Increased Prevalence of Human Papillomavirus in Hairs Plucked From Patients With Psoriasis Treated With Psoralen–UV-A

Background  Patients with psoriasis treated with psoralen–UV-A (PUVA) are at increased risk of skin cancer ; however, the exact causes of this increasedincidence are not well understood. It has been suggested that PUVA may increase expression of the tumorigenic agent human papillomavirus (HPV) in skin bydirectly stimulating virus replication, immune suppression, or both, thereby leading to skin cancer formation.

Objective  To determine whether HPV DNA prevalence in the skin is increased after long-term PUVA treatment.

Design  Screening for the presence of HPV sequences in DNA isolated from plucked body hairs of patients with psoriasis with a history of PUVA exposure anda history of skin cancer (group A), PUVA exposure and no history of skin cancer(group B), and no PUVA exposure and no history of skin cancer (group C).

Setting  University hospital.

Patients and Methods  Hair samples were obtained from 81 patients with psoriasis (56 men and 25 women; mean age, 52 years), including 16 in group A (mean number of PUVAexposures, 702), 35 in group B (mean number of PUVA exposures, 282), and 30 in group C. DNA was isolated from the hair samples and analyzed by polymerasechain reaction with the use of 2 nested primer systems specific for epidermodysplasia verruciformis–associated or related and genital or mucosal virus types,respectively.

Results  The rate of HPV DNA positivity was significantly higher in groups A (73% [11/15]) and B (69% [24/35]) than in group C (36% [10/28]) (A + B vsC, P = .009; χ² test; age adjusted).

Conclusion  The prevalence of HPV in the skin (hair follicles) is increased in patients with psoriasis who have a history of PUVA exposure.

Although psorlaen– UV-A (PUVA) treatment of psoriasis has been unambiguously linked with an increased risk of developing skin cancers includingsquamous cell carcinoma (SCC) (for review see Stern et al¹), the exact causes of the increased incidence are not well understood. Severalexplanations have been proposed. One theory is that PUVA, which is mutagenic and carcinogenic, may directly initiate skin cancer in patients with psoriasis,perhaps by mutating the tumor suppressor gene p53,^2-4 the INK-4a-ARF locus,⁵ and/or the proto-oncogene Ha-ras.⁶ A secondtheory is that PUVA treatment, being immunosuppressive,^7,8 may promote the growth of skin cancers that are induced either by itself or byother suspected or known carcinogenic treatment agents, including UV-B, ionizing radiation, methotrexate, topical tar, and arsenic (for review see Maier etal⁹). A third theory is that PUVA may promote tumorigenesis by acting as a (co)-factor for tumorigenic viral agents suchas human papillomavirus (HPV) (see several reviews^10-13). In such a scenario, skin carcinogenesis may depend on the ability of the earlyoncoprotein E6 of cutaneous HPVs to inhibit apoptosis on DNA damage^14,15 induced by agents such as UV-B or PUVA. Psoralen–UV-A may act similarly to UV light in stimulating HPVinfection via (1) direct stimulation of virus replication and/or (2) immune suppression, the result being increased amounts of viral DNA in the skin.¹⁶

The HPVs appear to be closely linked to skin cancers. In one study, a specific group of closely related HPV types, including HPV-5 and HPV-8,were isolated from more than 90% of SCCs from patients with epidermodysplasia verruciformis (EV).¹⁷ Types of HPV associatedwith EV have been detected in nonmelanoma skin cancers and normal skin from immunosuppressed renal transplant recipients^18-21 and immunocompetent individuals.^21-24 Of particular interest for the present study is that several reports have indicateda possible link between HPV and skin cancer formation in PUVA-treated patients with psoriasis.^25-30 Forinstance, Harwood et al²⁹ found HPV DNA in approximately 75% of nonmelanoma skin cancers from PUVA-treated patients andnoted that the majority of HPV-positive samples contained EV-related HPV, including types 5, 20, 21, 23, 24, and 38. Moreover, EV HPV DNA sequences have been found to be highly prevalent in lesional and nonlesional skin scrapings³¹ and lesional skin biopsy samples³² from patients with psoriasis. Especially important is a previous study in whichEV HPV DNA was detected in more than 90% of plucked hairs from immunosuppressed renal transplant recipients.^33,34 Inlight of these data, we used polymerase chain reaction (PCR) analysis, DNA sequencing, and in situ hybridization techniques to test the hypothesis thatimmunosuppressive PUVA treatment^7,8 may lead to an increased prevalence of HPV in the skin.

Three groups of patients (N = 81; 56 men and 25 women) were recruited from the networks of patients treated for psoriasis at the Department of Dermatologyof Karl-Franzens-University and at the Outpatient Dermatology Unit of the Regional Social Insurance Office of the State of Styria in Graz, Austria.Group A patients (n = 16; 9 men and 7 women) had a history of PUVA exposureand a history of at least 1 skin cancer; group B patients (n = 35; 24 menand 11 women) had a history of PUVA exposure and no history of skin cancer; and group C patients (n = 30; 23 men and 7 women) had no history of PUVA exposureand no history of skin cancer (Table 1). The mean age of patients in group A was 64 years (range, 55-76 years); ingroup B, 54 years (range, 22-83 years); in group C, 42 years (range, 20-78 years). The mean number of PUVA exposures was 702 (range, 82-1430) in groupA and 282 (range, 5-632) in group B. The mean total UV-A dose was 3823 J/cm² (range, 222-8580 J/cm²) in group A and 1298 J/cm² (range, 5-4384 J/cm²) in group B. All PUVA-treated patientshad received oral administration of methoxsalen and/or 5-methoxypsoralen forPUVA treatment. Group A included 8 patients with a history of SCC, 2 with a history of basal cell carcinoma (BCC), 5 with a history of combined SCCand BCC, and 1 with a history of combined SCC, BCC, and malignant melanoma. At the time of study, all patients were free of skin cancer and not receivingongoing immunosuppressive therapy. The patients' skin phototype and history of exposure to UV-B treatment and other potentially carcinogenic treatmentmodalities (ie, methotrexate, cyclosporine, arsenic, and x-ray) are listed in Table 1. Before hair sampleswere collected, informed consent was obtained from each subject.

Hairs were plucked from nondiseased skin (as distant as possible from psoriatic plaques if present) on the arms, legs, and trunk of each subject.New tweezers were used for each subject. Only hairs containing intact hair follicles (at least 2 hairs per body site, or 6 hairs per subject) were collected,pooled, snap-frozen, and stored at −70°C for further processing and analysis.

To disclose the exact location of HPV in the skin, we used in situ hybridizationand histologic techniques to examine archived paraffin-embedded skin tissuesamples from patients in our study whose hair DNA tested positive for the presence of HPV. By cross-checking our list of study patients against a computerizedarchival data bank, we were able to identify 7 tissue samples taken from patients after first PUVA treatment and stored in the archives of the HistopathologyUnit of the Department of Dermatology, University of Graz. Of those 7 samples, 3 (from patients A14, B12, and C7) harbored hairs, including 2 psoriatic lesionsand 1 seborrheic keratosis. As a normal control, we also obtained from the archives a paraffin-embedded, hair-harboring skin tissue sample adjacent toa BCC surgically excised from the temple of a 36-year-old Austrian woman with EV.

The DNA was extracted from the snap-frozen hairs by means of a commercial forensic DNA extraction kit (InViSorb Forensic Kit I; Invitek GmbH, Berlin,Germany). The DNA was extracted from 5-µm-thick sections of formalin-fixed, paraffin-embedded tissue specimens as follows: specimens were deparaffinizedby xylene and ethanol; scraped off after air drying; suspended in digestion buffer containing proteinase K, 1 µg/µL , in 0.1M Tris hydrochloride,pH 8.0; incubated overnight at 55°C; and then kept for 10 minutes at 95°C for heat-inactivating proteinase K. The DNA samples were stored at −20°Cuntil used.

Before HPV screening, all DNA preparations isolated from clinical specimens were tested for their quality by amplification of a 209–base pair fragmentof the cellular β-globin gene as described by de Roda Husman et al.³⁵ Only β-globin–positive samples were subjected to further analyses. The HPV sequences were detected by means of 2 nestedPCR approaches that used degenerated primer sets CP62/CP70:CP65/CP69 and A5/A10:A6/A8 specific for a broad range of cutaneous or EV-associated and mucosal or genitalvirus types as described by Boxman et al³³ and Wieland et al,³⁶ respectively. Standard anticontaminationprecautions were taken during all experiments.³⁷

The PCR products were resolved by electrophoresis in 2.5% agarose gels. The DNA fractions of the expected size were excised from the gel slabs. Theamplified sequences were then purified by means of a kit (QIAquick Gel Extraction Kit; Qiagen GmbH, Hilden, Germany) and cloned in a vector (pCR-Blunt II-TOPO)(Invitrogen, Breda, the Netherlands). Depending on size variations in the cloned PCR products caused by EcoRI digestion, 3to 6 recombinant plasmid clones were subjected to final sequence analysis in each case. Sequencing was performed with the Taq FSBigDye sequencing kit (PE Biosystems, Weiterstadt, Germany) terminator cycle system with the use of an automatic sequencer (ABI Prism 377; PE Biosystems).

Sequence analyses were performed with BLAST 2.1.3. software (National Center for Biotechnology Information, National Institutes of Health, Bethesda,Md)³⁸ and MacVector 7.0 software (Oxford Molecular Group PLC, Oxford, England). The sequence databases accessed for comparisonwere EMBL (European Molecular Biology Laboratory), GenBank, DDBJ, and PDB (Protein Data Bank). The HPV typing was performed according to the 90% borderlinerule to determine sequence homology between recognized HPV types and the putative new types within the amplified fragments of the viral L1 gene.³⁹

To generate labeled, patient-specific, double-stranded HPV probes, the PCR products generated from DNA extracted from paraffin-embedded skin samplesof individual patients with plucked hairs positive for HPV DNA were reamplified with the same pair of primers in a reaction cocktail that now included digoxigenin(DIG) 2′-deoxyuridine 5′-triphosphate (Boehringer Mannheim, Mannheim, Germany). For in situ hybridization, tissue sections mounted on silanizedglass slides were deparaffinized, encircled by means of a hydrophobic slide-marking pen (PAP PEN; DAKO, Vienna, Austria), rehydrated in phosphate-buffered saline,and then permeabilized by treatment with 0.1% protease type XXIV. For hybridization, the PCR-generated DIG-labeled probe in a buffer consisting of 2 × SSC,50% formamide, 10% dextran sulfate, 1-µg/µL transfer RNA, 0.1-µg/µLsalmon sperm DNA, and 1% DIG-blocking reagent (Boehringer Mannheim) was appliedto tissue sections. After denaturation of the sections for 3 minutes at 95°Cand subsequent hybridization overnight at room temperature, detection of hybridizedDIG-labeled probe was performed with a kit (DIG Nucleic Acid Detection kit; Boehringer Mannheim) according to the manufacturer's instructions. After colordevelopment with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate solution, tissue sections were counterstained with nuclear fast red (Kernechtrot;Merck & Co, Inc, Vienna, Austria). As negative controls, tissue sections were hybridized with different HPV-unrelated probes (to assess backgroundstaining) and with hybridization buffer containing no labeled probe.

The exact χ² test or Fisher exact test was used to determine the statistical significance of differences in HPV DNA hair positivity amongthe different groups of patients. Odds ratios and exact 95% confidence intervals were calculated for HPV DNA hair positivity for PUVA-treated vs untreatedpatients. Age adjustment of odds ratios was achieved by dividing the patients into 4 age groups according to the quartiles of age group C and includingthe groups as confounding factor into a logistic regression model. A 2-sided P<.05 was considered to indicate a statistically significant difference between groups.

The DNA of hair samples from 81 patients with psoriasis was screened for the occurrence of HPV-sequences by means of 2 nested primer systems specificfor EV-associated or related and genital or mucosal virus types, respectively. Regardless of skin cancer history, patients from the 2 PUVA-treated groups(groups A and B) had relatively higher rates of HPV DNA positivity (73% [11/15] and 69% [24/35]) than did non–PUVA-treated patients in group C (36%[10/28]) (Table 1, Figure 1). These differences were statistically significant both before and after age adjustment (A + B vs C, P =.003 and P = .009, respectively; χ² test), indicating that age was not a significant risk factor in the regression model(P = .84). The mean number of PUVA treatments and of UV-A doses was higher for HPV-positive, PUVA-treated patients than forHPV-negative, PUVA-treated patients (527 vs 436 treatments and 2766 vs 2033 J/cm² UV-A, respectively). The HPV positivity was not significantlyrelated to sex, skin phototype, UV-B treatment, or other systemic risk factors, including cyclosporine, methotrexate, and arsenic exposure (data of statisticalanalysis not shown). Samples from 1 patient in group A (patient A5) and 2 patients in group C (patients C11 and C13) did not show the presence of intactDNA after β-globin control PCR and thus were excluded from the statistical analysis of HPV DNA hair positivity. All HPV-positive samples were disclosedby DNA sequencing after PCR with the degenerated primer sets CP62/CP70:CP65/CP69.No single HPV-positive sample could be identified by PCR with the A5/A10:A6/A8(mucosal or genital HPV) primer sets.

Sequencing of PCR products and subsequent sequence analysis showed the presence of a variety of different HPV types in the analyzed hair samples(Table 1 and Table 2). Ten recognized HPV types (HPV-5, -14d, -17, -24, -25, -37, -38, -51, -61, and -80) and 20 presumably new HPV types could be identifiedamong sequences isolated from the tested samples. The majority of sequences (68% [19/28]) belonged to the phylogenetic B1 group of HPVs, which containsEV-associated and genetically related cutaneous papillomaviruses.⁴⁰ Known genital or mucosa–specific HPV types (HPV-51 and -61 of groups A5 and A3) were detected in only 2 cases, both involvingcoinfection with B1-virus types. Infection with 2 different HPV types was found in 14 (31%) of the 45 HPV-positive hair samples; infection with 3 HPVtypes was found in 1 (2%). However, there was no relationship between infection with 2 or more HPV types and history of PUVA treatment and dose. The mostprevalent HPV type in all tested samples (per total number of HPV DNA sequencespresent) was HPV-38 (15% [9/61]), followed by HPV-25 (8% [5/61]) and IA16(8% [5/61]) (Table 2). Interestingly, HPV-38 sequences were found only in hair samples from PUVA-treated patientsfrom groups A and B (33% [5/15] and 11% [4/35] of samples, respectively) and not at all in samples from the non–PUVA-treated patients of group C(0/28) (Table 1, Figure 1). This difference in HPV-38 DNA hair positivity was statistically significant (P = .02; Fisher exact test). However,among PUVA-treated patients (groups A and B) there was no statistically significant difference in the mean number of PUVA treatments and UV-A dose among HPV-38–positiveand –negative patients. Apart from the differences in HPV-38 prevalence, there were no significant differences in HPV type prevalences among the differentgroups of patients.

In the hope of determining the exact location of HPV in hair samples, 3 paraffin-embedded, hair-containing skin samples (including a psoriatic lesion[containing HPV-25] from patient A14 in a biopsy specimen taken 4 years before this study; a psoriatic lesion [containing HPV RTRX9] from patient B12 ina biopsy specimen taken 12 years before this study; and a seborrheic keratosis [containing RTRX7] from patient C7 in a biopsy specimen taken 6 months beforethis study) were subjected to in situ hybridization with PCR-generated labeled probes specific for HPV(s) residing in the hair samples. The experiments,however, failed to demonstrate HPV-specific nuclear staining in either hairs or lesional skin (data not shown). This was also the case after attempts atin situ hybridization with the use of mixtures of different patient-specific HPV probes. For control purposes, in situ hybridization was also performedon sections from a paraffin-embedded, hair-harboring normal skin tissue sampleadjacent to a BCC surgically excised from the temple of a 36-year-old Austrianwoman with EV. Sequencing of DNA extracted separately from the BCC as well as from the adjacent, hair-harboring normal skin of this patient had shownthe presence of HPV-5 and HPV-25 sequences. In this case, in situ hybridizationwith a PCR-generated DIG-labeled HPV-5 probe demonstrated focal nuclear staining of cells in the infundibular area of hairs and the adjacent peri-infundibular epidermis (Figure 2). Interestingly,skin samples from all patients with psoriasis and from the patient with EV exhibited strong cytoplasmic staining in cells of the glycogen-rich outerroot sheath in the stem and bulb area of hair follicles (data not shown). However, in situ hybridization with HPV-unrelated probes showed that thisstaining was most likely due to nonspecific binding (data not shown). In situ hybridizations that excluded labeled probes consistently gave negative resultsfor all tissue specimens examined.

It is well recognized that infections with cutaneous, EV-associated HPVs are widespread in humans and that body hair follicles may represent asilent reservoir for these viruses.^16,33,41,42 Premalignant skin lesions, cutaneous SCCs, and psoriatic skin scrapings and biopsy specimenshave all shown a particularly high prevalence of EV HPV DNA, and this points to the possible involvement of HPV in the etiology of psoriasis as well astumorigenesis.^32,43 In the present study, we found that the prevalence of HPV in plucked body hairs was significantlyhigher in PUVA-treated patients with psoriasis (with or without a history of skin cancer) than in non–PUVA-treated patients (Figure 1). We are tempted to interpret this observation in termsof PUVA-induced immunosuppression,^7,8 which, like UV exposure, may favor the survival of HPV-infected keratinocytes, orin terms of a direct stimulating influence of PUVA on HPV viral activities. In regard to the latter interpretation, UV light has recently been shown todifferentially regulate the promoters of a number of cutaneous HPVs.^44,45 This is especially interesting because the prevalence of HPV DNA in hairs of PUVA-treated patients with psoriasisin our study (73% in group A and 69% in group B) nicely corresponds to that of healthy Australian individuals (ie, 66%)¹⁶ whoare on average exposed to substantial cumulative UV doses, whereas the prevalence of HPV DNA in hairs of non–PUVA-treated patients in our study (36%)is similar to that in healthy Europeans (ie, 45%).³³ Treatment with PUVA seemed to coincide in particular with the elevated prevalence ofHPV-38 in plucked hairs. In contrast, the overall prevalence of other identified HPVs remained roughly constant in all patient groups (Table 1). This may suggest an especially high responsiveness ofHPV-38 to PUVA treatment. Interestingly, in a previous study,³² HPV-38, which was originally described in malignant melanoma,⁴⁶ wasfound to occur quite frequently (ie, at a rate of 24%) in lesional skin samples from patients with psoriasis. If differences between PUVA treatment in ourpatients and sun exposure in healthy individuals are taken into account, then the comparison of overall HPV DNA prevalence in plucked hair samples of patientswith psoriasis from this study and in healthy individuals from previous studies^16,33 tentatively argues against the idea that patients with psoriasis are more susceptible than healthy individualsto HPV infection of hair follicles.

The DNA sequencing showed that the majority of viruses infecting the tested samples belonged to the phylogenetic B1 group of papillomaviruses (EV-associatedor genetically related HPV types),⁴⁰ including not only recognized HPV genotypes but also a number of putative new types(Table 1 and Table 2). Only in 2 cases were known genital or mucosa-specific HPVs (HPV-51 and HPV-61) identified in hair samples from the tested patients.This finding is in line with results of the studies by Boxman et al,^33,41 who reported on the absence of HPV-6/11 from plucked eyebrow hairs but detected EV-associated and typical genitalpapillomaviruses in 62% and 43% of pubic and perianal hairs, respectively. These findings suggest that, in comparison with genital HPVs, EV viruses areclearly more able to colonize hair follicles latently at different areas of the human skin. The HPV type most frequently found in our present series wasHPV-38, representing 15% of all HPV sequences detected (Table 2, Figure 1). This prevalence of HPV-38 in hair samples corresponds very closely to that reportedby Boxman et al¹⁶ (17%) for the Australian population.

Whereas oncogenic HPV-5 and HPV-36 were the most prevalent virus types identified in psoriatic lesional skin specimens in 2 other studies,^31,32,47 they were relativelyrare in the present study. The reasons for this difference remain unclear, but they may involve the geographic origin of the patients, sample type (ie,plucked hairs vs scraped scales), and/or differences in the molecular detection techniques used. Indeed, the wide array of PCR-based techniques used in previousstudies of HPV DNA detection in skin makes it very difficult to directly compare results among different studies. For instance, a limitation of our study isthat the primer set CP62/CP70:CP65/CP69 used to detect EV HPV types does not sensitively detect HPV types of genuses A4 and A2 and subgroups B2 and E.¹⁹ However, these HPV types have been previously foundin PUVA-associated skin lesions as well.²⁹ Interestingly, in the study by Weissenborn et al³² cited atthe beginning of this paragraph, the overall HPV detection rate and the HPV-38 rate did not significantly differ between PUVA-treated and untreated patients,but infections with multiple HPV types could be detected almost 5 times more frequently in psoriatic lesions of PUVA-treated patients. In addition, thedetection rate of HPV-5 was significantly higher in the psoriatic lesions of PUVA-treated patients than in untreated patients (80% vs 42%, respectively).However, in the study by Favre et al,³¹ there was no significant difference in HPV-5 detection rates between PUVA-treatedand untreated patients. Weissenborn et al³² suggested that, given the high overall prevalence of HPV-5 of more than 90% in bothPUVA-treated and untreated patients, it probably would have been impossible for Favre et al³¹ to find a statistically significantincrease after PUVA treatment. Also, in the former study³² the total doses of PUVA were much higher than in the latter.³¹ TheHPV spectrum in the present study was similar to the spectrum of HPV typesobserved in healthy individuals in other studies. For instance, no instanceof HPV-5 infection and only 4 instances of HPV-36 infection could be identifiedin 93 PCR-positive hair samples from the Australian population.¹⁶ Moreover,a specific search for HPV-5 by means of a highly sensitive, type-specific nested PCR showed HPV sequences in only 16% of tested hair samples from immunocompetentsubjects.⁴¹

To localize exactly HPV in the skin of patients with psoriasis, we used in situ hybridization but were unable to detect specific signals (data notshown). In contrast, in situ hybridization with an HPV-5–specific probeperformed for control purposes on sections of a paraffin-embedded, hair-harboringnormal skin tissue sample from a patient with EV showed focal nuclear stainingof cells in the infundibular area of hairs and the adjacent peri-infundibularepidermis (Figure 2). This is, to the best of our knowledge, the first localization of HPV DNA to human hairfollicles. The failure to detect HPV sequences in hair-harboring psoriatic skin may well be due to low sensitivity of the DIG-labeling technique used.This assumption is supported by recent experiments on psoriatic skin lesions using a real-time PCR approach, the results of which point to HPV DNA loadvalues in the range of 1 genomic HPV copy per 1 to 12 000 cells (S. J. Weissenborn, PhD, unpublished observations, 2002).

Taken together, the results of the present study indicate that long-term PUVA exposure increases the prevalence of HPVs, and specifically HPV-38, innonlesional skin (hair follicles) of patients with psoriasis. This is an intriguing finding because HPV-38 has been previously detected in PUVA-associated SCCin a study²⁹ from the United Kingdom and has been found to be the most prevalent HPV type (50% [4/8]) in HPV-positive SCCof patients with psoriasis from our clinic in Austria (P.W., P.G.F., unpublished observations, 2002). Further studies are thus warranted to determine whetherthe increased prevalence of HPVs, including HPV-38, in PUVA-treated nonlesional skin and the presence in particular of this virus type in PUVA-associatedSCC is coincidental to or directly involved in tumorigenesis.

Corresponding author: Peter Wolf, MD, Department of Photodermatology, Department of Dermatology, Karl-Franzens-University, Auenbruggerplatz 8, A-8036,Graz, Austria (e-mail: peter.wolf@kfunigraz.ac.at).

Accepted for publication April 17, 2003.

This work was supported by grants 7285 and 8682 from the Austrian National Bank Jubilee Fund, Vienna (Dr Wolf), and from the Cologne Center for MolecularMedicine, Cologne, Germany (Drs Pfister and Fuchs). Dr Seidl was supported as a postdoctoral fellow by grant 12383-GEN from the Austrian Science Foundation,Vienna (Fonds zur Förderung der Wissen Schafflichen Forschung).

We thank Arnold Gerger, MD (Outpatient Dermatology Unit of the RegionalSocial Insurance Office of the State of Styria, Graz, Austria), for referringpatients and Alexandra van Mil (Cologne Center for Molecular Medicine) for her excellent technical assistance.