EXPRESSION OF SLUG IS REGULATED BY c-MYB AND IS REQUIRED FOR INVASION AND BONE MARROW HOMING OF CANCER CELLS OF DIFFERENT ORIGIN

In metastatic cancer cells, the process of invasion is regulated by several transcription factors which induce changes required for migration and resistance to apoptosis. Slug (SNAI2, Snail2) is involved in Epithelial Mesenchymal Transition (EMT) in physiological and in pathological contexts. We show here that, in embryonic kidney, chronic myeloid leukemia-blast crisis and in neuroblastoma cells, expression of Slug is transcriptionally regulated by c-Myb via Myb binding sites in the 5’ flanking region and in the first intron of the slug gene. In embryonic kidney and neuroblastoma cells, c-Myb induced vimentin, fibronectin and N-cadherin expression, and membrane ruffling via actin polymerization consistent with the acquisition of mesenchymal-like


Some of these changes are predominantly Slug-dependent since Slug silencing via RNA interference (RNAi) reverts the cells to a quasi-parental condition. Changes in gene expression and morphology induced by c-Myb-activated
Slug correlated with increased ability to migrate (embryonic kidney) and to invade through a Matrigel membrane (embryonic kidney, colon carcinoma, neuroblastoma). c-Mybdependent Slug expression was also essential for the homing of chronic myeloid leukemia K562 cells to the bone marrow. In summary, we show here that the proto-oncogene c-Myb controls Slug transcription in tumor cells of different origin. Such regulatory pathway contributes to the acquisition of invasive properties that are important for the metastatic process.
The main cause of death for cancer patients is metastasis, the formation of secondary tumors in organs distant from the site of the original cancer (1). A series of wellcoordinated and interconnected biological processes must occur to drive tumor cells from the site of the primary neoplasm to a distant location (2). To acquire an invasive phenotype, tumor cells need to interact with components of the extracellular matrix (ECM), such as collagen and hyaluronic acid, and with soluble growth factors, such as members of the transforming growth factor β (TGF-β) and fibroblast growth factor (FGF) families, epidermal growth factor (EGF) and SF/HGF (3). During local invasion, the first step of the metastatic process, tumor cells of epithelial origin undergo a trans-differentiation process termed Epithelial Mesenchymal Transition (EMT) (3). During EMT, tumor cells lose cell-cell interactions and apico-basal polarity and acquire mesenchymal and migratory properties (4). EMT is a reversible process: once cancer cells localize to a site of metastasis undergo a Mesenchymal Epithelial Transition (MET) through which they reacquire many of the characteristics of the tumor cells at the primary site (5). In EMT, loss of epithelial properties occurs through down-regulation of the expression of epithelial-specific proteins (e.g., E-cadherin and cytokeratins), and the acquisition of mesenchymal proteins (e.g., Ncadherin and vimentin) (6). In addition, cancer cells undergoing EMT activate proteinases that allow them to pass through the extracellular matrix and to become more resistant to anoikis, a form of cell death that occurs in cells detached from their stroma support (7). A central question for understanding the process of metastasis at the molecular level is: which are the regulators of the genes activated or repressed during EMT that occurs at the onset of invasion? Several transcription factors that strongly repress E-cadherin (such as members of the Snail, ZEB and basic helix-loop-helix (bHLH) families) are now thought to be inducers of the phenotypic changes required for EMT (4). Nevertheless, the specific role of these different repressors in tumorigenesis is not fully understood. In particular, the Snail family of transcription factors that in vertebrates includes Snail (SNAI1, Snail1), Slug (SNAI2, Snail2) and SNAI3, is involved in physiological (8) and pathological (9;10) EMT. Snail is expressed during mesoderm formation, gastrulation and neural crest development, and in the majority of developmental processes in which EMT occurs [reviewed by (4)]. Slug expression has been detected in mesoderm and migratory neural crest cells, as well as in other tissues not always associated with EMT [reviewed by (4)]. Slug contributes to invasion in melanomas (11) and in malignant mesotheliomas, where it is induced by stem cell factor (SCF) (12). We have recently demonstrated that Slug silencing inhibits neuroblastoma invasion in vitro and in vivo (13). Also Snai1 expression appears to be correlated with invasive growth potential in human cancer (14). Both Snail and Slug are zinc finger transcription factors that share a common organization (4). In most cases, they act as transcriptional repressors through a conserved domain (the SNAG domain) located in the N-terminus of the protein (4). Several direct and indirect targets of Snail genes are associated with survival, proliferation, cell shape and motility (8). However, knowledge of the mechanisms that control Snail and Slug expression is incomplete.
The c-Myb proto-oncogene is a transcription factor involved in leukemogenesis in several species including humans (15). It can be activated by overexpression or inappropriate expression, structural alteration and/or genomic rearrangements and is required in the bone marrow, colonic crypt and neurogenic niches as indicated by the phenotype of global or tissuespecific knockout mice [reviewed by (15)]. Abnormal c-Myb expression has been demonstrated in colon (16;17) and breast carcinomas (18;19). We have previously shown that c-Myb is expressed also in neuroblastoma cell lines and primary tumors (20;21). Of interest, simultaneous expression of c-Myb and Slug was detected in delaminating neuroblasts of the avian embryonic neural crest (22). In this context, c-Myb is necessary for the motile phenotype as demonstrated by analysis of the effects of c-Myb knockdown induced by morpholino oligos (22).
We show here that c-Myb activates Slug expression through a transcriptional mechanism in embryonic kidney, colon carcinoma, chronic myeloid leukemia and neuroblastoma cells. c-Myb-activated Slug expression is associated with increased in vitro invasion and the acquisition of mesenchymal markers and is required for bone marrow homing of K562 cells in mice. Together, these data support the existence of a functional relationship between the proto-oncogene c-myb and metastasis-associated slug in hematopoietic and non-hematopoietic cancer cells.

Experimental Procedures
Cell lines. LAN-5 (human neuroblastoma) (20), HEK-293 (human embryonic kidney cells transformed by adenovirus type 5 DNA) (23) and LoVo (human colon carcinoma) (24) cells were cultured in DMEM medium (Euroclone), COLO-205 (human colon carcinoma) (16) in RPMI 1640 medium (Euroclone) and K562 (chronic myeloid leukemia) (25) cells in IMDM medium (Euroclone) supplemented with 10% FCS at 37° C, 5% CO 2 , 100% humidity. Plasmids. Regulatory regions of the human slug gene, were cloned into pGL3 Basic (Slug prom) and pGL3 Promoter (intron 1 and intron 2) vectors (Promega, Madison, WI), after PCR amplification of the corresponding genomic regions with the Pfu enzyme (Promega), using these primers: Slug prom FORW:5'GTT GGC TGG TAC CGC TCC TGC GCC CCT C 3'; slug prom REV: 5'GAA GGA AAG ATC TCT TGC CAG CGG GTC TGG 3'; slug intron 1 FORW: 5'GTT GGC TGG TAC CGT AAA AAG AGA AAA ATA TAT CTA GAA CTA CGT3'; slug intron 1 REV: 5'GAA GGA AAG ATC TCT GGG AAA GAA AAG GGA GGG AG 3'; Slug intron 2 FORW: 5'GTT GGC TGG TAC CGT AAG AGA AAA ACC ATA GGC AGG 3'; slug intron 2 REV: 5'GAA GGA AAG ATC TCT GGG GGT GGA GTG GGA G 3'. The PCR-amplified products were digested with KpnI/BglII restriction enzymes and cloned into pGL3 vectors using standard cloning procedures. The sequence of each construct was verified by dideoxy sequencing. The mutants in A, B, C and D sites, and the double mutant A/B were obtained by using the QuikChange sitedirected mutagenesis kit (Stratagene, La Jolla, CA), according to the manufacturer's instructions. The primers used were: slug prom A FORW: For shRNA-mediated inhibition of Slug, we used lentiviral vector pLKO-Slug3 (Addgene, Cambridge, MA) which encodes an shRNA directed against human Slug (11). In control infections, we used lentiviral vector pLKO-GFP that encodes an shRNA directed against GFP (11). For shRNA-mediated c-Myb inhibition we used vector pLVsh-c-Myb (26). Lentiviruses were generated by transient cotransfection of DNA into 293FT cells. Briefly, 2.5x10 6 293FT cells were seeded in 10 cm plates and cotransfected with 21 μg of appropriate packaging plasmids pCMVd8.2: pCMV-VSVG (2.5:1) and 20μg of pLKO-SLUG or pLKO-GFP vectors using Calcium Phosphate Transfection Kit (Invitrogen). After 6-8 hours, medium was replaced with 6.0 ml of complete medium supplemented with 1.0 mM of Sodium Pyruvate (Sigma-Aldrich, St. Louis, MO). Lentiviral stocks were harvested 48 h later, centrifuged 5 min at 3,000 RPM, aliquoted and stored at -80 C. When needed, viral supernatants were supplemented with 8 μg/ml polybrene (Sigma-Aldrich) and used to infect indicated tumor cell lines; 18 hours after infection, lentivirus-transduced cells were selected with puromycin (2-3 μM;Sigma-Aldrich) or by flow cytometric sorting of GFPpositive cells . The human slug retroviral vector was generated by cloning the full-length cDNA of human slug obtained by reverse transcriptase-PCR into the pBABE-puro vector system using standard techniques. Slug sequence was confirmed by dideoxysequencing. Infectious retroviruses were obtained by transient cotransfection of vector DNA into amphotropic Phoenix cells. Briefly, 4x10 6 cells were seeded in 10 cm plates and the plates were cotransfected with 1.5 μg of packaging plasmid pCMV-VSVG and 13.5μg of pBABE-Slug or pBabe-puro vector by calcium phosphate precipitation. Transfection was carried out in the presence of 25 μM chloroquine (Sigma, St. Louis, MO). Viral supernatants were harvested 24 hours post-transfection by centrifugation at 3,000 rpm for 5 min, supplemented with 5 μg/ml polybrene (Sigma) prior to infection and used to infect K562 cells. Forty-eight hours after infection, cells were selected by puromycin (Sigma). Western blotting. Whole cell proteins were extracted from parental and derivative cell lines, separated on SDS-polyacrylamide gels and Western blot analyses were carried out as previously described (27). Antibodies used were: anti c-Myb (clone 1-1) from Millipore For ChIP experiments performed in K562 cells, samples were prepared as described previously (28). Immunoprecipitations (IPs) were carried out using 6μg of anti-c-Myb (clone 1-1 from Millipore), anti-p53 (DO-1 from Santa Cruz) or no antibody. Recovered DNA was analyzed by real-time quantitative PCR. For every promoter fragment analyzed, each sample was quantified in triplicate from 2 independent IPs. The relative proportions of immunoprecipitated promoter fragments were determined by relative quantification method: a standard curve was generated with serial dilution of the input sample and used as a reference standard for extrapolating quantitative information for ChIP sample. Relative enrichment was calculated as follows: relative enrichment = X-Y, where X= Quantity of sample/Quantity of input and Y= Quantity of control (immunoprecipitation with unrelated antibody)/Quantity of input. Primers were the same as those utilized for semiquantitative PCR analysis. Real-time PCR was carried out using SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) in an ABI PRISM® 7000 Sequence Detection System (Applied Biosystems) according to the manufacturer's instructions. Flow cytometric analysis. 1 x 10 6 cells were harvested and the pellets were washed twice with PBS. Cells were then fixed in cold 70% ethanol added dropwise while vortexing gently. Fixed cells were kept overnight at 4° C. Cells were centrifuged and pellets were resuspended in 1 ml of PI/RNase Staining Buffer (BD Biosciences, Franklin Lakes, NJ). Reactions were incubated for 20 min at 4°C, protected from the light. Samples were analyzed by flow cytometry using a FACScalibur flow cytometer (BD Biosciences). For each sample, at least 2x10 4 cells were analyzed. Cell cycle distribution and hypodiploid DNA content were calculated by Cell Quest software (BD Biosciences). Statistical significance (p) was calculated by two-tailed Student's t test. Migration and invasion assays. Wound healing assay (29) was carried out by plating cells to confluence. Cell surface was scratched using a pipette tip. Cells were allowed to repopulate the scratched area for 3 days and photographs were taken using a digital camera mounted on an inverted microscope. For migration and invasion assays, cells were plated (1.25 x 10 5 cell/well) in the BD migration or Matrigel®coated invasion chambers (BD Biosciences). Medium in the upper chamber was supplemented with 2% heat-inactivated FCS. In the lower chamber, FCS concentration was 20%. After 24 hours, cells migrated into the lower chamber were stained and counted. Experiments were carried out in triplicate and repeated twice. Homing assay of K562 cells in SCID mice. For the homing assay, NOD/SCID mice (NOD.CB17-Prkdc scid /J, Charles River Laboratories International, Inc. Wilmington, MA) (three/group) were treated with doxycycline in the drinking water (2mg/ml; 72 hours) prior to injection (2.5 x 10 6 cells/mouse) with untreated or DOX-treated (36 hours, 7.5 μg/ml) K562 cells (c-MybshRNA or c-MybshRNA/full-length Slug). 24 hours after injection, mice were sacrificed, bone marrow harvested and the number of K562 cells determined by GFP positivity and by methylcellulose colony formation assays of marrow cells plated in the absence of hemopoietic cytokines to allow only growth of K562 cells.  (Fig. 1C). In addition, colon carcinoma COLO-205 (16) and chronic myeloid leukemia K562 cells (25) that express high levels of endogenous c-Myb, were infected with a dicistronic lentiviral vector encoding a DOXinducible c-Myb shRNA and GFP (pLVsh-c-Myb). Upon DOX treatment (7.5 μg/ml for 24 hours), c-Myb and Slug expression were downregulated ( Fig. 1D) demonstrating that also endogenous c-Myb controls Slug levels.

c-Myb induces
c-Myb induces an increase of Slug expression through a transcriptional mechanism. To investigate whether the increase in Slug levels induced by c-Myb expression is due to a transcriptional mechanism, we first searched the 5'flanking (5' from the initial ATG) and intronic regions of the Slug gene for potential Myb-binding sites (MBS). By setting the homology threshold to 86% with the consensus MBS (TFsearch web site http://www.cbrc.jp/research/db/TFSEARCH.ht ml), we identified two putative MBS in the 5' flanking region : (A at nt -9 to -4 and B at nt +122 to +128 from the transcription start site), one in the first intron (C) and one in the second intron (D). A scheme of the human slug gene and the position of the potential MBSs is shown in Fig. 2A. We cloned a portion of the slug 5'flanking region (-280 to +128 from the transcription start site) in the reporter vector pGL3-basic (hereafter named Slug prom) and intron 1 (744 bp) and 2 (906 bp) in pGL3promoter vector. Co-transfection of HEK-293 and LAN-5 cells with the reporter vectors containing the slug prom or intron 1 and 2 and an expression vector for c-Myb yielded a significant increase in luciferase activity ( Fig  2B), demonstrating that Slug transcription is c-Myb-responsive. We assessed the functional relevance of these putative MBS by mutating each site (a scheme of the wt and mutant vectors is shown in Fig. 2C) and performing assays of luciferase activity in LAN-5 and HEK-293 cells co-transfected with the wt or mutant reporters and with an expression vector for c-Myb or with the empty vector. As shown in Figure 2 D, mutation of site A lowered the c-Myb-induced increase in transcription activity of the Slug prom from 5.19 and 4.76 to 3.64 and 3.73 folds in HEK-293 and in LAN-5, respectively, compared to the activity of the corresponding promoter co-transfected with the empty vector. A similar decrease was caused by mutation in site B (4.40 and 3.51 in HEK-293 and LAN-5 respectively) and even more by the double mutation of sites A and B (3.44 and 3.25 in HEK-293 and LAN-5 respectively). The mut C reporter had also lower activity than the corresponding wt reporter (3.01 and 3.27 folds compared to 2.51 and 2.77 folds in HEK-293 and LAN-5 respectively), suggesting that c-Myb binding to the MBS in intron 1 is also relevant for transcription (Fig. 2D). In contrast, in the presence of c-Myb, the transcription activity of the mut D reporter was not significantly different from that of the corresponding wt reporter (Fig. 2D). Together, these data indicate that the slug gene contains MBS in the 5' flanking region and in intron 1 important for c-Myb DNA binding-dependent transcription activity. It should be stressed that residual c-Myb-dependent transcription activity was detected in the assay with reporter plasmids mutated at sites A and/or B and in the first intron. To assess the relevance of endogenous c-Myb expression for transcription of the slug gene, K562 cells infected with the DOXregulated pLVsh-c-Myb were transfected with reporter vectors containing the wt or the mutated forms of the slug regulatory regions and luciferase activity was measured after DOX treatment (7.5 μg/ml for 36 hours). Consistent with the results of the luciferase assays in cells ectopically expressing c-Myb, down-regulation of endogenous c-Myb significantly reduced transcription from slug regulatory regions (Fig.  2E). Also in this setting, mutations in sites A and B (Slug prom) and in C (intron 1) affect c-Myb-dependent transcription while mutation of site D in intron 2 is dispensable (Fig 2E). Together, these studies support the importance of c-Myb expression in regulating transcription of the slug gene.
In vivo binding of c-Myb to MBS in the promoter and in intron 1 of the slug gene. To assess whether c-Myb interacts in vivo with the putative MBS identified in the regulatory regions of the slug gene, we performed ChIP assays in HEK-293 cells transfected with the pHT2-c-Myb-Halo expression plasmid. Halo-c-Myb expression in transfected cells was assessed by western blot (Fig. 3 A). After decrosslinking, DNA was amplified by PCR using primers specific for segments including sites A, B, C and D. As a positive control, we used a pair of primers which amplify a region which includes a MBS in the Adenosine deaminase (ADA) gene (30) validated for interaction with c-Myb in ChIP assays (31). After 30 PCR cycles, a clear band was detected using primers for sites A, B in the slug 5' upstream region and for the ADA gene positive control (Fig. 3B). A faint band was visualized after 35 cycles using the primers for site C in intron 1. No amplification was detected utilizing the primers for site D in intron 2 even after 35 PCR cycles (Fig. 3B). Together, these data indicate that c-Myb efficiently binds in vivo to the 5'regulatory region of the slug gene that includes sites A and B. Since these two putative MBSs are separated by 123 nucleotides and the average size of sonicated DNA during the ChIP procedure was approximately 500 bp, we cannot determine whether the PCR amplification products obtained with the A and B sets of primers (Fig.  3B) reflected immunoprecipitated DNA fragments with c-Myb bound to site A or to site B or to both. In addition, our results indicate that c-Myb binds weakly to site C in intron 1. In contrast, site D in intron 2 does not appear to interact with c-Myb, at least in the experimental conditions utilized.
Consistent with this, mutation of site D did not cause a change in transcriptional activity of the Slug intron 2 luciferase promoter (Fig 2D). To assess whether endogenous c-Myb binds to the regulatory regions of the slug gene, we carried out ChIP assays in K562 cells which express high levels of c-Myb. Cross-linked DNA was immunoprecipitated with a c-Myb monoclonal antibody or with an unrelated antibody of the same IgG subclass or with no antibody. After de-crosslinking, immunoprecipitated DNA was amplified by quantitative real-time PCR. Amplification from anti-c-Myb-precipitated DNA was detectable for the regions including sites A, B and C but not for site D (Fig. 3C). No significant amplification for any site was detectable using unrelated antibody-or no antibody-precipitated DNA (Fig. 3C). Thus, also endogenous c-Myb binds to sites A, B and C but not to site D in the regulatory regions of slug gene.
Slug expression driven by c-Myb induces lineage-specific changes consistent with an EMT process and increased resistance to apoptosis. To assess the functional significance of c-Myb regulation of Slug expression, HEK-293 embryonic kidney cells and LAN-5 neuroblastoma cells, of epithelial and neuroectoderm origin respectively (20), were transfected with c-Myb or with the pcDNA3 empty vector and puromycin-resistant mixed cell populations were expanded and transduced with a pLKO lentiviral vector encoding the slug or GFP shRNA (control infection). c-Myb and Slug expression was assessed in each derivative cell line (pcDNA3/GFPkd; c-Myb/GFPkd; and c-Myb/Slugkd). As expected, c-Myb/GFPkd cells expressed higher levels of both c-Myb and Slug than the pcDNA3/GFPkd control cells (Fig. 4A  and B). Slug levels in c-Myb/Slugkd cells were lower than in c-Myb/GFPkd and in pcDNA3/GFPkd control cells (Fig. 4A and B).
Because c-Myb and Slug are involved in regulation of EMT (22;32) and cell survival (33;34), we analyzed the expression of genes involved in these processes. Compared to the pcDNA3/GFPkd control cells, Snail expression was unchanged in c-Myb/GFPkd and c-Myb/Slugkd HEK-293 and LAN-5 cells (Fig.  4A and B). In contrast, vimentin and fibronectin expression slightly increased in c-Myb/GFPkd HEK-293 and LAN-5 cells compared to pcDNA3/GFPkd control cells, but it returned to the levels of control cells after Slug knockdown only in HEK-293 (see expression in c-Myb/Slugkd cells) (Fig. 4A and B). Expression of pro-apoptotic Bax decreased in c-Myb/GFPkd HEK-293 and LAN-5 cells compared to pcDNA3/GFPkd control cells, but it increased after Slug knockdown in c-Myb/Slugkd cells ( Fig. 4A and B). Expression of anti-apoptotic Bcl-2 moderately increased in c-Myb/GFPkd LAN-5 cells as compared to pcDNA3/GFPkd control cells and it returned to the levels of control cells in c-Myb/Slugkd transfected cells (Fig. 4A and B). In HEK-293 cells, Bcl-2 levels did not increase after c-Myb transfection, but a significant decrease was observed after Slug knockdown ( Fig. 4A and  B). Finally, levels of p53 were unchanged in all samples (Fig. 4A). Together, these results suggest that i) c-Myb overexpression induces changes in gene expression that may affect apoptosis and cell morphology; and ii) some of the detected changes are Slug-dependent. We also assessed whether levels of endogenous c-Myb could affect the expression of a prototypical epithelial marker such as Ecadherin and whether this effect was dependent on c-Myb-induced Slug expression. We used COLO-205 cells infected with pLVsh-c-Myb alone or with a constitutive vector encoding a shRNA directed against Slug (pLKO-Slug indicated as Slugkd). Cells were cultured in normal conditions or with DOX (7.5 μg/ml for upon DOX treatment of pLVsh-c-Mybtransduced cells and in Slug-silenced cells (Fig.  4C). After 3 days of DOX treatment we did not detect any change in vimentin expression in pLVsh-c-Myb-transduced and in Slug-silenced cells (not shown), but a partial and yet significant decrease was detected after 6 days of c-Myb or Slug down-regulation (30 and 20%, respectively) (Fig. 4C). Simultaneous inhibition of c-Myb and Slug expression caused a more pronounced decrease of vimentin levels (80%) (Fig. 4C). These data support the role of c-Myb, in part through its control of Slug expression, in promoting phenotypic changes consistent with an EMT process. Finally, the mRNA levels of desmoplakin and occludin, two genes whose expression is high in epithelial cells and readily decreases upon mesenchymal differentiation (3), were measured by quantitative real-time PCR in COLO-205 cells infected with pLVshc-Myb. Upon c-Myb downregulation induced by DOX treatment (7.5 μg/ml for 3 days), levels of desmoplakin and occludin mRNAs were significantly increased (Fig. S1).

c-Myb expression increases apoptosis resistance in HEK-293 cells in part through a Slug-dependent mechanism.
Apoptosis susceptibility of c-Myb/GFPkd (overexpressing c-Myb), c-Myb/Slugkd (overexpressing c-Myb and Slug-silenced) and pcDNA3/GFPkd (controls) HEK 293 cells was assessed after treatment with the chemotherapeutic drug etoposide. Cells were grown in the presence of 1 μM etoposide and at the indicated times (1, 2 and 3 days) were harvested, fixed in ethanol and stained with propidium iodine for flow cytometric analysis of hypodiploid DNA content (Table 1). c-Myb/GFPkd HEK-293 cells were more resistant to etoposide-induced apoptosis than pcDNA3/GFPkd controls (8.54, 10.93 and 15.82 compared to 15.33, 33.34 and 51.13% of sub G1 cells after 1, 2, 3 days of etoposide treatment). All these differences were highly significant (<0.0001, two-tailed Student's t test). Of interest, spontaneous apoptosis in untreated cells was also lower in Myb/GFPkd HEK-293 than in pcDNA3/GFPkd controls (2.56 vs 4.90% of sub G1 cells). The frequency of apoptotic cells was higher in c-Myb/Slugkd HEK-293 cells than in c-Myb/GFPkd HEK-293 cells (13.70, 17.35 and 22.60 compared to 8.54, 10.93 and 15.82% of sub G1 cells after 1, 2, 3 days of etoposide treatment), but not as high as in pcDNA3/GFPkd control cells. Together, these data indicate that etoposide-induced apoptosis is blocked by c-Myb overexpression in HEK-293 cells, but the effect is only in part due to c-Myb-induced Slug upregulation. Similar experiments carried out in LAN-5 cells overexpressing c-Myb (c-Myb/GFPkd) and after Slug silencing (c-Myb/Slugkd) failed to demonstrate significant changes of apoptosis after treatment with etoposide compared to controls (pcDNA3/GFPkd) (data not shown). However, it should be noted that, after transfection with the c-Myb expression vector, c-Myb levels in LAN-5 were markedly lower than in HEK-293 cells (see blots in Fig. 4).
Motility and invasion changes caused by c-Myb overexpression. The changes in vimentin and fibronectin expression induced by modulation of c-Myb levels in various tumor lines (Fig. 4) are consistent with a mesenchymal-like phenotype often associated with increased migration and invasion capabilities (32). This was substantiated by immunofluorescence analysis to detect E-and N-cadherin expression and formation of F-actin fibers which are associated with a motile phenotype (35). Levels of E-cadherin were not appreciably changed in c-Myb overexpressing cells and Slug knockdown did not affect Ecadherin levels in HEK-293 cells (Fig. 5, A, B, C), but it induced E-cadherin expression in LAN-5 cells (Fig. 5 compare panel L with J and K). Expression of N-cadherin was increased in c-Myb expressing HEK-293 and LAN-5 cells (Fig. 5 E and N). Of interest, N-cadherin expression was almost undetectable after Slug knockdown in HEK-293 (Fig. 5 F), but not in LAN-5 cells (Fig. 5 O). Detection of F-actin fibers showed membrane ruffling at the cell periphery in both HEK-293 and LAN-5 c-Mybexpressing cells (Fig. 5, panels H and Q). Membrane ruffling dimmed after Slug knock down (Fig. 5, panels I and R). Furthermore, we analyzed the expression of F-actin and the presence of stress fibers also in pLVsh-c-Myb COLO-205 cells after DOX-induced downregulation of c-Myb and/or Slug knockdown in pLKO-Slug shRNA-transduced cells. Upon DOX treatment (7.5 μg/ml for 3 days), F-actin levels decreased and cell morphology became more rounded (Fig. 5 panels S and T). A similar effect was obtained upon Slug knock down (Fig.5 panel U) and by simultaneous inhibition of c-Myb and Slug (Fig. 5 panel V). Although immunofluorescence analyses are qualitative, collectively these data further support the notion that c-Myb expression promotes changes consistent with a mesenchymal-like phenotype. Next, we carried out experiments directly aimed at evaluating motility and invasion in vitro. First, we utilized the wound healing assay (29) to qualitatively assess cellular motility. c-Myb/GFPkd, c-Myb/Slugkd and pcDNA3/GFPkd HEK-293 cells were grown to confluence on plastic dishes before scratching the cell surface using a pipette tip. After 3 days, cells were observed under the microscope. The scratch was almost completely filled by c-Myb/GFPkd HEK-293 cells (Fig 6 A). In contrast, after 3 days, control pcDNA3/GFPkd and c-Myb/Slugkd cells did not bridge the gap (Fig 6 A). We failed to carry out the wound healing assay with LAN-5 cells because of their tendency to detach from the plastic substrate once they reached confluence. Invasion was tested using Boyden chambers in which upper and lower chambers are separated by a porous membrane coated with extracellular matrix proteins (Matrigel (R) ). In HEK-293 and in LAN-5 cells, invasion significantly increased (p <0.05) in cells overexpressing c-Myb (c-Myb/GFPkd) compared to control pcDNA3/GFPkd and to Slug knockdown (c-Myb/Slugkd) cells (Fig 6B). Of interest, invasion decreased at levels similar to those of controls in c-Myb/Slugkd cells (Fig. 6B), indicating that Slug expression plays an important role in invasion in both cell lines. To investigate whether c-Myb-regulated Slug expression affects invasion in vitro, we used COLO-205 cells infected with pLVsh-c-Myb and with a constitutive vector encoding a shRNA directed against Slug (pLKO-Slug, indicated as Slugkd in Fig. 6C) or against GFP (pLKO-GFP, indicated as GFPkd in Fig. 6C). Cells were cultured in normal medium or pretreated for 1 day with DOX (7.5 μg/ml) before seeding on Boyden chambers for invasion assays. Invasion was significantly inhibited (p< 0.05, Student's t test) upon DOX treatment of c-Myb shRNA-transduced COLO-205 cells (Fig.  6C). Slug knockdown decreased invasion by a similar extent even in the presence of c-Myb expression (Fig. 6C), indicating that c-Myb regulates invasion of COLO-205 cells largely through Slug.
c-Myb affects bone marrow homing of K562 cells via Slug. To investigate the role of c-Myb-regulated Slug in an in vivo process required for metastasis, we performed a bone marrow "homing' assay using untreated or DOX (7.5 μg/ml, 36 hours)-treated pLVsh-c-Myb or pLVsh-c-Myb/full length Slug K562 cells injected in the tail vein of SCID mice (2.5 x 10 6 cells/animal).c-Myb and Slug expression were detected by western blot before cell injection (Fig. 7A). Twenty four hours after injection, mice were sacrificed and "homing" of K562 cells to the bone marrow was evaluated by flow cytometry analysis of GFP-positive cells. Bone marrow homing of untreated K562 cells was 2.84 % (± 0.16) but was reduced to 1.46 % (± 0.26) after DOX-induced c-Myb downregulation.
This difference was statistically significant (p = 0.002, Student's t test) (Fig. 7B) Of interest, bone marrow homing was completely rescued in DOX-treated pLVsh-c-Myb K562 cells ectopically expressing Slug (2.81 % ± 0.56) (Fig. 7B). K562 expressing endogenous c-Myb and ectopic Slug had a slightly higher (3.59 % ± 0.75) bone marrow homing compared to cells expressing c-Myb only (Fig.7B). Bone marrow homing of K562 cells was independently evaluated by methylcellulose colony formation assays in the absence of cytokines to allow only growth factor-independent colony formation of K562 cells. Thus, 10,000 and 50,000 marrow cells from each injected mouse were plated in methylcellulose and colonies scored 5 days later. Results of the colony assays were in agreement with the flow cytometric determination of GFP-positive cells in the bone marrow. Colony formation from bone marrow cells of mice injected with DOX-treated pLVsh-c-Myb K562 cells was markedly suppressed compared to that of untreated cells and ectopic Slug expression rescued almost completely the negative effect on colony formation induced by c-Myb downregulation ( Fig. 7C and D).

DISCUSSION
We show here that c-Myb activates the expression of the transcription repressor Slug in embryonic kidney, colon carcinoma, chronic myeloid leukemia and neuroblastoma cells. During kidney development, mesenchymal cells are initially formed by EMT, and subsequently some of these cells undergo MET to form the epithelia of the pronephros, mesonephros and metanephros (36). It has been recently shown that Snail genes are repressed during kidney differentiation and that they regulate Cadherin-16 expression by repressing HNF-1β (9). In adult life, Snail and Slug reactivation has been associated with renal fibrosis (9) and kidney carcinoma progression (10). Noteworthy, Slug rather than Snail is activated in kidney cancer cells in agreement with the prominent role of Slug in renal EMT (10). Snail and Slug expression is activated via the TGF-β axis signalling (8;32), although information is incomplete on the identity of activator(s) of Snail/Slug transcription. Our data suggest that c-Myb can be involved in the acquisition of the invasive phenotype of embryonic kidney, colon carcinoma and neuroblastoma cells for its ability to enhance Slug expression. Accordingly, cells overexpressing c-Myb invade more efficiently through Matrigel (R) membranes in in vitro assays. Moreover, our data show that c-Myb-induced Slug expression is essential for bone marrow homing of chronic myeloid leukaemia K562 cells in NOD/SCID mice. Furthermore, quantitative (western blotting) and qualitative (immunofluorescence) data on vimentin, fibronectin, E-cadherin expression and F-actin fibers formation suggest that cells overexpressing c-Myb acquire mesenchymal-like characteristics.
c-Myb overexpression in embryonic kidney cells causes an increase in resistance to etoposide-induced apoptosis. This process is largely Slug-dependent since tumor cells reacquire the sensitivity of parental cells after Slug silencing. In contrast with the results in HEK-293 cells, we did not observe protection from etoposide-induced apoptosis in LAN-5 cells overexpressing c-Myb. This discrepancy may be explained by the lower levels of c-Myb expressed in LAN-5 compared to HEK-293, although we cannot rule out cell-type specific differences. Thus, the effects of c-Myb-induced Slug on apoptosis protection need further characterization by using additional cell types and detection methods.
Of interest, in birds, c-Myb is able to promote EMT in neural crest cells activating the expression of msx-1 and slug mRNAs (22), supporting the notion that c-Myb function goes beyond its role as a key regulator of hematopoiesis. We show here that c-Myb expression bestows mesenchymal features (increase of vimentin and fibronectin, membrane ruffling via F-actin polymerization at the cell periphery) to neuroblastoma cells and increases their invasion capability. Some of these changes were not detected in neuroblastoma (LAN-5) and in embryonic kidney cells (HEK-293). These discrepancies are likely due to the different origin of these cells. In neuroblastoma cells, the observed changes, reminiscent of delaminating neuroblasts in the embryo, are mostly dependent on c-Myb-induced Slug activation as demonstrated by the rescue of the phenotype of parental cells after silencing Slug expression.
c-Myb is not alone among oncogenes in playing a role in neural crest formation. The protooncogene c-Myc is an essential early regulator of neural crest cell formation in Xenopus (37). Of interest, c-Myc is localized at the neural plate border prior to the expression of early neural crest markers, such as Slug (37). In the avian neural crest, c-Myb expression is activated by bone morphogenic protein 4 (BMP4), a cytokine that plays a central role in determining the fate of embryonic neuroectodermal cells (38). It is now widely recognized that several growth factors and cytokines act in the acquisition of the invasive phenotype of many tumors (39). Thus, it is conceivable that in neuroblastoma, a tumor of neural crest origin (40), cancer cells increase their motility by growth factor-or cytokinemediated signalling which activates the expression of c-Myb that, in turn, induces Slug expression.
A possible role of c-Myb in metastasis has been suggested in colon cancer by its ability to activate acetylcholinesterase that, in turn, promotes anchorage-independent growth (41). When inappropriately expressed, as in leukemia and epithelial cancers of the breast, colon and gastro-esophagus, c-Myb appears to activate gene targets of key importance for cancer progression and metastasis (42). Thus, our observation that c-Myb activates Slug expression not only in neuroblastoma and embryonic kidney cells, but also in colon carcinoma and in chronic myeloid leukemiablast crisis cells, is consistent with a broad role of c-Myb in tumor invasion.
In colon cancer cells (COLO-205), downregulation of c-Myb and/or Slug expression led to increased E-cadherin, desmoplakin and occludin and decreased Vimentin expression and reduced invasion capability indicating that the connection between c-Myb and Slug is important in regulating the invasive phenotype of tumors of different origin. In agreement with these data, we also demonstrated that c-Myb induced Slug expression is central for bone marrow homing of K562 cells in SCID mice. To our knowledge this is the first indication of a functional role of Slug in leukemic cells.
The role of c-Myb in controlling Slug expression is exerted through a transcriptional, partly DNA binding-dependent, mechanism as shown in HEK-293 cells ectopically expressing c-Myb and upon downregulation of endogenous c-Myb in K562 cells. We mapped 4 potential MBS, two located in the 5' region of the gene and the remaining two in intron 1 and intron 2. The latter does not bind c-Myb in ChIP experiments although intron 2 possesses c-Myb-dependent enhancer activity in functional assays. This suggests that c-Myb can exert transcriptional control on Slug also by indirect mechanisms such as induction of other transcription factors or inhibition of repressor(s) that act directly on intron 2. Nevertheless, the enhancer activity of intron 2 may derive from the presence of 3 additional MBS, detected with transcription factor search, which were not included in our analysis because their homology to the MBS consensus was lower than the utilized threshold (86%). These MBS in intron 2 can be low affinity sites for c-Myb that, although able to function as transcription enhancers in luciferase assays, were not bound by c-Myb in ChIP experiments. The MBS in intron 1 shows very weak binding in ChIP experiments leaving the two sites in the 5' regulatory region as the main c-Myb bindingdependent enhancer elements. We have previously demonstrated that c-Myb exerts transcriptional control on IGFBP-5, a member of the IGF axis, through direct and indirect mechanisms in neuroblastoma cells (27). Thus, the presence of c-Myb sensitive, bindingindependent sequences in the introns of Slug would be consistent with the possibility that c-Myb cooperates with other transcription factors or by mechanisms that do not require direct binding to DNA. Furthermore, the residual c-Myb-dependent transcription activity detected with luciferase plasmids carrying mutations in the 5' regulatory region MBS is in agreement with a DNA binding-independent role of c-Myb, although, we cannot rule out the possibility that the mutations decreased but did not abolish c-Myb binding.
c-Myb-dependent regulation of slug gene expression occurs similarly in cells as diverse as embryonic kidney, colon carcinoma, chronic myeloid leukemia and neuroblastoma. Of interest, twist-1 expression is associated with MYCN amplification, advanced (metastatic) stage an poor outcome in neuroblastoma patients (43). These data and our present work suggest a scenario in which two oncoproteins (MYCN and c-Myb) expressed in neuroblastoma control at the transcription level the expression of two invasion-associated genes (twist-1 and slug) in a network that contributes to the acquisition of the invasive phenotype.
In summary, we have identified c-Myb as a new transcriptional regulator of slug. c-Myb-dependent changes in migration/invasion capabilities, bone marrow homing and resistance to drug-induced apoptosis support its critical role, in part via transcriptional regulation of Slug expression, in tumors of different origin. Transcriptional activity of each reporter plasmid alone was set at 100. C: Scheme of wt or MBSmutated reporter plasmids. D: (left) luciferase assays were carried out by co-transfection of HEK-293 and LAN-5 cells with reporter plasmids containing the wt slug promoter (prom) or the promoter mutated in site A (mut A) or in site B (mut B) or at both sites (mut AB) and an expression vector for c-Myb or the empty vector; (right), luciferase assays were carried out by co-transfecting reporter plasmids containing wt intron 1 (intron 1) or intron 2 (intron 2) or the introns mutated in site C (mut C) or in site D (mut D) and an expression vector for c-Myb or the empty vector. The difference between the transcription activity of each reporter in the presence of c-Myb compared to that of the same reporter with the empty vector was always statistically significant (p<0.05, twotailed Student's t test) except for reporter mut D whose transcription activity did not vary in the presence of c-Myb. E: luciferase assays were carried out in K562 cells infected with pLVsh-c-Myb by transfection of wt and mutated reporter vectors described in D. Doxycycline (DOX) treatment (7.5 μg/ml; 36 hours) was started 18 hours after transfection. Transcriptional activities of the wt reporters (prom, intron 1 and intron 2) cotransfected with the empty vectors were set at 100. Experiments in B, D and E were carried out twice in triplicate. Values ± SE are reported.  expression in untreated or DOX-treated COLO-205 cells infected with pLVsh-c-Myb or with pLVsh-c-Myb and pLKO-Slug. c-Myb, Slug and E-cadherin were detected after 3 days and vimentin after 6 days of DOX treatment (7.5μg/ml). β-actin was detected for normalization.