Dual Role of Inactivating Lef1 Mutations in Epidermis: Tumor Promotion and Specification of Tumor Type

The NH 2 terminus of LEF1 is frequently mutated in human sebaceous tumors. To investigate how this contributes to cancer, we did two-stage chemical carcinogenesis on K14#NLef1 transgenic mice, which express NH 2-terminally truncated Lef1 in the epidermal basal layer. Transgenic mice developed more tumors, more rapidly than littermate controls , even without exposure to tumor promoter. They developed sebaceous tumors, whereas controls developed squamous cell carcinomas. K14#NLef1 epidermis failed to up-regulate p53 and p21 proteins during tumorigenesis or in response to UV irradiation, and this correlated with impaired p14ARF induction. We propose that LEF1 NH 2-terminal mutations play a dual role in skin cancer, specifying tumor type by inhibiting Wnt signaling and acting as a tumor promoter by preventing induction of p53.


Introduction
Aberrant activation of canonical Wnt signaling occurs in many tumors (1).Stabilizing mutations within the NH 2 terminus of h-catenin or genetic defects in molecules regulating h-catenin degradation lead to nuclear accumulation of h-catenin and activation of Tcf/Lef target genes.Transgenic mice overexpressing a stabilized mutant form of h-catenin in the epidermis develop hair follicle tumors (2,3), and activating h-catenin mutations are found in the corresponding human tumors (4).
Surprisingly, an association between tumors and inhibition of Wnt signaling has also been reported.Mutation or deletion of the NH 2 terminus of Lef1 prevents h-catenin binding.When DN32Lef1 is expressed in the basal layer of the epidermis via the K14 promoter, mice (K14DNLef1 transgenics) develop sebaceous tumors at high frequency (5).A high proportion of human sebaceous adenomas and sebeomas have double nucleotide substitutions in exon 1 of the LEF1 gene (6).These result in E45K and S61P amino acid substitutions in the NH 2 terminus, which impair binding to h-catenin and inhibit h-catenindependent transcription.
At present, it is unclear how Lef1 mutation contributes to tumor formation.One possibility is that its role is solely to specify the differentiated characteristics of the tumor because h-catenin signaling levels control lineage selection in normal epidermis (7)(8)(9)(10)(11)(12)(13). In K14DNLef1 transgenics, hair follicles convert into epidermal cysts with sebocyte and interfollicular differentiation (5,13).Another possibility is that, in addition to directing the differentiated characteristics of a tumor, Lef1 mutations increase epidermal susceptibility to tumor development.DNLef1 and E45K + S61P LEF1, although unable to bind h-catenin, retain other properties, such as the ability to bind Groucho corepressors and to increase expression of Indian hedgehog (6,12), and these might positively contribute to tumorigenesis (e.g., by stimulating proliferation of sebocyte progenitors; ref. 12).In culture, NH 2 -terminally truncated Tcf4 blocks induction of p14ARF by h-catenin (14).Thus, DNLef1 could potentially prevent accumulation of the p53 tumor suppressor protein by preventing induction of ARF.
In the present report, we set out to test the hypothesis that deletion of the NH 2 terminus of Lef1 both directs tumor type and stimulates tumor formation.

Materials and Methods
Experimental mice.All procedures conformed to Cancer Research UK ethical guidelines and the terms of a British Home Office license.K14DNLef1 transgenic mice ( founder line L; ref. 5) were maintained on a CBA Â C57Bl/6 F1 background.
Tumor experiments.Seven-week-old female K14DNLef1 transgenic mice and wild-type littermate mice were used (15).Typically, in each experiment, 20 wild-type and 20 transgenic mice received a subthreshold dose of carcinogen in 200 AL acetone [100 nmol 7,12-dimethylbenz (a)anthracene (DMBA) or 1.6 Amol benzo[a]pyrene (B[a]P); Sigma].Wildtype mice received 12-O-tetradecanoylphorbol-13-acetate (TPA; 6 nmol in 200 AL acetone) thrice per week for 25 weeks, whereas transgenic mice were treated twice per week for 10 weeks because of their high sensitivity to TPA.Control groups of 20 mice were treated with DMBA or B[a]P and acetone, acetone and TPA, or acetone alone.Experiments were done twice and results obtained were similar.Scoring of tumors was carried out once a week for up to 50 weeks after tumor initiation.Bromodeoxyuridine (BrdUrd) labeling and tissue harvesting were done as previously described (15).
Reverse transcription-PCR.RNA was isolated from skin and tumors using Tri-reagent (Helena BioSciences Ltd., Gateshead, United Kingdom) and transcribed into cDNA using Ready-to-go You-prime First-strand beads and oligo-d(T) primer (Amersham Biosciences, GE Healthcare).Primers are described in Supplementary data.

Results
#NLef1 increases sensitivity to chemical carcinogenesis and is a tumor promoter.K14DNLef1 transgenic and wild-type littermate control mice were subjected to classic two-stage carcinogenesis protocols (Fig. 1).The skin received one application of DMBA or B[a]P to induce Ha-Ras mutations (15) and repeated TPA treatments to stimulate tumor promotion.
K14DNLef1 transgenic mice treated with DMBA and TPA developed seven times as many tumors as their wild-type littermates (K14DNLef1: 22 tumors per mouse; wild-type: 3 tumors per mouse; P < 0.001, Student's t test; Fig. 1A).Transgenic animals treated with B[a]P and TPA developed 10 times more tumors than control mice (K14DNLef1: 16 tumors per mouse; wild-type: 1.6 tumors per mouse; P < 0.001; Fig. 1B).Of K14DNLef1 transgenic mice, 100% developed tumors with either protocol, whereas the proportion of tumor-bearing wild-type mice was 90% in response to DMBA (Fig. 1C) and 65% in response to B[a]P (Fig. 1D).In addition, tumors developed much faster in K14DNLef1 mice than in littermate controls (Fig. 1C and D).We conclude that K14DNLef1 transgenic mice were much more sensitive to twostage carcinogenesis than transgene-negative mice.
The classic controls in chemical carcinogenesis experiments are to treat animals with DMBA or TPA alone (16).As predicted, wildtype mice subjected to DMBA or TPA alone did not develop tumors (data not shown; ref. 15).TPA treatment did not affect the spontaneous tumor incidence (<1 tumor per mouse; 27% of mice  #NLef1 determines the differentiated characteristics of tumors with Ha-Ras mutations.In response to DMBA and TPA, wild-type mice first develop benign papillomas, some of which progress (1% in Fig. 1) to malignant squamous cell carcinomas (15,16).Papillomas and squamous cell carcinomas have characteristics of interfollicular epidermal differentiation, including accumulation of cornified layers (Fig. 2A).In contrast, all of the tumors induced in K14DNLef1 transgenic mice with DMBA or B[a]P F TPA exhibited a high degree of sebocyte differentiation, resembling the spontaneous tumors that lack Ras mutations (Fig. 2B and C).Macroscopically, some of the K14DNLef1 tumors consisted of a ''head'' on a stalk (''raised''), resembling wild-type papillomas, whereas others were flattened (''flat'').However, raised tumors were not precursors of flat tumors, and histologically they were indistinguishable (Fig. 2B and C).We conclude that DNLef1 overrides the genetic program for squamous differentiation in tumors bearing Ha-Ras mutations.
K14#NLef1 tumors are distinguished by reduced p53 and p21 protein levels.We did not detect any significant difference in the number of S-phase (BrdUrd-positive) undifferentiated cells in wild-type and K14DNLef1 tumors (Fig. 3A; Supplementary Table S1).This led us to consider other potential mechanisms for the tumor-promoting effect of DNLef1.In cell culture, activation of h-catenin induces accumulation of transcriptionally active p53 (14,17), which in turn could protect against neoplastic conversion (18).Because DNLef1 blocks h-catenin activation (5), we investigated whether it prevented up-regulation of p53 in skin tumors.p53 protein is induced by oncogenic Ras (18) and, as expected, papillomas and squamous cell carcinomas from wild-type mice showed nuclear accumulation of p53 (n = 3; Fig. 3B; Supplementary Table S1).Trichofolliculomas induced by prolonged activation of h-catenin in K14DNh-cateninER transgenics (3) also displayed p53 accumulation (n = 2; Fig. 3B; Supplementary Table S1).In contrast, the sebaceous tumors (n = 13) that developed in K14DNLef1 mice were either negative for p53 (n = 8) or had only a small number of positive cells (n = 5; Fig. 3B; Supplementary Table S1), regardless of whether they were spontaneous or induced by DMBA or B[a]P F TPA. Sequencing exons 4 to 9 of p53 in six of six DMBA/TPAinduced K14DNLef1 tumors revealed p53 to be wild-type.
To determine whether lack of p53 protein reflected transcriptional or posttranslational regulation, we isolated RNA from tumors and unaffected skin of wild-type and K14DNLef1 transgenic mice.We carried out reverse transcription-PCR (RT-PCR) with primers for p53, p21, and ARF, a protein that increases p53 protein stability (18).p53 and ARF mRNAs were undetectable in skin from wild-type and transgenic animals (Fig. 3D).p53 mRNA was readily detected in all tumors analyzed (Fig. 3D).However, wild-type and K14DNLef1 tumors differed in ARF mRNA levels (Fig. 3D), with wild-type tumors expressing more (Fig. 3D).In wild-type tumors the ARF band was more intense than the p53 band, whereas in transgenic tumors the p53 band was more intense.p21 levels were higher in tumors than in unaffected skin, but there were no differences between wild-type and transgenic tumors (Fig. 3D).
The reduction in ARF expression was confirmed by immunohistochemistry. Two of two wild-type squamous cell carcinomas and one of one wild-type papilloma contained a large number of cells with ARF-positive nuclei (Fig. 3D).In contrast, ARF was undetectable in three of three spontaneous tumors and 13 of 16 transgenic tumors induced by DMBA or B[a]P (Fig. 3D), and in 3 of 16 transgenic tumors there were fewer than 10 ARF-positive nuclei per section (data not shown).Taken together, the results suggest that the lack of detectable p53 protein in K14DNLef1 tumors reflects rapid degradation of the protein as a result of an impaired ability to up-regulate ARF.
Differential responsiveness of wild-type and #NLef1 epidermis to UV radiation.Up-regulation of p53 is a welldocumented response of mammalian epidermis to UV radiation (19,20).If induction of p53 protein is indeed impaired by expression of DN32Lef1, then the epidermis of transgenic mice should show increased susceptibility to UV irradiation.To test this, we exposed transgenic and wild-type mouse back skin to a combination of UVA and UVB (data not shown) or UVB alone and compared exposed and unexposed skin 24 h later (Fig. 4A-C).
Whereas the skin of wild-type mice had normal morphology after UV treatment, the skin of K14DNLef1 mice was dramatically altered, with regions of epidermal detachment from the dermis (Fig. 4A, arrows).UV irradiation increased apoptosis and proliferation in both wild-type and transgenic skin (Fig. 4B; data not shown).The fold increase in Ki67-positive cells was variable but generally higher in transgenic (mean, 2.6; range, 1.08-4.57)than in wild-type (mean, 1.3; range, 0.82-1.82)epidermis ( five independent experiments).
UV irradiation resulted in strong nuclear accumulation of p53 in wild-type skin (Fig. 4C).In contrast, there were very few cells with nuclear p53 in UV-treated K14DNLef1 skin (Fig. 4C).The percentage of p53-positive nuclei was >70% lower in K14DNLef1 compared with wild-type epidermis (2.5% versus 9.0%; five independent experiments; Fig. 4D).Thus, the sensitivity of K14DNLef1 epidermis to UV-induced damage correlates with a failure to up-regulate p53.Nuclear accumulation of ARF could not have been detected in skin from wild-type and K14DNLef1 mice with or without UV treatment (Fig. 4C and data not shown).

Discussion
Mutation or deletion of the NH 2 terminus of LEF1 prevents h-catenin-dependent transcriptional activation and is associated with human and mouse sebaceous tumors (5,6).Using K14DNLef1 transgenics, we now show that NH 2 -terminal deletion of Lef1 contributes to tumorigenesis in two ways.It specifies the type of tumor formed as a result of chemically induced Ras mutations and it acts as a tumor promoter.
Chemically induced tumors bearing Ras mutations in wild-type epidermis usually express markers of the interfollicular epidermal differentiation pathway (16).In contrast, chemically induced tumors in K14DNLef1 mice contained differentiated sebocytes.DN32Lef1 thus specifies tumor type independent of the presence (chemically induced) or absence (spontaneous) of Ras mutations.Our experiments do not distinguish between the alternative possibilities that tumor type depends on the target cell (sebocyte or interfollicular epidermal progenitor) or that a common target cell with multilineage differentiation potential (putative stem cell) forms different types of tumor in response to the specific signals it receives (16).However, it is striking that the types of tumor directed by DN32Lef1 and stabilized h-catenin reflect the differentiated lineages selected by normal epidermis in response to different levels of h-catenin activation (8,16).
UV-induced mutations in p53, resulting in elevated levels of the protein, are detected at high frequency in phenotypically normal epidermis and are common in skin tumors (20).Human sebaceous tumors express high levels of p53, consistent with sunlight-induced mutations, and these could potentially drive tumor formation, with mutant LEF1 solely specifying tumor type (6).However, our experiments show conclusively that DNLef1 plays a positive role in tumorigenesis, in particular acting as a strong tumor promoter in mice treated with DMBA alone.Sequencing of K14DNLef1 tumors did not reveal any p53 mutations and K14DNLef1 tumors lacked detectable p53 protein.The discrepancy between the p53 status of human and mouse sebaceous tumors may be explained by the different time scales involved: the mouse tumors developed within weeks (Fig. 1), whereas the average age of humans with LEF1 mutant tumors was 70 years, allowing time for additional oncogenic changes such as accumulation of p53 mutations and failure of DNA mismatch repair (6).
Activation of p53 by oncoproteins occurs mainly via ARF, which binds to murine double minute-2 and thereby suppresses p53 ubiquitination and degradation (18).We found that K14DNLef1 tumors had, like wild-type tumors, elevated p53 mRNA; however, the levels of ARF mRNA and protein were reduced.This provides in vivo validation of cell culture experiments showing that ARF transcription is blocked by DNTcf4 (14,17).We therefore propose that DNLef1 specifies tumor type by preventing h-catenin-dependent induction of hair follicle genes and acts as a tumor promoter by preventing accumulation of the tumor suppressor p53 (Fig. 4D).

Figure 1 .
Figure 1.Formation of benign and malignant skin tumors.Frequency (A and B) and incidence (C and D ) of tumors in wild-type and K14DNLef1 transgenic mice.Female mice received a single subthreshold dose of DMBA (A and C ) or B[a]P (B and D ; initiation).One week later, animals were promoted with topical application of TPA or acetone vehicle only.Mice were observed for 52 wk after initiation.The tumors in wild-type mice were a combination of papillomas and squamous cell carcinomas.The actual numbers of mice used in the experiments shown were as follows: 18 K14DNLef1 (DMBA + TPA), 15 K14DNLef1 (B[a]P + TPA), 15 K14DNLef1 (DMBA + acetone), 15 K14DNLef1 (B[a ]P + acetone), 20 wild-type (DMBA + TPA), and 20 wild-type (B[a]P + TPA).
>3 months old develop tumors; ref.5) in K14DNLef1 transgenics (data not shown; ref.5).However, in contrast to wild-type mice, K14DNLef1 transgenics treated with DMBA or B[a]P alone developed skin tumors; furthermore, the frequency and kinetics were similar whether transgenics received carcinogen alone or in combination with TPA (P = 0.388 for DMBA and TPA versus DMBA alone; P = 0.552 for B[a]P and TPA versus B[a]P alone; Fig.1).We conclude that DNLef1 functions as a tumor promoter, cooperating with Ha-Ras to induce tumor formation.Mutations in Ha-Ras codon 61 (CAA!CTA), the signature DMBA induced lesion(16), were detected in 10 of 10 tumors of K14DNLef1 mice treated with DMBA or DMBA and TPA.In contrast, no mutations in Ha-Ras exons 1 and 2 were found in five of five spontaneous K14DNLef1 tumors.Thus, although DNLef1 promotes development of tumors with Ha-Ras mutations, Ras mutations are not the underlying cause of the spontaneous tumors in K14DNLef1 mice.

Figure 2 .
Figure 2. Histology of skin tumors.H&E-stained sections of tumors from wild-type (wt; A ) and K14DNLef1 transgenic mice (DNLef1 ; B and C ).A and B, chemically induced tumors; C, spontaneous tumors.Pap, papilloma; SCC, squamous cell carcinoma.Sebaceous tumors (B and C ) were macroscopically raised or flat.Arrows, regions of extensive sebaceous differentiation.CE, accumulation of cornified layers, indicating squamous differentiation.Bar, 100 Am.

Figure 4 .
Figure 4. Mechanism of p53 regulation in vivo.A to C, back skin of adult wild-type and K14DNLef1 transgenic mice, either unirradiated (No UV ) or after a single dose of UVB radiation (0.36 J/cm 2 ; UV ).A, H&E staining; B, immunolocalization of Ki67; C, immunolocalization of p53 and ARF.Open arrows, detachment and destruction of cells in the interfollicular epidermis (A-C); closed arrows , positive nuclei (B and C ). Bar, 200 Am (A), 50 Am (B ), and 100 Am (C).D, left, quantitation of percent p53-positive nuclei in UV-irradiated wild-type and transgenic epidermis.Columns, mean from five independent experiments; bars, SD.A total of 7,281 transgenic and 8,265 wild-type nuclei were scored.Right, model of role of DNLef1 in tumor promotion and tumor type specification.