TAF4b and Jun/Activating Protein-1 Collaborate to Regulate the Expression of Integrin α6 and Cancer Cell Migration Properties

The binding of the transcription factor IID (TFIID) complex, composed of the TATA box-binding protein (TBP) and 14 TBP-associated factors (TAF), to promoter DNA is responsive to cellular signals and constitutes the first step in transcription. A number of different TFIID forms with functionally distinct properties exist, among which the TAF10-free TFIID, the TAF4b-containing TFIID, the TAF6δ-containing TFIID, the TBP-free TFIID, and the seven TAF complex have been described (1, 2). Notably, several studies suggest that TAFs are important in specific events like cell cycle regulation (3, 4), apoptosis (5, 6), cancer, and epithelial-to-mesenchymal transition (EMT; ref. 7). TAF4b was first identified as a tissue-specific TFIID subunit, present only in a limited number of complexes, and was later shown to be necessary for ovarian follicle development, proliferation, and function (8). TAF4b shares high homology with the COOH-terminal part of TAF4 in contrast to its coactivator NH₂-terminal domain (9). TAF4b contains a nuclear export signal allowing it to shuttle between the nucleus and the cytoplasm (10), although it displays DNA-binding capacity when incorporated into the TFIID (11). Even though there is no evidence for a direct sequence-specific contact between DNA and TAF4b, the involvement of TAF4b in direct promoter-selective recognition and subsequent recruitment of activators in a cell type–specific manner has been suggested (12). Indeed, transcriptional induction of the activating protein-1 (AP-1) family member c-Jun by TAF4b in granulosa cells has recently been proposed (13). As a member of the AP-1 transcription factor, c-Jun participates in the control of cellular responses, mainly by converting extracellular signals into specific gene expression profiles via the general transcription machinery. Altering the transcription of target genes, c-Jun has been shown to interact with the coactivator CBP (14) and with TAF7 in HEK293 and COS cells (15).

c-Jun follows a two-stage activation pattern including a step of phosphorylation by mitogen-activated protein kinases (ERK, JNK, p38) and a subsequent selective formation of dimers whose nature defines the activation of a specific subset of AP-1 binding site containing target genes (16, 17). Notably, AP-1 activity is frequently elevated in transformed cell lines due to an oncogene-specific upregulation of the AP-1 family members c-Jun, JunB, Fra-1, and Fra-2 (18, 19). Different types of tumors have been related to RAS-protein activation, which in turn, regulate the activity of AP-1 (20). For instance, c-Jun, which is frequently implicated in the acquisition of invasive properties in aggressive forms of cancer (21), is required for in vitro cellular transformation by oncogenic RAS partially via a phosphorylation mechanism (22, 23).

Colorectal carcinogenesis occurs through the accumulation of gene alterations in tumor suppressor genes and oncogenes including RAS (24), leading to invasion/metastasis (25). EMT, occurring during the last steps of cancer progression prior to metastasis, is controlled by a number of regulators resulting in a loss of cell-cell adhesion, mediated by repression of E-cadherin, whereas vimentin and other mesenchymal proteins like matrix metalloproteinases and fibronectin are upregulated (26). Importantly, activation and maintenance of EMT can be achieved by the signaling cascade of an oncogenic form of Harvey RAS (Ha-RAS; ref. 27). Even though the phenomenon of EMT reflects a transient state in vivo, by constitutively expressing the mutated Ha-RASV12 in the intermediate colon adenoma Caco-2 cell line, we have created a cell line (Caco-H) which adopts and maintains an EMT state (28).

In this study, focusing on the investigation of Jun family members and their interplay with TAFs in colon cancer and metastasis, we have identified an interaction between c-Jun and TAF4b and have evaluated its effect in the regulation of integrin α6, an EMT-related AP-1 target gene. The implication of other AP-1 family members in this mechanism suggests a dynamic switch between these proteins and their interacting partners in the control of transcription.

Caco-2, HT29, and HCT116 cells were obtained from American Type Culture Collection and cultured in DMEM supplemented with 10% fetal bovine serum, antibiotics, and nonessential amino acids (all from Invitrogen, Corp.). Caco-2 cells constitutively overexpressing HRASV12 (Caco-H) were cultured as mentioned above. For reasons of consistency, the name Caco-H will be used throughout the text when referring to Caco-2 cell lines overexpressing Ha-RASV12 in the case of presenting results of only one Caco-H cell line. Although in some experiments, two different Caco-H cell lines were presented (referred to as Caco-H1 and Caco-H2) for the validation of results.

Nuclear, cytoplasmic, and whole cell lysate extracts were prepared as described earlier (19, 29). Protein concentrations were determined by the Bradford method using a Bio-Rad protein assay kit. Extracts were subjected to SDS-PAGE and transferred to a nitrocellulose membrane (Pall Corporation). The antibodies used for immunoblotting are described in Supplementary Data. Signals were visualized using enhanced chemiluminescence (Amersham Biosciences) after exposure to Kodak Super RX film. All experiments were repeated at least three times. Representative images are shown.

RNA was prepared from sampled cells by the TRIzol reagent (Invitrogen). Reverse transcription was carried out using the SuperScript Reverse Transcriptase (Invitrogen) and oligo(15)-(dT), following the instructions of the manufacturer. Primers are described in the Supplementary Data. Values were measured using the Image-Quant software (Amersham Biosciences). All experiments were repeated at least three times. Representative images are shown.

Real-time quantification was carried out using a Bio-Rad iCycler and the iQ5 Multicolor Real-time PCR detection system (Bio-Rad). Cycling conditions included a denaturing step of 3 min at 95°C followed by 40 cycles at 95°C for 40 s and annealing/elongation at 60°C for 40 s. All genes were tested in triplicate. Values were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Results were analyzed on the iCycler software.

Nuclear protein extracts (100 μg) were incubated overnight at 4°C under rotation with 5 μg of c-Jun, JunB, JunD, and TAF4b (9) antibodies in a total volume of 500 μL of 100 mmol/L NaCl immunoprecipitation buffer, adding 25 μL of dry Protein A-Sepharose matrix CL-4B (Amersham Biosciences) over a period of 2 h, followed by three washing steps with 500 mmol/L of KCl immunoprecipitation buffer and 100 mmol/L of KCl immunoprecipitation buffer. To detect specific interactions with TAFs, Western blotting analysis of the immunoprecipitated complexes was done by immunoblotting with TAF antibodies.

The protocols used have been previously described (7). Chromatin was incubated with 5 μg of anti-TAF4b (9) or any other antibody, as indicated overnight at 4°C. For Re-ChIP experiments, complexes were eluted after the first round of immunoprecipitation by 30 min of incubation at 37°C in 10 mmol/L of DTT. The eluted chromatin was diluted 20 times in sonication buffer and again subjected to chromatin immunoprecipitation procedures with the indicated antibodies. De–cross-linking of eluted chromatin was done by the addition of 200 mmol/L of NaCl plus RNase A overnight at 65°C. The remaining proteins were digested with Proteinase K for 2 h at 42°C. DNA fragments were recovered by phenol/chloroform extraction.

Cells on glass coverslips were fixed with 4% paraformaldehyde for 10 min at room temperature. Cell membranes were permeabilized with 0.1% Triton X-100 in PBS for 15 min. Blocking of cells was done with 5% fetal bovine serum in PBS for 1 h at room temperature. The antibodies and dilutions used are listed in the Supplementary Data. Nuclei were stained with Hoechst 33258 (Sigma-Aldrich). The slides were viewed with a 40× objective, Leica TCS SPE confocal laser scanning microscope (Leica Leisertechnik). The objective lens used was 63×. LAS AF software was used for image acquisition.

Plasmid DNA was transfected into cells by the calcium phosphate method (30). For short interfering RNA (siRNA) treatment, cells were transfected with human TAF4b siRNA, c-Jun siRNA, TAF4 siRNA, or with siControl using the protocols of the manufacturer (Dharmacon). After transfection (48 and 72 h), cells were harvested and extraction of proteins and RNA was done.

To assess luciferase activity, the Promega Dual Luciferase Reporter Assay (Promega Corporation) was used according to the instructions of the manufacturer. Luminescence was measured using a Tecan Safire fluorescence plate reader (Tecan Group, Ltd.). The reporter plasmid used was 5xcoll-TRE-tata-luciferase (31), whereas the expression plasmids included TAF4b (9), c-Jun (32), and TAF4 (33).

The assays were done on transwell plates (Corning Costar, Co.). Twenty-four hours after transfection with siRNA, treated cells (1 × 10⁴) were trypsinized and migration ability was measured as described previously (7). Cells were visualized and counted by bright-field microscopy with a Nikon Eclipse TE200 inverted fluorescence/phase contrast microscope equipped with a Sony charge-coupled device camera with a 40× objective. Data were obtained from two independent experiments, each repeated twice.

Nuclear extracts (8 μg) were incubated for 60 min at room temperature with a ³²P-labeled double-stranded probe ([γ-³²P]ATP; Perkin-Elmer) in the presence of 1.5 μg of poly(deoxyinosinic-deoxycytidylic acid; Sigma-Aldrich). The reaction was loaded onto native 6% polyacrylamide gels containing 0.5× Tris-borate EDTA buffer. The gel was prerun for 30 min at 120 V, run at 250 V for 2 h, dried for 1.5 h, and exposed for 12 h to a storage phosphor screen (Amersham Biosciences). Scanning was done with a Storm 860 scanner (Amersham Biosciences) and values were measured using ImageQuant software. For competition, nonlabeled probe (100 molar–fold excess of the labeled probe) was incubated with nuclear extracts for 30 min before the addition of the labeled probe. Double-stranded oligos were blunt-ended (sequences of oligos in Supplementary Data).

Forty-eight hours after the transfection of cells with TAF4b siRNA or siRNA control, cells were harvested, washed with PBS, and 1 μg of integrin α6 antibody (Santa Cruz Biotechnology) was added to 1 × 10⁶ cells and kept on ice for 30 min. After washing twice with PBS, cells were incubated for 30 min in secondary antibody (anti-mouse Alexa-488) and again washed twice with PBS before analysis using FACScan CANTO II (Becton Dickinson).

Forty-eight hours after TAF4b and TAF4 siRNA treatment, cells in 12-well plates were fixed with methanol, stained with 0.5% crystal violet, and washed with PBS. Stained cells were extracted using 30% acetic acid. Absorbance was measured at 595 nm.

Data are represented throughout the text with ±SD error bars. Statistical significance was tested with unpaired Student's t test.

To unveil any interactions between Jun family members and human TAFs possibly playing a role in carcinogenesis, immunoprecipitations with anti–c-Jun, JunB, and JunD antibodies were done using nuclear extracts from the human colon adenocarcinoma cell line HT29. Among the interactions identified (Table 1) by Western blot analysis, a specific interaction between c-Jun and TAF4b was observed in the c-Jun immunoprecipitate, the specificity of which was tested by siRNA against c-Jun and subsequent immunoprecipitation with anti–c-Jun antibody (Fig. 1A). Verifying this interaction, in the inverse situation, c-Jun was detected in the TAF4b immunoprecipitate (Fig. 1A, bottom). We focused on this interaction as they have been shown to influence each other during the induction of specific transcription programs (13). To identify a potential role in cancer-related mechanisms, the same experiment was done using nuclear extracts from parental Caco-2 and Caco-H cells with EMT characteristics. The same interaction was confirmed and was slightly enhanced in Caco-H cells (Fig. 1B). Together, these experiments not only show a specific interaction between TAF4b and c-Jun but also stress out the possible significance of this interaction in EMT phenomena.

To further explore the two components involved in the interaction and identify any possible role in cancer-related mechanisms, we first analyzed the expression levels of the two proteins in Caco-2 and Caco-H cells. The levels of TAF4b in the total extracts showed small differences between the cell lines tested, whereas, as expected, c-Jun revealed an overexpression pattern in all cell lines as compared with Caco-2 (Fig. 2A). The levels of EMT markers, E-cadherin and vimentin, were also evaluated and showed that these cell lines were at different stages of cancer progression. Overexpression of the Ha-RAS protein in Caco-H cell lines (Caco-H1 and Caco-H2) was confirmed. In parallel, a quantification of the mRNA levels of c-Jun showed that they were highly increased in both Caco-H cell lines as compared with Caco-2, whereas the mRNA levels of TAF4b, in agreement with its protein expression pattern, were slightly increased in Caco-H cells compared with the increase of c-Jun in the same cell lines (Fig. 2B).

In an attempt to elucidate the specific role of TAF4b in cancer, we reduced the levels of TAF4b and its homologue TAF4 by using specific siRNA either against TAF4b or TAF4. Seventy-two hours after transfection, the reduction of the mRNA levels of TAF4b and TAF4 was measured by PCR analysis (Fig. 3A, left). Interestingly, the cells subjected to TAF4b siRNA treatment, but not those transfected with TAF4 siRNAs, acquired up to a 76% differentiation in appearance, forming prolonged cell protrusions (Fig. 3A, right and bottom). In both cases, cell proliferation was unaffected (Fig. 3B). Because this morphologic change was predominant in cells with EMT, we investigated the role of TAF4b in related cell characteristics. Interestingly, the migration ability of Caco-H cells in response to a chemoattractant gradient was increased by 80% after TAF4b siRNA treatment (Fig. 3C). Similarly, Caco-2 and colon cancer HCT116 cells with partial EMT characteristics in which c-Jun and TAF4n were also detected to interact (Supplementary Data) showed increased migration abilities of 50% and 30%, respectively, as compared with control cells or cells treated with siControl. Together, these results indicate a negative role for TAF4b in cell migration and metastasis-related phenomena.

Based on the proposed preferential regulation of AP-1 site–containing genes by TAF4b and its implication in cell motility, we focused our study on the search for a TAF4b-regulated gene playing a role in cell migration while bearing an AP-1 site in its promoter. Changes in the mRNA levels of some metastatic/EMT markers after knockdown of TAF4b by siRNA were analyzed revealing increased levels of only some genes (e.g., MMP-2 and integrin α2; Supplementary Data). Interestingly, a gene important for cell migration—also bearing a putative AP-1 site in its promoter (34), i.e., integrin α6, after TAF4b siRNA treatment in Caco-2, Caco-H, and HCT116 cells—showed a decrease in its mRNA levels (Fig. 4A), surmising a regulatory function of TAF4b on its transcriptional activation. Similarly, integrin α6 hemidesmosome partners integrin β4 (Fig. 4C) and integrin β1 (Fig. 4D) showed lower mRNA levels after TAF4b siRNA treatment.

Notably, integrin α6 showed higher mRNA levels in the EMT-like Caco-H cells, which have also shown a greater extent of TAF4b/c-Jun colocalization (Fig. 2C) and in the advanced colon cancer HCT116 cells (Fig. 4A, left) as compared with Caco-2 cells. Interestingly, integrin α6 also showed higher protein levels in Caco-H and HCT116 cells as compared with Caco-2 cells, and the highest overexpression rate as compared with integrin β4 and β1 in the same cell lines (Supplementary Data). Immunostaining of the integrin α6 protein confirmed the increased expression of integrin α6 in Caco-H cells, whereas visualizing its accumulation in cell protrusions (Fig. 4D, right). At the same time, underlining the specificity of TAF4b on the regulation of integrin α6, TAF4 siRNA treatment in the same cells did not affect the mRNA levels of integrin α6 or that of TAF4b (Supplementary Data). TAF4b-specific siRNA treatment, on the other hand, not only reduced the levels of integrin α6 but also induced a change in its localization. As in cells under siControl treatment, the integrin α6 protein was localized mainly on the cell surface of Caco-H cells. In cells treated with siRNA against TAF4b, integrin α6 was found localized throughout the cell (Fig. 4E). The same was shown with fluorescence-activated cell sorting analysis comparing the cell surface expression of integrin α6 in cells treated with siControl and cells treated with siRNA against TAF4b (Fig. 4F).

The role of integrin α6 in cell migration properties was verified in our cell system, as a reduction of integrin α6 levels through siRNA resulted in an increase of the migration ability of Caco-2 (an increase of 50%), Caco-H (70%), and HCT116 (90%) cells, as compared with cells treated with scrambled siRNA (siC; Fig. 5A). Investigating a possible coinvolvement of c-Jun in the regulation of integrin α6, we reduced c-Jun levels by siRNA and observed a decrease of its expression in the cell lines tested, similarly to TAF4b (Fig. 5B). In addition, fluorescence-activated cell sorting analysis showed that, similar to TAF4b (Fig. 5C), c-Jun changed the localization of integrin α6 in the cell surface after treatment of cells with c-Jun–specific siRNA. This data indicates that integrin α6, β1, and β4 are transcriptionally regulated by TAF4b, whereas a coinvolvement of AP-1 family member and TAF4b-interacting partner c-Jun in the transcriptional regulation of integrin α6 has been proposed and will be further studied.

To understand the importance of an AP-1 site for gene regulation by TAF4b, a luciferase reporter construct controlled by the AP-1 site (5xcoll-TRE-tata-luciferase) of the Collagenase gene promoter was activated after c-Jun and TAF4b overexpression in Caco-2 cells. On the other hand, an overexpression of TAF4 did not produce the same effects, suggesting a specific involvement of TAF4b in AP-1–binding complexes (Fig. 6A, left). Overexpression of the transfected genes was confirmed by Western blotting (Fig. 6A, right).

To investigate the c-Jun/TAF4b interaction and elucidate its possible effect on AP-1 site–containing promoters in vivo, ChIP was done with primers encompassing the AP-1 site of the integrin α6 promoter schematically represented in Fig. 6B. Interestingly, both c-Jun and TAF4b were found on the promoter of integrin α6 in Caco-2 and Caco-H cells whereas, surprisingly in HCT116 cells, only TAF4b was bound (Fig. 6C and D). Re-ChIP analysis of the c-Jun immunoprecipitate revealed that c-Jun and TAF4b simultaneously co-occupied the promoter of integrin α6 in Caco-2 as well as in Caco-H cells (Fig. 6E). In summary, TAF4b was found on the regulatory AP-1 site of the integrin α6 promoter whereas its interplay with c-Jun on this particular site, and only under a specific cellular context, suggests a differential regulation pattern.

To understand the importance of the cell context on the regulation pattern of integrin α6 by c-Jun and TAF4b, we knocked down c-Jun by siRNA in Caco-2 and Caco-H cells (Fig. 7A), and followed the integrin α6 promoter occupancy by TAF4b with ChIP analysis. Interestingly, the binding of TAF4b on this promoter was dependent on the binding of c-Jun on the AP-1 site in Caco-H cells; the anti-TAF4b antibody could not precipitate promoter chromatin when c-Jun was knocked down (Fig. 7A, bottom). On the contrary, in Caco-2 cells, TAF4b remained bound on the promoter even after the reduction of c-Jun (siRNA c-Jun) protein levels. In the inverse situation in which TAF4b levels were reduced by siRNA (Fig. 7B, top), ChIP experiments showed that the binding of c-Jun on the promoter of integrin α6 was independent of the presence (siControl) or absence (siTAF4b) of TAF4b (bottom). This enforces the assumption that c-Jun controls the binding of TAF4b on the promoter of integrin α6 in Caco-H cells with an EMT-like phenotype, whereas in Caco-2, this binding is independent of c-Jun. Therefore, a partial, cell type–specific c-Jun dependency of TAF4b-binding on this specific promoter is proposed.

To further clarify the AP-1 composition on the promoter of integrin α6 in relation with TAF4b, we did Re-ChIP experiments in Caco-H cells. Interestingly, the integrin α6 promoter was found to be occupied by FRA-2 and c-Fos in addition to c-Jun and TAF4b (Fig. 8A, i), indicating that c-Jun might dimerize with any of these proteins while in complex with TAF4b. To clarify the specificity of the complex formation between TAF4b and AP-1 family members on specific promoters regulated by TAF4b, we followed with the same experiment on another promoter, that of vimentin. Vimentin was not found to be regulated by TAF4b in our system (Fig. 4E), and interestingly, TAF4b was also not found on its promoter together with any AP-1 family member (Fig. 8A, ii).

On the other hand, in the Caco-2 cell line, in which TAF4b exhibited a c-Jun–independent binding, only FRA-1 was detected on the integrin α6 promoter after ChIP (Fig. 8B, top). Re-ChIP analysis (Fig. 8B, bottom) showed that in Caco-2 cells, in addition to c-Jun, FRA-1 was also present on the promoter. In the case of HCT116 cells, in which a c-Jun–independent binding of TAF4b was observed, FRA-1, ATF2, and c-Fos were identified on the promoter (Fig. 8C, top). Re-ChIP experiments further showed that TAF4b was able to interfere with FRA-1, ATF2, and c-Fos, but not FRA-2 (Fig. 8C, bottom). Because c-Jun was not detected within the TAF4b/AP-1 family complex formed in HCT116 cells, we examined whether it could be substituted by another Jun family member on the integrin α6 promoter. Re-ChIP analysis showed that in HCT116 cells, JunB was detected in TAF4b-immunoprecipitated chromatin whereas in Caco-2 and Caco-H cells, neither JunB nor JunD were present (Fig. 9A, top). Interestingly, JunB was not able to interact with TAF4b in the same cell line (Fig. 9A, bottom), indicating that the presence of both proteins, TAF4b and JunB, on the promoter of integrin α6 was promoter DNA–dependent. JunB siRNA treatment in HCT116 cells showed decreased levels of integrin α6 expression, whereas in Caco-H cells, under the same conditions, the expression of integrin α6 presented no significant changes (Fig. 9B). Therefore, to determine a possible JunB-dependence in the binding of TAF4b on the integrin α6-promoter in this particular cell line, HCT116, we followed with ChIP experiments after the knockdown of JunB by specific siRNA treatment (Fig. 9C, top). As a control, the same experiment was done in Caco-H cells in which JunB was not found on the promoter of integrin α6 together with TAF4b. Interestingly, the integrin α6 promoter occupancy in HCT116 cells by TAF4b showed a dependency on the presence of JunB because the anti-TAF4b antibody could not precipitate promoter chromatin after the exclusion of JunB (Fig. 9C, bottom). On the contrary, in Caco-H cells, TAF4b remained on the promoter unaffected by the reduced JunB protein levels from siRNA.

To assess both the preferential binding of specific AP-1 family members on the promoter of integrin α6 and their cell type–specific interplay with TAF4b, we analyzed the expression levels of particular AP-1 family members in total extracts of Caco-2, Caco-H, and HCT116 cells. As judged by Western blot analysis (Fig. 9D), the levels of these proteins were cell type–dependent. Interestingly, the expression levels of AP-1 factors c-Jun, FRA-2, and ATF2 correlated with the occupancy of integrin α6 promoter; whereas some of them, e.g., FRA-1, did not. The formation of complexes between AP-1 family members, together with TAF4b, might not be explained solely by the respective expression levels of each AP-1 member.

Validating the proposed preferential binding of Jun family members, together with TAF4b on the promoter of integrin α6, we analyzed the actual binding of protein complexes on this exact DNA segment (AP-1 site) by gel shift assays using nuclear extracts from Caco-2, Caco-H, and HCT116 cells. Indeed, the presence of the integrin α6 probe induced the specific binding of a protein complex in all cell lines (Fig. 10; lanes 5, 6, and 7). Interestingly, the DNA-binding activity in each cell line was relevant to the expression levels of integrin α6 in the respective cell line, as shown in Fig. 5C. Caco-H and HCT116 cell extracts (lanes 6 and 7) showed stronger protein complex signals as compared with Caco-2 (lane 5). To exclude any off-target effects, we added an excess of nonradiolabeled integrin α6 probe (lane 4), which almost abolished the protein complex formation on the AP-1 site of integrin α6, whereas with other nonspecific competitors (lanes 2 and 3), the complex formations on these sites remained intact. In agreement, a nonspecific probe (cold ETS 1.3) did not affect the intensity of the signal (lane 2) whereas a cold vimentin probe (Vim) containing an AP-1 site competed with the integrin α6 probe (lane 3). Moreover, an integrin α6–mutated (iα6 mut, lane 8) probe, bearing three point mutations, failed to build the same protein complex as the wild-type AP-1 site of integrin α6. These results indeed validate the protein complex formation around the AP-1 binding site of the integrin α6 promoter.

For many years, it had been proposed that TAFs were transmitters of information between activators and the core transcriptional machinery. Notably, it has been shown that individual TAFs are required for the expression of only a specific subset of genes (35-38), and that the TFIID, or any of its other forms, was recruited in core promoters by a direct interaction between TAFs and their gene-specific activators (39). Given that TAF4b target promoter selectivity is enhanced by activators such as c-Jun and Sp-1 (12), we propose a synergistic function of the two factors, c-Jun and TAF4b, through their interaction in which TAF4b is the coactivator of AP-1 during regulation of AP-1 target genes (Fig. 11). Importantly, because among all the human TAFs, only TAF4b was found to interact with c-Jun, we cannot exclude the possibility of this interaction taking place independently of the TFIID complex.

Interestingly, TAF12, which heterodimerizes with TAF4 or TAF4b, has also been shown to directly interact with AP-1 family member ATF7, controlling transcription (40). In the present study, we suggest that TAF4b functions as a coactivator of integrin α6 transcription in concert with activator c-Jun or, depending on the cell type, with other AP-1 family members. We propose that TAF4b together with AP-1 factors induces preinitiation complex formation and transcriptional regulation of certain genes containing an AP-1 binding site (Fig. 8).

The coactivator domain of TAF4 and TAF4b is not conserved, leading to the hypothesis of an independent sequence-specific coactivator function. Our analysis, using luciferase reporter constructs transfected to colorectal cancer cells, supports that TAF4b, but not TAF4, acts as a coactivator on specific AP-1 targets. Furthermore, in our study, we observed that an intracellular reduction of TAF4b might be compensated by induced expression levels of TAF4 (Supplementary Data), reinforcing the previous assumption that TAF4 was limiting the TAF4b-containing TFIID complexes present in cells (36). Interestingly, TAF4 and TAF4b, through their competitive equilibrium in TFIID, have been shown to regulate the expression of genes involved in cell proliferation and transforming growth factor-β signaling, known to drive tumor cells to EMT transformation (41, 42). This underlines the importance of further analyzing the gene- and signal-specific coactivator functions of both factors.

Based on our observations, the modulation of AP-1 composition in response to external signals, possibly by exchanging the most abundant AP-1 family member in each cell type, serves as a regulatory mechanism triggering the expression of specific genes. Indeed, we have shown that in the integrin α6 promoter, c-Jun is not the only Jun family member occupying the AP-1 site together with TAF4b but might as well be replaced by JunB, for example, within HCT116 cells. Other AP-1 family members were found on this site as well, probably building dimers in the absence or presence of c-Jun and in a cell type–dependent mechanism, to some extent, in an expression level–proportional manner (Fig. 8). Because the protein levels of transcription factors do not perfectly correlate with the pattern of occupancy of the integrin α6 promoter, other mechanisms that fine-tune the affinity of particular AP-1 family members on the cognate cis element could be considered. For instance, the interaction with TAF4b or the existence of particular posttranslational modifications could result in DNA-affinity alteration. In this vein, it is well known that the mutational profile of each cell line influences signaling pathways such as JNK and mitogen-activated protein kinase, which are crucial for AP-1 regulation. Notably, the HCT116 cell line bears, among others, endogenous mutated forms of K-RAS and PIK3CA oncogenes, whereas the Caco-H cell line overexpresses a mutated Ha-RAS oncogene leading to high levels of JNK activity (19). Our study illustrates the significance of screening different cell types as a way to link genetic alterations to particular patterns of gene regulation.

Jun proteins have diverse expression profiles and biological functions, even though they share high sequence homologies. Although, c-Jun and JunB may induce opposite effects (43), they also have gene targets in common (44). Interestingly, JunB has been suggested to substitute c-Jun during mouse development and cell proliferation when c-Jun is depleted (45), further supporting our suggestion that AP-1 family members take their promoter-specific interaction with TAF4b into their own hands and follow with regulation of transcription.

The involvement of other TAFs in cancer and/or metastasis has already been brought into light (7, 46), however, this is the first report linking TAF4b itself with the regulation of tumor cell migration. Herewith, we provide evidence that TAF4b is a suppressor of cell migration regulating, in concert with c-Jun, the expression of genes playing a pivotal role in EMT. Accordingly, in a microarray analysis done in granulosa cells stably overexpressing TAF4b, a number of AP-1 target genes were upregulated (13), including the EMT marker vimentin. In our hands, the expression of this gene was unaffected by the alteration of TAF4b levels in Caco-H cells, possibly due to an already highly activated state of this promoter in the EMT cells. Indeed, in these particular cells, in contrast with the epithelial Caco-2 cells, the vimentin gene was found to be predominantly regulated by FRA-1 (19), which did not interact with TAF4b in our experiments, suggesting a cell type–specific and TAF4b-independent mechanism regulated by AP-1 family member selectivity. On the other hand, TAF4b was shown to regulate integrin α6 in different cell lines: a colon adenocarcinoma cell line, its derivative Caco-H cell line constitutively overexpressing Ha-RAS, and in parallel, in an established colon cell line that has gained some EMT characteristics. Interestingly, in another study (36), integrin α6 was found to be upregulated in mouse fibroblasts after knockdown of TAF4 and its proven replacement by TAF4b.

The integrin subunits α6 and β4 together form the hemidesmosomal α6β4 integrin, a transmembrane receptor known to play a critical role in a number of cellular functions including cell migration and differentiation (47). Both integrins have been found elevated in several types of carcinomas, whereas increased expression levels correlate with invasive phenotypes and EMT (48, 49). Integrin α6 is a controversial member of the integrin family because it has been shown to have a dual role depending on the mutational oncogenic profile of the tumor tissue and its stage (50). Interestingly, in human mammary cancer and in mouse keratinocytes, reduced α6 expression has been linked to increased cell migration and metastasis (51, 52). On the onset of cell migration, cells must detach from the basal membrane, in part by reducing the expression of integrin α6, which is responsible for detachment and migration. On the other hand, when gaining a more mesenchymal phenotype, integrin α6 expression must again be upregulated to invade the basal membrane. Nevertheless, under conditions of induced migration ability (e.g., after TAF4b knockdown), integrin α6 expression must be downregulated so that migration could begin. Accordingly, integrin α6 expression was reduced in tumors formed after the injection of Caco-H cells in severe combined immunodeficiency mice, in comparison with cultured cells, supporting its role in invasion mechanisms (28).

In conclusion, we report a cell type–specific initiation of transcription controlled by TAF4b together with variant AP-1 family members in a promoter-dependent manner, emphasizing the importance of deciphering the independent roles of specific TFIID subunits as cofactors that govern transcription.

No potential conflicts of interest were disclosed.

Grant Support: Marie Curie Fellowship and EU Marie Curie Research Training Network “TAF-Chromatin” grant MRTN-CT-2004 504228 (L. Tora, A. Pintzas, and R. Dikstein).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.