p53 mutations appear to be early events in skin carcinogenesis induced by chronic UVB irradiation. Clusters of epidermal cells that express p53 in mutant conformation (“p53 positive foci”) are easily detected by immunohistochemical staining long before the appearance of skin carcinomas or their precursor lesions. In a hairless mouse model, we determined the dose-time dependency of the induction of these p53+ foci and investigated the relationship with the induction of skin carcinomas. The density of p53+foci may be a good direct indicator of tumor risk.
Hairless SKH1 mice were exposed to either of two regimens of daily UVB(500 or 250 J/m2 broadband UV from Philips TL12 lamps; 54%UVB 280–315 nm). With the high-dose regimen, the average number of p53+ foci in a dorsal skin area (7.2 cm2) increased rapidly from 9.0 ± 2.1 (SE) at 15 days to 470 ± 80 (SE) at 40 days. At half that daily dose, the induction of p53+ foci was slower by a factor of 1.49 ± 0.15, very similar to a previously observed slower induction of squamous cell carcinomas by a factor of 1.54 ± 0.02. In a double-log plot of the average number of p53+ foci versus time, the curves for the two exposure regimens ran parallel (slope, 3.7 ± 0.7), similar to the curves for the number of tumors versus time (slope, 6.9 ± 0.8). The difference in slopes (3.7 versus 6.9) is in line with the contention that more rate-limiting steps are needed to develop a tumor than a p53+ focus. By the time the first tumors appear (around 7–8 weeks with the high daily dose), the dorsal skin contains >100 p53+ foci/cm2.
To further validate the density of p53+ foci as a direct measure of tumor risk, we carried out experiments with transgenic mice with an enhanced susceptibility to UV carcinogenesis, homozygous Xpa knockout mice (deficient in nucleotide excision repair) and heterozygous p53 knockout mice (i.a. partially deficient in apoptosis). In both of these cancer-prone strains, the p53+ foci were induced at markedly increased rates,corresponding to increased rates of carcinoma formation. Therefore, the frequency of p53+ foci appears to correlate well with UVB-induced tumor risk.
The p53 gene plays a crucial role in the protection against carcinogenesis in a wide variety of tissues. Compared with other cancer genes, p53 is the most commonly (∼50%)mutated gene in human cancers. An extensive database, originally set up by Hollstein et al. (1), summarizes those mutations in >50 different cell and tissue types and is at present maintained at the IARC.3The timing of p53 somatic mutations in human cancers depends on the tumor type as well as on the nature of the mutation. For example, in sporadic colorectal cancer a complete loss of a gene (loss of heterozygosity) is a late event (2), whereas in skin cancers, p53 point mutations appear to occur early; they are already present in preneoplastic lesions (3, 4).
The p53 protein is involved in a great variety of regulatory pathways,mainly through transcriptional activation but also through protein-protein interactions (5). In response to DNA damage or cellular stress, p53 activates genes that induce cell cycle arrest to allow DNA repair to take place or induce apoptosis in the case of overly damaged DNA. Dysfunctional and/or loss of p53 enhances cell transformation and thereby carcinogenesis; therefore, p53 is typically referred to as a “guardian of the genome”(6). The “guardian of the tissue” function of p53 refers to its regulatory role in apoptosis, which is reduced in p53 knockout mice (3), thus resulting in cancer predisposition (7). Cells expressing mutant p53 protein can in some cases be more tumorigenic than cells lacking endogenous p53. This “gain of function” (8) is caused by particular missense mutations. The majority of sun exposure-related skin carcinomas in humans bear missense mutations in p53 at dipyrimidine sites. These C→T and CC→TT transitions are characteristic of UV and therefore called “UV signature mutations.”They are found in hotspots, clustered around codons 173–179, 235–250,and 273–278 (9, 10, 11, 12, 13). In most of these cases, the wild-type allele appeared to be still present, indicating that the p53 missense mutation is selected for in solar UV-induced skin tumorigenesis. Most hotspot point-mutated p53 proteins have a dominant-negative effect on wild-type p53. These mutant p53 proteins are able to drive cotranslated wild-type p53 into mutant conformation,resulting in tetramers that are inactive in DNA binding (14, 15). In this regard, it is interesting to note that the germ-line p53 mutations in (cancer-prone) Li-Fraumeni patients do not appear to predispose for skin carcinomas, i.e., no increased risk for skin carcinomas has been reported for these patients(16), which is in line with the fact that among the mutants that were tested only the codon 248 mutant (which is a Li-Fraumeni mutation) showed no dominant-negative effect(14).
UV signature mutations in the p53 gene are already present in benign precursor lesions of squamous cell carcinomas, i.e., actinic keratoses (3). Whether such a p53 mutation is the very first event or an auxiliary event in the development of a skin carcinoma is still not clear. It was found in humans that chromosomal aberrations are already abundantly present in actinic keratoses, and these aberrations also occur in the absence of any detectable p53 mutation (17). Surprisingly, these chromosomal aberrations are even more abundant in some actinic keratoses than in the squamous cell carcinomas, which indicates that only a certain subset of the actinic keratoses may progress to squamous cell carcinomas.
UV signature mutations in the p53 gene found in human skin carcinomas are also observed in the majority of experimentally UV-induced squamous cell carcinomas and actinic keratoses in several experimental mouse strains, including hairless mice (12, 18, 19, 20). In a first attempt to study the timing of these p53 mutations during skin carcinogenesis, we discovered clusters of epidermal cells, expressing a mutant conformation of p53 protein in the skin of chronically UVB irradiated hairless mice(21). These “p53-positive foci” also commonly referred to as “p53 patches,” arise long before tumors appear and are more abundant in size and number in mice exposed for 30 days than in those exposed for 17 days. p53+ foci were immunohistochemically stained in sections of skin biopsies using Pab240, an antibody specific for p53 in mutant conformation (22, 23, 24, 25). The epitope of Pab240 is cryptic in wild-type p53, but most mutations in the gene result in a change in the structure of the protein, allowing the antibody to bind specifically to p53 in mutant conformation.
In parallel to these murine data, p53+ foci are found in regularly sun-exposed normal human skin (4). Sequence analysis of the gene in microdissected foci revealed C→T and CC→TT transitions in 50% of the foci analyzed. Again, “UV signature mutations” at dipyrimidine sites strongly implicate UVB radiation as the cause. In comparable studies, mutations are detected in up to 80% of the p53+foci (11, 26, 27). Considering the aforementioned data, it is quite plausible to assume that p53+ foci are potential onsets of UV-driven tumor development.
The SKH1 hairless mouse is an established model for studying UV carcinogenesis. Experiments with this animal model under well-defined conditions have yielded reproducible quantitative data on how the development of squamous cell carcinoma depends on dose,time, and wavelength of the UV radiation (28).
The aim of this study was to measure the kinetics of the induction of p53+ foci by chronic UV exposure in the (SKH1) hairless mouse model and to determine whether the induction of p53+ foci can be quantitatively related to the subsequent induction of skin carcinomas. We wanted to investigate whether the density of p53+ foci in the skin could be a reliable predictor of skin cancer risk. To explore the causality of the relation between the p53+ foci and skin tumors, we also investigated UV-induced p53+ focus formation in transgenic mice that are more susceptible to UVB carcinogenesis. To this end, homozygous Xpa knockout mice (deficient in NER4; Refs. 29, 30, 31) and heterozygous p53 knockout mice (partially defective in apoptosis; Refs. 3 and32) were chronically exposed to UVB radiation to establish whether the rate of p53+ focus induction parallels the increased rate of tumor formation.
MATERIALS AND METHODS
Irradiation of Mice.
Hairless mice (SKH1; Charles River, Sulzfeld, Germany) with ages ranging from 6 to 9 weeks were housed individually in standard Macrolon cages (size, 23 × 13 × 13 cm). Standard mice chow (Hope Farms RMB-H) and tap water were available ad libitum. As legally required, approval for the experiments was obtained from the university’s ethical commission on animal experiments. Cages were kept stationary in automatically time-switched irradiation cabins (33). Hairless Xpa-deficient mice originated from the stock as described (34), and hairless p53-deficient mice were generated by breeding p53-deficient mice (32) with inbred hairless mice.5Philips TL-12/40W fluorescent tubes [54% output in UVB (280–315 nm)and 46% in UVA (315–400 nm)] were used to expose nontransgenic and p53-deficient mice dorsally 500 J/m2 or 250 J/m2 UV daily. These doses correspond approximately to 1 and 0.5 MED for these animals. A MED is the minimal dose causing just perceptible effects such as edema and erythema after 24–48 h. Our previous experiments on tumor induction in these mice were carried out with Philips/Westinghouse F40 lamps, which are no longer available, but the spectrum of the Philips TL12/40W is similar to that of the Philips/Westinghouse F40 lamp. In experiments on small groups of mice (n = 6; data not shown), we verified that biological effects, such as edema and carcinogenesis,evoked by 500 J/m2 UV from a TL12/40W or 1000 J/m2 UV from a F40 lamp do not significantly differ, as was predicted from an analysis based on action spectra(35). This difference in effective exposure is attributable to the fact that the TL-12 contains slightly more of the biologically effective short-wave UVB than the F40 lamp. Xpa-deficient mice are very sensitive to UVB (1 MED, ∼125 J/m2 from F40 lamps) and received a daily UV dose of 80 J/m2 from F40 lamps.
Nontransgenic mice from the group that received 1 MED/day were sacrificed by ether anesthesia and cervical dislocation at 12, 15, 17,25, 31, and 40 days and from the group that received 0.5 MED/day at 20,25, 31, 40, and 59 days (four mice/group/time point). Xpa−/− mice (low MED) were sacrificed after 2, 3, 4, 5,and 7 weeks; their littermates at 3, 4, 5, 7, 9, 11, and 18 weeks. Heterozygous p53 knockout mice (normal MED) and their littermates were sacrificed at days 15 and 20.
Preparation of Epidermal Sheets.
After final irradiation, four mice/time point were sacrificed. Rectangular parts (size, 40 × 25 mm2) of the dorsal skin were isolated and placed floating on a 100 μg/ml thermolysin solution (P-1512; Sigma Chemical Co., St. Louis, MO) in PBS (containing 8.20 g/l NaCl, 3.20 g/l Na2HPO4·12H2O,and 0.20 g NaH2PO4·2H2O/l distilled water) containing 5 mmCaCl2 (Merck, 2083, Darmstadt, Germany), pH 7.8. After overnight incubation at 4°C, each epidermis was rolled on a polyethylene tube to separate it from the dermis.
Immunohistochemical Staining of Epidermal Sheets.
The epidermal sheets were spread floating on PBS in a Petri dish and subsequently fixed in PBS buffered 4% formaldehyde solution (Merck,1.04003; 37%, Z.A.) for 10 min at room temperature. Treated under the same conditions, the epidermal sheets of unirradiated abdominal skin,served as negative controls, whereas microscopic sections of UVB-induced tumors (from other tumor induction experiment) were used as positive controls for p53 staining. After a brief PBS wash, antigen retrieval was performed by 5 min boiling in 10 mm citrate buffer (pH 6.0). The epidermal sheets were washed in polystyrene tubes filled with 4 ml PBS. Subsequently, endogenous peroxidase was blocked in methanol, containing 1.5%H2O2, during a 20-min incubation in an end-to-end rotor. Sheets were washed three times in a 5-min incubation of PBS, containing 0.5% Tween 20 (polyoxyethylene sorbitan monolaurate; P-1379; Sigma). Aspecific binding was blocked with 10% normal rabbit serum and 0.2% BSA in PBS containing 0.1%saponin for permeabilization.
To stain the p53 protein in mutant conformation, we used the Pab240 antibody (NCL-p53–240; Novocastra, New Castle, United Kingdom). To affirm the specificity, we showed in earlier experiments that this antibody does not recognize the wild-type p53 protein, even at high levels of overexpression after 6 MED of UV from an F40 sunlamp(21). Moreover, the p53 gene was found to be mutated in exon 8 in 9 of 10 Pab240-positive UV-induced skin tumors: 8 C→T transitions at codon 267 and 1 CC→TT tandem mutation at 272(hotspots for UVB radiation).
Pab240 was diluted 1:25 in PBS, containing 5% normal rabbit serum,0.2% BSA, and 0.1% saponin, and incubated overnight with epidermal sheets at 4°C. Unbound antibody was removed in a triple wash of 5 min in PBS/Tween 0.5%. The secondary antibody, rabbit antimouse(IgG1)-biotin (61-0140, Zymed; San Francisco, CA), diluted 1:50 in PBS,containing 0.2% BSA and 0.1% saponin, was used to incubate the sheet for 1 h at room temperature. Excess of this antibody was removed in a triple wash of 5 min in PBS. The sheets were incubated for 45 min in avidin-biotin-peroxidase complex (ABC complex, K0355; Dakopatts,Copenhagen, Denmark). After a triple wash of 5 min in PBS, the sheets were stained for 3 min in 50 ml of substrate solution, containing 40 mg of 3,3′-diaminobenzidine (D-5905; Sigma) and 100 μl of 30%H2O2. Peroxidase reaction was stopped by a triple wash of 5 min in PBS. The sheets were mounted basal side up in Paragon (7.0% gelatin and 50% glycerol in distilled water).
Scoring of p53+ Foci.
A grid, placed on top of each epidermal sheet preparation, was used to count p53+ foci in 20 squares (total area, 29.0 × 18.5 mm2), using a light microscope (Pl ×25/0.5 objective). Fixation and antigen retrieval caused an area shrinkage of the epidermal sheet by 25 ± 2 (SD)% (area before shrinkage was 33.5 × 21.4 mm2). A p53+ focus was defined as a cluster of at least 10 Pab240-positive epidermal cells.
The data on the average number of p53+ foci, (NF) versus time (t, in days) were fitted with the formula “log(NF) = (slope)log(t/30) + (intercept)” by error-weighted least-squares (see Fig. 3). In this formula, t is divided by 30 days to create a dimensionless argument for the log function. The displacement in time between the two lines was calculated from the difference in intercepts divided by the common slope.
Induction of p53+ Foci in Wild-Type Hairless Mice after Chronic UV Exposure.
Two groups of hairless mice were daily irradiated with either 1 or 0.5 MED (500 J/m2 or 250 J/m2UV from TL12/40W, respectively). At several time points, four mice were sacrificed, and dorsal and abdominal skin was used to prepare epidermal sheets. Pab240 immunostaining revealed distinct clusters of cells with nuclear p53 staining in the interfollicular areas of the dorsal epidermis (Fig. 1), but none were found in the abdominal epidermis. These “p53+ foci”increased in number and size in a time- and dose-dependent way. Eventually, from day 50 individual foci grew to confluency, limiting the maximum count to ∼500 foci/epidermal sheet.
We counted interfollicular p53+ foci (≥10 cells each) in the epidermal sheets in a grid of 20 squares of 27 mm2. The p53+ foci were not evenly distributed over the sheet: the spatial distribution of p53+ foci in a typical example is shown in Fig. 2 for 1 MED/day at 40 days. The distribution seems to follow the geometry of the dorsal skin in vivo. The central regions contained 40 or more foci, whereas the neck regions showed <3.
Dose-Time Response for the Induction of p53+ Foci.
After immune staining, total numbers of p53+ foci were scored in the epidermal sheets from different time points, i.e., in the grid area of 29 × 18.5 mm2, which corresponds to an original skin area of 33.5 × 21.4 mm2 on the mouse (see “Materials and Methods”). The results for the average numbers of p53+ foci in the two dose groups are presented in Fig. 3 in a double-log plot against time. The data points for the two groups appear to line up parallel to each other. We have fitted lines through these points by the method of weighted least-squares. These lines represent the yield of p53+ foci in each of the dose groups as a cumulative Weibull hazard function (36). The slopes in the two dose groups are equal: 3.7 ± 0.6 in the 1 MED/day group and 3.7 ± 0.7 in the 0.5 MED/day group. The induction of p53+ foci appears to be slowed down by a factor of 1.49 ± 0.15 by lowering the daily dose by a factor of 2.
Induction Kinetics: p53+ Foci Compared with Tumors.
Fig. 4 depicts the yields of the p53+ foci together with the yields of subsequent 1-mm squamous cell carcinomas [taken from de Gruijl et al. (37)] for 1 and 0.5 MED/day in one graph. The results on the p53+ foci are obtained from three independent experiments. The p53+ foci occur earlier and are far more numerous than tumors, but the relative increase with time is much steeper for the tumors.
From the previous extensive experiments on tumor induction in the SKH1 hairless mice (36, 37), we know that carcinoma induction is delayed by a factor of 1.54 ± 0.02 when the daily UV dose is lowered by a factor 2. This factor appears to correspond very well with the factor of 1.49 ± 0.15 by which the induction of p53+ foci is slowed down. [In fact, the relationship between the daily dose, D, and the time to reach a certain yield (e.g., 1 lesion/mouse), t1, can be written as Drt1 = constant, where r = 0.62 ± 0.02 for tumors and that r = 0.58 ± 0.15 for p53+ foci.]
As appears to be the case for the p53+ foci, the yield of carcinomas goes up linearly in a double-log plot against time, as shown in Fig. 4[i.e., the induction of p53+ foci and carcinomas can be described by Weibull statistics (37): the yield = (t/t1)p,where t1, p are constants]. However, the slope (p) for the tumors is 2-fold larger than for p53+ foci and equals 6.9 ± 0.8.
UV-induced p53+ Foci in Xpa-deficient Mice.
In previous experiments (30), we have established that hairless Xpa−/− mice who are completely deficient in NER develop UV-induced skin tumors faster than their heterozygous Xpa+/− and wild-type Xpa+/+, NER-proficient counterparts. To investigate whether this corresponds with a faster induction of p53+ foci, we daily irradiated Xpa−/− and Xpa+/− mice with 80 J/m2 UV from F40 lamps (comparable with 40 J/m2 UV from TL-12; see“Materials and Methods”). This low UV dose was chosen because Xpa−/− mice are very UV sensitive (1 MED is ∼125 J/m2). Xpa−/− mice were sacrificed after 2, 3, 4, 5, and 7 weeks and Xpa+/− mice after 3, 4,5, 7, 9, 11, and 18 weeks. Average numbers of p53+ foci are presented in Fig. 5. At this low UV dose level, p53+ foci could already be detected after 2 weeks in Xpa−/− mice. The number and size increased rapidly with time; after 7 weeks, the foci were so numerous and large that distinction between individual foci became difficult. The Xpa heterozygous mice showed virtually no p53+ foci over the 18-week period at this low level of daily UV exposure; only one focus was observed at 7 weeks.
p53+ Foci in p53+/− Knockout Mice.
Previous experiments (38) 5 have demonstrated that heterozygous p53 knockout mice are more susceptible to UV-induced skin carcinogenesis than their wild-type counterparts. We wanted to investigate whether this increased susceptibility corresponds to an increased induction of p53+ foci.
Heterozygous p53+/− knockout mice have a MED comparable with wild types. p53+/− and p53+/+ littermates received ∼1 MED (500 J/m2 UV from a TL-12 lamp)or 0.5 MED (250 J/m2 UV from a TL-12 lamp) per day. After 15 and 20 days of chronic irradiation, epidermal sheets were isolated and stained with Pab240. p53+ focus scores are plotted in Fig. 6.
At both 1 and 0.5 MED/day, the p53+/− mice showed increased numbers of p53+ foci, in comparison with their wild-type p53+/+ littermates. At 1 MED/day, the data from the wild-type p53+/+ mice were in fair agreement with the previous experiment (Fig. 3), but at 0.5 MED/day, the numbers of p53+foci seemed to be rather low. These low scores at 15 and 20 days are,however, very inaccurate: 0.5 ± 0.3 and 1.0 ± 0.7 foci/sheet at days 15 and 20, respectively.
The chronically UV-exposed epidermal sheets show distinct,compact clusters of cells with a dark nuclear staining for p53 protein in mutant conformation (Fig. 1). This indicates that all cells in such a p53+ focus expanded clonally from one cell with an altered p53 expression, e.g., caused by a mutation in the p53gene. Preliminary results obtained by mutation analysis (to be published elsewhere) confirm that at least 50% of individual foci contain a mutation in the p53 gene. This is in line with mutation analysis in human p53+ foci, which appeared to contain mutations in 50–80% of the cases (4, 26, 27).
The uneven distribution of p53+ foci over the dorsal skin (Fig. 2)corresponds closely to the convex shape of the mouse back; the numbers are found to peak in the optimally exposed mid-dorsal area and drop steadily toward the flank regions. This is in accordance with the fact that these flank regions receive less UV irradiation than the mid-dorsal part.
At the first time point (day 15), we found 9.0 ± 2.1(SE) foci/mouse at 1 MED/day (Fig. 3). To estimate how early the first p53+ foci were generated, we can extrapolate the yield of foci (Fig. 3)backward in time; an average of one focus/mouse is then found at day 7. According our definition, a p53+ focus should contain at least 10 cells, and therefore a minimum of four cell divisions must have taken place for a clonal expansion to that size. Hence, these p53 alterations appear to be very early events in UV carcinogenesis. This inference is in close accordance with experiments (39) that showed that specific p53 mutations could be detected by allele-specific PCR as soon as 1 week after the onset of daily UV exposure mice.
The hairless mouse model has allowed us to investigate the dose-time dependency of the induction of the p53+ foci and to relate these results quantitatively to the well-established subsequent induction of skin carcinomas. The induction of squamous cell carcinomas can be described appropriately by Weibull statistics (36, 37),and this also appears to hold for the induction of the p53+ foci. However, the slopes of the tumor yield curves (Fig. 4) are about twice as steep as those of the p53+ foci (6.9 ± 0.8 versus 3.7 ± 0.7). This difference in slopes can be mathematically attributed to more rate-limiting steps in the formation of tumors than in the formation of p53+ foci(36). In fact, the similarity in dose dependence and the 2-fold difference in slope implies that the tumor could arise from a p53+ focus by a second event that follows the same kinetics as the induction of a p53+ focus (see “Appendix”).
The data in Fig. 4 show that lowering the daily exposure by a factor of 2 delays the build-up of p53+ foci and tumors by a similar factor that is smaller than 2 (1.49 ± 0.15 for the foci and 1.54 ± 0.02 for squamous cell carcinoma). Hence, UV dose dependencies of p53+ foci and tumors appear to be very similar.
Our data indicate that the density of p53+ foci and the time at which it was measured are predictive of the tumor risk later. For example, 10 foci/sheet at 20 days (0.5 MED/day) in Fig. 4 leads to an average of one tumor/mouse around 110 days (5.5 times later). If 52 foci/sheet at 20 days (1 MED/day) were found, then one expects an average of one tumor/mouse around 75 days. The yield of one tumor/mouse corresponds with a 63% chance for an individual of having contracted a tumor[chance = 1 − e−yield (36)]. Because we are dealing with stochastic processes, these are, of course, not deterministic predictions but ones with inherent statistical variation(dependent on group size, number of observed foci, and others).
Although it has been established that p53+ foci also occur in high numbers in regularly sun-exposed skin of humans (4), it is not clear whether and how the above calculations can be translated to the human situation. Measurements on well-chosen groups (with known skin cancer risks) from populations with different levels of sun exposure may serve to adjust the parameters of the mouse model to a human model. A good starting point appears to be that the age dependency of squamous cell carcinomas in humans is very similar to that in mice; under roughly similar average UVB exposures (0.3 MED/day), the mice contract their tumors ∼250 times faster(40). Considering the difference in exposed skin area(∼1200 cm2 for face, neck, and back of the hands, i.e., ∼170-fold more than for a hairless mouse),the tumor yield/cm2 of skin in humans needs only to be 0.006 times of that in mice for a comparable chance for an individual to contract a squamous cell carcinoma. Correspondingly lower densities of p53+ foci would be expected in humans. A Swedish study(41) on normal skin adjacent to resected nonmelanoma skin tumors from sun-exposed sites showed that the prevalence of p53+ foci increased with age; 35% of the microscopic sections taken from people at ages ∼55 years contained p53+ foci, 65% at 65 years, and 58% at 75 years. [This corresponds with a slope for the log(yield) versus log(age) of three to four, in agreement with the slope of 3.7 that we found for p53+ foci in the mice]. However, the yield of p53+ foci/cm2 was not measured in this study, but a crude estimate indicates that the p53+ foci were detected at an average of 4–40 foci/cm2 at the age 75.(Section lengths were 1–2 cm, and the size of a typical p53+ focus varied from 0.2 to 1 mm across, and prevalence per section was 58%;with a sun-exposed skin area of 1200 cm2, we found 4,800–48,000 foci/person at 75 years of age; projected into Fig. 4, this corresponds with a predicted yield of 2–10 squamous cell carcinomas/individual). According to the present mouse data, this would correspond with a risk of 90–100% at 75 years for this Swedish patient population. This appears to be a considerably higher risk than in the general Swedish population (which is on the order of 0.5% at 75 years of age). A study in the United States (4) did report on the number of p53 foci/cm2 (20–50 in chronically exposed areas) in discarded tissue from cosmetic surgery,but these data showed no clear age dependency, which may have been attributable to a selection bias and the small size of the data set. A prediction of tumor risk from the density of p53+ foci in humans clearly needs to be validated by proper human data.
If properly validated, measurements on p53+ foci could potentially be used as indicators/surrogates of skin cancer, at least in short-term assays in mice. For more practical purposes like protection of sunscreens against skin cancer, investigators have already started to use the early detection of p53 mutations to this end(42).
Although after 40 days huge numbers (>100 per cm2 of surface area of skin) of p53+ foci are detected, only a few tumors arise at later time points. Thus, if p53+foci are potential precursors, it has to be concluded that only a very minor fraction progresses to become tumors. As the chance of progression may simply be low, it may theoretically be just a matter of time before a p53+ focus progresses. The mice may not live long enough for this to happen, and outgrowing tumors may eventually crowd out the p53+ foci. On the other hand, we cannot exclude that not all of the p53+ foci have “tumor precursor potential.” Clearly, this issue has to be addressed in further research, e.g., by extensive comparative analyses of p53 mutation spectra in p53+ foci and in tumors from hairless mice (currently in progress in our group).
Our data clearly show quantitative relationships between the p53+ foci and skin tumors in wild-type hairless mice. To establish further the causality of this relationship, we ascertained the p53+ focus induction in transgenic mice that are more susceptible to UV-induced carcinogenesis.
The early detection of p53+ foci in the Xpa−/− mice at a very low level of UV exposure and the virtually complete absence of foci in the Xpa+/− control mice confirms the suspected enhanced induction of p53+ foci through an increased mutagenicity in the DNA repair-deficient background. This is in line with the abundance of p53+ foci found in an 18-year-old xeroderma pigmentosum group C(global genome repair-deficient) patient (43). Hence, an increased tumor risk in a DNA repair-deficient background corresponds to the early detection of p53+ foci.
Significantly more p53+ foci were found in the p53+/−knockout, compared with wild-type littermates. In this genetic background with diminished DNA repair and apoptotic capacity, an increase in p53+ foci formation again “predicts” tumor proneness. Although the chance of a mutational hit in the p53 gene is less, because one allele is already missing, p53+ focus formation is enhanced and again indicates increased tumor risk.
The induction of p53+ foci appears to be far more strongly enhanced in Xpa−/− mice than in p53+/− mice. This corresponds with the much larger difference in carcinoma induction times between Xpa−/− and Xpa+/− mice (factor 4.2; Ref. 30) than between p53+/− and p53+/+ mice (at the most, a factor of 2 at 1 MED/day).5
In conclusion, we can state that the formation of a p53+ focus is a very early event in UVB-induced carcinogenesis. UV dose dependencies of the p53+ foci and carcinomas appear to be very similar. In mice that are more susceptible to UV carcinogenesis, the p53+ foci develop earlier and in increased numbers. This confirms a direct relationship between p53+ foci and the subsequent tumors. Our data show that p53+foci appear to be good and useful indicators of UV-induced skin carcinoma risk.
Tumor Induction as a Sequence of Two Events with the Dose and Time Kinetics of p53+ Foci.
We write the yield of p53+ foci/unit skin area U(e.g., U = 1 cm2 or 1 epidermal sheet) as:
where D stands for the daily UV dose, t for time; Do, to, a, and b are constants; a/b = r = 0.58 ± 0.15=0.6 (by choosing a certain yield, e.g., Yf = 1 at t = t1, one finds that Drt1 = constant). The power of time, b, is found to equal 3.7 ± 0.7, and for computational convenience, is taken to equal 3.5. If the number of primary target cells (e.g., germinative basal epidermal cells) per unit surface area of skin equals N/U, we write the yield/target cell as:
We take this as the kinetics for the induction of a first stage in tumor progression, i.e., the formation of p53+ foci(yf is a “feeding function” of target cells for the next stage). Next, we model the experimental results on tumors by assuming that the probability (≪1) per p53+ to progress to a tumor after a time, t, over a small time interval, dt, also follows the same kinetics, i.e.,
To account for the fact that clones of p53+ cells in the first stage expand, we introduced a net target size (number of cells) per clone of stage-1 cells, σ. The expected tumor yield per primary target cell (e.g., basal epidermal cell) then becomes:
i.e., a convolution of yf(t) and pT(t). Eq. A4 can be solved simply through Laplace transformation, L [i.e.,a function of t is transformed into a function of s, which is the variable in the Laplace domain; importantly,the Laplace transform of convoluted functions equals multiplication of the Laplace transforms of the functions, and L[tx] =Γ(x + 1)/sx+1, whereΓ(x + 1) denotes the gamma function, which equals x! for integer values for x(44)]. Thus, we find:
where L[yf(t)] =[Γ(b+1)/N][D/D0]a[1/(t0bsb+1)]and L[pT(t)] = [bΓ(b)/N][D/D0]a[1/(t0bsb)],and by applying the inverse Laplace transformation on 1/s2b+1, we solve equation A4:
Note that the powers for the UV dose, D, and time, t, in the tumor yield equal 2a and 2b,respectively, i.e., double the values in the yield of p53+foci. Thus, in the tumor yield the power of time p = 2b =7, which is in agreement with the value of 6.9 ± 0.8 that we found in earlier experiments. Furthermore, we find for the tumors that r = 2a/2b = a/b = 0.6, which implies a similar dose-time relationship for p53+foci and tumors.
For the tumor yield per unit surface area U, we find:
If we take the point in time for which YT with 1 MED/day reaches the value of 1 per epidermal sheet (Fig. 4), we find that the extrapolated value of Yf reaches a value of ∼11,000/sheet of 7.2 cm2. If we assume that all basal epidermal cells are potential targets and that each basal cell takes up 10 × 10 μm2, we find that σequals an average of about two cells, i.e., a very small effective target size per p53+ focus. This could either mean that only a few (stem) cells per p53+ focus are true targets for the second UV-induced tumorigenic step, or that most of the p53+ foci are not suitable targets, i.e., not true precursors, which then lowers the effective number of target cells per p53+ focus. Thus, these calculations show that the time-dose dependency of the UV-induced skin tumors can simply be described by two sequential processes, each of which follows the time-dose dependency of UV-induced p53+ foci.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Financed mainly by grant UU97-1531 from the Dutch Cancer Society.
Internet address: http://www.iarc.fr/p53.
The abbreviations used are: NER, nucleotide excision repair; MED, minimal erythema/edema dose.
A. Westerman-de Vries, R. J. W. Berg, C. F. Kreijl, P. W. Wester, and H. J. van Kranen. Alterations in latency time, tumor progression and p53 mutation spectrum in UVB-induced skin tumors of p53-deficient hairless mice,manuscript in preparation.
We thank Hans Sturkenboom for loyal maintenance of the animals and Kees Guikers for general assistance.