Induction of KLF4 in basal keratinocytes blocks the proliferation-differentiation switch and initiates squamous epithelial dysplasia

Induction of KLF4 in mouse skin

As transgenic animals carrying constitutive keratin promoters exhibited lethality, we used alternate inducible and constitutive strategies to direct KLF4 to basal cells. KLF4 was linked to a tetracycline response element (TRE) (Figure 1a), and seven founder lines were crossed to mice carrying a keratin 14 (K14)-reverse tetracycline-responsive transcriptional activator (rtTA) transgene (Figure 1b). This X chromosome-linked K14-rtTA was previously shown to direct expression of TRE-linked transgenes to K14-positive cell types (Xie et al., 1999). Induction with doxycycline (dox) yielded a skin phenotype in male animals of two lines, 32831 and 32812. No phenotype was observed in K14-rtTA control animals treated with dox. The results shown below were obtained using 32831 mice, carrying an autosomal TRE-KLF4, although similar results were obtained using 32812 (not shown). A phenotype in TRE-KLF4 females required homozygosity of the K14-rtTA (see below).

To detect human KLF4 transcripts, we utilized the size difference of mouse and human products of reverse transcriptase (RT)-PCR obtained using a pair of conserved primers (Figure 1c, upper panel). Expression of transgene-derived transcripts was comparable to that of endogenous KLF4 at two days (lane 3). By 21 days, a majority of the transcripts were derived from the transgene (lane 5). No product was obtained without addition of RT (lanes 2, 4, and 6), indicating successful removal of genomic DNA from the samples. Detectable transgene expression required both the rtTA (middle panel, lanes 3, 5) and dox (upper panel, lane 1). As for whole skin, Northern analysis of RNA from primary keratinocyte cultures revealed undetectable levels of transgene expression without dox (Figure 1d). Expression was readily detected after 3 hours of induction, and gradually increased until 24 hours.

KLF4 rapidly induces dysplasia and SCC-like skin lesions

Following 2 days of dox treatment, skin morphology was indistinguishable from uninduced animals or control mice (Figure 2a–b). In contrast, at 9 days the skin showed hyperkeratosis, atrophy of the sebaceous glands, cystic degeneration of hair follicles, and hyperplasia (Figure 2c). After 3–4 weeks, we observed crusted skin lesions and moistness and thinning of the pelage of the ventral skin, with hyperplastic, keratinizing epithelium extending into the dermis (Figure 2d). Dysplastic changes included loss of basal cell polarity and delayed maturation of squamous epithelium. By 42 days lesions resembled severe dysplasia or SCC in situ, with hyperchromatic, pleomorphic nuclei and mitotic figures (Figure 2e). Extension of cells into the dermis were similar to superficially invasive SCC, but lacked the invasive property of fully malignant cells and did not efface the adjacent skeletal muscle layer.

Progression of dysplastic lesions observed at 21 days to SCC-like lesions required continued induction. Withdrawal of dox on days 21–42 lead to resolution of the phenotype (Figure 2f). RT-PCR revealed no expression of the transgene following withdrawal (not shown).

KLF4 is predominantly nuclear in mouse skin lesions and human cutaneous SCC

mRNA in situ hybridization analysis detected KLF4 in dysplastic mouse epithelium, but not in adjacent mesenchymal cells, consistent with restriction of transgene expression to K14-positive cell types, as previously reported for this K14-rtTA line (Xie et al., 1999)(Figure 2g–h). KLF4 antibody stained a subset of nuclei in this epithelium, and stained more diffusely within the cytoplasm (Figure 2i). Little or no staining was observed in normal mouse skin (Figure 2j), perhaps because only part of the immunizing peptide is conserved.

We previously showed that KLF4 is upregulated in virtually all cases of HNSCC (Foster et al., 1999; Foster et al., 2000). To determine if KLF4 is expressed in cutaneous cancers, we stained 5 cases of SCC and one case of basal cell carcinoma (BCC). Two of the SCCs exhibited prominent nuclear staining, in contrast to weak staining of adjacent epithelium (Figure 2k–l). Three other SCCs and the BCC showed little or no staining, indicating that KLF4 is differentially regulated in skin tumors.

Dysplastic lesions exhibit similarities with SCC

To molecularly characterize the lesions, we analyzed the cytokeratins K14, K1, K16, and K17, and the proliferation marker PCNA in ventral skin (Figure 3). K14, normally confined to the basal cell layer (Figure 3, No dox), stained basal and many parabasal cells after 2 days of induction. By 9–21 days nearly all the epithelial cells were K14 positive.

Prior to induction, PCNA was prominent in basal cells of the interfollicular skin, but low in parabasal cells (Figure 3, No dox). As for K14, PCNA was rapidly induced in parabasal cells of interfollicular skin (Figure 3, 2 days), and largely mirrored K14 at later timepoints. PCNA persisted in basal and parabasal cells within spheres of epithelial cells deep within the dermis (Figure 3, PCNA, 21 days, asterisks).

K1, a marker of early differentiation in interfollicular cells (Figure 3, No dox), was rapidly induced by KLF4 in outer root sheath keratinocytes of the hair follicle, indicating an alteration of cell fate (Figure 3, 2 days). At day 9, K1 was expressed in the majority of epithelial cells, but was later restricted to more differentiated, suprabasal cells, and was largely negative in cells deeper in the dermis (Figure 3, 21 days, asterisks). Analysis of skin at 21 days using the follicle marker K17 revealed uniform staining of cystic follicles, but not of the dysplastic surface epithelium, consistent with derivation of dysplastic epithelium from both cell types (not shown).

K16 was low prior to induction as expected, but focally positive in basal keratinocytes of interfollicular epithelium by 2 days (Figure 3). At 9 days, expression was uniform in interfollicular cells, and focally positive in follicular cysts. K16 was prominent in suprabasal cells at day 21.

In summary, KLF4 induced outgrowth of dysplastic, squamous epithelial lesions composed of K14-, PCNA-, and K16-positive cells that gradually lost K1, similar to human cutaneous SCC (van der Velden et al., 1997; Horn and Bravo, 1998). Rapid induction of K14 and PCNA in parabasal cells at 2 days is consistent with inhibition of the proliferation-differentiation switch that normally occurs in developing epithelium.

KLF4 induces an apoptotic response

Induction of KLF4 in cultures of bitransgenic, primary keratinoctyes resulted in death of the vast majority of cells by 48 hrs (not shown). To determine whether KLF4 can induce apoptosis in vivo, we analyzed frozen sections by TUNEL. Prior to induction, TUNEL-positive squamous cells were rare (Figure 4). By day 2 parabasal cells exhibited staining. Apoptosis peaked at day 9, when both interfollicular cells and follicular cysts were strongly positive. In contrast, staining of severely dysplastic skin was less frequent (Figure 4, day 38). Thus, KLF4 induces a transitory apoptotic responses that diminishes in association with progression to dysplasia.

p53 deficiency promotes KLF4-induced dysplasia

The established role of p53 in the genesis of SCC suggested that deficiency of this tumor suppressor could promote the KLF4-induced phenotype (Boyle et al., 1993; Kemp et al., 1993; Ziegler et al., 1994; Jonason et al., 1996; Brash and Ponten, 1998; Jonkers et al., 2001). We identified settings in which p53 deficiency was important or essential for KLF4-induced dysplasia (Figures 5 and 6). Unlike p53 wild-type males (p53^+/+), in which the phenotype was largely confined to the ventral skin (see above), induction of KLF4 in p53 hemizygous knockout animals (p53^+/−) induced pronounced gross and microscopic changes of both the dorsal and ventral skin (not shown). Whereas the dorsal skin of p53^+/+ males exhibited only focal, mild dysplasia (Figure 5a, left panel), p53^+/− and p53^−/− animals similarly exhibited severe dysplasia in association with infiltration of the dermis by a well-vascularized, cellular and fibrotic stroma, or fibrovascular response (FVR) (Figure 5a, middle and right panels). No dysplasia was observed in p53-deficient mice in the absence of KLF4 induction.

The above studies utilized males, in which our expression strategy is expected to yield relatively uniform induction. However, human SCCs are clonally derived from small clusters of cells (Brash and Ponten, 1998; Mao et al., 2004). To examine whether induction of KLF4 in a mosaic pattern could lead to dysplasia, we examined females, in which random inactivation of the X chromosome leads to expression of X-linked, hemizygous alleles in patches measuring less than 1.0 mm (Deamant and Iannaccone, 1987; Ng et al., 1990; Gao et al., 2002). In K14-rtTA_X^+/−;TRE-KLF4^+/−;p53^+/+ females, mosaic KLF4 was insufficient to induce any gross phenotype. By microscopy there were only focal changes including cystic follicles and mild changes in the adjacent interfollicular epithelium (e.g., Figure 5b, left panel). However, the dysplastic phenotype was consistently observed in females that were either p53^+/− (Figure 5b, middle panel) or K14-rtTA_X^+/+ (Figure 5b, right panel). These results suggest that KLF4 must be expressed in a sufficient patch size to induce dysplasia, and that a smaller patch size is sufficient in p53-deficient skin.

Constitutive KLF4 induces dysplasia in a p53-dependent fashion

To determine the consequence of longer term expression of KLF4, we utilized the mouse mammary tumor virus promoter (MMTV) promoter, which is active in skin (Hennighausen et al., 1995; Jonkers et al., 2001). As no phenotype was observed by 18 months of age, we crossed the p53 knockout allele into each of four lines. By 8 months of age, MMTV-KLF4;p53^+/− animals of two lines developed dorsal skin lesions similar to those observed in the inducible model (above), with dysplastic skin overlying a vascular, fibrotic dermis (100% of 15 animals; Figure 6a). Ten p53^+/− littermate controls showed no abnormality by 8 months of age (Fisher’s exact test, P<0.001). Likewise, MMTV-KLF4;p53^+/− animals at 3–6 months of age exhibited no phenotype (Figure 6b). Thus, progression from grossly normal skin to dysplasia occurred between 6 and 8 months of age in MMTV-KLF4;p53^+/− animals.

The life-span of these animals was limited to 8 mo. due to the outgrowth of sarcoma within the dysplasia-associated FVR (100% of 15 animals) (not shown). These 1.0–1.5 cm diameter tumors were composed of spindled, anaplastic cells and scattered giant cells. KLF4, p63, or cytokeratin antibodies did not stain sarcoma cells, but did stain the overlying dyplastic epithelium (Figure 6c–f). The FVR may lead to sarcoma when additional defects are acquired in the responding p53^+/− mesenchymal cells, consistent with the known predisposition of p53^−/− animals to this neoplasm.

Three MMTV-KLF4;p53^+/+ animals that were >18 months of age developed severe dysplasia of the dorsal skin or ears, with delayed maturation, nuclear pleomorphism and hyperchromicity (Figure 6g–h). Although sarcoma was not observed, a FVR included blood vessels intercalated between basal and parabasal epithelial cells (Figure 6h). The skin phenotypes and the associated mesenchymal changes observed in KLF4 transgenic mice are summarized in Table 1.

Localization to the nucleus correlates with transforming activity

Although KLF4 can function in the nucleus to regulate the activity of several cellular or viral promoters (Jenkins et al., 1998; Zhang et al., 1998; Mahatan et al., 1999; Shie et al., 2000; Zhang et al., 2000; Hinnebusch et al., 2004), mechanistic insight into its role as an oncogene is lacking. To identify molecular changes following acute activation of KLF4, we generated a 4-OHT conditional allele by fusion of KLF4 to a portion of the estrogen receptor (KLF4-ER) (Littlewood et al., 1995). Upon addition of 4-OHT to culture media, the fusion protein underwent rapid translocation to the nucleus (Figure 7a). To test for transforming activity, we used a volume of KLF4-ER retroviral supernatant sufficient to transduce only a small fraction of cells (i.e., <1.0%), as indicated by resistance to puromycin. When 4-OHT was included in the growth media, KLF4-ER induced transformed foci on a background monolayer of wild-type RK3E cells (Figure 7b). The spindled, refractile appearance of cells composing these foci were typical of KLF4-transformed cells (Foster et al., 1999; Pandya et al., 2004). Few or no foci were observed in the absence of 4-OHT, indicating the strictly conditional nature of the allele.

KLF4 expression constructs

To generate a TRE-regulated expression construct, the protein coding region of the wild-type human KLF4 cDNA was excised from pCTV3K-SCC7-1 using SalI and Asp718 (Foster et al., 1999). The resulting 2.3 kb fragment was adapted with EcoRI sites and ligated to the vector fragment produced by EcoRI digestion of pXP2-TRE-hGFAT (Xie et al., 1999). A 4.4 kb BamHI fragment containing these elements was purified for microinjection.

To generate a constitutive construct under control of MMTV, bases 1–1780 of KLF4 were released from pXP2-TRE-KLF4 with EcoRI and then ligated to the vector fragment produced by EcoRI digestion of MMTV-pEV-TGFα (Matsui et al., 1990). A 4.7 kb XhoI fragment was purified for microinjection.

To generate a 4-OHT dependent allele, the protein-coding region of KLF4 was amplified by PCR, incorporating a myc epitope tag at the N-terminus and BamHI sites at the fragment ends. When cloned into pBluescript, the expected sequence was obtained and the insert was ligated to the vector fragment produced by BamHI digestion of pBpuro c-mycER^™ (Littlewood et al., 1995). The final construct, termed pBpuro-KLF4-ER, fuses the KLF4 protein-coding region to an estrogen-independent, 4-OHT-inducible fragment of the estrogen receptor (ER). Retroviral transduction of RK3E cells, selection in puromycin, and 4-OHT induction was performed as described (Louro et al., 2002). For induction of transformed foci, 0.2 ml of retroviral supernatant was added to subconfluent cultures of RK3E as described (Foster et al., 1999). Cells were fed twice weekly with non-selective growth media, containing either 4-OHT or vehicle, for a total period of 2.5 weeks. Cells were fixed and stained overnight with modified Wright’s stain (Sigma).

Generation and analysis of transgenic mice

C57BL/6 (B6) X SJL (J) F2 fertilized ova were treated by microinjection in the Transgenic Animal/Embryonic Stem Cell Core Facility at the University of Alabama at Birmingham. Transgenic founders were identified by PCR and confirmed by Southern analysis. PCR primers for genotyping were 5′ GGCAAGTTCGTGCTGAAGGCGT 3′ and 5′ CGATCGTCTTCCCCTCTTTGGCTT 3′, and were used as described (Foster et al., 1999). TRE-KLF4 founders were crossed to B6;J F2 animals, and these lines were subsequently crossed to mice transgenic for X chromosome-linked or autosomal alleles of K14-rtTA (Xie et al., 1999). For the inducible model, the illustrated results were obtained using the TRE-KLF4³²⁸³¹ autosomal allele and an X-linked K14-rtTA (rtTA_X) in the B6;J background of the founders. For the constitutive model, MMTV-KLF4 transgenes and the p53 loss-of-function allele (strain p53N5-T, Taconic) were backcrossed 3 times onto FVB/NJ (Jackson Labs) prior to analysis of skin phenotypes (Donehower et al., 1992). Following fixation and embedding in paraffin, analysis of hematoxylin- and eosin-stained slides was performed by a pathologist (A.R.F.). Unless otherwise noted, the described gross or microscopic results were observed in each of seven or more animals, and with complete penetrance.

To generate TRE-KLF4 mice with the genotypes indicated for K14-rtTA_X and p53, the three crosses used (male X female) were: TRE-KLF4^+/− X rtTA_X^+/+, TRE-KLF4^+/−;rtTA_X^+/Y X rtTA_X^+/+, and TRE-KLF4^+/−;p53^−/− X p53^+/−;rtTA_X^+/+. rtTA_X^+/+ females were first identified by consistent transgene transmission, and were expanded for breeding. Oligonucleotides, cycling parameters, and electrophoretic parameters for genotyping of K14-rtTA and p53 are available upon request.

Dox (Sigma) was administered in 5% sucrose water in amber bottles, and was changed three times per week. To limit toxicity, dox was initiated at 125 μg/ml and then increased at five day intervals to 250 μg/ml or 2.00 mg/ml as necessary to induce a skin phenotype.

mRNA expression studies

Total RNA was isolated from mouse skin or cultured cells (Chomczynski and Sacchi, 1987). RT reactions were performed using total RNA, oligo-dT(12–18) at 50 μg/ml, and SuperScript^™ II RT (Invitrogen). For RT-PCR analysis of mouse skin, cDNA was treated with DNAse I to remove genomic DNA (DNA-free^™, Ambion), and products corresponding to mouse and human KLF4 cDNA were resolved on a 5% polyacrylamide gel. Primer sequences and cycling parameters were as above for genotyping. Northerns were performed as described (Foster et al., 1999). mRNA in situ hybridization analysis of mouse skin cryosections was performed using digoxigenin (dig)-labeled transcripts as described (Louro et al., 2002). Templates for in vitro transcription were isolated from total mRNA of mouse NIH3T3 cells by RT-PCR. PCR products incorporated a T7 RNA polymerase binding site, and forward and reverse primers were: KLF4-AS, 5′ TGGATATGACCCACACTGCCAGA 3′, and 5′ GGATCCTAATACGACTCACTATAGGGAGAAATTCTGGTCTTCCCTCCCCCA 3′; KLF4-S, 5′ GGATCCTAATACGACTCACTATAGGGAGATGGATATGACCCACACTGCCAGA 3′, and 5′ AATTCTGGTCTTCCCTCCCCCA 3′; Hybridized transcripts were detected using enzyme-antibody conjugates and the alkaline phosphatase substrate Fast Red.

Immunodetection assays

Tissues were fixed overnight in 10% neutral buffered formalin and processed for paraffin embedding. Rabbits were immunized using a KLH-conjugated peptide (aa 74–91: CGGSNLAPLPRRETEEFN). Anti-KLF4 was purified by peptide immunoaffinity chromatography (Geneka/Active Motif), and used for paraffin IHC at 13μg/ml. Normal rabbit Immunoglobulin (Dako, X0903) was used a negative control. Anti-KLF4 specifically detected KLF4 by immunoblot, and similar staining of normal and tumor tissue was obtained in side-by-side comparisons to the monoclonal antibody we previously reported (see Supplementary Information) (Foster et al., 2000; Pandya et al., 2004).

For other primary antibodies, we utilized heat-induced epitope retrieval in 10mM sodium citrate (pH=6.0). Anti-p63 mouse monoclonal antibody was used at 1μg/ml (Ab-4, NeoMarkers). Rabbit anti-cytokeratin 1 (CK1) was used at 1 μg/ml (PRB165P, Covance Research Products). Monoclonal antibodies to CK14 and CK16 were used at 1:20 (v/v; NCL-LL002 and NCL-CK16, Novocastra Laboratories Ltd). CK17 polyclonal antibody, a gift from Pierre Coulombe, was used at 1:2000 (McGowan and Coulombe, 1998). Pan-cytokeratin monoclonal antibody was used at 2 μg/ml (AE1/AE3, NeoMarkers). Immunodetection was performed using a biotinylated secondary antibody, streptavidin-horseradish peroxidase (Signet Pathology Systems), and the substrate diaminobenzidine (BioGenex). Sections were counterstained with Harris’ hematoxylin (Surgipath). PCNA was detected using a biotinylated monoclonal antibody (clone PC10) and streptavidin-peroxidase (Zymed).

Immunofluorescence analysis of KLF4-ER was performed using the aminoterminal myc epitope antibody 9E10 (Sigma) diluted 1:100 in 50% goat serum (Pandya et al., 2004).

In situ apoptosis detection (TUNEL)

Frozen sections were fixed in PBS containing 1% paraformaldehyde for 10 minutes at room temperature, and then in ethanol/acetic acid (2:1 vol/vol) for 5 minutes at −20°C. Dig-labeled nucleotides were added to DNA ends with terminal deoxynucleotidyl transferase (Chemicon). Sections were incubated in blocking buffer for 20 min. at room temperature (blocking buffer is 5.0 % goat serum (vol/vol; Sigma), 10 mg/ml casein (ICN), 100 mM Tris-HCl (pH 7.5), 150 mM NaCl). Following incubation with alkaline phosphatase-conjugated antibody (Dako D5105; diluted 1:60 in blocking buffer), sections were developed with the chromogenic substrate Fast Red (Sigma), counterstained with Mayers Hematoxylin (Dako), and mounted using Supermount (Biogenex).