Fos cooperation with PTEN loss elicits keratoacanthoma not carcinoma, owing to p53/p21 WAF-induced differentiation triggered by GSK3beta inactivation and reduced AKT activity.

To investigate gene synergism in multistage skin carcinogenesis, the RU486-inducible cre/lox system was employed to ablate Pten function (K14.cre/Δ5Ptenflx) in mouse epidermis expressing activated Fos (HK1.Fos). RU486-treated HK1.Fos/Δ5Ptenflx mice exhibited hyperplasia, hyperkeratosis and tumours that progressed to highly differentiated keratoacanthomas, rather than to carcinomas, owing to re-expression of high p53 and p21WAF levels. Despite elevated MAP kinase activity, cyclin D1 and cyclin E2 overexpression, and increased AKT activity that produced areas of highly proliferative papillomatous keratinocytes, increasing levels of GSK3β inactivation induced a novel p53/p21WAF expression profile, which subsequently halted proliferation and accelerated differentiation to give the hallmark keratosis of keratoacanthomas. A pivotal facet to this GSK3β-triggered mechanism centred on increasing p53 expression in basal layer keratinocytes. This increase in expression reduced activated AKT expression and released inhibition of p21WAF, which accelerated keratinocyte differentiation, as indicated by unique basal layer expression of differentiation-specific keratin K1 alongside premature filaggrin and loricrin expression. Thus, Fos synergism with Pten loss elicited a benign tumour context where GSK3β-induced p53/p21WAF expression continually switched AKT-associated proliferation into differentiation, preventing further progression. This putative compensatory mechanism required the critical availability of normal p53 and/or p21WAF, otherwise deregulated Fos, Akt and Gsk3β associate with malignant progression.


Introduction
PTEN is a tumour-suppresser gene that has attracted significant interest given its high mutation frequency in human cancers and its roles in apoptosis/proliferation via negative regulation of AKT/PKB activity (Downward, 2004;Parsons, 2004). Consistent with the direct protein-protein interactions that regulate p53 function (Freeman et al., 2003;Lei et al., 2006), Pten mutation in individuals with Cowden Disease results in cancer predisposition (Liaw et al., 1997) associated with cutaneous hyperkeratosis (Fistarol et al., 2002), suggesting that roles in keratinocyte differentiation can be added to PTEN activities that are essential for normal development. In transgenic mice, Pten hetrozygotes (Stambolic et al., 2000) or conditional knockouts (Li et al., 2002;Suzuki et al., 2003) exhibit neoplasia associated with increased anti-apoptotic AKT activities, cell migration/adhesion anomalies (Masahito et al., 1998;Subauste et al., 2005) and cell cycle control failure (Di Cristofano et al., 2001;Weng et al., 2001). In addition, recent models demonstrate that GSK3β, which integrates WNT and β-catenin signalling (Karim et al., 2004), cooperates with PTEN loss in prostate carcinogenesis (Mulholland et al., 2006) when p53 is also compromised (Chen et al., 2005), and bladder cancer models have identified compensatory roles for p21 WAF that counter initial Pten null -mediated hyperplasia (Yoo et al., 2006).
Multistage skin carcinogenesis studies also implicate these molecules. Roles for p53 are well established (Brash, 2006), if sometimes paradoxical (Greenhalgh et al., 1996;Wahl, 2006), as are those for p21 WAF (Topley et al., 1999;Devgan et al., 2006). In classic two-stage DMBA/TPA chemical carcinogenesis, AKT activation and GSK3β inactivation typically correlate with tumour progression (Leis et al., 2002;Segrelles et al., 2006). Furthermore, employing conditional PTEN knockouts, studies showed that DMBA-initiated c-ras Ha (Hras1) activation achieved increased malignancy after TPA promotion (Suzuki et al., 2003). However, two-stage chemical carcinogenesis using heterozygous Pten knockouts identified a mutual exclusivity between Pten loss and c-ras Ha activation (Mao et al., 2004). This was partly resolved on finding that synergism of c-ras Ha with Pten loss (Li et al., 2002) produced benign papillomas, but required TPA for malignant conversion, which involved a separate Pten-mediated mechanism of cell cycle deregulation that superseded initial Pten/ras Ha synergism .
Given that the oncogene Fos is a major effecter of TPA promotion (Schlingemann et al., 2003) and cooperates with c-ras Ha during papillomatogenesis and malignant conversion (Greenhalgh et al., 1990;Greenhalgh et al., 1993a;Greenhalgh et al., 1995;Saez et To investigate gene synergism in multistage skin carcinogenesis, the RU486-inducible cre/lox system was employed to ablate Pten function (K14.cre/Δ5Pten flx ) in mouse epidermis expressing activated Fos (HK1.Fos). RU486-treated HK1.Fos/Δ5Pten flx mice exhibited hyperplasia, hyperkeratosis and tumours that progressed to highly differentiated keratoacanthomas, rather than to carcinomas, owing to re-expression of high p53 and p21 WAF levels. Despite elevated MAP kinase activity, cyclin D1 and cyclin E2 overexpression, and increased AKT activity that produced areas of highly proliferative papillomatous keratinocytes, increasing levels of GSK3β inactivation induced a novel p53/p21 WAF expression profile, which subsequently halted proliferation and accelerated differentiation to give the hallmark keratosis of keratoacanthomas. A pivotal facet to this GSK3β-triggered mechanism centred on increasing p53 expression in basal layer keratinocytes. This increase in expression reduced activated AKT expression and released inhibition of p21 WAF , which accelerated keratinocyte differentiation, as indicated by unique basal layer expression of differentiation-specific keratin K1 alongside premature filaggrin and loricrin expression. Thus, Fos synergism with Pten loss elicited a benign tumour context where GSK3β-induced al., 1995), this study investigated whether activated Fos would cooperate with PTEN loss in papillomatogenesis and drive this cras Ha -independent Δ5Pten-mediated mechanism of malignant progression. Indirect links between Fos and PTEN deregulation already exist, as Δ5Pten could substitute for activated c-ras Ha during TPA promotion , and Fos-mediated photocarcinogenesis associates with both AKT activation and GSK3β inactivation (Gonzales and Bowden, 2002). Furthermore, UVBmediated p53 mutation and subsequent PTEN loss induces AP1 expression (Wang et al., 2005), whereas, in reverse, PTEN specifically targets Fos expression via AKT signalling to downregulate AP1 activity (Koul et al., 2007).
Direct cooperation between activated Fos and inducible Pten loss in adult skin resulted in an unanticipated keratoacanthoma (KA) aetiology, rather than malignant progression to squamous cell carcinoma (SCC). Analysis of the underlying mechanism demonstrated that compensatory p53 and p21 WAF expression prevented progression via switching highly mitotic, papilloma keratinocytes into a programme of accelerated differentiation, manifest by unique, novel basal layer expression of early differentiation-specific keratin K1. This p53/p21 WAF expression profile was apparently induced by progressively increasing levels of GSK3β inactivation. In addition, pivotal roles for AKT were identified, where p53/p21 WAF -mediated reduction of AKT activity in basal layer keratinocytes of benign tumours appeared to be a key facet underlying the switch in progression to KA, not SSC.
The KA outcome of Fos cooperation with PTEN contrasts with induction of malignant conversion in cooperation with c-ras Ha - (Greenhalgh et al., 1990;Greenhalgh et al., 1995;Saez et al., 1995) or TPA-mediated, i.e. Fos-associated (Schlingemann et al., 2003), conversion of Δ5Pten/ras Ha papillomas . This difference may centre on inherent abilities of an epidermis to cope with specific genetic insults, as reflected by the histotypes produced. For instance, the histotype of HK1.fos skin (Fig. 1A) was indistinguishable from normal, despite the fact that HK1.fos expression (supplementary material Fig. S1, lane 5) doubled the mitotic index [below Fig. 3 (Greenhalgh et al., 1993b)]. Hence, prior to wound promotion, this HK1.fos-induced proliferation was possibly counterbalanced by an increase in keratinocyte turnover and/or differentiation, functions regulated in part by endogenous Fos (Angel et al., 2001;Mehic et al., 2005), and by downregulation of AKT activity. Similarly, as observed in cancer-prone individuals with Cowden Disease (Stambolic et al., 2000;Fistarol et al., 2002), treated K14.cre/Δ5Pten flx epidermal histotypes exhibited a relatively mild hyperplasia dominated by hyperkeratosis (Fig. 1B), with blooms of 'ghost' cells indicative of incomplete stratification. This suggests that proliferation in response to PTEN loss was rapidly translated into hyperkeratosis to eliminate potentially neoplastic cells at an early stage, an observation consistent with recent findings on roles for AKT activation during the initial stages of terminal differentiation (Calautti et al., 2005). Furthermore, Δ5Pten expression would compromise PTEN-mediated functions in cellcell adhesion and cell-matrix interactions (Masahito et al., 1998;Subauste et al., 2005) that threaten epidermal barrier function; hence, this hyperkeratotic response may also conscript epidermal homeostasis mechanisms in order to maintain epidermal integrity.
At ~5-6 months, when HK1.fos mice displayed only wounddependent hyperplasia/early papillomatogenesis (Fig. 1C), HK1.fos/Δ5Pten flx mice possessed mature KAs (Fig. 1D). These tumours comprised two distinct histotypes: one of significant differentiation, with 'fronds' of keratinocytes interspaced within massive areas of keratosis; and a second papillomatous area comprising highly proliferative keratinocytes, similar to late-stage aggressive papillomas or possibly carcinoma in situ. Furthermore, whereas keratinocyte differentiation in HK1.fos phenotypes displayed an ordered nature with sequential expansion of each cellular compartment (Fig. 1C), keratotic, but not papillomatous, HK1.fos/Δ5Pten flx KA histotypes displayed a distinctly disordered differentiation pattern ( Fig. 1E-G). Here, cornified and granular cells co-existed alongside proliferative basal cells, culminating in the appearance of micro-cysts (Fig. 1E,G: arrows) and a prominent stratum lucidum (Fig. 1F, arrows) that is indicative of incorrect cornification. This confusion of differentiated and proliferative cell subtypes in each epidermal compartment suggests that HK1.fos/ Δ5Pten flx keratinocytes within the keratotic/differentiated histotype received abruptly conflicting proliferation and differentiation signals.
Premature differentiation marker expression in keratoacanthomas associates with reduced progression marker expression and decreased proliferation Tumours were analysed for expression of keratin K1 (an early-stage differentiation marker), for the late-stage differentiation markers filaggrin and loricrin (both proteins that typically become lost during carcinogenesis), and also for keratin K13 [a simple epithelia keratin employed as a marker of papilloma progression, the expression of which typically becomes uniform prior to malignant conversion (Greenhalgh et al., 1995)]. As observed previously, HK1.fos papillomas exhibit a delay in the onset of K1 expression owing to expansion of the proliferative basal layer compartment ( Fig. 2A, indicated by the K14 keratin counterstain). This result was also observed in papillomatous HK1.fos/Δ5Pten flx KA histotypes (not shown); however, highly differentiated HK1.fos/Δ5Pten flx KA histotypes exhibited novel, K1 expression in the proliferative basal layers. K1 expression was very strong, given the lack of yellow colour from (red) K14 co-expression, although K14 expression itself remained unchanged (see Fig. 3). Typically, AP1-regulated keratin K1 is expressed as differentiating keratinocytes commit to leave the basal layer (Rothnagel et al., 1993) and this result suggests that HK1.fos/Δ5Pten flx keratinocytes accelerated their commitment to differentiation in these keratotic areas.

HK1.fos/Δ5Pten
KAs express high levels of normal p53 whereas control HK1.fos and K14.cre/Δ5Pten phenotypes loose p53 expression Given the close relationship between PTEN and p53 regulation (Freeman et al., 2003;Lei et al., 2006;Wang et al., 2005), p53 status during HK1.fos/Δ5Pten flx KA aetiology was determined by western analysis of normal epidermis, pre-neoplastic phenotypes and tumours taken from separate animals (Fig. 4), or from the same animals ( Fig. 5), to compare KAs with similar keratosis/papilloma ratios and control phenotypes from their age-matched littermates. Hyperkeratotic K14.cre/ Δ5Pten flx epidermis exhibited little detectable p53 expression (Figs 4, 5: HK lanes) compared with normal epidermis (Fig. 4: aN,N,NE lanes). Similarly, hyperplastic HK1.fos epidermis and papillomas also lost p53 expression and p53 levels were undetectable in 'normal' appearing HK1.fos skin (Figs 4, 5: N, PAP, HP lanes). This latter result was consistent with the doubled mitotic index but inconsistent with the normal histotype. On rare occasions, low-level p53 expression was recorded in HK1.fos phenotypes owing to inflammation or presence of anagen follicles ( Fig. 4: aN lane) where HK1.fos was not expressed.
Conversely, in both homozygous and heterozygous Δ5Pten animals, significantly high levels of p53 expression were recorded in HK1.fos/Δ5Pten KAs (Figs 4, 5: KA lanes). Expression levels varied among randomly selected KAs ( Fig. 4) but were usually high, and p53 expression increased with KA maturity/size ( , suggesting that p53 expression was an early feedback response to HK1.fos synergism with Pten loss. This high p53 expression in benign tumours was consistent with the reduced BrdU labelling in keratotic differentiated KA histotypes compared with high labelling indices of papillomatous regions. As indicated by decreased K13 tumour marker expression, elevated p53 would inhibit further tumour progression. This idea was supported by sequence analysis of p53 cDNAs from HK1.fos/Δ5Pten flx KAs (n=5), which found full-length transcripts without detectable mutation or alternate splicing (not shown); hence, normal p53 tumour suppressor functions appeared intact (Nister et al., 2005). HK1.fos/Δ5Pten KAs also lacked spontaneous c-ras Ha activation (Corominas et al., 1989;Greenhalgh et al., 1990;Greenhalgh et al., 1995;Lieu et al., 1991) (n=5; not shown). Thus, high expression of normal p53 in KA aetiology may be rendered impotent by c-ras Ha activation, which leads to SCC, an idea currently under investigation in triple HK1.ras/Fos/Δ5Pten flx mice.

Regulation of AKT activation is a pivotal target of tumour progression and epidermal homeostasis
Consistent with loss of PTEN phosphatase function following ablation of exon 5 (Parsons, 2004), levels of activated AKT ser473 phosphorylation (P-AKT) rose in RU486-treated K14.cre/Δ5Pten flx epidermis ( Fig. 4: HK lanes). HK1.fos/Δ5Pten flx KAs also exhibited increased P-AKT expression (Figs 4, 5); however, levels were not as high as expected and, compared with total AKT expression levels, P-AKT expression varied significantly with the degree of keratosis and hyperproliferation (Fig. 4, KA*) or with KA size/maturity (Fig. 4, lanes 8898 and 9593). Analysis of histologymatched KAs (Fig. 5) found only moderate increases in P-AKT expression compared with hyperplastic epidermis taken from the same animal. Moreover, P-AKT levels in pre-neoplastic HK1.fos/Δ5Pten flx epidermis were consistently lower than in agematched K14.cre/Δ5Pten flx littermate epidermis (Fig. 5, K14.cre/ Δ5Pten flx HK lanes 1, 2; versus HK1.fos/Δ5Pten flx HK lanes 5, 6 and 9). This suggests that P-AKT inhibition was a target of the early low-level p53 feedback response, consistent with Pten null prostate carcinogenesis, where NKX3.1 inhibits P-AKT to stabilise p53 expression (Lei et al., 2006). The fact that the P-AKT expression increase in HK1.fos/Δ5Pten flx KAs was lower than that of comparable c-ras Ha /Δ5Pten synergism  was also consistent with inhibition of AKT by high p53 levels. However, this moderate P-AKT expression profile masked a significant expression level in HK1.fos/Δ5Pten flx papillomatous areas, as detected by immunohistochemical analysis (below, Fig.  6; supplementary material Fig. S2), suggesting that AKT played significant roles in papillomatogenesis and continuation of this activity was essential for further malignant progression (Segrelles et al., 2006;Yao et al., 2006).
An earlier role for AKT regulation was identified in HK1.fos 'normal'-appearing or hyperplastic epidermis, which exhibited little P-AKT expression compared with total AKT expression levels, and levels remained relatively low until overt papillomas appeared (Fig. 4 HK1.fos, lanes N, HP, P; Fig. 5, lanes N, P). Thus, P-AKT downregulation may be an element of the epidermal resistance to early carcinogenesis. This observation could explain the delay in papilloma appearance and the longstanding puzzle that a p53negative HK1.fos epidermis exhibits a normal histotype, despite a mitotic index that produces hyperplasia/hyperkeratosis in K14.cre/Δ5Pten flx skin (Figs 1, 3). Given the direct links between Fos and PTEN (this study) (Koul et al., 2007;Wang et al., 2005), coupled to the intimate interactions between p53 and PTEN (Freeman et al., 2003), HK1.fos-mediated p53 loss may be countered, in part, by a PTEN-mediated feedback involving P-AKT Journal of Cell Science 121 (10) downregulation, which facilitates keratinocyte turnover and differentiation (Angel et al., 2001;Calautti et al., 2005). This is currently under investigation. Hence, HK1.fos phenotypes required a wound-promotion stimulus, eliciting high P-ERK1/2 and increased cyclin D1 and cyclin E2 expression (below), to antagonise/interdict such putative countermeasures and restore P-AKT expression in HK1.fos papillomas (Figs 4, 5, lane PAP). Adding further complexity to AKT oncogenicity, in p53-negative K14.cre/Δ5Pten flx epidermis, where AKT would be released from PTEN control, elevated P-AKT expression (Fig. 4, HK lanes; Fig. 5, HK lanes 1 and 2) was accompanied by a rapid translation of hyperplasia into hyperkeratosis (Fig. 1B), as observed in Cowden Disease, but no papillomas [unless promoted by TPA ], demonstrating that AKT regulation in keratinocyte differentiation can dictate differing outcomes depending on the context(s) of gene expression.
Inactivation of GSK3β was found to be instrumental to the eventual KA outcome, as increased P-GSK3β expression correlated to elevated p53 expression (Figs 4,5). This association was initially unclear, owing to differing keratosis/papilloma ratios (Fig. 4, lanes: KA vs KA*); however, analysis of KAs with similar keratosis/ papilloma ratios consistently demonstrated high levels of inactivated P-GSK3β, concomitant with high p53, but not P-AKT, expression, which remained similar to that in K14.cre/Δ5Pten flx epidermis (Fig.  5, KA versus HK lanes). Hyperplastic HK1.fos/Δ5Pten flx epidermis also possessed moderately elevated P-GSK3β levels, associated with low-level p53 expression, again uncoupled from that of P-AKT, which was downregulated (Fig. 5, lanes HK 5, 6 and 9). The moderate P-GSK3β expression associated with low-level p53 expression in HK1.fos/Δ5Pten flx epidermis, coupled with the major increases in P-GSK3β expression alongside the burst of p53 expression in KAs, suggests that inactivation of GSK3β function triggered p53 re-expression (Ghosh and Altieri, 2005). Furthermore, the burst of p53 and abrupt reduction in keratinocyte proliferation that prevented further progression required a high threshold level of GSK3β inactivation; this may have been achieved from the moderate AKT-independent P-GSK3β expression in HK1.fos/ Δ5Pten flx observed in early preneoplastic hyperplasia (above), coupled with that derived from increasing P-AKT activity in papillomatous areas (Fig. 5, KA lanes).

HK1.fos/Δ5Pten flx KAs exhibit novel p21 WAF expression deregulated cell cycle control and elevated MAP kinase signalling
The mechanism underlying KA aetiology was extended to investigate cell cycle deregulation via western analysis of p21 WAF , cyclin D1 and cyclin E2 expression, together with MAP kinase signalling via analysis of ERK1/2 activation. Analysis of p21 WAF was doubly attractive, as p21 WAF possesses roles in keratinocyte differentiation (Topley et al., 1999) separate to that of cell cycle regulation (Devgan et al., 2006) and can be an early response to PTEN loss (Yoo et al., 2006). All HK1.fos/Δ5Pten flx KAs exhibiting P-GSK3β hyper-inactivation and high p53 expression also exhibited novel high p21 WAF expression levels (Fig. 5, KA lanes). However, unlike HK1.fos/Δ5Pten flx hyperplasia where moderate levels of GSK3β inactivation induced a low level of p53 expression (Figs 4, 5), a high level of GSK3β inactivation was required to induce p21 WAF expression (Fig. 5, lanes HK and P versus KA) and below this threshold level of GSK3β phosphorylation, p21 was not expressed. Furthermore, p53-negative HK1.fos papillomas and K14.cre/Δ5Pten flx phenotypes, which have lower GSK3β inactivation levels, were also negative for p21 WAF expression (Fig.  5). Thus, p21 WAF expression was specific to mature KAs, and the data suggest that p21 WAF expression arose following induction of p53, possibly as a consequence of p53-mediated downregulation of AKT activity (Zhou et al., 2001). Moreover, this temporal p21 WAF expression indicated that the crucial changes in progression occurred at the overt benign tumour stage, a result consistent with the K13/BrdU labelling data and with previous roles for p21 WAF that are associated with inhibition of malignant conversion (Topley et al., 1999).
Analysis of cyclin D1, cyclin E2 and MAP kinase signalling in HK1.fos/Δ5Pten flx KAs (Figs 4,5) was also consistent with the idea that persistent keratinocyte hyperproliferation was continually switched into differentiation. Increasing hyperplasia in .fos/Δ5Pten KAs expressed high but variable P-GSK3β levels, depending on tumour maturity (Δ5Pten heterozygous KA 8898 versus homozygous KA 9593). However, KAs exhibited lower increases in P-AKT expression, which varied extensively with the degree of keratosis (Ka*). Compared with total (t-) protein levels, HK1.fos epidermis exhibited low P-AKT and P-GSK3β expression (first panel), whereas P-GSK3β expression but not that of P-AKT increased in HK1.fos papillomas. Control K14.cre/Δ5Pten flx epidermis possessed elevated P-GSK3β and P-AKT expression (end panel). All hyperplastic phenotypes expressed elevated P-ERK1 and P-ERK2, including 'normal' HK1.fos epidermis and HK1.fos papillomas in particular, which remained steady, if slightly reduced, in all KAs. β-Actin served as a loading control.
Immunohistochemical analysis identified P-GSK3β-associated p53/p21 WAF expression and downregulation of P-AKT activity in basal layer keratinocytes To further clarify these molecular interactions, the in situ expression profiles of p53, p21 WAF , P-GSK3β and P-AKT were determined via immunohistochemical analysis of differentiated, transitional and papillomatous HK1.fos/ΔPten flx KA histotypes (Fig. 6, see supplementary material Fig. S2). Analysis of HK1.fos and K14.cre/ΔPten flx control phenotypes are given in supplementary material Fig. S3. In all differentiated KA histotypes, p53 was strongly expressed throughout each epidermal compartment, including proliferative basal layer keratinocytes (Fig. 6A). In transitional areas, initially p53 expression was low and predominantly suprabasal, but expression became increasingly stronger and appeared in the basal layer ( Fig. 6B; supplementary material Fig. S2). Conversely, papillomatous areas possessed little detectable p53 protein ( Fig. 6C; supplementary material Fig. S2). However, low-level, suprabasal/granular p53 expression was observed in hyperplastic HK1.fos/ΔPten flx epidermis and occasional papillomatous areas; both associated with elevated suprabasal expression of P-GSK3β (not shown). Differentiated KA histotypes exhibited strong p21 WAF expression in all layers ( Fig. 6D; supplementary material Fig. S2). Again, this began in transitional histotypes with a low-level suprabasal and cytoplasmic p21 WAF expression profile, until elevated expression appeared in the nuclei of basal cells associated with increased differentiation (Fig. 6E), prior to becoming strong and uniform in all compartments. This expression profile appeared to trail the wave of high p53 expression (supplementary material Fig. S2), as all papillomatous KA histotypes always lacked detectable p21 WAF expression, even if low levels of p53 were detectable ( Fig. 6F; supplementary material Fig. S2), and Journal of Cell Science 121 (10) p21 WAF was undetectable in hyperplastic HK1.fos/ΔPten flx epidermis (not shown) or HK1.fos and K14.cre/ΔPten flx control phenotypes (supplementary material Fig. S3). Given the roles for p21 WAF in epidermal differentiation (Topley et al., 1999), this basal layer expression of p21 WAF would be consistent with the premature commitment of HK1.fos/ΔPten flx keratinocytes to terminal differentiation, as indicated by novel basal layer K1 expression (above). The confused atypical nature of epidermal differentiation, however, may be due to continued, p21 WAF expression in the suprabasal/granular layers when normally p21 expression shuts down (Devgan et al., 2006).

fos/Δ5Pten
KAs exhibit high p53 and novel p21 WAF expression associated with a threshold level of GSK3β inactivation. Western analysis of p53, total (t-) and phosphorylated (p-) GSK3β and AKT were compared with p21 WAF , cyclin D1 and cyclin E2 expression in pathology-matched KAs (similar keratosis/papilloma ratios) and age-matched preneoplastic phenotypes. Hyperkeratotic (HK) K14.cre/Δ5Pten flx ear epidermis displayed little detectable p21 WAF or p53, and slightly increased expression of P-AKT, cyclin D1 and cyclin E2, with P-GSK3β being higher in the tagged (T) wound-promoted biopsy. Normal (N) appearing HK1.fos epidermis was negative for p21 WAF and p53 expression, with low P-GSK3β and decreased P-AKT levels, alongside slightly elevated cyclin D1. HK1.fos papillomas (PAP) expressed little p21 WAF and p53, but displayed increased P-GSK3β expression compared with P-AKT, together with elevated cyclins. HK1.fos/Δ5Pten flx epidermis (HK) expressed barely detectable p21 WAF , limited p53 and moderate P-GSK3β expression, whereas P-AKT expression was less than K14.cre/Δ5Pten flx controls. All HK1.fos/Δ5Pten flx KAs expressed high levels of p21 WAF and p53 that mirrored significant increases in P-GSK3β inactivation. However, P-AKT expression remained similar to K14.cre/Δ5Pten flx epidermis. All KAs exhibited elevated cyclin D1 and cyclin E2 expression, particularly in ear-tagged samples (KA T ). β-Actin served as a loading control. of hyperproliferation; however, this was insufficient to induce p21 WAF . Thus, increasingly high and basal layer expression of P-GSK3β in the transitional areas induced a corresponding increase in basal layer expression of, first, p53, which halts proliferation, and, later, p21 WAF , which increases differentiation rate.
Analysis of P-AKT in HK1.fos/ΔPten flx KAs (Fig. 6J-L) demonstrated a reverse of these expression profiles, as differentiated or transitional p53/p21 WAF -positive areas expressed decreasing levels of P-AKT (Fig. 6J,K). Conversely, p53/p21 WAFnegative papillomatous histotypes exhibited high P-AKT expression levels (Fig. 6L), a result masked in western analysis, as P-AKT expression faded with increasing differentiation and KA maturity ( Fig. 6J; supplementary material Fig. S2). Moreover, P-AKT expression consistently appeared in the basal layers of papillomatous areas (Fig. 6L), suggesting that AKT activity helped provide a continuous supply of hyperproliferative keratinocytes, hence the lack of KA regression. In increasingly p53/p21 WAF -positive transitional areas, however, P-AKT expression became suprabasal (Fig. 6K) following the appearance of high basal layer p53 expression (Fig. 6B) that culminated in reduced suprabasal P-AKT expression in differentiated histotypes (Fig. 6J). Analysis of consecutive sections found that co-expression of p21 WAF and P-AKT appeared particularly antagonistic, with high P-AKT expression being almost mutually exclusive to that of p21 WAF (Fig. 6E,K). In composite micrographs, a uniform increase in p21 WAF expression paralleled downregulation of P-AKT (supplementary material Fig. S2). Collectively, it is possible that p53mediated reduced P-AKT expression in basal keratinocytes is instrumental to releasing p21 WAF activity (Zhou et al., 2001) and to the commitment to premature differentiation (Topley et al., 1999).

Discussion
HK1.fos/Δ5Pten flx mice demonstrated direct cooperation between inducible PTEN loss and activated FOS expression, which resulted in preneoplastic hyperplasia/hyperkeratosis and a rapid development of overt benign tumours that progressed to KA not SCC. Importantly, this study found that in the context of HK1.fos/Δ5Pten flx benign tumours, significant reexpression of p53 and p21 WAF , previously lost in control phenotypes, now inhibited further malignant progression. This compensatory p53/p21 WAF expression profile was triggered by increasing levels of GSK3β inactivation (Ghosh and Altieri, 2005), which inhibited P-AKT activity in basal layer keratinocytes (Lei et al., 2006) to reduce proliferation, as indicated by BrdU labelling, and initiate p21 WAF -mediated differentiation (Devgan et al., 2006;Topley et al., 1999;Zhou et al., 2001) that is associated with novel basal layer expression of keratin K1, and with premature loricrin and filaggrin expression. This potential sentinel mechanism, which is deployed at the benign tumour stage and crucially dependent on normal p53 and p21 WAF functions, was able to block malignant progression by continually switching keratinocyte hyperproliferation into differentiation, resulting in the hallmark keratosis of KA.
This outcome of KA rather than SCC, was in sharp contrast to the high frequency of TPA-promoted [i.e. Fos-associated (Schlingemann et al., 2003)] carcinomas observed in Δ5Pten flx / c-ras Ha mice  or in c-ras Ha -activated DMBA/TPA carcinogenesis studies involving PTEN knockouts (Mao et al., 2004;Suzuki et al., 2003). Nonetheless, early HK1.fos/Δ5Pten flx synergism was consistent with promotion roles assigned to Fos (Greenhalgh et al., 1993a;Greenhalgh et al., 1993b;Saez et al., 1995) and the fact that PTEN loss could act as a weak initiator for TPA promotion . Indeed, although rapid, papillomatogenesis presented few surprises, as HK1.fos-Δ5Pten flx apparently substituted for c-ras Ha activation observed in earlier KA studies (Corominas et al., 1989), exhibiting moderate elevation in MAP kinase signalling (Parsons, 2004;Downward, 2004;Karin, 1995) and overexpression of cyclin D1 (Bamberger et al., 2001;Burnworth et al., 2006) or cyclin E2 (Di Cristofano et al., 2001;Weng et al., 2001). Incremental increases in keratinocyte proliferation culminated in very high BrdU labelling indices in papillomatous KA histotypes, with typical delays in expression of differentiation markers and the appearance of focal keratin K13 expression, an early marker of tumour progression (Greenhalgh et al., 1995). However, the initial appearance of K13 and high BrdU labelling abruptly diminished in transitional and differentiated KA histotypes, indicating a potent inhibition of proliferation appeared at the benign tumour stage that accelerated terminal differentiation rather than apoptosis, given the premature expression of keratin K1, loricrin and filaggrin. The resulting disorder in keratinocyte differentiation, which was also observed in cyclin D1-transformed HaCaT keratoacanthomas (Burnworth et al., 2006), highlighted a clash between proliferative/oncogenic and compensatory/differentiation pathways. Here, novel basal layer expression of keratin K1 was perhaps a major contributor to the KA outcome, as it indicated a sudden accelerated commitment to differentiation (Rothnagel et al., 1993) and basal layer K1 expression would itself significantly inhibit further tumour progression, as introduction of K1, or its partner K10, into carcinoma cells reverses the malignant phenotype via enforced differentiation (Kartasova et al., 1992;Santos et al., 2002).
Human KA aetiology is also typified by an initial rapid growth phase, followed by arrest and regression. In several respects, murine HK1.fos/Δ5Pten flx KA aetiology mimics that of humans, producing a tumour with a highly proliferative papillomatous/ carcinoma in situ histotype, underlying areas of massive keratosis. However, whether Fos/Pten null synergism drives human KA aetiology remains to be confirmed, although roles for Fos in hyperproliferative disease and keratinocyte differentiation/turnover (Angel et al., 2001;Mehic et al., 2005) and the hyperkeratosis following PTEN loss (Fistarol et al., 2002;Stambolic et al., 2000;Yao et al., 2006) would be consistent with the increased differentiation in KAs. In addition, most human KAs are devoid of p53 mutations and exhibit increased p21 WAF expression (Ahmed et al., 1997;Perez et al., 1997;Ren et al., 1996). These data fuel the debate on whether KA represents a differentiated extreme of SCC or a class of benign tumour in their own right, with a separate molecular aetiology. Given the contrasting results for activated Fos or c-ras Ha synergism with PTEN in KA versus previous SCC aetiology , and the relative lack of typical initiating c-ras Ha or p53 mutations (Ahmed et al., 1997;Lieu et al., 1991;Perez et al., 1997;Ren et al., 1996), these murine data suggest a separate molecular aetiology. However, this idea, again, awaits analysis of whether additional/appropriate mutations of c-ras Ha or p53 interdict a murine KA aetiology mediated by Fos, PTEN and the p53/p21 WAF switch.
Against this background, high levels of p53 expression in basal layers of differentiated KA histotypes was unexpected and identified p53 re-expression as being a key facet underlying a HK1.fos/Δ5Pten flx KA aetiology. Earlier HK1.fos/Δ5Pten flx hyperplasia had exhibited a low-level p53 expression response, associated with moderate GSK3β inactivation, which was subsequently lost (giving rise to the hyperproliferative p53-negative papillomatous KA histotype with elevated MAP kinase/cyclin D1/E2 activities). Hence, when re-expressed in the proliferative basal layers of transitional areas, increasing p53 expression abruptly reduced BrdU labelling and K13 expression, and demonstrated that inhibition of tumour progression depended upon both intensity and locality of gene expression (Wahl, 2006). In human KAs, p53 expression also becomes increased (Perez et al., 1997) and is seldom mutated (Ren et al., 1996), consistent with the lack of p53 or alternate splicing observed in HK1.fos/Δ5Pten flx KAs, which suggested that normal p53 functions were intact (Nister et al., 2005). As with compensatory p53 expression in Pten-mediated prostate carcinogenesis (Chen et al., 2005;Lei et al., 2006), these data predict that a KA aetiology requires fully functional p53 pathways. Hence, permanent loss of p53 in ΔPten/c-ras Ha cooperation or chemical carcinogenesis results in SCC (Suzuki et al., 2003;Mao et al., 2004;Yao et al., 2006). Indeed, the relative rarity of human KA compared with SCC may reflect the high frequency of UV-B-induced p53 mutations (Brash, 2006) that would interdict this putative compensatory mechanism.
High p21 WAF expression levels were observed in mature HK1.fos/Δ5Pten flx KAs but post the appearance of overt benign tumours, as pre-neoplastic or papillomatous histotypes displayed little detectable p21 WAF and these data suggest that p21 WAF inhibited malignant conversion (Topley et al., 1999). Consistent with this idea, western analysis of p21 WAF expression in KAs trailed that of p53 and this may be a consequence of activated AKT expression in papillomatous histotypes (below), as P-AKT inhibits expression, nuclear localisation and function of p21 WAF (Zhou et al., 2001). Logically, therefore, induction of high p53 re-expression would reduce P-AKT expression (Miyauchi et al., 2004) and facilitate p21 WAF escape from P-AKT inhibition. The resultant basal layer expression of p21 WAF would reduce proliferation; however, the roles of p21 WAF in differentiation, separate to that of cell cycle control (Devgan et al., 2006), may be of greater significance. In normal epidermal differentiation, p21 WAF expression increases when postmitotic keratinocytes commit to differentiate (Topley et al., 1999;Devgan et al., 2006), echoing the normal keratin K1 expression profile (Rothnagel et al., 1993) and suggesting that p21 WAF functions in early decisions to commit to terminal differentiation. Therefore, high basal layer p21 WAF expression would accelerate this commitment to differentiate, indicated by basal layer K1 expression, and establish a mechanism that continually inhibited progression via terminal differentiation (Kartasova et al., 1992;Topley et al., 1999;Santos et al., 2002). In addition, as p21 WAF has both positive and negative roles in keratinocyte differentiation, and actually inhibits the latter stages, when p21 WAF is normally downregulated (Devgan et al., 2006). Thus, intense p21 WAF expression in each epidermal compartment may explain the general disorder to keratinocyte differentiation in HK1.fos/Δ5Pten flx KA histotypes, manifest by premature loricrin/filaggrin expression and the appearance of microcysts, a problem further compounded by an increasing lack of P-AKT (Calautti et al., 2005), which would add to the failure to downregulate p21 WAF function (Devgan et al., 2006;Zhou et al., 2001).
Human KAs also exhibit elevated p21 WAF in two distinct patterns: one associated with reduced proliferation and one with increased differentiation (Ahmed et al., 1997). In a tissue continually exposed to environmental carcinogens, the ability to exert resistance to tumour progression at each stage is logical and may involve common components. In classic Ras/Myc cooperation, p21 WAF induction inhibited c-ras Ha -activated skin carcinogenesis in Myc-null cells, until p21 WAF was itself compromised by re-introduction of oncogenic Myc (Oskarsson et al., 2006). Similar compensatory effects of p21 WAF were observed in Δ5Pten-mediated bladder carcinogenesis, where initial hyperplasia was countered by p21 WAF expression (Yoo et al., 2006). However, p21 WAF expression was not induced in Δ5Pten-mediated prostate carcinogenesis (Mulholland et al., 2006), which relied on p53 interactions (Chen et al., 2005), whereas the reduced numbers of DMBA/TPA skin tumours in AKT knockout mice was independent of p53 (Skeen et al., 2006), highlighting multi-layered redundancies in these systems (Wahl, 2006). Perhaps in epithelia concerned with barrier functions, where terminal differentiation to eliminate pre-malignant cells is preferable to widespread apoptosis/senescence, induction of p21 WAF -mediated differentiation (Topley et al., 1999;Devgan et al., 2006;Yoo et al., 2006) provides a necessary adjunct to p53-mediated apoptosis.
Alternately, in K14.cre/Δ5Pten epidermis, elevated P-AKT expression increased differentiation to produce hyperkeratosis (Fistarol et al., 2002;Stambolic et al., 2000), consistent with negative roles in reduction of endothelial cell lifespan (Miyauchi et al., 2004). In normal epidermis, P-AKT expression is mainly suprabasal and in vitro its activities prevent p53-mediated apoptosis. This may provide a protected interval for keratinocytes to fully commit to terminal differentiation (Calautti et al., 2005). In pre-neoplastic K14.cre/Δ5Pten flx epidermis, elevated basal cell P-AKT expression disrupted this balance and increased proliferation owing to concurrent loss of p53/p21 WAF cell-cycle regulation. However, instead of papillomatogenesis, resultant P-AKT-mediated hyperplasia was rapidly translated into hyperkeratosis, suggesting that basal expression of normally suprabasal P-AKT activity induced an early differentiation response. If correct, this elegant mechanism thus serves the dual purpose of rapidly eliminating potentially highly cancerous cells when PTEN tumour suppressor regulation and compensatory p53/p21 WAF -mediated apoptosis are interdicted (Brash, 2006). At the same time, it maintains epidermal tissue integrity and barrier functions under pathological conditions such as Cowden Disease, where Pten functions in adhesion signalling (Masahito et al., 1998;Subauste et al., 2005) are potentially compromised and cutaneous keratinocytes lack normal p21 WAF functions to initiate differentiation.
In HK1.fos/Δ5Pten flx KA aetiology, initial pre-neoplastic HK1.fos/Δ5Pten flx hyperplasia exhibited reduced P-AKT expression, alongside low-level p53 feedback, consistent with PTEN loss in prostate cancer where compensatory NKX3.1 inhibited P-AKT expression to stabilise p53 (Lei et al., 2006). With time, increased MAP kinase signalling, and cyclin D1 and cyclin E2 expression interdicted this early p53 countermeasure, resulting in high P-AKT expression in p53/p21 WAF -negative papillomatous histotypes. As outlined above, subsequently high p53 co-expression fed back to reduce P-AKT activity in basal layers (Lei et al., 2006), inducing increasingly suprabasal P-AKT expression that, in turn, facilitated basal layer expression of p21 WAF (Zhou et al., 2001) and accelerated differentiation. This reduction in proliferative basal layer P-AKT expression appeared crucial to inhibition of benign tumour progression, i.e. unless significant p53/p21 WAF co-expression induced a basal-to-suprabasal P-AKT expression switch to prevent sustained basal layer P-AKT activities, hyperproliferative benign tumour keratinocytes would be at risk for conversion. This is demonstrated by the ability of constitutively active AKT to indict the malignant transformation of DMBA-initiated papilloma keratinocytes (Segrelles et al., 2006), possibly via corruption of the anti-apoptotic AKT roles observed in normal differentiation (Calautti et al., 2005).
As levels of P-GSK3β expression increased, possibly from a combination of moderate P-AKT-independent expression (observed in HK1.fos/Δ5Pten flx hyperplasia) and from increasing P-AKT expression during papillomatogenesis, it achieved a threshold of GSK3β inactivation that triggered the high sustained p53/p21 WAF response. Again, a key component centred on the switch of moderate suprabasal P-GSK3β expression in papillomatous histotypes to one of high basal expression in transitional areas that induced p53, reduced P-AKT and initiated p21 WAF -mediated differentiation (above). This attractive scenario thus explains why induction of high p53, and of p21 WAF in particular, abruptly appeared in benign tumours, as the mechanism required substantial increases in P-GSK3β expression. Temporal GSK3β inactivation thus provided the sensory component of the mechanism geared to induce compensatory p53/p21 WAF responses, and actually required/exploited HK1.fos/P-AKT synergism in papillomatogenesis to increase P-GSK3β levels, which continually blocked further progression. As this GSK3β-associated mechanism of compensatory p53/p21 WAF may also induce apoptosis in alternate tumours (Ghosh and Altieri, 2005;Miyauchi et al., 2004;Yoo et al., 2006), it makes GSK3β inhibitors attractive for therapeutic intervention (Smalley et al., 2007;Tan et al., 2005). However, this should be interpreted with caution, given that GSK3β-inactivated inhibition of skin tumour progression directly contrasts with GSK3β-inactivated cooperation with PTEN loss that accelerates prostate carcinogenesis (Mulholland et al., 2006). Hence, potential efficacy may require tumour aetiologies where intact p53/p21 WAF response pathways (Nister et al., 2005;Wahl, 2006) can be induced (Smalley et al., 2007;Tan et al., 2005), as chemical carcinogenesis (Leis et al., 2002) and alternate models of AKT activation (Segrelles et al., 2006) show that should p53 and/or p21 WAF pathways become compromised, GSK3β inhibition could prove to be a double-edged sword.
In summary, this HK1.fos/Δ5Pten flx model links PTEN-PI3K-AKT signalling, Ras-MAPK-Fos pathways and the GSK3β-βcatenin-WNT axis, and demonstrates that when deregulated by Fos activation and/or Pten loss, benign tumour progression can be inhibited by induction of p53 and/or p21 WAF pathways that limit oncogenic AKT activities. Collectively, these findings highlight the worth of inducible, transgenic models that allow mice to develop normally and thus yield valuable insights into the molecular relationships that regulate normal tissue homeostasis. This carcinogenesis study also stressed the importance of context to the biological outcome of temporal, stage-specific gene expression, where common molecular expression profiles combined to give an unanticipated outcome that provides new insights into the capacity of the epidermis to cope with specific oncogenic insults.