Structural Basis for Ligand-Regulated Oligomerization of AraC

See allHide authors and affiliations

Science  18 Apr 1997:
Vol. 276, Issue 5311, pp. 421-425
DOI: 10.1126/science.276.5311.421


The crystal structure of the arabinose-binding and dimerization domain of the Escherchia coli gene regulatory protein AraC was determined in the presence and absence of l-arabinose. The 1.5 angstrom structure of the arabinose-bound molecule shows that the protein adopts an unusual fold, binding sugar within a β barrel and completely burying the arabinose with the amino-terminal arm of the protein. Dimer contacts in the presence of arabinose are mediated by an antiparallel coiled-coil. In the 2.8 angstrom structure of the uncomplexed protein, the amino-terminal arm is disordered, uncovering the sugar-binding pocket and allowing it to serve as an oligomerization interface. The ligand-gated oligomerization as seen in AraC provides the basis of a plausible mechanism for modulating the protein’s DNA-looping properties.

AraC is a regulator of transcription that changes the way in which it binds DNA when the protein forms a complex with its monosaccharide ligand, l-arabinose. In E. coli, the AraC protein controls expression of genes necessary for uptake and catabolism of arabinose (1). In the absence of arabinose, a single AraC dimer contacts the two widely separatedI 1 and O 2 half-sites, forming a 210–base pair DNA loop and repressing transcription from thep BAD and p C promoters (2-5) (Fig. 1). The binding of arabinose to AraC causes the protein to cease DNA looping and favor binding to the adjacent I 1 and I 2half-sites (4, 6), resulting in activation of transcription from the p BAD promoter (7). DNA looping, which was discovered in the arabinose operon (2), plays an important role in both eukaryotic and prokaryotic transcription and has been shown to be a mechanism for allowing control elements distally located from the start site of transcription to participate in gene regulation (8).

Figure 1

Regulation of the araBAD operon by the AraC protein. Relative to the start site of transcription, theI 1 and I 2 half-sites are positioned at −35, the O 1 site at −125, and the O 2 site at −285. Binding of the CAP protein is not essential, but is required to achieve the maximum level of transcriptional activation from the araBAD promoter. N, NH2-terminal arabinose-binding and dimerization domain of AraC; C, COOH-terminal DNA-binding domain.

The 292-residue AraC protein consists of an NH2-terminal domain that binds arabinose and mediates dimerization (residues 1 to 170) joined to a COOH-terminal DNA-binding domain (residues 178 to 292) (9, 10) by a linker of at least five residues (10-12). The two domains of AraC appear to be functionally independent, as they retain their respective activities when fused to other DNA-binding or dimerization domains (9). At protein concentrations near those found in the cell, AraC is a dimer in solution both in the presence and absence of arabinose (13). It is also a single AraC dimer that binds to theI1-I 2 site in the presence of arabinose and to the widely separatedI 1 and O 2 sites in the absence of arabinose (4, 5). DNA-binding and footprinting experiments indicate that the binding of arabinose restricts the freedom of the DNA-binding domains of AraC to contact nonadjacent DNA half-sites, probably by reducing the distance that the DNA-binding domains can be separated (14). The binding of arabinose to AraC must therefore alter the protein dimer in a way that changes its preference for binding site spacing.

To learn the basis for the pronounced effect of arabinose on the dimerization domain of AraC, we have determined its structure in the presence and absence of arabinose. Crystals of the dimerization domain of AraC complexed with l-arabinose were prepared from a tryptic fragment of the protein consisting of the NH2-terminal domain of AraC (residues 2 to 178) (15). The crystals form in space groupP21 with unit cell dimensionsa = 39.75 Å, b = 93.84 Å,c = 50.33 Å, β = 95.62o and contain two monomers in the asymmetric unit. The structure of the AraC-arabinose complex was determined at 1.5 Å resolution (Table 1) and contains residues 7 to 167 of monomer A and residues 7 to 170 of monomer B, two molecules of l-arabinose, and 412 water molecules. The structure of the unliganded AraC dimerization domain was determined at 2.8 Å resolution (Table 1) from crystals that form in space group P3121 with unit cell dimensionsa = b = 57.37 Å, c = 83.45 Å (16) and contain one monomer in the asymmetric unit. The model of the unliganded dimerization domain contains residues 19 to 167 of AraC.

Table 1

A native data set from the arabinose-bound crystals was collected to a resolution of 1.8 Å on an RAXIS-IIC detector with CuKα radiation. All data were processed and reduced with DENZO and SCALEPACK (26). A mercury derivative with two atoms per asymmetric unit was prepared by soaking crystals in stabilizing solution containing 0.9 M ethylmercurithiosalicylate for 3 days. These crystals were transferred to the standard cryosolution for 20 min and flash-frozen (15). A complete derivative data set was collected at beamline X4A of the National Synchrotron Light Source at Brookhaven National Laboratories (NSLS, Upton, New York) with the use of a Fuji BAS-2000 image plate detection system. Data were collected at a wavelength of 1.006 Å, just above the mercury LIIIabsorption edge, in order to maximize the anomalous scattering from the mercury atoms in the crystal. These data were scaled to the native data set with SCALEIT (27), and single isomorphous replacement plus anomalous scattering (SIRAS) phases were calculated with MLPHARE (27). The initial phases were improved further by applying solvent flattening and histogram matching with DM (27), resulting in a high-quality 1.8 Å electron density map. The model was built with the interactive graphics program O (28), and refinement proceeded with X-PLOR (29) through use of a 1.5 Å native data set recorded at the NSLS and standard refinement protocols (30). The structure of the unliganded AraC sugar-binding and dimerization domain was determined by molecular replacement with AMoRe with data collected at beamline X4A (31). One monomer of the plus arabinose AraC structure with the sugar and waters deleted was used as a search model. After rigid-body refinement of the structure, manual corrections of the model were made, including deletion of the NH2-terminal 12 residues. The structure was then refined by torsion angle dynamics refinement (32, 33) and conventional conjugate gradient positional refinement in X-PLOR. Both the arabinose-bound and unliganded AraC models show good stereochemistry and no Ramachandran plot outliers.

View this table:

The structure of AraC complexed with arabinose shows that the fold of the sugar-binding and dimerization domain of AraC contains an eight-stranded antiparallel β barrel (β1 to β8) with jelly-roll topology (Fig. 2A). The β barrel is followed by a long linker that contains two turns of 310 helix, followed by a ninth β strand (β9) that forms part of one sheet of the β barrel. The last β strand is followed by two α helices (α1 and α2), each approximately 20 amino acids in length, that pack against the outer surface of the barrel. One molecule of l-arabinose binds in a pocket within the β barrel (Fig. 2, A and B). Residues 7 to 18 of AraC constitute an NH2-terminal arm that lies across the sugar-binding pocket, fully enclosing the arabinose molecule within the β-barrel domain (Fig. 2A). The overall fold of the NH2-terminal domain of AraC and the manner in which it binds ligand has not, to our knowledge, been previously observed. Residues 2 to 6 are not visible at the NH2-terminus of either monomer in the asymmetric unit and are presumably disordered, as are the COOH-terminal residues 168 to 178 of monomer A and residues 171 to 178 of monomer B. The disorder at the COOH-terminus is in agreement with previous experiments indicating that these residues are part of a flexible and mutatable linker that joins the sugar-binding and dimerization domain to the DNA-binding domain (9,10, 12).

Figure 2

The AraC sugar-binding and dimerization domain monomer. (A) Monomer in the presence of arabinose. The protein is shown as a blue ribbon trace, and the sugar as a ball-and-stick model. The NH2- and COOH-termini, as well as secondary-structure elements referred to in the text, are indicated. This figure, as well as Figs. 2, B and D, and 3, A and B, were prepared with the program SETOR (34). (B) The arabinose-binding site in AraC (35).The NH2-terminal arm is shown with the side chains removed as a backbone trace, and waters that mediate hydrogen bonds between the arm, the sugar, and the sugar-binding site are shown as red spheres. Arg38, which makes a bidentate interaction with O4 and O5 of arabinose, is shown in yellow, as is Trp95, on which the arabinose ring stacks. Other side chains involved in binding arabinose are depicted in blue. (C) Hydrogen bond interactions with arabinose. Waters and side chains making direct hydrogen bonds to the sugar are in boxes, whereas those making indirect hydrogen bonds to the sugar are in ovals. Several waters have been omitted from the network of water-mediated hydrogen bonds along the right side of the figure for clarity. (D) Superposition of the AraC structures in the presence (blue) and absence (yellow) of arabinose.

Each monomer of AraC binds one molecule of α-l-arabinose (17) in the full-chair conformation within the open end of the β barrel, occupying a small pocket lined with both polar and nonpolar side chains (Fig. 2, A to C). The sugar stacks against the indole ring of Trp95 and is stabilized by hydrogen bonds formed by side chains and water molecules within the binding pocket with the sugar hydroxyls and ring oxygen (Fig. 2C). The binding site is completed by the NH2-terminal arm of the protein (residues 7 to 18), which loops around to close off the end of the β barrel in which arabinose is bound. The NH2-terminal arm forms both direct and indirect contacts with the sugar, resulting in complete burial of arabinose within the protein (18). The position of the arm is stabilized by the main chain carbonyl of Pro8, which makes a direct hydrogen bond with the anomeric hydroxyl (OH-1) of the bound sugar. An extensive network of bound water molecules in the sugar-binding pocket mediates hydrogen bonds between the sugar and protein main chain atoms of residues Leu9, Leu10, Gly12, Tyr13, and Phe15 in the NH2-terminal arm (Fig. 2, B and C). Because the NH2-terminal arm blocks access to the arabinose-binding pocket, the arm must move aside when arabinose enters or leaves.

The structure of the unliganded AraC monomer is very similar to the structure of the liganded monomer, with residues 19 to 167 of the two structures superimposable with a root-mean-square difference (rmsd) of 0.6 Å for main chain atoms (Fig. 2D). There are, however, several notable differences between the liganded and unliganded molecules. Most prominent of these is that, in the absence of sugar, the NH2-terminal arm (residues 7 to 18) that formerly closed off the opening to the ligand-binding site is disordered. Evidence for the disorder comes both from the absence of electron density corresponding to the NH2-terminal arm and from a packing arrangement in the crystal (described below) that is made possible by the absence of the NH2-terminal arm from the opening of the sugar-binding pocket. In the presence of arabinose, both direct and water-mediated hydrogen bonds to the sugar help stabilize the NH2-terminal arm. It is therefore not surprising that the arm is unstructured in the absence of arabinose.

At intracellular concentrations, the AraC protein exists as a dimer in both the presence and absence of arabinose (13). In the presence of arabinose, the NH2-terminal domain of AraC crystallizes with a dimer in the asymmetric unit (Fig.3A). This arrangement presumably reflects the solution dimer form of the protein, as the crystals contain no other twofold axes relating AraC monomers. The two monomers in the crystal associate by an antiparallel coiled-coil formed between the terminal α helix, helix 2, of each monomer (Fig. 3A). Although a few additional contacts are made between the short 310 helices, most of the 1200 Å2 dimer interface consists of the coiled-coil interaction. Each end of the coiled-coil is anchored by a triad of leucine residues, Leu150 and Leu151 from one monomer, and Leu161 from the second, that pack together in a knobs-into-holes manner. This packing scheme persists along the length of the helices, consistent with the model proposed by Crick for a coiled-coil (19).

Figure 3

Structures of AraC dimers. (A) The dimer observed in the presence of arabinose. A distance of 37 Å separates the COOH-termini of the coiled-coil dimer, which are the attachment points for the DNA-binding domains. (B) The β-barrel dimer that forms only in the absence of arabinose. The Tyr31:Trp95 pair from one monomer is colored pink, and in the other monomer the pair is colored red. The distance between the COOH-termini of the β-barrel dimer is 60 Å.

The two nonpolar leucine triads flank a hydrophilic central core of the coiled-coil consisting of polar side chains that form a network of hydrogen bonds with each other and with a molecule presumed to be a water (20) that is buried at the interface between the two helices. The buried water stabilizes the central portion of the coiled-coil interface by forming four hydrogen bonds, one each to Asn154 and Gln158 from both monomers. Glu157 from each helix further stabilizes Asn154 through a hydrogen bond. Although two-helix antiparallel coiled-coils between two monomers are relatively rare, the coiled-coil seen in the AraC arabinose dimer is remarkably similar to the coiled-coil seen in the MetJ-DNA complex (21). In that case, MetJ dimers bound to adjacent sites on the DNA interact by way of an antiparallel coiled-coil that contains hydrophobic interactions at its ends flanking a hydrophilic core that buries a water molecule.

In the absence of arabinose, the dimerization domain of AraC crystallizes with a monomer in the crystallographic asymmetric unit. The crystal-packing interactions must therefore be examined to identify the solution dimer interface. Each unliganded AraC monomer in the crystal has two neighboring monomers related to it by independent crystallographic twofold axes, giving rise to two potential choices of dimerization interface. One of the two dimer interfaces in the crystals of the unliganded protein reproduces nearly the same coiled-coil interaction seen in the crystals of arabinose-bound AraC. In the unliganded protein, however, a distortion in the top of helix 2 slightly reorients the leucine triad at the ends of the coiled-coil, resulting in an approximate 13° rotation of one monomer relative to the other as compared with the coiled-coil dimer formed by AraC with bound arabinose. The rotation is a hinge-like motion about the coiled-coil axis, giving rise to a slight rearrangement of the packing interactions at the coiled-coil interface.

The second dimer interface in the uncomplexed AraC crystals (Fig. 3B) brings together two β barrels, in a face-to-face manner, and buries 75% more area than the coiled-coil interface, giving a total of 2100 Å2 of buried surface area. This seems likely to be the dimerization interface in the absence of arabinose, as in crystals of virtually all multimeric proteins, the biologically relevant interface is that which buries the most surface area (22). It is the absence of arabinose and the disordering of the NH2-terminal arm that exposes the β-barrel surface, allowing it to function as an oligomerization interface. A striking feature of the β-barrel interface is the insertion of Tyr31 from each monomer into the sugar-binding site of the opposing monomer where the tyrosine side chain takes the place of arabinose and packs against the indole ring of Trp95 (Fig.3B). In the arabinose-bound AraC molecule, Tyr31 is in a slightly different conformation and is solvent exposed. The position of Tyr31 in each arabinose-free monomer is stabilized by a hydrogen bond to Tyr82 from the opposite monomer, as well as by van der Waals interactions with Trp95, Met42, and Ile36. Association of two AraC monomers by way of the β-barrel interface is further stabilized by hydrophobic interactions mediated by Val20, Leu32, Phe34, and Ile46.

The ability of unliganded AraC to engage simultaneously in both the β-barrel and the coiled-coil interactions in the crystal lattice suggests that, at elevated concentrations, both interactions should occur in solution. If they do, the uncomplexed protein should aggregate. In contrast, the sugar-bound form should not aggregate because in the presence of arabinose, the NH2-terminal arms fold over the arabinose-binding pockets and block the β-barrel interaction, leaving the protein capable of coiled-coil dimerization only. We tested this prediction with velocity sedimentation of the AraC NH2-terminal domain at 0.5 mg/ml (Fig. 4). The domain sediments in the absence of arabinose as a range of species with sedimentation coefficients in the range of 3.2 to 6S, consistent with the formation of dimers and higher ordered oligomers. In contrast, the domain in the presence of arabinose behaves as a homogeneous dimer, sedimenting at 2.9 to 3.1S and showing no signs of aggregation. Small-angle x-ray scattering data (23) also lead to the same conclusions.

Figure 4

Sedimentation velocity data for the AraC sugar-binding and dimerization domain in the presence (○) and absence (•) of arabinose. Samples were prepared in 15 mM tris-HCl (pH 7.5) and 75 mM KCl in the presence and absence of 0.2% (w/v)l-arabinose and analyzed at 42,000 rpm in a Beckman XL-A analytical ultracentrifuge. Van Holde–Weischet analysis of the resulting data (36) produced the distribution plot shown.

The structural and solution studies of AraC presented here show that a new oligomerization interface is opened when arabinose is absent from its binding pocket and the NH2-terminal arm of the protein consequently becomes disordered. Association by way of this interface is completely disrupted when arabinose binds within the sugar-binding pocket, causing occlusion of the β-barrel oligomerization interface by displacing the tyrosine residue inserted by the opposing monomer and triggering the folding of the NH2-terminal arm. This mechanism of ligand-regulated oligomerization contrasts with that of another structurally characterized regulatory protein, E. coli arginine repressor (ArgR), whose oligomeric state is also controlled by ligand. In that case, arginine binding directly promotes the association of ArgR trimers to form a hexamer (24).

How might the arabinose-induced change in the ability of AraC to self-associate through the β-barrel interface alter the protein’s DNA-binding properties? An intriguing possibility is that the AraC dimerization interface may change in response to the presence or absence of arabinose. By this mechanism, the protein would dimerize by the antiparallel coiled-coil in the presence of arabinose and, because AraC always functions as a dimer (4, 5,13), by the more extensive β-barrel interface in the absence of arabinose.

In the β-barrel dimer, the DNA-binding domain attachment points to the COOH-terminus of the dimerization domain are 60 Å apart (Fig.3A), whereas in the coiled-coil dimer the distance is 37 Å (Fig. 3B). The reduction in distance could make DNA looping energetically unfavorable in the presence of arabinose and simultaneously increase the affinity of the protein for adjacently located DNA half-sites, as is observed experimentally (6, 14,25). The shorter distance between the DNA-binding domains in the liganded versus unliganded protein would also account for the observation that AraC in the absence of arabinose can bind to DNA with an additional 10 base pairs inserted between half-sites, as compared with AraC with bound arabinose (14). Unclear at this time is how, in the absence of arabinose, formation of higher ordered oligomers is inhibited because the coiled-coil interface is still available for interactions; however, it is possible that the structural rearrangements at the coiled-coil interface observed in the absence of arabinose could decrease its affinity.

Dissociation of arabinose from AraC, in addition to exposing a new oligomerization interface, also frees the NH2-terminal arm. Although studies with protein chimeras do not indicate the presence of a significant interaction between the sugar-binding and dimerization domain and DNA-binding domain of AraC (9), it remains possible that the freed NH2-terminal arm may modulate the effective distance or orientation between DNA-binding domains.

The regulation by arabinose of the NH2-terminal arm’s conformation and of the availability of the sugar-binding pocket for protein-protein interactions represents a versatile way to couple ligand binding with changes in conformation and multimeric state. Definitive determination of the mechanism by which the binding of arabinose changes the way AraC interacts with DNA awaits further testing by mutagenesis experiments, solution studies, and additional structural studies of the AraC protein.

  • * To whom correspondence should be addressed. E-mail: bob{at} (R.S) or cynthia{at} (C.W.)


View Abstract

Navigate This Article