Transcriptional Activities of the Zinc Finger Protein Zac Are Differentially Controlled by DNA Binding

Abstract

Zac encodes a zinc finger protein that promotes apoptosis and cell cycle arrest and is maternally imprinted. Here, we show that Zac contains transactivation and repressor activities and that these transcriptional activities are differentially controlled by DNA binding. Zac transactivation mapped to two distinct domains. One of these contained multiple repeats of the peptide PLE, which behaved as an autonomous activation unit. More importantly, we identified two related high-affinity DNA-binding sites which were differentially bound by seven Zac C₂H₂ zinc fingers. Zac bound as a monomer through zinc fingers 6 and 7 to the palindromic DNA element to confer transactivation. In contrast, binding as a monomer to one half-site of the repeat element turned Zac into a repressor. Conversely, Zac dimerization at properly spaced direct and reverse repeat elements enabled transactivation, which strictly correlated with DNA-dependent and -independent contacts of key residues within the recognition helix of zinc finger 7. The later ones support specific functional connections between Zac DNA binding and transcriptional-regulatory surfaces. Both classes of DNA elements were identified in a new Zac target gene and confirmed that the zinc fingers communicate with the transactivation function. Together, our data demonstrate a role for Zac as a transcription factor in addition to its role as coactivator for nuclear receptors and p53.

Zac is a seven-zinc-finger protein which potently promotes apoptosis and cell cycle arrest upon expression in mesenchymal and epithelial cell lines. Conversely, ablation of Zac gene expression increased cell proliferation, further supporting Zac's antiproliferative role (19, 24). Concordant with these findings, the rat ortholog Lot1 was isolated from ovarian surface epithelial cells in a screen for genes showing reduced expression upon spontaneous transformation in vitro, hence the designation “lost on transformation.” Moreover, treatment with epidermal growth factor receptor ligands rapidly downregulated Lot1 expression in these cells. In this view, Lot1 may control mitogenic signal transduction pathways by counteracting cellular transformation (1, 2).

The human ortholog ZAC/LOT maps on chromosome 6q24, a region known to contain a tumor suppressor gene for several types of neoplasms, including breast cancer (28). We recently showed that ZAC is expressed in normal mammary and anterior pituitary epithelial cells and that its expression is downregulated in primary breast and pituitary tumors (4, 20). Treatment of some mammary cell lines with 2-deoxyazacytidine restored ZAC expression and suggested that ZAC may be inactivated by methylation of regulatory promoter sequences. The recent finding that ZAC/Zac is imprinted (see below) (3, 21) supports our observations and suggests an additional mechanism for inactivation of ZAC in tumors with loss of heterozygosity at 6q24-25, a condition observed in up to 80% of breast tumors.

Recent studies (7, 13) evidenced an additional role for ZAC in the etiology of transient neonatal diabetes mellitus, a rare form of childhood diabetes which usually resolves in the first 6 months of life and strongly predisposes to type 2 diabetes of adult onset (OMIM *601410; Online Mendelian Inheritance in Man; NCBI). Transient neonatal diabetes mellitus is associated with intrauterine growth retardation, dehydration, and lack of insulin due to overexpression of a maternally imprinted gene localized on 6q24 that interferes with pancreatic development and/or glucose homeostasis in general.

Patients with transient neonatal diabetes mellitus show paternal uniparental disomy of chromosome 6, paternal duplications of the critical region, or a defect in a methylation imprint on chromosome 6 (3, 7). This methylation imprint mapped to a CpG island within the promoter region of ZAC, and its methylated state silenced promoter activity (27). In fibroblasts from patients with transient neonatal diabetes mellitus, the monoallelic expression of ZAC is relaxed, providing strong supportive evidence that two unmethylated alleles of this locus are indeed associated with transient neonatal diabetes mellitus (16).

Taken together, accumulating evidence demonstrates a role for Zac/ZAC in the regulation of proliferation and differentiation, and altered expression most likely contributes to cancer and diabetes. These findings agree with the notion that imprinted genes are intricately involved in fetal development and underlie numerous human diseases (6).

Interestingly, Zac has recently again been isolated in a screen for proteins that bind the nuclear receptor coactivator GRIP1 (12). More importantly, Zac potently coactivated or corepressed the hormone-dependent activity of nuclear receptors, including androgen, estrogen, glucocorticoid, and thyroid hormone receptors, which are key regulators of cell growth and differentiation, homeostasis, and development in a cell-specific manner. In further support of Zac's role in coactivation, the authors additionally observed that Zac enhanced transactivation of p53 on synthetic and endogenous promoters and strongly reversed E6 inhibition of p53 (11). Similarly, Zac coactivated p53 transactivation of the Apaf-1 promoter (23) but in both studies was unable to confer transactivation by itself.

At odds with these findings, we previously suggested that Zac/ZAC DNA binding is indicative of a role as a transcription factor (5, 28). Therefore, it appeared mandatory for us to further elucidate Zac's status in transcriptional regulation. In particular, we reasoned that Zac's status as transcription factor will be instrumental to placing Zac/ZAC within the network of genes regulating cell proliferation, pancreatic development, and/or glucose metabolism.

MATERIALS AND METHODS

Cell culture and transfection.

Saos-2 cells were kept as described previously (24). For immunoblots, 100 ng of each Gal4-Zac fusion construct was transfected; results represent three to four independent experiments. Gal4-Zac (10 ng), adjusted amounts of deletion constructs, and Gal4-(PXX)_n plasmids (100 ng) were cotransfected with the reporter p(GAL4)5E1BLUC (0.5 μg) into 2 × 10⁶ cells. To prepare nuclear extracts (5), 50 ng of wild-type or mutated Zac cDNAs were transfected. Half-maximal transactivation of the palindromic and repeat Zac DNA-binding site, their derivatives, and the CK14 promoter motifs by wild-type and mutated Zac cDNAs was determined from dose-response curves obtained from three independent cotransfections.

Plasmids, GST pulldown, and in vitro translation assays.

Zac, synthetic reporter, and CK14 promoter constructs were generated and sequenced by standard methods; details are available on request. Oligonucleotides encoding Zac DNA-binding sites were inserted into the vector pGL2TATA at the SacI and KpnI sites. Restriction fragments used in electrophoretic mobility shift assays were released by digestion with XbaI (present as an internal site) and XhoI. Glutathione S-transferase (GST)-Zac fusion proteins were expressed from pGEX2TK and purified as described previously (28). Equivalent amounts of protein, as judged by Coomassie blue staining, were used. In vitro translations were done with the TNT kit (Promega). Results represent at least three independent experiments.

Electrophoretic mobility shift assays.

Conditions for electrophoretic mobility shift assays were as described previously (5). Additionally, Zac proteins in nuclear extracts were measured by immunoblotting, and aliquots were adjusted and incubated with 20,000 cpm of labeled probe. Results represent three to four independent experiments quantified in a scintillation counter.

Immunoblots and immunohistochemistry.

Western blots were done with anti-Zac (24) and anti-HA (Roche) antibodies. Immunostaining with anti-CK14 antibodies (BioGenex) was carried out as described previously (19).

Tetracycline-regulated Zac.

Stable cell clones for tetracycline-regulated expression of ZacΔZF were generated as described previously (24). Total RNAs were purified and hybridized with a cDNA expression macroarray according to the manufacturer's instructions (Clontech 7742). Reverse transcription-PCR experiments were done as reported before (4).

RESULTS

Zac transactivation and autonomous minidomains.

The Zac N-terminal zinc finger domain contains seven canonical zinc fingers of the C₂H₂ type (amino acids 1 to 208). This region is flanked by 65 amino acids (amino acids 209 to 274) without known protein motifs. The central part of Zac (amino acids 275 to 382) harbors 34 repeats of the motifs PLE, PMQ, PML, and PLQ. The C terminus (amino acids 383 to 667) is rich in single and clustered P and Q residues, which commonly occur in the transactivation domains of transcription factors. The extreme C terminus includes several PE repeats and E clusters (amino acids 583 to 654), which cause high acidity (29%). We refer here to these regions as the zinc finger, linker, proline repeat, and C-terminal domains (Fig. 1A).

FIG. 1.

Zac transactivation and autonomous minidomain. (A) Scheme of the Zac protein. Numbers indicate amino acid residues. ZF, zinc fingers (amino acids 1 to 208); L, linker (amino acids 209 to 274); PR, proline repeat (amino acids 275 to 382); C, C terminus (amino acids 383 to 667). (B) Zac confers transactivation in the Gal4 fusion system. Scheme of the fusion proteins; transactivation (TA) for concentrations of plasmids adjusted to the expression of G-Zac was set to 100%. (C) Representative immunoblot of fusion proteins for doses of 100 ng each. (D) Zac transactivation of cognate DNA-binding site depends on the linker and proline repeat domains. The reporter plasmid contains two tandem palindromic DNA elements. Activation by Zac (50 ng) was set to 100% and compared to adjusted doses of ZacΔL, ZacΔPR, and ZacΔLPR (inset). (E) Amino acid sequence of the proline repeat region. Amino acid triplets are 13 PLE (red), 15 PMQ (green), 4 PML (blue), and 1 PLQ (black). Residues 343 to 362 are absent in a variant of Zac (AF324471). (F) Amino acid triplets selectively transactivate. The indicated peptide motifs were serially fused in steps of four (PLE, PLK, PMQ, PLQ, and PPP) or two (LEPPLE) copies to Gal4 and tested for transactivation (TA). (G) Representative immunoblot of Gal4 alone (Gal) and the indicated Gal4 fusion proteins. The number of amino acid triplets is noted. Note the aberrant migration of the PLK polypeptides.

To investigate Zac transactivation potency, we fused different segments to the heterologous DNA-binding domain of Gal4 (Fig. 1B), which we additionally tagged at its C terminus with the HA epitope to compare expression of the corresponding proteins (Fig. 1C). Absence of the zinc finger domain (G-ZacΔZF) reduced protein levels fivefold compared to Zac (G-Zac), whereas the remaining constructs were similarly expressed. Note the aberrant migration of the proline repeat domain at a higher molecular weight than predicted. Doses of plasmids adjusted to obtain equal levels of protein expression were cotransfected with a luciferase reporter gene, which contained five GAL4 DNA-binding sites in front of the TATA element derived from the E1B promoter. Zac (G-Zac) moderately (21-fold) activated transactivation, whereas absence of the zinc finger domain (G-ZacΔZF) strongly improved transactivation (908-fold) (Fig. 1B). The entire C terminus (G-Zac-C) was inactive, which was also the case when shorter segments (G-Zac-1572/2004 and G-Zac-1141/1572) were separately expressed (data not shown). The isolated linker (G-Zac-L) and proline repeat region (G-Zac-PR) each weakly activated (7- and 11-fold, respectively), while joint expression (G-Zac-LPR) multiplied (384-fold) transactivation. Equal DNA-binding activities were measured for the entire set of constructs in an electrophoretic mobility shift assay with an oligonucleotide encoding one GAL4 DNA-binding site as the probe (data not shown). Therefore, transactivation by the joint linker and proline repeat region demonstrated strong cooperativity between two independent and unrelated transactivation domains.

To confirm the relevance of these putative transactivation domains for Zac function, we deleted them individually within the full-length Zac fusion protein. Absence of either the linker (G-ZacΔL) or the proline repeat (G-ZacΔPR) region diminished transactivation by 5-fold and 10-fold, respectively. In support of their role, the absence of both domains (G-ZacΔLPR) completely abolished transactivation (Fig. 1B). To further validate these results, we additionally studied native Zac bound to a reporter plasmid containing two copies of a cognate Zac DNA-binding site (see below). In agreement with the data obtained from the Gal4 fusion system, transactivation was clearly diminished or abolished in the absence of either or both of the identified transactivation domains (Fig. 1D). Amounts of Zac, ZacΔL, ZacΔPR, and ZacΔLPR were adjusted to similar levels of protein expression in this experiment (inset).

The structural basis for the ability of activators to stimulate transcription, the importance of particular residues, and their position within the transactivation domain are only partly understood (9). The Zac proline repeat domain has a characteristic modular structure which consists of 13 PLE, 15 PMQ, 5 PML, and 1 PLQ amino acid triplet (Fig. 1E). In view of this unusual organization, we asked whether individual motifs autonomously confer transactivation. In order to test this, we serially fused oligonucleotides encoding four copies of each PXX triplet to the heterologous DNA-binding domain of Gal4. Four PLE triplets transactivated 4-fold, whereas insertion of a second and third oligonucleotide showed 69-fold and 596-fold transactivation, respectively (Fig. 1F).

To test the positional restraints of the proline residue within this motif, we designed oligonucleotides in which they faced each other (LEPPLELEPPLE). Fusion proteins including two, four, and six copies of this motif transactivated 3-fold, 92-fold, and 539-fold, respectively, thus behaving indistinguishably from the parent PLE polypeptide (Fig. 1F). PLE and LEPPLE fusion proteins containing multiple copies of each motif were less expressed, emphasizing their actual activities (Fig. 1G). These results led us to test whether negatively charged residues are critical to function. When lysine replaced glutamic acid (PLK), transactivation was abolished despite efficient expression of this fusion protein. Likewise, the motif PMQ, which contains the polar residue glutamine, conferred no transactivation. Also, the construct PLQ, in which leucine substituted for methionine, activated only weakly, although both the PMQ and PLQ fusion proteins were well expressed.

Finally, we compared transactivation of the PLE minidomain to the one reported for homopolymeric prolines (8). In agreement with this study, we noted a threefold transactivation by PPP polypeptides, which was, however, at a fraction of the activity obtained for the PLE and LEPPLE constructs and did not significantly change with the number of proline residues present or the amount of DNA cotransfected. We conclude that the proline repeat domain of Zac contains a genuine transactivation function based on the strong transcriptional synergism of discrete PLE minidomains which strictly depended on the acidic charge of the glutamic acid residue. Our findings further suggest that a recently reported Zac variant (AF324471), which lacked amino acids 343 to 362 representing five PLQ and two PML motifs, is unlikely to differ in transactivation potency.

Zac DNA-binding and transcriptional activities.

With a random oligonucleotide selection assay we previously (28) recovered two related consensus sequences: (i) a G-rich motif (5′-GGGGGGnnnnnnGGGGGG-3′) following three sequential cycles of gel shift purification, and (ii) a GC-rich motif (5′-nacGGGGGGCCCCtttan-3′) which prevailed after additional rounds of selection. To assess DNA binding in vitro and transactivation in vivo in parallel, oligonucleotides derived from these consensus sequences were inserted in front of the TATA element of a luciferase reporter plasmid. For DNA binding in vitro, restriction fragments containing individual motifs were excised from the respective reporter plasmids, fractionated on polyacrylamide gels, and purified. Labeled fragments were tested in electrophoretic mobility shift assays with nuclear extracts of Zac-transfected Saos-2 cells, which were adjusted to the results from the respective Zac immunoblots. Results are representative of at least three experiments varying by less than 10%. Half-maximal transactivation of these reporter plasmids by cotransfected Zac (varying by less than 5%) was determined from three dose-response curves and referred to the activity of the parent motif, which was set to 100%.

Zac binding to a GC-rich DNA element confers transactivation.

In order to define the core for Zac DNA binding and transactivation, we equilaterally reduced the number of G and C nucleotides within the GC-rich consensus. The motif G₄C₃ showed about half of the DNA-binding and transactivation activity measured for the motif G₄C₄, whereas the motif G₃C₃ was inactive (Fig. 2A). The sequence G₄C₂ retained some DNA-binding and transactivation activity, which was not the case for the motif G₄C₁.

FIG. 2.

Zac transactivates the palindromic DNA element dependent on zinc fingers 6 and 7. (A) Determination of the core sequence within the GC-rich consensus motif. Zac DNA-binding activity (DB%) was measured by electrophoretic mobility shift assays with the motifs indicated at the top. Half-maximal transactivation (TA%) was measured with the corresponding reporter plasmids. Activities are referred to the consensus motif, which was set to 100%. (B) Zac transactivates in an orientation-dependent manner. Single or four tandem copies of G₄C₄ or of the reverse motif C₄G₄ were tested for transactivation (TA) with increasing doses of Zac. (C) Zac binds specifically to G₄C₄. Zac-DNA complexes were inhibited by unlabeled self-oligonucleotide (100× cold) or by anti-Zac serum (α-Zac). The mutated motif G₄AC₄ (100× cold mt) and preimmune and anti-p53 serum (α-p53) were inactive. (D) The palindrome G₄C₄ is a high-affinity DNA-binding site. A constant amount of Zac protein was incubated with increasing concentrations of G₄C₄ until equilibrium occurred. The dissociation constant (K_D, ≈2.7 nM) was estimated as the negative reciprocal of the regressed slope by Scatchard analysis (inset). (E) The central nucleotide positions of the palindrome G₄C₄ are conserved. Substitutions in the primary sequence or the spacing at central positions of G₄C₄ prevented Zac DNA binding and transactivation. Motifs are noted at the top. (F) Zac N-terminal zinc fingers are dispensable for DNA binding and transactivation. Successive truncations C-terminal to each zinc finger as noted at the top were tested for DNA binding (DB%) and transactivation (TA%). Protein expression levels of these cDNAs are shown in the bottom panel. (G) Zac DNA binding and transactivation depend on zinc fingers 6 and 7. Zac proteins containing single broken zinc fingers as noted at the top were tested for DNA binding (DB%), half-maximal transactivation (TA%), and protein expression levels. (H) Scheme of the C₂H₂ zinc finger. Residues forming the β-sheet and the α-helix are shown in dark and light gray, respectively. Key positions within the α-helix are numbered (−1, 2, 3, and 6). Conserved residues are cysteine (Cys) and histidine (His) coordinated by one zinc ion (Zn), phenylalanine (F) or tyrosine (Y), and any hydrophobic amino acid (Φ). (I) Key residues within the recognition helix of zinc fingers 6 and 7 participate in DNA binding. Mutated key residues within the α-helices (positions −1, 2, 3, and 6) of zinc fingers (ZF) 6 and 7 were tested for DNA binding (DB%), half-maximal transactivation (TA%), and protein expression. (J) Zinc fingers 6 and 7 are sufficient for G₄C₄ DNA binding and transactivation. Native Zac, Zac containing solely zinc fingers 6 and 7 (ZacΔZF1-5), or additionally single broken zinc fingers (ZacΔZF1-5/6mt and ZacΔZF1-5/7mt) or α-helical mutations (ZacΔZF1-5/R173N and ZacΔZF1-5/R201N) were tested for DNA binding, transactivation, and protein expression. (K) Scheme comparing predicted amino acid (AA) base (nt) contacts of key residues (−1, 2, 3, and 6) within the α-helices of zinc fingers (ZF) 6 and 7 to the palindrome G₄C₄.

We also tested whether Zac transactivation was orientation dependent. Indeed, transactivation conferred by one copy of the reverse motif C₄G₄ was only half of that measured for the motif G₄C₄ (Fig. 2B). This difference grew larger in the case of four tandem copies of each motif and was also observed for cotransfection of human ZAC or the related zinc finger proteins KIAA and PLAG (28) (data not shown).

We also investigated the authenticity of the G₄C₄ binding site by determining its specificity and affinity. Addition of a 100-fold excess of unlabeled self-oligonucleotide but not of the mutated oligonucleotide G₄AC₄ (see below) completely inhibited the Zac-DNA complex (Fig. 2C). Moreover, preincubation with a polyclonal Zac antiserum efficiently prevented the formation of the complex observed in the presence of preimmune serum or an unrelated p53 antibody. Lastly, the Zac-specific DNA-protein complex was not observed for nuclear extracts from mock-transfected cells in any of these motifs (data not shown).

We determined Zac affinity for the motif G₄C₄ by Scatchard analysis of saturation binding isotherms and measured an equilibrium binding dissociation constant (K_D) of ≈2.7 nM (Fig. 2D). Taken together, these results demonstrated specific and high-affinity binding of Zac to the palindrome G₄C₄, which conferred transactivation in an orientation-dependent manner.

The specificity of DNA-binding sites involves primary nucleotide sequences as well as spacing of half-sites. To assess both possibilities at the same time, we inserted two complementary base pairs at the center of the palindrome G₄C₄ to obtain the motif G₄CGC₄. This inversion strongly impaired Zac transactivation and DNA binding (Fig. 2E). To analyze spacing restraints alone, we tested the motif G₄AC₄, in which the complementary half-sites were separated by one unrelated nucleotide. This insertion strongly weakened Zac DNA-binding and transactivation activity, which declined further and progressively in the case of two or three intervening nucleotides (Fig. 2E). These results demonstrated that the central positions within the palindrome G₄C₄ were conserved and that changes either in primary sequence or in spacing interfered with Zac function.

Although similarities in DNA recognition by related zinc fingers have been pointed out, these do not provide rules which could explain DNA-binding specificity in general. The prototypical C₂H₂ zinc finger binds a single zinc ion that is sandwiched between the two-stranded antiparallel β-sheet and the α-helix (Fig. 2H, dark and light grey residues, respectively), and at least two zinc fingers are necessary to establish high-affinity and stable contacts with DNA (30). In this view, Zac could recognize G₄C₄ (i) by more than two zinc fingers, (ii) by two zinc fingers with key residues within each α-helix contacting single or multiple bases, and (iii) by at least one zinc finger in case Zac binds as a dimer or as an oligomer.

To distinguish between these possibilities, we studied progressive N-terminal deletions of the zinc finger domain. Stepwise truncation of zinc fingers 1 to 5 (ΔZF1 to ΔZF1-5) reduced neither Zac DNA binding nor transactivation. In the latter case, we observed a twofold enhancement in the presence of unchanged protein levels (Fig. 2F, bottom). In contrast, Zac DNA binding and transactivation abruptly disappeared upon deletion of zinc fingers 6 and 7. These proteins were undetectable in nuclear extracts (Fig. 2F, bottom) but were even more highly expressed in whole-cell extracts (data not shown), demonstrating a defect in nuclear localization that precluded further analysis of the role of zinc fingers 6 and 7 at this step.

Therefore, we generated a second set of constructs with subtler defects in DNA binding by replacing the first cysteine residue with alanine, which destroys the zinc-dependent tetrahedral coordination of the zinc finger structure (broken zinc finger, Fig. 2H). Analysis of these mutated Zac cDNAs confirmed that zinc fingers 1 to 5 were dispensable to DNA-binding and transactivation of the palindrome G₄C₄ (Fig. 2G). Importantly, if either zinc finger 6 or 7 was broken, this completely abolished binding of the G₄C₄ element in the presence of protein levels undistinguishable from those of wild-type Zac. In conclusion, Zac zinc fingers 6 and 7 were necessary to recognize and transactivate the palindrome G₄C₄.

Amino acid residues at positions −1, 2, 3, and 6 (numbering with respect to the start site of the α-helix, Fig. 2H) typically make key base contacts in the major groove that define sequence specificity. Possible pairings between the 20 amino acids and the four DNA bases and stereochemical rules which describe the base and amino acid positions in contact have been compiled into a “recognition code” that describes the preferred side chain base interactions of key residues within the recognition helix (30). The scheme of amino acids at position −1, 2, 3, and 6 within the α-helices of zinc fingers 6 and 7 compares anticipated nucleotides to the palindrome G₄C₄ (Fig. 2K). Provided that position 2 of zinc finger 7 and positions 2 and 6 of zinc finger 6 contact bases in the secondary strand, a code/consensus correlation of 100% is achieved. The conserved center of the palindrome G₄C₄ indicated a limited flexibility of zinc fingers 6 and 7, compatible with position 6 of zinc finger 6 and position 2 of zinc finger 7 reciprocally contacting bases on the primary strand of the respective up- and downstream subsites.

To investigate this code/consensus correlation, we individually replaced each key residue within the α-helices with asparagine, a structurally related amino acid containing a side chain with a very low affinity for guanosine or cytidine residues. Replacement at position −1 within zinc finger 6 (T167N) weakly reduced G₄C₄ binding and transactivation, suggesting an auxiliary role of this residue in base recognition (Fig. 2I). In contrast, mutation of position 2, 3, or 6 of zinc finger 6 (K169N, D170N, and R173N) or of position −1, 2, 3, or 6 of zinc finger 7 (R195N, D197N, H198N, and R201N) each strongly reduced DNA binding. All key residues within the α-helices of zinc fingers 6 and 7 were important to transactivation except position 3 of zinc finger 6, which preserved some activity. Probably, conditions for DNA binding in vitro were more stringent than those for transactivation in vivo, or additional protein interactions occurring in vivo mitigated the phenotype. Taken together, our data conclusively showed that Zac DNA binding and transactivation of the palindrome G₄C₄ involved all key residues within the α-helices of zinc fingers 6 and 7.

Lastly, to investigate whether zinc fingers 6 and 7 were sufficient for G₄C₄ binding, we studied broken zinc finger and α-helix mutations within the context of a Zac protein having only these two zinc fingers. As shown in Fig. 2J, a single broken zinc finger (6 or 7) abolished G₄C₄ DNA binding and transactivation. Noticeably, however, nuclear protein levels of these constructs (ZacΔZF1-5/6mt and ZacΔZF1-5/7mt) were weakly or strongly reduced compared to the corresponding mutations contained within the entire zinc finger domain (Fig. 2G). This finding points to a second nuclear translocation signal in the N-terminal half of the zinc finger domain that could compensate for the impaired nuclear localization in case of broken zinc fingers 6 and 7, respectively. In contrast, single α-helical mutations in either zinc finger 6 or 7 (ZacΔZF1-5/R173N and ZacΔZF1-5/R201N) led to a complete loss of G₄C₄ DNA binding without changes in nuclear protein expression levels and demonstrated that each was necessary and sufficient for DNA binding and transactivation.

Zac binding to a G-rich DNA element differentially regulates transcriptional activities.

With respect to the G-rich consensus motif, we equilaterally decreased the number of guanosine residues to determine bases essential to Zac DNA binding. Formation of Zac-DNA complexes in the case of the motif G₄N₆G₄ were undistinguishable from the consensus motif but disappeared in the case of the motifs G₃N₆G₃ and G₂N₆G₂ (Fig. 3A, left). The asymmetric motifs G₄N₆G₃, G₄N₆G₂, G₄N₆G₁, and G₄N₆G₀ showed similar behaviors, defining G₄N₆G₄ as the recognition core (Fig. 3A, right). Unexpectedly, none of these motifs conferred transactivation, which led us to test the G₄ clusters adjacent to each other or separated by an increasing number of unrelated nucleotides. In comparison to the parent motif G₄N₆G₄, Zac-DNA complexes were reduced twofold in the case of G₄N₃G₄. In contrast, they were enhanced twofold in the case of G₄N₁₂G₄ and unaltered in the case of the motifs G₄N₀G₄ and G₄N₂₄G₄ (Fig. 3B). Again, none of the corresponding reporter plasmids conferred transactivation by Zac.

FIG. 3.

Zac DNA binding to a single or a repeat G-rich DNA element confers repression and transactivation, respectively. (A) Determination of the core sequence within the G-rich consensus as described above. DNA-binding activity (DB%) is indicated. None of these motifs conferred transactivation. (B) Distinct spacing of G clusters maintains DNA binding but enables no transactivation. G clusters (G₄) were separated by 0, 3, 6, 12, or 24 unrelated nucleotides and tested for DNA-binding activity (DB%). None of these motifs conferred transactivation. (C) Zac binds specifically to a single G₄N₆G₄ motif. Zac-DNA complexes were abolished by unlabeled self-oligonucleotide G₄N₆G₄ (100× cold) or by anti-Zac serum (α-Zac). The mutated motif G₄N₆G₃ (100× cold mt) and preimmune and anti-p53 (α-p53) serum were inactive. (D) The G₄N₆G₄ motif is a high-affinity DNA-binding site. The dissociation constant (K_D, ≈4.2 nM) was measured as described above. (E) Schematic drawing of the simian virus 40 (SV40) early promoter showing the two 72-bp enhancer regions, each containing a complementary Zac DNA-binding site and the three 21-bp repeats encoding six canonical Sp1 sites. The sequence of the complementary Zac DNA-binding site is shown. (F) Zac dose-dependently represses simian virus 40 early promoter activity (SV40 wt). Repression was largely reduced for a simian virus 40 promoter containing a single substitution (5′-CCACAGGCTCCCC-3′) in each of the cDNA binding sites (SV40 mt). (G) The direct repeat element (G₄N₆G₄)₂ confers transactivation. Two copies of the motif G₄N₆G₄ separated by six unrelated nucleotides conferred robust transactivation for increasing doses of Zac. (H) Reverse direct elements discriminate between Zac DNA binding (DB%) and transactivation (TA%). Zac binds to two copies of G₃N₆G₄ and of G₂N₆G₄ with similar or even higher affinity than to G₄N₆G₄ but not to the direct repeat G₃N₆G₃, whereas transactivation progressively declines. (I) Multiple zinc finger contacts are necessary to bind the repeat element. Successive truncations C-terminal to each zinc finger were tested for DNA binding (DB%) and transactivation (TA%). (J) Zac N- and C-terminal zinc fingers participate in binding of the direct repeat element. Single broken zinc fingers were tested with (G₄N₆G₄)₂ for DNA binding (DB%) and transactivation (TA%), which depends on zinc fingers 2 to 4, 6, and 7.

To validate the core motif G₄N₆G₄ as an authentic DNA-binding site, we determined its specificity and affinity. A 100-fold excess of unlabeled self-oligonucleotide completely abolished the Zac-DNA complex, whereas the motif G₄N₆G₃ was inactive (Fig. 3C). Furthermore, preincubation with Zac antiserum but not with preimmune serum or an unrelated p53 antibody completely abolished the shift, demonstrating the specificity of the Zac-DNA complex. Finally, Zac-specific DNA-protein complexes were not observed for any of these motifs when nuclear extracts from mock-transfected cells were used (data not shown). By Scatchard analysis, we measured for the motif G₄N₆G₄ an equilibrium binding dissociation constant (K_D) of ≈4.2 nM (Fig. 3D), which is similar to the one obtained for the palindromic Zac element (K_D, ≈2.7 nM).

During the early phase of this study, we observed that Zac strongly repressed a simian virus 40 early promoter-driven β-galactosidase reporter plasmid which we initially used to standardize transfection efficiencies. The simian virus 40 promoter contains within each of its two 72-bp repeat enhancer regions the sequence 5′-CCCCAGGCTCCCC-3′, which perfectly matched a complementary Zac DNA-binding site (Fig. 3E). In fact, Zac efficiently bound the 72-bp fragment in electrophoretic mobility shift assay experiments (K_D ≈4.1 nM), whereas a single nucleotide substitution in the upstream half-site (5′-CCACAGGCTCCCC-3′) completely abolished Zac DNA binding (A. Hoffmann and D. Spengler, unpublished observations). Importantly, the presence of this mutation in each of the 72-bp enhancer regions largely reduced Zac-mediated repression of the simian virus 40 promoter in vivo (Fig. 3F). Moreover, we also observed Zac-dependent repression when a number of G-rich and/or C-rich cellular promoters were additionally tested (A. Hoffmann and D. Spengler, unpublished observations). Taken together, these findings demonstrated that Zac possessed, in addition to transcriptional activation, a repressor function, which became disclosed upon binding to a single G-rich DNA element.

In view of these observations, we speculated that Zac dimerization on two neighboring DNA elements might be necessary to unlock transactivation. In order to test this idea, we studied Zac behavior on two adjacent G₄N₆G₄ motifs separated by six unrelated nucleotides. In fact, the presence of a second copy of this motif enabled vigorous transactivation (Fig. 3G). In the following experiments, we refer to this direct and related repeat motifs as (G₄N₆G₄)₂ and (G_nN₆G_n)₂, respectively. Next, we reanalyzed the motif G₄N₆G₄ and its derivatives in the context of two copies which were each placed tail to tail and spaced by six unrelated nucleotides. Zac strongly bound to the reverse repeats (G₃N₆G₄)₂ and (G₂N₆G₄)₂ with similar or even higher affinity than observed for the direct repeat (G₄N₆G₄)₂, but weakly to the direct repeat (G₃N₆G₃)₂ (Fig. 3H). In sharp contrast, Zac transactivation of the reverse repeats (G₃N₆G₄)₂ and (G₂N₆G₄)₂ was 2-fold and 10-fold, respectively, diminished compared to the direct repeat (G₄N₆G₄)₂.

Because loss of transactivation occurred despite unchanged DNA binding, we hypothesized that Zac transactivation is site-specifically controlled. Moreover, dimerization, although partly necessary for DNA binding, was apparently not sufficient for transactivation. In this view, different mechanisms of DNA binding and/or mechanisms distinct from DNA binding regulate Zac transactivation. To test these assumptions for this class of repeat elements, we studied, first, DNA binding by each of the Zac zinc fingers and, second, amino acid base contacts by key residues within each DNA-binding recognition helix. Deletion of zinc finger 1 reduced transactivation twofold compared to wild-type Zac, although DNA binding of (G₄N₆G₄)₂ was unaltered (Fig. 3I). Deletions downstream of zinc finger 1 collectively abolished DNA binding and transactivation.

Therefore, we next tested Zac cDNA constructs encoding single broken zinc fingers. In these experiments, zinc fingers 2 to 4, 6, and 7 proved necessary to (G₄N₆G₄)₂ DNA binding and transactivation, whereas zinc fingers 1 and 5 were dispensable (Fig. 3J). Importantly, we obtained identical results for a single G₄N₆G₄ element (data not shown). Based on these data, we concluded that the absence or presence of transactivation in the case of a single or a repeat element did not result from numerical differences in zinc finger contacts but, alternatively, from mechanistic differences in the amino acid base contacts established by each DNA-bound zinc finger.

Because Zac binding to the repeat elements depended on zinc fingers 2 to 4 in addition to zinc fingers 6 and 7, we considered the former prime candidates for site-specific regulation of transactivation. The scheme of amino acids at positions −1, 2, 3, and 6 within the α-helices of zinc fingers 2, 3 and 4 compares anticipated nucleotides to the repeat element (G₄N₆G₄)₂ under the assumption that zinc fingers 2 to 4 bind to one half-site and zinc fingers 6 and 7 bind to the second one (Fig. 4A). We noticed a good code/consensus correlation, which in addition predicted that one base might be contacted by multiple amino acids.

FIG. 4.

Differential control of Zac transactivation. (A) Scheme comparing predicted amino acid base contacts of key residues (position −1, 2, 3, and 6) within the α-helices of zinc fingers 2, 3, and 4 to one half-site of G₄N₆G₄. (B) Zac binding to the repeat element involves multiple amino acid base contacts of zinc fingers 2 to 4. Mutated key residues within the α-helices of zinc fingers 2, 3, and 4 as noted at the top were tested with (G₄N₆G₄)₂ for DNA binding (DB%) and transactivation (TA%). Immunoblot shows the corresponding protein expression levels. (C to F) Site-specific control of transactivation occurs through zinc fingers 6 and 7. Mutated key residues within zinc fingers 6 and 7 as noted at the top were tested for DNA binding (DB%) and transactivation (TA%) with the single element G₄N₆G₄ (C) or the repeat elements (G₄N₆G₄)₂ (D), (G₃N₆G₄)₂ (E), and (G₂N₆G₄)₂ (F). Positive control phenotypes (PC) are shaded and show a value of >1 for DNA binding divided by transactivation. (G) Zac hybrids are selectively impaired in transactivation of the repeat element. The ectopic transactivation domain of VP16 replaces that of Zac in the hybrid ZacΔLPR/VP and confers higher transactivation on (G₄C₄)₂ compared to wild-type Zac, whereas transactivation on (G₄N₆G₄)₂ is strongly reduced. (H) Scheme summarizing the behavior of mutated key residues within the α-helices of zinc fingers 6 and 7 on a single and the different repeat elements for DNA binding (DB) and transactivation (TA). DNA motifs, zinc fingers (ZF), numbered and abbreviated amino acids, and their respective position within the α-helix are indicated. Activity scores are defined in the inset. Positive control phenotypes are shaded.

To evaluate this correlation, we individually replaced each key residue within the α-helices of zinc fingers 2 to 4 with asparagine. Briefly, position 3 of zinc finger 2 (K48N) showed no base contacts, whereas substitution of position 6 (R51N) strongly reduced DNA binding and moderately reduced transactivation (Fig. 4B). Similar results were obtained for position −1 (R73N) within zinc finger 3, whereas replacement of positions 2 (D75N) and 3 (H76N) caused moderate defects in DNA binding and small impairments in transactivation. Mutation of position 6 within zinc finger 4 again showed an intermediate phenotype for DNA binding and transactivation. Immunoblots demonstrated that this set of constructs were expressed at similar protein levels (Fig. 4B, bottom).

In general, changes in DNA binding and transactivation of these α-helical mutations of zinc fingers 2 to 4 were less than those obtained from the corresponding mutations of zinc fingers 6 and 7 when measured on the palindromic DNA element. These data suggest that multiple amino acids within the α-helices of zinc fingers 2 to 4 contact a single nucleotide base. More importantly, we again obtained identical results when this set of constructs were tested with a single G₄N₆G₄ element (data not shown), which led us to conclude that base contacts established by zinc fingers 2 to 4 did not participate in site-specific control of transactivation.

In light of these findings, we went on to evaluate the effect of individually substituted key residues within the recognition helix of zinc fingers 6 and 7 for DNA binding to the single or the repeat G₄N₆G₄ element. Replacement of positions −1 and 2 by asparagine within zinc finger 6 (T167N and K169N) did not alter binding to a single G₄N₆G₄ element, whereas a twofold reduction occurred for mutation of positions 3 and 6 (D170N and R173N) (Fig. 4C). In contrast, substitution of positions −1, 2, and 6 within zinc finger 7 (R195N, D197N, and R201N) each strongly reduced DNA binding, whereas the effect of position 3 (H198N) was about twofold.

A strikingly different pattern of amino acid base contacts emerged when we tested this class of mutations of zinc fingers 6 and 7 with the repeat element (G₄N₆G₄)₂. Of note, none of the mutations within the recognition helix of zinc finger 6 interfered with efficient DNA binding (Fig. 4D). With respect to zinc finger 7, the function of positions −1, 2, and 3 within the recognition helix were conserved, with positions −1 and 2 (R195N and D197N, respectively) being necessary and position 3 (H198N) being less required for DNA binding. Surprisingly, position 6 (R201N) within the α-helix of zinc finger 7 was completely dispensable for DNA binding to the repeat element (G₄N₆G₄)₂, in sharp contrast to the behavior on a single element. Interestingly again, this zinc finger mutation conferred only weak transactivation (10%) compared to wild-type Zac. A similar, albeit blunted phenotype was noted for mutation of positions 2 and 6 (K169N and R173N, respectively) of zinc finger 6, which reduced transactivation by 40% and 70%, respectively.

Based on these results, we concluded that this class of mutations distinguished contacts essential for transactivation from those involved in DNA binding. This phenotype, known as a positive control mutant (10), was clearly illustrated by dividing values for DNA binding by those for transactivation. Mutation of positions 2 and 6 within the α-helix of zinc finger 6 and position 6 within the α-helix of zinc finger 7 gave results of 2.2, 3.3, and 12, respectively. In contrast, all other α-helical mutations within the Zac zinc finger domain showed values of 1 or less, probably because analysis of transactivation in vivo was less stringent than analysis of DNA binding in vitro. Taken together, these results evidenced a clear mechanistic difference in Zac binding to a single versus a repeat element. Thus, DNA sequences specifically controlled transactivation, first at the level of amino acid base contacts necessary for DNA binding and second at the level of contacts distinct from DNA binding but necessary for transactivation. Based on these findings, we hypothesized that the different degrees of transactivation conferred by Zac on the individual repeat elements result from the differential use of DNA-dependent and -independent contacts in response to the specific nucleotide sequences.

To further test this prediction, we studied this class of mutations on the reverse repeat element (G₃N₆G₄)₂, which conferred half of the Zac transactivation noted for the direct repeat element (G₄N₆G₄)₂. Changes in DNA binding and transactivation preferentially mapped to the positive control phenotypes (Fig. 4E). Specifically, position 2 of zinc finger 6 was dispensable for transactivation of the reverse repeat motif (G₃N₆G₄)₂. In contrast, position 6 of zinc finger 6 and zinc finger 7 was partly or strongly required for DNA binding. These results marked a transition in the function of position 6 within the α-helix of zinc finger 7 with an exclusive role in transactivation on the direct repeat element (G₄N₆G₄)₂ and an additional role in DNA binding on the reverse repeat element (G₃N₆G₄)₂. We reasoned that this transition should advance further if Zac is DNA bound but deficient in transactivation. In fact, Zac binding to the reverse repeat element (G₂N₆G₄)₂ strongly depended on position 6 within the α-helix of zinc finger 7 (Fig. 4F), whereas minor changes occurred for positions −1, 2, and 3. These results on site-specific control of Zac transactivation are schematically summarized in Fig. 4H. Taken together, they imply that the DNA-binding domain of Zac specifically communicates with its transactivation function and that this communication is tightly regulated by the underlying amino acid base contacts within the α-helices of zinc finger 6 and, more importantly, zinc finger 7.

These findings led us to ask whether an unrelated transactivation domain could respond to this mode of regulation. To address this issue, we replaced the Zac transactivation domain with herpesvirus VP16 to create ZacΔLPR/VP and compared Zac to this hybrid protein on both classes of reporter plasmids. Remarkably, ZacΔLPR/VP strongly induced two copies of the palindromic DNA element (G₄C₄)₂, exceeding wild-type Zac activity, but was about 10-fold less active than wild-type Zac on the direct repeat element (G₄N₆G₄)₂, although it bound similarly to both sites (Fig. 4G and data not shown). Likewise, the hybrid proteins ZacΔLPR/p53 and ZacΔLPR/E1A, containing the activation domains of human p53 (amino acids 1 to 64) and of adenovirus E1A (amino acids 121 to 223), respectively, were selectively about 15-fold less active than wild-type Zac on the repeat element but not on the palindromic one (data not shown).

We suggest that DNA binding to the palindromic element imposed no or few restraints on these Zac hybrids, whereas they behaved as dysfunctional activators on the repeat element. These results further support the idea that Zac DNA binding and transactivation of the repeat elements are carefully coordinated and suggest that conformational restraints or regulatory protein interactions of the zinc finger domain interfered with the function of the unrelated transactivation domains.

Cooperative DNA binding and dimerization.

Our results strongly indicated that Zac molecules interact with each other on adjacent DNA binding sites and raised the possibility that such interactions also increase the affinity of DNA binding. Concordant with this prediction, we observed that Zac bound about eightfold more to the repeat motif than to a single site, corresponding to a fourfold cooperativity in DNA binding (Fig. 5A). Preincubation with Zac antiserum completely abolished these shifts and proved their specificity. We further tested whether an enlarged spacing between the two half-sites of the repeat element could disrupt dimerization. Indeed, Zac cooperative DNA binding was strongly reduced when the two half-sites were separated by 12 instead of 6 unrelated nucleotides [(G₄N₆G₄N₁₂)₂]. Again, Zac antiserum completely prevented this shift.

FIG. 5.

Cooperative DNA binding and dimerization. (A) Zac dimerization enhances DNA binding. Fragments containing one or two (G₄N₆G₄) sites were separated by 6 (N₆) or 12 (N₁₂) unrelated nucleotides, as noted at the top. In case of N₆, fragments were of equal size. Electrophoretic mobility shift assays were done in the absence (−) or presence (+) of Zac antiserum (α-Zac). DNA-binding activity (DB%) for (G₄N₆G₄N₆)₂ and (G₄N₆G₄N₁₂)₂ was set to 100% to calculate cooperativity (CO). Arrows mark Zac-DNA complexes. Note enhanced migration of the dimer DNA complex. (B) Zac dimerizes through its zinc finger domains. In the GST pulldown assay, equal amounts of in vitro-translated Zac, ZacΔZF, ZacΔLPR, and ZacΔC were incubated with GST-Zac, GST-ZacΔZF, and GST alone. Lanes −, 10% of the input; lanes +, fraction of the input (100%) bound by each GST protein (BD%). (C) Mapping of zinc fingers participating in dimerization. Mutants with deletions C-terminal to each zinc finger were incubated with GST-Zac-ZF containing zinc fingers 1 to 7. Sequences upstream of zinc finger 2 are necessary for binding. (D) Top, scheme of prototypical C₂H₂ zinc finger structure; β1 and β2 strand and α-helix are outlined. Bottom, Zac sequence of zinc fingers 1 and 2; amino acids are numbered. (E) Fine mapping of interactive subregions of zinc finger 2. Refined N-terminal deletions of zinc finger 2 (numbers as above) were incubated with GST-Zac-ZF1-2 containing solely zinc fingers 1 and 2. The first H-C link and the β1 strand of ZF2 are important to binding. (F) Zac dimerization in vivo depends on zinc finger 2. In the mammalian two-hybrid system, G-ZF (10 ng) containing the Zac zinc finger domain fused to Gal4 was cotransfected with increasing doses of the hybrid ZF/VP, which contains the zinc finger domain fused to the VP16 transactivator. The transactivator VP alone served as a control (left). To map the interactive subregions of zinc finger 2, the hybrids ZF/VP, ZFΔ31/VP, ZFΔ34/VP, and VP alone (200 ng each) were cotransfected with G-ZF (right).

Additional experiments with solely the Zac DNA-binding domain obtained from nuclear extracts, purified from Escherichia coli or produced by in vitro translation failed due to either undetectable expression levels, strongly reduced DNA-binding activity, or the formation of multiple nonspecific complexes. Because the Zac-DNA complex on the repeat element migrated near the one on a single G₄N₆G₄ site and because the DNA fragments tested were of equal size, we suggest that dimerization of native Zac led to the formation of complexes with enhanced mobility. Moreover, we propose that the aberrant migration of this Zac-DNA complex could point to distinct conformations of native Zac assembled on a half-site versus the repeat motif and that subtle changes in Zac DNA-binding domain structure might be translated to other Zac regions, resulting in more global conformational changes that control transcriptional activities.

We then sought to detect Zac dimerization in vivo by coimmunoprecipitation experiments with cellular extracts obtained from cotransfection of HA- or vesicular stomatitis virus G protein epitope-tagged Zac. p53 which was similarly tagged served as a positive control in these studies. Immunoblots from Zac or p53 cotransfections evidenced no Zac immunoreactivity, whereas p53 could be readily detected (data not shown). Based on these results, Zac dimerization appeared to be too weak to persist under these experimental conditions.

Therefore, to study Zac interaction in more detail, we employed the GST pulldown assay. These experiments clearly demonstrated that in vitro-translated Zac bound efficiently to a GST-Zac fusion protein. Remarkably, deletion of the zinc finger domain completely abolished retention (Fig. 5B, lanes 3 to 4), whereas absence of Zac transactivation domain (ZacΔLPR) or of the whole C terminus (ZacΔC) maintained binding to GST-Zac. In the reverse experiment, GST-Zac lacking the zinc finger domain (GST-ZacΔZF) bound none of these proteins, and none of them bound to GST alone. Moreover, a GST fusion protein containing solely zinc fingers 1 to 7 (GST-Zac-ZF) efficiently retained Zac (Fig. 5C, lanes 1 to 2). Zac binding resisted the inclusion of ethidium bromide (100 μg ml⁻¹) in the incubation step, demonstrating that it was not due to the presence of bacterial DNA (data not shown). Therefore, the Zac zinc finger domain was necessary and sufficient for dimerization.

To narrow the interacting region within the zinc finger domain, successive N-terminal deletions of Zac were tested with a GST fusion protein containing the entire zinc finger domain. Retention of Zac was reduced by one-third and a striking ninefold for deletion of zinc fingers 1 and 2, respectively (Fig. 5C, lanes 3 to 6). Further downstream truncations gradually declined in GST-Zac binding, which was completely abolished upon deletion of zinc finger 6, indicative of a second, albeit weak interaction interface. In turn, a GST fusion protein containing solely zinc fingers 1 and 2 (GST-Zac-ZF1-2) efficiently retained full-length Zac, demonstrating that this region was necessary and sufficient for binding (Fig. 5E, lanes 1 to 2).

This finding raised the question of which structural domains of the second zinc finger participate in dimerization. Amino acids of zinc fingers 1 and 2 in relation to the structure of a prototypical C₂H₂ zinc finger are shown in Fig. 5D. Additional refined analysis of zinc finger 2 demonstrated that Zac binding required amino acids 28 to 40, whereas the putative β2 strand and the α-helix of zinc finger 2 were dispensable (Fig. 5E). Within this interactive region, mainly the first H-C link and β1 strand, but less so the amino acid loop connecting the cysteine ligands, were required for Zac binding.

To evaluate these findings for Zac interaction in vivo, we employed the mammalian two-hybrid system and cotransfected the Zac zinc finger domain fused to Gal4 (Gal4-ZF) with a plasmid containing the zinc finger domain fused to the transactivation domain of VP16 (ZF/VP). This hybrid dose-dependently increased transactivation, whereas VP16 alone was inactive (Fig. 5F, left). In the reverse experiment, Gal4-ZF but not Gal4 alone restored transactivation by ZF/VP (data not shown). These findings were further strengthened by electrophoretic mobility shift assay studies with Zac constructs containing N-terminal truncations of zinc finger 2 and the repeat element (G₄N₆G₄)₂. We noted in the absence of the first H-C link (ZacΔ31) a twofold impairment in DNA binding, whereas absence of the β1 strand of zinc finger 2 abolished any DNA binding, precluding detailed analysis at this step (data not shown).

Lastly, we evaluated the importance of these amino acids for Zac dimerization by means of the two-hybrid assay. These experiments showed substantial reductions in transactivation upon deletion of the first H-C link and the β1 strand (Fig. 5F, right). In sum, our results from electrophoretic mobility shift assays, GST pulldown experiments, and the two-hybrid system gave strong and concurrent evidence for Zac dimerization through zinc finger 2.

Zac differentially regulates the cytokeratin gene family and transactivates through the palindromic and repeat element.

To address the importance of our present findings for the regulation of endogenous genes, we used tetracycline-regulated Zac expression (24). Zac protein levels reached a plateau 12 h after tetracycline removal, and RNAs from cells grown for an additional 12 h were probed with a cDNA expression array. These results were compared to signals obtained from cells kept for 24 h with tetracycline (A. Hoffmann and D. Spengler, unpublished observations). Zac regulated the family of cytokeratin (CK) genes in a differential manner by inducing CK4, CK6, and CK14 gene expression and repressing CK2, CK7, CK12, and CK19 gene expression (Fig. 6A).

FIG. 6.

Zac differentially regulates the cytokeratin gene family and transactivates through the palindromic and repeat elements. (A) Zac differentially regulates cytokeratin (CK) gene expression. Zac induces CK4, CK6, and CK14 gene expression but represses CK2, CK7, CK12, and CK19 gene expression (fold regulation). (B) Time course of CK14 gene induction. CK14 gene expression was measured by reverse transcription-PCR in Zac- and ZacΔZF-expressing cell clones at the indicated times (hours) after tetracycline removal. (C) Zac expression enhances CK14 immunoreactivity. Immunostaining for CK14 protein (brown) in Zac- and ZacΔZF-expressing cells grown with (+) or without (−) tetracycline for 24 h. Counterstaining was done with toluidine blue. (D) Scheme of CK14 promoter sequences matching the palindromic or repeat Zac DNA-binding site. Elements represented in the distal SmaI-BstXI fragment are shown in black, and the locations of the different classes of motifs within the intact promoter are schematically indicated. (E) Zac transactivates the distal CK14 promoter plasmid. Cotransfection of Zac and ZacΔZF1-5 dose-dependently transactivated the SmaI-BstXI fragment derived from the CK14 promoter in front of a TATA element but not the vector alone. In contrast, Zac cDNAs encoding a broken zinc finger 7 (ZacZF7mt) or lacking the transactivation domain (ZacΔLPR) were inactive. (F) Zac differentially binds and transactivates motifs from the CK14 promoter. Zac DNA binding of the perfect repeat (rep) was set to 100% and compared to that of the imperfect repeat motifs M1 to M4 (left). Similarly, the perfect palindrome (pal) was compared to the incomplete palindromic motifs M5 and M6 (right). Half-maximal transactivation of the corresponding reporter plasmids is given at the bottom. (G) Mutated CK14 promoter motifs distinguish Zac DNA binding (DB%) from transactivation (TA%). Structural scheme of motifs M2, M3, and M5. Color code: green, orange, and red symbolize nucleotides predicted to promote or to interfere slightly or strongly with transactivation, respectively. Actual DNA binding (DB%) and transactivation (TA%) were measured as above and are indicated. (H) Mutated repeat and palindromic elements in the distal CK14 promoter selectively interfere with Zac transactivation. CK14 promoter fragments containing wild-type (−) or mutated sequences (+) of M2, M3, and M5 as shown above were cotransfected with Zac, ZacΔZF1-5, and ZacZF7mt (50 ng of each).

We analyzed CK14 gene regulation in more detail by reverse transcription-PCR experiments, which we performed at different time points following tetracycline removal. Upregulation of CK14 expression occurred as early as 6 h after tetracycline removal, compatible with a direct transcriptional effect of Zac, and peaked at 24 h (Fig. 6B). In contrast, a tetracycline-regulated cell clone containing Zac lacking the zinc finger domain (ZacΔZF) showed no significant change in CK14 mRNA levels despite similar steady-state levels of Zac and ZacΔZF proteins (data not shown). We observed faint immunostaining for CK14 protein in the repressed state of both cell clones, whereas robust immunoreactivity resulted from wild-type Zac expression in the absence of tetracycline (Fig. 6C).

The human CK14 promoter contains several G- and C-rich sequences matching both classes of Zac DNA elements, which were designated M1 to M6 (Fig. 6D). Two imperfect repeat motifs (M2 and M3) and one incomplete palindromic site (M5) localized to a SmaI-BstXI fragment present in the distal CK14 promoter region (Fig. 6D, lower scheme). To study regulation by Zac, we transferred this fragment in front of a TATA element. Despite somewhat lower constitutive activity compared to the intact CK14 promoter, expression levels of this construct still largely exceeded that of the minimal promoter (>5,000-fold). Cotransfection of Zac, however, further strongly activated (eightfold) this promoter plasmid but not the parent vector (Fig. 6E). Moreover, a Zac cDNA containing solely zinc fingers 6 and 7 (ZacΔZF1-5) conferred significant transactivation, albeit less than that obtained with wild-type Zac, whereas Zac proteins containing a broken zinc finger 7 (ZacZF7mt) or lacking the transactivation domain (ZacΔLPR) were inactive in CK14 promoter regulation.

These findings led us to analyze Zac DNA binding and transactivation in the context of the CK14 promoter and to compare the results with those described in the previous section. To this end, we first studied the behavior of the individual motifs, then identified mutations that abolished their activities, and finally analyzed the effect of these mutations in the context of the distal CK14 promoter. Oligonucleotides encoding motifs M1 to M6 were inserted separately into the minimal reporter plasmid, and DNA binding and transactivation were studied as described in the previous sections. Zac bound with similar or even higher affinity to the imperfect repeat motifs M1, M2, and M4 compared to the perfect repeat, whereas motif M3, which contained an interrupted central G cluster in one half-site, showed clearly reduced complex formation (Fig. 6F, left). Addition of the Zac antibody abolished these Zac-DNA complexes and demonstrated their specificity.

Importantly, Zac transactivation measured for the imperfect repeat motifs M1, M2, and M4 was consistently less than that obtained for the perfect repeat motif concurrent with the differential use of zinc finger contacts for DNA binding and transactivation (Fig. 6F, bottom). Moreover, we studied the imperfect palindromic motifs M5 and M6, which both showed specific, albeit reduced Zac DNA binding and transactivation (Fig. 6F, right). In further agreement with our previous results, transactivation of these imperfect palindromes was orientation dependent.

To identify mutations that could abolish Zac transactivation by the motifs M2, M3, and M5, we schematically compared them to the respective perfect repeat and palindromic motifs as depicted in Fig. 6G. The repeat motif M2 contains the sequence G₃N₁G₄, which confers moderate transactivation in the context of a repeat motif, in one half-site, whereas the second half-site corresponds to the sequence G₂N₆G₄, which is inactive in this respect. Substitution of one G residue in the outward cluster of the functional half-site led to the motif M2mt, largely deprived of transactivation despite unchanged Zac DNA binding (Fig. 6G). In contrast, motif M3 showed concomitant reductions in DNA binding and transactivation due to disruption of one of the central G clusters which regulate Zac dimerization. We reasoned that additional disruption of the outward G clusters (M3mt) should interfere with the residual transactivation but not with DNA binding, which was in fact the case (Fig. 6G). Furthermore, insertion of one nucleotide at the center of the imperfect palindrome (M5mt) completely prevented Zac DNA binding and transactivation (Fig. 6G).

Having identified mutations that interfered with transactivation of the isolated motifs, we studied their effects within the context of the distal CK14 promoter fragment. We stepwise mutated either both repeat motifs or the palindrome or both classes of motifs. The corresponding distal CK14 promoter constructs were cotransfected with either native Zac, a deletion of the N-terminal five zinc fingers (ZacΔZF1-5), or a broken zinc finger 7 (ZacZF7mt). In the presence of the mutated repeat motifs, transactivation by wild-type Zac was reduced by 60% compared to the intact promoter plasmid, whereas that of ZacΔZF1-5 was unchanged (Fig. 6H). In contrast, the mutated palindrome (M5mt) additionally prevented transactivation by ZacΔZF1-5 (Fig. 6H). Because wild-type Zac similarly transactivated both mutated promoter plasmids, albeit at a reduced level compared to the intact promoter, transactivation apparently occurred alternatively through the repeat and palindromic binding sites. Lastly, a CK14 fragment containing mutations in both classes of binding sites failed to significantly confer Zac transactivation, as was the case for a DNA-binding-defective Zac protein (Fig. 6H).

Taken together, these results strongly support Zac's role as transcription factor regulating gene repression and transactivation and demonstrate the presence of the palindromic and repeat DNA elements in the context of a complex promoter of a newly identified Zac target gene. Moreover, Zac DNA binding and transactivation of a natural gene confirmed the critical role of the zinc finger domain in communicating with the transactivation function.

DISCUSSION

We report here that Zac contains transactivation and repressor activities which are differentially controlled by DNA binding (Fig. 7). Moreover, we demonstrate for Zac a role as a transcription factor which confers transactivation through two related DNA elements in a differential manner (Fig. 7).

FIG. 7.

Molecular bar code of Zac DNA binding and transactivation. (A) Zac binds as a monomer to the palindrome G₄C₄ to transactivate in an orientation-dependent manner. All key residues within the α-helix of zinc fingers (ZF) 6 and 7 (except position −1 of zinc finger 6) participate in DNA binding (DB) and transactivation (TA). The function of each key residue is shown in color, as defined in the inset. (B) Zac binding to one half-site of the direct repeat element confers repression instead of transactivation. DNA binding occurs by multiple zinc finger contacts, including zinc fingers 2 to 4, 6, and 7. (C) Zac binding to the direct repeat element (G₄N₆G₄)₂ promotes dimerization through zinc finger 2 (open arrow) and transactivation. Multiple zinc fingers, including 2 to 4, 6, and 7, contribute to DNA binding. Amino acid base contacts of key residues from the recognition helix of zinc fingers 6 and 7 involve DNA-dependent and DNA-independent interactions. The latter, as highlighted by position 6 of zinc finger 7, are essential to transactivation. Dots mark the positions of the corresponding positive control phenotypes. (D) Zac binding to the reverse repeat element (G₂N₆G₄)₂ depends on dimerization through zinc finger 2 (solid arrow). Transactivation is largely concealed, which correlates with a transition in the amino acid base contacts of key residues within the α-helices of zinc fingers 6 and 7. Note that the amino acid base contacts of zinc fingers 2 to 4 are conserved between C and D, arguing against a direct role in regulation of transactivation.

Zac transactivation.

We used the heterologous Gal4 fusion system to map two adjacent structurally distinct domains which conferred synergistically potent transactivation. The zinc finger domain strongly silenced transactivation of the full-length Zac fusion protein, although it did not interfere with DNA binding. This could indicate that the native Zac zinc finger domain exerts an inhibitory effect on cognate transcriptional activation function through conformational restraints and/or heterologous proteins (15). Our results on differential control of Zac transcriptional activities support such a possibility, and additional studies are needed to resolve this issue. Importantly, however, separate expression of each of the candidate domains showed transactivation. Moreover, the absence of either or both of them partly and completely, respectively, prevented Zac transactivation either in the heterologous Gal4 system or on the palindromic DNA element.

The Zac proline repeat domain contains a characteristic structural motif which consists of proline in conjunction with an acidic residue. Acidic residues are crucial to the interaction of acidic activators with their target(s), though they do not depend on the precise structural complementarity usually associated with the formation of specific protein-DNA complexes (9). The amino acid triplet PLE behaved as an autonomous unit, which conferred synergistic and strong enhancement of transactivation upon concatenation. We coined the term PLE minidomain to emphasize the distinction of the general concept of the modular nature of transactivation domains. The PLE minidomain was independent of the order of prolines but strictly required the negatively charged glutamic acid. Recent evidence suggests that acidic and hydrophobic residues play an important role in acidic transactivation domains, though the position of these residues appears to be generally unimportant (26). The PLE minidomain exemplified both principles at the level of a simple triplet motif.

Differential control of transcriptional activities.

It has been increasingly recognized that although some factors are pure activators or pure repressors, transcriptional regulators may also be modified in an allosteric manner by response elements themselves to generate the pattern of regulation that is appropriate to an individual gene (reviewed in references 14 and 15). Zac bound with high affinity to a single half-site of the direct repeat element and potently repressed constitutively active promoters containing this site. Because a Zac protein lacking its C-terminal repressor domain remained transcriptionally inactive when bound under this condition (A. Hoffmann and D. Spengler, unpublished observations), we additionally propose that Zac transactivation was not simply inhibited by repression but was actively concealed. In this regard, the nature of the DNA sequence to which Zac bound seemed to regulate the transcriptional response, as has been reported previously for steroid and thyroid hormone receptors (15). Distinct Zac conformations could promote selective interactions with the transcriptional machinery, leading to either repression or transactivation. Since deletion of the repressor domain did not restore transactivation of half-site-bound Zac and since dimerization was not sufficient for transactivation (see below), various conformational transitions could exist depending on each DNA element, and further studies are necessary to resolve this issue.

We identified here two related high-affinity DNA-binding sites that operated through different mechanisms to control Zac transcriptional activities. Zinc fingers 6 and 7 were necessary and sufficient to recognize the palindromic DNA element and to confer transactivation. Single-amino-acid base contacts involved all potential key residues of zinc fingers 6 and 7 and were indistinguishably required for DNA binding and transactivation, in contrast to their differential regulatory role on the repeat elements (Fig. 7). These data, together with the orientation-dependent transactivation, strongly support the idea that Zac binds as a monomer to the palindromic DNA-binding site.

Zac dimerization on the different repeat elements conferred transactivation, which strictly correlated with the pattern of amino acid base contacts by key residues within the α-helices of zinc finger 6 and, more importantly, zinc finger 7. Position 6 within the α-helix of zinc finger 7 played a key role in distinguishing contacts necessary for DNA binding from those necessary for transactivation (Fig. 7). In the case of the reverse repeats (G₃N₆G₄)₂ and (G₂N₆G₄)₂, we observed a transition in the function of these amino acid residues which tightly correlated with a parallel decrease in transactivation. The positive control phenotype of these residues was most evidently disclosed by the direct repeat element (G₄N₆G₄)₂ and suggested that a surface of the DNA-binding domain may be directly involved in transcriptional regulation, by regulating either protein-protein contacts, DNA structure, or a combination of these mechanisms. These positive control residues play a critical role in interpreting and signaling the information provided by the response element at which Zac binds to its transactivation domain.

In the well-characterized C₂H₂ zinc finger motif, the recognition helix binds in an antiparallel manner to DNA, with the N terminus directed into the major groove and the C terminus directed toward the surrounding environment. The positive control subregion itself or a region affected by these changes could define a site that interacts with other proteins in transcriptional regulation. To our knowledge, this mode of regulation is unprecedented among transactivators, since key residues within the recognition helix alternate between base contacts necessary for transactivation and DNA binding, and specific zinc finger dimer conformations may preclude or enhance recruitment of factors necessary to transactivation. In support of this view, specific dimer configurations of the DNA-binding domain of the transcription factor Oct-1 have recently been reported to differentially synergies with the coactivator OBF-1 (Oct-binding factor 1) (25). These observations strongly emphasize the connection between DNA sequence, protein conformation, multimerization state, and transcriptional activity, although the dimer interface appeared to play no direct role in Zac transcriptional activation (see below).

Importantly, we demonstrated that Zac differentially regulates gene expression in vivo, as exemplified by the family of cytokeratin genes. We identified both classes of DNA-binding sites within the distal CK14 promoter which specifically conferred transactivation by Zac and further supported the differential role of Zac zinc finger contacts in DNA binding and transactivation in the regulation of a natural promoter. Numerous reports employed the common DNA-binding site selection assay to identify DNA elements for transcription factors from distinct families. In many cases, in vitro selection led to the isolation of optimal binding sites, whereas naturally occurring response elements displayed nucleotide divergence that lowered binding affinity and/or transcriptional activities. For instance, all steroid receptor-specific elements characterized from natural promoters are of lower affinity than the canonical binding site (18). Therefore, not surprisingly, the Zac DNA-binding sites identified in vivo in the CK14 gene promoter conferred less transactivation than the in vitro-selected optimized sites due to differences in symmetry. Importantly, however, mutation of the structural core of these natural Zac elements was poorly tolerated, if at all (Fig. 6G).

Zac dimerization and transcriptional regulation.

Zac interacted through amino acid residues localized in the first H-C link and the β1 strand of zinc finger 2. Protein-protein interactions can be mediated via C₂H₂ zinc fingers (17), and genetic selection experiments to search libraries for peptides identified sequences as short as six residues to support in vivo dimerization on DNA (31). Relatively weak interactions that are not sufficient to give stable dimers in solution can still dramatically stabilize the corresponding protein-DNA complexes (29). Under these conditions, both the peptide and the surface that it recognizes are present at high local concentrations when these proteins bind to adjacent DNA sites. Concordant with this view, our coimmunoprecipitation experiments showed no Zac interaction. More importantly, we obtained concurrent evidence for Zac dimerization by electrophoretic mobility shift assays, GST pulldown experiments, and the two-hybrid system.

Zac dimerization was dispensable for DNA binding to the half-site element G₄N₆G₄₂, which conferred transcriptional repression. In contrast, cooperative DNA binding at the direct repeat element involved dimerization and conferred potent transactivation. Moreover, Zac dimerization was prerequisite for DNA binding to the reverse repeat elements (G₂N₆G₄)₂ and (G₃N₆G₄)₂, although it was not sufficient to confer robust transactivation. Based on these findings, we conclude that Zac DNA binding and dimerization mutually depended on each other to unlock transactivation. Additional analysis of the reverse and direct repeat elements with Zac proteins containing α-helical mutations within zinc fingers 2, 3, and 4 showed similar and mostly parallel reductions in DNA binding and transactivation (A. Hoffmann and D. Spengler, unpublished observations). Therefore, the zinc fingers engaged in Zac dimerization appeared to play no direct role in transcriptional regulation, in contrast to zinc fingers 6 and 7 (Fig. 7).

DNA binding distinguishes between Zac biological activities.

Our results strongly support a role for Zac as a transcription factor in addition to recent reports pointing to a role as coactivator and corepressor (11, 12, 23). Numerous nuclear receptor-interacting proteins have been identified in the past few years (22) and raised questions about the definition of a coactivator. A real coactivator first must interact directly and/or indirectly with the activation domain of a nuclear receptor in an agonist-dependent manner, leading to enhancement of the receptor activation function, and second must not enhance the basal transcriptional activity on its own, although it contains an autonomous activation function. Our present work can potentially shed new light on Zac's regulatory function for nuclear receptors (12), and additional investigations to understand the mode of these interactions in greater detail are warranted. Regardless of whether Zac is eventually found to have a dual role as a transcription factor and coactivator, our present report emphasizes the possibility that the identification of Zac-dependent target genes could advance our insight into its role in proliferation and metabolism and may help to elucidate the consequences of altered Zac/ZAC expression in neoplasia and diabetes.

The existence of two classes of Zac DNA-binding sites was anticipated from our recent studies (5). Here, we identified ZACΔ2, an alternatively spliced variant of human ZAC lacking the sequence encoding the two N-terminal zinc fingers. Interestingly, although both proteins were equally potent in antiproliferation, their activities ensued from a differential regulation of apoptosis versus cell cycle progression, since ZACΔ2 was more efficient at induction of cell cycle arrest than ZAC, whereas the reverse was the case for the induction of apoptosis. We obtained similar data for tetracycline-regulated expression of Zac and ZacΔZF1-5 proteins (A. Hoffmann and D. Spengler, unpublished observations). Based on these results, we had previously predicted (5) the possibility of ZAC DNA binding to elements requiring zinc finger 1 and/or 2, as illustrated now by the repeat motifs.

Collectively, we propose that Zac binding to genes containing the repeat element in their regulatory region preferably induces apoptosis, whereas those containing the palindromic element preferably control G₁ arrest (Fig. 8). It is noteworthy that the ratio of ZAC to ZACΔ2 mRNA varies among individuals (5), and further studies will be necessary to determine whether this heterogeneity is due to polymorphic variations among human populations or whether it is linked to different physiological conditions between individuals at the time the samples were collected. In either case, alternative splicing could serve to distinguish between ZAC antiproliferative activities due to differential DNA binding and might contribute to ZAC-related pathologies.

FIG. 8.

Model of Zac biological activities based on differential DNA binding. Zac binding to the palindromic DNA element preferably induces G₁ arrest rather than apoptosis, as exemplified by the human splice variant ZACΔ2, which lacks zinc fingers 1 and 2 (right) (5). In contrast, Zac binding to the repeat motif preferably promotes apoptosis (middle). Zac-mediated repression due to half-site binding enhances apoptosis (left) (A. Hoffmann and D. Spengler, unpublished observations).