Mutagenesis and Molecular Modeling Reveal Three Key Extracellular Loops of the Membrane Receptor HasR That Are Involved in Hemophore HasA Binding

Abstract

On the basis of the three-dimensional model of the heme/hemophore TonB-dependent outer membrane receptor HasR, mutants with six-residue deletions in the 11 putative extracellular loops were generated. Although all mutants continued to be active TonB-dependent heme transporters, mutations in three loops abolished hemophore HasA binding both in vivo and in vitro.

HasR is a specific TonB-dependent outer membrane receptor from Serratia marcescens that internalizes heme in the periplasm from either free heme or a secreted heme carrier, the HasA hemophore of known structure (1, 7, 14). Both HasA and HasR are monomeric heme binding proteins with 1:1 stoichiometry between heme and the protein (10, 11). A 1:1 stoichiometric complex is formed between HasA and HasR in vitro; upon complex formation between holo-HasA and apo-HasR, heme is transferred from its binding site on HasA to its binding site on HasR, despite an unfavorable drop in the respective heme affinities between the two (5.5 × 10¹⁰ M⁻¹ for HasA and 5 × 10⁶ M⁻¹ for HasR) (11). Two conserved histidine residues present in the other heme/hemoprotein receptors are involved in heme binding and heme transfer from HasA to HasR (2, 11). HasA binding to HasR involves two specific short independent regions, with mutations in both regions being necessary in order to abolish fixation (12), opening up the possibility that two “complementary” extracellular regions on HasR may also be involved in HasA binding. Heme acquisition via the has system has been reconstituted in Escherichia coli (7). HasA binding to HasR is of high affinity, and the release of apo-HasA from HasR requires heme, a high proton motive force, and a high TonB complex concentration, such as that induced under conditions of iron starvation (13). At a low TonB complex concentration, HasA binds HasR irreversibly and inhibits heme uptake via HasR.

Since the high-resolution HasR structure is not presently elucidated (9), we built here a three-dimensional model of HasR. Based on this, we constructed mutants with six-residue deletions in the 11 putative extracellular loops of HasR and analyzed their phenotypes with respect to complex formation with HasA in vitro as well as heme uptake from free heme or holohemophore in vivo. We identified three putative HasR loops in which six-amino-acid deletions unexpectedly abolished HasA binding both in vivo and in vitro.

Building of the three-dimensional model of HasR and mutant generation.

In a previous study, we generated a three-dimensional model of HasR, which was used to delineate the boundaries between the N-terminal extension, the plug, and the beta-barrel (15). Here, we built a model that takes into account new parameters, namely, two recently elucidated templates and the existence of two disulfide bridges (see below), that greatly changed loop positions. This model was constructed using the structure Protein Data Bank files for the six TonB-dependent outer membrane transporters (TBDTs) homologous to HasR that are now currently available under accession numbers 1FEP (FepA), 1KMO (FecA), 1NQE (BtuB), 2FCP (FhuA), 1XKW (FptA), and 1XKH (FpvA). Despite a low level of sequence identity, these proteins display extensive structural similarities, with a common plug-barrel architecture of similar dimensions. The main differences lie in the loop conformations. Structure-based sequence alignment between these Protein Data Bank files was generated using CEMC (8). Multiple sequence alignment of these proteins with HasR was then performed with ClustalW. In the region whose primary sequences are the most divergent, several manual adjustments were necessary to optimize the alignment using hydrophobic cluster analysis and local motif conservation in the heme receptors belonging to TBDTs and in HasR proteins identified in other bacteria. Four cysteine residues were particularly conspicuous in HasR; they are found in highly variable regions of the protein, and two of them are completely conserved in the HasR subfamily, whereas the other two are found only in Erwinia and Yersinia in addition to Serratia. Those cysteine residues were likely to be involved in disulfide bonds, since purified denatured HasR did not react with 5,5′-dithiobis-(2-nitrobenzoic acid) (not shown). Supplementary local readjustments were consequently made around the four cysteines to accommodate the disulfide bridges. The homology model of HasR was then generated with the program MODELLER (16). Finally, the SCWRL 3.0 program (3) and energy minimizations were used for predictions of protein side chain conformations.

The model of HasR showed a typical TBDT receptor fold in two domains (Fig. 1): the C-terminal barrel, containing residues 240 to 865 folded into 22 antiparallel β-strands connected by 11 loops at the extracellular side and 10 short turns in the periplasmic side, and the N-terminal plug, folded into six β-strands and two α-helices. The side chain localization of the highly conserved C-terminal phenylalanine enabled insertion into the outer membrane, as observed in all crystallized TBDTs. The putative TonB box motif was located in the plug, at the periplasmic opening of the barrel. The plug was anchored to the barrel shell by hydrogen bonds and maintained within the barrel by electrostatic interactions between conserved residues of the barrel and plug. All conserved charge cluster motifs occurring at the plug-barrel interface reported in the other TBDTs may exist: R165 of the conserved IRG motif of the plug interacted with two conserved residues, E623 (β14) and E689 (β16), of the β-barrel; the latter was also able to form another salt bridge with K210 of the plug (4).

FIG. 1.

The three-dimensional HasR model presents an elliptical cross-section of 30 Å by 50 Å by 54 Å. The plug, in pink, with a diameter of 35 Å, completely occludes the large pore of the barrel (in gray). The two conserved histidines (in violet) H189 (at the top of the plug) and H603 (L7 of the barrel), involved in heme entry through HasR, are localized at the apex of the receptor. Despite being separated by 413 amino acids in the primary sequence, these two histidines are sufficiently close to each other in the three-dimensional model to be the potential heme iron ligands. The three loops 6, 8, and 9 in which deletions abolish HasA binding are represented by blue, amber, and red, respectively.

In order to identify extracellular loops that are potentially involved in HasA binding by HasR, we generated, by PCR mutagenesis, 11 six-amino-acid deletions in all loops including loop 7, containing H603, involved in heme binding and heme transfer from HasA to HasR (see the supplemental material). The mutant plasmids were named pFR2ΔL1 to pFR2ΔL11, and the mutant proteins were named HasRΔL1 to HasRΔL11.

In vitro formation of complexes between His-HasA and mutant receptors.

The E. coli C600ΔhemA strain was transformed with wild-type pFR2 and the respective mutant pFR2ΔL1 to ΔL11 plasmids. The mutant proteins were all expressed at levels close to those of the wild type in whole cells upon arabinose induction (3 h at 20 μg/ml starting at an optical density at 600 nm of 0.5), as determined by immunodetection with anti-HasR antibodies (Fig. 2), and were localized in the outer membrane, as determined by Triton-EDTA solubilization (not shown), indicating that the overall structure was not grossly altered. The main purpose of our study was to identify extracellular HasR loops that are potentially involved in HasA binding. To this end, we checked the in vitro formation of HasA-HasR complexes with solubilized HasR mutants and His-HasA as an affinity handle (15). Briefly, crude membranes from C600ΔhemA harboring various recombinant plasmids were prepared from 500 ml of culture after French press treatment. Membrane solubilization with ZW3-14 (n-tetradecyl-N,N-dimethyl-3-ammonio-1-propanesulfonate) detergent was carried out as previously reported for wild-type HasR (11). After affinity purification of the complexes, we found that complexes were formed between His-HasA and wild-type HasR, HasRΔL1, HasRΔL2, HasRΔL3, HasRΔL4, HasRΔL5, HasRΔL7, HasRΔL10, and HasRΔL11 but not HasRΔL6, HasRΔL8, and HasRΔL9 (Fig. 3, top), although the corresponding proteins were found in normal amounts in the detergent extracts (Fig. 3, bottom). Loops 6 and 9 were the two loops that contained two cysteine residues involved in disulfide bond formation, and they were potentially involved in HasA binding. We thus tested, using the wild-type receptor, whether disulfide bond reduction on HasR would affect HasA binding. Using two criteria, we observed the following: the absence of an isothermal titration calorimetry signal between HasA and HasR in the presence of the reducing agent TCEP [Tris(2-carboxyethyl)phosphine hydrochloride], the absence of complex formation as detected by gel filtration in the presence of TCEP, and, as a consequence, the absence of spectral changes accompanying complex formation (not shown). In the presence of TCEP, both disulfide bonds were reduced, as verified by mass spectrometry after modification by AMS (4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid). Indeed, four molecules of AMS were added to HasR in the presence of TCEP (not shown). Last, HasRΔL4 and HasRΔL5 were found in slightly reduced amounts in the complex compared to the control. This might originate from a lower affinity between His-HasA and mutant HasR and/or reduced complex stability, but this has not been investigated further.

FIG. 2.

Whole-cell detection of HasR and its variants from strain C600ΔhemA harboring the various recombinant plasmids. The cells were grown in LB medium in the presence of arabinose (20 μg/ml), and the equivalent of 0.15 optical density units at 600 nm was loaded onto each lane. C indicates empty vector. HasRWT, wild-type HasR.

FIG. 3.

Complex formation between His-HasA and HasR and its mutants. In all cases, about 20 μg of purified His-HasA (holo) was mixed with 1.6 ml of ZW3-14-solubilized membranes from strain C600ΔhemA harboring the various recombinant plasmids grown for 3 h in the presence of arabinose (20 μg/ml). After 1 h of incubation at 4°C, Ni-nitrilotriacetic acid agarose beads (50 μl; QIAGEN) were added, and the incubation continued for 1 h. The beads were pelleted, washed three times with a solution containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 20 mM imidazole, and 0.02% ZW3-14, and finally eluted with the same buffer containing 0.5 M imidazole before being loaded onto a gel. The top shows the stained gel of the various bound fractions, whereas the bottom represents the immunodetection of HasR proteins in solubilized fractions. C indicates empty vector.

Heme uptake activity of wild-type and mutant receptors with free heme and holo-HasA as heme sources.

Growth of the transformants around wells containing either 10 μM heme or 10 μM holo-HasA after overnight incubation in the presence of arabinose (20 μg/ml) in LB medium in the presence or absence of 0.2 mM dipyridyl was also tested. Dipyridyl induced increased formation of the TonB complex and hence enabled HasA-bound heme uptake via the wild-type receptor (13). Results are shown in Table 1. In the absence of dipyridyl, all mutants were able to take up free heme as efficiently as the wild type except for HasRΔL1, HasRΔL5, and HasRΔL7 (Table 1); none of the mutants, like the HasR wild type, was able to take up heme from holo-HasA (Table 1). In the presence of dipyridyl, all mutants were able to take up free heme as efficiently as the HasR wild type except for HasRΔL5 and HasRΔL7, for which the efficiency, as measured by the diameter of growth around the well, was severely reduced (Table 1); only HasRΔL1 and HasRΔL4 mutants were able to use holo-HasA as the heme source, like the wild-type receptor (Table 1). The TonB complex was required for heme uptake by all the mutants, as C600ΔhemA exbB expressing the various mutant receptors did not grow in the presence of free heme (not shown). We also checked inhibition by wild-type holo-HasA on HasR heme uptake in conditions at low TonB complex concentrations, indicative of HasA binding to HasR (13). HasRΔL1, HasRΔL5, and HasRΔL7 were not used in this test, since they allow little heme uptake in the absence of dipyridyl. HasRΔL2, HasRΔL3, HasRΔL4, HasRΔL10, and HasRΔL11 behaved like wild-type HasR, since holo-HasA inhibited growth on heme at low TonB complex concentrations. On the contrary, mutants HasRΔL6, HasRΔL8, and HasRΔL9 did not present this growth inhibition phenotype by holo-HasA, indicating that heme was still taken up under those conditions (data not shown). This is consistent with in vitro tests, since His-HasA formed complexes with all mutants except for HasRΔL6, HasRΔL8, and HasRΔL9.

TABLE 1.

Growth around wells containing either holo-HasA (10 μM) or heme (10 μM) of strain C600ΔhemA harboring pFR2 derivatives encoding wild-type HasR and its mutants^a

Protein	Growth
	Without dipyridyl		With dipyridyl
	Holo-HasA	Heme	Holo-HasA	Heme
HasRΔL1	−	−/+	+	+
HasRΔL2	−	+	−	+
HasRΔL3	−	+	−	+
HasRΔL4	−	+	+	+
HasRΔL5	−	−	−	−/+
HasRΔL6	−	+	−	+
HasRΔL7	−	−/+	−	−/+
HasRΔL8	−	+	−	+
HasRΔL9	−	+	−	+
HasRΔL10	−	+	−	+
HasRΔL11	−	+	−	+
HasR	−	+	+	+

^aHolo-HasA was used at 10 μM, and heme was used at 10 μM. In all cases, plates contained arabinose (20 μg/ml) without or with 0.2mM dipyridyl. Growth was observed after overnight incubation at 37°C. + indicates growth, − indicates no growth, and −/+ indicates very reduced growth.

Since the three-dimensional structure of HasR is not yet available, we generated a three-dimensional model that served as a basis for mutations. This study was undertaken for TBDT receptors BtuB (6), FepA (17), FecA (18), and FhuA (5); in those cases, except for BtuB, the structure was known, and precise deletions were made. As in the case of BtuB, instead of making complete loop deletions, we chose to make six-residue deletions. All mutants were synthesized at roughly wild-type levels and correctly localized in the outer membrane. In terms of heme uptake, under certain conditions, all mutants were as active as the wild-type receptor, except for HasRΔL5, for which heme entry was very limited under all conditions tested, and HasRΔL7, which, as expected, behaved like a barrel histidine mutant (11); this indicated that the overall architecture of the receptor was not affected by the mutations. The most striking feature uncovered in our study was the phenotype of ΔL6, ΔL8, and ΔL9 mutants: these three HasR mutants behaved like wild-type HasR in terms of free heme uptake; they were inactive with holo-HasA but did not bind HasA either in vitro or in solid-phase assays on whole cells (not shown). We had previously shown that HasA binds HasR via two independent parts (positions 51 to 60 and 95 to 106) (12), and it was thus possible that mutations could identify independent loops of HasR that would bind those two identified HasA regions. This was not what we observed with HasR deletion mutants. Given the phenotype of HasR mutants, it is likely that the three loops L6, L8, and L9 contributed either directly or indirectly to one or several HasA binding sites. Those three loops are positioned on the same side of the receptor around loop 7, which contains the conserved barrel histidine, and they are likely to collectively contain at least one binding site for HasA. Given the respective dimensions of HasA and HasR, HasA might thus be positioned so as to occlude the potential opening of the HasR barrel. It is possible that other loops are also implicated in HasA binding and that the deletions in loops 6, 8, and 9 have a stronger effect than point mutations in the context of the receptor. The precise function of the other loops awaits further experiments.