Independent Repressor Domains in ZEB Regulate Muscle and T-Cell Differentiation

Cell culture.

The HT1080 fibrosarcoma (American Type Culture Collection [ATCC], Rockville, Md.) and C33A cervical carcinoma cell lines (ATCC) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Life Technologies) containing 5% fetal calf serum and 5% calf serum (Life Technologies). C3H10T1/2 fibroblasts (ATCC) were grown in DMEM containing 13% fetal calf serum. The T-cell line Jurkat (ATCC) was grown in RPMI 1640 supplemented with 10% FCS. 293T cells were obtained from S. J. Korsmeyer (Dana Farber Cancer Institute, Boston, Mass.). CV1-MLP-CAT cells were obtained by stably cotransfecting CV1 cells (ATCC) with a chloramphenicol acetyltransferase (CAT) reporter driven by the major late promoter (MLP) and a neomycin-resistant vector.

Plasmid construction.

Gal4 and LexA fusion proteins of different ZEB fragments were obtained by PCR of the corresponding regions and in-frame cloning of the product in the BamHI-XbaI site of PM1 and PBXL3, respectively. RD (repressor domain)-ZEB refers to the region between the two zinc finger clusters (from nucleotides 906 to 2706). Region 1 comprises the cDNA between nucleotides 906 and 1626; region 2 comprises the cDNA between nucleotides 1626 and 2280; and region 3 comprises the cDNA between nucleotides 2280 and 2706. Out-of-frame GAL4 and LexA constructs for RD-ZEB were obtained by out-of-frame insertion of RD-ZEB in PM2 and PBXL3 and used as controls (data not shown). Most Gal4 activators were either previously described (67) (Gal4-CTF, Gal4-VP16, Gal-Sp1, Gal4-MEF2C [amino acids 1 to 465], Gal4-NFκB-p65, Gal4-PU.1, Gal4–E2F-1 [amino acids 285 to 437], Gal4–c-fos [amino acids 21 to 380], Gal4–ITF-1 [amino acids 1 to 427], and Gal4–TFE-III [amino acids 2 to 216]) or obtained from the following investigators: Gal4–c-myb–CD (J. Lipsick, Stanford University, Stanford, Calif.), Gal4-myoD (amino acids 1 to 318) (G. Tomaselli, Johns Hopkins University, Baltimore, Md.), and Gal4–c-jun (C. Caelles, University of Barcelona, Barcelona, Spain).

pETS contains three ets sites from the α₄ gene promoter as previously described (50). pETS-ZEB contains two ZEB binding sites (CACCTG) upstream of pETS; pETS-ZEB-MYB contains four myb binding sites upstream of pETS-ZEB. pETSmut-ZEB-MYB is as pETS-ZEB-MYB but with the ets sites mutated. Details on these constructs have been previously described (reference 50 and references therein). The transcriptional elements included in the deletional constructs 76α4-CAT, 400α4-CAT, and 2.0α4-CAT as well as details on their construction were previously reported (reference 50 and references therein). An expression vector for c-myb (pSV–c-myb) was obtained from B. Calabretta (Jefferson University, Philadelphia, Pa.).

Fusion proteins between the DNA binding domain of ZEB and the different regions in RD-ZEB were designed as follows. The C-terminal zinc fingers of ZEB (DB-ZEB) were cloned in the MluI-XbaI sites of pCI-neo (Promega Corp. Madison, Wis.). Annealed oligonucleotides of a Kozak sequence, an ATG codon, and a FLAG sequence (GCC ACC ATG GAC TAC AAG GAC GAC GAT GAC AAG) were cloned upstream of and in frame with DB-ZEB in the XhoI-MluI sites. PCR fragments of regions 1, 2, and 3 were then cloned in frame with DB-ZEB in the XbaI-NotI sites of pCI-neo. A simian virus 40 nuclear localization signal and an in-frame stop codon are immediately downstream of the cDNA sequence (CAA ATG GAA CCT AAG AAG AAG AGG AAG GTT TAA).

pGL contains six LexA sites 30 bp upstream of two (or five) Gal4 sites and was described previously (67). A new version of pGL (where the LexA sites are 1.7 kb upstream of the Gal4 sites) was constructed by cloning a BglII-HindIII 1.7-kb fragment corresponding to the Drosophila cDNA Nautilus (cloned into the EcoRI site of pSP73 [Promega Corp.] and obtained from A. Michelson, Brigham and Womens Hospital, Boston, Mass.) into the BglII-HindIII site of pGL.

CMVp300 was obtained from Dr. D. Livingston (Dana-Farber Cancer Institute, Boston, Mass.). Gal4 p300 1737–2414 was obtained from A. Giordano (Jefferson University). CMV-E1A-12S and E1A-12SΔ2–36 were obtained from E. Harlow (Massachusetts General Hospital, Charlestown, Mass.).

A CAT reporter driven by the MLP was obtained from D. E. Ayer (University of Utah, Salk Lake City, Utah).

An expression vector for mitogen-activated protein (MAP) p38 kinase (pSRα2-p38) was obtained from S. Chellappan (College of Physicians and Surgeons, Columbia University, New York, N.Y.).

Transient transfections and CAT assays.

HT1080 cells were transfected by the calcium phosphate method, and Jurkat cells were transfected by electroporation as previously described (49, 50). After 48 h, lysates were collected and the transfection efficiency was corrected as previously described by cotransfection of a thymidine kinase-driven luciferase vector (49, 50). CAT assays were performed as described previously (49). DNA was brought to a total of 6 μg for 60-mm dishes (or to 20 μg for 100-mm dishes) with pBluescript (Stratagene, La Jolla, Calif.).

For transfection of CV1-MLP-CAT cells, cells were plated into 150-mm dishes, cotransfected with 60 μg of Gal4-ZEB constructs and 5 μg of puro-BABE (a puromycin-resistant vector), and maintained in growth medium supplemented with 800 μg of G418 (Life Technologies) and 1.5 μg of puromycin for 4 days to both select transfected cells and assess newly synthesized CAT activity. In the last 48 h, 300 nM trichostatin A (TSA) (Wako Chemicals USA, Inc., Richmond, Va.) was added to the medium.

CAT results are means of duplicate assays and are all representative of at least five separate experiments with standard deviations below 15%.

Myogenic conversion assays and immunostaining.

For 10T1/2 cell myogenic conversion assays, 10T1/2 cells were transfected by the Lipofectamine method in Optimem medium (Life Technologies). After 12 h, medium was removed and replaced with differentiation medium (2% horse serum–DMEM). After 4 to 5 days, the cells were fixed in methanol and stained with anti-myosin heavy chain MF-20 (R. Kopan, Washington University, St. Louis, Mo.) monoclonal antibody as previously described (49). After being washed to remove unbound antibody, the cultures were incubated sequentially with horseradish peroxidase-conjugated anti-mouse IgG antibody (Jackson ImmunoResearch, West Grove, Pa.) and with diaminobenzidine substrate (Vector Laboratories, Burlingame, Calif.). Nuclei were counterstained with hematoxylin (Zymed, South San Francisco, Calif.).

Western blot analysis.

At 48 h after transfection with the appropriate plasmids, C33a cells were lysed in ELB (150 mM NaCl, 50 mM HEPES [pH 7.0], 5 mM EDTA, 0.1% Nonidet P-40), sonicated briefly, and centrifuged to remove nuclear debris as described previously (51). The lysates were then loaded onto a sodium dodecyl sulfate–4 to 15% polyacrylamide gradient gel (Bio-Rad, Hercules, Calif.) and transferred overnight to a polyvinylidene difluoride membrane (Immobilon P; Millipore Corp., Bedford, Mass.) in 10 mM CAPS [3(cyclohexylamino) 1-propanesulfonic acid] buffer (pH 11). The membrane was serially incubated with anti-Flag polyclonal antibody (Santa Cruz Biotechnologies, Santa Cruz, Calif.) and anti-rabbit IgG-horseradish peroxidase secondary antibody. After several washes, the membrane was then developed by the chemiluminescence method (Renaissance; NEN Life Sciences Products, Boston, Mass.) as specified by the manufacturer.

ZEB is a selective transcriptional repressor.

In previous studies, we found that the central region of ZEB, between the N-terminal and C-terminal zinc finger DNA binding domains, contains the repressor domain (RD) (schematized below in Fig. 2A) (49). We found that this RD could function independently to repress transcription when fused to the DNA binding domain from the yeast transcription factor Gal4 (49). In this study, we asked which transcription factors are repressed by ZEB. We found that ZEB blocked the activity of some transcription factors but not others (Fig. 1). It is of note that the set of transcription factors blocked by ZEB includes factors that are important for both myogenesis (MEF2C) and lymphoid differentiation (i.e., c-myb, ets family members, TFE-III, and the NF-κB protein p65).

ZEB contains two independent repressor domains.

To further localize sequences important for repressor activity, we divided the RD into three parts (Fig. 2A). Region 1 includes sequences N-terminal of the central homeodomain; region 2 contains the homeodomain, and region 3 contains the C-terminal region of the RD. Region 2 did not show any repressor activity in these assays (Fig. 2B). We found that region 1 was sufficient to repress all of the transcription factors repressed by RD, as shown in Fig. 1, with the exception of the myogenic factor, MEF2C (Fig. 2). Factors repressed by region 1 include hematopoietic factors (e.g., c-myb, ets family members, and TFE-III) as well as several more general and ubiquitous factors (Fig. 2B and data not shown). Interestingly, region 3 blocked the activity of MEF2C but had no effect on the activity of any other transcription factor we tested (Fig. 2B and data not shown). These results indicate that ZEB has two independent repressor domains, which target different subsets of transcription factors.

Recent reports indicate that in Drosophila, transcriptional repressors fall into two categories according to their ability to repress at short (<50 bp) or long (>300 bp) distances (22, 23). This distinction between long- and short-range repressors is thought to be important in establishing patterns of gene expression in Drosophila (21–23). Short-range repressors permit enhancer autonomy in modular promoters—repressors would affect the activity only of enhancers nearby. Long-range repressors function to block expression without regard to promoter organization. We tested the two repressor regions of RD-ZEB to determine whether they act as long- or short-range repressors (Fig. 3A). We constructed reporter plasmids in which the distance between the transcriptional activator binding site and the ZEB binding site was either 1.7 kb (long range) or 30 bp (short range). Region 1 repressed only at short range, whereas region 3 was equally active at both short and long ranges (Fig. 3A). These results provide additional evidence that regions 1 and 3 repress through distinct mechanisms.

Region 1 of the ZEB RD regulates the activity of the hematopoietic transcription factors ets and c-myb.

c-myb is essential for normal hematopoietic differentiation (45). The ets family contains a number of members, some of which are hematopoiesis specific and some of which are more widely acting (2). Mutations in hematopoietic ets genes have demonstrated that these proteins, as with c-myb, are essential for hematopoietic differentiation (55, 63). A number of hematopoietic genes contain binding sites for both c-myb and ets, and it has been demonstrated that the two factors synergize to activate transcription (40). In fact, the E26 virus encodes a v-myb–v-ets fusion protein, and it has been shown that the synergy between these two factors is essential in the generation of erythroleukemias in animal models (41, 42).

ZEB sites are found in a number of c-myb- and ets-dependent genes including those encoding α₄-integrin, p56^lck, and CD4, where they function as silencers (40, 50, 54). ZEB blocks the activity of c-myb or ets alone, but together these factors synergize to resist repression by ZEB (50). This arrangement may have two consequences: (i) it would ensure that some c-myb- and ets-dependent genes are not expressed until both factors appear during hematopoietic differentiation, and (ii) since ets factors contain a common binding domain and some are ubiquitous, ZEB would also ensure that c-myb- and ets-dependent genes are not ectopically expressed in nonhematopoietic cells. This is illustrated in Fig. 3B, where the activity of ets sites located in bp 1 to −76 of the α₄-integrin gene promoter is repressed in nonhematopoietic cells (lacking c-myb expression) by the activity of ZEB sites located at bp −361 and −399 (50 and Figure 3B). Expression of c-myb is able to restore the activity of the α4 promoter. The c-myb sites in the α₄ gene promoter are also sensitive to ZEB repression when the ets sites are not present or are mutated (see below and data not shown).

We wondered whether region 1 of the ZEB repressor domain is sufficient to recapitulate the activity of ZEB toward c-myb and ets. The results shown in Fig. 3C indicated that region 1 inhibits the activity of c-myb and ets products individually. To demonstrate this regulation in the context of a more normal promoter setting than that shown in Fig. 2B, individual regions of RD-ZEB were fused to the DNA binding region of ZEB (C-terminal zinc fingers, which have the highest DNA binding affinity [56]). Expression of these C-terminal zinc fingers (DB-ZEB) alone did not have any repressor activity and released repression by endogenous ZEB (50) (Fig. 3C). Neither, region 2 nor 3 (fused to DB-ZEB) had any effect on c-myb or ets activity. In contrast, region 1 efficiently blocked the activity of c-myb and ets sites individually. Also, as with full-length ZEB, region 1 was not able to repress the combination of c-myb and ets sites. The level of expression of these fusion proteins is shown below in Fig. 5C. These results suggest that region 1 is responsible for repression of α₄-integrin (and probably other hematopoietic genes) by ZEB in lymphoid cells and thus mediates its role in lymphoid differentiation and function.

Region 1 of ZEB represses transcription by a mechanism involving histone deacetylase activity.

p300/CREB binding protein (CBP) interacts with a number of transcription factors and activates transcription by acetylation of histones and disruption of the nucleosome structure (reviewed in reference 38). While p300/CBP itself has histone acetyltransferase (HAT) activity on its own, the C-terminal region of the protein interacts with PCAF, which contains HAT activity essential for p300/CBP function (47). p300/CBP was initially discovered by its interaction with the adenovirus protein E1A (62). The amino-terminal region of E1A (amino acids 2 to 36) also binds to the C-terminal region of p300/CBP and displaces PCAF (62, 72). In addition, E1A can directly block p300/CBP HAT activity (9).

Factors repressed by ZEB (including c-myb, ets, and MEF2C) interact with p300/CBP and synergize with it in transcriptional activation (references 14, 70, and 71) and unpublished results. That prompted us to investigate whether ZEB may repress transcription by a mechanism involving p300/CBP. Inclusion of ZEB binding sites upstream of the ets or c-myb sites blocked the activity of both transcription factors (Fig. 3C) (50). We reasoned that if region 1 of ZEB were acting to block p300/CBP activity, overexpression of p300 should overcome repression by ZEB. Indeed, we found that overexpression of p300 completely reversed the repression of c-myb or ets sites by ZEB in hematopoietic cells (Fig. 4A), further suggesting that region 1 of ZEB could be targeting the p300 co-activator. Similar rescue of ZEB repression was obtained for NF-κB (Fig. 3B). We also found that repression of MEF2C by region 3 was fully rescued by overexpressing p300 (Fig. 3B).

Next, we wanted to determine whether ZEB domains could block p300/CBP transcriptional activity. Since p300/CBP does not bind directly to DNA, we used a fusion protein where the C-terminal region of p300, which binds PCAF, was linked to the DNA binding domain of Gal4. This fusion protein is a potent transcriptional activator (74) (Fig. 4C). When region 1 of ZEB was targeted to the promoter, transcriptional activation by p300/CBP was blocked (Fig. 4C). Region 3 of ZEB (or the retinoblastoma protein, which is also a transcriptional repressor [67]) did not have significant effect on p300 activity (Fig. 4C and data not shown). This experiment suggested that although p300/CBP was able to overcome ZEB repression by both regions 1 and 3, only region 1 is actively repressing p300 transcriptional activity.

Since we were not able to detect interaction between ZEB and p300/CBP (data not shown), we wondered whether region 1 of ZEB may offset p300 activity by recruitment of histone deacetylase activity. To investigate this, we studied the ability of regions 1 and 3 of ZEB to inhibit the activity of the major late promoter (MLP). It has been shown that repression of this promoter by the Mad and retinoblastoma repressors depends completely upon recruitment of histone deacetylase activity (28, 37, 38). As shown in Fig. 4D, region 1 but not region 3 repressed the activity of the MLP reporter. Moreover, the histone deacetylase inhibitor trichostatin A rescued repression by region 1, further suggesting that this region of ZEB offsets p300 activity by a histone deacetylase mechanism.

Region 3 of RD-ZEB blocks myogenesis.

Expression of myoD is sufficient to trigger the myogenic pathway in a number of cell types (66). MyoD activates a cascade of transcription factors, including members of the MEF2 family, which collaborate with myoD to amplify the muscle differentiation program (44). A dominant negative form of MEF2 blocks vertebrate myogenesis (48), and somatic and cardiac myogenic differentiation is disrupted in Drosophila embryos with mef2 null mutations (6, 34), indicating that MEF2 function is essential in myogenesis from Drosophila to mammals. Negative regulation of muscle by ZEB is not due to ZEB binding E boxes and displacing myogenic factors (49). ZEB binds only a small subset of E boxes (and it also binds a subset of non-E-box sequences), and thus it cannot efficiently displace MRF proteins. Indeed, DB-ZEB has no effect on myogenesis (49). Instead, the entire molecule was necessary for inhibition of muscle differentiation (49); thus, ZEB is functioning as an active repressor when bound to promoters.

We wondered which region of RD-ZEB is important for inhibition of myogenesis. For these studies, we fused RB-ZEB and each of the three regions of the RD to the DNA binding domain of ZEB (these are the same constructs used in the lymphocyte experiments in Fig. 3C). Expression vectors for these constructs were then tested in myogenesis assays, where myoD was used to trigger myogenesis in 10T1/2 cells. Myogenic conversion was followed by expression of the muscle-differentiation specific myosin heavy chain (Fig. 5A). We found that the RD is sufficient to block myogenesis. While neither region 1 nor region 2 had any effect on myogenic differentiation, region 3 blocked it efficiently (Fig. 5A and B). Western blots of these ZEB constructs are shown in Figure 5C. Even though DB-ZEB had no effect in these assays, it was expressed at higher levels than the other constructs were. Note that even though region 1 had no effect in these assays, the same construct efficiently repressed transcription in lymphoid cells (where the region 3 construct was inactive) (Fig. 3B). Together, these results suggest that region 1 of RD-ZEB blocks selectively the activity of several hematopoietic factors and that it is responsible for the regulation of hematopoietic genes and therefore for lymphoid differentiation and function. In contrast, region 3 functions in muscle cells, where it appears to block myogenesis and specifically inhibits the activity of MEF2.

p300/CBP interacts and transcriptionally synergizes with MEF2C (52). The MAP kinase p38 is required for MEF2C activity, as shown by experiments examining mutants for phosphorylation sites on MEF2C or overexpression of a p38 dominant negative (25). Phosphorylation of cyclic AMP response element binding protein (CREB) by the calcium calmodulin CAMIV kinase is required for efficient recruitment of CBP/p300 by CREB (10, 66a). A similar mechanism could be postulated for MEF2C. We therefore investigated whether p38 kinase was able to rescue repression of MEF2C by region 3 of ZEB. Indeed, and as shown in Fig. 6, overexpression of p38 overcomes repression by region 3 of ZEB. We found that CAMIV kinase also overcomes repression by region 3 of ZEB (see Discussion) (data not shown). However, repression of c-myb by region 1 of ZEB remained unaffected by overexpression of p38 (or CAMIV kinase) (Fig. 6 and data not shown), further suggesting different mechanisms of repression for regions 1 and 3 of ZEB.