The proximal promoter of the human transglutaminase 3 gene. Stratified squamous epithelial-specific expression in cultured cells is mediated by binding of Sp1 and ets transcription factors to a proximal promoter element.

The transglutaminase 3 enzyme is expressed during the late stages of the terminal differentiation of the epidermis and in certain cell types of the hair follicle. The enzyme is thought to be critically involved in the cross-linking of structural proteins and in the formation of the cornified cell envelope, thereby contributing to rigid structures that play vital roles in shape determination and/or barrier functions. To explore the mechanisms regulating the expression of the transglutaminase 3 gene (TGM3), 3.0 kilobase pairs of sequences upstream from the transcription start site were assessed for their ability to control the expression of a chloramphenicol acetyltransferase reporter gene. Deletion analyses in transiently transfected epidermal keratinocytes defined sequences between -126 and -91 as the proximal promoter region of the gene, and which can confer epithelial-specific expression to the TGM3 gene in vitro. Mutation and DNA-protein binding analyses indicated that a complex interaction between adjacent Sp1- and ets-like recognition motifs with their cognate binding factors is required for the function of the TGM3 promoter. As these TGM3 sequences can confer promoter/enhancer activity to reporter genes at a level comparable to the powerful SV40 promoter, they may be useful for gene therapy in keratinocytes.

The transglutaminase 3 enzyme is expressed during the late stages of the terminal differentiation of the epidermis and in certain cell types of the hair follicle. The enzyme is thought to be critically involved in the crosslinking of structural proteins and in the formation of the cornified cell envelope, thereby contributing to rigid structures that play vital roles in shape determination and/or barrier functions. To explore the mechanisms regulating the expression of the transglutaminase 3 gene (TGM3), 3.0 kilobase pairs of sequences upstream from the transcription start site were assessed for their ability to control the expression of a chloramphenicol acetyltransferase reporter gene. Deletion analyses in transiently transfected epidermal keratinocytes defined sequences between ؊126 and ؊91 as the proximal promoter region of the gene, and which can confer epithelial-specific expression to the TGM3 gene in vitro. Mutation and DNA-protein binding analyses indicated that a complex interaction between adjacent Sp1-and ets-like recognition motifs with their cognate binding factors is required for the function of the TGM3 promoter. As these TGM3 sequences can confer promoter/enhancer activity to reporter genes at a level comparable to the powerful SV40 promoter, they may be useful for gene therapy in keratinocytes.
The role of each of these enzymes in the fate of differentiating epidermal and hair follicle cells is not yet clear. The TGase 2 enzyme has been implicated in apoptosis, cell adhesion, and signal transduction (1)(2)(3)32). The TGase 1 and TGase 3 enzymes are thought to be required for the orderly assembly of specific structural proteins to form a specialized structure termed the cornified cell envelope which provides vital barrier functions for the organism (1)(2)(3)(33)(34)(35)(36)(37). In addition, the TGase 3 enzyme is thought to be required for the cross-linking of the structural protein trichohyalin and the keratin intermediate filaments to form a rigid structure within the inner root sheath cells (38 -42), and thereby participate in shape determination of the developing hair cortical cells internal to the sheath structure (38,39,43). In the medulla cells internal to the hair fiber, the trichohyalin is cross-linked to itself to form an insoluble vacuolized lattice-like structure which in turn is essential entrapment of air for thermal regulation in mammals (38,43). The involvement of TGases in cell envelope formation is supported by three types of observations. First, a large body of in vivo studies have documented that the TGases and the cell envelope structural proteins are co-expressed (reviewed in Refs. 1-3 and 33). Second, many in vitro cross-linking studies have shown that the cell envelope proteins or model peptides derived from them are efficiently used as substrates by these enzymes (33,34,44,45). Third, very recent studies have shown that mutations in the TGM1 gene that result in an inactive TGase 1 enzyme are the cause of lamellar ichthyosis, an autosomal recessive disorder of the cornification (46 -48). Thus it is quite likely that mutations in the TGM3 gene will also cause autosomal recessive ichthyosis-like diseases (46,49).
The ability of the three epidermally expressed TGases to cross-link the same substrates, although with different efficiencies (44), creates a potential to provide for a missing TGase activity by appropriately modulating the expression of the two other enzymes. Hence, it is imperative to understand the mechanisms that control and modulate the expression of these TGase genes in the epidermis. TGM1, TGM2, and TGM3 genes have distinctly different patterns of expression. TGM2 is expressed in a variety of tissues (18 -21, 32) but in the epidermis it is largely restricted to proliferative basal keratinocytes (17). The expression is induced both in vivo and in vitro by retinoids (50), most probably through the retinoic acid receptor RAR␣dependent signaling pathway (51). The retinoic acid-induced TGase 2 activity is inhibited by the phorbol ester TPA and to a lesser extent by Ca 2ϩ (50). TGM1 is expressed in all tissues of epithelial origin (4 -16). Its transcription is regulated negatively by retinoids and is induced by TPA and calcium, primarily in suprabasal cells (28,50,52,53). TGM3, as far as is currently known, is expressed only during the last stages of terminal differentiation of the epidermis and epidermal appendage cell types such as the inner root sheath and medulla of the hair follicle (22)(23)(24)(25)(26)(27)(28)(29)(30)(31). Its expression is initiated well after the transcription of the genes encoding the earlier differentiation markers, such as the keratin proteins K1 and K10, have been repressed (17) and approximately coincident with expression of the profilaggrin (54) and loricrin (55) genes. Although TGase 3 mRNA represents less than 2% of the TGase transcripts, the activated TGase 3 accounts for up to 75% of the total TGase activity in mammalian epidermis (17). In submerged cultures of undifferentiated normal human epidermal keratinocytes (NHEK) the abundance of TGase 3 mRNA is greatly diminished compared to the foreskin epidermis. In contrast, TGase 1 and TGase 2 mRNA levels are increased. Induction of NHEK cell differentiation with Ca 2ϩ leads to an increase in the TGase 1 and TGase 3 mRNA expression, whereas TGase 2 mRNA is down-regulated (28).
To date, very little is known about the recognition elements and the protein factors that regulate the transcription of these three TGase genes. In a recently published study on TGM 2 gene (56), about 1.6 kb of the 5Ј-upstream region conferred low levels of constitutive activity that could not be modulated by retinoids, so that the sequences which control the high level and the retinoic acid inducibility of TGM2 transcription must be located elsewhere. In the case of the TGM1 gene, 0.82 kb of upstream sequences were found to induce expression in epithelial cells in the presence of TPA, which could be suppressed by retinoids or protein kinase C (57). Co-transfection experiments indicated that c-jun and c-fos transcription factors were involved, presumably through an AP-1 site in this region, but none of the regulatory sequences have been defined functionally. In the case of the TGM3 gene, no data are available on the factors which control its expression.
The aim of this study has been to explore the mechanisms which regulate the expression of the human TGM3 gene. As an initial step, we have defined sequences in the vicinity of the mRNA start site which provide high promoter activity and which restrict expression to epithelial cells in vitro. Our data indicate that this epithelial-specific activity is provided by cooperative interactions between Sp1 and ets transcription factors.

MATERIALS AND METHODS
All recombinant DNA technology was done according to standard procedures (58,59).
Construction of Recombinant Clones-Previously, we have isolated a genomic clone gTGM3-6 which extends about 3 kb above the functional transcriptional start site of the human TGM3 gene (30). In order to identify sequences which confer TGM3 promoter/enhance function, recombinant plasmids were constructed by subcloning various portions of these sequences into the HindIII/XbaI sites of the reporter vector pCAT-Basic (Promega Corp., Madison, WI) which contains the chloramphenicol acetyltransferase (CAT) gene but does not contain any regulatory elements. For generation of the TGM3 fragments, the following forward primers (5Ј to 3Ј notation) were used in a polymerase chain reaction with gTGM3-6 as template (HindIII and XbaI sites are shown in lower case letters): Ϫ3050: aagcttGCCCTACTGCTGGTCAG; Ϫ821: aagct-tCAGTGAGGGTCAGTG; Ϫ280: aagcttCCTACAATCCAGGA; Ϫ130: aagcttACAGGCACTACAGG; Ϫ126: aagcttCAGGCACTACAGGAATG; Ϫ116: aagcttAGGAATGACCTGGTGCC; Ϫ106: aagcttGGTGCCTCGC-CCACT; and Ϫ91: aagcttCATTAGAATTCTAAT. All utilized at the 3Јend the reverse primer ϩ10 to Ϫ7: tctagaCTCTGGGAATGGCACGG. The polymerase chain reaction products were gel-purified, digested, and subcloned into the HindIII/XbaI sites of pCAT-Basic vector. The nucleotide sequence of the inserts was verified by dideoxynucleotide chain termination sequencing with a Sequenase kit 2.0 (U. S. Biochemical Corp.).
Two other commercial vectors were used for comparisons with the eight pCAT-Basic constructs. These were the pCAT-Promoter and pCAT-Control vectors (Promega), which are the same as the pCAT-Basic vector except that in the former the CAT gene is driven by the powerful SV40 promoter, and the latter contains SV40 early promoter and enhancer sequences.
In addition, we used a second heterologous promoter derived from the herpes simplex virus thymidine kinase gene. Sequences between Ϫ50 and ϩ50 of the thymidine kinase gene from the pBLCAT2 vector (60) were used to construct a ptk-Promoter vector. TGM3 gene sequences between Ϫ126 and Ϫ73, derived from synthetic oligonucleotides, were inserted in front of this minimal promoter. The sequence of the resulting construct ptk-TGM3 was verified by sequencing.
Cell Cultures, Transfections, and Protein Assays-Cryopreserved NHEK were obtained from Clonetics (San Diego, CA) and grown in calf skin collagen (Sigma) coated dishes in serum-free keratinocyte growth medium (KGM, Clonetics) at 0.05 mM Ca 2ϩ , supplemented with 60 g/ml bovine pituitary extract. Third passage NHEK cells were used for transfection experiments, and preparation of nuclear extracts. A431, Cos-7, HeLa, MCF-7, HepG2, and NIH-3T3 cells were purchased from the American Tissue Culture Collection (ATCC, Rockville, MD) and were grown and maintained following the recommended procedures. HaCaT cells were a gift from Dr. Norbert E. Fusening, and were grown in Dulbecco's modified Eagle's medium supplemented with 4.5 g/liter glucose, 10% fetal bovine serum, and non-essential amino acids (Life Technologies, Inc., Bethesda, MD). Neuroblastoma cells (SK-N-AS) were a gift from Dr. Carol Thiele and were grown in RMPI 1640 medium supplemented with 10% fetal calf serum (Life Technologies, Inc.).
Transient transfections were performed in duplicate using Lipofectin reagent (Life Technologies) for NHEK, HeLa, A431, HepG2, and MCF-7 cells, or DOTAP (Boehringer Meinheim) for all other cells lines following the manufacturer's recommendations. Typically, 2-3 ϫ 10 5 cells were plated in 6-well culture plates 16 -20 h before transfection. Transfections were done when cultures reached 60 -70% confluency. Transfection efficiencies were always monitored by use of a thymidine kinase ␤-galactosidase construct (tk-␤-gal) (Clontech, Palo Alto, CA). For Lipofectin transfections, cell cultures were washed once at 37°C with phosphate-buffered saline, and then were preincubated for 30 min at 37°C with either Keratinocyte-SFM (Life Technologies, Inc.) for NHEK cells, or Opti-MEMI media (Life Technologies, Inc.) for HeLa, A431, and HepG2 cells. For each well, 1.5 g of reporter plasmids and 0.5 g of tk-␤-gal was mixed with 6 g of Lipofectin, and incubated for 20 min at room temperature. The lipid/DNA mixture was then added into each well and incubated for 16 -18 h in the case of NHEK and A431 cells, or 3-4 h in the case of HeLa and HepG2 cells. At the end of the transfection period the medium was replaced with the medium in which the cells normally grow, and for NHEK cells the concentration of the Ca 2ϩ in the KGM medium was adjusted to 1.2 mM. Cells were harvested 50 -60 h post-transfection. For DOTAP transfections, the plasmids (1.5 g of reporter constructs and 0.5 g of tk-␤-gal) were mixed with 14 g of DOTAP in HEPES buffer solution (pH 7.3) and incubated for 20 min at room temperature. The transfections were carried out in the medium of each cell type for 16 -18 h. The media were then replaced and cells cultured for another 50 -60 h. Some NHEK cultures were co-transfected with the TGM3 Ϫ126/ϩ10 sequences with the pECE vector or this vector containing mouse ets-2 cDNA (gift of Dr. Richard Maki, La Jolla Cancer Research Foundation). Similar co-transfection experiments used the human pRSV-Sp1 cDNA or the RSV-vector alone (gift of Dr. Robert Tjian, University of California, Berkeley).
Cellular extracts were prepared through at least three freeze-thaw cycles as described (61). Aliquots were used for CAT assays, ␤-gal assays, and total protein quantitation (62). Cellular extracts of untransfected cells and of cells transfected with the pCAT-Basic vector alone without the TGM3 inserted sequences were used as negative controls, while the pCAT-Promoter and pCAT-Control vectors served as positive controls. CAT activities were determined (63) using chloramphenicol and [ 3 H]acetyl-CoA (DuPont NEN) as substrates. ␤-Gal activity was assayed by using a commercial enzyme assay system (Promega). The values for CAT were normalized by protein content and ␤-gal activity. The relative CAT values are the average of at least three independent experiments, each with duplicate samples.
Nuclear Extracts and Mobility Shift Assays-Nuclear extracts were prepared according to Schreiber et al. (64) with slight modifications. The cell pellets were resuspended and left to swell on ice for 15 min in ice-cold buffer containing 10 mM HEPES (pH 7.9), 10 mM KCl, 0.1 mM EGTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and a mixture of protease inhibitors which included 5 g/ml benzamidine, 5 g/ml pepstatin, 5 g/ml leupeptin, and 5 g/ml aprotinin. Nonidet P-40 was then added to 0.6% and the suspension was homogenized with 20 strokes in a tightly fitting glass homogenizer. After centrifugation, the nuclear pellets were resuspended in ice-cold buffer containing 20 mM HEPES (pH 7.9), 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 20% glycerol, and the mixture of protease inhibitors, and vigorously shaken at 4°C for 15 min. The nuclear debris was discarded by centrifugation for 15 min at 14,000 ϫ g and the extracts were aliquoted and stored at Ϫ70°C until used.
Mobility shift experiments were performed with 5 g of nuclear extracts and 2-4 ϫ 10 4 cpm (about 1 ng) of gel-purified 5Ј end-labeled double-stranded oligonucleotides containing the desired TGM3 sequences (see figures for sequences). Typically, the binding reactions were carried out in 20 l containing 10 mM Tris-HCl (pH 7.5), 65 mM NaCl, 5 mM dithiothreitol, 5 mM MgCl 2 , 0.05% Nonidet P-40, 10% glycerol, 1 mg/ml bovine serum albumin, and 25 g/ml poly(dI-dC) as a carrier for 30 min at 4°C. In competition experiments, a 100-fold molar excess of the cold competitor was preincubated with the extracts for 30 min at 4°C before the labeled DNA fragment were added. Irrelevant control competitor oligonucleotides for consensus AP1 and AP2 sequences (Santa Cruz Biotechnology) were also used. Recombinant human Sp1 protein was from Promega and used as per the manufacturer's recommendations. The complexes were resolved on nondenaturing 6% polyacrylamide gels in 0.5 ϫ TBE buffer for 1 h at 14 V/cm, and viewed following overnight autoradiography.

RESULTS
The Proximal Promoter of the TGM3 Gene is Highly Active in NHEK Cells-It has been previously shown (15) that NHEK cells grown in 1.2 mM Ca 2ϩ express the TGM3 gene, although at a lower level than in the epidermis. Therefore these cells can be used as an in vitro system to explore the transcriptional control of the TGM3 gene. Recently the TGM3 gene was cloned and the primary transcript was characterized (30). In this study we concentrated our interest on the sequences upstream of the transcription initiation site. Several fragments of the 5Ј-region were subcloned into a promoterless vector containing the CAT gene (pCAT-Basic), of which eight (Fig. 1A) proved to be informative. The constructs were analyzed in transiently transfected NHEK cells grown in media containing 0.05 or 1.2 mM Ca 2ϩ . The sequences located in the vicinity of the transcription initiation site (construct Ϫ126/ϩ10) showed an activity similar to that conferred by the powerful SV40 promoter/enhancer region in pCAT-Control vector (Fig. 1B). There was no significant difference in the CAT activities in cells cultured under conditions promoting differentiation (1.2 mM Ca 2ϩ ) (Fig.  1B) or proliferation (0.05 mM Ca 2ϩ ) (data not shown). The level of expression of the constructs containing the sequences upstream of position Ϫ126 was reduced to equal or less than the activity of pCAT-Promoter construct. Thus the positive effect of the proximal promoter region was markedly reduced by the presence of upstream elements. The sequences extending up to position Ϫ3050 were unable to overcome this negative effect.
The Activity of TGM3 Proximal Promoter Is Specific to Stratified Squamous Epithelial Keratinocytes-To explore the cell type specificity, the regulatory potential of TGM3 proximal promoter sequences was also analyzed in a variety of other cell types (Table I)  squamous epithelial tissues A431 and HeLa; two cell lines derived from simple epithelial cells MCF-7 and HepG2; the fibroblast cell lines NIH 3T3 and Cos-7; and the SK-N-AS neuroblastoma cell line. The level of expression of the Ϫ91/ϩ10 TGM3 construct was low in all cell types tested. However, the Ϫ126/ϩ10 TGM3 construct showed high levels of expression in the two epidermal and two stratified squamous epithelial cell lines tested, at levels Ն10 times that of the Ϫ91/ϩ10 construct. In contrast, its expression level in the two cell lines derived from simple epithelial tissues was very low, comparable to that seen in the two other non-epithelial cell types tested. These data suggest the sequences between Ϫ91 and Ϫ126 confer stratified squamous epithelial specificity of expression to the TGM3 gene, at least in vitro, under the experimental conditions defined here. On the other hand, the expression of the Ϫ130/ϩ10 TGM3 construct was much lower than the Ϫ126/ ϩ10 construct in all the epithelial cells tested, indicating the presence of a silencer which may be epithelial specific.
DNA-Protein Interactions in the Proximal Promoter Region of the TGM3 Gene-To define the regulatory sequences which confer the cell type specific expression to the TGM3 proximal promoter, we explored the DNA-protein interactions at a fragment encompassing the sequences between Ϫ186 and ϩ10 of the TGM3 5Ј-flanking region.
In a pilot DNase I footprinting experiment using keratinocyte nuclear extract we detected two weak footprints on both DNA strands within the TGM3 promoter (data not shown). The protected region which extended from Ϫ73 to Ϫ102 contained an Sp1-like binding site. The second protected region was located between positions Ϫ111 and Ϫ128 and contained two direct copies of the sequence (T/C)TACAGG(C/A)A, which encompass ets-like recognition motifs.
Mobility shift experiments were aimed at characterizing in more detail these DNA-protein interactions. For each labeled probe the specificity of the binding was established in competition experiments with a 100-fold molar excess of either the corresponding unlabeled oligonucleotides (specific competitors), or of poly(dI-dC) or unrelated oligonucleotides as unspecific competitors.
In the presence of a 100-fold molar excess of poly(dI-dC) (lane 1) or of the two irrelevant oligomers, consensus AP1 (lane 6) or consensus AP2 (lane 10) sequences, five retarded complexes were discerned. Specific competition with the unlabeled probe completely prevented the formation of complexes A, B, and C (lane 2). Competition experiments with wild type oligonucleo-tides derived from the probe revealed that complexes A and C involved the sequences of the DNase I protected region which encompassed the Sp1 recognition site (lanes 3 and 7), whereas complex B originated from interactions in the ets-like protected region which extended between Ϫ111 and Ϫ118 (lanes 4 and 8). The contribution of the putative Sp1-and ets-recognition motifs to the formation of these complexes was evaluated in competitive binding with mutant variants of the corresponding sequences. Mutations in the Sp1 binding motif (lanes 4 and 5) prevented competion for complexes A and C. Likewise, mutations in the core ets motif (lanes 7 and 9) strongly interfered with the ability of the mutant oligonucleotide to compete for complex B formation. However, a 100-fold molar excess of the oligonucleotides containing the intact Sp1-binding motif were not able to compete for ets binding (lanes 3 and 7) and, likewise, ets oligomers did not affect the formation of the Sp1 complexes (lanes 4 and 8). Two other bands marked with arrowheads were relatively resistant to self-competition (lane 2) and were formed both with double-and single-stranded oligonucleotides (data not shown), and have not been investigated further.
Binding to the Sp1 Motif-Because the Sp1 complex C was always weaker than complex A, additional experiments were performed to explore whether complexes A and C were due to multicomponent interactions in the NHEK nuclear extracts, or due to protein degradation. First, a probe spanning the sequences between Ϫ105 and Ϫ70 encompassing the Sp1 motif was used (Fig. 3A). Two retarded bands with mobilities corresponding exactly to complexes A and C of Fig. 2 were observed, but in this case of equal intensity (Fig. 4A, lane 1). While an

TABLE I Expression of TGM3 constructs in different cell types
The relative CAT activities of the three TGM3 constructs were normalized with respect to the activity of the tk-␤-gal construct and are expressed as multiples of the activity of the pCAT-Promoter construct. The pCAT-Promoter activity was comparable in all cell types studied.  Likewise, incubation of the binding reactions with an Sp1 specific antibody prevented the formation of both complexes (data not shown). Second, we incubated the same probe with recombinant Sp1 protein. In this case, a single retarded band which migrated like complex C was observed (Fig. 3B, lane 1). This recombinant Sp1 complex was successfully competed by the oligonucleotide containing the wild type Sp1 motif of the TGM3 promoter (lane 2), while the mutant (lane 3) and the deletion oligonucleotide (lane 4) could not compete.
Thus, whereas the Sp1 motif interacted with purified recombinant Sp1 protein to form the faster migrating complex C only, the interaction with the multicomponent NHEK extract resulted in the formation of both complex C and the slower migrating complex A. Moreover, when a probe carrying only the Sp1 motif was used, complexes A and C were of equal intensity (Fig. 3A), but when a longer probe carrying both the Sp1 and ets-like motifs was used, the intensity of complex C was always weaker (Fig. 2). Taken together these data suggest that complex C was not due to protein degradation of complex A, but rather, results from interactions of the Sp1 motif with Sp1 transcription factor alone. Complex A is likely to be due to a multicomponent interaction involving both Sp1, ets and/or other as yet unidentified proteins.
Binding to the ets Motif-To define precisely the nucleotides involved in the ets binding, the wild type and a series of mutant variants of the double stranded sequence between Ϫ134 and Ϫ102 were used in mobility shift experiments. A strong specific band, corresponding to the ets-containing complex B was discerned (Fig. 4, lane 1). To determine which of the two ets-like repeats was responsible for the binding, competition with oligomers spanning overlapping portions of the initial probe were used. Of those, complex B was successfully competed by a 100-fold molar excess of the oligomers containing the wild type ets recognition motif between Ϫ111 and Ϫ118 (lanes 3 and 4), but not by an oligonucleotide spanning the sequences of the ets-like motif between Ϫ119 and Ϫ127 (lane 2). Mutations in any of the nucleotides of the sequence ACAGGAAT, encompassing the core ets motif, compromised the interaction with the nuclear protein(s) (lanes 5 and 7-12), whereas mutations in the adjacent nucleotides (lanes 6 and 13) did not have an effect. Accordingly, only the sequence between Ϫ111 and Ϫ118 represents the functional ets binding site in the TGM3 promoter region. Comparison with the known ets binding motifs (65) indicates that this recognition sequence is most similar to the binding site for ets-2, but another ets-like factor with a similar binding specificity cannot be excluded. In this regard, an at-  2-13). The arrow denotes the ets-specific complex B. The letters in bold are the nucleotides in or adjacent to the ets motif (boxed) that were mutated; dots represent unchanged nucleotides; the arrowheads mark the complexes of unknown origin; P, position of the free probe.
tempt was made to elucidate the nature of the ets protein(s) in complex B. The binding reactions were performed in the presence of antibodies recognizing either specific ets-domain transcription factors or broadly reactive with many of them. Under various experimental conditions, we were not able to detect interference with the binding or the mobility (data not shown). However, a report in another gene system (66) has shown that these antibodies do not always interfere with the formation of transcriptionally active ets complexes. In co-transfection experiments, these authors (66) established that an ets regulatory motif was interacting with the factor ets-2 and that this interaction, in conjunction with an AP1 binding activity, was essential in conferring a TPA inducibility to the macrophage scavenger promoter. Incubation of an oligonucleotide carrying the crucial AP1/ets binding sequences with the same antibodies used in our study, failed to interfere with the binding pattern. Whether this is due to the nature of the antibodies or is an indication of a specific configuration of the DNA-protein complexes is not clear.
Functional Analyses of the Proximal Promoter in Vivo-These mobility shift experiments revealed that several DNAprotein complexes could be formed in vitro over the sequences of the proximal promoter region. To explore whether the observed DNA-protein binding affects the function of the TGM3 promoter in vivo, we assessed a series of deletion and mutant fragments (Fig. 5A) encompassing the proximal promoter region. The activity of the resulting constructs was tested by CAT assays in transiently transfected keratinocytes (Fig. 5B). The highest activity was detected with construct Ϫ126/ϩ10, which contained both ets and Sp1 binding sites. The activity of the construct containing only the Sp1 site (Ϫ106/ϩ10) amounted to 12% of the activity of construct Ϫ126/ϩ10 and, therefore, an intact Sp1 site alone was not enough to ensure high levels of expression. Its presence, however, was indispensable, since altering the nucleotides crucial for binding of the Sp1 recognition motif to NHEK nuclear protein(s) (Fig. 5B, mutations A and B) reduced the transcription from the otherwise highly active constructs Ϫ116/ϩ10 (to Ͻ15%) and Ϫ126/ϩ10 (to Ͻ40%). Likewise, mutations in the ets binding motif were deleterious for the promoter function (construct 116C to Ͻ20%; construct 126C to Ͻ5%). Moreover, the effect of these DNAprotein interactions was obviously synergistic: mutations in either Sp1 or ets binding sites essentially abolished the transcription, and simultaneous mutations in both recognition sequences did not have a greater effect (constructs 116D, 126D). However, a C to A mutation in the ets-like motif between Ϫ119 and Ϫ127 (construct 126E) showed a level of expression comparable to the wild type construct, thus confirming the observation from the gel shift experiments that this ets-like sequence may not be involved in the regulation of the TGM3 gene.
The involvement of ets and Sp1 transcription factors in the control of TGM3 promoter activity was further confirmed in co-transfection experiments. Simultaneous introduction into NHEK cells of the TGM3 construct Ϫ126/ϩ10 with the pECE expression vector containing mouse ets-2 cDNA resulted in a 100% increase in the level of CAT compared to the level obtained on co-transfection of the same TGM3 construct with the pECE expression vector alone (Fig. 6). Co-transfection of the TGM3 construct with an Sp1 expression vector did not alter the activity of CAT, presumably because of high endogenous levels of this transcription factor in keratinocytes (67). However, a further 30% increase in expression was obtained when vectors encoding both the ets-2 and Sp1 factors were simultaneously co-transfected (Fig. 6), thereby supporting the notion that ets-2 and Sp1 factors cooperate to regulate the expression of the TGM3 gene.
TGM3 Sequences between Ϫ126 and Ϫ91 Constitute a Powerful Promoter/Enhancer-The data of Fig. 1 demonstrate that the TGM3 sequences between Ϫ126 and ϩ10 confer high constitutive activity in a pCAT-Basic construct in epithelial cells. The data of Fig. 5B show that this promoter activity is confined to sequences between Ϫ126 and Ϫ91. To establish whether these sequences can also serve as an enhancer for a heterologous promoter, they were cloned into a CAT reporter vector (ptk-Promoter) containing the minimal thymidine kinase promoter (see "Materials and Methods"). In transiently transfected NHEK cells this construct (ptk-TGM3) showed about a 4-fold increase in the level of CAT activity compared to the activity of the ptk-Promoter (Fig. 7). By way of comparison, a similar enhancement was exerted in NHEK cells by the SV40 enhancer sequences on the SV40 promoter (Fig. 7, pCAT-Control). These data establish that the TGM3 sequences Ϫ126 to Ϫ91 are capable of enhancing the transcription from a heterologous promoter and in vitro, in keratinocytes, are as powerful as the potent SV40 enhancer. DISCUSSION Our study demonstrates that the correct transcription of the human TGM3 gene, at least in vitro, depends on the simultaneous effect of elements residing in both the proximal promoter and in regions distal from it. The distal elements which are not present within the region extending 3.05 kb upstream of the transcription initiation site are required for late epidermal differentiation-specific transcription. The proximal promoter of the TGM3 gene, however, contains the information sufficient to direct high levels of expression in stratified squamous epithelial cells in culture. We have mapped this region to sequences between Ϫ126 and Ϫ91, which encompass Sp1 and ets-like binding sites. The overall activity of the proximal promoter region results from the cooperative interactions between these positively acting motifs.
Keratinocytes contain relatively high levels of Sp1 (67). This transcription factor is important for the regulation of a variety of other epidermally-expressed genes (68 -70). On the other hand, the TGM3 gene is the first epidermally-expressed gene in which the ets transcription factors have been shown to be important. These transcription factors have been implicated in the regulation of gene expression during a variety of biological processes, including growth control, developmental or transformation programs, and usually function as components of larger transcription complexes (65). Indeed, our transient transfection data (Figs. 5 and 6) demonstrated cooperative interactions between ets and Sp1 transcription factors, as has been previously reported in other gene systems (70 -73). However, in the TGM3 gene proximal promoter region, the interactions clearly involved not only Sp1, but also additional proteins in the NHEK nuclear extracts (Fig. 3). Thus the cooperative relationship between the ets and Sp1 factors in the activity of the TGM3 proximal promoter is modulated by interactions with additional so far unidentified nuclear proteins.
Our present data on the expression of the TGM3 gene in keratinocytes represent the first detailed study on the regulatory elements involved in the transcription of either of the three transglutaminase genes expressed in epithelia. To date in the TGM1 gene, a 0.82-kb fragment could confer epithelial specific expression only in TPA-treated cells (57). In the case of the TGM2 gene, an initial study showed that 1.6 kb of flanking DNA sequences contain low core promoter activity in both epithelial and non-epithelial cell lines, at a level comparable to our construct Ϫ91/ϩ10 ( Fig. 1; Table I). Deletion analyses indicated that putative Sp1 binding motifs may be responsible in the TGM2 gene (56). In this regard we have examined the upstream sequences of the three transglutaminase genes for common regulatory elements. All three contain consensus Sp1 recognition motifs near the transcription start site. Interestingly, the TGM1 gene 2 also possesses an ets-like motif at position Ϫ190. Thus it will be interesting to determine whether the Sp1 and ets motifs can also confer high levels of epithelial specific expression to the TGM1 gene. Furthermore, it raises the possibility that members of the ets family of transcription factors may be important in the regulation of late differentiation genes in the epidermis.
To date, only two of the several genes involved in the latest stages of epidermal differentiation have been studied, the TGM3 gene, as reported here, and the loricrin gene. About 2.5 kb of upstream sequences are required for epithelial specific expression of the loricrin gene (74), and 10 kb are required for correct temporal expression (74,75). The proximal promoter region which confers high levels of epithelial specific expression of the TGM3 gene in cultured keratinocytes is located within a narrow window of Ϫ126 to Ϫ91 base pairs of the transcription start site. This promoter should now be very useful for further studies on the expression of this and other genes in keratinocytes. Furthermore, it may aid in devising strategies to manage lamellar ichthyosis (46 -48) and other related recessive genodermatoses involving mutations in the transglutaminase genes. FIG. 7. The proximal promoter element located between ؊126 and ؊91 of the TGM3 gene confers high activity to a heterologous promoter in NHEK cells. The TGM3 sequences between Ϫ126 and Ϫ91 were cloned into a CAT reporter vector (ptk-TGM3) based on the ptk-Promoter (as described under "Materials and Methods"). The data are expressed as percentages of the activity of the ptk-Promoter and are the averages of three independent experiments.