Human Papillomavirus Infection and the Multistage Carcinogenesis of Cervical Cancer

The association of risk with sexual behavior has been posited since the mid-1800s, but the central causal role of human papillomavirus (HPV) infection was identified just 35 years ago (1). Thus, the important preventive impact of cervical cytologic screening (Papanicolaou tests) in the second half of the 20th century preceded and even advanced etiologic understanding. The accessibility of the cervix for population-wide tissue sampling, the delimited ring of tissue at risk (the cervical transformation zone), and the uniform causal pathway centered on HPV infection fostered the last few decades of productive interdisciplinary studies and improved preventive strategies (2).

The cervix is the lower third of the uterus, projecting into the anterior aspect of the vagina; with regard to carcinogenesis, it can be viewed topologically as a 2-dimensional ring of epithelium. The cervical transformation zone is an area of metaplastic tissue between the squamous epithelium of the vagina and the glandular tissue of the endocervical canal. While the entire anogenital epithelium can be infected by HPV, the cervical transformation zone is especially susceptible to carcinogenesis. Recently, a cell population in the transformation zone with specific morphologic and molecular features was described that may represent the cells of origin of most cervical precancers and cancers (3). Destruction or excision of the entire susceptible cell population when precursor lesions are found is still the mainstay of prevention.

As currently conceived (Fig. 1), the stages in cervical carcinogenesis include HPV infection; persistence, rather than clearance of the virus, linked to the development of a high-grade precursor lesion or “precancer”; and invasion. These are necessary stages; cervical cancer is virtually impossible in the absence of sexually transmitted HPV infection (4) and in the absence of intermediate progression to precancer. New prevention strategies reviewed below are based on this strong causal model and the established chain of surrogate endpoints.

It is interesting to review the historic development of nomenclature for precursors from our current perspective; as the changing terms have slowly adapted to understanding of the key role of HPV infection (ref. 5; Fig. 2). The nomenclature for the much more common squamous precursors (glandular lesions will be considered only briefly below) has evolved somewhat separately for histology than for cytology. Histology is what defines the underlying neoplastic process and guides treatment, whereas exfoliative cytology is used in screening that assesses the probable underlying histologic state.

Fifty years ago, for squamous histology, the cervical cellular abnormalities viewed as the precursor of cervical cancer were termed mild, moderate, or severe dysplasia; severe dysplasia was distinguished from the more severe diagnosis of carcinoma in situ. In the late 1960s, Richart proposed the concept of intraepithelial neoplasia (6). CIN3 encompassed severe dysplasia and carcinoma in situ, CIN2 replaced moderate dysplasia, and CIN1 later came to include both the microscopic evidence of HPV infection (koilocytotic atypia) and mild dysplasia. The severity of the diagnosis was based on the degree of replacement of the normal stratified epithelium with mitotically active basal-like epithelium (≤1/3 = CIN1, ≤2/3 = CIN2, >2/3 = CIN3). CIN was viewed as a stepwise progression, with a high probability of transition from the more minor to more serious cancer precursors.

As HPV research showed the high prevalence and transient nature of most cervical HPV infections, it became clear that the notion of inexorable CIN progression was not correct. CIN1 was found to be a poorly reproducible and insensitive histologic diagnosis of acute and mostly transient HPV infection (7). CIN2 was reconsidered as a heterogeneous borderline category between acute HPV infection and the more likely cancer precursor lesions (CIN3). The risk factor profiles (8–10) and HPV genotype distributions (11) in CIN2 and CIN3 are different, and CIN2 is more likely to regress spontaneously compared to CIN3 (12), but current clinical management of CIN2 and CIN3 diagnoses is very similar. The histologic nomenclature did not formally change, however, to a 2-stage system (low-grade lesion reflecting acute HPV infection, high-grade lesion representing cancer precursor to be treated) until the Lower Anogenital Squamous Terminology (LAST) conference in 2012 (13). The LAST nomenclature relies on p16 staining to triage CIN2; p16 is a biomarker of disruption by HPV of the Rb pathway (14). CIN2 that is p16-positive is combined with CIN3 to form high-grade squamous intraepithelial lesion (HSIL), representing the immediate precursor to cervical cancer. CIN2 negative for p16 is combined with CIN1 to form low-grade squamous intraepithelial lesion (LSIL), representing the histologic sign of HPV infection (Fig. 2). The LAST nomenclature is new for histology, somewhat controversial, and has not been implemented widely as of yet. We will use LAST terminology for this review, although many centers still report histology using the CIN, or even the dysplasia, scale.

Cytology nomenclature evolved separately and was focused practically on which women to refer for colposcopically directed biopsy. The early Pap classification (I, II, III, IV, V) gave the probabilistic prediction of underlying invasive cancer. As experience with cytology increased, an effort was made to describe and predict the actual precursor lesions. The dysplasia/CIS nomenclature was applied and then replaced by CIN terminology. Importantly, in 1990 and with subsequent revisions, the Bethesda System recognized the low-grade/high-grade distinction, with explicit distinction between acute HPV effect (LSIL) and high-grade precursor (HSIL; ref. 15).

Therefore, if LAST terminology is accepted, squamous histology and cytology will eventually be indistinguishable in the United States.

The much less common glandular lesions are not as well-studied but are also caused by HPV infection (16). The putative stages, in cervical cytology, are atypical glandular cells (AGC), adenocarcinoma in situ (AIS), and adenocarcinoma. There is no defined precursor pathway delineated for even rarer histologies, for example, adenosquamous.

In describing the epidemiology of cervical cancer precursors, there are more easily accessible statistics for cytologic results than for histologic precursors because SEER sites generally do not record CIN3 or lesser lesions, whereas there are large published databases of screening cytologic results (17). It is important to note that the distribution of precursors is affected not just by the prevalence of HPV in the population but also by previous rounds of screening and treatment of HSIL and cancer (HSIL+). Both LSIL and HSIL are detected only when screening is conducted. In well-screened populations in the United States, approximately 0.5% of screening cytology results are HSIL (18). Much more common are the results (LSIL and the half of equivocal or ASC-US lesions) that reflect the cytologic evidence of acute HPV infection, accounting together for approximately 5% of cytologic results. HPV infection, detected by molecular tests, is much more common still, that is, only a minority of infections of infections produce even equivocal cytologic abnormalities (17, 19).

At the precursor level, squamous LSIL/HSIL lesions are overwhelmingly more prevalent than glandular AGC/AIS lesions (0.2% of screening results), whereas in the United States, about 15% of cancers now are adenocarcinoma (20). This reflects that most screen detected precursors are squamous, which are easier to detect, that is, it shows that squamous cancer precursors are better caught and prevented by screening.

The etiology and descriptive epidemiology of each carcinogenic stage can be viewed separately. With regard to the first stage, acute HPV infection, its epidemiology as just stated is that of a sexually transmitted infectious agent. The behavioral factors influencing risk are those that increase the chance of encountering an infected partner. There is no known innately immune state, although one could theoretically exist in a few individuals, without our knowledge.

HPV infection is a necessary precursor state to cervical cancer. Most HPV infections as detected by molecular (DNA or RNA) assays become undetectable after several months (21, 22). It is not known to which extent the lack of detectability represents viral clearance or persistence in some kind of latent stage (23). The subsequent necessary and proximal precursor state is HSIL linked to viral persistence. It is not known, by the way, whether long-term persistence precedes the first clonal expansion of (possibly regressing and never diagnosed) HSIL or vice versa. This is a theoretical point that is difficult to study; at the level of detection, persistence is more common than and precedes HSIL. Etiologic cofactors for persistence and progression to HSIL include viral, behavioral, and genetic host factors (Fig. 1).

By far, the most important viral factor is HPV type (24). The carcinogenic types of HPV are genetically related and found in several species of the alpha HPV genus. The established carcinogenic types include HPV16, HPV18, HPV31, HPV33, HPV35, HPV39, HPV45, HPV51, HPV52, HPV56, HPV58, HPV59, and HPV68 (25). Another dozen HPV types are possible cervical carcinogens and might account for a tiny fraction of cases of cancer. The most important HPV type is HPV16, which is responsible for only 20% of infections but which causes 40% of HSIL and half of cervical cancer (see below). HPV18 is next most important and is also preferentially responsible for adenocarcinoma. HPV18 is underrepresented in cancer precursors compared with its importance in cervical cancer (11). Viral genomic variation is so important for etiology that even subtle variations within viral types (called viral variants) influence risk of progression and invasion, with relative risks stronger than for behavior and genetic factors (26). While many other anogenital HPV types can cause LSIL and even a fraction of HSIL, the types listed above are the ones that can cause cervical cancer (Fig. 3).

Behavioral factors that approximately double the risk of HSIL among HPV-infected women include multiparity, long-term oral contraceptive use, and smoking (27, 28). These cofactors are firmly established by large case–control studies and prospective evidence, but it is not known why they are cofactors. Accordingly, multiparity might be a risk factor because of hormonal, trauma, or other mechanisms. The risk-inducing (hormonal) component of oral contraceptives is likewise unknown (29). It is even uncertain whether smoking acts via an immunosuppressing or genotoxic pathway. Less well-established cofactors for progression to HSIL given HPV infection include chronic cervical inflammation and immunosuppression (e.g., HIV).

Host genetic factors influencing control of infection certainly exist but are poorly understood. The only consistent association is with human leukocyte antigen (HLA), supporting the importance of T-cell responses in control of HPV infections and cervical precancers (30).

The natural history of CIN3 in terms of risk factors for invasion is not well-understood. An old natural history study of CIN3 progression shows that a minority of CIN3 invaded over the lifetime of patients (31). The factors associated with invasion remain elusive and cannot be directly studied, as prospective observation of CIN3 is unethical. Similarly, it is not understood which mechanisms lead to containment or even regression of CIN3, although immunological factors are presumed to play an important role. There are few identified risk factors for the transition between HSIL and cancer, besides time (32).

The natural history of a cervical cancer starts with a sexually transmitted carcinogenic HPV infection (Fig. 4; ref. 22). In the minority of infections, an ASC-US or LSIL cytologic abnormality is evident. Clearance of the HPV infection and associated LSIL is often rapid, with more than half of infections clearing (undetectable using standard DNA/RNA detection methods) within a year, and 90% of infections within approximately 2 years of acquisition. As the rate of clearance slows, the chance of development of HSIL gradually increases, representing the growth of a clonal high-grade lesion. In cohort studies, HSIL is diagnosed up to several years following HPV acquisition. HSIL lesions typically grow laterally around the circumference of the transformation zone, taking many years to decades before invasion, and accounting for the success of screening and secondary prevention of cancer.

The typical time course of the natural history leads to typical ages of each stage (Fig. 4) with the peak of HPV acquisition in adolescence and early adulthood, the peak of HSIL around 25 to 35 years and the peak of cancer from 45 to 60 years (33).

There are biomarkers associated with and reflecting each stage in natural history. Biomarkers for HPV-related diseases measure viral nucleic acids, viral proteins, or cellular factors altered by viral oncogenes (Fig. 5). The most important biomarker of HPV infection is the detection of HPV DNA. DNA tests, based on hybridization or PCR of a pool of carcinogenic types, have been the mainstay of HPV-related prevention efforts (34). Individual genotyping can predict risk of HSIL, as progression differs strongly by type. HPV16, in particular, predicts a greater prospective risk than other carcinogenic types. It tends to persist slightly longer (23), and predicts a substantially higher risk of HSIL and cancer (35–37). However, for clinical purposes, genotyping alone does not allow sufficient discrimination between transient infections and HSILs or cancers. Increased expression of HPV oncogene mRNA has been evaluated as a marker for the transition from a productive to a transforming infection. p16 is a cellular marker of HPV oncogene activity and has been evaluated for adjudication of ambiguous histology and as a cytological test for triage of women with abnormal screening results (14). Other cellular markers associated with HPV transformation are proliferation markers such as Ki-67, mcm-2, and top2a (38). Expression of HPV oncogenes leads to increasing chromosomal instability, even at the HSIL stage. Several regions, most importantly 3q and 5p, are often altered in HPV-related cancer precursors and can be detected by in situ hybridization assays. Recently, methylation of late HPV genes has been described in the transition from HPV infection to HSIL (39–41). Disease-specific biomarkers will be important to decide who among women testing positive for HPV needs referral to colposcopy. While there is increasing evidence that HPV oncogene mRNA and p16 could serve these needs, data for other markers are very limited.

The steps in a conventional cervical screening program are cervical cytologic screening, triage by HPV testing of equivocal (ASC-US) cytologic results, colposcopic referral of women with definitely abnormal or HPV-positive ASC-US results, biopsies of acetowhite lesions (those that turn white on application of acetic acid), and treatment by loop electrical excision procedure (LEEP) of the transformation zone if histologic HSIL is found. HSIL is treated to prevent cervical cancer; there is no specific treatment yet for HPV infection itself. Cervical cancer treatment is beyond the scope of this review of precursors.

New knowledge of the central etiologic role of HPV has led to 2 preventive strategies, vaccines and HPV-based screening that both act at the cancer precursor stage. Already available HPV vaccines are highly effective prophylactic agents against HPV infection and associated ASC-US, LSIL, and HSIL but are not therapeutic (42). Therefore, they are best administered to girls before onset of sexual activity (e.g., age, 11 years; range, 9–14 years). The bivalent vaccine (GSK) targets the 2 most important HPV types (HPV16 and HPV18), which together account for approximately 70% of cervical cancers and slightly more than half of HSILs (43). The quadrivalent vaccine (Merck) also prevents infections with HPV6 and HPV11, which together cause 90% of genital warts (condyloma acuminatum; ref. 44). It is important to realize that it is not essential that HPV vaccines (whose prophylactic efficacy has already been shown prospectively to last a decade) confer lifelong immunity. Cervical cancers in later life are mainly the result of HPV infections in earlier life. If sufficient population coverage is achieved, the early peak of HPV will be removed, the secondary peak in HSIL will not occur, and cervical cancer will inevitably be substantially reduced.

Despite the promise, vaccine uptake has been variable in wealthy nations and limited in the low-resource regions that are most in need. The available vaccines are expensive, require a cold chain, and are administered in 3 doses spanning 6 months. Thus, for a variety of practical and societal reasons (e.g., opposition to vaccination of young girls against a sexually transmitted agent, fear of vaccination), coverage in the United States has been lower than would be optimal from a public health perspective.

Next-generation vaccines are being tested. The closest to release is a nonavalent vaccine (HPV16, HPV18, HPV31, HPV33, HPV45, HPV52, HPV58, HPV6, and HPV11) that, if as efficacious as the current vaccines, would protect against 90% of cervical cancers as well as condyloma (45). Other novel vaccines are in development and have been reviewed elsewhere (46). It would be ideal to have a safe, therapeutic anti-HPV vaccine or other treatment but that is not yet available.

Until prophylactic HPV vaccine coverage is much more extensive, and time has passed to create an immune cohort of women (i.e., primary prevention), screening will remain the mainstay of cervical cancer (secondary) prevention. The main intent of screening is to identify HSIL and treat it to prevent cervical cancer morbidity and mortality.

Screening programs previously built around cytology alone are changing, prompted by our improved understanding of HPV natural history and the advent of HPV testing (17). We have already mentioned the nearly universal use of HPV testing in the United States to triage equivocal (ASC-US) cytologic results, the most common cytologic abnormality (∼5% of screens). The next phase is the incorporation of HPV testing into general screening, permitting the extension of screening intervals (47).

In Fig. 4, we show the evolution of screening strategies. In the era before the etiology of cervical cancer was understood, the public health community successfully convinced most U.S. women to participate in annual cytology screening. The annual Pap smear was effective programmatically, given the typically slow growth of HSILs, because repetition overcame the relative insensitivity of each round of screening. Cervical cytology remains a successful example of screening and can be credited with greatly reducing cervical cancer incidence and mortality. However, more focused screening is now possible.

Specifically, we now know that it typically takes many years for an HPV infection, even if persistent, to cause cervical cancer. We understand that there is a peak in HPV acquisition and LSILs in adolescent and young adult women, a secondary peak in HSILs some years later, and a rise in cancer many years later. Thus, it is not optimal to screen adolescent women, when HPV infection and LSIL are extremely common but the risk of cervical cancer is extremely low. Accordingly, the age of initiation of cytologic screening has been raised to 21 years, which is still conservative because cervical cancer rates do not start to increase appreciably until ages 25 to 30 years. The interval between cytologic screens has been increased to every 3 years. Beginning at the age of 30, past the peak of acute HPV infection, cotesting with HPV assays and cytology is preferred in the United States over cytology alone (47). In some other countries, for example, the Netherlands, there is a move toward primary screening with HPV testing alone, followed by cytology restricted to the HPV screen–positive women (48). For women testing HPV-negative, Pap-negative in the US, the recommended repeat screening interval is 5 years. This extension recognizes that too-frequent screening with sensitive HPV testing will mainly and nonspecifically pick up new HPV infections and associated LSILs, rather than HSILs. Screening now stops for women with normal screening histories at the age of 65 because new infections are rarely acquired at that age and, even if they are, the latency period until invasive cancer would typically exceed a woman's lifespan (47).

Because of the great sensitivity and negative predictive value (reassurance) provided by HPV testing, it is possible in low-resource regions to consider even more extended and cost-effective HPV-based screening (49). One or 2 rounds of HPV screening at the ages of 30 to 45 could theoretically reduce the burden of cervical cancer, with the proviso that there must be clinical resources available to manage the 5% to 20% or more of women, depending on region, who will test HPV-positive even at those ages.

To summarize, we now have the knowledge of cervical cancer precursors and tools to approach the virtual eradication of cervical cancer efficiently. From a primary prevention standpoint, girls should be vaccinated early to prevent the initial peak of carcinogenic HPV infections. For complementary secondary prevention, screening using HPV and cytology co-testing should be focused especially on the age of maximal incidence of easily treatable HSILs (ages 25–40 are critical), before the age of highest cervical cancer rates. Eventually, for women who follow the vaccination screening prevention strategy, it might be possible to assure extremely low risk of cervical cancer in later life, possibly permitting (and this would require the kind of medical record and patient follow-up that does not yet exist in most places) later initiation of screening, lengthening of screening intervals and cessation of prevention efforts even earlier than the age of 65 years (Fig. 4).

M. Schiffman has commercial research support from Qiagen and Roche. No potential conflicts of interest were disclosed by the other author.

Conception and design: M. Schiffman, N. Wentzensen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Schiffman

Writing, review, and/or revision of the manuscript: M. Schiffman, N. Wentzensen