NF-κB Activation, Dependent on Acetylation/Deacetylation, Contributes to HIF-1 Activity and Migration of Bone Metastatic Breast Carcinoma Cells

Metastases represent the main cause of death in cancer patients, but how tumors spread and kill the organism is still largely unknown (1, 2). The key molecular events in breast cancer metastasis are poorly understood. Some models can be used to study the phenotypic and epigenetic changes responsible for bone colonization, and recent advances in the molecular profiling of breast cancer MDA-MB231 clones metastasizing to bone have made it possible to establish the presence of genes that closely correlate with the metastatic phenotype (3, 4). However, few studies have considered how the activities of transcription factor networks are regulated, which would clarify the pathogenesis and behavior of metastases, and identify potential molecular targets for therapeutic strategies.

Increasing evidence indicates that NF-κB activity is critical for metastatic colonization and survival, and osteolysis (5-7). The mechanisms regulating NF-κB activity involve subunit acetylation and interaction with IκBα, which plays a unique role in actively removing the p65-NF-κB subunit from the nucleus by means of nucleocytoplasmic shuttling (8-10).

NF-κB regulates epithelial/mesenchymal transition (EMT; refs. 6, 11), but metastasis does not seem to require complete transition to the mesenchymal phenotype (12). Despite their opposite significance in EMT, the E-box–containing genes E- and N-cadherins and vimentin, undergo indirect via Twist, or direct NF-κB regulation (12). E-box overlaps with the HIF-1 responsive element (HRE; ref. 13). The promoters of the genes specifically regulated in bone metastasis, such as endothelin and CXCR4, contain NF-κB and/or HIF-1 consensus sequence(s) [refs. 4, 14, 15].

We hypothesized the possible involvement of complex transcription factor networks in the bone metastasis phenotype and the role of NF-κB and HIF-1 transcription factors, considering also that metastases seem to maintain some aspects of the epithelial phenotype (12). Molecular gene therapy may take a real advantage from the identification of altered synergic interactions and control mechanisms of NF-κB and HIF-1 activities in an attempt to affect gene expression, prevent metastatic growth and improve patient outcome (5, 16, 17).

NF-κB and HIF-1 are involved in cancer progression (5, 18). The interactions of these transcription factors may link inflammation/tumorigenesis, and tumor/microenvironment (19, 20). HGF and vascular endothelial growth factor of stromal and/or tumor origin may activate HIF-1 in carcinoma cells, also in combination with hypoxic stimulus (21, 22). HIF-1 is the master regulator of oxygen homeostasis, but may be activated independently of oxygen concentrations as a result of genetic alterations and the dysregulation of growth factor signaling (18, 20, 23). Under normoxic conditions, biological stimuli such as growth factors transcriptionally regulate HIF-1α, the inducible subunit of HIF-1 heterodimer. HIF-1α promoter contains numerous consensus sequences for NF-κB and HIF-1, which allow reciprocal control to maintain an autoregulatory loop in the case of stimuli activating NF-κB in normoxia (21, 24). However, the exact mechanism(s) underpinning NF-κB/HIF-1 interactions in metastatic cells, and the signaling pathways downstream of up-regulated NF-κB have not yet been defined. Twist plays an essential role in tumor metastasis and is regulated by NF-κB and HIF-1 activities (12, 25, 26).

The aim of this study was to examine the molecular and biological characteristics of the human bone metastatic 1833 clone derived from parental highly invasive MDA-MB231 breast cancer cells (3), in an attempt to define some aspects of the metastatic phenotype, and the possible retention of some epithelial characteristics leading to an incomplete EMT. To give value to our working hypothesis and to arrive to a broad interpretation, another metastasis cancer model was used. Mouse 4T1 mammary metastatic cells show an epithelioid phenotype and seem to acquire EMT after transforming growth factor (TGF)-β1 treatment (27). We studied the reciprocal regulation of NF-κB and HIF-1, and its impact on migration in human and mouse models, because complex transcription factor interactions might cause disparate responses to identical stimuli, characterizing the molecular differences between prevalently mesenchymal and prevalently epithelial phenotypes (12). In addition to the different patterns of the transcription factors, their regulation at epigenetic level via acetylation/deacetylation might be important to define the metastatic phenotype (28). So, we investigated lysine acetylation of p50 and p65 NF-κB subunits, and the role of histone deacetylases (HDAC) under conditions of NF-κB activation and c-Src expression. The level of c-Src transducer is elevated in the most aggressive breast carcinomas (29), and seems to regulate NF-κB activity, to increase HDAC3 phosphorylation and to affect HDAC3 nuclear translocation, which may also depend on a decrease in interaction with IκBα (30-32). c-Src overexpression, which is sometimes c-Jun dependent, may represent a system regulating cell migration (33). HDACs function as corepressors/coactivators of transcription factors by means of as yet unclear mechanisms, although HDAC3 is a corepressor of NF-κB through the deacetylation of p65 (34, 35).

We found that 1833 cells retained some characteristics of differentiated epithelial cells, i.e., E-cadherins and organized structures in collagen type I, while also showing the moderate expression of N-cadherins and vimentin such as mesenchymal cells. The 1833 cells, endowed with the incomplete EMT, had more similarities with TGF-β1–treated 4T1 cells than with MDA-MB231 mesenchymal cells (36). The enhanced migratory capacity of 1833 cells after p50/p65 transfection, and of TGF-β1–treated 4T1 cells, required NF-κB-HIF-1 interaction increasing Twist levels and changing the ratio of EMT markers. By affecting these critical molecular mechanisms, c-Src expression increased, and HDAC knockdown decreased 1833 cell migration. HDAC-dependent epigenetic mechanisms and p65-lysine acetylation regulated, indeed, NF-κB and HIF-1 activities. None of these regulatory mechanisms was observed in MDA-MB231 cells.

We studied NF-κB transactivation, DNA binding activity, subunit composition, and acetylation at lysine in bone metastatic 1833 in respect to parental MDA-MB231 breast carcinoma cells (Fig. 1). The luciferase activity of the gene reporter driven by the NF-κB multimer was similar in both cell lines. Cotransfection with the super-repressor (ssNFκB), a specific and effective NF-κB inhibitor mutated at IκB-conserved serine residues typically targeted for phosphorylation (37), reduced luciferase activity by 90% in MDA-MB231 cells and by 65% in 1833 cells (Fig. 1A). The NF-κB DNA binding and the transactivating activities were reciprocally consistent (Fig. 1B): super-gelshift experiments revealed p50/50 (a), p50/65 (b), and p65/65 (c) dimers in both cell lines (Fig. 1C). ssNFκB transfection largely reduced the DNA binding of the three dimers in the MDA-MB231 cells, whereas p65/65 DNA binding was less affected in the 1833 cells, thus explaining the decrease in transactivating activity after ssNFκB. Octamer-1 DNA binding was unchanged by the treatments (data not shown).

Regardless of IκB degradation, some posttranslational modifications seemed to be involved in regulating NF-κB activity. The partial insensitivity to ssNFκB of NF-κB-DNA binding and transactivation indicated the possible role of phosphorylation. Moreover, the absence of the p52/RelB complex in super-gelshift carried out with p52 and RelB antibodies (data not shown) excluded the noncanonical pathway (9, 10).

We next examined the effects of p50 and p65 (p50/p65) cotransfection, an experimental condition that may enhance intracellular subunit levels and their posttranslational changes by increasing NF-κB activity, as occurs after hypoxia and/or treatment with cytokines such as HGF (38). In both cell lines, p50 protein levels increased in the cytosol and nuclei, whereas cytosolic p65 protein levels remained the same as in the corresponding controls (Fig. 1D). However, p65 accumulated (2-fold) in the nuclei of 1833 but not of MDA-MB231 cells. Furthermore, p65-lysine acetylation increased (1.5-fold) in the nuclei of the 1833 cells, but decreased by 40% in the MDA-MB231 nuclei. In the absence of a specific antibody that recognizes acetylated p65, we used the antibody against acetylated lysine, by establishing with immunoprecipitation experiments that the acetylated band at 64 kDa corresponded to p65 (data not shown; ref. 39). The 51-kDa lysine acetylated band coincided with p50, and increased after p50/p65 cotransfection in the 1833 nuclei (Fig. 1D).

Taken together, these findings clearly indicate that endogenous and exogenous p50 and p65 NF-κB subunits were acetylated in the nuclei of the 1833 clone, thus showing that the regulation of NF-κB activity at posttranslational level was different in invasive and bone metastatic cells.

We studied the involvement of HDAC epigenetic mechanism(s) in the control of NF-κB activity by transfecting the HDAC3 expression vector. HDAC3 or p50/p65 transfection reduced the DNA binding of all of the dimers in the MDA-MB231 cells, and the cotransfection of the three expression vectors caused a reversion that mainly enhanced p50/65 (b) and p65/65 (c) DNA binding (Fig. 1E). One possible explanation is that HDAC3 overexpression partly reduced the IκBα-Luc promoter activity stimulated by p65, thus decreasing the activity of IκBα, the negative regulator of NF-κB (40). In the 1833 cells, p50/p65 transfection enhanced p65/65 (c) DNA binding, whereas HDAC3 had no effect (Fig. 1E). The acetylation status of p65, which decreased in MDA-MB231 and increased in 1833 cells after p50/p65 transfection (see Fig. 1D), might cause the opposite changes in NF-κB DNA binding.

To evaluate the association of the endogenous NF-κB subunits and HDAC3 with IκBα, total cell extracts (500 μg of protein) were immunoprecipitated with anti-p65 or anti-IκBα antibody. In the MDA-MB231 cells, p65 coimmunoprecipitated with p100, and the derived p52 was twice that observed in the 1833 cells (Fig. 1F). Also, p65 as HDAC3 seemed to be less associated with IκBα in MDA-MD231 than in 1833 cells (Fig. 1G).

The association of p65 and IκBα was, therefore, regulated by different mechanisms in the two cell lines: (a) competition of p65 with p100 in MDA-MB231 cells (41) and (b) p65 acetylation in 1833 cells (40). These mechanisms contributed to the constitutively high level of NF-κB activity, probably because of p65 nuclear retention in both cell lines. It is worth noting that NF-κB transactivating activity was ∼20-fold higher in the aggressive breast cancer cells than in the low invasive MCF-7 cells (data not shown).

To clarify the functional role of NF-κB activity, we studied the expression of HIF-1α and c-Jun target genes, and the activity of HIF-1. As shown in Fig. 2A and B, HIF-1α mRNA and protein levels increased (2- to 2.5-fold) in 1833 cells cotransfected with the p50 and p65 expression vectors, c-Jun mRNA level tripled, and c-Jun protein level almost doubled in the nuclei but decreased by 70% in the cytosol. Nuclear c-Jun seemed to be phosphorylated as shown by the presence of a slower band. In the cotransfected MDA-MB231 cells, c-Jun mRNA and nuclear-HIF-1α protein levels diminished. The phosphoAkt (pAkt)/Akt ratio increased after p50/p65 transfection only in the 1833 cells (data not shown).

To investigate the effect of c-Jun on c-Src promoter activity, 1833 cells were cotransfected with the 0.38SRCLuc and c-Jun or c-Jun-NH₂-terminal kinase (JNK1) expression vectors (Fig. 2C). 0.38SRCLuc activity was reduced by approximately 70% and 50% after c-Jun and JNK1 transfection, respectively. In the same experimental conditions, c-Src protein levels decreased by ∼40% (data not shown).

Figure 2D shows that HIF-1 was constitutively active only in the 1833 control cells. Using the labeled HIF-1 oligonucleotide, we observed the specific binding of the dimer containing HIF-1α, and the constitutive binding possibly due to the presence of members of the activating transcription factor/cAMP-responsive element binding protein family (42, 43). The transfection of HDAC3 and p50/p65 respectively reduced and enhanced specific (HIF-1α) DNA binding. Cotransfection of the three gene reporters reduced constitutive binding, possibly because HDAC3 compete for the p300 coactivator (35). The super-gelshift assay using anti–HIF-1α antibody, which gave immunodepletion of the complex, showed the presence of the HIF-1α subunit in the specific DNA binding of HIF-1 (Fig. 2E). Unlabelled activator protein-1 (AP-1) oligonucleotide competed with HIF-1 oligonucleotide binding, but only with regard to the constitutive factor (data not shown). Octamer-1 DNA binding was unchanged by the treatments (data not shown).

To evaluate the role of c-Jun in regulating HIF-1 activity, experiments were done using TAM67 (44), a transactivation-domain deletion mutant of c-Jun that sequesters Jun and Fos family proteins in low activity complexes (45). In 1833 cells, p50/p65 transfection enhanced HRELuc activity 1.6-fold (Fig. 2F). Both basal and induced luciferase activities were largely prevented by TAM67 transfection, thus indicating a possible role of c-Jun in specific and constitutive HIF-1 transactivating activities. These findings are in line with the following: (a) possible c-Jun/HIF-1 functional cooperation; (b) overlapping of the AP-1 consensus site (first two bases) and the ACGT motif of CRE with HRE (CGTG). In addition to Jun and Fos family members, activating transcription factor/cAMP-responsive element binding protein members bind to AP-1 or other closely related motifs such as CRE (42), and there may be competition for binding to gene promoters (46).

In conclusion, p50/p65 cotransfection induced HIF-1 only in 1833 cells, thus indicating a possible regulatory interaction between NF-κB and HIF-1 activities that may have functional effects on, for example, cell migration.

On the basis of the above findings, 1833 cells were used to evaluate the functional role of HIF-1 and c-Jun in cell invasiveness after NF-κB induction. The experiments were done using a Matrigel invasion chamber, which is considered to be an in vitro model system for metastasis (38, 47). Representative images of migrated cells are shown in Fig. 3A. The transfection of p50/p65 enhanced cell invasion (∼8-fold) through Matrigel in the absence of serum, known to contain growth and chemotactic factors. TAM67 or ΔARNT transfection almost completely prevented the p50/65-dependent increase in the number of migrated cells, but ssNFκB alone or in combination with ΔARNT was less effective, leading to reductions of ∼60%. ΔARNT codes for a mutant of the HIF-1β subunit that lacks the basic domain and, although still capable of heterodimerizing with the α subunit, cannot bind DNA (13). TAM67, ssNFκB, or ΔARNT slightly decreased basal cell invasion (data not shown). The ssNFκB data concerning 1833 cell invasion were consistent with the partial inhibition of NF-κB transactivating activity and DNA binding (see Fig. 1).

The studies were extended to another model consisting of metastatic mouse mammary 4T1 cells exposed or not to TGF-β1. As shown in Fig. 3B, the activity of the gene reporter driven by the NF-κB multimer was higher in the 1833 than in the 4T1 cells, whether or not they were cotransfected with p50/p65. In the 4T1 cells treated for 48 hours with TGF-β1, luciferase activity increased 2.8-fold in comparison with the respective control (Fig. 3C).

The 4T1 cells spontaneously migrated through Matrigel, and TGF-β1 increased the migration ∼4.5-fold, but concomitant p50/65 transfection did not change the number of migrated cells (Fig. 3D). TAM67 and ΔARNT prevented about 50% and 70% of the TGF-β1–induced migration. It is worth noting that HRELuc activity in the 4T1 cells was similar to that reported for 1833 cells (data not shown).

The data obtained from the two models of metastatic cells, i.e., 1833 transfected with p50/p65 and 4T1 exposed to TGF-β1, confirmed that c-Jun protein and HIF-1 activity downstream of NF-κB were responsible for enhancing motility, although permissive molecular changes might be involved and they were analyzed in the following experiments.

As shown in Fig. 3E, the EMT markers influencing the motile phenotype of metastatic cells were analyzed in human and mouse breast carcinoma cells. For the mouse model, 4T1 metastatic cells were compared with the completely nonmetastatic tumorigenic 68H cells, with an epithelial phenotype, which did not spontaneously migrate through Matrigel (data not shown). The E-cadherin epithelial marker was markedly expressed in the 68H, less in 4T1 and 1833 cells, and not expressed in the MDA-MB231 cells. The N-cadherin and vimentin mesenchymal markers were expressed in all but not the 68H cells, with the MDA-MB231 cells showing the highest vimentin level.

Figure 3F shows that all of the cell lines expressed Twist protein. The levels were 4-fold higher in the 68H cells than in the 4T1 cells (data not shown). After p50/p65 transfection, Twist protein levels increased 2.5-fold in the 1833 cells, and decreased by 40% in the 4T1 cells. It is worth noting that p50/p65-induced Twist appeared as a diffuse pattern of multiple bands in 1833 cells, possibly due to phosphorylation (48). The opposite changes in Twist protein levels might explain why 4T1 cell migration was unchanged by p50/p65 transfection, unlike that of 1833 cells.

TGF-β1 influenced the expression of EMT markers in 4T1 cells, increasing N-cadherin and vimentin 2.5- to 2.8-fold, and decreasing E-cadherin by 50% (Fig. 3G). It is known that TGF-β1 enhances Twist mRNA expression in 4T1 cells (27), and we also observed an increase in protein level (data not shown).

The activity of the gene reporter driven by the multimer of NFκB or HRE was measured after the transfection of c-Src wild-type (Srcwt; Fig. 4), the expression vector for c-Src, which produces the protein and the phosphorylated form depending on cell conditions (31, 49).

Srcwt transfection increased c-Src protein levels ∼5-fold in MDA-MB231 and 1833 cells (Fig. 4A), and reduced NFκBLuc activity in both cell lines by about 70% (Fig. 4B). It enhanced basal HRELuc activity (1.7-fold) only in 1833 cells, a stimulatory effect that was prevented by TAM67 (Fig. 4C). These findings confirm the possible role of c-Jun in specific and constitutive HIF-1 transactivating activity in Srcwt-transfected 1833 cells.

To better define c-Src involvement in controlling NF-κB activity, coimmunoprecipitates were obtained using anti-IκBα antibody after Srcwt transfection. As shown in Fig. 4D, Srcwt decreased nuclear IκBα by 40% in the MDA-MB231 cells, and increased it 1.7-fold in 1833 cells: nuclear p65/IκBα association tripled in the MDA-MB231 cells, but did not change in the 1833 cells that, however, showed marked p65-lysine acetylation. IκBα/p65 association also considerably increased in the cytosol of Srcwt-transfected MDA-MB231 cells.

Srcwt enhanced nuclear IκBα/p65 association in MDA-MB231 cells, but caused p65 acetylation in 1833 cells: NF-κB activity therefore diminished, but by means of different mechanisms in the two cell types overexpressing c-Src.

We investigated the involvement of HDACs 1 and 3 in NF-κB and HIF-1 DNA binding under basal conditions and after Srcwt expression by knocking-down the HDACs with the specific siRNAs, which caused the almost complete disappearance of HDAC proteins (Fig. 5A). To check the reproducibility of knockdown, we used two different oligonucleotides for HDAC1: HDAC1.1 and HDAC1.3 (50). HDAC siRNA transfection increased p52 lysine acetylation in the MDA-MB231 cells, which corresponded to the 53-kDa band (Fig. 5B).

In comparison with controls, Srcwt transfection reduced NF-κB DNA binding (except for p65/65; c) in the MDA-MB231 cells, whereas the HDAC siRNAs partially reduced all of the bands (Fig. 5C). In the 1833 cells, Srcwt and both HDAC siRNAs largely decreased the NF-κB DNA binding of all of the dimers. These inhibitory effects were partially reversed by Srcwt cotransfection with HDAC siRNAs. The data were confirmed in super-gelshift experiments (Fig. 5D).

As shown in Fig. 5E, Srcwt transfection did not modify practically the specific DNA binding of HIF-1 in 1833 cells. However, the HDAC siRNAs decreased HIF-1 activity (the specific binding of the dimer containing HIF-1α and the constitutive binding) in the controls and after Srcwt transfection of 1833 cells. The presence of the HIF-1α subunit in the DNA binding of the control sample was verified by means of super-gelshift experiments using the specific antibody (Fig. 5E). HDAC1 and HDAC3 knockdown reduced c-Src protein levels by ∼50% after Srcwt cotransfection, probably by acting at promoter level (Fig. 5F).

The data relating to NF-κB and HIF-1 transactivating activities in the Srcwt-transfected 1833 cells (Figs. 4 and 5) were largely consistent with those of DNA binding, whereas the data obtained from Srcwt-transfected MDA-MB231 cells suggest that p65/65 was inactive.

Figure 6A shows the cells grown as three-dimensional cultures in collagen type I gel. MDA-MB231 cells seemed to be separate and spindle shaped (mesenchymal type; ref. 36), rarely forming colonies, whereas the 1833 cells formed more organized colonies similar to tissue structures, probably related to the expression of E-cadherins, and their morphology seemed to be more epithelial than that of the parental cells.

To clarify the influence of the biochemical changes on the biological characteristics of MDA-MB231 and 1833 cells, we did Matrigel invasion assays after Srcwt or HDAC 1 and HDAC 3 siRNA transfection. Srcwt increased the number of migrated 1833 cells by ∼3.5-fold, whereas HDAC knockdown was inhibitory (−50%; Fig. 6B). In the Srcwt-transfected 1833 cells, HIF-1α and c-Jun protein levels increased (Fig. 6C); E-cadherin protein level decreased by 50%; and N-cadherin protein level doubled (Fig. 6D). HDAC knockdown in 1833 cells showed opposite effects: it reduced migration (Fig. 6B), increased E-cadherin protein level 4-fold, and decreased N-cadherin protein level by 40% (Fig. 6D). The treatments did not change MDA-MB231 cell invasiveness through Matrigel, in line with the c-Jun reduction by Srcwt and E-cadherin appearance after HDAC knockdown and the generally unchanged levels of the other studied proteins (Fig. 6B, C, and D).

The results of this study show the NF-κB-HIF-1 reciprocal regulation that was biologically relevant for migration of human 1833 metastatic clone with bone tropism (3). Tumor cell motility is the hallmark of invasion and an essential step in metastasis, whereas EMT involvement is more controversial and continues to be debated (51).

The following molecular characteristics possibly mediated the 1833 metastatic phenotype, and lacked in parental MDA-MB231 cells: (a) the constitutively high level of NF-κB activity, which was up-regulated by the overexpression of the p50 and p65 subunits, with consequent (b) HIF-1α induction and HIF-1 activity increase above basal value; (c) the p65 lysine acetylation, which increased NF-κB-DNA binding; and (d) nuclear c-Jun level enhancement. The up-regulated NF-κB activity in the 1833 cells transactivated HIF-1α and c-Jun, which cooperatively regulated HIF-1 activity and invasiveness. This regulatory interplay was confirmed by experiments using the dominant negative of c-Jun, which largely inhibited HIF-1 transactivating activity. However, the c-Jun/c-Src system did not seem to be relevant to enhance migration of 1833 cells under our experimental conditions. HIF-1α induction implied transcriptional and posttranscriptional mechanisms, possibly favored by the increased pAkt/Akt ratio (Fig. 7; ref. 52).

What may be relevant for the markedly enhanced motile phenotype acquired by 1833 cells after p50/p65 transfection is the possible control of the expression of genes such as matrix metalloproteinase (MMP)1 by HIF-1 (53), probably in cooperation with other important transcription factors for the metastatic phenotype such as Ets1 (54), and the regulation of MMP9 and MMP2 by NF-κB (11).

We examined the validity and the sharing of this complex transcription factor interaction, critical for motile phenotype, and we considered the relationship of the metastatic ability with EMT by using a second experimental model. The fully metastatic 4T1 cells seemed invasive and largely epithelial as they expressed E-cadherins but very little N-cadherins and vimentin; however, exposure to TGF-β1 caused the partial loss of the epithelial phenotype, with reduction of E-cadherin and increase of N-cadherin and vimentin levels. The TGF-β1–induced 4T1 phenotype was therefore quite similar to that of the 1833 metastatic cells, which we can consider an incomplete EMT. The metastatic cells retain, therefore, certain epithelial characteristics (12). In line with this, invasiveness increased in the p50/p65-transfected 1833 cells and in the TGF-β1–treated 4T1 cells. The strong up-regulation of NF-κB activity participated in the enhancement of the motile phenotype, and possibly required the permissive role of Twist and the opposite changes in E-box–regulated proteins of different significance in EMT. The necessity of these concomitant events might explain the different ability of the p50/p65 and TGF-β1 treatments to increase 4T1 cell motility. Interestingly, the NF-κB/HIF-1/c-Jun interaction participated in both the spontaneous and TGF-β1–induced 4T1 motile phenotype, as it did in the p50/p65-transfected 1833 cells.

Our data concerning c-Src effects and HDAC epigenetic regulation add mechanistic insights into the role of NF-κB and HIF-1 in tumor progression and metastasis.

Under our experimental conditions, c-Src overexpression in 1833 cells increased migration and elicited the phenotypic changes associated with incomplete EMT. In relation to the signaling pathways triggered downstream of c-Src, HIF-1 activity remained high and HIF-1α/c-Jun/N-cadherins were induced although NF-κB-DNA binding decreased, probably because of the inhibitory p65 acetylation at Lys¹²² and Lys¹²³ (40). The HIF-1 target gene Twist might affect E-box–regulated proteins, which is probably why N-cadherin/E-cadherin ratio increased in the Srcwt-transfected 1833 cells exacerbating motility.

HDAC knockdown did exactly the opposite to the system consisting of NF-κB and HIF-1 activities and E-box–regulated proteins, by reducing the invasiveness of 1833 cells. We observed important effects of HDACs as corepressors or coactivators depending on the transcription factor and cell type. This was particularly clear in the 1833 cells, in which endogenous HDAC1 and HDAC3 coactivated basal NF-κB and HIF-1 activities, thus possibly participating in Srcwt-induced migration. c-Src expression may make 1833 cells responsive to microenvironmental stimuli acting through tyrosine kinase receptors (12).

Altogether our findings permit to make a comprehensive conclusion on the biochemical changes with biological relevance, underlying the differences between 1833 and MDA-MB231 cells with partial and complete EMT phenotype. EMT probably promotes early events of metastatization, such as cell migration and invasion of the basement membrane and dissemination into the circulation (51). Unlike MDA-MB231 cells, the 1833 cells formed almost organized structures in three-dimensional collagen gel, and the cohesive forces associated with multicellular aggregates might offer advantages for metastasis survival. The expansion of carcinoma cells into surrounding tissue and, even more, the colonization of secondary growth sites are probably not prevented by the presence of E-cadherins, but what is known as mesenchymal to epithelial transition and E-cadherin re-establishment may occur in vivo after cell migration to the site of metastasis. New methylations might be favored in 1833 cells with epigenetic plasticity that, for example, regulate Twist and E-cadherin expression (55). Twist seemed to play a role in the progression of breast cancer, and phosphorylation was probably important in controlling its activity under stimuli that enhanced NF-κB and HIF-1 activities.

The 1833 cells were more plastic than the MDA-MB231 cells also because their NF-κB/HIF-1–regulated gene profile might be influenced by different stimuli and epigenetic mechanisms, thus confirming that metastatic cells have the genome instability necessary to adapt to different environments during metastasis spreading (56). Clinicopathologic studies have shown that increased levels of HIF-1α are associated with tumor progression and a poor prognosis in breast cancer patients (17, 57, 58). We suggest the prevalent importance of HIF-1 transcription factor activity, which does not always correspond to HIF-1α protein levels. Only the 1833 bone metastatic cells showed an active HIF-1, whereas constitutively expressed HIF-1α cannot form a functional heterodimer with the largely mutated HIF-1β in MDA-MB231 cells (59). Thus, in MDA-MB231 cells the expression of NF-κB/HIF-1 target genes, which are typical of 1833 bone metastatic cells (3), could be impaired.

Not only HRELuc activity was very low in MDA-MB231 cells, but also p50/65 transfection reduced nuclear NF-κB subunit acetylation at lysine and NF-κB activity. The posttranslational regulation of NF-κB via the lysine acetylation of p65, which primarily occurs in the nuclear compartment, seems to determine the strength and duration of the NF-κB transcriptional response (9) and trigger downstream signaling pathways, which suggests that the persistence of NF-κB activation is critical for HIF-1–dependent transactivation of target genes, not occurring in MDA-MB231 cells.

HDAC knockdown was ineffective on spontaneous invasiveness of MDA-MB231 cells probably because of HDAC3 corepressor function on NF-κB activity. A similar mechanism might be involved in the inhibitory response to Srcwt. As consequence, HIF-1α and c-Jun levels diminished and did not contribute to changes in invasiveness. The p52 lysine acetylation in MDA-MB231 cells knocked down for HDACs, might participate in the reduction of p65/NF-κB DNA binding (39), suggesting that p52 may be a nonhistone substrate of HDACs (Fig. 8).

Our data underline the fact that the skepticism associated with the targeting of transcription factors needs to be reconsidered: complex regulation of NF-κB and HIF-1 may really influence the molecular signature of bone metastases from human breast cancer, which is different from the invasive phenotype.

Anti–c-Jun (D), anti-NFκB p65 (A), anti-NFκB p50 (NLS), anti-NFκB p52 (C-5), anti-Rel B (C-19), anti–IκB-α (C-21), anti-B23, anti-Twist (H-81), and anti-vinculin antibodies were from Santa Cruz Biotechnology. Anti–HIF-1α antibody for Western blot, anti–E-cadherin (clone 36), anti–N-cadherin (clone 32), and anti-vimentin antibodies, collagen type I, were from BD Pharmingen-Transduction Laboratories. Anti–HIF-1α antibody (OZ15) for super-gelshift was from NeoMarkers-LabVision Corporation. Anti–acetylated-lysine and anti-HDAC3 antibodies were from Cell Signaling. Anti-HDAC1 was from Abcam. Anti-p52 and anti-p100 antibodies were from G. Bonizzi (IEO, Milan, Italy). Recombinant human TGF-β1 was from R&D System. siRNA sequences (a generous gift of S. Chiocca, IFOM, Milan, Italy) were as follows: 5′-CGUACGCGGAAUACUUCGATT-3′ (siRNA LUC, control); 5′-CAGCGACUGUUUGAGAACCTT-3′ (HDAC1.1) and 5′-CUAAUGAGCUUCCAUACAATT-3′ (HDAC1.3) for siRNA HDAC1; and 5′-GAUGCUGAACCAUGCACCUTT-3′ for siRNA HDAC3.

The MDA-MB231 human breast carcinoma cells and the derived bone metastatic 1833 clone were a kind gift of J. Massagué (Memorial Sloan-Kettering Cancer Center, New York, NY; ref. 3). Mouse breast carcinoma 68H and 4T1 cells were generously given by F. Miller (Karmanos Cancer Institute, Detroit, MI) and M.P. Colombo (Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan, Italy). Some 4T1 cells were starved overnight and treated with 5 ng/mL TGF-β1 (60).

The cells seeded in 24-multiwell plates were transfected (31, 61). The gene reporter plasmid NFκBLuc containing three NF-κB consensus sequences was from M. Hung (Anderson Cancer Center, Houston, TX) and pGL3PGK6TKp containing six HRE was from P.J. Ratcliffe (Welcome Trust Center for Human Genetics, Oxford, United Kingdom); 0.38SRCLuc is the ScaI-HindIII fragment of the gene reporter construct p0.38SRCA1-CAT (62), subcloned in PGL2-enhancer vector, previously cut with SmaI and HindIII. For normalization, the cells were transfected with pRL-TK (Renilla luciferase) vector, and Firefly/Renilla luciferase activity ratios were calculated by the software. The expression vector for HDAC3 was from M. Kovacs (H. Lee Moffitt Cancer Center & Research Institute, Tampa, FL); pCMV-p50 and pCMV-p65 were from W.C. Greene (University of California, San Francisco, CA); Srcwt was from S. Parson (University of Virginia, Charlottesville, VA); pCDNA3-Flag-JNK1 was from R. Davis (Howard Hughes Medical Institute, Worcester, MA); RSVIκBαMSS, coding ssNFκB, was from N.D. Perkins (University of Dundee, Dundee, United Kingdom); pCMVc-Jun and pCMVTAM-67 were from R. Pope (Northwestern University, Chicago, IL); pcDNA3ARNTδ_b (ΔARNT) was from M. Schwarz (Institute for Toxicology, University of Tübingen, Tübingen, Germany). Cells in T75 flasks, transfected with 12.5 μg of Srcwt, were used for the immunoprecipitation experiments; cells in T25 flasks were transfected with 2.5 μg p50, p65, Srcwt, or TAM67, 6.25 μg ΔARNT, and/or 12.5 μg ssNFκB and were used for Northern blots, Western blots, or Matrigel invasion assay.

Cells were transfected with 150 nmol/L siRNAs (control, HDAC1, and HDAC3) and/or cotransfected with 2.5 μg Srcwt expression vector using Lipofectamine 2000 in six-multiwell plates. Two cycles of siRNA transfection were done at 24-h interval, and the cells harvested at 72 h were lysed in urea buffer [8 mol/L urea, 0.1 mol/L NaH₂PO₄, 0.01 mol/L Tris (pH 8), and protease inhibitors] to obtain proteins for Western blotting analysis (50), or processed to obtain nuclear extracts for Electrophoretic mobility shift assay (EMSA; ref. 31), or used to perform Matrigel assay.

The sequences of the oligonucleotides for EMSA and super-gelshift were as follows: 5′-GGATCCTCAACAGAGGGGACTTTCCGAGGCCA-3′, containing the NF-κB consensus site; 5′-TTCTGATTCGTGCCAAAGCT-3′, containing HRE; and 5′-TGCGAATGCAAATCACTAGAA-3′ containing the Octamer-1 binding site, for loading control (31).

Total RNA (30 μg), or cytosolic (100 μg) and nuclear (50 μg) proteins were used for Northern or Western blots. Total and cytosol extracts (500 μg of protein) or nuclear extracts (100 μg of proteins) were immunoprecipitated with 2 μg of anti-p65 or anti-IκBα antibody (49).

Invasion assays were carried out essentially as previously described using Matrigel invasion chambers from BD Biocoat Cellware (Becton Dickinson Labware). Transfected cells and TGF-β1–treated cells were harvested, resuspended in medium without serum, added (8 × 10⁴ per well) in triplicate to the upper chambers, and allowed to invade through Matrigel in the absence of serum. The cells that had invaded the lower surface of the membrane were fixed and stained (63).

Three-dimensional collage-type I gels were prepared on ice using collagen solution (2.5 mg/mL)/0.5 mol/L HEPES/10× DMEM/cell suspension (1 × 10⁵ cells/mL) 7/1/1/1, and allowed to copolymerize at 37°C. Cells (1 × 10⁴) were added to each well of the 96-well plate, and cell culture medium was added above (100 μL containing 20% fetal bovine serum). The plates were incubated at 37°C for 6 to 8 d, at which times complex structures were found. The three-dimensional structures were visualized by phase contrast microscopy (Olympus CKX41).

Luciferase activity and densitometric values were analyzed using ANOVA, with P value of <0.05 considered significant.

No potential conflicts of interest were disclosed.

We thank J. Massagué for 1833 and MDA-MB231 cells (Memorial Sloan-Kettering Cancer Center, New York, NY), S. Pennacchietti (Institute of Cancer Research and Treatment, University of Torino, Italy) for her helpful suggestions concerning the collagen type I experiments, and E. Erba (Istituto di Ricerche Farmacologiche Mario Negri, Milano, Italy) for his microscopy assistance.