The most common malignancy in the worldwide Caucasian population is nonmelanoma skin cancer with the 2 main histological types squamous cell carcinoma (SCC) and basal cell carcinoma (BCC).1 BCC is the more prevalent of the two and metastasizes extremely rarely2, 3 while SCC is more prone to metastasize and cause death.4, 5 Keratoacanthoma (KA) is a common neoplasm of the skin that has such strong resemblance to SCC that it is controversial whether it is a separate benign entity or a benign variant of SCC.6 KA is typically characterized by rapid growth (6 weeks) that self resolves within a year.6 In spite that most of the lesions diagnosed as KA are of benign nature, aggressive behavior such as metastasizing has been documented.7 As there are currently no biomarkers to distinguish SCC from KA, the KA lesion is commonly treated as a SCC.7 Advantages in distinguishing the two diagnoses would bring about fewer unnecessary surgeries, lower burden on the healthcare system and most importantly less anxiety for the patients. Over the years, elaborate work of investigating possible biomarkers for the distinction of KA from SCC has been performed, including studies of p53,8 proliferating cell nuclear antigen,9 angiotensin type I receptor,10 vascular cell adhesion molecule, intracellular adhesion molecule,11 telomerase, cyclo-oxygenase-2,12 KI-67,13 p16 tumour suppressor antigen,14 matrix metalloproteinases,15 syndecan-116 and Bcl-2.17 Still, none of these studies have provided specific or reliable distinguishing criteria.
The major risk factor for development of skin cancer is exposure to the UV light.18 However, other factors such as immunosuppresion, fair complexion, older age, male gender and exposure to certain chemical carcinogens are also implicated.19
Furthermore, tumors at sun-exposed sites of particularly immunocompromised patients commonly harbor human papillomavirus (HPV).20 HPV is a common virus that is also frequently detected in normal skin.21 Some HPV types of the Betapapillomavirus species 2 seem to be more common in SCC.22, 23 HPV is also found in KA lesions, although there are hitherto no specific associations between HPV and KA reported.24, 25
The tissue microarray (TMA)26 technique enables simultaneous immunohistochemical staining of a large number of patient tissue specimens. As all specimens are arrayed on the same slide, the interassay variability of immunohistochemical stainings is minimized. In this study, we applied the TMA technique in order to screen for possible biomarkers distinguishing KA from SCC. Nineteen potential biomarkers (Table I) were selected for study, due to their involvement in different aspects of carcinogenesis such as the immune response, apoptotic pathways, proliferation and cell cycle regulation.
Table I. Overview of the Investigated Antigens
Marker commonly used to distinguish T cells
Marker commonly used to distinguish B cells
Marker commonly used to distinguish mononuclear phagocytes
Induce proliferative and antiapoptotic gene products
Inhibitor of p65
Induce proliferative and antiapoptotic gene products
Tumor suppressor, induce cell cycle arrest and apoptosis
TGFβ receptor associated protein with role as inhibitor of the growth factor TGFβ
Tumor suppressor, inhibits transcription factor important for cell cycle progression
Inhibitor of cellular apoptosis by interaction with initiators
Involved in execution of apoptosis
Initiator of apoptosis
Growth of endothelial cells
Material and methods
Patients and tissues
From the Swedish Cancer Registry, 131 patients with SCC tumor specimens diagnosed during the period 1979–2005 and stored in the Pathology Department in Malmö, Malmö University Hospital, Sweden, were selected.
The original glass-slides of paraffin-embedded tissues were retrieved and the most representative blocks were selected for each patient. A new hematoxylin-eosin stained glass-slide was prepared from the selected blocks. These slides were reviewed by two expert pathologists (KB and CK) to evaluate whether the SCC diagnosis was confirmed or whether it was reclassified. In the evaluation, the presence of obvious epithelial lips at the periphery and a sharp delimitation between stroma and epithelium was criteria in favour of KA, while large numbers of mitotes and strong cellular and nuclear atypia favoured SCC (Fig. 1a and 1b). After the initial review, 12 patients were excluded from the study due to the poor condition of the blocks. An additional 18 patients were excluded from the study after the second review. One specimen was reclassified as a verruca, 12 tissues to dysplasia without evidence of invasive cancer and 5 were classified as BCC. After construction of the TMA, an additional 3 patients had to be excluded because of failure in transfer to the recipient block. In total, 98 patients were included in the study, 27 specimens with confirmed SCC and 71 specimens reclassified as KA. The mean age of these patients was 70 years. For KA, 66% were men and 34% women and for SCC 72% were men and 28% women.
Tissue array and immunohistochemistry
All glass-slides were reviewed by a single pathologist (K.B.), who selected a representative area of lesion and adjacent normal tissue for the construction of TMA as previously described.26 One mm punch was taken from the selected areas in each donor block and mounted in a recipient block using a manual arrayer (MTA-1, Beecher WI). The first section from TMA was stained with hematoxylin-eosin in order to confirm that punches were representative, i.e. contained squamous epithelial cells of both tumor and normal tissue. In total, 25 SCC tissues, 64 KA tissues and 49 normal tissues were obtained. From 9 patients only normal tissues were representative.
Four μm sections of the tissue arrays were dried, deparaffinised, rehydrated and microwave treated for 10 min in high pH target retrieval solution (DAKO, Glosrup, Denmark) before being processed in automatic Techmate 500 immunohistochemistry staining machine (DAKO). Duplicate sections were stained for each antibody. The following antibodies covering different aspects of carcinogenesis - the immune response, apoptotic pathways, proliferation and cell cycle regulation were used for the staining: CD3+ (1:40), CD20+ (1:1500, CY) CD68+ (1:1500, KP-1), p53 (1:100, 318-6-11), Ki-67 (1:200, M0722) and p21WAF1/Clp1(1:25, SX118) from DAKO; pRb (1:50, IF8), Bcl-xL (1:50, H-5), NFκB p65 (1:500, F-6), IκB-α (1:10, H-4), Trap1 (1:5,C-8), Survivin (1:25, FL-142), Flk1 (1:1000, A-3) and Stat3 (1:100, F-2) from Santa Cruz Biotechnology (Santa Cruz, CA); phosphorylated pRb (1:100, Ser807/811) and Caspase 3 (1:100) from Cell Signaling Technology (Beverly, MA); p16 (1:100) from BD Pharmingen (San Jose, CA); Cyld (1:500, rabbit polyclonal antibody courtesy of Dr. R. Massoumi) and Bak AB-2 (1:500, RB-1520-P) from Lab Vision Corp. (Fremont, CA). The antibodies stained the expected cellular compartments as stated in the manufacturer's manuals.
The stained TMAs were evaluated by light microscopy without knowledge of clinical details or HPV status. The TMAs were scored in 3 categories as negative staining, weak but detectable staining of some or all cells or strong staining of some or all cells.
Due to the fact that some punches in the recipient block did not contain sufficient tissue for all the sections, there were varying numbers of patients in the different groups evaluated for the different antibodies.
From all the selected patient blocks, a slide of about 30 μm was cut for the HPV analysis. These sections were deparaffinised, dried and DNA extracted by phenol-free method.27 Briefly, sections were incubated in xylen twice at 50°C for 30 min, centrifuged and then washed twice in ethanol (99%). Samples were air dried and incubated in Digestion buffer (50 mM Tris HCl, 1 mM EDTA 0.5% Tween 20 and 200 μg/ml Proteinase K) for 24 hr. Next day, samples were heated to 100°C for 10 min and stored in −20°C. To ensure the quality of the extracted DNA, all samples were analysed for the presence of the β-globin gene by PCR.28 This quality control excluded 2 SCCs and one KA specimen. For detection of HPV, two separate single tube “hanging droplet” nested PCRs with FAP primer pairs29 or E1/E2 primer pairs30 were applied. Five microliter of each PCR mix was examined by gel electrophoresis and the amplicons cloned using TOPO TA Cloning kit (Invitrogen, Leek, The Netherlands) and sequenced according to ABI Prism® Big Dye™ Terminator Cycle Sequencing Ready Reaction Kits (Applied Biosystems).
Multivariate exact logistic regression models were applied using LogXact version 6 (Cytel software corporation). A p-value of <0.05 was considered significant.
This study was approved by the Ethical Review Board of Lund University, Sweden.
The TMAs were stained with 19 different antibodies (Table I). The percentage of stained tissues is summarized in Table II. Since the study aimed to find qualitatively different biomarkers, the results were scored as negative, positive or strong staining. The immunological response markers for T-cells (CD3) and macrophages (CD68) were more often scored as strong staining among the KA than in the normal tissues (p = 0.004 and p = 0.003, respectively, Table II). A tendency for larger proportions of strong staining for T-cell and B-cell (CD20) markers was noticed for KA compared with that of SCC (Table II). Among the transcription factors there were no differences between SCC and KA, but a larger proportion of the tumor tissues SCC (p = 0.04) and KA (p = 0.03) was strongly stained for p53 than that of normal tissues (Table II).
Table II. Proportion of Staining Scores in SCC, KA and Normal Tissues
Positive and strong positive
Positive and strong positive
Positive and strong positive
Differences between the groups was calculated using multivariate exact logistic regression, adjusting for sex, age and all other biomarkers tested. The significant differences are shown in italics.
A similar difference was observed with the tumor suppressor proteins p21 (SCC p = 0.02; KA p < 0.001), p16 (SCC p = 0.01; KA p < 0.001) and retinoblastoma protein (pRb) (SCC p = 0.008; KA p < 0.001) (Table II). The receptor protein FLK-1 was found at approximately the same proportions in all the tissue types, as were the proteins involved in apoptosis: Bak, Cyld and survivin. The antiapoptotic protein Bcl-xL showed positive staining in 84% of the SCC compared to only 15% of the KA tissues, OR, 22.2 (95% CI, [4.5-157]) (Table II). Bcl-xL staining was found only in 3% of the normal tissue (Table II). Representative staining patterns of each tissue type with the Bcl-xL antibody are shown in Figure 1c.
The staining with the proliferation marker Ki67 indicated that all the tissues had proliferating cells. The Ki67 staining was similar for KA and SCC, but was stronger in the lesions than in the normal tissues (SCC p < 0.001; KA p < 0.001) (Table II).
The combination of the 2 nested PCR techniques detected HPV DNA in 21.7% of SCC and 27.0% of KA lesions (Table III). Eighty percent of the detected HPV types were grouped to Betapapillomavirus species 2. One SCC patient and 9 KA patients harbored multiple HPV types. Only one mucosal type, HPV16, was detected (in a KA patient with a quadruple infection) (Table III).
Table III. HPV Types and Putative HPV Types Found in Keratoacanthoma and SCC
The staining for the different biomarkers was compared with the HPV status of the lesion, but no significant differences were observed (data not shown).
There are currently no reliable markers to distinguish KA and SCC and the vast proportion are likely misdiagnosed as SCC cases out of precaution,7 as was demonstrated by the major reclassification of specimens in our study. This tissue-microarray-based study identified a difference in expression of the antiapoptotic protein Bcl-xL between KA and SCC. Overexpression of Bcl-xL is associated as a negative prognostic marker for several cancer types, such as breast, prostate and colorectal cancers.32–34 Because Bcl-xL was only infrequently found in KA, it is tempting to speculate that apoptotic pathways are activated in KA which in turn leads to the regressive nature of the KA lesion. However, the difference was essentially specific only for Bcl-xL, and there was only a tendency for higher number of the strongly stained tissues for the pro-apoptotic protein Bak in KA and no difference in staining for caspase-3, a protein involved in execution of apoptosis. Other studies have reported indicative differences between KA and SCC concerning expression of proteins involved in apoptosis, such as cytolytic calcium channel receptor P2X7, Bcl-2 and Bak.17, 35, 36 Also, a small study found no difference in expression of Bcl-xL between SCC and KA.37
The fact that both T cells (CD3) and phagocytes (CD68) were statistically more common in the KA than in the normal tissues, is in line with the notion that the immune system is activated in skin carcinogenesis.38 Furthermore, cytotoxic T lymphocytes and helper T cells have previously been identified as inducers of tumour regression.39 None of the tissues in our study stained negative for T-cells and the highest number of strongly stained tissues was observed in KA lesions (37%) followed by SCC (23%). Previous reports have also shown larger amounts of T-cells in KA tissues compared to that of both normal and SCC tissues.40, 41
The transcriptional factor NFκB, its inhibitor IκB-α, STAT3, tumor suppressors TRAP-1, pRb, phosphorylated pRb, Cyld, antiapoptotic protein survivin and the receptor FLK-1 have not been investigated as possible distinguishing markers between KA and SCC before. None of these were significantly different between SCC and KA, but tendencies were observed for phosphorylated pRb, survivin and NFκB.
Recent studies have indicated p53 as distinguishing marker between KA and SCC, being expressed at higher rates in SCC than KA.17, 36 p53 is an important tumor suppressor, which is commonly mutated and accumulated in cancers.42 p53 has rapid turnover and should as a rule not be detected in normal tissue.17 However, our finding that half of the normal tissues surrounding the tumors were positive for p53 agrees with data showing that p53 mutations are common in normal skin.43
We found predominantly strong staining in the tumors of the important tumor suppressors, p16 and p21, but no difference between SCC and KA. Another study had similar results regarding p16 staining in SCC and KA as our study,14 but a small study (20 subjects) reported a difference in p21 between the lesions.37
Our study had several strengths that may explain discrepancies with previous reports. The present study is rather large, based on 98 patients with stored SCC tumor specimens and is based on blinded rereview of all specimens by two expert pathologists. Furthermore, the use of the TMA technique enabled minimal interassay variability in the immunohistochemical stainings. All our results were also adjusted for age and sex, factors known to be associated with development of skin lesions.19 Limitations include the fact that not all tissues were representatively present on all sections and that we did not study differentiation stage of the tumors.
Patients in our study were predominantly infected by HPVs of Betapapillomavirus species 2, in agreement with previous studies reporting that Betapapillomavirus species 2 is associated with SCC.22, 23 HPV types or species of HPV did not differ between SCC and KA. However, the pro-apoptotic protein Bak is a target for degradation by cutaneous HPV44 and it has been reported reduced levels of Bak in HPV positive SCC compared to HPV negative SCC45, which was not confirmed by us. As there are usually only low viral loads of cutaneous HPVs in infected tissue,46–48 visible viral degradation of Bak proteins seems unlikely.
In summary, staining for Bcl-xL differs in SCC and KA but further evaluation of whether routine Bcl-xL staining, possibly in combination with other markers, would be useful in histopathological diagnosis is necessary.
The National Tissue Array Center was supported by the Swedish National Biobanking Program (part of the functional genomics initiatives SWEGENE and Wallenberg Consortium North). The authors thank Dr. Massoumi for the kind gift of Cyld antibody. The assistance of Mr. Ivo Filinic in retrieving the paraffin embedded blocks is gratefully acknowledged.