Chapter 111. Chemical Carcinogenesis

According to the American Cancer Society, it is estimated that over 2,000,000 cases of basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), 46,770 cases of melanoma in situ, and 68,130 cases of invasive melanoma will be diagnosed that in the year 2010.1 The precise number of nonmelanoma skin cancers is unknown since they are not routinely reported to registries, but recent data suggest that previous approximations are likely to have been underestimates.2 Although only a small fraction of patients with nonmelanoma skin cancer will die as a result of their cancer, which is SCC in nearly all cases, the frequency of these cancers nonetheless results in an estimated 1,000 and 2,000 deaths per year. In contrast, although much less common than nonmelanoma skin cancer, melanoma has a continually rising death rate now estimated at 8,700 per year, and it is currently predicted that the lifetime risk in Caucasians of developing melanoma is a staggering 1 in 37 for males and 1 in 56 for females.1 Because they are so common, cutaneous cancers have a major impact on health care costs: in addition to the mortality burden, treatment is associated with considerable morbidity and cosmetic defects. For these reasons, understanding the etiology and pathogenesis of these malignancies is a significant public health goal, and development of rational nondeforming therapies to reduce morbidity and mortality is urgently needed. The high prevalence of skin cancer, the external location of the tumors, and well-defined preneoplastic lesions all provide an excellent opportunity for studying the factors regulating cutaneous cancer induction in humans. Those qualities that facilitate the study of cutaneous neoplasms in human populations have also been useful in establishing relevant animal models. Advances in molecular genetics, keratinocyte cell culture, and development of genetically altered mice and reconstructed human skin models have greatly facilitated the analysis of basic mechanisms of cutaneous carcinogenesis. Our main focus in this chapter will be on nonmelanoma skin cancer: the reader is referred to Chapter 124 for further discussion of melanoma.

The ultraviolet radiation (UVR) in sunlight is the primary etiologic agent for all skin cancers, and thus UVR is the major carcinogen in the human environment. The powerful carcinogenic activity of UVR is attributable to its ability to damage DNA and cause mutations, its capacity to clonally expand incipient neoplastic cells whose altered signaling pathways provide a survival advantage in the face of ultraviolet-induced cytotoxicity, its ability to induce reactive oxygen species, and its activity as an immune suppressant (see Chapter 112). The association of UVR with skin cancer is so strongly supported by clinical, epidemiologic, and experimental data that it represents the most clear-cut etiologic factor in human malignancy.

Also implicated in the development of human skin cancer are various chemicals, as a result of environmental, occupational, or medicinal exposures (Table 111-1). In 1775, Sir Percivall Pott3 attributed the increased incidence of scrotal cancer in chimney sweeps to repeated exposure to soot. This report provided the first link between occupational exposure and the development of cancer as well as the first example of chemical carcinogenesis. According to the National Toxicology Program's 11th Report on Carcinogens, published in 2005 (http://ntp.niehs.nih.gov/ntp/roc/toc11.html), 246 agents are listed as known or likely human carcinogens, and the great majority of these are chemicals. In the last 35 years, the mechanisms by which chemicals cause cancer have been unraveled, and reveal striking similarities to the properties responsible for UVR carcinogenicity, namely DNA damage, selective cytotoxicity, and immune suppression.

The carcinogenic potential of coal and petroleum derivatives is now firmly established as a result of experimental animal studies and epidemiologic reports. Petroleum products, grease as well as insecticides, herbicides, and fungicides are particularly pathogenic for SCC, while fiberglass and dry-cleaning agents increase the incidence of BCC.4 Cigarette and pipe smokers have an overall twofold increased risk for cutaneous SCC, and the risk increases with the intensity of the tobacco use.5 Arsenic exposure is associated with the development of premalignant keratoses, Bowen's disease, SCC, and BCC, as well as a number of internal malignancies.6 While Fowler's solution (1% potassium arsenite) is no longer used in medical practice, certain herbal medicines are still a source of arsenic exposure. Occupational exposures to arsenic as a component of agricultural pesticides, sheep and cattle dip, mining and smelting, glass manufacturing, and other industries are well documented. A more insidious source of environmental arsenic exposure is contaminated drinking water or shellfish, and analysis of affected populations shows a dose-dependent increase in skin cancer.7 Mouse models have provided evidence for an interaction of ingested arsenic acting as a cocarcinogen with solar radiation to increase the frequency and size of cutaneous carcinomas and reduce the latency period.8,9

A variety of medications have been associated with the development of skin cancer. Systemic treatment with immunosuppressive agents results in an increased incidence of both benign and malignant skin lesions,10 which is generally attributed to reduced immune-surveillance of nascent tumor cells. However, a more specific mechanism predisposing to skin cancer has been proposed for azathioprine, which may sensitize DNA to ultraviolet A (UVA) radiation via its metabolite 6-thioguanine, leading to the production of mutagenic reactive oxygen species.11 In addition, cyclosporine A inhibits DNA repair and cell death in UVB-irradiated keratinocytes,12 raising the possibility that this compound may also increase photocarcinogenesis. Topical nitrogen mustard also increases the risk of developing skin cancer.13 Despite the clear-cut association between occupational exposure to tars and skin cancer, the use of coal tar in patients with psoriasis or eczema does not appear to increase the risk of developing skin cancer, based on a large study including over 13,000 patients.14 In contrast, the use of ionizing radiation to treat a variety of skin diseases increases the risk of BCC in all treated patients, and SCC in individuals with sun-sensitive skin.15

Systemic administration of 8-methoxypsoralen combined with UVA treatments (PUVA) is associated with a dose-dependent increase in the risk of developing cutaneous SCC that persists following discontinuation of therapy.16 Patients receiving high-level exposures (≥337 PUVA treatments) had over a 100-fold increase in SCC incidence while those receiving <100 treatments had a fivefold higher than expected incidence. 3.8% of individuals with PUVA-associated SCC developed metastases, and SCCs developed on the male genitalia much more frequently than in the general population.17 The incidence of BCC and melanoma is also increased in PUVA patients.18,19 PUVA is thus a potent stimulus for the induction of epithelial skin cancer in humans and also induces SCC in mice.20 While this implies that PUVA is a complete carcinogen by virtue of its DNA damaging properties, the frequent detection of human papilloma virus DNA in PUVA-induced cancers suggests that immunosuppression may also contribute to carcinogenesis in this setting.21 In contrast to PUVA, UVB phototherapy does not appear to be associated with an increased risk of developing nonmelanoma skin cancer.22 There is strong evidence implicating the use of tanning beds in the development of both nonmelanoma skin cancers and melanoma.23 The World Health Organization now places tanning beds in the category of group I carcinogens,24 which will hopefully help bring about policy changes leading to strict regulation of the tanning bed industry.25

How can such a diverse group of chemical and physical agents contribute to cutaneous cancer when the great majority of environmental agents to which humans are exposed are not carcinogenic? We now know that carcinogens can be genotoxic, nongenotoxic, or both.26 Genotoxic carcinogens have high chemical reactivity (such as alkylating agents like nitrogen mustard) or can be metabolized to reactive intermediates by the host (such as petroleum products). They form covalent adducts with macromolecules and target DNA in the nucleus and mitochondria.27 Since there is a good correlation between the ability to form covalent DNA adducts and the potency to induce tumors in laboratory animals, DNA is considered the ultimate target for most genotoxic carcinogens. The interaction with DNA is not random, and each class of agents reacts selectively with purine and pyrimidine targets.27 Furthermore, targeting of carcinogens to particular sites in DNA is determined by nucleotide sequence, by host cell, and by selective DNA repair processes making some genetic material at risk over others. As expected from this chemistry, genotoxic carcinogens are potent mutagens, particularly adept at causing base mispairing or small deletions, leading to missense or nonsense mutations. Others may cause macrogenetic damage such as chromosome breaks and large deletions. In all cases, mutations detected in tumors represent a combination of the effect of the mutagenic change on the function of the protein product and the effect of the functional alteration on the behavior of the specific host cell type.

A number of chemicals that cause cancers in laboratory rodents and contribute to human skin cancer incidence are not demonstrably genotoxic.28 Synthetic pesticides and herbicides, dry-cleaning reagents, and arsenic may fall within this group. The mechanism of action by nongenotoxic carcinogens is controversial and may be related in some cases to toxic cell death and regenerative hyperplasia. Induction of endogenous mutagenic mechanisms such as DNA oxyradical damage,29 depurination of DNA, and deamination of 5-methylcytosine may contribute to carcinogenicity of these agents. Nongenotoxic carcinogens may interfere with host protective mechanisms as has been suggested for the action of arsenic in suppressing DNA repair or inhibiting the activity of tumor suppressor genes.8 Thus, nongenotoxic carcinogens may serve as modifiers in concert with genotoxic agents such as UVR.

While a number of carcinogens can directly interact with DNA, many require bioactivation by cellular metabolic enzymes to form a compound that can react and form adducts with DNA. In general, these reactive intermediates are inadvertent byproducts of xenobiotic detoxification pathways. These pathways are complex and interactive,30 and genetic polymorphisms both in animal models and humans contribute to cancer susceptibility.31 This genetic variation in metabolic enzymes leading to differing rates of biotransformation and detoxification among individuals provides an approach to estimate individual risk profiles for particular chemical exposures. Initial stages in metabolic activation are most often carried out by cytochrome P450s, a class of heme containing monooxygenases, although other enzymes can be involved.32,33 Oxidation of carcinogens at carbon–carbon double bonds or saturated carbon atoms, or oxidation or reduction at nitrogen moieties, yields reactive intermediates that can be further metabolized and detoxified or can bind to DNA. Further biotransformation involves enzymatic conjugation of one of several different chemical groups such as glucuronide, glutathione, acetyl, or sulfate to reactive intermediates to enhance elimination. However, these conjugation reactions can also activate as well as detoxify carcinogens.30 A number of metabolic pathways and enzymes such as cytochrome P450s are induced by carcinogen or other chemical exposures through interaction of these xenobiotics with the aryl hydrocarbon receptor (AhR) a cytoplasmic receptor that functions as a transcription factor for detoxifying enzymes.34 Thus, carcinogen exposure induces detoxifying enzymes, which in turn increase cancer risk through biotransformation. Expression of these metabolic pathways can be modified by diet and hormones, thus, adding further complexity to the process and determination of relative risk of carcinogenesis.

If DNA is the target for carcinogens and mutagenesis is the underlying mechanism of cancer pathogenesis, how do we identify the genes that when mutated are responsible for tumor formation? Studies on the molecular basis of cancer development in the last 30 years have revealed two classes of genes, oncogenes and tumor suppressor genes that play a key role in the pathogenesis of cancer. An oncogene is any gene that can transform normal cells in culture and induce cancer in animals. Most oncogenes are derived from proto-oncogenes: normal cellular genes that are critical positive regulators of cell proliferation or inhibitors of apoptosis. Conversion to an oncogene can occur through point mutations resulting in a constitutively active protein, through DNA amplification, or through chromosomal translocations that link a highly active promoter with the proto-oncogene. Both of these later two mechanisms cause increased or inappropriate expression of a proto-oncogene and altered growth regulation of the normal cell. Tumor-suppressor genes normally function to negatively regulate cell proliferation, cause apoptosis, repair damaged DNA, or induce cellular differentiation. In contrast to oncogenes, both copies of a tumor suppressor gene must be inactivated to promote tumor development. Frequently, inactivating point mutations occur in one copy of a tumor suppressor gene, and the remaining normal copy is lost through a process of chromosomal missegregation during mitosis that leads to loss of heterozygosity.35,36

Which oncogenes and tumor suppressors contribute to development of cutaneous neoplasms? Considerable insight into the genetic basis of sporadic skin cancers has come from the elucidation of specific genes or genetic loci that define hereditary skin tumor syndromes (Table 111-2).37 In particular, the importance of DNA as a target for carcinogenesis was strongly supported by the discovery of defects in DNA repair genes that comprise the complementation groups of skin cancer prone xeroderma pigmentosum (XP) patients (Chapters 110 and 139).38,39 At least six independent genes, on distinct chromosomal loci, define proteins involved in nucleotide excision repair, a process critical for the protection of DNA against both UVR and chemical carcinogen-induced damage.40 Among these are proteins that recognize and bind to sites of DNA damage (XPA, XPC, XPE), helicases (XPB, XPD) and endonuclease components (XPG, XPF, defects in any of which give a skin cancer-prone phenotype. Potential polymorphisms with functional consequences in these and other DNA repair genes may contribute to susceptibility states in the general population as well.41–43 Examination of SCCs and BCCs from XP patients has also revealed potential target genes for cancer induction since signature mutations in PTCH1, RAS, p53, and INK4a-ARF (p16INK4a and ARF) have been found with high frequency.44

Chromosomal mapping studies in the basal cell nevus syndrome (Chapter 116), coupled with genetic and functional studies in Drosophila and mice, implicated mutations in the PTCH1 gene and other defects in the Hedgehog (Hh) signaling pathway in the development of hereditary and sporadic BCCs45 and a variety of internal malignancies.46 All of these tumors are characterized by elevated Hh signaling activity (Chapter 115). The INK4a-ARF locus, specifically mutations in the p16(INK4A) gene, was identified in the etiology of melanoma47 through mapping of the inheritance pattern of familial melanoma. Oncogenic mutations in BRAF were discovered in a high proportion of melanoma by systematically screening for alterations in genes in the Ras/MAP kinase signaling pathway,48 an important growth regulator in many cell types. Detection of specific mutations in Cowden syndrome (PTEN),49,50 Muir–Torre syndrome (MSH2, MLH1),50,51 multiple cylindroma (CYLD1),52 multiple trichoepithelioma (CYLD1),53 Birt–Hogg–Dube syndrome (FLCN),54 and sporadic pilomatricomas (CTNNB)55 and sebaceous tumors (LEF1)56 have illuminated genetic defects associated with the development of adnexal tumors. The delineation of the specific genes mutated in other syndromes where locus mapping is confirmed (Table 111-2) should provide even more insight into the molecular pathogenesis of a broader spectrum of skin neoplasms.

Using human hereditary cancer syndromes as a guide, genetically based animal models have been developed that validate the contribution of individual genes to human cancer susceptibility by prospective experimental analysis.57 These models have also contributed to the understanding of gene–environment interactions by testing the influence of chemical and physical carcinogens on tumor development in the setting of an altered genetic makeup. The discovery of inactivating PTCH1 mutations and activating SMO mutations in BCCs (see Chapters 115 and 116) led to the development of several mouse models exploring the role of deregulated Hh signaling in BCC tumorigenesis.45 Keratinocyte-targeted overexpression of SHH58,59 or an activated form of the Hh pathway signaling molecule SMO (M2SMO)60,61 resulted in the development of basal cell-like proliferations or follicular hamartomas in mouse skin. Overexpression of SHH in human keratinocytes grafted onto immune-deficient mice also resulted in BCC-like lesions.62 Mice with heterozygous disruption of the Ptch1 gene have features in common with basal cell nevus syndrome patients,63,64 including microscopic hair follicle-derived tumors and various macroscopic skin tumors, including BCCs, after treatment with UVR or ionizing radiation.65,66 When GLI1 or Gli2 transcription factors, which are nuclear effectors of Hh signaling are targeted to mouse keratinocytes, multiple skin tumors develop. GLI1 mice display a variety of follicle-derived tumors with relatively few BCCs,67 while Gli2 mice develop BCCs exclusively.68 Interestingly, when Gli2 expression was extinguished in gene-switch mice, BCCs regress but leave behind a subset of residual tumor cells, at least some of which may be able to form growing tumors when the Hh pathway is reactivated.69 Collectively, these findings provide strong evidence supporting the notion that constitutive Hh signaling plays a key role in, and may be sufficient for, BCC development.

The multistage induction of cutaneous SCC (Fig. 111-1)70 and the absence of a hereditary syndrome that specifically imparts uniquely high susceptibility to this lesion have made the delineation of specific genetic loci involved in SCC induction more complex. Nevertheless, the analysis of genetic defects in human SCCs has focused attention on specific targets such as DNA repair genes, RAS, p53, p16(INK4A), and the epidermal growth factor receptor (EGFR) pathway. Significant progress has been made in defining pathways capable of driving nonmelanoma skin cancer development using reconstructed human skin, comprising genetically altered primary human keratinocytes on architecturally intact dermis, grafted onto immune-deficient mice.71 Using this powerful system for modeling human skin cancer in mice, it has been shown that coexpression of oncogenic Ras and either Cyclin-dependent kinase 4 (Cdk4), or the NF-κB inhibitor IκB, in human keratinocytes, rapidly leads to invasive SCC development in reconstructed skin grafts.72,73

Mouse models for XPA, XPC, and XPE deficiency have been constructed, and all show sensitivity to UVR and enhanced development of SCC after either UVR or chemical carcinogen exposure.39 A number of other DNA repair-deficient mutant mice are also being tested for cancer susceptibility and may reveal the mechanistic basis for new clinical syndromes by reverse genetics.74 Mouse mutants targeting the Ras gene have revealed important information concerning the contribution of the Ras pathway to particular stages of SCC development. Heterozygous activating Ras gene mutations are sufficient to induce a benign squamous papilloma, the precursor to SCC, and the yield of chemically induced tumors in mice genetically deleted in the H-ras allele is markedly reduced.75 Homozygosity of a mutant ras gene is associated with progression to SCC,76 suggesting that high penetrance of the Ras pathway can recruit additional changes required for progression. Transgenic targeting of oncogenic Ras to suprabasal epidermis produces terminally benign tumors, while mosaic targeting to basal cells or hair follicle outer root sheath cells encourages progression to SCC,77 indicating the target cell for early mutations may also determine the tumor phenotype.78 While heterozygous p53 inactivation is detected early in human SCC, p53 deletion in mice enhances malignant progression but does not increase benign tumor formation.79 This may be analogous to the common loss of heterozygosity seen at the p53 locus later in human SCC development.80 Of particular interest for human tumor development is the frequent observation that double heterozygotes for p53 deletion and other cancer-prone phenotypes (such as DNA repair deficiency) further sensitize mouse skin to tumor induction by UVR or chemical carcinogens. Mice with genetically defined defects in the p16Ink4a locus or its downstream target Cdk4 are sensitive to both squamous tumor and melanoma induction after treatment with carcinogenic chemicals, consistent with defects in this pathway detected in both melanoma and nonmelanoma human skin cancers.81,82 Elegant mouse modeling studies have confirmed the importance of Braf mutations, when combined with alterations in the tumor suppressor Pten, in the development of metastatic melanoma in mice.83

General Principles

Malignant tumors exhibit fundamental alterations in behavior that distinguish them from the normal tissues in which they arise. These differences include a reduced requirement for growth stimuli, impaired response to growth inhibitory/differentiation signals, alterations in apoptosis, delayed or blocked senescence, prolonged angiogenesis, and the capacity for invasion and metastasis.84 Although one or more of these abnormalities can be detected at different stages of tumor progression and may thus be seen in premalignant lesions, all are present in advanced cancers. The driving force behind many of these changes is genomic instability, which facilitates the accumulation of mutations in both oncogenes and tumor suppressor genes that contribute to the observed aberrations in cell function.85 While some of these changes are cell-autonomous and can be studied in purified populations of tumor cells, others depend on various additional cell types recruited to participate in the development and progression of cancer in intact organisms. Both the intrinsic alterations in neoplastic keratinocytes and the influence of collaborating cell types in skin tumor biology are being elucidated through the use of powerful experimental models.

From analyses of both human SCC pathogenesis and experimental skin tumor induction by chemical carcinogens, specific premalignant and malignant stages have been identified that have typical phenotypic, genetic, and biochemical characteristics (Fig. 111-1).86,87 Note that in contrast to SCCs and most other cancers, BCCs do not appear to have precursor lesions or give rise to more aggressive cancers. This is likely related to the observation that the genomic instability that fuels neoplastic progression in other tumor types is largely missing in BCCs, even though a substantial proportion of these tumors harbor p53 mutations that would be expected to facilitate the accumulation of additional genetic defects. SCC starts as a single or small number of specific mutations in a cell that has the capacity to cycle, a change understood as the “initiation” of carcinogenesis. Upon clonal expansion, initiated cells form a premalignant lesion, such as a squamous papilloma in the mouse or an actinic keratosis in the human. Agents that enhance clonal expansion of initiated cells are called tumor promoters. Tumor promotion may occur as a consequence of exogenous exposures such as UVR, topical chemicals or medications, infections, or wound healing. Promotion may be an endogenous process influenced by diet, smoking, or immune suppression. The acquisition of additional mutations that provide a growth advantage to the incipient cancer cell may also serve as an autonomous promoting stimulus. Premalignant lesions undergo further phenotypic changes, often in a predictable sequence and commonly multifocal within a single lesion.88 Some foci or lesions progress at a faster rate than others, and these are at highest risk for malignant conversion.89 Premalignant progression encompasses the majority of the tumor latency period prior to malignant conversion, when the lesion shows invasive properties.

Initiation is usually a low frequency genetic event and is directly dependent on carcinogen dose. A large variety of carcinogen classes can initiate skin tumors in rodents (Table 111-3). The phenotype of initiated epidermal cells as defined from in vitro analysis includes a defect in maturation, escape from senescence, and an enhanced growth potential, but initiated cells are still responsive to negative regulation, for example by TGFβ. At the molecular level, initiation involves an alteration in signal transduction pathways that regulate cellular responses to extracellular signals, and these are internally regulated by proto-oncogenes and tumor suppressor genes.84,90 Tumor promotion is generally a nongenotoxic stimulus that results in a disturbance of tissue homeostasis.88 Multiple classes of agents have promoting properties in experimental skin carcinogenesis (Table 111-3). The mechanisms of tumor promotion include activation of cell surface receptors, activation or inhibition of cytosolic enzymes, and nuclear transcription factors, stimulation of proliferation, inhibition of apoptotic cell death, and direct cytotoxicity. Chronic inflammation is linked with tumor promotion, and agents that suppress inflammation reduce tumor formation in experimental models.91 Genotoxic carcinogens also can have promoting properties, and repeated exposures to low concentrations of genotoxic carcinogens can induce tumors more effectively than fewer exposures to the same total dose in experimental carcinogenesis.92 Exogenous tumor promoters can determine the target site for tumor formation, and the promoting action of UVR may in part contribute to skin targeting for tumors when germ-line mutations produce an initiated state in multiple cell types. In general, initiated cells respond differently to promoters than normal cells, allowing for clonal selection of an initiated population.93 Premalignant progression in an initiated cell clone occurs spontaneously but is accelerated by additional exposures to genotoxic agents including some cancer chemotherapeutic drugs.94 Premalignant progression in chemically induced mouse skin carcinogenesis is associated with nonrandom, sequential chromosomal aberrations, including amplifications, duplications, deletions, and loss of heterozygosity,95 suggesting repeated episodes of cell selection, producing specific chromosomal aberrations that become modally dominant within the progressing focus. Thus, at least one function of the relevant genetic events in premalignant progression must result in a growth advantage for the affected cell. Epigenetic changes are also associated with malignant conversion, and upregulation of AP-1 transcriptional activity, alterations in cell cycle regulatory genes and secreted proteases, changes in gene splicing and expression of modified cell surface molecules have all been reported. Together these changes could facilitate migration and invasion that characterize the malignant phenotypes. Alterations in methylation of DNA in tumor cells could also contribute to this stage of carcinogenesis.96

The Pathogenesis of Basal Cell Carcinomas

Animal models for induction of skin tumors by chemical and physical agents have been extremely valuable for exploring mechanisms of carcinogenesis in an experimental setting (Table 111-4). Rats exposed to chemical carcinogens (methylcholanthrene, dimethylbenzanthracene) or ionizing radiation preferentially develop BCCs97 that are phenotypically similar to human BCCs. In contrast, mice are resistant to BCC induction unless they are genetically modified in the laboratory, although one study reported mouse BCC development following either subcutaneous or topical exposure to dehydroretronecine.98

As noted in Section “General Principles,” BCCs represent an intriguing exception to the rule of multistage cancer development. Individual tumors typically consist of homogeneous, slowly growing cells that very rarely metastasize, although they can cause extensive local damage. Even those rare BCCs exhibiting cellular atypia generally have a benign course,99 suggesting that the full complement of factors driving malignant progression in SCCs and other neoplasms is not operating in these tumors. BCC cells exhibit abnormalities in growth control and terminal differentiation, induce clinically apparent angiogenesis, and invade their immediate surroundings. However, transplantation studies using human BCCs or transgenic mouse skin with BCC-like lesions58 suggest that the tumor cells are critically dependent on adjacent stroma for survival. This property may help explain both the lack of BCC metastases and the difficulty in establishing BCC cell lines. Interestingly, in a recent study in which global gene expression profiling was performed on stromal cells derived from tumors versus normal tissue, the bone morphogenetic protein (BMP) antagonist, GREMLIN 1, was found to be highly expressed in BCC stroma but not normal skin.100 BCCs express BMP 2 and 4, and GREMLIN 1 stimulates growth of cultured BCC cells in vitro, suggesting that GREMLIN 1 may be a stromal factor needed to support BCC proliferation in vivo, by antagonizing growth-inhibitory effects of BMPs.

Studies performed using human and rodent BCCs have identified additional candidate molecules that may be important in the biology of these tumors. Upregulation of various matrix metalloproteases (MMPs), both in tumor cells and stroma, is likely to be important in the local invasion of BCC, and may impact on growth control and other cellular functions given the emerging role of MMPs in regulating signaling at multiple levels.101 Expression of several adhesion molecules is downregulated in BCCs, and this may account for the characteristic appearance of artifactual clefts between tumor cells and stroma in histological sections, but the functional significance of this finding in BCC biology has not been established. The antiapoptotic molecule BCL2 is consistently upregulated in BCCs probably because BCL2 is an Hh target gene,102,103 but this does not appear sufficient to block apoptosis in these tumors as it does in other settings. Indeed, the remarkably slow growth rate of BCCs has been attributed in part to baseline apoptotic cell loss. Expression studies in human BCCs and experimental models suggest the involvement of growth factor signaling pathways such as PDGF104,105 and EGFR.106 Several lines of evidence point to components of the cell cycle machinery as potentially critical targets in mitogenic responses to Hh signaling. Keratinocytes overexpressing Shh have enhanced proliferative potential and fail to undergo growth arrest in response to the cell cycle inhibitor p21107; in the absence of Shh, Ptch1 interacts with and may sequester Cyclin B1, a component of M-phase promoting factor which is required for mitosis and cell cycle progression108; and Shh can drive proliferation of immature cerebellar cells by causing prolonged induction of Cyclins D1, D2, and E.109 In skin, Hh-driven proliferation of hair follicle epithelium is likely to be mediated by Cyclin D2 and N-Myc.110 Oncogenic Hh signaling in skin leads to activation of the Wnt/β-catenin pathway; remarkably, signaling via this additional pathway is required for Hh pathway-driven skin tumor development.111 The discovery of key molecules and other pathways required for Hh-driven tumorigenesis will provide additional targets for therapy.

The Pathogenesis of Squamous Cell Carcinomas

Both mice and rats are sensitive to SCC induction by a variety of chemical and physical agents (Table 111-4), and the induction of SCC in mice by UVR mimics the pathogenesis of SCC in human skin (Chapter 114). Multistage chemical carcinogenesis in mouse skin has been a powerful tool for delineating many of the fundamental concepts underlying epithelial carcinogenesis in other organs. Inbred strains of mice differ in their sensitivity to tumor induction using chemical carcinogens and tumor promoters, and extensive genetic analysis has led to identification of loci that confer sensitivity or resistance to induction of either benign and malignant tumors.112–114 Interestingly, benign tumors that develop in FVB/n mice have a high frequency of conversion to SCC, and this is due to a polymorphism in the FVB/n Ptch allele which enhances Ras-induced SCC development.115

Building on the association of deregulated EGFR signaling in human cancer, mouse models suggest that constitutive EGFR activation is an important component of skin tumor formation. Transgenic epidermal targeting of TGFα, which activates the EGFR, can produce a benign tumor phenotype in the absence of activating Ras mutations.116,117 However, activation of the EGFR by ligand overexpression alone is not an efficient stimulus for autonomous tumor formation, since many of these tumors regressed. Loss of function studies also point to an important role for EGFR in squamous tumorigenesis in skin. Transformation of Egfr null keratinocytes with oncogenic ras, followed by transplantation onto immune-deficient mice, results in the formation of smaller benign tumors that those arising from ras-expressing control keratinocytes. These data imply point to an important role for EGFR signaling in squamous tumors, but suggests that an alternative, ras-induced pathway for tumor growth exists in the early, premalignant tumor.18

Additional evidence implicating enhanced growth factor signaling in SCC is provided by transgenic studies using a dominant form of SOS, an adapter molecule involved in transducing growth-stimulatory signals from receptor tyrosine kinases like EGFR, which resulted in spontaneous development of papillomas in mouse skin.119 Overexpression of ErbB2, a receptor tyrosine kinase related to and capable of interacting with EGFR, results in spontaneous skin tumor development in transgenic mice.120 Insulin-like growth factor 1 (Igf1), when overexpressed in mouse skin, also triggers development of squamous papillomas, some of which progressed to carcinomas.121 While most of these studies invoke an autocrine mechanism for growth stimulation of tumor cells, factors produced by cells in the tumor stroma may supply mitogenic signals to keratinocytes or angiogenic signals as well.122 Additional studies performed in mice suggest that specific components of the cell cycle machinery, including E2F1,123 Cyclin D1,124 and Cdk4,125 can contribute to SCC development and/or progression.

The TGFβ signaling pathway is an additional growth factor pathway that contributes to SCC development but in contrast to the EGFR it functions as a negative regulator of epidermal proliferation and to maintain epidermal homeostasis.126 Reduced levels of intracellular mediators of TGFβ signaling, Smad2 and Smad4, are observed in human SCC,127,128 suggesting that loss or disruption of this growth control pathway is important in tumor development. Interestingly, SCC that develop in organ transplant recipients receiving antirejection drugs exhibit significant activation of the TGFβ pathway. Thus, in this context the TGFβ pathway may enhance tumor susceptibility.128

In addition to deregulated growth control, there is a progressive loss in the capacity for terminal differentiation during SCC progression (Fig. 111-1), culminating in a tumor with a spindle cell morphology that is indistinguishable from malignancies originating in mesenchymal tissues. Although the mechanisms responsible for defective differentiation at different stages of human skin cancer are not known, there is evidence that deregulation of the protein kinase C (PKC) family of enzymes plays a role in aberrant differentiation of mouse skin keratinocytes expressing the Ras oncogene.88 PKCδ, which has been implicated in terminal differentiation of normal epidermal keratinocytes, is rendered inactive as a result of tyrosine-phosphorylation in ras-transformed keratinocytes. Restoring PKCδ to its native, nontyrosine phosphorylated state reversed the block to terminal differentiation in ras-transformed keratinocytes. In addition, overexpression of PKCδ in the skin of transgenic mice inhibits the development of squamous papillomas and carcinomas.129 In contrast, skin-targeted overexpression of PKC ϵ resulted in less differentiated SCCs that rapidly metastasized to regional lymph nodes.130 Additional work is needed to better understand the mechanism by which SCC tumor cells evade signals that trigger differentiation of normal keratinocytes.

Prostaglandin metabolism is activated by UVR and is constitutively induced in human SCCs, probably due to overexpression of the enzyme cyclooxygenase 2 (COX-2). Although the mechanism by which changes in prostaglandins influence skin cancer is not clear, they appear to operate at the tumor promotion stage. Telomerase activity is detected in a substantial proportion of human skin cancers, suggesting that these tumor cells are capable of evading cellular senescence, and telomerase-deficient mice are resistant to chemical carcinogenesis.131 Additional contributors that are likely to be involved in SCC development and progression include mediators of angiogenesis, MMPs, and integrins, as well as other molecules involved in cellular adhesion and migration.

Spontaneous or carcinogen-induced tumor formation in genetically modified mice has revealed genes and pathways that appear to be important in skin cancer induction but would not have been apparent from hereditary cancer syndromes or analysis of human skin cancers. Suprabasal targeting of c-Myc in transgenic mice permits suprabasal cells to cycle and produces the papilloma/actinic keratosis phenotype while basal cell targeting of c-Myc is not oncogenic.132,133 Deletion of the Cyclin/CDK inhibitor p21waf1, a downstream effector of p53, increases the number of benign tumors but not the rate of premalignant progression.134,135 Inactivation of TGFβ signaling enhances premalignant progression while overexpression blocks papilloma outgrowth and promotes invasion and metastasis of established tumors and progression from SCC to a spindle cell phenotype.136–139 Surprisingly, reduction of cutaneous TGFβ1 also suppresses papilloma formation indicating a critical role for physiological levels of this growth factor in supporting tumor outgrowth.140 Two AP-1 transcription factors influence distinct stages of skin tumor development, where c-Jun is essential for papilloma and c-Fos is essential for SCC development.141,142 Inhibition of AP-1 activity in benign papillomas prevents progression to carcinomas, but instead, converts these lesions into benign sebaceous adenomas,143 supporting the concept that squamous tumors are derived from progenitor cells with multilineage potential. Additional molecules now implicated in SCC development are ornithine decarboxylase, p16Ink4a, p15Ink4b, E-cadherin, STAT3, c-Myc, Notch, α−catenin, NF-κB, and Smad3.144–152

Constitutional Modifiers of Carcinogenesis

Inbred mouse strains differ in susceptibility to particular carcinogenic exposures by several orders of magnitude, and transplantation studies indicate sensitivity resides in the target tissue rather than systemically. Genomic scans of backcrossed mice among sensitive and resistant mouse strains indicate that constitutional determinants are multigenic and distinct for squamous tumor formation or progression.113,153 Similar studies indicate genetic loci determine the survival potential for tumor bearing animals.154 Classical genetic approaches are frequently difficult and complex because multiple interacting loci are typically involved in defining cancer susceptibility.155 However, the study of skin cancer susceptibility may uncover genes of broad relevance to oncology. For example, the identification of a low-penetrance susceptibility gene for mouse skin cancer, Stk6,156 fueled investigation of the human homologue AURKA which is now implicated in the development of multiple tumor types. More recently, studies examining the susceptibility of different mouse strains to cutaneous SCC uncovered, surprisingly, an alteration in the Ptch1 gene,115 which is strongly implicated in the development of a BCC.45 Thus, alterations in the same gene can promote either SCC or BCC development, and this may in part be determined by influencing a cell's decision to enter either the squamous or basal cell lineage.

Analysis of genetically modified mouse models has also revealed pathways associated with skin tumor susceptibility. Alterations in p53,79 drug metabolizing enzymes,157 T cell function158 and DNA repair74 modify risk for experimental skin cancer induction. Remarkably, polymorphisms in drug metabolizing enzymes and DNA repair pathways,159,160 and variable responses in the p53 tumor suppressor pathway or in T cell immunity after exposure to skin carcinogens161,162 modify the human risk for skin cancer.

Exogenous Modifiers of Carcinogenesis

In experimental models, antioxidants and agents that alter microsomal metabolism of carcinogens can reduce or prevent tumor initiation by inhibiting the formation of ultimate carcinogens or accelerating their detoxification (Table 111-5). Similar mechanisms may influence ultraviolet light or ionizing radiation mutagenesis through inhibition of mutagenic oxyradicals produced endogenously following exposure. Scavengers prevent reactive mutagens from reaching critical targets. Cell cycle inhibitors prevent fixation of mutations, allowing for DNA repair while cytotoxic agents kill initiated cells prior to their expansion into a tumor mass.

A number of agents are effective in the postinitiation phase of experimental tumor development, and some are being clinically evaluated.163 Inhibitors of COX-2 (celecoxib, indomethacin) and ornithine decarboxylase (difluoromethylornithine) prevent both UVR and chemically induced tumors.164,165 While the mechanism of action of retinoids as modifiers of tumor development is not clear, retinoids are effective inhibitors of benign tumor formation in mouse skin carcinogenesis studies. Low-dose systemic retinoids also reduce SCC development in organ transplant recipients,166 but additional studies are needed to assess whether they will be useful in long-term chemoprevention in certain high-risk populations. Predictable and parallel alterations in retinoid receptors in mouse and human cutaneous SCC indicate that the retinoid pathway must be central to cancer development in epidermis.167,168 Considerable interest has developed in dietary factors that modify skin carcinogenesis. Restricted fat/energy diets reduce papilloma incidence in mouse models169 and decrease the frequency of actinic keratoses in human populations.170 Diverse dietary or natural factors such as green tea, caffeic acid phenethyl ester from honey bee hives, resveratrol from grapes, silymarin from milk thistle, ginger extract, crocetin, and ursolic acid from the rosemary plant are among natural products shown to be effective inhibitors of skin tumor formation by topical application or systemic exposure.171 Several of these agents are believed to be antioxidants, but the precise mechanism of inhibition and their effectiveness as chemopreventive agents for human skin cancer remain to be established. Finally, immune response modifiers such as imiquimod may be effective for treating some types of skin cancer by modulating cytokine expression.172

The substantial progress in our understanding of factors involved in skin tumor pathogenesis holds the promise of new approaches to treatment and prevention. In addition to the identification of new therapeutic targets, improved strategies are being developed for altering gene and protein function in skin with great selectivity. The abundance and unrivaled accessibility of skin cancers and precancers makes them prime targets for rigorous translational studies and clinical trials. On the basis of the basic research studies outlined above, potential therapeutic targets include p53; COX-2; telomerase; EGFR or other receptor tyrosine kinases, and intracellular signaling elements such as SOS; Ras; ornithine decarboxylase; the DNA repair machinery; MMPs; PKC; molecules involved in cell cycle progression, such as Cyclins D1 and D2, CDK4, p16INK4a, and E2F1; retinoid receptors; and c-Fos. In all cases, the potential efficacy of the proposed treatments would need to be carefully evaluated in animal models or human tissues grown in immune-deficient mice. Although the great majority of cutaneous SCCs are effectively treated using surgery or radiation therapy it is likely that dermatology patients would benefit from agents, which could prevent appearance of premalignant lesions, block neoplastic progression to SCC or cause tumor ablation with minimal invasive methods especially in cosmetically sensitive sites. This is likely to be especially true for genetically predisposed individuals or organ transplant recipients that are prone to multiple aggressive SCC. Along these lines, a double-blind study reported a significant reduction in the development of actinic keratoses and BCCs in XP patients treated topically with the DNA repair enzyme T4 endonuclease V, administered over a 1-year period.173 Pathogenic mutations in BRAF have recently been exploited for the treatment of melanoma, and in a promising phase I trial using a mutated BRAF inhibitor, the majority of individuals with melanomas carrying the V600E BRAF mutation exhibited either a partial or complete response.174

In contrast to the multiple genetic and biochemical changes associated with SCC development, the majority of BCCs have mutations in PTCH1 or SMO and essentially all of these tumors exhibit uncontrolled activation of the Hh pathway. Since several experimental models suggest that deregulated Hh signaling plays a central role in BCC development and maintenance, this pathway is a prime target for mechanism-based drug development. Cyclopamine is a natural product which blocks Hh signaling by inhibiting the function of SMO,175 a pivotal effector of Hh signaling. Cyclopamine and other Hh pathway inhibitors targeting SMO have been shown to effectively inhibit growth of BCC-like cancers and other Hh-activated tumors in pre-clinical studies.176 In a phase I clinical trial, a systemic Hh pathway antagonist showed promising results in patients with advanced and metastatic BCC,177 raising hopes that these cancers, and multiple other malignancies linked to deregulated Hh signaling, may be amenable to treatment with Hh pathway antagonists.178 BCC is rarely lethal, but its common occurrence on sun-exposed, cosmetically-sensitive sites makes a medical approach to treatment highly desirable, particularly if a topical Hh pathway antagonist is effective in treating tumors or preventing their appearance in high-risk individuals, including those with Nevoid BCC syndrome. Additional trials are currently underway in BCC patients using both systemic and topical Hh pathway antagonists, and if they establish efficacy with an acceptable safety profile, they may change our approach to treating these extremely common cancers. Given the fundamental advances in our understanding of nonmelanoma skin cancer, we can look forward to additional, rationally designed therapeutics to complement and perhaps in some cases replace, surgical management.