|Home | About | Journals | Submit | Contact Us | Français|
TonB-dependent transporters (TBDTs) are bacterial outer membrane proteins that bind and transport ferric chelates called siderophores, as well as vitamin B12, nickel complexes, and carbohydrates. The transport process requires energy in the form of protonmotive force and a complex of three inner membrane proteins, TonB-ExbB-ExbD, to transduce this energy to the outer membrane. The siderophore substrates range in complexity from simple small molecules such as citrate to large proteins like serum transferrin and haemoglobin. Because iron uptake is vital for almost all bacteria, expression of TBDTs is regulated in a number of ways that include metal-dependent regulators, σ/anti-σ factor systems, small RNAs, and even a riboswitch. In recent years many new structures of TBDTs have been solved in various states, resulting in a more complete picture of siderophore selectivity and binding, signal transduction across the outer membrane, and interaction with TonB-ExbB-ExbD. However, the transport mechanism is still unclear. In this review, we summarize recent progress in understanding regulation, structure and function in TBDTs and questions remaining to be answered.
Transport into Gram-negative organisms is initiated by passage of the transported species across the outer membrane and into the periplasmic space prior to inner membrane translocation. The uptake of iron is particularly important for bacterial growth (71) and synthesis of outer membrane iron transporters (called TonB-dependent transporters, TBDTs) is therefore regulated in a variety of ways. While iron complexes constitute the majority of substrates for TBDTs, vitamin B12, nickel chelates, and carbohydrates are also transported by this mechanism (75). These transporters show high affinity and specificity for metal chelates called siderophores and require energy derived from the protonmotive force across the inner membrane to transport them (33, 87). To tap this energy source, TBDTs must interact with an inner membrane protein complex consisting of TonB, ExbB, and ExbD (72).
The first crystal structures of two Escherichia coli TonB-dependent transporters, ferrichrome transporter (FhuA) (34, 54) and ferric enterobactin transporter (FepA) (10), showed that TBDTs use a 22-stranded β-barrel to span the outer membrane with an unanticipated ‘plug’ domain folded into the barrel interior. The plug domain functions to bind a specific metal chelate at the extracellular side of the membrane and to interact with TonB-ExbB-ExbD at the periplasmic side of the outer membrane. In these ‘ground state’ structures, the plug domain completely occludes the barrel pore, revealing an unexpected complexity for siderophore transport. There has been significant recent progress in structure determination of TBDTs, with a total of 45 structures solved to date, representing 12 unique transporters. In this review, we summarize new data on the complex regulation of TBDTs, structural similarities and differences, and new functional data pertaining to the transport mechanism. We will focus on E. coli, but include information on other Gram negative bacteria where appropriate.
Genes encoding the seven TBDTs in E. coli are scattered throughout the chromosome. Several of them, btuB, fhuE, and cirA, are transcribed as monocistronic units. In contrast, fecA and fhuA are the first genes of multicistronic operons, fecABCDE and fhuACDB respectively. In these cases, the downstream genes encode ABC transporters that transport the siderophores across the cytoplasmic membrane. Downstream of fepA is the gene entD, involved in the synthesis of the enterobactin siderophore. Finally, fiu is followed by the ybiX and ybiI genes, but whether or not they form an operon has not been investigated. Expression of all of these genes is highly regulated both at the transcriptional and post-transcriptional levels. These controls can limit the synthesis of TBDTs when they are not needed, which could be beneficial since some of these outer membrane proteins are also used by phages and colicins to enter the bacterial cell (12).
Although iron is essential for most living organisms, iron accumulation can be toxic because it may lead to production of reactive radicals (46) and it is therefore crucial to keep cellular iron levels under tight control. In E. coli, the Fur (Ferric Uptake Regulator) transcriptional repressor plays a key role in this process by regulating expression of genes involved in iron homeostasis as a function of cellular iron concentration (43). In the presence of iron, Fur binds DNA sequences referred to as Fur boxes using Fe2+ as a cofactor and thereby represses expression of dozens of genes (3). When iron is limiting, Fur cannot bind DNA, leading to derepression of genes that encode iron transporters and proteins involved in siderophore biosynthesis and iron metabolism, but also other cellular functions (13, 80). Somewhat surprisingly, several genes were reported to be positively regulated by Fur. This apparent puzzle was solved when Massé and Gottesman identified a regulatory small RNA, RyhB, whose transcription is repressed by Fur and which, in turn, negatively regulates expression of numerous genes (58).
Consistent with their role in iron transport, all TBDTs for ferric siderophores are controlled by Fur and their expression is therefore repressed when iron reaches a certain level (Figure 1). Fur binds in vitro to the promoter regions of fepA-entD (45), fecABCDE (2), fhuACDB (13) and cirA (39). In addition, Fur boxes were identified not only in these promoter regions, but also upstream of fhuE and fiugenes (13 and references therein, 64). Interestingly, Fur also directly represses transcription of the tonB gene (1, 89) as well as the exbB-exbD operon by binding upstream of exbB (13).
Transcription of the fecABCDE operon is dependent on the minor σ factor FecI (σ19). FecI belongs to the group of ECF (Extracytoplasmic Function) σ factors, also known as group IV. ECF σ factors are present in virtually all bacteria and regulate the expression of many genes, including genes for periplasmic or outer membrane proteins, hence the name ECF (79). A common feature of the ECF σ factors is that their activity is regulated by an anti-σ factor, which is usually coexpressed with its cognate σ factor. In general, the anti-σ factor sequesters the σ factor under non-activating conditions and this inhibition is relieved under specific activating conditions.
There is a signal transduction cascade that leads from extracellular siderophore binding to FecA to the activation of FecI and subsequent transcription of fecABCDE genes (44). Upon binding ferric citrate, FecA transduces a signal across the outer membrane to FecR, an inner-membrane protein. FecR then transmits the signal across the inner membrane to FecI which directs RNA polymerase to transcribe the fecABCDE operon. As for other ECF σ factors, the activity of FecI is regulated by its anti-σ factor, which is FecR. However, FecR is not a ‘classical’ anti-σ factor because FecR is required for activation of fecABCDE by FecI (68). Interactions between FecA and FecR, and FecR and FecI were analyzed in vivo : experimental data support a model where the periplasmic N-terminal region of FecA and C-terminal region of FecR interact, while the cytoplasmic region of FecR interacts with FecI (Figure 1) (26). While structures exist for the FecA transporter (32, 90) and the FecA signalling domain that interacts with FecR (36), the details of the signal transduction cascade are not understood.
Fur and FecI are probably not the only regulators affecting the transcription of genes for TBDTs. Expression of fecA, fepA, cirA and fiu was found to be increased in a mutant for the global transcriptional regulator Crp (cAMP receptor protein), both by a transcriptomic approach and by RT-PCR (92). Synthesis of these 4 TBDTs is therefore expected to be modulated not only in response to iron availability, but also to the carbon status of the cell. However, even though these effects are independent of Fur since these experiments were done in a fur mutant, experimental data are still lacking to discriminate between direct or indirect effects.
In addition, fecA expression is also increased by pyruvate, because it is regulated by dhR, a transcriptional regulator of genes involved in the energy production pathway (69). In the absence of pyruvate, PdhR represses the expression of its target genes by binding to the promoter region; when the level of pyruvate is sufficiently high however, PdhR no longer binds to DNA and repression is relieved (69, 73). PdhR was identified as a potential regulator of fecA expression by an algorithm developed to determine transcriptional regulatory interactions in E. coli on the basis of multiple microarray expression profiles (27). Furthermore, ChIP experiments indicate that PdhR could directly bind to the fecA promoter. Consistent with this dependence of fecA expression on PdhR, fecA expression was found to be highest in the presence of both pyruvate and citrate by real-time quantitative PCR (27).
Riboswitches are RNA elements that can change conformation upon specific binding of a small molecule (57, 60). Typically, they are located at the 5′ end of mRNAs and the ligand-induced conformational change directly affects, either positively or negatively, transcription or translation of the downstream gene(s). Genes controlled by riboswitches are often involved in the uptake or metabolism of the ligand, which could be a vitamin, amino acid, or nucleotide (67). Over the years, a number of experimental data suggested the existence of a riboswitch controlling the synthesis of BtuB, the transporter for vitamin B12 (cyanocobalamin (CnCbl)). Indeed, it has long been known that btuB expression is repressed when cells are grown in the presence of vitamin B12 (47), yet no repressor protein has been identified. Mapping of the 5′ end of btuBmRNA revealed a 241 nt untranslated leader (55) that is necessary for repression of btuB expression by vitamin B12. However, mutants defective in the production of adenosylcobalamin (AdoCbl), a downstream product of vitamin B12 metabolism, were known to constitutively express btuB, suggesting that AdoCbl, not vitamin B12, may have a direct role in repression. AdoCbl was later shown to inhibit ribosome binding to btuB mRNA (66). Further experiments demonstrated that AdoCbl binding to the btuB leader induced a structural change (62). This conformational change is likely to be responsible for translational control of btuB expression by stabilizing a structure inhibitory for translation.
Some RNA-mediated post-transcriptional controls are exerted by regulatory small RNAs (sRNAs). In the last decade, sRNAs have been recognized as major regulators of gene expression (37, 85). In most cases, bacterial sRNAs are short RNA molecules (< 250 nt), which are synthesized as discrete transcripts and act as post-transcriptional regulators. A large group of sRNAs bind the RNA chaperone Hfq, that can, among other roles, facilitate RNA-RNA interactions (6, 83). Accordingly, all sRNAs of this group have the ability to pair with one or several mRNAs, and thereby regulate their translation and/or stability.
OmrA and OmrB are two Hfq-binding sRNAs conserved in several enterobacteria. They are encoded by two adjacent genes and display almost identical 5′ and 3′ ends but a rather distinct central region. Even though OmrA and OmrB could theoretically bind different targets through their central regions, only targets that are common to both OmrA and OmrB have been identified so far and are all negatively regulated by OmrA/B. They encode several outer membrane proteins (OmpT, Cir, FecA and FepA) as well as the GntP inner membrane transporter and the EnvZ-OmpR two-component system, which itself activates transcription of omrA and omrB (40, 41, 82). Preliminary data from microarray analyses also indicate that genes downstream of fecA or fepA (ie fecBCDE and entD respectively) could be regulated by OmrA/B. This may be due to a change in mRNA stability in the presence of OmrA/B and/or to translational coupling between the different genes of a single operon.
As mentioned above, OmrA/B repress the synthesis of at least three TBDTs, Cir, FecA and FepA. cirA was shown to be a direct target of OmrA/B, but whether this is true as well for fecA and fepA remains to be experimentally tested (41).
OmrA/B are synthesized in response to the activation of the EnvZ-OmpR two component system (40). Although the physiological signal for this activation remains unclear, the levels of phosphorylated OmpR change as a function of the osmolarity of the external medium. Consequently, several genes regulated by EnvZ-OmpR, such as the ones for the major porins OmpC and OmpF, as well as omrA and omrB, are differentially expressed at different osmolarities. The importance of down-regulating several TBDTs for siderophores in response to changes in osmolarity is not entirely clear.
Virtually all Gram negative bacteria have TBDTs that are involved in the uptake of iron and vitamin B12, as well as nickel, carbohydrates, and probably other substrates (75). The total number of TBDTs is highly variable among bacterial genomes: while E. coli synthesizes just 7 TBDTs, Pseudomonas aeruginosa makes 34 TBDTs (81) and Caulobacter crescentus makes 65 TBDTs (65). The current knowledge about the regulation of TBDT synthesis in these and other bacteria is limited.
Expression of genes for TBDTs involved in iron uptake is regulated by Fur in numerous bacteria (53). Somewhat similarly to this regulation of iron uptake genes by Fe2+ via Fur, synthesis of Helicobacter pylori FrpB4, a TonB-dependent nickel transporter (74), is repressed by the nickel-sensing transcriptional regulator NikR (23). In C. crescentus, the outer membrane protein MalA, which is likely a TBDT and required for maltose uptake, is induced in presence of maltose, but the mechanism for this regulation is still unknown (63).
FecIR-type regulation is also present in numerous bacteria (5, 50). Interestingly, several of these ‘anti-σ’ factors were shown to be required for full activation of their cognate σ factors, just like FecR (see above). This is the case for Bordetella avium RhuR (49) and Pseudomonas aeruginosa FoxR and FiuR involved in the regulation of the uptake of desferrioxamine and ferrichrome siderophores, respectively (59).
The control of btuB expression by a vitamin B12-responsive riboswitch is most likely widespread in Gram negative bacteria. Indeed, two independent phylogenetic analyses identified a similar conserved RNA motif not only in the 5′ UTR of btuB homologs from numerous Gram negative bacteria, but more generally in the 5′ UTR of genes involved in the metabolism or transport of vitamin B12, as well as some other genes, both in Gram positive and Gram negative bacteria (61, 84). When this was looked at, for instance for the elements upstream of btuB and cob genes of Salmonella typhimurium, these RNA motifs were shown to efficiently and selectively bind vitamin B12 (61).
OmrA and OmrB are conserved in most enterobacteria, even though one or even both of them can be absent in some species (40). A direct base-pairing interaction between cirA mRNA and OmrA/B was shown to control cirA expression in E. coli and a similar interaction is predicted in other enterobacteria (41). However, whether cirA (and also fecA and fepA) are really regulated by OmrA/B in other species remains to be investigated. In addition, it would also be interesting to determine whether other post-transcriptional events control the synthesis of these TBDTs in bacteria lacking OmrA/B.
In 2005, Chimento et al. published a comprehensive structural analysis (17) of the four TBDT structures published at that time (10, 18, 32, 34, 54, 90). Since then, the structures of eight more TBDTs have been determined (8, 9, 19-21, 51). In addition, structures were solved for TBDTs with various ligands bound (7, 30, 35, 38, 52, 76, 88), in complex with the periplasmic domain of TonB (70, 77), and one TBDT crystallized from a lipidic cubic phase (15), giving us 45 crystal structures to compare now (Table 1; Supplemental Figure 1). An analysis of the original four TBDTs showed that all of them have the same domain architecture: a 22-stranded transmembrane β-barrel encloses a globular plug domain (Figure 2). Ligand binding sites are formed from residues on the extracellular side of the plug domain, as well as from residues on the walls and extracellular loops of the β-barrel. The TonB box is found at the N-terminus of the plug domain, and in some structures protrudes into the periplasm. In others, the TonB box is tucked up into the plug domain within the barrel or is disordered and not visible in the structures. A structure-based sequence alignment revealed conserved motifs in the plug and barrel which are close to one another and interact. Finally, an analysis of water molecules located at the plug-barrel interface revealed that the plug is highly solvated, resembling a transient protein complex and suggesting conformational change and/or movement of the plug within the barrel during transport. In the following sections, we will outline some of the significant structural and functional studies done with TBDTs in recent years.
The 22-stranded β-barrel with inserted plug domain is conserved for all known transporters and very likely represents the architecture of all TBDTs. The ligand binding sites are customized for the cognate siderophore or colicin. For example FhuA uses aromatic residues to bind ferrichrome (31, 54) while the binding pocket on FecA contains several arginine residues to bind the negatively charged diferric dicitrate (32, 90). Two heme transporters coordinate their ‘siderophore’ through conserved histidine residues residing in the plug and an extracellular loop (21, 51). Only three structures have been solved for TBDTs bound to the receptor binding domain of various colicins (9, 52, 76), and binding differs substantially from siderophores, although the binding sites for colicins and siderophores appear to overlap (12). Interfacial water molecules were analyzed for Cir (9) and agreed with previous results reported by Chimento et al. in 2005 for BtuB, FepA, FecA, and FhuA (17) – that the plug is highly solvated inside the barrel pore. For this review, we performed a structure-based sequence alignment for the twelve unique TBDTs to ask how many of the conserved features identified by Chimento et al. (17) remain conserved in this larger group representing TBDTs from a variety of Gram negative bacteria (Supplemental Figure 2). Interestingly, many of the conserved motifs identified in the four original TBDT crystal structures in 2005 are also observed in the twelve currently known TDBT crystal structures. Conserved motifs include the TEE, PGV, IRG box, LIDG box, RP box, and the Hβ4 motifs, which are all located within the plug domain. Additionally, we observed significant conservation for most of the β-strand sequences with many of the β-strands having one or more signature residues that were found completely conserved, which may have implications for structure prediction of other TBDTs and possibly even other families of β-barrel proteins. Figure 2d shows FhuA (colored gold) with residues which were found to be at least 50% conserved among all 12 unique TBDTs with known structures indicated in blue. We found that 32% of these conserved residues were located within the plug domain (27% of total plug domain residues) and the other 68% were located within the core β-strands of the β-barrel domain (16% of total β-barrel domain residues). None of the indicated conserved residues were observed within the extracellular loops, further emphasizing their evolutionary divergence.
While almost all TBDT structures were crystallized from detergent/precipitant mixtures (4), Caffrey and colleagues crystallized the apo form of BtuB from a lipidic cubic phase (11) instead of detergent, yielding the highest resolution structure for this family of transporters (Table 1) (15). Crystals grown in meso exhibit denser packing than those grown in detergent, resulting in conformational differences in extracellular loops between the two apo BtuB structures (15, 16). Since these loops are unrestrained in vivo and probably move continuously, both structures depict physiologically relevant states of the protein. Otherwise the two structures are remarkably similar, with backbone RMSD values of less than 1.5 Å over 82% of all residues. This work confirms that lipidic cubic phase crystallization could be as useful for the crystallization of β-barrel outer membrane proteins as it is for α-helical inner membrane proteins (14).
The binding of a siderophore to its TBDT transduces a signal across the outer membrane that results in a disordering (also called unfolding or undocking) of the TonB box, as described below. The nature of the transduced signal is not completely clear, but for some TBDTs it appears to involve large conformational changes in extracellular loops which fold in over the top of the TBDT when siderophore binds, sequestering the ligand and contributing new residues to the binding site. This type of induced fit mechanism has been observed for FecA (32, 90), ShuA (21), and FyuA (Lukacik et al., unpublished). Ligand binding also induces smaller conformational changes in the plug domain (observed in many TBDT structures) but exactly how binding of a small molecule at the extracellular surface results in disordering of the TonB box is not completely clear.
Table 1 shows that the TonB box, generally located near the N-terminus of TBDTs, adopts a variety of conformations ranging from ordered to disordered that does not seem to correlate well with siderophore binding. Since this stretch of five residues is an essential part of the transporter (transport will not happen without it), it is important to understand its location and mobility in apo and siderophore bound TBDTs. The most definitive work in this area has been done by Cafiso and colleagues using site-directed spin labelling and electron paramagnetic resonance spectroscopy (EPR) to determine position and mobility of the TonB box. They showed that siderophore binding to BtuB results in an unfolded TonB box (termed disordered by crystallographers), whereas the apo structure exhibits a folded (or ordered) TonB box (28). They also showed that reagents used in protein crystallization can inhibit this transition (29), explaining the highly variable results seen in the crystal structures. It now seems clear that siderophore binding transduces a signal across the outer membrane that ultimately results in unfolding (or increased mobility) of the TonB box, which signals to TonB-ExbB-ExbD that a particular transporter is ligand-loaded and primed for transport. However, while FecA undergoes the same order/disorder transition seen for BtuB, the TonB box of FhuA was found to be constitutively unfolded (48), suggesting either that interactions between FhuA and TonB are constitutive or not regulated by the TonB box configuration. Clearly tools in addition to crystallography and EPR will be required to elucidate the signal transduction and transport mechanisms.
When a TBDT has bound its siderophore and signalled to TonB-ExbB-ExbD, the next step appears to be a physical association between the TonB box of the TBDT and the C-terminal (periplasmic) domain of TonB. Structures of this complex have been described by Wiener and colleagues for BtuB-TonB (77) and by Coulton and colleagues for FhuA-TonB (70). In both structures TonB assumes an alpha-beta fold containing a 3-stranded β-sheet. The TonB box of either transporter adopts a β-strand conformation that pairs with the existing β-sheet of TonB. Association through strand pairing has been observed for many protein complexes and although details differ for the two complexes described here, we can conclude that the binding interface is relatively small. In both structures the plug domain still resides inside the β-barrel just like in all the other ground state structures of TBDTs. Presumably energy in the form of protonmotive force, as well as a full-length TonB-ExbB-ExbD complex, would be needed to visualize the transporter in action.
It is widely accepted that the plug domain of TBDTs must undergo some form of conformational change in order to transport either siderophores or larger cargo such as colicins (12, 33, 87). However, the extent of the conformational change and whether or not the plug domain completely exits the β-barrel is a topic of debate. It has been postulated that upon binding of siderophores, the plug domain could undergo a conformational change that creates a small pore between the plug domain and the inner wall of the barrel whereby transport may occur (10, 34, 54). The observations that the plug domain is highly solvated (9, 17) and fairly loosely packed inside the β-barrel suggest that minimal rearrangement of the plug domain could in fact lead to a pore capable of allowing siderophore passage.
While pore formation through TBDTs via conformational change in the plug domain could be a feasible mechanism for transport of smaller ligands such as siderophores, this mechanism does not explain how much larger protein cargo such as colicins, which range in size from 29 kDa (91) to 69 kDa (86), are transported across the outer membrane (Supplementary Figure 3). A narrow pore might allow siderophore transport but near complete unfolding would be required in order for a colicin to pass through the same pore, which would be highly unlikely given the energy barrier for such a mechanism.
Several experiments to determine whether the plug exits the barrel during substrate transport used pairs of cysteine mutants to tether the plug to the β-barrel (Figure 3). When the tether was located near the N-terminus of the plug domain, both FhuA (25) and FepA (56) were inactivated. In the case of FhuA, transport was restored upon reduction of the disulfide. However, when disulfides tethered the middle of the plug domain to the β-barrel, siderophore transport still occurred, albeit at a reduced rate (24). One explanation for the discrepancy could be that disulfides were formed less efficiently in the middle of the plug domain compared to those located at the N-terminus, which is exposed to the oxidizing environment of the periplasm.
Recently, two groups attempted to label cysteine residues in the plug domain with reagents located in the periplasm to demonstrate plug domain movement. Li et al. introduced cysteine residues into the plug domain of FepA and observed differential labelling with fluorescein maleimide for G54C during transport of ferric enterobactin (56). G54C is located in the middle of the plug domain and is weakly labelled by the periplasmically located fluor in the ground state, but is more strongly labelled during transport. This suggests that the plug may partially exit the barrel during transport of the siderophore. Similarly, Devanathan and Postle introduced cysteine residues into the FepA plug domain and used biotin maleimide to probe conformational changes occurring upon translocation of colicin B (22). They observed increased labelling for N-terminal regions of the plug domain, particularly S46C, with much smaller increases in labelling residues in the C-terminal portion of the plug domain, also suggesting plug domain movement out of the β-barrel. However, Smallwood et al. found the opposite result for FepA and colicin B; they did not detect structural changes in the FepA plug domain upon interaction with colicin B using a different labelling reagent, different bacterial strains, and different colicin concentrations (78). Because of the variations in experimental approaches used, it may not be possible yet to determine whether the plug domain exits the barrel (or becomes more exposed to the periplasm) when colicin B interacts with FepA.
Taking a computational approach, Gumbart et al. used steered molecular dynamics (42) to simulate what happens when force is applied to the BtuB-TonB crystal structure (77). They found that force can be transmitted from TonB to BtuB without disruption of the β-strand interactions linking the two proteins, supporting a mechanical mode of coupling. When pulling simulations were performed, part of the BtuB plug domain unfolded, corresponding to periplasmic exposure of those residues. These results, and most of the experiments described above, suggest that some movement of the plug domain occurs upon interaction with TonB. However, the details and extent of this domain movement (or unfolding) and the precise transport mechanism remain to be elucidated.
We thank B. Canagarajah for reading the manuscript. NN, TJB, and SKB are supported by the Intramural Research Program of the NIH, National Institute of Diabetes and Digestive and Kidney Diseases. MG is supported by the CNRS and the University of Paris 7-Denis Diderot.