Structural Basis of Trans-Inhibition in a Molybdate/Tungstate ABC Transporter

See allHide authors and affiliations

Science  11 Jul 2008:
Vol. 321, Issue 5886, pp. 246-250
DOI: 10.1126/science.1156213


Transport across cellular membranes is an essential process that is catalyzed by diverse membrane transport proteins. The turnover rates of certain transporters are inhibited by their substrates in a process termed trans-inhibition, whose structural basis is poorly understood. We present the crystal structure of a molybdate/tungstate ABC transporter (ModBC) from Methanosarcina acetivorans in a trans-inhibited state. The regulatory domains of the nucleotide-binding subunits are in close contact and provide two oxyanion binding pockets at the shared interface. By specifically binding to these pockets, molybdate or tungstate prevent adenosine triphosphatase activity and lock the transporter in an inward-facing conformation, with the catalytic motifs of the nucleotide-binding domains separated. This allosteric effect prevents the transporter from switching between the inward-facing and the outward-facing states, thus interfering with the alternating access and release mechanism.

Active transport proteins consume cellular energy to move substrates across biological membranes against their (electro)chemical gradients. Transport processes are regulated at various stages; this includes genetic regulation of the expression levels or control of the transporters by inhibitory cellular signals. Another mechanism is trans-inhibition, which occurs when substrates exert a concentration-dependent, inhibitory effect on the transporter after the translocation has occurred, that is, on the target side (trans side) of the membrane (1). This type of inhibition results in the decrease of the transport rates as the concentration of substrate increases (2). It is therefore a functional equivalent to product inhibition of soluble enzymes. Trans-inhibition has been reported for ion transporters (3) and various amino acid transporters (48), including glutamate transporters in astrocytes (9). It has also been described for binding protein–dependent ATP-binding cassette (ABC) transporters, such as those specific for methionine [met system in Escherichia coli (10)], glycine/betaine [OpuA from Lactobacillus plantarum or Listeria monocytogenes (11, 12)], or spermidine/putrescine [pot system in E. coli (13)]. Even though the functional effect of trans-inhibition has been known for a long time, its structural basis is not well understood.

We have studied Methanosarcina acetivorans ModBC (MaModBC), a binding protein–dependent ABC transporter specific for molybdate/tungstate. It consists of two transmembrane domains (TMDs, ModB subunits) that form a translocation pathway for the substrate, and two cytoplasmic nucleotide-binding domains (NBDs, ModC subunits) that bind and hydrolyze adenosine triphosphate (ATP) and power the transport reaction. What distinguishes MaModBC from other ABC importers such as the B12 transporter BtuCD (14, 15) or the molybdate/tungstate transporter ModBC from Archaeoglobus fulgidus [AfModBC (16)] is the presence of a regulatory domain appended to the C terminus of the NBDs. This domain has some 120 amino acid residues, and similar extensions have been found in other ABC importers such as the methionine importer. Even though these domains had been suspected to be involved in regulation, their exact function has been somewhat of a mystery in most cases (17). While investigating its adenosine triphosphatase (ATPase) activity, we found that binding of molybdate or tungstate had an inhibitory effect on the ATP hydrolysis rate of the intact transporter MaModBC (Fig. 1A). The ATPase rate of MaModBC was already inhibited at low micromolar concentrations of molybdate (or tungstate), with an apparent inhibitory constant of ∼5 μM. By contrast, the ATPase activity of AfModBC, which does not contain regulatory domains appended to the NBDs, was insensitive even to high concentrations of molybdate or tungstate (Table 1). Because ATP hydrolysis is a strict requirement for transport in ABC transporters, these results suggested that the regulatory domains present in MaModBC, but absent in AfModBC, mediate trans-inhibition.

Fig. 1.

ATPase activity and crystal structure of M. acetivorans ModBC. (A) Relative ATP hydrolysis rates of MaModBC in the presence of the oxyanions molybdate (open circles), tungstate (solid diamonds), and sulfate (solid squares). Only molybdate and tungstate are substrates of MaModBC. (B) Side view of MaModBC in ribbon representation illustrating the arrangement of the protein subunits. The gray box represents the approximate position of the lipid membrane.

Table 1.

Relative ATP hydrolysis rates of A. fulgidus ModBC free or in the presence of various oxyanions.

View this table:

We then proceeded to determine the crystal structure of the complete MaModBC transporter in a trans-inhibited conformation after cocrystallization with bound tungstate. We found that the regulatory domains form two oxyanion binding pockets that bind the substrate and lock the transporter in an inward-facing conformation, with the catalytic motifs of the NBDs separated. The structure of tungstate-bound MaModBC was determined with experimental phases from a three-wavelength MAD (multiwavelength anomalous dispersion) data set collected around the tungsten edge (18). The diffraction data (fig. S1 and table S1) were anisotropic and were truncated to 3.0, 3.3, and 3.5 Å resolution in the three diffraction directions (19). Because of the high accuracy of the MAD phases, which were further improved by solvent flattening and noncrystallographic symmetry averaging, the experimental electron density was of excellent quality (fig. S2) and allowed unambiguous building of the entire protein structure.

The TMDs and NBDs of MaModBC have folds similar to those of the previously determined AfModBC (16). However, the conformations of MaModBC and AfModBC are quite different. Whereas the NBDs of AfModBC have a shared interface, those of MaModBC contact each other exclusively through their regulatory domains. The catalytic (RecA-like and helical) subdomains of MaModBC have no direct contacts (fig. S3). Although separated, these domains nonetheless exhibit a “head-to-tail” arrangement, with the conserved P-loops juxtaposed to the LSGGQ motifs of the opposite subunit. Yet the distance between these conserved motifs increases from ∼10 Å in AfModBC to ∼23 Å in MaModBC, leaving a large gap between the NBDs in the latter. The separation is even more pronounced than that previously observed in the “open” conformation of the nucleotide-free NBD of the maltose transporter MalK (20).

The regulatory domains of MaModBC bind two tungstate ions at their shared interface, with both subunits contributing to both tungstate binding pockets. This is reminiscent of how ATP is bound between the NBDs of ABC transporters, where two ATP binding sites exist at the shared interface of opposite NBDs, with each domain contributing to both binding sites. Each regulatory domain contains two similar and consecutive subdomains that feature “Greek key” motifs. Similar folds have been observed for other molybdate-binding proteins, among them the C-terminal domain of the molybdate-dependent transcription regulator ModE (21). In the presence of molybdate, homodimeric ModE changes its conformation to form a tight, molybdate-bound interface (22), whereupon its N-terminal DNA binding domain acts as a repressor of the molybdate transporter and as an enhancer of molybdenum-dependent enzymes (23). The regulatory domains of MaModC (residues 231 to 348) and E. coli ModE (residues 124 to 262) are structurally similar (Fig. 2, A and B), with an overall root mean square deviation (RMSD) of 2.1 Å. However, a circular permutation places the two Greek key motifs in an opposite order, with the N and C termini of the ModE domain located at the linker between the Greek key motifs of MaModC (fig. S4). Two oxyanions are bound at the shared interface in both proteins (Fig. 2B), and the location of the binding sites, as well as the specific amino acid side chains mediating contacts (mostly hydrogen bonds) to the oxyanions, are conserved (Fig. 2C). The similarity in sequence and structure extends to proteins involved in molybdate storage and homeostasis, including the Mop protein from Sporomusa ovata [RMSD = 2.3 Å (24)] and ModG from Azotobacter vinelandii [RMSD = 1.3 Å (25)].

Fig. 2.

Comparison of the regulatory domain of MaModC with that of the transcriptional regulator ModE. (A) Side view of MaModBC and E. coli ModE (PDB code 1o7l) after superposition of the substrate-bound regulatory domains. The MaModBC backbones are colored blue and green, whereas those of ModE are red and purple. (B) Stereoview of the regulatory domains of MaModC and ModE, with colors as in (A). Tungstate bound to MaModC is shown in ball-and-stick and colored blue, whereas molybdate bound to ModE is in yellow. (C) Structure-based sequence alignment of ModC with ModE and other molybdate-binding proteins of a similar fold. The secondary structure elements above the amino acid sequences (in single-letter code) were derived from MaModC, and residues shaded red and green contribute to the two distinct oxyanion binding pockets. Gray-shaded residues contribute to additional molybdate binding sites in Mop and ModG.

The architectural similarity of the regulatory domains of ModE and MaModC creates an unexpected structural link between genetic regulation and trans-inhibition in ABC transporters. In both processes, tight dimers of regulatory domains with two oxyanions (molybdate or tungstate) sandwiched at the shared interface affect the function of the fused partner domains. ModE exploits this to promote the binding of its N-terminal domains to DNA (23), whereas in the ABC transporter MaModBC, binding of molybdate keeps the catalytic motifs of the NBDs separated, thus disrupting the ATP hydrolysis cycle. Elevated concentrations of free cytoplasmic molybdate will thus lead to an inhibition of the transport activity (and hence to a decrease in molybdate uptake) through both processes. However, not all bacteria contain ModE, and in particular, there does not seem to be a homolog in the genome of M. acetivorans. This may indicate that trans-inhibition can substitute for the genetic regulation of the molybdate transporter. Many other organisms such as E. coli, Photorhabdus luminescens, and Klebsiella pneumoniae do contain genes coding for ModE and at the same time feature regulatory domains attached to their ModC subunits (fig. S5), which suggests that molybdate import in these organisms may be regulated both by trans-inhibition and by genetic control of the expression levels.

Another ABC transporter, the maltose importer MalFGK from E. coli (26), also contains a regulatory domain with a similar fold. However, maltose has not been reported to bind to MalK, nor does it seem to affect the ATPase activity of the isolated NBDs or the full maltose transporter. Instead, the regulatory domains of MalK have been reported to interact with the transcriptional activator MalT (27) and with the enzyme IIA component of the glucose phosphotransferase system (28). The arrangement of the two regulatory domains of MalK in the crystal structures is distinct from what we observe in MaModBC, and an arrangement similar to that in MaModBC is precluded by a C-terminal extension of 15 amino acid residues in MalK that would cause a steric clash (fig. S4). This difference in arrangement may explain why the activity of MalK is not trans-inhibited by elevated levels of maltose.

Because ATPase activity is a strict requirement for transport in ABC transporters, the inhibition of the ATPase activity of MaModBC by molybdate or tungstate will result in inhibited transport activity. To demonstrate that the inhibitory effect was specific, we identified four conserved residues in MaModBC (Ser286, Thr320, Ser323, and Lys340; Fig. 2C) that serve as hydrogen bond donors in the oxyanion binding pockets. We individually mutated these residues to alanines (except for Thr320, which was mutated to the isoelectronic valine) and tested the sensitivity of the ATPase activity of the resulting mutants to molybdate (Fig. 3). The mutants revealed ATPase activities similar to that of the wild type, but the sensitivity of the hydrolysis rates to molybdate was strongly reduced in two mutants (Ser323 → Ala and Thr320 → Val) and abolished in the others (Ser286 → Ala and Lys340 → Ala). A control mutation (Ser342 → Ala) of a non-conserved serine residue that is in the vicinity of, but does not provide a hydrogen bond to, the oxyanion, did not have any effect on the inhibitory concentration (Fig. 3). Combined, our functional data demonstrate that molybdate or tungstate binding to the oxyanion pocket at the interface of the regulatory domains allosterically inhibit the ATPase activity, and hence the reaction cycle, of MaModBC.

Fig. 3.

Inhibition of the ATPase activity of wild-type or mutant MaModBC by molybdate. Relative hydrolysis rates are shown at various concentrations of molybdate and in percent of the uninhibited rates for wild type (solid circles), Ser286 → Ala (open circles), Thr320 → Val (solid squares), Ser323 → Ala (open diamonds), Lys340 → Ala (solid diamonds), and the control mutant Ser342 → Ala (open squares). All experiments were carried out twice with independent protein preparations.

The transmembrane domains of AfModBC, MaModBC, and MalFGK have similar folds and TM topologies but reveal differences in conformation (Fig. 4A). Whereas MaModBC and AfModBC reveal nucleotide-free, inward-facing conformations, MalFGK is ATP-bound and outward-facing. The two ModBC structures differ in that the trans-inhibited MaModBC reveals increased angles of the two TMDs in a more pronounced inward-facing conformation, which is reflected in a larger distance of the coupling helices relative to AfModBC (Fig. 4A).

Fig. 4.

TMD conformations as observed in the crystal structures of MaModBC, AfModBC, and MalFGK. (A) Comparison of the TMDs MaModB, AfModB, and MalFG. The key TM helices 4 and the coupling helices of each transporter are colored yellow and blue, respectively. The Cα atoms of residues MalG 183 and MalF 394 and the equivalent residues in MaModB (165) and AfModB (153) are depicted as red spheres, with the distances indicated. For clarity, only the cores of the TMDs are shown for MalFG, and the NBDs have been removed. (B) Chemical cross-linking of engineered cysteine residues at position 153 in AfModB. Cross-linking was performed by CuCl2 in detergent solution, with or without ATP and o-vanadate (VO4). No Cu was added to the control reaction. Protein markers are shown in the left lane, with molecular masses indicated.

On the basis of the crystal structures of full ABC transporters, we have previously proposed a conserved coupling mechanism for ABC transporters (16, 29), which is an adaptation of Jardetzky's alternating access and release mechanism (30) and also agrees with other mechanistic models (31). The basic two-state schematic suggests that the TMD-formed translocation pathways of ABC transporters adopt an inward-facing conformation when the NBDs are nucleotide-free, and an outward-facing conformation when the NBDs form a head-to-tail sandwich with tightly bound ATP. If correct, this scheme predicts that the different conformations of the TMDs may be converted into one another, and then trapped using chemical cross-linking, depending on whether ATP is bound to the NBDs. We have thus engineered cysteines into AfModB (the TMD) at the cytoplasmic end of TM 4 at positions that are very close in the structure of MalFGK, but far apart in both ModBC structures (Fig. 4A). The engineered cysteines had no effect on the stability and ATPase activity of AfModBC, which suggests that the mutation had not induced any conformational change in the protein. The cysteines were subsequently reacted with copper, which can promote disulfide formation; this technique was introduced for membrane proteins during studies of the conformational changes in the aspartate receptor (32). In contrast to longer chemical cross-linkers, the use of Cu as a catalyst requires the cysteine side chains to be in the immediate vicinity and in an optimal geometrical arrangement for successful disulfide formation. One of our AfModBC mutants (Ser153 → Cys) revealed pronounced cross-linking that was dependent on the addition of ATP (Fig. 4B), as evidenced by the increase of a dimeric ModB band and the simultaneous decrease of the monomeric ModB band. The efficiency of the cross-linking was estimated to be above 50%, as assessed from the band intensities of several SDS gels. Such an efficiency is remarkable given the stringent requirements on geometry and the competition of the cross-linking reaction with air oxidation of the cysteines.

The results suggest that the TMDs of AfModBC can adopt a conformation similar to that of MalFGK only when ATP is bound to the NBDs. This demonstrates that binding of ATP to the NBDs is coupled to a conformational change in the TMDs, forcing the cytoplasmic TM helices together and likely converting the entire translocation pathway into an outward-facing conformation. Similar experiments with MaModBC were not successful, possibly because the relative arrangement of the introduced cysteines is not as favorable for efficient disulfide formation or because air oxidation of the cysteines was faster than the cross-linking reaction. Our results nonetheless demonstrate that in the ATP-bound state, the cytoplasmic ends of the transmembrane helices of molybdate ABC transporters approach each other and adopt a conformation similar to the outward-facing state of the maltose transporter.

Our cross-linking data lend further biochemical support to the above-mentioned two-state model of the mechanism of ABC transporters (29), which is in agreement with several ABC transporter structures that have been determined in recent years (33, 34). The structure of MaModBC we describe here fits this two-state scenario but adds an important new element. Even though we do not know how the absence of bound molybdate or tungstate will affect the conformation of MaModBC, it will likely remain inward-facing in the absence of nucleotide and will only adopt an outward-facing conformation, similar to that observed in the maltose transporter intermediate, upon binding of ATP. The trans-inhibited MaModBC structure demonstrates that regulatory domains can keep the catalytic domains of the NBDs apart, thereby preventing efficient ATP hydrolysis. Because the NBDs are connected to the TMDs via the coupling helices, the cytoplasmic parts of the TMDs are spread apart as well, which locks the transporter in an inward-facing conformation and prevents the ATP-triggered switch to the outward-facing state. Thus, high cytoplasmic concentrations of free substrate can prevent the transporter from cycling through the alternating access and release mechanism. Even though the details of substrate binding may differ in other families of transporters, it is likely that trans-inhibition in such proteins is based on similar effects.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1


References and Notes

View Abstract

Navigate This Article