Early Events in UV Carcinogenesis—DNA Damage, Target Cells and Mutant p53 Foci


  • This paper is part of a special issue dedicated to Professor Hasan Mukhtar on the occasion of his 60th birthday.

*Corresponding author email: f.r.de_gruijl@lumc.nl (Frank R. de Gruijl)


Skin carcinomas are the most common cancers in fair-skinned populations of North West European descent. The risk is closely related to sun (UV) exposure and susceptibility to sunburn. Induction of squamous cell carcinomas (SCCs) in the skin of hairless mice by daily UVB exposure appears to emulate the genesis of these tumors in humans quite well. The carcinomas, and the UVB signature mutations that they carry in their p53 genes, can be linked most specifically to the induction of cyclobutane pyrimidine dimers (CPDs). The wavelength dependence of the induction of carcinomas parallels that of CPD induction over the UVB and UVA2 spectral regions. Microscopic clusters of cells overexpressing p53 with UVB signature mutations (“p53 patches”) can be detected in the interfollicular epidermis long before the skin tumors arise. DNA repair—more precisely nucleotide excision repair—is a crucial line of defense against UV-induced p53 patches and skin carcinomas. Although chemoprevention of UV carcinogenesis, e.g. with difluoromethylornithine, may be successful by inhibiting the outgrowth of tumors, it may be better to counter the initial steps in tumor development. As the p53 patches appear to be potential precursors of SCCs, regression of p53 patches in unexposed skin should lower subsequent development of SCCs. However, “holoclonal” p53 patches might persist. Ablation of the interfollicular epidermis would be expected to abrogate development of SCC, and negation of this expectation [Faurschou A. et al., Exp. Dermatol. 2007;16:485–489] would indicate that SCCs stem from deep-seated cells in the hair follicles. Careful examination of archival material showed that although most small p53 patches arise interfollicularly, some may actually arise high up in a follicle, in the infundibulum.


Cells exposed to sunlight have to cope with cytotoxic UV radiation. In the early 20th century action spectral analyses showed that UV absorption by DNA was responsible for cell killing and mutagenic effects, with a corresponding peak efficacy around 265 nm in the UVC (wavelengths 200–280 nm) (1,2). In the anoxic primordial atmosphere the short wavelength UVC radiation from the sun had free access to the earth’s surface, imposing a highly cytotoxic regimen under which it would be impossible for cells to survive outside special niches, e.g. in oceans under layers of organic material. With the introduction of oxygen into the atmosphere, the “Great Oxidation” some 2.4 billion years ago (3), the ozone layer was formed high up in the stratosphere, where molecular oxygen (O2) was split up by short wavelength UV (UVC and vacuum UV, <200 nm) to recombine to ozone (O3). The stratospheric ozone layer blocked off the UVC radiation and filtered out a substantial part of the UVB radiation (280–315 nm). The remaining UV radiation that reached ground level was still cytotoxic but life on earth adapted itself most impressively to fend off the residual cytotoxic effects from sunlight. The human skin is a fine example of this adaptation—despite persistent insults imposed by solar UV radiation for decades, the skin remains viable by continuous “high fidelity” repair and renewal processes in its outermost and most vulnerable layer, the epidermis, which also forms a protective barrier of cornified cell remnants at its outer surface, the stratum corneum. That solar UV radiation is still a cytotoxic factor to reckon with for the skin becomes manifest in severe sunburn reactions after acute overexposure, and in the long run, a few “sun-crazed and perverted” cells may even slip through the skin’s robust defenses and get a foothold from which cancers may develop.

With a wide appreciation of a skin tan, sun exposure behavior appears to have changed dramatically over the last century throughout white Caucasian populations in the Western world (4), and the strong increases in skin cancer incidences (5,6)—and to a lesser extent increases in mortality—have been attributed to this change in behavior (4,7). With this tremendous increase, skin cancer is now by far the most common cancer in the white non-Hispanic population in the United States, and although the mortality is very low for the skin carcinomas, the morbidity and burden to the healthcare system is substantial (8). Furthermore, in certain high-risk groups, most notably in organ transplant recipients (9,10), the rate of sun-related skin carcinoma development can rise excessively. Thus, skin carcinomas pose a serious health problem, and novel therapies or interventions that are effective and less burdening than conventional surgical removal are highly desirable. A fundamental understanding of UV-driven skin carcinogenesis is quintessential for the developments of such novel therapies or interventions. The group of Hasan Mukhtar has extensively tested chemoprevention by naturally occurring compounds (e.g. phenols in teas and lupeol in fruits) (11,12). Besides underpinning the merits of certain drinks or foods, this approach with active ingredients may identify certain crucial oncogenic processes susceptible to intervention. Here, we will discuss some of our explorations into the early steps of UV-driven skin carcinogenesis as studied in models with hairless mice which were either wild type or genetically modified to identify the crucial steps. In these studies we focused on the UV-induced DNA damage and its repair in relation to formation of clones of (p53) mutated cells, which may then grow out into tumors. Identification of crucial early steps and the target cells in which these steps occur may direct future highly effective interventions in high-risk individuals, such as organ transplant recipients.

UV-induced CPDs and skin carcinomas

Adjacent pyrimidines in a DNA strand are particularly vulnerable to dimerization by UVB radiation, with the cyclobutane pyrimidine dimer (CPD) as the predominant lesion and three to four times more frequent than the pyrimidine 6-4 pyrimidone photoproduct (6-4PP). The 6-4PPs with larger DNA helix distortions are generally repaired much faster than CPDs. The action spectrum of CPD induction in mouse skin peaks around 290 nm (13), as does the action spectrum of sunburn (14) (radiation of wavelengths below 290 nm do not penetrate very deeply because of strong absorption in the superficial layers of the epidermis). Similarly, we found the efficacy in induction of squamous cell carcinomas (SCCs) in hairless mice to peak at 293 nm (15). By correcting this wavelength dependency in mice for differences in epidermal transmission, we estimated an action spectrum of carcinoma induction in humans that peaked at 298 nm (16) and resembled the action spectra of sunburn (16–18) and CPD induction in human skin (19,20).

The causal role of CPDs in UV carcinogenesis was substantiated by specific removal of these DNA lesions through photorepair in transgenic mice (21). In contrast to photorepair of 6-4PPs, the photorepair of CPDs suppressed the induction of skin carcinomas by chronic UVB exposure.

UVB radiation was found to induce characteristic mutations (C to T transitions) at di-pyrimidine sites (22), the typical target for UVB-induced dimerization. These UVB signature mutations were subsequently detected in p53 genes from human SCCs (23), clearly attesting to solar UVB radiation as the culprit. In extensive analyses of p53 in experimentally UV-induced SCCs in hairless mice we found such UVB signature mutations in a majority of the tumors (24), with a bias towards associations with di-pyrimidine sites on the nontranscribed strand—a bias not found in human SCCs. This bias is attributable to a slow repair of CPDs in the nontranscribed DNA strand in mice (25). At the base sequence level, sites with a low rate of pyrimidine dimer removal appeared to correspond with mutational hotspots in the p53 gene (26).

With long wavelength UVA1 radiation (340–400 nm) other types of DNA lesions (e.g. oxidation of bases) than CPDs become relatively more important than with UVB radiation (27). We therefore expected to see a shift in types of mutations in the p53 gene between SCCs induced by UVB radiation versus those induced by UVA1 radiation. Surprisingly, most of the UVA1-induced SCCs harbored no mutations in the p53 genes, and the few p53 mutations that occurred were UVB-like hotspot mutations (28).

UV-induced skin tumors in NER and p53-deficient mice

Loss of p53 or mutation of the gene renders the skin more susceptible to UV carcinogenesis (29–32), indicating that p53 plays an important protective role and that mutation of p53 can be an important step in the development of UV-induced SCCs. Mutagenesis is greatly enhanced if repair of UV-damaged DNA is compromised. The UVB-induced pyrimidine dimers are mainly removed by nucleotide excision repair (NER), and the human syndrome xeroderma pigmentosum (XP) with defects in NER is associated with a dramatic increase in risk of skin cancer. Two main pathways can be distinguished in NER—transcription coupled repair (TCR) of actively transcribed DNA strands, and global genome repair (GGR) of nontranscribed strands. TCR is commonly more rapid than GGR, particularly in rodents, which explains the bias in p53 mutations in UV-induced SCCs, mentioned earlier. Mice with defects in NER were found to develop the skin carcinomas much faster upon UV exposure than wild type littermates—a defect in TCR alone (Csb mice) speeded up tumor development by a factor of 2, a defect in GGR alone (Xpc mice) by a factor of 3, and defects in both TCR and GGR (Xpa mice) by a factor of 4 (33,34). Interestingly, the strand bias in p53 mutations in UV-induced skin carcinomas was reversed in TCR-deficient mice, i.e. C to T transitions occurred in association with dipyrimidine sites on the transcribed strand instead of the nontranscribed strand (35). A TCR defect (as in Csb and Xpa mice) resulted in the induction of many benign papillomas which—in contrast to SCCs—carried H-ras mutations and no p53 mutations (36). The H-ras mutations occurred in codon 12 with a CC site in the transcribed strand. This site apparently becomes a mutational target of UVB radiation only if TCR is defective. Also, in two-stage chemocarcinogenesis (e.g. 7,12-dimethylbenz[a]antracene [DMBA] tumor initiation followed by 12-O-tetradecanoylphorbol-13-acetate [TPA] tumor promotion) mainly papillomas are induced which carry H-ras mutations (in codon 61) and no p53 mutations (37).

ODC inhibition suppresses SCC development

A subset of XPA patients did not appear to develop skin tumors even at advanced age, despite a severe NER deficiency. Fibroblasts from these patients were deficient in UV-induced enhanced recovery (ER) of genotoxically damaged viruses, and this was associated with a deficiency in UV induction of ornithine decarboxylase (ODC) (38). ODC is a regulator of polyamine biosynthesis, and plays a key role in the development of (skin) tumors. Difluoromethylornithine (DFMO) inhibits ODC and was shown to abrogate ER in proficient fibroblast (39). Hence, we inferred that DFMO could protect ER-proficient XPA patients from developing skin cancers. We tested this hypothesis in an experimental model with daily UV exposure of Xpa-null mice who received DFMO in their drinking water (40). DFMO was indeed found to suppress UV carcinogenesis very strongly. However, discontinuation of the DFMO treatment after 14 weeks resulted in a rapid appearance of skin carcinomas, and after 4 weeks their number equaled that observed in the control group which was also exposed to UV radiation but not treated with DFMO. If DFMO treatment started late, after 14 weeks, the number of carcinomas could still be reduced two- to threefold in comparison with the control group. Interestingly, the tumors that did grow under late-stage DFMO treatment appeared to be selected for high ODC activity. These results indicate that DFMO does not protect from tumor initiation, i.e. induction of persistent latent tumor foci, but does suppress the outgrowth of tumors. This tumor suppression is, however, not complete, and DFMO should therefore probably be combined with another tumor-suppressive drug, e.g. a COX-2 inhibitor (41,42); thus, tumors which are resistant to one drug may be stopped by the other drug.

Early mutant p53 foci

In normal UV-exposed murine and human skin, clusters of p53 over-expressing keratinocytes were found long before tumors arose (43,44). Hairless mice that daily received a threshold dose for sunburn reactions (one minimal edemal dose) developed ∼10 tumors within 100 days while the same irradiation regimen yielded ∼10 p53 patches at 15 days, i.e. seven times earlier (45). These p53 patches appear to be interfollicular clonal outgrowths (see below, and Fig. 1), and can therefore also be referred to as p53 clones (44,46,47). As the immunopositive p53 patches were scored as clusters of at least 10 cells, p53 patches must have been induced in the very first week of irradiation (45).

Figure 1.

 p53 patch in an epidermal sheet from a UV-exposed hairless mouse—viewed from the basal side. A mutant p53-specific antibody (PAb240) was used for this staining. Arrows point at some of the hair follicles which appear to stay free from the interfollicularly expanding p53 patch.

The majority of the p53 patches were found to carry the UVB-signature mutations in the p53 gene (44,45,48). This mutation spectrum is comparable with that of SCCs, which is in line with the hypothesis that p53 patches are precursors of skin carcinomas (45). Deregulation of the major functions of p53, e.g. apoptosis and cell cycle arrest, would therefore appear to be an early step in skin carcinogenesis as high percentages (54–88%) of SCCs contain p53 mutations (23,24,49–51). Mice with defects in NER were also found to be more sensitive to UV induction of p53 patches, in correspondence with their sensitivity to develop UV-induced SCCs (52). When p53 patches and SCCs were compared no shift in mutation spectrum was found, indicating that there was no selection of certain mutations in the progression from p53 patch to SCC (52,53). Based on p53 mutation spectra all p53 patches can be considered potential precursors of SCCs. However, only a few of the numerous p53 patches develop into SCCs—it is estimated that at the time the first tumor arises, an individual has 8000–40 000 patches (52).

Origin of p53 patches and SCCs

Daily exposure of hairless mice to a dose just below the threshold for sunburn reactions leads to the formation of actinic keratoses (AKs, around 1 mm in diameter) in 2–3 months, and these lesions may grow out into frank SCCs (diameters > 3 mm) in a matter of weeks (54). Early discontinuation of the daily doses after about 1 month, well before the appearance of any AK, will still give rise to AKs and SCCs, although at a much lower rate and with far longer latency times (55). The small AKs then show a high tendency of regression. Likewise, at the microscopic level the number of p53 mutant foci tends to drop after discontinuation of UV exposure (43,56). If these p53 patches are the precursors of SCC after discontinuation of UV exposure, they should persist in the epidermis, i.e. they should not be lost in the turnover and renewal of the epidermis (in mice the turnover time of the viable epidermis equals about 1 week) (57,58). This raises the notion that persistent p53 mutant clones should be “holoclonal,”i.e. they should necessarily originate and be maintained from a long residing germinative epidermal cell, a stem cell. As the p53 patches are located in the interfollicular epidermis (Figs. 1 and 2b), one would infer that they and the ensuing SCCs are derived from interfollicular stem cells. However, it has recently been reported that UV carcinogenesis in hairless mice was not affected by removal of the interfollicular epidermis by ablation at week 7 in the course of chronic UV exposure (59,60) (a similar result was obtained for chemocarcinogenesis in older experiments [61]). Assuming that the ablation was not only adequate on naïve unexposed epidermis, but also on UV-exposed hyperplastic epidermis, one would be led to agree with the authors, and conclude that the target cells for the UV-induced SCCs do not reside in the interfollicular epidermis, but somewhere deeper in the hair follicles. This would imply that the interfollicular p53 patches may not be the precursors of AK and SCC at all. This prompted us to go back to archival material and carry out an exploratory, but thorough, examination of an epidermal sheet from a UV-exposed hairless mouse. Restricting our analysis to early small p53 patches (<10 cells) we measured the distance for each p53 patch to the orifice of the nearest hair follicle, and for reference, also the distance of this orifice to the next nearest neighbor (see Fig. 2a). It turned out that—as expected—most of the small p53 patches were located well between the follicles (Fig. 2b). However, we did find 1 in 10 patches that was actually located out off the plane of the epidermal sheet—it appeared to be seated in the upper part of the follicle, above the sebaceous gland in the infundibulum (Fig. 2c). Whether p53 mutant foci from this latter location are the only ones which are truly “holoclonal” in a hyperplastic epidermis under chronic UV exposure, and consequently, the only foci from which AKs and SCCs may develop, is an interesting concept which clearly needs further study.

Figure 2.

 Positions of early developing p53 patches (<10 cells) in relation to hair follicles—(a) a plot of the percentage of p53 patches within a certain distance “x” from the nearest hair shaft (solid line, = 10) and for comparison a similar plot for nearest neighbor distances between hair shafts (dashed line, = 10); (b) a bottom view of a sheet with interfollicular p53 patches, some orifices of hair shafts in the plane of the interfollicular epidermis are highlighted by dashed white circles (the dormant follicles rise up and are out of focus with aspecifically stained sebaceous glands); (c) a p53 patch in a hair follicle, out off the focal plane of the interfollicular epidermis.

Following experiments by Mitchell et al. (62), our group showed that interfollicular stem and progenitor cells accumulated CPDs with daily UV exposures low enough to avoid a hyperplastic response (63), and that upon growth stimulation by TPA these damaged basal cells gave rise to an abundance of clusters of cells overexpressing wild type p53 (64). Whether these wild type p53 foci evolve into tumors is not yet known. By comparing the low-dose with the high-dose regimen, it may be inferred that the hyperplastic state and the growth stimuli involved (e.g. EGF-R MAPK signaling [65]) may be essential for the rapid development of SCCs; the amplification of H-ras genes and overexpression of the protein may also be instrumental in this respect (66). Even with low daily doses epidermal hyperplasia will ultimately develop (67). The importance of hyperplasia to SCC development appeared to be underlined by comparing the susceptibilities of two mouse strains to UV carcinogenesis at equal genotoxic dosages (68).


UV radiation is arguably the most abundant carcinogen in our environment (next to oxygen), and the skin is necessarily equipped with very effective defenses. However, skin cancers do develop in the long run as we age. And as we enjoy the sun more and more, the risk has grown substantially. Developing practical and adequate preventive strategies and targeted interventions in high-risk groups appear to be feasible, but require a detailed knowledge of UV carcinogenesis in order to be effective. Skin carcinogenesis can be halted in different stages, e.g. DFMO appears to arrest a late stage of tumor development, most likely the outgrowth of a tumor, but it can best be combined with another agent, e.g. a COX-2 inhibitor, to lower the chance of selection for drug-resistant tumors. Abrogating early stages of tumor development may be more effective and robust. However, the earliest stages of tumor transformation still remain somewhat veiled—are the interfollicular p53 patches genuine precursors of AKs and SCCs, or is there a subfraction of follicle-derived p53 patches which are the only foci persisting long enough to evolve into AKs and SCCs? Or is the p53 patch not an essential early step in SCC development, and is some other early oncogenic event more important, e.g. increased sensitivity to UV-induced epidermal growth stimuli? Resolving these uncertainties in early UV carcinogenesis is quintessential for adequate interventions in tumor progression.


Acknowledgements— We thank the PhD students and postdocs who were involved in this work, in particular Rob Berg, Annemarie de Laat and Henk van Kranen. The close collaboration with Leon Mullenders, Bert van der Horst and Harry van Steeg and their expertise on DNA repair were instrumental to our research and is very much appreciated. The Dutch Cancer Society has supported the research by our group with successive grants.