The DP-1 Transcription Factor Is Required for Keratinocyte Growth and Epidermal Stratification*

The epidermis is a stratified epithelium constantly replenished through the ability of keratinocytes in its basal layer to proliferate and self-renew. The epidermis arises from a single-cell layer ectoderm during embryogenesis. Large proliferative capacity is central to ectodermal cell and basal keratinocyte function. DP-1, a heterodimeric partner of E2F transcription factors, is highly expressed in the ectoderm and all epidermal layers during embryogenesis. To investigate the role of DP-1 in epidermal morphogenesis, we inhibited DP-1 activity through exogenous expression of a dominant-negative mutant (dnDP-1). Expression of the dnDP-1 mutant interferes with binding of E2F/DP-1 heterodimers to DNA and inhibits DNA replication, as well as cyclin A mRNA and protein expression. Chromatin immunoprecipitation analysis demonstrated that the cyclin A promoter is predominantly bound in proliferating keratinocytes by complexes containing E2F-3 and E2F-4. Thus, the mechanisms of decreased expression of cyclin A in the presence of dnDP-1 seem to involve inactivation of DP-1 complexes containing E2F-3 and E2F-4. To assess the consequences on epidermal morphogenesis of inhibiting DP-1 activity, we expressed dnDP-1 in rat epithelial keratinocytes in organotypic culture and observed that DP-1 inhibition negatively affected stratification of these cells. Likewise, expression of dnDP-1 in embryonic ectoderm explants produced extensive disorganization of subsequently formed epidermal basal and suprabasal layers, interfering with normal epidermal formation. We conclude that DP-1 activity is required for normal epidermal morphogenesis and ectoderm-to-epidermis transition.

The skin performs many functions essential for life. It prevents loss of water and electrolytes, provides a barrier against the external environment, and is an integral component of the endocrine, immune, and nervous systems (for review, see Refs. 1 and 2). Two layers form the skin, the outer epidermis and the inner dermis. The epidermis is a complex stratified squamous epithelium formed by one basal and several suprabasal keratinocyte layers (for review, see Ref. 3). The epidermis is constantly subjected to mechanical and chemical insults; as a result, its upper layers are constantly shed and replenished by underlying keratinocytes. This process allows the removal of damaged cells without compromising barrier function. The basal layer contains undifferentiated keratinocytes with high proliferative capacity, which allows them to continuously replenish the suprabasal layers. The stem cells necessary for this process reside in various regions, including the basal layer, and bulge areas in the hair follicles (2,4).
Early in development, the embryo is covered by a single cell-layered ectoderm. In the mouse, as early as 9.5 days postcoitus (dpc), 1 the single-layered ectoderm begins to express epithelial markers such as keratins 8 and 18 (5). Between 9 and 12 dpc, the ectoderm undergoes stratification, producing a second transitory layer termed the periderm (for review, see Ref. 6). At this stage, the epidermal keratins 14 and 5 are expressed (7), indicating the commitment of the embryonic ectoderm to form a mature epidermis. Further stratification of the ectoderm gives rise to the first signs of hair follicle induction (8) and to an intermediate layer termed the stratum intermedium, which differentiates and later gives rise to the first spinous and granular cell layers of the mature epidermis by 15-16 dpc. The final stages of differentiation include the formation of a stratum corneum and acquisition of barrier function, just before birth.
An essential feature of both the ectoderm and the basal epidermal layer is their capacity to proliferate. The ectoderm needs to rapidly expand coordinately with the growing embryo, whereas basal keratinocytes must continuously replenish the epidermis and, in the event of wounding, generate enough new cells to efficiently repair the damage. Central to the control of cell proliferation is the E2F family of transcription factors. This family consists of two subgroups of proteins, termed E2F and DP. To date, seven E2f and two Dp genes have been identified (9,10). In most cases, the functional E2F unit consists of a heterodimer composed of one E2F and one DP protein. The E2F/DP network regulates the expression of genes necessary for transit through G 1 to S phase of the cell cycle, and for DNA replication (for review, see Refs. 10 -13). E2F/DP factors fulfill other functions unrelated to G 1 /S progression, such as modulation of apoptosis, DNA repair, mitosis, and expression of homeobox genes (12,14). Targeted inactivation of individual E2f genes in mice has confirmed the existence of complex biological functions for these factors, including an essential role in epidermal regeneration after injury (15), thymic selection and lymphocyte proliferation (16 -18), cardiac function (19), intestinal and hematopoietic maturation (20,21), and choroid plexus function (22). DP-1 is a broadly expressed member of the E2F family and is present at high levels in a variety of cell lines and tissues (23,24). It efficiently dimerizes with E2F-1 through -6 and is essential for proper DNA binding of these E2F transcriptional units. DP-1 also mediates interactions with other factors that are probably involved in chromatin modulation (25,26). Targeted inactivation of the Dp-1 locus in mice causes severe abnormalities in the development of extraembryonic tissues, which lead to embryonic lethality between 10.5 and 11.5 dpc (27), although the precise mechanisms leading to this phenotype remain unclear.
The early embryonic lethality of DP-1 Ϫ/Ϫ mice precludes the investigation of the role that DP-1 proteins play in the development of the epidermis. To address this issue, we examined the alterations in keratinocyte growth and epidermal morphogenesis after functional inactivation of DP-1, using primary keratinocyte monolayers and organotypic cultures exogenously expressing a dominant-negative DP-1 mutant. We demonstrate that interference with DP-1 activity abrogates normal proliferation of basal keratinocytes, development of the epidermis from the embryonic ectoderm, and keratinocyte stratification.
Plasmids and Recombinant Adenoviruses-In situ hybridization experiments were conducted using a cDNA probe corresponding to the C terminus and 3Ј-untranslated region of murine DP-1 (28). A cDNA encoding a C-terminal HA-tagged murine DP-1 mutant lacking amino acid residues 103-126, herein termed dnDP-1, was generated by PCR. The dnDP-1 or ␤-gal cDNAs were cloned into pAdtrack-CMV, which were used to generate the corresponding recombinant adenoviruses using the pAdEasy system (29). All pAdEasy adenoviruses generated also encode green fluorescent protein (GFP). Viral stocks were amplified in human embryonic kidney 293 cells, purified through CsCl gradients, and titered by dilution assay in human embryonic kidney 293 cells. Titers were expressed as colony-forming units, based on the number of GFP-positive cells observed with each dilution.
Embryonic Ectoderm Explant Cultures and Histology-Vibrissae explant cultures were established from CD-1 embryos, modifying proce-dures described previously (31)(32)(33)(34). In particular, 12.5-13.5 embryos were harvested, and vibrissa pads were isolated and placed on sterile gelatin sponges (Gelfoam; Upjohn) with the ectoderm side up. The sponges were transferred to culture inserts in a two-chamber tissue culture system; the lower chamber contained DMEM supplemented with 2% FBS, penicillin (50 units/ml), and streptomycin (50 g/ml; Invitrogen). Explants were cultured at 37°C, with growth medium changes every 2 days. For adenovirus-mediated gene transfer experiments, 1-day-old explants were immersed for 5 h at 37°C in serum-free DMEM containing the appropriate adenovirus used at 10 8 colony-forming units/ml. The infection medium in the upper and lower insert chambers was removed, and fresh FBS-containing DMEM was added to the lower chamber. At the end of the culture periods, explants were washed with PBS, embedded in Tissue-Tek (OCT compund; Sakura) or 7.5% gelatin, and flash-frozen in liquid nitrogen. Frozen sections (7 m) were obtained, fixed in 4% PFA, stained with hematoxylin and eosin, and mounted using Permount (Fisher Scientific). For ␤-gal staining, PFA-fixed cryosections were washed twice with PBS, incubated in the dark for 16 -20 h at 37°C in a solution containing 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 2 mM MgCl 2 , 1 mg/ml X-gal, and 0.02% Nonidet P-40 in PBS (35). The stained tissues were washed twice in PBS, fixed in 4% PFA (4°C, 16 h), dehydrated with 2-propanol as the final step instead of xylene (to prevent leaching of the blue precipitate), and mounted using Permount.
REK Organotypic Cultures-REK cells were plated on a basal lamina-containing collagen substrate and cultured to allow stratification as described previously (36). In brief, to first generate a basal lamina, confluent MDCK monolayers were cultured onto collagen gels in a Transwell insert for 22 days. The cells were lysed (Buffer H: 10 mM Tris-HCl, 0.1% bovine serum albumin, and 0.1 mM CaCl 2 , pH 7.5) and solubilized (0.2% deoxycholate in buffer H, followed by 0.5% Nonidet P-40 in buffer H), and the basal lamina-coated collagen was extensively rinsed with DMEM. Confluent REK monolayers seeded onto the prepared collagen were cultured in DMEM supplemented with 8% FBS at the air-liquid interface for up to 9 days to allow stratification.
Direct and Immunofluorescence Microscopy-Frozen sections from embryonic explants or REK organotypic cultures (7 m) were fixed with 4% PFA and permeabilized (0.2% Triton X-100 in PBS, 4°C, 15 min). After two PBS washes, the sections were blocked in PBS containing 5% nonfat dry milk and 5% goat serum (2 h, 22°C), followed by incubation with the appropriate primary antibody (1 h, 22°C). After three 20-min PBS washes, the cells were probed with a Cy3-conjugated secondary antibody at 22°C for 1 h. After removal of the secondary antibody, the cells were rinsed twice, incubated with Hoescht 33258 (10 g/ml) in PBS for 5 min at 22°C, rinsed five times with PBS, and mounted in anti-fade medium (DakoCytomation). GFP was visualized in fixed sections by direct fluorescence. Photomicrographs were obtained with a Leica DMIRBE microscope equipped with an Orca II digital camera (Hamamatsu, Japan), using Openlab 3.0 software (Improvision, Coventry, UK).
Cell Culture and Adenovirus Infection-Primary epidermal keratinocytes were isolated from 2-day-old CD-1 mice and cultured in Ca 2ϩfree Eagle's minimum essential medium (Cambrex Bio Science) supplemented with 8% FBS previously treated with Chelex resin (Bio-Rad) and epidermal growth factor (10 ng/ml, Invitrogen), as described previously (37). All experiments were conducted 3-6 days after keratinocyte isolation. REK cells, a line of rat neonatal epidermal keratinocytes (38,39), and Madin-Darby canine kidney (MDCK) cells were a generous gift from Drs. E. Maytin and V. Hascall, and were cultured as monolayers in DMEM (Invitrogen) supplemented with 8% FBS. Human embryonic kidney 293 cells (American Type Culture Collection) were cultured in DMEM containing 8% FBS. Viral infections were conducted by incubating primary keratinocytes in serum-and Ca 2ϩ -free Eagle's minimum essential medium with the appropriate adenovirus at a multiplicity of infection of 50 for 3-4 h, followed by culture in normal growth medium. Under these conditions, Ն90% of primary mouse keratinocytes were infected, as confirmed by X-gal staining or GFP fluorescence, respectively, in cells infected with ␤-galor GFP-encoding virus. Adenovirus infection of REK cells was conducted in serum-free DMEM, at a multiplicity of infection of 200, followed by culture in normal growth medium for 24 h before fluorescence-activated cell sorting.
Immunoblotting and Immunoprecipitation Assays-Whole-cell keratinocyte extracts were prepared for immunoblot and analyzed as described previously (40). For immunoprecipitation experiments, keratinocytes were harvested and incubated for 30 min at 4°C in lysis buffer A (50 mM HEPES, pH 7.7, 250 mM KCl, 10% glycerol, 0.1% Nonidet P-40, 0.4 mM NaF, 0.4 mM Na 3 VO 4 , 0.5 mM phenylmethylsulfonyl flu-oride, 1 mM dithiothreitol, 2 g/ml each leupeptin, pepstatin, and aprotinin, and 0.1 mM EDTA). Lysates were pre-cleared by incubation with protein A and protein G-Sepharose (2 h, 4°C; Amersham Biosciences), before being incubated at 4°C for 16 h with the appropriate primary antibody for immunoprecipitation. Incubation with secondary antibodies (2 h, 22°C) was followed by protein A or protein G-Sepharose, as appropriate (2 h, 22°C), and extensive washes with lysis buffer A. Immune complexes bound to the Sepharose beads were dissociated with sample buffer (5 min, 95°C) and resolved by SDS-PAGE followed by immunoblot analysis.
Electrophoretic Mobility Shift Assays-Whole-cell extracts were prepared in lysis buffer A. Each binding reaction was prepared and analyzed using 5-25 g of cellular protein as described previously (37).
[ 3 H]dThd Incorporation into DNA, Flow Cytometry, and Fluorescence-activated Cell Sorting-Incorporation of [ 3 H]dThd into DNA was measured from TCA-insoluble keratinocyte fractions as described previously (37). For flow cytometry analysis, keratinocytes were infected at 0 or 48 h after isolation and harvested at the indicated times after infection. Cells were rinsed twice with PBS, trypsinized, fixed using cold methanol (Ϫ20°C), and analyzed immediately or stored in methanol at Ϫ20°C. Just before analysis, cells were washed with PBS, and suspended for 30 min in 50 mM HEPES with 10 mM EDTA, pH 8.0, to prevent clumping. The cells were treated with RNaseA (50 g/ml, final) and propidium iodide (10 g/ml), and analyzed on a FACScan flow cytometer (BD Biosciences) using the CellQuest program. The data were analyzed using ModFit software. REK cells were infected with adenovirus encoding GFP and ␤-galactosidase or GFP and dnDP-1, and 24 h later, they were harvested and suspended in PBS with 10% FBS (1-2 ϫ 10 6 cells/ml). Cells were sorted using a FACS Vantage sorter (BD Biosciences), using an Enterprise II laser (488 nm). Viable cells with GFP fluorescence were isolated and plated on collagen substrates (300,000 cells/sample) to form organotypic cultures.
Cyclin A mRNA Abundance-Cyclin A mRNA levels were measured using a quantitative reverse transcription PCR method (15) from total RNA prepared from cultured keratinocytes infected with recombinant adenovirus encoding ␤-gal or dnDP-1 or from uninfected controls. Reverse transcription followed by polymerase chain reaction amplification was conducted on at least two different RNA samples (1 g of RNA was reverse-transcribed, and cDNA equivalent to 0.1 g RNA was used in each PCR reaction). The primers used were, for cyclin A, 5Ј-CCCCCA-GAAGTAGCAGAGTTTGTG and 5Ј-CATGTTGTGGCGCTTTGAGGT-AGG, which amplify a 400-bp fragment, and for GAPDH, 5Ј-CAAAGT-TGTCATGGATGACC and 5Ј-GTTGCCATCAACGACCCCTT, or 5Ј-GC-TTCACCACCTTCTTGATGTCATC and 5Ј-GTTGCCATCAACGACCC-CTT. PCR fragments obtained at 18, 20, 22, and 24 amplification cycles were resolved by electrophoresis, transferred to nylon membranes, and hybridized to appropriate 32 P-labeled probes. The signals were detected and quantified using filmless autoradiographic analysis (Canberra Industries) and normalized to GAPDH products amplified in the same reactions. For each cDNA, amplification was quantified by filmless autoradiographic analysis at multiple PCR cycles (18 -24 cycles). Amplification of all cDNAs tested was logarithmic within these parameters.

DP-1 Expression in Embryonic Ectoderm and Stratified
Epidermis-To begin to understand the role of DP-1 during epidermal morphogenesis, we first analyzed its patterns of expression in embryonic ectoderm and epidermis. In situ FIG. 1. Expression of DP-1 during epidermal development. Bright-field and corresponding dark-ground micrographs of sagittal embryo sections of the indicated gestational ages hybridized to a DP-1 antisense (as) or sense (s) probe. Arrows in 10.5-and 14.5-dpc embryo sections indicate the single-cell and multilayered ectoderm, respectively. DP-1 expression is observed in ectoderm and fully stratified epidermis. E, epidermis; HF, hair follicle. hybridization analysis of the single-cell layered ectoderm in 10.5-dpc embryos revealed DP-1 expression (Fig. 1). DP-1 transcripts were also detected in 14.5-dpc embryos, in which the ectoderm had given rise to the stratum intermedium and stratum basale (Fig. 1). DP-1 expression increased substantially in fully stratified epidermis in 17.5-dpc embryos. At this stage, DP-1 transcripts were detected in all epidermal layers and epidermal appendages, including the hair follicles (Fig. 1). Thus, DP-1 is likely to play an important role in the ectodermto-epidermis transition, as well as in keratinocyte stratification in the fully developed epidermis.
Inhibition of E2F/DP DNA Binding Activity and Interaction with Cellular E2F Proteins by a Dominant-negative DP-1 Mutant-To determine the alterations in keratinocyte proliferation and differentiation consequent to DP-1 inactivation, we developed a model that allowed us to express a dominantnegative DP-1 mutant in primary cultured keratinocytes, in keratinocyte organotypic cultures, and in murine embryonic ectoderm. We first generated a recombinant adenovirus encoding a mutant mouse DP-1 protein lacking residues 103-126, which we termed dnDP-1 ( Fig. 2A). This mutant lacks residues essential for binding to DNA (42). Heterodimers containing the human orthologue of this DP-1 mutant can associate with cellular E2F proteins, greatly reducing capacity to bind DNA and activate transcription when expressed in Saos-2 osteosarcoma cells (42).
We initially examined the time course of expression of adenovirus-encoded dnDP-1 in cultured primary keratinocytes and determined that peak expression levels occurred 72 h after infection (Fig. 2B). High dnDP-1 expression levels were maintained as late as 5 days after infection.
We also determined the effect of dnDP-1 on the ability of E2F/DP complexes to bind DNA. Results from electrophoretic mobility shift assay using lysates prepared from untreated keratinocytes, or keratinocytes infected with the ␤-gal-encoding adenovirus, showed several complexes with different mobilities, which correspond to either "free" E2F/DP dimers or higher order complexes containing E2F/DP species associated with pRB family proteins, as reported previously (37). Lysates prepared from keratinocytes expressing dnDP-1 showed a marked reduction in the abundance of all E2F/DP species capable of binding to DNA, compared with lysates from keratinocytes infected with a ␤-gal-encoding adenovirus (Fig. 2C) or from uninfected cells (data not shown). Thus, dnDP-1 interferes with the ability of E2F/DP species to bind DNA, irrespective of the presence or absence or pRB family proteins in the complexes. These observations are consistent with the concept that the dnDP-1 mutant exhibits dominant-negative behavior with respect to DNA binding and, consequently, transcriptional regulation in keratinocytes.
FIG. 2. dnDP-1 inhibits E2F/DP binding to DNA and associates with cellular E2F proteins. A, schematic of dnDP-1 mutant showing the deletion of amino acid residues 103-126 and the location of the HA tag. The regions in the wild-type (wt) DP-1 protein corresponding to the DNA-binding domain and the dimerization domain that mediates complex formation with E2F proteins are shown. B, expression of dnDP-1 in cultured primary mouse keratinocytes. Primary mouse keratinocyte cultures were established. Forty-eight hours later, the cells were cultured in the presence of an adenovirus encoding the HA-tagged dnDP-1 for 5 h. Cell lysates were prepared at the indicated times after the end of adenoviral infection and analyzed by immunoblot using an anti-HA antibody. The 60-kDa protein corresponds to full-length dnDP-1, and the species migrating at 43 kDa corresponds to a protein translated from position 127, generated by the presence of an NcoI site, which acts as an internal ribosome entry site present in the dnDP-1 cDNA se-quence. Accumulation of dnDP-1 reaches a peak 72 h after infection but is maintained as late as 120 h after infection. C, electrophoretic mobility shift assay showing E2F/DP binding to an oligonucleotide probe corresponding to the E2F-binding site in the dihydrofolate reductase promoter. Primary keratinocytes were infected with adenovirus encoding ␤-gal or dnDP-1; 24 h later, lysates were prepared. Binding to DNA in lysates from keratinocytes expressing dnDP-1 is inhibited relative to that observed in cells expressing ␤-gal. Assays were conducted in the absence (-) or presence (ϩ) of a 50-fold molar excess of an unlabelled competing oligonucleotide containing sequences corresponding to the E2F site in the adenovirus E2 promoter. Loss of binding in the presence of competing oligonucleotide confirms the specificity of E2F binding. D, dnDP-1 associates with endogenous E2F proteins in keratinocytes. Lysates were prepared from keratinocytes infected with adenovirus encoding ␤-gal or dnDP-1, as indicated at the top of the figure. The lysates were subjected to immunoprecipitation with the E2F antibodies indicated, and the presence of dnDP-1 in the immunoprecipitates was examined from immunoblots probed with an anti-HA antibody.
Next, we investigated whether the mutant dnDP-1 showed selectivity in its association with individual E2F forms and identified individual E2F proteins with which dnDP-1 interacts. Keratinocyte lysates were incubated with antibodies directed against E2F-1 through -5. The immunoprecipitates were resolved by SDS-PAGE and analyzed by immunoblotting for the presence of the HA-tagged dnDP-1. We readily detected HA immunoreactivity in the immunoprecipitates of each E2F protein in lysates from keratinocytes expressing dnDP-1, but not in those cells infected with adenovirus encoding ␤-gal (Fig. 2D), demonstrating the pleiotropic ability of this mutant to form complexes with endogenous E2F forms present in keratinocytes.
Deficient Transit through S Phase in Keratinocytes Expressing dnDP-1-E2F/DP activity is essential for normal cell proliferation. Therefore, we evaluated the effects of dnDP-1 expression on the ability of primary murine keratinocytes to synthesize DNA and progress through the cell cycle. We found that keratinocytes expressing dnDP-1 had a substantially impaired DNA synthetic capacity. Indeed, incorporation of [ 3 H]dThd into DNA in dnDP-1-expressing cells was ϳ55% of that observed in cells expressing ␤-gal (Fig. 3A). To better characterize this phenomenon, we subjected dnDP-1-or ␤-galexpressing keratinocytes to flow cytometric analysis. At the time of isolation from the epidermis, 70 -85% of keratinocytes exhibit a 2N DNA content and are in the G 0 /G 1 phase of the cell cycle (15). Upon culture in medium containing serum and growth factors, these cells enter the cell cycle, reaching the S phase within 2-3 days of being placed in culture. We infected keratinocyte cultures with adenoviruses encoding either dnDP-1 or ␤-gal 48 h after isolation. At the time of infection, the distribution of the cell population was about 65% in G 1 phase, 20% in S phase, and 14% in G 2 /M (Fig. 3B). Over a 72-h time course after infection, the fraction of cells with 2N DNA content that expressed either ␤-gal or dnDP-1 decreased to about 40%, indicating a similar capacity in both cell populations to transit through the G 1 phase (Fig. 3B). The fraction of ␤-gal-expressing keratinocytes in S phase, with a DNA content between 2N and 4N, showed an initial increase 24 h after infection to about 22% and did not substantially increase thereafter, because some cells moved from G 1 into S phase and others moved out of S and into G 2 /M (Fig. 3B). In contrast, the fraction of keratinocytes expressing dnDP-1 with DNA content between 2N and 4N accumulated throughout the 72-h time course, reaching ϳ30% of the cell population (Fig. 3B). The increase in this cell fraction was associated with a reduced proportion of cells with a 4N DNA content (corresponding to cells in G 2 /M; Fig. 3B). Thus, cells that expressed dnDP-1 seemed to transit more slowly through S phase, or to synthesize DNA at lower rates, than control, ␤-gal-expressing keratinocytes. The increased fraction of cells expressing dnDP-1with DNA content between 2N and 4N, together with their reduced ability to synthesize DNA, is consistent with the concept that inhibition of DP-1 activity in keratinocytes disrupts normal rates of DNA replication, whereas transit through the G 1 phase remains unaffected. Similar results were obtained when keratinocytes were infected with the adenovirus encoding dnDP-1 at the time of isolation and 24 h before stimulation with serum and EGF (data not shown). Thus, DP-1 activity is essential for normal S phase, but not G 1 , progression in primary murine keratinocytes. It is noteworthy that flow cytometric analyses showed less than 1% of cells with a sub-G 1 DNA content (a marker of apoptosis), irrespective of whether they expressed ␤-gal or dnDP-1 (data not shown), demonstrating that dnDP-1 expression did not induce apoptosis.

Inhibition of E2F/DP-1 Activity Results in Decreased Cyclin
A Protein Expression-To investigate the molecular basis for the impaired transit through S phase in keratinocytes ex-  's t test). B, primary mouse keratinocytes were infected with adenovirus encoding ␤-gal (f) or dnDP-1 (E). Cells were harvested at the indicated times after infection, fixed, and stained with propidium iodide for flow cytometry analysis. The results are expressed as the mean of nine experiments. For each experimental point in the charts, the standard deviation value was Յ8% of the mean; error bars have been omitted from the graphs for clarity. *, p Ͻ 0.05; **, p Ͻ 0.01 (paired Student's t test).
pressing dnDP-1, we examined the effect of DP-1 inhibition on known E2F/DP target genes that are critical for normal DNA replication and S phase progression. Two of these targets are cyclins E and A, which fulfill multiple functions (43). For example, whereas cyclin E contributes to the formation of DNA replication complexes, cyclin A activates DNA synthesis by those pre-formed complexes (44). Cyclin A is essential for DNA replication in primary epithelial and other somatic cell types (45)(46)(47).
We first determined whether the presence of dnDP-1 altered the abundance of cyclins E and A. Immunoblot analyses of keratinocyte lysates revealed a substantial decrease in cyclin A protein levels in dnDP-1-expressing keratinocytes, which was most pronounced 24 and 72 h after (Fig. 4). In contrast, and consistent with our previous observation that dnDP-1-expressing cells are able to reach S phase, but fail to progress normally through it, cyclin E levels were not significantly altered by dnDP-1 (Fig. 4).
Cyclin-dependent kinase inhibitors also play an important role in progression through G 1 and S phases. In particular, p21 WAF/CIP1 can retard S phase progression by interfering with activation of cdk2 by cyclin E (48 -50). We observed modest increases in p21 levels in keratinocytes expressing dnDP-1 but not in other cyclin-dependent kinase inhibitors, such as p27 Kip1 (Fig. 5).
Modulation by dnDP-1 of in Vivo Binding of E2F/DP Complexes to the Cyclin A Promoter in Keratinocytes-We next assessed whether the marked decreases in cyclin A protein levels induced by dnDP-1 were associated with reductions in cyclin A mRNA abundance, expected by interference of dnDP-1 with E2F/DP-regulated cyclin A transcription. To this end, we used a reverse transcription PCR method that allowed us to quantify cyclin A mRNA from uninfected keratinocyte cultures or cells infected with adenovirus encoding ␤-gal or dnDP-1 (15), and observed that keratinocytes expressing dnDP-1 exhibited a 50 -60% reduction in cyclin A mRNA abundance (Fig. 6A).
The transcriptional activation capacity of the E2F/DP dimers is imparted by the E2F subunit, and individual E2F forms seem to be able to activate transcription to different extents (10). In particular, E2F-1, -2 and -3 seem to be stronger activators than E2F-4 and -5 under certain circumstances. Furthermore, although several promoters seem to be bound in live cells equally well by various E2F proteins, others are selectively bound by only a subset of these transcription factors (51-53). Having determined that dnDP-1 associates with E2F-1 through -5, we wished to determine whether one or several E2F/DP-1 forms bind to transcriptional regulatory sites on the cyclin A promoter in live keratinocytes. To this end, we first conducted ChIP assays, using chromatin prepared from proliferating keratinocytes. We used appropriate antibodies to isolate immunoprecipitates containing each E2F protein from E2F-1 through -5. The DNA was then analyzed for the presence of cyclin A promoter sequences containing the E2F-binding consensus, using polymerase chain reaction amplification with specific primers that amplify a 300-bp fragment that encompasses the E2F binding element in this promoter. We observed clear amplification above background of cyclin A promoter sequences in chromatin immunoprecipitated with antibodies against E2F-3 and E2F-4 (Fig. 6B). In contrast, we did not observe amplification above background in material isolated from E2F-1, E2F-2, or E2F-5 immunoprecipitates, even though we were able to amplify other E2F target promoters, such as N-myc, from the same samples ( Fig. 6B and data not shown). The results of these experiments are consistent with the concept that E2F-3 and -4 collaboratively modulate the cyclin A promoter in proliferating keratinocytes. This modulation results in overall transcriptional activation and cyclin A expression.
Our observations that dnDP-1 associates with E2F-3 and E2F-4 ( Fig. 2D) prompted us to examine whether expression of dnDP-1 decreased their ability to associate with the cyclin A promoter in live keratinocytes. We conducted ChIP experiments with chromatin isolated from cells infected with adenovirus encoding ␤-gal or dnDP-1. We consistently observed that the abundance of cyclin A promoter amplicons from E2F-3 immunoprecipitates was similar to that found in amplicons from E2F-4 precipitates in either uninfected cells or cells infected with the adenovirus encoding ␤-gal (Fig. 6, B and C). In contrast, expression of dnDP-1 resulted in decreased abundance of cyclin A amplicons, especially those from E2F-3 immunoprecipitates (Fig. 6C). We interpret this finding to mean that dnDP-1 interferes with E2F-3 and E2F-4 binding to the cyclin A promoter. Together, our data strongly suggest that activation of cyclin A gene expression by complexes containing DP-1 and E2F-3 and/or E2F-4 is a determinant for proper keratinocyte division and transit through the cell cycle.

DP-1 Is Required for Normal Epidermal Keratinocyte
Stratification-The experiments described above demonstrated the importance of DP-1 for keratinocyte progression through the cell cycle, but did not address specifically the role of DP-1 in the maturation of undifferentiated basal keratinocytes and their capacity to form a multilayered tissue. To determine whether DP-1 is necessary for proper keratinocyte stratifica- Lysates from primary mouse keratinocytes infected with the indicated recombinant adenovirus were prepared at timed intervals after infection, and analyzed by immunoblot with the indicated antibodies. The blots were also probed with an antibody against GAPDH to provide a loading control.

FIG. 5. Expression of cyclin-dependent kinase inhibitors in keratinocytes expressing dnDP-1.
Lysates from primary mouse keratinocytes infected with the indicated recombinant adenovirus were prepared at timed intervals after infection and analyzed by immunoblot. The blots were also probed with antibodies against ␤-tubulin or GAPDH to provide a loading control. tion, we used a line of rat epidermal keratinocytes (REK), which retain the capacity to form a stratified tissue in organotypic culture (36,39,54). REK cells form a confluent basal monolayer when cultured on a basal lamina-covered collagen fibrillar matrix. Such a monolayer differentiates and stratifies if it is lifted to an air-liquid interface. Forty-eight hours after being lifted, REK cultures begin to show stratification, which continues over time. By 5-6 days of culture, the stratified epithelium shows a smooth, highly organized basal layer, covered by several suprabasal layers (Fig. 7A) that are characterized by normal expression of differentiation markers such as filaggrin and keratin 10 (36,54,55). These organotypic cultures also exhibit other typical epidermal characteristics, including suprabasal involucrin expression, a well developed stratum corneum with tightly packed keratin filaments (which increases in thickness in a time-dependent fashion (Fig. 7A)), and a permeability barrier that mimics that of normal skin in vivo (36,54,55).
We infected REK monolayers with adenovirus encoding ␤-gal or dnDP-1. These viruses also encode GFP, to allow identification of infected cells. Twenty-four hours after infection, we subjected the cells to fluorescence-activated cell sorting to ensure that all cells used for the organotypic cultures had been infected. This step was taken because only 60 -70% of REK cells were infected by our adenoviruses under conditions that precluded virus-induced toxicity (data not shown). We maintained the organotypic cultures for 6 days, at which time we evaluated their morphology, extent of stratification and expression of differentiation markers. In a manner analogous to that observed in uninfected controls, cultures infected with the ␤-gal-encoding virus showed the presence of three to four suprabasal cell layers, which included a well defined cornified layer containing keratin filaments (Fig. 7B, H&E). In contrast, cultures that expressed dnDP-1 showed a substantially lower degree of stratification, and the stratum corneum in these organotypic cultures was absent or severely reduced in thickness (Fig. 7B, H&E). The basal layer in these cultures also showed occasional gaps and a degree of disorganization on the cell surface in contact with the basal lamina not observed in ␤-gal-infected cells. To determine the effect of dnDP-1 on the differentiation status of the REK cells in the organotypic culture, we also analyzed the expression of keratin14, keratin 10, and involucrin. Keratin 14 is normally expressed in undifferentiated keratinocytes, whereas keratin 10 and involucrin are markers of differentiated cells. Analysis of uninfected cells or cells expressing ␤-gal indicated expression of keratin 14 in the basal layer and keratin 10 in the suprabasal layers adjacent to the basal stratum ( Fig. 7B and data not shown). It is notable that keratin 14 was present in dnDP-1-expressing basal cells, and those scant areas with a suprabasal layer showed keratin 10 and involucrin expression ( Fig. 7D; data not shown). This indicates that inactivation of DP-1 delays the formation of multiple layers in the organotypic cultures; once the layers are formed, however, induction of at least some differentiation markers occurs.
Disruption of Epidermis Formation in Embryonic Ectoderm Explants in the Presence of dnDP-1-During early embryogenesis, the surface epithelium consists of a single-layered, highly proliferative ectoderm. A crucial period in epidermal morphogenesis is the time of ectodermal cell commitment to form interfollicular epidermis or appendages (6,55). In the mouse, between 9 and 12 dpc, the ectoderm undergoes stratification to produce a transitory peridermal layer, which is shed later. Additional stratification of the ectoderm between 12 and 15 dpc produces basal and intermediate layers. The cells in the intermediate layer differentiate at 15-16 dpc, ultimately giving rise to the upper (suprabasal) spinous and granular layers of the epidermis, whereas those in the basal layer include keratinocyte stem cells and their committed but still undifferentiated progeny. Formation of epidermal appendages involves fate decisions in ectodermal stem cell lineages partly directed by signals originating in the underlying mesenchyme (for review, see Ref. 4).
To investigate the role of DP-1 in the transition of the embryonic ectoderm to fully stratified epidermis, we first established vibrissae explant cultures from 12.5-dpc embryos. In these explants, some regions of the surface ectoderm have already formed the intermediate layer. It is noteworthy that these cultured explants give rise to a fully stratified epidermis and normal vibrissae with a time course that parallels epidermal and hair follicle morphogenesis in vivo (Fig. 8A, H&E, no virus) (33,34). We initially incubated these explants with an adenovirus encoding GFP and ␤-gal to determine the susceptibility to infection of the cultured ectoderm, as well as the duration of exogenous protein expression in the explants. We

FIG. 6. Expression of cyclin A mRNA and binding of E2F-3/DP and E2F-4/DP to the cyclin A promoter in keratinocytes.
A, expression of cyclin A mRNA in keratinocytes infected with adenovirus encoding ␤-gal or dnDP-1, as indicated. Cyclin A mRNA levels were estimated in RNA harvested 24 h after adenovirus infection of primary keratinocytes, using quantitative reverse transcription-PCR assays, detected by hybridization to a cyclin A 32 P-labeled probe. For each cell population, band intensities were measured using filmless autoradiographic analysis and normalized to those obtained from GAPDH. The results are expressed as mean Ϯ S.D. of two experiments with duplicate estimates. B, ChIP assays to detect E2F binding to the cyclin A promoter were conducted with the indicated anti-E2F antibodies or with normal IgG to control for nonspecific enrichment of chromatin. A sample corresponding to 0.2% of chromatin input was used as a positive control for the PCR amplification. Samples were amplified using specific primers that flank the E2F sites in the murine cyclin A promoter. C, ChiP assays were conducted as above, using chromatin from keratinocytes infected 24 h before harvesting with adenovirus encoding ␤-gal or dnDP-1 and immunoprecipitated with either E2F-3 or E2F-4 antibodies.
were able to efficiently infect explants from 12.5-or 13.5-dpc embryos, but not those from later developmental stages, and determined that GFP expression was maintained for at least 72 h after infection (Fig. 8A, ␤-gal/GFP, and data not shown). Expression of GFP was observed across all cell layers in some areas of the explants and limited to suprabasal layers in other areas of the tissue. It is noteworthy that adenovirus infection of the ectoderm did not perturb normal stratification or hair follicle formation at later stages (Fig. 8A, ␤-gal/GFP, and data not shown). The high GFP expression in the upper layers is consistent with the concept that the intermediate layer, which is the precursor of the first epidermal suprabasal layers, was a major target of adenovirus infection in these explants.
In contrast to the well developed epidermal tissue in ␤-galexpressing explants, adenovirus-mediated expression of dnDP-1 interfered with normal development (Fig. 8A, dnDP-1/ GFP). In particular, as early as 24 -30 h, and up to 72 h after infection, proper stratification in these explants did not occur. In place of the highly structured basal layer containing packed cells with cuboidal or columnar shape, dnDP-1-infected explants exhibited a poorly structured, disorganized basal layer characterized by the presence of large gaps, potentially suggestive of abnormal patterns of cell division. Furthermore, suprabasal region rudiments consisted of scarce and irregularly shaped columns, similarly interspersed by large gaps in the tissue, and a substantial reduction in epidermal thickness. Similar degrees of disorganization and presence of gaps were evident in the developing hair follicles. We verified in these explants the expression of the virally induced proteins and found that GFP was expressed in a pattern similar to that observed in explants infected with the control ␤-gal-encoding adenovirus (Fig. 8A, GFP). We also examined the explants for presence of condensed chromatin, indicative of apoptosis, and found no evidence of apoptosis in any of the explants (Fig. 8B, Hoescht, and data not shown).
To better understand the mechanisms of epidermal morphogenesis disruption induced by dnDP-1 in these explants, we investigated whether cells expressing dnDP-1 retained the capacity to differentiate. To this end, we analyzed serial sections for the presence of suprabasal keratinocytes that expressed the differentiation markers involucrin and keratin 10. We determined that cells in the uppermost layers of the explants were able to express these two proteins, irrespective of the presence of dnDP-1 or ␤-gal (Fig. 8B, Keratin 10, and data not shown), indicating that dnDP-1 does not abolish the ability of keratinocytes to differentiate once they reach the suprabasal epidermal layers. Together, our observations indicate that DP-1 activity is essential for normal ectoderm transition to epidermis, keratinocyte stratification, and proliferation. It is noteworthy that interference with DP-1 activity does not seem to impair the ability of keratinocytes to respond to differentiation signals once they reach the suprabasal layers and to express differentiation markers. DISCUSSION A key event during embryonic development is the extensive expansion of the surface ectoderm needed to maintain the external surface of the growing embryo. Essential for this process is the ability of ectodermal cells to rapidly divide. Numerous factors are involved in cell division and progression through the cell cycle, including cyclins, cyclin-dependent kinases and their inhibitors, E2F transcription factors, and the retinoblastoma family of proteins (56,57). Differential expression of E2F factors occurs in the ectoderm and the fully formed epidermis (30,37). In particular, E2F-4 mRNA is first detected in the 11.5-dpc single cell-layered ectoderm and later in the basal epidermal layer, and E2F-4 protein is expressed in primary cultured keratinocytes, irrespective of their differentia-FIG. 7. Inhibition of DP-1 activity interferes with keratinocyte stratification. A, REK organotypic cultures were established on a basal laminacoated collagen matrix and cultured at the air-liquid interface for the indicated number of days to allow keratinocyte stratification and formation of suprabasal layers. Organotypic cultures were fixed, sectioned and stained with hematoxylin and eosin (H&E) (4d, 6d, 9d) or processed for immunofluorescence microscopy and probed for involucrin (9d-Inv), using Hoescht 33258 to stain the cell nuclei. Arrowheads indicate the presence of eosin-stained orthokeratotic stratum corneum overlaying the suprabasal layers in 6d and 9d cultures. B, delayed REK stratification by inhibition of DP-1 activity. REK cells were infected with adenovirus encoding GFP and either ␤-gal or dnDP-1. GFP-expressing cells were sorted 24 h after infection and cultured on a basal lamina-coated collagen matrix for 6 days before fixation and embedding. Sections stained with H&E show a well developed stratum corneum in uninfected or ␤-galinfected cells but not in cultures expressing dnDP-1. Adjacent tissue sections were processed for immunofluorescence microscopy and probed for keratin 10. Cells expressing the adenovirus-encoded proteins are identifiable by their expression of GFP. Cell nuclei were visualized with Hoescht 33258. Note that the positive keratin 10 staining is confined to suprabasal cells in all samples. The results shown are representative of three experiments. tion status. In contrast, E2F-2 mRNA is first detected in some areas of the 13.5-dpc ectoderm and later in the basal layer of the epidermis and in the epithelium surrounding the dermal papilla. E2F-5 expression, first detected in the suprabasal layers of the 15.5-dpc epidermis, occurs preferentially in differentiated keratinocytes, whereas E2F-1, -2, and -3 proteins are mainly expressed in undifferentiated keratinocytes. Finally, the presence of DP-1 in the ectoderm and in all epidermal layers indicates its central role in E2F/DP function in this tissue. E2F and the retinoblastoma family of proteins are centrally involved in keratinocyte proliferation, epidermal morphogenesis, and tumorigenesis (15, 58 -65). DP-1 is required for extraembryonic development, and Dp1null mice die at approximately 10 days of gestation because of abnormalities in trophectoderm-derived tissues (27). The early death of these mutant mice precludes the investigation of how DP-1 modulates organogenesis and epidermal formation at later stages. However, the broad expression of DP-1, together with the more restricted distribution of DP-2 (23,28), the other member of the DP subfamily, is consistent with the idea that DP-1 fulfills a central function in the development and physiology of many tissues.
E2F/DP activity is very important for transition from quiescence (G 0 ) to growth, as well as progression from the G 1 to the S phase in a variety of cell types. Exogenous expression of a dominant-negative human DP-1 mutant incapable of binding to DNA has been reported to block G 1 progression in human osteosarcoma or lymphoid cells, without significant effects on S phase transit (42,66). In stark contrast, the principal alteration we observed in keratinocytes that expressed dnDP-1 consisted of S phase lengthening, without significant effect on transit through G 1 . This may be caused by incomplete inactivation of DP-1 and/or DP-2 activity by dnDP-1, or through E2F-7, an E2F form that does not bind DP proteins (9), in keeping with the low residual DNA binding activity we observed in keratinocyte lysates containing dnDP-1 and used in electrophoretic mobility shift assay experiments.
Expression of cyclin A is a well established requirement for normal DNA replication, whereas cyclin E activity is essential for G 1 3 S progression (45)(46)(47). Thus, the S phase lengthening observed in keratinocytes expressing dnDP-1 is consistent with the observed reductions in cyclin A expression without significant effect on cyclin E abundance. The cyclin A promoter contains functionally important E2F/DP-binding sites, and transcription of this gene is activated by E2F/DP dimers (67). However, the exact identity of the E2F and DP proteins in physiological complexes that regulate this gene has never been determined. Our ChIP assays demonstrate that E2F-3 and FIG. 8. Inhibition of DP-1 activity precludes normal embryonic epidermis formation. A, vibrissae explants were established from 12.5-dpc embryos. Twenty-four hours after isolation, the explants were maintained uninfected or were infected with adenovirus encoding GFP and ␤-gal or GFP and dnDP-1, as indicated, and cultured for an additional 60-h interval. The explants were embedded, frozen, and sectioned. Adjacent tissue sections were either stained with H&E, or examined by direct fluorescence to localize GFP expression (GFP), indicative of adenovirus-mediated gene transfer. The arrows indicate hair follicles. B, the expression of differentiation markers was also examined in explants infected with adenovirus encoding ␤-gal and GFP or dnDP-1 and GFP as indicated. DNA in these tissues was visualized with Hoescht 33258, whereas infected cells were identified by direct GFP fluorescence. Keratin 10 expressed in suprabasal layers was visualized by indirect immunofluorescence. Note the large proportion of GFP-positive cells that also express keratin 10. The results shown are representative of nine experiments done in duplicate or triplicate samples. E2F-4 occupy this promoter in live, exponentially proliferating epidermal keratinocytes. Furthermore , E2F-3 activity is necessary for normal DNA replication (68 -70). The binding of E2F-3/DP and E2F-4/DP dimers to the cyclin A promoter exhibits selectivity, in that other E2F-regulated promoters in these cells do not seem to be bound by these factors. 2 Binding selectivity to several E2F-activated promoters was not observed in asynchronous NIH 3T3 fibroblast cultures but was demonstrated for the b-myb and the dihydrofolate reductase promoters in serum-stimulated cells (71). The net effect of E2F-3/DP and E2F-4/DP binding on the cyclin A promoter in keratinocytes is transcriptional activation. The result of inhibiting binding of these factors by dnDP-1 is repression of cyclin A expression. These two complexes probably mediate the crucial functions of DP-1-containing species in cyclin A expression, which would contribute to normal DNA replication, essential for early ectoderm development and epidermal morphogenesis.
To explore the role of DP-1 in formation of the epidermis, we exogenously expressed a dominant-negative DP-1 mutant and used several experimental models that would allow us to examine distinct events involved in epidermal morphogenesis. Thus, we used 12.5-13.5-dpc embryonic vibrissae explants, in which ectodermal precursors have yet to give rise to keratinocytes, to determine the effect of inactivating DP-1 on the transition of the ectoderm into epidermis. We found that loss of DP-1 activity in the ectoderm interfered with further development and formation of the stratified epidermis and hair follicles, without increases in apoptosis or ability to express keratin 10 or involucrin in suprabasal cells, two markers of differentiation. This phenotype probably results from inactivation of multiple E2F/DP dimers in ectodermal cells, in that targeted inactivation of single E2f1 through E2f5 genes in mutant mice does not recapitulate the more severe perturbation of epidermal morphogenesis in our explants and is consistent with the existence of functionally redundant roles among E2F/DP species.
Inhibition of DP-1 activity by expression of dnDP-1 in REK cultures also interfered with their ability to form a stratified epithelium under organotypic culture conditions. It is noteworthy that although the formation of a multilayered tissue was significantly reduced in the absence of normal DP-1 activity in these cells, the ability to express involucrin and keratin 10 in those scarce areas that exhibited two-or three-cell layers was not abolished. Likewise, normal basal expression of keratin 14 was present in the tissues. Several mechanisms can be responsible for the observed effect of the dnDP-1 mutant, including a proliferation defect and that prolonged transit through S phase of dnDP-1-expressing keratinocytes that would preclude the generation of enough cells to create a normal multilayered tissue. Thus, although DP-1 activity does not seem to be required for normal implementation of all differentiation programs, it is required for normal proliferative responses in epidermal keratinocytes, which, in turn, affect the ability of these cells to form a three-dimensional tissue.