Expression of Mad, an antagonist of Myc oncoprotein function, in differentiating keratinocytes during tumorigenesis of the skin.

The Myc oncoprotein is associated with cell proliferation and is often down-regulated during cell differentiation. The related Mad transcription factor, which antagonises Myc activity, is highly expressed in epidermal keratinocytes. Mad also inhibits cell proliferation in vitro. To study Mad expression in keratinocyte proliferation and differentiation, we have analysed Mad RNA expression in regenerating and hyperproliferative epidermal lesions and epidermal tumours of varying degrees of differentiation using the RNA in situ hybridisation and RNAase protection techniques. Mad was strongly expressed in differentiating suprabasal keratinocytes in healing dermal wounds and in benign hyperproliferative conditions, but also in squamous cell carcinomas, in which the keratinocytes retain their differentiation potential. However, Mad expression was lost in palisading basal carcinoma cells and poorly differentiated squamous cell carcinomas, which lacked the epithelial differentiation marker syndecan-1. We therefore suggest that Mad expression is closely associated with epithelial cell differentiation, and that this association is retained in epithelial tumours of the skin. ImagesFigure 1Figure 2Figure 3Figure 4Figure 5Figure 6

Epithelial cell proliferation and differentiation is a complex process. The regulation of genes encoding structural proteins, such as keratins, during epithelial cell growth and differentiation is relatively well known (Fuchs, 1993). However, less is known about the roles of specific transcription factors, for example in keratinocyte proliferation and differentiation (Fuchs, 1990). The Myc oncoprotein and transcription factor regulates cell growth and apoptosis (Alitalo et al., 1992;Amati et al., 1993;Cantley et al., 1991;Morgenbesser and Depinho, 1994;Vastrik et al., 1995). Induction of cell differentiation is in general associated with down-regulation of Myc mRNA, although the expression of the myc genes in many cases is compatible with differentiation (Liischer and Eisenman, 1990). Mad is a recently cloned basic region helix-loop -helix -leucine zipper (bHLHZip) transcription factor that competes with Myc for dimerisation with Max, a constitutively expressed bHLHZip protein (Blackwood and Eisenman, 1991;Vaistrik et al., 1993). In contrast to the Myc-Max complex, which transactivates gene expression, the Mad -Max complex suppresses transcription from promoter constructs containing the Myc target sequence CACGTG (Ayer et al., 1993). The relative abundance of Mad and Myc could thus determine the activity of Myc target genes involved in the control of cell proliferation and differentiation . In particular, the expression of Mad may be required for normal differentiation by counterbalancing the growth-promoting effects of Myc. Changes in the differentiation of epidermal keratinocytes and in the maturation of epidermal cell layers occur during wound healing and tumorigenesis in the epidermis. Wound healing includes responses involving increased expression of several growth factors and cytokines, keratinocyte migration, proliferation and modulation of pericellular matrix biosynthesis and deposition. For example, syndecan-1, which represents a family of cell-surface proteoglycans that influence cellular proliferation and differentiatiion is increased during wound healing (Mali et al., 1990), but lost in carcinomas and during the development of severe dysplasia (Inki et al., 1994).
Skin tumour formation in the mouse induced by repeated applications of polycyclic hydrocarbons, such as 9,10dimethylbenz(a)anthracene (DMBA) or UV irradiation can be used as an in vivo model in the study of different stages of neoplastic disease (Stenback, 1978). This tumour model progresses in a series of steps, from the formation of hyperplastic, regressing lesions to dysplasia and papillomas, and ultimately highly malignant squamous cell carcinomas (Yuspa et al., 1991;Bjelogrlic et al., 1994;Stenback, 1978). The neoplastic cell populations are characterised by altered keratin expression, increased expression of proliferating cell nuclear antigen (PCNA), decrease in the basement membrane constituents laminin and collagen IV, and an increase in the p53 oncoprotein in the malignant cells (Yuspa et al., 1991;Bjelogrlic et al., 1994).
We have recently observed that Mad mRNA is strongly expressed in the differentiating epithelia of mouse embryos (Vastrik et al., 1995). Mad signals are abundant in suprabasal differentiating cell layers, but not in the basal proliferating cells of either skin or gut epithelium. This suggests that during re-epithelialisation, Mad expression may be sequentially altered. This selective expression may then be disrupted or completely lost during malignant epithelial progression. In this study we have analysed Mad expression in adult mouse where hyperproliferative and malignant states were induced by full-thickness wounds and by DMBA treatment of the skin, respectively. We have also studied human epidermal tumours such as basal cell and squamous cell carcinomas and melanomas. We have localised Mad mRNA to the upper epidermal cell layers, where the keratinocytes differentiate irreversibly. We observed a similar expression pattern in carcinomas, where the malignant cells retain differentiation capacity, whereas anaplastic tumours consisting of proliferating cells without signs of differentiation are negative for Mad mRNA. Materials and methods Generation of skin tumours in mice NMRI mice were exposed to DMBA, 50 mg twice a week for 10 weeks and then sacrificed at the termination of the study. Animal treatment, maintenance and conduction of the study followed standard protocols (Bremner et al., 1994;Bjelogrlic et al., 1994;Yuspa, 1994). The facility was supervised by the University of Oulu Animal Welfare Committee and followed established guidelines. Samples for histological, immunohistochemical and RNA in situ studies were excised from the skin of normal mice as a control, and from the skin lesions obtained after wounding or carcinogen treatment (Werner et al., 1992). The RNAase protection assay was carried out as described by Vastrik et al. (1995). The mouse Mad antisense cRNA probe was synthesised from nucleotides 1-297 of the published cDNA sequence (Vastrik et al., 1995) using T7 polymerase and [32P]UTP. The mouse fl-actin cRNA was similarly synthesised from nucleotides 1188-1279 of the published cDNA sequence (Tokunaga et al., 1986). After purification in a 6% polyacrylamide/7 M urea gel, the labelled transcripts were hybridised with 30,g of total RNA overnight at 55°C. Single-stranded RNA was then digested with RNAase Tl and RNAase A at 30°C and the purified protected fragments were analysed in a 6% polyacrylamide/7M urea gel. Total RNA was isolated from wound tissue in the mice by guanidium thiocyanatephenolchloroform extraction (Chomczynski and Sacchi, 1987).
In situ hybridisation The mouse Mad antisense and sense cRNA probes were synthesised from linearised pBluescript II SK + plasmid (Stratagene, La Jolla, CA, USA), containing an ApaI-PstI fragment from mouse Mad cDNA (nucleotides 301-1001), via incorporation of [35S]UTP using T3 and T7 polymerases (Amersham, Little Chalfont, UK). Mouse c-Myc antisense and sense cRNA probes were synthesised in a similar manner from linearised pmcxs plasmid (a kind gift from Drs Ronald DePinho and Nicole Schreiber Agus), containing a 750 bp XbaI-Sacl fragment from mouse c-Myc cDNA in pBluescript SK+ (Stratagene). The human Mad antisense and sense cRNA probes were synthesised from linearised pGEM3Zf( +) plasmid containing a PCR-amplified human Mad cDNA insert (Vastrik et al., 1995), using T7 and SP6 polymerases and [35S]UTP.
In situ hybridisation of paraffin sections was performed as Mad expression in keratinocytes A Lymboussaki et al previously described (Wilkinson et al., 1987a,b) with the following modifications: (1) instead of toluene, xylene was used before embedding in paraffin wax; (2) cut sections were placed on a layer of diethylpyrocarbonate-treated water on the surface of glass slides pretreated with 2% 3-triethoxysilylpropylamine; (3) alkaline hydrolysis of the probes was omitted; (4) the hybridisation mixture contained 60% deionised formamide; (5) the high-stringency wash was for 105 min at 65°C in a solution containing 50 mm DTT and 1 x SSC. The sections were coated with NTB-2 emulsion (Kodak) and stored at 4°C. The slides were exposed for 21 days, developed and stained with haematoxylin. Control hybridisations with sense strand and RNAase A-treated sections did not give a specific signal above background.

Immunohistochemistry
Immunohistochemical examination of skin and tumour specimens with antibodies against the core protein of mouse syndecan-1, which served as an epithelial cell differentiation marker was carried out using the avidinbiotin immunoperoxidase method and paraformaldehydefixed, paraffin-embedded material (Korhonen et al., 1984;Elenius et al., 1991). After deparaffinisation and rehydration of the tissue sections, endogenous peroxidase activity -was blocked by incubating the slides in methanol containing 0.3% hydrogen peroxide for 30 min. The sections were then incubated with normal rabbit serum diluted in phosphatebuffered saline (PBS), for 30 min at room temperature. The primary monoclonal antibody 281-2, which specifically recognises the core protein of mouse syndecan-1, was used (Jalkanen et al., 1985) at a concentration of 20 mg ml-' in PBS and incubated overnight at 4°C. The slides were then incubated with biotinylated rabbit anti-rat IgG at a 1:200 dilution in PBS for 40 min at room temperature and then with avidin-biotin-peroxidase complex (Vectastain kit, Vector Laboratories). Between each antibody incubation, the slides were washed three times with PBS. Immobilised peroxidase was visualised by incubation with 0.25 mg ml-' of its substrate 3,3'-diaminobenzidine tetrahydrochloride (DAB; Polysciences, Northampton, UK) in 0.05 M Tris-HCl buffer, pH 8, containing 0.03% hydrogen peroxide for 5 min. Finally, the sections were counterstained with haematoxylin and mounted (Aquamount; BDH, Poole, UK). Antibodies against PCNA (Dako, Glostrup, Denmark) and p53 (a gift from Dr Allan Balmain, Glasgow, UK) were used as described previously (Korhonen et al., 1984).

Results
Mad is highly expressed in the suprabasal layer of mouse epidermis As shown in the in situ hybridisation analysis of Figure 1, Mad is highly expressed in the suprabasal layers of newborn mouse skin (arrows in Figure la and b), and to a lesser extent in adult mouse skin ( Figure Ic and d). Basal epidermal cells adjoining the basement membrane were consistently negative as were the basal cell layers of the hair follicle and hair germ cells (less than 10 grains per cell). No specific signal was detected in stromal fibroblasts, vessels or other supporting structures. Control sections hybridised with the Mad sense strand did not give a specific signal above background either. The degree of expression correlated directly with the relative thickness of the epidermis in the various anatomical regions. Therefore, in regions where the epidermis was thick, such as dorsal skin, the signal was strong ( Analysis of healing skin of full-thickness wounds via in situ hybridisation showed an initial up-regulation of Mad mRNA 3 days after wounding (Figure 2a and b) and a subsequent strong expression at the edges of the wound on days 5 (Figure 2c and d) and 7. By day 13, the expression had decreased to levels equivalent to the unwounded adult mouse epidermis (Figure 2e and f). Consistent with these results, RNAase protection analysis of the wounds showed increased accumulation of Mad mRNA with increasing thickness of the differentiating epidermal keratinocyte layer, reaching a maximum on day 7 of healing ( Figure 3).
Mad in carcinogen-induced mouse skin tumours Our previous studies suggested that Mad causes an inhibition of cell growth in vitro (Vastrik et al., 1995). In order to study if Mad expression is altered during skin carcinogenesis, samples were taken from normal skin and from carcinogenexposed, hyperplastic and dysplastic lesions and squamous cell carcinomas with varying degrees of differentiation. The tumours were generated by repeated administration of DMBA and samples were taken during the neoplastic development and after the animal was sacrificed owing to extensive neoplastic involvement. The lesions analysed represented the entire spectrum of tumour progression, from uninvolved skin to reversible hyperplasia and dysplasia increasing in severity ultimately resulting in the formation of squamous cell carcinomas, varying in differentiation from well-differentiated, keratin-producing tumours to sarcomalike spindle cell neoplasms. Different morphological changes varying in extent and severity were observed in single samples.
Epidermal hyperplasia of carcinogen-treated skin consisting of 5-15 layers of histologically regular, stratified and polarised cells exhibited a distinct, enhanced signal in the upper half of the epidermis (about 50-100 grains per cell) (Figure 4a and b). The suprainfundibular part of the hair follicles also had a distinct signal in the superficial cell layers, whereas the basal cells were consistently negative. Epidermal dysplasia with cytological irregularities and disturbed stratification and polarisation showed a less intense signal in the keratinocytes (arrows in Figure 4c and d). Only the most superficial cells exhibited a strong Mad signal. Downward extensions of cells surrounded by basement membrane were negative (arrowhead in Figure 4c and d).  Figure 4e and f). PCNA-positive proliferating cells and cell layers adjoining the basement membrane were consistently negative for Mad (data not shown). In contrast, poorly differentiated sarcoma-like squamous cell carcinomas exhibited no Mad mRNA, although the adjoining normal epidermis contained cells with a distinct signal (arrows in Figure 4g and h). Weak Mad signal was also detected inside the sebaceous glands (arrowhead in Figure 4g and h). The anaplastic cells, showing only focal keratin expression, were abundantly positive for PCNA and p53 (data not shown).
In order to monitor keratinocyte differentiation in the tumours, the expression of epithelial syndecan-1, which is known to decrease during tumorigenesis, was analysed in the same skin lesions via immunoperoxidase staining. Syndecan-1, like Mad, was most abundantly expressed in the differentiating suprabasal layers of epidermis and in the dermal hair follicle cysts, whereas the dermis was negative. No specific signal was observed in the sections stained with control antibody (data not shown). Hyperplastic DMBAtreated epidermis showing elongated rete ridges, exhibited distinct syndecan-1 staining that was slightly discontinuous and irregular, whereas the basal cell layer and the superficial differentiated epidermal cell layer were negative (Figure 5a and b). In dysplastic areas, atypical keratinocytes showed intensive staining, while the basal cell layer was negative ( Figure Sc). Some syndecan-1 expression was seen, in the well-differentiated areas of the squamous cell carcinomas where keratin horn cysts were formed (Figure 5c), whereas the expression was lost in the poorly differentiated areas of these tumours (Figure 5d).
Mad expression in human squamous cell carcinoma, basal cell carcinoma and melanoma The human squamous cell carcinoma biopsies showed hyperkeratosis, parakeratosis, hyperplastic epidermis with cell dysplasia and various degrees of differentiation including horn cyst formation and invasion in the dermis. Strong Mad signals were detected in the thickened stratum granulosum and in the cells around the horn cysts (Figure 6a and b). Superficial spreading-type basal cell carcinomas showed distinct Mad expression in the stratum granulosum but the peripheral palisading basal carcinoma cells were negative. Autoradiographic grains were present in keratotic basal cell carcinomas forming horn cysts and also in the cells surrounding these cysts, which are considered to represent incomplete hair shaft formation (Figure 6c and d). No Mad expression was detected in nodular melanoma (data not shown). The biopsies also contained areas of adjacent normal healthy skin, where the Mad expression pattern was observed to be similar to that of normal adult mouse skin (Figure 6e and f). Mad expression was also noted inside the sebaceous glands and in the keratinocytes surrounding the lumen of the hair shafts (data not shown). Sections of normal human skin hybridised with the Mad sense strand gave only background unspecific signal from the most superficial keratin layers (inset Figure 6e).

Discussion
In this study we have observed that Mad mRNA is highly expressed in differentiating epidermal keratinocytes in normal epidermis, healing skin wounds and epidermal tumours, whereas the proliferating basal epidermal cells are negative. This , 1995). Differentiating cells in hyperplastic lesions and cells surrounding horn cysts in well-differentiated squamous cell carcinomas strongly expressed Mad mRNA. Dysplastic epidermal cells, although benign and with a preserved basement membrane, expressed only low levels of Mad. Mad was therefore observed in benign as well as in malignant lesions. A few of the most malignant lesions that had lost the structure of epidermal stratification and differentiation were negative for Mad. However, even in the presence of a Mad mRNA signal, the possibility remains that the mad gene has suffered small mutations in critical regions, for example in the region encoding the 25 first amino acid residues, which are important for its function as a transcriptional repressor (Ayer et al., 1995) and presumably as an inhibitor of cell growth (Vastrik et al., 1995).
The expression pattern of Mad was different from that of syndecan-1, which began in a deeper epidermal layer and also extended to the rete ridges. However, there was significant correlation of Mad and syndecan-1 expression in the skin tumours in relation to their degree of malignancy. Syndecan-1 is known to be down-regulated in the most malignant skin lesions (Inki et al., 1994), possibly owing to the loss of epidermal layered cytoarchitecture. Thus, syndecan-1 provides a marker for the loss of epidermal cell differentiation associated with malignant progression, and our studies show that the expression of this marker is correlated well with that of Mad. interestingly, the syndecan-l gene promoter also contains several Myc target sequences (Hinkes et al., 1993), which may be regulated by members of the Myc oncoprotein transcription factor family.
These studies indicate that Mad mRNA expression is associated with epithelial keratinocyte differentiation, but that it is not expressed in rapidly dividing basal epidermal cells. Furthermore, Mad expression is up-regulated in hyperproliferative epidermis, increasing with the thickness of the stratum granulosum. Our results also suggest that Mad expression can occur in both benign and malignant hyperproliferative lesions as long as the cells retain differentiation potential. Loss of expression was seen only in the most anaplastic areas of the tumours. A similar expression pattern was evident in murine and human skin and in chemically induced and naturally occurring skin tumours, suggesting an important role for this transcription factor in the regulation of keratinocyte differentiation.