Mapping Ultra-weak Protein-Protein Interactions between Heme Transporters of Staphylococcus aureus

Background: Isd proteins convey heme molecules across the bacterial cell wall by means of sequential and transient protein-protein complexes. Results: Photo-cross-linking experiments revealed the contact regions between the IsdC transporter and other Isd proteins. Conclusion: Transient interactions are governed by distinct structural elements around the heme-binding pocket of IsdC. Significance: Targeting this epitope could lead to successful therapeutic strategies against Staphylococcus aureus. Iron is an essential nutrient for the proliferation of Staphylococcus aureus during bacterial infections. The iron-regulated surface determinant (Isd) system of S. aureus transports and metabolizes iron porphyrin (heme) captured from the host organism. Transportation of heme across the thick cell wall of this bacterium requires multiple relay points. The mechanism by which heme is physically transferred between Isd transporters is largely unknown because of the transient nature of the interactions involved. Herein, we show that the IsdC transporter not only passes heme ligand to another class of Isd transporter, as previously known, but can also perform self-transfer reactions. IsdA shows a similar ability. A genetically encoded photoreactive probe was used to survey the regions of IsdC involved in self-dimerization. We propose an updated model that explicitly considers self-transfer reactions to explain heme delivery across the cell wall. An analogous photo-cross-linking strategy was employed to map transient interactions between IsdC and IsdE transporters. These experiments identified a key structural element involved in the rapid and specific transfer of heme from IsdC to IsdE. The resulting structural model was validated with a chimeric version of the homologous transporter IsdA. Overall, our results show that the ultra-weak interactions between Isd transporters are governed by bona fide protein structural motifs.

Iron is an essential nutrient for the proliferation of Staphylococcus aureus during bacterial infections. The iron-regulated surface determinant (Isd) system of S. aureus transports and metabolizes iron porphyrin (heme) captured from the host organism. Transportation of heme across the thick cell wall of this bacterium requires multiple relay points. The mechanism by which heme is physically transferred between Isd transporters is largely unknown because of the transient nature of the interactions involved. Herein, we show that the IsdC transporter not only passes heme ligand to another class of Isd transporter, as previously known, but can also perform self-transfer reactions. IsdA shows a similar ability. A genetically encoded photoreactive probe was used to survey the regions of IsdC involved in self-dimerization. We propose an updated model that explicitly considers self-transfer reactions to explain heme delivery across the cell wall. An analogous photo-cross-linking strategy was employed to map transient interactions between IsdC and IsdE transporters. These experiments identified a key structural element involved in the rapid and specific transfer of heme from IsdC to IsdE. The resulting structural model was validated with a chimeric version of the homologous transporter IsdA. Overall, our results show that the ultra-weak interactions between Isd transporters are governed by bona fide protein structural motifs.
Staphylococcus aureus is a bacterium that causes life-threatening infections (1)(2)(3)(4). In recent years, the number of strains of S. aureus resistant to antibiotics of last resort ("superbugs") has increased at an alarming pace (5)(6)(7). It is critical to develop new drugs to reduce the risks posed by this bacterium to public health.
The so-called Isd (iron-regulated surface determinant) system is an attractive target to battle S. aureus. Inactivation of the Isd system would interfere with the uptake of nutritional iron necessary for the proliferation of this bacterium during infection (8 -10). The Isd system comprises 12 proteins that sequester heme molecules from the host organism to extract the iron atom contained in it (supplemental Fig. S1) (11)(12)(13).
Iron acquisition is initiated with the capture of heme from hemoglobin by the extracellular receptors IsdH and IsdB. Heme molecules are subsequently transferred to the intermediate transporters IsdA and IsdC, both of which are anchored to the cell wall via specific sorting motifs (11,14). IsdA and IsdC must relay their cargo throughout the thick cell wall of the bacterium to membrane-anchored IsdE. IsdE, together with IsdF (and possibly other proteins), catalyzes the passage of heme across the plasma membrane. Finally, the cytoplasmic enzymes IsdG and IsdI release the iron atom from the porphyrin ring.
Structural studies have showed that the IsdA and IsdC transporters, as well as the C-terminal domains of IsdH and IsdB, share a common heme-binding domain known as NEAT (near transporter) (supplemental Movie 1) (15)(16)(17)(18)(19)(20)(21). This domain is characterized by a rather hydrophobic pocket and displays a conserved tyrosine residue that coordinates the metal atom of the porphyrin ring. NEAT domains have been identified also in other bacterial pathogens such as Bacillus anthracis and Streptococcus pyogenes, thus expanding the therapeutic potential of this class of proteins (22,23).
On the other hand, the heme transporter IsdE does not possess a NEAT domain (24). Instead, IsdE belongs to class III of the periplasmic heme-binding proteins, which are characterized by a distinctive bilobular architecture (25,26). The crystal structure of IsdE with heme bound shows that the iron atom is coordinated by the axial ligands His-229 and Met-78 of the protein, although only His-229 is essential for heme uptake in vivo (24).
A key aspect of heme relay and transport in the Isd system involves the recognition mechanism between NEAT domains. A recent NMR study has proposed that heme is transferred from IsdA to IsdC via an ultra-weak affinity "handclasp" complex (27). The term ultra-weak affinity indicates transient interactions with binding constants in the millimolar range (27). According to the handclasp model, NEAT domains juxtapose their 3 10 -helices and ␤7/␤8 strands during heme transfer. This mechanism involves a transient stereospecific complex between IsdA and IsdC. A separate study has suggested that NEAT domains may undergo self-exchange heme transfer reactions, although no experimental evidence has been reported (21). This hypothesis could explain heme relay across the entire length of the cell wall, estimated at ϳ30 nm (28), with only two classes of anchored transporters, IsdA and IsdC.
The transfer of heme from IsdC to IsdE constitutes the last step in the heme pathway across the cell wall before passage through the membrane (29,30). Interestingly, heme is transferred from IsdC to IsdE at a markedly slower rate than between NEAT domains (29). This step shows exquisite selectivity because IsdE cannot receive heme from the homologous NEAT transporter IsdA. Kinetic data suggest that the transfer reaction is activated via a transient IsdC-heme-IsdE complex (29), although direct evidence of their physical interaction at the molecular level has not been demonstrated.
Herein, we reveal the mechanism of self-dimerization between IsdC transporters, demonstrating the concept of the self-transfer reaction in the Isd system. We also mapped the contact interface between IsdC and IsdE. These data show that the ␤7/␤8 strand of IsdC stands in close proximity to IsdE during their short-lived encounter. Overall, our results demonstrate that ultra-weak protein-protein interactions between Isd proteins are governed by specific structural motifs.

EXPERIMENTAL PROCEDURES
Materials-Vector pEVOL, encoding a Tyr-tRNA synthetase suitable for the incorporation of photoreactive amino acid p-benzoylphenylalanine (pBPA), 4 was a gift from Prof. P. G. Schultz (The Scripps Research Institute). pBPA was purchased from Bachem (Bubendorf, Switzerland). Hemin and thrombin were obtained from Sigma-Aldrich. Murine horseradish peroxidase-conjugated anti-His IgG 1 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA), and ECL reagents were from GE Healthcare. Other chemicals were purchased from Wako (Osaka, Japan). Primers were synthesized by Operon, and their sequences are listed below or in supplemental Table S1.
Cloning, Expression, and Purification of IsdE-IsdE (residues 32-289) was cloned into a modified pET26 vector (containing the sequence of the tobacco etch virus cleavage site) using forward primer 5Ј-GGAATTCCATATGGGCGAATTCAGA-ATCGTACC-3Ј and reverse primer 5Ј-ACCGCTCGAGCTT-ATAAAATAAATCATATAATTGAGTCATTGCC-3Ј. Expression and purification were performed as described above. When necessary, the His 6 tag was cleaved off with tobacco etch virus protease (32).
Cloning, Expression, and Purification of IsdA-IsdA (residues 58 -187) was cloned into the pET26b vector using forward primer 5Ј-GGAATTCCATATGAGCACACAAGTTTCTCA-AGC-3Ј and reverse primer 5Ј-ACCGCTCGAGTGCTGCGT-CAGCTAATG-3. Expression and purification were carried out as described above. When necessary, the His 6 tag was cleaved off with thrombin protease.
Preparation of Holoproteins-A solution of hemin dissolved in Me 2 SO was mixed in stoichiometric excess with IsdA (or IsdC) for 10 h at 4°C. Excess heme was eliminated in a DEAE column (GE Healthcare) (19). Flow-through solution containing holoprotein was loaded onto a Superdex 16/60 column (GE Healthcare) equilibrated with buffer A. The concentration of holoprotein was determined by the pyridine hemochrome assay at three different wavelengths using extinction coefficients ⑀ 418.5 ϭ 170 mM Ϫ1 cm Ϫ1 , ⑀ 526 ϭ 17.5 mM Ϫ1 cm Ϫ1 , and ⑀ 555 ϭ 34.4 mM Ϫ1 cm Ϫ1 as described previously (33).
Heme Transfer Assay-Untagged holo-IsdC (or holo-IsdA) and His 6 -tagged apo-IsdC (or apo-IsdA) were mixed for 10 min. Proteins were subsequently separated by IMAC. Their UV-visible spectra were recorded in a Jasco spectrophotometer before and after protein treatment.
Incorporation of Photoreactive pBPA-pBPA was incorporated in the DNA sequence of IsdC or IsdE as an amber codon. Procedures to produce muteins containing pBPA were adapted from previous reports (34,35). Protein expression was induced with 1.0 mM isopropyl ␤-D-thiogalactopyranoside, 4% arabinose, and medium supplemented with 1 mM pBPA. IsdC and IsdE muteins were purified as described above.
Photo-cross-linking Assay-Apo-IsdC containing photoreactive pBPA at 20 M was irradiated for 30 min at 4°C with UV light at 365 nm using a UVP B-100AP lamp. Samples were separated by SDS-PAGE and subjected to immunoblotting with anti-His 6 IgG 1 antibody. Photo-cross-linked products were monitored using an LAS-4000 image analyzer (GE Healthcare). Experiments involving IsdE contained this transporter at 20 M.
Preparation of IsdA Long Loop-The IsdA long loop containing the sequence of the ␤7/␤8 strand of IsdC was prepared with a Takara Bio mutagenesis kit using forward primer 5Ј-GATG-AAAAAGTAAATGGAAAGCCATTCAAATACAATCATA-GATATACACGCATTTGG-3Ј and reverse primer 5Ј-CACG-ACAATATGTACTTTAGTAGTTAAG-3Ј). Expression and purification were carried out essentially as described above.
Heme Transfer to IsdE-Holo-IsdC or holo-IsdA (wild-type) at 1.5 M was incubated with 15 M apo-IsdE (wild-type or variant) at room temperature, and the UV-visible spectra were collected. Alternatively, we monitored the kinetics of transfer by following changes in absorption at a wavelength of 411 nm. Time courses were fit to a single exponential equation with GNUPLOT 4.2 to obtain rate constants (Equation 1), where A is absorbance, k obs is the observed rate constant, and t is time.
Preparation of Samples for MS/MS-Photo-cross-linked dimers composed of mutein F124X (where X denotes pBPA) of IsdC and wild-type IsdE were separated by SDS-PAGE. The band containing the photo-cross-linked product was digested overnight with trypsin (Promega) at 37°C. Peptides were desalted with ZipTip C 18 microcolumns (Millipore) and concentrated to 20 l prior to analysis by nano-LC-MS/MS.
Characterization by MS/MS-Peptide analyses were performed using a linear ion trap/orbitrap mass spectrometer (LTQ Velos Orbitrap, Thermo Scientific) coupled to a nanoflow DiNa-2A LC system (KYA TECH Corp., Tokyo, Japan). Peptides were injected into a 75-m reversed-phase C 18 column at a flow rate of 10 l/min and eluted with a linear gradient of 98% solvent A (2% acetonitrile and 0.1% formic acid in H 2 O) to 50% solvent B (80% acetonitrile and 0.1% formic acid in H 2 O) at 300 nl/min. The separated peptides were sequentially sprayed from a nanoelectrospray ion source (KYA TECH Corp.) and analyzed by collision-induced dissociation, which was performed in the data-dependent mode, switching automatically between MS and MS/MS acquisition. Full-scan mass spectra (from m/z 380 to 2000) were acquired in the orbitrap at a resolution of 60,000 at m/z 400 after ion count accumulation to a target value of 500,000. The 20 most intense ions at a threshold above 2000 were fragmented in the orbitrap at a resolution of 7500 at m/z 400. The orbitrap analyzer was operated with the "lock mass" option to perform highly accurate detection (36).
Analysis of MS Data-Identification of photo-cross-linked peptides was performed as reported (37). Briefly, we built a database with virtual protein sequences in which the pBPAcontaining peptide derived from IsdC was concatenated with peptides of IsdE digested in silico. We included the NCBI Reference Sequence human protein database for internal consistency. Peptide identification was performed with Mascot Version 2.3.02 (Matrix Science, Tokyo, Japan).

RESULTS
Self-transfer Reactions-We examined the possibility of selftransfer reactions in IsdC (Fig. 1). Exchange of heme between untagged IsdC loaded with heme (holo-form) and the His 6tagged form of IsdC without ligand bound (apo-form) was monitored by UV-visible spectroscopy (Fig. 1A).
The UV-visible spectrum of holo-IsdC was characterized by two absorption peaks centered at 280 and 404 nm (Soret band) that corresponded to protein and heme, respectively (Fig. 1B). As expected, His 6 -apo-IsdC did not exhibit the characteristic Soret band of heme (Fig. 1C). Incubation of holoprotein with apoprotein for 10 min, followed by a short separation by IMAC, led to a reduction in the amount of heme attached to holo-IsdC. In relative terms, the A Soret /A 280 ratio decreased from 2.1 to 1.3. A control experiment showed no spontaneous discharge of heme during the IMAC separation step (supplemental Fig. S2).
The heme molecules lost from holo-IsdC were therefore transferred to His 6 -apo-IsdC (Fig. 1C). After the incubation step, a strong Soret band (not present before incubation) appeared in samples of His 6 -apo-IsdC. The A Soret /A 280 ratio increased from 0.07 to 0.73. The fraction of heme transferred to the apo-form during the 10-min incubation step was higher than that caused by dissociative transport (29, 38 -40). These results are consistent with a heme transfer mechanism involving activated protein complexes.
Similarly, incubation of holo-IsdA with His 6 -apo-IsdA was conducive to a net transfer of heme from its holo-to its apoform (Fig. 1, D and E). The A Soret /A 280 ratio in samples of holo-IsdA decreased from 2.3 to 0.84, whereas that in samples of apo-IsdA increased to 0.47. Overall, these two experiments demonstrated self-transfer reactions in IsdC and IsdA transporters.
Detection of IsdC-IsdC Complexes-IsdC modified with the photoreactive residue pBPA was expressed in E. coli cells. This non-natural residue reacts with C-H bonds of proteins (except vinylic or aromatic C-H bonds) under UV illumination conditions, forming carbon-carbon covalent bonds (supplemental Fig. S3) (41)(42)(43).
Transient IsdC-IsdC interactions were monitored by immunoblotting following a period of 30-min irradiation with UV light at 365 nm (Fig. 2). In these experiments, it was assumed that the appearance of cross-linked dimers demonstrates proximity between pBPA and a neighboring residue of the partner protein (41,42,44). It follows that the mutated residue should be located in the interaction site of the unmodified (wild-type) version of the protein (35,(45)(46)(47)(48)(49).
UV irradiation of mutein I48X led to the appearance of a weak band at ϳ45 kDa corresponding to the photo-crosslinked homodimer ( Fig. 2A). This band was absent in samples not irradiated. The intensity of this band was significantly lower than that of the monomeric form, suggesting a low photocross-linking yield. This is a consequence of the transient nature of IsdC-IsdC encounter complexes.
We note that this experiment was carried out with the apoform of IsdC. The presence of heme during the irradiation step considerably degraded the samples, rendering them unsuitable for analysis by Western blotting (data not shown). The use of the apo-form of IsdC as a surrogate of the holo-form in the photo-cross-linking assay assumes that the structure of the apoprotein is largely maintained in the heme-free form. This proposition is supported by a previous NMR study concluding that the structure of apo-IsdC in solution is largely preserved (17). Similar observations were reported in the crystal structures of the NEAT transporters IsdH and IsdA in their holoand apo-forms (16,18).
Overall, photo-cross-linked products were observed in eight different muteins ( Fig. 2B and Table 1). The intensity of the photo-cross-linked bands varied among the eight dimerizing muteins, which suggested an additional level of specificity during IsdC self-interaction. Importantly, all of the mutations producing dimers were located in the vicinity of the heme-binding pocket (Fig. 2C). For example, positions 48 and 52 belong to the 3 10 -helix, whereas Lys-116, Lys-124, Lys-128, and Tyr-120 belong to the ␤7/␤8 strand. Mutations far from the heme-binding pocket did not yield the photo-cross-linked dimer (Fig. 2C).
Detection of Transient IsdC-IsdE Complex-Analogous photo-cross-linking experiments were carried out to examine the transient interactions between IsdC and IsdE (27). Irradiation of apo-IsdC mutein F130X with wild-type apo-IsdE produced a new band with an apparent molecular mass of ϳ55 kDa (Fig. 3). The position of the band was consistent with the molecular mass of the IsdC-IsdE dimer (ϳ10 kDa heavier than the IsdC-IsdC self-dimer shown in Fig. 2A).
This observation was recapitulated with the IsdE mutein D101X (Fig. 3B), which addressed a residue in the vicinity of the heme-binding pocket of IsdE (supplemental Fig. S4). In con-trast, IsdE mutein Y115X, with replacement of a residue farther from the heme-binding site, did not result in a photo-crosslinked product (Fig. 3B).
We carried out a more complete mapping of the interaction surface of IsdC with 17 amber mutants ( Fig. 3C and Table 1). Eight single muteins containing pBPA at positions 52, 77, 96, 120, 124, 128, 130, and 135 gave rise to the characteristic band of the photo-cross-linked dimer. In contrast, we could not detect dimerization in muteins at positions 40, 48, 60, 71, 82, 87, 112, 116, and 141. When mapped on the crystal structure of IsdC, the mutated residues were closely clustered around the heme-binding pocket ( Fig. 3D and supplemental Fig. S5). Compared with the self-dimerization of IsdC (Fig. 2C), the transient dimerization between IsdC and IsdE showed a more obvious segregation of the residues involved in the interactions.
The SDS-PAGE band of the photo-cross-linked dimer obtained by mixing IsdC mutein K124X and wild-type IsdE was digested in-gel with trypsin protease and analyzed by nano-LC-MS/MS. A search of photo-cross-linked peptides in the mass spectra with the program Mascot (37) revealed two peaks consistent with a photo-cross-linked peptide (supplemental Fig.  S6). These two peaks were absent in samples not irradiated with UV light (supplemental Fig. S6). Although the relative abun-FIGURE 1. Self-transfer of heme between Isd transporters. A, experimental design. Untagged holoprotein and His 6 -tagged apoprotein were mixed. After a brief incubation step, the tagged and untagged proteins were separated by IMAC. Hm, heme iron. Shown are spectra of holo-IsdC (B), His 6 -apo-IsdC (C), holo-IsdA (D), and His 6 -apo-IsdA (E) before (dashed lines) and after (solid lines) a 10-min incubation, followed by IMAC. Holo-IsdC and holo-IsdA were mixed with His 6 -apo-IsdC and His 6 -apo-IsdA, respectively. Apoprotein was added in excess to a solution containing 3 M holo-form. In the spectra, the downward-pointing arrows indicate heme depletion, and the upward-pointing arrows indicate heme acquisition. dance of these peaks was small, we could confirm their identity by collision-induced dissociation (supplemental Fig. S7). The sequence of this peptide comprised residues 117-128 of IsdC (except for pBPA at position 124) and residues 50 -62 of IsdE (supplemental Table S2).
Role of ␤7/␤8 Strand of IsdC in Heme Transfer to IsdE-The above results suggested that the ␤7/␤8 strand of IsdC is essential for self-dimerization and transient interaction with IsdE. Interestingly, the ␤7/␤8 strand of IsdC is six residues longer than that of IsdA (Fig. 4, A and B). Because IsdA cannot transfer heme efficiently to IsdE (29,30), we examined the hypothesis that the length and identity of this strand are critical factors affecting the transfer of heme to IsdE. Thus, the sequence of the ␤7/␤8 strand of IsdC was grafted into IsdA by site-directed mutagenesis. We called this construct IsdA long loop.
Heme transfer assays showed a dramatic increase in the ability of IsdA long loop to relay heme molecules to IsdE compared with wild-type IsdA (Fig. 4, C and D). When incubated with apo-IsdE, the max of the Soret band shifted from 404 to 412 nm in samples of holo-IsdA long loop but not in samples of wildtype holo-IsdA (Fig. 4, C and D). In fact, after a 60-min treat- Heme is shown in dark blue. The magenta sphere corresponds to iron. The majority of residues involved in dimerization are clustered around the distal 3 10 -helix and the ␤7/␤8 strand. Molecular graphics images were produced with the UCSF Chimera package (54) using the coordinates of holo-IsdC (Protein Data Bank code 2O6P) (15).

IsdC-IsdC
The letter X indicates photoreactive pBPA. This residue is encoded by the amber codon in the pEVOL system used to produce pBPA-labeled protein (34). b Intensity was qualitatively evaluated based on visual inspection of the photocross-linked bands in the Western blot. Ϫ, no reaction; ϩ, visible band; ϩϩ, intense band. MAY 11, 2012 • VOLUME 287 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 16481
ment of holo-IsdA long loop with IsdE, the Soret band became indistinguishable from that in a control experiment with only holo-IsdE. This result indicated that all heme had been transferred to IsdE. The observed transfer rate of heme (k obs ) calculated from Equation 1 jumped by Ͼ300-fold when the loop of IsdC was grafted into IsdA long loop. Specifically, the k obs increased from 2.1 ϫ 10 Ϫ5 s Ϫ1 for samples of wild-type holo-IsdA to 6.6 ϫ 10 Ϫ3 s Ϫ1 for samples of holo-IsdA long loop (Fig. 4E). We note that the k obs for IsdA long loop was still 6-fold smaller than that for IsdC (3.8 ϫ 10 Ϫ2 s Ϫ1 ) (Fig. 4E), indicating that some additional factors affected the transfer reaction. Importantly, the k obs for holo-IsdA long loop more than doubled as the concentration of IsdE increased (Fig. 4F), demonstrating that heme transfer was accelerated via activated protein-protein complexes (29, 30, 38 -40).
Role of Iron-coordinating Residues Met-78 and His-229 of IsdE-It has been shown that the axial ligand His-229 (but not Met-78) is essential for heme uptake in vivo (24). We examined their function with our experimental setup (Fig. 5). Homogeneous holo-IsdE protein (wild-type and muteins M78A and H229A) was prepared by incubation of the apoprotein with heme, followed by ion exchange chromatography to remove unbound ligand. We observed that the Soret band of holo-M78A was very similar to that of the wild-type protein, except for a small shift in the position of the maximum (⌬ max ϭ Ϫ3 nm) (Fig. 5A). In contrast, holo-H229A displayed a much lower absorption in the Soret region, suggesting a reduced ability to bind heme.
The transfer of heme between holo-IsdC and apo-IsdE was also examined (Fig. 5B). The spectrum of a control sample containing only holo-IsdC was characterized by a max at 403 nm,  in close agreement with previous reports (29,38). Incubation of holo-IsdC with wild-type apo-IsdE caused a pronounced shift in the Soret band (⌬ max ϭ 13 nm), consistent with the complete transfer of heme to IsdE. Similarly, mutein M78A showed heme transfer ability as demonstrated by the large shift in max upon incubation with IsdC (⌬ max ϭ 8 nm). In contrast, we did not observe significant changes in the Soret band of samples incubated with mutein H229A, suggesting very little transfer reaction.
To clarify why mutein H229A cannot receive heme from IsdC, we carried out a photo-cross-linking experiment by mixing mutein A96X of IsdC and mutein H229A of IsdE. Mutein A96X was selected for its ability to produce cross-linked dimers with IsdE (Fig. 3C). The result of this experiment indicated that mutein H229A could also engage in the dimerization reaction with IsdC under UV irradiation (Fig. 5C). This observation strongly suggested that the inability of mutein H229A to accept heme from IsdC was caused by a heme-binding deficiency of the mutein.

DISCUSSION
In this study, UV-visible spectrophotometry, column chromatography, kinetic analysis, and cross-linking with a genetically encoded photoreactive amino acid (pBPA) demonstrated direct and specific interactions between heme transporters of S. aureus.
Self-transfer Reaction in IsdA and IsdC-We demonstrated that the IsdA and IsdC transporters are capable of self-transfer reactions (Fig. 1). This is a novel finding that expands our understanding of heme transport by the Isd system across the thick cell wall (ϳ30 nm) of S. aureus (28). Previous models did not explicitly address how Isd proteins, being anchored to the cell wall, could convey heme molecules from the extracellular space to the plasma membrane without breaking their anchoring contacts (Fig. 6A) (29,30,51).
In the updated model (Fig. 6B), there is no need for Isd proteins to move along the cell wall because self-transfer is permitted. IsdA is depicted near the extracellular space, whereas IsdC is placed closer to the membrane, in agreement with their cell-sorting signals (11,14). Accordingly, the selftransfer reaction permits the movement of heme along the cell wall without invoking large protein translations. However, the model in Fig. 6B may require a higher concentration of Isd proteins at the cell wall than previously anticipated (12,52). The overall driving force in this model is the increasing affinity among transporters (as suggested between IsdC and IsdE in Fig. 5) and internalization of heme across the plasma membrane.
Physical interaction between IsdC molecules was confirmed by the photo-cross-linking technique (Fig. 2). Mapping of 17 single muteins containing the photoreactive amino acid pBPA indicated that IsdC molecules approached each other through specific structural elements. Specifically, most cross-linked muteins displayed mutations at the 3 10 -helix and the ␤7/␤8 strand. These elements are also involved in transfer of Zn(II)porphyrin from IsdA to IsdC as graphically represented in the so-called handclasp model (27). Importantly, our photo-crosslinking scan showed traits of the handclasp model in the absence of heme (photo-cross-linking experiments were performed with the apo-form of IsdC). The similarity between the handclasp model and our analysis suggests that the molecular basis of heme transfer is already contained in the tertiary structure of the apoprotein.
Transient Interaction between IsdC and IsdE-Previous kinetic analysis indicated that IsdC transfers heme molecules to IsdE through an activated and transient protein-protein complex (29,30). However, our exhaustive efforts to detect this interaction by size exclusion chromatography, isothermal titration calorimetry, and Biacore analysis failed repeatedly (data not shown). These observations clearly suggest that the affinity between IsdC and IsdE is very low (Ն millimolar range) (53).
We thus examined the physical interaction between IsdC and IsdE by the photo-cross-linking technique (Fig. 3). The picture that emerged from these experiments was similar to that described above for self-dimerization of IsdC, i.e. direct protein-protein interactions involved residues in the proximity of the heme-binding pocket of IsdC ( Fig. 3 and Table 1).
We detected additional evidence of the photo-cross-linked product from a pool of tryptic peptides by nano-LC-MS/MS. The peptide identified was present at low concentration relative to other peptides (supplemental Figs. S5 and S6). These observations suggest that photo-cross-linking reactions occur at more than one location or through more than one reaction mechanism (41). These results contrast with the more uniform and abundant photodimerization product obtained between partners displaying high affinity (42)(43)(44)49).
The model depicted in Fig. 7 represents our view of the transient IsdC-IsdE complex based on the data gathered. Each protein is shown as a block of a LEGO that docks into each other around the heme-binding pocket. In this model, the "arms" of IsdC are of particular importance for the interaction. Residues near the heme-binding pocket of IsdE also participate in the transient association as demonstrated in mutein D101X. Importantly, the axial ligands His-229 and Met-78 of IsdE did not participate in IsdC-IsdE interactions (Fig. 5).
Site-directed mutagenesis of IsdA confirmed the functional relevance of the model proposed in Fig. 7. Insertion of the ␤7/␤8 strand of IsdC into the homologous NEAT transporter IsdA accelerated heme transfer to IsdE by Ͼ300-fold compared with wild-type IsdA. From the concentration dependence of k obs , we concluded that enhanced transfer in IsdA long loop was achieved via activated protein-protein interactions (27,29,30,38). Overall, our results demonstrate that the specificity of heme transfer is largely mediated by a discrete structural motif.
Conclusions-In this study, we proved the concept of selfdimerization in Isd NEAT transporters. A revised mechanism of heme transport across the cell wall now takes into account these findings. Photo-cross-linking analysis revealed specific structural elements involved in self-dimerization of IsdC. Using an analogous strategy, we also showed that the ␤7/␤8 strand of IsdC is necessary for the rapid and specific transfer of heme between IsdC and IsdE.