A Cytochrome b562 Variant with a c-Type Cytochrome CXXCH Heme-binding Motif as a Probe of the Escherichia coli Cytochrome c Maturation System*

Cytochrome b562 is a periplasmic Escherichia coli protein; previous work has shown that heme can be attached covalently in vivo as a consequence of introduction of one or two cysteines into the heme-binding pocket. A heterogeneous mixture of products was obtained, and it was not established whether the covalent bond formation was catalyzed or spontaneous. Here, we show that coexpression from plasmids of a variant of cytochrome b562 containing a CXXCH heme-binding motif with the E. coli cytochrome c maturation (Ccm) proteins results in an essentially homogeneous product that is a correctly matured c-type cytochrome. Formation of the holocytochrome was accompanied by substantial production of its apo form, in which, for the protein as isolated, there is a disulfide bond between the two cysteines in the CXXCH motif. Following addition of heme to reduced CXXCH apoprotein, spontaneous covalent addition of heme to polypeptide occurred in vitro. Strikingly, the spectral properties were very similar to those of the material obtained from cells in which presumed uncatalyzed addition of heme (i.e. in the absence of Ccm) had been observed. The major product from uncatalyzed heme attachment was an incorrectly matured cytochrome with the heme rotated by 180° relative to its normal orientation. The contrast between Ccm-dependent and Ccm-independent covalent attachment of heme indicates that the Ccm apparatus presents heme to the protein only in the orientation that results in formation of the correct product and also that heme does not become covalently attached to the apocytochrome b562 CXXCH variant without being handled by the Ccm system in the periplasm. The CXXCH variant of cytochrome b562 was also expressed in E. coli strains deficient in the periplasmic reductant DsbD or oxidant DsbA. In the DsbA- strain under aerobic conditions, c-type cytochromes were made abundantly and correctly when the Ccm proteins were expressed. This contrasts with previous reports indicating that DsbA is essential for cytochrome c biogenesis in E. coli.

Cytochrome b 562 is a hemoprotein expressed in the periplasm of the Gram-negative bacterium Escherichia coli (1). It is the only b-type cytochrome (a complex of protein and heme in which the heme is not attached covalently to the polypeptide) known to occur naturally in that compartment of the E. coli cell. Cytochrome b 562 has a four ␣-helix bundle structure (2)(3)(4), and its apoprotein (the heme-free form) is also stable and structurally relatively ordered (5). The other periplasmic cytochromes in E. coli are of the c-type, i.e. they contain heme covalently attached to polypeptide through thioether bonds between the vinyl groups of heme and the cysteine sulfurs of a CXXCH peptide motif (6 -9). Biogenesis of c-type cytochromes in E. coli and many other Gram-negative bacteria is a complex process thought to require at least 12 proteins. These are the cytochrome c maturation (Ccm) 1 system (CcmABCDEFGH) and the disulfide bond-oxidizing and disulfide bond-reducing proteins DsbA, DsbB, DsbD, and TrxA (6, 9 -11). A disulfide bond is commonly thought to form between the cysteines of the c-type cytochrome CXXCH heme-binding motif, and this has been shown to occur in vitro in the apoproteins of several monoheme cytochromes c (12,13). The Dsb and Ccm systems are believed to work in series during cytochrome c biogenesis such that as the unfolded apocytochrome is exported to the periplasm, it is first oxidized by DsbA (whose oxidant is DsbB). The apocytochrome disulfide bond must be reduced before heme can be attached, and this reduction is achieved by CcmG and/or CcmH, which is, in turn, reduced by the transmembrane protein DsbD, whose electron donor is TrxA (Refs. 14 and 15; reviewed in Refs. 8 and 10).
It has been shown that E. coli cytochrome b 562 can be altered by site-directed mutagenesis such that it will form c-type cytochromes in the periplasm (16). In cytochrome b 562 , a histidine at position 102 is the proximal ligand to the heme iron; an analogous histidine is a virtually universal feature of both band c-type cytochromes. Barker et al. (16) constructed R98C, Y101C, and R98C/Y101C variants of b 562 , resulting in CXXXH, XXXCH, and CXXCH heme-binding motifs respectively. Each of these proteins could be detected in the periplasm of E. coli as covalent complexes of apoprotein and heme. The R98C/Y101C variant did not form as a homogeneous c-type cytochrome product, but rather as two major and two minor products. From these studies, it was unclear whether the covalent attachment of heme was catalyzed or spontaneous. For each variant, large amounts of apoprotein were also detected (16).
When those previous experiments (16) were conducted, little was known about the cytochrome c biogenesis apparatus used by E. coli. Subsequent work identified the Ccm system (CcmABCDEFGH) (17,18), which is expressed physiologically only under certain anaerobic growth conditions. However, these gene products, overexpressed constitutively from a plasmid lacking anaerobic control elements, are now widely used for overproduction of c-type cytochromes when coexpressed with a cytochrome c gene (17, 19 -21). The Ccm apparatus has been shown to process cytochromes c from diverse prokaryotic and eukaryotic origins (e.g. Refs. 17 and 19 -21). In light of these recent data, we have investigated the action of the Ccm system on the R98C/Y101C (CXXCH) variant of E. coli cytochrome b 562 . To our knowledge, this is the first study of the Ccm system in conjunction with an "artificial" cytochrome c. Furthermore, we have investigated the effect of mutations in the Dsb system with respect to production of this cytochrome in the E. coli periplasm.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-E. coli strain JCB387 (22) was used in all experiments unless stated. Other E. coli strains used had mutations in the genes for the periplasmic disulfideoxidizing and disulfide-reducing proteins DsbA and DsbD; these strains were JCB3510 (dsbA Ϫ ) (23), JCB571 (dsbA Ϫ ) (24), and JCB606 (dsbD Ϫ ) (14,25). Note that JCB387, JCB606, and JCB3510 are isogenic; all were the kind gift of Professor Jeff Cole (University of Birmingham, Birmingham, United Kingdom). JCB571 was the gift of Professor James Bardwell (University of Michigan). Plasmids for wild-type E. coli cytochrome b 562 and the R98C/Y101C (CXXCH heme-binding motif) variant were as described (16). Plasmid pKPD1, encoding the gene for Paracoccus denitrificans cytochrome c 550 , was as described by Sambongi and Ferguson (26). As required, cells were also transformed with plasmid pEC86 (17), encoding the E. coli ccm genes (ccmABCDEFGH). Transformants were initially grown on LB agar plates with the appropriate antibiotics (100 g/ml ampicillin in each case plus 34 g/ml chloramphenicol when pEC86 was cotransformed and 25 g/ml kanamycin for the dsbA Ϫ strains). Single colonies were picked into 500 ml of 2ϫ TY medium containing 16 g/liter peptone, 10 g/liter yeast extract, and 5 g/liter NaCl supplemented with 1 mM isopropyl-␤-D-thiogalactopyranoside in 2.5-liter flasks. Cultures were grown at 37°C with shaking at 200 rpm for 20 -24 h before harvesting.
Cell Fractionation and Biochemical Methods-Periplasmic fractions were obtained from cells using the procedures described previously (27). Cytochrome content was determined by recording the absorption spectra of the crude or purified periplasmic fractions. E. coli produces small amounts of endogenous c-type cytochromes in the periplasm under our growth conditions (see, for example, Ref. 27). To provide reference data to correct for the level of this background cytochrome expression, E. coli JCB387 cells were transformed and grown with pTZ19r (an ampicillin resistance-conferring cloning vector; Fermentas) and pEC86. These cells were fractionated, and the absorbances of the periplasmic fractions (normalized in relation to volume and wet cell weight and averaged from several cultures) were taken to represent "typical" endogenous (background) cytochrome expression levels. This cytochrome yield cannot be expressed directly because it represents a mixture of components with different (and in some cases, undetermined) extinction coefficients. However, the endogenous cytochrome production can be expressed in terms of absorbance (which is directly proportional to the concentrations of the species present). The absorbance of the periplasmic fraction from such cells (prepared using our routine growth and fractionation procedures) was 0.026 A units (mean average for the maximum at 418 nm, S.D. of 0.004 A units)/g of cells (wet weight). When yields of periplasmic exogenous cytochromes are expressed in this work, this endogenous cytochrome absorbance has been subtracted from the actual observed absorbance to determine the expression level.
Apo-and holocytochromes b 562 R98C/Y101C were purified on a DEAE-Sepharose anion-exchange column (Amersham Biosciences XK26/20 with a resin bed volume of 70 ml). Chromatography was conducted at 4°C in 50 mM Tris-HCl, pH 8.0. The column was eluted with a 0 -0.5 M NaCl gradient at a flow rate of 9 ml/min, and 7-ml fractions were collected. Apocytochrome b 562 R98C/Y101C clearly separated from the holocytochrome following this chromatographic step. Apocytochrome-containing fractions were identified from their absorption spectra following addition of dithiothreitol (DTT) and heme and were nearly pure as judged by SDS-PAGE. The spectrophotometrically purest holocytochrome fractions were pooled, diluted 4-fold, and loaded onto a Q-Sepharose column (XK26/20), which was eluted with a 0 -0.25 M NaCl gradient in the same buffer and under the same conditions. After this step, holocytochrome b 562 R98C/Y101C was essentially pure. Wild-type cytochrome b 562 was purified using a protocol modified from that of Nikkila et al. (28). The crude periplasmic extract from fractionated E. coli cells containing the cytochrome was taken to pH 4.55 with concentrated phosphoric acid, stirred on ice for 30 min, and centrifuged at 40,000 ϫ g for 1 h. The cytochrome-containing solution was then adjusted to pH 8.0 and purified on DEAE-Sepharose and Q-Sepharose columns as described above for the R98C/Y101C variant. Wild-type apocytochrome b 562 was prepared by extraction of heme from the holoprotein with imidazole using the procedure described by Tomlinson and Ferguson (29).
Hemin (Fe-protoporphyrin IX-Cl) was added to protein solutions from a stock solution in Me 2 SO. NMR spectra were collected as described previously (16,30,31) using a Bruker AMX 500 spectrometer or a Bruker Avance 700 spectrometer. The NMR methods used with the different instruments were similar, except that WATERGATE sequences were used on the 700-MHz instrument to remove the residual water signal. All spectra were acquired at 300 K. Data processing was as described previously (16,30,31), except that it was performed using XWINNMR (Bruker). All samples were between 0.4 and 1 mM protein in 20 mM potassium phosphate buffer in D 2 O, pH* 6.6, and 0.5 M KCl. Where appropriate, the proteins were reduced by addition of sodium dithionite to ϳ5 mM from anaerobic stock solutions in the same buffer. Electrospray ionization mass spectrometry was performed on a Micromass Bio-Q II-2S triple quadrupole atmospheric pressure instrument equipped with an electrospray interface. Samples were introduced via a loop injector as a solution (20 pmol/l in 1:1 water/acetonitrile and 0.2%  (38).

RESULTS AND DISCUSSION
Formation of the R98C/Y101C Variant of E. coli Cytochrome b 562 with and without Expression of the Ccm Proteins-E. coli strain JCB387 was transformed with a plasmid encoding the R98C/Y101C variant of cytochrome b 562 (16), which has the CXXCH heme-binding motif that is the general characteristic of cytochromes c. This construct includes the natural signal sequence of cytochrome b 562 , which directs the apoprotein to the periplasm, where both wild-type b 562 and all Gram-negative bacterial cytochromes c are naturally assembled. Cells were cotransformed with plasmid pEC86, which encodes the E. coli ccm genes (17). They were grown aerobically (which represses expression of the endogenous E. coli Ccm proteins and also production of the endogenous c-type cytochromes) (34), and periplasmic fractions were prepared. The absorption spectrum of the dithionite-treated periplasmic extract from these cells had resolved peaks characteristic of a low spin, six-coordinate cytochrome. The absorbance in the Soret band region (normalized in relation to fraction volume and wet cell weight), was 5-30-fold in excess of that in control (background expression) experiments (see "Experimental Procedures" for details of these controls). The ␣-band of the dominant observed cytochrome was at 556 nm ( Fig. 1 and Table I). This value is red-shifted relative to most cytochromes c (e.g. Refs. 35 and 36), but is blue-shifted by 6 nm relative to the ␣-band observed for wild-type cytochrome b 562 at 562 nm. Such a shift to higher energy is consistent with saturation of the heme vinyl groups and hence covalent attachment of heme to polypeptide. Cytochrome c 556 , which shares the same fold and heme coordination as cytochrome b 562 , but is a natural c-type cytochrome (16,37), also has its ␣-band at 556 nm. The other main absorption peaks in the spectrum of the dithionite-reduced periplasmic extract from cells expressing b 562 R98C/Y101C in the presence of the Ccm proteins were at 526 and 421 nm ( Fig. 1 and Table I). Treatment of an aliquot of a cytochrome with an alkaline pyridine solution and dithionite produces a spectrum that is highly characteristic of the nature of attachment of heme to protein (38). In the present case, the ␣-band of such a spectrum of the pyridine-and hydroxide-treated crude periplasmic extract was at 550 nm ( Fig. 1, inset), the wavelength expected for covalent attachment of heme to protein through two thioether linkages to a CXXCH peptide motif, as found in a typical cytochrome c.
To assess the homogeneity of the product, cytochrome-containing fractions were purified from the crude periplasmic extract of cells that had expressed b 562 R98C/Y101C in the presence of the Ccm proteins. The periplasmic extract was chromatographed on a DEAE-Sepharose column (see "Experimental Procedures" for details). Cytochrome-containing fractions (identified by their red color) eluted in a continuum that consisted of three parts upon spectral analysis. The first group of these fractions had absorption and pyridine hemochrome maxima characteristic of c-type cytochromes and similar to those expected for the endogenous soluble E. coli cytochrome c NrfA (39). The second and clearly spectroscopically dominant group was the c-type cytochrome form of b 562 R98C/Y101C, with its ␣-band absorption maximum at 556 nm and its pyridine hemochrome ␣-band maximum at 550 nm. The third (small) set of fractions contained a mixture of at least two species, one of which was the b 562 R98C/Y101C c-type cytochrome. The absorption spectra of these later fractions were red-shifted by 1-1.5 nm, and the pyridine hemochrome spectra were red-shifted by up to 1 nm relative to the uncontaminated b 562 R98C/Y101C c-type cytochrome fractions. Thus, in addition to the major b 562 R98C/Y101C c-type cytochrome product, these fractions may contain some malformed (e.g. single cysteine-attached) cytochrome c and/or some b-type cytochrome from lysed cytoplasm contaminating the periplasmic cell extract, either of which would have the observed effect on the spectra. Quantitatively, our chromatographic analysis showed that at least 95% of the holocytochrome b 562 R98C/Y101C produced in the presence of the Ccm proteins was in the major cytochrome-containing band (i.e. in the second group of fractions) and thus was a c-type cytochrome with spectra wholly characteristic of covalent attachment of heme to protein through two thioether bonds.
Barker et al. (16) purified "component II" of the mixture of two major and two minor products that was observed when b 562 R98C/Y101C was expressed in the periplasm of E. coli under different growth conditions and with little control over the expression of the Ccm system, which had not then been identified. The purified holocytochrome was studied by NMR spectroscopy; it was shown to be homogeneous, and the heme was attached covalently to the cytochrome with the same stereochemistry as observed for all naturally occurring cytochromes c (7,16). Because the absorption spectra of component II are essentially identical to those observed in the present work for cytochrome b 562 R98C/Y101C produced in the presence of the Ccm proteins, we have also analyzed the Ccm-matured holocytochrome by NMR. The main cytochrome species (accounting for Ͼ95% of the total holocytochrome b 562 R98C/Y101C produced (see above)) was further purified to remove any noncytochrome contaminants. It was found by NMR to be homogeneous (i.e. a single cytochrome species) and identical in both the oxidized and reduced states to component II from the earlier work (16) and thus a properly matured c-type cytochrome. The chemical shifts of the heme, aromatic, and distal methionine heme ligand protons were within 0.01 ppm of those previously reported (see Supplemental Figs. S1-S3 for the NMR data). We therefore conclude that component II was holocytochrome matured by the Ccm system operating at a background level under the growth conditions that were used earlier (16). When apparently uncatalyzed covalent heme attachment to b 562 R98C/Y101C occurred, three products that did not have the correct heme attachment were apparent (Ref. 16 and see also data below). In the similarly well characterized case of uncatalyzed covalent heme attachment to Thermus thermophilus cytochrome c 552 , at least two distinct products have been obtained (40,41).
In contrast to the experiments in which the Ccm proteins were expressed from plasmid pEC86, in the absence of this plasmid, we detected a heterogeneous mixture of forms of holocytochrome b 562 R98C/Y101C in the periplasmic fraction. These observations are therefore similar to those reported previously (16). The absorption spectrum of our dithionite-reduced crude periplasmic extract was notably red-shifted relative to the case with the Ccm system present ( Fig. 1 and Table I). The ␣-band maximum of the pyridine hemochrome spectrum of the product obtained in the absence of Ccm was at ϳ552.5 nm (Table I), consistent with the presence of improperly matured forms of this c-type cytochrome. A b-type cytochrome such as wild-type b 562 (in which the heme and protein form a noncova-lent complex) has a pyridine hemochrome maximum at 556 nm, and a properly matured c-type cytochrome with double cysteine attachment (through two thioether bonds) has it at 550 nm (38); a c-type cytochrome with only one thioether bond between the heme and peptide, such as the naturally occurring Euglena gracilis cytochrome c 558 , has it at ϳ553 nm (42).
Our data show unambiguously that the E. coli Ccm system can act on a variant of cytochrome b 562 with a CXXCH hemebinding motif and produce a c-type cytochrome with the correct covalent attachment of heme to protein. The Ccm proteins act as a very effective quality control system for the product cytochrome (clearly illustrated by its homogeneity and NMR spectra). The yield of holocytochrome b 562 R98C/Y101C produced in the presence of the plasmid-expressed Ccm proteins varied from 0.3 to 1.8 mg/g of cells (wet weight). To our knowledge, this is the first convincing evidence that the Ccm apparatus can act on an "artificial" apocytochrome c. The Ccm system, found in some Gram-negative bacteria and in plant mitochondria (6,8,43), has previously been shown to act on eukaryotic cytochromes c (19,21) and thus to be versatile in terms of substrate. However, there is no reason why apocytochrome b 562 R98C/Y101C, based on the scaffold of a b-type cytochrome, should have any recognition factors required for cytochrome c biogenesis (i.e. the action of the Ccm proteins) apart from, simply, an artificially created CXXCH peptide motif and a heme-binding pocket.
The yield of the homogeneous c-type holocytochrome b 562 R98C/Y101C produced in the presence of the Ccm system (i.e. with pEC86) varied between cell growths from 25 to 100% of that of the combined yield of all the holocytochrome species produced without expression of the Ccm proteins (given that the Soret band extinction coefficients of all the relevant species are the same to within 10% (16)). Thus, despite increasing the quality of the product, the Ccm system did not catalyze the formation of increased levels of cytochrome with covalently bound heme and often lowered the overall holocytochrome yield. However, one cannot formally exclude that other factors, e.g. the nature of heme provision (see below), the presence of a second antibiotic in the growth medium, or additional competition for the cellular transcription/translation machinery, affected the relative cytochrome yields.
Presence of Apocytochrome b 562 R98C/Y101C-In the previous study of expression of R98C, Y101C, and R98C/Y101C variants of cytochrome b 562 in the E. coli periplasm (16), a significant amount of apocytochrome (i.e. heme-free protein) was observed in each case. We therefore assessed whether the same was true under our (different) growth conditions. Aliquots of the periplasmic extract were treated with DTT and dithionite to reduce any potential disulfide bond formed between the cysteines of the CXXCH motif of apocytochrome b 562 R98C/Y101C. The periplasmic material was then titrated with heme, and the change in the absorption ␣-band (around 560 nm) was observed. A sharp ␣-band is characteristic of a reduced six-coordinate cytochrome heme iron, but not of free heme. To normalize for the spectral contribution of free and nonspecifically bound heme, we analyzed the change in intensity of the ␣-band (i.e. peak maximum minus trough minimum). Heme was added until this value did not change significantly, regarded as the point at which apocytochrome b 562 R98C/Y101C was saturated with heme. As reported (16), addition of heme to this variant of apocytochrome b 562 results initially in a spectrum very similar to that of wild-type cytochrome b 562 ; thus, for quantitation purposes, we used the wild-type extinction coefficients for this protein (16). The extinction coefficients used for the initial (untitrated) periplasmic cytochromes were as published (16). Analysis of this type showed that when b 562 R98C/ Y101C was expressed in either the presence or absence of the Ccm proteins (from pEC86), at least 75% of the b 562 R98C/ Y101C polypeptide was present as apoprotein (i.e. without heme bound). Apocytochrome b 562 R98C/Y101C could be purified from the periplasmic extract with relative ease; the apo form eluted from an anion-exchange column well before the holo form and was retained for use in experiments when it was desirable to avoid the complication of background cytochrome absorbance.
The presence of a large amount of apocytochrome b 562 R98C/ Y101C in the E. coli periplasm raises issues about the provision of heme to that compartment of the cell. It is not known how heme is transported across the cytoplasmic membrane (from its site of synthesis in the cytoplasm) for formation of either cytochromes c or cytochrome b 562 in the periplasm (see Ref. 8 for discussion). However, our results indicate, as do previous observations (16,44), that heme synthesis and transport across the membrane are not governed (only) by the availability of a periplasmic receptor apocytochrome. The similar ratio of apoto holocytochrome b 562 R98C/Y101C in both the presence and absence of the Ccm proteins in this work, particularly when considered with the generally lower total holocytochrome yield when the Ccm proteins are expressed, adds further weight to the argument that the Ccm system does not act as (or include) a heme transporter from the cytoplasm to the periplasm for cytochrome c production. However, the central question of how heme gets to the periplasm at all remains extant and is a crucial issue for cytochrome c biogenesis.
Disulfide Bond in the CXXCH Heme-binding Motif of the Apocytochrome-In light of the recent data showing that apocytochromes c can form a disulfide bond in the CXXCH motif in vitro (12,13), it was important to determine, at least qualitatively, the oxidation state of apocytochrome b 562 R98C/Y101C in the periplasmic fractions of our cells. To obtain reference spectra, we first analyzed the purified b 562 R98C/Y101C apoprotein (Fig. 2). Upon addition of oxidized (Fe 3ϩ ) heme to this protein under oxidizing (atmospheric) conditions without addition of reductant, there were increases in the absorption bands in the region at 350 -390 nm. These wavelengths are where free heme and heme nonspecifically bound to protein in a high spin complex absorb. However, upon subsequent addition of DTT to reduce any disulfide bond in the apocytochrome, a spectrum characteristic of a reduced, low spin, six-coordinate b-type cytochrome appeared rapidly (Fig. 2). In contrast, when oxidized heme was added to pure wild-type apocytochrome b 562 under oxidizing conditions, the heme bound readily, resulting in the characteristic spectrum of an oxidized low spin cytochrome (Fig. 2). Clearly, no disulfide bond can form in the wild-type protein, as it does not have any cysteine residues (2). The present results indicate the presence of a disulfide bond, which may be reduced by DTT, in the b 562 R98C/Y101C apoprotein. They were replicated with freshly prepared crude periplasmic extracts, indicating that a disulfide bond had formed in the periplasmic apocytochrome b 562 R98C/Y101C either in the cell or during the cell fractionation procedure. The purified apocytochrome showed no reaction with Ellman's reagent under oxidizing conditions, also indicating the absence of free thiol groups (and hence the presence of a disulfide bond between the cysteines of the CXXCH motif).
Electrospray mass spectrometric analysis of the b 562 CXXCH apoprotein under oxidizing conditions indicated that the protein was present almost entirely (Ͼ95%) as a monomer, rather than as a dimer or oligomer (with intermolecular disulfides). The mass observed for the monomeric protein was 11,664.54 Ϯ 0.02 Da, compared with a value of 11,665.1 Da calculated from the peptide sequence for apocytochrome b 562 R98C/Y101C with the two thiols linked by a disulfide bond. When a sample of this protein, after reduction overnight with DTT, was analyzed by electrospray ionization mass spectrometry and compared immediately, and under identical conditions, with a sample of the oxidized protein (both calibrated against the same equine myoglobin standard), a mass increase of 1.73 Ϯ 0.07 Da was observed. These data are highly indicative of reduction of a disulfide bond in apocytochrome b 562 R98C/Y101C to produce free thiol groups and thus rigorously confirm the presence of an intramolecular disulfide bond in the oxidized material. The accuracy and errors of such electrospray ionization mass spectrometric methodology have been established recently in a study of human hemoglobin heterozygotes differing by as little as 1 Da (32). Note that because of the presence of the reductants required to reduce the disulfide bond in apocytochrome b 562 R98C/Y101C, the resulting cytochrome is reduced (-----), giving rise to a spectrum distinct from that of wild-type cytochrome b 562 (OO), which is oxidized. These spectra were normalized based on the extinction coefficients given in Ref. 16 for oxidized and reduced wild-type holocytochrome b 562 . Spectra were recorded ϳ2 min after addition of heme to the cuvette with the proteins in 50 mM potassium phosphate buffer, pH 7.0.
Taken together, these different types of experiments show that apocytochrome b 562 R98C/Y101C can form between its cysteines a disulfide bond that prevents heme from binding correctly. This could be due to the disulfide sterically blocking the heme-binding pocket or, alternatively, to restriction of the conformation of the C-terminal helix, altering the side chain packing, including that of the important proximal heme-ligating residue His-102. Upon addition of a suitable reducing agent, the disulfide is broken, and heme may be quickly bound to the apocytochrome, just as it is in oxidized wild-type holocytochrome b 562 . Note that it is necessary for the disulfide to be reduced for correct heme binding to occur irrespective of the oxidation state of the heme iron. After fractionation of cells over ϳ1-2 h following initial harvesting, the apocytochrome b 562 R98C/Y101C obtained was essentially oxidized. Given this, we assessed the rate of formation of the disulfide bond in vitro.
Samples of purified apoprotein were reduced by incubation overnight at room temperature with a large molar excess of dithionite and either DTT or tris(carboxyethyl)phosphine. The excess reductant was removed using a desalting column containing P6-DG resin (Bio-Rad). The column had been calibrated with a mixture of horse heart cytochrome c and potassium ferricyanide, which were observed by their distinct colors to separate clearly. The column was pre-equilibrated and eluted with 0.5 M sucrose, 1 mM EDTA, and 200 mM Tris-HCl, pH 7.3, under normal (atmospheric) oxygen tension to replicate the conditions used for cell fractionation in this work. The eluted protein was assayed by addition of ferric heme; the first spectrum was recorded immediately after the protein eluted from the desalting column, i.e. ϳ20 min after the reduced protein was loaded onto the column. In each case, the resulting spectrum was essentially characteristic of unbound heme. These data suggest that the disulfide bond in the CXXCH motif of b 562 R98C/Y101C formed, almost completely, in Ͻ20 min in a nor-mally aerated solution. As shown above (Fig. 2), addition of oxidized heme to DTT-reduced apocytochrome b 562 R98C/ Y101C did initially give rise to a cytochrome spectrum with absorbance maxima characteristic of cytochrome b 562 . That the eluted apocytochrome b 562 R98C/Y101C was mainly reoxidized after ϳ20 min was also demonstrated with Ellman's reagent. The spectra obtained after reduction of protein with DTT, passage down the P6-DG column, and treatment with Ellman's reagent indicated that not more than 25% of the thiols in the sample remained in the reduced state after ϳ20 min (i.e. Ն75% of the protein sample was oxidized).
Incorrect Holocytochrome Formation in Vitro from the Reaction of Apocytochrome b 562 R98C/Y101C with Heme-It is apparent that in the absence of the Ccm proteins expressed from pEC86, some correct product cytochrome c may be formed from the reaction between b 562 R98C/Y101C and heme (this work and Ref. 16). This might be due to a low level background activity of the endogenous E. coli Ccm proteins, e.g. if the cell cultures are not fully aerobic (27). Alternatively, it may be that this is the result of spontaneous (uncatalyzed) thioether bond formation, as has been observed recently in vitro for three cytochromes c (12,13). Thus, we investigated the extent of the in vitro (i.e. non-enzyme-catalyzed) reaction between the apo form of b 562 R98C/Y101C and heme. An aliquot of purified apoprotein (ϳ20 M) was mixed with dithionite and a large excess of DTT in a sealed nitrogen-sparged cuvette. Heme was added to a final concentration of ϳ5 M. A spectrum characteristic of that expected for a b-type cytochrome was recorded (absorption maxima at 560.8, 530.4, and 425.3 nm) (Fig. 3 and Table I) bonds between heme and protein (12,13). After 48 h, the spectrum of the apocytochrome b 562 R98C/Y101C and heme mixture had changed; all the absorption bands had blue-shifted relative to their starting positions (maxima now at 559.4, 529.0 and 424.1 nm) (Fig. 3 and Table I). The pyridine hemochrome ␣-band maximum of the product after 48 h was at 552 nm (compared with 556 nm for noncovalently bound heme). These data indicate that covalent attachment of heme to protein had occurred, but that the product was not a properly matured c-type cytochrome. Qualitatively, the spectral characteristics of the product mixture were similar to those of the crude periplasmic extract obtained when b 562 R98C/Y101C was expressed without the plasmid-expressed Ccm proteins (Fig. 3 and Table  I; cf. Fig. 1).
The products of the spontaneous in vitro reaction, prepared on a larger scale, were purified by ion-exchange chromatography. One major fraction containing covalently attached heme could be identified, but small amounts of other such species were observed in the purification. The major species eluted from anion-exchange resin earlier than expected (based upon the elution of wild-type holocytochrome b 562 ) and had optical spectra essentially indistinguishable from those of component I of the b 562 R98C/Y101C mixture in the earlier work (16); it was analyzed by 1 H NMR spectroscopy as described previously (16,30,31). The NMR spectra of the reduced (Fe 2ϩ ) form of the in vitro prepared cytochrome were qualitatively identical to those of the in vivo derived component I as reported previously (16); we conclude that this in vitro protein fraction is identical to component I. Thus, the nonenzymatic reaction between heme and apoprotein that occurs in the cuvette is similar to the mode of (presumably uncatalyzed) covalent attachment of heme to apocytochrome b 562 R98C/Y101C in the E. coli periplasm when the Ccm proteins are not appreciably expressed. The NMR spectra of the reduced in vitro produced holoprotein are consistent with the presence of at least two species that are very closely related and cannot be separated from each other. They may represent conformational isomers that are not interchanging on the NMR time scale or may be related to the presence of species with additional mass (ϩ16 or ϩ32 Da) that have been observed previously in samples of this protein (16). Notably, there were two signals from the heme 4-vinyl ␣-proton, two signals from the ␣-meso-proton, three signals from the ␤-mesoproton, and three signals from the Met-7 methyl resonance (Supplemental Fig. S4). However, only one NMR signal was observed from each of the methine and methyl groups generated by the covalent (thioether) attachment at the 2-position of heme to protein, suggesting that the conformational differences between the different species were remote from this point of attachment.
Despite heterogeneity in the nuclear Overhauser effect correlation and total correlation spectra, analysis of the Fe 2ϩ form of component I (16) revealed the presence of signals consistent with one thioether covalent linkage as the heme 2-substituent and one intact heme vinyl group as the 4-substituent in all of the species present. Because component I is spectroscopically identical to the major product with covalently attached heme formed in vitro from the reaction of heme and apocytochrome b 562 R98C/Y101C, these signals are also apparent in the spectra of the latter material (Supplemental Fig. S4). The loss of only one vinyl group is consistent with the pyridine hemochrome spectra (Table I) (16). More strikingly, NMR analysis clearly indicated that the heme was in an alternative (so-called "minor") orientation, i.e. rotated 180°about its ␣,␥-meso-axis (e.g. Refs. 12 and 45), within the protein in all of the species present. Thus, both in vivo in the absence of the ccm gene products and in vitro, Cys-101 has reacted with the 2-vinyl group of heme. In contrast, in all structurally characterized examples of the products of heme attachment catalyzed by any known biogenesis machinery, including the Ccm system in this work, the 2-vinyl group reacts with the first cysteine in the CXXCH motif (in the present case, residue 98).
Two main sets of paramagnetically shifted signals were also observed in the 1 H NMR spectra of the oxidized (Fe 3ϩ ) in vitro derived protein with covalently bound heme (Supplemental Fig. S5). Again, this is in stark contrast to the holocytochrome matured in vivo by the Ccm apparatus that was found to be homogeneous by NMR (Supplemental Fig. S2).
This work establishes that whereas Ն95% of the holocytochrome b 562 R98C/Y101C produced in vivo in the presence of the Ccm proteins was a properly matured c-type cytochrome, nearly all of the holocytochrome with covalently bound heme that was formed via an uncatalyzed reaction (in vitro) had the heme misattached. These data highlight the requirement for, and the importance of, the very strict, universally conserved, stereo-and regiospecific control of heme attachment that must be enforced by nature's cytochrome c biogenesis systems, including the Ccm apparatus. Recent work has shown that cytochromes c can sometimes be formed without the assistance of enzymatic catalysis either in vitro or in the cytoplasm of E. coli. The apparent stereospecificity of heme attachment in the product(s) varies with the (apo)cytochrome studied. In the case of H. thermophilus cytochrome c 552 , the product is, as far as may be determined from the available data, properly matured (12,46,47). 2 In the present case, essentially none of the in vitro prepared holocytochrome b 562 R98C/Y101C was correctly formed. Horse heart mitochondrial cytochrome c (13), P. denitrificans cytochrome c 550 (13), and T. thermophilus cytochrome c 552 (40,41) are all intermediate. Presumably, the observed differences between these proteins reflect differences in the structures of the apocytochromes, in the respective heme-binding pockets, and possibly (for the in vitro work) in the method of preparation of the apocytochrome c. The stability of the cytochrome b 562 R98C/Y101C heme-protein complex prepared in vitro, with the heme bound in the alternative minor (rotated) orientation, clearly provides an efficient reaction pathway that could, in principle, compete with the Ccm-catalyzed reaction pathway. The fact that this alternative product is not observed when the ccm gene products are expressed suggests that (i) the Ccm apparatus presents heme to the protein only in the orientation that results in formation of the correct product and (ii) that heme does not become covalently attached to apocytochrome b 562 R98C/Y101C without being handled by the Ccm system in the periplasm. Both these factors are likely to be applicable to, and important for, the maturation of all Ccmmade c-type cytochromes.
Effect of Mutations in the E. coli Periplasmic Disulfide Bondoxidizing/reducing Apparatus-Cytochrome b 562 R98C/Y101C was also expressed in E. coli mutants in which the periplasmic strong disulfide oxidant DsbA or reductant DsbD had been deleted by genetic manipulation. The dsbA strain is JCB3510 (23), and the dsbD strain is JCB606 (14), both isogenic with the parental strain JCB387 used throughout this work. A different (non-isogenic) dsbA strain, JCB571 (24), was used to verify several of the qualitative results. Both DsbA and DsbD are regarded as essential for biogenesis of c-type cytochromes by the E. coli Ccm system. E. coli is unable to synthesize any of its endogenous cytochromes c in mutants with either of these proteins deleted (8,10,14,23).
In the DsbD deletion mutant in the absence of the Ccm proteins expressed from pEC86, periplasmic expression of b 562 R98C/Y101C resulted in a mean 50% decrease in holocytochrome formation relative to the DsbD ϩ (parental) strain. Qualitatively, the absorption and pyridine hemochrome spectra were similar to those obtained for the periplasmic extract from the DsbD ϩ strain, indicating that the mixture of product cytochromes formed was probably similar (Table I). Note, however, that there may be a small contribution to this holocytochrome production from the endogenous E. coli Ccm proteins, the expression of which is not quite zero under our growth conditions (see, for example, Refs. 27 and 48). In the DsbD Ϫ strain, the ratio of apocytochrome b 562 R98C/Y101C to holocytochrome increased typically 4-fold relative to the parental strain (DsbD ϩ ). DsbD provides reductant to the periplasm; its absence correlates with a lower yield of holocytochrome and a higher ratio of apocytochrome to holocytochrome than is observed for the parental strain. These data are consistent with the idea that covalent attachment of heme to apocytochrome c requires the cysteine thiols of the CXXCH motif to be reduced. In the absence of DsbD, less of the apocytochrome c can be reduced, and so less holocytochrome forms. However, the observation that a still significant amount of holocytochrome can form even in the absence of DsbD suggests that, at least under our growth conditions, there is an alternative source of reductant in the periplasm. These electrons may be relayed by a protein electron transport chain or may, for example, be from a component of the rich growth medium that was used. It has been shown that addition of exogenous thiols to growth medium can complement for the DsbD deletion in cytochrome c biogenesis (49). Alternatively, heme acquisition by apocytochrome b 562 R98C/Y101C and oxidation of the cysteine thiols to a disulfide bond may be a kinetically competitive process in the E. coli periplasm. 3 In the dsbA deletion mutant strain JCB3510 in the absence of the Ccm proteins, periplasmic expression of b 562 R98C/ Y101C resulted in total holocytochrome expression similar to that of the parental strain (DsbA ϩ ). Qualitatively, the absorption and pyridine hemochrome spectra were similar to those obtained for the periplasmic extract from the DsbA ϩ strain, indicating that the mixture of product cytochromes formed was probably similar (Table I). More remarkably, when b 562 R98C/ Y101C was coexpressed with the Ccm proteins in E. coli JCB3510 (dsbA Ϫ ), properly matured c-type cytochrome formed with a yield comparable with that in the parental strain (0.4 -1.0 mg of holocytochrome/g of cells (wet weight)). The product was indistinguishable from that when b 562 R98C/Y101C was made in the parental E. coli strain by the Ccm system, as judged by the absorption and pyridine hemochrome spectra (Table I; cf. Fig. 1). We did not anticipate this because it is commonly believed that DsbA is an essential protein for c-type cytochrome biogenesis in E. coli (10,14). However, the data were qualitatively the same for both of our (non-isogenic) DsbA Ϫ strains (JCB3510 and JCB571). One might argue that the result arose because b 562 R98C/Y101C is not a "real" c-type cytochrome. Thus, as a control, the experiments were replicated with the plasmid for b 562 R98C/Y101C replaced with pKPD1 (25), which encodes the gene for P. denitrificans cytochrome c 550 , a typical bacterial monoheme cytochrome c 2 . Once again, large amounts of correctly formed cytochrome c were made in the DsbA Ϫ mutant strain JCB3510 (Fig. 4 and Table  I); the mean holocytochrome c 550 yield was ϳ65% of that in the isogenic parental DsbA ϩ strain (JCB387) (means of 1.5 and 2.3 mg of cytochrome produced per g of cells (wet weight) for the variant and wild-type cells, respectively).
These observations, although unexpected, can be rationalized. Sambongi and Ferguson (50) showed that a DsbA deletion could be chemically complemented for cytochrome c biogenesis by addition of exogenous oxidant to the growth medium. In their experiments, cells were grown anaerobically (necessary to induce expression of the endogenous E. coli Ccm system), and cystine or glutathione was added to complement for the dsbA mutation. In the present work, our cells were grown aerobically (with shaking of the growth flasks). Under our conditions, the Ccm proteins were constitutively expressed from plasmid pEC86. It may be that, in our case, it was oxygen, provided by the growth conditions, that acted as the exogenous oxidant to complement for the DsbA deletion. Alternatively, under our growth conditions, an additional oxidant may be transported to or made in the periplasm. In either case, these observations have potentially important implications for c-type cytochrome biogenesis in Gram-negative bacteria.
E. coli is atypical in that its endogenous c-type cytochromes are expressed only anaerobically. It does not have a cytochrome bc 1 complex (respiratory complex III), nor does it have a cytochrome c implicated in electron transport to a cytochrome oxidase (51). However, other Gram-negative bacteria such as Pseudomonas aeruginosa and Shewanella putrefaciens do make c-type cytochromes aerobically and have more typical aerobic respiratory chains. The implication of the present work is that DsbA may not be needed in these organisms for such aerobic Ccm system-dependent cytochrome c formation. It is presumed that the role of DsbA in cytochrome c biogenesis in E. coli is to form a disulfide bond between the cysteines of the apocytochrome CXXCH motif. At least, in the present case, it appears that an exogenous oxidant can substitute in this role (or, in principle, that the disulfide bond is not required at all). However, it must be considered that in the case of multiheme c-type cytochromes (such as pentaheme NrfA and tetraheme NapC), an additional role of DsbA is to ensure disulfide bond formation within each particular apocytochrome c CXXCH motif. At present, it is not clear whether an oxidant such as exogenous oxygen would be sufficient for this, i.e. whether the tendency for cysteines to form a disulfide within a CXXCH motif (this work and Refs. 12 and 13) is so strong (and fast) as to make formation of mixed disulfides with the cysteines of other CXXCH motifs very unlikely. Even if such mixed disulfides do form quite readily, it may be that DsbA is still not required if the disulfide isomerase DsbC is present and active  and thus able to readily "correct" any inter-CXXCH motif disulfide bonds. Nevertheless, a requirement for DsbA in the biogenesis of multiheme cytochromes c may account for many of the previous observations on its function in E. coli, all of whose natural c-type cytochromes have multiple hemes.
Our observations with the E. coli DsbA mutants in some respects complement those of Deshmukh et al. (52) published during the preparation of this manuscript. They showed that in Rhodobacter capsulatus, CcdA (a homolog of DsbD) is essential for c-type cytochrome biogenesis. However, neither DsbA nor DsbB was essential, at least for the biosynthesis of monoheme cytochromes c, under any growth conditions they investigated.
It is, in fact, becoming apparent that the Ccm/Dsb c-type cytochrome biogenesis system as a whole is somewhat modular, i.e. those components used vary with the particular organism. In E. coli, by far the best studied case, CcmABCDEFGH are all required along with DsbD and the thioredoxin TrxA; DsbA and DsbB are also required under some conditions, but see the results discussed above (6,10). In Bradyrhizobium japonicum and others, an additional protein, CcmI, albeit a homolog of the C terminus of E. coli CcmH, is required (6). In land plant mitochondria, CcmA, CcmB, CcmC, CcmE, CcmF, and possibly CcmH have been identified (43), but CcmD and CcmG have not been to date. CcmG and CcmH are thioredoxin-like components of the Ccm system, and we have suggested that the apocytochrome oxidation state in the intermembrane space of the plant mitochondrion may be such that they are not necessary in this case (8). R. capsulatus and some other organisms use CcdA, an analog of the core transmembrane domain of DsbD, rather than the whole of DsbD as a periplasmic reductant (53), and R. capsulatus does not require DsbA or DsbB for c-type cytochrome biogenesis (52). Thus, flexibility is clearly apparent in the type I c-type cytochrome biogenesis machinery.