Persistence of Betapapillomavirus Infections as a Risk Factor for Actinic Keratoses, Precursor to Cutaneous Squamous Cell Carcinoma

Human papillomaviruses (HPV) have oncogenic capacities and persistent infection with a high-risk mucosal HPV type from the α genus causes the development of cervical cancer (1, 2). HPV from the β genus (βPV) have cutaneous rather than mucosal tropism. They were first found to cause flat, wart-like lesions in epidermodysplasia verruciformis patients and so were formerly known as epidermodysplasia verruciformis types. Currently, 25 βPV types have been fully sequenced (HPV types 5, 8, 9, 12, 14, 15, 17, 19-25, 36-38, 47, 49, 75, 76, 80, 92, 93, and 96; ref. 3), but many more new types can probably be added to this list based on partial sequences (4).

βPVs are detected frequently in actinic keratoses and cutaneous squamous cell carcinomas (SCC) but have also been found in biopsies from normal skin, eyebrow hairs, and skin swabs from persons with and without skin cancer (5–7). Several studies have shown statistically significant associations between markers of βPV infection (viral DNA in eyebrow hairs or skin biopsies and antibodies in serum) and SCC (8–12), but no βPV types have been found to cause SCC in experimental systems.

There is limited knowledge about persistence of cutaneous βPV infection over extended periods. Studies of βPV persistence in eyebrow hairs thus far have involved relatively few participants followed over limited periods (2-6 years; refs. 13, 14), and no studies have examined βPV persistence in relation to future occurrence of actinic keratosis or SCC. Furthermore, most epidemiologic studies have been cross-sectional (11, 15), assessing both βPV detection in an individual's eyebrow hairs and their skin cancer status at the same point in time.

One of the few cohorts available for estimation of the risk of skin cancer in regard to HPV infection has been the Nambour Skin Cancer Study cohort (16, 17). Boxman and colleagues performed a cross-sectional study with 518 individuals sampled in 1996 and found a strong association between βPV infection and actinic keratoses in men but not in women (17). McBride and colleagues studied the prevalence of actinic keratoses in 291 participants of the Nambour Skin Cancer Study 7 years after single measurement of βPV DNA in eyebrow hairs. They found that βPV positivity was associated cross-sectionally with the presence of >10 actinic keratoses, especially in those ages >60 years, those who had fair skin, or those who had experienced high sun exposure (16).

In the present more detailed and longer-term study of the Nambour Skin Cancer Study cohort, we assessed the persistence of 25 different βPV types in eyebrow hairs over an 8-year period and then examined the longitudinal association between βPV persistence and the disease outcome, actinic keratoses, over a total period of 11 years.

Participants comprised an unselected subset of the study population of the Nambour Skin Cancer Study (Supplementary Figure) as described in detail previously (18–20). Briefly, in 1986, 2,095 of 3,000 randomly selected residents ages 20 to 69 years from Nambour, a subtropical township in Australia (latitude 26°S), participated in a skin cancer prevalence survey. From 1992 to 1996, 1,621 of these participants then took part in a trial of sunscreen application and β-carotene supplementation for the prevention of skin cancer, and follow-up continued until 2007.

In 1996, 507 participants took part in a substudy aiming to understand the association between HPV and skin cancer (17). Ten eyebrow hairs were plucked from each participant and processed as described below. In 2003, 291 of these 507 who were still actively participating in the follow-up of the Nambour Skin Cancer Study took part in a second phase of the HPV study when eyebrow hairs were plucked again (16). At baseline and in 2003, standardized questionnaires were used to elicit information about skin type and sun exposure.

In 1992, 1994, and 1996, specialist dermatologists conducted full-body skin examinations of study participants and documented the degree of telangiectasia and nuchal elastosis and the number of actinic keratoses on each anatomic site. A proportion of participants were fully examined for actinic keratoses again in 2007 (Supplementary Figure).

Ethical approval for all aspects of the study was obtained through Bancroft Centre Ethics Committee, Queensland Institute of Medical Research.

DNA from eyebrow hairs collected in 1996 was isolated according to a method described previously (21). The 2003 eyebrow hairs were treated using the QIAamp DNA mini kit (Qiagen). Briefly, hairs were pretreated overnight with proteinase K solution according to the manufacturer's instructions. After lysis with 200 μL AL buffer, half of the volume was stored at −70°C, whereas the other half was processed according to the manufacturer's instructions.

βPV detection and genotyping of 1996 and 2003 samples was done using a reversed hybridization assay as described previously (22). Briefly, PM-PCR was done in a final reaction volume of 50 μL, containing 10 μL of the isolated DNA, 2.5 mmol/L MgCl₂, 1× GeneAmp PCR buffer II, 0.2 deoxynucleotide triphosphates, 1.5 units AmpliTaq Gold DNA polymerase, and 10 μL of the PM primer mix. The PCR was done by a 9 min preheating step at 94°C followed by 35 cycles of amplification comprising 30 s at 94°C, 45 s at 52°C, and 45 s at 72°C. The PCR was ended by a final elongation step at 5 min at 72°C. As the positive PCR control, a βPV plasmid clone was included at an amount ∼100 times the limit of detection of the assay. All positive controls were detected. Each 10th sample was a negative isolation control that was processed parallel to the other samples, of which 12% was positive with no specific βPV type standing out.

All amplimers were subsequently analyzed with a reverse hybridization assay that permitted specific detection and identification of the 25 established βPV genotypes (5, 8, 9, 12, 14, 15, 17, 19-25, 36-38, 47, 49, 75, 76, 80, 92, 93, and 96). The reverse hybridization assay was done according to the manufacturer's instructions [RHA Kit Skin (β) HPV; Diassay].

χ² tests were used to analyze univariate associations. To examine factors associated with persistence, and the association between βPV infection/persistence and actinic keratoses, binomial models were used to compute relative risks (RR) and 95% confidence intervals (95% CI). A range of possible confounding factors were explored, but none of the factors (skin type, occupational sun exposure, sunburns, smoking, use of sunglasses, and randomized sunscreen/β-carotene interventions) altered the effect estimates and therefore were not included in the models. The final model describing associations with βPV detection over time was adjusted for age; the final models for assessing association with actinic keratoses were adjusted for age, sex, and the presence of actinic keratosis at baseline. Statistical analyses were done in SAS 9.1.

βPV DNA detection and typing information for the participants from 2003 was available for 171 of the 507 (34%) participants whose data were first analyzed in 1996, and these therefore formed the basis of this study. The characteristics of the 171 participants, whose mean age in 1996 was 50 years, were not significantly different from the remainder seen at baseline (data not shown). Half of the participants reported their skin type as fair and most had mainly indoor occupations. On dermatologic examination in 1996, moderate to extensive photo aging of the skin of the neck (nuchal elastosis) was prevalent in at least 50% of participants. Forty percent of participants had actinic keratosis present at baseline, ranging from 23% in the ages <40 years group to 61% in the ages >60 years group (Table 1). Of the 171, 138 (81%) were still active in the study in 2007 and underwent the final skin examination.

Of the 25 who were βPV negative in 2003, 11 (6% of all participants) were also βPV negative in 1996. Of the 171 subjects whose βPV was measured, 11 (6%) were βPV negative in 1996 and 2003. In 1996, 47 of 171 (27%) participants had no βPV DNA detected and were therefore deemed to be βPV negative. The 124 participants who were βPV positive (73%) had between 1 and 12 different types detected (median, 2), with 100 persons (81% of those positive) having multiple βPV types (Table 2).

In 2003, βPV was detected in 146 of 171 (85%) participants and the median number of detections was 2 (range, 1-14). Of those positive, 109 (75%) had more than one virus type detected. The total number of βPV type-specific detections in all 171 participants in 1996 was 413. Of these, 211 were also detected in 2003 (Table 2) and 202 were no longer detectable. In 2003, 279 viruses were detected, which had not been found in 1996.

The most prevalent types in 1996 were HPV23 (19%), HPV17 (18%), HPV38 (18%), and HPV93 (18%). HPV47 was not found in any samples from 1996 or 2003. The types detected most frequently in 2003 were HPV38 (26%), HPV17 (24%), HPV23 (19%), and HPV24 (19%; Table 2), and for the persistent types, HPV38 (11%), HPV24 (9%), and HPV93 (9%) were among the most prevalent types. HPV17, with prevalences of 18% and 24% in 1996 and 2003, respectively, was detected on both occasions in only 6% of participants. HPV types 92, 96, 80, 5, and 9 had the highest probabilities of persistence (Table 2).

Type-specific detection of βPV DNA in 1996 and 2003 was seen in 83 of 171 (49%) of participants. Of these, 32 (39%) had one type persisting and 51 (61%) more than one (range, 2-11).

The presence of βPV DNA in 1996 and type-specific persistence both increased with age, with those ages >60 years having a 1.5-fold (95% CI, 1.1-1.9) increased risk to have βPV DNA detected in 1996 and a 1.6-fold (95% CI, 1.0-2.8) increased risk to have at least one persistent βPV type as those ages <40 years. Sex, skin type, occupational sun exposure, number of lifetime sunburns, use of sunglasses, elastosis of the neck, and telangiectasia of the face were not associated with the presence of βPV DNA in 1996 or with persistence in 2003 (Table 3). No interaction was observed between βPV persistence and traditional risk factors for actinic keratosis, although with the modest sample size the power to detect interactions was inadequate.

In 2007, 76 of 138 (55%) participants who underwent a skin examination had at least one actinic keratosis on the head and neck, and 94 of the 138 (68%) participants had at least one actinic keratosis on the whole body. Twenty-seven of 138 (20%) participants had >10 actinic keratoses on the whole body. The presence of actinic keratoses in 2007 was associated with being ages >60 years (RR, 1.7; 95% CI, 1.2-2.4), male sex (RR, 1.4; 95% CI, 1.1-1.8), the presence of actinic keratosis at baseline (RR, 1.9; 95% CI, 1.5-2.5), having moderate (RR, 1.8; 95% CI, 1.1-2.9) or extensive (RR, 2.4; 95% CI, 1.5-3.8) solar elastosis of the neck, and high degree of telangiectasia of the face (RR, 1.5; 95% CI, 1.0-2.3).

Table 4 shows the associations between actinic keratoses diagnosed in 2007 and the presence of βPV DNA in 1996, 2003, and at both time points, adjusted for age, sex, and actinic keratosis diagnosed at baseline. Having βPV detected at a single time point (in either 1996 or 2003) was not associated with actinic keratosis in 2007. βPV type-specific persistence, however, was associated with actinic keratoses on head and neck (RR, 1.4; 95% CI, 1.0-1.9). When we repeated the analyses with only the 77 actinic keratosis–free persons at baseline, the results were essentially identical although with slightly wider confidence intervals. Associations between actinic keratoses diagnosed in 2003 and βPV DNA persistence were of similar magnitude but nonsignificant (data not shown). Strength of associations did not vary with increasing numbers of prevalent actinic keratosis (data not shown).

We studied the persistence of βPV DNA in eyebrow hairs over an 8-year period in a broadly representative sample of adults from an Australian community. We then assessed prospectively whether such long-term βPV persistence was a risk factor for the development of actinic keratoses after adjustment for the presence of actinic keratoses at baseline. Among 171 persons ages 30 to 79 years in 1996, 73% were βPV-positive for at least one type in 1996 and 85% at follow-up in 2003. These prevalences were higher than found in earlier studies in this cohort (16, 17) probably because of the newer, more sensitive method of βPV detection used here. Half of the participants had at least one βPV type-specific persistent infection, 67% of the persons who tested positive in 1996. Of the total βPV infections at baseline in all participants, half of the infections appeared to persist over the 8-year follow-up period. The likelihood of βPV DNA being both present and persisting 8 years later increased with age. Except for older age, no other factors were associated with βPV persistence. We postulate that this association of βPV infection and persistence with older age reflects natural deterioration of the immune system with age, known as immune senescence (23). Alternatively, contraction of de novo βPV infections with increasing age might cause the observed association, but we consider this option less likely.

McBride and colleagues performed a study in the same population, assessing actinic keratosis risk 7 years after baseline βPV status had been established (16). We have not only extended this period to 11 years but also assessed association between persistent βPV infection and actinic keratoses. In the current study, we did not find the effect of βPV presence in 1996 on the risk of actinic keratoses that McBride and colleagues found nor associations between actinic keratosis and risk factors other than age. We also did not find different associations between βPV infection and actinic keratoses in men and women, as Boxman and colleagues found in 1996 (17). Again, these differences may be due to our use of a βPV test that was able to detect the common low copy number infections. The different cohort sizes and time intervals measured may also have played a role.

Indeed, a potential limitation of this study was the different methods used for DNA isolation of the 1996 and the 2003 samples. Because recent research shows that different isolation methods can give different end-results (24), it is possible that the newer and more sensitive method used in 2003 partly also explains the higher number of infections that were found in 2003 than in 1996 (490 versus 413). On the other hand, based on previous findings, it was to be expected that the prevalence of βPV increased with age (15, 25). Further, although we have referred to the presence of a virus at both time points as “persistent,” it is possible that infections are labile and the virus was not present through the entire period. The fact that 12% of the negative control samples tested βPV positive, with no dominant type, did not affect our conclusions, as this would have resulted in underestimation of βPV persistence. Finally, βPV can occur in very low copy numbers (26), so it is conceivable that some viruses were still present in 2003 but were below the detection limit.

There is little knowledge about the long-term persistence of βPV infections. A previous study on βPV persistence in eyebrow hairs among 23 healthy volunteers ages 21 to 64 years in The Netherlands showed that 48% of βPV type-specific infections in 74% of the study participants were persistent over a period of 6 months (13). We found lower persistence (30% of βPV infections in 49% of the participants) after 8 years of follow-up. Apart from our longer follow-up time, the smaller number of βPV measurements, the different DNA purification method, and the different geographic region of the Australian study group, compared with the Dutch, are all likely to have influenced the comparability of these results.

A small study in 42 immunocompetent persons found that 48% who were positive at baseline had a persistent infection after 5 to 7 years of follow-up (14). In our study, the viral persistence among those who were βPV-positive at baseline was 67% (83 of 124). This difference again might be explained by our larger study group, the different geographic region, and the different method used for βPV detection.

With regard to specific βPV types, we observed that the more prevalent a βPV type was, the more likely it was to persist, although one frequently occurring type (HPV17) was only found to persist in 6% of participants. Our overall finding accords with studies concerning high-risk mucosal HPV persistence in women (27), where the more prevalent HPVs persist more often. Although comparisons between mucosal and βPV are interesting, the lack of knowledge about the pathophysiology of βPV infections and about the role of the immune system in clearance of these infections does limit conclusions that can be drawn. In general, however, βPV persistence appears much more common than persistence of mucosal HPV infections in the (ano)genital tract (27). To what extent βPV persistence increases the risk of cutaneous SCC was not determined in this study, because the number of incident SCC cases was too small. Similarly, we could not identify any βPV types that unequivocally increased the risk of developing actinic keratosis.

The finding that long-term βPV infection in eyebrow hairs is related to subsequently developing actinic keratoses on the skin of the head and neck has not been reported previously. We observed no association with whole-body actinic keratoses in those with the long-term persistent βPV infections. The association between βPV persistence and actinic keratoses on the head and neck but not with actinic keratoses on the whole body might indicate that βPV in eyebrow hairs is a better marker of infection on the face, neck, and scalp than on the rest of the body, despite a recent study showing that eyebrow hairs are a good marker of cutaneous infection on the body (5, 28–30). Previous studies of the association between actinic keratosis and βPV did not examine head and neck actinic keratosis in isolation; therefore, comparisons cannot be made.

Overall, we conclude that persistence of βPV infection in eyebrow hairs over a long period is common. Our results indicate that persistent infection with βPV is a risk factor for the development of actinic keratosis; whether it is a risk factor for SCC needs to be determined. In contrast, transient infection did not alter the risk of actinic keratosis. Although these findings accord with persistent mucosal HPV infections and increased risk for cervical intraepithelial neoplasia and cervical cancer, there are many uncertainties and dissimilarities between mucosal and cutaneous HPVs and their role in tumorigenesis. The role of βPV in the development of cutaneous tumors and the importance of duration of βPV infection require ongoing investigation.

No potential conflicts of interest were disclosed.

Grant support: Leiden University Fund/Van Walsem and the Foundation “De Drie Lichten” travel grants (E.I. Plasmeijer), National Health and Medical Research Council (Australia) Career Development Award (R.E. Neale), and The Netherlands Organization for Health Research and Development, ZonMW, clinical fellowship grant 907-00-150 (M.C.W. Feltkamp).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Peter O'Rourke and Kylie-Ann Mallit for providing statistical advice.