Loss of BRMS1 Promotes a Mesenchymal Phenotype through NF-κB-Dependent Regulation of Twist1

Abstract

Breast cancer metastasis suppressor 1 (BRMS1) is downregulated in non-small cell lung cancer (NSCLC), and its reduction correlates with disease progression. Herein, we investigate the mechanisms through which loss of the BRMS1 gene contributes to epithelial-to-mesenchymal transition (EMT). Using a short hairpin RNA (shRNA) system, we show that loss of BRMS1 promotes basal and transforming growth factor beta-induced EMT in NSCLC cells. NSCLC cells expressing BRMS1 shRNAs (BRMS1 knockdown [BRMS1^KD]) display mesenchymal characteristics, including enhanced cell migration and differential regulation of the EMT markers. Mesenchymal phenotypes observed in BRMS1^KD cells are dependent on RelA/p65, the transcriptionally active subunit of nuclear factor kappa B (NF-κB). In addition, chromatin immunoprecipitation analysis demonstrates that loss of BRMS1 increases Twist1 promoter occupancy of RelA/p65 K310—a key histone modification associated with increased transcription. Knockdown of Twist1 results in reversal of BRMS1^KD-mediated EMT phenotypic changes. Moreover, in our animal model, BRMS1^KD/Twist1^KD double knockdown cells were less efficient in establishing lung tumors than BRMS1^KD cells. Collectively, this study demonstrates that loss of BRMS1 promotes malignant phenotypes that are dependent on NF-κB-dependent regulation of Twist1. These observations offer fresh insight into the mechanisms through which BRMS1 regulates the development of metastases in NSCLC.

INTRODUCTION

Although there have been significant advances in the identification of specific driver mutations in non-small cell lung cancer (NSCLC) oncogenesis, the majority of patients continue to die of metastatic disease (1, 2). Epithelial-to-mesenchymal transition (EMT) is the major mechanism in epithelial tumor progression and is a prerequisite for tumor dissemination and metastasis (3). A hallmark of EMT is the transcriptional repression of E-cadherin, which results in loss of epithelial cell-cell contacts, whereby cells become detached and migrate away from the organized epithelial tissue (4). In addition, with the initiation of the EMT cascade, there is a loss of cell polarity complex components, such as lethal giant larvae homolog 2 (LLGL2) and crumbs homolog 3 (5). In parallel, cells acquire mesenchymal features through upregulation of specific markers, including N-cadherin, vimentin, and fibronectin (6).

Although multiple growth factors and cytokines contribute to the induction of EMT, transforming growth factor β (TGF-β) is the primary mediator of this process (7–9). In response to TGF-β, cells upregulate the EMT-inducing transcription factor (EMT-TF) genes Twist 1, Snail, Slug, and ZEB, which in turn differentially regulate downstream genes responsible for promoting the mesenchymal phenotype (5, 10, 11). In a previous study, we showed that inhibition of the transcription factor nuclear factor kappa B (NF-κB) completely represses TGF-β- and tumor necrosis factor (TNF)-induced formation of EMT in lung cancer cells (12). The mechanisms by which NF-κB induces EMT include upregulation of EMT-TFs and stabilization of Snail (13–16).

The breast cancer metastasis suppressor 1 gene (BRMS1) is a metastasis suppressor gene that significantly reduces intratumoral expression in NSCLC, compared with that in adjacent tissue (17, 18). Importantly, we have shown that the reduction of BRMS1 expression at protein and mRNA levels significantly correlates with NSCLC progression and poor prognosis (18). We recently showed, in both NSCLC cell lines and human specimens, that the reduction of BRMS1 in NSCLC is secondary to RelA/p65-mediated methylation of the BRMS1 promoter (19, 20). The importance of BRMS1 in suppressing metastases has been demonstrated in melanoma, breast, hepatocellular, and ovarian carcinomas (18, 21–24). BRMS1 has been shown to function as a corepressor to inhibit NF-κB transactivation through deacetylation of the RelA/p65 subunit at K310 (25). Additional mechanisms by which BRMS1 functions include the regulation of phosphoinositide signaling (26), expression of microRNA (miRNA) (27), angiogenesis (28), and p300 histone acetyltransferase levels (29). Whereas metastasis suppressor family members NM23, CD44, MKK4, and Kiss1 have been shown to regulate EMT, the role of BRMS1 in EMT has not been previously explored.

In this study, we show that loss of endogenous BRMS1 significantly promotes basal and TGF-β-induced EMT in NSCLC cells, as evidenced by an epithelial-to-mesenchymal phenotypic switch, enhanced cell migration and invasion, and differential regulation of EMT markers. The phenotypic changes associated with loss of BRMS1 require NF-κB transcriptional activity and functional RelA/p65. Finally, we demonstrate that loss of BRMS1 expression promotes mesenchymal and malignant phenotypes through NF-κB-dependent expression of Twist1.

MATERIALS AND METHODS

Cell culture, antibodies, and reagents.

Human NSCLC cell lines (A549 and H1993) were obtained from the American Type Culture Collection and were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum (Life Technologies, Carlsbad, CA). Human embryonic kidney cells (HEK293T) were obtained from the American Type Culture Collection and were grown as described previously (25). Following standard surgical resection, human NSCLC specimens and adjacent noncancerous lung tissue were obtained at the Division of Thoracic Surgery, University of Virginia, and preserved in liquid nitrogen. Written informed consent was obtained from all patients, and the Human Investigations Committee at Memorial Sloan Kettering approved the study. The primary antibodies used were those to BRMS1, E-cadherin, vimentin, RNA polymerase II, RNA polymerase II phosphor-Ser.2, paxillin, anti-hemagglutinin (anti-HA)-epitope, and LLGL2 from Abcam, Cambridge, MA, those to β-actin, Twist1, and RelA/p65 from Santa Cruz Biotechnology, Dallas, TX, those to Smad3 and phosphor-Smad3 from Cell Signaling, Danvers, MA, and antibody to acetyl-p65 K310, which was described previously (25). Human TGF-β1 and collagen type IV were purchased from Sigma-Aldrich (St. Louis, MO). Adenoviral-Cre recombinase (Ad-Cre) was purchased from the University of Iowa (Iowa City, IA).

Plasmid construction.

pSicoR short hairpin RNA (shRNA) for BRMS1 and scramble shRNA were generated as described previously (34). The packaging plasmid DNAs (pMDLg/pRRE, pRSV-Rev, and pMD2.G) were purchased from Addgene (Cambridge, MA). pLKO-shRNA Twist1 and pLKO-shRNA scramble were kindly provided by Phuoc T. Tran (The Johns Hopkins School of Medicine, Baltimore, MD). pLKO-shRNA Zeb1 (TRCN0000017567) was purchased from Sigma-Aldrich. The pcDNA plasmids encoding the FLAG-RelA/p65 wild type and K310R mutant and the pCMV plasmid encoding hemagglutinin (HA) tagged-BRMS1 have been described previously (25). Twist1 promoter reporter genes were provided by Mien-Chie Hung (M. D. Anderson Cancer Center, Houston, TX).

Virus production and infection methodology.

For production of virus, 10 μg of lentiviral plasmid DNA pSicoR or pLKO and 5 μg each of the packaging plasmid DNAs (pMDLg/pRRE, pRSV-Rev, and pMD2.G) were cotransfected into HEK293T cells using Polyfect (Qiagen, Valencia, CA), as described previously (35). NSCLC A549 and H1993 cells were then infected with lentiviruses. A single green fluorescent protein (GFP)-positive cell was selected and expanded (34). For pLKO lentiviral infection, single cell clone selection was performed using puromycin (1 μg/ml) (Life Technologies, Carlsbad, CA) for 2 weeks, and the clones were screened for the level of Twist1 expression. Only clones with >80% of target gene knockdown (KD) were selected. NSCLC A549 and H1993 cells were treated with or without Ad-Cre at a multiplicity of infection (MOI) of 100:1, as described previously (34).

Total RNA isolation and qRT-PCR.

Total RNA was isolated using the RNeasy kit in accordance with the manufacturer's protocol (Qiagen, Valencia, CA). Quantitative reverse transcription-PCRs (qRT-PCRs) were performed as described previously (25). The primers used in this study are listed in Table 1.

TABLE 1.

Primers used in qRT-PCR in this study

Gene	Sequence
	Forward	Reverse
E-cadherin	5′-TTCCTCCCAATACATCTCCCTTCACAGCAG-3′	5′-CGAAGAAACAGCAAGAGCAGCAGAATCAGA-3′
Vimentin	5′-GGAGCTACGTGACTACGTCCA-3′	5′-GAGAAGTCCACCGAGTCCTG-3′
LLGL2	5′-GTTTAACAAGACGGTGGAGCA-3′	5′-AGAGCTTGATGGCTCCAGAA-3′
Twist1	5′-AGTCCGCAGTCTTACGAGGA-3′	5′-CCAGCTTGAGGGTCTGAATC-3′
BRMS1	5′-TGCAGCGGAGCCTCAAG-3′	5′-TCACATCCAGACAGAAGCCCT-3′
HPRT	5′-TTGGAAAGGGTGTTTATTCCTCA-3′	5′-TCCAGCAGGTCAGCAAAGAA-3′
Human ERV3	5′-ATGGGAAGCAAGGGAACTAAT-3′	5′-CCCAGCGAGCAATACAGAATTT-3′

Microarray analysis and GO analysis.

Total RNA was extracted from H1993 stable cell lines using the RNeasy kit (Qiagen) in accordance with the manufacturer's instructions. Illumina BeadArray analysis was performed in duplicate, using the HumanHT-12 v4 Expression BeadChip (Illumina, San Diego, CA), by the Integrated Genomics Operation core facility at Memorial Sloan Kettering Cancer Center. The raw data were extracted and analyzed using Partek Genomics Suite 6.6 (Partek, St. Louis, MO). Array data were filtered using a P value of <0.05. Gene signal values were logarithm transformed and normalized using the quantile method (36). Comparative analysis between control and knockdown cells was performed on the basis of fold change in expression levels. Gene ontology (GO) analysis was performed using Partek Genomics Suite 6.6.

Western blot analysis.

Western blotting was conducted as described previously (25). The primary antibodies were used at dilutions of 1:200 to 1:1,000, and the secondary antibodies (Santa Cruz Biotechnology, Dallas, TX) were used at a dilution of 1:5,000. In select experiments, the densitometry of specific immunoblots was measured using the ChemlDoc MP system (Bio-Rad, Hercules, CA), and the expression of target proteins was quantified by normalization with actin.

Luciferase reporter gene assays.

Twist1-luciferase reporter activity assays were performed as described previously (37). In brief, A549V and -I cells were plated at 40% confluence 24 h before transfection. On the second and third days, cells were cotransfected with individual Twist1-promoter reporter genes and plasmids encoding β-galactosidase and HA-tagged BRMS1 or an empty vector. All transfections were normalized with cytomegalovirus (CMV) β-galactosidase activity; the data represent the means ± standard errors of the means (SEM) of values from 3 independent experiments performed in triplicate.

ChIP assays.

Chromatin immunoprecipitation (ChIP) assays were performed as described previously (25). DNA was immunoprecipitated with 4 μg of antibody (anti-BRMS1, anti-p65, anti-acetyl-p65 K310, anti-phospho-S2 of RNA polymerase II [Pol II], or normal rabbit IgG) and purified. The region of the human Twist1 promoter containing the functional κB binding site (38) was targeted for amplification. The human GAPDH (glyceraldehyde-3-phosphate dehydrogenase gene) promoter was amplified as a control (19). The primers used for amplification of the Twist1 promoter include the following: forward, 5′ TTTGGGAGGACGAATTGTTAGACC-3′ and reverse, 5′-TGGGCGAGAGCTGCAGACTTGG-3′.

Immunofluorescence.

Cells were plated into chamber slides at 2.5 × 10⁴ cells per chamber. For TGF-β stimulation, cells were starved overnight using serum-free medium, followed by stimulation with TGF-β (2 ng/ml) for an additional 24 h. Sixty hours after plating, cells were fixed with 4% paraformaldehyde. For detection of vimentin, LLGL2, and paxillin, cells were permeabilized with 0.25 Triton X-100 for 10 min. After incubation with 1% bovine serum albumin in phosphate-buffered saline–Tween 20 (PBST) for 30 min, rhodamine-phalloidin (Life Technologies, Carlsbad, CA) was added, for detection of F-actin, at a 1:40 dilution for 30 min. Primary antibodies for detection of E-cadherin, vimentin, LLGL2, or Twist1 were used at a 1:100 dilution overnight at 4°C. Secondary antibodies labeled with red-fluorescent Alexa Fluor 594 dye (Life Technologies, Carlsbad, CA) were used at 1:2,000 dilutions for an additional 1 h. Stained coverslips were washed and mounted onto glass slides using UltraCruz mounting medium (Santa Cruz Biotechnology, Dallas, TX). The projected cell area was measured using ImageJ software (NIH, Bethesda, MD).

Transwell invasion assays.

Cell invasion was assessed as previously described (18). In brief, the chambers were precoated with type IV collagen (0.25 mg/ml) in 0.25% acetic acid overnight, and a 2.5 × 10⁴-cell suspension was plated into the chamber. After incubation for 36 h, cells were fixed and stained with 0.1% crystal violet. For TGF-β stimulation, cells were pretreated with or without TGF-β (2 ng/ml) for 24 h and plated into the chamber with TGF-β for an additional 24 h.

Wound-healing assays.

Twenty-four-well plates were seeded with 5 × 10⁴ cells per well. At 72 h, each well was scraped using a sterile wound scratcher. For TGF-β stimulation, cells were pretreated with or without TGF-β (2 ng/ml) for 24 h before scratching and then were incubated with or without TGF-β (2 ng/ml) for an additional 24 h. The wounds were observed, and the cells were photographed at the initial scratch and 24 h later. The migration of cells toward the wound was analyzed using ImageJ software and expressed as the percentage of wound closure (39).

Cell viability assays.

Cells were seeded into 96-well plates at 1 × 10³ cells per well. At 0 h, 24 h, 48 h, and 72 h, cell viability was determined using the CellTiter-Glo luminescent cell viability assay kit (Promega, Madison, WI) in accordance with the manufacturer's instructions. For TGF-β stimulation, cells were pretreated with or without TGF-β (2 ng/ml) for 24 h before seeding and then were incubated with or without TGF-β (2 ng/ml) for additional periods, as described above.

Cell attachment assays.

Cell attachment assays were performed as described previously (40). In brief, cells were seeded into 96-well plates at 5 × 10⁵ cells per well. At 120 min, nonadherent cells were washed with PBS, and attached cells were fixed with 5% glutaraldehyde for 20 min. Cells were stained using 0.1% crystal violet and counted under the microscope.

Soft agar colony formation.

A cell suspension containing 5 × 10³ cells in 0.3 ml of 0.4% agar medium was layered on top of 0.5 ml of 0.8% agar medium in a 24-well plate. After 24 h, RPMI 1640 medium (Life Technologies) supplemented with 10% fetal bovine serum (FBS) was added to the top of the agar. After 3 weeks of incubation, the colonies were photographed with a microscope at a ×5 magnification. Colonies were quantified using ImageJ software (http://rsb.info.nih.gov).

Tail vein injection assay.

All animal experiments were approved by the University of Virginia Animal Care and Use Committee. In brief, 4-week-old athymic nude mice (Taconic, Hudson, NY) underwent tail vein injection with 0.1 ml of PBS containing 1 × 10⁶ H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD; n = 10/group). Eight weeks after injection, the mice were euthanized, and necropsy was performed. To quantify tumor burden and verify the human origin of tumors, genomic DNA was extracted from 20 mg of the mouse lung tissue sample using a DNase blood and tissue kit (Qiagen, Valencia CA). The ratio of human genomic material to mouse genomic material was assessed by use of quantitative PCR (qPCR) primers specific to human endogenous retrovirus 3 (ERV3) and mouse tumor necrosis factor alpha (TNF-α) (Table 1) (12, 30).

Histopathologic examination.

After the mice were sacrificed, lungs were excised and fixed for 24 h in 10% buffered formalin (Richard Allan Scientific, Kalamazoo, MI) for histologic analysis. The tissues were embedded in paraffin blocks, and 5-μm paraffin tissue sections were cut and stained with hematoxylin and eosin.

Statistical analysis.

The results of all experiments represent the means ± standard errors of the means (SEM) of values from 3 separate experiments conducted in triplicate. Statistical differences were determined using Student's t test, when appropriate. P < 0.05 was considered to indicate statistical significance.

Microarray data accession number.

The Illumina microarray data have been submitted to the Gene Expression Omnibus (GEO [http://www.ncbi.nlm.nih.gov/geo]) under accession no. GSE62359.

RESULTS

Loss of BRMS1 promotes a mesenchymal cellular phenotype and enhances invasion and migration.

To determine whether loss of endogenous BRMS1 affects the NSCLC phenotype and cell migration, we used a lentivirus-based shRNA expression system (pSicoR) to knock down endogenous BRMS1 expression in 2 NSCLC lines, H1993 and A549. Compared with the control shRNA cells, cells with reduction of BRMS1 expression had morphological changes consistent with mesenchymal phenotypes. NSCLC cells expressing BRMS1 shRNA displayed a spindle shape, increased lamellipodia, increased cell spreading, and decreased cell polarity, compared with control cells (Fig. 1A). Moreover, consistent with our previous finding (34), knockdown of BRMS1 resulted in subcellular redistribution of the cell polarity marker paxillin (31) from the cytosol such that it accumulated at the leading edge of cells (Fig. 1B). BRMS1 knockdown (BRMS1^KD) cells had a 3- to 7-fold increase in invasion, compared with control shRNA cells, as measured by Transwell invasion assays (Fig. 1C). Differences in cell migration were observed in wound-healing assays, in which BRMS1^KD cells had increased wound surface closure compared with control shRNA cells (Fig. 1D). These data support the hypothesis that a reduction of endogenous BRMS1 promotes NSCLC cell migration and invasion.

FIG 1.

Knockdown of endogenous BRMS1 induces cell morphological changes and enhances invasion and migration. (A) Immunofluorescent (F-actin) micrographs show the morphological appearance of NSCLC H1993 and A549 cells infected with pSicoR vector containing BRMS1 shRNA or control shRNA sequence. The bar graph shows quantification of 30 cellular projected areas in each group. The results represent the means ± standard errors of the means (SEM). *, P < 0.05, compared with the control shRNA group. (B) Immunofluorescent micrographs show paxillin staining in NSCLC H1993 and A549 cells infected with BRMS1 or control shRNA sequence. Western blots show expression of endogenous BRMS1 and actin as controls. (C) NSCLC H1993 and A549 cells were infected with BRMS1 or control shRNA sequence. Invasion assays were conducted. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, compared with the control shRNA group. (D) The cell GFP micrographs show wound-scratch assays performed using NSCLC H1993 cells with BRMS1 or control shRNA sequence. The results of all experiments represent the means ± SEM from 3 separate experiments. *, P < 0.01, compared with the control shRNA group.

BRMS1 expression correlates with changes in select EMT markers.

Given the dramatic phenotypic changes observed in our BRMS1^KD cells, we sought to evaluate whether these differences were associated with changes in prototypical markers of EMT. Illumina microarray analysis was performed to determine changes in gene expression between control and BRMS1 shRNA-infected cells. We identified 2,256 genes that were differentially expressed (fold change of more than 2 or less than −2; P < 0.05) between control and BRMS1^KD cells; of these, 1,002 genes were upregulated and 1,254 were downregulated in BRMS1^KD cells. Comparative analysis indicated that the highest and lowest expression changes were 130-fold and −423-fold, respectively. To explore the biological processes that BRMS1^KD is involved in, we functionally classified these genes using GO terms (Fig. 2A). Five of these processes—multicellular-organismal (6.1%), developmental (20.95%), single-organismal (26.99%), signaling (0.48%), and cellular (14.64%)—were involved in EMT signaling in lung cancer (GO: 0061111 [http://www.ebi.ac.uk]). Next, we analyzed the expression changes of 25 selected EMT-related genes (32) in control and BRMS1^KD cells (Fig. 2B); 15 of these genes were differentially regulated (P < 0.05). Furthermore, qRT-PCR data confirmed that BRMS1^KD cells displayed a downregulation of E-cadherin and LLGL2 transcripts and an upregulation of the mesenchymal marker gene Vimentin (Fig. 2C). Correspondingly, there were decreased levels of E-cadherin and LLGL2 and increased levels of vimentin in BRMS1^KD cells, compared with those in the control shRNA cells (Fig. 2D and E). Thus, loss of BRMS1 expression alone correlates with basal changes in gene expression of mesenchymal markers commonly associated with EMT.

FIG 2.

Reduction of BRMS1 expression alters the expression of EMT marker genes. (A) Illumina microarrays were conducted using H1993 cells with BRMS1 shRNA or control shRNA. The chart diagram represents the GO analysis of the differentially regulated genes. (B) Heat map of the expression of selected EMT-related genes in H1993 cells with BRMS1 or control shRNA sequence. (C) qRT-PCR was performed to detect the indicated mRNA levels in NSCLC cells (A549 and H1993) infected with BRMS1 or control shRNA sequence. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with the control shRNA group. (D) Expression of the indicated proteins was detected by Western blots in NSCLC cells (A549 and H1993) infected with BRMS1 or control shRNA sequence. Actin was probed as a control. (E) Immunofluorescent merged micrographs show the expression of the indicated proteins (red) and nuclear counterstain (DAPI [4′,6-diamidino-2-phenylindole]; blue) in NSCLC H1993 cells infected with BRMS1 or control shRNA sequence.

Loss of endogenous BRMS1 magnifies TGF-β-induced EMT.

TGF-β signaling is the primary inducer of EMT in cancer progression and metastasis (33). To determine whether loss of BRMS1 enhances TGF-β-induced EMT, BRMS1^KD cells were stimulated with TGF-β and examined for induction of mesenchymal markers. NSCLC cells treated with TGF-β had reduced expression of E-cadherin and LLGL2 at both the transcript and protein levels (Fig. 3A and B). As expected, TGF-β increased mRNA and protein expression of vimentin. Importantly, the effect of TGF-β on the expression of these EMT markers was significantly enhanced in BRMS1^KD cells compared with control cells (Fig. 3A and B). To confirm these observations at an individual cell level, immunofluorescence was performed. Protein expression patterns were similar to results obtained by Western blotting (Fig. 3C). Moreover, BRMS1^KD cells had significantly increased TGF-β-induced migration and invasion, as determined by wound-healing and Transwell assays, respectively (Fig. 3D and E). Thus, loss of BRMS1 expression magnifies TGF-β-induced EMT to promote invasion into and migration to NSCLC cells.

FIG 3.

Loss of BRMS1 enhances TGF-β-induced EMT. (A) qRT-PCR was performed to detect the indicated mRNA levels in NSCLC cells (A549 and H1993) infected with BRMS1 shRNA or control shRNA sequence, after stimulation with or without TGF-β (2 ng/ml) for 12 h. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with vehicle treatment in the same shRNA group; #, P < 0.05, and ##, P < 0.01, compared with the same treatment in the control shRNA group. (B) Western blots were performed to access the indicated proteins in NSCLC cells (A549 and H1993) infected with BRMS1 or control shRNA sequence, after stimulation with or without TGF-β (2 ng/ml) for 24 h. (C) Immunofluorescent merged micrographs show indicated protein expression levels (red) and nuclear counterstain (DAPI; blue) in NSCLC H1993 cells infected with BRMS1 or control shRNA sequence ± TGF-β (2 ng/ml) for 24 h. (D) Cell GFP micrographs show wound-scratch assays performed using NSCLC H1993 cells with BRMS1 or control shRNA sequence with or without TGF-β stimulation (2 ng/ml) for 24 h. The results of all experiments represent the means ± SEM from 3 separate experiments. *, P < 0.01, compared with the vehicle treatment in the same shRNA group; #, P < 0.01, compared with the vehicle treatment in the control shRNA group. (E) Invasion assays were performed in NSCLC H1993 and A549 cells infected with BRMS1 or control shRNA pretreated with or without TGF-β (2 ng/ml) for 24 h. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with the vehicle treatment in the same shRNA group; #, P < 0.05, compared with the same treatment in the control shRNA group.

Twist1 mediates EMT following loss of BRMS1.

Ectopic overexpression of BRMS1 in breast cancer cells has been shown to downregulate expression of Twist1 (27). Moreover, our microarray data indicate that knockdown of endogenous BRMS1 significantly induced transcription of Twist1 (Fig. 2B). To explore the mechanisms through which BRMS1 regulates EMT, we confirmed the expression of Twist1 in our NSCLC model system by qRT-PCR. As shown in Fig. 4A, BRMS1^KD cells had increased basal levels of Twist1 mRNA, compared with those in control shRNA cells. Subsequent stimulation with TGF-β resulted in enhanced expression of Twist1 mRNA in BRMS1^KD cells, compared with controls (Fig. 4A). These data confirm that BRMS1 is associated with decreased expression of Twist1 transcript. Alternatively, loss of BRMS1 was associated with increased basal and TGF-β-induced transcription of Twist1, suggesting that BRMS1 modulates EMT processes, in part, through Twist1-dependent pathways.

FIG 4.

Upregulation of Twist1 mediates EMT following loss of BRMS1. (A) qRT-PCR was performed to detect Twist1 mRNA in NSCLC cells (A549 and H1993) infected with BRMS1 shRNA or control shRNA sequence with or without TGF-β (2 ng/ml) for 12 h. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with the vehicle treatment group with the same shRNA; #, P < 0.05, and ##, P < 0.01, compared with the control shRNA group with the same treatment. (B) Immunoblots show the expression of indicated proteins in NSCLC A549 and H1993 cells infected with BRMS1 or control shRNA ± Ad-Cre. (C) Immunofluorescent merged micrographs show Twist1 expression levels (red), nuclear counterstain (DAPI; blue), and GFP (green) in NSCLC H1993 cells infected with BRMS1 or control shRNA sequences ±Ad-Cre. (D) Heat map of the expression of selected EMT-related genes in H1993 BRMS1^WT/Twist1^WT, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD cells. (E) Quantitative RT-PCR was performed to detect mRNA levels of the indicated genes in NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD). The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with Twist1^WT with the same BRMS1 status; #, P < 0.05, compared with BRMS1^WT with the same Twist1 status. Western blots show that Twist1 and BRMS1 were specifically knocked down in these NSCLC H1993 cells. (F) qRT-PCR was performed to detect mRNA levels of the indicated genes in NSCLC H1993 cells (BRMS1^WT/Zeb1^WT, BRMS1^WT/Zeb1^KD, BRMS1^KD/Zeb1^WT, and BRMS1^KD/Zeb1^KD). #, P < 0.05, compared with BRMS1^WT with the same Zeb1 status. Western blots show that Zeb1 and BRMS1 were specifically knocked down in these NSCLC H1993 cells.

To confirm that the increases in Twist1 transcription following shRNA knockdown of BRMS1 were specific to BRMS1 and not related to possible off-targeting effects, NSCLC BRMS1^KD cells were treated with Ad-Cre, which excises the LoxP shRNA sequence from the pSicoR vector and rescues the expression of endogenous BRMS1. As shown in Fig. 4B and C, the addition of Ad-Cre suppressed the expression of Twist1 in NSCLC BRMS1^KD cells. Furthermore, following exposure of cells to Ad-Cre, immunoblots (Fig. 4B) and immunofluorescence assays (Fig. 4C) confirmed recovery of BRMS1 expression in BRMS1^KD cells compared with that in control cells. The recovery of BRMS1 expression results in reexpression of E-cadherin and LLGL2 and inhibition of Twist1 and vimentin levels (Fig. 4B). Collectively, these data confirm that changes in Twist1 expression and EMT marker expression following BRMS1 shRNA expression are BRMS1 specific and are not related to potential off-targeting effects.

To investigate the importance of Twist1 in BRMS1-regulated EMT, we generated stable single knockdown (Twist1^KD) and double knockdown (BRMS1^KD/Twist1^KD) NSCLC cells. To determine whether knockdown of Twist1 restores BRMS1^KD-induced EMT, we performed an Illumina microarray analysis to assess for EMT gene expression. As shown in Fig. 4D, knockdown of Twist1 significantly recovered most of the selected EMT-related genes that were differentially regulated by BRMS1^KD. To confirm expression of EMT markers in BRMS1^KD/Twist1^KD cells, qRT-PCR analysis was performed. Knockdown of Twist1 alone significantly increased the expression of E-cadherin and LLGL2 and decreased the expression of Vimentin, compared with levels in control cells (Fig. 4E). Since we observed that EMT-TF Zeb1 was also upregulated in BRMS1^KD cells (Fig. 2B), we sought to investigate whether Zeb1 was involved in BRMS1-regulated EMT. To experimentally address this question, we generated NSCLC H1993 stable single knockdown (Zeb1^KD) and double knockdown (BRMS1^KD/Zeb1^KD) cells. Additional knockdown of Zeb1 was not sufficient to recover the expression of the tested EMT-related genes that were regulated by BRMS1^KD (Fig. 4F). Therefore, these observations indicate that Twist1 is a key transcription factor responsible for EMT following loss of BRMS1, and knockdown of Twist1 is sufficient to restore BRMS1^KD-induced deregulation of EMT genes.

Knockdown of Twist1 abrogates the BRMS1^KD-induced EMT phenotype in vitro.

Consistent with our previous observations (34), BRMS1^KD cells exhibited a dramatically altered cell morphological appearance, with increases in lamellipodia and cell spreading and a loss of cell polarity, compared with those in vector controls (Fig. 1A). Importantly, the cell morphological appearance of the double knockdown BRMS1^KD/Twist1^KD cells closely resembled epithelial cells and featured less lamellipodia and cell spreading than in BRMS1^KD cells (Fig. 5A). Surprisingly, Twist1^KD cells exhibited a cobblestone epithelial cell morphological appearance (Fig. 5A, top row, second image), suggesting that basal expression of Twist1 is required for the dedifferentiated state of H1993 cells. In addition, double knockdown of BRMS1^KD/Twist1^KD cells restored the distribution of paxillin in a similar pattern to BRMS1^WT/Twist1^WT cells (Fig. 5B). To better characterize the biological functions of the double knockdown BRMS1^KD/Twist1^KD cells, we performed a series of biological function assays. As expected, the knockdown of Twist1 alone inhibited NSCLC migration and invasion compared with those in vector control cells. Furthermore, knockdown of BRMS1 alone resulted in a significant increase in cell invasion and migration (Fig. 5C and D). Whereas these cells did not show any difference in cell viability (Fig. 5E), we observed that single knockdown of BRMS1 (BRMS1^KD) significantly increased anchorage-independent cell growth, with more colonies formed in soft agar colony formation assays, and BRMS1^KD/Twist1^KD double knockdown slightly reduced colony formation, compared with of BRMS1^KD single knockdown (Fig. 5F). Furthermore, single knockdown of either BRMS1 or Twist1 slightly increased cell attachment, but BRMS1^KD/Twist1^KD double knockdown cells demonstrated a strong ability to attach (Fig. 5G). Collectively, these biological function assays indicated that knockdown of Twist1 was sufficient to restore epithelial characteristics but not all of the BRMS1^KD-induced biological changes.

FIG 5.

Knockdown of Twist1 abolishes BRMS1^KD-induced EMT in vitro. (A) Immunofluorescent (F-actin) micrographs show the morphological appearance of NSCLC H1993 cells infected with BRMS1 (BRMS1^KD) or control (BRMS1^WT) shRNA and Twist1 (Twist1^KD) or scramble (Twist1^WT) shRNA. The bar graph shows quantification of 30 cellular projected areas in each group. The results represent the means ± SEM. *, P < 0.05, compared with Twist1^WT with the same BRMS1 status; #, P < 0.05, compared with BRMS1^WT with the same Twist1 status. (B) Immunofluorescent micrographs show paxillin staining in the indicated NSCLC H1993 stable cells. (C) Invasion assays were conducted using NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD). The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.01, compared with Twist1^WT with the same BRMS1 status; #, P < 0.01, compared with BRMS1^WT with the same Twist1 status. (D) Cell GFP micrographs show wound-scratch assays performed using NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD). The results of all experiments represent the means ± SEM from 3 separate experiments. *, P < 0.001, compared with BRMS1^WT with the same Twist1 status; #, P < 0.001, compared with Twist1^WT with the same BRMS1 status. (E) NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD) were pretreated with or without TGF-β (2 ng/ml) for 24 h, and cell viability assays were conducted ± TGF-β (2 ng/ml) at the indicated time points. (F) Soft agar colony formation assays were conducted using NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD). The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.01, compared with BRMS1^WT with the same Twist1 status. (G) Cell attachment assays were conducted 120 min after seeding with NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD). *, P < 0.05, compared with BRMS1^WT/Twist1^WT with vehicle treatment; #, P < 0.05, compared with BRMS1^KD/Twist1^WT with vehicle treatment.

RelA/p65 is required for BRMS1-mediated changes in EMT markers.

Given that TGF-β-induced phosphorylation of Smad complexes (Smad3 and Smad4) plays a key role during induction of EMT (41, 42), we sought to clarify whether loss of BRMS1 affects TGFβ-induced Smad3 phosphorylation. As shown in Fig. 6A, knockdown of BRMS1 significantly increased basal phospho-Smad3 but not TGF-β-stimulated phosphorylation of Smad3. Inconsistently, we previously demonstrated that knockdown of BRMS1 significantly magnified the TGF-β-induced EMT phenotype. Therefore, this finding suggests that, in addition to Smad3, other signaling pathways perhaps participate in this process. We have shown that BRMS1 inactivates the transactivation potential of RelA/p65 by functioning as a corepressor to promote histone deacetylase 1 (HDAC-1)-mediated deacetylation of lysine 310 on RelA/p65 (25). A recent study by Li et al. showed that NF-κB binds the Twist1 promoter and transcriptionally regulates Twist1-mediated EMT in breast cancer cells (38). These two observations suggest that the BRMS1-mediated effects on EMT may be modulated through RelA/p65. To experimentally address this hypothesis, A549 cells that stably express SR-IκB, a dominant-negative inhibitor of NF-κB (A549I), or vector controls (A549V) (43) were infected with the BRMS1 shRNA and control shRNA. As expected, BRMS1 knockdown induced expression of Twist1 in A549V cells but not in A549I cells (Fig. 6B and C). Moreover, knockdown of BRMS1 in A549V control cells significantly decreased mRNA and protein levels of E-cadherin and LLGL2 but increased expression of vimentin (Fig. 6B and C). In contrast, A549I cells expressing the NF-κB inhibitor did not exhibit changes in mesenchymal markers following loss of BRMS1 expression (Fig. 6B and C). Stimulation with TGF-β significantly reduced expression of E-cadherin in A549V cells, although not in A549I cells, which lack functional NF-κB activity, indicating that the active form of NF-κB is critical for TGF-β-induced EMT. Furthermore, loss of BRMS1 abrogated the TGF-β-induced decrease of E-cadherin in both A549V and A549I cells. (Fig. 6D). Consistent with previous findings (38), we observed that, in the A549V cell line, Twist1 WT, D1, D2, and D3 promoter reporters were activated and that overexpression of BRMS1 significantly inhibited their activity (Fig. 6E). However, in the A549I cell line, none of these reporters was activated and the addition of BRMS1 did not affect their activity (Fig. 6E), indicating that (i) activated NF-κB is required for activation of Twist1 promoter, (ii) deletion of the first 3 κB binding sites failed to affect NF-κB-mediated activation of Twist1 promoter, (iii) the fourth κB binding site is the functional site, and (iv) BRMS1-regulated Twist1 promoter activity is NF-κB dependent. These results indicate that loss of BRMS1 potentiates the induction of EMT through an NF-κB-dependent process.

Fig 6.

Acetylation of RelA/p65 K310 is required to induce EMT following loss of BRMS1. (A) NSCLC H1993 and A549 cells were infected with BRMS1 shRNA or the control shRNA and stimulated with TGF-β (2 ng/ml). Cells were harvested at the indicated time points, and the levels of phospho-Smad3 and Smad3 were assessed by Western blotting. (B) NSCLC A549V and -I cells were infected with BRMS1 or control shRNA sequence. qRT-PCR was performed to detect the mRNA levels of the indicated genes. The results of all experiments represent the means ± SEM from 3 separate experiments conducted in triplicate. *, P < 0.05, and **, P < 0.01, compared with control shRNA in the same cell line; #, P < 0.05, and ##, P < 0.01, compared with the same shRNA sequence in A549V cells. (C) Expression of indicated proteins was assessed in NSCLC A549V and -I cells infected with BRMS1 or control shRNA by Western blots. (D) Immunofluorescent micrographs show E-cadherin staining in A549V and -I cells infected with BRMS1 or control shRNA after stimulation with or without TGF-β (2 ng/ml) for 24 h. (E) NSCLC A549V and -I cells were cotransfected with the indicated Twist1 promoter reporters and HA-BRMS1 or empty vector. Luciferase activity was determined. (Top) Schematic illustration of the Twist1 promoter: the black squares indicate 4 putative κB binding sites (D1, deletion of κB site I; D2, deletion of κB sites I and II; D3, deletion of κB sites I, II, and III; D4, deletion of all 4 κB sites). (Middle) The bar graph represents the activity of individual Twist1 promoter reporters. (Bottom) Western blots indicate the expression of HA-BRMS1 in each cell line. (F) NSCLC H1993 cells were cotransfected with RelA/p65 wild-type, K310R mutant, or empty vector control and HA-BRMS1 or empty vector control. Expression of the indicated proteins was detected by immunoblots. (G) NSCLC A549 and H1993 cells were infected with BRMS1 shRNA or the control shRNA. ChIP assays were performed, indicating chromatin-associated proteins were detected across the functional κB binding site in the Twist1 promoter.

We next asked whether ectopic expression of a construct encoding RelA/p65 K310R mutant protein would affect protein expression of mesenchymal markers. As shown in Fig. 6F, ectopic expression of RelA/p65 suppressed protein expression of epithelial markers (E-cadherin and LLGL2) and increased the expression of Twist1 and vimentin. Conversely, cells expressing the RelA/p65 K310R mutant were unable to suppress E-cadherin and LLGL2 expression or increase levels of Twist1 and vimentin. Importantly, ectopic expression of BRMS1 abolished the ability of wild-type RelA/p65 to regulate Twist1 and EMT markers but had no effect on cells expressing the mutant RelA/p65 K310R (Fig. 6F).

To confirm that loss of BRMS1 induces the transcription of Twist1 via acetylation of endogenous RelA/p65, ChIP assays were performed across the Twist1 promoter containing the functional κB binding site (38) and the GAPDH promoter as a control. We observed that RelA/p65 physically interacted with the Twist1 promoter. In addition, BRMS1 is associated with the Twist1 promoter (Fig. 6G). However, knockdown of BRMS1 dramatically increased the occupancy of acetyl-RelA/p65 K310 and phospho-Ser.2 of RNA polymerase II (Pol II) but not the total RelA/p65 on the Twist1 promoter (Fig. 6G). Thus, BRMS1 does not affect the interaction between RelA/p65 and DNA. However, loss of BRMS1 results in increased occupancy of acetyl-RelA/p65 K310 (a histone mark commonly associated with increased transcription) and phospho-Ser.2 of RNA Pol II (a marker of transcription initiation) (44) on the Twist1 promoter.

Twist1 drives the development of lung metastases in the absence of BRMS1 expression.

To evaluate the correlation between downregulation of BRMS1 and activation of NF-κB/Twist1, we assessed the protein levels of BRMS1, Twist1, and RelA/p65 in 10 pairs of human NSCLC samples and matched adjacent noncancerous tissue. Of these tumor samples, 5 had lymph node metastasis, and 5 did not. Consistent with our previous findings (18, 25), BRMS1 was downregulated in all 10 tested NSCLC samples compared with levels in adjacent tissue (Fig. 7A). Moreover, Twist1 and RelA/p65 were increased in most of the tested NSCLC samples (9 of 10), compared with levels in adjacent tissue (Fig. 7A). Interestingly, the increased levels of Twist1 were significantly higher in NSCLC samples with node metastasis than in samples without node metastasis (approximately 13-fold) (Fig. 7B). This finding confirms the previous reports that Twist1 in the primary tumor is a key factor driving metastasis in NSCLC (45, 46).

FIG 7.

Twist1 mediates BRMS1^KD-induced tumor metastasis in vivo. (A) The protein levels of BRMS1, Twist1, RelA/p65, and actin in 10 pairs of patient NSCLC tumor samples (T) and matched adjacent noncancerous tissues (N) were assessed by Western blotting. The immunoblot bands were quantified by densitometry and normalized with actin. Fold changes of tumor versus adjacent tissue were calculated and are labeled under each corresponding blot. (B) The bar graph demonstrates the fold changes of indicated proteins in tumors versus adjacent tissue. *, P < 0.01, compared with the relative expression of the same protein in tumors without nodal metastases. (C to F) NSCLC H1993 cells (BRMS1^WT/Twist1^WT, BRMS1^WT/Twist1^KD, BRMS1^KD/Twist1^WT, and BRMS1^KD/Twist1^KD, respectively) were injected into the tail veins of mice. The mice were euthanized after injection. (C) The bar graph shows the difference in lung weight. *, P < 0.05, compared with BRMS1^WT/Twist1^WT; #, P < 0.05, compared with BRMS1^KD/Twist1^WT. (D) The photographs illustrate the lungs of mice (upper), with hematoxylin and eosin–stained lung histology slides (lower), for each group. (E) Lung tumor burden was evaluated using genomic qPCR. The bar graph shows the ratio of human (ERV3) to mouse (TNF-α) genomic material in mouse lungs (n = 6). The results represent means ± SEM. *, P < 0.0005, compared with BRMS1^WT/Twist1^WT; #, P < 0.0001, compared with the groups with Twist1^KD. (F) Human-specific E-cadherin and LLGL2 in mouse lungs were evaluated by qRT-PCR and normalized using the ratio of human to mouse genomic material in each group (n = 4). *, P < 0.005, compared with BRMS1^WT/Twist1^WT; #, P < 0.001, compared with the groups with Twist1^KD.

To determine the impact of loss of BRMS1 and Twist1 on tumor EMT in vivo, we chose the tail vein injection model (47). Mice injected with BRMS1^KD/Twist1^KD cells had decreased tumor burden and metastatic foci in lungs, compared with mice injected with BRMS1^KD/Twist1^WT cells. This confirms that loss of BRMS1 expression in H1993 cells requires an intact Twist1 to develop lung metastases (Fig. 7C and D). Analysis of genomic DNA isolated from mouse lungs confirmed that (i) these tumor foci were of human origin and (ii) significantly more human genomic material is present in the BRMS1^KD group than in the control (BRMS1^WT/Twist1^WT [P = 0.0003]) or Twist1^KD (BRMS1^WT/Twist1^KD [P = 0.00001] or BRMS1^KD/Twist1^KD [P = 0.0001]) groups (Fig. 7E). In addition, to verify that knockdown of Twist1 can restore the epithelial phenotype in vivo, we checked the expression of E-cadherin and LLGL2 using human gene-specific primers. Consistent with our observations in vitro, BRMS1^KD reduced mRNA of both genes, and Twist1^KD significantly promoted the expression of these two genes (Fig. 7F). Importantly, BRMS1^KD/Twist1^KD double knockdown restored the expression of both genes (Fig. 7F). Collectively, these data indicated that Twist1 plays key roles to drive the EMT phenotype in tumor cells with decreased BRMS1.

DISCUSSION

In this study, we have shown that loss of BRMS1 promotes both basal and TGF-β-induced EMT in NSCLC cells. Knockdown of BRMS1 in NSCLC cells promotes the development of mesenchymal features, including morphological cell shape changes and cell spreading, increased lamellipodia, and loss of cell polarity. Consistent with efficient induction of EMT, BRMS1^KD cells display loss of epithelial markers (E-cadherin and LLGL2) and an upregulation of the mesenchymal marker vimentin, as well as an increase in cell migration and invasion.

Recently, our group reported that knockdown of BRMS1 in p53 null human immortalized bronchial epithelial cells that express oncogenic K-Ras^V12 promotes a mesenchymal phenotype and enhances cell migration (34). Consistent with these observations, we have demonstrated here that knockdown of BRMS1 also induces a mesenchymal phenotype in NSCLC cells, as evidenced by deregulation of EMT markers and increased cell migration and invasion. Therefore, our studies highlight a new function of BRMS1 as a key mediator for maintenance of an epithelial phenotype and repression of cell migration and invasion in premalignant (34) and malignant lung epithelial cells.

The connection between expression of BRMS1 and activation of NF-κB has been established by our group and by others (19, 25, 48). BRMS1 is a member of the mSin3 corepressor complex and functions as a corepressor to inhibit NF-κB transactivation through deacetylation of the RelA/p65 subunit at K310 (25). BRMS1 also indirectly regulates NF-κB transcription by inhibiting IκBα phosphorylation and NF-κB nuclear translocation (48). Our group recently showed that activation of RelA/p65 decreases BRMS1 expression through methylation of the BRMS1 promoter (19). Thus, BRMS1 and NF-κB antagonize each other's activity through negative-feedback mechanisms (19, 25). This negative feedback establishes a paradigm through which constitutive activation of NF-κB (49) results in loss of BRMS1 in human solid tumors. This loss of BRMS1 further enhances aberrant activation of NF-κB and deregulates NF-κB-dependent genes. NF-κB plays an essential role in the induction and maintenance of EMT in malignant epithelial cells in response to stimulation by the inflammatory cytokines TGF-β and TNF (12, 50, 51). We observed that knockdown of BRMS1 significantly magnified TGF-β-stimulated EMT in NSCLC cells. Moreover, activation of NF-κB was required for BRMS1 knockdown-induced EMT and cell invasion. Whereas previous studies demonstrated an EMT regulatory circuit between the metastasis suppressor Raf-kinase inhibitor protein, NF-κB, and Snail (52), we have demonstrated here that loss of BRMS1 alone promotes basal and TGF-β-induced mesenchymal characteristics through novel mechanisms that are dependent on NF-κB transcriptional activity.

The mechanisms through which NF-κB induces and maintains the mesenchymal state of cancer cells involve both direct transcriptional upregulation of EMT-TFs (such as Twist1, Snail, Zeb, and Slug) and stabilization of Snail protein (13–16). Our ChIP data from the present study indicate that both RelA/p65 and BRMS1 physically interact with the functional κB binding site in the Twist1 promoter. Whereas knockdown of BRMS1 significantly enhances the transcription of Twist1, total RelA/p65 associated with the Twist1 promoter is unaffected. Moreover, loss of BRMS1 increases the level of acetyl-RelA/p65 K310 and phospho-Ser.2 of RNA polymerase II on the Twist1 promoter, indicating that (i) acetylation of RelA/p65 is critical for NF-κB-mediated transcriptional upregulation of Twist1 and (ii) BRMS1 is a key transcriptional regulator of Twist1, which it accomplishes by coordinating the deacetylation of RelA/p65 on chromatin.

We observed that knockdown of BRMS1, as a transcriptional regulator, upregulated transcription of Zeb1 and Twist1 in NSCLC cells. However, in our model system, knockdown of Zeb1 could not fully restore BRMS1^KD-induced changes in EMT-related genes. Surprisingly, inactivation of NF-κB blocked BRMS1^KD-induced transcriptional activation of Twist1, suggesting that loss of BRMS1 drives NF-κB to specifically induce EMT through the upregulation of Twist1. Furthermore, by using Ad-Cre to reexpress BRMS1 in BRMS1^KD cells, we confirmed that loss of BRMS1 is responsible for upregulation of Twist1 and subsequent alternations of Twist1 downstream EMT markers. In support of our observation regarding activation of selective EMT-TFs, Pantuck and colleagues demonstrated that loss of VHL activity stimulated renal cell carcinoma cells to activate NF-κB and promote EMT through selective upregulation of Twist1 and Slug but not Snail, Zeb1, or Zeb2 (53). Interestingly, both BRMS1 and VHL can function as E3 ubiquitin ligases to suppress tumor progression (29, 54). Therefore, one plausible explanation of these observations is that the targets of E3 ligases BRMS1 and VHL, such as p300 and HIF1α, respectively, contribute to NF-κB-mediated selective upregulation of Twist1.

Welch et al. demonstrated a correlation between BRMS1 overexpression and decreased expression of Twist1 protein (27). We have extended those observations to show that loss of BRMS1 expression drives mesenchymal phenotypes and the development of lung metastases through the ability of NF-κB to upregulate Twist1 expression. Moreover, silencing of Twist1 expression not only inhibits the ability of BRMS1^KD cells to undergo EMT and increase cell invasion but also promotes a phenotypic switch from EMT to mesenchymal-to-epithelial transition (MET) in NSCLC cells. Consistent with previous findings (46), our data reveal that Twist1 is a primary factor driving the EMT process in NSCLC. Tsai et al. reported that Twist1 reversible regulation of EMT plays an important role during the metastatic processes of skin squamous carcinoma (45). Using a tail vein injection model, they showed that tumor cells must precede EMT in blood to be capable of extravasation from the lung vasculature. Once the tumor cells reach the distant site (lungs), reversion of EMT to MET promotes tumor cell colonization. In our mouse model, knockdown of Twist1 abrogated BRMS1^KD-induced EMT in tumor cells. Collectively, our findings (i) indicate that, following loss of BRMS1, NSCLC cells undergo EMT due to dysregulated NF-κB–dependent gene expression and (ii) identify Twist1 as the primary EMT-TF driving this process.