A novel heme protein, the Cu,Zn-superoxide dismutase from Haemophilus ducreyi.

Haemophilus ducreyi, the causative agent of the genital ulcerative disease known as chancroid, is unable to synthesize heme, which it acquires from humans, its only known host. Here we provide evidence that the periplasmic Cu,Zn-superoxide dismutase from this organism is a heme-binding protein, unlike all the other known Cu,Zn-superoxide dismutases from bacterial and eukaryotic species. When the H. ducreyi enzyme was expressed in Escherichia coli cells grown in standard LB medium, it contained only limited amounts of heme covalently bound to the polypeptide but was able efficiently to bind exogenously added hemin. Resonance Raman and electronic spectra at neutral pH indicate that H. ducreyi Cu,Zn-superoxide dismutase contains a 6-coordinated low spin heme, with two histidines as the most likely axial ligands. By site-directed mutagenesis and analysis of a structural model of the enzyme, we identified as a putative axial ligand a histidine residue (His-64) that is present only in the H. ducreyi enzyme and that was located at the bottom of the dimer interface. The introduction of a histidine residue in the corresponding position of the Cu,Zn-superoxide dismutase from Haemophilus parainfluenzae was not sufficient to confer the ability to bind heme, indicating that other residues neighboring His-64 are involved in the formation of the heme-binding pocket. Our results suggest that periplasmic Cu,Zn-superoxide dismutase plays a role in heme metabolism of H. ducreyi and provide further evidence for the structural flexibility of bacterial enzymes of this class.

Bacteria have evolved different strategies to acquire iron, an essential element for all organisms, and to protect themselves from its potential toxic effects (1). The problem of iron availability is particularly important for those microorganisms that colonize animal hosts, where most of the extracellular iron is tightly bound to plasma proteins. Strategies to overcome limitation in iron availability include the production of sid-erophores and the use of host compounds containing iron, such as heme, hemoglobin, transferrin, and lactoferrin (2).
Haemophilus ducreyi, the etiological agent of the sexually transmitted human genital ulcerative disease known as chancroid, is unable to synthesize heme, and therefore must obtain this from its host. Although chancroid is relatively uncommon in the United States and Western Europe, it is one of the most prevalent sexually transmitted diseases worldwide and is an important cause of morbidity in the developing world. Recent studies have provided evidence that chancroid is a significant cofactor for the heterosexual spread of the human immunodeficiency virus, type 1, increasing the risk of acquiring human immunodeficiency virus infection 2-5-fold (3,4). This observation, together with the emergence of antibiotic-resistant strains in areas where this disease is widespread, has stimulated increasing research effort aimed at understanding the pathogenesis of chancroid and characterizing H. ducreyi virulence factors, including proteins involved in heme uptake, that might be exploited in developing new strategies to prevent infection.
Free heme, hemoglobin, or catalase can supply the iron and heme requirement for H. ducreyi in vitro (5,6). Little is known about mechanisms of heme uptake in this bacterium, but a few proteins involved in heme and hemoglobin utilization have been cloned (7,8). Recently, it has been observed that an H. ducreyi hemoglobin receptor mutant is unable to initiate disease, thus suggesting that the main source of heme in vivo is hemoglobin (9).
A few other putative virulence determinants of H. ducreyi have been described, including a hemolytic cytotoxin (10), a soluble cytolethal distending toxin (11), a lipooligosaccharide molecule (12), the unique pili which may favor cell attachment (13), and Cu,Zn-superoxide dismutase (Cu,Zn-SOD) 1 (14,15). This latter virulence factor is shared by several Gram-negative bacteria, including some important pathogens (16). In view of its periplasmic location, it has been proposed that bacterial Cu,Zn-SOD, by its ability to scavenge the superoxide radical at diffusion-controlled rates, could provide protection from the oxygen free radicals produced in the respiratory burst of the phagocytic host defense. This hypothesis is supported by recent studies that have shown that periplasmic Cu,Zn-SOD could confer protection from oxidative damage caused by exposure to phagocytic cells (17,18) and that sodC null mutants of Brucella abortus, Salmonella typhimurium, and Neisseria meningitidis are attenuated in animal infection studies (19 -23). In the case of H. ducreyi, it has been shown that Cu,Zn-SOD confers some protection against superoxide generated in vitro (15) and that a Cu,Zn-SOD-deficient strain is more sensitive than wild type to killing by neutrophils in a swine model of chancroid (24).
Here we report that the recombinant enzyme expressed in Escherichia coli is able to bind heme with efficiency and specificity, and we suggest that a single heme moiety is bound at the enzyme-dimer interface. Our data suggest that, beside its ability to protect the bacterium from oxidative damage, Cu,Zn-SOD might also play a role in heme transport or detoxification in H. ducreyi.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Media-E. coli K12 strains and plasmids used in this work are listed in Table I. All the strains were grown in standard Luria-Bertani (LB) medium, with the exception of the H500 strain that required the addition of 0.2% glucose. Heme binding by Cu,Zn-SOD expressed in 71/18 cells was also tested in M9 minimal medium (40).
Plasmid pPHSOD directing the expression of wild type human Cu,Zn-SOD in the bacterial periplasmic space was obtained as follows: the sequence coding for the enzyme was amplified by PCR using the oligonucleotides 5Ј-ATCCATGGCCGCGACGAAGGCCGTCTGC and 5Ј-CAGAATTCTTATTGGGCGATCCCAATTAC and plasmid pKH (39) as a template. Amplified DNA was digested with NcoI and EcoRI and inserted in the corresponding sites of plasmid pHEN-1 (34).
Protein Expression and Purification-Overexpression of wild type and mutants Cu,Zn-SODs was carried out in E. coli 71/18 cells grown at 37°C in LB medium containing 100 g/ml ampicillin. When cells reached an A 600 value of 0.5, Cu,Zn-SOD expression was induced by the addition of 0.1 mM isopropyl-␤-D-thiogalactopyranoside, 0.25 mM CuSO 4 , and 10 M ZnSO 4 . Cells were harvested by centrifugation 16 h after induction. Wild type H. ducreyi Cu,Zn-SOD was purified from periplasmic extracts by immobilized metal affinity chromatography on a nickel-nitrilotriacetic acid (NTA) resin (Qiagen), as described (35). The affinity purified enzyme was further subjected to gel filtration chromatography on a High Load 16/60 Superdex 75 gel filtration FPLC (Amersham Pharmacia Biotech). Samples of H. ducreyi Cu,Zn-SOD were subjected to different ion exchange chromatography tests on Mono Q HR 5/5 or Mono S HR FPLC columns (Amersham Pharmacia Biotech) with different buffers and pH conditions. Partial separation of the heme-containing Cu,Zn-SOD from the heme-devoid enzyme was obtained using the cationic exchange column Mono S, equilibrated with 20 mM potassium phosphate, pH 6.5, eluting the enzyme with a 0 -0.1 M NaCl gradient.
Pure heme-devoid enzyme was obtained by E. coli H500 cells grown for 24 h in LB medium supplemented with 0.2% glucose, 0.05 mM isopropyl-␤-D-thiogalactopyranoside, 0.125 mM CuSO 4 , and 5 M ZnSO 4 . It should be noted that H500 cells grow very poorly, and after 24 h of growth at 37°C the A 600 of cultures never exceeded 0.5.
Polyacrylamide Gel Electrophoresis and Heme Staining-Polyacrylamide gel electrophoresis under denaturing conditions (SDS-PAGE) was performed according to Laemmli (42) using a 12.5% resolving gel and a 4.5% stacking gel. Samples were heated at 100°C for 5 min in a buffer containing 0.1% SDS and 10% ␤-mercaptoethanol. Gels were stained with Coomassie Brilliant Blue. In order to detect covalent heme binding by Cu,Zn-SOD, SDS-PAGE gels were stained for heme-dependent peroxidase activity using 3,3Ј,5,5Ј-tetramethylbenzidine as an oxi-  (43). Before electrophoresis samples were heated for 5 min at 100°C in a sample buffer containing 0.1% SDS but devoid of ␤-mercaptoethanol. Electrophoresis was carried out using the buffer system described by Laemmli (42), with the addition of 2 mM EDTA. Gels were immersed in a 30:70 methanol, 0.25 M sodium acetate, pH 5.0 solution, containing 1.25 mM 3,3Ј,5,5Јtetramethylbenzidine and kept in constant shaking at room temperature. After 30 min H 2 O 2 was added to 26 mM. After shaking for a further 15 min, the same quantity of H 2 O 2 was again added. Gels were photographed immediately.
Proteins and Heme Analyses-Protein concentration was determined by the Lowry method, using bovine serum albumin as a standard (44). Cu,Zn-SOD molar concentrations were determined assuming a molecular mass of 38,172 and 32,972 Da for dimeric wild type and N-terminal deleted H. ducreyi enzyme, respectively. Superoxide dismutase activity was assayed by the pyrogallol method (45).
Pyridine hemochromes of Cu,Zn-SOD samples were obtained by diluting the enzyme with alkaline pyridine and then adding solid sodium dithionite to the protein solution. Absorption spectra were recorded in a PerkinElmer Life Sciences Lambda 2 spectrophotometer thermostated at 25°C, using 1-cm path length cuvettes. Heme content in purified Cu,Zn-SOD solutions was determined from the absorbance of the ␣-peak (at 551-556 nm) of the pyridine hemochrome (46), using the ⑀ value of 34500 M Ϫ1 cm Ϫ1 for samples obtained from hemin-supplemented LB (which contain only limited amounts of covalently bound heme) or the ⑀ value of 29100 M Ϫ1 cm Ϫ1 for samples obtained from standard LB medium. It should be noted, however, that such extinction coefficients provide only an approximate value of heme concentration in Cu,Zn-SOD, due to the variable amount of covalently bound heme in each protein sample.
Acid-acetone heme extraction was carried out as described by Di Iorio (47), by dropping small amounts of Cu,Zn-SOD into a large volume excess of cold (Ϫ20°C) acid/acetone, prepared by adding 2 ml of 12 N HCl to 1 liter of acetone. After 10 min of centrifugation at 17,000 ϫ g, the supernatant was discarded, and the same operation was repeated. Once clear of all traces of acetone, the final precipitate was dissolved in 100 mM phosphate buffer, pH 7.0, and the heme content was reanalyzed.
In Vitro Heme Binding Assay-0.5 mg of heme-deficient wild type (from H500 E. coli cells) or His-64 3 Glu mutant H. ducreyi Cu,Zn-SOD (Х0.013 mol of enzyme) was dissolved in 400 l of 10 mM phosphate buffer, pH 8.0, placed in dialysis bags, and dialyzed against 1 liter of the same buffer containing 8.48 mg of hemin (13 mol) with constant stirring. After 15 h of dialysis at room temperature, the samples were extensively dialyzed against 10 mM phosphate buffer, pH 8.0 (with several changes of the buffer), to remove unbound hemin. The heme content of each sample was assessed by recording the enzyme absorption spectra in the visible region, whereas the presence of covalently bound heme was evaluated by acid/acetone extraction and by the staining of SDS-PAGE gels for heme-dependent peroxidase activity.
Spectroscopy-Resonance Raman (RR) and electronic absorption spectra were carried out in 10 mM sodium phosphate buffer at pH 7.0. The protein samples (N-del 1 and N-del 2) used for spectral analysis were obtained from cells grown in hemin-enriched LB or in standard LB, respectively; N-del 1 contained less than 10% covalently bound heme, whereas N-del 2 contained about 90% covalent heme. The ferrous forms were prepared by adding a minimum volume of fresh sodium dithionite solution to a deoxygenated buffered solution. Protein concentration was about 0.04 -0.15 mM for RR spectroscopy and about 0.05-0.1 mM for the UV-visible absorption measurements. The absorption spectra were recorded with a Cary 5 spectrophotometer. The RR spectra were obtained with excitation from the 406.7-nm line of a Kr ϩ laser (Coherent Radiation) and the 514.5-nm line of an Ar ϩ laser (Coherent Radiation, Innova/5). The back-scattered light from a slowly rotating NMR tube was collected and focused into a computer-controlled double monochromator (Jobin-Yvon HG2S) equipped with a cooled photomultiplier (RCA C31034 A) and photon-counting electronics. Polarized spectra were obtained by inserting a Polaroid analyzer between the sample and the entrance slit of the spectrometer. The depolarization ratios of the bands at 314 and 460 cm Ϫ1 of CCl 4 were measured to check the reliability of the polarization measurements using a rotating NMR tube with a back-scattered geometry. The values obtained, 0.73 and 0.0, compare favorably with the theoretical values, 0.75 and 0.0, respectively. To minimize the heating effects induced on the protein by the laser beam, RR spectra were collected using a rotating NMR tube cooled by a gentle flow of N 2 gas passed through liquid N 2 .
RR spectra were calibrated with indene as standard. The frequencies were accurate to Ϯ1 cm Ϫ1 for the intense isolated bands and to about Ϯ2 cm Ϫ1 for overlapped bands or shoulders.
Computer-assisted Molecular Modeling-A molecular model of H. ducreyi Cu,Zn-SOD was generated using the Swiss PDB-Viewer (48) starting from the x-ray structure of the highly homologous Cu,Zn-SOD from A. pleuropneumoniae (see Ref. 26, Protein Data Bank code 2aps.pdb). Eighteen residues (out of the 155 present in each monomer) were mutated in order to account for all the differences between the two enzymes.

Heme Accumulation in the Periplasm of E. coli Cells Expressing H. ducreyi Cu,Zn-SOD-
The periplasmic extracts from E. coli 71/18 cells expressing H. ducreyi Cu,Zn-SOD were of a reddish color, unlike extracts from control cells. The electronic absorption spectrum of a periplasmic extract containing this enzyme in comparison with periplasmic extracts from control cells is shown in Fig. 1. It shows a pattern typical of hemoproteins with a main band in the Soret region around 410 nm and weaker bands close to 530 and 555 nm, indicative of the presence of relatively large amounts of heme. Initial expression experiments were carried out with E. coli cells bearing plasmid pIT4. As this plasmid contains a rather large H. ducreyi chromosomal insert containing genes other than sodC, we performed other expression experiments using the plasmid pPHduSOD (35). Periplasmic extracts from cells bearing plasmid pPHduSOD or pIT4 contained comparable amounts of heme (data not shown). On the contrary, the periplasmic extracts from cells bearing the control plasmids pEMBL18 or overexpressing the Cu,Zn-SODs from E. coli, Photobacterium leiognathi, Haemophilus influenzae, Haemophilus parainfluenzae, Xenopus laevis B, and Homo sapiens (some of which are reported in Fig. 1) contained only barely detectable amounts of heme, thus indicating that heme accumulation in the periplasmic space is specifically related to the expression of H. ducreyi Cu,Zn-SOD. This phenomenon is independent of the E. coli strain used to express the H. ducreyi enzyme, as nearly identical spectra were recorded in extracts from the E. coli strains 71/18, GC4468, and MC1061 (data not shown). The histidine-rich N-terminal region of H. ducreyi Cu,Zn-SOD (25) is not involved in the accumulation of heme, as the periplasmic fraction from cells expressing a mutant devoid of the first 22 amino acid residues (henceforth referred to as N-terminal deleted Cu,Zn-SOD) showed an absorption spectrum identical to that of the wild type enzyme (Fig. 1).
Periplasmic Heme Is Specifically Associated with Cu,Zn-SOD- Fig. 2 shows the absorption spectra of a sample of H. ducreyi Cu,Zn-SOD purified by affinity chromatography on a Ni-NTA column (spectrum 3) (35), the periplasmic extract (spectrum 1), and the flow-through eluate (spectrum 2). Nearly all the heme present in the periplasmic fraction is associated with the purified enzyme, which, on the basis of SDS-PAGE analysis, appears to be more than 95% pure (inset, Fig. 2). These findings suggested that heme could be bound to Cu,Zn-SOD. However, purified enzyme samples showed very high Cu,Zn-SOD/heme molar ratios (between 80:1 and 5:1, depending on the enzyme preparations), indicating that there was significantly less than one heme molecule/Cu,Zn-SOD dimer. So, given the very high extinction coefficient of the heme cofactor, SDS-PAGE analysis was not sufficient to exclude the presence of low amounts of contaminating hemoproteins. The affinity-purified enzyme was therefore further subjected to gel filtration chromatography. Fig. 3 shows the elution profile of the enzyme. It shows two main peaks at 38 and 76 kDa and another, less resolved, peak at higher molecular weight (ϳ110 kDa). All these peaks were characterized by SOD activity and by the same electronic spectra, typical of hemoproteins. SDS-PAGE analysis of the different fractions (Fig. 3, inset) revealed the presence of a main band of 19 kDa and a minor band of 38 kDa. Sequencing of the N-terminal region of the two bands revealed that they both corresponded to H. ducreyi Cu,Zn-SOD and did not reveal the presence of any significant amounts of contaminating proteins. Therefore, the 76-kDa protein is probably a tetrameric form of Cu,Zn-SOD, and the peak of higher molecular weight is a higher order aggregate of the same protein.
Attempts to obtain separation of a putative heme-containing contaminant protein from Cu,Zn-SOD by ion exchange chromatography (both on anionic and cationic resins) were unsuccessful, although we obtained a partial separation of a fraction of Cu,Zn-SOD without heme from the heme-containing enzyme.
In no case could we detect other contaminant hemoproteins in our Cu,Zn-SOD preparations. However, our attempts to obtain purified wild type H. ducreyi Cu,Zn-SOD without heme were severely hampered by progressive degradation of the N-terminal domain, due to its high protease sensitivity (35), during the different chromatographic steps.
Since the enzyme lacking the 22-residue N-terminal fragment was more stable, the heme-containing N-terminal deleted enzyme was purified and analyzed further. Purification of the N-terminal deleted enzyme was carried out as described under "Experimental Procedures." Heme co-eluted with Cu,Zn-SOD in all the chromatographic steps (ion exchange chromatography on anionic and cationic resins and gel filtration). Chromatography on a Mono S column at slightly acidic pH allowed the separation of heme-free Cu,Zn-SOD from the heme-containing enzyme. Heme-containing Cu,Zn-SOD samples purified by this procedure showed a Cu,Zn-SOD/heme molar ratio ranging from 1:0.7 to 1:0.95. These values suggest that one Cu,Zn-SOD dimer binds one heme molecule. SDS analysis and N-terminal sequencing of the purified heme-containing sample did not reveal the presence of contaminating proteins.
When cells were grown in standard LB medium, H. ducreyi Cu,Zn-SOD-associated heme was endogenously produced by E. coli. In fact, we observed that when E. coli 71/18 cells expressing the enzyme were grown in M9 minimal medium, the purified enzyme contained heme (data not shown). However, when the enzyme was purified from E. coli 71/18 cells grown in hemin-enriched media, the Cu,Zn-SOD/heme molar ratio was raised to values between 5:1 and 2:1. Again, no heme-containing contaminating proteins could be detected. Therefore, we conclude that the Cu,Zn-SOD of H. ducreyi is able to bind heme in a highly specific way. Further evidence for heme binding by H. ducreyi Cu,Zn-SOD was obtained by studies carried out with the E. coli strain H500, where hemA, encoding glutamyl-tRNA reductase, has been deleted and which is, therefore, unable to supply heme (32). H. ducreyi Cu,Zn-SOD purified from E. coli H500 cells did not contain significant amounts of heme. However, the addition of hemin to cultures of E. coli hemA cells resulted in efficient heme uptake by H. ducreyi Cu,Zn-SOD. This also occurred when hemin was added after the addition of chloramphenicol to block protein synthesis.
Studies on the Heme Nature and the Bond Type in H. ducreyi Cu,Zn-SOD-The pyridine hemochrome spectra give a clear indication of the nature and number of the electron-attracting groups on the heme chromophore. In the present case, the hemochrome of the enzyme purified from cells grown in hemin (heme b)-enriched media showed maxima close to 555, 525, and 417 nm (small variability was observed in different samples), which are almost superimposable on those expected for free heme. On the other hand the pyridine hemochrome spectrum of the H. ducreyi enzyme isolated from cells grown in LB medium showed peaks at about 551, 522 and 414 nm. These values, which are blue-shifted with respect to those of heme b, are close to those typical of several proteins with covalently bound heme, containing a c-type heme with one or both vinyl groups involved in thioether linkages (46).
To investigate the interaction between heme and Cu,Zn-SOD, the enzyme (both the wild type and the N-terminal deleted form) purified from standard LB (without hemin supplementation) was subjected to acidic acetone extraction. Most of the heme (between 70 and 95%, depending on the preparation) precipitated with the protein, suggesting that the interaction between heme and Cu,Zn-SOD is very tight. In a different preparative protocol, purification of the enzyme was carried out by metal affinity chromatography after overnight incubation of the periplasmic extract with 6 M guanidine HCl (a treatment that completely inactivates H. ducreyi Cu,Zn-SOD). The guanidine-denatured Cu,Zn-SOD showed an affinity for Ni-NTA similar to that of the native enzyme and eluted with buffers containing 80 mM imidazole. In this case too, heme co-eluted with the enzyme, and no contaminants could be detected by SDS-PAGE analysis of the purified protein. Moreover, the association between Cu,Zn-SOD and heme was not affected by boiling the enzyme in the presence of 2% SDS prior to loading on Ni-NTA columns or acidic acetone extraction. All these experiments indicate that there is a tight association between heme and Cu,Zn-SOD and suggest that heme is covalently bound to the enzyme. This hypothesis was further supported by staining of denaturing gels for heme-dependent peroxidase activity, a technique commonly used to discriminate c-type cytochromes from the other hemoproteins that lose heme after denaturation (43). As is shown in Fig. 4, the purified enzyme occurs in two distinct bands, both of which possess peroxidase activity. Therefore, we conclude that H. ducreyi Cu,Zn-SOD binds heme and that at least a fraction of this cofactor is covalently bound to the enzyme.
Covalent attachment of heme to the polypeptide chain is typically observed for cytochromes containing heme c. However, H. ducreyi Cu,Zn-SOD produced in an E. coli strain that, lacking the whole ccm operon, is unable to synthesize mature c-type cytochromes (31, 49) still contained covalently bound heme (Fig. 4, lane 7), suggesting that the enzyme does not bind heme c.
Most (60 -90%) of the heme associated with Cu,Zn-SOD purified from hemin-enriched medium could be separated from the polypeptide by acidic acetone extraction, and peroxidase activity staining showed an abundant band corresponding to free heme. The enzyme isolated from hemA E. coli H500 cells grown in standard LB medium did not possess peroxidase activity (Fig. 4, lane 4), since they do not synthesize the heme cofactor. However, when extracts from hemA cells grown in LB supplemented with hemin were analyzed for peroxidase activity on SDS-PAGE, Cu,Zn-SOD showed high heme-dependent peroxidase activity (Fig. 4, lane 5), although a large fraction of the heme molecules bound to Cu,Zn-SOD detached from the enzyme. This observation suggests that H. ducreyi Cu,Zn-SOD binds heme b and that this cofactor is subsequently covalently attached to the enzyme. This hypothesis was also supported by results obtained by an in vitro hemin binding assay. In fact, we have observed that after a 15-h dialysis against a buffer containing hemin, the wild type enzyme isolated from H500 E. coli cells exhibits an electronic absorption spectrum (not shown) identical to that typical of the enzyme isolated from cells grown in hemin-enriched media. Interestingly, both acidic acetone extractions and staining of SDS-PAGE gels for heme-dependent peroxidase activity demonstrated that a fraction (close to the 10%) of the heme bound to the enzyme was covalently bound to the polypeptide. Fig. 5 shows the electronic absorption spectra of N-terminal deleted H. ducreyi Cu,Zn-SOD samples isolated from E. coli 71/18 cells grown in hemin-enriched media (N-del 1) (Fig. 5A) and from E. coli cells grown in standard LB medium (containing more than 80% covalent heme) (N-del 2) (Fig. 5B) in both the oxidized (bottom) and the reduced (top) forms. The spectra of N-del 1 are 1-3 nm red-shifted with respect to those of N-del 2 and are very similar to those of cytochrome b 5 (50). The spectra are characteristic of 6-coordinated low spin (6-c LS) hemes. Moreover, the shape and the wavelength maxima of the ␣and ␤-bands of the ferric forms indicate the presence of histidine residues as fifth and sixth ligands of the heme iron. In fact, it has been suggested that maxima at about 530 and 560 nm together with their relative intensity are characteristic of a bis-imidazole complex (51).   (52). In agreement with the electronic absorption spectra, both proteins have core size marker bands typical of 6-c LS hemes ( 4 , 1374 cm Ϫ1 ; 3 , 1504 cm Ϫ1 ; 11 , 1564 -1565 cm Ϫ1 ; 2 , 1578 -1583 cm Ϫ1 ; 19 , 1584 -1586 cm Ϫ1 ; and 10 , 1638 -1639 cm Ϫ1 ). However, they markedly differ in the region 1600 -1630 cm Ϫ1 . N-del 1 exhibits two polarized bands, present only in the Soret excitation, at 1619 and 1628 cm Ϫ1 , assigned to the (CϭC) stretching modes of the vinyl groups. N-del 2 shows only one polarized band at 1619 cm Ϫ1 . This might be due to only one vinyl stretch or the two vinyl substituents that give rise to a coincident band at 1619 cm Ϫ1 , as in the case of myoglobin (53). In general, a down-shift of the vinyl band frequency from 1628 -1619 cm Ϫ1 indicates a strong conjugation of the vinyl double bond with the heme. As a consequence, a red-shift of the electronic absorption spectra is expected. On the contrary, compared with N-del 1, N-del 2 is characterized by an overall blue-shift of the electronic absorption spectra of both the ferric and ferrous forms. Moreover, the 2 mode of N-del 2 up-shifts by 5 cm Ϫ1 with respect to the frequency observed for N-del 1. It is known that the 2 mode is coupled to the vinyl (CϭC) mode in protohemes, and saturation of the vinyl groups raises the 2 frequencies by about 10 cm Ϫ1 (52). The RR spectra of the ferrous forms show a similar behavior. N-del 1 is characterized by two (CϭC) stretches, and the 2 mode at 1583 cm Ϫ1 , N-del 2 displays one (CϭC) stretch and the 2 mode at 1586 cm Ϫ1 (data not shown). Therefore, the combined analysis of the electronic and RR spectra indicates that the two proteins have different heme types as follows: a b-type (with two vinyl groups) in N-del 1, and a heme with only one vinyl group in N-del 2.
Histidine 64 Is Required for Heme Binding-The spectroscopic data suggest as putative ligands for the heme iron two histidine residues. H. ducreyi Cu,Zn-SOD is characterized by an unusually high content of histidine residues (35). In order to identify the possible histidine ligands, we compared the amino acids sequence of the H. ducreyi enzyme with those of H. parainfluenzae and Actinobacillus pleuropneumoniae, two enzymes that are not able to bind heme (Fig. 7). Apart from the N-terminal histidines, which are not required for heme binding, H. ducreyi Cu,Zn-SOD has two histidine residues (His-60 and His-64) that are not present in the H. parainfluenzae enzyme. These two residues were, therefore, considered as potential candidates for site-directed mutagenesis studies. It should be noted that the alignment of all known bacterial Cu,Zn-SODs indicates that a histidine in position 60 is present in the enzyme from A. pleuropneumoniae, whereas histidine 64 is unique to H. ducreyi Cu,Zn-SOD (26,35,54).
Two single His mutants of H. ducreyi Cu,Zn-SOD, in which His-60 and His-64 were replaced by glutamine or aspartic acid, respectively, and the double mutant, were produced and expressed in E. coli 71/18 cells grown in standard LB medium. The electronic spectra of the periplasmic extracts from cells expressing the His-60 3 Gln mutant enzyme were identical to those of cells expressing the wild type enzyme, thus indicating that His-60 is not required for heme binding. In contrast, extracts from cells expressing the mutants His-64 3 Glu and His-60 3 Gln, His-64 3 Glu did not contain appreciable amounts of heme. Wild type and mutant enzymes were purified by immobilized metal affinity chromatography on Ni-NTA columns, and their ability to bind heme was assessed by hemestaining in SDS gels. As shown in Fig. 8, mutants lacking His-64 did not show heme-dependent peroxidase activity, indicating that this residue is required for heme binding. Similar results were obtained when the mutant enzymes were purified from cells grown in LB supplemented with hemin. Moreover, unlike the wild type enzyme, upon dialysis against a buffer containing a 1000-fold molar excess of hemin the His-64 3 Glu mutant enzyme contained only a barely detectable amount of heme.
Recently, the x-ray structure of A. pleuropneumoniae Cu,Zn-SOD has been determined at 2.1-Å resolution (26). This enzyme shows a very high degree of homology with the H. ducreyi enzyme (88.4% identity excluding the N-terminal extension). Moreover, 18 of the 19 residues involved in dimer formation in A. pleuropneumoniae Cu,Zn-SOD (26, 27, 54) are conserved in the H. ducreyi enzyme, suggesting a nearly identical subunit assembly in the two enzymes (the only mutation being the substitution of an Asn residue with Ile at position 130 of H. ducreyi Cu,Zn-SOD). In view of these considerations, we used the A. pleuropneumoniae enzyme as a template to model the three-dimensional structure of H. ducreyi Cu,Zn-SOD employing the Swiss PDB-Viewer facility. Fig. 9 shows the hypothetical dimeric structure of H. ducreyi Cu,Zn-SOD and highlights the position of His-60 (green), His-64 (red), and the disulfide bond (blue). His-60 is located on the molecule surface and is solvent-exposed. In contrast, His-64 is located just below the surface of association between the two subunits and lies on the 2-fold symmetry axis that crosses the dimer interface, in a region that appears to be only partially accessible to the solvent. The side chains of His-64 in the two subunits point toward one another at a distance that, in the model generated by the Swiss PDB-Viewer, is close to 3 Å. Although such a distance is too short to accommodate a heme cofactor having His-64 as axial ligands, it must be considered that this distance comes from a model in which the orientation of the main chain was left identical to that of the template. Only small conforma-tional changes are required to increase the distance between the two axial ligands to about 4 Å, in order to easily accommodate the heme group. The only two cysteine residues present in each subunit and which form a disulfide bond in A. pleuropneumoniae Cu,Zn-SOD, as well as in all the other characterized Cu,Zn-SOD, are quite far from His-64. Therefore, it is very unlikely that these cysteine residues are involved in a covalent linkage through two thioether bonds if the heme is coordinated to His-64.
We have also inspected the model in search of other residues possibly involved in heme binding. The side chains closest to His-64 in each enzyme subunit are those from Phe-121 and Val-122, which are both involved in dimer formation, and from His-124 and Met-123. Although His-126 is present also in the Cu,Zn-SOD from all the other Haemophilus species, Met-123 is not present in the Cu,Zn-SODs from A. pleuropneumoniae and H. parainfluenzae nor in any other bacterial Cu,Zn-SOD (in Fig. 9 Met-123 is shown in yellow). To evaluate if mutation of this residue might affect heme binding, we have engineered a mutant enzyme carrying the Met-123 3 Leu substitution. This mutant is still able to bind heme and is still a 6-c LS heme, but the affinity for the cofactor is more than 10-fold decreased with respect to the wild type enzyme (data not shown).
In order to evaluate whether a His residue at the dimer interface is sufficient to allow heme binding by other Cu,Zn-SODs, the enzyme from H. parainfluenzae was engineered to introduce a histidine in position 64 (Gln-64 3 His mutant). Wild type and mutant enzymes expressed in E. coli 71/18 cells were purified by metal affinity chromatography and analyzed FIG. 7. Alignment of amino acid sequences of H. ducreyi, A. pleuropneumoniae, and H. parainfluenzae Cu,Zn-SODs. The amino acid numbering of H. ducreyi Cu,Zn-SOD, shown on the top line, is used as reference. All histidine residues are highlighted in red, and black arrows mark copper and zinc ligands, indicated by C and Z, respectively. Boxes and red arrows mark the amino acids that have been mutated in this work. respectively. This enzyme view emphasizes the highly symmetrical structure of bacterial Cu,Zn-SOD. It should be noted, however, that while the two His-64 residues are at a distance of about 3 Å, the two Met-123 residues are more than 10 Å from each other. for their heme content by peroxidase staining. Both the electronic spectra (data not shown) and heme staining of denaturing polyacrylamide gels (Fig. 8) failed to identify heme in the purified proteins, even in extracts from cells grown in media enriched with hemin. DISCUSSION X-ray structures of four different bacterial Cu,Zn-SODs have shown that prokaryotic and eukaryotic Cu,Zn-SODs share a similar monomer fold, based on a flattened Greek-key eightstranded ␤-barrel and a similar organization of the metal cluster (26, 27, 54 -56). Bacterial Cu,Zn-SODs are, however, characterized by a different architecture of the active site channel and a distinctive quaternary assembly (54). Whereas all the eukaryotic enzymes are characterized by a very high degree of structural conservation (57), among prokaryotic Cu,Zn-SODs significant structural variations have been observed or may be inferred from their amino acid sequences. Most of these differences concern the quaternary structure. Although most prokaryotic Cu,Zn-SODs have been shown to be dimeric, some are monomeric (36,54,56,58,60), and the subunit:subunit recognition modality may change in some dimeric variants (27,54). Other peculiar features that may be observed in specific bacterial Cu,Zn-SOD include the substitution of some of the metal ligands (25,54,59), insertion or deletion in the major loops connecting the ␤-strands (54), and the presence of metal-binding extensions at the N terminus (35). The finding that the Cu,Zn-SOD from H. ducreyi is able to bind heme further emphasizes the structural and functional variability of this class of enzymes. The spectral features of this protein suggest that histidine ligands could be involved in the heme iron coordination, as also indicated by site-directed mutagenesis experiments showing that the His-64 residue is required for heme binding. These observations have been rationalized by inspection of a structural model of the enzyme, which has been generated starting from the structure of the highly homologous enzyme from A. pleuropneumoniae. We are aware that molecular details suggested by modeling could in part differ from those found in the real enzyme and that heme binding could induce structural rearrangements that we have not taken into account. However, such a model provides a clear explanation for the pivotal role of His-64 in heme binding. This residue has been located at the center of symmetry at the dimer interface, so that the side chains of the His-64 residues on the two subunits point toward each other, in such an arrangement as to be suitable for direct coordination of iron. Recent studies have suggested a role for the dimeric structure in modulating enzyme stability and activity both in eukaryotic and prokaryotic Cu,Zn-SODs (27,(61)(62)(63). In this case the dimer interface builds up an additional cofactor-binding site. However, the observation that a mutant H. parainfluenzae Cu,Zn-SOD containing a histidine residue in the position corresponding to that of H. ducreyi His-64 is unable to bind heme indicates that other residues must be involved in the formation of a heme-binding pocket. In agreement with this hypothesis we have observed that the affinity for heme of H. ducreyi Cu,Zn-SOD is dramatically decreased upon mutation of Met-123, a residue close to His-64.
Comparison of the sequences of all known bacterial Cu,Zn-SODs shows that His-64 is present only in the enzyme from H. ducreyi, and none of the other Cu,Zn-SODs (prokaryotic and eukaryotic) that we have tested show any affinity for heme. This finding raises the question of whether Cu,Zn-SOD hemebinding ability is related to one of the most striking features of H. ducreyi, namely its absolute requirement to obtain heme from the host. In principle, heme could serve as a cofactor for additional catalytic activities involved in protecting the orga-nism from reactive nitrogen or oxygen intermediates other than superoxide, which are produced during the phagocytic oxidative burst. An alternative possibility could relate to the potential toxicity of "free" heme for the bacterial cell. How H. ducreyi obtains sufficient amounts of this essential cofactor in vivo is not known, but it produces toxins that can induce hemolysis, and it is possible that at some stage of infection the bacterium could be exposed to a potentially dangerous heme overload. Under such conditions the binding of heme to Cu,Zn-SOD could protect the bacterial cell from damage due to oxyradicals formed upon the reaction of heme iron with oxygen. In this perspective, Cu,Zn-SOD may act as a periplasmic protein able to store heme when its availability exceeds the needs of the cell. However, as heme is bound stoichiometrically to Cu,Zn-SOD, this hypothesis raises the question of how this might serve the bacterial cell under conditions (should they ever occur) of saturating heme concentrations. Alternatively, it is possible to hypothesize that Cu,Zn-SOD might participate, via the transient binding of heme, in the mechanisms of heme transport from membrane-localized receptors to other cellular targets. This hypothesis is obviously in contrast with the possibility that heme should be covalently attached to Cu,Zn-SOD when this enzyme is expressed in H. ducreyi. Further work is required to explain why H. ducreyi Cu,Zn-SOD is able to bind heme and to understand whether covalent binding of heme to Cu,Zn-SOD reflects a functionally important property of the enzyme, an artifact of the E. coli expression system, or some other not yet defined possibility.
An interesting question raised by the present results is the nature of the heme cofactor bound to the H. ducreyi Cu,Zn-SOD when it is produced in E. coli. The enzyme is able to bind heme b, as demonstrated by the UV-visible, RR, and pyridine hemochrome spectra of N-del 1. Although heme b is expected to be the most important heme source for H. ducreyi, we observed that a large part of heme bound by the recombinant enzyme expressed in E. coli is covalently attached to the protein, as the case of N-del 2 protein. Covalent attachment of heme is frequently observed in c-type cytochromes, where two thioether linkages are formed between the vinyl groups of the heme and two cysteine residues, usually within a CXXC motif (64). H. ducreyi Cu,Zn-SOD does not contain such a motif, since it contains only two cysteine residues that are involved in the formation of a disulfide bond, essential for enzyme stability and activity (38). The structural model of the enzyme indicates that these two cysteine residues are very distant from the putative iron ligand His-64. As a consequence, this residue could coordinate to the heme iron linked to the apoprotein via thioether bonds only after an extensive loss of quaternary and, perhaps, tertiary structures. This is not compatible with our findings that the enzyme containing covalently bound heme exhibits catalytic activity and spectroscopic properties similar to the enzyme containing mainly non-covalent heme.
Further evidence that heme c does not bind to the enzyme has come from expression studies in an E. coli strain lacking the whole ccm operon. A ccm strain is unable to synthesize mature c-type cytochromes (31,49), but recombinant H. ducreyi Cu,Zn-SOD expressed in this strain does contain high amounts of covalently bound heme. The spectroscopic data of the N-del 2 sample together with its pyridine ferrohemochrome suggest that the cofactor is a heme containing one vinyl group. In particular, the RR spectra clearly show only one (CϭC) stretch, and the bands in the UV-visible spectrum of the ferrous forms are 4 -8 nm red-shifted with respect to ferrous cytochrome c (65), which does not contain any vinyl groups, and about 2-3 nm blue-shift with respect to N-del 1 (which contains two vinyl substituents). Therefore, although the present inves-tigation does not provide information about the protein residues involved in covalent binding of heme, we suggest that heme is covalently bound to the protein via the saturation of the double bonds of one of the two vinyl substituents. However, it is difficult, at the present stage, to give a conclusive definition of the nature of the heme. It is very unlikely that H. ducreyi Cu,Zn-SOD binds heme c, but the growing body of evidence indicating that covalent attachment of chemically modified heme b, such as that found in lactoperoxidase (66), eosinophil peroxidase (67), thyroperoxidase (68), and myeloperoxidase (69), is not uncommon leads us to suggest that the covalently bound heme cofactor associated with recombinant H. ducreyi Cu,Zn-SOD could be a derivative of heme b.