The p65 subunit of NF-κB and PARP1 assist Snail1 in activating fibronectin transcription

Epithelial–mesenchymal transition (EMT) is a morphogenetic process accompanied by epithelial gene repression and mesenchymal gene activation (Hay, 1995) that is essential for embryogenesis (Nieto, 2002) as well as cancer development (Thiery, 2002; Mani et al., 2008). Snail1 triggers EMT when expressed in epithelial cells (Batlle et al., 2000; Cano et al., 2000). It is well established that Snail1 acts as a transcriptional repressor of epithelial genes, such as E-cadherin (CDH1). Snail1 binds to consensus sequences (E-boxes) in their proximal promoters and recruits cofactors, such as histone deacetylases (Peinado et al., 2004), Ajuba (Hou et al., 2008), the Polycomb repressive complex 2 (Herranz et al., 2008) and lysyl oxidase-like 2 (Peinado et al., 2005). Although Snail1 expression also affects the transcription of mesenchymal genes, the molecular mechanism mediating activation is poorly understood.

Growing evidence indicates that NF-κB plays a central role in EMT and metastasis by affecting the expression of mesenchymal genes (Huber et al., 2004; Min et al., 2008). In epithelial cells, a pool of the NF-κB p65 subunit (p65NF-κB) associates with E-cadherin and other cell-adhesion components. By contrast, in cells undergoing EMT, E-cadherin repression results in the release of the p65NF-κB that is attached to the membranes. This pool then enters the nucleus and binds to the promoters of mesenchymal genes, such as fibronectin (FN1), that are activated during EMT (Solanas et al., 2008).

Here, we describe how a molecular cooperation between p65NF-κB and Snail1 activates FN1 transcription. p65NF-κB anchors Snail1 to the FN1 promoter, where these components form a DNA–protein complex, together with PARP1, that regulates mesenchymal gene transcription.

In order to study the molecular mechanism by which mesenchymal genes are activated during EMT, we tested whether Snail1 binds to the proximal FN1 promoter, a paradigmatic mesenchymal gene activated during EMT. We performed chromatin immunoprecipitation (ChIP) assays in well-characterized cell culture models for EMT, such as HT29 M6 cells. These cells have a clearly epithelial phenotype but acquire a mesenchymal-like phenotype and express FN1 following Snail1–HA expression (Guaita et al., 2002). Although the proximal FN1 promoter does not contain a consensus Snail1-binding site (the E-box), a significant level of ectopic Snail1 bound to the proximal FN1 promoter was detected when ChIP was performed with an HA antibody (Fig. 1A).

At subconfluency, control and SW480 cells express fibronectin and adopt a mesenchymal-like phenotype; however, forced expression of E-cadherin imposes an epithelial phenotype on them and reduces the expression of fibronectin in both cell populations (Solanas et al., 2008). We detected that the Snail1 antibody precipitated more of the proximal FN1 promoter from control or Snail1-expressing SW80 cells than from E-cadherin-expressing or E-cadherin- and Snail1-expressing SW480 cells (Fig. 1A). This observation indicates that endogenous Snail1 also binds to the FN1 promoter. To confirm endogenous Snail1 binding, we analyzed the SW620 cell line, which is derived from a metastasis of the same tumor as the SW480 cell line but expresses higher amounts of Snail1 and lower amounts of E-cadherin (Batlle et al., 2002). In these cells, more than four times more FN1 promoter was precipitated with the Snail1 antibody than with IgG (Fig. 1B).

Epithelial NMuMG cells treated with TGF-β1 undergo EMT (Miettinen et al., 1994), accompanied by an accumulation of nuclear Snail1 and an increase in fibronectin levels (Fig. 1C). Therefore, we analyzed whether TGF-β1 treatment induces Snail1 to interact with the FN1 promoter, and how Snail1 binding to FN1 compares with its binding to the CDH1 promoter. After 1 hour of treatment, we detected nuclear accumulation of Snail1 (Fig. 1C) as well as Snail1 binding to the CDH1, but not to the FN1, promoter (Fig. 1D). By contrast, treating cells with TGF-β1 for 8 hours (Fig. 1C), which leads to fibronectin upregulation, reduced Snail1 binding to the CDH1 promoter but increased Snail1 binding to the FN1 promoter (Fig. 1D). These data indicate that endogenous Snail1 binds to the FN1 promoter when FN1 is activated during EMT.

We previously reported that a proximal nucleotide fragment from bp −341 to +265 (−341/+265; numbers are given relative to the transcription start site) of the FN1 promoter can be activated by Snail1 (Solanas et al., 2008), and we observed here that a shorter region (−341/+72) was likewise activated by Snail1 (Fig. 2A). Thus, we analyzed whether the proximal FN1 promoter mediates Snail1 binding. Using a biotinylated dsDNA sequence corresponding to the −341/+265 region of the promoter as bait, we precipitated Snail1–HA from a nuclear protein extract (Fig. 2B, lane 3). This result confirms that elements other than E-boxes are involved in the recruitment of Snail1. By contrast, we did not detect Snail1 binding to this sequence in a pull-down assay using a recombinant Snail1 protein (Fig. 2B). This finding suggests that other nuclear proteins assist Snail1 in binding to the promoter. In pull-down assays performed in parallel, we observed that a CDH1 promoter fragment (−92/−64) that included an E-box was able to precipitate Snail1–HA both in the absence and the presence of nuclear proteins (Fig. 2B, lane 6). These data indicate that the interaction of Snail1 with the FN1 and CDH1 promoters differs at the molecular level.

To narrow down the sequence of the FN1 promoter to which Snail1 binds, we repeated the pull-down assay using each of the two halves of the proximal promoter. We observed that the 3′ fragment of −36/+265, but not the 5′ fragment of −341/−37, precipitated Snail1–HA (Fig. 2B, lanes 4 and 5). On the basis of the data from the pull-down and reporter assays, we concluded that the Snail1 binding site was located in the −36/+72 region of the FN1 promoter. These results prompted us to check whether Snail1 interacted with a region containing a functional NF-κB binding site (+33/+50) required for Snail1-induced FN1 activation (Solanas et al., 2008; Fig. 2A). In an electrophoresis mobility shift assay (EMSA) performed with a +24/+53 FN1 promoter sequence as a probe (Fig. 2C), retarded bands were observed when the probe was incubated with nuclear extracts of Snail1–HA (lane 2, arrowheads), but not with E-cadherin and Snail1–HA (lane 3) cells. These specific retarded bands were altered by the addition of the Snail1 antibody, which prevented the formation of the faster band and shifted the mobility of the middle one (lane 6). These data indicate that Snail1 is present in a protein complex that binds to a FN1 promoter sequence containing an active p65NF-κB site. Accordingly, in vivo interaction of both transcription factors with the FN1 promoter was found by ChIP analysis in SW620 cells (Fig. 2D).

To study whether p65NF-κB has a role in Snail1 binding to the FN1 promoter, we analyzed the subcellular localization of this protein in TGF-β1-treated cells. Immunofluorescence (Fig. 3A) and cell fractionation (Fig. 3B) showed a clear increase of nuclear p65NF-κB at 8 hours of TGF-β1 treatment compared with no treatment (0 hours) or 1 hour of treatment, mimicking that observed for fibronectin expression (Fig. 3A, immunofluorescence; Fig. 1C, western blot). In contrast to p65NF-κB, a nuclear accumulation of Snail1 was detected earlier, after the first hour of TGF-β1 treatment (Fig. 3A, immunofluorescence; Fig. 1C, western blot). On the basis of these kinetics, we propose that by accumulating in the nucleus, p65NF-κB recruits nuclear Snail1 to the FN1 promoter, thereby contributing to the activation of this promoter.

To confirm that Snail1 and p65NF-κB interact simultaneously with the FN1 promoter in vivo, we performed reChIP assays (see Materials and Methods) using NMuMG cells treated with TGF-β1. The FN1 promoter was found in the double-immunoprecipitates of cells treated with TGF-β1 for 8 hours, but not in untreated cells or cells treated for only 1 hour. This interaction was detected independently of the order in which Snail1 and p65NF-κB were immunoprecipitated, but it was not detected when control IgGs were used (Fig. 3C). This result confirms that Snail1 and p65NF-κB interact simultaneously with the FN1 promoter at 8 hours, a timepoint at which fibronectin is actively expressed.

We next analyzed whether Snail1 and p65NF-κB were present in the same protein complex. Using co-immunoprecipitation we did indeed detect p65NF-κB in the anti-HA immunoprecipitate obtained from nuclei of Snail1-transfected HEK-293T cells, but not from those of mock-transfected cells (Fig. 4A). In addition to p65NF-κB, its interacting cofactor, PARP1 (Hassa and Hottiger, 1999), was found to co-immunoprecipitate with Snail1 (Fig. 4A). This suggested that the protein complex also included chromatin structure modulators that might contribute to the modification of the transcription activity. These three proteins also co-immunoprecipitated from nuclear extracts from HT29 M6 cells that stably expressed Snail1–HA (data not shown). Additionally, in this cell model, we detected colocalization of Snail1 with p65NF-κB and PARP1 in the nuclei, using immunofluorescence (supplementary material Fig. S1).

It is noteworthy that p65NF-κB and PARP1 co-precipitated with endogenous Snail1 from SW620 cells (Fig. 4B), as well as from NMuMG cells treated for 8 hours with TGF-β1 (Fig. 4C). Likewise, complexes that co-immunoprecipitated with p65NF-κB from TGF-β1-treated NMuMG cells contained Snail1 and PARP1 (Fig. 4C); note that, although only low amounts of PARP1 co-precipitated after 8 hours of TGF-β1 treatment, this was detectably higher than that for the non-treated cells. By contrast, neither PARP2 nor β-catenin were detectable in the complex (supplementary material Fig. S2). Immunofluorescence revealed that Snail1, p65NF-κB and PARP1 colocalized in the nuclei of NMuMG cells that had been treated for 8 hours with TGF-β1 (Fig. 4D), further supporting an interaction between PARP1, p65NF-κB and Snail1. Therefore, we conclude that the endogenous proteins PARP1, p65NF-κB and Snail1 form a ternary complex in the nuclei of cells that are actively expressing fibronectin.

To determine the extent to which p65NF-κB and PARP1 are involved in activating fibronectin expression, their expression was interfered with by using shRNAs in NMuMG cells. TGF-β1 failed to activate FN1 expression in cells depleted of either PARP1 or p65NF-κB (Fig. 4E; supplementary material Fig. S3), indicating that these two proteins are required to activate FN1. In these shRNA-expressing cells, TGF-β1 promoted effective nuclear accumulation of Snail1 (Fig. 4E; supplementary material Fig. S3) and E-cadherin repression (not shown). By contrast, TGF-β1 did not induce nuclear accumulation of Snail1 in NMuMG cells that stably expressed RELA (the gene encoding p65NF-κB) shRNAs by puromycin selection (data not shown).

Because PARP1 co-precipitated with p65NF-κB and Snail1, and was required for FN1 activation by TGF-β1, we next investigated whether it also interacts with the proximal FN1 promoter. For this, we performed EMSA using the FN1 promoter probe (+35/+48) that was shown to bind Snail1 (Fig. 2C) and p65NF-κB (Solanas et al., 2008) (Fig. 5A). Incubating the EMSA mixture with anti-PARP1 antibody inhibited the formation of the same band that was inhibited by anti-Snail1 and anti-p65NF-κB antibodies (Fig. 5A). Additionally, ChIP assays of SW620 cells detected the binding of not only Snail1 and p65NF-kB to the FN1 promoter but also PARP1 (Fig. 2D). Likewise, PARP1 bound to the FN1 promoter in HT29 M6 and SW480 cells that ectopically expressed Snail1 (Fig. 5B). PARP1 is an abundant nuclear protein, and its levels in the nucleus remained unchanged irrespective of the presence of Snail1 (Fig. 4A) or treatment with TGF-β1 (Fig. 3B). PARP1 was also found bound to the FN1 promoter in HT29 M6 cells and SW480 cells expressing E-cadherin and Snail1–HA, which have an epithelial phenotype (Fig. 5B; see the Discussion).

We obtained a list of candidate genes from a gene expression microarray that might also be upregulated by Snail1 and NF-κB. Similar to FN1, the expression levels of 168 genes increased by more than twofold (supplementary material Table S4) after treatment of cells with TGF-β1 for 8 hours as compared with 1 hour. Of those genes, 32 were related to cell movement, according to ingenuity pathway analysis (Fig. 6A). We further analyzed HAS2, THBS1 and LAMB3, three genes that, like FN1, encode extracellular matrix (ECM) proteins and that contain NF-κB binding motives near to their transcription start site (Fig. 6A). The expression of these genes was also activated by ectopic Snail1 in epithelial cells (Fig. 6B). In reChIP assays, binding of Snail1 and p65NF-κB to the promoter of these genes was detected only in cells treated with TGF-β1 for 8 hours (Fig. 6D,E). Therefore, we conclude that, in collaboration with p65NF-κB, Snail1 regulates other ECM genes related to cell motility during EMT.

In order to determine whether our findings about FN1 transcription can be generalized for other cells expressing fibronectin, we analyzed FN1 regulation in fibroblasts. We measured FN1 mRNA expression in mouse embryonic fibroblasts (MEFs) deficient for one of the following genes: SNAI1, RELA or PARP1. Depletion of any one of these genes resulted in a decrease in FN1, suggesting that the Snail1–p65NF-κB–PARP1 complex also controls FN1 transcription in MEFs (Fig. 7A). Accordingly, the FN1 promoter activity in PARP1^−/− MEFs was decreased by 60% compared with PARP1^+/+ MEFs (supplementary material Fig. S4). However, the activity of an FN1 promoter carrying a mutation in the p65NF-κB binding motif (+35/+48) was similar in the two MEF populations (supplementary material Fig. S4). These findings stress the importance of NF-κB and PARP1 for FN1 transcription in MEFs. In agreement with this, Snail1, p65NF-κB and PARP1 in these fibroblasts also co-immunoprecipitated (Fig. 7B), suggesting that PARP1 might be necessary for the assembly and/or the stability of the protein complex, because the co-immunoprecipitation of this protein complex was lost in PARP1^−/− cells (Fig. 7B). In adult 1BR3-G fibroblasts, the levels of both Snail1 and fibronectin increased in response to TGF-β1 treatment (supplementary material Fig. S5). The increase of fibronectin, but not of Snail, was prevented by transducing the cells with RELA shRNAs (supplementary material Fig. S5), highlighting the requirement of p65NF-κB for regulation of fibronectin in these fibroblasts.

Expression of ectopic Snail1 in 1BR3-G adult fibroblasts increased the expression levels of fibronectin (Fig. 7C) as well as of the ECM genes that we observed to bind to Snail1 and p65NF-κB after 8 hours of TGF-β1 treatment (Fig. 6C). Depleting PARP1 expression with specific shRNAs interfered with the observed FN1 increase (Fig. 7C), indicating that PARP1 is also required for FN1 expression in Snail1-activated fibroblasts. To further characterize the mechanism by which PARP1 affected FN1 transcription, we examined whether PARP1 influenced the binding of Snail1 and p65NF-κB to the FN1 promoter. Although both p65NF-κB and Snail1 interacted with the promoter in PARP1^+/+ MEFs, this interaction did not occur in PARP1^−/−MEFs (Fig. 7D). Therefore, PARP1 promotes the formation and/or the stabilization of a p65NF-κB–Snail1 complex that is bound to the FN1 promoter.

Taken together, our results indicate that Snail1, p65NF-κB and PARP1 interact to activate the expression of fibronectin and other ECM genes involved in cell movement. This mechanism is functional not only in epithelial cells undergoing EMT but also in fibroblasts. Although PARP1 appears to promote the formation and/or stabilization of a promoter-bound complex, p65NF-κB could be the factor that anchors Snail1 to mesenchymal promoters rather than to epithelial ones. Because repression of epithelial genes, such as E-cadherin, must be maintained throughout EMT, we expect that other repressors in addition to Snail1 are involved. In fact, ZEB1 was found to be activated downstream of Snail1 (Guaita et al., 2002), and it was upregulated in response to TGF-β1 (microarray data, supplementary material Table S4) (Dave et al., 2011).

Despite the fact that we have reported that β-catenin is required for Snail1 to activate mesenchymal genes (Solanas et al., 2008), and that others have reported an interaction between β-catenin and Snail1 (Stemmer et al., 2008), we did not detect a nuclear in vivo interaction between β-catenin and Snail1 or p65NF-κB (supplementary material Fig. S2; data not shown). Our data suggest that the β-catenin interaction with the FN1 promoter is independent of Snail1 and p65NF-κB.

We have recently reported that, upon Snail1 induction, the transcription factor TFCP2c binds to the −341/−323 region of the FN1 promoter (Porta-de-la-Riva et al., 2011). However, we found no evidence linking this molecular event to the molecular complex we describe here; specifically, TFCP2c did not interact with Snail1, and anti-Snail1 did not supershift a specific EMSA band when nuclear extracts from Snail1-expressing cells were incubated with a −341/−323 probe (data not shown). TFCP2c was not required for the transcription of other mesenchymal genes, such as THBS1 or LEF1, suggesting that although the p65NF-κB–Snail1 mechanism affects a cluster of genes, the TFCP2c mechanism is probably restricted to FN1.

In contrast to the induced binding of Snail1 and p65NF-kB to the FN1 promoter, we detected constitutive PARP1 binding to FN1 in epithelial HT-29 M6 and SW480 E-cadherin–Snail1-expressing cells. Similarly, we detected PARP1 binding to FN1 and THBS1 promoters in non-stimulated NMuMG cells that was not significantly altered by TGF-β1 treatment (data not shown). Consistent with the role of PARP1 as a nucleosome-binding protein that regulates chromatin architecture and the transcriptional outcome (Krishnakumar et al., 2008), a pool of PARP1 bound to these promoters in epithelial cells could promote an inactive transcriptional chromatin structure. Upon EMT-inducing signals, the interaction of PARP1 with p65NF-κB and Snail1 at specific promoter sites might provide a switch that facilitates an active structure. Ménissier de Murcia et al. demonstrated that PARP1/2-double-null mice die at the onset of gastrulation, and they concluded that DNA-dependent poly (ADP-ribosyl)ation is essential during early embryogenesis (Ménissier de Murcia et al., 2003). However, we observed that FN1 upregulation in Snail1 cells was not modified by PAR inhibitors (data not shown); therefore, it is possible that poly(ADP-ribosyl)ation mediated by PARP1 (and by PARP2) regulates events in early embryogenesis rather than mesenchymal gene expression.

SNAI1-null mice are not viable as a consequence of undergoing incomplete EMT. In the SNAI1^−/− embryos, E-cadherin is anomalously expressed in mesodermal cells, which retain epithelial characteristics (Carver et al., 2001), suggesting that the phenotype of the null mice is a consequence of a failure of epithelial gene repression. The fact that both PARP1-null (Wang et al., 1997) and RELA-null (Beg et al., 1995) mice form normal mesodermal cells indicates that the mechanism we describe is dispensable for gastrulation. It is possible that, in this specific cellular context, other transcriptional factors contribute to FN1 expression. Alternatively, in PARP1- and RELA-null embryos, basal ECM components might be supplied by neighboring cells through a mechanism that does not require these factors. Nonetheless, the importance of Snail1 for the acquisition of invasive characteristics in tumors of different origins (Rowe et al., 2009; Wells et al., 2008) suggests that pharmacological intervention of this pathway might have therapeutic relevance.

Cells were grown in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 4.5 g/l glucose (Life Technologies), 2 mM glutamine, 56 IU/ml penicillin, 56 μg/l streptomycin and 10% fetal bovine serum (FBS; GIBCO). Cells were maintained at 37°C in a humid atmosphere containing 5% CO₂. Where indicated, cells were treated with TGF-β1 (Peprotech). SW620, 1BR3G, NMuMG, HEK-293T and HEK-293 Phoenix cells were acquired from the repository stock of our center. Stable HT29 M6 clones expressing mouse (m) Snail1–HA were generated (Batlle et al., 2000) and maintained in our laboratory. Stable expression of mSnail1–HA was conserved by adding the antibiotics G418 (500 μg/ml) and hygromycin (200 μg/ml) to the culture medium. PARP1^−/− and PARP1^+/+ MEFs were kindly provided by Zhao-Qi Wang, Fritz Lipmann Institute, Germany, and RELA^−/− and RELA^+/+ MEFs were provided by A. Hofmann, University of California, USA. SNAI1^−/− and SNAI^+/+ MEFs were established in our laboratory from a conditional knockout mouse (manuscript in preparation). SW480 cells expressing Snail1–HA, E-cadherin or Snail1–HA and E-cadherin were previously described (Solanas et al., 2008; Pàlmer et al., 2001).

1BR3G human fibroblasts were infected with retroviruses using pBABE mSnail1–HA (cDNA) and GFP vectors and then selected with puromycin (2 μg/ml). The cell line was grown in DMEM (Invitrogen) supplemented with 10% FBS (Biological Industries), 1 mM L-glutamine and 100 U/ml penicillin–streptomycin, at 37°C in 5% CO₂.

For transient transfection of HEK-293T or PARP1^−/− and PARP1^+/+ MEFs, cells were grown to 60–80% confluency and transfected with the pcDNA or pcDNA Snail1–HA vector (Batlle et al., 2000) with PEI reagents (Polysciences Inc., cat. no. 23966). Cells were kept in complete medium for 48 hours before immunoprecipitation or ChIP analysis.

For lentiviral infection, HEK-293 Phoenix cells were used to produce viral particles. Cells were grown to 90–110% confluency and then transfected (day 0) by adding, drop-wise, a mixture of NaCl, DNA (50% pLKO shRNA, 10% pCMV-VSV-G, 30% pMDLg/pRRE and 10% pRSV rev) and polyetherimide polymer that had been pre-incubated for 15 minutes at room temperature. The transfection medium was replaced with fresh medium after 24 hours (day 1) and the cell-conditioned medium at day 2 was filtered and used to infect target cells with 8 μg/ml polybrene. HEK-293 cells were incubated with fresh medium for a further 24 hours and on day 3, a second infection with the conditioned medium and polybrene was performed. Infected cells were lysed on day 4 for protein analysis. PARP1 shRNAs [PARP1₁ MISSION TRCN0000007931 (Fig. 4; Fig. 7) and PARP1₂ MISSION TRCN0000007930 (Fig. 7; supplementary material Fig. S3)], RELA (p65NF-κB) shRNAs [MISSION TRCN0000235832 (Fig. 4; supplementary material Fig. S5) and TRCN0000235833 (supplementary material Figs S3, S5)] and a scrambled irrelevant shRNA were obtained from Sigma. The three other shRNAs for RELA (MISSION TRCN0000235834, TRCN0000244319 and TRC0000055346) that were also tested were less efficient in depleting p65NF-κB.

The −341/+265 FN1 promoter was amplified from the −341/+370 fragment (Domínguez et al., 2003) with Pfx polymerase and the specific sense and antisense oligonucleotides: 5′-CCCCACGCGTACACAAGTCCAGCCACTCCC-3′ and 5′-GTTGAGACGGTGGGGAGAG-3′. The sense primer contained the restriction site for MluI, and the amplified promoter was digested with MluI and cloned into a PGL3*-luciferase vector (Domínguez et al., 2003) that had been previously digested with MluI and SmaI. A NF-κB box-mutated FN1 promoter was synthesized following the QuickChange TM site-directed mutagenesis protocol (Stratagene) using the cloned −342/+265 promoter as a DNA template and the primers of 5′-CTGCACAGGGGGAGGAGAGAGATCTGGAGGCGCGAGCGGG-3′ and its reverse complementary oligonucleotide. The −341/+72 FN1 promoter was generated by digesting the −341/+370 FN1 promoter with PstI and XhoI and then religating it.

RNA was extracted with the GenElute TM Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). For quantitative analysis, 1 μg RNA was retrotranscribed with the Transcriptor First Strand cDNA Synthesis Kit (Roche), and 100 ng of the cDNA obtained was used as the template for quantitative SYBR Green-based PCR with specific oligonucleotides (see supplementary material Table S1). Correct product size was confirmed in agarose gels. The amount of RNA calculated was systematically normalized to the amount of HPRT RNA. For microarray analyses, amplification, labeling and hybridizations were performed according to the protocols from Ambion and Affymetrix. Briefly, 250 ng total RNA was amplified using the Ambion® WT Expression Kit (Ambion/Applied Biosystems), labeled using the WT Terminal Labeling Kit (Affymetrix Inc.), and then hybridized to Mouse Gene 1.0 ST Array (Affymetrix) in a GeneChip® Hybridization Oven 640. Washing and scanning were performed using the Hybridization Wash and Stain Kit and the GeneChip® System of Affymetrix (GeneChip® Fluidics Station 450 and GeneChip® Scanner 3000 7G). The ingenuity pathway analysis software, IPA (Ingenuity Systems) was used to analyse diferential gene expression and protein functions.

Nuclear extracts were prepared using a variation of the Dignam procotol for preparing transcriptionally active protein extracts from HeLa cell nuclei (Dignam et al., 1983). Cells were washed twice with PBS and then scraped into 500 μl of cytoplasmic extraction buffer [buffer A, consisting of 10 mM Hepes pH 7.8, 1.5 mM MgCl₂, 10 mM KCl and 0.5 mM dithiothreitol (DTT)]. After 10 minutes incubation on ice, Triton X-100 was added to a volume of 1/30 of the extract, and the samples were vortexed for 30 seconds. Subsequently, samples were centrifuged at 11,500 g for 1 minute, and the cytosolic extract was purified from the supernatant while the nuclear extracts were prepared (see below). The supernatant, which contained the cytosolic fraction, was recovered and treated with buffer B (0.3 mM Hepes, pH 7.8, 1.4 M KCl, 30 mM MgCl₂), added at 1.1× the volume of the cytosolic fraction, and the mixture was incubated for 30 minutes under agitation. The mixture was then centrifuged at maximum speed for 30 minutes, and the supernatant was reserved. To prepare the nuclear extract, the pellets containing nuclei were washed three times in the cytosolic extraction buffer, to remove contamination, and then resuspended in 100 μl nuclear extraction buffer (buffer C, consisting of 20 mM Hepes, pH 7.8, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA and 0.5 mM DTT). Samples were incubated at 4°C for 30 minutes with agitation and then centrifuged at maximum speed for 15 minutes. The supernatant contained the nuclear fraction.

For immunoprecipitation assays, 2.5 μg of the indicated antibody was added to 0.5–1 mg of nuclear extracts pre-cleared with protein-A- or -G-combined magnetic beads and incubated overnight at 4°C. Antibody-bound proteins were pulled-down with 15 μl of protein-A- or -G-combined magnetic beads, washed three times with buffer C, and then three times with buffer A with 0.1% Triton X-100. The washed beads were resuspended in 15 μl sample buffer and used for western blot analysis.

For immunofluorescence (IF) analysis, cells were grown for at least 48 hours on ethanol-sterilized glass coverslips following a standard IF protocol. All steps were carried out at room temperature. Cells were fixed with 4% paraformaldehyde (PFA) for 10 minutes. For nuclear colocalization experiments, coverslips were washed with PBS plus 0.025% NP-40 before fixation, in order to decrease the cytosolic staining of p65NF-κB. PFA autofluorescence was quenched by incubating with 50 mM NH₄Cl in PBS for 5 minutes, and cells were then permeabilized with 0.2% Triton X-100 for 5 minutes. Blocking was carried out for 1 hour with PBS containing 3% bovine serum albumin (BSA). Coverslips were incubated for either 4 hours or overnight with a specific primary antibody in blocking solution, and then for 45 minutes with a secondary antibody. For colocalization, cells were initially incubated with a solution containing all of the primary antibodies and then with a solution containing all of the secondary antibodies. Negative controls were performed in parallel, in which the cells that were incubated in a solution lacking the indicated primary antibodies. Where indicated, nuclei were counterstained with DAPI. Coverslips were mounted with fluoromont G, and fluorescence was viewed and captured using a Leica TCS-SP2 confocal microscope. The following secondary antibodies were used: anti-rabbit Alexa Fluor 647, anti-rabbit or anti-mouse Alexa Fluor 555 and anti-rabbit or anti-mouse Alexa Fluor 488.

Nuclear extracts and EMSAs were performed as described by Solanas et al. (Solanas et al., 2008). Forward and reverse oligonucleotides of 5′-gggggaggagaGGGAACCCCAGGcgcgagc-3′, corresponding to the +24/+53 FN1 promoter sequence (the NF-κB consensus is shown in capital letters), were annealed in TEN buffer (10 mM Tris, pH 7.5, 50 Mm NaCl, 1 mM EDTA) for 10 minutes at 70°C and allowed to cool to room temperature (for 3 hours or overnight). dsDNA was then labeled with γ-³²P using T4 polynucleotide kinase (Invitrogen) for later use as a probe, following the manufacturer's instructions. Excess unincorporated radioactive ATP was removed using Microspin TM G-25 columns (Amersham Pharmacia Biotech Inc.). Radioactivity was quantified using 1 μl samples.

EMSA reactions were incubated for 30 minutes on ice in a final volume of 15 μl, containing 10 μg nuclear extract and 100,000 c.p.m. ³²P-labeled probe in binding buffer (20 mM Hepes, pH 7.9, 100 mM KCl, 3 mM MgCl₂, 4% Ficoll, 0.1% Nonidet P-40, 1.5 mM ZnCl₂, 0.5 mg/ml BSA, 10 μg poly(dI-dC), 1 mM DTT protease and phosphatase inhibitors). For competition experiments, the stated amount of non-radiolabeled probe was added to the reaction. Where indicated, antibody was added to the reaction 15 minutes before the radiolabeled probe.

Non-denaturing Tris–borate–EDTA (TBE)-polyacrylamide gels were prepared and polymerized overnight at 4°C, pre-run for at least 1 hour at 100 V, and then loaded with the samples. Gels were run at constant voltage (125 V), dried and exposed to autoradiography.

Recombinant GST–mSnail1–HA protein (10 ng) was incubated overnight at 4°C with 200 ng FN1 promoter probe (−341/+265) in binding buffer, under agitation. 20 μl total volume of streptavidin-conjugated beads (Roche), blocked with 1% BSA for 1 hour at room temperature, was added to each sample, and the slurrys were incubated for 10 minutes at room temperature with agitation. Biotinylated streptavidin-conjugated probes were precipitated by centrifuging for 5 minutes at maximum speed. The beads were then washed three times, for 10 minutes each, with washing buffer and resuspended in 20 μl sample buffer for western blot analyses.

Nuclear extract (250 μg) was pre-incubated with binding buffer and 15 μl of effective streptavidin-combined magnetic beads (New England Biolabs) for 3 hours at 4°C. Samples were then incubated in a Dynal MPC®-S Magnet (Invitrogen) and the supernatant was recovered, with 10% of the volume stored at 4°C to use as the input. DNA probe (250 ng) was added to each sample, and the samples were incubated overnight at 4°C with agitation. 15 μl of effective streptavidin-conjugated magnetic beads (New England Biolabs) was added to each tube and the samples were incubated under agitation for 10 minutes at 4°C. Biotinylated probes were pulled down with the streptavidin-combined magnetic beads. Three washes of 10 minutes each were performed at 4°C with agitation in the same buffer used for extracting the nuclear fraction. The remaining beads were resuspended in 15 μl sample buffer for western blot analysis.

Assays were performed as described by Peiro et al. (Peiro et al., 2006). Cells, seeded in 150-mm plates, were washed twice with PBS pre-warmed to 37°C, and then cross-linked with 1% formaldehyde for 10 minutes at 30°C in DMEM. Reactions were stopped by adding 250 μl of 2.5 M glycine (to a final concentration of 0.125 M) and incubating for an additional 2 minutes. Cells were washed twice with cold PBS, and 1 ml of soft lysis buffer (50 mM Tris, pH 8.0, 2 mM EDTA, 0.1% Nonidet P-40, 10% glycerol protease and phosphatase inhibitors) was added to the plates on ice. After scraping, lysates were transferred to Eppendorf tubes, incubated for 10 minutes on ice, and centrifuged for 15 minutes at 850 g. Pellets were resuspended in SDS lysis buffer (1% SDS, 10 mM EDTA and 50 mM Tris, pH 8.0) and sonicated to generate DNA fragments ranging from 200 to 1000 bp (using a 40% amplitude in Branson DIGITAL Sonifier® UNIT Model S-450D sonicator, 10 pulses of 10 seconds each). Lysates were incubated for 20 minutes on ice and subsequently centrifuged at maximum speed for 10 minutes.

The protein concentration of supernatants was determined by the Lowry assay, and 500 μg of protein per immunoprecipitation (IP) was diluted in dilution buffer (0.001% SDS, 1.1% Triton X-100, 16.7 mM Tris, pH 8.0, 2 mM EDTA and 167 mM NaCl). To reduce background, samples were pre-cleared for 3 hours at 4°C with agitation, using agarose–protein G (Upstate) blocked with mouse IgG (Dako) and salmon sperm–BSA. Samples were then centrifuged at 400 g to remove the beads. A portion (10%) of each sample was stored as input, and the rest of the samples were divided for IP and incubated with either a specific antibody (supplementary material Table S2), or a non-specific antibody of the same species, overnight at 4°C under agitation.

Blocked beads were added to each sample, and the samples were incubated for an additional 1 hour at 4°C. Five washes were performed on MoBiTec columns in each of the buffers: low salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris, pH 8.0, 150 mM NaCl), high salt buffer (the same as low salt but with 500 mM NaCl), LiCl buffer (250 mM LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA and 10 mM Tris, pH 8.0) and TE buffer. Columns were centrifuged to remove traces of buffer (2 minutes, 400 g). The remaining beads were incubated with elution buffer (100 mM Na₂CO₃ and 1% SDS) at 37°C for 30 minutes, and the precipitate was recovered by centrifugation (5 minutes, 400 g). In reChIP assays, a second IP was performed: eluates were brought to a final volume of 1 ml with the dilution buffer and re-incubated with either specific antibody or the proper control, overnight with agitation. The precipitation process was then repeated as described for the first IP.

To decrosslink, samples were supplemented with NaCl to 200 μM and incubated at 65°C overnight. After 2–4 hours of digestion with proteinase K (55°C), DNA was purified using the GFX PCR DNA and Gel Band Purification kit (Amersham Biosciences), and finally analyzed by semi-quantitative or quantitative SYBR Green PCR (supplementary material Table S3). A minimum of three independent experiments were performed, and the results are shown as mean values with the standard deviations.

Cells grown to 60–80% confluency were trypsinized and seeded in 24-well culture plates (1×10⁵ cells/well). After 24 hours, they were transfected with 500 ng of the indicated FN1 promoter sequence cloned into the PGL3* reporter plasmid and 10 ng of thymidine kinase–Renilla luciferase plasmid (Promega). Transfection was performed with LipofectAMINE Plus (Life Technologies) following the manufacturer's instructions. The activities of firefly and Renilla luciferases were measured as relative luminescence units (RLU) using the Dual Luciferase Reporter Assay System (Promega) 48 hours after transfection with an FB12 luminometer (Berthold Detection Systems, Pforzheim, Germany). Firefly RLU values were normalized to Renilla RLU values and an empty reporter vector. Triplicate samples were systematically included, and experiments were repeated at least three times. The results are shown as mean values with the standard deviations.

We thank Raúl Peña for technical assistance, José Yelamos (IMIM-Hospital del Mar, Spain) for providing the monoclonal PARP1 antibody, Zhao-Qi Wang (Fritz Lipmann Institute, Germany) for providing the PARP1^−/− and PARP1^+/+ MEFs, and A. Hofmann (University of California, USA) for the RELA^−/− and RELA^+/+ MEFs.