MHC class I antigens are expressed ubiquitously on human cells, including cervical keratinocytes. They present peptides that are derived from the processing of endogenous antigens of both host and viral origin to CD8⁺ T cells. When a vigorous CD8⁺ T-cell response is generated through stimulation by activated dendritic cells and by CD40 signalling pathways, cytolytic CD8⁺ T cells can kill cells that present viral peptides on the MHC class I antigen.

Early results on the expression of class I antigens gave conflicting results, probably due to the antibodies that were used for immunostaining. However, more recent studies consistently observed down-regulation of class I expression in cervical cancers (IARC, 1995). HLA class I A and B genes are highly polymorphic and inherited alleles for each gene are expressed. Examination of cervical tumours revealed that down-regulation of MHC class I expression was variable (Keating et al., 1995). Certain alleles were more frequently down-regulated including A2, A3, the A9 group, the B5 group, B7, B8 and B44. Several studies have examined whether certain HLA class I genotypes confer a risk for progression to cancer (see Section 4.2.5). The risk associated with some of these alleles is probably due to their more frequent down-regulation, although it is unclear whether this is an early event that would affect progression. Down-regulation of MHC class I has also been observed in laryngeal papillomas (Vambutas et al., 2000).

Many viruses target various steps in the class I antigen presentation pathway to escape immune detection. The mechanism by which class I down-regulation occurs in HPV infection is not clear and probably involves multiple mechanisms. Cromme et al. (1994) showed that directly diminished expression of class I genes was a rare event. The defect in class I expression in some tumours was found to be due to a loss of the peptide transporter 1 associated with antigen processing (TAP1) which occurred through down-regulation of the transcription of TAP1. HPV 16 and 18 E7 have been reported to down-regulate the promoters of both the TAP1 and MHC class I heavy chain (Georgopoulos et al., 2000). Other studies have reported that the HPV 11 E7 binds to the TAP transporter protein and blocks the loading of peptides onto class I antigens (Vambutas et al., 2001). HPV 16 E5 was recently reported to cause the retention of HLA class I complexes in the Golgi apparatus and impede their transport to the cell surface; this was rectified by treatment with IFN. Unlike BPV E5, HPV 16 E5 did not affect the synthesis of HLA class I heavy chains or the expression of the transporter associated with the antigen that processes TAP (Ashrafi et al., 2005).

(b) MHC class II

MHC class II antigens are generally only expressed on antigen-presenting cells, and present antigens derived from an exogenous antigen-presenting pathway to CD4⁺ T cells. However, in the majority of cervical premalignant and malignant neoplasias, HLA class II molecules are up-regulated on keratinocytes (Glew et al., 1992; Cromme et al., 1993). Normal ectocervical epithelium does not express detectable levels of class II antigens, nor did expression of HPV genes in cultured epithelium result in class II expression (Coleman & Stanley, 1994), which suggests that the up-regulation of class II was indirect and was mediated by the pro-inflammatory cytokine IFNγ and in part by TNFα (Majewski et al., 1991). Consistent with this explanation is the finding that cultured keratinocytes that expressed HPVs also expressed class II antigens when treated with IFNγ and the fact that class II expression is increased in various inflammatory skin diseases. Also consistent with this hypothesis is the correlation between increased inflammatory response in the stroma and increased infiltrating inflammatory cells and the extent of class II upregulation (Coleman & Stanley, 1994). The DR, DP and DQ class II MHC subloci are differentially expressed on keratinocytes within cervical squamous tumours, which suggests independent regulation (Glew et al., 1992). It has also been reported that the HPV 16 E5 protein can block INF-induced surface expression of MHC class II by preventing degradation of the invariant chain (Zhang et al., 2003).