Expression of the α5 Integrin Subunit Gene Promoter Is Positively Regulated by the Extracellular Matrix Component Fibronectin through the Transcription Factor Sp1 in Corneal Epithelial Cells in Vitro *

The accumulation of fibronectin (FN) in response to corneal epithelium injury has been postulated to turn on expression of the FN-binding integrin α5β1. In this work, we determined whether the activity directed by the α5 gene promoter can be modulated by FN in rabbit corneal epithelial cells (RCEC). The activity driven by chloramphenicol acetyltransferase/α5 promoter-bearing plasmids was drastically increased when transfected into RCEC grown on FN-coated culture dishes. The promoter sequence mediating FN responsiveness was shown to bear a perfect inverted repeat that we designated the fibronectin-responsive element (FRE). Analyses in electrophoretic mobility shift assays provided evidence that Sp1 is the predominant transcription factor binding the FRE. Its DNA binding affinity was found to be increased when RCEC are grown on FN-coated dishes. The addition of the MEK kinase inhibitor PD98059 abolished FN responsiveness suggesting that alteration in the state of phosphorylation of Sp1 likely accounts for its increased binding to the α5 FRE. The FRE also proved sufficient to confer FN responsiveness to an otherwise unresponsive heterologous promoter. However, site-directed mutagenesis indicated that only the 3′ half-site of the FRE was required to direct FN responsiveness. Collectively, binding of FN to its α5β1 integrin activates a signal transduction pathway that results in the transcriptional activation of the α5 gene likely through altering the phosphorylation state of Sp1.

Corneal wounds account for a substantial proportion of all visual disabilities and medical consultations for ocular problems in North America. They can be superficial with damage limited to the epithelium or associated with a deeper involvement of the epithelial basement membrane and of the stromal lamella. Severe recurrent and persistent corneal wounds are most commonly secondary to ocular diseases and damage such as recurrent erosion, mild chemical burns, superficial herpetic infections, neuroparalytic cornea, autoimmune diseases, and stromal ulcerations due to viral or bacterial infections or to severe burns (1). Despite currently available treatments, many of these corneal wounds persist for weeks and months or else recur frequently and can progress to corneal perforation.
Tissue repair requires cell migration, proliferation, and adhesion. Cell adhesion and migration in turn require extracellular matrix (ECM) 1 synthesis and assembly. ECM is a complex, cross-linked structure of proteins and polysaccharides. It organizes the geometry of normal tissues. Fibronectin (FN) is an ECM adhesion protein identified as a potential wound healing agent because of its cell attachment, migration, differentiation, and orientation properties (for a review see Refs. [2][3][4]. In the unwounded rat eye, FN is observed by immunohistological staining at the level of the corneal epithelium basement membrane (5)(6)(7). Shortly after corneal injury, the basal cells that border the injured area and stromal keratocytes start producing massive amounts of FN (5, 8 -11). FN promotes corneal cell migration both in vivo (12,13) and in vitro (14) by acting as a temporary extracellular matrix to which corneal epithelial cells attach as they migrate over the wounded area (13,15). Once the wound is re-epithelialized, the subepithelial immunohistological staining of FN progressively decreases (5, 16 -18).
The increase in FN expression that has been reported to occur during corneal wound healing was postulated to be coordinated with the expression of its major integrin receptor ␣ 5 ␤ 1 (5), as has also been shown for laminin and tenascin and their corresponding integrin receptor subunits ␣ 6 and ␣ 9 , respectively (19 -21). For instance, the integrin ␣ 5 ␤ 1 was shown to be present during corneal wound healing after radial keratectomy (22). Direct evidence that FN can positively alter ␣ 5 ␤ 1 integrin expression at both the protein and mRNA levels has been provided through FN antisense expression studies performed in the epithelium-derived human colon carcinoma cell line Moser (23) as well as in murine AKR-2B fibroblasts (24). Other indirect evidence linking expression of ␣ 5 ␤ 1 to that of FN has also emerged from recent studies (25,26).
As a consequence, it is not surprising that ECM, through its interactions with membrane-bound integrins, exerts profound influences on the major cellular program of growth, differentiation, and apoptosis by altering, through a number of signal transduction pathways, the transcription of genes whose specific functions are linked to these cellular functions. Binding of ECM components, such as FN, with their corresponding integrin receptors will trigger the activation of intracellular signaling mediators such as focal adhesion kinase, mitogen-activated protein kinases (MAPKs), and Rho family GTPases (for a review see Ref. 27). Activation of the MAPK signal transduction pathway is of particular interest since it links integrin-mediated signaling to transcriptional regulation of genes that are crucial for cell growth and differentiation. The results presented hereby provided evidence that, by acting on ␣ 5 gene expression, such a route of signal transduction might alter cell adhesion properties as well. The downstream cascade of family members that are activated following transient activation of Ras GTP-binding proteins through receptor tyrosine kinases include MAPK/ERK kinase (designated MEK) and ERK1 (p44)/ ERK2 (p42) (28). Activation of ERK1/ERK2 through phosphorylation causes their translocation to the nucleus, where they have been reported to phosphorylate and activate distinct transcription factors, such as ELK, c-Jun, and c-Myc (29 -31), as well as members of the ETS family (such as PEA3) (32).
In the present study, we demonstrated that FN can alter the transcription of the ␣ 5 integrin subunit gene at the promoter level. Such a FN responsiveness was shown to be determined by the binding of the transcription factor Sp1 to a target site that is part of a perfect inverted repeat which, by itself, can confer FN responsiveness to an otherwise unresponsive heterologous promoter. Most of all, the FN-activated, integrin-mediated signal transduction pathway appears to require activation of ERK1/ERK2 since the Sp1 DNA binding affinity, and, as a consequence, the FN responsiveness of the ␣ 5 promoter were both found to be diminished by blocking their activation with the MEK kinase inhibitor PD98059. Together, these results demonstrate the novel finding that the ␣ 5 integrin subunit, through activation of the MAPK pathway, can autoregulate its own synthesis in a manner that is dependent on the extracellular concentration of FN.

EXPERIMENTAL PROCEDURES
Cell Culture and Media-Rabbit corneal epithelial cells (RCECs) were obtained from the central area of freshly dissected rabbit corneas as described previously (33) and then grown to low (near 15% coverage of the plates), intermediate (near 75% coverage), or high cell density (100% coverage for more than 48 h) under 5% CO 2 in SHEM medium supplemented with 5% FBS and 20 g/ml gentamicin. When indicated, human plasma FN (obtained as described previously (34)) or ECM gel (basement membrane matrice from Engelbreth-Holm-Swarm mouse sarcoma, Fisher) was coated for 18 h at 37°C on the culture dishes at varying concentrations (FN, 1-16 g per cm 2 ; ECM, 10 g per cm 2 ). Coated Petri dishes were washed twice with phosphate-buffered saline and blocked at 37°C with 2% bovine serum albumin in phosphatebuffered saline. Cells were then seeded and grown as above. Inhibition of ERK1/ERK2 was performed by culturing subconfluent RCEC in the presence of 10 M of the MEK/kinase inhibitor PD98059 (Sigma) for 48 h before cells were harvested. Drosophila Schneider cells (ATCC CRL-1963) were cultured at 28°C without CO 2 in Schneider medium (Sigma) supplemented with 10% FBS and 20 g/ml gentamicin.
Plasmids and Oligonucleotides-The plasmids ␣ 5 Ϫ41, ␣ 5 Ϫ92, ␣ 5 Ϫ178, and ␣ 5 Ϫ954, which all bear the chloramphenicol acetyltransferase (CAT) reporter gene fused to DNA fragments from the human ␣ 5 gene upstream regulatory sequence extending up to 5Ј positions Ϫ41, Ϫ92, Ϫ178, and Ϫ954, respectively, but all sharing a common 3Ј end located at position ϩ23, have been described previously (35). The recombinant plasmids bearing one or two sense copies of either the ␣ 5 FRE or its mutant derivatives were created by inserting the corresponding double-stranded oligomers upstream from the basal promoter of the mouse p12 gene (into the unique BamHI site) that has been previously mutated into its Sp1-binding site (and designated p12.108/M (36)). The Sp1 expression vector pPacSp1 was generously provided by Dr. Guntram Suske (Institute fü r Molecular Biology und Tumorforschung, Philipps Universitä t Marburg, Germany), whereas the LacZ expression plasmid pAC5/V5-His/LacZ was obtained from Invitrogen (Carlsbad, CA).
Transient Transfection and CAT Assay-RCEC plated at either low (5 ϫ 10 4 cells per 35-mm tissue culture plates), intermediate (5 ϫ 10 5 cells per 35-mm tissue culture plates), or high (1, 5 ϫ 10 6 cells per 35-mm tissue culture plates) cell density were transiently transfected using the polycationic detergent LipofectAMINE (Life Technologies, Inc.) as recommended by the manufacturer. Each LipofectAMINEtransfected plate received 1.5 g of the test plasmid and 0.5 g of the human growth hormone (hGH)-encoding plasmid pXGH5 (39). Drosophila Schneider cells were transfected according to the calcium phosphate precipitation procedure (36,40) at a density of 1 ϫ 10 6 cells per 60-mm culture plate.
Levels of CAT activity for all transfected cells were determined as described (40) and normalized to the amount of hGH secreted into the culture media and assayed using a kit for quantitative measurement of hGH (Immunocorp, Montréal, Québec, Canada). Because the metallothionein-I promoter, which directs expression of hGH from the pXGH5 plasmid, proved to be highly inefficient in Drosophila cells, CAT activities from transfected Schneider cells were normalized to the amount of ␤-galactosidase encoded by the plasmid pAC5/V5-His/LacZ and cotransfected along with the CAT recombinant constructs. Each cell-containing plate therefore received 15 g of the test plasmid, 4 g of pAC5/V5-His/ LacZ, and 1 g of pPAC (empty vector). In the cotransfection experiments performed with the Sp1 expression plasmid, the empty pPAC CAT activities were measured and normalized to hGH as described under "Experimental Procedures." Each value is expressed as the ratio of the CAT activity from RCEC grown on FN-coated Petri dishes over that of RCEC grown solely on plastic. Standard deviation is provided for each individual value. B, dose-dependent activation of the ␣ 5 promoter FN responsiveness in RCEC. RCEC (3 ϫ 10 5 cells/Petri) were plated on culture dishes that have been coated with either none or increasing concentrations of FN (1-16 g/cm 2 ). RCEC were then transfected with the ␣ 5 Ϫ954 recombinant construct and harvested 48 h later. CAT activity was determined and normalized as described under "Experimental Procedures." Each value is expressed as detailed in A.
was substituted for 1 g of pPacSp1. The value presented for each individual test plasmid transfected corresponds to the mean of at least three separate transfections done in triplicate. To be considered significant, each individual value needed to be at least three times over the background level caused by the reaction buffer used (usually corresponding to 0.15% chloramphenicol conversion). Standard deviation is also provided for each transfected CAT plasmid.
Electrophoretic Mobility Shift Assays (EMSA) and Supershift Experiments-EMSAs were carried out using either the 27-bp ␣ 5 FRE or the high affinity Sp1 oligomer as 5Ј end-labeled probes. Approximately 2 ϫ 10 4 cpm labeled DNA was incubated with crude nuclear proteins (as specified in the figure legends) from RCEC grown on either untreated or FN-coated culture dishes in the presence of 500 ng of poly(dI-dC)⅐poly(dI-dC) (Amersham Pharmacia Biotech) in buffer D (5 mM HEPES (pH 7.9), 10% glycerol (v/v), 25 mM KCl, 0.05 mM EDTA, 0.5 mM dithiothreitol, 0.125 mM phenyl methosulfonyl fluoride). Occasionally, crude nuclear extracts from human HeLa cells were also used in EMSA as a positive control for comparison purposes. Incubation proceeded at room temperature for 10 min upon which time DNA-protein complexes were separated by gel electrophoresis through 6% native polyacrylamide gels run against Tris glycine buffer as described (42). Gels were dried and autoradiographed at Ϫ80°C to reveal the position of the shifted DNA-protein complexes generated. Competitions in EMSA were performed using 10 g of crude nuclear proteins from RCEC grown in the presence of FN at 8 g per cm 2 as above except that molar excesses (100-and 500-fold) of synthetic double-stranded oligonucleotides bearing the DNA sequence of the ␣ 5 FRE, the DNA-binding site for human HeLa CTF/NF-I in adenovirus type 2 (37), the high affinity binding site for the positive transcription factor Sp1 (38), or the p12.A Sp1-binding site from the mouse p12 gene (36) were added to the binding reaction prior to loading on the gel. Supershift experiments in EMSA were conducted by first incubating varying amounts (as specified in the figure legends) of crude nuclear proteins from RCEC grown either with (8 g per cm 2 ) or without FN, in the presence of 250 ng of poly(dI-dC)⅐poly(dI-dC), with either none or 1 l (corresponding to 1 g) of a commercially engineered rabbit antiserum raised against the transcrip-tion factor Sp1 (Santa Cruz Biotechnology, Inc.) in buffer D. Then, 2 ϫ 10 4 cpm FRE-labeled probe was added, and incubation was extended for another 15 min at room temperature. Samples were finally loaded on high ionic strength, 6% native polyacrylamide gels and run at 4°C against Tris glycine buffer as above. Formation of DNA-protein complexes was revealed following autoradiography at Ϫ70°C.

SDS-PAGE and Western
Blot-Crude nuclear proteins were obtained from either HeLa cells (used as a positive control) or from RCEC grown on culture dishes coated or not with FN (8 g per cm 2 ) as detailed above. Protein concentration was evaluated by the Bradford procedure and further validated following Coomassie Blue staining of SDS-polyacrylamide fractionated nuclear proteins. One volume of sample buffer (6 M urea, 63 mM Tris (pH 6.8), 10% (v/v) glycerol, 1% SDS, 0,00125% (w/v) bromphenol blue, 300 mM ␤-mercaptoethanol) was added to 20 g of proteins before they were size-fractionated on a 10% SDS-polyacrylamide minigel and transferred onto a nitrocellulose filter. A full set of protein molecular mass markers (Life Technologies, Inc.) was also loaded as a control to evaluate protein sizes. The blot was then washed once in TS buffer (150 mM NaCl, 10 mM Tris-HCl (pH 7.4)) and 4 times (5 min each at 22°C) in TSM buffer (TS buffer plus 5% (w/v) fat free Carnation milk and 0.1% Tween 20). Then, a 1:500 dilution of a rabbit monoclonal antibody raised against the transcription factor Sp1 (Santa Cruz Biotechnology, Inc.) was added to the membrane-containing TSM buffer and incubation proceeded further for 4 h at 22°C. The blot was then washed in TSM buffer and incubated an additional 1 h at 22°C in a 1:1000 dilution of a peroxidase-conjugated goat anti-mouse immunoglobulin G (Jackson ImmunoResearch). The membrane was successively washed in TSM (4 times, 5 min each) and TS (twice, 5 min each) buffers before immunoreactive complexes were revealed using Western blot chemiluminescence reagents (Renaissance, PerkinElmer Life Sciences) and autoradiographed. A, CAT activity directed by recombinant plasmids bearing 5Ј deletions of the ␣ 5 promoter in RCEC grown with or without FN. RCEC plated at an intermediate cell density (5 ϫ 10 5 cells per Petri) on either regular or FN-coated (2 g/cm 2 ) culture dishes were transiently transfected, 24 h later, with CAT recombinant plasmids bearing various lengths (up to positions Ϫ954, Ϫ178, Ϫ92, and Ϫ41 relative to the ␣ 5 mRNA start site) from the human ␣ 5 gene promoter. CAT activity was measured and expressed as the ratio of CAT ϩ FN over CAT Ϫ FN. Standard deviation is provided for each value. B, identification of a perfect inverted repeat in the ␣ 5 promoter segment that mediates FN responsiveness. The DNA sequence of a perfect inverted repeat that has been designated as the ␣ 5 FRE and identified between positions Ϫ77 and Ϫ61 from the human ␣ 5 gene promoter is indicated in bold capital letters. Arrows indicate the position of each ␣ 5 FRE half-sites.

FIG. 3. Influence exerted by both ECM and FN on ␣ 5 promoter function.
A, ␣ 5 promoter activity in RCEC grown on FN-, ECM-, or both FN ϩ ECM-coated dishes. RCEC were grown to mid-confluency either solely on plastic (Ϫ) or on coated ((ECM (10 g/cm 2 ), FN (2 g/cm 2 ) or both ECM ϩ FN; (10 and 2 g/cm 2 , respectively)) culture dishes prior to their transfection with the ␣ 5 recombinant plasmid ␣ 5 Ϫ954. Transfected cells were harvested 48 h later, and CAT activity was determined and normalized as described under "Experimental Procedures." Each value is expressed as detailed in the legend to Fig. 1. B, CAT activity directed by recombinant plasmids bearing 5Ј deletions of the ␣ 5 promoter in RCEC grown with or without ECM. RCEC plated at an intermediate cell density (5 ϫ 10 5 cells per Petri) for 24 h on either regular or ECM-coated (10 g/cm 2 ) culture dishes were transiently transfected with the various 5Ј deletion constructs from the ␣ 5 promoter (see Fig. 2A). CAT activity was measured and expressed as the ratio of CAT ϩ ECM over CAT Ϫ ECM. Standard deviation is provided for each value.
gene promoter (35) in order to evaluate whether such an FNdependent increase in ␣ 5 mRNA could be determined by discrete cis-acting elements from the ␣ 5 gene upstream regulatory region. For this purpose, a recombinant plasmid bearing the ␣ 5 promoter up to position Ϫ954 (␣ 5 -954) inserted upstream from the CAT reporter gene was transfected into RCEC plated either on plastic or FN-coated culture dishes (2 g/cm 2 ) at varying cell densities. As Fig. 1A indicates, culturing RCEC on FN-coated Petris did not alter the activity driven by the ␣ 5 Ϫ954 plasmid when transfected at low cell density (near 15% coverage of the plates). However, at both intermediate (near 75% coverage) and high (100% coverage for more than 48 h) cell density, the activity of the ␣ 5 promoter was found to be 6.1-and 6.4-fold, respectively, higher when cells are grown on FN-coated culture plates rather than solely on plastic. Withdrawal of the serum contained into the culture medium (which normally contains 5% FBS) prior to cell seeding on FN-coated culture dishes had no statistical effect on the CAT activity directed by ␣ 5 Ϫ954 (results not presented).
The dose dependence of the ␣ 5 promoter FN responsiveness was next evaluated by transfecting RCEC plated at an intermediate cell density on culture dishes coated with either none or increasing concentrations of FN (from 1 to 16 g/cm 2 ). As shown on Fig. 1B, the activity directed by the ␣ 5 Ϫ954 plasmid increased proportionally to the amount of FN coated on the culture dishes, reaching a drastic 18-fold stimulation at 16 g/cm 2 FN. No further increase in ␣ 5 promoter function was observed at FN concentrations above 16 g/cm 2 (results not presented). We therefore conclude that the activity of the human ␣ 5 promoter can be drastically increased when RCEC are grown on FN-coated culture dishes and that such a positive influence is obviously cell density-dependent.
A Distinct Cis-acting Element from the Basal Promoter of the Human ␣ 5 Gene Mediates FN Responsiveness in RCEC-Discrete cis-acting regulatory elements are known to mediate many of the regulatory effects that are triggered through signal transduction pathways by binding trans-acting nuclear proteins with distinctive regulatory properties. To determine more precisely the minimal ␣ 5 promoter sequence required to confer FN responsiveness, CAT recombinant plasmids bearing various 5Ј deletions of the ␣ 5 promoter were transfected into RCEC grown at intermediate density on both plastic and FN-coated (2 g/cm 2 ) culture dishes. Neither the deletion of the ␣ 5 promoter down to position Ϫ178 nor Ϫ92 could prevent the average 5-fold increase in ␣ 5 promoter activity observed when RCEC are grown on FN-coated plates. However, the further deletion of the ␣ 5 sequences down to position Ϫ41 almost totally abolished the FN responsiveness of the ␣ 5 promoter. A detailed examination of this 41-bp sequence revealed the presence of a perfect inverted repeat of the following sequence, 5Ј-GGAGTTTG-3Ј (Fig. 2B). Therefore, FN responsiveness of the ␣ 5 promoter appears to be determined by a short stretch of DNA sequence contained between positions Ϫ41 and Ϫ92 relative to the ␣ 5 mRNA start site.
Influence of ECM Components Other Than FN on the Activity of the ␣ 5 Promoter-Apart from FN, proteins such as collagen IV, vitronectin, entactin, and laminin are also commonly found in the extracellular matrix. We examined whether components from the ECM other than FN can also alter the expression directed by the ␣ 5 promoter in RCEC. For this purpose, RCEC were grown to intermediate cell density on either untreated or ECM-coated culture dishes before they were transiently transfected with the ␣ 5 Ϫ954 plasmid. The ECM gel (basement membrane matrice from Engelbreth-Holm-Swarm mouse sarcoma, Fisher) contains laminin, collagen IV, entactin, and heparin sulfate proteoglycans but no FN. As shown on Fig. 3A, the CAT activity driven by ␣ 5 Ϫ954 is increased by only 2.5-fold when RCEC are grown on ECM-coated dishes (10 g/cm 2 ). The CAT activity directed by the ␣ 5 promoter was raised to 4-fold when both FN (2 g/cm 2 ) and the ECM gel (10 g/cm 2 ) are coated together on the culture dishes. However, optimal promoter activation was obtained when FN (2 g/cm 2 ) was coated alone on the culture plates (7.6-fold activation). Transient transfection of RCEC plated on ECM-coated culture dishes with the recombinant plasmids bearing the various 5Ј deletions of the ␣ 5 promoter identified the ECM-responsive element somewhere between positions Ϫ178 and Ϫ954 (Fig. 3B). We conclude that components from the ECM other than FN had only a moderate effect on the ␣ 5 promoter activity and that their action is mediated through a cis-acting element distinct from that which determines FN responsiveness in RCEC.
The Transcription Factor Sp1 Binds Specifically to the ␣ 5

FIG. 4. Binding of nuclear proteins from RCEC to the ␣ 5 FRE in vitro.
A, EMSA analysis of the nuclear proteins from RCEC interacting with the ␣ 5 FRE. The double-stranded oligonucleotide bearing the ␣ 5 FRE was 5Ј end-labeled and incubated with varying concentrations (2-20 g) of crude nuclear proteins from RCEC grown on either not coated (FNϪ) or FN-coated (FNϩ) culture dishes (8 g/cm 2 ). The position of three DNA-protein complexes is shown (a-c) along with that of the free probe (U). P, labeled probe alone. B, competitions in EMSA. The 5Ј end-labeled ␣ 5 FRE was incubated with 10 g of crude nuclear proteins from RCEC grown on FN-coated culture dishes in the presence of either 100-or 500-fold molar excess of various unlabeled double-stranded oligonucleotide competitors (FRE, Sp1, NF1, and p12.A). U, free labeled probe; P, labeled probe alone; C, labeled probe with nuclear proteins but without unlabeled competitor.
FRE in Vitro-In order to determine whether the FN responsiveness mediated by the Ϫ41/Ϫ92 ␣ 5 promoter segment (which also contains the Ϫ82 to Ϫ56 inverted repeat that has been designated as the fibronectin-responsive element (FRE)) depends on its recognition by nuclear transcription factors, EM-SAs were performed. For this purpose, the synthetic oligomer bearing the ␣ 5 FRE was 5Ј end-labeled and incubated with increasing amounts of crude nuclear proteins (2, 5, 10, and 20 g) from RCEC grown either on plastic or FN-coated culture flasks (8 g/cm 2 ). As shown in Fig. 4A, three distinct DNAprotein complexes (designated a, b, and c) were observed upon autoradiography, complex a being the most abundant at 5, 10, and 20 g of proteins (a few other fast-migrating complexes were also occasionally observed in EMSA but their formation proved to be highly inconsistent). The signal corresponding to both complexes a and c was usually found to be much stronger in the crude extract prepared from RCEC grown on FN-coated culture dishes. Specificity for the formation of these complexes was then evaluated by competition experiments in EMSA using, as unlabeled competitors, various double-stranded oligonucleotides bearing target sequences for known transcription factors. Formation of both complexes a and b could easily be competed off by a 100-fold molar excess of unlabeled FRE, whereas that of complex c was partly prevented at a 100-fold molar excess but nearly completely abolished at a 500-fold excess (Fig. 4B). Formation of both complexes a and b could not be prevented by an unrelated oligomer bearing the target sequence for HeLa CTF/NF-I in adenovirus type 2. However, that of complex c was efficiently prevented when a 500-fold molar excess of the NF1 oligomer was used suggesting that binding of a member of the NF1 family of transcription factors likely accounts for the formation of this complex. Most of all, an oligomer bearing the high affinity binding site for the positive transcription factor Sp1 could compete for formation of complexes a and b even as efficiently as the FRE itself, a 100-fold molar excess being sufficient to almost totally prevent their formation in EMSA (Fig. 4B). As further evidence that Sp1 or any other member of this family (43) is the major transcription factor binding the ␣ 5 FRE, a synthetic oligomer bearing the target sequence for Sp1 that we identified in the basal promoter from the mouse p12 gene (and designated p12.A) (36) was also used as unlabeled competitor. This Sp1 site diverges from the Sp1 consensus by the lack of the central C residue (Fig. 2B) which is substituted by a T in the p12.A element. It is also relatively well preserved with the 3Ј half-site of the ␣ 5 FRE (9 out of 12 residues) since the five G residues identified as critical for recognition of the p12.A element by Sp1 are also preserved in the ␣ 5 FRE (36) (see Fig. 10). As shown in Fig. 4B, FIG. 5. In vitro binding of Sp1 to the ␣ 5 FRE as revealed by supershift analyses in EMSA and Western blots. A, the nuclear protein yielding complex a corresponds to Sp1. Varying amounts (5-20 g) of crude nuclear proteins from RCEC grown with (8 g/cm 2 ) or without FN were incubated with the labeled FRE either alone or in the presence of 1 l of the Sp1 antiserum. Formation of the shifted DNA-protein complexes was examined by the EMSA as above. The position of both the Sp1-FRE (corresponding to complex a) and the supershifted Sp1Ab-Sp1-FRE (identified as a/Sp1Ab) complexes is provided, along with that of the free probe (U). B, as controls, the FRE labeled probe was incubated in the presence of either 1 l of a non-immune serum (NIS) or 1 l of the Sp1 antiserum (Sp1Ab) with or without crude nuclear proteins (NP), and formation of DNA-protein complexes was evaluated as above. U, unbound fraction of the labeled probe. C, Western blot analysis of Sp1 in RCEC. Crude nuclear proteins were obtained from RCEC grown on culture dishes coated with (ϩFN, 8 g/cm 2 ) or without (ϪFN) fibronectin and tested in Western blot analyses using the Sp1 antiserum as detailed under "Experimental Procedures." As a positive control, increasing concentrations (5-20 g) of a crude nuclear extract from HeLa cells were also loaded next to those from RCEC on the SDS gel. The molecular mass markers shown correspond to myosin (200 kDa) and phosphorylase b (97.4 kDa). The position of the endogenous Sp1 is indicated (Sp1 (95/106)). the p12.A element competed nearly as well as the FRE for the formation of both complexes a and b in EMSA. These results suggest that formation of complexes a and b likely results from the recognition of the labeled ␣ 5 FRE by distinct members of the Sp1 family of transcription factors and that complex c might result from the recognition of that same probe by a member of the NF1 family. A detailed examination of the DNA sequence from the ␣ 5 FRE indeed revealed the presence of a perfect half-palindromic site for NF1 (TGGCA; see Figs. 2B and 10) that has been previously reported to bind this transcription factor (44,45).
We next performed supershift experiments in EMSA to establish clearly whether Sp1 was truly binding the FRE to yield complexes a and b in EMSA. The experiment was conducted at three different protein concentrations (5, 10, and 20 g of crude nuclear proteins) in the presence of either none or 1 l (corresponding to 1 g) of an antiserum raised against human Sp1. Again, formation of complex a but not that of complex c was found to be much stronger when the extract from RCEC grown on FN-coated Petri dishes was used at either 10 or 20 g (but not at 5 g; Fig. 5A), suggesting that Sp1 expression (or its corresponding DNA binding affinity) is increased when RCEC are cultured on FN-coated dishes. The further addition of the Sp1 antiserum resulted in a strong reduction of complex a formation and yielded a new complex (a-Sp1Ab) with a lower electrophoretic mobility resulting from the recognition of complex a by the Sp1 antibody. The proportion of the signal supershifted by the Sp1 antibody was much stronger in the extract from RCEC grown on FN-coated culture dishes (at both 10 and 20 g but not at 5 g of proteins) than with RCEC grown solely on plastic, providing further evidence that either Sp1 expression, or its DNA binding affinity, is indeed increased in cells grown on FN-coated culture dishes. As Fig. 5B indicates, no supershifted complex could be obtained when the Sp1 antiserum was substituted with the non-immune serum, which is used as a negative control in such experiments. Western blot analysis using the Sp1 antiserum as the source of primary antibody revealed that RCEC express nearly the same amount of Sp1 regardless of whether they are cultured on plastic or FN-coated culture dishes (Fig. 5C), therefore providing evidence that an improved DNA binding affinity, rather than a variation in the level of expression of Sp1, likely accounts for the increased binding of Sp1 to the ␣ 5 FRE when RCEC are cultured on FN-coated dishes. However, nearly five times more proteins from RCEC were required in order to detect an Sp1 signal of equal strength to that obtained using proteins from human HeLa cells (often used as a positive control for Sp1 expression). This substantial difference did not arise from a reduced affinity of the anti-human Sp1 antibody directed against rabbit Sp1 in our experiments since EMSAs performed using the high affinity Sp1 oligomer as labeled probe (Fig. 6A) also revealed a much weaker shifted signal when crude nuclear extracts from RCEC are selected, despite that equal amounts of proteins from both RCEC and HeLa cells were used. Therefore, the reduced Sp1 binding observed in nuclear extracts from RCEC likely suggests that Sp1 is expressed at a much lower level in RCEC than in HeLa cells. Furthermore, Sp1 clearly possesses a higher affinity for the Sp1 oligomer than for its target site in the ␣ 5 FRE, since only a 100-fold molar excess of the high affinity Sp1 oligomer is sufficient to totally prevent binding of Sp1 to the Sp1-labeled probe, whereas a 500-fold excess of the ␣ 5 FRE was required to almost totally prevent formation of this complex in EMSA (Fig. 6B).
The Inverted Repeat from the ␣ 5 Basal Promoter Mediates Sp1-dependent FN Responsiveness-To answer whether the inverted repeat from the ␣ 5 FRE (Fig. 2B) is sufficient to confer FN responsiveness to a heterologous promoter, a synthetic, double-stranded oligonucleotide bearing the ␣ 5 sequence from Ϫ82 to Ϫ56 (designated as FRE) was inserted upstream from the basal promoter of the mouse p12 gene. The p12 gene encodes a 12-kDa secretory protease inhibitor whose expression is mainly restricted to the ventral prostate, the coagulating gland, and the seminal vesicle (46). We have previously shown that the basal promoter from the p12 gene, which extends from position Ϫ108 to ϩ7 in plasmid p12.108, is constitutively expressed to relatively high levels in most transfected cell types (36,47). However, to avoid any interference by the Sp1 site identified in the middle of the p12 basal promoter, the FRE was inserted in a derivative from p12.108 that bears mutations into the p12 Sp1 target site (p12.108/M (36)) (Fig. 7A). When transfected into mid-confluent RCEC, only a weak difference (1.6fold activation) was observed in the CAT activity directed by the parental plasmid p12.108/M when 8 g/cm 2 FN was coated on the culture plates (Fig. 7A). However, insertion of either one (in plasmid p12/FRE) or two sense copies (in plasmid p12/ 2xFRE) of an oligomer bearing the Ϫ82/Ϫ56 ␣ 5 FRE immediately upstream from the p12 basal promoter resulted in 6.1and 8.8-fold increase in CAT activity, respectively. Mutations introduced in the 5Ј half-site of the inverted repeat contained on the FRE (see under "Experimental Procedures") had no statistically significant effect on either the basal p12 promoterdriven activity when cells are grown on plastic or on the FN responsiveness when they are cultured on FN-coated culture dishes (32% reduction when compared with the level directed by the wild-type p12/FRE) (Fig. 7B). On the other hand, mutations that altered part of the 3Ј half-site of the FRE and most of its downstream GC-rich sequence had no affect on the unstimulated, p12 promoter basal activity but totally abolished FN responsiveness when RCEC were cultured on FN-coated dishes. As expected, mutating both the 3Ј-and 5Ј half-sites from the ␣ 5 FRE had the same effect as mutating the 3Ј halfsite alone. To confirm that the lack of FN responsiveness resulting from mutating the 3Ј half-site of the ␣ 5 FRE was the consequence of preventing Sp1 from properly interacting with its target sequence in the FRE, competition experiments in EMSA were performed. Crude nuclear proteins were prepared from mid-confluent RCEC grown on FN-coated culture dishes and incubated with the ␣ 5 FRE-labeled probe in the presence of varying concentrations of unlabeled oligonucleotides bearing the sequence from either the wild-type FRE or any of its mutated derivatives. As shown in Fig. 7C, incubation of the labeled probe with nuclear proteins from RCEC yielded the typical Sp1-FRE complex observed above (also denoted complex a in both Figs. 4 and 5). As expected, as little as a 100-fold molar excess of unlabeled wild-type FRE totally prevented formation of this complex. Similarly, a 100-fold molar excess of the 5Ј half-site-mutated FRE competed as well the unmutated FRE for the formation of the Sp1 complex providing evidence that these mutated positions did not interfere with the recognition of the oligomer by Sp1. On the other hand, derivatives of the FRE bearing mutations in the 3Ј half-site (altering either the 3Ј half-site alone or in combination with the 5Ј half-site) were totally inefficient in preventing formation of the Sp1-FRE com-plex, even when used at a 500-fold molar excess, therefore providing evidence that both mutated oligomers are unable to bind Sp1. We therefore conclude that the inverted repeat identified in the basal promoter of the ␣ 5 gene can confer Sp1-dependent FN responsiveness to an otherwise unresponsive heterologous promoter and that only the 3Ј repeat of the FRE, along with its downstream GC-rich sequence, is required for this effect to occur.
As further evidence that FN responsiveness mediated by the ␣ 5 FRE was determined through its recognition by Sp1, cotransfection experiments were therefore conducted into Drosophila Schneider cells. These cells have been reported to be deficient in producing this transcription factor, as well as many others expressed in higher eukaryotes, which make them an ideal system for studying gene expression or transcription factor functions (for a review see Ref. 48). Both the FRE-bearing ␣ 5 Ϫ92 and the FRE-depleted ␣ 5 Ϫ42 plasmids were cotransfected into Schneider cells either alone or with a recombinant plasmid (pPacSp1, a generous gift from Dr. Guntram Suske, Institute fü r Molecular Biology und Tumorforschung, Philipps Universitä t Marburg, Germany) containing the Sp1 cDNA under the control of the Drosophila actin gene promoter and therefore ensuring high levels of Sp1 expression in Schneider cells. Neither ␣ 5 Ϫ41 nor ␣ 5 Ϫ92 could determine high basal promoter activity when individually transfected in Schneider FIG. 7. The ␣ 5 FRE can confer FN responsiveness to the basal promoter of the heterologous p12 gene. A, a synthetic oligonucleotide bearing the ␣ 5 inverted repeat from position Ϫ82 to Ϫ56 (␣ 5 FRE) was inserted in either one (in plasmid p12/FRE) or two sense copies (in plasmid p12/2xFRE) upstream from the basal promoter of the mouse p12 gene (p12.108/M) (36). The recombinant constructs were transiently transfected into midconfluent RCEC grown on tissue culture dishes coated with either none (FNϪ) or 8 g/cm 2 FN (FNϩ). CAT activity was measured and expressed relative to the unstimulated level directed by the parental plasmid p12.108/M. Arrows indicate the position of each of the wild-type, unmutated ␣ 5 FRE inverted repeats. B, double-stranded oligonucleotides bearing mutations in either the 3Ј (in plasmid p12/FREm3Ј) or the 5Ј (in plasmid p12/FREm5Ј) half-site from the ␣ 5 FRE, or both (in plasmid p12/FREm5Јϩ3Ј), were inserted in single-sense copies upstream from the p12 promoter in p12.108/M. The CAT activity directed by each recombinant construct was assessed following transient transfections in mid-confluent RCEC grown on culture dishes coated or not with FN (8 g/cm 2 ) as in A. Arrows indicate the position of each unmutated ␣ 5 FRE inverted repeat, whereas mutation-bearing repeats are indicated by a ϫ. CAT activity was measured as in A. C, formation of the Sp1-␣5FRE DNA-protein complex (identified as complex a in Fig. 5) was evaluated in EMSA by incubating the ␣ 5 FRE-labeled probe with crude nuclear proteins from mid-confluent RCEC grown on FN-coated dishes (8 g/cm 2 ), in the presence of unlabeled double-stranded oligonucleotides (both 100-and 500-fold molar excess) bearing the DNA sequence of either the wild-type, unmutated FRE or that of any of its mutated derivatives (FREm3Ј; FREm5Ј; FREm3Јϩ5Ј). The position of the Sp1-FRE complex is provided, along with that of the free probe (U). P, probe without nuclear proteins. cells. However, when cotransfected along with pPacSp1, a dramatic 75-fold increase in promoter activity was observed with the FRE-containing plasmid ␣ 5 Ϫ92 but not with the FREdeleted plasmid ␣ 5 Ϫ41 (Fig. 8A). The recombinant ␣ 5 FRE/p12 promoter constructs were then transfected either alone or with pPacSp1 into Schneider cells (Fig. 8B). The parental plasmid p12.108/M, although encoding substantial amounts of CAT in Schneider cells, responded only weakly (3.5-fold activation) to the presence of Sp1. However, the further addition of the ␣ 5 FRE in p12/FRE resulted in a strong increase (52-fold) in the CAT activity normally directed by p12.108/M. As with the transfection experiments conducted in RCEC (see Fig. 7, A and B), mutations introduced in the 5Ј half-site of the FRE (in plasmid p12/FREm5Ј) had only a modest effect on the Sp1mediated activation of the p12 promoter (Fig. 8B). On the other hand, no Sp1-mediated activation could be observed upon mutating either the 3Ј half-site alone or both the 3Ј and 5Ј halfsites of the FRE (in the plasmids p12/FREm3Ј and p12/FRE3Ј ϩ 5Ј, respectively). These results are consistent with those of the competition experiment shown in Fig. 7C and provide clear evidence that Sp1 does bind to the 3Ј half-site of the FRE in order to influence positively the activity of its downstream promoter (in this case, either the ␣ 5 or the p12 promoter).
Activation of ERK1/ERK2 Mediates the FN Responsiveness of the ␣ 5 Promoter-Culturing murine Swiss 3T3 or rat REF52 fibroblasts on substrata coated with either FN or with a synthetic peptide containing the RGD sequence has been shown to result in the activation of mitogen-activated protein kinases (MAPK) (49), such as extracellular signal-regulated kinases (ERKs), which have been shown to be recruited to the ECM ligand/integrin-binding site (50). Sp1 has been recently recognized as one of the few target transcription factors phosphorylated by ERK kinases (51)(52)(53). We have shown above that its ability to interact with the ␣ 5 FRE is strongly increased upon activation of the FN/␣ 5 ␤ 1 integrin-mediated signal transduction. To determine whether the Sp1-mediated FN responsiveness directed by the ␣ 5 FRE was due to the activation of the Ras-Erk signaling pathway (27), RCEC grown to intermediate cell density on culture dishes coated (8 g/cm 2 ) or not with FN were transiently transfected with either the recombinant plasmids ␣ 5 Ϫ92 or p12/FRE and cultured with either none or 10 M of the MEK-1 kinase inhibitor PD98059. As expected, the CAT activity directed by the transfected plasmid ␣ 5 Ϫ92 was strongly increased (9.9-fold increase) when RCEC were cultured on FN-coated dishes (Fig. 9A). However, the addition of as little as 10 M of the PD98059 inhibitor (many studies have used doses 5-10-fold higher of this inhibitor (51-53)) totally abolished this FN responsiveness, the level of CAT activity returning to the unstimulated level. Identical results were also obtained with the recombinant plasmid p12/FRE, which bears one sense copy of the ␣ 5 FRE inserted upstream from the basal promoter of the p12 gene (see Fig. 7, A and B). Again, the nearly 3-fold increase in the ␣ 5 FRE-mediated FN responsiveness of the p12 promoter was totally abolished when cells were cultured in the presence of the inhibitor (Fig. 9B). Crude nuclear extracts were prepared from RCEC grown either on plastic or FN-coated culture dishes in the presence of either none or 10 M PD98059 and then used in EMSAs. Upon incubation with the ␣ 5 FRE-labeled probe, a clear Sp1 signal that increased severalfold when RCEC were grown on FN could be observed with nuclear extracts from RCEC that have not been exposed to the MEK-1 inhibitor (compare 1st and 3rd lanes in Fig. 9C). However, culturing RCEC in the presence of the inhibitor totally abolished formation of the Sp1-␣ 5 FRE complex (Fig. 9C) even when cells were grown solely on plastic. We therefore conclude that activation of the Ras-Erk signaling pathway through the interaction of the ␣ 5 ␤ 1 integrin with its ECM ligand FN accounts for the increase in ␣ 5 promoter activity when RCEC are grown on FN-coated culture dishes and that this effect is most likely dependent on the altered phosphorylation state of Sp1 by activated ERK1/ERK2. DISCUSSION Debridement of the corneal epithelium is known to promote wound healing by stimulating migration and differentiation of both the basal corneal epithelial cells that border the injured area and the precursor cells from the corneal limbus. Induction of the migration process is known to be influenced by the FIG. 8. Transient transfection experiments in Sp1-deficient Drosophila Schneider cells. A, both the recombinant plasmids ␣ 5 Ϫ41 and ␣ 5 Ϫ92 were transiently transfected either alone or in combination with the Sp1 expression plasmid pPacSp1 into Drosophila Schneider cells. Cells were harvested 48 h later, and CAT activity was determined and normalized as detailed under "Experimental Procedures." B, same as in A except that the recombinant plasmids p12.108/M and p12/FRE, or its mutated derivatives p12/FREm3Ј, p12/FREm5Ј, and p12/FREm3Јϩm5Ј (see under "Experimental Procedures"), were selected for the cotransfection experiments. massive production of FN by both the stromal keratinocytes and the basal cells that border the injured area (5, 8 -15). Moreover, recent studies provided evidence that the level of expression for the mRNA encoding the ␣ 5 integrin subunit was positively modulated by the presence of such FN (23,24). The present study was therefore conducted in order to investigate whether FN, through its ␣ 5 ␤ 1 receptor-mediated signal transduction pathway, can alter the transcriptional activity directed by the promoter of the ␣ 5 integrin subunit gene. We provided clear evidence that FN can indeed alter the transcriptional activity of the ␣ 5 promoter by altering the DNA binding affinity of the positive transcription factor Sp1 for a short ␣ 5 promoter segment located between positions Ϫ77 and Ϫ61 that has been designated as the ␣ 5 FRE. The ␣ 5 FRE bears a perfect inverted repeat of the following sequence, 5Ј-GGAGTTTG-3Ј. However, site-directed mutagenesis provided evidence that only the 3Ј half-site (along with its nearby 3Ј GC-rich base pairs (TCCCC)) was required for FN responsiveness to occur. This short stretch of sequence from the ␣ 5 promoter was found to be highly homologous to a 12-bp sequence from the murine acetylcholine receptor (AChR) ␦-subunit gene that was reported to be absolutely required for muscle-specific expression of AChR-␦ (54) (see Fig. 10). This cis-acting element, which is comprised between positions Ϫ106 and Ϫ95, was postulated as being the target sequence for the transcription factor myogenin (54). However, myogenin has not been reported as a target protein that might be subjected to differential phosphorylation by protein kinases.
The substantial variations we have observed in the ability of Sp1 to bind the ␣ 5 FRE in RCEC grown with or without FN might either result from alteration in the binding affinity of Sp1 or from modification in the amount of Sp1 protein produced by RCEC under both culture conditions. However, our inability to detect any significant variations in the absolute amount of Sp1 between RCEC grown with or without FN in Western blot analyses rather favors the former hypothesis. Our results are consistent with those recently reported by Alroy et al. (55) who could not see any variation in Sp1 protein levels despite an increased binding of Sp1 to the Neu differentiation factor response element from the promoter of the acetylcholine receptor ⑀ upon stimulation with Neu differentiation factors. Variations in the ␣ 5 promoter activity might then be triggered by modifying the affinity of Sp1 for the FRE target site by altering its state of phosphorylation through nuclear proteins that belong to the MAPK family, such as ERK-1 (p44) and ERK-2 (p42). Alteration of the state of phosphorylation for the transcription factor Sp1 has been reported to alter, either positively or negatively, its DNA-binding properties in vitro (52,55,56). Li et al. (51) recently reported that Sp1 might also be a target for ERK proteins. Indeed, they identified a 16-bp sequence (repeat 3) that mediates responsiveness of the human low density lipoprotein receptor (LDLR) to oncostatin M (OM) through a FIG. 9. The integrin-mediated FN responsiveness is abolished by the MEK/kinase inhibitor PD98059. A, the recombinant plasmid ␣ 5 Ϫ92 was transiently transfected into RCEC grown on plastic (ϪFN) or FN-coated (ϩFN; 8 g/ cm 2 ) culture dishes with either none or 10 M of the MEK/kinase inhibitor PD98059. Cells were harvested 48 h later, and CAT activity was determined and normalized as detailed under "Experimental Procedures." B, same as in A except that the recombinant plasmid p12/FRE (see Fig. 7) was substituted to ␣ 5 Ϫ92 for the transfection experiments. C, the double-stranded oligonucleotide bearing the ␣ 5 FRE was 5Ј end-labeled and incubated with crude nuclear proteins (5 g) from RCEC grown on either none (Ϫ) or FN-coated (ϩ) culture dishes (8 g/cm 2 ), in the presence of either none or 10 M PD98059. Formation of the Sp1-FRE complex was then monitored by EMSA as detailed in Fig. 4 except that the concentration of the polyacrylamide gel was lowered to 4%. The position of the Sp1-FRE complex is shown (Sp1) along with that of the free probe (U). P, labeled probe alone. signal transduction pathway that involves phosphorylation of ERK-1 and ERK-2 by the upstream kinases MEK-1 and MEK-2. The sequence from the LDLR repeat 3 also bears an intact copy of the GGAGTTT motif (on the non-coding strand) identified in the ␣ 5 FRE (see Fig. 10). Most of all, it also contains the GC-rich sequence (TCCCC) located downstream of the 3Ј repeat that proved to be required for the FN responsiveness directed by the ␣ 5 FRE. Interestingly, only Sp1 and the Sp1-related protein Sp3 have been shown to bind repeat 3 (51). However, these authors have been unable to detect any OMinduced alterations in Sp1 binding by EMSA or in the ratio of hyper-versus hypophosphorylated Sp1 by Western blot analyses (51). As Fig. 5A reveals (and to some extent also Fig. 4A), no significant changes in Sp1 binding to the ␣ 5 FRE could be observed between nuclear extracts obtained from RCEC grown with or without FN when only 5 g of crude nuclear proteins were used. However, raising the amount of proteins to either 10 or 20 g clearly revealed a much stronger binding of Sp1 to the FRE when RCEC are grown on FN-coated culture dishes, which is also supported by a more intense supershift of the Sp1/FRE DNA-protein complex. Formation of this DNA-protein complex is therefore likely to be concentration-dependent and might explain why these authors (57) could not detect any alterations in Sp1 binding. A recent study conducted by Milanini et al. (53) also identified Sp1 as a target of the p42/p44 MAPK pathway in the activation of the vascular endothelial growth factor (VEGF) gene. However, transcriptional activation of VEGF through this p42/p44 transduction pathway not only requires binding of Sp1 to a GC-rich region from the VEGF promoter located between positions Ϫ88 and Ϫ66 but also that of AP-2, which synergizes with the former to ensure the proper regulatory response (53). Both the Sp1 and AP-2 binding activities have been shown by EMSA to be increased when the p42/p44 MAPK pathway is activated. However, no clear evidence has been provided yet as to whether this effect is dependent on an altered state of phosphorylation of both factors or by an increase in the amount of both proteins. Recently, Merchant et al. (52) have shown, by blocking the EGF-induced Ras-Erk pathway with the MEK-1 kinase inhibitor PD98059, that phosphorylation of Sp1 through Erk2 is indeed required in order to induce Sp1 binding to the gastrin gene promoter. Therefore, and as also supported by our results, the DNA binding ability of Sp1 and, as a consequence, its transactivation properties can be enhanced likely through phosphorylation by activated ERK1/ERK2. Not all Sp1/Sp3-binding sites are subjected to regulation by the MAPK pathway. Proper positioning of the Sp1/Sp3 target site relative to the TATA box has been postulated as being particularly critical for OM-mediated transcriptional activation of the LDLR gene (57). Indeed, although LDLR repeat 1 also bears an Sp1-binding site that is critical for basal transcriptional activity, it is not affected by the presence of OM, unlike the more proximal Sp1 site from repeat 3 (57). Liu et al. (57) postulated that OM may induce the expression of an Sp1dependent coactivator that would bridge Sp1 bound to a properly positioned target site to the general transcriptional machinery. Alternatively, OM-mediated transduction pathway, through activation of ERK1/ERK2, might also lead to posttranslational phosphorylation of Sp1 and account for LDLR transcriptional activation (57), which has been recently shown to occur for the gastrin gene promoter (52). Although the ␣ 5 promoter has no TATA box, it has been shown to contain an initiator site located at position Ϫ45 that is likely sufficient to bind the general transcriptional machinery (35). The ␣ 5 FRE is therefore located approximately 15 bp upstream from the pu-tative initiator site, a positioning nearly identical to that observed for the LDLR Repeat 3 Sp1 site and the LDLR TATAlike sequence (17 bp from the most 5Ј located TATA-like sequence) (57,58). As for the LDLR repeat 1, the putative high affinity Sp1-binding sites identified in the ␣ 5 promoter between positions Ϫ178 and Ϫ92 did not contribute much to the FNmediated responsiveness of the ␣ 5 promoter since they could be deleted without any significant effect on the CAT activity. Based on the results shown in the present study, we suggest that both phosphorylation of Sp1 and proper positioning of its target site relative to the general transcriptional machinery are required to transduce properly the signal triggered by extracellular FN. Whether the need for an Sp1-dependent coactivator is required to produce the proper response remains to be demonstrated.
Expression of integrin subunit genes other than ␣ 5 has also been reported to be positively influenced by the binding of Sp1 to their promoter sequence. Indeed, the transcriptional activity directed by the ␣ 6 integrin subunit gene promoter was reported to be dependent on its recognition by both Sp1 and AP2 (59). Transcription directed by the promoter of both the ␤ 2 /CD18 and the ␣IIb integrin subunits has recently been reported to be positively regulated through the synergistic action of both Sp1 and members of the Ets family of transcription factors, such as GABP (60 -61). Two tandemly repeated Sp1-binding sites were also identified in the promoter of the ␣ 2 integrin subunit gene (62). Interestingly, phosphorylation of Sp1 appears to be required for formation of the Sp1 DNA-protein complex in vitro.
Characterization of an FN-responsive element such as the ␣ 5 FRE is a first step toward the understanding of the nuclear events taking place upon activation of the signal transduction pathway normally triggered by membrane-bound FN integrins. Our results provide a link between the extracellular ligand/ integrin-mediated signal transduction and the nuclear events leading to expression of the ␣ 5 integrin subunit gene. Most of all, it also provides further support to the major signaling pathway activated by the ␣ 5 ␤ 1 integrin and for which a model was recently proposed (63). FIG. 10. Sequence homology between the ␣ 5 FRE and other gene regulatory target sequences. The DNA sequence from the human ␣ 5 FRE is aligned with the Sp1-binding site identified in the promoter of the mouse p12 gene (p12.A), the 16-bp repeat 3 element from the human LDLR, and sequences from the murine AChR-␦ subunit gene promoter that also bears a binding site for the transcription factor myogenin (MG) (underlined). Arrows indicate the position of each ␣ 5 FRE half-inverted repeats, and black dots indicate the position of those G residues whose methylation by dimethyl sulfate interferes with the recognition of the p12.A element by Sp1. The DNA sequences that show homology to both the NF1 and Sp1 target sites are indicated.