UV-induced DNA Damage and Mutations in Hupki (Human p53 Knock-in) Mice Recapitulate p53 Hotspot Alterations in Sun-exposed Human Skin¹ Free

The p53 tumor suppressor protein is at the hub of multiple cellular control networks that regulate growth arrest, angiogenesis, DNA repair and apoptosis (1). This master-switch role is underscored by the fact that most human cancers have a malfunctioning p53 (2, 3). It is clear that the predominant mechanism by which p53 is inactivated during human carcinogenesis is acquisition of a single point mutation in the DBD³ that renders the mutant p53 protein unable to carry out its function in gene transcription regulation. Far less well understood are the causes of these mutations and the relative contributions of endogenous promutagenic factors and exogenous agents to the cellular mutation load.

Although a yeast-based system for examination of mutations in human p53 sequences on a plasmid is available (4, 5), there is currently no laboratory animal model to generate mutations in the human p53 gene experimentally in vivo. Such a test system would be useful not only for exploring hypotheses on the origins of mutations in cancer but also for in vivo screening of pharmaceuticals designed to restore DNA binding to human mutant p53 protein. We designed a transgenic animal model, the Hupki mouse to address this need. In this strain (p53^KI/KI), in which the endogenous murine p53 gene harbors the human instead of the murine DBD sequence, expression levels of the chimeric gene are normal, and various p53 functions are retained (6). The experiments we report here indicate that UVB-exposed Hupki mice show important molecular pathology features of sun-exposed human skin (7, 8): clusters of p53-immunoreactive cells with intensely staining nuclei develop; UV photoproducts map in Hupki mice to the same p53 gene dipyrimidine locations as in UVB-exposed human keratinocytes; and C to T transitions, including tandem mutations, arise at the p53 mutation hotspots identified in human basal and squamous cell carcinomas (9, 10, 11). Thus, the Hupki mouse allows for the first time experimental in vivo investigation of DNA damage sites and tissue hotspot mutation burden involving human p53 gene sequences.

For single acute exposure for DNA damage mapping by LM-PCR, the shaved dorsal areas (2 cm²) of four Hupki (p53 ^KI/KI) mice (6 weeks old; genetic background: 129/Sv) were exposed to a single acute dose of 50 kJ/m² UVB in a BioSpectra System (Vilber Lourmat, Marne-La Vallee, France) comprising an irradiation chamber equipped with 312-nm illumination lamps, a sensor/dosimeter, calibrator, and software. As chronic exposure for mutation analysis, shaved mice (five mice per exposure group) were exposed to 4.5 kJ/m² UVB radiation 5 × per week for 4 weeks with the Biospectra System. After sacrifice, epidermis was separated from dermis and divided into two segments. One segment was transferred to ABI lysis buffer [4 m urea, 0.5% n-laurylsarcosine, 0.2 M NaCl, in 100 mm Tris (pH 8.0)/10 mm CDTA, Applied Biosystems International], digested for 2 h at 56°C with proteinase K, and extracted with phenol-chloroform. DNA was precipitated in an ethanol/sodium acetate solution and resuspended in 10 mm Tris (pH 8.0)/0.1 mm EDTA. The remaining portion was fixed in ethanol, embedded in paraffin, and prepared for histology. Epidermal sections (4 microns) were stained with H&E, and images were recorded with Leica Quantimet 500 software (Leica, Bensheim, Germany). Immunoreactivity to anti-p53 antibody was examined using polyclonal antihuman p53 antiserum CM1 (Novacastra; dilution 1:1000) and peroxidase-conjugated secondary antibody according to standard protocols.

For mapping UV photoproduct distribution along the Hupki p53 gene after in vivo acute exposure of 6 week-old Hupki mice, DNA was prepared from epidermis as indicated above except that protein digestion was performed overnight at room temperature; DNA was cleaved with T4 endonuclease V and used as template for LM-PCR of exons 7 and 8 as described (12, 13). PCR products were resolved on 8% polyacrylamide gels and hybridized to ³²P-labeled single-stranded p53 gene exon 7- or 8-specific probes to visualize fragments.

Detection of (unselected) p53 mutations in 2-cm² areas of chronically exposed Hupki epidermis was achieved with BAS-PCR, a biotin/restriction enzyme PCR method of mutation enrichment (Fig. 4,A). Genomic DNA (200 ng) was digested with the restriction enzyme that cleaves the wild-type sequence of the major human skin tumor p53 mutation hotspot at codons 247–248 in exon 7 (MspI/HpaII) or at the hotspot codons 278–279 in exon 8 (MvaI). DNA was then amplified by 19 cycles of PCR (each cycle: 95° 1 min/62° l min/72° 30 s) with Pfu polymerase and P53 GeneChip primer pairs (Affymetrix, Inc.) for exon 7 and 8, one of each pair biotinylated. Streptavidin bead-bound PCR product was redigested with restriction enzyme before an additional 31 PCR cycles and separated on 3% agarose gels by electrophoresis after final restriction enzyme digestion. Enzyme-resistant PCR products for exon 7 or exon 8 were sequenced directly and also after cloning into plasmid vectors to resolve individual mutations present in the enzyme-resistant PCR product fragment. DNA sequencing was performed using BigDye fluorescent dye dideoxy reagents and an ABI Model 310 Genetic Analyzer as described (14). Reconstruction experiments with DNA from unirradiated nonmutated Hupki cells spiked with serial dilutions of DNA from a mutant Hupki embryonic stem cell strain harboring a codon 248 G to A hotspot transition were performed to estimate the sensitivity of the assay (Fig. 4 B).

The human and mouse domains of the Hupki chimeric p53 are depicted in Fig. 1. The Hupki gene is under normal endogenous transcriptional regulation at the murine locus, and homozygous p53^KI/KI mice show typical wild-type p53 responses to UV- and γ-irradiation-induced DNA damage: nuclear accumulation of Hupki p53 protein; enhanced expression of genes regulated by wild-type p53 including WAF1/p21, cyclin G, and noxa; and p53-dependent apoptosis of γ-irradiated thymocytes (6). Hupki mice develop normally; have no apparent physiological defects; and unlike p53-deficient or null mice, are not prone to spontaneous lymphomas, sarcomas, or other neoplasms.

Sun-exposed individuals have clones of epidermal cells with nuclei strongly immunoreactive to anti-p53 antibody (“p53 patches”) as do mice with a normal murine p53 genotype exposed for 17 or more days to UVB (8, 15). The staining from nuclear p53 protein accumulation is clearly distinguishable from the faint and diffuse reactivity to anti-p53 antibody observed shortly after a single exposure, which is attributed to transient stabilization of wild-type p53 (15). To determine whether Hupki mice show the characteristic precancerous p53 patch-response to chronic UVB irradiation, we exposed a 2-cm² dorsal area on each of five Hupki mice to UVB irradiation (4.5 kJ/m² 5 ×/week) for 4 weeks. At the end of treatment, patches of cells with enlarged nuclei that stained intensely with antihuman p53 CM1 antiserum were detected in tissue sections prepared from ethanol-fixed, paraffin-embedded Hupki epidermis. In two of the five exposed Hupki mice we noted epidermal hyperplastic areas 3–8 cells deep that were strongly immunoreactive (Fig. 2). With antimouse p53 CM5 antiserum, nuclear staining also can be detected in UVB-exposed wild-type mice (data not shown), as reported previously by Berg et al. (15). In an ongoing long-term UVB cancer test, skin hyperplasia, dysplasia, and carcinomas have been observed both in Hupki and wild-type exposed mice; thus far a difference between the genotypes in neoplastic response to exposure has not been apparent,⁴ although possible differences in sensitivity could surface when the long-term experiments are complete.

LM-PCR has been used to investigate the distribution of UV-induced photoproducts along the p53 gene of human cells and to explore correlations between formation of UV-induced DNA adducts and sites of frequent mutation in human skin cancer (13, 16). In UVB-exposed human keratinocytes and Hupki mouse skin irradiated in situ with UVB, the UV photoproduct distributions are essentially superimposable (Fig. 3); signals from cyclobutane pyrimidine dimers are seen almost exclusively at adjacent pyrimidines. A particularly strong signal is seen at codon 248, which is a prominent mutation hotspot in human skin cancers but not in UV-induced murine skin tumors. The normal mouse p53 gene sequence lacks the transcribed strand dipyrimidine at the equivalent exon 7 codon (murine codon 245; numbering system as in Ref. 17). Photoproducts were also mapped in exon 8 of the p53 gene in Hupki mouse skin, and the distribution pattern again matched that of human cells (data not shown).

Sun-exposed areas of skin from healthy individuals harbor p53 mutations that are typical signatures of UV irradiation mutagenesis: they are predominantly C to T transitions at sites of adjacent pyrimidines in the DBD of the p53 gene; they include CC to TT tandem mutations; and the mutations arise at sites that are frequently mutated in human skin cancers (7, 8). The two most prominent skin cancer mutation hotspots in the human p53 gene, codons 247–248 in exon 7 and at codons 278–279 in exon 8, together account for one-eighth of all of the p53 mutations identified to date in human skin cancers (IARC mutation database, version R4). Therefore, we tested for their presence in skin DNA of five Hupki mice exposed for 4 weeks to UVB irradiation. To detect mutants among the predominantly normal epidermal cell population, a protocol was developed that used restriction enzyme digestion of wild-type p53 sequences at the two hotspot locations (HpaII/MspI for hotspot in exon 7 and MvaI for hotspot in exon 8) and PCR amplification with one primer biotin-labeled to remove subsequently the enzyme-cleaved PCR product (Fig. 4,A). The procedure allows detection of one hotspot mutant among ∼100,000 unmutated cells (Fig. 4,B). We identified the presence of mutant p53 sequences at both hotspots in the epidermis of the treated Hupki mice but not in the untreated mice, as shown for the hotspot in exon 7 at codons 247–248 in Fig. 5 (data not shown for the hotspot at codon 278–279). These data indicate a good correspondence between cyclobutane pyrimidine dimer damage, which is especially prominent in exon 7 at codon 248 in both human and Hupki epidermal cells (Fig. 3) and the presence of UV-specific mutations at the exon 7 hotspot in chronically irradiated Hupki skin (Table 1). To resolve each mutation and to estimate the relative proportions of the various hotspot mutations in the enzyme-resistant PCR product, we then cloned the noncleaved DNA fragments into plasmid vectors and sequenced 59 clones, 39 harboring exon 7 fragments and 20 with exon 8 fragments. Fifty-eight of the 59 mutant clones harbored a transition mutation (Table 1), including 9 (15%) with tandem CC to TT mutations. At hotspot 278–279, all of the mutations (20/20) were at sites where the dipyrimidine resided on the nontranscribed strand, whereas at the exon 7 hotspot, a strand bias was less pronounced and shows the presence of transitions (28 of 39 clones) at locations where premutated dipyrimidine sequences are on the transcribed strand. The induction of mutations at the two human hotspots in UV-exposed Hupki mice and the very high preponderance of transition mutations over transversions (Table 1), as well as the presence of tandem CC to TT mutations and the more prominent nontranscribed strand orientation of the premutated pyrimidine seen for mutations at the exon 8 hotspot compared with the exon 7 site (codons 247–248; Table 1) are features that are characteristic of p53 mutations at these sequences in sun-exposed human skin tumors.

Individual mutagens generate characteristic changes at specific locations along a given DNA sequence. The murine p53 gene is not an optimal target sequence to explore origins and modulators of p53 mutation patterns found in human tissues, because the mouse p53 gene differs from the human coding sequence at 15% of residues, and base sequence context is a primary determinant of a UV-irradiation or chemically induced mutation spectrum. For example, codons 248 and 249 of the human gene together account for ∼10% of the mutations recorded in the IARC p53 mutation database,⁵ yet 3 of the 6 bases composing these two heavily mutated codons are dissimilar in the two species (human: CGG AGG and mouse: CGC CGA). Whereas UVB- and aflatoxin B₁-associated human tumor p53 mutations accumulate at this sequence, not unexpectedly, UVB-induced mouse skin tumors and aflatoxin B₁-induced murine lung tumors were not found to have p53 mutations at the equivalent location (18, 19). Particularly notable is the absence of UVB-induced murine p53 mutations in the mouse p53 gene position equivalent to human codon 247–248 (18), because this is the most frequently mutated site in human skin tumors.

Regarding applications of the Hupki model, tissue inflammatory responses involve a complexity of different cell and molecular interactions, generating free radicals such as nitric oxide or lipid peroxidation products that contribute to the mutation load in human carcinogenesis (20, 21, 22). The Hupki mouse offers an in vivo approach to such investigations, allowing comparison with human mutation data. It remains to be seen whether major differences in human and murine metabolism and metabolic rate will be reflected in mutations generated from oxidative damage. Some modulation in mutation profiles because of species-specific features of DNA repair may become apparent in mutagenesis studies with the model, because premutagenic lesions in the Hupki p53 gene are subject to the murine DNA repair system. Interbreeding of the Hupki strain with DNA repair-deficient transgenic mouse strains can be undertaken to investigate the role of different DNA repair pathways in shaping p53 mutation patterns observed in humans (23, 24). Measurement of p53 hotspot mutation load in clinical material (e.g., target tissue and plasma) from cancer patients or from individuals heavily exposed to carcinogens is a powerful new technique (22, 25) that can be explored experimentally in the Hupki mice at the identical p53 sequence. Given species and strain differences in carcinogen metabolism and tissue tumor response, carcinogenic agents that are direct-acting mutagens would be the most appropriate choice for tumor induction and p53 mutation profiling in Hupki mice to additionally validate the model. For some research questions Hupki epidermis can serve as surrogate tissue for induction of neoplastic lesions, an approach that has been used in mice with a wild-type murine p53 genotype to demonstrate G to T mutation specificity of the human carcinogen benzo(a)pyrene (26).

We have shown that UV-treated Hupki epidermis harbors specific p53 DBD mutants typical of sun-exposed human skin of healthy individuals and of sunlight-associated human tumors (Table 1). Although we have not addressed pharmacological applications of the Hupki mouse experimentally in the present study, our data imply that this strain will lend itself to in vivo preclinical testing of drugs designed to modulate the human p53 DBD and restore transcriptional and tumor suppressive functions to human p53 mutants (27, 28, 29). The remarkable vulnerability of p53 biological function to single amino acid residue substitutions has been appreciated for some time in view of the variety and abundance of single missense p53 mutations found in human tumors. More recent findings have broadened the implications of this sensitivity. Intragenic suppressors of common human tumor mutations have been discovered, and these second site amino acid substitutions restore DNA binding affinity to some mutants (30, 31). A common germ-line p53 variant in human populations at codon 72 also has been shown to be an intragenic modifier of tumor somatic DBD mutants: the presence of the arginine residue variant on an allele that in addition harbors a conformational mutation in the core domain invests certain of these mutant p53 molecules with the capacity to bind to proapoptotic p73, thereby compromising function of the latter (32). These observations underscore the potential value in testing p53-targeting drugs with an in vivo experimental model that presents the precise mutant p53 DBDs found in human tumors.