GATA-1 bends DNA in a site-independent fashion.

The DNA binding domain of GATA-1 consists of two adjacent homologous zinc fingers, of which only the C-terminal finger binds DNA independently. Solution structure studies have shown that the DNA is bent by about 15 degrees in the complex formed with the single C-terminal finger of GATA-1. The N-terminal finger stabilizes DNA binding at some sites. To determine whether it contributes to DNA bending, we have performed circular permutation DNA bending experiments with a variety of DNA-binding sites recognized by GATA-1. By using a series of full-length GATA-1, double zinc finger, and single C-terminal finger constructs, we show that GATA-1 bends DNA by about 24 degrees, irrespective of the DNA-binding site. We propose that the N- and C-terminal fingers of GATA-1 adopt different orientations when bound to different cognate DNA sites. Furthermore, we characterize circular permutation bending artifacts arising from the reduced gel mobility of the protein-DNA complexes.

3 10, 11). These high affinity DNA sites are usually characterized by double GATA motifs, albeit arranged in diverse orientations and spacing. Since GATA-1 binds as a monomer to such sites, it has been proposed that both the N-and C-terminal fingers are involved in DNA recognition. Among these sites is an overlapping palindromic GATA sequence (ATCTGATA, referred to as GATApal), that is necessary for the activity of at least three vertebrate hematopoietic GATA-1 promoters, and requires both zinc fingers for high affinity interaction (11). Similarly, both zinc fingers of GATA-1 are requisite for the interaction with the double AGATA sites of the γ-globin promoter (7). In the accompanying manuscript we show that both the N-and C-terminal fingers of GATA-1 are necessary for the high affinity recognition of the overlapping AGATA sites of the ε-globin silencer (12). We also present similar results for an alternate binding site that contains a GATC consensus sequence.
Solution NMR studies of the complex formed between the C-terminal finger of GATA-1 and its cognate DNA sequence indicate that the DNA is bent by an overall angle of about 15° (3). This kink probably results from the insertion of the C-terminal basic residues required for DNA binding (6) into the minor groove. Unlike the single C-terminal zinc finger, it has been proposed that the slow migrating complex is indicative of binding through only the C-terminal finger (11). The migration anomalies observed may therefore be due to the introduction of bends into the DNA target by the binding of the N-terminal finger. In order to evaluate the contribution of the N-terminal finger to DNA bending, we have performed gel mobilityshift assays using circular permuted probes containing different double and single GATA binding motifs. Using various GATA-1 constructs we show that the C-terminal finger is the sole contributor to DNA bending and that the bend angle previously reported to be induced by full length GATA-1 was overestimated. Because the bend angle is identical among double GATA sites separated by 0 to 9 base pairs, we propose that the N-and Cterminal fingers of GATA-1 necessarily adopt different relative orientations when bound to distinct cognate double GATA sites.

Constructions.
Bending vectors. Oligonucleotides were synthesized on an Applied Biosystems 394 DNA/RNA synthesizer and purified by denaturing PAGE.
The EcoRV (GATATC) and BglII (AGATCT) restriction sites on pBEND5 (14) were modified into SfcI (CTCGAG) and BsrI (ACCAGT) sites, respectively, to remove potential GATA binding sites. Furthermore, the central SalI (GTCGAC) cloning site was modified into an AatII (GACGTC) site. The modified pBEND5, pB5RG, was prepared sequentially via pB5R. pB5R was prepared by insertion of the modified dsDNA containing the circularly permuted restriction sites into pBEND5 digested with EcoRI and XbaI; pB5RG was prepared by insertion of a similar dsDNA into pB5R digested with XbaI and HindIII ( Figure 1). Bending vectors containing various double and single GATA binding sites (

Expression and purification of GATA-1 proteins.
GST chicken GATA-1. Full-length chicken GATA-1 with an Nterminal glutathione-S-transferase (GST) fusion was prepared as described (16). purification was carried out as described above. Peptides corresponding to the human GATA-1 double finger and chicken GATA-1 C-terminal single finger were prepared as previously described (6,11). Full-length GATA-1 nuclear extracts were made by standard procedures from adult chicken erythrocytes (17).

DNA Bending experiments.
Preparation of the bending probes. Radioactively labeled, circularly permuted probes were prepared by PCR amplification of the EcoRI to HindIII portion of the pB5RG vector and restriction enzyme treatment as described (18). PCR was carried out using CCCGGGCTGCAGGAATTCACG and GACGGTATGCATAAGCTTGGA as forward and reverse primers, respectively.
Restriction digests to yield the circularly permuted products were performed with BsrI, NheI, ClaI, SpeI, DraI, MspI, NruI, KpnI, HinfI and BamHI. The DNA probes were gel purified on 5% native acrylamide gels and adjusted to the same concentration based on their specific activity.

GATA-1 bends DNA in a site independent manner.
Gel mobility-shift assays using circularly permuted probes containing varying GATA binding motifs ( Table 2) were carried out to determine the bend angle introduced by both chicken and human double zinc finger peptides. The GATA binding sites evaluated consisted of a series of double sites separated by 0 (GATApal (G2)), 1 (ε-globin silencer (εS) and GATC consensus site (GC)), 4 (GATA-1 promoter (G3)) or 9 (γ-globin promoter (γP)) base pairs, allowing us to evaluate how the combination of the two zinc fingers may contribute to the overall bend. It has already been shown that both the zinc fingers of GATA-1 interact with DNA when bound to the G2 and γP sites (7,11). In the accompanying manuscript, we show that the same holds true for the εS and GC binding sites (12). In addition to these double GATA sites, two single GATA binding sites were analyzed in order to evaluate the contribution of the C-terminal finger to the overall DNA bend. One site (G1) represents the doubly mutated mouse GATA-1 promoter site, whereas the other (G0) is the consensus site used in obtaining the solution structure of the DNA complex formed with the C-terminal finger (6).
Circular permutation bending experiments carried out with the human and chicken double finger peptides show, within the experimental precision of 12 the method, that the DNA bend induced by 1:1 complex formation is independent of the GATA binding site. The average bend angle of 24 ± 3°o btained for these complexes is identical to the average bending angle obtained with the chicken C-terminal finger peptide (Table 3). These data demonstrate that GATA-1 bends DNA in a site independent manner and that this bend arises solely from the binding of the C-terminal finger of GATA-1 to the DNA (Table 3, sites G1 and G0).

GATA-1 bends DNA by 24°. The bend angle of 24° is significantly
different from the value of 64° published for the full-length chicken GATA-1 (13), and we have just shown that this difference cannot be ascribed to the contribution of the N-terminal finger. To explain this difference we performed a series of circular permutation mobility shift assays with full-length chicken GATA-1 obtained from chicken erythrocytes. Data obtained for the 1:1 complex formed with sites from the mouse GATA-1 promoter (Table 3, site G3, essentially identical to the chicken GATA-1 promoter used in (13)) lead to an angle of 59 ± 4° ( Figure   2), a value identical to that previously published. As in the case of the double and single C-terminal finger peptides, the angle was site independent. Circular permutation experiments were also carried out with 13 an MBP 2 fusion to the human double finger. As in the case of the double finger peptide, the bending angle obtained did not depend on the binding site. Unlike the peptide, however, an average bending angle of 83 ± 3° is noted (Table 3). Similar experiments carried out with a bacterially expressed GST 1 fusion to the full-length chicken GATA-1 lead to a larger bending angle vis-à-vis the chicken GATA-1 (i.e. 104° versus 63°; Table   3). As neither the GST 1 nor the MBP 2 domains interact with DNA, these data indicate that the 63° bend observed with GATA-1 includes a contribution resulting from the decreased migration of the 1:1 complex.
Indeed, a plot of the bending angle (α) as a function of the relative mobility of the complex (R bound /R free ) suggests a direct relation between these parameters ( Figure 3A). In the reptation model describing DNA migration through a gel it is assumed that the presence of a single intrinsic bend imposes a large barrier to the motion of the DNA chain. Furthermore, it is assumed that the elastic force constant, B eff , describing the 'interaction' of the DNA chain with the acrylamide pores does not change within a particular set of probes (20). Measuring the bending angle as a function of the acrylamide concentration readily tests this assumption, which is critical to the 14 derivation of the quadratic equation relating mobility to bending. Bending of the mouse GATA-1 promoter (G3) by the double finger peptide of chicken GATA-1 leads to bending angles that are slightly dependent on the acrylamide concentration in a manner similar to the A-tracts used as controls ( Figure 3B). This slight variation in the angle is therefore an intrinsic DNA property. However, both the full-length GATA-1 and the MBP-chicken double finger lead to bending angles that show a marked dependence on the acrylamide concentration ( Figure 3B). In these cases, the elastic force constant of the complex varies as a function of the circularly permuted probe, leading to an overestimation of the true bend angle. Therefore, together with the observation that the double and single C-terminal zinc finger peptides lead to identical bending angles, we conclude that GATA-1 bends DNA by 24° in a site independent manner.

GATA-1 bends DNA in a site independent manner.
We have shown that the GATA-1 proteins bend DNA by 24° in a site independent manner and that the C-terminal finger is sufficient to account for this bend. The has been documented that circular permutation analysis usually overestimates the bend angle induced by DNA binding proteins (21,22). For example, bending angles of 81° were obtained with this method for the thyroid hormone and retinoid X receptor dimers (TR/TR and TR/RXR) (23).
A large contribution to this bend must be due to anomalous migration as phasing analysis leads to induced bends of about 10° (23). Similar conclusions have been reached in a study of the GCN4-DNA complex, where it was shown that the full-length GCN4 leads to an anomalous circular permutation analysis due to its size (24).
In a comparative study we have shown that the non-bending migration anomaly, due in part to 'trailing' portions of the protein, contributes significantly to the apparent bending angle in the case of the full-length GATA-1, the MBP and GST fusion proteins. ( Figure 3A). Because their apparent bending angles vary significantly as a function of the acrylamide concentration, the angles consequently are not a reflection of the true bend induced.

Implied properties of GATA-1.
In the accompanying manuscript we demonstrate that the N-and C-terminal zinc fingers of GATA-1 influence one another in DNA binding (12). This change is most likely the result of an intramolecular interaction between the GATA fingers and (or) linkers.
Indeed, it has been shown that GATA-1 dimerizes with low affinity on DNA. This occurs through an association between the N-and C-terminal zinc fingers of separate molecules, indicating that the fingers associate specifically with one another (25). For similar interactions to occur intramolecularly, the fingers must be capable of movement relative to one another through the amino acids linking them. This linker region of GATA-  Table 2 are inserted into the XbaI and AatII sites.       Table 3