Recent publicity in the USA has been designed to suggest that ultraviolet (UV) radiation-emitting sunbeds are safe, presumably meaning safe for everyone including teenagers and children. This is directly counter to a current call by the World Health Organization for sunbed use by under-18s to be banned by national governments on safety grounds, echoed by French law under which it is already banned (World Health Organization, 2003). Similar views are expressed by numerous other national and international bodies and health-professional associations, such as the International Agency for Research on Cancer and the British Government’s Health and Safety Executive (Al-Ani, 2005; Green et al., 2007; Health and Safety Executive, 1995). Accordingly, it seems timely to examine the present state of evidence regarding the safety of the longer (UVA) wavelengths that commonly predominate in sunbed lamps [although not all: lamp spectra vary widely (Berwick, 2008; Miller et al., 1998)]. Is it true that “the harmful wavelengths have been removed” from such lamps, as is sometimes suggested? This question of harmful wavelengths also has broader implications regarding appropriate sun protection, including the choice and design of sunscreens (Haywood et al., 2003). The question is examined here from a biological viewpoint, with emphasis on the most dangerous form of skin cancer, melanoma, and under the two categories of mutagenicity and carcinogenicity. [Other kinds of harm by UVA such as skin ageing, and by UVB such as non-melanoma skin cancer, are considered in the accompanying review by Tran et al. (2008), and elsewhere, e.g. Gallagher and Lee (2006).]
In view of claims that ultraviolet radiation-emitting sunbeds are safe, or safe when they emit only longer wavelengths, research findings are reviewed here on the effects of ultraviolet wavebands A and B (UVA, 315–400 nm and UVB, 290–315 nm) on mutagenesis and carcinogenesis in skin, with particular reference to melanocytes and melanoma. Both UVA and UVB radiation have been shown to induce mutations, as well as mutagenic photoproducts such as cyclobutane pyrimidine dimers, in human skin. UVB can induce melanoma in susceptible mice and in xenografted human skin engineered to express melanocyte growth factors. There is evidence for photosensitization of melanocytes by melanin, especially pheomelanin. UVA can induce melanoma in pigmented fish, and melanocytic hyperplasia in pigmented opossums, but has not generally been tested for melanoma induction in pigmented mammals or in human skin. There is no experimental basis for a claim that UVA is safe, and recreational exposure to this known mutagen should be discouraged.
Mutagenicity and ultraviolet wavebands
A wide range of epidemiological data implicate exposure to natural sunlight as the principal cause of skin cancer generally, and melanoma in particular. This area has often been reviewed and is not detailed here (Gallagher and Lee 2006; Garland et al., 2003; Moan et al., 2008; Tran et al., 2008; Tucker and Goldstein, 2003). Sunlight reaching terrestrial ground level contains predominantly the UVA (315–400 nm) and UVB (280–315 nm) wavebands (Garland et al., 2003; Seckmeyer et al., 2008). (Note that some authors use UVB borderlines of 290 or 295 nm, and/or 320 nm instead; it is unclear which if any criteria are preferable, but the choice may affect research outcomes.) UVB is absorbed by DNA markedly more efficiently than is UVA, with about a 10-fold difference between 300 nm and 330 nm (Young et al., 1998). It has thus been widely assumed to induce more DNA damage in vivo, and to be the predominant carcinogen in sunlight for melanoma (Agar et al., 2004; Tucker and Goldstein 2003).
However, this assumption fails to take into account several relevant factors. One is the widely different amounts of UVA and UVB reaching the target cells. UV in equinoctial sunlight at noon has been calculated from solar angles and latitudes to contain about 96% UVA (Singapore) through 97.5–98% (Europe, USA, Australia) to 99% (Iceland), with the rest UVB (Garland et al., 2003). The proportion of UVA will be lower in summer and higher in winter, or at times of day other than noon, or if clouds are present – the calculations assumed no clouds. An alternative model gives about 4%–6% UVB (defined as <320 nm) in mid-latitude summer conditions (De Fabo et al., 2004, and sources cited). Measured values of noonday UVB (280–315 nm) to UVA ratios in sites at latitudes between 70°N and 35°N gave about 1%–3% UVB in September, 2%–3.5% in June; these figures (Seckmeyer et al., 2008) are used in the following. Adult human Caucasian epidermis absorbs on average about twice as much UVB as UVA as a proportion of incident (Agar et al., 2004; Garland et al., 2003), so in the USA etc. the UV radiation reaching the basal layer of the epidermis – site of the melanocytes and the keratinocyte stem cells – will typically be about 99% UVA. The target cells for melanoma carcinogenesis may be these basal epidermal melanocytes, or it is alternatively possible that specific melanocytic stem cells may be important targets (Fang et al., 2005). Such cells are found in the “bulge” region part-way down hair follicles in mice (Nishimura et al., 2002), and probably in an equivalent region in humans (Grichnik et al., 1996; Tiede et al., 2007), though this is unclear at present. Any UV light reaching this even deeper level would comprise well over 99% UVA.
Garland et al. (2003) also reported that male mortality from melanoma, after multiple adjustments including for skin color, correlated among different countries with UVA flux and not with UVB flux. This finding should be taken with caution because the correlation was not seen in females in this analysis. Nonetheless, in another study comparing fewer countries, a better association of general melanoma incidence with UVA than with UVB flux was likewise noted (Moan et al., 2008).
Consistent with the known differential absorption by epidermis is the finding that UVB fingerprint mutations (GC to AT) and characteristic photoproducts (cyclobutane pyrimidine dimers, CPD) were largely confined to non-basal cells of human squamous-cell carcinomas and solar keratoses, while UVA fingerprint mutations (AT to CG) and a characteristic adduct (8-hydroxy-2’-deoxyguanine) were at least as common in the deeper (dermal) parts of the lesions as in the distal layers (numerically more common in the data shown) (Agar et al., 2004). This adduct 8-hydroxy-2’-deoxyguanine is associated with oxidative DNA damage; relevant because this is caused largely by UVA rather than by UVB, via the generation of reactive oxygen species (ROS) (review: Brenner and Hearing, 2008). Clear gradients in distributions of photolesions with depth were likewise observed in normal human epidermis irradiated with UVC or shorter-wave UVB (290 nm), and shallower gradients with UVB at 300 nm, detectable only by image analysis (note: normal skin is thinner than the lesions studied by Agar et al.) (Chadwick et al., 1995; Young et al., 1998). No gradient with depth was seen with UVA-induced CPD photolesions (Young et al., 1998). UVA induces CPD in skin at rates 2–3 orders of magnitude lower per dose than UVB (Young et al., 1998). However, a substantial fraction of the total CPD caused by sunlight in basal epidermis may thus still be due to the ∼99% UVA, especially in the presence of UVB-selective sunscreens. Douki et al. (2003) reported that the majority (2/3) of initial UVA-induced photolesions (in cultured, unpigmented cells) were in fact CPD, not oxidative lesions as previously thought. They later presented evidence that CPDs were likewise the prevalent UVA-induced lesions in whole normal human skin (Mouret et al., 2006). These appeared more persistent in skin when induced by UVA than by UVB. UVA was also found to be responsible for observed conversion (probably oxidative) of some readily-repaired initial 6–4 pyrimidine dimers into “Dewar isomers” which are slowly repaired and highly mutagenic (Chadwick et al., 1995; Douki et al., 2003). More generally, there are several reports that UVA is mutagenic to cultured cells at physiological fluxes (brief review: Douki et al., 2003). It was also recently reported that UVA at high physiological doses can induce DNA single- and double-strand breaks as well as tumorigenicity, in human keratinocytes (Wischermann et al., 2008).
Incidentally, both UVB and UVA can induce some CPDs from both CT and TC as well as TT dipyrimidines (Douki et al., 2003; Mouret et al., 2006). Several such dipyrimidines are found in the non-coding sequence (TCACTTT) opposite the GTG (valine) codon that participates in the V600E mutation of BRAF, common and important in melanoma. CPDs are quite slowly repaired (by nucleotide excision repair, NER), and hence are mutagenic (van Steeg and Kraemer, 1999). Thomas et al. (2006) propose that error-prone DNA replication past such a CPD could lead to V600 mutations following UV damage, contrary to previous thinking.
Another factor relevant to the balance between UVA and UVB carcinogenicity in skin is melanin. There are clear photoprotective effects of melanin in Black and Asian skin, but the role of the smaller amount of melanin in Caucasian skin is less clear (Brenner and Hearing 2008). There is a good deal of evidence conversely that melanin within the target cell, as opposed to distal to it, may act as a photosensitizer to UV light, by photodegradation of melanin, generating ROS (Moan et al., 1999; Wenczl et al., 1998). Such evidence seems stronger for pheomelanin, the red or yellow, sulfur-containing subtype of melanin found in human red and fair hair, especially when irradiated with UVA (Chedekel et al., 1978; Hill and Hill, 2000; Wenczl et al., 1998; review: Brenner and Hearing 2008). The latter is important for the present issue, as it would increase the relative mutagenicity of UVA in the skin of many Caucasians.
In summary, there is clear evidence that UVA is mutagenic in human skin, and some evidence that it may be more mutagenic for melanocytes in Caucasian skin than for other, unpigmented cells.
Melanoma carcinogenicity and ultraviolet wavebands
The epidemiological evidence directly linking human melanoma and specific wavebands of UV radiation is weaker than for non-melanoma skin cancer, as discussed above. This does not necessarily imply a lesser mechanistic link, because statistical significance depends on numbers and melanoma is overall substantially less common (although it causes more deaths). Data from experimental models of melanoma or its benign precursors are clearer, but not plentiful. These models will be reviewed, noting the pigmentation of the animals – to be discussed.
Setlow et al. (1993) examined wavelength-dependency of the induction of melanoma in pigmented swordtail-platyfish hybrids (genus Xiphophorus), finding that melanoma could be induced not only by UVB but also by UVA (>320 nm) and even blue visible light, with the rate of induction falling much more shallowly with increasing wavelength than the action spectrum for absorption by DNA, or the mutagenicity in non-pigmented human cells. This raised the then-surprising possibility that UVA might be an important carcinogen for melanoma. A subsequent study showed that the action spectrum for generation of ROS from melanin in Xiphophorus skin was very similar to that for melanoma induction (Wood et al., 2006). Mitchell et al. (2007) found that the melanoma susceptibility of different Xiphophorus strains correlated negatively with their efficiency of NER. This implied that DNA lesions predominantly repaired by NER contribute significantly to Xiphophorus melanoma. This is also known for humans, since susceptibility to melanoma as well as other skin cancers is highly elevated in people with the NER deficiency xeroderma pigmentosum (Hakem, 2008; Mitchell et al., 2007; van Steeg and Kraemer 1999). This is consistent with known pathways of UVA (and UVB) mutagenesis, since lesions repaired by NER include CPD, 6–4 photoproducts and base adducts typical of oxidative damage (Hakem 2008; van Steeg and Kraemer 1999).
There are growing numbers of mammalian models of melanoma or its benign precursors, nevi, but few studies to date have compared melanoma genesis or nevogenesis by UVA and UVB. Opossums (pigmented) can develop malignant melanoma following UV irradiation of shaved skin, but not with UVA alone. UVA was found to induce only focal melanocytic hyperplasia, and only at higher doses per unit area than for Xiphophorus melanoma (Ley, 2001). Menzies et al. (2004) described a white-haired guinea-pig strain that acquired pigmented benign nevi on UV irradiation of shaved skin; the melanocytic basis for the white hair is evidently not true albinism. At any rate, the authors used UVA and UVB in the ratios and fluxes found in noonday equatorial sunlight, and reported that UVB but not UVA or visible light would augment nevus formation (Menzies et al., 2004).
A striking counterpoint to the Xiphophorus findings was provided by a report that in transgenic mice overexpressing hepatocyte growth-factor in melanocytes, neonatal exposure to UVB (<320 nm) but not UVA could induce malignant melanoma (De Fabo et al., 2004). UVA was used at up to 10x the maximum UVB dose, or 33x a UVB dose that produced some melanomas. These mice were albino. Carcinogenicity was age-dependent; melanomas were not seen in adult mice with the same regimen. Several other transgenic mouse models are reported to show induction or increase of melanoma development by neonatal UVB; these have been well reviewed by de Gruijl et al. (2005). These are mice with defective tumor suppressor genes or expressing oncogenes in melanocytes. Remarkably, 7/8 of the listed mouse genotypes in which UVB alone was found to initiate melanoma were albino, and the 8th (Tyr-SV40LT) was hypopigmented (de Gruijl et al., 2005). Three black or agouti genotypes were listed as inducible for melanoma by DMBA (no information on UV), and two as not inducible by UVB. One additional pigmented transgenic strain has since been reported (Xpc−/−, Cdkn2a−/−) where neonatal UVB could induce melanoma-like unpigmented tumors (Yang et al., 2007). Together, the findings suggest that pigmented mice are protected against UVB induction of (typical) melanoma, although only one transgenic (Tyr-Hras) was directly compared in both pigmented and unpigmented littermates (Broome Powell et al., 1999). Unfortunately there appear to have been no other studies of UVA in this context, and from the relative fluxes in sunlight it could be argued that doses up to 50x the active UVB doses should be tested, and pigmented mice should be included. Also there appear to be no data available on pheomelanic mice. More work is evidently needed on the relations between UV wavebands, melanin and melanoma in transgenic mice.
Some care is needed in comparing animal models of photocarcinogenesis with the human case, since animals vary in (for example) the distribution of melanocytes between the epidermis, dermis and hair follicles; the pigment color, the epidermal thickness (affecting UV penetration), the efficiency of DNA repair, and so on. Comparisons even between albino mice and albino humans are puzzling, since there is evidence that albino humans (at least in Tanzania, and anecdotal reports suggest the same in Africa) have surprisingly low susceptibility to melanoma despite high susceptibility to non-melanoma skin cancer (Lookingbill et al., 1995). This is consistent with the idea of melanin as a photosensitizer to sunlight in melanocytes.
There are some reports on experimental models of human melanoma initiation. Herlyn and colleagues have used adenoviral vectors to express exogenous genes in organ-cultured human skin or skin reconstructs, which were then xenografted to SCID mice. Melanocytic hyperplasia was found when FGF2 overexpression was combined with UVB irradiation of grafts (Berking et al., 2001), while in skin overexpressing three melanocyte growth-factors, repeated UVB irradiation of xenografts led to the development of overt melanomas. With adult skin these remained in situ but with neonatal skin they became invasive, further evidence for age-dependency of UV carcinogenicity (Berking et al., 2004). No experiments on UVA in this system were reported however.
In conclusion, it is clear that both UVA and UVB are mutagenic for skin, and for melanocytes in particular. UVA is much less mutagenic in unpigmented cells, but UVA flux to the basal epidermis from sunlight is typically around 50-100-fold higher than UVB flux. Moreover there is evidence that melanin and especially pheomelanin can photosensitize cells to UVA mutagenicity. UVA can initiate melanomas in fish and melanocytic hyperplasia in pigmented opossums, while UVB can induce melanoma in susceptible mice and UVA has generally not been tested. Pending better experimental data on whether UVA can indeed cause melanoma in mammals, and given that it is mutagenic and cell mutations can cause cancer, much stronger steps should be taken internationally to warn users of sunbeds – even those emitting UVA only – that this activity may be hazardous, such as mandatory warning notices. An example of such a notice has been made available in the UK (Health and Safety Executive 1995), also at http://www.hse.gov.uk/radiation/nonionising/sunbeds.htm. Specifically use by those under 18 should be banned, and publicity claiming that UVA sunbeds are safe should not be permitted.
Thanks to a number of colleagues for helpful discussions. Work in the author’s laboratory on melanoma is supported by the Wellcome Trust and Cancer Research UK.