Spontaneous squamous cell carcinoma induced by the somatic inactivation of retinoblastoma and Trp53 tumor suppressors. Cancer Res

Squamous cell carcinomas (SCC) represent the most aggressive type of nonmelanoma skin cancer. Although little is known about the causal alterations of SCCs, in organ-transplanted patients the E7 and E6 oncogenes of human papillomavirus, targeting the p53- and pRb-dependent pathways, have been widely involved. Here, we report the functional consequences of the simultaneous elimination of Trp53 and retinoblastoma (Rb) genes in epidermis using Cre-loxP system. Loss of p53, but not pRb, produces spontaneous tumor development, indicating that p53 is the predominant tumor suppressor acting in mouse epidermis. Although the simultaneous inactivation of pRb and p53 does not aggravate the phenotype observed in Rb-deficient epidermis in terms of proliferation and/or differentiation, spontaneous SCC development is severely accelerated in doubly deficient mice. The tumors are aggressive and undifferentiated and display a hair follicle origin. Detailed analysis indicates that the acceleration is mediated by premature activation of the epidermal growth factor receptor/Akt pathway, resulting in increased proliferation in normal and dysplastic hair follicles and augmented tumor angiogenesis. The molecular characteristics of this model provide valuable tools to understand epidermal tumor formation and may ultimately contribute to the development of therapies for the treatment of aggressive squamous cancer.


Introduction
The nonmelanoma skin cancer (NMSC), the most common cancer among humans, comprises two major types: squamous cell carcinoma (SCC) and basal cell carcinoma (BCC).Several factors, including the clinical characteristics, differentiate between BCCs and SCCs, being the latter more aggressive with increased metastatic properties.The major factor responsible for development of NMSC is UV radiation (1,2).However, whereas for BCC development the genetic hallmark is abrogation of the patched-sonic hedgehog pathway, little is known about the causal alterations of SCCs.Mutations in Trp53 tumor suppressor gene are frequently observed in both types of tumors.Interestingly, organ transplant recipients have an up to 100-fold increased risk of SCC and a 10-fold increased risk of BCC, resulting in a reversal of the normal ratio of SCC to BCC (2).Several studies have suggested the involvement of cutaneous human papillomaviruses (HPV) in the development of NMSCs, in part through alterations in p53 and pRb pathways mediated by E6 and E7 oncoproteins, respectively (2).
The Trp53 gene product, p53, coordinates the cellular response to stress, including DNA damage, hypoxia, and oncogenic stress through transcriptional mechanisms, resulting in cell cycle arrest, senescence, or apoptosis (3).The tumor suppressor retinoblastoma (Rb) also acts through transcriptional mechanisms to regulate cell cycle progression, differentiation, and apoptosis (4).Numerous evidences have shown the functional connection between pRb-and p53-dependent pathways (5,6).For instance, p21 waf1 is induced by p53 and negatively regulates cyclin-dependent kinase (cdk) function, leading to the inhibition of mechanisms that functionally inactivate pRb by phosphorylation.In addition, p21 waf1 can inhibit cell proliferation in cells lacking pRb through direct binding to cdks and E2F transcription factors (7)(8)(9).On the other hand, the inactivation of pRb leads to activation of E2F transcription factors, which produces sustained p53 induction through p19 arf -dependent (10,11) and p19 arf -independent mechanisms (12,13).In support of these cooperative functions, different transgenic mouse models bearing inactivation of both tumor suppressor genes have been generated, providing useful prototypes of human tumors of lung (14), medulloblastoma (15), and prostate (16).
In epidermis, the specific ablation of Rb gene does not result in spontaneous tumors (17), unless other pRb family members, such as p107, are also inactivated (18). 1 In addition, chemical carcinogenesis protocols in these mice result in reduced number of tumors, which also show smaller size and increased malignant conversion (19).This is mediated, at least in part, by the induction of p53 and the consequent apoptosis at early stages (20).Importantly, the subsequent inactivation of p107 abolishes p53dependent apoptotic processes, leading to spontaneous tumor formation (18). 1 It has been previously reported that the specific ablation of Trp53 gene in stratified epithelia leads to spontaneous tumor development (21,22).However, to our knowledge, there are no studies about the possible cooperation between pRb and p53 in epidermis in vivo.
To analyze all these aspects, we sought to create a mouse skin cancer model based on joint inactivation of Rb and Trp53 in stratified epithelia.To this, we bred mice bearing floxed alleles of Rb gene (Rb F19/F19 ; ref. 17) with mice bearing floxed alleles of Trp53 (Trp53 F2/F10 ; refs.22,23) and with transgenic K14cre mice (22).The Rb F19/F19 ;Trp53 F2/F10 ;K14cre (thereafter pRb À/À ;p53 À/À ) mice, losing these tumor suppressor genes in stratified epithelia (17,22), and their different allelic combinations, were generated.We show that somatic inactivation of Trp53 does not produce any evident phenotypic changes nor it aggravates the phenotypic consequences of Rb ablation in epidermis.On the other hand, p53 loss leads to spontaneous tumor formation, a process that is dramatically accelerated by simultaneous pRb loss.Such cooperative activities seem to be promoted by the early activation of the epidermal growth factor receptor (EGFR)/Akt axis, leading to an early angiogenic response.

Materials and Methods
Experimental mice and genotyping.Mice of the different genotypes in FVB background [Rb F19/F19 (17), Trp53 F2/F10 (22,23), and K14cre mice (22)] were genotyped by PCR using specific primers as previously described (17,22,23).Mice of all genotypes were monitored for tumor formation over a period of 20 months.Tumor volume was calculated as p(4/3) Â (width/2) 2 Â (length/2) using isolated tumor specimens.All animals showing obvious tumors or morbidity were sacrificed for necropsy and histopathologic analyses.The animal experiments were approved by the Animal Ethical Committee and conducted in compliance with Centro de Investigaciones Energe ´ticas, Medioambientales y Tecnolo ´gicas (CIEMAT) Guidelines.
RNA purification and Affymetrix mouse GeneChip 430A analysis.Newborn mouse skin tissue was preserved in RNAlater (Ambion) and disrupted and homogenized using Mixer Mill MM301 (Retsch).Total RNA was extracted and purified from 30 mg of skin using RNeasy Fibrous Tissue Mini kit (Qiagen) following the manufacturers' recommendations.The integrity of the RNA populations was tested in the Bioanalyzer (Agilent) showing at least 1.4 28S to 18S ratio.Hybridization of pooled samples from three animals of each genotype into Affymetrix MOE 430A microarrays was performed per duplicate at the Genomic Facility of the Cancer Research Center (Salamanca, Spain).We exported CEL files from Affymetrix GeneChip Operating Software and performed background substraction with robust multichip average (25) and normalized the chips with the quantile method (26) using the Gene Expression Pattern Analysis Suite (27).The signal intensity values of each probe set were log 2 transformed.Further analyses were performed using the MeV software (28).The selection of genes differentially expressed between control and mutant mice (1,435 gene IDs) was made using ANOVA analysis (P V 0.01).The list of significant genes implicated in phosphatidylinositol 3-kinase/Akt pathway and angiogenesis process was searched using Gene Ontology Biological Process.
Epidermal proliferation, analyzed by BrdUrd incorporation, showed no significant increase in the Trp53-deficient epidermis compared with wild-type epidermis (Supplementary Fig. S1; see also Fig. 4A).In agreement with our previous data (17), we found increased proliferation in pRb-deficient epidermis, which was nonaugmented by the simultaneous absence of both tumor suppressors in the interfollicular epidermis.Interestingly, we observed increased proliferation in hair follicles of pRb À/À ;p53 À/À mice (see below).Finally, we also observed a significant BrdUrd incorporation in some scattered dermal cells in the dermis (denoted by arrows in Supplementary Fig. S1) in pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À samples.
Collectively, these observations show that the epidermal-specific ablation of p53 does not produce any obvious phenotypic abnormalities nor aggravate those produced by pRb ablation with respect to proliferation and differentiation, being the only alteration detected the abnormal proliferation of scattered cells in the dermis.
Spontaneous tumors in mice bearing the Rb and Trp53 ablation in epidermis.Despite the fact that p53 loss in epidermis did not cause any evident phenotype in nonlesional skin, we observed that all the mice lacking p53 in epidermis developed spontaneous tumors (Fig. 2A).The onset of appearance of these lesions is 5 months of age and all the mice developed tumors by 12 months of age (median, 8.3 F 1.1 months).The tumor susceptibility is accelerated in double-deficient mice (Fig. 2A), where we observed that the onset of tumor appearance was 3 months and all mice developed tumors by 8 months (median, 5.5 F 1.4 months; P V 0.0009).The multiplicity of tumors is also increased in double-deficient mice compared with Trp53deficient mice (3.4 F 0.7 versus 1.1 F 0.8, respectively; P V 0.0006).Of note, the type and localization of the tumors were similar between pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À mice and almost exclusively affected different areas of the skin, such as neck, snout, perianal, face, and back skin, which in most cases do not produce external wounds.Histology analysis revealed that all these tumors were predominantly epidermal SCCs (84.5% of 197 tumors analyzed).These results showed that p53 is an essential tumor suppressor in epidermis, in agreement with others (21,22,29).In addition, our data also indicate that pRb and p53 cooperate to suppress mouse skin tumorigenesis.
To further analyze the cooperation between both tumor suppressors, we performed a detailed study of tumor development in a large cohort of mice bearing different combinations of Trp53 and Rb alleles.In these analyses, we found that Rb heterozygosity did not increase the rate of appearance of tumors in mice lacking Trp53 (Fig. 2B).On the other hand, Trp53 heterozygosity in pRb À/À mice leads to spontaneous tumor development, although it is significantly delayed compared with complete Trp53 loss (Fig. 2C).We also detected loss of the remaining Trp53 allele in some tumors of pRb +/À ;p53 +/À and pRb À/À ;p53 +/À mice (data not shown).The rate of appearance of tumors was similar between pRb +/À ;p53 +/À and pRb À/À ;p53 +/À (Fig. 2C).Consequently, the tumor susceptibility attributable to Trp53 loss, either complete or in heterozygosis, is only accelerated by the complete loss of Rb.
Hair follicle origin of spontaneous tumors.The vast majority of the tumors observed in all the genotypes correspond to poorly differentiated SCCs with loss or reduced expression of suprabasal keratins K10, K6, and K17 (Fig. 3A).Interestingly, the study of skin of tumor-bearing mice revealed multiple dysplastic hair follicles, which in most cases are also characterized by the presence of tumoral cell masses arising from the tip of these structures (Fig. 3B).This suggests that the tumors might have a hair follicle origin.This observation was further supported by the finding that early-stage tumors, associated with dysplastic hair follicles, express markers such as keratins K15 and K17 in addition to the basal keratin K5 (Fig. 3B) and suprabasal K6 (data not shown).Importantly, the incidence of these type of lesions is higher in pRb À/À ;p53 À/À (41.6%; n = 36) compared with pRb +/+ ;p53 À/À (23.5%; n = 17) mice.
Collectively, these data indicate that the increased incidence and early appearance of tumors in pRb À/À ;p53 À/À mice is attributable to increased proliferation but not to altered apoptosis in hair follicles, leading to early dysplasia and malignant conversion.
We also monitored the growth rate of the tumors.However, as the dysplastic and early-stage tumors were only detectable during histology analyses, we focused only in advanced isolated tumors.We observed increased volume in double-deficient mice tumors compared with p53-deficient tumors in age-matched samples (Fig. 4B).This result indicates that, in spite of similar proliferation rates, double-deficient tumors might have alternative mechanisms leading to augmented tumor growth.
Finally, as the SCC can undergo epithelial mesenchymal transition (EMT) leading to the formation of highly aggressive spindle cell carcinomas, we analyzed whether EMT is increased in the absence of pRb and p53.A significant number of spindle cell carcinomas was observed in pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À mice (Supplementary Fig. S4).We found that the rate of appearance of spindle cell carcinomas is slightly accelerated in pRb À/À ;p53 À/À compared with pRb +/+ ;p53 À/À (Supplementary Fig. S4).However, the rate of conversion of SCC into spindle cell carcinoma was similar between the two genotypes (Supplementary Fig. S4), suggesting that EMT is not significantly altered in double-deficient tumors.
of Spry2 (Fig. 5C), whose gene product, Sprouty 2, inhibits the activation of MAPK pathway induced by activated EGFR (32).We observed increased expression of Sprouty 2 in the skin of mice lacking p53, which was more evident in pRb À/À ;p53 À/À skin (Fig. 5B).This finding may help to explain why the activation of EGFR leads to increased Akt, but not ERK activity, in the pretumoral stages of pRb À/À ;p53 À/À skin.
The absence of Rb and Trp53 induces a premature increased angiogenic response.The increased Akt activity found in tumors and pretumoral skin would be responsible for the increased susceptibility to spontaneous tumor development.Akt signaling is involved in increased proliferation and decreased apoptosis (33)(34)(35).Moreover, we have recently observed that the expression of a constitutively active Akt in mouse epidermis leads to spontane-ous tumor formation and heightened susceptibility to chemical carcinogenesis mainly through the increase in proliferative signals (36).This may be responsible for the accelerated appearance of tumors in pRb À/À ;p53 À/À skin.However, we did not observe alterations in proliferation or apoptosis in advanced tumors lacking Rb and Trp53 compared with those observed in mice lacking Trp53 (Fig. 4A; Supplementary Figs.S2 and S3), whereas a significant increase in growth was observed in pRb À/À ;p53 À/À tumors (Fig. 4B), indicating that other processes can be involved.There is broad evidence on the involvement of PTEN/Akt signaling in tumor angiogenesis (37,38), and we have observed that Akt may specifically mediate an angiogenic switch in keratinocytes through transcriptional and posttranscriptional mechanisms (24,39).In support of this, the appearance of pRb À/À ;p53 À/À tumors suggests Figure 3. Hair follicle origin of spontaneous tumors.A, histology and immunofluorescence analysis of poorly differentiated SCCs from a pRb À/À ;p53 À/À (Rb F19/F19 ;Trp53F2 /F10 ; K14cre ) mouse showing the sustained expression of K5 and reduced expression of K10, K6, and K17 (colors in tabs denote the specific keratin).B, H&Estained section showing a representative example of s.c.tumor mass arising from the hair follicle in a pRb À/À ;p53 À/À (Rb F19/F19 ;Trp53 F2/F10 ;K14cre ) mouse.The expression of K5 (red ) and K15 (green ) and K17 (red) and K10 (green ) supports the possible follicular origin of these lesions.Bar, 100 Am. increased vascularity (data not shown).Consequently, we monitored the possible changes in angiogenesis in tumors.
We analyzed the tumor-associated vascularization in frozen sections of the tumors, which were stained for the endothelial junction molecule CD31 (40).In pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À tumor samples (Fig. 6A), the vessels seemed similar to those observed in poorly differentiated carcinomas (40).However, an increased number of vessels were observed in doubly deficient tumors.To verify this observation in a quantitative manner, computer-assisted morphometric image analysis was performed in the CD31-stained sections.This revealed that both vessel density and the relative area occupied by tumor blood vessels were increased in pRb À/À ;p53 À/À tumors compared with pRb +/+ ;p53 À/À samples (Fig. 6B).In support of this augmented angiogenic switch, we also detected that the number of mature vessels, coexpressing CD31 and the pericyte marker aSma (Fig. 6A), was reduced in double-deficient tumors compared with pRb +/+ ;p53 À/À samples (Fig. 6B).Finally, the altered angiogenesis was further supported by microarray analysis, which also showed increased expression of multiple angiogenic factors, such as Pdgfa, Pitx2, and Ptgs2 (cyclooxygenase 2), and decreased expression of antiangiogenic molecules, such as Foxo1, Col18a1 (endostatin), and Timp3, in double-deficient newborn skin (Fig. 6C).
Collectively, these results indicate that the increased growth of pRb À/À ;p53 À/À tumors is also supported by increased angiogenesis.

Discussion
SCCs are, together with BCC, the prevalent tumor types of NMSC (2).SCCs are characterized by increased aggressiveness and metastatic properties, but the molecular basis of their development is still poorly understood.This is a relevant issue as the immunosuppressed population of organ-transplanted patients displays severely increased susceptibility to SCC occurrence (2), mediated .Proliferation and growth in spontaneous tumors.A, percentage of BrdUrd incorporation in interfollicular epidermis, nonlesional and dysplastic hair follicles, and tumors at early or advanced stage.Data come from the analysis of at least 10 different samples.Columns, mean; bars, SE.B, median volume (FSE) of the tumors isolated from pRb À/À ;p53 À/À (Rb F19/F19 ;Trp53 F2 /F10 ; K14cre ) and pRb +/+ ;p53 À/À (Rb wt/wt ; Trp53 F2/F10 ;K14cre ) at different ages (months).The values of P obtained after t test of paired samples are provided.in part by cutaneous HPV infection.The relevance of SCC in human tumorigenesis is also highlighted because it is the most frequent type of tumor in head and neck and cervix and also occurs as a minor metaplastic type in breast and lung.All these aspects reinforce the necessity of developing more accurate models of SCC that can be used as preclinical tools.
The tumorigenic properties of HPV have been associated with the expression of E6 and E7 oncoproteins, which target the p53and the pRb-dependent pathways, respectively (2).This has led us to hypothesize that the ablation of p53 and pRb genes would recapitulate SCCs.Previous work has shown that the specific ablation of p53 gene in stratified epithelia confers high susceptibility to skin SCC development (21,22,29).Our present data agree with these observations.On the contrary, the epidermal-specific ablation of Rb gene does not confer susceptibility to spontaneous skin tumor development (17), and on chemical carcinogenesis experiments, the absence of pRb in epidermis produced increased resistance to tumorigenesis but also higher malignant rate of conversion (19).These aspects have been previously associated to the induction of p53, which causes a selective pressure leading to the premature loss of p53 functions (19,20).In support of these findings, we have also observed that the simultaneous absence of pRb and p107 leads to spontaneous tumor formation in part through the abrogation of p53-dependent apoptotic processes (18). 1 These data suggest that the simultaneous absence of p53 and pRb would cooperate in the skin tumor Figure 5. Activation studies of Akt and MAPK/ERK signaling pathways in pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À spontaneous tumors.A, immunohistochemical detection of phosphorylated Akt (Akt-P ), phosphorylated ERK (ERK-P ), cyclin D1 (CycD1 ), and Foxo3a in pRb +/+ ;p53 À/À (Rb wt/wt ;Trp53 F2/F10 ;K14cre ) and pRb À/À ;p53 À/À (Rb F19/F19 ;Trp53 F2/F10 ;K14cre ) spontaneous tumors.Insets, higher magnification of representative fields.Bar, 150 Am.B, biochemical analysis of pooled whole nonlesional skin extracts obtained from mice of the quoted genotypes used to examine the status of Akt, ERK2, EGFR, Errfi1, and Sprouty 2. Actin was used as a loading control.C, heat map graphical representation showing the relative expression of MAPK and Akt pathway-related gene IDs after microarray analysis of newborn skin of mice of the quoted genotypes.
The present data, besides confirming the relevance of p53 as a tumor suppressor in epidermis (41), also show that the simultaneous absence of pRb and p53 accelerates the rate of tumor appearance.The importance of p53 in suppressing mouse tumor development has been previously shown in multiple tissues, including epidermis (42,43).In this last case, it has been shown that depletion of p53 gene, or generation of epidermal-specific knock in, produces spontaneous tumors and increases the susceptibility to carcinogenic treatments (21,22,29).Therefore, we have studied in depth the mechanisms that might confer the increased susceptibility in the absence of Rb and Trp53.

Cancer Research
Cancer Res 2008; 68: (3).February 1, 2008 the observed increased tumor development.Indeed, the tumors were initially detected as s.c.masses, and histology revealed pretumoral alterations in hair follicle structures.It is worth mentioning that the putative epidermal stem cells are located in these structures (44), which could indicate that the absence of these tumor suppressors might have a functional effect on this cell population.A similar observation has been recently shown in the prostate in close relationship with carcinogenesis (45).In support of this suggestion, we found an increased number of cells expressing K15, a putative marker of epidermal stem cells (46) in the early lesions.However, we did not find K15-positive cells in advanced tumors.This may indicate that the K15-positive cells either are lost or change their phenotypic characteristics during tumor progression.
On the other hand, we did not find significant apoptosis in skin or differences in apoptotic rates between pRb +/+ ;p53 À/À and pRb À/À ;p53 À/À tumors.This is in contrast with the increased apoptosis observed in the tumors generated by chemical carcinogenesis in pRb-deficient mice (19) and reinforces the previous suggestion that, on oncogenic stress, the absence of pRb can cause apoptosis in a p53-dependent manner (20). 1  The Akt and MAPK signaling pathways have been previously involved in the genesis of SCC in mouse skin (30,36,39).We thus investigated whether these pathways act differentially in pRb +/+ ; p53 À/À and pRb À/À ;p53 À/À tumors.The data clearly showed an increased Akt activity leading to increased nuclear cyclin D1 and cytoplasmic Foxo3a localization in pRb À/À ;p53 À/À tumors.These data are in close parallelism with the findings obtained in human head and neck SCC and Akt-transformed mouse keratinocytes (39,47).On the other hand, the apparent increase in MAPK was significantly confined to the stromal component of the pRb À/À ;p53 À/À tumors.This would imply that this signaling pathway is not responsible for the increased tumorigenicity, at least in a cell autonomous process, and suggests that several changes occur in the stroma of these tumors.Whether these changes may affect the tumor development remains unknown at present.The absence of active ERK1/2 in nonlesional mouse skin extracts, in contrast with the increased Akt activity, supports the hypothesis that MAPK signaling contributes to the maintenance but not to the generation of the tumors.Notably, such increase in Akt activity seems to be mediated by increased EGFR signaling.This pathway has a dramatic relevance in SCC in human and mice.Our data indicate that, in p53 À/À ;pRb À/À skin, this increased signaling does not proceed through increased EGFR expression but rather down-regulation of negative regulators of this pathway, such as Errfi1.In this regard, it has been shown that the Errfi1 À/À mice develop spontaneous tumors in various organs and are highly susceptible to chemically induced formation of skin tumors (48).
EGFR activation can trigger both MAPK/ERK and Akt activity; however, in the absence of Egfr amplification, activating mutations of EGFR selectively activate Akt but have no effect on ERK signaling (49).Our data completely fit into this aspect.In addition, the ectopic Sprouty 2 expression observed in doubly deficient skin can contribute to this finding.This putative tumor suppressor can be induced by EGFR/MAPK signaling and produces the feedback inhibition of MAPK/ERK (50).
Akt activities in oncogenesis are diverse and have been mainly associated with the inhibition of apoptosis and increased proliferation (35,37).However, as mentioned above, we did not detect decreased apoptosis or increased proliferation in pRb À/À ;p53 À/À compared with pRb À/À ;p53 +/+ tumors at advanced stage.Moreover, in spite of these similar proliferation rates, there is increased tumor growth in pRb À/À ;p53 À/À mice, suggesting an alternative mechanism.In this regard, the PTEN/Akt pathway has also been involved in tumorigenesis through nonautonomous cell mechanisms that include angiogenic switch (37).In support of this, we found increased number of blood vessels, which also display immaturity features, in pRb À/À ;p53 À/À compared with pRb +/+ ;p53 À/À tumors and altered expression of genes involved in angiogenesis.These observations agree with the reported increased angiogenesis in HPV16 E6-and E7-transduced human primary keratinocytes (51) and our previous data showing that increased Akt activity may mediate an increase in skin tumorigenesis through increased angiogenesis (24).
Collectively, we report the generation of a mouse model of SCC through the epidermal ablation of Rb and Trp53 genes.We have also carried out a detailed molecular characterization of the signaling pathways responsible for the increased susceptibility of epidermal tumor development in this model.Functional comparative studies for the validation of the model as a preclinical tool are currently being performed.These will clearly open new possibilities of pharmacologic intervention for the treatment of these tumors.