The Cytochrome c Maturation Components CcmF, CcmH, and CcmI Form a Membrane-integral Multisubunit Heme Ligation Complex*

Cytochrome c maturation (Ccm) is a post-translational and post-export protein modification process that involves ten (CcmABCDEFGHI and CcdA or DsbD) components in most Gram-negative bacteria. The absence of any of these components abolishes the ability of cells to form cytochrome c, leading in the case of Rhodobacter capsulatus to the loss of photosynthetic proficiency and respiratory cytochrome oxidase activity. Based on earlier molecular genetic studies, we inferred that R. capsulatus CcmF, CcmH, and CcmI interact with each other to perform heme-apocytochrome c ligation. Here, using functional epitope-tagged derivatives of these components coproduced in appropriate mutant strains, we determined protein-protein interactions between them in detergent-dispersed membranes. Reciprocal affinity purification as well as tandem size exclusion and affinity chromatography analyses provided the first biochemical evidence that CcmF, CcmH, and CcmI associate stably with each other, indicating that these Ccm components form a membrane-integral complex. Under the conditions used, the CcmFHI complex does not contain CcmG, suggesting that the latter thio-reduction component is not always associated with the heme ligation components. The findings are discussed with respect to defining the obligatory components of a minimalistic heme-apocytochrome c ligation complex in R. capsulatus.

and CcmI-2 domains play distinct roles during Ccm with the former being functionally interconnected with CcmF and CcmH, and the latter with CcmG (14,19).
In this work, using combinations of reciprocal affinity and size exclusion chromatography, we provide the first direct biochemical evidence that R. capsulatus CcmF, CcmH, and CcmI interact with each other to form a stable, multisubunit membrane protein complex. Implications of this CcmFHI-containing heme ligation complex lacking CcmG for the heme-apocyt c ligation process during Ccm are discussed.
Molecular Genetic Techniques-Molecular genetic techniques were performed using standard procedures (25). All constructs were confirmed by DNA sequencing. Sequence analyses and comparisons were conducted using MacVector (Accelrys, San Diego, CA) and BLAST software packages (26). Constructions of various epitope-tagged Ccm derivatives were as follows. A Strep-tag sequence was added in-frame to the 5Ј-end of ccmF in pYZ1 (23) via the QuikChange TM site-directed mutagenesis kit according to the manufacturer's instructions (Stratagene, La Jolla, CA) using the primers CcmF-StrepN-Fwd (5Ј-GGA GGA CCC CGC ATG ATC AGC TGG AGC CAC CCG CAG TTC GAA AAA GGC GTC GAG ACC GGC CAT-3Ј) and CcmF-StrepN-Rev (5Ј-GAA ATC GCC GGT CTC GAC GCC TTT TTC GAA CTG CGG GTG GCT CCA GCT GAT CAT GCG GGG TCC-3Ј) to create pYZ5. The 2.95-kb XbaI and KpnI-restricted fragment of pYZ5 carrying Strep::ccmF, ccmH and the G488A promoter-up mutation overexpressing the ccmF-ccmH gene cluster (23) was then cloned into the respective sites of pRK415 to yield pYZ6. Plasmids pST6, pST7, and pST8 were constructed as described in Table 1 (39). The plasmids pCS1581 (ccmH::Strep ccmI::FLAG) and pCS1582 (Strep::ccmF-ccmH ccmI::FLAG) were constructed by cloning the 1.7-kb KpnI fragment of pCS1564 (19) containing ccmI::FLAG expressed via its own promoter into the unique KpnI site of pST6 and pYZ6, respectively. To construct pCS1718, the R. capsulatus cycA mat , corresponding to the matured form of cyt c 2 protein encoded by cycA, was PCRamplified using pHM14 (27) as a template and RccycAmat/ NdeI-Fwd (5Ј-CAT ATG GGC GAC GCC GCG AAG GGC GA-3Ј) and RccycAmat/NdeI-Fwd (5Ј-GGA TCC TAT TTC ACG ACC GAG GCC AG-3Ј) as primers. The generated 0.37-kb fragment was phosphorylated and cloned into EcoRVrestricted pBSK. Plasmid pCS1718 was then digested with NdeI and BamHI sites, which were introduced during the PCR amplification from pHM14, and the DNA fragment corresponding to cycA mat cloned into the same sites of pCS1302, a derivative of pCS905 (28), to yield pCS1726. This plasmid contains an inframe Strep-tag sequence fused at the 5Ј-end of cycA mat expressed from a Ptac-lac promoter-operator system in E. coli.
Detergent-solubilized Membrane Protein Preparation-R. capsulatus cells grown by respiration were resuspended in TNE1 buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 25 mM EDTA, 0.1 mM ⑀-amino-caproic acid, and 0.1 mM Pefabloc SC) at a ratio of 5 ml per g cell wet weight, and intracytoplasmic membrane vesicles (chromatophores) prepared using a French pressure cell as described in Ref. 21. Chromatophores were homogenized in TNE1 buffer at a protein concentration of 8 mg/ml, solubilized by addition of n-dodecyl ␤-D-maltoside (DDM; Sigma-Aldrich) at a protein:detergent ratio of 1:1 (0.8% w/v DDM) from a 20% (w/v) stock solution under continuous stirring for 1 h at 4°C, and then centrifuged for 2 h at 4°C and 120,000 ϫ g to collect solubilized membrane proteins in the supernatants for further use.
Various Protein Chromatography-For size exclusion chromatography, solubilized membrane proteins were loaded onto a Sephacryl S-400 HR column (GE Healthcare Biosciences, Piscataway, NJ) pre-equilibrated with five column volumes of TNED1 buffer (TNE1 plus 0.01% (w/v) DDM), which was also used as elution buffer. The flow rate was adjusted to 0.8 ml/min, the absorption of the eluates monitored at 280 nm, and 2.4 ml per fraction were collected. For each Ccm component monitored, proteins present in 400-l aliquots of desired fractions were precipitated and subjected to SDS-PAGE and immunodetection. The size exclusion column was calibrated using blue dextran (2000 kDa), thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), and aldolase (158 kDa) to estimate the molecular mass ranges across the fractions collected. As needed, appropriate high molecular mass fractions were pooled for further analyses.
For tag-affinity chromatography, the TNED1 buffer of the solubilized membrane proteins was exchanged with TNED2 (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, and 0.05% (w/v) DDM), to not damage the tag-affinity matrices with high EDTA amounts, using a PD-10 column (GE Healthcare Biosciences, Piscataway, NJ). Protein concentrations were adjusted with the same buffer to a final concentration of 2 mg/ml, and loaded by three successive passages to appropriate affinity columns (StrepTactin-Sepharose from IBA or ANTI-FLAG-Agarose from Sigma-Aldrich).
For tandem size exclusion and affinity chromatography, high molecular weight fractions separated by the size exclusion column were pooled, the TNED1 buffer was exchanged with TNED2, and proteins were concentrated to 1 mg/ml protein using 10-kDa cut-off Centriplus YM-10 centrifugal filter units (Millipore, Billerica, MA). Affinity columns contained 1-ml matrix volume and were pre-equilibrated with 20 ml of TNED2 buffer. Following sample loading, the respective columns were washed twice with 5 ml of TNED2 buffer, and Strep-tagged (i.e. CcmF and CcmH) or FLAG-tagged (i.e. CcmI) proteins were eluted using 5ϫ matrix volumes (E1-E5 fractions, 1 ml each) with TNED2 buffer containing 5 mM desthiobiotin (DTB) or 0.1 mg/ml FLAG peptide, respectively. For each Ccm component, samples corresponding to 100 g of total proteins of non-solubilized or DDM dispersed chromatophore membranes, 25 g of flow-through or column wash solutions, and 5 g of elution fractions (E2-E5) were precipitated, and analyzed via SDS-PAGE and immunodetection. Protein concentrations were determined using the Bicinchoninic Acid kit from Sigma-Aldrich, and SDS-PAGE and immunoblot analyses were performed as described below.
Production of CcmF Antisera-Antisera generated toward predicted small (ϳ10 amino acids) soluble antigenic CcmF peptides did not detect CcmF in crude extracts by immunoblotting, even when CcmF was overproduced. Therefore, longer polypeptides (ϳseveral tens of amino acids) from soluble domains were generated and purified for antisera production. The following was determined to give the best results in immunoblots and was used in these studies. The R. capsulatus ccmF gene was used as template for synthesis of a PCR product that encoded the sixth periplasmic domain of R. capsulatus CcmF (P6, as designated in Ref. 16) was amplified. The forward primer sequence was 5Ј-GGG CCC ATG GAG GAT ATC CGC GTG GCG AAG, beginning at the amino acid residue 512 of CcmF with the sequence EDIRV (the ATG in the primer encodes the initiating methionine). The reverse primer sequence was 5Ј-ATC CCA AAG CTT GTT CGC GAA AGG CTT GAC, whereby the amino acid sequence KPFN represents the final residues within the CcmF fragment (ending at residue 609). NcoI and HindIII sites (underlined) were engineered into the forward and revers primers for cloning purposes, respectively. This NcoI and HindIII restricted ccmF fragment cloned into the cytoplasmic expression vector pRSETB (Invitrogen, Carlsbad, CA) produced high yields of an N-terminally hexahistidinetagged polypeptide of about 20 kDa. Rabbits were immunized either directly with via nickel affinity chromatography purified CcmF-P6 fragment, or after its separation via SDS-polyacrylamide gel and subsequent electro-elution. The antisera toward the polypeptide (called CcmFP6D) obtained from the latter procedure yielded the best results, and were used in Ref. 23, and in this study. SDS-PAGE and Immunoblot Analyses-Proteins were precipitated at Ϫ20°C with 90% (v/v) ice-cold acetone overnight and centrifuged for 30 min at 20,000 ϫ g and 4°C. The obtained pellets were air-dried and then re-solubilized in SDS loading buffer (62.5 mM Tris-HCl pH 6.8, 2% (w/v) SDS, 0.1 M dithiothreitol, 25% (v/v) glycerol, and 0.01% (w/v) bromphenol blue) by incubation at 42°C for 45 min prior to loading. SDS-PAGE was performed according to Ref. 29) using 15% (T) polyacrylamide gels. Gel-separated proteins were transferred onto polyvinylidene difluoride membranes (Millipore, Billerica, MA) at 1 mA/cm 2 for 2 h using a Trans-Blot SD semi-dry transfer cell (Bio-Rad). Membranes were then washed twice for 5 min at room temperature with TBS buffer (25 mM Tris/HCl, pH 7.5, and 150 mM NaCl), saturated with TTBS buffer (TBS ϩ 0.05% (v/v) Triton X-100 and 0.05% (v/v) Tween 20) containing 5% (w/v) nonfat dry milk for 1 h at room temperature, washed twice for 5 min at room temperature with TTBS buffer, and subsequently probed with rabbit antisera against CcmG (1:5000), CcmH (1:5000), and CcmF (1:1000), or rabbit ANTI-FLAG polyclonal antibodies (Sigma) (1:2000) for 16 h at 4°C. Thereafter, membranes were washed three times for 5 min with TTBS buffer and reprobed with either monoclonal anti-rabbit (immuno-globulins clone RG-16) alkaline phosphatase conjugate (Sigma-Aldrich) (1:2000) or stabilized goat anti-rabbit IgG horseradish peroxidase conjugates (GE Healthcare Bio-Sciences, Piscataway, NJ) (each 1:10,000) for 1 h at room temperature. Antibody-antigen complexes were visualized by chromogenic detection using the BCIP/NBT liquid substrate (Sigma-Aldrich) or by enhanced chemiluminescence (ECL) detection using the SuperSignal West Pico chemiluminescent substrate (Pierce).
Chemicals-All chemicals were of reagent grade and obtained from commercial sources.
Epitope-tagged Ccm components were probed for proteinprotein interactions by affinity chromatography using chromatophore membranes prepared from semi-aerobically grown cells and subsequently solubilized with DDM. Ccm components thus purified or copurified were monitored in flowthrough, wash, and ligand-eluted fractions by immunodetection using specific antibodies against R. capsulatus CcmH (11), CcmF ("Experimental Procedures") and CcmG (11) or the FLAG epitope fused to CcmI. Additional coimmunoprecipitation and cross-linking experiments were also conducted to verify the affinity purification results (data not shown).
CcmF and CcmH Interact with Each Other, but Not with CcmG-In E. coli, CcmF and CcmH are known to coimmunoprecipitate (13). To test whether similar interactions also occur in R. capsulatus, affinity purification of Strep-CcmF using StrepTactin-Sepharose was carried out with an appropriately complemented CcmF-null mutant (MD12 ϫ pYZ6, Table 1). DTB-eluted fractions contained Strep-CcmF and CcmH, but not CcmG (Fig. 1A), suggesting that at least a fraction of CcmH copurified with Strep-CcmF, and hence associated with each other. Reciprocal affinity purification was next performed to confirm this finding. Membranes from a CcmH-null mutant complemented with a CcmH-Strep derivative (MD14 ϫ pST6) were used, and the data showed that at least a fraction of CcmF, but again not CcmG, copurified with CcmH-Strep (ϳ3 kDa larger than native CcmH) under the same experimental conditions (Fig. 1B). These physical interactions therefore indicated that CcmF and CcmH form a stable complex, devoid of CcmG.
Interactions between CcmI and CcmH Appears to Be Stronger Than Those between CcmI and CcmF-Affinity purification of the CcmH-Strep derivative from membranes of a CcmI-null CcmH-null double mutant, complemented with coexpressed CcmI-FLAG and CcmH-Strep (MD15 ϫ pCS1581, Table 1), indicated that the DTB-eluted fractions contained, in addition to CcmH-Strep (M r of ϳ20,000), both CcmI-FLAG and CcmF but not CcmG (Fig. 3A). This result confirmed the close interactions of CcmI-FLAG and CcmF with CcmH-Strep, as seen earlier (Figs. 1 and 2A). Moreover, it also suggested that CcmF, CcmH, and CcmI are part of a stable multisubunit membraneintegral complex.
Finally, to further probe whether CcmF and CcmI interacted directly, affinity purification of the Strep-CcmF derivative was repeated using membranes from the CcmF-null/CcmI-null double mutant complemented with coexpressed Strep-CcmF and CcmI-FLAG derivatives (MD13 ϫ pCS1582, Table 1). DTB-eluted fractions containing Strep-CcmF (M r of ϳ58,000) had copious amounts of CcmH as shown above (Fig. 1), but included only trace amounts of CcmI-FLAG, and no CcmG (Fig. 3B). This observation suggested that, under the conditions used, the Strep-CcmF and CcmI-FLAG derivatives interacted either weakly with each other, or indirectly via CcmH, which was further assessed as described below.
CcmF, CcmH, and CcmI Form a Multisubunit Membraneintegral Complex-A different purification approach, which consisted of tandem size exclusion and affinity chromatography, was next used to establish that CcmF, CcmH, and CcmI are part of a multisubunit membrane-integral protein complex. DDM-dispersed proteins from a strain overproducing CcmF, CcmH, and CcmI-FLAG (MT-SRP1.r1 ϫ pNJ2, Table 1) were fractionated by FPLC within 8000 -20 kDa ranges using a Sephacryl S-400 HR column (180 ml). The CcmF, CcmH, and CcmI-FLAG contents of the elution fractions were determined using SDS-PAGE and immunoblot analyses. Large amounts of  Fig. 1 except that chromatophore membranes from R. capsulatus strain pNJ2 (CcmI-FLAG) ϫ MT-SRP1.r1 (CcmI-null overproducing CcmH and CcmF), ANTI-FLAGா-Agarose column for affinity chromatography, and FLAG peptide for elution were used as described under "Experimental Procedures." Aliquots from different steps of purification were analyzed by SDS-PAGE and immunoblots. Polyclonal antibodies against the FLAG tag (fused to CcmI), CcmH, CcmF, and CcmG as indicated on the right of each panel, and molecular mass markers (in kDa) are shown on the left. In each case, lane 1 contained 100 g of protein of CM from appropriate mutant strains (Table 1) used as negative controls for the respective immunoblots, as depicted on the top left. Panel B illustrates the plasmids pCS1581 coproducing CcmH-Strep and CcmI-FLAG, and pCS1582 coproducing Strep-CcmF, CcmH, and CcmI-FLAG. The corresponding genes and their epitope sequence fusions are represented as rectangles, and the promoters (P) of ccmI, cycA, and up-regulated ccmFH indicated with arrows. See "Experimental Procedures" and Table 1 for the construction of these plasmids.
CcmF, CcmH, and CcmI-FLAG were detected in the fractions 36 -48 and 66 -78, 33-51 and 54 -78, and 36 -45 and 54 -63, respectively (Fig. 4A). Additional protein bands of unknown identities with molecular weights larger than those of CcmH and CcmI-FLAG but still reacting with CcmH-and CcmI-FLAG-specific antibodies were also observed in fractions 51-60. Some of these bands might reflect the different thioredox states of CcmH. Calibration of the size exclusion chromatography column with proteins of known molecular masses ("Experimental Procedures") indicated that CcmF, CcmH, and CcmI-FLAG colocalized mainly in the elution fractions 30 -50, corresponding to molecular masses of protein complexes larger than 400 kDa with detection peaks for all three components around ϳ800 kDa (Fig. 4A). These fractions were pooled after eight independent FPLC runs (ϳ40 mg total proteins), concentrated, and subjected to ANTI-FLAG-Agarose affinity chromatography to purify CcmI-FLAG. Fractions eluted with the FLAG peptide contained CcmI-FLAG, CcmH, and CcmF, but not CcmG (Fig. 4B). Thus, CcmH and CcmF copurified readily with CcmI-FLAG from the high molecular weight fractions pool, demonstrating that they formed a multisubunit membrane protein complex, which was devoid of CcmG under the conditions used here.

DISCUSSION
Based on our earlier studies using the suppressors of CcmInull mutants, we proposed that CcmF, CcmH, and CcmI form a multisubunit membrane-integral complex in R. capsulatus (14,19). Here, we sought direct biochemical evidence to substantiate this hypothesis. At the onset, this task appeared daunting as most of the Ccm components are poorly characterized mem-brane proteins (2,3). Purification of membrane-integral protein complexes, especially those lacking optically detectable cofactors, is challenging because it requires empirical definition of adequate conditions for membrane lipid dispersion while keeping the subunit interactions intact. In our case, reliable detection means for individual Ccm components were also restricted (2). First we generated functional epitope-tagged derivatives of CcmF, CcmH, and CcmI, and expressed them individually or in pairs in appropriate mutants to supplement available Ccm antibodies. After establishing that the tagged Ccm derivatives were functional, we initiated purification of various Ccm components. Lack of CcmI-specific antibodies initially restrained our ability to monitor it in some instances, but this difficulty was surmounted upon coproduction of a CcmI-FLAG derivative with Strep-CcmF and CcmH-Strep components. Reciprocal affinity purification as well as size exclusion fractionation followed by affinity chromatography yielded for the first time strong supporting evidence that CcmF, CcmH, and CcmI associate with each other to form a multisubunit membrane protein complex. Pairwise copurification results suggested that CcmF and CcmI interact poorly or indirectly with each other despite their strong and presumably direct association with CcmH (Fig. 5). In addition, in the early experiments we noted that the flow-through fractions contained CcmH not associated with CcmF and vice versa, CcmF not associated with CcmH (Fig. 1). This observed sub-stoichiometry suggests that either free pools of these components are also present in membranes or that the solubilization conditions used here are not optimal to fully preserve the intactness of the CcmFHI-containing complex. Thus, the experiments reported here do not address the degree of purity or stoichiometry of the components.
Clearly, the Ccm complex identified here does not contain CcmG, but whether it includes CcmE or other components awaits the availability of specific antibodies. CcmG was present in readily detectable amounts in solubilized membranes of all utilized R. capsulatus strains, and its absence of interaction with CcmH was surprising, because CcmG reduces oxidized CcmH in vitro (11). However, no mixed disulfides between CcmG and CcmH have so far been trapped in vivo. Furthermore, the compensatory thio-redox interactions between DsbA and CcdA or CcmG apparently do not include CcmH (39). It therefore seems that CcmG associates either too transiently or indirectly with CcmF, CcmH, or CcmI under our experimental conditions. Interestingly, our preliminary data indicate that addition of E. coli-produced and purified apocyt c 2  (Table 1) as a negative control for the respective immunoblot analysis, as depicted on top left of each panel.
("Experimental Procedures" and Table 1) to detergent-dispersed membranes leads to its colocalization with CcmG in a subcomplex of ϳ100 kDa and also with the ϳ800 kDa complex including CcmFHI, 3 in agreement with our current thoughts about how the apocyt c might interact with the Ccm components in R. capsulatus (14,19,31). In any event, establishing that the CcmFHI components are parts of a multisubunit complex supports the idea that cyt c maturation process is carried out by a well-defined membrane-integral apparatus (Fig. 5).
Furthermore, using R. capsulatus as a model for Ccm-system I, we report here the first biochemical evidence for physical association of CcmH and CcmI. We found that CcmF and CcmI interact weakly with each other, or possibly indirectly via CcmH, to form a complex, which does not contain CcmG. Previously available coimmunoprecipitation data with E. coli indicated that only CcmF and CcmH (13,32) (as well as Arabidopsis thaliana CcmH (32)) interact with each other. Remarkably, colocalization of CcmH, CcmF, and CcmI-FLAG peaking in high molecular mass fractions of ϳ800 kDa during size exclusion chromatography is intriguing. This observation suggested that the Ccm components either form large aggregates escaping solubilization or are inherent parts of a large multisubunit complex with additional proteins. In mitochondria from T. aestivum (33) and A. thaliana (32), high molecular mass complexes (ϳ700 and ϳ500 kDa, respectively) containing the bacterial CcmF or CcmH homologs have also been detected. Using yeast two-hybrid assays, the A. thaliana CcmH homolog was shown to interact with apocyt c (32). In bacteria, coimmunoprecipitation data indicated that E. coli CcmF interacts with CcmE, but not with apocyt c (13). Similarly, the C-terminal helix and its adjacent loop of the pentaheme cyt c NrfA was found to interact with the CcmI ortholog NrfG via a TPR domain (34). Currently, to what extent or under which conditions, the CcmFHI-containing complex also associates with other Ccm components (such as CcmG and CcmE) or with apocyts c is unknown, precluding the conclusion that it is composed of solely three subunits (Fig. 5). Indeed, future identification of the high Proteins from every third elution fraction (from 30 to 78) were precipitated, subjected to SDS-PAGE, and stained with Coomassie (A, middle section). Immunoblot analyses were carried out using polyclonal antibodies against CcmF, CcmH, and the FLAG epitope (fused to CcmI) (A, lower section). In each case, the molecular mass markers (in kDa) are indicated on the left and the specific antibodies as probes are shown on the right. B, elution fractions 30 -49 containing high molecular weight materials were pooled, concentrated, and utilized for the purification of FLAG-tagged CcmI as described under "Experimental Procedures." Samples from different purification steps (P, concentrated fraction pool 30 -49; FT, flow-through after the ANTI-FLAGா-Agarose column; W, column wash; E2-E5, FLAG peptide eluted fractions 2-5) were analyzed by SDS-PAGE (100 g of protein in lanes 3 and 4, 25 g of protein in lane 5, and 5 g of protein in lanes 6 -9) and immunoblots using polyclonal antibodies as indicated on the right. Lanes 1 and 2 correspond to 100 g of protein of CM from a mutant strain used as a negative control for the respective immunoblot analysis (lane 1) or from R. capsulatus strain pNJ2 (pccmI::FLAG) ϫ MT-SRP1.r1 (⌬ccmI ccmFH up ) (lane 2), as depicted on the top, left side. Molecular mass markers (in kDa) are shown on the left.

FIGURE 5.
A minimalistic heme-apocyt c ligation process. The known ten (CcmABCDEFGHI and CcdA) components required for the Ccm-system I could be reduced to a set of five components as a minimalistic heme-apocyt c ligation apparatus per se, because in the absence of a thio-redox pathway (i.e. DsbA, CcdA) cyt c production still occurs, and CcmABCD is involved in loading heme to the holoCcmE. Thus, the apocyt c thioreductase/holdase CcmG, the heme chaperone CcmE and the heme ligation complex involving CcmFHI are the three major partners of this membrane confined process that is essential for cellular energy production. OCTOBER 31, 2008 • VOLUME 283 • NUMBER 44 molecular weight derivatives of CcmH and CcmI seen among the size exclusion chromatography fractions 51-60 will be informative. In summary, establishment of a Ccm-FHI-containing complex will hopefully enable us to further define the molecular events occurring during heme-apocyt c ligation in organisms using Ccm-system I for cyt c production.