The Escherichia coli cytochrome c maturation (Ccm) system does not detectably attach heme to single cysteine variants of an apocytochrome c.

Cytochromes c are typically characterized by the covalent attachment of heme to polypeptide through two thioether bonds with the cysteine residues of a Cys-Xaa-Xaa-Cys-His peptide motif. In many Gram-negative bacteria, the heme is attached to the polypeptide by the periplasmically functioning cytochrome c maturation (Ccm) proteins. Exceptionally, Hydrogenobacter thermophilus cytochrome c(552), which has a normal CXXCH heme-binding motif, and variants with AXXCH, CXXAH, and AXXAH motifs, can be expressed as stable holocytochromes in the cytoplasm of Escherichia coli. By targeting these proteins to the periplasm using a signal peptide, with or without co-expression of the Ccm proteins, we have assessed the ability of the Ccm system to attach heme to proteins with no, one, or two cysteine residues in the heme-binding motif. Only the wild-type protein, with two cysteines, was effectively processed and thus accumulated in the periplasm as a holocytochrome. This is strong evidence for disulfide bond formation involving the two cysteine residues of apocytochrome c as an intermediate in Ccm-type Gram-negative bacterial cytochrome c biogenesis and/or that only a pair of cysteines can be recognized by the heme attachment apparatus.

Cytochromes c are typically characterized by the covalent attachment of heme to polypeptide through two thioether bonds with the cysteine residues of a Cys-Xaa-Xaa-Cys-His peptide motif (1,2). They are ubiquitous in nature; their primary function is in electron transfer, but they may also be found at the active sites of enzymes, and in higher cells they play a role in apoptosis (3)(4)(5)(6)(7)(8). Despite the essential roles of these proteins, it is not clear either how or why the heme becomes covalently attached to the polypeptide; it is not covalently bound in most hemoproteins (9). Remarkably, three different c-type cytochrome biogenesis systems have been identified to date (1,2,10). Gram-negative bacteria use either the cytochrome c maturation (Ccm) 1 proteins (also called system I) or a distinct set of proteins known as system II (10). The Ccm proteins (in Escherichia coli, CcmABCDEFGH; Ref. 11) are located in the periplasm and/or cytoplasmic membrane; all bacterial cytochromes c are either periplasmic or on the periplasmic face of this membrane (1,2,10,12).
It is commonly assumed, but it has not been proven, that formation of a disulfide bond between the two cysteine residues of apocytochrome c is an intermediate in bacterial cytochrome c biogenesis by the Ccm system. The cysteine thiols are thought to be oxidized to a disulfide by DsbA, one of the proteins of the Dsb disulfide isomerase system, which actively forms disulfide bonds in the periplasm (2,13). We have recently shown that a bacterial apocytochrome c can spontaneously form an intramolecular disulfide bond in oxidizing conditions even in the absence of DsbA (14). It would be necessary for any intramolecular disulfide in apocytochrome c to be reduced before the thiols can react with the vinyl groups of heme. CcmG and CcmH both contain the thioredoxin motif (CXXC) of disulfide reductases (13); at least one of them has been argued to reduce the cysteines of apocytochrome c before heme attachment, ultimately by transferring electrons from DsbD (also known as DipZ) (13,15). DsbD receives electrons from the cytoplasmic TrxA and transfers them to the periplasm. Mutants of E. coli deficient in DsbA, DsbB (the oxidant for DsbA), DsbD, or TrxA were all unable to synthesize c-type cytochromes (16 -18). However, it is not clear whether this was because of the inability of these mutants to process a disulfide involving apocytochrome c, one or more of the Ccm proteins, or indeed a combination of these proteins. The importance of investigating whether the substrate, i.e. cytochrome c, or the biogenesis apparatus requires disulfide bonds is emphasized by work on the type II secretion system in Gram-negative bacteria (19). The requirement for DsbA was initially taken to mean that a secreted protein needed a disulfide bond; later it was shown that the disulfide was required in the biogenesis machinery and that a modified substrate could be secreted with a single cysteine thiol instead of a disulfide between two cysteines (19).
It is important for understanding the operation of the Ccm system to determine whether it can assemble cytochromes c in which the heme is attached to the apoprotein by only one thioether bond. Among other things, this would provide strong evidence as to whether an intramolecular disulfide bond in an apocytochrome c is indeed an intermediate. Ríos-Velá zquez et al. (20) made single and double cysteine to alanine substitutions in the heme-binding motif of Rhodobacter sphaeroides cytochrome c 2 . Such variant proteins were unable to support growth via photosynthesis where functional cytochrome c 2 is required. However, in their experiments, no product cytochromes with these substitutions, either holo or apo forms, could be isolated. This indicates that the proteins produced were unstable and rapidly degraded in vivo; thus it is not possible to say definitely whether or not the heme was covalently attached to the polypeptides by the Ccm proteins of R. sphaeroides before degradation occurred. Sambongi et al. (21) conducted similar experiments using heme-binding motif variants of Paracoccus denitrificans cytochrome c 550 , but again the products were unstable. Unusually heme may be covalently (and correctly) attached to cytochrome c 552 from Hydrogenobacter thermophilus in the cytoplasm of E. coli apparently without the action of any specialized biosynthesis proteins (22)(23)(24). Replacement of one or both of the heme-binding cysteine residues by alanines (C11A, C14A, and C11A/C14A) also results in formation of stable cytochromes (25,26). In the two former cases, the heme is covalently attached to the polypeptide through a single thioether bond; in the latter the product is a b-type cytochrome whose heme is noncovalently bound to the polypeptide. Therefore, we have used H. thermophilus cytochrome c 552 and the C11A, C14A, and C11A/C14A variants to investigate the functioning of the Ccm system. Each of the proteins has been targeted to the periplasm of E. coli where the Ccm proteins act and expressed with or without plasmid-borne ccm genes.

EXPERIMENTAL PROCEDURES
E. coli strain JCB387 (27) was transformed with an appropriate plasmid encoding for H. thermophilus cytochrome c 552 , C11A, C14A, or C11A/C14A variants. Two forms of these plasmids were used; plasmids pKHC12 (22), pEST201, pEST202, and pEST203 (25,26) with no periplasmic signal sequence, resulting in cytoplasmic expression of the cytochromes, have been described previously. New plasmids were constructed with the signal sequence of Pseudomonas aeruginosa cytochrome c 551 to target the apocytochromes to the periplasm. PCR was used to amplify and fuse the signal sequence gene to the cytochrome gene on plasmids pKHC12, pEST201, pEST202, and pEST203 using the same method as Zhang et al. (28). pEST210, which carries wild-type cytochrome c 552 and the signal sequence, was the kind gift of Dr. Y. Sambongi. Each of these new plasmids was sequenced to ensure that the signal sequence was present and that there were no secondary mutations in the cytochrome gene. Cells could also be co-transformed with pEC86 (29), which carries the E. coli cytochrome c maturation genes ccmABCDEFGH. E. coli cytochrome b 562 with a periplasmic targeting sequence was produced using a plasmid described previously (30); b 562 cell cultures were grown identically to those transformed with the other plasmids, and the plasmid-borne ccm genes were co-transformed as required.
Cells were initially grown on LB-agar plates with the appropriate antibiotics (100 g ml Ϫ1 ampicillin in each case plus 34 g ml Ϫ1 chloramphenicol where pEC86 was co-transformed). Single colonies were picked into 500-ml 2ϫ TY medium (16 g liter Ϫ1 peptone, 10 g liter Ϫ1 yeast extract, 5 g liter Ϫ1 NaCl) supplemented with 1 mM isopropyl-1thio-␤-D-galactopyranoside in 3-liter flasks. Cultures were grown at 37°C with shaking at 200 rpm for 20 -24 h before harvesting. Cell pellets were washed with cold 10 mM Tris-HCl, pH 7.3, 150 mM NaCl. The cells were then centrifuged again, and the cell pellets were weighed and resuspended in 15 ml of SET buffer (0.5 M sucrose, 200 mM Tris-HCl, pH 7.3, 1 mM EDTA). 15 mg of lysozyme was added, immediately followed by 15 ml of ice-cold water to administer a mild osmotic shock. This mixture was incubated at 37°C for 30 min and then centrifuged at 9000 ϫ g for 15 min. The supernatant was retained as the periplasmic fraction. The pellet was resuspended in 15 ml of water, sonicated vigorously on ice, and centrifuged at 39,000 ϫ g for 45 min. The supernatant at this stage was taken to be the cytoplasmic fraction.
Cytochrome content was determined by recording absorption spectra of the crude periplasmic and cytoplasmic fractions to which a few grains of solid disodium dithionite had been added. Concentrations (and hence the quantity of cytochrome per gram of wet cells) were determined using the characteristic absorbance wavelengths and extinction coefficients described below for reduced H. thermophilus cytochrome c 552 and variants. Note, however, that is was necessary to correct absorbance measurements of cell extracts because of the presence of endogenous E. coli cytochromes (see "Results"). For the wild-type H. thermophilus cytochrome c 552 , ⑀ ϭ 182 or 27.7 mM Ϫ1 cm Ϫ1 at 417 or 552 nm, respectively; for the C11A variant, ⑀ ϭ 179.5 or 27.7 mM Ϫ1 cm Ϫ1 at 422 or 557 nm, respectively; for the C14A variant, ⑀ ϭ 174.5 or 29.8 mM Ϫ1 cm Ϫ1 at 420 or 556 nm, respectively; for the C11A/C14A variant, ⑀ ϭ 145 or 27.3 mM Ϫ1 cm Ϫ1 at 425 or 560 nm, respectively. The extinction coefficients for the wild-type and single cysteine cytochromes were determined by recording absorption spectra of highly purified protein whose absolute concentration had been determined by total amino acid analysis. The extinction coefficients for the C11A/C14A protein were determined from pyridine hemochrome assays (31). Activity staining of SDS-polyacrylamide gels for covalently bound heme was conducted using the method of Goodhew et al. (32).
Since all the cytochrome-encoding plasmids conferred ampicillin resistance, cell fractions were assayed for periplasmic proteins using ␤-lactamase activity (33). To 980 l of 50 mM potassium phosphate buffer, pH 7.0 in a quartz cuvette, 10 l of 100 mg ml Ϫ1 ampicillin and 10 l of cell extract were added. Absorbance change at 244 nm was recorded; note that it was necessary (and appropriate) to directly compare only the activity, normalized for volume, observed for the periplasmic and cytoplasmic fractions from a particular culture fractionation. Cytoplasmic proteins were assayed using malate dehydrogenase activity (34,35). To 970 l of 50 mM potassium phosphate buffer, pH 7.0, 10 l of 25 mM ␤-NADH and 10 l of cell extract were added. The reaction was initiated by the addition of 10 l of 20 mM oxaloacetate as substrate. Absorbance decrease at 340 nm was monitored to relate the activity observed for the volume-normalized periplasmic and cytoplasmic fractions from a particular culture.

RESULTS
E. coli strain JCB387 was transformed with various plasmids to assess the ability of the E. coli cytochrome c maturation proteins to process wild-type H. thermophilus cytochrome c 552 and C11A, C14A, and C11A/C14A variants; these have CXXCH, AXXCH, CXXAH, and AXXAH heme-binding motifs, respectively. Each of these four proteins was available with or without a periplasmic targeting sequence. The latter enables translocation of the polypeptide to the periplasm where the Ccm system operates. Thus, in our investigations, 16 combinations were used: cytochrome c 552 , C11A, C14A, and C11A/ C14A Ϯ a periplasmic signal sequence, each Ϯ the plasmidborne ccm genes. E. coli strain JCB387 (27) was chosen because preliminary experiments indicated that it produced only moderate yields of each of the four H. thermophilus cytochromes cytoplasmically (i.e. in the absence of the targeting sequence the cytochrome production was detectable but not such as to swamp any periplasmic cytochromes that may be produced with a targeting sequence). Strain JCB387 also produced a poor yield of periplasmically targeted wild-type cytochrome c 552 in the absence of the co-transformed ccm plasmid. Cells were grown aerobically, so that expression of the endogenous Ccm system of E. coli would be minimized, and were fractionated into periplasmic and cytoplasmic components; each fraction was assayed for contamination by the other using enzyme marker assays. The fractionation protocol was carefully optimized to result in minimal contamination of the periplasmic fraction with cytoplasmic proteins and vice versa (see "Experimental Procedures"). Quantities of cytochrome produced were determined spectrophotometrically.
In the absence of a periplasmic signal sequence, effectively 100% of each of the cytochromes was made cytoplasmically (Table I). 2 These data were as anticipated and serve both as a control experiment and as a reference against which expression with a periplasmic signal sequence can be assessed. In the presence of the signal sequence, of cytochrome c 552 and each of the three variants studied, only the wild-type holoprotein was found in significant quantities in the periplasm. With the signal sequence and the ccm plasmid present, essentially 100% of the wild-type cytochrome was periplasmic (Table II). In contrast, the variant cytochromes were found in the periplasm only in small amounts even when the Ccm proteins were coexpressed (in each case less than 5% of the total cytochrome after subtraction for cytoplasmic contamination and endogenous cytochrome production by E. coli (see below)). In the presence of the periplasmic targeting sequence, with or without the ccm plasmid, there was no significant quantitative difference in the total expression level or periplasmic quantities of the AXXAH, CXXAH, or AXXCH proteins (Table II). Small quantities of variant proteins in the periplasm can be accounted for by self-assembly; if protein with a signal sequence is translocated to the periplasm and encounters heme, it can be expected to spontaneously form a cytochrome as it does in the cytoplasm.
Within experimental error, targeting the apocytochromes to the periplasm caused a decrease in total holocytochrome production in every case except for the wild-type cytochrome c 552 when co-expressed with the Ccm proteins (compare data in Table II with data in Table I). Moreover, the results in Table II (and described above) show that the vast majority of the periplasmically targeted holocytochromes with substitutions in the heme-binding motif were made cytoplasmically despite the presence of the signal sequence. Staining SDS-polyacrylamide gels for covalently bound heme for such periplasmically targeted c 552 variants resulted in a band at a higher molecular weight than for cytoplasmically produced protein with no targeting sequence, reflecting the presence of both heme and the targeting sequence in the higher molecular weight material. It might be anticipated that the availability of the periplasmic signal sequence would cause rapid translocation of the apoprotein before heme attachment could occur in the cytoplasm. However, our data indicate that there is competition between the rate of heme binding to apoprotein in the cytoplasm and the rate of apoprotein translocation to the periplasm by the general type II secretion (Sec) proteins. Translocation may be slowed because H. thermophilus apocytochromes c 552 are quite structured (36), and the Sec system transports unfolded proteins (37).
The observation that the periplasmically targeted double alanine (C11A/C14A) variant protein, which forms a b-type cytochrome (25), was not made in significant quantities in the periplasm, either with or without the Ccm system, is an apparent anomaly. In our experimental conditions, E. coli cytochrome b 562 with a periplasmic signal sequence was made effectively 100% periplasmically with or without co-expression of the ccm plasmid (Table II). It is well established that disruption of the Ccm proteins does not inhibit b 562 formation in the periplasm of E. coli (38,39); hence we anticipated that C11A/C14A c 552 would be made equally well. One plausible explanation for our observations is that, when in the periplasm, the apo form of the C11A/C14A protein, which is not a naturally occurring b-type cytochrome, acquires its heme less quickly than and/or is more susceptible to proteolysis than apocytochrome b 562 . This would not affect the wild-type cytochrome c 552 , which can rapidly acquire heme from the Ccm system. The means by which b-type cytochromes acquire heme in the bacterial periplasm, a compartment of the cell in which they are rare, is far from understood, so the specific nature of the apocytochrome may, therefore, also be important for the process.
When wild-type cytochrome c 552 was expressed with the periplasmic signal sequence but no ccm plasmid, a significant fraction of the c 552 was found in the periplasmic fraction (on  552 and variants with altered heme-binding motifs expressed from genes lacking signal sequences These control data are the averages of two or three experiments. Ccm ϩ indicates that the plasmid-borne cytochrome c maturation genes were co-transformed with the plasmid carrying the cytochrome gene. The periplasmic and cytoplasmic marker proteins were ␤-lactamase and malate dehydrogenase, respectively. The percentage of the total cytochrome in the periplasm and the yield of total cytochrome per gram of wet cells are corrected from the raw measurements by subtracting a constant background level of endogenous E. coli cytochromes (see "Results"). expressed from genes with signal sequences Ccm ϩ indicates that the plasmid-borne cytochrome c maturation genes were co-transformed with the plasmid carrying the cytochrome gene. Data are presented as the mean of multiple experiments; standard deviations are shown in parentheses. The periplasmic and cytoplasmic marker proteins were ␤-lactamase and malate dehydrogenase, respectively. The percentage of the total cytochrome in the periplasm and the yield of total cytochrome per gram of wet cells are corrected from the raw measurements by subtracting a constant background level of endogenous E. coli cytochromes (see "Results"). cyt., cytochrome. average ϳ60%). This suggests that the endogenous E. coli Ccm system, which is maximally expressed under anaerobic conditions (40), was being expressed to some extent during growth (i.e. our cultures were not fully aerobic). However, even with this level of expression of periplasmic cytochrome c 552 , the yield in mg of periplasmic cytochrome produced per g of wet cells was only ϳ14% of that produced when the ccm plasmid was present; thus this effect is relatively minor. As a corollary, wild-type c 552 with a periplasmic signal sequence but no ccm plasmid produced ϳ40% cytoplasmic cytochrome c, whereas in the equivalent case with the ccm plasmid, the latter value was ϳ0%. Furthermore, any low level expression of the cells' own Ccm proteins has not affected our data for the cytochromes with substitutions in the heme-binding motif since the introduction of the ccm plasmid (which implies a much higher level of expression) makes no significant quantitative difference to the periplasmic cytochrome expression levels of these variant proteins.
To determine background absorbances in our measurements, we assessed the level of endogenous cytochrome production by E. coli strain JCB387 not transformed with any of our plasmids. In the periplasmic fractions of such cells, we detected a c-type cytochrome ( max , 551 and 419 nm), probably NapB, a soluble subunit of the periplasmic nitrate reductase (41). In the cytoplasm we detected absorbance from one or more hemoproteins (observed max , ϳ560 and ϳ423 nm), e.g. catalase. The periplasmic cytochrome accounted for 33% of the total absorbance, and the cytoplasmic cytochrome(s) accounted for 67% with the periplasmic and cytoplasmic fractions normalized for volume. These proportions were essentially the same (32 and 68%) if E. coli JCB387 was transformed with the ccm plasmid pEC86 but no exogenous cytochrome plasmid. One might expect that expression of the Ccm proteins from the plasmid would stimulate production of the endogenous periplasmic c-type cytochrome. However, the nap and ccm operons are co-regulated in E. coli (11,42), so in a given set of growth conditions, the expression of NapB would be limited by the same factor(s) with or without co-expression of the Ccm proteins from pEC86. We have corrected for these background hemoprotein absorbances in Tables I and II assuming that the endogenous cytochromes were produced at the same levels per gram of wet cells in all of our growth experiments; the effect is to reinforce the conclusions drawn in the paragraphs above. The percentages of each of the variant cytochromes (AXXCH, CXXAH, or AXXAH heme-binding motifs) in our periplasmic fractions are lowered because some of the absorbance attributed to these proteins in the raw measurements was in fact due to endogenous cytochrome(s).

DISCUSSION
The principal result of this study is that the type I cytochrome c maturation system as found in many Gram-negative bacteria can only effectively process covalent heme attachment to apocytochrome c with two cysteine residues in the hemebinding motif. The nature of our experiments is such that we cannot say with certainty that the activity of the Ccm system toward the substrate apoproteins with single or no cysteine heme-binding motifs is 0%, rather than Ͻ2%, of that with the double cysteine apocytochrome; nevertheless, the latter value is a reasonable estimate of the upper limit of any such activity. The implication is that an intramolecular disulfide bond within apocytochrome c is, as postulated on the basis of less direct evidence, an intermediate in this type of cytochrome c biogenesis. This idea finds support in the observation that H. thermophilus apocytochrome c 552 forms an intramolecular disulfide bond under oxidizing conditions following removal of the heme in vitro (14). An alternative explanation for the data presented in the present work is that that both apocytochrome cysteine thiols play specific, and required, roles in the maturation pathway, possibly involving intermolecular disulfide bonds. A further interpretation, which may or may not be combined with the obligate formation of one or more disulfide bonds, is that the ultimate highly specific recognition determinant for heme attachment by the Ccm system is the two cysteine thiols in the apocytochrome heme-binding motif. Note, however, that any rationalization of our data must allow for the fact that the Ccm proteins are active with substrate apocytochromes that either have naturally (43), or have as the result of amino acid substitutions (20), CXXXXCH or CXXXCH heme binding-motifs, and thus it is the two cysteines that are important rather than their precise spatial arrangement.
It is very possible that the CXXCH motif of apocytochrome c cannot avoid being oxidized to a disulfide in the bacterial periplasm, e.g. by the active oxidant DsbA, and thus that the Ccm system has evolved to handle such oxidized apoproteins. Indeed it has been suggested (13) that formation of a disulfide in the apocytochrome may pre-fold the polypeptide and that this facilitates covalent heme attachment. The likelihood that a disulfide bond is an intermediate in holocytochrome c formation also implies that covalent heme attachment, or formation of a mixed disulfide between one of the apocytochrome cysteines and a cysteine from one of the Ccm proteins, is concerted with reduction of the apocytochrome disulfide. Otherwise the reduced disulfide is susceptible to rapid reoxidation by, for example, DsbA.
The crucial difference between our failure to observe covalent attachment of heme to apocytochromes with a CXXAH or AXXCH heme-binding motif and previous related studies (20,21) is that in the earlier cases the products were unstable with respect to degradation. Thus is was not possible to determine unequivocally whether heme was in fact covalently attached by the bacterial Ccm system before the protein degraded. Moreover, there is no evidence that single cysteine proteins with covalently bound heme could ever actually form from the variant apocytochromes tested in the earlier studies. In the present work, we have used H. thermophilus cytochrome c 552 together with mutants that are known to form holocytochromes and to be stable when expressed in the cytoplasm of E. coli (Table I and Refs. 25 and 26); the wild-type cytochrome is clearly also stable in the periplasm (Table II). Thus, our inability to observe variant holocytochromes in significant quantities in the periplasm is convincing evidence that the Ccm system cannot covalently attach heme to apocytochromes that do not have two cysteine residues in the heme-binding motif. Previous studies have shown that apocytochrome c with a periplasmic signal sequence, even with substitutions in the heme-binding motif, is translocated into the periplasm independent of the heme attachment process (21). Single cysteine-attached cytochromes (XXXCH motif) have been isolated from some eukaryotic sources (44 -47). In two cases (46,47) the formation of these cytochromes was catalyzed by a specific enzyme, cytochrome c heme lyase from yeast mitochondria, but the yields were low relative to protein with a CXXCH heme-binding motif. It may be that other mitochondrial heme lyases have evolved where necessary to cope with only having one thiol for heme attachment. No single cysteine-attached cytochromes c have been observed to date in bacteria, which is consistent with the inability of the Ccm system, one of two known bacterial cytochrome c biogenesis systems (1,2,10), to process them. Parts of the Ccm system appear to function in mitochondria from at least some plants and from protists such as Reclinomonas americana (48). If any eukaryote were found to have both the Ccm system and single cysteine attachment of heme in a cytochrome c, then the present work would imply that the