The thioreduction component CcmG confers efficiency and the heme ligation component CcmH ensures stereo-specificity during cytochrome c maturation

In many Gram-negative bacteria, including Rhodobacter capsulatus, cytochrome c maturation (Ccm) is carried out by a membrane-integral machinery composed of nine proteins (CcmA to I). During this process, the periplasmic thiol-disulfide oxidoreductase DsbA is thought to catalyze the formation of a disulfide bond between the Cys residues at the apocytochrome c heme-binding site (CXXCH). Subsequently, a Ccm-specific thioreductive pathway involving CcmG and CcmH reduces this disulfide bond to allow covalent heme ligation. Currently, the sequence of thioredox reactions occurring between these components and apocytochrome c and the identity of their active Cys residues are unknown. In this work, we first investigated protein–protein interactions among the apocytochrome c, CcmG, and the heme-ligation components CcmF, CcmH, and CcmI. We found that they all interact with each other, forming a CcmFGHI–apocytochrome c complex. Using purified wild-type CcmG, CcmH, and apocytochrome c, as well as their respective Cys mutant variants, we determined the rates of thiol-disulfide exchange reactions between selected pairs of Cys residues from these proteins. We established that CcmG can efficiently reduce the disulfide bond of apocytochrome c and also resolve a mixed disulfide bond formed between apocytochrome c and CcmH. We further show that Cys-45 of CcmH and Cys-34 of apocytochrome c are most likely to form this mixed disulfide bond, which is consistent with the stereo-specificity of the heme–apocytochrome c ligation reaction. We conclude that CcmG confers efficiency, and CcmH ensures stereo-specificity during Ccm and present a comprehensive model for thioreduction reactions that lead to heme–apocytochrome c ligation.

CcmH has a single C-terminal TM helix and a periplasmic domain containing a conserved LRCXXCQ active site. The 3D structure of CcmH differs from the canonical thioredoxin fold showing a three-helical bundle structure and has its N-terminal Cys residue buried while its C-terminal Cys is solvent-exposed ( Fig. 1A) (19 -21).
Earlier genetic studies showed that in the absence of thiooxidation by DsbA, thioreduction involving CcdA (22) or CcmG (14) is not required. In contrast, this thioredox compensation is not observed in mutants lacking both CcmH and DsbA (14), suggesting that CcmH plays another role in addition to thioreduction of the disulfide bond at the HBS of apocyts c. Indeed, CcmH together with CcmI and CcmF form the heme ligation complex CcmFHI (23).
Although the involvement of both CcmG and CcmH in thioreduction of the apocyts c is established, the sequence of the reactions between these components and their active Cys residues remain unknown. Early experiments using purified R. capsulatus CcmG and CcmH suggested a linear thiol-disulfide cascade based on the ability of CcmH to oxidize CcmG and to reduce a short peptide mimicking apocyt c HBS (Fig. 1B) (13, 21, 24 -27). Consistent with this model were the observations that the Arabidopsis thaliana CcmH homologue is able to reduce a peptide mimicking the apocyt c HBS (13,26), and P. aeruginosa CcmH interacts with a similar peptide at low (micromolar range) affinity (19). The unusual fold and biochemical properties of P. aeruginosa CcmH, together with the inability of CcmG to reduce the disulfide bond of CcmH, led to a different proposal in which CcmG is responsible for resolving a CcmH-apocyt c mixed disulfide formed during Ccm (Fig. 1C) (12,19,25,27).
The establishment of the order of thioreduction reactions between CcmG, CcmH, and apocyt c is essential for elucidating the mechanism of heme ligation. In this work, we first investigated protein-protein interactions between CcmG, CcmH, and class I apocyts c (using apocyt c 1 or c 2 as model substrates) via co-purification assays, using native and Cys-less variants of purified proteins, as well as detergent-dispersed membrane fractions. We found that CcmG binds tightly to CcmH (and the other heme ligation components) and to apocyt c 1 , but with lower affinity (micromolar range), forming a CcmFGHIapocyt c complex. Then, using purified mutant proteins we determined the rate constants of thiol-disulfide exchange reactions between selected Cys residues from CcmG, CcmH, and apocyt c 1 . Based on these rate constants and the interactions between these proteins, we propose a model for apocyt c thioreduction, suggesting that CcmG is required for efficient cyt c maturation and CcmH for stereo-specific heme-apocyt c ligation during Ccm.

Purification of His 6 -CcmG WT , FLAG-CcmH WT , and Strep-apocyt c 1 WT and their derivatives, and protein-protein interactions between apocyt c 1 and other Ccm components
CcmG and CcmH are Ccm-specific components that are responsible for thioreduction of the disulfide bond at the HBS of apocyts c. Currently, neither the order of the thiol-disulfide exchange reactions that occur during this process between the three components nor the identity of their active Cys residues that participate in these reactions are well defined. To address this issue, we overproduced in the cytoplasm of E. coli, and then purified by affinity chromatography (Ͼ95% purity) soluble derivatives of native (WT) (Fig. 2) as well as single and double Cys (*) mutant variants of His 6 -CcmG (CcmG WT , CcmG Cys-75 , CcmG Cys-78 , and CcmG*), FLAG-CcmH (CcmH WT , CcmH Cys-42 , CcmH Cys-45 , and CcmH*), and Strep-apocyt c 1 (apocyt c 1 WT, apocyt c 1 Cys-34 , apocyt c 1 Cys-37 , and apocyt c 1 *). Fig. 2 shows as an example the purified wild-type derivatives of His 6 -CcmG, FLAG-CcmH, and Strep-apocyt c 1. The Cys mutant derivatives of these Figure 1. R. capsulatus Ccm system I and earlier proposed models for thioreduction of apocyt c HBS disulfide. A, nine membrane integral proteins (CcmABCDEFGHI) with different functions are responsible for covalent heme ligation to the apocyts to produce mature c-type cyts. Apocyts c are translocated via the SEC system to the periplasm, where the Cys residues at their HBS are oxidized by the DsbA-DsbB system (thio-oxidation). CcdA receives electrons from the cytoplasmic thioredoxin TrxA and reduces CcmG. The thiol-disulfide oxidoreductases CcmG and CcmH reduce the intramolecular disulfide bond at the apocyt HBS to allow heme ligation (thioreduction). CcmH together with CcmF and CcmI form the heme ligation core, whereas CcmABCD is an ATP-dependent ABC-type transporter that loads heme to CcmE to produce holo-CcmE. CcmI traps the C termini of the apocyt c substrates, whereas the heme delivered by holo-CcmE is covalently ligated to the apocyts c heme-binding sites by the CcmFHI core complex. B, CcmG-CcmH-apocyt c linear cascade of thiol-disulfide exchange. This proposal suggests that reduced CcmH (recycled by CcmG and CcdA) reduces directly apocyt c HBS disulfide bond (13, 24 -27). C, CcmG reduces a mixed disulfide between CcmH and apocyt c. This proposal suggests that CcmG reduces a mixed disulfide formed between CcmH and apocyt c instead of oxidized CcmH or apocyt c HBS disulfide (12,19,25,27).

Thioreduction branch of the Ccm pathway
proteins were also purified to the same degree of homogeneity (supplemental Fig. S1).
In thioredoxins and thiol-disulfide oxidoreductases, substrate recognition relies mainly on non-covalent electrostatic and hydrophobic interactions as well as hydrogen bonding within the substrate-enzyme complex (28). Given that CcmG is a Ccm-specific thioredoxin, we investigated its interactions with apocyt c and other Ccm components, in particular CcmH, via co-purification assays using purified CcmG*, apocyt c 1 *, and CcmH* (Fig. 3). We chose the Cys-less variants of these proteins to avoid increased complexity that could emerge from inter-molecular disulfide bond formation during these assays. The data showed that FLAG-CcmH* co-purified with His 6 -CcmG* using a nickel-Sepharose HP resin (anti-His), indicating that they interact strongly with each other in vitro despite the absence of their Cys residues (Fig. 3A, right panel). As a control, we showed that in the absence of His 6 -CcmG*, FLAG-CcmH* was not retained by the anti-His resin (Fig. 3A, left panel). Next, anionic exchange chromatography (Q-Sepharose) was carried out using n-dodecyl ␤-D-maltoside (DDM)-dispersed membranes from R. capsulatus MTSRP1.r1 (a strain lacking CcmI but overproducing CcmF and CcmH) ( Table 1). The fraction eluted at 150 mM NaCl contained many proteins (SDS-PAGE data not shown), but it was highly enriched in CcmF, CcmH, and CcmG, as evidenced using appropriate specific antibodies, as done before (Fig. 3B) (23). The presence of CcmG in this fraction was detected using the newly produced rabbit polyclonal antibodies (see under "Experimental procedures"). The specificity of anti-CcmG antibodies was confirmed using purified His 6 -CcmG* and appropriate R. capsulatus wild-type and mutant strains (supplemental Fig. S2). The fraction eluted at 150 mM NaCl was incubated with purified His 10 -CcmI WT and loaded into a nickel-Sepharose resin, as done previously (29).
Analysis of the elution fraction using anti-CcmG antibodies revealed that CcmG WT was co-purified with CcmI WT (Fig. 3C). Thus, CcmG interacted with His 10 -CcmI either directly or indirectly via CcmH (Fig. 3A). CcmI is known to interact strongly with both CcmH and CcmF (23), and the data indicated that CcmG might also be associated with the heme ligation complex CcmFHI.  Figure 3. Protein-protein interactions between the apocyt c, thioredoxin CcmG, and heme ligation components CcmF, CcmH, and CcmI. A, co-purification of Cys-less FLAG-CcmH* with Cys-less His 6 -CcmG* using nickel-Sepharose resin is shown on the right panel. Note that FLAG-CcmH* does not bind to the resin in the absence of His 6 -CcmG* (left panel). In all panels, FT and E refer to flow-through and elution fractions, respectively. B, DDM-dispersed membrane proteins from R. capsulatus strain MTSRP1.r1 were separated using a Q-Sepharose column, and a fraction containing CcmF, CcmG, and CcmH co-eluting together was collected. 40 g of total membrane proteins before fractionation (left lane) and 10 g of the fraction containing CcmF, CcmG, and CcmH together (right lane) are shown. All proteins were detected by immunoblots using appropriate specific antibodies as indicated. C, co-purification of native CcmG from the CcmFGH-enriched fraction with purified CcmI. 10 g of purified His 10 -CcmI was incubated with 100 g of CcmFGH-containing fraction and re-purified in a nickel-Sepharose resin. Only in the presence of His 10 -CcmI is CcmG retained by the resin and found in the elution fraction (right lane). D, co-purification of equimolar concentration (ϳ1.5 M) of His 6 -CcmG WT with Strep-apocyt c 2 WT using Strep-Tactin-Sepharose resin (middle lane). In the absence of apocyt c 2 , CcmG was not retained by the resin (left lane), and in the presence of His 10 -CcmI (ϳ0.5 M) more His 6 -CcmG co-eluted with Strep-apocyt c 2 (right lane). C and D, CcmG was detected by immunoblots using anti-CcmG-specific antibodies. E, 500 g of DDM-dispersed membrane proteins from R. capsulatus strain MTSRP1.r1/ pNJ2, containing CcmF, CcmH, and FLAG-CcmI were incubated with 10 g of Strep-apocyt c 2 WT , which was then re-purified using Strep-Tactin-Sepharose resin. Binding to the Strep-Tactin column of the heme ligation components CcmF, CcmH, and FLAG-CcmI occurs only when apocyt c 2 WT is present (compare left and right lanes). Immunodetection was done with anti-CcmF, anti-CcmH, and anti-FLAG (for FLAG-CcmI detection) polyclonal antibodies, as appropriate. F, schematic representation of a hypothetical CcmFGHI-apocyt c complex that might occur during Ccm. Our data showing that apocyt c 2 interacts with CcmG, CcmF, CcmH, and CcmI, together with CcmG interacting with the heme ligation core components CcmH, CcmI, and CcmF in the absence of apocyt c 2 , support the occurrence of such a multisubunit complex binding the apocyt substrates.

Thioreduction branch of the Ccm pathway
Despite many attempts, co-purification of apocyt c 1 * with either CcmG* or CcmH* was not observed. Because the use of apocyt c 1 WT for these assays was not suitable due to its high tendency to dimerize in the presence of oxygen, we used biolayer interferometry to study these interactions. Real-time binding kinetics between purified CcmG WT and apocyt c 1 WT were determined, as done earlier (30). The association (k on of 9.97 Ϯ 0.11 ϫ 10 2 M Ϫ1 s Ϫ1 ) and dissociation (k off of 7.31 Ϯ 1.16 ϫ 10 Ϫ3 s Ϫ1 ) rates thus determined yielded a K D value of 7.2 Ϯ 1.8 M using a 1:1 homogeneous kinetic model describing bimolecular interactions (30). Similarly, when native apocyt c 2 WT (another class I cyt c, known to interact with CcmI and CcmE (29 -31)) was used instead of apocyt c 1 WT , its co-purification with CcmG WT was readily seen using specific antibodies (Fig. 3D). Furthermore, the amount of CcmG WT that co-purified with apocyt c 2 WT was higher upon addition of purified CcmI, a specific apocyt c chaperone, suggesting that the interactions between CcmG and apocyts c also involved additional Ccm partners (29). Finally, upon incubation with detergentdispersed membrane fractions from R. capsulatus strain MTSRP1.r1 complemented with plasmid pNJ2 carrying FLAGtagged CcmI (Table 1), native apocyt c 2 WT co-purified with not only FLAG-CcmI but also CcmF and CcmH (Fig. 3E), indicating that besides interacting with CcmG (above), apocyt c 2 also interacts with CcmI (29), CcmF, and CcmH, forming a CcmF-GHI-apocyt c complex (Fig. 3F). The need for the HBS Cys residues of apocyt c, similar to what was seen here with CcmG or CcmH and apocyt c 1 , has also been observed for the interactions of apocyt c 2 and apoCcmE (31). Our earlier data, showing a ternary complex composed by apocyt c 2 , CcmI, and apoCcmE (31), together with the data presented here, are consistent with the existence of a multisubunit maturase supercomplex, composed of at least the CcmEFGHI-apocyt c components, as proposed previously (2).

Thioreduction branch of the Ccm pathway
quantitate the thiol groups in proteins, as it readily forms mixed disulfide bonds with accessible thiols, and the TNB 2Ϫ ions released during this reaction can be easily monitored by visible spectroscopy at 412 nm (12,32,33). The protein-Cys-S-S-TNB (protein-TNB) adducts formed are good proxies for inter-molecular mixed disulfide bonds between a given protein and another Cys-containing partner protein (12,32,33). The faster a specific thiolate carries out a nucleophilic attack to the TNB-conjugated Cys residue (i.e. higher rate constant k), the higher is the likelihood of these two Cys residues to engage in thiol-disulfide exchange reactions (Fig. 4A). Purified single Cys mutant variants of CcmH and apocyt c 1 were reacted with DTNB, and their protein-TNB adducts were isolated (see under "Experimental procedures"). Because CcmG is a bona fide thioredoxin (17,22), we surmised that it would preferentially initiate a nucleophilic attack to an existing disulfide bond in oxidized CcmH or apocyt c 1 , or a mixed disulfide between them, hence its TNB adducts were not prepared. The rates of reduction of CcmH-TNB and apocyt c 1 -TNB adducts Cys-34 and apocyt c 1 Cys-37 ) were measured under pseudo-first order kinetics (i.e. excess of reducing partner versus the protein-TNB adduct), and their corresponding bimolecular rate constants were determined ( Table 2). As an example, a set of data showing the release of TNB 2Ϫ ions (increase in A 412 nm ) when 1 M CcmH Cys-45 -TNB reacted with different amounts (up to 30 M) of reduced CcmG Cys-78 are shown in Fig. 4B. In a control experiment when CcmG Cys-78 treated with iodoacetamide (IOA) was used, no signal increase at A 412 nm was observed. For each concentration of reducing protein used (e.g. CcmG Cys-78 ), the corresponding k obs value was determined (see under "Experimental procedures"). Plotting these k obs values against the concentrations of the reducing partner (e.g. CcmG Cys-78 ) yielded the bimolecular rate constant (k, M Ϫ1 s Ϫ1 ) of the thiol-disulfide exchange reaction for this Cys pair (Fig. 4C, e.g. slope of the top line). In similar ways, the k values for all Cys pairs between CcmG, apocyt c 1 , and CcmH were determined (Table 2). We inferred from these k values the likelihood of occurrence of the corresponding thiol-disulfide exchange reactions as follows.

Formation of a mixed disulfide in vitro between CcmG and CcmH
To further substantiate the DTNB-based assays, we also measured the Cys reactivity of the single mutants by testing the formation in vitro of mixed disulfide bonds between the proteins examined. Based on the data presented in Table 2, we chose to react reduced CcmG Cys-75 with CcmH Cys-45 -TNB (in a molar ratio of 2:1) to obtain a mixed disulfide in vitro under the DTNB assay conditions used (see under "Experimental procedures"). After 16 h of incubation at room temperature, the reaction mixture was analyzed using SDS-PAGE under non-reducing conditions. Fig. 5, lane 3, shows a faint band of ϳ33 kDa, containing both CcmG and CcmH, as identified by nLC-MS/MS spectrometry (Fig. 5, right panel). Similar data confirmed that the band of ϳ26 kDa corresponded to the CcmH Cys-45 dimer and the lower bands to CcmG and CcmH monomers ( Fig. 5 and supplemental Table S1). Overall data clearly showed the formation of a CcmG Cys-75 -Cys-45 CcmH mixed disulfide upon incubation of these two proteins. However, compared with the amounts of reduced CcmG Cys-75 (17 kDa, Fig. 5, lane 2) and CcmH Cys-45 -TNB (13.5 kDa, Fig. 5, lane 1) used, the yield of this reaction remained very low despite our many attempts using different buffers and incubation times. A possibility is that the pK a of CcmG Cys-75 is unexpectedly high and that at pH 7.5 only a small amount of thiolate is produced to carry out efficiently a nucleophilic attack to the CcmH Cys-45 -TNB mixed disulfide. Another possibility is that the reactions are not being quenched at low pH, reverse reactions might have occurred under oxic conditions. Recent studies have indicated that the chemistry underlying DTNB-based reactions is complex (34). Other studies have also reported similar very low yields of mixed disulfides formed between other thiol-disulfide oxidoreductases (32,(35)(36)(37). In each case, the data points are average of at least two assays, and the linear curve is the best fit to the data points. The slope of this line represents the bimolecular rate constant (k) of the thiol-disulfide exchange reactions between the indicated Cys residues. Cumulative data obtained with all tested Cys pairs between CcmG, CcmH, and apocyt c 1 are presented in Table 2.

Thioreduction branch of the Ccm pathway Determination of the redox states of CcmG and CcmH in actively growing cells
Information about the steady-state redox states of CcmG and CcmH in actively growing wild-type cells is crucial for correctly attributing specific roles to the above-identified Cys residues during the thioreduction of apocyts c in vivo. For this purpose, we used the thiol-alkylating reagent 4-acetamido-4Ј-maleimidylstilbene-2,2Ј-disulfonic acid (AMS) that reacts covalently with free thiolates in proteins and allows identification of their redox states. Proteins containing AMS-modified Cys residues exhibit migration shifts toward higher molecular weights during SDS-PAGE. Upon treatment of cell extracts with AMS, CcmG was shifted to a higher molecular weight (Fig. 6A, lane 2) compared with untreated samples (Fig. 6A, lane 1) when subjected to SDS-PAGE. This molecular weight shift was identical to that seen when cell cultures were reduced with DTT prior to AMS modification (Fig. 6A, lane 4). Thus, CcmG was mostly in a reduced state. On the contrary, CcmH showed no shift with or without AMS treatment, indicating that it was mainly in oxidized state (Fig. 6B, lanes 1 and 2). A molecular weight increase due to the modification of CcmH thiolates by AMS was seen only when cell cultures were reduced with DTT prior to AMS addition (Fig. 6B, lanes 3 and 4). We therefore concluded that, in actively growing R. capsulatus cells, CcmG and CcmH were mainly in the reduced and oxidized states, respectively, in agreement with earlier reports (13).

Discussion
Under oxic or anoxic growth conditions, and despite the presence of DsbA, which is an efficient periplasmic thiol oxi-dant, cells must keep the thiol groups of the apocyts c HBS Cys residues in the reduced state for heme ligation to occur (9). CcmG and CcmH are Ccm-dedicated thiol-disulfide oxidoreductases that carry out this process, but the mechanisms governing these reactions are not well defined (Fig. 1) (12,13,19,25,27). In this work, we addressed this issue systematically, using wild-type as well as single or double Cys mutant derivatives of CcmG (CcmG WT , CcmG Cys-75 , CcmG Cys-78 , and CcmG*), CcmH (CcmH WT , CcmH Cys-42 , CcmH Cys-45 , and CcmH*), and apocyt c 1 (apocyt c 1 WT, apocyt c 1 Cys-34 , apocyt c 1 Cys-37 , and apocyt c 1 *). First, we examined the proteinprotein interactions between these proteins, and then defined the catalytic abilities of their Cys residues to engage in thioldisulfide exchange reactions.
Protein-protein interaction studies indicated that CcmG and CcmH interacted strongly with each other and that these interactions did not require their Cys residues. Moreover, CcmG WT also interacted, but weakly with the class I apocyts c  Table S1. The ϳ26-kDa band was also identified as CcmH Cys-45 dimer (supplemental Table S1). nd indicates not found, and * indicates oxidized methionine or carbamidomethylated cysteine residues. Immunoblot analysis of cell extracts from appropriately complemented R. capsulatus mutants MD11/pCS1566 and MD14/pST6 (Table 1) using anti-CcmG-and anti-CcmH-specific polyclonal antibodies, respectively. Aliquots from cell cultures were TCA-precipitated with or without prior DTT reduction. Cell pellets were solubilized in buffer containing SDS with or without AMS. AMS alkylates free Cys thiolates, adding to the total protein molecular mass of ϳ0.5 kDa per AMS-modified thiolate. Reduced and AMS-modified proteins show higher molecular weights in SDS-PAGE. The data show that in actively growing cells, CcmG is reduced (A) and CcmH is oxidized (B). NT refers to untreated cell cultures solubilized in buffer without AMS; ϩ AMS indicates untreated cells solubilized in buffer containing AMS; ϩ DTT indicates DTTreduced cell cultures solubilized in buffer without AMS, and ϩDDT/AMS indicates cells first reduced with DTT and then reacted with AMS.

Thioreduction branch of the Ccm pathway
(e.g. approximately micromolar K D for apocyt c 1 WT ), and both the presence of the HBS Cys residues of apocyts c or the chaperone CcmI further enhanced these interactions. Finally, using dispersed membranes, we showed that apocyt c 2 WT interacted with the heme ligation components (CcmF-CcmH-CcmI), and altogether they yielded a CcmFGHI-apocyt c complex (Fig.  3F). Our earlier studies established that CcmI is an apocyt c chaperone that binds tightly the C terminus of class I apocyts (29,30,38) and that CcmG has oxidoreductase and chaperone (holdase) activities to assist the apocyts c and enhance cyt c maturation (14). Similar cooperation of CcmG with the heme ligation complex was reported with an engineered E. coli CcmFGH complex (equivalent of R. capsulatus CcmFGHI) that can carry out Ccm in the absence of CcmABCDE (39). Furthermore, genetic studies showed that the absence of CcmI can be bypassed by overexpression of CcmFH and CcmG and that CcmG is functionally related with CcmI (40). These data are consistent with the occurrence of a CcmFGHIapocyt c complex as seen in this study (Fig. 3). They also support our recent proposal that the entire Ccm machinery might form an even larger supercomplex also including the CcmABCDE complex (2).
Next, to probe the catalytic abilities of the Cys residues of CcmG, CcmH, and apocyt c 1 , the bimolecular rate constants (k) of thiol-disulfide exchange reactions between a protein-TNB adduct and a thiolate of a partner protein were determined ( Table 2). The k values obtained in this work are lower than those observed with the thiol-disulfide oxidoreductases of the thioredoxin family (10 6 to 10 7 M Ϫ1 s Ϫ1 ) (28). However, these rates are higher than those reported for the reactions involving CcmG WT and single Cys mutant derivatives of P. aeruginosa CcmH-TNB adducts (23 and 4.1 M Ϫ1 s Ϫ1 ) (12). We note that P. aeruginosa CcmG and CcmH have similar (Ϫ215 and Ϫ213 mV, respectively) redox midpoint potential (E m ) values, whereas a larger difference exists between those of R. capsulatus CcmG (Ϫ300 mV) and CcmH (Ϫ210 mV) (12,19,24). Furthermore, by using the single Cys mutants of both CcmG and CcmH, we could visualize the occurrence of a mixed disulfide intermediate in vitro, and AMS labeling in vivo indicated that CcmG is mainly reduced, in agreement with its faster rate of reduction by CcdA (10 5 M Ϫ1 s Ϫ1 ) (41) than its oxidation during apocyt c thioreduction (Table 2). Similarly, CcmH was found mainly oxidized in actively growing R. capsulatus cells, which would be expected if it is indeed oxidized by DsbA, which exhibits high rates of oxidation for its substrates (ϳ 10 5 M Ϫ1 s Ϫ1 ) (42).
The high k values observed with either CcmG Cys-75 or CcmG Cys-78 and apocyt c 1 Cys-34 (Table 2) suggested that CcmG can reduce directly and efficiently the disulfide bond at the HBS of apocyt c 1 (its Cys-34 residue being the site of nucleophilic attack). During the thiol-disulfide exchange reactions mediated by thioredoxins, the target disulfide is first attacked by an active thiolate (N-terminal Cys), forming a mixed disulfide bond. Subsequent attack of this disulfide bond by the remaining thiolate (C-terminal Cys) of thioredoxin reduces the target protein, leaving thioredoxin oxidized. In members of this superfamily the N-terminal Cys residue is more solvent-exposed (i.e. acidic) than the C-terminal Cys residue. This residue is kept buried in a hydrophobic environment until the occurrence of conformational changes induced by the formation of the mixed disulfide bond (43,44). Like a bona fide thioredoxin, it is likely that the Cys-75 (rather than Cys-78) of CcmG carries out the nucleophilic attack on Cys-34 of apocyt c 1 . Earlier studies attributed reduction of the apocyt c HBS disulfide bond to CcmH (13,26,27), and we also observed that CcmH Cys-45 could carry out this task (k of 2.9 ϫ 10 2 M Ϫ1 s Ϫ1 ) but slower than CcmG (k of 6.7 ϫ 10 2 M Ϫ1 s Ϫ1 ) ( Table 2). Hence, in wildtype cells (i.e. in the presence of DsbA), the N-terminal Cys-75 thiolate from reduced CcmG rapidly attacks oxidized apocyt c 1 WT HBS via Cys-34, followed by resolution of the mixed disulfide thus formed by its C-terminal Cys-78, yielding oxidized CcmG that is subsequently reduced by CcdA (Fig. 7, left side).
Next, we observed that both Cys-75 and Cys-78 of CcmG could rapidly resolve a mixed disulfide between CcmH Cys-45 and apocyt c 1 Cys-34 (Table 2), in agreement with another earlier proposal that a mixed disulfide bond involving CcmH might be the substrate of CcmG (12). The k values observed with CcmG Cys-78 are slightly higher than those obtained with CcmG Cys-75 in the attack of CcmH Cys-45 possibly due to conformation changes in CcmG inflicted by single Cys mutations.
The 3D structure of CcmH shows that its Cys-42 residue is buried inside the protein and is less accessible, which is consistent with its lower reactivity seen here (12,19). Similarly, the observed higher reactivity for the thiol-disulfide exchange reactions of apocyt c 1 Cys-34 compared with apocyt c 1 Cys-37 suggests that Cys-34 of apocyt c 1 might be more solvent-exposed. However no structural information is available.
Conceivably, the mixed disulfide containing CcmH Cys-45 -Cys-34apocyt c 1 intermediate described above (Fig. 7, step 7) could also be formed at a slower rate by an alternative route, via reduction of oxidized apocyt c with reduced CcmH (if available) (Fig. 7, steps 6a-7a, k of 2.9 ϫ 10 2 M Ϫ1 s Ϫ1 ). However, if under steadystate growth conditions DsbA keeps both CcmH and apocyt c oxidized then in the absence of reduced CcmG, this alternative pathway would be unlikely to occur. Similarly, in the absence of reduced CcmG, the mixed disulfide containing CcmH Cys-45 -Cys-34 apocyt c 1 intermediate (if formed) could also be resolved intra-molecularly (Fig. 7, steps 9a-10a) via Cys-42 of CcmH. However, this alternative would occur at a much lower rate compared with CcmG, as Cys-42 thiolate is not very reactive (reduced CcmH Cys-42 reacts with TNBapocyt c 1 at a k of 2.2 ϫ 10 2 M Ϫ1 s Ϫ1 ). Furthermore, the model depicted in Fig. 7 also accounts for low amounts of cyt c observed in several Ccm-related R. capsulatus mutants (14,22). In mutants lacking DsbA, reduced apocyt c would be rapidly degraded, and moreover, CcmH would remain reduced until its oxidation by non-Ccm-dedicated components (e.g. O 2 ), creating a rate-limiting step and reducing Ccm efficiency (ϳ 50%) (14). Similarly, in the absence of both DsbA and CcdA or DsbA, CcdA, and CcmG, a severe decrease in cyt c amounts (to ϳ10% of wild-type levels) would occur due to cumulative effects of the absence of DsbA and CcmG, which are required for stability of apocyt c and efficient resolution of CcmH Cys-45 -Cys-34 apocyt c 1 mixed disulfide, respectively.
In summary, this study provided a comprehensive description of thioreduction of apocyt c HBS disulfide bond, which is essential for Ccm to occur. Importantly, it identified the probable catalytic Cys residues of the Ccm components involved in these events. Future investigations will establish the occurrence

Experimental procedures
Bacterial strains, growth conditions, plasmid, and mutant constructions E. coli and R. capsulatus strains used in this work are defined in Table 1. E. coli strains were grown aerobically at 37°C, shaken at 200 rpm in Luria-Bertani (LB) broth medium, and supplemented with ampicillin (100 g/ml) or chloramphenicol (50 g/ml), as needed. Cultures were induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside as described elsewhere (29). R. capsulatus strains were grown chemoheterotrophically in the dark (i.e. by aerobic respiration) at 35°C, shaking at 150 rpm on enriched medium (MPYE) supplemented with tetracycline (2.5 g/ml), as appropriate.
R. capsulatus apocyt c 1 mutants were produced using the QuikChange site-directed mutagenesis kit and the plasmid pMAM1 as a template. pMAM1 encodes a variant of apocyt c 1 missing its last C-terminal 39 amino acids that constitute the TM helix and lacking the non-heme ligating Cys-144 and Cys-167 that form a structural disulfide bridge (Strep-apocyt c 1 WT ) (30). The plasmids pMAM1C37S, pMAM1C34S, and pMAM1C34SC37S obtained by site-directed mutagenesis produced the single and double (indicated by *) Cys mutant derivatives Strep-apocyt c 1 Cys-34 , Strep-apocyt c 1 Cys-37 , and Strepapocyt c 1 * of Strep-apocyt c 1 WT , respectively (Table 1). A soluble variant of CcmH was constructed by digesting the plasmid pCS1604 with NdeI and BamHI, isolating the insert carrying a truncated CcmH and ligating it into the plasmid pFLAG-1 using the same sites to obtain the plasmid pAV10 (Table 1). This construct encodes an N-terminal FLAGfused CcmH that lacks both its signal sequence and its 46 N-terminal amino acid residues acting as its membrane anchor (FLAG-CcmH WT ). Plasmids pAV10C45S, pAV10C42S, and pAV10C42SC45S, producing the single FLAG-CcmH Cys-42 and FLAG-CcmH Cys-45 and the double FLAG-CcmH* Cys mutant derivatives of FLAG-CcmH WT , respectively, were obtained by site-directed mutagenesis as above, with plasmid pAV10 serving as a template (Table 1).
A soluble variant of CcmG was prepared by digesting with EcoRI and HindIII the plasmid pQE60-helX, isolating the DNA fragment corresponding to the C-terminal His 6 -tagged CcmG without its TM helix, and ligating it into pBSK using the same restriction sites to yield the plasmid pCS1555 producing His 6 -CcmG WT (Table 1). Plasmids pCS1553, pCS1552, and pCS1554 producing the single His 6 -CcmG Cys-75 and His 6 -CcmG Cys-78 and double His 6 -CcmG * Cys mutant derivatives of His 6 -CcmG WT , respectively, were obtained by site-directed mutagenesis as above, using plasmid pQE60-helX as a template. Accordingly, plasmids pCS1556, pCS1557, and pCS1558 were constructed like pCS1555 described above, by cloning into the EcoRI-HindIII sites of pBSK the His 6 -CcmG mutant derivatives carried by the plasmids pCS1552, pCS1553, and pCS1554, respectively (Table 1). All constructs were confirmed by DNA sequencing. The nomenclature used for the plasmids refers to the Cys residues that were mutated to Ser. Mutant proteins produced by these plasmids are referred to by the Cys residue that remained after mutagenesis (e.g. plasmid pAV10C45S specifies that Cys-45 of CcmH was mutated to Ser, thus it produces the FLAG-CcmH Cys-42 mutant protein).

Protein purification and production of anti-CcmG polyclonal antibodies
E. coli cells overproducing each desired protein were resuspended in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl buffer containing protease inhibitors mixture (Pierce), and disrupted using a French pressure cell. Crude cell extracts were centrifuged at 138,000 ϫ g for 90 min at 4°C. Purification of Strepapocyt c 1 WT , Strep-apocyt c 2 WT , and their derivatives was performed using Strep-Tactin-Sepharose resin (IBA, Inc.) as done earlier (29). Purification of His 6 -CcmG WT and its derivatives was carried out using a nickel-Sepharose high performance column (GE Healthcare) equilibrated with 25 mM Tris-HCl, pH 7.5, 500 mM NaCl, 10 mM imidazole buffer. After extensive washing with the same buffer, elution was done at 100 mM imidazole. Purification of His 10 -CcmI used similar buffer conditions, except that the buffers contained 0.01% DDM, and elution was done at 500 mM imidazole, as described previously (29). Purification of FLAG-CcmH WT and its mutant derivatives used an anti-FLAG (DYKDDDDK) affinity gel (Biotool, Inc.) and 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.2 mM 4-(2-aminoethyl)benzenesulfonyl fluoride (AEBSF) buffer according to the manufacturer's instructions. Elution was carried out with 100 mM glycine, pH 2.75, 0.2 mM AEBSF buffer, and eluents were collected into tubes containing 1 M Tris-HCl, pH 8.0, buffer for immediate neutralization. All purified proteins were concentrated using Amicon-YM 3 (Millipore, Inc), desalted via a PD-10 column (GE Healthcare) equilibrated with 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA buffer, and kept at Ϫ20°C until further use.
Purified R. capsulatus His 6 -CcmG WT (ϳ3 mg) was subjected to preparative SDS-PAGE, electro-eluted from the gel matrix, and used as an antigen for production of rabbit polyclonal antibodies, which was performed by Thermo Fisher Scientific.

Protein-protein interaction studies using co-purification assays
Protein-protein interactions between the double Cys mutants His 6 -CcmG*, FLAG-CcmH* , and Strep-apocyt c 1 * were determined in vitro using co-purification assays, as described below. Equimolar amounts (ϳ10 M) of purified His 6 -CcmG* were mixed with FLAG-CcmH* or Strep-apocyt Thioreduction branch of the Ccm pathway c 1 * in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM imidazole buffer (final volume 60 l) and incubated for 60 min at room temperature, with gentle shaking. Then, 10 l of nickel-Sepharose HP resin previously equilibrated with the assay buffer was added, and incubation was continued for another 60 min with gentle shaking. At the end of incubation, the assay mixture was centrifuged at 8000 ϫ g for 1 min, and the supernatant (i.e. flow-through (FT)) was removed. The resin was washed twice with 100 l (10 resin volumes) of buffer supplemented with 80 mM imidazole, and the supernatants (i.e. washes (W)) were collected by centrifugation. His 6 -CcmG* and its partners were eluted with 50 l of elution buffer containing 250 mM imidazole (i.e. elution (E)). FT and E fractions were concentrated and analyzed by SDS-PAGE. Protein-protein interaction assays using purified His 6 -CcmG WT , Strep-apocyt c 2 WT , and His 10 -CcmI were done, as described previously (29), using equimolar amounts of CcmG and apocyt c 2 (1.5 M) and substoichiometric amounts of CcmI (0.5 M) in a final volume of 400 l. Copurification of Strep-apocyt c 2 WT with R. capsulatus-solubilized membranes was done as described previously (29). Co-purification of native CcmG from CcmFHG-enriched fractions (100 g) was done using His 10 -CcmI (10 g) and nickel-Sepharose resin as described elsewhere (29).

Protein-protein interactions determined by biolayer interferometry
Binding kinetics of CcmG and apocyt c 1 were monitored quantitatively in real time by biolayer interferometry using an Octet RED96 instrument (ForteBio, Inc.) as described elsewhere (30). Briefly, streptavidin (SA)-coated biosensors were loaded with biotinylated Strep-apocyt c 1 WT (400 nM) in 50 mM Tris, pH 8.0, 100 mM NaCl, 0.05% DDM, 1% BSA buffer at 30°C and at 1000 rpm. After washing, the SA sensors were incubated with increasing concentrations of His 6 -CcmG WT (from 1.28 to 20.4 M) (association step). Subsequent washing of the biosensors with the assay buffer released CcmG WT from the immobilized apocyt c 1 WT (dissociation step). SA sensors without immobilized apocyt c 1 WT were incubated with CcmG WT as a control for unspecific binding of this protein to the sensors. In addition, an assay lacking CcmG WT was used as a negative control to confirm that the monitored shifts were due to the formation of CcmG WT -apocyt c 1 WT complexes. The assays were done in duplicate, and the k on and k off rates of binding measured and the K D values were determined by fitting the experimental data to a 1:1 homogeneous kinetic model describing bimolecular interactions, as done earlier (30).

Preparation of TNB adducts of purified proteins
Single Cys mutant derivatives CcmH Cys-42 , CcmH Cys-45 , apocyt c 1 Cys-34 , and apocyt c 1 Cys-37 (50 -100 M) were reduced with large excess of DTT (25 mM) at room temperature for 60 min. Excess DTT was removed using a PD-10 desalting column and 100 mM potassium phosphate, pH 8.0 buffer. Reduced proteins were incubated with 15 mM DTNB in the dark for 120 min at room temperature, and excess of DTNB was also removed by desalting with a PD-10 column, using 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA buffer. The yield of the protein-TNB adduct formed was determined by reducing a small aliquot with excess of DTT for a few minutes and measuring spectroscopically at 412 nm the amount of 2-nitro-5-thiobenzoic acid (TNB 2Ϫ ) ions released. The concentrations of the protein-TNB adducts were calculated using an extinction coefficient ⑀ 412 of 14,500 M Ϫ1 cm Ϫ1 for the TNB 2Ϫ ions. For all protein-TNB adducts, the ratio of released TNB 2Ϫ ion per protein was between 0.90 and 1.03, as expected for a single thiol group, confirming their full modification. Reduced forms of the CcmG Cys-75 , CcmG Cys-78 , CcmH Cys-42 , and CcmH Cys-45 single Cys mutant derivatives were incubated immediately before use with an excess of DTT for 30 min at room temperature, and an excess of DTT was removed using a PD-10 column equilibrated with the assay buffer. In the case of the apocyt c 1 Cys-34 and apocyt c 1 Cys-37 derivatives, both reduction and DTNB conjugation were carried out in an anoxic chamber (COY Lab Products, Inc) to minimize their high tendency to form inter-molecular disulfide bonds in the presence of oxygen. Reduced samples of apocyt c 1 Cys-34 and apocyt c 1 Cys-37 were kept under anoxic conditions in sealed glass vials until use. An aliquot of each reduced sample was also incubated with 10 mM IOA for 30 min in the dark to control the extent of reduction and used as a negative control for the TNB 2Ϫ release assays, to ensure that no reductant other than the thiol of the chosen protein was present during the assay. As needed, full reduction of the samples was confirmed by determining the amount of free thiol groups available just prior to the DTNB assays, using the Ellman's reagent protocol (47). All protein samples were concentrated using Amicon YM10 after desalting, and their final concentrations were determined using the BCA assay (Sigma).

Determination of bimolecular rate constants of thiol-disulfide exchange reactions between a TNB-protein adduct and another reduced protein with a single Cys residue
The bimolecular rate constant (k) of thiol-disulfide exchange reaction between a chosen pair of Cys residues from two proteins was determined using the DTNB assay, as described previously (12,32,33). Briefly, a given amount (1 M) of TNB-protein (CcmH Cys-42 , CcmH Cys-45 , apocyt c 1 Cys-34 , or apocyt c 1 Cys-37 ) adduct was added to a stirring cuvette containing variable concentrations (1-30 M) of another fully reduced protein (CcmG Cys-75 , CcmG Cys-78 , CcmH Cys-42 , CcmH Cys-45 , apocyt c 1 Cys-34 , and apocyt c 1 Cys-37 ) in 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA buffer. For each case, the time-dependent increase in A 412 nm corresponding to release of the TNB 2Ϫ ions was monitored. The reaction was carried out under pseudo-first order kinetics conditions, where the reduced protein was in excess relative to the TNB-protein adduct. All assays were performed at least in duplicate. For each Cys pair, the initial rate of release of TNB 2Ϫ ions (product formation) was determined for a range of concentrations of reducing protein (e.g. CcmG 78 ). These rates were converted to k obs values using the initial concentration of the protein-TNB adduct (1 M CcmH Cys-45 -TNB in this case) and the absorption coefficient at 412 nm of TNB 2Ϫ . Similar k obs values were also obtained by exponential fit to the saturation kinetics. The k obs values were then plotted against the concentrations of the reducing protein used (e.g. CcmG Cys-78 ), where the slope of the curve is the bimolecular rate constant k (e.g. 23 ϫ 10 2 M Ϫ1 s Ϫ1 ) Thioreduction branch of the Ccm pathway of thiol-disulfide exchange reaction between the TNB-protein adduct (e.g. CcmH Cys-45 -TNB) and a fully reduced partner protein with a single Cys residue (e.g. CcmG Cys-78 ).

Identification of mixed disulfides between CcmG and CcmH and nLC-MS/MS analysis
The stability and efficiency of mixed disulfide bond formation between each Cys pair of two proteins were tested by incubating 15 M of a protein-TNB adduct with 2-fold (30 M) excess of fully reduced single Cys mutant derivative of another protein in a final volume of 20 l in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA buffer at room temperature for 2-16 h. Upon mixing of the two proteins, a slight yellow color was formed, indicating release of the TNB 2Ϫ ions. At the end of incubation, the reaction was stopped by blocking the remaining free thiols with 5 mM IOA at room temperature in the dark for 15 min. SDS-PAGE loading buffer without any reducing agent was added to the samples, which were then submitted to 15% SDS-PAGE. SDS-polyacrylamide gel bands were excised and subjected to alkylation with iodoacetamide, followed by in-gel trypsin digestion (Promega, sequencing grade modified trypsin) overnight at 37°C. Peptides eluted from the gel samples were dried and resuspended in 10 l of 5% acetonitrile in 0.1% formic acid and analyzed with a nanospray LC-MS/MS Thermo LCQ Deca XPϩ ion trap mass spectrometer coupled to a Thermo-Dionex LC Packings Ultimate Nano HPLC system controlled by Thermo Xcalibur version 2.0 software. A C18 nanocolumn (Thermo-Dionex, NAN-75-15-03-PM) was used to fractionate peptides, using a 60-min elution gradient (5-75% acetonitrile in 0.1% formic acid). MS/MS data were acquired in data-dependent analysis mode, where the top three precursor ions were trapped and fragmented using dynamic exclusion to maximize the detection of unique peptides. Collected spectra were searched against the R. capsulatus protein database, which included the mutated CcmH and CcmG sequences using Thermo Proteome Discoverer 1.4 software with standard settings.

Determination of the redox states of CcmG and CcmH in actively growing cells
To determine the in vivo redox state of CcmG and CcmH, we used a simple approach that consists of alkylating the free thiols of the Cys residues of the proteins of interest using AMS as described previously (48). Ten-ml cultures of R. capsulatus strains MD14 (lacking CcmH) and MD11 (lacking CcmG), complemented with plasmids pST6 (Strep-CcmH WT ) and PCS1555 (His 6 -CcmG WT ) (Table 1), respectively, were grown until reaching early exponential phase (A 630 ϳ0.3) at 35°C, 150 rpm in enriched medium. Two 1.8-ml aliquots were collected (tubes A and B) and precipitated on ice with 10% ice-cold trichloroacetic acid (TCA) for 30 min. The remaining culture was reduced by addition of 10 mM DTT for 10 min at 35°C and shaking at 150 rpm. Two additional aliquots of 1.8 ml were collected (tubes C and D) and precipitated with TCA, as before. After centrifugation for 5 min at 16,000 ϫ g and 4°C, the four pellets were washed twice with ice-cold acetone, and dried. Dry pellets of the tubes A and C were solubilized in 45 l of 50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.1% SDS, and 2 mM AEBSF buffer, and those of the tubes B and D were solubilized in the same buffer supplemented with 20 mM AMS, for 1 h at 37°C, with shaking. SDS-PAGE loading buffer was added to the tubes and incubated at 95°C for 15 min. Protein extracts were resolved in a 18% gel and analyzed by immunoblots using anti-CcmG-and anti-CcmH-specific polyclonal antibodies, as appropriate.