Dual Roles of the Transcription Factor Grainyhead-like 2 (GRHL2) in Breast Cancer*

Background: The role of the developmental transcription factor GRHL2 in breast carcinogenesis is ill defined. Results: Loss of GRHL2 expression induces an epithelial-to-mesenchymal transition and a reduction in cancer cell proliferation. GRHL2 and ZEB1 transcription factors form a negative feedback loop. Conclusion: GRHL2 exhibits dual roles in breast cancer. Significance: This study suggests the significance of GRHL2 in breast carcinogenesis. Using a retrovirus-mediated cDNA expression cloning approach, we identified the grainyhead-like 2 (GRHL2) transcription factor as novel protooncogene. Overexpression of GRHL2 in NIH3T3 cells induced striking morphological changes, an increase in cell proliferation, anchorage-independent growth, and tumor growth in vivo. By combining a microarray analysis and a phylogenetic footprinting analysis with various biochemical assays, we identified the epidermal growth factor receptor family member Erbb3 as a novel GRHL2 target gene. In breast cancer cell lines, shRNA-mediated knockdown of GRHL2 expression or functional inactivation of GRHL2 using dominant negative GRHL2 proteins induces down-regulation of ERBB3 gene expression, a striking reduction in cell proliferation, and morphological and phenotypical alterations characteristic of an epithelial-to-mesenchymal transition (EMT), thus implying contradictory roles of GRHL2 in breast carcinogenesis. Interestingly, we could further demonstrate that expression of GRHL2 is directly suppressed by the transcription factor zinc finger enhancer-binding protein 1 (ZEB1), which in turn is a direct target for repression by GRHL2, suggesting that the EMT transcription factors GRHL2 and ZEB1 form a double negative regulatory feedback loop in breast cancer cells. Finally, a comprehensive immunohistochemical analysis of GRHL2 expression in primary breast cancers showed loss of GRHL2 expression at the invasive front of primary tumors. A pathophysiological relevance of GRHL2 in breast cancer metastasis is further demonstrated by our finding of a statistically significant association between loss of GRHL2 expression in primary breast cancers and lymph node metastasis. We thus demonstrate a crucial role of GRHL2 in breast carcinogenesis.

Using a retrovirus-mediated cDNA expression cloning approach, we identified the grainyhead-like 2 (GRHL2) transcription factor as novel protooncogene. Overexpression of GRHL2 in NIH3T3 cells induced striking morphological changes, an increase in cell proliferation, anchorage-independent growth, and tumor growth in vivo. By combining a microarray analysis and a phylogenetic footprinting analysis with various biochemical assays, we identified the epidermal growth factor receptor family member Erbb3 as a novel GRHL2 target gene. In breast cancer cell lines, shRNA-mediated knockdown of GRHL2 expression or functional inactivation of GRHL2 using dominant negative GRHL2 proteins induces down-regulation of ERBB3 gene expression, a striking reduction in cell proliferation, and morphological and phenotypical alterations characteristic of an epithelial-to-mesenchymal transition (EMT), thus implying contradictory roles of GRHL2 in breast carcinogenesis. Interestingly, we could further demonstrate that expression of GRHL2 is directly suppressed by the transcription factor zinc finger enhancer-binding protein 1 (ZEB1), which in turn is a direct target for repression by GRHL2, suggesting that the EMT transcription factors GRHL2 and ZEB1 form a double negative regulatory feedback loop in breast cancer cells. Finally, a comprehensive immunohistochemical analysis of GRHL2 expression in primary breast cancers showed loss of GRHL2 expression at the invasive front of primary tumors. A pathophysiological relevance of GRHL2 in breast cancer metastasis is further demonstrated by our finding of a statistically significant association between loss of GRHL2 expression in primary breast cancers and lymph node metastasis. We thus demonstrate a crucial role of GRHL2 in breast carcinogenesis.
The Drosophila gene grainyhead (dGrh) is the founding member of a large family of genes encoding evolutionarily conserved transcription factors that play key regulatory roles during embryonic development. Based on sequence similarity, members of this gene family are grouped into two different branches (1). One phylogenetic arm, with Drosophila CP2 (dCP2) as an ancestral gene, includes the transcription factors CP2, LBP-1a, and LBP-9. These genes exhibit a wide tissue distribution and regulate diverse cellular processes. The other phylogenetic branch, with dGrh as an ancestral gene, encompasses the closely related grainyhead-like (GRHL) transcription factors GRHL1-3 (1)(2)(3). In contrast to dCP2-related transcription factors, GRHL1-3 proteins exhibit much more restricted and distinct expression patterns during murine development, suggesting that they are not functionally redundant proteins (4). Among the GRHL family members, GRHL2 probably represents the least characterized transcription factor. A fundamental role of GRHL2 proteins in both neural and non-neural developmental processes is reflected by the fact that Grhl2-null mutant mice die by embryonic day 11.5 (5). Grhl2-deficient mice exhibit neural tube closure defects and severe defects in the formation of apical junctional complexes in several types of epithelia (5). A more detailed molecular analysis revealed that GRHL2 regulates epithelial differentiation by directly modulating expression of the adherens junction gene Cdh1 (E-cadherin) and the tight junction gene Cldn4 (claudin 4) (5). Likewise, mice with an N-ethyl-N-nitrosourea-induced nonsense mutation in the Grhl2 gene die by embryonic day 12.5 due to defects in neural tube closure and heart development (6). Although these and several other developmental studies (4,(7)(8)(9) clearly established a crucial role of GRHL2 in embryonic development, an implication of GRHL2 in other physiological processes, such as, for example, wound healing and cancer, is less well defined. This is surprising because two members of the grainyhead family of transcription factors, namely dGrh and GRHL3, have attracted considerable interest in that these genes could be identified as important regulators in epithelial barrier formation and wound healing in flies and vertebrates, respec-tively (10 -12). It has been known for a long time that wound healing and carcinogenesis represent closely related physiological processes characterized by an increased cell proliferation, extensive tissue remodeling, blood vessel formation, and an inflammatory response (13). Despite fundamental differences between both pathological processes (14), it has been hypothesized that factors involved in wound healing potentially also could play a crucial role in cancer, and vice versa.
To date, however, evidence has been reported for both tumor-promoting and -suppressing activities of the GRHL2 transcription factor in tumorigenesis. For example, GRHL2 has been demonstrated to positively regulate expression of the human telomerase reverse transcriptase (hTERT) gene during cellular immortalization of oral squamous cell carcinoma cells (15,16) and to inhibit apoptosis by suppressing death receptor (FAS and DR5) expression in breast cancer cells (17). GRHL2 also has been shown to regulate proliferation of hepatocellular carcinoma cells (18) while acting as a suppressor of epithelialto-mesenchymal transition (EMT) 3 in breast cancer (19,20). Collectively, these observations strongly suggest a fundamental but probably tissue type-specific function of GRHL2 in carcinogenesis.
Using a genetic screen, we identified the GRHL2 transcription factor as a novel protooncogene capable of transforming NIH3T3 fibroblasts. We also provide evidence for dual roles of GRHL2 in breast cancer in that loss of GRHL2 induces a striking reduction in cell proliferation and EMT-like alterations. Moreover, our results further suggest the existence of a highly complex, interconnected GRHL2/ZEB1/miR-200 regulatory network in breast cancer cells. Finally, a comprehensive immunohistochemical analysis of GRHL2 expression in primary breast cancers further demonstrates the pathophysiological relevance of GRHL2 in breast carcinogenesis.
Retrovirus-mediated cDNA Expression Cloning of Protooncogenes-Poly(A) ϩ RNA from GI-101 cancer cells prepared with a FastTrack mRNA extraction kit (Invitrogen) was converted into an random-primed, non-directional retroviral cDNA expression library (Ͼ1 ϫ 10 7 independent clones, average insert size 1.5 kb) in the retroviral expression plasmid pMXs using the SuperScript plasmid system (Invitrogen). The retroviral cDNA expression library was converted to retroviruses by transient transfection into Phoenix-eco packaging cells. Fortyeight hours post-transfection, the infectious retroviral supernatant was harvested and filtered and was then used to infect NIH3T3 fibroblasts in the presence of 4 g/ml Polybrene (Sigma). The medium containing retroviruses was replaced with fresh culture medium 6 -8 h after the beginning of the infection, after which the NIH3T3 cell culture was fed at 3-day intervals with fresh medium. Foci, appearing 14 days after reaching cell confluence, were picked from infected NIH3T3 cultures and expanded, and genomic DNA was isolated using a DNeasy tissue kit (Qiagen). Proviral cDNA inserts were recovered with oligonucleotides annealing to retroviral vector sequences (pMXs-1, 5Ј-GGGTGGACCATCCTCTAGACTGC-3Ј; pMXs-2, 5Ј-AACCTACAGGTGGGGTCTTTCATTCC-3Ј) using Pfu DNA polymerase. PCR amplification products were then reintegrated into EcoRI/NotI or BamHI/NotI sites of the pMXs plasmid. Following conversion to retroviruses, individual plasmids were subjected to a second round of selection using the NIH3T3 focus assay. Plasmid clones tested positive for transformation were sequenced, and the identity of cDNA fragments was determined by a BLAST search (25).
Transformation Assays-Determination of growth rate, anchorage-independent growth (using soft agar assays), and tumorigenicity in athymic nude (nu/nu) mice of parental and transfected NIH3T3 cell clones was performed as described (26).
Microarray Analysis-Total RNA was extracted from parental and GRHL2-expressing NIH3T3 cells using the RNeasy kit (Qiagen). For each cell culture, three samples taken from different passages were utilized for expression profiling using Agilent Whole Mouse Genome 4 ϫ 44,000 microarrays (Agilent Technologies) according to the manufacturer's instructions using quantile normalization (27). Differential expression was calculated via fitting expression levels from parental and transfected cell lines to a linear model with moderated t statistics. Genes that were at least 2-fold (log 2 scale) up-or down-regulated at an adjusted p value of Յ1EϪ5 were considered to be differentially expressed. Microarray data sets are available at the NCBI Gene Expression Omnibus (GEO) Web site under accession code GSE43610.
Quantitative Real-time RT-PCR Analysis (qRT-PCR)-Differential mRNA expression was analyzed following extraction of total RNA from cells and reverse transcription using Superscript II (Invitrogen) and random hexamers. First strand reverse transcribed cDNA was then diluted 1:20 in water before use in real-time PCR. Primers were used together with the Quantitect TM -SYBR Green-Mastermix (Qiagen) in a Realplex 4 -PCR system (Eppendorf) according to the manufacturer's instructions. Primer sequences and PCR conditions are available upon request. Real-time PCR data analysis was performed using the ⌬⌬CT method with RPLP0 or Gapdh as an endogenous reference.
Luciferase Reporter Gene Assays-Promoter elements from the mouse Erbb3 gene were PCR-amplified and ligated to the firefly luciferase reporter gene by insertion into the XhoI/ HindIII cloning sites of the pGL4.10 vector (Promega). HEK 293T cells were transiently cotransfected with individual reporter plasmids and either the empty pIRES-N1 vector or the pIRES-N1 vector encoding the FLAG-tagged GRHL2 using Lipofectamine 2000. Cotransfection with the pGL4.74 plasmid containing the Renilla luciferase cDNA under the control of the HSV-TK promoter served as a control for the different transfection efficiencies. Forty-eight hours post-transfection, cell extracts were assayed in triplicate for luciferase activities using the Dual-Luciferase reporter assay system (Promega) and a GloMax TM 20/20 luminometer (Promega). Before calculating the -fold activation value, the luciferase activity of each sample was normalized to the Renilla luciferase activity.
Lentivirus-mediated Knockdown of Gene Expression-Lentiviral pLKO.1 shRNA vectors targeted against human GRHL2 (GRHL2#1, TRCN0000015810; GRHL2#2, TRCN0000015812) and human ZEB1 (ZEB1#1, TRCN0000017563; ZEB1#2, TRCN0000017565) were obtained from the RNAi Consortium (Broad Institute of Harvard and Massachusetts Institute of Technology) (30). pLKO.1 vectors harboring a scrambled nontarget shRNA sequence or a GFP-specific shRNA sequence (Addgene) served as negative controls. Lentiviruses were made by transfection of HEK 293T packaging cells with these constructs by using a three-plasmid system according to standard protocols. Supernatants were harvested, sterile filtered, and then used to infect target cells in the presence of 8 g/ml Polybrene. Pooled stable transfectants were established using puromycin selection. Stable transfectant cells were maintained in medium containing 1 g/ml puromycin.
In Silico Promoter Analyses-For in silico analysis of promoter sequences, we focused on significantly differently expressed genes (adjusted p value Յ 1EϪ5) for which mousehuman ortholog pairs were available from Ensembl Biomart (release 69) (32). Briefly, promoter sequences (1 kb upstream) for transcripts flagged as belonging to the consensus coding sequence (CCDS) at respective gene loci were obtained. Sequences were analyzed for individual matches of the GRHL2 motif defined by a frequency matrix (9), using a regulatory sequence analysis tool (RSAT, matrix-scan, upper threshold 1EϪ3, organism-specific background) (33). Reported normalized weights for unique sites were averaged per gene per species and across all hits. Gene ontology terms were also obtained from the Ensembl Biomart service. Putative transcription factor binding sites located within the proximal promoter of GRHL2 (1 kb upstream of the transcription start site) were extracted from the UCSC track "Transcription Factor ChIPseq" from ENCODE (34) using an in-house Ruby script. The ZEB1 ChIP-seq peak (chr8: 102504446 -102504669) was then screened for exact matches of putative ZEB1 binding sites (35). The software MUSCLE (version 3.8.31) (36) was employed to calculate a multiple-sequence alignment using the ZEB1 ChIPseq peak sequence and homologous sequences from chimpanzee, mouse, cattle, and chicken. Sequence alignment was visualized using Unipro UGENE software (37).
Chromatin Immunoprecipitation (ChIP) Assay-ChIP assays were performed using a ChIP-IT assay kit (Active Motif) according to the manufacturer's instructions with slight modifications. Briefly, cells were cross-linked with 1% formaldehyde in culture medium for 5 min at room temperature. After cell lysis, chromatin was fragmented by sonication. A fraction of soluble chromatin was saved for measurement of total chromatin input. ChIP assays were performed using antibodies specific for GRHL2 (anti-GRHL2-P2 and anti-GRHL2 (A01); Abnova) or ZEB1 (clone H-102 (Santa Cruz Biotechnology, Inc.) and clone HPA027524 (Sigma)). In control experiments, no antiserum or negative and positive control antibodies provided by the manufacturer (Active Motif) were used. DNA was purified from the immunoprecipitates using a PCR purification kit (Qiagen). Subsequently, promoter sequences were PCR-amplified using primers specific for the Erbb3 (5Ј-GAGGTGGCAT-GAACGCTAG-3Ј and 5Ј-GATGCGCGTAAGTGATGGA-AAG-3Ј) or GRHL2 (5Ј-CCACAATCCCTAGTGTTTGCAG-GTGTTGC-3Ј and 5Ј-CTAAAGGGTACAAGCCCGAGGG-ACGAGC-3Ј) regulatory regions, respectively. Murine and human GAPDH promoter-specific primers (Active Motif) were used in negative control reactions. PCR products were separated on 2% agarose gels and were visualized by ethidium bromide staining.
Immunohistochemistry-Paraffin-embedded tissue specimens of 55 breast cancer patients and a high density breast cancer prognosis tissue micoarray containing 36 control tissues, 34 normal breast tissues, and 2271 formalin-fixed, paraffin-embedded primary breast tumor specimens with available clinical follow-up and histopathologic data (38) were evaluated immunohistochemically for GRHL2 expression. Tissue sections were deparaffinized and then were subjected to pressure cooker pretreatment in citrate buffer (BioGenex) for 5 min at 125°C. After blocking of endogenous peroxidase activity using DAKO REAL TM peroxidase blocking solution (Dako), the polyclonal anti-GRHL2-P2 antibody (1:500 dilution) was applied overnight at 4°C. Subsequently, slides were incubated with peroxidase-labeled EnVision TM polymer coupled with goat antirabbit/mouse immunoglobulins (Dako) for 15 min at room temperature. GRHL2 expression was visualized using 3,3Ј-diaminobenzidine as chromogen in substrate buffer containing hydrogen peroxide. Finally, sections were counterstained with Mayer's hemalaun solution (Merck) and permanently mounted. For negative controls, the addition of primary antibodies was omitted. Staining intensity was estimated on a four-step scale (0, 1, 2, 3). The intensity of the nuclear GRHL2-specific immunoreaction was grouped into low (negative or weak) and high (moderate and strong).
Statistical Analyses-Calculations were performed using SPSS for Windows, version 18.0 (SPSS Inc., Chicago, IL). Correlations between GRHL2 expression in breast cancer tissues and clinicopathological parameters were analyzed by 2 or Fisher's exact tests. Two-tailed p values below 0.05 were considered statistically significant. Data from functional assays represent an average of at least triplicate samples. Error bars represent S.D. All experiments were repeated at least three times. Statistical significance was determined by the Mann-Whitney U test, and p Ͻ 0.05 was considered significant.

Transforming Activities of the GRHL2 Transcription Factor-
To identify genes possibly involved in breast carcinogenesis, a retrovirus-mediated cDNA expression cloning approach for the identification of protooncogenes, utilizing preneoplastic NIH3T3 fibroblasts as a recipient cell line, was employed. Briefly, a retroviral cDNA expression library prepared from GI-101 mammary carcinoma cells was converted to retroviral particles for the infection of contact-inhibited NIH3T3 fibroblasts. Foci appearing 2 weeks after cell confluence were isolated and expanded to isolate genomic DNA. To achieve a plasmid rescue, proviral cDNA inserts, recovered by PCR using primers derived from the retroviral expression plasmid, were reintegrated into the retroviral expression vector. Subsequently, individual expression constructs were subjected to a second round of selection using the NIH3T3 focus assay. Finally, the identities of cDNA clones exhibiting transforming activity were determined by DNA sequencing and BLAST search. Using this approach, a large number of cDNA clones with transforming activity was retrieved. As a proof-of-principle, the list of identified protooncogenes encompassed several well characterized oncogenes (e.g. RAF1, VAV2, and ECT2). Interestingly, two independent cDNA clones (positions 1-2390 and 1-3459 of the GRHL2 cDNA sequence (GenBank TM accession number NM_024915]) coding for the canonical, fulllength GRHL2 isoform were also isolated. Deviations from the published GRHL2 cDNA could not be identified by DNA sequencing, suggesting that transforming activity is associated with wild-type GRHL2 proteins and is not related to the presence of mutations within the GRHL2 molecule. To investigate the transforming potential of GRHL2, we stably expressed GRHL2 proteins in NIH3T3 cells (Fig. 1A) and subjected transfectants to various in vitro and in vivo transformation assays. Parental NIH3T3 fibroblasts and cells harboring an empty expression vector or NIH3T3 fibroblasts stably expressing an activated c-RAF kinase served as negative or positive controls, respectively. Overexpression of GRHL2 induced striking morphological changes (Fig. 1B) and a significant increase in cell proliferation as compared with parental or vector-transfected NIH3T3 cells (Fig. 1C). Furthermore, GRHL2 was capable of inducing and sustaining in vitro anchorage-independent growth of NIH3T3 cells (Fig. 1D). Most importantly, GRHL2 also promoted rapid tumor growth in nude mice (tumor latency of 13-31 days; Table 1). Thus, our results clearly demonstrate that GRHL2 represents a novel protooncogene.
Bioinformatic analysis suggested the existence of another GRHL2 gene transcript coding for a yet fully uncharacterized, var-iant GRHL2 isoform (GRHL2v). In contrast to canonical GRHL2, GRHL2v exhibited no oncogenic potential in NIH3T3 cells. Remarkably, in GRHL2v-transfected NIH3T3 cells, multiple GRHL2v gene products with apparent molecular masses of 54 kDa (major band) and 65-70 kDa (minor bands) could be detected by Western blot analysis using a GRHL2-specific antibody (Fig. 1E).
Sequence alignment of both GRHL2 mRNAs (Ensembl Transcript ID ENST00000251808 and ENST00000395927) showed that an internal 5Ј splicing site (AGGT; positions 146 -150) within the first exon of the GRHL2 gene is used to generate an alternate GRHL2-002 mRNA with a 203-base deletion containing the first AUG codon in its 5Ј-end. Therefore, initiation of transla-FIGURE 1. Overexpression of GRHL2 induces oncogenic transformation of NIH3T3 fibroblasts. A, NIH3T3 fibroblasts were stably transfected with an expression plasmid encoding GRHL2 or activated c-RAF kinase as confirmed by Western blot analysis using antibodies specific for GRHL2 (top) or c-RAF kinase (middle). Equal loading was demonstrated using an antibody recognizing HSC70 protein (bottom). B, GRHL2 mediates a morphological transformation of NIH3T3 fibroblasts. Parental cells, NIH3T3 cells transfected with empty vector, and NIH3T3 cells expressing an activated c-RAF kinase served as negative and positive controls, respectively. C, growth characteristics of NIH3T3 cells expressing GRHL2. Cells were plated at an initial density of 1 ϫ 10 4 cells/well in 6-well plates in growth medium on day 0. Cells from replicate plates were counted at the intervals indicated. D, GRHL2 induces anchorage-independent growth of NIH3T3 fibroblasts. Cells were analyzed for their ability to grow in soft agar. Growth of colonies in semisolid medium was visualized under a light microscope. Representative colony sizes after 14 days are shown. Quantitative data are included in Table 1. E, detection of GRHL2 gene products in NIH3T3 cells stably transfected with expression constructs encoding canonical (GRHL2) or variant GRHL2 proteins (GRHL2v) by Western blot analysis using a GRHL2-specific antibody. F, generation of alternate GRHL2 mRNAs by alternative splicing of the GRHL2 primary gene transcript (illustration is not to scale). G, expression of dominant negative GRHL2 (GRHL2-DN) proteins lacking the first 167 residues (GRHLv(168 -625)) in NIH3T3 (left) or T47-D cells (right). GRHL2 proteins were detected by Western blot analysis using a GRHL2-specific antibody. H, functional inactivation of GRHL2 by GRHL2-DN. GRHL2-transfected NIH3T3 cells were retrovirally transduced with an empty vector or an expression plasmid encoding GRHL2-DN proteins as verified by Western blot analysis using a GRHL2-specific antibody (left). The dominant negative effect of GRHL2-DN proteins on GRHL2 activity was confirmed by inhibition of GRHL2-induced anchorage-independent growth of NIH3T3 fibroblasts in soft agar (right) and by qRT-PCR analysis of GRHL2 target gene expression (bottom). -Fold change indicates alterations in expression of positively (Argef19) or negatively regulated (Fas) GRHL2-dependent genes relative to parental cells. Error bars, S.D.
tion of the GRHL2-002 transcript predictably occurs at the second AUG within exon 2 and results in an N-terminally truncated (Ϫ16 residues) GRHL2v isoform (GRHL2v(17-625)) with a disrupted transactivation domain (Fig. 1F). However, a computational analysis of the translation initiation site showed that the first AUG codon of the alternate GRHL2 mRNA is in a highly unfavorable sequence context for initiating translation (39), suggesting that translation of the GRHL2-002 mRNA could be initiated from downstream in-frame AUG codons. In fact, the pattern of GRHL2v gene products detected by Western blot analysis corresponded well with the calculated sizes of predicted GRHL2v(17-625), GRHL2v(51-625), GRHL2v(52-625), and GRHL2v(168 -625) proteins. To verify our hypothesis, we generated an expression plasmid encoding the putative major GRHL2v gene product (GRHL2v(168 -625); hereafter designated as GRHL2-DN) and retrovirally introduced it into NIH3T3 cells. A Western blot analysis showed that GRHL2-DN proteins co-migrate exactly with the major GRHL2v gene product and also with a smaller sized GRHL2 protein detectable in breast cancer cell extracts ( Fig. 1G; see also Fig. 3), suggesting a physiological relevance of GRHL2-DN proteins. To test whether GRHL2-DN proteins lacking the entire transactivation domain may exhibit dominant negative activity, we retrovirally transduced GRHL2-transformed NIH3T3 cells with a GRHL2-DN expression plasmid (Fig. 1H). GRHL2-DN was found to reverse morphological transformation of cells, to drastically reduce the ability of GRHL2-transformed NIH3T3 cells to grow in semisolid medium, and to significantly inhibit GRHL2-mediated gene expression (Fig. 1H). Collectively, these results led us to conclude that only the canonical GRHL2 isoform exhibits oncogenic potential, whereas the major GRHL2v gene product (GRHL2-DN) acts in a dominant negative fashion.
Identification of Erbb3 as a GRHL2 Target Gene-To better understand GRHL2-dependent regulatory networks leading to oncogenic transformation, it was important to identify potential GRHL2 target genes. To this end, a dual strategy combining gene expression analysis and a phylogenetic footprinting analysis was pursued. A comprehensive overview on GRHL2-responsive genes in our cell system was obtained by extracting total RNA from parental and GRHL2-expressing NIH3T3 cells and performing an expression profiling through high density oligonucleotide microarray analysis. A total of 637 genes were identified as being differentially expressed in GRHL2-expressing cells (253 up-regulated and 384 down-regulated), taking a cut-off of an at least 2-fold change in expression relative to parental NIH3T3 fibroblasts at the generally accepted significance threshold (multiple test corrected p value Յ0.05). To identify potential GRHL2 target genes, an in silico promoter analysis of regulatory regions of genes differentially expressed between parental and GRHL2-expressing NIH3T3 cells was performed. We also considered the phylogenetic conservation of these binding sites, assuming that motifs with a high degree of interspecies conservation are those that are most likely to have a physiologic relevance. We confined our analysis therefore to genes for which Ensembl lists a 1:1 ortholog mapping (n ϭ 121). The GRHL2 motif was represented by a position frequency matrix constructed from sequences recognized by the GRHL2 transcription factor (9). Using these criteria, we were able to identify 73 genes among our microarray candidate genes predicted to include a putative high scoring GRHL2 recognition sites in their promoter region. The validity of our computational analysis is demonstrated by the fact that the Rho guanine nucleotide exchange factor 19 (Arhgef19; mean score 0.97) and the death receptor Fas (mean score 0.92), both of which already have been shown to be directly regulated by GRHL2 (17), top the list of highest scoring, putative GRHL2 target genes (supplemental Table 1). Restricting the list of putative GRHL2 target genes to those flagged with the Gene Ontology term "regulation of cell proliferation," Erbb3, Fgfr2, Igf2, Fas, and Stat5a were considered relevant targets.
To experimentally validate the results obtained by phylogenetic footprinting analysis, we selected the Erbb3 gene, which is known to play a crucial role in carcinogenesis, for further studies. Computational analysis of the murine Erbb3 promoter region demonstrated the presence of a single conserved GRHL2 DNA-binding site (AACCGGCT) upstream of the transcription start site (from base pair Ϫ371 to Ϫ364). Binding of GRHL2 to the predicted DNA-binding site was demonstrated by EMSA ( Fig. 2A). Additionally, reporter gene assays using various wild-type or mutant murine Erbb3 promoter fragments linked to the firefly luciferase gene showed that activation of the murine Erbb3 promoter through GRHL2 is strictly dependent on the presence of an intact GRHL2 DNA-binding motif (Fig.  2B). Results obtained by phylogenetic footprinting analysis also were confirmed by ChIP assays showing occupancy of the Erbb3 promoter by GRHL2 proteins in vivo (Fig. 2C). In summary, our results strongly suggest that Erbb3 represents a novel GRHL2 target gene.  Fig. 1A. b Cells (3 ϫ 10 3 ) were suspended in a 0.35% agar solution and overlaid onto a 0.5% agar solution in growth medium in 60-mm dishes. One day after incubation, 2 ml of growth medium was added, and colonies were counted 14 days after plating. Each value represents the averages of three experiments with S.D. Representative colony sizes are shown in Fig. 1C. c Cells (1 ϫ 10 6 ) in 0.2 ml of serum-free DMEM were injected subcutaneously into the scapular region of 6 -8-week-old female nude mice (5 mice/group). Mice were checked daily for tumor formation at the site of inoculation for up to 70 days. Latency period was the time in days to produce visible tumors. d NA, not applicable.

Growth-promoting Activities of GRHL2 in Human Breast
Cancer Cells-To investigate whether GRHL2 also exhibits growth-promoting activities in human breast cancer cells, we first analyzed the expression of GRHL2 gene transcripts and proteins in a panel of human breast cancer cell lines. GRHL2 expression could be detected in eight of ten breast cancer cell lines at highly variable levels, as determined by Western blot analysis using an in-house GRHL2-specific polyclonal antibody and Northern blot hybridization, respectively (Fig. 3, A and B). The level of GRHL2 protein expression in each individual cell line corresponded well with the level of GRHL2 mRNA detected in these cells. Interestingly, GRHL2 isoform expression could be detected in all GRHL2-positive breast cancer cell lines by Western blot analysis (Fig. 3A) and non-quantitative RT-PCR analysis (Fig. 3C). It is important to note that the poorly differentiated, highly metastatic breast cancer cell lines MDA-MB-231 and BT-549 belonging to the "basal-B" subclass of breast cancer cell lines (40) were completely negative for GRHL2.
To examine whether GRHL2 promotes tumor growth, we first performed a series of lentivirus-mediated RNAi experiments to suppress GRHL2 expression in a variety of GRHL2expressing breast cancer cell lines (e.g. MCF-7, ZR-75-1, T47-D, and MDA-MB-468 cells). Two hairpin shRNAs that efficiently knocked down GRHL2 expression in breast cancer cells as compared with control cells (up to 90% silencing, as confirmed by Western blot analysis; Fig. 4A) could be identified. Interestingly, shRNA-mediated knockdown of GRHL2 expression in breast cancer cell lines caused cells to adopt a flat and significantly enlarged morphology. GRHL2-depleted breast cancer cells formed less prominent cell-cell contacts and exhibited increased cell scattering. In addition to these EMT-   AUGUST 9, 2013 • VOLUME 288 • NUMBER 32 like morphological changes, a marked down-regulation of ERBB3 protein expression and also a striking reduction in cell growth was observed. Representative results obtained for MDA-MB-468 breast cancer cells are shown in Fig. 4A. To further verify these results, we also retrovirally transduced a panel of human breast cancer cell lines with an expression construct coding for dominant negative GRHL2 proteins. We noticed that expression of GRHL2-DN proteins in breast cancer cells provoked a significant suppression of endogenous GRHL2 protein expression (Fig. 4B). Functional inactivation of GRHL2 using a dominant negative approach also resulted in EMT-like morphological changes, suppression of ERBB3 expression, and a remarkable reduction in cell proliferation (Fig. 4B). Next, MDA-MB-468 breast cancer cells expressing GRHL2-DN proteins were subjected to various in vitro assays. Functional inactivation of GRHL2 using a dominant negative approach significantly increased the ability of individual MDA-MB-468 cells to form colonies in a clonogenic assay (Fig. 4C) and to exhibit anchorage-independent growth in soft agar assays (Fig. 4D). Moreover, loss of GRHL2 expression also promotes breast cancer cell migration (Fig. 4E) and invasion (Fig. 4F) in Boyden chamber assays.

Dual Roles of GRHL2 Transcription Factor in Breast Cancer
We also stably expressed GRHL2 in GRHL2-deficient MDA-MB-231 and BT-549 cells and analyzed its influence on cell proliferation using an 3-(4,5-dimethylthiazol-2-yl)-2,5-diphe-nyltetrazolium bromide assay. Although forced overexpression of GRHL2 induced morphological changes reminiscent of a mesenchymal-to-epithelial transition (MET), no influence on cell proliferation of either basal-B type of breast cancer cell line could be detected (Fig. 5A). ERBB3 gene expression was found to be unchanged in MDA-MB-231 cells (Fig. 5A) or dramatically increased (about 35-fold) in BT-549 cells (data not shown). Overexpression of GRHL2 in MDA-MB-231 cells significantly suppressed cell migration (Fig. 5B) and invasion (Fig. 5C) in Boyden chamber assays, thus further supporting an EMT suppressor function of GRHL2 in breast cancer cells.
Additionally, we analyzed alterations in expression of EMTassociated genes and selected GRHL2 target genes (e.g. CDH1 (E-cadherin) and FAS) in all three experimental systems by qRT-PCR analysis. Data presented in Table 2 clearly indicate significant changes in a variety of EMT-related genes, including CDH1 (E-cadherin), KRT genes (cytokeratins), VIM (vimentin), CLDN genes, DSC2, ZEB1, and CD24 stem cell marker. GRHL2-induced changes in expression of selected genes were also confirmed by Western blot analysis (Fig. 5D). It is interesting to note that alterations in expression of selected GRHL2 target genes were not dose-dependent. For example, induction of E-cadherin and P-cadherin expression is significantly more pronounced in MDA-MB-231 cells expressing relatively low levels of GRHL2 (GRHL2-low) than in the GRHL2-high expres-  Table 2, were confirmed by Western blot analysis using antibodies as indicated on the right. Equal loading was demonstrated using an antibody recognizing HSC70 protein (A and D). AUGUST 9, 2013 • VOLUME 288 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 23001
sors. Most importantly, we also found evidence that GRHL2, which was identified as a repressor of death receptor (FAS, DR4) expression (17), caused a striking up-regulation of FAS and DR4 gene expression in GRHL2-transfected MDA-MB-231 breast cancer cells (Table 2). Collectively, our data strongly suggest that GRHL2 exhibits EMT suppressor activity and regulates ERBB3 gene expression and cell proliferation of breast cancer cells.
GRHL2 Antagonizes TGF-␤-induced EMT in MCF-10A Cells-The non-tumorigenic epithelial cell line MCF-10A probably represents the best characterized experimental system for the analysis of TGF-␤-induced EMT in human breast cancer (41). To obtain further evidence for a role of GRHL2 in suppressing EMT, we examined GRHL2 mRNA and protein expression in this experimental system. Although GRHL2 gene transcripts could be detected in MCF-10A cells by non-quantitative RT-PCR analysis at high cycle numbers (n ϭ 35), no GRHL2 protein expression could be detected by Western blot analysis using a GRHL2-specific antibody, suggesting a low level expression of GRHL2 in MCF-10A cells (Fig. 6, A and B). These observations prompted us to stably express GRHL2 in MCF-10A cells by means of retroviral gene transfer as demonstrated by Western blot analysis (Fig. 6A). Although no gross changes in the morphology of infected cells could be observed, we noticed slight morphological changes in subconfluent cultures in that GRHL2-transfected cells showed an increased tendency to grow as even more tightly packed clusters of cells. A qRT-PCR analysis of EMT marker gene expression revealed a striking down-regulation of selected mesenchymal marker proteins (e.g. vimentin) and a concomitant up-regulation of epithelial markers (e.g. CLDN4), indicating a reinforcement of the epithelial phenotype of MCF-10A cells by GRHL2 (Fig. 6D). We next analyzed the effect of TGF-␤ on parental, empty vectortransfected, and GRHL2-expressing MCF-10A cells and observed striking morphological changes indicative of a fully fledged EMT in parental and vector-transfected but not in GRHL2-expressing MCF-10A cells (Fig. 6C). These cells showed only very weak EMT-like morphological changes, and the scattering of cells could hardly be observed. A qRT-PCR analysis of EMT marker gene expression demonstrated a significant suppressive effect of GRHL2 on TGF-␤-induced gene expression changes in MCF-10A cells as compared with vectortransfected cells (Fig. 6D). GRHL2-induced changes in expression of selected genes were also confirmed by Western blot analysis (Fig. 6A). It is important to note, however, that the antagonizing effect of GRHL2 on TGF-␤-induced EMT in the MCF-10A experimental system was detectable but less evident when cells were grown in medium containing higher concentrations of TGF-␤ (10 ng/ml) for extended periods of time (Ͼ4 days). Still, our observations further substantiate an EMT suppressor function of GRHL2 in breast cancer cells.

GRHL2 and ZEB1 Transcription Factors Form a Double Negative Regulatory Feedback Loop-Little is known about the regulation of GRHL2 gene expression in cancer cells.
This prompted us to screen the GRHL2 regulatory region (1 kb upstream of the transcription start site) for transcription factor binding sites, especially for those known to mediate recruitment of EMT transcription factors. Using in silico promoter analysis, we identified three potential binding sites for the ZEB1 transcription factor, which itself very recently has been shown to be regulated by GRHL2 (19), in the human GRHL2 promoter sequence. ZEB1 has been demonstrated to interact with either E-box elements (especially 5Ј-CAGGTG-3Ј, 5Ј-CATGTG-3Ј, or 5Ј-CACCTG-3Ј) or Z-box elements (especially 5Ј-CAG-GTA-3Ј or 5Ј-TACCTG-3Ј) in the proximal promoters of target genes, such as, for example, the CDH1 (E-cadherin) gene (35). A cluster of putative ZEB1 binding sites consisting of two Z-box elements (at positions Ϫ129 and Ϫ106) and one E-box element (at position Ϫ76) was detected (Fig. 7A). All three repressive elements evolutionarily were highly conserved, suggesting a physiological relevance.
In fact, analysis of GRHL2 and ZEB1 protein expression in our panel of human breast cancer cell lines by Western blot analysis revealed an inverse pattern of expression (i.e. cell lines positive for GRHL2 were negative for ZEB1, and vice versa) (Fig. 7B). Next, we performed lentivirus-mediated RNAi experiments to knock down ZEB1 gene expression in MDA-MB-231 breast cancer cells. Results shown in Fig. 7C clearly demonstrate that repression of ZEB1 expression induces a striking up-regulation of GRHL2 mRNA and protein expression in MDA-MB-231 cells. To demonstrate involvement of the E/Zbox elements in the recruitment of ZEB1 to the GRHL2 regulatory region, we also conducted EMSAs using radiolabeled oligonucleotides harboring Z-box 1 and 2-or E-box 1-specific sequences, respectively. Our results clearly indicate significant binding of ZEB1 proteins to all three repressive E/Z-box ele-

GRHL2-induced alterations in gene expression in experimental breast cancer systems
Gene expression changes of at least 1.5-fold (up-or down-regulated) relative to controls are considered as significant and are highlighted in boldface type.   ments of the GRHL2 promoter region. Representative results obtained for Z-box 2-derived oligonucleotides and nuclear extracts from MDA-MB-231 cells are shown in Fig. 7D. Finally, occupancy of the GRHL2 promoter by ZEB1 proteins in vivo was confirmed by ChIP assays using two ZEB1-specific antibodies (Fig. 7E), suggesting that GRHL2 expression is repressed by the EMT transcription factor ZEB1. Because ZEB1 previously has been shown to be transcriptionally repressed by GRHL2 (19,20) (see also Table 2), our results suggest the existence of a double-negative ZEB1/GRHL2 transcriptional feedback loop in human breast cancer cells.
Expression of GRHL2 in Human Primary Breast Carcinomas-To investigate a pathophysiological relevance of a dysregulation of GRHL2 gene expression in breast cancer, we also evaluated GRHL2 expression in normal and malignant breast tissues by immunohistochemical analysis of a panel of formalin-fixed, paraffin-embedded tissue sections using an in-house GRHL2specific polyclonal antibody. Representative staining results obtained from a small cohort of tissue specimens (n ϭ 55) are shown in Fig. 8. Consistent with its proposed role in epithelial morphogenesis, nuclear GRHL2 expression could be detected in cells of normal mammary glands and in the majority of tumor cells in breast cancer tissues. Interestingly, we also identified a small proportion of primary breast carcinomas without or with only weak GRHL2 expression, suggesting a down-regulation of GRHL2 expression in these tumors during carcinogenesis. Moreover, we observed a striking loss of GRHL2 expression in tumor cells at the invasive front and in individual tumor cells with no discernible continuity with the parent tumor (Fig. 8, B  and D). To correlate GRHL2 results obtained by immunhistochemical analysis with clinico-pathological parameters, including age, nodal status, tumor stage, grade of differentiation, and histological tumor type, we also stained a breast cancer prognosis tissue microarray containing about 2000 tissue samples. The results summarized in supplemental Table 2 show that GRHL2 expression is significantly associated with estrogen and progester-one receptor expression (p Ͻ 0.0001). Also, we found statistically significant inverse correlations between an increased intensity of GRHL2 expression and increased tumor stages (p Ͻ 0.0001) or the presence of lymph node metastases (p ϭ 0.023). Thus, the results FIGURE 7. Direct suppression of GRHL2 expression by the transcription factor ZEB1. A, schematic representation of the human GRHL2 promoter region from positions Ϫ214 to ϩ10 relative to the transcription start site (TSS) corresponding to the ZEB1 ChIP-seq peak (chr8: 102504446 -102504669) as extracted from ENCODE. The relative positions and sequences of E/Z-boxes representing putative binding sites for the transcription factor ZEB1 are indicated. Sequence alignment of GRHL2 regulatory sequences with the corresponding sequences from chimpanzee, mouse, cattle, and chicken showed a high degree of evolutionary conservation of putative ZEB1 binding sites. B, inverse expression pattern of ZEB1 and GRHL2 in human breast carcinoma cell lines, as demonstrated by Western blot analysis using antibodies specific for ZEB1 and GRHL2, respectively. Equal loading was demonstrated using an antibody recognizing HSC70 protein. C, repression of ZEB1 expression induces up-regulation of GRHL2 expression. A stable knockdown of ZEB1 expression in MDA-MB-231 cells was achieved by lentiviral transduction of cells with two different ZEB1-specific shRNAs (ZEB1#1 and ZEB1#2). Non-infected, parental cells and cells lentivirally transduced with a non-target shRNA served as a control in these experiments. ZEB1 and GRHL2 protein expression levels were assessed by Western blot analysis using antibodies specific for ZEB1 and GRHL2. Equal loading was demonstrated using an antibody recognizing HSC70 protein (top). Levels of ZEB1 and GRHL2 mRNA expression in pooled stable MDA-MB-231 breast cancer cell lines were determined by qRT-PCR analysis (bottom). D, binding of ZEB1 to E/Z-box elements within the regulatory region of human GRHL2 was demonstrated by EMSA. Nuclear extracts prepared from MDA-MB-231 cells were incubated with a radiolabeled Z-box 2-derived oligonucleotide, and binding was analyzed by EMSA. Positions of free probe and shifted and supershifted ZEB1-containing complexes are indicated on the right. A binding reaction without nuclear extract served as a negative control. The specificity of the shift in migration of the labeled probe was demonstrated by competition EMSA using a 50-fold molar excess of unlabeled consensus (GRHL2-WT) or mutant (GRHL2-MT) oligonucleotides, respectively. The specificity of binding was further demonstrated by a supershift assay using anti-ZEB1 or unrelated anti-NFB p65 antibodies. Retarded complexes were separated from free probe by electrophoresis on native 4% polyacrylamide gels and were then visualized by autoradiography of the dried gels.

DISCUSSION
The identification and structural and functional characterization of novel cancer-related genes are a prerequisite for a better understanding of the molecular mechanisms driving carcinogenesis and for the development of innovative therapeutic approaches for the treatment of cancer patients. To isolate novel protooncogenes, a retrovirus-mediated cDNA expression cloning approach utilizing preneoplastic, murine NIH3T3 fibroblasts was employed. Using this genetic screen, we identified the GRHL2 transcription factor as a novel protooncogene capable of transforming NIH3T3 fibroblasts. GRHL2 therefore represents the first member of the grainyhead family of transcription factors known to induce malignant transformation of cells. We also describe the existence of GRHL2 isoforms generated by alternative splicing of the primary GRHL2 gene transcript and provide evidence that the corresponding gene products exhibit dominant negative effects on canonical GRHL2 proteins. These findings potentially could represent a novel mechanism of regulating GRHL2 activity in cancer cells.
To molecularly define the tumor-promoting activities of GRHL2, it was pivotal to identify potential GRHL2 target genes. A dual strategy combining gene expression analysis and a phylogenetic footprinting analysis was pursued and resulted in the identification of a large number of genes predicted to be directly regulated by GRHL2 (n ϭ 73), including the accepted GRHL2 target genes Arhgef19 (7) and Fas (17) and, most importantly, also the EGFR family member ERBB3. The identification of ERBB3 as a GRHL2 target gene in breast cancer cells has very important implications for breast cancer in that aberrant acti-vation of ERBB3, which forms oncogenic heterodimers with ERBB2 (HER2), has been linked to decreased survival in breast cancer (42) and has been functionally involved in resistance to targeted therapies, such as tamoxifen (43,44), trastuzumab (45,46), lapatinib (47), and gefitinib (48,49).
Our findings, however, are based on studies performed using murine NIH3T3 fibroblasts as a model system. A key question, therefore, is the pathophysiological relevance of our observations for human breast carcinogenesis. To address this issue, we stably overexpressed GRHL2 transcription factor in GRHL2deficient human breast cancer cell lines and analyzed its influence on cell proliferation. Consistent with results recently published by Cieply et al. (19), no significant impact on the rate of cell proliferation of any basal-B type of breast cancer cell line (e.g. MDA-MB-231 and BT-549 cells) could be observed. By contrast, shRNA-mediated knockdown of GRHL2 expression or functional inactivation of GRHL2 using GRHL2-DN proteins in a variety of GRHL2-expressing breast cancer cell lines resulted in a dramatic reduction of cell proliferation, suggesting that GRHL2 indeed exhibits strong tumor-promoting activities in a breast cancer subtype-specific fashion. The activity of a transcription factor in a cell depends on a variety of factors, including its expression level, the state of activity (which often is determined by various post-translational modifications), the presence of essential co-factors, and the occupancy of the regulatory region of target genes by other transcriptional regulators, to name a few. It is likely that essential co-factors are missing in the poorly differentiated, basal-B type human breast cancer cell lines to support GRHL2-driven regulation of epithelial cell proliferation.
Importantly, loss of GRHL2 expression in breast cancer cells is associated with EMT-like morphological changes and increased cell survival and migratory and invasive behavior of cells. Accordingly, reexpression of GRHL2 in GRHL2-deficient basal-B type breast cancer cell lines caused phenotypical alterations indicative of the inverse process, the MET. Thus, our data in conjunction with results recently published by others (19,20) certainly allow GRHL2 to be classified as an EMT suppressor molecule. The analysis of GRHL2-induced gene expression changes in our experimental model systems showing significant changes of selected EMT marker genes (e.g. CDH1, DSC2, ZO-1, KRT genes, CLDN genes, VIM, ZEB1, and CD24 stem cell marker) further support an important role of GRHL2 in regulating EMT. Overall, the GRHL2-dependent phenotypical and gene expression changes in our experimental systems were more pronounced in GRHL2-overexpressing cells than in GRHL2 knockdown experiments. Presumably, lentivirally delivered shRNAs may not lower the level of GRHL2 expression below that needed for full binding site occupancy of target genes; therefore, expression of GRHL2 target genes will be less affected. The experimental limitations of the siRNA technology in transcription factor biology are widely recognized (50).
The concept of coupling EMT with reduced cell proliferation is not novel, and other molecules controlling both processes, such as the YB-1 factor, could be identified (51,52). Although hyperproliferation generally is considered to be a hallmark of cancer cells, it has been hypothesized that hyperproliferation may in fact be detrimental to survival during tumor cell dissem- ination and metastasis (52). Shutdown of cell proliferation to preserve genomic integrity and to reduce the metabolic demands of cellular proliferation and, additionally, an increasing capability of anchorage-independent growth is beneficial for survival of carcinoma cells in the circulation and in the hostile microenvironment of secondary organs. A low proliferative activity may represent an integral part of the mesenchymal phenotype, allowing cancer cells to migrate and disseminate to secondary organs before reinitiating tumor cell growth (51). A frequent clinical observation in breast cancer is that recurrence of overt metastasis may occur many years after the removal of or seemingly successful therapy for the primary tumor. Presumably malignant cells, shed from the primary tumor mass, remain dormant but still viable all the time, only to express their tumorigenic potential at a later time point. The presence of micrometastatic, dormant cells is an accepted indicator of recurrent disease and poor prognosis (53,54). However, the molecular mechanisms allowing carcinoma cells to escape from dormancy still remain enigmatic. It is entirely conceivable that genes regulating EMT/MET processes as well as tumor cell proliferation and cancer cell survival, such as the EMT suppressor molecule GRHL2, represent prime candidate genes crucial for the establishment of clinically detectable metastases.
Another very important finding of our study is that expression of GRHL2 is repressed by the ZEB1 transcription factor in breast cancer cells. On the other hand, we also provide evidence that repression of GRHL2 function leads to an up-regulation of ZEB1 expression in breast cancer cells. In fact, the ZEB1 transcription factor very recently has convincingly been shown to be a direct target for repression by GRHL2 (19), suggesting that GRHL2 and ZEB1 transcription factors form a double negative regulatory feedback loop. It is important to note that ZEB1 also transcriptionally represses genes of the miR-200 family, which in turn post-transcriptionally inhibit expression of ZEB1 by binding to a highly conserved target site within the 3Ј-untranslated region (55) and that GRHL2 has been reported to up-regulate the expression of miR-200b/c family members in breast cancer cells (19). Collectively, our data therefore suggest the existence of a highly complex, interconnected GRHL2/ZEB1/ miR-200 regulatory system. Interestingly, a feedback loop consisting of GRHL2 and Nkx2-1 transcription factors has also been demonstrated to exist in lung epithelium (56), implying that feedback loop regulation of GRHL2 expression may emerge as a common theme in epithelial cell morphogenesis and differentiation. Autoregulatory transcriptional feedback loops result in a fine tuning of gene expression, stabilize transcriptional activity to withstand changes in microenvironmental cues, or reinforce changes in gene expression upon stimulation of cells by extracellular stimuli. Therefore, disruption of the GRHL2/ZEB1/miR-200 transcriptional regulatory network controlling EMT may ultimately contribute to breast cancer metastasis.
Conflicting results regarding GRHL2 expression in primary breast cancers and also its activity in breast cancer model systems were reported very recently (19,20,57). The results of our study, which represents the first comprehensive immunohistochemical analysis of GRHL2 protein expression in a certain tumor, point to an important role of GRHL2 during EMT in breast carcinogenesis. Although based on relatively small numbers, we found a statistically significant association between loss of GRHL2 expression and the presence of lymph node metastases in breast cancer (p ϭ 0.023). Furthermore, we also observed a striking loss of GRHL2 expression in tumor cells that bud off from the main tumor mass and in individual tumor cells with no discernible continuity with the parent tumor. Cells that have undergone EMT typically constitute a minor proportion of the primary tumor and are primarily found at tumor margins and have a low proliferative potential. In light of this, the results of our immunohistochemical analysis further substantiate an important role of GRHL2 during EMT in breast cancer. On the other hand, evidence was also reported for an up-regulation of GRHL2 expression in the majority of breast cancers except for those belonging to the "claudin-low" subclass as compared with normal breast tissues (19,20,57). Consistent with its fundamental role in preserving epithelial cell integrity, we observed strong GRHL2 expression in normal breast epithelium, which appeared not to be substantially lower than in tumor tissues. As discussed by Cieply et al. (19), an apparent up-regulation of GRHL2 mRNA, as determined by analysis of gene expression data sets, could represent an artifactual result of expansion of the epithelial cell compartment relative to normal mammary gland during tumor outgrowth. Still, an elevated GRHL2 expression also has been reported to correlate with a shorter relapse-free survival interval and increased risk of metastasis in breast cancer patients (20). These findings are supported by functional studies showing that overexpression of Grhl2 in murine 4T1 breast cancer cells significantly promotes tumor growth and metastasis (20). The discrepancies in results became even more apparent when GRHL2 expression and activity was analyzed in the non-tumorigenic epithelial cell line MCF-10A. Consistent with data published by Cieply et al. (19), we observed only very low levels of GRHL2 expression in this experimental system. Xiang et al. (20) reported significant GRHL2 expression in these cells, as determined by non-quantitative RT-PCR analysis, and demonstrated that knockdown of GRHL2 in MCF-10A cells induced EMT-like changes. Conversely, Yang et al. (57) very recently showed that overexpression of Grhl2 in MCF-10A cells provoked a mesenchymal phenotype, whereas we provide evidence herein that overexpression of GRHL2 rather preserves epithelial cell integrity and antagonizes TGF-␤-induced EMT in MCF-10A cells. At present, the reasons for obvious discrepancies in results from different studies are completely unclear. Further studies are therefore needed to unravel the still enigmatic and even contradictory roles of the EMT suppressor molecule GRHL2 in breast carcinogenesis.