The interaction of covalently bound heme with the cytochrome c maturation protein CcmE.

The heme chaperone CcmE is a novel protein that binds heme covalently via a histidine residue as part of its essential function in the process of cytochrome c biogenesis in many bacteria as well as plant mitochondria. In the continued absence of a structure of the holoform of CcmE, identification of the heme ligands is an important step in understanding the molecular function of this protein and the role of covalent heme binding to CcmE during the maturation of c-type cytochromes. In this work, we present spectroscopic data that provide insight into the ligation of the heme iron in the soluble domain of CcmE from Escherichia coli. Resonance Raman spectra demonstrated that one of the heme axial ligands is a histidine residue and that the other is likely to be Tyr134. In addition, the properties of the heme resonances of the holo-protein as compared with those of a form of CcmE with non-covalently bound heme provide evidence for the modification of one of the heme vinyl side chains by the protein, most likely the 2-vinyl group.

The biogenesis of c-type cytochromes is a complex post-translational modification process that occurs in different ways in different organisms (1). In many Gram-negative bacteria, including Escherichia coli (2), and in plant mitochondria (3), a complex multiprotein system, known as the cytochrome c maturation (Ccm) 1 system, exists to perform this function. The maturation process involves the transport of the apocytochrome into the periplasm by the Sec pathway (4), the transmembrane transport of heme to the periplasm from its site of synthesis in the cytoplasm, and the stereochemically controlled step of covalent heme attachment to the apocytochrome. Thioether bonds are formed between the cysteines of a CXXCH motif in the protein and the vinyl groups of heme. A disulfide isomerization system is also present to ensure that the cysteine side chains in the apocytochrome are reduced prior to heme attachment (5).
A central protein in the maturation process is the heme chaperone CcmE, which has been shown to bind heme covalently via a histidine residue before transferring the heme to the apocytochrome (6). The structures of two apo-forms of CcmE have been solved (7,8), revealing a two-domain structure with a ␤-barrel core and a flexible C-terminal domain. The heme-binding histidine residue lies at the junction of these two domains, and its position suggests that heme binds to the surface of the protein. This is an unusual feature when compared with classical heme-binding proteins, most of which bind heme in a solvent-shielded pocket within the protein. Heme binding on the surface of CcmE is, however, consistent with its dynamic function as a heme chaperone in that it is required to bind heme transiently and to release it to apocytochromes.
CcmE is a member of the oligo-binding protein family (9) in terms of its structure. These oligo-binding proteins have a common function in binding and releasing non-protein molecules, which is consistent with the known functions of CcmE. Covalent heme attachment to a histidine side chain of a protein has been described for an algal Hb variant that contains a bis-His coordinated heme covalently bound to the protein through the N⑀2 of a histidine and the 2-vinyl group of the heme (10).
Efforts to solve the structure of the heme-bound form of CcmE have not yet succeeded. Complications due to insolubility of the reduced holo-protein and paramagnetism of the ferric form may have hampered NMR experiments. The apparent conformational flexibility of the protein may explain why crystallization of CcmE has not been reported. Therefore, the only information at present on the nature of the heme ligation in holo-CcmE is limited to UV-visible spectra (6,11,12), which suggest that the heme adopts a high spin state in the ferric form and a 6-coordinate low spin state in the ferrous form. The difference in the heme ligation between the two redox states suggests a working hypothesis for the function of the protein, particularly regarding the control of heme binding and release according to the redox state of the heme iron.
In this work, we present the first detailed spectroscopic analysis of CcmE, which provides insight into the structure of the protein at the heme binding site. Studies of a variant of the protein with Tyr 134 mutated to Phe suggest that the tyrosine residue is involved in heme ligation. We also present evidence for the presence of the heme-histidine covalent bond comparing the resonance Raman (RR) spectra of holo-CcmE with those of a non-covalent complex of CcmE (expressed in the apo-form in the E. coli cytoplasm) and Fe-protoporphyrin IX (12).

EXPERIMENTAL PROCEDURES
Plasmid Construction-A pET22b-based (Novagen) CcmE expression vector used previously (11) was modified to include a thrombin cleavage site between the protein and the C-terminal His tag. This was performed using the ExSite mutagenesis system (Stratagene) with the primers 5Ј-CGCGGCAGCGGCCACCACCACCACCACCACTGAGATC-CGG-3Ј (forward) and 5Ј-CGGCACCAGGCCTGATGCTGGGTCCTTA-TAAACACTCGCC-3Ј (reverse). The insertion included an additional Gly on either end of the protease recognition site and produced the plasmid pE225. The Y134F variant of CcmE was made using the Quik-Change mutagenesis method (Stratagene) with the primers 5Ј-CACG-ATGAAAACTTTACGCCGCCAGAAG-3Ј (forward) and 5Ј-CTTCTGGC-GGCGTAAAGTTTTCATCGTG-3Ј (reverse) and the plasmid pE225 (described above) as the template to produce the plasmid pE226. The plasmids were sequenced to confirm that the constructs were correct.
Protein Expression and Purification-The E. coli strain JM109 (DE3) was used for protein expression. Holo-CcmE proteins were produced by co-expressing the plasmids described above with the plasmid pEC86, which carries the entire ccm operon (13), which was kindly provided by Prof L. Thöny-Meyer (Zü rich, Switzerland). The expression of the other Ccm proteins is necessary for the expression of holo-CcmE. Cultures were first grown to mid-log phase at 37°C following inoculation and then induced with 1 mM isopropyl-thio-␤-D-galactopyranoside and grown overnight at 28°C. Proteins were purified as described previously (11). Thrombin digestion was performed using a CleanCleave kit (Sigma) according to the manufacturer's instructions, and the uncleaved protein was separated from the cleaved protein by reapplying the protein mixture to a Ni 2ϩ -chelating Sepharose column. The presence of the covalently bound heme was found to inhibit the digestion step with thrombin as the process was much slower than that observed for the apo-form of the protein. ApoCcmE was produced as described previously (12), and the non-covalent complex with heme was produced by the addition of a 5-fold excess of protein to ϳ50 M hemin (Sigma; from a 1 mM stock in dimethylsulphoxide). Protein concentrations were determined using the Bradford method (Bio-Rad), and the protein was stored in 50 mM Tris-HCl, pH 7.4, 300 mM NaCl throughout. The protein referred to in this work as CcmE refers to the soluble form of CcmE that lacks its membrane anchor, as described (11).
Protein Characterization-Discontinuous SDS-PAGE (10% acrylamide) (NuPAGE, Invitrogen) was used to analyze the proteins, and staining for covalently bound heme was performed according to the method of Goodhew et al. (14). Visible absorption spectra were recorded on a PerkinElmer Lambda 2 spectrophotometer using 2-5 M protein solutions in 50 mM Tris-HCl buffer, pH 7.4, 300 mM NaCl. Protein samples were reduced by the addition of a small amount of solid sodium dithionite. Pyridine hemochrome spectra were obtained according to the method of Bartsch (15) using 5 M protein in 19% (v/v) pyridine and 0.15 M NaOH. Electrospray ionization mass spectrometry was performed using a Micromass Bio-Q II-ZS triple quadrupole atmospheric pressure mass spectrometer into which 10-l protein samples in 1:1 water: acetonitrile, 1% formic acid at a concentration of 20 pmol/l were injected at a flow rate of 10 l/min.
Resonance Raman Spectroscopy-Visible RR spectra were obtained with a single polychromator (Jobin Yvon, SPEX750M) equipped with a liquid N 2 -cooled CCD detector (Roper Scientific, Spec10:400B/LN). The excitation wavelengths used were 413.1 and 568.2 nm from a krypton ion laser (Spectra Physics, BeamLok 2060) and 363.8 nm from an argon ion laser (Spectra Physics, BeamLok 2080). The laser power at the sample point was adjusted to ϳ5 milliwatts for the reduced and airoxidized samples and to 0.1 milliwatts for the CO-bound form to prevent photodissociation of the CO. Raman shifts were calibrated with indene, CCl 4 , acetone, and an aqueous solution of ferrocyanide, and the accuracy of the peak positions of the well defined Raman bands was Ϯ 1 cm Ϫ1 . The UV RR spectra were obtained as described previously (16). UV light at 244 nm was produced from an intracavity frequency doubled argon ion laser (Coherent, Inova300C FreD). The laser power at the sample point was about 0.2 milliwatts. The scattered light was dispersed with a single monochromator (Jobin Yvon, SPEX1269M) equipped with an intensified CCD camera (Roper Scientific, PI-MAX 1024Cs-Te). All measurements were performed at room temperature with a spinning cell. The protein concentrations for visible and UV RR measurements were about 20 and 50 M, respectively, in 50 mM Tris-HCl, pH 7.4, 300 mM NaCl.

RESULTS
The wild-type and variant proteins were shown to be pure and lacking the affinity tags by SDS-PAGE analysis (data not shown) and of expected masses (for apo-protein) determined with electrospray ionization mass spectrometry (wild type, observed 15,001 Da, expected 15,002 Da; Y134F, observed 14,985 Da, expected 14,986 Da). Purified CcmE produced under the conditions described contained a substantial amount of apo-protein; usually about 20% of the protein contained covalently attached heme as determined by comparing the ratio of the absorbance value at 280 nm (apoprotein ϩ holo-protein) with the intensity of the Soret band (holo-protein). The co-production of large amounts of apoprotein with holo-protein has been observed previously with CcmE (6,11). Only the apo-forms of the proteins produced were observed in the mass spectra; the heme-containing form does not appear in the spectra under these conditions. The covalent attachment of heme to these proteins was confirmed by heme-staining SDS-PAGE gels (not shown).
UV-visible Spectroscopic Analysis- Fig. 1 shows spectra of the wild-type (A) and Y134F variant (B) of holo-CcmE in both the ferric and the ferrous forms. It is clear that the spectra of the Y134F variant are significantly different from those of the wild type. The ferrous form of the wild type shows a characteristic spectrum of a c-type cytochrome, whereas that of the Y134F protein does not have the characteristic ␣and ␤-bands. This indicates that a significant change has occurred in the heme ligation in the absence of Tyr 134 . The origin of the peak observed at about 650 nm for the Y134F protein has not been elucidated. The ␣-band in the pyridine hemochrome spectra of both the wild-type and the Y134F proteins was observed at 551 nm (data not shown). This wavelength would normally suggest that both of the vinyl groups of the heme were modified (as in a c-type cytochrome), although the uncertainty of this observation was noted previously for the heme-attached form of CcmE (6,11).
Resonance Raman Spectra of Wild-type Holo-CcmE- Fig. 2 shows RR spectra of holo-CcmE with excitation at 413.1 nm. The spectrum of the ferric protein shows a distinct 3 band at 1492 cm Ϫ1 , which is characteristic of a five-coordinate (5c) high spin (HS) heme (spectrum a). When the holo-CcmE protein is reduced, however, the spectrum shows different spin state marker bands ( 3 ϭ 1493 cm Ϫ1 and 2 ϭ 1588 cm Ϫ1 , spectrum b), which suggest that the heme iron is in a 6c low spin (LS) state. These observations are consistent with the UV-visible data shown above. The frequencies for the ferrous form are close to those of the CO-bound form (1497 and 1593 cm Ϫ1 , respectively) shown at the top of the figure.
To confirm the coordination of the CO-form, we observed the Fe-CO and C-O stretching modes for CO-bound holo-CcmE. Fig. 3 shows RR spectra in the low (left panel) and high frequency (right panel) regions. Two CO-isotope sensitive bands were observed at 496 cm Ϫ1 (487 cm Ϫ1 with 13 C 18 O) and 1966 cm Ϫ1 (1871 cm Ϫ1 with 13 C 18 O). Since these frequencies and isotope shifts are consistent with those observed for other heme proteins, we assigned the two bands at 496 and 1966 cm Ϫ1 to the Fe-CO and C-O stretching modes, respectively. The weak isotope-sensitive band observed at 577/557 cm Ϫ1 for the 12  Q-band Excited Resonance Raman Spectra-Q-band excitation selectively enhances the Raman intensity of the B g and A 2g modes of heme (20), and three modes ( 10 , 11 , and 19 ) are expected in the 1500 -1700 cm Ϫ1 region. The polarized RR spectra of ferrous holo-CcmE excited at 568.2 nm are shown in Fig. 5. Since 19 is the sole anomalously polarized band in this region, we can easily distinguish it by polarization analysis. The band at 1587 cm Ϫ1 is anomalously polarized and assigned to 19 , whereas the bands at 1536 and 1629 cm Ϫ1 are depolarized and assigned to 11 and 10 , respectively. The 10 frequency is significantly upshifted as compared with that of reduced cytochrome c 3 , which has a bis-His coordinated heme (1619 cm Ϫ1 ) (21). On the other hand, the 11 frequency of holo-CcmE is significantly lower. The 11 band is known to serve as a -electron marker and is sensitive to the donor strength of the axial ligand, especially in the ferrous state (22)(23)(24). The lower frequency of 11 for holo-CcmE as compared with that of cytochrome c 3 (1540 cm Ϫ1 ) suggests that an anionic ligand such as imidazolate or tyrosinate is coordinated to the heme in the ferrous form of holo-CcmE.
Near UV-excited Resonance Raman Spectra- Fig. 6 shows the low frequency region of the UV RR spectra of holo-CcmE (spectrum a) and the Y134F variant (spectrum b) excited at 363.8 nm. It has been reported that the Fe-Tyr stretching mode is observed at ϳ590 cm Ϫ1 upon 488-nm excitation for Hb and myoglobin (Mb) mutants in which the proximal His is replaced with Tyr (25)(26)(27). When 488-and 514-nm excitation wavelengths were applied to holo-CcmE, such a vibration was not observed (data not shown). However, on excitation at 363.8 nm, a band was detected at 600 cm Ϫ1 for wild-type holo-CcmE (Fig. 6, spectrum a) that was absent from the spectrum of the Y134F variant (spectrum b). Although it is known that 363.8-nm excitation gives an Fe-Cys stretching mode around 350 cm Ϫ1 for Cys-bound ferric 5cHS hemes (28,29), the observed frequency (600 cm Ϫ1 ) is too high to assign it to the Fe-S stretching mode, and the soluble domain of CcmE studied in this work contains no Cys residues. Considering that the 600 cm Ϫ1 band was not observed for the holo-Y134F variant and is in good agreement with those observed for proximal Tyr variants of Hb and Mb (578 -603 cm Ϫ1 ), it is likely that this band arises from the Fe-Tyr stretching mode.
Resonance Raman Spectra of the Y134F Variant of Holo-CcmE-RR spectra of the Y134F variant excited at 413.1 nm are shown in Fig. 7. The quality of the spectra is hampered by strong fluorescence. The spin state marker band, 3 , was observed faintly at ϳ1468 cm Ϫ1 for the reduced protein (Fig. 7,  spectrum b), although it was clearly observed at 1493 cm Ϫ1 for the wild-type protein (Fig. 2). The low frequency shift of 3 in the reduced Y134F protein is due to the presence of 5cHS heme, which suggests that the replacement of Tyr 134 with Phe induces a change in heme coordination from 6c to 5c. 2 This result suggests that Tyr 134 could be an axial ligand of the heme in the wild type. The RR spectra of the ferric (spectrum a) and CObound (spectrum c) forms of the Y134F variant are also distinctly different from those of wild-type protein shown in Fig. 2, meaning that the heme coordination is significantly perturbed by this mutation.
UV Resonance Raman Spectra-To confirm the coordination of Tyr 134 to the heme in holo-CcmE, we measured the 244 nm-excited UV RR spectra. It is known that free deprotonated Tyr has an absorption maximum at ϳ250 nm, and accordingly, the excitation wavelength of 244 nm is suitable for the obser- vation of tyrosinate coordination to a heme (30). In fact, in the UV RR spectra, the Raman bands of tyrosine are observed at 1619 (Y8a, ring C-C stretch), 1208 (Y7a, C ␤ C ␣ stretch), and 1177 (Y9a, ring C-H bend) cm Ϫ1 , whereas its deprotonation causes frequency shifts of the bands from 1619 and 1177 to 1603 and 1173 cm Ϫ1 , respectively (31,32). Fig. 8 shows the 244-nm excited UV RR spectra of holo wild-type (spectrum a) and Y134F (spectrum b) proteins. Three intense bands derived from Tyr were observed at 1177, 1207, and 1615 cm Ϫ1 with a shoulder at 1603 cm Ϫ1 for wild-type holo-CcmE. There are 6 Tyr residues in wild-type holo-CcmE, and all of them contribute to the three bands. In the Y134F variant, only the contribution from Tyr 134 would be absent if other regions of spectra remained unchanged. In the observed spectrum of Y134F, the intensity of the small shoulder at 1603 cm Ϫ1 was decreased. This is more clearly seen in the difference spectrum, wild type minus Y134F, depicted by spectrum c. This suggests that tyrosinate coordinates to the heme in holo-CcmE but that it is lost in the Y134F variant.
Heme Peripheral Modes in Wild-type Holo-CcmE-In the holo-protein, the presence of a covalent bond between a heme peripheral group and protein has been proposed (6,11). We anticipated that the RR properties of the heme in holo-CcmE might provide evidence for the nature of this covalent modification. If covalent modification has not taken place, heme vinyl bending and stretching modes would be expected to appear in the low and high frequency regions, respectively (33). Fig. 9 (left panel) illustrates the low frequency RR spectra of reduced holo-CcmE (spectrum a) and of a noncovalent complex of CcmE and heme (spectrum b). The vinyl bending mode, ␦(C ␤ C a C b ), appeared at 411 cm Ϫ1 for holo-CcmE. Although 2-and 4-vinyl modes are observed separately at 440 and 409 cm Ϫ1 , respectively, for deoxy Mb and at similar frequencies for other heme proteins (33), only one vinyl mode was observed for holo-CcmE. Since the RR spec- trum of the non-covalent complex of CcmE and heme also contains only one vinyl mode at 419 cm Ϫ1 as shown by spectrum b in Fig. 9, it is not clear which vinyl was modified. On the other hand, the vinyl CϭC stretching mode, CC , can be observed in the high frequency region. Fig. 9 (right panel) shows the high frequency region of RR spectra of reduced holo-CcmE (spectrum c) and the non-covalent CcmE-heme complex (spectrum d). The polarized bands of the latter at 1609 and 1621 cm Ϫ1 presumably arise from the CC modes, and the 1621 cm Ϫ1 band is missing in the spectrum of holo-CcmE. These data suggest that one vinyl group is converted to a covalent linkage in holo-CcmE. DISCUSSION Previous studies of CcmE have yielded some insight into the way in which heme is bound in this protein (6,11). It was clear that the heme was in a high spin form in the ferric state (the form in which CcmE was purified when overexpressed in the E. coli periplasm) but had a 6cLS structure upon reduction. Two structures of apo-forms of the CcmE proteins from Shewanella putrefaciens and E. coli (7,8) did not show typical heme binding sites. However, it was possible to model a heme molecule on the surface of the E. coli protein (7). The residue known to bind heme covalently in CcmE, His 130 , was exposed to the solvent, and its environment did not provide obvious clues to the identity of the heme ligands. It has been suggested, however, that Tyr 134 could be a heme ligand in CcmE (11). Its position in the protein, which is 4 residues away from the heme-binding histidine, puts it in the same position (relative to the first Cys) as the His in c-type cytochromes in the CXXCH motif that is always an axial heme ligand; in addition, its mutation to Ala affected the heme binding ability of CcmE in vivo (34). In this study, we employed RR spectroscopy to provide insight into the nature of the heme ligation in CcmE.
Coordination Structure of Heme in holo-CcmE-We applied several Raman excitation wavelengths to examine the heme ligation in wild-type holo-CcmE. In general, the RR spectra of reduced 5cHS heme with a histidine as a proximal ligand contain the iron-histidine stretching mode, (Fe-His), in the 200 -250 cm Ϫ1 region when 441.6-nm excitation is used (35). However, the excitation of holo-CcmE at 441.6, 413.1, and 428.7 nm provided no band in this frequency region due to the presence of predominant 6cLS species (data not shown). Even when a 5-ns pulse at 435.7 nm was applied to the CO-bound species to observe a transiently formed photodissociated species, rebinding of the photodissociated CO or coordination of an intrinsic amino acid residue occurred within the pulse duration, and as a result, no (Fe-His)-derived modes were observed.
However, these results do not rule out His coordination in holo-CcmE as described below. It is well known that the frequency of (Fe-CO) is correlated with that of (C-O) due to the electron back donation from the iron d (d xz and d yz ) orbitals to the CO * orbital, and this correlation provides information on the proximal ligand of the heme (17)(18)(19). The CO derivatives of hemoproteins or porphyrins with a similar proximal ligand (e.g. imidazole, imidazolate) fall on the same correlation line, whereas those for proteins with a strong -donating ligand such as thiolate (as in P450, chloroperoxidase, and thromboxane synthase) lie along a different line. RR spectra of CO-bound holo-CcmE showed (Fe-CO) and (C-O) frequencies that are typical of those of proteins with neutral His coordination and fall on the center of the histidine line (Fig. 4). In the case of catalase, which has Tyr as a proximal ligand, the (Fe-CO) and (C-O) frequencies are observed at 542 and 1908 cm Ϫ1 , respectively (36). These values are similar to those observed for many peroxidases, which have an imidazolate (not imidazole) as a proximal ligand and fall on the same line as that of proximal His-ligated proteins. However, they are located on the left side of the line. This clearly indicates that Tyr is not a trans ligand to the bound CO in holo-CcmE. Although a possibility of coordination by Lys or Pro is not excluded, it is most likely from the present experiment that neutral His is a proximal ligand in the CO-bound form, and probably also in the reduced form.
Contrasting with the argument presented above, there is evidence for coordination of an anionic ligand in ferric and ferrous holo-CcmE. The frequency of the -electron marker band ( 11 ) is sensitive to the donor strength of the axial ligands. The empirical correlation of the 11 frequency with the electronegativity of axial imidazole ligands has been investigated using the bis(imidazole) complex of iron(II)-protoporphyrin [Fe(II)PP(ImH) 2 ] (22). The 11 mode was observed at 1533 cm Ϫ1 for Fe(II)PP(ImH) 2 but was downshifted to 1526 cm Ϫ1 on deprotonation of one imidazole [Fe(II)PP(ImH)(Im Ϫ )] and to 1517 cm Ϫ1 on deprotonation of both imidazoles [Fe(II)PP(Im Ϫ ) 2 ]. This arises presumably because the deprotonated imidazole increases the back donation from the d orbital of the heme iron to the porphyrin e g * orbital. For holo-CcmE, the 11 frequency (1536 cm Ϫ1 ) is lower than that of cytochrome c 3 (1540 cm Ϫ1 ), which has two neutral histidines as axial ligands (21), suggesting that one of the axial ligands in ferrous CcmE has an anionic character (Fig. 5). As shown in Fig. 2, the ferrous heme of holo-CcmE is in the 6c form. Since one of the two axial ligands appears to be a neutral histidine as discussed above, the other axial ligand trans to the His may account for the negative charge.
To further characterize the heme ligands of CcmE, we measured RR spectra with excitation at 363.8, 488.0, and 514.5 nm. The Fe-Tyr stretching mode is known to be detectable with excitation at 488.0 or 514.5 nm (37). For Hb M Iwate and Hb M Hyde Park, which have Tyr as a proximal ligand, and for Mb and heme oxygenase mutants, in which proximal His was replaced by Tyr, the Fe-O(Tyr Ϫ ) stretching modes were observed around ϳ590 cm Ϫ1 when they were excited at 488 or 514.5 nm (25)(26)(27)38). However, no such vibrational mode was observed for holo-CcmE with 488-and 514.5-nm excitation (data not shown). However, with excitation at 363.8 nm, a new band appeared at 600 cm Ϫ1 ; this band was not present in the spectrum of the Y134F variant excited at 363.8 nm (Fig. 6). Since this frequency is close to previously observed Fe-Tyr(O Ϫ ) stretching modes (588 cm Ϫ1 for Hb M Iwate and Hyde Park, 585 cm Ϫ1 for Mb H93Y mutant, and 591 cm Ϫ1 for H25Y HO-1 heme complex), and moreover, the 585 cm Ϫ1 band of Mb H93Y mutant was also observed when it was excited at 363.8 nm (27), it is likely that the 600 cm Ϫ1 band observed for holo-CcmE indicates Tyr coordination to the heme iron. Furthermore, UV RR spectra excited at 244 nm provided additional evidence in favor of this. The band at 1605 cm Ϫ1 , which is derived from the ring C-C stretching of tyrosinate (30), was observed for wildtype holo-CcmE but was missing from the spectra of the Y134F variant, suggesting tyrosinate coordination in the wild-type holo-protein. The Fe-Tyr bond presumably yields no chargetransfer band around 500 nm as seen for other Tyr-coordinated heme proteins, and this is the reason why the Fe-Tyr Ϫ stretching was not detected upon excitation at 488.0 or 514.5 nm.
In summary, the present observations suggest that the covalently bound heme in holo-CcmE is ligated by a His and Tyr 134 . Interestingly, the same heme coordination was observed in the bacterial hemophore, HasA; His/Tyr heme coordination was observed in the crystal structure of the protein (39). HasA has been shown in some Gram-negative bacteria to function in extracellular heme binding in an iron acquisition pathway (40), which involves initial heme binding and subsequent release to target proteins. It is striking that this function is similar to that of CcmE and raises the possibility that this type of coordination is appropriate for controllable and transient heme binding.
Cross-linkage between Heme and holo-CcmE-As shown in Fig. 9, two vinyl stretching bands were observed at 1609 and 1621 cm Ϫ1 for the non-covalent complex of CcmE and heme, whereas the latter band was not present in the spectra of wildtype holo-CcmE. This clearly indicates that one of the two vinyl modes is absent from holo-CcmE and supports the hypothesis that heme is covalently attached to the conserved His residue (His 130 in E. coli) (34) of CcmE via a heme vinyl group (11). The 6 cm Ϫ1 upshift of the 2 mode for holo-CcmE (1588 cm Ϫ1 ) as compared with that of the reconstituted CcmE-heme non-covalent complex (1582 cm Ϫ1 , Fig. 9) also supports this idea. Since 2 , which primarily involves C b -C b stretching, is coupled with vinyl CϭC stretching (41,42), removal of the vinyl group leads to an increase in the frequency of 2 due to decoupling of the 2 and vinyl stretching modes (41)(42)(43). In fact, horseradish peroxidase reconstituted with mesoheme, which has ethyl groups instead of vinyl groups in the 2-and 4-positions of heme, showed an upshifted 2 at 1582 cm Ϫ1 as compared with that observed at 1573 cm Ϫ1 for native horseradish peroxidase (44). Previous studies have shown that the 2-vinyl mode has a more significant effect on the 2 frequency than the 4-vinyl (45). When the 2-or 4-vinyl was substituted with a formyl group, a substantial effect was observed when a formyl group was introduced at position 2. Therefore, it is likely that the 2-vinyl rather than 4-vinyl group of heme is modified in holo-CcmE.
Tryptic digestion of holo-CcmE revealed that heme is covalently bound to the 130 HDENYTPPEVEK 141 peptide (6). In c-type cytochromes, heme is bound to a CXXCH motif in the protein. The first and second cysteines in this motif form thioether bonds with the 2-and 4-vinyl groups, respectively; the His always serves as an axial ligand to the heme. If Tyr 134 is an axial heme ligand in CcmE, as suggested by this work, His 130 , which is 4 residues away from Tyr 134 in the amino acid sequence, is located in the same position as the first Cys in the CXXCH of cytochromes c, i.e. in position to react with the 2-vinyl group. All these results are consistent with the proposal that His 130 forms a covalent bond with the 2-vinyl group of heme in CcmE.
Interestingly, the RR spectra of the oxidized and reduced Y134F variant showed intense bands at 1654 and 1651 cm Ϫ1 , respectively, which are not present in wild-type holo-CcmE. A formyl CϭO stretching mode of heme a has been observed in the 1600 -1700 cm Ϫ1 region. In a previous RR study using reconstituted Mb with 2-or 4-formylprotoporphyrin IX, the formyl CϭO stretching mode was observed at 1648 cm Ϫ1 (position 2) and 1660 cm Ϫ1 (position 4) in the ferrous form (45). The x-ray structure of a truncated cytochrome c expressed in E. coli, from which the N-terminal signal sequence was removed, showed that the 2-vinyl group of heme was oxidized by an unknown route to a formyl group (46). On the basis of these previous observations, we cannot exclude the possibility that a formyl group is present on the heme of the Y134F variant. Although this point will have to be further investigated, it is possible that modification of the heme might have occurred in the Y134F variant as a consequence of mutation of the heme axial ligand.
The Heme Binding Site in Holo-CcmE-In contrast to what would be expected for a heme transfer protein, the RR spectra of the CO-bound state of CcmE suggest that the heme of holo-CcmE is not exposed to solvent. When bound CO is exposed to solvent, which was observed in Mb mutants in which the distal histidine was replaced with glycine or alanine, the C-O stretching mode displays line broadening due to multiple components caused by the introduced water molecules (47). However, such broadening was not observed for holo-CcmE. The half-width (⌫1 ⁄2 ) of the C-O stretching mode is ϳ20 cm Ϫ1 and almost identical to that of Mb (⌫1 ⁄2 ϭ ϳ20 cm Ϫ1 ). Since the C-terminal domain of apoCcmE has been observed by NMR to be flexible (7), this region of the protein could change its conformation to cover the heme upon heme binding. It could be that this conformational change is redox-dependent since the heme coordination appears to change from 5c to 6c upon reduction.
The propionate bending mode, ␦(C ␤ C c C d ), was observed at 391 cm Ϫ1 for holo-CcmE in the ferrous form (Fig. 9, spectrum  a). The propionate bending mode frequency correlated with the strength of the hydrogen bond between the propionate and the surrounding amino acid residues (48,49). For instance, the heme-7-propionate group of Mb is hydrogen-bonded to His 97 and Ser 92 , and the ␦(C ␤ C c C d ) mode appeared at 376 cm Ϫ1 . The disruption of this hydrogen bonding in H97F and H97A/S92A mutants resulted in downshifts to 366 and 365 cm Ϫ1 , respectively (50). Therefore, the 7 cm Ϫ1 upshift of the ␦(C ␤ C c C d ) mode observed in holo-CcmE as compared with that of the heme-CcmE non-covalent complex suggests the presence of stronger hydrogen bonding between the heme propionate groups and nearby amino acids and also indicates that the environment of the propionate groups is altered by the covalent bond formation in the holo-protein.
It has been observed that proteins that bind, transport, and release iron or heme have common structural features (51). One of these principles is that the proteins consist of two domains connected by a flexible linker region, which acts as a hinge between the two domains and facilitates their movement. This domain arrangement allows the protein to completely enclose the heme (or iron), protecting the surroundings from their harmful effects while providing it with the flexibility to allow release at an appropriate time. CcmE seems to have a similar arrangement that is consistent with the flexibility required for heme binding and release.