Inflammatory Doses of UV May Not Be Necessary for Skin Carcinogenesis

Abstract

The UV wavelengths in sunlight are the main cause of skin cancer in humans. Sunlight causes gene mutations, immunosuppression and, at higher doses, inflammation. While it is clear that immunosuppression and gene mutations are essential biologic events via which UV causes skin cancer, the requirement for UV-induced inflammation is less certain. Both the UVB (290–320 nm) and UVA (320–400 nm) wavebands within sunlight can cause skin cancer, gene mutations and immunosuppression. However, UVB, but not UVA, at realistic doses can cause inflammation, and UVB induces skin cancer, immunosuppression and gene mutations at doses much lower than those required to cause inflammation. Inflammation enhances skin carcinogenesis, but may not be UV induced, and inflammatory mediators at doses too low to cause inflammation may be required. UV-induced mutations can cause epidermal cells to make proinflammatory factors or to induce them in the surrounding stroma, creating an oxidizing environment in which additional oncogenic mutations are likely to take place, even in the absence of UV. Our hypothesis is therefore that subinflammatory doses of both UVA and UVB cause benign skin tumors. One of the effects of sunlight-induced mutations may be the production of inflammatory mediators that enhance carcinogenesis.

Introduction

This review will explore the issue of whether UV-induced inflammation is essential to cause skin carcinogenesis, or whether UV-induced immunosuppression and gene mutations are sufficient. Professor Hasan Mukhtar, whose work is being commemorated by this series of articles, has made a large contribution to this field. Determination of the cellular and molecular events involved will enable approaches such as chemoprevention to reduce the incidence and development of skin cancers (1–5).

UV radiation can induce the formation of the benign skin cancers actinic keratosis (AK) and in mice, papillomas which arise from transformed keratinocytes. These can develop into squamous cell carcinomas (SCC) which are malignant skin cancers. UV can also cause the formation of basal cell carcinomas (BCC) in humans, which also arise from keratinocytes and are invasive but rarely metastatic.

The importance of UV-induced immunosuppression in enabling the outgrowth of primary skin tumors was shown by transferring lymphocytes from chronically UV-irradiated or unirradiated mice into syngeneic recipients who had been given lethal doses of X-radiation to destroy their own immune systems. These recipients were also given skin grafts from the UV-irradiated donors. By 60 weeks after grafting, primary skin cancers only arose on about 15% of the UV-irradiated grafts on mice reconstituted with an unirradiated immune system, while primary skin cancers arose on about 60% of the grafts when the immune system had been reconstituted with lymphocytes from a UV irradiated donor (6). This experiment clearly shows that suppressor T lymphocytes (now called T regulatory cells or Treg), induced by UV irradiation, suppress immunity to primary skin cancers, enabling them to grow. Presumably the irradiated grafts contained cells with gene mutations that enabled them to develop into primary tumors in the presence of the UV-induced Treg. UV effects on the immune system can also inhibit the rejection of transplanted melanoma and nonmelanoma skin cancer cell lines (7–9).

It is well established that human skin tumors can be recognized and destroyed by the immune system, with immune mediated regression of primary human melanoma and nonmelanoma skin cancers being commonly observed (10–12). Many antigens expressed by human skin tumors that can be recognized by lymphocytes, enabling the immune system to cause tumor regression, have been identified, and lymphocytes specific for these tumor rejection antigens can be identified using tetramer technology (13). This provides proof that the immune system is able to reject primary skin cancers in humans, and UV-immunosuppression enables the outgrowth of skin cancer. However, it needs to be noted that immunosuppressive drugs can also inhibit repair of UV-damaged DNA (14).

Clear evidence that the immune system is important for eradicating skin tumors in humans can also be seen by studies showing an association between individual susceptibility to glucocorticoids and SCC in users of oral steroids to suppress immunity (15). Pharmacologically induced immunosuppression in organ transplant recipients also enhances the incidence of skin cancer (16).

UV radiation-induced gene mutations are also essential for the development of skin cancer. Cancer is primarily the result of genetic or epigenetic changes in normal cells that give them a growth advantage. It is also generally recognized that a series of genetic changes are required in order for a cancer to develop; a single mutation is not sufficient for a normal cell to escape all of the regulatory processes that restrict it to a normal growth pattern (17). Stem cells are quiescent or slowly cycling undifferentiated cells that, when stimulated, can maintain and regenerate the tissue. It has been proposed that it is these epidermal stem cells that give rise to nonmelanoma skin cancer when they acquire accumulated mutations (18). The most commonly known UV-induced mutations occur in the tumor suppressor gene p53 (19) and mutations in this gene have been shown to be present in solar-simulated UV-induced skin cancers in mice (20). Clusters of p53-positive epidermal keratinocytes (p53 patches) are clonal outgrowths of keratinocytes that contain mutations in p53 and are related to the formation of carcinomas (21). Other genes are also involved in at least some types of skin cancer, such as the Patched pathway in BCC (22) and BRAF in melanoma (23), although the role of UV in forming these mutations is not clear.

While there is clear evidence that UV-induced immunosuppression and gene mutations are important for skin carcinogenesis, it is less clear whether UV-induced inflammation is also important for skin cancer development.

Skin Cancer can be Induced with Subinflammatory Doses of UV

The minimal erythemal or edemal dose (MED) is the amount of UV radiation that causes just perceptible erythema or edema. This is generally called sunburn. However, the general population does not perceive sunburn until a threshold of about 3 MED is reached, as the procedures for laboratory-based assessment of MED are considerably more sensitive than the broad community-noticing sunburn. Edema, or skin swelling, is generally used to assess the MED in mice, as erythema is difficult to accurately determine in this species. Edema is the skin swelling resulting from inflammation and extravasation of fluids and hence the MED is also the minimum dose of UV that becomes inflammatory. Doses of UV greater than an MED are required to induce migration of macrophages into the skin (24), although it cannot be discounted that low levels of inflammatory cells are present with smaller doses of UV. For the purposes of this review we therefore regard inflammation as being the detectable influx of inflammatory cells that occurs in response to 1 MED or higher doses of UV. We acknowledge that some of the mediators of inflammation are likely to be induced by UV doses that are too low to cause observable inflammation, in the same way that some parameters associated with immunosuppression or genetic damage may occur in response to UV doses lower than those that result in detectable changes to these biologic endpoints.

Sunlight-induced sunburn is primarily caused by UVB, with UVA being 200–300 times less effective than UVB, although UVA can augment the erythemic effect of UVB (25). It has been suggested that DNA is the chromophore for sunburn (26). The MED is very dependent on spectrum (27), with lower UVB wavelengths being more effective. Therefore, both physical dose of UV and spectra need to be taken into account when assessing the inflammatory effect of irradiation.

Epithelial skin cancer

The issue we wish to address here is whether UV-induced inflammation is a prerequisite for skin carcinogenesis. UV has been shown to induce SCC in SKH hairless mice with doses as low as 0.032 MED per day, although tumors arose more rapidly with increasing UV doses (21). Exposure to 8 mJ cm⁻² UV (about 0.08 of an MED) is sufficient to induce SCC and AK with a latency time of 234–238 days, with nearly all mice developing multiple skin tumors (28,29). Therefore, with these low doses the tumors are relatively slow to develop, requiring about 0.5 of the life span of a mouse to form. This is similar to that observed in humans, as the majority of skin tumors in our species do not develop until half way through our life span. Exposure of opossums (Monodelphis domestica) to 25 mJ cm⁻² UV from FS40 sunlamps three times per week for 81 weeks has also been shown to induce nonmelanoma skin cancer in about 75% of animals. This dose is about half an MED in the opossums, again showing that subinflammatory doses of UV are able to induce skin cancers in this animal (30).

The Phillips FS40 sunlamps used in the above experiments emit primarily UVB, but contain a small peak in the UVC region at 250 nm and a disproportionately large amount of low wavelength UVB compared to sunlight. Hence, while these experiments clearly show that very low doses of UV are able to induce SCC in the absence of inflammation in mice and opossums, they do not show whether a spectrum that matches sunlight would be this effective.

SKH mice exposed to 250 mJ cm⁻² UVB per day from FS40 lamps filtered with Kodacel cellulose tri-acetate to remove wavelengths below 290 nm and attenuate the low wavelength UVB have been observed to develop skin cancers in 100% of mice after 24 weeks of exposure (31). While not directly stated by the authors, this dose should be just below the MED in these mice and therefore there may have been a small amount of inflammation with the repetitive irradiations. In a study from our group, UV lamps filtered with Kodacel that delivered a solar-simulated spectrum comprising UVB and UVA found the MED to be 6500 mJ cm⁻², which comprised 290 mJ cm⁻² UVB with 6210 mJ cm⁻² UVA. Irradiation with 1 MED per day induced skin tumors in the majority of mice by 20 weeks of irradiation (32). Hence, again in this study the UV dose would have induced a mild degree of inflammation with repetitive irradiations.

C3H/HeN mice exposed to 454 mJ cm⁻² UVB and 3026 mJ cm⁻² UVA per day from a solar simulator developed skin cancer. While the spectrum of this source closely matched that of sunlight, this was an inflammatory dose of UV that caused sunburn (33). C57BL/6 mice irradiated with 1000 mJ cm⁻² UVB and accompanying UVA from a solar simulator developed skin tumors in 100% of mice by 45 weeks of irradiation (34). This irradiation dose induced sunburn cell formation, and would be inflammatory.

UVB is far more effective at inducing inflammation than UVA, and in murine studies has been estimated to be 1000–10 000 times more effective for inducing skin cancer (35,36). In a study of UVA on skin carcinogenesis it was found that while both UVB and UVA induce SCC and AK (benign skin lesion, precursor to SCC), UVA additionally caused the formation of papillomas. It was found that 5600 mJ cm⁻² UVA induced skin tumors in 50% of mice after 431 days (37). When papillomas were excluded from the analysis it was found that the rate of formation of UVB- and UVA-induced skin tumors was similar, and additive. There was no indication of interactive effects between UVB and UVA either augmenting or reducing carcinogenesis based on the expected effects of the individual wavebands (38). The authors concluded that the additive effects of UVB and UVA need to be considered when estimating the skin cancer risk from UV radiation sources. The average MED in Caucasians is within the order of 5000 mJ cm⁻² solar-simulated UV or sunlight (39). This is greatly dependent upon skin type and is used here only as a guide. While the UVB component is largely responsible for sunburn, there is about 20-fold more UVA than UVB within sunlight. Hence the study described above demonstrates that, in the main, the amount of UVA found in about an MED of sunlight is photocarcinogenic in mice. This required exposure for 431 days, however. As skin cancer is induced over a long period of cumulative exposure in humans, they could easily be exposed to these UVA doses. It would require 200- to 300-fold higher doses of UVA to be inflammatory. Subinflammatory UVA (24 500 mJ cm⁻² day⁻¹) has also been shown to cause SCC in Sencar mice (40). UVA (2500 mJ cm⁻²) has also been shown to induce a single nonmelanoma skin cancer in 30 irradiated opossums (30). Thus, these experiments on UVA photocarcinogenesis clearly show that skin cancer can be induced in the absence of UV-induced inflammation. Inflammation from another cause may however be required for skin cancer development.

It is much more difficult to obtain information on the dose or spectrum of UV responsible for causing skin cancer in humans, and the issue being reviewed here, whether inflammation is involved in this process. Skin obtained from white skinned human donors who underwent cosmetic surgery was grafted onto immunodeficient SCID or Rag-1 knockout mice, and exposed two to three times per week for up to 10 months with 40 mJ cm⁻² UVB or solar-simulated UV consisting of 30 mJ cm⁻² UVB and 3–6000 mJ cm⁻² UVA. The UVB was filtered to remove wavelengths below 295 nm. UVB induced AK in 10% of grafts, and solar-simulated UV in 18% of grafts (41). The data demonstrate that, at least in the absence of an immune system, UV can induce skin cancers in human skin with very low doses, within the order of 10–20% of an average human MED. Also, UVA contributes to the formation of tumors, as there was an increased incidence when UVA was included in the spectrum. Hence this suggests that human skin tumors can be induced with subinflammatory doses of UV.

Melanoma

Tyr-SV40E transgenic 4-day-old mice were exposed to 328 mJ cm⁻² UV, 70% within the UVB range, for four consecutive days, resulting in the development of early melanomas. This transgenic mouse strain contains SV40 oncogenic sequences under the control of the mouse tyrosinase gene promoter, so that SV40 is expressed in pigment cells. This irradiation did not induce erythema in these mice (42) indicating that it is possible to induce melanoma in the absence of UV-induced inflammation. UVA is also able to induce melanoma in heavily pigmented back-cross hybrid fish of the Xiphophorus genus (43) again indicating the induction of melanoma in the absence of UV-induced inflammation. Both UVA (2500 mJ cm⁻²) and UVB (25 mJ cm⁻², about 0.5 MED) have been shown to induce melanocytic hyperplasia in opossums (M. domestica) (30).

However, in some studies inflammatory doses of UV are needed to induce melanoma. A transgenic mouse with metallothioniein gene promoter forced overexpression of hepatocyte growth factor/scatter factor (HGF/SF) required a single sunburning dose of UV to 3.5-day-old mice in order to induce melanoma (44). In a subsequent study irradiation of 3.5-day-old HGF/SF+/+ Ink4a/arf−/− mice with an erythemal dose of UV (624 mJ cm⁻² UVB, 331 mJ cm⁻² UVA and 3 mJ cm⁻² UVC) induced melanomas with a median onset of 50 days, which metastasized to the lymph node and liver (45).

Irradiation of human skin grafted onto Rag-1 mice three times per week for 12 months with 50 mJ cm⁻² UVB (290–320 nm, peak 314 nm) caused the development of solar lentigo in 23%, and melanocytic hyperplasia in 68% of human skin grafts (46). While melanoma was not induced in these studies, the changes in melanocytes are associated with atypia or early malignancy. The induction of these lesions with such a low dose of UV is consistent with subinflammatory UVB being able to induce melanoma in human skin.

In summary, the evidence indicates that skin cancers, both epithelial and melanoma, can be induced with subinflammatory doses of UV. Doses of UVB more than 10-fold lower than are required to cause inflammation can induce SCC in animal models. Benign skin neoplasms also form on human skin grafted onto immunodeficient mice in response to subinflammatory UVB. Additionally, UVA at doses several hundred fold too low to be inflammatory can induce skin cancers in animal models and at least contribute to photocarcinogenesis in human skin grafted onto immunodeficient mice. Similarly, subinflammatory UVB or UVA can cause melanocytic lesions in a variety of animal models and in human skin grafted onto immunoincompetent mice. This shows that it is possible for subinflammatory doses of sunlight to case skin cancer. However, the relative importance of subinflammatory compared to erythemal doses of sunlight in causing skin cancer in humans remains unknown. It is also a large leap from these animal models to understanding sunlight-induced skin cancer in humans.

Subinflammatory Doses of UV are Immunosuppressive in Humans

While it cannot be directly experimentally determined whether or not inflammatory doses of UV are required for photocarcinogenesis in humans, photoimmunosuppression and gene mutations can be studied as surrogate markers because these two biologic events are essential for skin cancer induction. Photocarcinogenesis in humans could only occur in response to UV doses too low to cause inflammation if these low doses are able to suppress immunity and mutate genes in humans.

There are several different models to describe UV-induced immunosuppression. The local model is when the antigen is applied to UV-irradiated skin of unimmunized individuals. Systemic immunosuppression is when the antigen is applied to an unirradiated skin site of an unimmunized individual who has been exposed to UV at a different skin site. An established or memory immune response can also be suppressed by UV irradiation. UV can suppress contact hypersensitivity (CHS) where the antigen, usually a chemically reactive hapten, is applied topically to the skin surface, and delayed type hypersensitivity (DTH) where the antigen, usually proteins or whole cells, are injected into the dermis or subcutaneously.

Exposure to 144 mJ cm⁻² UVB daily for four consecutive days, which is too low to cause inflammation, is able to induce local immunosuppression in humans (47). A single exposure to 4 MED, but not lower doses, also inhibited the induction of immunity induced by antigen application to a skin site distal to that which received the UVB (48). The reactivation of memory immunity is suppressed by subinflammatory doses of UVB in humans. Daily irradiations with the suberythemal dose of 144 mJ cm⁻² UVB caused increasing levels of immunosuppression with increasing numbers of daily irradiations (49). Increasing doses of single exposures to UVB also cause greater levels of immunosuppression in humans with the suppression being discernible within 48 h of irradiation with 300 mJ cm⁻² UVB, which is subinflammatory (50). Thus, subinflammatory doses of UVB can cause local immunosuppression. Inflammatory doses may be required for systemic immunosuppression in humans, while doses of UVB as low as 12.5% of an inflammatory dose can cause systemic immunosuppression in mice (51). However, skin cancers arise locally on UV-irradiated skin and therefore the local model is more likely to be applicable to skin cancer development in humans.

UVA is also immunosuppressive at doses much lower than are required for this waveband to cause inflammation. Irradiation with 1900 mJ cm⁻² UVA suppresses reactivation of memory immunity to nickel in humans (49). In mice, UVA suppresses the induction of local immunosuppression (52) and the reactivation of memory immunity (53). Systemic suppression of the induction of primary CHS can occur with three daily exposures to the relatively low UVA dose of 1680 mJ cm⁻² (51). UVA also causes systemic suppression of recall immunity to the Multitest kit Merieux in humans. This is a DTH to seven antigens to which most humans are immune (54). UVA has been shown to suppress this response in humans, both locally and systemically (55). Thus, UVA at doses far too low to be inflammatory can cause local and systemic immunosuppression in humans and mice.

Humans are not exposed to pure UVB or UVA, but to sunlight, which consists of a mixture of these two wavebands. The first study to show that such a mixture is immunosuppressive in humans used a commercial solarium with UV containing a mixture of about 90% UVA and 10% UVB (56). In this study Hersey et al. irradiated human volunteers for 12 × 0.5 h exposures on consecutive days, which reduced the induction of CHS to dinitrochlorobenzene. While the doses were unable to be accurately defined, the UVA component was about 10 000 mJ cm⁻² and therefore the UVB portion of this radiation could have been inflammatory. Solar-simulated UV doses as low as 0.25 of those required to cause an MED suppresses the induction of immunity at the irradiated site to dinitrochlorobenzene in humans of skin phototype I/II, although higher doses were required to suppress individuals with a higher skin phototype (57). A single exposure to 1740 mJ cm⁻² solar-simulated UV, which is less than half of what is required to cause inflammation has also been shown to suppress the reactivation of memory immunity to nickel in humans (50). A progressively increasing solar-simulated UV dose regime of 10 daily exposures to an average of 1.45 MED per day suppresses recall DTH to Merieux (55). A dose of 1820 mJ cm⁻² solar-simulated UV which is only half that required to cause barely detectable inflammation can cause systemic immunosuppression in mice (51). Both subinflammatory and inflammatory doses of solar-simulated UV can suppress the induction of primary contact sensitivity and secondary immunity where mice are resensitized 3 weeks after the initial sensitization (24). This clearly shows that UV that mimics the UV portion of the solar spectrum is immunosuppressive in humans and mice at subinflammatory doses.

Subinflammatory Doses of UV are Mutagenic

Mutations that give cells a growth advantage—inhibit cell death by apoptosis or other mechanisms, establish genomic instability or overcome senescence—can all contribute to the development of cancer. Such mutations are an essential step in the formation of skin cancer.

UV at sufficiently high doses induces inflammation with a concomitant accumulation of inflammatory cells in the skin. Increased production of nitric oxide and prostaglandins are two of the mediators of these events (58–60). The inflammatory cells produce reactive oxygen species (ROS) that cause damage to DNA. As ROS from inflammatory cells can cause gene mutations, it seems to be a reasonable hypothesis that UV-induced inflammation results in genetic damage that exacerbates UV-induced DNA mutations. While photodamage to DNA is important for photocarcinogenesis, it is less clear whether DNA damage resulting from UV-induced inflammation, or inflammation resulting from a cause other than UV, contributes to photocarcinogenesis.

UVB absorbed by two adjacent cytosine (C) or thymine (T) residues in DNA causes the formation of cyclobutane pyrimidine dimers. This photodamage can result in GC to AT mutations at dipyrimidine sites if it is not repaired prior to cell division. These mutations only occur in response to UVB and can be regarded as UVB fingerprints (61,62). UVA induces the fingerprint mutations AT to CG at high frequency (63,64). Guanine has the lowest oxidation potential, and therefore is the most sensitive DNA base to oxidation and ROS-mediated mutagenesis (65,66). Oxidation of guanine and resulting mutations can occur either as a direct result of UV-induced production of ROS, or indirectly from ROS produced by inflammatory cells if the UV dose is high enough to be inflammatory.

We recently microdissected groups of about 20 keratinocytes from human SK and SCC and analyzed mutations in the p53 gene. Using the criteria described above, we grouped the mutations into those most likely caused by ROS, UVB or UVA. About one third of the mutations in AK were found likely to be caused by sunlight, with an equal number resulting from UVA and UVB. The major difference between SCC and their precursor lesion, AK, was a large increase in SCCs in mutations likely to be caused by ROS, with sunlight-related mutations being similar in the two tumor groups (67,68). This suggests that the mixture of UVA and UVB in sunlight is largely responsible for the mutations that lead to AK, but the main factor driving progression to malignancy is ROS arising from a source other than UV, possibly inflammatory cells.

This is consistent with our hypothesis that sunlight causes gene mutations in normal keratinocytes that result in the formation of benign skin lesions, while inflammatory responses, developing for unknown reasons as a postsunlight exposure event, drive progression toward malignant SCC.

However, this hypothesis could only be correct if doses of sunlight too low to cause inflammation are able to cause gene mutations. Doses as low as 50 mJ cm⁻² UVB induce photolesions in DNA of human skin (69). Chronic irradiation with 0.5 MED UV (25 mJ cm⁻² with a spectra comprising predominantly UVB) for 59 or 75 days has been shown to cause the formation of p53 patches which are groups of cells with increased expression of the p53 protein. The majority of these patches contained mutant p53 (21). We have shown that doses of UVB and solar-simulated UV as low as those found in 0.25 of an MED, are able to induce mutations in the p53 gene of human keratinocytes in engineered human skin that resembles human skin structurally and biochemically (X.X. Huang, F. Bernerd and G.M. Halliday, unpublished). Cell lines established from tumors that arose in mice irradiated daily 3 times per week with 16.8 mJ cm⁻² UV from FS40 sunlamps have been observed to contain mutations in N-ras (70).

UVA (18 000 mJ cm⁻²), which is too low a dose to cause inflammation, induces mutations in irradiated fibroblasts (71). The lower dose of 7680 mJ cm⁻² UVA has also been shown to induce genetic damage to fibroblasts in the presence of riboflavin which has been proposed to be a natural chromophore for UVA (72). In other studies, 20 000 mJ cm⁻² UVA has been shown to induce mutations in cultured fibroblasts (73). A UVA dose that was less than half the dose that was able to cause inflammation (40 000 mJ cm⁻²) has been shown to induce mutations in human skin (74).

It is therefore clear that doses of UVB, UVA and solar-simulated UV too low to cause inflammation are able to induce mutations. However, this does not exclude a role for ROS from inflammatory cells contributing to skin carcinogenesis. Indeed ROS from inflammatory cells may be important for tumor progression.

Inflammation and Skin Carcinogenesis

Despite UV being able to induce skin cancer at doses high enough to cause immunosuppression and gene mutations, but too low to cause inflammation, inflammation is likely to be important for skin carcinogenesis. UV may not directly cause the inflammation. Tumors, even those not caused by UV, often produce an inflamed environment (75,76). Proinflammatory molecules have been shown to have important roles in the pathogenesis of skin cancer, as determined from clinical or experimental animal studies, and the activity or quantities of many of these are affected by UV radiation (77,78). Cancer cells can themselves produce these proinflammatory molecules, such as chemokines, interleukins (ILs), tumor necrosis factor (TNF), colony-stimulating factors and prostaglandins. They can cause vascular changes that allow accumulation of fluid in the tissues and the margination and extravasation of leukocytes via postcapillary venules. Depending upon the microenvironment of the extravascular tissues into which they enter, monocytes can differentiate into macrophages, dendritic cells or Langerhans cells. The cytokines produced by the cancer cells can then act directly upon the leukocytes, modulating their immune responsiveness and their ability to degrade the extracellular matrix. For example, we have recently shown that transforming growth factor-β produced by skin tumors prevents migration of Langerhans cells from skin tumors to draining lymph nodes, thus inhibiting the development of antitumor immunity (79,80). The leukocytes themselves release soluble factors that affect the behavior of the epidermal cells, as well as endothelial and other cells of the dermal stroma. Thus, inflammation can arise as a result of tumor progression, and tumors of the skin are no exception. However, the question arises as to whether an external source of inflammation such as high-dose UV is required to promote the formation of tumors from initiated skin cells or to induce progression of benign lesions to malignancy. A second, related question is whether this inflammation is necessary for the tumor to retain malignancy, or is merely a by-product of malignancy.

Inflammatory mediators are involved in melanoma, as well as epithelial skin cancer. Eicosanoids and endothelins enhance melanocyte growth and production of melanin, while a number of inflammatory cytokines including IL-1, TNF and IL-6 inhibit these effects on melanocytes. A review of these effects concluded that inflammatory mediators and cytokines alter melanocytes so that they resemble nevus cells and melanoma at least in vitro (81).

Evidence that supports a requirement for inflammation or inflammatory mediators for the development of skin cancer comes from both clinical studies and experimental mouse models. In a clinical study by our group (82), skin lesions that were clinically identified as AKs, but were also inflamed and painful, presented as SCCs in about 50% of cases when examined histologically, implying that the onset of inflammation coincided with progression. The remaining 50% of cases were AK with an increased level of dysplasia. While difficult to assess, there was no indication that the inflammation was caused by recent sun exposure. Immunohistochemical analysis of differentiation and other markers of tumor progression were consistent with inflammation-induced progression into SCC. However, support for UV-induced inflammation as a contributing factor to skin carcinogenesis comes from experimental mouse studies in which UV radiation of skin facilitates SCC growth at inflammatory, but not subinflammatory doses (9,24).

Interestingly, a recent study showed that male mice are more sensitive to UVB-induced skin carcinogenesis than female mice, which is consistent with men having a higher incidence of skin cancer than women (83). This study used an inflammatory dose of UVB to induce the cancers, and while women developed a larger inflammatory response to UVB, men were found to have lower antioxidant protection in the skin resulting in a higher level of oxidative damage to DNA. It has also been shown that men are more sensitive to UV immunosuppression than women (84). It therefore appears that UV-induced immunosuppression and DNA damage but not inflammation were associated with this increased pattern of skin carcinogenesis in men more than in women. This is consistent with UV-induced immunosuppression and gene mutations, but inflammation from a different source, driving skin carcinogenesis.

UV-Regulated Proinflammatory Mediators in Skin

The expression of the inflammatory mediators, IL-1α and IL-1β, which bind to a common receptor, is induced in keratinocytes within 6 h of exposure to UVB doses of 50–100 mJ cm⁻² (85) or 200–300 mJ cm⁻² (86). In contrast, UVA (10–20 000 mJ cm⁻²) suppresses basal and UVB-induced IL-1 activity (85). IL-1 induces other proinflammatory molecules and extracellular matrix-degrading proteinases in keratinocytes (87). TNF mRNA is induced strongly within 4 h of exposure to subinflammatory doses of both UVA and UVB, at least in part through an increase in mRNA stability, and a higher dose of UVB induces a second wave of synthesis 48 h after irradiation (85,88,89). Low-dose UVB also induces the CXCR2 ligands, IL-8 and Gro-α in keratinocytes (7 mJ cm⁻²) (90) and melanoma cells (20–40 mJ cm⁻²) (91), thereby increasing the tumorigenic capacity of melanoma cells. The expression of CXCR2 ligands can be induced in keratinocytes by other UV-inducible cytokines, including TNF, via an NF-kappa B-dependent mechanism and mediate protein kinase C-induced cutaneous inflammation (92). Macrophage migration inhibitory factor (MIF) is produced by human keratinocytes when as little as 25 mJ cm⁻² UVB is administered, although maximal levels require 100 mJ cm⁻² (93). MIF released by T lymphocytes and macrophages represses p53 transcription (94), resulting in enhanced proinflammatory responses in macrophages, including the up-regulation of expression of cyclo-oxygenase-2 (COX-2) (95), and presumably the MIF released by keratinocytes would also contribute to these effects. It also confers susceptibility to transformation via mechanisms that require the cell cycle regulatory transcription factor family, E2F (96). UVA (60–10 000 mJ cm⁻²) stimulates MIF production by keratinocytes that bypasses cytotoxicity while maintaining granzyme B-mediated extracellular matrix degradation (97).

Matrix metalloproteinase-9 (MMP-9) is an extracellular proteinase whose substrates include basement membrane proteins and chemokines (98,99). MMP-9 is produced at high levels by keratinocytes in response to proinflammatory cytokines and growth factors (87), allowing for constitutive MMP-9 production by epithelial cells that produce these cytokines or have active receptors. MMP-9 processes IL-8, increasing its activity by an order of magnitude and thereby enhancing the attraction of leukocytes and endothelial cells. Whereas single low-dose exposure to UVA inhibits MMP-9 production in keratinocytes (100), chronic UVA irradiation of HaCaT cells results in a cell line that is resistant to apoptosis and results in constitutively high levels of MMP-9 (101). Suberythemal doses of UVB also stimulate MMP-9 in human epidermis, via induction of AP-1 and NF-kappa B activities (102). Cells of the monocytic lineage are another, potentially abundant, source of MMP-9. As well as proinflammatory cytokines, lipopolysaccharide (LPS) from pathogenic bacteria induces its expression, and LPS-associated bacterial proteinases activate its proenzyme form, enabling it to degrade extracellular matrix proteins and serine proteinase inhibitors (103).

UV-Regulated Signaling Pathways Stimulate Production of Inflammatory Mediators

Although UVA and UVB are both carcinogenic, they induce different signaling pathways (104,105). Many of the signaling pathways activated by UV are triggered as a consequence of ROS production (4). However, cis-urocanic acid, produced by UV in the skin when it is absorbed by the trans-isomer, both immunosuppresses and potentiates inflammatory edema, in part through the activity of neuropeptides (106). A low dose of UVB (2–40 mJ cm⁻²) activates mitogen-activated protein kinase (MAPK) pathways in multiple cell types by causing a rapid clustering of the receptors for epidermal growth factor, IL-1α and TNF (107) through the action of ROS (108). ROS generated by low-dose UVB (25 mJ cm⁻²) induce c-fos transcription in keratinocytes through cAMP response element binding protein-binding sites in its proximal promoter (109), via Akt (protein kinase B) (110) and p38 MAPK pathways, with the extracellular signal-regulated kinase (ERK) MAPK pathway playing a minor role (111). The latter is important for maintaining cell viability following UVB irradiation (108). UVB (10 mJ cm⁻²), but not UVA (30 000 mJ cm⁻²) also produces tryptophan derivatives that can act as ligands for the arylhydrocarbon receptor. This can directly activate the transcription of genes such as cytochrome P450 1A1 and, by mechanisms that remain to be fully elucidated, cause internalization of the epidermal growth factor receptor and activation of the ERK MAPK pathways (112). Low-dose UVB (25 mJ cm⁻²) induces COX-2 transcription and mRNA stability in keratinocytes via p38 MAPK (113) and Akt (114), both of which are activated by epidermal growth factor receptor.

NF-kappa B activity is up-regulated in many types of cancer, including skin cancers, and contributes to their pathogenesis through the induction of proinflammatory mediators and inhibition of apoptosis (115). Keratinocytes respond rapidly to 40–60 mJ cm⁻² UVB by up-regulating RelA-mediated NF-kappa B activity (116) by a mechanism that is independent of I-kappa B degradation and Akt (117). The response of keratinocytes to NF-kappa B pathway components is distinct from other cell types. Overexpression of the inhibitor of NF-kappa B, I-kappa B alpha, in the epidermis leads to the spontaneous formation of cutaneous SCC, despite an increase in spontaneous apoptosis and apoptosis induced by irradiation with 50–100 mJ cm⁻² UVB (118). For many years UVB was considered to be the main contributor toward skin cancers, based largely on the DNA action spectrum of UV radiation (119,35), but UVA has more recently been appreciated as having an important role in skin carcinogenesis. This is in keeping with the growing body of evidence that direct molecular targets of UV other than DNA are important to carcinogenesis. UVA, although it does not produce an inflammatory response at doses normally encountered by people, also produces ROS and so, unsurprisingly, activates many of the same signaling pathways as UVB.

UVA (25 000–30 000 mJ cm⁻²) generates intracellular singlet oxygen in keratinocytes, which in turn activates the c-Jun N-terminal kinase (JNK) and p38 MAPK pathways (120) and transcription of c-fos (121). Unlike the cases with UVB and UVC, the ERK1/2 MAPK pathway is not activated. This dose of UVA liberates bound Fe²⁺ in keratinocytes, which causes oxidative damage to membrane lipids and subsequent activation of NF-kappa B (122,123). It also promotes keratinocyte death by caspase-dependent apoptosis and other pathways, but at the same time activates the JNK pathway, which confers at least partial protection against apoptotic cell death (124).

Thus, many of these UV-regulated signal transduction pathways result in the production of inflammatory mediators at low UV doses. This does not result in gross signs of inflammation, such as edema, at these UV doses. Therefore, there needs to be a distinction between detectable levels of inflammation and production of inflammatory mediators at doses or combinations that do not induce obvious inflammation. Whether clinically evident inflammation is necessary for skin cancer is unknown.

Inhibition of Inflammation Reduces Skin Carcinogenesis

A number of plant-derived chemicals show promise as chemopreventative agents for UV-induced damage to skin, particularly the promotion and progression stages of carcinogenesis (4). A common property of many of these chemicals is an anti-inflammatory effect. Studies have shown that green tea polyphenols inhibit production of ROS in human skin (125), thereby preventing the activation of downstream signaling pathways. Green tea polyphenols also have the dual biologic effects of inhibiting the migration of inflammatory leukocytes into UVB-irradiated (90 mJ cm⁻²) mouse skin (126) and inhibiting UVB carcinogenesis (127).

The importance of inflammation for skin cancer is illustrated by the utility of Solaraze (3% diclofenac), a topical COX-2 inhibitor, for successful treatment of AK (128,129). This reduces downstream products of arachidonic acid production such as prostaglandins. COX-2 is highly expressed in AK and SCC (130). COX-2 inhibition has also been reported to cause regression of metastatic melanoma (131). As COX-2 is an important proinflammatory enzyme, this implies that inflammation is necessary for the survival of at least some of these skin tumors. Inhibition of COX-2 with diclofenac results in skin cancer cells dying by apoptosis (132), suggesting that inflammatory mediators are required for survival of skin cancer cells rather than induction of inflammation. Hence the activation of proinflammatory genes may be important for skin cancer survival and progression.

Human epithelial skin cancer cells have high COX-2 levels, which are increased in response to UVB (240 mJ cm⁻²) (133). The use of nonsteroidal anti-inflammatory drugs (NSAIDs) by patients significantly reduces the risk of BCCs, SCCs and number of AKs (134,135), indicating a contributory role for COX-2 in the development of those lesions. Analysis of single nucleotide polymorphisms (SNPs) has also implicated a role for COX-2 in skin cancer. Inheritance of the COX-2 allele that has a G at the SNP at position −1195 in the COX-2 promoter is protective from BCC (135). This allele cannot bind the transcription factor myb, whereas the A allele can, and NSAID use may be protective for the A allele. Inheritance of the C allele at SNP T8473 in the 3′ untranslated region conferred a two-fold increased relative risk of BCC that was not affected by NSAID use.

Support for a skin carcinogenic role for COX-2 also comes from animal models. COX-2 inhibitors are skin tumor preventative, reducing the yield of UVB-induced tumors in hairless mice (136,137), and when co-administered with an ornithine decarboxylase (ODC) inhibitor can also cause pre-existing tumors to regress (138). Topical treatment with COX-2 inhibitors prevented UVB-induced (224 mJ cm⁻²) oxidation of guanine, and the mice had fewer papillomas that were slower to appear (137). Sulindac, a COX inhibitor, reduces inflammatory responses to UVB in SKH mice (139) and chemically induced skin neoplasia, independently of peroxisome proliferator-activated receptor beta (140).

Knockdown of the COX-2 gene in mice reduces photocarcinogenesis while overexpression of this gene under the control of a keratin 14 promoter enhanced photocarcinogenesis, showing that COX-2 is required for the development of skin cancers (141). In these experiments the mice were given an incremental irradiation regime from an initial dose of 90 to a final dose of 175 mJ cm⁻² UVB from FS20 sunlamps. This UV dose was suberythemic in these mice, and doses of 220 mJ cm⁻², which were high enough to induce sunburn cell formation, were required to increase prostaglandin E₂ production. This suggests that inflammation is not required during tumor induction with UV, but is important for tumor survival and development at the post-UV stage. Conversely, K14-COX-2 mice on an FVB background are more resistant to dimethylbenzanthracene (DMBA)-initiated skin carcinogenesis than wild types (142), although this effect has subsequently been shown to be specific for the use of phorbol esters as a tumor promoter; tumors developing in DMBA-initiated mice without any promoter treatment or following promotion with anthriline were more numerous (143).

Constitutive expression of IL-1α in mouse epidermis is sufficient to cause a chronic inflammatory disease involving recruitment of macrophages (144). Interestingly, when these mice were subjected to the DMBA/12-O-tetradecanoyl phorbol-13-acetate (TPA) initiator/promoter protocol for producing skin tumors, the incidence of papillomas and papilloma-derived carcinomas fell dramatically, despite the mice retaining a hyperproliferative response to TPA (145). Recruitment of lymphocytes was not involved in the response, because mice with IL-1α overexpression on a Rag−/− background also developed no papillomas. In contrast, when the IL-1α-overexpressing mice were subjected to prolonged treatment solely with the mutagen, DMBA, the mice developed SCCs more early than control mice and they developed de novo, rather than from pre-existing papillomas, which were absent from these mice. TNF-α is required for chemical carcinogenesis of the skin, as determined by reduced tumor incidence in mice knocked out for this gene (146,147), TNF receptor genes (148) or TNF activity, using antibodies (149).

Cells of the monocyte lineage require MMP-9 for their trafficking to the skin. Mice whose macrophages lack this molecule have diminished tumor formation in a transgenic model of skin cancer, providing evidence for the importance of both MMP-9 and infiltrating macrophages to skin carcinogenesis (150). Mast cells are required for activating pro-MMP-9 and providing an angiogenic stimulus in early-stage skin tumorigenesis (151). Mice overexpressing ODC in the epidermis under the control of a keratin-5 promoter develop tumors from an initiation dose of UVB without the need for exogenous tumor promoters (152). Consumption of a synthetic ODC inhibitor inhibited the promotion effect of ODC overexpression.

Thus, there is good evidence that inflammatory mediators are important for skin cancer, and inhibition of inflammation reduces skin carcinogenesis, but it is not clear whether or not observable levels of inflammation are required. The inflammatory mediators may work via mechanisms other than induction of inflammation.

Conclusions

The clinical correlation between inflammation and AK progression to SCC seen in human skin, where UV is the main driver of promotion and progression (82), and the increase in ROS-mediated genetic damage in SCC compared to AK (67,68) suggests an association between inflammation and carcinogenesis. While the inflammation might be a by-product of carcinogenesis and not be required for its development or maintenance, the success of anti-inflammatory strategies at reducing skin cancer argues against this. There is an abundance of evidence from studies with genetically modified animals that genes that contribute to inflammation (TNF-α, TNF receptors, MMP-9, COX-2, etc.) also contribute to skin cancer. However, it remains a possibility that the procarcinogenic role of these mediators of inflammation can be maintained at levels lower than those required to cause inflammation, as is the case for UV itself. Therefore, these mutations may not cause inflammation, but drive carcinogenesis via a different mechanism.

The tumor-promoting ability of proinflammatory molecules such as TPA and the behavior of the K5-ODC transgenic mice to initiating doses of UV (152) are suggestive of a role for inflammation in promotion. A systematic experimental comparison of the tumor promotion efficacies of well-known substances such as TPA with those of other nonmutagenic inducers of inflammation that either activate different specific pathways or act as nonspecific inducers of inflammation might be enlightening as to the real role of inflammation in tumor promotion.

If it is not actually necessary, does UV-induced inflammation nevertheless contribute to carcinogenesis when it does occur? Although mutations and immune suppression can occur at subinflammatory doses of UV (Fig. 1), the influx of inflammatory leukocytes into the skin provoked by inflammatory doses of UV would be expected to produce prolonged damage from ROS that they produce. The sheer extra dose or duration of ROS provided by the infiltrating inflammatory cells might be sufficient to overwhelm the DNA repair mechanisms of epidermal cells, creating a “tipping point” toward an abundance of extra mutations. Particularly vulnerable would be cells in which prior mutations (such as p53) have already perturbed programmed cell death pathways. If inflammation is required for progression to, or maintenance of, the malignant phenotype, then mutations that induce a proinflammatory gene expression signature in the cancer cells themselves could provide the necessary inflamed microenvironment and aid progression toward malignancy.

Figure 1.  Skin cancer occurs in response to lower doses of UV than are required to cause inflammation. Skin cancer, immunosuppression and gene mutations can all occur in response to lower doses of UV radiation than are required to cause inflammation. See text for details and references.

For this reason chemoprevention with agents that have anti-inflammatory properties are likely to be beneficial for reducing the incidence of skin cancer, even when UV-induced inflammation is not required for photocarcinogenesis. Prof. Mukhtar’s research has highlighted a number of photochemopreventive compounds with such activity, and has helped elucidate these complex and as yet unresolved issues (1–5).