PTEN Loss Promotes ras^Ha-Mediated Papillomatogenesis via Dual Up-Regulation of AKT Activity and Cell Cycle Deregulation but Malignant Conversion Proceeds via PTEN-Associated Pathways

The roles of PTEN tumor suppressor gene signaling have been under intense scrutiny given that loss of function via mutation, genetic silencing, and chromosomal deletion have been identified in human cancers at frequencies that rival p53 (1–3). Germ line PTEN mutations occur in hereditary cancer-associated diseases, such as Cowden syndrome (4), which exhibits cutaneous lesions, suggesting that PTEN may play an important role in keratinocyte differentiation (5). PTEN functions as a lipid and protein phosphatase that targets phosphatidylinositol(3,4,5)triphosphate (PIP3; refs. 6, 7), the product of phosphatidylinositol 3-kinase, to negatively regulate growth factor signaling, including ras (8). This serves to antagonize PIP3-mediated activation of PKB/AKT, a serine-threonine kinase that phosphorylates key intermediate signaling molecules leading to increased proliferation and cell survival (6, 7). Hence, loss of PTEN function results in accumulation of activated AKT that plays a key role in PTEN-mediated tumorigenesis via anti-apoptosis mechanisms (1–3, 9), overexpression of cyclin D (10), and interactions with MDM2 that lead to increased p53 degradation (1–3, 9, 11). Alternate mechanisms of PTEN-mediated tumorigenesis, possibly independent of AKT (12, 13), show that PTEN directly associates with p53 increasing stability, protein levels, and transcriptional activity (14). Similarly, PTEN regulates cell cycle arrest via protein phosphatase-dependent interaction with cyclin D and phospholipase-dependent p27^kip mechanisms (15), consistent with inhibition of S phase via recruitment of p27^kip into cyclin E complexes following adenovirus-mediated delivery of PTEN (16) and cooperation between PTEN loss and p27^kip in prostate cancer (17). The oncogenic potential of PTEN is further highlighted by roles in integrin signaling and an ability to dephosphorylate focal adhesion kinase that reduce cell adhesion and enhance migration (12, 18).

Transgenic studies showed that although knockout of PTEN is embryonically lethal, heterozygous mice develop several tumor types (19, 20). Moreover, in conditional mammary or prostate models, PTEN loss also resulted in developmental defects, highlighting the importance of PTEN to normal differentiation (21, 22). In conditional epidermal knockouts, 7,12-dimethylbenz(a)anthracene (DMBA)/12-O-tetradecanoylphorbol-13-acetate (TPA) chemical carcinogenesis elicited tumors that rapidly progressed to carcinoma (23) via mechanisms centered on failed apoptosis and increased mitogen-activated protein kinase (MAPK) activity, consistent with AKT activation in classic two-stage chemical carcinogenesis (10). However, DMBA/TPA chemical carcinogenesis using heterozygous PTEN knockout mice (24) found that whereas papillomas exhibited ras^Ha activation (25), squamous cell carcinomas failed to exhibit any ras mutations, suggesting that mutually exclusive pathways existed between ras^Ha-mediated and PTEN-mediated carcinogenesis (24).

Thus, to investigate PTEN loss in ras^Ha-mediated skin carcinogenesis, an RU486-inducible keratin K14–driven cre (26) was used to ablate lox-P flanked PTEN exon 5 functions (Δ5PTEN; refs. 14, 21) at localized epidermal sites, including stem cells (26), the presumed targets for carcinogens (27). Previously, using a human keratin K1 promoter, ras^Ha expression in transit amplifying keratinocytes (HK1.ras) resulted in benign papillomas (28), whereas mating experiments to model ras/fos or ras/p53 mechanisms (29–31) resulted in papillomas but without conversion that required additional events (32, 33). In this study, ras^Ha and Δ5PTEN cooperated at the onset of papillomatogenesis but progression via ras^Ha-associated marker mechanisms gave cells with only limited malignant potential, whereas TPA promotion selects a population of Δ5PTEN-associated cells that converted to malignancy with high frequency.

RU486 treatment of transgenic mice. Characterization of HK1.ras (28, 31), inducible K14.creP regulator (26), and PTEN^flx/flx lines (14, 21) have been described previously and breeding strategies maintained HK1.ras and K14.creP transgenes as heterozygotes with wild-type (PTEN^wt/wt), heterozygous (PTEN^wt/flx), or homozygous (PTEN^flx/flx) floxed Δ5PTEN alleles, respectively. To achieve epidermal PTEN ablation, cre recombinase was activated in dorsal skin or ears via topical treatment with 2.5 μg RU486/50 μL ethanol per week (mefipristone) for 3 to 4 weeks (26), with controls receiving ethanol alone. In TPA promotion experiments, 8- to 10-week-old 1276 HK1.ras/K14.cre/PTEN^{wt/wt, wt/flx, flx/flx} or K14.cre/PTEN^{wt/wt, wt/flx, flx/flx} cohorts were treated with 2.5 μg TPA/50 μL acetone per week (50 μL of 1.6 × 10⁻⁴ mol/L TPA) for up to 25 weeks with negative controls receiving acetone alone. Additional TPA controls included RU486-treated HK1.ras/Δ5PTEN or HK1.ras/K14.creP cohorts. HK1.ras 1205 mice were not used given their high sensitivity to TPA (UK Scientific Procedures License PPL 60/2929 to D.A. Greenhalgh).

Genotype analysis and transgene expression. Skin and tumor biopsies were frozen in liquid nitrogen before DNA or total RNA isolation using the Rneasy Midi kit (Qiagen, Sussex, United Kingdom). HK1.ras mice were genotyped as described previously (28) and expression was confirmed via reverse transcription-PCR (RT-PCR). cDNA was generated using SuperScript II reverse transcriptase (Invitrogen, Paisley, United Kingdom) and amplified using HK1.ras primers that spanned the HK1 intron to rule out DNA contamination (forward: 5-GGATCCGATGACAGAATACAAGC-3; reverse 5-ATCGATCAGGACAGCACACTTGCA-3) under conditions described previously (30). K14.creP mice were identified by a 410 bp band using primer pairs forward 5′-TCATTTGGAACGCCCACT-3 and reverse 5′-GATCCCGAATAACTACCTGTTTTG-3′, and reaction conditions of 95°C for 5 minutes, 35 cycles of 45-second denaturation at 95°C, 1-minute annealing at 60.7°C, 1-minute extension at 72°C, and final 15-minute extension at 72°C. PTEN^flx/flx mice were identified by a floxed allele–specific band at ∼1,100 bp and wild-type allele at ∼900 bp (14, 21) using primer pairs P1 forward 5′-ACTCAAGGCAGGGATGAGC-3′ and P2 (exon 5-specific) reverse 5′-GTCATCTTCACTTAGCCATTGG-3′, and reaction conditions of 94°C (2 minutes), 35 cycles of 30-second denaturation at 94°C, 1-minute annealing at 64°C, 1.5-minute extension at 72°C, and 10-minute extension at 72°C. PTEN exon 5 ablation was detected by a nested PCR using identical reaction conditions and a second reverse primer P3, 5′-GCTTGATATCGAATTCCTGCAGC-3′, sited 3′ distal to the downstream lox P site (21), which gave an additional 400 bp P1 to P3 band (Δ5PTEN).

Histology, immunofluorescence, and 5-bromo-4-deoxyuridine labeling analysis. Skin and tumor biopsies were fixed in 10% formalin overnight at 4°C, stored in 70% ethanol, and stained with H&E. Biopsy samples were also frozen in optimum cutting temperature solution and stored at −70°C. For immunofluorescence, frozen sections (5-7 μm) were incubated overnight with rabbit anti-K13 (a gift from D. Roop, Cell Biology, Baylor College of Medicine, Houston, TX), anti-K1, antiloricrin, antifilaggrin (diluted 1:500; Covance/Cambridge Bioscience, Cambridge, United Kingdom), or guinea pig anti-K14 antibodies (dilution 1:2,000; Research Diagnostics, Flanders, NJ) in 12% bovine serum albumin-PBS (v/v), visualized by secondary biotinylated goat anti-guinea pig/streptavidin-Texas red (diluted 1:100; Vector Labs, Burlingame, CA) or FITC-labeled anti-rabbit IgG (diluted 1:100; The Jackson Laboratory, Bar Harbor, ME) antibodies (28–33). For 5-bromo-4-deoxyuridine (BrdUrd) labeling, mice were injected i.p. with 125 mg/kg BrdUrd (Sigma, Dorset, United Kingdom) in sterile 0.9% salt solution 2 hours before biopsy. Deparaffinized sections were subjected to antigen retrieval (boiling for 10 minutes in 10 mmol/L sodium citrate buffer; Lab Vision, Fremont, CA) and immunofluorescence was done by overnight incubation with FITC-conjugated anti-BrdUrd (Becton Dickinson, Oxford, United Kingdom) counterstained with anti-K14. Mitotic index was determined by assaying the number of BrdUrd-labeled nuclei per millimeter of basement membrane (separate counts of three areas per section, three sections per tumor, and five tumors per genotype).

Western blot analysis. Primary transgenic keratinocytes isolated from newborn epidermis (34) were seeded at 5 × 10⁶/60 mm dish and cultured in DMEM/10% chelexed FCS/0.05 mmol/L Ca²⁺ with or without 5 × 10⁻⁹ mol/L RU486 (Sigma) for 7 to 10 days. Whole cell extracts were prepared in duplicate by lysis in 50 to 200 μL buffer [10 mmol/L Na-phosphate (pH 8.0), 2 mmol/L MgCl₂, 1 mmol/L EDTA, 2 mmol/L fresh DTT, and 1:100 protease inhibitor cocktail-Sigma P-8340], mixed with an equal volume of 2× SDS gel loading buffer [100 mmol/L Tris-Cl (pH 6.8), 4% SDS, 0.1% bromophenol blue, 10% glycerol, and 200 mmol/L DTT] and homogenized through a fine gauge needle. Proteins were extracted from biopsy tissue thawed into 200 μL lysis buffer, sonicated for 10 minutes, and centrifuged at 13,000 rpm. Supernatant was transferred into an equal volume of 2× SDS gel loading buffer and homogenized. Protein samples were subjected to 10% SDS-PAGE gel electrophoresis followed by Western blot analysis using primary antibodies to PTEN, total AKT, phosphorylated AKT (phospho-AKT), total extracellular signal-regulated kinase (ERK) p42/44, phospho-ERK p42/44, cyclin D1, cyclin E2 (Cell Signaling Technology, Danvers, MA), and β-actin (Santa Cruz Biotechnology, Santa Cruz, CA) as a loading control. Signals were detected with horseradish peroxidase–conjugated secondary antibodies (DAKO, Glostrup, Denmark) and ECL detection (Amersham Biosciences, Little Chalfont, United Kingdom).

Phenotypes of RU486-treated compound genotypes. To study the effects of PTEN loss in ras^Ha-mediated skin carcinogenesis yet limit PTEN ablation to RU486-treated cutaneous keratinocytes, K14.cre/Δ5PTEN mice were crossed with two lines of HK1.ras mice—line 1276, which exhibits an initiated skin without papillomas (28), and line 1205, which exhibits regression-prone papillomas (31). RU486-treated K14.cre/PTEN^flx/flx cohorts, and, to a lesser extent, K14.cre/PTEN^wt/flx, displayed a thickened skin, mild alopecia, and a prominent hyperkeratosis most noticeably on ears (100% mice, n = 58; Fig. 1A; Table 1). Despite continued hyperplasia/hyperkeratosis by virtue of PTEN ablation in stem cells (26), K14.cre/PTEN^flx/flx did not exhibit tumors for up to 10 months (n = 23). Conversely, all 1276 HK1.ras/K14.cre/PTEN^wt/flx (n = 26) developed novel papillomas at RU486-treated ear tag sites by 8 to 10 weeks and occasional dorsal papillomas (from a scratching/biting wound promotion stimulus), whereas 1276 HK1.ras/K14.cre/PTEN^flx/flx developed bilateral ear and dorsal papillomas by 6 to 8 weeks, apparently independent of wound promotion (n = 26). Such 1276 HK1.ras/Δ5PTEN papillomas were autonomous, suggesting that PTEN loss imparted a constitutive promotion role. This was shown in HK1.ras line 1205 mice, where untreated 1205 HK1.ras/K14.cre/PTEN^wt/flx heterozygotes (n = 15) exhibited papillomas at 10 weeks (Fig. 1B; ref. 31), whereas RU486 treatment accelerated promotion, eliciting larger papillomas in 100% of animals (n = 16; Fig. 1C). Rapid appearance of large papillomas precluded analysis of papilloma progression, which was addressed in 1276 HK1.ras/K14.cre/PTEN^flx/flx progeny. Although sibling K14.cre/PTEN^flx/flx cohorts remained tumor free (Fig. 1D), 1276 HK1.ras/K14.cre/PTEN^flx/flx papillomas exhibited an increasing degree of keratosis (Fig. 1E) instead of progression. Of the eventual 126 HK1.ras/K14.cre/PTEN^flx/flx tumors analyzed from eventually 60 animals, only 6 (∼5%) spontaneously progressed to squamous cell carcinoma (below). RU486-mediated exon 5 ablation was confirmed in vivo by the presence of a 400 bp truncated Δ5PTEN PTEN^flx/flx allele and mRNA analyzed by RT-PCR to confirm HK1.ras expression (Fig. 1F). To confirm PTEN protein loss and identify effects on target molecules, primary keratinocytes were analyzed for expression of total PTEN, phospho-AKT, and total AKT (Fig. 1G). Total PTEN expression levels were lower as Δ5PTEN protein is unstable (14, 21) and loss of phosphatase activity resulted in elevated AKT phosphorylation, also observed in HK1.ras keratinocytes, and increased further in HK1.ras/K14.cre/PTEN^flx/flx keratinocytes.

Histopathology of HK1-Δ5PTEN epidermis and tumors. Untreated K14.cre/PTEN^{−wt/flx −flx/flx} epidermis appeared normal (Fig. 2A) and untreated 1276 HK1.ras/K14.cre/PTEN^{−wt/flx,−flx/flx} controls were identical to parental 1276, exhibiting mild keratinocyte hyperplasia, little hyperkeratosis, and a relative order to keratinocyte differentiation (Fig. 2B). RU486-treated K14.cre/PTEN^flx/flx epidermis also gave hyperplasia but with a prominent, granular layer, a folded (papillomatous) appearance and moderate hyperkeratosis (Fig. 2C). In treated HK1.ras/K14.cre/PTEN^flx/flx cohorts, both hyperplasia and hyperkeratosis increased, giving rise to a characteristic acanthotic, highly tortuous papillomatous histotype (Fig. 2D), with novel squamous cell cysts in the dermis (Fig. 2E), which, alongside sebocyte clusters, probably derive from K14-driven, Δ5PTEN-mediated failed hair follicle differentiation, resulting in alopecia. Papilloma histology in untreated HK1.ras/K14.cre/PTEN^wt/flx was indistinguishable from typical HK1.ras papillomas (28–31), exhibiting an organized, well-differentiated nature, with little dysplasia, moderate dermal infiltrate, and few signs of progression (Fig. 2F). RU486-treated HK1.ras/Δ5PTEN papillomas revealed a similar histotype but with a more dysplastic appearance and increased areas of keratin (Fig. 2G). Eventually, occasional well-differentiated carcinomas appeared (Fig. 2H), histologically indistinguishable from squamous cell carcinomas derived from TPA promotion (compare Fig. 2H with 3F); however, unlike TPA-promoted equivalents, spontaneous squamous cell carcinomas displayed an etiology (i.e., rare/latency >6 months) and differentiation marker profiles identical to all previous HK1.ras cooperation models (32, 33).

TPA promotion of HK1-Δ5PTEN transgenic mice. TPA promotion of untreated 1276 HK1.ras/PTEN^{wt/wt, wt/flx, flx/flx} cohorts was identical to parental HK1.ras 1276 mice (31, 33), giving increased epidermal hyperplasia but an ordered differentiation pattern to the expanded epidermal compartments (Fig. 3A). At prepapilloma stages, TPA-promoted, RU486-treated K14.cre/PTEN^flx/flx cohorts exhibited increased hyperplasia dominated by hyperkeratosis (Fig. 3B). Similarly, TPA promotion of RU486-treated HK1.ras/K14.cre/PTEN^flx/flx cohorts increased epidermal hyperplasia and papillomatosis (Fig. 3C), but at higher magnification (Fig. 3D) the appearance of multiple microcyst-like structures with a greater prominence of the granular layer suggests that the epidermis attempts to compensate for TPA-mediated proliferation following PTEN loss via accelerated differentiation clearly indicated on analysis of atypical differentiation marker expression (Fig. 5) concomitant with appearance of basal cells in postmitotic suprabasal layers (BrdUrd; Supplementary Data).

TPA promotion of RU486-treated K14.cre/PTEN^flx/flx cohorts resulted in novel papillomas in ∼70% of animals following long latency (16-20 weeks). At 24 weeks, papillomas had not progressed to squamous cell carcinoma but did possess the Δ5PTEN-associated, highly keratotic histotype (Fig. 3E) when compared with typical HK1.ras papillomas. Conversely, TPA promotion of HK1.ras/K14.cre/PTEN^flx/flx cohorts exhibited papillomas that progressed to well-differentiated squamous cell carcinoma (Fig. 3F) in most (90%) animals by 12 to 16 weeks, unless culled earlier due to tumor burden. However, there was little difference in histotype between spontaneous and TPA-induced squamous cell carcinomas (Fig. 2F versus 3G) despite a major difference in keratin marker profiles between the two types of squamous cell carcinomas. As spontaneous ras^Ha activation can be a conversion event (33, 34) and was observed in papillomatogenesis following chronic TPA promotion (35, 36), tumors were analyzed for endogenous c-ras^Ha activation [via PCR and XbaI digestion (24) or direct sequencing] without success in TPA-promoted K14.cre/PTEN^flx/flx papillomas (n = 6), spontaneous HK1.ras/K14.cre/PTEN^flx/flx squamous cell carcinoma (n = 4), or TPA-induced squamous cell carcinomas (n = 8; not shown).

Analysis of mitotic activity in HK1.ras/Δ5PTEN epidermis and tumors. The mitotic index was determined via the numbers of BrdUrd-labeled nuclei per millimeter of basement membrane (Table 2; Supplementary Data). Mildly hyperplastic epidermis exhibited a 3- to 4-fold increase in labeling over normal epidermis (3.3 ± 1.6) in HK1.ras (13.8 ± 2.2), K14.cre/PTEN^flx/flx (15.8 ± 3.7), and HK1.ras/K14.cre/PTEN^wt/flx (14.6 ± 2.6) cohorts with an additional doubling in HK1.ras/K14.cre/PTEN^flx/flx (36.6 ± 6.7). In tumors, this trend of increased mitotic index attributable to Δ5PTEN homozygosity continued. Untreated HK1.ras/K14.cre/PTEN^wt/flx papillomas possessed a lower index (39.6 ± 7.5) than RU486-treated littermates (50.1 ± 6.6), whereas HK1.ras/K14.cre/PTEN^flx/flx papillomas exhibited a further doubling (90.3 ± 9.3) to levels observed in HK1.ras/K14.cre/PTEN^flx/flx squamous cell carcinomas (99.1 ± 14.6), including numerous BrdUrd-labeled cells in postmitotic suprabasal layers (Supplementary Fig. S1F versus S1H).

Following TPA promotion, normal skin hyperplasia gave a mitotic index (11.6 ± 3.3) similar to that of K14.cre/PTEN^wt/flx (10.9 ± 2.1), whereas TPA-promoted 1276 HK1.ras epidermis was increased (23.2 ± 2.3). TPA promotion of RU486-treated K14.cre/PTEN^flx/flx skin showed that Δ5PTEN homozygosity in the absence of HK1.ras gave a doubling of mitotic index (42.6 ± 2.3) to levels approaching TPA-promoted, RU486-treated HK1.ras/K14.cre/PTEN^flx/flx epidermis (53.4 ± 8.2). Although consistent with increased cyclins D1 and E2 expression (below), it was noteworthy that K14.cre/PTEN^flx/flx epidermis countered TPA-induced proliferation via increased hyperkeratosis (Fig. 3B). As above, complete loss of PTEN in TPA-promoted HK1.ras/K14.cre/PTEN^flx/flx papillomas doubled the index (98.1 ± 11.9) to levels again approaching squamous cell carcinomas (108.1 + 14.2), although one additional striking facet was the level of labeling recorded in benign K14.cre/PTEN^flx/flx papillomas (69.9 ± 8.3), reminiscent of HK1.p53 gain-of-function papillomas (37).

Expression of tumor progression marker keratins K1 and K13 in HK1.ras/Δ5PTEN carcinogenesis. In all previous HK1 conversion models (28–33), the gradual loss of keratin K1 expression, a marker of early terminal differentiation, and uniform appearance of keratin K13, a marker of simple internal epithelia, was indicative of progression to carcinoma associated with ras^Ha activation as a transgene or via endogenous mutation (32, 33, 37). Consistent with this progression profile, 100% of HK1.ras/K14.cre/PTEN^{wt/wt, wt/flx flx/flx} papillomas exhibited strong K1 expression that became reduced with increasing papilloma aggression (n = 26; Fig. 4A) and lost on spontaneous conversion to squamous cell carcinomas (n = 6; Fig. 4B). Conversely, TPA-promoted, RU486-treated K14.cre/PTEN^flx/flx papillomas (n = 12) exhibited strong K1 expression (Fig. 4C), which lacked a delay in onset of K1 expression, observed in HK1.ras-mediated papillomas, due to proliferative basal layer expansion (Fig. 5; refs. 28–33). Moreover, TPA-promoted HK1.ras/K14.cre/PTEN^flx/flx papillomas (taken early, n = 6) and carcinomas (n = 15) possessed this atypical retention of K1 expression (Fig. 4D).

A novel keratin K13 expression profile was also observed. In spontaneous HK1.ras/K14.cre/PTEN^flx/flx papillomas (Fig. 4E) and initially in TPA-promoted K14.cre/PTEN^flx/flx papillomas (Fig. 4F), K13 was sporadically expressed, increasing in HK1.ras/K14.cre/PTEN^flx/flx papillomas to become uniform in (rare) spontaneous squamous cell carcinomas (Fig. 4G). However, following TPA promotion, all HK1.ras/K14.cre/PTEN^flx/flx carcinomas tested (n = 15) were completely devoid of K13 expression (Fig. 4H). Moreover, during TPA-promoted K14.cre/PTEN^flx/flx papillomatogenesis, the initial sporadic K13 expression pattern, exemplified in HK1.ras-mediated progression, became confined to ever-decreasing areas (Fig. 4F), consistent with the overgrowth of K13-negative keratinocytes. Hence, this novel, unique K1/K13 expression profile would seem to be associated with a TPA/Δ5PTEN-mediated pathway of progression.

Atypical expression of early- and late-stage differentiation markers in HK1.ras/Δ5PTEN epidermis and papillomas. To assess for changes in keratinocyte differentiation, immunofluorescence was done to analyze early-stage keratin K1 and late-stage loricrin and filaggrin expression (Fig. 5). As cited above, expansion of the proliferative basal layer creates an apparent delay in onset of keratin K1 expression (28–33), a result observed in unpromoted HK1.ras/K14.cre/PTEN^flx/flx hyperplasia (Fig. 5A). However, this hallmark of keratinocyte hyperproliferation was absent in TPA-promoted HK1.ras/K14.cre/PTEN^flx/flx epidermis (Fig. 5B) despite high proliferation rates and suprabasal BrdUrd labeling (see Supplementary Data). This subtle effect was associated with TPA promotion, as unlike TPA-promoted K14.cre/PTEN^flx/flx counterparts (Fig. 4C), spontaneous HK1.ras/K14.cre/PTEN^flx/flx papillomas retained the typical delay in onset of K1 expression (Fig. 5C).

Atypical, early loricrin (Fig. 5D-F) and filaggrin (Fig. 5G-I) expression was also observed in TPA-treated epidermis following PTEN loss. Hyperplastic HK1.ras epidermis exhibits loricrin (Fig. 5D) and filaggrin (Fig. 5G) expression restricted to the granular layers. However, following TPA/RU486 treatment, HK1.ras/K14.cre/PTEN^flx/flx epidermis exhibits a diffuse, strong loricrin expression in cells of the spinous and basal compartments (Fig. 5E), whereas filaggrin expression was observed from the spinous layer upward (Fig. 5H) and both were expressed at microcyst sites (Fig. 5E and H). In addition, instead of the usual reduction in expression during HK1.ras-mediated tumor progression (31–33), HK1.ras/K14.cre/PTEN^flx/flx papillomas retained loricrin (Fig. 5F) and filaggrin (Fig. 5I) at sites of high mitotic activity (e.g., Fig. 5F versus Supplementary Fig. S1F). This expression profile was also observed in TPA-promoted K14.cre/PTEN^flx/flx papillomas; however, in carcinomas, loricrin and filaggrin expression were lost (not shown).

Analysis of activated AKT, MAPK signaling, and cyclin expression. Given the atypical K1 and K13 expression data, the relative contributions of ras^Ha or PTEN signaling were determined in tumor progression (Fig. 6). Total AKT and activated phospho-AKT, major targets for PTEN regulation (1–3, 9), were initially analyzed in vitro to confirm functional PTEN exon 5 loss (Fig. 1G). In vivo, RU486 treatment increased total AKT protein in all hyperplastic epidermis and tumors regardless of histotype, etiology, or genotype (lane 1 versus lanes 2-13). Following the sequential loss of each functional PTEN allele, an increase in phospho-AKT expression was observed, increased further by TPA promotion (lanes 2-4). HK1.ras expression also gave a small rise in phospho-AKT levels (lane 5 versus lane 1), whereas HK1.ras/Δ5PTEN synergism in severely hyperplastic epidermis (lane 6) gave high phospho-AKT levels, similar to that of TPA-promoted K14.cre/PTEN^flx/flx skin (lane 4). In overt tumors, HK1.ras papillomas exhibited slightly elevated phospho-AKT levels (lane 7), whereas HK1.ras/Δ5PTEN cooperation increased levels further in both heterozygous (lane 8) and homozygous papillomas (lanes 9 and 10). However, the highest levels of activated phospho-AKT expression were observed in TPA-promoted carcinomas (lanes 11-13).

MAPK signaling, a target of both positive ras (8) and negative PTEN regulation (9), was determined by analysis of total and activated phospho-ERK 1 and 2. Oddly, given an obvious increased proliferation, total ERK 1 and 2 protein expression in hyperplastic K14.cre/PTEN^wt/flx and K14.cre/PTEN^flx/flx was lower than normal epidermis (lanes 1-3), becoming elevated following TPA promotion. Conversely, HK1.ras expression induced high levels of total ERK proteins in hyperplastic epidermis and papillomas (lanes 5-8), regardless of PTEN status, until late-stage papillomas (lanes 9 and 10), where PTEN loss significantly down-regulated expression of total ERK proteins, which then remained surprisingly low in TPA-promoted squamous cell carcinomas (lanes 11-13). Activated phospho-ERK expression presented a more complex picture. In preneoplastic K14.cre/PTEN^flx/flx hyperplasia, phospho-ERK levels were elevated (lane 3) despite the down-regulation of total ERK proteins and TPA promotion increased this activity further (lane 4), consistent with elevation of total ERK proteins. As expected, high levels of phospho-ERK were detected in HK1.ras hyperplasia (lane 5), but following PTEN loss in HK1.ras/K14.cre/PTEN^flx/flx hyperplasia and despite high total ERK protein levels, phospho-ERK levels were then down-regulated (lane 6). Similarly, in HK1.ras papillomas, elevated phospho-ERK levels (lane 7) were sequentially lowered in HK1.ras/K14.cre/PTEN^wt/flx papillomas (lane 8) becoming virtually undetectable in HK1.ras/K14.cre/PTEN^flx/flx papillomas (lanes 9 and 10). However, phospho-ERK levels became elevated again in squamous cell carcinomas following TPA promotion.

The increased mitotic activity in all Δ5PTEN homozygotes (Table 2; Supplementary Data) prompted assessment of cell cycle deregulation via analysis of cyclin E2 (15–17) and cyclin D1 (8, 10, 16). Epidermal expression of cyclin E2 increased following sequential PTEN loss and gave high levels in TPA-promoted K14.cre/PTEN^flx/flx epidermis (lanes 2-4). HK1.ras-induced hyperplasia gave only slightly elevated levels of cyclin E2 (lane 5 versus lane 1) and in hyperplastic HK1.ras/K14.cre/PTEN^flx/flx epidermis, cyclin E2 expression increased to levels similar to, rather than above, the levels observed in K14.cre/PTEN^flx/flx hyperplasia (lane 6 versus lane 3) and lower than TPA-treated K14.cre/PTEN^flx/flx skin (lane 6 versus lane 4). That cyclin E2 levels were associated with Δ5PTEN expression was consistent with the small increase in cyclin E2 expression observed in HK1.ras papillomas (lane 7) compared with high cyclin E2 levels in HK1.ras/Δ5PTEN papillomas (lanes 8-10), which became significantly increased in TPA-treated squamous cell carcinomas (lanes 11-13).

Analysis of cyclin D1 expression gave similar elevated results but with a synergism between HK1.ras- and Δ5PTEN-induced deregulation in early papillomatogenesis. In Δ5PTEN-induced hyperplasia, cyclin D1 increased over normal (lanes 1-3), with a further increase following TPA promotion (lanes 4). Hyperplastic HK1.ras epidermis also expressed elevated cyclin D1, which further increased in HK1.ras/K14.cre/PTEN^flx/flx hyperplasia (lanes 5 and 6). Heterozygous HK1.ras/Δ5PTEN papillomas consistently expressed high levels compared with HK1.ras alone (lanes 7 and 8), approaching that observed in TPA-promoted squamous cell carcinomas (lanes 11-13). However, although still elevated, HK1.ras/K14.cre/PTEN^flx/flx papillomas exhibited a decrease in cyclin D1 expression (lanes 9 and 10). Given the ERK data, it is tempting to speculate that this decrease in cyclin D1 reflects the change in relative contributions provided by HK1.ras versus PTEN loss in progression and that TPA interactions with Δ5PTEN (observed in lane 4) mediate the return to high levels in squamous cell carcinomas.

This transgenic model of inducible PTEN loss and ras^Ha activation identified an early synergism in papillomatogenesis and highlighted important roles for PTEN in maintenance of cutaneous homeostasis, as its loss induced an apparent compensatory keratinocyte differentiation program. Initially, RU486-treated K14.cre/PTEN^flx/flx cohorts did not exhibit spontaneous tumors despite sustained long-term keratinocyte hyperplasia, but exhibited increased epidermal hyperkeratosis consistent with Cowden disease (6). This was in sharp contrast to conditional K5.cre/PTENfloxed mice where papillomas appeared by 3 to 4 months and were converted to carcinoma (23). Logically, these differences center on the inducible versus conditional approach and, as PTEN loss compromises, e.g., p53 guardian functions (1–3, 11, 14), reflect the number (hyperplasia) and type (stem cells/embryonic epidermis) of target keratinocytes susceptible to additional mutations. Unlike K14.cre/PTEN^flx/flx cohorts, effects of PTEN loss in K5.cre/PTENfloxed mice would begin immediately on development of the embryonic epidermis (E 13/14; ref. 38) and continue throughout development on a global rather than localized scale (23). Moreover, K5.cre/PTENfloxed epidermis exhibited precocious hair follicle development during embryogenesis (23), similar to that reported for β-catenin transgenics (39, 40). This identified links between PTEN loss and E-cadherin/β-cat/wnt/Lef1 signaling (39–41) that effectively increased the numbers of PTEN-null bulge region stem cells at risk for additional mutations (27), hence a predisposition to carcinogenesis. Conversely, RU486-treated K14.cre/PTEN^flx/flx cohorts exhibited alopecia, as RU486-treated adult follicles formed cyst-like structures similar to Lef 1 mutants that interdict β-catenin signaling (42), which would decrease stem cell targets. Nonetheless, given K14 promoter-targeting specificity elicits the same Δ5PTEN/p53 mutation susceptibility (14) in long-term keratinocyte hyperplasia (6-10 months), the lack of K14.cre/PTEN^flx/flx papillomas remained unclear.

Clues to this lack of tumors came from analysis of keratinocyte differentiation, which highlighted complex mechanisms sensitive to PTEN loss that countered excess proliferation via the induction of a terminal differentiation program. For instance, following TPA promotion in Δ5PTEN epidermis, the increase in mitotic index, elevation of ERK signaling, AKT activity, and cyclins D1 and E2, was actively countered by a massive degree of hyperkeratosis, hence the latency period in appearance of TPA-promoted K14.cre/PTEN^flx/flx papillomas compared with HK1.ras (28, 31). Analysis of keratin K1, an early marker of keratinocyte differentiation together with late-stage markers filaggrin and loricrin in TPA-promoted HK1.ras/Δ5PTEN epidermis, revealed an early anomalous expression amidst a confusion of granular, spinous, and proliferative suprabasal cells together with epidermal microcysts. Indeed, when this putative protective mechanism was overcome in TPA-promoted K14.cre/PTEN^flx/flx or HK1.ras K14.cre/PTEN^flx/flx papillomas, this differentiation was echoed by lack of a delay in keratin K1 expression and retention, not loss, of loricrin and filaggrin. Given that TPA induces c-fos (43), it is noteworthy that this novel differentiation mechanism was observed in experiments exploring Δ5PTEN synergism with fos activation.³

To maintain cutaneous homeostasis, such a sentinel differentiation program probably evolved to accommodate environmental mutations and yet prevent disruption of the paramount epidermal role in barrier maintenance. Given the links between PTEN and cell-to-cell adhesion signaling (12, 18, 39–41), together with precocious follicle development (23), this epidermal sensitivity to Δ5PTEN may reflect PTEN functions that recruit the necessary molecular scaffold for E-cadherin/β-catenin in maintenance of adherens junctions (44, 45). Failure of E-cadherin-mediated adhesion resulted in a lethal neonatal phenotype of barrier dysfunction (46); thus, if PTEN was critical for E-cadherin function(s), then a default terminal differentiation program would be a logical mechanism to deploy following PTEN loss to counter proliferation and the dangers of a compromised epidermal barrier. Testing of these intriguing ideas await transgenic models of inducible E-cadherin and β-catenin knockout.

In papillomatogenesis, inducible Δ5PTEN expression provided a lesion-eliciting event for 1276 HK1.ras mice and accelerated papilloma appearance in promotion-sensitive HK1.ras 1205 mice, suggesting a constitutive promotion role. Due to loss of Δ5PTEN phosphatase activity in HK1.ras/K14.cre/PTEN^flx/flx epidermis and tumors, an early synergism involved dual increase in active phospho-AKT expression in vitro and in vivo. Although consistent with previous chemical carcinogenesis studies (10, 24) and resistance to UV-induced apoptosis in vitro (23), this occurred without changes in total AKT protein levels (24), as general keratinocyte hyperplasia (or tissue culture) gave a continuous, elevated expression. Interestingly, at these early preneoplastic stages, total ERK 1 and 2 protein expression in RU486-treated K14.cre/PTEN^wt/flx and K14.cre/PTEN^flx/flx hyperplasia was lower than normal epidermis, returning to elevated levels following TPA promotion or HK1.ras expression (Fig. 6, lanes 1-4). Also, despite PTEN being a negative regulatory element to MAPK signaling via down-regulation of AKT (1–3), Δ5PTEN did not increase activated phospho-ERK levels to that of TPA promotion or HK1.ras expression. These data may reflect the increased differentiation associated with Δ5PTEN expression in a feedback loop of decreased MAPK signaling. Thus, overt papillomatogenesis may require ras^Ha activation (or TPA promotion in K14.cre/PTEN^flx/flx cohorts) to reactivate MAPK signaling and escape this in vivo counter to PTEN loss, whereas Δ5PTEN would inhibit the potential apoptotic response to ras^Ha activation (47) via increased antiapoptotic AKT expression. Nonetheless, in progression, all late-stage HK1.ras/Δ5PTEN papillomas consistently exhibited reduced total and phospho-ERK expression (Fig. 6, lanes 8-10), suggesting that once papillomas were established, HK1.ras signaling became less significant and the elevated phospho-ERK levels observed on conversion (lanes 11-13) was due to TPA-mediated synergism with Δ5PTEN (below).

One important mechanism underlying papillomatogenesis was cell cycle deregulation, typified by the strikingly high mitotic indices following Δ5PTEN homozygosity, consistent with down-regulation of p53 (11, 14). One arm involved dual elevation of cyclin D1 by both Δ5PTEN and HK1.ras, which culminated in high cyclin D1 levels in heterozygous HK1.ras/K14.cre/PTEN^wt/flx papillomas. Previous ras^Ha-activated chemical carcinogenesis studies also linked increased AKT activity with elevated cyclin D1 (10, 15) and an early promotion role was supported by increased AKT/cyclin D1 activity in preneoplastic TPA-promoted K14.cre/PTEN^flx/flx epidermis or HK1.ras/K14.cre/PTEN^flx/flx hyperplasia (Fig. 6). However, as with reduced ERK expression, in HK1.ras/K14.cre/PTEN^flx/flx papillomas, cyclin D1 levels were lower, possibly reflecting a reduced HK1.ras/MAPK signaling contribution to this papilloma progression mechanism, subsequently restored to high levels by TPA promotion resulting in a high mitotic index with a propensity to conversion given the compromised p53 guardian functions (11, 14). A second arm of cell cycle–mediated promotion involved failure of PTEN-mediated p27^kip regulation, which resulted in cyclin E2 overexpression (15–17, 22). This promotion arm appeared independent of HK1.ras signaling, as elevation of cyclin E2 in hyperplastic K14.cre/PTEN^wt/flx and K14.cre/PTEN^flx/flx epidermis occurred in parallel to increased phospho-AKT levels but decreased ERK, whereas PTEN null-mediated cyclin E2 elevation gave a 50% increase in BrdUrd labeling regardless of HK1.ras expression. Furthermore, cyclin E2 levels remained relatively low in HK1.ras epidermis and papillomas and, in progression, cyclin E2 expression was maintained at high levels, unlike the transient decrease in cyclin D1 when phospho-ERK was reduced, becoming significantly increased on conversion to malignancy. Thus, one important mechanism underlying progression (or propensity for conversion) involved a PTEN/p27^kip/cyclin E2 pathway of failed cell cycle regulation and deregulated cyclin E2 expression maybe an important key as to why TPA-mediated progression seems to be accelerated via Δ5PTEN pathways (15–17), consistent with reports that PTEN loss and p27^kip mutations induce progression in prostate carcinogenesis (17, 22) and are exploitable prognostic markers (48).

With respect to tumor progression, these HK1 data assumed an intriguing aspect compared with earlier DMBA/TPA chemical carcinogenesis studies in conditional PTEN knockouts that achieved malignancy via mechanisms involving (presumably ras^Ha activated; ref. 25) elevation of MAPK signaling (23) and increased AKT activity, whereas similar experiments using heterozygous knockouts found a distinct lack of MAPK activity and mutual exclusivity between PTEN loss and DMBA-induced ras^Ha activation in carcinomas (24). In HK1.ras/Δ5PTEN mice, two distinct pathways of progression arose. The first spontaneous conversion pathway was less aggressive, relatively rare (within this time frame), and expressed keratins K1 and K13 progression markers in an identical fashion to conversion mediated by HK1.ras or TPA-promoted endogenous c-ras^Ha activation (refs. 31–33, 37 and references therein). The second, dependent on TPA promotion, was rapid, occurred at high frequency, and expressed K1 and K13 in a unique, novel fashion that may be a marker of this conversion pathway. Furthermore, this progression mechanism may involve a greater contribution of Δ5PTEN as opposed to HK1.ras. As outlined above in late-stage papillomas, both total and phospho-ERK, together with cyclin D1, were down-regulated although Δ5PTEN-associated AKT and cyclin E2 were consistently elevated. In addition, TPA promotion of K14.cre/PTEN^flx/flx epidermis did not induce spontaneous c-ras^Ha activation, unlike all previous TPA-promoted HK1 transgenic or SENCAR carcinogenesis models (32–36), which showed that Δ5PTEN can substitute for ras^Ha activation in initiation, albeit as a much weaker initiator given the delay in onset of TPA-promoted K14.cre/PTEN^flx/flx papillomas. Such TPA-promoted K14.cre/PTEN^flx/flx papillomas were not recorded within the time frame tested in K5.cre/PTENfloxed experiments, but Mao et al. (28) reported appearance of heterozygous PTEN knockout papillomas devoid of ras^Ha mutations that arose via loss of the normal PTEN allele (29). Moreover, in TPA-promoted K14.cre/PTEN^flx/flx papillomas, following an initial expression of K13 in a typical HK1.ras-associated sporadic profile (Fig. 4E), a subsequent decrease in keratin K13 was observed consistent with the gradual overgrowth of a separate population of papilloma cells (Fig. 4F). Because TPA-derived squamous cell carcinomas were characterized by this novel absence of K13, TPA-promoted K14.cre/PTEN^flx/flx papillomas may represent an intermediate stage where a possibly ras^Ha-independent TPA/Δ5PTEN mechanism begin to overlay the foundations of earlier HK1.ras/Δ5PTEN synergism and give rise to cells with increased malignant potential.

Hence, HK1.ras/Δ5PTEN data suggest that, in agreement with the heterozygous knockout study, although unable to conclude a mutual exclusivity between ras activation and Δ5PTEN (24), following initial synergism in early papillomatogenesis, TPA-mediated tumor progression seems to involve Δ5PTEN-associated pathways more so than HK1.ras, with a significant degree of cell cycle perturbation underlying progression due to cyclin E2 deregulation in particular. In the K5.cre/PTENfloxed study, the roles assigned to MAPK signaling may be due to in vitro analysis of ERK expression in epidermal growth factor–stimulated keratinocytes (23) versus tissues (this study; ref. 24). Equally, this result may reflect the HK1.ras/Δ5PTEN spontaneous progression pathway, which is less aggressive in the inducible model as progression depends on acquired (unknown) mutations probably derived from PTEN/p53 guardian dysfunction (14), which would be a more significant defect in TPA-promoted cohorts. Given that ras remains a promising target in the design of novel therapies (8), these data may have important implications where PTEN is also compromised, as their action may eliminate a benign population following ras-dependent mechanisms with limited progression capability and select PTENnull-dependent keratinocytes with increased malignant potential.

Grant support: Cancer Research UK (C1361/GA2395; D.A. Greenhalgh).

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.

We thank Profs. Dennis Roop (Baylor College of Medicine, Houston, TX) and Xiao-Jing Wang (Oregon Health and Science University, Portland, OR) for the gift of K14.CreP transgenic mice and helpful advice; Prof. Rona M. MacKie for assistance with histological analysis, and Mr. Graham Chadwick for help with figure preparation.