Yale School of Medicine, New Haven, Connecticut, USA

Reductionism – the search for the ultimate building blocks of life – comes with an obligation: to eventually put the pieces back together. This is harder than it looks, because it involves finding the relationships between the pieces. These usually went through the tissue grinder on the way down. Albert Szent-Györgyi once reflected, "In my hunt for the secret of life, I started my research in histology. Unsatisfied, I turned to physiology ... pharmacology ... bacteriology. But bacteria were even too complex, so I descended to the molecular level. But electrons are just electrons, and have no life at all. Evidently on the way I had lost life; it had run out between my fingers" (^{Wilson et al, 1973}).

In the case of skin cancer, attention has focused on mutated genes. The multiple genetic hit model for cancer has been a spectacular success, yielding "gatekeeper" genes like APC in colon cancer and PTCH in basal cell carcinoma of the skin. Gatekeepers single out a particular tissue for cancer development by acting as a tissue specific rate-limiting step. Genetics has also identified "caretaker" genes that prevent mutations in gatekeepers, including MLH1 and MSH2 mismatch repair genes in hereditary colon cancer, the BRCA putative repair genes in hereditary breast cancer, and the XP and CS repair gene families that underlie the sun-sensitive and cancer-prone disorders xeroderma pigmentosum and Cockayne syndrome. In contrast to these caretaker genes, P53– which is involved in DNA repair and in apoptosis of dam-aged cells – is mutated not only in rare hereditary cancers but also in the majority of sporadic tumors, including basal and squamous cell carcinomas of the skin (^{Brash and Bale, 2001}).

But mutated genes are embedded in a larger process (Figure 1). It is now feasible to begin reassembling these pieces. The initial steps, such as mutation by sunlight, are well-understood for skin cancer though poorly understood for other cancers. An essential point is that mutation frequencies are known, and range from 10^–8 per gene per cell generation for spontaneous polymerase errors to 10^–3 for moderate doses of UVB radiation (^{McGregor et al, 1991}). Even genomic instability occurs at 10^–3 or less (^{Wang et al, 2002}), and usually only in about 15% of sporadic cancers. So if 6 genes must be mutated for a tumor to arise (a figure at the conservative end of estimates from human age-incidence data (^{Stein, 1991})), and some are tumor suppressor genes that must be mutated in both alleles, then approximately 10 hits are required. The likelihood of a single cell sustaining all the required mutations is approximately 10^–30. Since the number of proliferating keratinocytes in human skin is only about 10⁶ per cm² (^{Bergstresser et al, 1978}), with about 1000 cm² of skin exposed, this is not going to happen. For this reason, the multiple genetic hit model of cancer is critically dependent on a physiological trick: after the first mutation is made, the mutant cell clonally expands. The target size increases by 100–1000 fold, and only one of these cells needs to be hit the second time. If the mutation frequency is 10^–3 and clonal expansion is 10³-fold, the subsequent mutation is a certainty. (Gene amplification could conceivably achieve the same end.) Thousands of clones of P53-mutant keratinocytes indeed reside in sun-exposed human epidermis and UVB-irradiated mouse skin (^{Brash and Bale, 2001;Zhang et al, 2001}). But what drives clonal expansion? Is it sunlight, and, if so, how?

Figure 1.

The multiple genetic hit model of cancer relies on physiology. Cancer typically begins when a carcinogen such as sunlight or body temperature alters DNA's chemical structure, creating an abnormal nucleotide such as a cyclobutane dimer or abasic site. During DNA replication, an incorrect guess by the polymerase leads to insertion of a normal, but incorrect, base at this site – a mutation. Imperfect polymerases and viral insertion cause mutations without a carcinogen. Because point mutation frequencies are only 10^-8–10^-3 per gene per cell generation, even after carcinogen exposure or with genomic instability, the likelihood of mutating in a single cell the 6–12 genes needed for an adult onset cancer is 10^-30 or less. This paradox is relieved by clonal expansion, in which a cell carrying the first mutation is copied to make a larger target of 100 cells or more, only 1 of which must sustain the next mutation. ^{Mudgil et al} show that UVB induces apoptosis and suppresses proliferation only in normal human keratinocytes, allowing premalignant keratinocytes to clonally expand.

Full figure and legend (6K)

Into this breach step^{Mudgil et al} (p. 191). The group has constructed organotypic cultures in which normal human keratinocytes (NHK) are grown on a collagen matrix overlaid with dermal fibroblasts. The NHK are allowed to differentiate, so that cells in the basal layer proliferate and differentiate upward to make a keratinized layer at the air–liquid interface. The NHK are then mixed with II-4 cells, a cell line of early stage malignant human keratinocytes carrying mutations in each P53 allele and carrying an activated HRAS oncogene; the cells are also genetically marked with the bacterial -galactosidase gene to enable easy visualization. In previous work, the group showed that a normal:premalignant ratio of 4:1 led to elimination of the premalignant cells by inducing their cell cycle withdrawal and terminal differentiation (^{Javaherian et al, 1998}). At a 1:1 ratio, the premalignant cells survived, proliferated, and invaded. The tumor promoter TPA facilitated clonal expansion of the premalignant cells, by inhibiting NHK proliferation and inducing their differentiation (^{Karen et al, 1999}).

The effect of ultraviolet B radiation (UVB) on a mixture of normal and premalignant cells was striking. Using doses corresponding to a minimal erythemal dose (MED) or less, the authors first found that UVB allows the premalignant cells to form clones even at normal:premalignant cell ratios that would otherwise have led to their disappearance. The mutants eventually occupied 28% of the tissue area. Double labeling with bromodeoxyuridine revealed that during this occupation only the premalignant cells were proliferating. Was UVB aiding the premalignant cells or restraining the normal NHK? Pure NHK cultures showed a dose-dependent increase in apoptotic sunburn cells and soundbreak-positive cells, while their proliferation was suppressed 10-fold. In contrast, pure II-4 cells showed neither UVB-induced apoptosis nor suppression of proliferation. The defective apoptosis was expected from the P53 mutations (^{Brash and Bale, 2001}), but the effect was even greater in the presence of activated RAS, which suppresses apoptosis in many systems. Activated RAS also interferes with the G1/S cell cycle checkpoint that restrains proliferation. The opposite reactions of normal and premalignant keratinocytes to UVB would lead to apoptosis of normal keratinocytes, after which the premalignant cells would preferentially repopulate the "microscopic burn wound", as proposed a number of years ago (^{Brash and Bale, 2001}). End result: each trip to the beach becomes a selection pressure favoring mutant cells. Although UVB-induced apoptosis appears to be a powerful protective mechanism preventing sunlight-mutated cells from accumulating in the epidermis, the system backfires when a P53 mutation is present.

These molecular results complement recent visual observations on UVB-induced clonal expansion in mouse skin in vivo (^{Zhang et al, 2001}). Chronic irradiation leads to more and larger P53-mutant clones. A geneticist might expect the effect on clone size to come from a UV-induced mutation in a second gene, one controlling cell proliferation. Alternatively, expansion might be due to a physiological effect of UVB. In the genetic mechanism, a clone would continue to irreversibly expand once mutated, whether or not UVB exposure continued; in the physiological version, each increment of clonal expansion would depend on additional exposure. The latter was the case: clonal expansion stopped when UVB ceased. UVB-induced apoptosis and differential proliferation would fulfil this physiological role. A second visual finding was that clone size was quantized, occuring in multiples of one murine stem cell compartment. UVB's role was to allow escape into an adjacent stem cell compartment. UVB-induced apoptosis would accomplish this by providing a steady source of vacated compartments.

Clonal expansion is evidently so important that it is normally suppressed. In the organotypic system, neighboring NHK induced premalignant keratinocytes to differentiate. Other work has found tumor-suppressive effects from dermal fibroblasts (^{Dotto et al, 1988}), prostate fibroblasts, and tracheal epithelium (reviewed in^{Rubin, 2002}). An important step in tumor development is the escape from this suppression. Preneoplastic tracheal epithelial cells no longer inhibited the proliferation of neoplastic epithelial cells, and fibroblasts derived from human prostate carcinoma allowed immortalized prostate epithelial cells to progress to a tumor. These phenomena may reflect mutations or physiological changes in the cells surrounding the tumor's cell of origin, something of a terra incognita in cancer biology. In the organotypic system, as in vivo (^{Rubin, 2002}) tumor promoters such as TPA and UVB handicap the normal keratinocytes by inducing their differentiation or apoptosis, and inhibiting their proliferation, while allowing malignant cells to proliferate. Chemical carcinogens can facilitate clonal expansion of pre-existing spontaneously mutant cells, recognized as such because their mutation signature does not match that of the carcinogen itself (^{Cha et al, 1994;Rubin, 2002}). Since many gatekeeper tumor suppressor genes have roles in embryonic development, losing them may achieve escape by genetic means. Given the drama of suppression and escape, it is well to remember that everything in life is a selection pressure. The cycle of suppression and escape selects – even in the absence of an exogenous carcinogen or promoter – for cells that resist suppression. This fact is visible during serial passages of 3T3 fibroblasts or chicken embryo cells infected with Rous sarcoma virus (^{Rubin, 2002}), where the cell's need for transformation-permissive conditions diminishes with passage.

So as we work our way back up from photon to tumor, we see that mutant genes can be tied together by the physiology of the cells that contain them. In contrast, phosphorylate-and-bind stories remain parked at the molecular level. Some caveats remain. As with an earlier study showing an association between hypoxia-induced apoptosis and clonal expansion of P53-mutant tumor cells in the hypoxic tumor core, it remains to be shown that the observed apoptosis is actually responsible for the observed clonal expansion. For example, UVB's key effect might be to suppress whatever signal allows NHK to induce differentiation in the premalignant cells. These are harder experiments, but they become feasible in the elegant organotypic culture system of ^{Mudgil et al}. Even now, however, it seems clear that UVB will have two roles in skin cancer: mutation and clonal expansion; initiation and promotion.