Based on sequence homology analyses, the bacterial SBP superfamily has been classified into 8 clusters, with cluster 5 comprising dipeptide binders (DppA family), oligopeptide binders (OppA family) and nickel specific SBP's (NikA family) [7]. Continuous family updates by the Transporter Classification Database http://www.tcdb.org has now led to a cluster 5 SBPs containing up to 27 different subfamilies that are associated with translocation cargos as diverse as - in addition to di-and oligo-α-peptides and nickel substrates - antimicrobial peptides, δ-aminolevulinic acid, heme, plant opines, carbohydrates, the osmoprotective proline betaine, and the metal-chelater ethylene diamine tetraacetate. The most recent addition to the SBP family is termed GbpA (TCID: 3.A.1.5.27), a lipoprotein from Haemophilus influenzae, which binds reduced (GSH) and oxidized (GSSG) forms of glutathione to prime the dipeptide-DppBCDF ABC-transporter for glutathione translocation across the inner membrane [8]. Structural studies of the highly homologous GbpA from H. parasuis in complex with GSSG have revealed structural features that typify cluster 5 SBPs, namely, a pear-shaped, two-domain α/β-fold that collapses about the hinge region connecting the N-and C-terminal domains to sandwich the molecular cargo, in this case a single GSSG molecule [8]. In the absence of ligand, SBP's are flexible with the two domains rotating around the hinge and existing largely in the open conformation with both domains separated. Substrate binding induces the closed conformation, and the ligand is trapped at the interface between the two domains, according to what has been termed the "Venus Fly-trap" mechanism [9]. The structural analysis of GbpA in complex with GSSG has identified many specific interactions between GSSG and its cognate SBP that may be helpful in the delineation of the entire GbpA family [8]. The discovery that GbpA mediates glutathione transport in H. influenza came as a complete surprise as this protein was previously thought to be a heme-binding protein, accordingly annotated HbpA, and was implicated as a binding-platform for heme [5,10]. Nonetheless, GbpA does bind hemin, albeit weakly with an apparent K_d of 655 μM [8], and a possible role for GbpA and DppBCDF in heme acquisition has been described [10,11]. In this regard, GbpA presents itself as a good example of the high degree of substrate promiscuity especially common among cluster 5 SBPs [12-15].

Interestingly, the HbpA2 clade diverged significantly from both the GbpA and DppA signature sequences. In fact, some of the strictly conserved residues that contact the ligand's charged N- and C-termini in either the GbpA or the DppA family are replaced by physicochemically dissimilar residues in the HbpA2 sequences thereby virtually disrupting critical ligand-stabilizing salt bridges (In case of GbpA-GSSG binding, Arg33 substituted by a Thr, and Asp432 substituted by an Arg; in case of DppA-dipeptide binding, Asp408 substituted by an Arg). The ligand specificity of the HbpA2 clade is therefore difficult to predict, but it is highly unlikely that glutathione or dipeptides are the natural molecular cargos. Given the auxotrophic nature of Pasteurellaceae for heme and the fact that the dpp-architecture is a proven hemin-binding scaffold (E. coli DppA binds hemin with a 10 μM affinity [14] and also GbpA_Hi displays an, albeit low, affinity for hemin [8]) it is tempting to speculate that HbpA2 proteins may play a role in heme transport.

To document the heme-binding characteristics of the GbpA family, to verify the role of the posttranslational 1,2-diacylglycerol-modification of GbpA proteins in terms of glutathione and heme binding, and to establish the ligand-preferences of the HbpA2 family, we selected in addition to GbpA_Hi yet another GbpA lipoprotein (from A. pleuropneumoniae, GbpA_Ap), 2 non-lipoprotein GbpA's (GbpA_Hp and GbpA_Pm from H. parasuis and P. multocida, respectively), and 2 HbpA2 proteins (HbpA2_Hp and HBPA2_Ap from H. parasuis and A. pleuropneumoniae, respectively), for further study.

In a previous report, we had employed a hemin-binding assay based on native-PAGE to show that GbpA_Hi has a physiologically irrelevant affinity for hemin [8]. To probe heme-binding among our protein test panel, the purified recombinant soluble forms of the test proteins were subjected to our native-PAGE-based assay in the presence and absence of 0.5 mM hemin (Figure ​(Figure3).3). As this hemin concentration approaches its K_d-value, the GbpA_Hi band splits up, with about half of it migrating faster because of complexation with hemin (as judged by visual inspection (red-brownish bands) and heme-staining with 2,3',5,5'-tetramethylbenzidine/H₂O₂). Although all tested proteins displayed the split migration pattern, the fraction of the faster running hemin-complexed bands is much lower compared to that of GbpA_Hi and therefore indicative of an extremely low affinity for hemin. These results strongly suggest that heme-binding is not a general feature of the GbpA and HbpA2 family. Notably, the second best binder of hemin is GbpA_Ap, the other lipoprotein in our test panel.

A fluorescence-based thermal shift assay (Thermofluor assay [17]) was subsequently employed to screen for potential allocrites. The set of putative ligands amounted to 15 different ligands comprising of glutathione, some of its derivatives, and already established allocrites of the type 5 SBP superfamily such as di- and tripeptides, δ-aminolevulinic acid, nickel, and proline-betaine. The temperature-induced changes in relative fluorescence of 100 μg of test protein as a function of candidate ligands at 1 mM were recorded and those ligands that significantly affected the transition midpoint temperature (T_m) of the apo-form (threshold was set at 1.5°C) are shown in Table ​Table11 as a function of the corresponding ΔT_m-values (apparent T_m-differences in °C for the ligand-protein complexes relative to the uncomplexed proteins). Our analysis revealed an affinity of all tested GbpA proteins for (ranked according to descending ΔT_m): GSSG > GSH > S-methylglutathione glutathione-cysteine disulfide. In addition, GbpA_Hp and GbpA_Pm also showed a minor but significant T_m-shift in the presence of the bulky S-alkylated glutathione derivatives, S-hexylglutathione and S-decylglutathione. Although certain S-modifications were tolerated, fragments of glutathione such as γ-glutamylcysteine or cysteinylglycine or a slightly elongated form of glutathione (homoglutathione) did not influence the melting behavior of any of the tested proteins, showing that the GbpA family carries a specificity for the glutathione backbone. Interestingly, in contrast to the notion that increasingly bulkier S-alkylations abrogate binding, the disulfide of glutathione with another glutathione molecule or with cysteine appear to be good allocrites for the entire GbpA family, strongly suggesting that the GbpA-fold evolved to bind these types of glutathione derivatives in vivo. This observation makes sense as many Pasteurellaceae are glutathione as well as cysteine auxotrophs and glutathione-cysteine disulfide reaches levels similar to those of glutathione in human plasma (up to 10 μM [18]).

Isothermal titration calorimetry (ITC) was subsequently used to determine the equilibrium dissociation constants for the interaction of our GbpA proteins with GSSG, GSH, and S-me-GSH. Typical ITC thermograms, showing the raw and integrated data for the interaction of GbpA_Hp with these allocrites are shown in Figure ​Figure4,4, and all respective calculated K_d-values are summarized in Table ​Table2.2. Except for GbpA_Hp, the ranking of binding strength according to the thermal shift ΔT_m-values was recapitulated by the ITC-derived K_d-values. Notably, affinities for the natural allocrites, GSSG and GSH, varied 200-fold, and for the artificial ligand S-me-GSH ~ 400-fold among the selected GbpA-family members, with GbpA_Hi being the worst binder for all tested putative ligands. Interestingly, GbpA from H. influenzae, which naturally exists in a membrane anchored form, takes a unique position within the GbpA-family as the best binder of hemin, and the worst binder of glutathione. On the other hand, the soluble form of the predicted lipoprotein GbpA from A. pleuropneumoniae displays affinities for the tested glutathione derivatives that are similar to the two non-lipoprotein GbpA's (see Table ​Table2).2). Therefore, membrane-anchoring of GbpA proteins appears not to impose any functional implications. Interestingly, the best hemin-binders from our test proteins were the lipoprotein GbpAs (Figure ​(Figure3),3), amongst which the one of H. influenzae was shown to be biologically significant for heme acquisition [10,11]. Therefore, membrane-anchoring may influence the role of GbpAs in heme acquisition by increasing their intrinsic affinity for hemin. Nonetheless, GbpA-mediated heme import appears to be of minor importance under laboratory conditions as yet another family 5 SBP, the antimicrobial peptide binder SapA, has recently been shown to be essential for heme utilization by iron-starved nontypeable H. influenzae cells [19].

The lack of conservation of consensus sequence fingerprints important for ligand-binding by GbpA- and DppA-family proteins (Figure ​(Figure2)2) had already suggested that the HbpA2 proteins in our test set would fail to bind the glutathione- and dipeptide-types of ligands. Indeed, our thermofluor analyses showed that none of the two HbpA2 proteins under study were able to interact with any of the tested type 5 SBP superfamily allocrites (Table ​(Table1).1). Because of the possibility that the HbpA2 proteins would co-purify with their natural ligands, as observed for some other structurally characterized SBP's, such as e. g. the cysteine-complexed CjaA from Campylobacter jejuni [20] or the oligopeptide-binder AppA from Bacillus subtilis in complex with a nonapeptide [21], we sought to determine the crystal structure of HbpA2_Hp hoping to elucidate an interaction with a possible ligand. The crystal structure of HbpA2_Hp was determined to 2.0 Å resolution by maximum-likelihood molecular replacement (Figure ​(Figure5;5; additional file 1, Table S1). The structure reveals the two-lobe α/β-fold architecture and β-strand topology typical for SBP-like proteins, and is essentially identical to that of the structurally characterized GbpA and DppA proteins. Crystallographic refinement and exhaustive examination of residual difference electron density maps failed to provide any evidence for a bound ligand to HbpA2. Moreover, the N- and C-terminal domains were opened by about 33 degrees with respect to the GSSG-bound GbpA and glycylleucine-bound DppA reported previously [8,16] (Figure ​(Figure5A),5A), again indicating we crystallized apo-HbpA2_Hp. Importantly, the crystal structure of HbpA2_Hp offers an explanation for its inability to bind peptide-like ligands. Figure ​Figure5B5B shows a structural superposition of residues of the GbpA ligand-binding site with only those corresponding residues in HbpA2_Hp of which the physicochemical properties are significantly different as revealed by our sequence alignments (Figure ​(Figure2).2). This analysis focuses on the C-terminal lobe, because it comprised the majority of the ligand-interacting residues as shown by the GbpA_Hp-GSSG complex [8] (13 out of 18 interactions), and due to the fact that it is believed to drive formation of the SBP-ligand-encounter complex [22]. Out of the 13 GSSG-contacting residues, 3 were not strictly conserved in HbpA2_Hp, i.e. A380P, S430T, and D432R. All of these residues appear to be critical for GSSG-binding by GbpA_Hp: the peptide-nitrogen of A380 hydrogen-bonds with the carbonyl oxygen of GlyI of one of the glutathione legs (GS-I); the D432 side chain carboxylate forms a salt bridge with the amino terminus of GS-I as well as H-bonds with the side chain hydroxyl groups of Y138 and Y521 thereby positioning these residues for favorable hydrophobic interactions, the side chain of S430 is involved in H-bonding with both the carboxylate- and amino-groups of the γ-glutamyl-moiety of GS-I [8]. The structural superposition in Figure ​Figure5B5B shows that S430 and D432 in the HbpA2_Hp structure occupy the exact same position as the corresponding active site residues in GbpA_Hp. At the same time, A380 takes a slightly different position which would be expected due to the elimination of the special structural role of a proline residue in maintaining loop structure at this position. Our structural analysis offers direct evidence that the A380P, S430T, and D432R substitutions would be grossly incompatible with GSH and GSSG binding as they would abolish electrostatic, H-bonding and hydrophobic interactions contributions critical for binding of such ligands. A similar analysis, this time against E. coli DppA, shows that R415 in HbpA2_Hp takes the exact same position as the active site residue D408 in E. coli DppA, a residue that makes a salt-bridge with the amino-terminus of the bound dipeptide ligand. Thus, R415 would prevent dipeptide ligand binding. Finally, we note that the inability of the HbpA2 proteins to interact with either glutathione or dipeptides is correlated by looking at the interspecies occurrence: 2 out of 3 species with HbpA2 genes also carry genes for both a GbpA and a DppA family member.