The Structure of GABPα/β: An ETS Domain- Ankyrin Repeat Heterodimer Bound to DNA

See allHide authors and affiliations

Science  13 Feb 1998:
Vol. 279, Issue 5353, pp. 1037-1041
DOI: 10.1126/science.279.5353.1037


GA-binding protein (GABP) is a transcriptional regulator composed of two structurally dissimilar subunits. The α subunit contains a DNA-binding domain that is a member of the ETS family, whereas the β subunit contains a series of ankyrin repeats. The crystal structure of a ternary complex containing a GABPα/β ETS domain–ankyrin repeat heterodimer bound to DNA was determined at 2.15 angstrom resolution. The structure shows how an ETS domain protein can recruit a partner protein using both the ETS domain and a carboxyl-terminal extension and provides a view of an extensive protein-protein interface formed by a set of ankyrin repeats. The structure also reveals how the GABPα ETS domain binds to its core GGA DNA-recognition motif.

Gene expression in eukaryotes is frequently mediated by multiprotein complexes that bind DNA in a sequence-specific manner. This type of transcriptional regulation, termed combinatorial control, is a hallmark of gene regulation in eukaryotic cells. The multiprotein complexes that control eukaryotic gene expression may be composed of structurally similar proteins, such as the Fos and Jun bZIP heterodimer (1) or the yeast MATa1/MATα2 homeodomain complex (2). In many other cases, genes are regulated by complexes composed of proteins from different structural families. Examples include the complex formed by the MATα2 homeodomain protein with the MCM1 MADS box protein (3), and by the herpes simplex VP16 transactivator protein with both the Oct-1 POU-domain protein and host cell factor (4). Understanding how transcriptional regulators recruit their partners to form tight, highly specific complexes is central to an understanding of combinatorial control of transcription.

ETS domain proteins make up a large family of DNA-binding proteins found in organisms ranging from fruit flies to humans that play a role in a variety of developmental pathways, in oncogenesis, and in viral gene expression (5). These proteins have in common a conserved DNA-binding domain whose structure, as determined for the ETS proteins Fli-1, Ets-1, and PU.1, has an overall topology similar to that of the “winged helix-turn-helix” family of proteins (6-9). ETS domains bind DNA as monomers and recognize a consensus sequence that contains a core GGA motif. In many cases, greater DNA target specificity is achieved by the cooperative binding of ETS family members with partner proteins (5). For example, the related ETS proteins Elk-1, SAP-1, and SAP-2 interact with the serum response factor at the serum response element in the c-Fos promoter (10), and the ETS protein PU.1 interacts with Pip on several immunoglobulin light-chain enhancers (11).

GA-binding protein (GABP) is a cellular heteromeric DNA-binding protein involved in the activation of nuclear genes encoding mitochondrial proteins (12), adenovirus early genes (13), and herpes simplex virus immediate-early genes (14). The GABP complex is composed of two subunits: an ETS family member, GABPα, and an ankyrin repeat–containing protein, GABPβ (13, 15). Ankyrin repeats, typically 33 amino acids in length, occur in multiple copies in a functionally diverse array of proteins that includes the yeast cell cycle control proteins cdc10/SWI6; the Notch transmembrane protein ofDrosophila melanogaster; the erythrocyte membrane-associated protein, ankyrin; and IκB, an inhibitor of the transcription factor NF-κB (16). GABPβ contains four-and-a-half ankyrin repeats at its NH2-terminus that mediate heterodimerization with GABPα (17). Formation of the GABPα/β heterodimer requires both the GABPα ETS domain and 31 amino acids immediately COOH-terminal to the ETS domain (17). The GABPα/β heterodimer binds to DNA sequences containing a core GGA motif with greater affinity than the GABPα subunit alone (17,18). Two GABPα/β heterodimers associate via the COOH-terminal residues of GABPβ, resulting in a heterotetramer that binds to DNA sequences containing two tandem repeats of the GGA motif (19).

To investigate how the structurally dissimilar GABP α and β subunits form a tight heterodimer with enhanced DNA-binding affinity, we determined the crystal structure of the GABPα/β ETS domain–ankyrin repeat heterodimer bound to DNA. Recombinant fragments of mouse GABPα and GABPβ were expressed in Escherichia coli, purified as a heterodimer, and crystallized bound to a 21–base pair (bp) DNA fragment (20). The structure was solved to 2.15 Å by a combination of multiple isomorphous replacement (MIR) and multiwavelength anomalous dispersion (MAD) methods (Table1). The model of the GABPα/β- DNA ternary complex presented here contains residues 320 to 429 of GABPα, residues 5 to 157 of GABPβ, and all 21 bp of the DNA.

Table 1

Crystals of the GABPα/β-DNA complex formed in space group C2 with unit cell dimensionsa = 201 Å, b = 34.6 Å, c = 59.5 Å, β = 99.9°, with a single ternary complex in the crystallographic asymmetric unit and a solvent content of 50%. DNA derivatives were synthesized containing five bromine atoms (5Br) and one (1I) or two (2I.1, 2I.2) iodine atoms. The mercury derivative (2Hg) was prepared by soaking crystals in EMTS (20). Diffraction data sets were collected with an R-axis IIc detector and Cu Kα. A MAD data set was collected at beamline X-4A of the National Synchrotron Light Source at Brookhaven National Laboratory. Diffraction images were processed with DENZO and SCALEPACK (35). Data sets were scaled with SCALEIT, phases calculated with MLPHARE, and solvent-flattening carried out with DM (36). For the MAD data sets, phases were determined independently and combined with MIR phases with SIGMAA (36). Inclusion of MAD phases resulted in a slight improvement in the 2.8 Å experimental map. An initial model was built with O (37). The model was refined with X-PLOR (23) to 2.15 Å against the 5Br (λ1) data set, incorporating constrained individual B factor and anisotropic overall B factor refinements. Ten percent of the data were excluded from refinement calculations forR free determination. Refined Bfactors for the DNA were relatively high in regions not contacting GABPα. It was therefore necessary to constrain DNA backbone torsion angles at certain residues at nonprotein contacted positions. The model was confirmed with simulated-annealing omit maps. rms, root mean square.

View this table:

An overview of the complex (Fig. 1) shows GABPα interacting with both the 21-bp DNA fragment and GABPβ. The DNA is B form with a slight curve in the region of contact with GABPα. Although GABPβ lies close to the DNA, there are no direct GABPβ-DNA interactions. The GABPα ETS domain contains four antiparallel β sheets that pack against three α helices and is essentially identical in topology to the nuclear magnetic resonance (NMR) structures of Fli-1 (6) and Ets-1 (8,9) and the crystal structure of PU.1 (7). The additional 31 COOH-terminal residues, required for interaction with GABPβ, form a short α helix (helix 4) connected by a turn to a second α helix (helix 5) that extends away from the GABPα-DNA interface (Figs. 1 and 2).

Figure 1

The structure of the GABPα/β-DNA ternary complex. Ribbon diagrams of GABPα (gold) and GABPβ (green) are shown together with a stick model of the DNA (gray). This figure as well as Figs. 3B, 4, and 5 were prepared with SETOR (33).

Figure 2

Sequence of the GABPα ETS domain and COOH-terminal helices. The sequence of GABPα is shown together with the sequences of ETS proteins Ets-1, Fli-1, and PU.1. The numbers across the top refer to the GABPα sequence, with the numbers of the initial residue given for other sequences. Highly conserved residues in all ETS domains are colored red. Residues that are identical COOH-terminal to the ETS domain in GABPα and Ets-1 are colored blue. Secondary-structure elements are indicated by cylinders (α helices) and arrows (β sheets). The third helix terminates with a short region of 310 helix. Amino acid side chains that hydrogen bond to the DNA (Δ) as well as residues at the GABPβ interface (‡) are indicated. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

The GABPβ NH2-terminal domain consists of four-and-a-half ankyrin repeats arranged in tandem. The overall structure of the ankyrin repeats is very similar to that reported for 53BP2 and p19Ink4d (21, 22). Each ankyrin repeat consists of a pair of α helices that form an antiparallel coiled-coil, followed by an extended loop that lies perpendicular to the helices and contains a type I β turn at its tip. Adjacent loops are held together by a series of side-chain and main-chain hydrogen bonds (21). Coiled-coils from neighboring repeats associate via hydrophobic interactions to form a four-helix bundle. The helix adjacent to the loop presents mostly short residues to the helix bundle, whereas the residues from the  helix on the face opposite to the loop are longer (not shown). This asymmetry gives rise to a distinct curvature in the packing arrangement of adjacent ankyrin-repeat helices. As a result, the tips of the loops form a concave surface that curves around one side of GABPα (Fig.3A).

Figure 3

The interaction interface between GABPα and GABPβ. (A) The ankyrin-repeat loops of GABPβ insert into a depression on the side of GABPα. The molecular surface of GABPα is shown together with backbone traces of GABPα (gold) or GABPβ (green). This figure was prepared with GRASP (34). (B) Residues at the GABP α/β interface. Buried water molecules are shown as red spheres. α1, α4, and α5 indicate the numbering of the helices in GABPα (Fig. 2).

The ankyrin repeats in GABPβ contact GABPα by inserting the tip of each loop into a depression in the α subunit that lies between the first helix of the ETS domain and the two COOH-terminal helices (Fig. 3A). Residues in the NH2- and COOH-terminal helices of GABPα, as well as in the loop joining ETS domain helix 3 to β strand 3, form direct contacts with the β subunit (Figs. 2 and 3B). The interface between GABPα and β has a total buried surface area of 1600 Å2 (23) and consists primarily of complementary hydrophobic surfaces with several direct and water- mediated hydrogen bonds that presumably add specificity to the interaction. It is principally the two residues at the tip of each ankyrin repeat loop that interact with GABPα (Fig. 3B). Additional contacts with GABPα are mediated by GABPβ residues adjacent to the tips of the loops and in the ankyrin-repeat helices (Fig. 3B). A comparison with the 53BP2-p53 complex structure shows that, whereas the interaction between 53BP2 and p53 is mediated largely by the Src homology 3 (SH3) domain of 53BP2, the one region of contact with the ankyrin repeats is with the tip of the COOH-terminal ankyrin repeat loop (21). The mediation of protein-protein interactions by the tips of the loops in both the GABPβ and 53BP2 structures suggests that the tips of ankyrin-repeat loops may form the principal protein-protein interaction surface of other ankyrin-repeat proteins.

The extensive involvement of the COOH-terminal extension of GABPα in the interaction interface with GABPβ explains why this portion of GABPα is required along with the ETS domain to obtain heterodimerization (17). The GABPβ interface is highly specific for GABPα because even Ets-1, which has a similar COOH-terminal extension (Fig. 2), fails to form a complex with GABPβ (24, 25). Although nearly all residues at the GABPα interface are conserved in Ets-1, the few differences would give rise to less-than-optimal contacts between Ets-1 and GABPβ (26). However, the conservation of most of the GABPβ-interacting residues in Ets-1 raises the possibility that Ets-1 may bind to an as yet unidentified GABPβ-like protein.

GABPα contacts DNA bases with amino acid side chains in helix 3, which lies in the major groove of the DNA and forms both direct and water-mediated hydrogen bonds with the sequence GGAA at the center of the GABPα recognition motif (Fig. 4A). The DNA bends toward the recognition helix with an overall curvature of 18° (27), thereby maximizing the region of contact with GABPα. The arginine residues that are conserved among ETS family members, Arg379 and Arg376 (Fig. 2), participate in bidentate hydrogen bonding with the guanines in the core GGA motif (G8 and G9, respectively) (Fig. 4A). These contacts readily explain why the two guanine residues are essential for most ETS proteins to bind to DNA (5). The preference for an adenine residue in the core GGA motif may arise from favorable interactions mediated by the methyl group of the thymine base on the opposing strand. The methyl group of thymine 34 fills a pocket at the surface of GABPα that results in a series of favorable van der Waals contacts with side-chain and main-chain atoms in Lys373 and Arg376 (not shown). GABPα makes DNA backbone contacts 3′ of the GGA recognition motif with helices 1 and 2 and contacts phosphates 5′ of the GGA motif with β strands 3 and 4 (Fig. 2, not shown). The observed phosphate contacts are consistent with ethylation interference footprinting of GABPα/β on DNA (28).

Figure 4

GABPα-DNA interactions. (A) GABPα-DNA interactions in the major groove. Only bases making hydrogen bond contacts with GABPα are shown. (B) Interactions between the recognition helix of PU.1 and the major groove (7). (C) Indirect GABPβ-DNA interaction mediated by Lys69 of GABPβ and Gln321 of GABPα.

The contacts formed by GABPα with DNA differ in some notable respects from those observed in the x-ray crystal structure of the minimal ETS domain of PU.1 bound to DNA (7). PU.1 is one of the most evolutionarily divergent members the ETS family (Fig. 2), and some of the PU.1-DNA–contacting residues are not conserved in GABPα. Nevertheless, most DNA backbone contacts observed for GABPα are identical in the PU.1 structure (Fig. 2, not shown). It is therefore surprising that a comparison of the two complexes reveals significant differences in the positions of the respective ETS domain recognition helices relative to the bases in the major groove. A 1.5 Å shift in the relation of the bases to the protein in the PU.1-DNA structure makes possible only a single hydrogen bond between Arg235and guanine 8 (Fig. 4B). In addition, Arg232 in the PU.1 complex is positioned such that the NH1 atom is between the guanine and adenine residues of the GGA motif. This results in a hydrogen bond with guanine 9 (O6) and a close contact with adenine 10 (N6). Although these differences are surprising, they are partly consistent with the observation that PU.1 is unusual amongst ETS proteins in that it is able to recognize an AGA motif in addition to the GGA core motif (29).

The structures that have been reported for PU.1 (7) and Ets-1 (8) complexed with DNA contain only the minimal ETS domain. However, the structure of uncomplexed Ets-1 containing the COOH-terminal extension that is conserved in both GABPα and Ets-1 has been determined by solution NMR (9). Comparison of the backbone topology of GABPα with Ets-1 reveals a marked difference between the two proteins in the positioning of the COOH-terminal helix relative to the ETS domain (Fig. 5). In the GABPα/β complex, helix 5 of GABPα forms only a few contacts with the ETS domain and extends away from the protein-DNA interface (Fig. 1). In contrast, the structure of uncomplexed Ets-1 shows that the corresponding COOH-terminal extension points in the opposite direction, allowing the COOH-terminal helix to pack against helix 1 of the ETS domain. This packing buries hydrophobic residues in both the COOH-terminal tail and the ETS domain of Ets-1. These residues correspond to hydrophobic residues in GABPα that are contacted by GABPβ. The fact that these residues are conserved in both Ets-1 and GABPα (Fig. 2) suggests that, in the absence of GABPβ, the COOH-terminal extension of GABPα may also pack against the ETS domain in the manner observed for Ets-1. Such a shift in structure would result in helix 5 of GABPα pointing toward, rather than away from, the region of contact with the DNA where it could interfere with DNA binding. It is therefore possible that GABPβ may augment the binding of GABPα to DNA by reorientating helix 5 of GABPα.

Figure 5

Comparison of the GABPα crystal structure and Ets-1 NMR structure (9). Each structure is positioned with its ETS domain in the same orientation.

The structure of GABPα/β bound to DNA suggests another possible mechanism for stabilization of the GABPα-DNA complex by GABPβ (17, 18). Gln321 of GABPα hydrogen bonds to Lys69 of GABPβ and to the DNA sugar-phosphate backbone (Fig. 4C). This indirect contact between GABPβ and the DNA, in addition to the reorientation of the COOH-terminal helix of GABPα, may explain the slower rate of dissociation from DNA observed for GABPα/β as compared to GABPα alone (17,18).

Full-length Ets-1 binds to DNA less efficiently than truncated Ets-1 variants lacking residues on either side of the ETS domain. This inhibition of DNA binding is a consequence of the association of the COOH-terminal residues with a helix-containing domain that lies NH2-terminal to the ETS domain (30,31). Whereas the COOH-terminal extension is conserved in both Ets-1 and GABPα, the NH2-terminal helix domain is not. However, comparison of the GABPα/β and Ets-1 structures is highly suggestive of a mechanism by which an auxiliary factor could relieve the inhibition of DNA binding observed for Ets-1. Binding of an auxiliary factor in a manner analogous to GABPβ would result in a shift in the COOH-terminal helix of Ets-1, thereby disrupting the interaction with the NH2-terminal domain and relieving inhibition of DNA binding.

Complex formation between the GABPα ETS domain protein and the GABPβ ankyrin-repeat protein is just one of a number of known examples of ternary complexes involving ETS domain proteins and structurally different partners. Biochemical evidence would suggest that other ETS proteins have evolved different mechanisms by which they recruit other proteins (10, 32), which may be a consequence of the structural diversity of their protein partners. The GABPα/β complex also reveals a mechanism by which a series of ankyrin repeats can interact with their partner protein. The modular manner by which successive loops interact with GABPα is highly suggestive of an evolutionary process in which the duplication of a repeat followed by mutation of the residues at the tips of the loops would allow for the progressive alteration of the specificity or affinity of an interaction. The facility by which an ankyrin repeat domain could evolve new protein target specificities may explain why ankyrin repeats are frequently involved in protein-protein interactions.

  • * To whom correspondence should be addressed. E-mail: cynthia{at}


View Abstract

Navigate This Article