Genetics and regulation of heme iron transport in Shigella dysenteriae and detection of an analogous system in Escherichia coli O157:H7

Shigella species can use heme as the sole source of iron. In this work, the heme utilization locus of Shigella dysenteriae was cloned and characterized. A cosmid bank of S. dysenteriae serotype 1 DNA was constructed in an Escherichia coli siderophore synthesis mutant incapable of heme transport. A recombinant clone, pSHU12, carrying the heme utilization system of S. dysenteriae was isolated by screening on iron-poor medium supplemented with hemin. Transposon insertional mutagenesis and subcloning identified the region of DNA in pSHU12 responsible for the phenotype of heme utilization. Minicell analysis indicated that a 70-kDa protein encoded by this region was sufficient to allow heme utilization in E. coli. Synthesis of this protein, designated Shu (Shigella heme uptake), was induced by iron limitation. The 70-kDa protein is located in the outer membrane and binds heme, suggesting it is the S. dysenteriae heme receptor. Heme iron uptake was found to be TonB dependent in E. coli. Transformation of an E. coli hemA mutant with the heme utilization subclone, pSHU262, showed that heme could serve as a source of porphyrin as well as iron, indicating that the entire heme molecule is transported into the bacterial cell. DNA sequences homologous to shu were detected in strains of S. dysenteriae serotype 1 and E. coli O157:H7.

The Shigella species have the ability to thrive in a variety of different habitats. They can grow extra-or intracellularly in the host as well as in the environment (8). This ability to occupy multiple niches implies that these bacteria can respond to changes in the environment and adapt to a variety of conditions, including changes in temperature, osmolarity, oxygen content, and concentrations of important nutrients (8).
One nutrient which is essential for all pathogenic bacteria is iron. Although iron is abundant in nature, its availability is restricted because of its poor solubility in aerobic environments (33). In mammalian hosts, most iron is sequestered. The majority of iron is found in erythrocytes as heme in the protein hemoglobin (3,4). Host cells also contain ferritin, a storage form of iron. High-affinity iron-binding proteins, such as transferrin and lactoferrin, complex any extracellular iron (25,28). Thus, the intracellular and extracellular pools of free iron are extremely small.
Enteric pathogens have several mechanisms to acquire iron. One method is the production of siderophores (5). Under conditions of iron starvation, these low-molecular-weight iron chelators are synthesized and secreted from the cell. Gramnegative bacteria transport ferric siderophores into the cell using specific outer membrane protein receptors, periplasmic proteins, and inner membrane protein complexes (26).
Alternatively, pathogenic bacteria may obtain iron directly from host iron sources. Pathogenic Neisseria spp. use specific receptors to obtain iron from human transferrin and lactoferrin (2). Heme proteins also provide an iron source for some pathogens. Haemophilus influenzae has been shown to acquire iron from hemoglobin, hemoglobin bound to haptoglobin, and heme-hemopexin (29), and mesophilic Aeromonas species grow well with hemoglobin, alone or complexed to haptoglobin, as the only iron source (22).
The free heme molecule itself may be a potential source of iron for pathogens. Vibrio cholerae (10), Plesiomonas shigelloides (6), and Yersinia enterocolitica (36) have been shown to possess heme transport systems. A number of other bacterial species, including Aeromonas species (22) and Vibrio anguillarum (23), are able to grow in iron-deficient media with heme as the only iron source. Shigella spp. produce one or more siderophores (28). However, mutants unable to make siderophores are able to invade and multiply intracellularly, indicating the ability to acquire iron by other methods (16). Assays for utilization of potential iron sources found in the host indicate that the Shigella species are able to grow with heme as a sole iron source and that the ability to use heme as an iron source is siderophore independent (16). This investigation was undertaken to characterize the Shigella dysenteriae heme utilization system. Colonies were isolated on L agar containing Tc and then screened for growth on L agar containing EDDA and hemin. Restriction enzymes, calf intestinal phosphatase, and T4 DNA ligase were purchased from New England Biolabs, Inc. (Beverly, Mass.), and used according to the supplier's directions.
Transposon insertional mutagenesis. Heme utilization clones were mutagenized with a Cm r mini-Tn10 construct, derivative 105 (13), which was amplified in NK5012. Mutagenesis was performed in MM51 by the method of Kleckner et al. (13). Cm r Tc r colonies were pooled, and plasmid DNA was isolated by alkaline lysis (32). E. coli 1017 was transformed (18) with the pooled plasmids and screened for heme utilization on L agar containing Cm, Tc, EDDA, and hemin.
Crystal violet sensitivity. L agar plates were seeded with 100 l of an overnight culture. A sterile 6-mm-diameter disk spotted with 10 l of a 1-mg/ml crystal violet solution was applied to the center of the agar. The plates were incubated overnight and observed for growth inhibition. 35 S-minicell analysis. Volumes of L broth (300 ml each) containing EDDA, hemin, and the appropriate antibiotics were inoculated 1:50 with overnight cultures of E. coli P678-54 transformed with the indicated plasmids. After an overnight incubation, the minicells were purified on two sucrose step gradients by the method of Meagher et al. (24). Minicells were labeled with 50 Ci of 35 S Trans-label (ICN Biomedicals, Costa Mesa, Calif.) per ml. Labeling was performed in the presence of hemin and EDDA for heme utilization clones and mini-Tn10 derivatives.
Cell fractionation. For the cellular localization of proteins, cultures were grown overnight in iron-restricted (L plus EDDA) or iron-replete (L plus FeSO 4 ) medium. The cultures were centrifuged at 8,000 ϫ g, resuspended in sodium phosphate buffer (10 mM NaPO 4 , 5 mM MgSO 4 [pH 7.0]), and frozen at Ϫ80ЊC. The cells were thawed and lysed by passage through a French press (Aminco, Urbana, Ill.). Total membranes were prepared by the method of Inouye and Guthrie (12) and treated with 0.5% N-laurylsarcosine (CIBA-GEIGY Corporation, Ardsley, N.Y.) to selectively solubilize the inner membrane (7). Inner and outer membrane fractions were then separated by centrifugation at 100,000 ϫ g, and the outer membrane pellet was resuspended in Laemmli solubilization buffer (15). For hemin affinity chromatography, sphero-plasts from mid-log, iron-starved cultures were lysed by sonication and total membranes were isolated. Membrane fractions were then separated by isopycnic centrifugation, and the outer membranes were isolated by the method of Osborn and Munson (27). The final pellets were resuspended in 50 mM Tris (pH 8.0)-1 M NaCl. The protein concentration was determined by the Bio-Rad Protein Assay (Bio-Rad Laboratories, Hercules, Calif.).
Hemin affinity chromatography. Outer membranes (150 g of protein) were preincubated with or without hemin (40 to 400 nmol) in a volume of 100 l with gentle mixing for 1 h at 37ЊC. Preincubated membranes were then added to 50 l of packed hemin-agarose (235 nmol of hemin; Sigma Chemical Company) which had been equilibrated in 50 mM Tris (pH 8.0)-1 M NaCl and incubated with mixing for 1 h at 37ЊC. After binding, solubilization buffer (9) was added to give a 250-l total volume; final solubilization buffer concentrations were 10 mM Tris (pH 7.8), 150 mM NaCl, 10 mM EDTA, 1% (vol/vol) Triton X-100, 0.2% (wt/vol) sodium deoxycholate, and 0.1% (wt/vol) sodium dodecyl sulfate (SDS). Samples were solubilized for 90 min at 37ЊC with mixing. Nonadherent proteins were removed by centrifugation at 750 ϫ g for 5 min, and the supernatants were discarded. Resin beds were then washed three times with a high-salt solubilization solution (50 mM Tris, pH 8.0; 1 M NaCl; 10 mM EDTA; 0.75% Triton X-100; 0.15% sodium deoxycholate; 0.075% SDS) and once with a final wash of 50 mM Tris (pH 8.0)-1 M NaCl. The wash volumes were 1 ml, and the samples were centrifuged between washes as described above. Heme-binding proteins were eluted from the resin by adding 100 l of Laemmli solubilization buffer (15), boiling for 5 min, chilling on ice for 1 min, and centrifuging as described above.
TonB bioassays. Cultures of RK4321 and RK4338 with and without heme utilization plasmids were grown to log phase and used to inoculate pour plates of L agar plus EDDA. The inoculum for RK4321 was 0.5 ϫ 10 4 to 1 ϫ 10 4 CFU/ml, and that for RK4338 was 1 ϫ 10 5 to 5 ϫ 10 5 CFU/ml. After solidification, each plate was spotted with 5 l of 80 M hemin and 5 l of 200 M ferrichrome. A sterile disk containing 20 l of 10 mM FeSO 4 was also applied to each plate. The zones of growth around the FeSO 4 , ferrichrome, and hemin were measured after 18 to 24 h of incubation at 37ЊC.
Growth of hemA mutant. E. coli RK1065 was transformed with the heme utilization plasmid pSHU262 and maintained on L agar plus Cm plus 24 M aminolevulenic acid (ALA). Individual colonies were grown in L broth plus 48 M ALA plus Cm and then were streaked onto L agar plus Cm with or without 24 M ALA and with or without hemin. Growth and colony size were determined after 18 to 24 h of incubation.

Cloning of the heme utilization locus from S. dysenteriae.
The heme utilization gene from S. dysenteriae was cloned on the basis of its ability to confer the use of hemin as an iron source upon a laboratory strain of E. coli. To achieve this, a cosmid library of S. dysenteriae chromosomal DNA was constructed in E. coli 1017, a siderophore synthesis mutant of a strain incapable of hemin transport. Colonies were screened on iron-restricted medium in the presence or absence of hemin; of 4,000 clones screened in this manner, 14 displayed the desired phenotype of healthy growth on iron-restricted medium supplemented with hemin. Two of these clones were able to thrive also on iron-poor medium without hemin because of complementation of the siderophore synthesis mutation (data not shown) and were not characterized further.
Restriction endonuclease digestion of the 12 putative heme utilization clones showed them to contain overlapping DNA fragments (data not shown). One clone, containing a 27-kb insert and designated pSHU12 (Table 1), was selected for further study. The insert was subcloned to obtain the smallest DNA fragment conferring growth on heme (Fig. 1).
To rule out the possibility that permeability changes in the cell envelope were permitting nonspecific diffusion of hemin into the cell, E. coli 1017 alone or transformed with heme utilization plasmid pSHU12, pSHU18, or pSHU9 was tested for crystal violet sensitivity. No significant differences between the clones and the original strain, 1017, were observed; all strains were observed growing within 2 to 4 mm of the disk.
Insertional mutagenesis of pSHU9. pSHU9 was mutagenized with a mini-Tn10 derivative (13). Mutants that had lost the ability to grow with hemin as the iron source were isolated, and the insertion sites were mapped (Fig. 1). These insertions, four of which are depicted in Fig. 1, were clustered in a 1.2-kb region which spanned a ClaI site located within a 3.7-kb KpnI fragment. The 3.7-kb KpnI fragment was subcloned to create pSHU37 (Fig. 1). However, pSHU37 did not allow growth with hemin as the only iron source in E. coli 1017, indicating that sequences located outside this KpnI fragment are required for heme transport. An overlapping 2.6-kb EcoRV fragment was isolated and cloned into the EcoRV site of pACYC184. This subclone, designated pSHU262, did yield the phenotype of heme utilization (Fig. 1).
Identification of proteins encoded by the heme utilization plasmid. The proteins encoded by pSHU9, by its heme utilization-negative derivatives pSHU37 and pSHU912 (pSHU9:: mini-Tn10), and by the minimal heme utilization clone pSHU 262 were identified by minicell analysis. pSHU9 encodes at least two proteins, a 70-kDa protein and a 36-kDa protein (Fig.  2, lanes 2 and 6). The mini-Tn10 insertion mutant pSHU912, which had lost the ability to grow with hemin as the sole iron source, no longer produced the 70-kDa protein but did produce the 36-kDa protein (Fig. 2, lane 3). Upon longer exposure of the autoradiogram, a 50-kDa band, which may represent a truncated form of the 70-kDa protein, could be seen (Fig. 2, middle arrow). The EcoRV subclone, pSHU262, which does allow the transport of hemin, encodes only the 70-kDa protein (Fig. 2, lane 8). This indicates that the 70-kDa protein designated Shu (Shigella heme uptake) is essential and sufficient to yield the phenotype of heme utilization. The KpnI subclone, pSHU37, which contained part of the presumed heme utilization region but did not restore the phenotype, encoded only the 36-kDa protein (Fig. 2, lane 5).

Cellular localization of the 70-kDa protein and demonstration of iron regulation.
To determine the cellular location of the 70-kDa protein and whether it showed the iron-stressinducible phenotype common to many iron uptake systems, E. coli containing either the vector pAT153 or the cloned heme utilization system pSHU9 was grown in both iron-replete medium and iron-poor medium. The 70-kDa protein was found in the outer membrane compartment of the cell (Fig. 3). Expression of the 70-kDa outer membrane protein was induced in iron-poor media (Fig. 3, lane 3). To verify that the 70-kDa protein observed in minicells (Fig. 2, lanes 2, 6, and 8) and the 70-kDa iron-regulated outer membrane protein noted here were the same protein, an additional protein gel was run in which aliquots of both outer membrane proteins from pSHU9 grown in iron-restricted medium and the [ 35 S]methionine proteins encoded by pSHU9 were loaded into the same well. Results showed that these two proteins comigrated (data not shown).
The 70-kDa protein is a heme-binding protein. If this 70-kDa iron-regulated protein serves as the outer membrane receptor for heme transport, it should be a specific heme-binding protein. The ability of the 70-kDa protein to bind heme was assessed by hemin-agarose batch affinity chromatography. Outer membranes from E. coli P678-54 carrying either a heme utilization plasmid, pSHU9, or the vector pAT153 were isolated and combined with hemin-agarose resin as described in Materials and Methods. The primary hemin-binding protein in outer membranes from cells carrying pSHU9 (Fig. 4, lane 3) comigrated with the 70-kDa Shu protein (Fig. 4, lane 1). No protein of this size was detected in eluants from the same strain carrying the vector alone (Fig. 4, lane 2). Specific binding of the protein to heme was confirmed by competition experiments in which the membranes were preincubated with a range of hemin concentrations. Binding of the 70-kDa protein was drastically reduced by preincubation with hemin; binding was inversely proportional to the amount of free hemin added (Fig.  4, lane 4, and data not shown). Even at the highest concentration of added hemin, however, some binding of the 70-kDa protein to the resin was observed. This incomplete competition with exogenous hemin is consistent with the observation of Lee (17) in studies of a heme-binding protein from H. influenzae type b. Because of solubility constraints, the highest concentration of free hemin used in these assays was only 1.7-fold higher than the concentration of hemin in the hemin-agarose resin. Heme utilization reconstituted in E. coli K-12 is TonB dependent. In a number of systems, the E. coli TonB protein has been found to be essential for transport of iron complexes across the outer membrane. To ascertain if TonB was necessary for the transport of heme iron by the S. dysenteriae Shu protein, an isogenic pair of E. coli strains, RK4321 (entA) and RK4338 (entA ⌬tonB), was transformed with pSHU262. Because of the defect in enterobactin synthesis, neither strain is capable of growth on iron-restricted medium. RK4321/ pSHU262, the TonB ϩ strain harboring the heme utilization plasmid, was able to grow on iron-limited medium supplemented with hemin (Table 2). In contrast, the tonB mutant strain, RK4338, carrying the same heme utilization clone showed no growth on this medium ( Table 2).
The transport of ferrichrome into the E. coli cell is known to be dependent on the TonB protein. The TonB ϩ strain, in the presence or absence of pSHU262, was able to grow on the same iron-chelated medium supplemented with ferrichrome, whereas the tonB mutant, with or without pSHU262, was unable to grow, thus confirming the TonB phenotype of the strains ( Table 2). As expected, all strains grew in this iron-poor medium in the presence of excess FeSO 4 ( Table 2). These results indicated that the heme utilization system of S. dysenteriae, when reconstituted in E. coli, requires a functioning TonB protein.
The entire heme moiety is transported into the bacterial cell. The iron requirements of S. dysenteriae could be satisfied either by extracellular removal of iron from heme followed by transport of the iron into the bacterium or by transport of the entire heme moiety into the cell. To distinguish between these mechanisms, the ability of pSHU262 to permit hemin to satisfy the porphyrin requirement of an E. coli hemA mutant was determined. RK1065 (hemA) is unable to synthesize heme because of a defect in production of ␦-ALA, an intermediate in heme biosynthesis. This strain showed no growth aerobically (no isolated colonies) on L agar unless ALA was added (colony diameter, 2 mm). No growth of RK1065 was observed on L plus hemin, presumably because this strain cannot transport the compound. RK1065 transformed with pSHU262 still failed to grow on unsupplemented medium, but it grew on either medium supplemented with hemin (colony diameter, 1 mm) or ALA (colony diameter, 2 mm). This indicated that the heme utilization clone enabled the transport of the intact heme molecule into the hemA mutant cell, bypassing the defective heme biosynthesis pathway.
Detection of homologous DNA sequences. An 800-bp internal KpnI-ClaI fragment (Fig. 1) from the gene encoding the 70-kDa protein was used to probe KpnI-digested chromosomal DNA from strains of S. dysenteriae serotypes 1 to 4 and E. coli serotypes O157:H7 and O26:H11. Chromosomal DNA from S. dysenteriae O-4576 was used as a positive control and showed the 3.7-kb KpnI fragment from which the probe was derived (Fig. 5). Hybridization of the probe to DNA of other serotype 1 strains of S. dysenteriae (Fig. 5) but not to DNA of S. dysenteriae serotypes 2, 3, or 4 ( Table 3) was observed. This internal fragment was also used to probe KpnI-digested DNA from the naturally occurring plasmids of S. dysenteriae O-4576, the strain from which shu was cloned. No hybridization with plasmid DNA was detected, verifying that the heme utilization gene was located on the chromosome (data not shown).
The production of a hemolytic uremic syndrome by some E. coli strains suggested that these strains might have a similar heme transport system. Nine strains of E. coli O157:H7 were assayed for heme transport and for the presence of DNA homologous to the shu gene. All isolates tested were able to grow with heme as a sole iron source (Table 3). Additionally, DNA from each of the isolates, one of which is shown in Fig.  5, hybridized to the KpnI-ClaI probe, while no homology was seen with DNA from a laboratory E. coli strain, HB101 (data not shown). This suggests that a heme transport system related to the one in S. dysenteriae exists in E. coli O157:H7. The size of the KpnI fragment detected by hybridization was approximately 20 kb in all of the O157:H7 strains, compared with 3.7 kb in S. dysenteriae serotype 1. An enterohemorrhagic E. coli of a different serotype, O26:H11, tested negative for both the utilization of heme iron and the presence of a DNA sequence homologous to the heme transport gene of S. dysenteriae (Table 3).

DISCUSSION
The adaptation of bacteria to their environment is well illustrated in the variety of mechanisms employed for acquisition of the essential element iron. These range from highaffinity siderophore-mediated systems which solubilize and transport iron in aerobic environments to expression of receptors for binding of host iron carrier proteins. The largest potential pool of iron within a mammalian host is the heme molecule. Thus, it is not surprising that pathogens which encounter heme in their host have transport systems to take advantage of this abundant iron source.
The Shigella species, which produce a bloody diarrhea, are able to use heme as their sole iron source. In order to determine the mechanism of heme utilization, a DNA fragment conferring this phenotype was cloned in E. coli. Subcloning and transposon insertional mutagenesis accompanied by minicell analysis of the encoded proteins demonstrated a correlation between heme utilization and expression of a 70-kDa protein.
Mutations which disrupt expression of this protein abolish the microbe's ability to use heme as an iron source, indicating that this protein is essential for heme utilization. Transformation of E. coli with a heme utilization plasmid which encodes only the 70-kDa protein allows heme transport, which confirms that this protein is sufficient for the phenotype of heme utilization.
The 70-kDa protein was observed in outer membrane fractions isolated by either Sarkosyl extraction or sucrose density gradient centrifugation. This protein has the ability to bind heme and could be affinity purified on hemin-agarose. Binding to hemin-agarose was specific; the binding was inhibited by the addition of free hemin. The location of the protein in the outer membrane and its ability to bind heme indicate that it could serve as the receptor for heme in this heme uptake system. As has been noted for many well-characterized receptors for bacterial iron transport systems, the expression of the Shu protein is regulated by iron. Maximum production of the heme receptor occurs when iron levels are limiting.
The mechanism of transport in S. dysenteriae appears to involve uptake of the entire heme molecule into the bacterial cell. The presence of the heme utilization plasmid, pSHU262, permitted an E. coli hemA mutant to use hemin as a source of porphyrin. This DNA, which encodes only the Shu protein, appears to allow the transport of the entire heme moiety, providing both an iron source and a porphyrin source for E. coli.
The heme receptor of S. dysenteriae appears to be dependent on the TonB protein for the transport of hemin in E. coli. It may be that the S. dysenteriae TonB analog will be required for heme transport in that organism. Comparison with other TonB-dependent iron transport systems suggests that additional proteins may be required for transport of heme into the cell, but additional genes required for heme utilization have not been identified in this system. A gene encoding a 36-kDa protein of unknown function and cellular location is linked to shu. It has not been determined whether this protein plays a role in heme transport in S. dysenteriae, but its presence is unnecessary to reconstitute the phenotype of heme utilization in E. coli.
Heme transport proteins with molecular masses ranging from 70 to 80 kDa have been identified in V. cholerae, P. shigelloides, and Y. enterocolitica. The V. cholerae HutA heme receptor is 77 kDa (10), while the HemR protein of Y. enterocolitica is 78 kDa (36). A 77-kDa protein has been observed by minicell analysis of a cosmid clone containing the heme utilization system of P. shigelloides (35). Each of these heme uptake proteins has been shown to exist in the outer membrane of the respective bacterium, and expression of the proteins is regulated by the concentration of iron in the environment. Transport of the intact heme molecule is another instance of similarity among these heme transport systems. The HutA heme receptor (10), the HemR protein (36), and the cloned sequences containing the heme-iron uptake system of P. shigelloides (6) have been shown to allow the transport of hemin as a porphyrin source. To utilize hemin as an iron source, the heme receptor of Y. enterocolitica, HemR, requires the presence of the Y. enterocolitica Ton B protein (36). In contrast, the HutA heme receptor of V. cholerae is able to transport hemin normally in an E. coli tonB mutant (10). This may not represent a significant difference between HutA and the other transport proteins, as there is some evidence that HutA requires a V. cholerae gene which may serve the same function as E. coli tonB (11). Thus, these heme utilization proteins and the Shu protein of S. dysenteriae may belong to a family of iron-regulated, outer membrane protein receptors.
Because the Shigella species and E. coli are closely related genetically, the possible presence of this heme utilization system in selected strains of E. coli was determined. A homologous DNA sequence was detected in E. coli O157:H7 strains which cause hemorrhagic colitis that, as is the case with S.