Krüppel-Like Factor 8 Induces Epithelial to Mesenchymal Transition and Epithelial Cell Invasion

Tumor invasion and metastasis are the main causes of death from cancer. Epithelial to mesenchymal transition (EMT) is a determining step for a cancer cell to progress from a noninvasive to invasive state. Krüppel-like factor 8 (KLF8) plays a key role in oncogenic transformation and is highly overexpressed in several types of invasive human cancer, including breast cancer. To understand the role of KLF8 in regulating the progression of human breast cancer, we first established stable expression of KLF8 in an immortalized normal human breast epithelial cell line. We found that KLF8 strongly induced EMT and enhanced motility and invasiveness in the cells, by analyzing changes in cell morphology and epithelial and mesenchymal marker proteins, and using cell migration and Matrigel invasion assays. Chromatin immunoprecipitations (ChIP), oligonucleotide precipitations, and promoter-reporter assays showed that KLF8 directly bound and repressed the promoter of E-cadherin independent of E boxes in the promoter and Snail expression. Aberrant elevation of KLF8 expression is highly correlated with the decrease in E-cadherin expression in the invasive human breast cancer. Blocking KLF8 expression by RNA interference restored E-cadherin expression in the cancer cells and strongly inhibited the cell invasiveness. This work identifies KLF8 as a novel EMT-regulating transcription factor that opens a new avenue in EMT research and suggests an important role for KLF8 in human breast cancer invasion and metastasis. [Cancer Res 2007;67(15):7184–93]


Introduction
More than 90% of human cancers are carcinomas that arise from epithelial origin.The vast majority of cancer deaths result from carcinoma cell invasion and metastasis.Thus, understanding the underlying mechanisms is a key to cut the cancer deaths.Increased motility and invasiveness of cancer cells is reminiscent of epithelial to mesenchymal transition (EMT) that normally takes place during embryonic development and wound healing (1).The defining event for EMT is disruption of E-cadherin-mediated intercellular adhesion, which leads to loss of epithelial morphology and gain of a motile and invasive fibroblast-like mesenchymal phenotype.This process is associated with the functional loss of E-cadherin due largely to the repression of its transcription (2).Downregulation of E-cadherin is the key step toward the invasive phase of carcinoma.An increasing number of transcription factors has been implicated in the repression of E-cadherin expression, including zinc-finger proteins of the Snail/Slug family (3)(4)(5), yEF1/ZEB1 (6), SIP1 (7), twist (8), nuclear factor-nB (9), and the basic helix-loop-helix E12/E47 factor (10), although their functional links to human cancer invasion and metastasis remain largely unknown.
KLF8 was initially identified as a transcription repressor of Kru ¨ppel-like C2H2 zinc-finger transcription factor family proteins.Our previous studies identified KLF8 as a downstream effector of focal adhesion kinase (FAK;ref. 11).FAK regulates many types of cellular events, including cell proliferation, survival, migration, invasion, and EMT (12)(13)(14), and is implicated to play a crucial role in metastatic progression of human carcinoma given its overexpression in many types of invasive human cancer (15).Recently, we found that KLF8 plays an important role in oncogenic transformation and is highly overexpressed in several types of invasive human cancer (16).However, whether and how KLF8 might play a role in carcinoma cell invasion and metastasis is completely uninvestigated.
Here, we report that KLF8 is a potent inducer of EMT and invasion and a novel repressor of E-cadherin in epithelial cells, and its up-regulation plays a large part in the loss of E-cadherin expression in human breast carcinoma cells and their invasiveness.Our results provide strong evidence that KLF8 may play a critical role in EMT-associated diseases, including breast cancer metastasis.
pBabe-puro-HA-KLF8 was constructed by inserting the HA-KLF8 fragment excised from pKH3-KLF8 (11) into the EcoRI site in pBabe-puro retroviral vector.MCF-10A cells were grown based on the protocols established by Dr. Joan S. Brugge's laboratory (19).The retroviral vector and packaging cells were gifts from Dr. Jinsong Liu (University of Texas M. D. Anderson Cancer Center).Polyclonal MCF-10A cell lines expressing HA-KLF8 (MCF-10A-KLF8 cells) or the control vector (MCF-10A-mock cells) were generated using the retroviral infection protocol as previously described (20).
Western blotting and immunofluorescence staining.These assays were done essentially as described previously (21).For Western blotting, 1:2,000 primary antibodies and 1:10,000 secondary antibodies were used.For fluorescent staining, the primary and secondary antibodies were usually diluted to 1:200.
Promoter reporter assay.Luciferase reporter assays were done essentially as described previously (11,22).The wild-type and E box mutant human E-cadherin promoter-luciferase reporter constructs were generous gifts from Dr. Eric R. Fearon (University of Michigan, Ann Arbor, MI; ref. 23).The GT box mutant was made by overlapping PCR mutagenesis.The reporter constructs were cotransfected into MCF-10A cells with expression vectors encoding KLF8, KLF8DC38, Snail, and the control pKH3 vector.Twenty-four hours after transfection, luciferase activity was determined using Dual-luciferase reporter assay system according to the manufacturer's instructions (Promega).
Chromatin immunoprecipitation assay.MCF-10A-KLF8 cells and anti-HA antibody were used in the assays to determine interaction between KLF8 and endogenous E-cadherin promoter.MCF-10A-mock cells, anti-HAfree precipitation, as well as primer-free PCR were all included as negative controls.The assays were done as we described (24).PCR primers for amplifying human E-cadherin promoter are 5 ¶-AACTCCAGGCTA-GAGGGTCA ( forward) and 5 ¶-GGGCTGGAGTCTGAACTGA (reverse).
RNA interference.siRNA was transfected into MCF-10A, MDA-MB-231, or Hs578Tcells using Oligofectamine according to manufacturer's (Invitrogen) instructions.Two to three days later, cells were either further treated with TGF-h or not and processed for invasion assays, quantitative real-time PCR, Western blotting, etc.
Quantitative real-time PCR.Total cellular RNA was prepared using Nucleospin RNA II kit (BD Biosciences) and cDNA was synthesized using the Superscript First-Strand Synthesis system (Invitrogen).cDNAs were used for quantitative reverse transcription-PCR (qRT-PCR) analysis using SYBR-Green Master PCR mix on iCycler (Bio-Rad).Glyceraldehyde-3phosphate dehydrogenase (GAPDH) was included as a normalizing control.The relative expression value for each target gene compared with the calibrator for that target is expressed as 2 -(Ct-Cc) , where Ct and Cc are the mean threshold cycle differences after normalizing to the expression of GAPDH in the same cDNA pool using the standard curve method described by the manufacturer.Semiquantitative RT-PCR was designed based on qRT-PCR results so that the PCR reaction was stopped within the linear range of production.The amplified DNA fragments were visualized by agarose gel electrophoresis.Primers used were as follows: for E-cadherin, forward 5 ¶-CGGGAATGCAGTTGAGGATC, reverse 5 ¶-AGGATGGTGTAAGCGATGGC; for KLF8, forward 5 ¶-TCTGCAGGGACTACAGCAAG, reverse 5 ¶-TCA-CATTGGTGAATCCGTCT; and for GAPDH, forward 5 ¶-TGGTATCGTG-GAAGGACTCA, reverse 5 ¶-CCAGTAGAGGCAGGGATGAT.
Biotinylated oligonucleotide precipitations.HEK293 cells were cotransfected with pKH3-KLF8 (11) and pMT-Snail (25) constructs expressing HA-tagged KLF8 and flag-tagged Snail for 24 h and processed for biotinylated oligonucleotide precipitation (BOP) assays as we previously described (24).DNA-bound KLF8 and Snail proteins were identified using anti-HA and anti-flag antibodies, respectively, in Western blotting.
Migration and invasion assays.For trans-well migration assay, 2.5 Â 10 4 cells were added to the top chambers of 24-well trans-well plates (BD Biosciences; 8-Am pore size), and complete media were added to the bottom chambers.After 6 h incubation, top (nonmigrated) cells were removed, and bottom (migrated) cells were fixed, stained, and counted.For assays with MDCK-KLF8 cells, cells were incubated under uninduced or induced conditions for 24 h before being loaded into the chambers.Relative motility was normalized to the uninduced MDCK-KLF8 cells or the MCF-10A-mock cells under serum-free conditions.Data are presented as the average number of migrated cells in nine random fields of view from three independent experiments.
For scratch wound closure assays, cells were seeded in 6-cm tissue culture dishes at a density of 5 Â 10 5 .A wound was incised 24 h later in the central area of the confluent culture, which was incubated for a further 2 h after careful washing to remove detached cells and addition of fresh medium.Time-lapse phase-contrast pictures were taken of the wounded area at 10-min intervals for 16 h using Image-Pro Plus software (MediaCybernetics) and an inverted Olympus IÂ70 microscope with an incubator that maintains 37jC and 5% CO2.
Matrigel invasion assays were done using BD BioCoat invasion chambers and serum in the complete medium as the chemoattractant.Except for the presence of the Matrigel, the procedures and the analyses were the same as those for the trans-well migration assays.In some experiments, cells were transfected with siRNA 48 h before being processed for the invasion assays.
Immunohistochemical analyses.Information about the patients and tumor specimens of the infiltrating breast carcinomas was previously described (26,27).Immunoperoxidase staining was done based on avidinbiotin complex techniques as previously described (28).Briefly, tissue sections were deparaffinized and dehydrated, antigens were retrieved, and endogenous peroxidase activity was blocked.After sequential incubation with primary antibody, biotinylated secondary antibody, and avidin-biotin-HRP, detection was done using the HRP chromogen substrate solution.Counterstaining was done using Mayer's hematoxylin and nuclei were stained with ToPro.A two-headed microscope was used for staining evaluation.

Results
KLF8 is a strong EMT inducer.We have previously found that there is little or no expression of endogenous KLF8 in both MDCK canine renal and MCF-10A human mammary epithelial cells (16).We first asked whether overexpression of KLF8 in MDCK cells could induce EMT.We chose this cell line because (a) this is the cell line used in the original experiments from which the EMT phenomenon was discovered; (b) this cell line has since been widely used as a typical cell model in EMT studies; (c) FAK has been shown to promote EMT formation in this cell line (14,29); and (d) this cell line was derived from renal epithelium where renal fibrosis, one of the typical EMT-associated diseases, takes place.
We generated Tet-off MDCK cell lines that express inducible expression of HA-tagged KLF8 (MDCK-KLF8 cells) or its dominantnegative mutant KLF8DC38 (ref.16; MDCK-KLF8DC38 cells) and mock control vectors (MDCK-mock cells) as previously described (11,16,21).In the presence of tetracycline (uninduced) in the medium, expression of the KLF8 proteins was completely repressed (Fig. 1A, lanes 1 and 3).In contrast, in the absence of tetracycline (induced), the expression was strongly induced (Fig. 1A, lanes 2 and 4) and the induced expression could last at least for several weeks in culture (data not shown).After 3 days of incubation, the uninduced MDCK-KLF8 cells remained in their original cobblestone-like typical epithelial morphology in a confluent monolayer or islands of grouped cells with tight cell-cell contacts when replated at either a high or low density (Fig. 1B, a and c).However, the induced MDCK-KLF8 cells have undergone a strikingly morphologic change that is characterized by loss of cell-cell contact and the cobblestone-like phenotype, failure to form a confluent monolayer or islands, and being elongated and spindleshaped and scattered (Fig. 1B, b and d).These phenotypes are typical of the fibroblastoid cells formed during the EMT process (1).In contrast, the induced MDCK-KLF8DC38 cells showed no difference in morphology from their uninduced counterparts (Fig. 1B, compare f with e), the MDCK-mock cells, or the parental MDCK cells (data not shown).
We then extended our studies to the well-characterized normal human epithelial cell line MCF-10A cells because (a) this cell line is different from MDCK in both species and organ origins and (b) this cell line represents the normal origin from which human breast cancer cells develop and metastasize.We generated MCF-10A cell lines that constitutively express either HA-tagged KLF8 (MCF-10A-KLF8 cells) or empty vectors (MCF-10A-mock cells) by retroviral infection.The stable expression of KLF8 was confirmed by anti-HA Western blotting (Fig. 1C, lane 2).Similarly to those observed for the MDCK cells, whereas the typical polarized monolayer epithelial phenotype of the cells was not affected by expression of the control vector (Fig. 1D, a), MCF-10A-KLF8 cells lost their ability to grow as a monolayer and to show contact-mediated growth inhibition, formed networks with cells crossing over each other with extremely long membrane extension, and the cobble-stone-like appearance of epithelial cells was replaced by a spindle-like, fibroblastic morphology (Fig. 1D, b).
Taken together, these observations suggested that the expression of KLF8 induced EMT in both MDCK and MCF-10A epithelial cells.
In addition to morphologic changes at the cellular level, expression switch from epithelial to mesenchymal (or fibroblastoid) marker genes is another hallmark of EMT.To analyze the molecular changes induced by KLF8 in the cells, we examined the expression of epithelial and mesenchymal marker proteins by Western blotting.Three days after KLF8 expression was induced in the MDCK-KLF8 cells, we detected a remarkable decrease, if not a complete loss, of the epithelial marker protein E-cadherin although the decrease in another epithelial marker protein h-catenin seemed modest; in contrast, the expression of fibroblast markers including N-cadherin, vimentin, and fibronectin was strongly induced (Fig. 2A, compare lane 2 with 1).Time course observation clearly showed a causative effect of the induced expression of KLF8 on the expression switch of E-to N-cadherin in the MDCK-KLF8 cells (Supplementary Fig. S1).Similar expression switch of the marker proteins took place in the MCF-10A-KLF8 cells compared with the MCF-10A-mock cells (Fig. 2A, compare lane 4 with 3).We further verified the switch of the marker expression by examining the subcellular presence of the proteins using indirect immunofluorescent staining.As expected, abundant E-cadherin protein was detected in the boundaries between the uninduced MDCK-KLF8 cells (Fig. 2B, a-c); however, the E-cadherin protein disappeared from the induced MDCK-KLF8 cells (Fig. 2B, d-f ).On the contrary, there was little or no detectable expression of vimentin in the uninduced MDCK-KLF8 cells (Fig. 2B, g-i), but the induced MDCK-KLF8 cells showed strong staining for vimentin in the cytoplasm (Fig. 2B, j-l).Similarly, strong cell-cell boundary staining for E-cadherin and h-catenin and little or no staining for Ncadherin, vimentin, or fibronectin were found in the MCF-10Amock cells (Fig. 2C, a-f ).In contrast, expression of KLF8 in the MCF-10A cells (Fig. 2C, g-l) resulted in marked loss of E-cadherin expression, the translocation of h-catenin from the periphery to the cytoplasm (Fig. 2C, h and I), and the gain of expression of the fibroblast marker proteins including N-cadherin, vimentin, and fibronectin (Fig. 2C, j-l).
Collectively, these findings strongly shown that KLF8 plays a crucial role in the regulation of the EMT.
The promoter of E-cadherin gene is a direct trans-repression target of KLF8.E-cadherin plays a central role in maintaining epithelial cell-cell adhesion and polarity.Down regulation of its transcription is thought to be a primary mechanism that contributes to the onset of EMT (1, 2).Interestingly, database search revealed several potential KLF8-binding consensus sequences (i.e., GT boxes) in both human and canine E-cadherin gene promoters, suggesting a possible direct regulation of E-cadherin transcription by KLF8.
To test whether KLF8 regulates E-cadherin at the promoter level of the gene, we did promoter luciferase reporter assays.We found that, indeed, KLF8 strongly repressed the activity of the E-cadherin promoter in a dose-dependent manner, whereas the KLF8DC38 mutant did not affect the promoter activity (Fig. 3A).
Snail has been identified as another main transcriptional repressor of E-cadherin gene promoter in both MDCK and MCF-10A cells (30)(31)(32) through directly binding to the E boxes in the promoter (3,5).It is possible that KLF8 represses the promoter through directly binding to the GT box(es) in the promoter or by influencing Snail binding to the E boxes in the promoter or by a combination of both mechanisms.To determine the KLF8responsive sites in the promoter, we disabled the E boxes or GT boxes either individually or simultaneously (Fig. 3B, top) and examined the response of the promoter to KLF8.Coexpression with Snail was done as a control.We found that KLF8 repressed the E-cadherin wild-type promoter to a similar degree to that by Snail and combination of KLF8 and Snail showed a slightly enhanced repression (Fig. 3B, bottom, WT).Interestingly, disruption of one of the GT boxes resulted in an entire resistance to the repressive effect of KLF8, but not of Snail (Fig. 3B, bottom, mGT box).The other A, loss of the epithelial markers and gain of the mesenchymal markers determined by Western blot.The MDCK-KLF8 cells and the MCF-10A stable cell lines were grown for 72 h under conditions as described in Fig. 1.Whole-cell lysates were prepared for Western blotting using antibody against the epithelial marker (E-cadherin and h-catenin) and mesenchymal marker (N-cadherin, vimentin, and fibronectin) proteins.Anti-ERK blotting was done as loading controls.B, fluorescent view of the marker protein expression in MDCK cells.The MDCK-KLF8 cells without (a-c; g to i) or with (d-f ; j-l ) induced expression of KLF8 for 3 d were stained with antibodies for E-cadherin (a, d) and vimentin (g, i).The nuclei were stained with Hoechst dye (b, e, h, k ).C, fluorescent view of the marker protein expression in MCF-10A cells.Phase-contrast photography (a, g) and immunofluorescent staining were done as described in Fig. 1A and in (B ).
GT boxes were not functional as determined by luciferase assays (data not shown).Conversely, mutations of the E boxes only abolished the repressive function of Snail but not of KLF8 (Fig. 3B, bottom, mE boxes).Notably, when both the GT and E boxes were disabled, the promoter completely lost its responsiveness to both KLF8-and Snail-mediated repression (Fig. 3B, bottom, mE-/GT boxes).These results suggested that KLF8 directly represses E-cadherin promoter independently of Snail by binding to the GT box in the promoter.
To determine the direct interaction between KLF8 protein and the GT box sequence in the E-cadherin promoter, we carried out BOP assays (Fig. 3C).In agreement with the luciferase results described above, we found that the wild-type oligonucleotide that includes the GT box bound to KLF8 as efficiently as its binding to Snail (Fig. 3C, lane 1).However, when the GT box was disabled, the mutant oligonucleotide could no longer bind to KLF8 although its binding to Snail was not changed (Fig. 3C, lane 2).Conversely, mutation of the E boxes in the oligonucleotide did not affect its binding to KLF8 but its binding to Snail was completely abolished (Fig. 3C, lane 3).The oligonucleotide with both the GT and E boxes mutated lost its binding to both KLF8 and Snail.These results strongly suggested a specific binding of KLF8 to the GT box in the E-cadherin promoter.To further verify this binding in vivo, we did ChIP assays.We first tested the possible in vivo binding of the HA-KLF8 to the endogenous E-cadherin promoter region spanning the GT box in MCF-10A-KLF8 cells (Fig. 3D, top).We found that the HA-KLF8 directly and specifically interacted with the E-cadherin promoter region (Fig. 3D, top, compare lane 4 with 1-3).We then examined the interaction between endogenous KLF8 and the endogenous E-cadherin promoter in either TGF-h stimulated MCF-10A cells (Fig. 3D, middle) or the invasive breast cancer cell MDA-MB-231 (Fig. 3D, bottom) in both of which we have shown the up-regulation of KLF8 (see below; ref. 16).We found a clear association between the endogenous KLF8 and E-cadherin promoter in the cells (Fig. 3D, middle, lane 3 and bottom, lane 2).The association was specific to KLF8 as it was undetectable in the unstimulated MCF-10A cells (Fig. 3D, middle, lane 1) as well as the KLF8 knockdown cells (Fig. 3D, middle, lane 2 and bottom, lane 1).
Taken together, these results showed that KLF8 functions as an endogenous regulator of E-cadherin transcription by repressing E-cadherin gene promoter independently of Snail through direct binding to the GT box in the promoter.Importantly, the identified KLF8 responsive GT box region in the promoter is highly conserved among several species (not shown), suggesting a broadly repressive Figure 3.The E-cadherin promoter is a direct repression target of KLF8.A, KLF8 represses the E-cadherin promoter reporter.MCF-10A cells were cotransfected with the indicated amounts of expression vectors (0.5 Ag of the insertless vector was used in column 1) encoding KLF8 or its mutant KLF8DC38 and the E-cadherin promoter reporter.After 16 h, luciferase assays were done as described in Materials and Methods.Columns, mean of at least three independent experiments; bars, SE.B, inhibition of the E-cadherin promoter by KLF8 depends on a GT box site in the promoter.Top, diagrams of the E-cadherin promoter reporter and its mutants (WT, wild-type; mGT-box, the GT box was disabled; mE-box, all the E boxes were disabled; mE-/GT-boxes, both the E and GT boxes were disabled).Bottom, the E-cadherin promoter reporter or its mutants shown in (B) were cotransfected into MCF-10A cells with expression vector encoding KLF8, snail, their combination, or the insertless vector for luciferase assays.C, direct interaction in vitro between KLF8 and the E-cadherin promoter at the GT box.Lysates from 293T cells cotransfected with vector expressing HA-KLF8 or Flag-Snail were subjected to BOP assays (see Materials and Methods) with oligonucleotides spanning the E-cadherin promoter region containing the GT or E boxes or their mutants.The precipitated KLF8 or Snail was detected using antibodies for the epitopes.Whole-cell lysates (WCL ) were used in the blot to confirm the expression.D, direct interaction in vivo between KLF8 and the E-cadherin promoter at the chromatin level.Top, ChIP assays were done as described in Materials and Methods using the MCF-10A-mock (lane 1 ) and MCF-10A-KLF8 (lanes 2-5 ) cells.Anti-HA coprecipitated chromatin was subjected to E-cadherin promoter-specific PCR amplification.PCR amplification from the total chromatin (Input, lane 5 ) was used as a positive control and those lacking HA-KLF8 expression (lane 1), PCR primers (lane 2 ), or anti-HA (lane 3) served as diverse negative controls.Middle, MCF-10A cells were transfected with KLF8 siRNA (lane 2) or control siRNA (lane 3) for 2 d and then grown with (+) or without (À) TGF-h (5 ng/mL) in the medium for 3 d.ChIP assays were carried out similarly except that anti-KLF8 was used to coprecipitate the chromatin.PCR amplifications from untreated MCF-10A precipitates (lane 1) and from the total chromatin (Input, lane 4) were used as negative and positive controls, respectively.Bottom, MDA-MB-231 cells were used in the similar ChIP assays.

Cancer Research
Cancer Res 2007; 67: (15).August 1, 2007 role for KLF8 in the regulation of E-cadherin transcription.The results also suggested an important role of KLF8 in the direct repression of E-cadherin transcription in invasive breast cancer cells.
The expression of KLF8 and its regulation of EMT are independent of Snail.Although the above results strongly suggested that both KLF8 and Snail directly repress E-cadherin gene promoter by binding to independent cis-elements in the promoter, we could not rule out that these two transcription factors might also modify the expression of each other to regulate the transcription of E-cadherin and possibly other EMT-related target genes.To test this possibility, we first examined whether overexpression of ectopic Snail in MCF-10A cells affects KLF8 expression or vice versa.We found that overexpression of Snail in MCF-10A cells, as confirmed by Western blotting (Fig. 4A, inset), did not influence KLF8 expression (Fig. 4A, compare column 2 with 1).Similarly, Snail expression level in the MCF-10A-KLF8 cells did not change compared with that in the MCF-10A-mock cells (Fig. 4B, compare column 2 with 1).
Snail expression can be induced by TGF-h treatment of epithelial cell lines including MCF-10A cells (33).To test whether TGF-h also up-regulates KLF8 expression in the same cell line, we compared KLF8 expression level in MCF-10A cells before and after TGF-h treatment by qRT-PCR.We found that, similar to Snail, the expression of KLF8 was induced by more than four times (Fig. 4C).Time course experiments revealed a similar pattern for the induction of both KLF8 and Snail by TGF-h (Supplementary Fig. S2A).To test whether the up-regulation of KLF8 and Snail expression by TGF-h is a sequential or mutually independent parallel events, we selectively blocked KLF8 or Snail expression by RNA interference and then analyzed the induction of Snail or KLF8 expression in response to TGF-h in MCF-10A cells by qRT-PCR.Despite the fact that expression of KLF8 or Snail was knocked down by approximately four times (Fig. 4D, compare column 2 with 1 or 5 with 4), the TGF-h-induced expression of Snail or KLF8 remained unchanged (Fig. 4D, compare column 6 with 4 or 3 with 1).To explore TGF-h downstream proteins responsible for the up-regulation of KLF8, we selectively blocked several potential signaling proteins and found that phosphatidylinositol 3-kinase (PI3K), but not Rho, p38, or Smad3, seemed to play a major role (Supplementary Fig. S2B).
To determine the role of KLF8 in TGF-h-induced EMT, KLF8 expression was inhibited by RNA interference from MCF-10A cells.We found that the KLF8 knockdown dramatically slowed down the TGF-h-induced EMT (Supplementary Fig. S2C).
Altogether, these data provided strong evidence that the expression of KLF8 and Snail are independent of each other at least in MCF-10A cells, suggesting KLF8 and Snail may play parallel and/or mutually supplemental roles in the regulation of EMT.
KLF8 promotes epithelial migration and the gain of invasiveness.Epithelial cells that have undergone EMT are characterized by their increased motility and gain of invasiveness.We asked whether overexpression of ectopic KLF8 results in increased motility in MDCK as well as MCF-10A cells.We analyzed directional cell motility using trans-well migration assays and wound closure assays coupled with time-lapse image recording.As shown in Fig. 5A, uninduced MDCK-KLF8 cells and the MCF-10Amock cells (Mock) migrated twice to thrice faster toward serumcontaining medium than serum-free medium (compare Fig. 5A, column 2 with 1, or 6 with 5).In contrast, the induced MDCK-KLF8 cells and the MCF-10A-KLF8 cells migrated >40 times faster toward serum-containing medium than serum-free medium (compare Expression of the KLF8 protein was confirmed by Western blot (inset ).C, TGF-h induces increased expression of both KLF8 and Snail.Expression of KLF8 and Snail was analyzed by qRT-PCR in MCF-10A cells with or without TGF-h treatment (5 ng/mL for 24 h).D, inductions of KLF8 and Snail by TGF-h are separate events.MCF-10A cells were transfected with either KLF8 siRNA or Snail siRNA.Forty-eight hours later, the cells were treated with TGF-h for 24 h.Expression of KLF8 and Snail was then analyzed by qRT-PCR.Fig. 5A, column 4 with 3, or 8 with 7), or approximately five times faster than the uninduced MDCK-KLF8 or MCF-10A-mock cells (compare Fig. 5A, column 4 with 2 or 8 with 6).Consistently, within 16 h of wound closure, whereas the uninduced MDCK-KLF8 cells (Fig. 5B, a-c) or MCF-10A-mock cells (Fig. 5C, a-c) had barely moved, the induced MDCK-KLF8 cells (Fig. 5B, d-f ) or MCF-10A-KLF8 cells (Fig. 5C, d-f ) were able to migrate and almost completely close the wounds.These results show that KLF8 plays a critical role in promoting the migratory function of the epithelial cells undergoing EMT.
To examine role of KLF8 in the regulation of the cell invasiveness, we did trans-well Matrigel invasion assays to assess the ability of the cells to invade through the Matrigel layer toward serum-containing medium.As shown in Fig. 5D, the uninduced MDCK-KLF8 or MCF-10A-mock cells were barely able to invade regardless of the presence of serum (Fig. 5D, compare column 2 with 1 or 6 with 5); in contrast, both induced MDCK-KLF8 and MCF-10A-KLF8 cells gained strikingly high invasiveness (Fig. 5D, compare column 4 with 2, or 8 with 6).These findings suggest that KLF8 play a central role in the gain of invasiveness of epithelial cells during EMT.
Up-regulated expression of KLF8 is highly responsible for the invasive potential of the human breast cancer cells.We have previously shown that KLF8 is highly overexpressed in invasive human breast cancer cell lines ( 16).E-cadherin expression has been inversely correlated to the invasive potential of several types of human cancer including breast cancer (2).To determine whether correlation exists between expression patterns of KLF8 and E-cadherin in the human breast cancer cell lines, equal amounts of total RNA recovered from the cell lines were used for qRT-PCR analyses.We found that there was clearly an inverse correlation of the expression of KLF8 with that of E-cadherin (Fig. 6A) and a positive correlation of KLF8 expression with the invasive potential of the cells determined by Matrigel invasive assays (Fig. 6B).Combined with the results showing that KLF8 directly represses E-cadherin gene promoter (see Fig. 3) and promotes cell invasiveness (see Fig. 5D), these data suggest that the elevated KLF8 expression in the highly invasive cell lines (Fig. 6A, lanes 4-6, and B columns 4-6) may be responsible for both the loss of E-cadherin expression and the high invasiveness.To test this possibility, the expression of endogenous KLF8 was knocked down in two of the most invasive cancer cell lines (i.e., MDA-MB-231 and Hs579T).qRT-PCR confirmed a decrease in KLF8 expression by f60% (Fig. 6C, top).We found that the decreased KLF8 expression was accompanied by a remarkable increase in the expression of E-cadherin in the cells as determined by Western blotting (Fig. 6C, middle) and by a decrease in the cell invasiveness by almost a half as examined by Matrigel invasive assays (Fig. 6C, bottom).To evaluate direct clinical relevance of KLF8 versus E-cadherin expression, we did immunohistochemical staining of KLF8 and E-cadherin proteins in surgical specimens of human breast cancer tumor tissues.Again, we found a strongly aberrant up-regulation of KLF8 in 50% of E-cadherin-negative breast tumors with lymph node metastases (Fig. 6D; Supplementary Table ).Collectively, these results strongly suggest a potentially important role of KLF8 in the regulation of progression of human breast cancer invasion and metastasis to which repression of E-cadherin transcription by KLF8 is critical.

Discussion
In this report, we showed a novel role and mechanisms for KLF8 in the regulation of EMT and this finding is possibly clinically relevant.First, we showed that KLF8 is a potent inducer of EMT.Second, we identified E-cadherin gene promoter as a novel in vivo target of direct transcriptional repression by KLF8 and such repression is via a GT box sequence in the promoter.Third, this study revealed that KLF8 plays a critical role in inducing the epithelial invasiveness and maintaining the invasive potential of human breast cancer cells.Finally, we showed that the aberrant upregulation of KLF8 is strongly correlated to the loss of E-cadherin expression in both invasive human breast cancer cell lines and,

Cancer Research
Cancer Res 2007; 67: (15).August 1, 2007 more importantly, in surgical specimens of metastatic breast tumors recovered from patients.This study has opened an area of study into a novel role for KLF8 in EMT-associated physiologic and pathologic processes such as embryonic development, wound healing, fibrosis, and particularly cancer invasion and metastasis.
Snail has been described as a primary player in the regulation of EMT by repressing E-cadherin transcription.Indeed, this role has linked Snail to the early development of metastatic breast cancer progression as well as recurrence (34,35).Another E-cadherin repressor, Twist, has also been shown to play an important role in EMT and tumor metastasis in a murine breast cancer model (8).However, although Snail and Twist repress E-cadherin gene promoter independent of each other, they both act on the E box elements in the promoter (3,8).We identified the GT box, instead of the E boxes, as the responsive site to KLF8 (Fig. 3).Importantly, the KLF8 responsive element is strikingly conserved among different species (data not shown).These findings provide new insights into mechanisms by which the expression of E-cadherin is regulated during the processes of EMT and invasion.First, an invasive cell may have all the transcription repressors up-regulated to ensure an effectively constitutive inhibition of E-cadherin expression.For this reason, such cancer is likely resistant to gene therapy that targets to any single one of the individual repressors considering of the potential compensating effects by the other repressors not targeted.However, such individual gene targeting strategy may still work well for cancer that has gained overexpression of only one of the repressors.Indeed, our result showed that 50% of the E-cadherin-negative tumors have aberrant elevation of KLF8, whereas the other half does not (Supplementary Table ).This interesting result suggests that the EMT regulating transcription repressors may each play a dominant role in the repression of E-cadherin transcription in different subgroups of invasive breast tumors.Second, the Kru ¨ppellike transcription factor family may emerge as a new type of transcription factors that control the transcription of E-cadherin and thus EMT and invasion as the KLF members target similar consensus elements in their target gene promoters.Indeed, KLF6, another KLF member and tumor suppressor, has been shown to directly bind and activate E-cadherin gene promoter (36).
In addition to E-cadherin-mediated adherens junction, other junctions including tight junction and desmosomes are also dissociated during EMT.Disruption of these intercellular adhesions is also thought to be critical to EMT progression given the fact that overexpression of E-cadherin alone cannot usually reverse the fibroblastoid cells back to epithelial phenotype (1).Indeed, all the tight junction proteins such as occludins, claudins, and ZO-1 as well as desmosomal proteins such as desmoglein and desmocollin have been found transcriptionally repressed during EMT induced by Snail as well as other EMT inducers (37)(38)(39)(40).It will be interesting to test whether KLF8 also plays a role in the regulation of the dissociation of tight junction and desmosomes and whether such regulation by KLF8 is through direct repression of the promoters of the genes that encode the proteins at tight junction and desmosomes.
KLF8 seems to function downstream of TGF-h/PI3K signaling (see Fig. 4; Supplementary Fig. S2).In agreement with this, a previous report has shown that PI3K function is required for TGFh-mediated EMT in the mouse mammary epithelial cell line Figure 6.Up-regulated KLF8 expression in human breast cancer cells is highly correlated with the decrease in E-cadherin expression and is responsible for the invasive potential.A, correlation between the increase in KLF8 expression and the decrease in E-cadherin expression in human breast cancer cell lines.Expression of E-cadherin and KLF8 was compared in the indicated cells by qRT-PCR, relative expression was represented as normalized level of KLF8 or E-cadherin divided by the sum of them (top ).Semiquantitative RT-PCR was done as described in Materials and Methods and representative gel images were shown (bottom ).B, the invasive ability of the breast cancer cells.The same set of cells as shown in (A) were used in Matrigel invasion assays.C, knockdown of KLF8 expression results in a gain of E-cadherin expression and reduced invasiveness of the highly invasive cancer cells.The indicated invasive cancer cells were transfected with KLF8 siRNA or control siRNA for 72 hr.The knock down effect was confirmed by KLF8-specific qRT-PCR (top ).E-cadherin expression was evaluated by Western blot with ERK as the loading control (middle ).The knock down cells were subjected to Matrigel invasion assay as described above (bottom ).Columns, mean of at least three independent experiments; bars, SE.D, inverse correlation between KLF8 and E-cadherin expression in human breast cancer tumors.E-cadherin positive (Case 1) and negative (Case 2) surgical specimens of human breast tumors were stained immunohistochemically as described in Materials and Methods with anti-KLF8 (a and c ) or anti-E-cadherin (b and d).The data represent 10 pairs of E-cadherin positive/ noninvasive and negative/invasive breast tumors (see Supplementary Table ).
NMuMG (41).Interestingly, the same group also showed that TGF-h-mediated EMT in NMuMG requires the downstream activation of both p38 mitogen-activated protein (MAP) kinase and RhoA (42,43).However, p38 MAP kinase and RhoA do not seem to play a role in mediating TGF-h induction of KLF8 in MCF-10A cells (Supplementary Fig. S2B), whereas KLF8 clearly plays a role in mediating TGF-h-induced EMT (Supplementary Fig. S2C).Regardless of potential species variation in TGF-h signaling, it is possible that either KLF8 acts upstream of p38 MAP kinase and RhoA or TGF-h regulates the expression of KLF8 through PI3K and that of other EMT regulators such as Snail through MAP kinase and/or RhoA.This notion fits our observation that KLF8 knockdown slowed down TGF-h-induced EMT but could not completely prevent it, although we should keep in mind that the knockdown effect might be partial and transient.It is possible that TGF-h or its receptors are transcriptional activating targets of KLF8 in a potentially positive feedback loop.However, this possibility is slight as Snail expression in response to TGF-h stimulus is independent of KLF8 (see Fig. 4).It is likely that KLF8 regulates the expression and/or secretion of TGF-h signaling components other than Snail such as IL-11 (44,45) and ILEI (interleukin-like EMT inducer; ref. 46) that participate in EMTassociated cellular processes.Clearly, future experiments, for example, generating inducible KLF8 knockdown MCF-10A cell lines, are needed to further delineate the TGF-h signaling through KLF8 in the regulation of EMT.
Exactly how KLF8 promotes the cell motility remains to be elucidated.It is possible that KLF8-promoted cell motility is an effect secondary to its repression of E-cadherin.It is also possible that such an effect is due to KLF8 direct regulation of other proteins that are critical to cell migration.Notably, during EMT, cells undergo profound changes in cytoskeletal dynamics of which the Rho family proteins are critical regulators (13,47,48).As discussed above, RhoA may work downstream of KLF8 in the regulation of migration.Interestingly, the increased motility in MCF-10A cells during TGF-h-induced EMT can be blocked by inhibiting N-cadherin expression (33).Similarly to the cell motility, the molecular mechanisms by which KLF8 grants the cell ability to invade deserve further investigation.Although sufficient motility is essential for the cells to invade, the cells must also gain the ability to degrade the matrix proteins that form the Matrigel.In other words, the activation of some secreted matrix-degrading proteases including the protease families of matrix metalloprotenases (MMP) and plasminogen activators must accompany EMT.Does KLF8 activate such protease(s) by directly promoting their transcription (indeed, database search shows that all EMT-associated MMPs have some potential KLF8 binding sequences in their gene promoters), through up-regulating their activating proteins, or via inhibiting their inactivating proteins?
Our data clearly showed that KLF8 plays a critical role in maintaining the invasiveness of breast cancer cells and KLF8 is highly up-regulated in 50% of E-cadherin-negative metastatic breast tumors of human patients (Supplementary Table ).In agreement with this, FAK, the upstream regulator of KLF8, has been proven to be a key regulator of metastasis of breast cancer in both a rat and a mouse model of human breast cancer (49,50).It will be important to determine the role of KLF8 in breast cancer metastasis in vivo.Notably, unlike FAK that is normally expressed at modest levels, the expression of KLF8 is barely detectable (16) in normal epithelial cells, including MCF-10A, the human mammary epithelial cells (Fig. 6A).This makes KLF8 a potentially more specific therapeutic target than FAK against breast cancer invasion and metastasis.
In summary, we have identified KLF8 as a representative of a new class of transcriptional factors in the regulation of EMT and epithelial cancer invasiveness.Further studies into KLF8 target genes and their expression profiles as well as the role of KLF8 in vivo using animal models will add important new insights into the mechanisms underlying EMT-associated physiologic and pathologic processes and provide novel information useful for combating cancer metastases.

Figure 1 .
Figure 1.Ectopic expression of KLF8 transforms epithelial cells into fibroblastoid cells.A, tetracycline-controlled inducible expression of KLF8 and its dominantnegative mutant in MDCK cells.The Tet-off MDCK cell lines that express inducible KLF8 (MDCK-KLF8 cells) and KLF8DC38 (MDCK-KLF8DC38 cells) were generated as described in Materials and Methods.Cell lysates prepared from the cells with (I ) or without (U ) induction of the expression of KLF8 or the mutant were used for Western blotting with anti-HA.Anti-ERK blotting was done as loading controls.B, morphologic change in MDCK cells by KLF8 expression.The MDCK-KLF8 cells (a-d ) or MDCK-KLF8DC38 cells (e and f ) of high (a and b) or low (c-f ) densities were grown under uninduced conditions for 72 h before phase-contrast cell images were taken.C, stable expression of KLF8 in MCF-10A cells.Stable cell line expressing KLF8 and the mock cell line generated by retroviral infection are described in Materials and Methods.KLF8 expression was confirmed by Western blotting with anti-HA.Anti-ERK blotting was done as loading controls.D, morphologic change in MCF-10A cells by KLF8 expression.The MCF-10A stable cell lines expressing KLF8 (b ) and the mock cell line (a ) were grown to subconfluent density before phase contrast cell images were taken.

Figure 2 .
Figure 2. Ectopic expression of KLF8 in MDCK and MCF-10A cells induces a switch of expression from epithelial to mesenchymal specific marker proteins.A, loss of the epithelial markers and gain of the mesenchymal markers determined by Western blot.The MDCK-KLF8 cells and the MCF-10A stable cell lines were grown for 72 h under conditions as described in Fig.1.Whole-cell lysates were prepared for Western blotting using antibody against the epithelial marker (E-cadherin and h-catenin) and mesenchymal marker (N-cadherin, vimentin, and fibronectin) proteins.Anti-ERK blotting was done as loading controls.B, fluorescent view of the marker protein expression in MDCK cells.The MDCK-KLF8 cells without (a-c; g to i) or with (d-f ; j-l ) induced expression of KLF8 for 3 d were stained with antibodies for E-cadherin (a, d) and vimentin (g, i).The nuclei were stained with Hoechst dye (b, e, h, k ).C, fluorescent view of the marker protein expression in MCF-10A cells.Phase-contrast photography (a, g) and immunofluorescent staining were done as described in Fig.1Aand in (B ).

Figure 4 .
Figure 4. Expression of KLF8 and Snail in MCF-10A cells is mutually independent.A, Snail does not regulate KLF8 expression.Expression vectors encoding Flag-Snail (S) or the insertless vector (V ) was transfected in MCF-10A cells.KLF8 expression was then analyzed by qRT-PCR.Expression of the Snail protein was confirmed by Western blot (inset ).B, KLF8 does not regulate Snail expression.Snail expression was compared by qRT-PCR in the MCF-10A-mock (M ) and MCF-10A-KLF8 (K) cells.Expression of the KLF8 protein was confirmed by Western blot (inset ).C, TGF-h induces increased expression of both KLF8 and Snail.Expression of KLF8 and Snail was analyzed by qRT-PCR in MCF-10A cells with or without TGF-h treatment (5 ng/mL for 24 h).D, inductions of KLF8 and Snail by TGF-h are separate events.MCF-10A cells were transfected with either KLF8 siRNA or Snail siRNA.Forty-eight hours later, the cells were treated with TGF-h for 24 h.Expression of KLF8 and Snail was then analyzed by qRT-PCR.

Figure 5 .
Figure 5. KLF8 promotes epithelial migration and invasion.A to C, migration assays.The MDCK-KLF8 cells with or without induced expression of KLF8 and MCF-10A-mock (Mock ) and MCF-10A-KLF8 (KLF8 ) cells were evaluated for their motility by Trans-well migration assays (A ) or scratch wound closure assays (B and C ) as described in the Materials and Methods.D, Matrigel invasion assays.The MDCK-KLF8 cells with or without induced expression of KLF8 for 16 h and MCF-10A-mock and MCF-10A-KLF8 cells were assayed for their ability to invade through the Matrigel (see Materials and Methods).Data are representatives of at least three independent experiments.