Kruppel-like Factor 4 Regulates Laminin α3A Expression in Mammary Epithelial Cells*

Laminin-5, the major extracellular matrix protein produced by mammary epithelial cells, is composed of three chains (designated α3A, β3, and γ2), each encoded by a separate gene. Laminin-5 is markedly down-regulated in breast cancer cells. Little is known about the regulation of laminin gene transcription in normal breast cells, nor about the mechanism underlying the down-regulation seen in cancer. In the present study, we cloned the promoter of the gene for the human laminin α3A chain (LAMA3A) and investigated its regulation in functionally normal MCF10A breast epithelial cells and several breast cancer cell lines. Using site-directed mutagenesis of promoter-reporter constructs in transient transfection assays in MCF10A cells, we find that two binding sites for Kruppel-like factor 4 (KLF4/GKLF/EZF) are required for expression driven by the LAMA3A promoter. Electrophoretic mobility shift assays reveal absence of KLF4 binding activity in extracts from T47D, MDA-MB 231, ZR75-1, MDA-MB 436, and MCF7 breast cancer cells. Transient transfection of a plasmid expressing KLF4 activates transcription from the LAMA3A promoter in breast cancer cells. A reporter vector containing duplicate KLF4-binding sites in its promoter is expressed at high levels in MCF10A cells but at negligible levels in breast cancer cells. Thus, KLF4 is required forLAMA3A expression and absence of laminin α3A in breast cancer cells appears, at least in part, attributable to the lack of KLF4 activity.

The laminin family of extracellular matrix glycoproteins is heterotrimeric proteins consisting of three distinct subunits, designated ␣, ␤, and ␥ that are encoded by the LAMA, LAMB, and LAMC genes, respectively. To date there are five ␣ chains, three ␤ chains, and two ␥ chains that assemble into 12 laminins. Laminin-5 (␣3A, ␤3, and ␥2) is the major extracellular matrix protein produced in mammary epithelial cells. In these cells laminin-5 functions as a ligand for the ␣ 3 ␤ 1 and ␣ 6 ␤ 4 integrins to regulate adhesion, migration, and morphogenesis (1). Loss of laminin-5 has been found in breast cancer progression. Henning et al. (2) demonstrated a loss of laminin-5 protein expression by immunostaining in malignant lesions while benign ductal and lobular proliferations and fibroadenomas show continuous laminin-5 staining at the epithelial-stromal interface. Martin et al. (3) used a molecular approach to analyze mRNA expression of the laminin-5 subunits and found no expression in late stage tumors and decreased expression in early stage tumors compared with normal breast epithelial cells.
The murine LAMA3A promoter has been studied and found to contain three binding sites for the complex dimeric transcription factor, AP-1, one site of which is essential for basal expression of LAMA3A in keratinocytes (4). Mutation of this single key AP-1 site reduced promoter activity by ϳ90% while mutation of the other two sites had much less effect (4). In the present study we analyze the human LAMA3A promoter in the MCF10A mammary epithelial cell line and the T47D breast cancer cell line. We sought to find a mechanism that would explain the LAMA3A down-regulation in the nonexpressing cells. In doing this, we demonstrated a key role for the transcription factor Kruppel-like factor 4 (KLF4) 1 in regulating this gene.
KLF4, also known as gut-enriched Kruppel-like factor (GKLF) and epithelial zinc finger (EZF), is a member of the Kruppel-type zinc finger transcription factors (5)(6)(7)(8). One of the best known members of this family is the erythroid Kruppellike factor that is involved in the activation of the ␤-globin promoter in red blood cells (9). The family also includes LKLF (10), UKLF (11), IKLF (12), BTEB2 (13), and BKLF (14), many of which are tissue specific in their expression. Members of the Kruppel-type family are highly conserved in the carboxyl-terminal region that contains three zinc fingers and they bind similar GC-rich recognition sequences. KLF4 is a nuclear protein shown to contain both transcriptional activation and repression domains (8). In keratinocytes, KLF4 activates the keratin 4 promoter and may be important in the transition toward differentiation (7). KLF4 has also been shown to be involved in the regulation of the CYP1A1 gene (15). This report describes the novel role of this factor in the regulation of the LAMA3A promoter. Through transient transfections and EMSA we show that KLF4 activates expression of the LAMA3A gene in MCF10A cells and that loss of KLF4 activity is in part responsible for the loss of LAMA3A expression in breast cells.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-MCF10A (a nontransformed, spontaneously immortalized human breast epithelial cell line) and M-H cells (a hygromycin resistant subclone of MCF10A cells) were maintained as previously described (16). T47D cells were cultured in RPMI 1640 medium supplemented with 0.01 mg/ml insulin and 10% heat inactivated fetal bovine serum. MDA-MB 231, MCF7, and ZR75-1 cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 300 g/ml glutamine. MDA-MB 436 cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum, 0.01 mg/ml insulin, and 16 g/ml glutathione. MCF10A and T47D cells were transfected in log growth phase at a density of 1 ϫ 10 5 cells per well in 6-well plates using the Superfect Transfection Reagent (Qiagen, Inc.) per the manufacturer's protocol. Quantities transfected of the respective vectors are indicated in the figure legends. Cells were harvested in Reporter Lysis Buffer (Promega Corp.) between 24 and 48 h. Luciferase activity was assayed by mixing aliquots of cell extracts with luciferin reaction mixture (Promega Luciferase Assay Kit) and emission of light was quantitated with a Microlumat luminometer. ␤-Galactosidase activity was determined using the ␤-Galactosidase Luminescence Kit (CLONTECH Laboratories, Inc., Palo Alto, CA) per the manufacturer's instruction. Transfections were normalized for efficiency by normalizing relative light units from the ␤-galactosidase assay to the luciferase light units for each individual sample.
Plasmids-The sequence for the intron between LAMA3B exon 1 and LAMA3A exon 1 which has been shown to constitute the LAMA3A promoter in mouse was kindly provided by Dr. Angela Christiano (Columbia University, New York). Based on this sequence, primers were designed to amplify 1410 base pairs of this intronal promoter. The 5Ј primer 5Ј-CAGGTACCAAGTTTTCCCATCCGCAACATTTCC-3Ј that has a KpnI site engineered into the 5Ј end and the 3Ј primer 5Ј-CAGCTAGCAGGCTGACCGCCTCACTGCTGGAGG-3Ј that has an NheI site engineered were used in polymerase chain reaction of genomic DNA from MCF10A cells. The polymerase chain reaction fragment was TA cloned into the pGEMT vector (Promega), then excised by KpnI and NheI, and cloned into the pGL2 basic vector (Promega) that was digested with KpnI and NheI. The resulting vector was transformed into XL1 Gold Supercompetent cells (Stratagene), named pL3A and contains the human LAMA3A promoter upstream of a luciferase reporter gene. Three separate colonies were sequenced and the consensus sequence was submitted to GenBank TM (accession number AF279435). Restriction sites were identified using MAP functions of GCG software. The adenine of the initiator ATG codon was designated base ϩ1.
EMSA-Nuclear extracts were prepared by collecting cells in cold Buffer A (10 mM HEPES pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride) and incubated on ice for 10 min. Cells were then frozen and thawed once and vortexed for 10 s. Nuclei were pelleted by brief centrifugation and resuspended in cold Buffer C (20 mM HEPES pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride), incubated for 30 min, briefly centrifuged to remove debris, and the nuclear extract supernatant was collected and stored at Ϫ70°C. Primers to the pL3A vector were used to amplify probe A, the 103-base pair fragment spanning bases Ϫ199 to Ϫ97, for gel shift analysis, 5Ј-CGCTCTGGCACAGG-3Ј 5Ј-TTCCTGCTCAGTGCC-3Ј. Probe A was radiolabeled with [␣-32 P]dCTP (Amersham Pharmacia Biotech) using T4 polynucleotide kinase. Probe B (Ϫ84 to Ϫ62) was created by annealing two oligonucleotides, 5Ј-AGAGGAAGAGGCA-GAGGTTC-3Ј and 5Ј-CAGGAACCTCTGCCTCTTCCT-3Ј, that contain base pair overhangs that were radiolabeled using DNA polymerase in the presence of [␣-32 P]dCTP (Amersham Pharmacia Biotech). The oligonucleotide containing the KLF4 consensus sequence designated TDA by Shields and Yang (15) and an oligonucleotide based on the TDA with 2 base pairs mutated that does not bind KLF4 designated TDAmut were created as described by Shields and Yang (17) and end labeled as Probe A. Labeled probes were then separated from free [␣-32 P]dCTP using Sephadex ProbeQuant G-50 micro columns (Amersham Pharmacia Biotech). Binding reactions included: ␣-32 P-labeled DNA probes (1-5 ng, 100,000 cpm); 4 g of poly(dI-dC); 10 g of nuclear extract; 5 g of bovine serum albumin; 6 l of 5 ϫ binding buffer (50 mM HEPES pH 7.9, 5 mM dithiothreitol, 0.5% Triton X-100, and 2.5% glycerol); and the appropriate volume of H 2 O to a final volume of 30 l. Binding reaction mixtures were then incubated at 30°C for 30 min. Bound products were resolved by electrophoresis through a 7 or 8% native polyacrylamide gel in 1 ϫ running buffer (50 mM Tris, 0.38 M glycine, 2.0 mM EDTA, pH 8.5). Gels were dried with a Bio-Rad Gel Dryer for 45 min at 85°C, followed by exposure to Hyperfilm-MP (Amersham Pharmacia Biotech) at Ϫ70°C. Film was then developed using an x-ray film processor (Fuji).
To deplete extracts of KLF4, 50 l of nuclear extracts from an MCF10A subclone were incubated for 4 h at 4°C with either 20 l of rabbit preimmune serum or 10 l each of KLF4 antibodies from Dr. Yang (6) and Dr. Tseng (18) and 40 l of protein G-agarose beads (Life Technologies). Samples were briefly centrifuged, and supernatants were used in gel shifts as described above.
Competition experiments consisted of 150 ϫ molar excess for the TDA, AP-1, and AP2 oligonucleotides that were preincubated with nuclear extracts for 30 min at room temperature before proceeding to the binding reactions with the 103-base pair DNA fragment. AP-1 and AP2 duplex oligonucleotides were obtained from Stratagene, the AP2 oligonucleotide serving as a nonspecific control. Competition of the labeled TDA probe consisted of 125 ϫ molar excess of TDA(WT) and TDA(M6-mut) oligonucleotides in Fig. 6A and 250 ϫ molar excess in Fig. 6B.

RESULTS
The ␣3 Subunit of Laminin-5 Is Not Expressed in Breast Cancer Cell Lines-There is previous evidence that laminin-5 is down-regulated in breast cancer cell lines. Examination of mRNA in breast tumor cells showed that laminin ␣3 was greatly reduced or not present at all in breast tumor tissue and 15 breast cancer cell lines including T47D, MCF7, ZR75-1, MDA-MB 436, BT474, MDA-MB 361, and MDA-MB 231 (3). Long exposure of Northern blots revealed very low level expression in breast cancer cell lines, ruling out systematic deletion of this gene in breast cancer (3). The main transcript of the laminin-5 ␣3 subunit in mammary epithelia is the LAMA3A isoform (19). In the mouse the genetic structure of the mLAMA3 gene consists of the mLAMA3A and mLAMA3B isoforms that each contain a unique exon 1 that is expressed using alternative promoters. Exon 1 of mLAMA3B is upstream of the mLAMA3A exon and the intron region between these exons contains the promoter for the mLAMA3A isoform. The genetic structure for the human LAMA3 was previously described by Pulkinnen et al. (20) and is very similar to the mouse organization. We cloned the human intron region between the LAMA3B and LAMA3A exons into the pGL2 reporter vector and tested the transcriptional activity in MCF10A cells and T47D cells. We found that the resulting vector, pL3A, contained potent transcriptional activity when transfected into MCF10A cells but had no activity in T47D cells when compared with the activity of empty vector (Fig. 1A).
KLF4 Regulates LAMA3A Promoter Activity-The murine LAMA3A promoter contains three AP-1 sites (AP-1A, AP-1B, and AP-1C) that are involved in the up-regulation of this gene by transforming growth factor-␤ in keratinocytes (4). One of these sites, designated AP-1-B is critical for basal expression as well (4). Computer analysis of the human LAMA3A promoter sequence using the Matinspector software (21) revealed several putative transcription factor-binding sites. As in the mouse promoter, there are three AP-1 sites identified in the human promoter at Ϫ387, Ϫ185, and Ϫ127. Ten putative KLF4-binding sites were noted throughout the 1410-base pair promoter and are located at positions Ϫ1327, Ϫ921, Ϫ892, Ϫ572, Ϫ412, Ϫ407, Ϫ367, Ϫ167, Ϫ87, and Ϫ81. KLF4 sites at Ϫ412 and Ϫ407 overlap as do the sites at Ϫ87 and Ϫ81. Bases upstream of position Ϫ589 lacked promoter activity (data not shown) and we further examined transcription factor-binding sites between bases Ϫ589 and Ϫ6. To analyze which specific transcription factors may be involved in the regulation of this promoter, site-directed mutations were created in the full-length pL3A vector. We found that mutation of the AP-1 site homologous to the key mouse AP-1B site (4) at position Ϫ185 in vector pL3A-AP-1mut3, nearly abolished transcription in MCF10A cells, while a mutation targeting a nearby putative USF1 site in vector pL3A-USFmut4 at position Ϫ180 had no effect (Fig. 1B). We were interested in the KLF4-binding sites and found that mutations in the KLF4-binding sites closest to the translational start site most dramatically decreased transcriptional activity in MCF10A cells (Fig. 1B). Mutation of the single KLF4 site at Ϫ167 in vector pL3A-KLF4mut5 abolished 63% of tran- scriptional activity while vector pL3A-KLF4 mut6 with mutations in the two overlapping KLF4 sites at Ϫ87 and Ϫ81 abolished 79% of activity. Mutating the overlapping sites at Ϫ412 and Ϫ407, vector pL3A-KLF4mut1, as well as mutating the site at Ϫ367 in vector pL3A-KLF4mut2 had no significant effect on transcription (Fig. 1B). None of the site-directed mutations had any effect on the lack of transcription in T47D cells (data not shown).
The transfection results suggest that both AP-1 and KLF4 are important in the regulation of LAMA3A. A 103-base pair DNA fragment, Probe A, containing the key AP-1 site and the Ϫ167 KLF4 site was used for gel shift analysis. Extracts from MCF10A, T47D, MCF7, MDA-MB 231, and MDA-MB 436 cells all show AP-1 binding that is competed with an AP-1-specific oligonucleotide ( Fig. 2A, lanes 3, 7, 11, 15, and 19) but not a nonspecific competitor ( Fig. 2A, lanes 4, 8, 12, 16, and 20). This results confirms a previous report that AP-1 binds the human LAMA3A promoter (22). A complex near the free probe was noticed only in the MCF10A cell samples that was competed with a KLF4 consensus binding oligonucleotide (lane 5) designated TDA by Shields and Yang (15). To further prove that this shift was in fact KLF4, we show that depleting the nuclear extracts of KLF4 with anti-KLF4 antibody resulted in loss of the complex that was competed by the KLF4 consensus binding oligonucleotide (Fig. 2B). By transfection the KLF4 site at Ϫ81 was also important in LAMA3A transcription. Gel shift with Probe B, an oligonucleotide that spans bases Ϫ84 to Ϫ62, containing the Ϫ81 KL4-binding site showed a shift in MCF10A cells that was identical to a shift in a sample containing purified KLF4 (Fig. 2C). A band shift with Probe B was absent in breast cancer cells. To further prove that the shift seen with Probe B in MCF10A cells was in fact KLF4, a supershift experiment was performed with either preimmune serum or an anti-KLF4 antibody. The band shift of Probe B with MCF10A cell extract was specifically supershifted with the KLF4 antibody (Fig. 2D). One of the reasons that a band shift was not seen in the breast cancer cells with LAMA3A promoter sequence probes is that these cells do not express KLF4. In fact we found that to be the case by Western analysis. MCF10A cells clearly express KLF4 while no detectable KLF4 was found in whole cell extracts of MCF7, T47D, or MDA-MB 231 cells (Fig. 3).
KLF4 Activity Is Decreased in Breast Cancer Cells-To further confirm the lack of KLF4 protein in the breast cancer cells, we compared general KLF4 activity of MCF10A to the activity of T47D cells. Shields and Yang (17) have published a consensus binding sequence for KLF4. The vector TDA(WT)x2-pGL2-TATA-Luc basic contains two tandem copies of this consensus sequence upstream of a TATA element in the pGL2 basic vector. This vector has been shown to be activated by KLF4 (17). The TDA(M6-Mut)x2-pGL2-TATA-Luc vector has mutations in the KLF4 consensus and is not KLF4 responsive. We tested activity of the TDA(WT) vector compared with the TDA(M6-Mut) vector to analyze the overall KLF4 transcriptional activity in MCF10A and T47D cells. When the TDA vector was transfected in MCF10A cells, there was a 6-fold increase in transcriptional activity compared with the mutant vector that does not bind KLF4 (Fig. 4). Activation in the T47D cells, on the other hand was negligible (Fig. 4). We next analyzed KLF4 DNA binding activity by EMSA analysis in MCF10A and five different breast cancer cell lines. An oligonucleotide for EMSA analysis was created that contained the TDA consensus sequence as previously described (17). We found specific binding to the TDA-labeled oligonucleotide that was competed with cold TDA oligonucleotide but not the TDA mutant oligonucleotide in MCF10A cells (Fig. 5A). Specific binding was lacking in the five  lanes 17-19) cells. An oligonucleotide containing the KLF4 consensus binding sequence was end labeled with T4 polynucleotide kinase, and binding reactions were electrophoresed in an 8% native polyacrylamide gel. The KLF4 specific complex was seen only in MCF10A cells (lane 2). The KLF4 binding complex was partially competed with 125 ϫ cold specific (S) oligonucleotide (lane 3) but not a 125 ϫ excess of mutated (NS) oligonucleotide that does not bind KLF4 (lane 4). None of the breast cancer cell lines tested had any KLF4 binding activity. Lane 1 shows free probe (FP). Arrow points to the KLF4 specific complex. B, to better demonstrate the specific competition of the TDA oligonucleotide in MCF10A cells, competition with 250X molar excess of cold specific (S) and cold nonspecific (NS) oligonucleotide as in part A were used. Binding reactions were electrophoresed in a 7% native polyacrylamide gel. Arrow points to the KLF4 specific complex. breast cancer cell lines tested (Fig. 5A). The competition in Fig.  5A was at 125 ϫ molar excess that did not completely compete the TDA oligonucleotide. In Fig. 5B the competition was performed at 250 ϫ molar excess that demonstrated complete complex inhibition with TDA(WT) but not TDA(Mut) oligonucleotides in MCF10A cells. The high level of competitor required for competition of binding to the labeled TDA oligonucleotide suggested a high level of KLF4 DNA binding activity in MCF10A cells. These data demonstrated that the KLF4 transcriptional activity and DNA binding activity was greater in the MCF10A cells than the breast cancer cells.
Expression of KLF4 in Breast Cancer Cells Stimulates LAMA3A Transcription-Lack of KLF4 protein expression and activity in breast cancer cells lead us to hypothesize that expressing KLF4 in the breast cancer cells would have a positive effect on pL3A activity in these cells. Expression of the pL3A vector was increased 15-fold in T47D cells when co-transfected with 10 ng of the pcDNA3-hEZF vector obtained from Dr. Mu-En Lee (8) (Fig. 6B). In MDA-MB-231 cells (Fig. 6C) the fold induction with 10 ng of pcDNA3-hEZF was 39-fold the basal expression of pL3A in these cells and in MCF7 cells (Fig.  6D) the induction was 20-fold. Thus, expression of KLF4 alone was sufficient to see promoter activity in LAMA3A nonexpressing cells. Transfection of the plasmids with mutated KLF4-binding sites into MCF7 cells (pL3A-Mut5 and pL3A-Mut6) showed no activation in the presence of co-transfected KLF4 (data not shown). We were unsure what the results would be of overexpressing KLF4 in the MCF10A cells that already express KLF4 protein. Co-transfection of pL3A with low doses of the pcDNA3-hEZF vector in MCF10A cells resulted in an increase in luciferase activity (Fig. 6A). On the other hand, a higher dose of KLF4 in MCF10A cells resulted in repression of the promoter by about 5-fold, suggesting that optimal levels of KLF4 are necessary for LAMA3A expression. DISCUSSION Laminin-5 is a highly tissue-specific gene with expression found only in epithelial cell populations. A recent study shows that the human LAMA3A promoter is also driven by AP-1 (22). Virolle et al. (22) suggest that the conformation of the AP-1 sites is critical in determining whether LAMA3A expression occurs or not. They suggest that under normal circumstances, a repressor binds AP-1 in the fibroblasts but this repressor is absent in keratinocytes which allows expression of LAMA3A. Our results confirm a role of AP-1 in the regulation of the human LAMA3A promoter. Mutation of the AP-1 site decreases transcriptional activity of pL3A in MCF10A cells. Unlike the previous report, our gel shift analysis does not suggest the presence of a repressor complex that is unique to the breast cancer cells. Instead we find that the presence of KLF4 is involved. In MCF10A cells that contain KLF4, mutation of KLF4-binding sites also results in decreased transcription of pL3A. We show that breast cancer cells lack KLF4 expression and activity which likely results in the lack of LAMA3A expression in these cells. Expression of the LAMA3A promoter in the breast cancer cells by expression of KLF4 further supports this claim. This is not the first time that KLF4 has been shown to be involved in tissue specific expression. Presence of KLF4 has been shown to confer tissue specific expression of keratin 19 in pancreatic ductal cells (23). Acinar cells that do not express keratin 19 lack KLF4. We suggest that like keratin 19, LAMA3A expression relies on the presence of KLF4; cells that express KLF4 will express LAMA3A and those that do not express KLF4 will not express LAMA3A.
These studies do not address the mechanism of loss of KLF4 in the breast cancer cells. Further analysis is necessary to determine whether lack of KLF4 is due to gene loss, transcriptional repression, or protein instability. Studies in colon cells suggest that transcription of KLF4 is reduced in cancer cells. It has been shown that KLF4 mRNA is decreased in colon cancer cells and has an effect on the proliferation and differentiation of these cells (24,25). Dang et al. (26) find that expression of KLF4 is decreased in patients with familial adenomatous polyposis. The actual regulation of KLF4 is not very well understood at this time. Several factors including Cdx2, Sp1, Sp3, and KLF4 itself may be involved in its regulation (27).
Higher doses of KLF4 results in declining LAMA3A promoter activation in both MCF10A and cancer cells. The mechanism for this is unknown. KLF4 has been shown to have both activation and repression domains (8). Also, the Cyclin D1 promoter has been shown to be repressed by KLF4 by competing away the positive acting transcription factor SP1 (18). Further studies will be necessary to determine the mechanism by which large amounts of KLF4 can result in inhibition of LAMA3A transcription.
It is unclear whether AP-1 and KLF4 form a complex in MCF10A cells. In the gel shift of Probe A that contains both AP-1-and KLF4-binding sites, we see that an AP-1 competitor competes only the AP-1 band while a KLF4 competitor competes both the AP-1 and KLF4 bands. However, there is no effect on AP-1 binding with the KLF4 oligo in the breast cancer cells. It is possible that the competition of AP-1 in the MCF10A cells with KLF4 is due to an interaction of the two transcription factors. Further studies would be necessary to confirm this finding.
In summary, this study clearly demonstrates a relationship between LAMA3A expression and KLF4. The MCF10A cells which express LAMA3A express KLF4 and have high levels of KLF4 binding activity while all the breast cancer cell lines tested lacked KLF4 expression and activity. Laminin-5 in conjunction with many other pathways mediates the differentiation of mammary epithelial cells. Therefore, lack of KLF4 may attribute to the undifferentiated phenotype of breast cancer cells.