The Rieske protein from Paracoccus denitrificans is inserted into the cytoplasmic membrane by the twin-arginine translocase Julie Bachmann 1 , Brigitte Bauer 1 , Klaus Zwicker 2 , Bernd Ludwig 1 and Oliver Anderka 1, * 1 Institut fu ¨ r Biochemie, Johann Wolfgang Goethe-Universita ¨ t, Frankfurt, Germany 2 Zentrum der Biologischen Chemie, Institut fu ¨ r Molekulare Bioenergetik, Universita ¨ ts-Klinikum, Frankfurt, Germany Mitochondrial respiratory complex III ⁄ cytochrome bc 1 is among the best-characterized membrane proteins, with structures elucidated from several species [1–4]. These structures revealed the organization of three cat- alytic subunits (SU) in the homodimeric complex; these are cytochrome b, cytochrome c 1 , and the Rieske iron– sulfur protein (ISP). Cytochrome b forms eight trans- membrane (TM) helices that bind two hemes and is largely contained within the membrane bilayer. Both cytochrome c 1 and the ISP are single-spanning TM proteins with globular hydrophilic domains located in the periplasmic space; these ecto-domains carry the heme and [2Fe)2S] cofactors, respectively. As the first cytochrome bc 1 structures were characterized, two fea- tures of the ISP came as a major surprise: (a) the ISP intertwines between the monomeric halves of the enzyme, such that the N-terminal TM helix of a given ISP is anchored within one monomer, whereas the Keywords cytochrome bc 1 complex; membrane targeting; Paracoccus denitrificans; Rieske iron–sulfur protein; twin-arginine translocation Correspondence O. Anderka, Institut fu ¨ r Biochemie, Johann Wolfgang Goethe-Universita ¨ t, D-60438 Frankfurt, Germany Fax: +49 69 3058 1901 Tel: +49 69 3051 2418 E-mail: oliver.anderka@sanofi-aventis.com *Present address Sanofi-aventis, TD Metabolism, Frankfurt, Germany (Received 16 June 2006, revised 23 August 2006, accepted 24 August 2006) doi:10.1111/j.1742-4658.2006.05480.x The Rieske [2Fe)2S] protein (ISP) is an essential subunit of cyto- chrome bc 1 complexes in mitochondrial and bacterial respiratory chains. Based on the presence of two consecutive arginines, it was argued that the ISP of Paracoccus denitrificans, a Gram-negative soil bacterium, is inserted into the cytoplasmic membrane via the twin-arginine translocation (Tat) pathway. Here, we provide experimental evidence that membrane integra- tion of the bacterial ISP indeed relies on the Tat translocon. We show that targeting of the ISP depends on the twin-arginine motif. A strict require- ment is established particularly for the second arginine residue (R16); con- servative replacement of the first arginine (R15K) still permits substantial ISP transport. Comparative sequence analysis reveals characteristics com- mon to Tat signal peptides in several bacterial ISPs; however, there are distinctive features relating to the fact that the presumed ISP Tat signal simultaneously serves as a membrane anchor. These differences include an elevated hydrophobicity of the h-region compared with generic Tat signals and the absence of an otherwise well-conserved ‘+5’-consensus motif lysine residue. Substitution of the +5 lysine (Y20K) compromises ISP export and ⁄ or cytochrome bc 1 stability to some extent and points to a specific role for this deviation from the canonical Tat motif. EPR spectroscopy confirms cytosolic insertion of the [2Fe)2S] cofactor. Mutation of an essential cofac- tor binding residue (C152S) decreases the ISP membrane levels, possibly indicating that cofactor insertion is a prerequisite for efficient translocation along the Tat pathway. Abbreviations EPR, electron paramagnetic resonance spectroscopy; ISF, Rieske iron–sulfur protein soluble fragment; ISP, Rieske iron–sulfur protein; SU, subunit; Tat, twin-arginine translocation; TM, transmembrane. FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4817 periplasmic domain structurally and functionally inter- acts with the other monomer; (b) the periplasmic domain seems to undergo large-scale motion in order to shuffle electrons between cytochromes b and c 1 .Up to eight accessory subunits surround the catalytic core of the enzyme; they are probably required for assembly and ⁄ or stability of the complex, but their precise func- tion is largely unknown. Cytochrome bc 1 complexes from bacterial respirat- ory chains, e.g. from Paracoccus denitificans, are made up of only the catalytically essential subunits, and show high sequence identity towards their mitochond- rial counterparts [5,6]. There is considerable interest in studying these minimal complexes as model systems; they are readily amenable to genetic manipulation and therefore allow unsolved issues of mechanism or bio- genesis to be tackled. However, structures of such ‘minimal’ bc 1 complexes cannot currently be solved to high resolution. In the case of the related b 6 f complex of oxygenic photosynthesis, a structure of prokaryotic origin has recently been characterized [7]. Despite its relatively simple composition, there is currently little information about biogenesis and assembly of the prokaryotic bc 1 complex. It is not known how cytochrome b as the central and largest subunit is inserted into the membrane. The cyto- chrome c 1 precursor is translocated along the Sec translocon; its heme cofactor is exported to the peri- plasm and attached to the apo-protein by the c-type cytochrome maturation machinery [8,9]. A twin-argin- ine-dependent translocation (Tat) was first proposed for the Rieske iron–sulfur protein by Berks [10], based on the occurrence of a specific consensus motif in its N-terminal region. Since then, considerable informa- tion has been obtained about the Tat system [11–13]. Its hallmarks are: (a) the occurrence of and export dependence on a S ⁄ T-R-R-x-F-L-K consensus motif within a tripartite signal peptide; (b) proton-motive force-dependent and ATP-independent transport; (c) insertion of cofactors and ⁄ or assembly of different subunits at a cytosolic stage; and (d) export in a fully folded conformation, which is probably the most remarkable feature. Components of the Tat trans- locon are TM proteins TatA, B, and C. TatBC seems to form the initial receptor [14]. Electron microscopy reveals that multiple copies of TatA form ring-like structures which are thought to represent the translo- cation pore [15]. Recently, specific chaperones have been identified which seem to exert ‘proof-reading’ or ‘quality control’ on the Tat translocon substrates [16,17]. In thylakoids, a ‘DpH pathway’ has been des- cribed that is homologous to the bacterial Tat pathway [18]. Currently known Tat substrates are almost exclu- sively soluble periplasmic proteins; to date only five Escherichia coli proteins containing a C-terminal mem- brane anchor have been shown to be transported along the Tat pathway. In contrast, the Rieske ISP is N-ter- minally anchored, which is novel and unique for a putative Tat substrate: The N-terminus would serve a dual role of export signal and membrane anchor. In the thylakoid system, it has been already shown that the ISP is transported via the DpH ⁄ Tat pathway [19,20]. Interestingly, the chloroplast ISP displays a KR motif, which is only the second known example of natural deviation from the otherwise invariant RR motif [21]. We examined membrane translocation of the ISP from P. denitrificans. Experimental evidence is provi- ded that the bacterial ISP is indeed a substrate of the Tat translocon, and transport depends on the presence of the Tat consensus motif. However, as in the case of the thylakoid ISP and in contrast to the majority of other Tat substrates, transport of the P. denitrificans ISP shows more relaxed requirements regarding the conserved RR motif. Furthermore, bioinformatic ana- lysis and site-directed mutagenesis reveal distinctive features of a Tat signal that simultaneously serves as a membrane anchor. Results Bacterial Rieske proteins contain signal sequences that deviate from the canonical Tat consensus In an early review on double-arginine signal sequences, the ISP of P. denitrificans was listed as a potential sub- strate for what was later named the Tat pathway [10]. To substantiate this assignment, the P. denitrificans ISP was initially analysed using bioinformatic tools. Its primary sequence was aligned to ISP sequences from other proteobacteria. Sequences were selected accord- ing to a phylogenetic study on Rieske proteins [22]. All chosen ISPs are subunits of respiratory cytochrome bc 1 complexes. The main part of the sequence representing the cluster-binding periplasmic domain was omitted from the comparison to avoid biasing the alignment towards this highly conserved protein region, which might mask the similarities of interest within the N-terminal part. The alignment shown in Fig. 1 reveals that all selec- ted sequences contain the indicative twin-arginine. Comparison with the canonical Tat consensus (S ⁄ T)- R-R-x-F-L-K [10] shows good agreement in the other positions of the motif, with the remarkable exception Rieske protein from Paracoccus denitrificans J. Bachmann et al. 4818 FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS of the C-terminal lysine residue, which is not found in any of the ISP sequences examined. On average, a lysine residue appears in this position in > 60% of general Tat signal sequences [10]. This position is num- bered ‘+5’, relative to the first invariant arginine; it corresponds to Y20 in the P. denitrificans sequence and is discussed in more detail later. Upstream of the consensus motif, a mean of 11 residues is found, con- sistent with the frequently observed extended n-region of Tat signal sequences relative to Sec signals [23]. The upstream sequences do not exhibit sequence conserva- tion; in contrast, it has been observed that signal pep- tides for proteins binding a given cofactor (e.g. the [Ni–Fe] hydrogenase small subunits) often show marked sequence conservation within this region [11]. For the n-region of cofactor-containing Tat substrates, an a-helical structure was proposed [12]. Using jpred (see Experimental procedures), a corresponding secon- dary structure could not be predicted for bacterial Rieske proteins (data not shown). The h-region was defined according to a set of rules given by Cristobal et al. [23]; it consists of 19 residues, in good agreement with a length of 15–20 residues found within known Tat signals. It is in remarkable contrast to established Tat substrate proteins that a number of conserved resi- dues can be found within the ISP h-region (for discus- sion, see below). The c-region of the putative ISP Tat signal predominantly displays an initial proline residue which has been described for other Tat signals as a helix breaker following the a-helical h-region [23]. However, Tat signal peptides characteristically contain basic amino acids within the h-region that serve as a ‘Sec-avoidance signal’; these basic residues are not observed in the analysed ISP sequences. All Rieske sequences lack the AxA cleavage site at the end of the c-region for obvious reasons, as this part of the ISP serves as a membrane anchor. The c-region overlaps with the flexible hinge-region observed in the crystal structures of the mitochondrial enzyme which allows for movement of the Rieske ecto-domain within the cytochrome bc 1 complex [1,4,24]. Taken together, the N-terminal domain of Rieske proteins displays several hallmarks of Tat signal peptides, such as the invariant twin-arginine and the tripartite structure. However, it also deviates in important aspects, missing for example the consensus lysine or a ‘Sec-avoidance’ signal. The N-terminal part of Rieske proteins serves as a membrane anchor, whereas the majority of known Tat substrates are exported to the periplasm where their export signals are cleaved. In order to examine this difference in detail, the corresponding h-regions were compared. Kyte–Doolittle analysis was performed with a sequence window size of 19, appropriate for detect- ing potential TM helices [25]. ISP sequences were selec- ted according to the sequence alignment in Fig. 1. For known Tat substrates, used as a comparison group, three to five sequences were taken from five different classes each: [NiFe] hydrogenase small subunits, MauM family ferredoxins, NapA periplasmic nitrate reductases, NosZ nitrous oxide reductases, and TorA Trimethylamine-N-oxide reductases (for details, see Experimental procedures). The resulting Kyte–Doolit- tle data were aligned relative to the Tat consensus motif. An averaged hydropathy index was calculated for the ISP sequences and the comparison group (Fig. 2). Both curves display a positive score in the h-region which corresponds to relative hydrophobicity. The data show that the ISP group is distinctly more hydrophobic than the comparison group; this differ- ence is statistically significant (P<0.001, two-tailed Fig. 1. Bacterial ISP sequences contain the twin-arginine consensus specific of the Tat translocation pathway. Sequence alignment of Rieske proteins that are subunits of proteobacterial cytochrome bc 1 complexes. The C-terminal portion of the sequences representing the cluster- binding hydrophilic domain is removed to avoid the alignment being biased towards this highly conserved protein region. Sequences were retrieved from the SwissProt server and the alignment performed with CLUSTAL X, as detailed in the Experimental procedures. Star symbols denote invariant residues, colons highly conserved and dots conserved positions. Limits of the h-region were determined following rules given previously [23]. The start of the so-called hinge region is indicated [4,57]. (Lower) Canonical Tat consensus motif. J. Bachmann et al. Rieske protein from Paracoccus denitrificans FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4819 Mann–Whitney U-test with pooled data for the corres- ponding h-regions). However, the ISP group h-region shows relatively weak hydrophobicity with hydropathy values < 1.5 compared with TM helices of multispan- ning membrane proteins which typically reach hydro- pathy values of > 1.8 in the Kyte–Doolittle analysis, using a window size of 19 [25]. This was confirmed with a small selection of single-spanning membrane proteins from P. denitrificans; here, hydropathy values for the TM helices ranged between 2 and 3 (data not shown). It was observed that the h-region of Sec signal peptides is significantly more hydrophobic than the h-region of common Tat peptides [23]; this also holds true when the ISP signal sequence is compared with Sec signal peptides. For a set of 20 predicted Sec sub- strates from P. denitrificans, a mean hydropathy value of 1.8 (± 0.1 SEM) was obtained for the h-region; the set of ISP h-regions showed a mean hydropathy value of only 1.0 (± 0.1 SEM) in the Kyte–Doolittle analysis (details not shown; here, a window size of 9 was applied to both data sets). To obtain comparative information from a different method, TM helix prediction was performed using the program tmap [26]. As an input, multiple sequence alignments were used that were generated with clu- stal x [27]. The algorithm predicts a TM helix for the ISP group only, not for the five different classes of Tat substrates mentioned above. Most remarkably, predic- tion of the ISP TM helix was absolutely dependent on the natural deviation from the canonical consensus motif described above: When a ‘+5’ lysine residue of the Tat consensus was introduced in silico (e.g. a Y20K mutation in the P. denitrificans sequence, see also below), TM prediction failed in all examined ISP sequences. In conclusion, slightly higher mean hydro- phobicity compared with average Tat signal peptides and the exchange of the canonical ‘+5’ lysine residue against a more hydrophobic amino acid (isoleucine, phenylalanine, tyrosine) provide a clear discrimination and an initial evidence for the ISP signal sequence to simultaneously serve as a membrane anchor. Finally, to predict whether the Tat translocation machinery is operating in P. denitrificans, the draft version of the genome was inspected (Joint Genome Institute Microbial Sequencing Program). Three genes annotated as TatA, TatB, and TatC homologues could be found on contig 67; TatB and TatC are adjacent genes and might form a transcriptional unit, whereas the TatA homologue is found in a separate locus. Specific mutations demonstrate membrane insertion of the P. denitrificans Rieske protein via the Tat pathway In order to analyse membrane insertion of the ISP in P. denitrificans, a number of mutants was generated. Individually and in combination, the invariant arginine residues of the consensus motif were conservatively exchanged for lysine. In addition, a Y20K mutation introduces the ‘+5’ lysine residue that is ‘missing’ in the ISP sequences. As export via the Tat pathway clas- sically requires previous cofactor-insertion in the cyto- plasm, a mutation C152S was introduced that conservatively replaces one of the cluster-binding lig- ands. Site-directed mutagenesis and cloning procedures were performed as described in the Experimental pro- cedures, and mutations were confirmed by sequencing. Wild-type and ISP mutants of the complete fbc operon coding for the three-subunit cytochrome bc 1 complex under control of its native promotor were cloned into a broad host-range vector and introduced into a P. denitrificans Dfbc::Km strain [28] via conjugation. Expression of the ISP subunit was probed by western blotting of whole-cell samples (not shown). For subcellular fractionation of P. denitrificans cells, a protocol originally developed for E. coli was adapted (Experimental procedures). To check the effectiveness of the process, three markers characteristic for each subcellular fraction were assayed (Table 1). Redox- difference spectra were recorded and the amount of soluble c-type cytochromes determined, these are Fig. 2. The signal sequence h-region in general Tat substrates is significantly less hydrophobic compared with Rieske proteins. Kyte–Doolittle plot comparing the hydropathy of ISPs and general Tat substrate proteins. Values fall within a range of +4 to )4, with hydrophilic residues having a negative score. Each data point represents an averaged hydropathy value derived from analysis of multiple sequences, as detailed in the Experimental procedures. A relative sequence numbering is given, with position 0 represent- ing the first invariant arginine residue of the consensus motif. Boundaries of the h-region are indicated as defined in Fig. 1. Rieske protein from Paracoccus denitrificans J. Bachmann et al. 4820 FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS normally found only in the periplasm, but not in the cytosolic fraction [29]. Enzymatic activities of cytosolic malate dehydrogenase and membrane-bound cyto- chrome c oxidase were the two other markers that allowed any cross-contamination to be assessed. The periplasm was isolated efficiently, as demonstrated by the high relative yield of c-type cytochromes given in Table 1. Within the periplasmic fraction, no malate dehydrogenase activity was detectable; this confirms that practically no cell lysis occurred during extraction of the periplasm. The enzymatic activites of malate dehydrogenase and cytochrome c oxidase show that there is little cross-contamination between the cytosolic and membrane fractions. However, good separation, as demonstrated here, could be achieved only after repeated ultracentrifugation. Small differences between the total activity in the nonfractionated cell lysate and the sum of the individual fractions can be easily explained by either loss of material or protein degrada- tion during the procedure. Taken together, the sub- cellular fractionation method applied here resulted in essentially quantitative separation with cross-contamin- ation of a few per cent at most. Localization of the ISP variants was analysed by western blotting of subcellular fractions derived from small-scale cultures of P. denitrificans in the exponen- tial growth phase (50 mL, D 600  1.5). The result of this experiment is given in Fig. 3. The deletion strain with the wild-type protein expressed in trans showed a dominant ISP signal in the membrane fraction; how- ever, substantial amounts of the protein could be found in the cytoplasm. It should be mentioned that the Rieske protein typically separates into two bands on SDS ⁄ PAGE; the lower band can be seen here only in the fractions with elevated ISP amounts. Exchange of the first invariant arginine (R15K) leads to a com- parable distribution, but the total ISP level is clearly diminished. In contrast, both the R16K mutation and the R15K⁄ R16K double mutation result in almost complete loss of the signal in the membrane fraction. The bands for the Y20K mutant resemble the wild- type, with slightly decreased levels. A much weaker sig- nal was obtained for the C152S mutant, which should abolish cofactor binding. This is in line with earlier observations by Davidson et al. [30]; they prepared an equivalent mutation in the closely related Rhodobacter capsulatus complex and observed a strongly decreased membrane level of the apo-ISP that was between one and two orders of magnitude less than the wild-type overproducer. Interestingly, a decrease or loss of the membrane-bound form in the mutant strains was not accompanied by ISP accumulation in the cytosol, pointing to rapid degradation of ISP that cannot be targeted to the membrane. Table 1. P. denitrificans cells are efficiently fractionated. Cell cultures at exponential growth (D 600 )1.5) were fractionated as described in the Experimental procedures. For each cell fraction, a marker protein was assayed. Periplasmically located c-type cytochromes were quantified using redox difference spectra. Activities of the cytosolic malate dehydrogenase and the membrane-integral cytochrome c oxidase were assayed as described in the Experimental procedures, and the total activities of the cell fractions were compared. Three separate fractiona- tions gave consistent results; values given here are from a single representative experiment. ND, values not determined. Marker Cell fraction (%) Cell lysate Periplasm Cytoplasm Membrane c-type cytochromes ND 95 5 nd Malate dehydrogenase activity 100 < 1 92 3 Cytochrome c oxidase activity 100 nd 1 95 Fig. 3. Specific mutations strongly inhibit membrane insertion of the P. denitrificans Rieske protein. Detection of ISP by western blotting in cell fractions from different P. denitrificans strains. Strain variants are indicated above the corresponding lanes, as follows: wt, P. denitrifi- cans bc 1 deletion strain MK6 expressing the wild-type fbc operon from plasmid pAN42; R15K, R16K, R15 ⁄ R16K, Y20K, and C152S, MK6 strain bearing pAN42 derivatives with the respective mutation(s) in the fbcF gene. Total amounts of 15 lg protein were loaded in each lane. The applied cell fractions are indicated: C, cytoplasm; M, membrane; periplasmic samples did not show any ISP signal and therefore are omitted. The Rieske protein typically separates into two bands on SDS ⁄ PAGE, as indicated by the two arrows; the lower band shows up only with higher protein amounts. J. Bachmann et al. Rieske protein from Paracoccus denitrificans FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4821 Cytochrome bc 1 enzymatic activity was assayed to obtain a more quantitative measure of decreased ISP amounts in the mutant membranes. To this end, we assume that mutations in the signal sequence do not exert an effect on the intrinsic activity of the enzyme; this seems plausible as the structures of mitochondrial homologues show that the N-terminal part of the ISP is far from the catalytic centres and is separated by the membrane dielectric in the native environment [3,4]. Obviously, this argument does not apply for the C152S mutant. Specific QH 2 :cytochrome c oxidoreduc- tase activities of membrane samples are given in Table 2. The data show that mutant membranes R16K and R15K ⁄ R16K were essentially inactive, whereas the R15K and the Y20K mutant membranes contained considerable amounts of fully assembled and active enzyme. As expected, the C152S mutant was fully inac- tive. For the Y20K mutant, an interesting observation was made when the membranes were subjected to sodium carbonate treatment prior to activity measure- ments. Although this treatment had only a minor effect on wild-type membranes, pretreated Y20K mem- branes showed severe instability when recording steady-state activities spectroscopically: traces with ini- tially normal slopes became a flat line a few seconds after the addition of substrate (data not shown). In conclusion, enzymatic data from Table 2 and the western blot results given in Fig. 3 provide a consistent picture with strong evidence for a Tat-dependent trans- location of the P. denitrificans ISP. Conservative exchanges of the twin-arginine motif for a lysine pair blocked membrane insertion; the single mutations on the arginine pair showed differential effects, with a more important role for the second arginine residue. As anticipated by the above sequence analysis, ‘restor- ation’ of the canonical Tat consensus with the Y20K mutant has a negative impact on membrane insertion; somewhat surprisingly, this effect is rather mild, and to a considerable extent the mutant ISP is still func- tionally incorporated into the cytoplasmic membrane. In addition, the effect of sodium carbonate treatment points to a structurally destabilizing effect of this mutation. Finally, removal of cofactor binding capa- city with the C152S mutant strongly impairs mem- brane insertion; this result might be explained by the postulated ‘cofactor-proof-reading’ operating in the Tat translocation scheme [11,31]. However, certain amounts of the apo-ISP are still found in the mem- brane; there is obviously no strict requirement for cofactor insertion prior to transport. Furthermore, a potential drawback of our data is that the effects of site-specific mutants are deduced only from the steady- state distribution of the ISP; no kinetic data for mem- brane translocation were obtained. Particularly in case of the C152S cofactor insertion mutant and the Y20K mutation in the ‘+5’ position, it is conceivable that secondary effects such as proteolytic degradation due to improper folding or an assembly defect of the bc 1 complex with subsequent proteolysis might also account for the reduced amounts of ISP in the mem- brane. Further experiments will be needed to fully exclude such side effects. The [2Fe)2S] cluster of the Rieske protein is inserted in the cytoplasm As cytosolic cofactor insertion is one of the hallmarks of Tat translocation, the cofactor loading status of the ISP in P. denitrificans cytosolic fractions was examined using electron paramagnetic resonance (EPR). Cells from a 0.5 L culture in the exponential growth phase were harvested; cytosolic fractions of the complemented wild-type, the export mutant R15K ⁄ R16K, and the cofactor binding mutant C152S were isolated and concentrated by ultrafiltration. Membranes from the complemented wild-type and the C152S mutant were used as positive and negative controls for the presence of the Rieske [2Fe )2S] clus- ter EPR signal. A reference spectrum was recorded with a purified sample of the Rieske protein fragment (ISF). EPR samples were reduced with 5 mm sodium ascorbate (Fig. 4). Spectrum I shows the reference spectrum of the purified ISF; the complemented wild-type in spec- trum II gave a clear Rieske signal in the membrane fraction, with shifted peak positions relative to Table 2. Cytochrome bc 1 activity reflects the decreased ISP con- tent in mutant membranes. Membranes were isolated from the parental fbc::Km deletion strain that was complemented with a wild-type copy of the fbc operon in trans or mutants thereof. Val- ues given are the average of three to five measurements. To elim- inate unspecific activity effects, activity data were corrected for the slope measured when the enzyme was inhibited with 10 l M anti- mycin A. Relative activity refers to the wild-type complemented strain. Strain Specific activity of membrane fraction (mUÆmg )1 ) Relative activity (%) fbc::Km deletion strain 3 < 1 Complemented wild-type 4018 100 R15K 1397 35 R16K 67 2 R15K ⁄ R16K 29 1 Y20K 929 23 C152S 12 < 1 Rieske protein from Paracoccus denitrificans J. Bachmann et al. 4822 FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS spectrum I. This shift most probably arises from h-bond interactions of a histidine cluster ligand with the quinone substrate bound to the membrane integral cytochrome bc 1 complex [32]. No [2Fe)2S] cluster sig- nal was visible in the cytosolic and membrane fractions of the C152S mutant, demonstrating: (a) the inability of the mutant to insert a cofactor, and (b) the specific origin of the signal in the other samples. The cytosolic fraction of the complemented wild-type clearly showed the EPR signature of the Rieske cluster (spectrum V), which provides strong evidence for the cytosolic assem- bly of the holoprotein. Likewise, the cytosol of the R15K ⁄ R16K double mutant contained the Rieske clus- ter, albeit at lower concentration compared with the wild-type cytosol (spectrum VI). In order to demon- strate the presence of the cluster in this strain more convincingly, the cytosolic fraction obtained from a 2.5 L culture was further enriched. It was applied to a Q Sepharose column, and the 150–250 mm NaCl salt gradient eluate was pooled and concentrated; the ISP is known to elute at  200 mm NaCl [33]. The enriched cytosolic fraction clearly shows the indicative Rieske spectrum (VII). The existence of the Rieske cluster in the cytosol of the R15K ⁄ R16K which is incompetent of membrane insertion (Fig. 3) clearly rules out the possibility that the signal could arise from membrane remnants in the cytosolic fraction. Furthermore, no Rieske cluster signal was observed in membrane samples of the R15K ⁄ R16K double mutant (not shown). Thus, the EPR results clearly demon- strate cytosolic cofactor insertion which is a typical feature of Tat substrate proteins. Discussion The aim of this study was to obtain experimental evi- dence of whether the Rieske [2Fe)2S] protein subunit of bacterial cytochrome bc 1 complexes is targeted to the cytoplasmic membrane by means of twin-arginine- dependent translocation. Studying the ISP from P. denitrificans with sequence analysis tools, site-direc- ted mutagenesis, and EPR spectroscopy, we found the key requirements of Tat translocation fulfilled: the characteristic features of the signal sequence, the export dependence on the conserved arginine pair, Fig. 4. The [2Fe)2S] cluster is inserted into the Rieske apoprotein in the cytoplasm. EPR spectra of: I, purified Rieske protein frag- ment (ISF); II, membranes of complemented wild-type strain; III, membranes of C152S mutant; IV, cytosol of C152 mutant; V, cyto- sol of complemented wild-type; VI, cytosol of R15K ⁄ R16K double mutant; VII, IEX chromatography-enriched cytosol fraction of R15K ⁄ R16K double mutant (150–250 m M NaCl eluate). Samples were reduced with 5 m M sodium ascorbate. For representation pur- poses, spectra are scaled differently on the y-axis: Spectra II–VII are magnified by a scaling factor of 500 relative to spectrum I. This scaling factor includes differences in sample concentration, spectral accumulation, and graphical scaling. Hence, spectral intensities do not reflect the concentration ratio between membrane and cytosol fractions. Peaks in the g x and g y region are indicated by vertical lines; the g z region is omitted due to overlaps with EPR signals from other proteins. The positions of g x signals in samples I and VII (g x1 ) are shifted to higher magnetic field because of the occurrence of ligand-free iron–sulfur protein. Conditions for EPR spectroscopy are given in the Experimental procedures. J. Bachmann et al. Rieske protein from Paracoccus denitrificans FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4823 and cytosolic cofactor insertion as a prerequisite for membrane targeting. To our knowledge, the only experimental evidence for Tat dependence of the bacterial ISP to date is the finding that a DtatBC deletion mutant of R. leguminos- arum lacks a functional bc 1 complex [34]. In contrast, insertion of the chloroplast ISP into the thylakoidal membrane via the Tat ⁄ DpH-pathway is well documen- ted [19,20]. This protein was the first Tat substrate shown to be an integral membrane polypeptide with a signal sequence that is not cleaved after translocation. Another interesting feature is the lack of the ‘invariant’ twin-arginine; instead, a KR sequence is found in the corresponding position. In contrast, cyanobacteria as supposed ancestors of chlorplasts contain ISPs with a perfect twin-arginine motif. It was argued that the RR to KR transition has a functional role in slowing import of the now nucleus-encoded ISP to allow for proper cofactor insertion in the stroma [19]. Associ- ation of the ISP with stromal chaperonin Cpn60 and ⁄ or Hsp70 was observed [19,35]. Furthermore, evi- dence was found for interplay with components of the Sec system and it was hypothesized that the ISP is an ‘intermediate’ substrate in the evolution of the chloro- plast export pathways [19]. The second conserved arginine residue plays a critical role in ISP translocation The relative amounts of ISP, detected by immunologi- cal means in the membrane fractions of the variant strains R15K, R16K, and R15K ⁄ R16K, show that both arginine residues are important for membrane targeting. However, the second arginine appears the most critical, and even a conservative mutation in this position leads to an essentially complete block, whereas replacement of the first arginine allows sub- stantial membrane insertion. This is a remarkable find- ing in the light of the naturally occurring KR motif in plant ISPs. This observation raises the question whether the Tat translocon is especially ‘permissive’ towards the Rieske protein, allowing for variation at the first arginine position, or whether the stricter role of the second arginine is a general feature of Tat sub- strates. Originally, an absolute requirement for both arginines was stated [11,36,37]. A gain-of-function mutant screen with a Tat-targeted GFP reporter con- struct, however, indicated that both positions tolerate variation, with the second position even being more flexible. Similarly, the E. coli multi-copper oxidase superfamily homologue SufI allowed single conserva- tive substitutions at both positions. In contrast to this, several authors report a critical role only for the second arginine [21,38–40]. Furthermore, apart from the plant ISPs, another example of natural ‘KR’ vari- ation of the Tat motif is known, interestingly, also in case of a protein carrying an iron–sulfur cofactor [21]. Taken together, these examples show that in some sig- nal peptides at least, conservative variation especially in the first position of the twin-arginine is possible, and this idea is corroborated by this study. The ISP Tat signal serves as a membrane anchor Sequence analysis of the N-terminal ISP portion is in good overall agreement with the structure of general Tat signal peptides. However, it shows some clear dis- tinctions that may account for its dual role as a Tat signal and as a membrane anchor. The ISP h-region exhibits a significantly higher hydrophobicity than average Tat signal peptides (Fig. 2). Likewise, it is well established that Sec signal peptides show higher hydro- phobicity than typical Tat signals. An engineered increase in hydrophobicity in a Tat signal peptide even leads to a (nonphysiological and functionally inexpedi- ent) re-routing of a precursor protein to the Sec trans- locon [23]. We wondered if the ISP h-region has a similar degree of hydrophobicity as the corresponding portion of Sec signal peptides. However, when we compared the h-region hydropathy values of the ISPs and a set of Sec substrates, we found the ISP h-region to be considerably less hydrophobic (not shown). Therefore, a clear ranking of hydropathy values becomes apparent, with Sec h-regions as the most hydrophobic, followed by ISP signal h-regions, which again are distinctly more hydrophobic than generic Tat h-regions. Together with the fact that we do not find a basic ‘Sec-avoidance’ signal [41] in the c-region, the elevated hydrophobicity of the h-region raises the intriguing question whether the bacterial ISP may also interact with Sec components. Evidence for such inter- play exists in case of the chloroplast ISP, where it was suggested that soluble components involved in Sec- targeting also deliver the Rieske protein to the Tat translocon [19]. It will be interesting to see if future experiments show a similar linkage in case of the bacterial ISP. As another critical determinant for membrane anchorage, the moderately hydrophobic residue at the ‘+5’ position was identified in silico, which was found in place of the consensus lysine in the ISP Tat motifs. Genetic substitution by the canonical lysine residue leads to slightly decreased ISP levels in the cytoplasmic membrane. This could be interpreted in line with the observation made by Stanley et al. [42] that a ‘+5’ lysine slows export of Tat substrates. Alternatively, Rieske protein from Paracoccus denitrificans J. Bachmann et al. 4824 FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS our results may be explained by secondary effects of the Y20K mutation which might destabilize the bc 1 complex and lead to proteolysis. At any rate, the hypothesis of these authors that this slowing has the physiological role of allowing for proper cofactor insertion is not substantiated here; the ISP is a cofac- tor-containing protein but lacks the ‘+5’ lysine in its native sequence. Probably, retardation is needed for other cofactor classes or in the case of heterodimeric proteins, where the ‘hitch-hiking’ subunit is granted time to associate with the subunit containing the Tat signal peptide [12]. Activity measurements with Y20K mutant mem- branes pretreated with carbonate show apparent rapid loss of cytochrome bc 1 activity; this can be tentatively interpreted as a less stable insertion of the ISP into the hydrophobic core of the enzyme complex. It has been a frequent observation that the Rieske subunit appears only poorly associated with the bc 1 complex, is easily lost during purification of the bc 1 complex, and can be extracted by high detergent concentrations or chao- tropic salt treatment [43]; before the crystal structure information emerged, it was still a matter of debate whether the ISP is a true integral membrane protein [44,45]. This weak association may be explained by the h-region hydrophobicity as given in Fig. 2, which, albeit being higher as for the Tat substrate average, is still rather low compared with TM helices of other membrane-anchored proteins. Possibly, the lacking ‘+5’ lysine, on the one hand, and the comparatively low hydrophobicity, on the other hand, represent a compromise between the conflicting requirements of TM helix formation and acceptance as a substrate by the Tat translocon. Recently, the crystal structure of a cytochrome bc 1 homologue, the cytochrome b 6 f complex from the cyanobacterium Mastigocladus laminosus was solved [7]. As discussed by Berks et al. [13], this enzyme also contains an ISP subunit with a putative Tat signal. Here, an asparagine residue is in the ‘+5’-position; sequence alignments of ISP subunits from bacterial b 6 f complexes show that for this subgroup asparagine is the most frequent amino acid in this position (data not shown), whereas the canonical lysine residue cannot be found, as in the case of bacterial bc 1 complexes. In the enzyme structure, no major interactions were found for the ‘+5’-Asn side chain. However, whereas the invariant Arg residues are located at the membrane– water interface, the ‘+5’-residue lies well within the TM region of the enzyme. It is a reasonable assump- tion that a Lys residue cannot be accommodated in this hydropohobic environment and is therefore absent from the Tat motif of bacterial ISPs. Potential steps during ISP biogenesis The observation that cofactor-containing periplasmic proteins carry a conserved twin-arginine motif led to discovery of the Tat translocation pathway [10]. Cyto- solic cofactor insertion is a key feature of Tat sub- strates and was experimentally confirmed by a number of studies for the bacterial and the homologous thylak- oid system, as reviewed by Berks et al. [11]. Our EPR data demonstrate the presence of holo-ISP in cytosolic fractions, thereby confirming that cluster assembly takes place in the cytosol. Impaired membrane inser- tion in the C152S mutant may indicate that cofactor insertion is a prerequisite for efficient export of the ISP. It was convincingly shown by different authors that a lack of the cofactor prevents export of Tat sub- strates [31,36]. However, because the kinetics of mem- brane insertion were not examined experimentally, the possibility exists that the apo-ISP is targeted to the membrane perfectly normally and only secondary pro- teolysis depletes the membrane fraction. It is an inter- esting observation that the cluster content of the R15K ⁄ R16K mutant cytosol was much lower than in the complemented wild-type (Fig. 4); for an export- deficient strain, rather an accumulation of the signal might be expected. It is therefore tempting to speculate on a potential interplay of the cluster insertion machinery and the Tat translocation process. With a closer look at the spectra in Fig. 4, it is remarkable to see that the chromatographically enriched cytosol frac- tion of R15K ⁄ R16K (spectrum VII) shows the same g-value positions as the purified reference protein (I), whereas cytosolic fractions taken directly for EPR measurements (spectra V + VI) resemble spectrum II from intact bc 1 complex where the [2Fe)2S] cluster histidine ligand is involved in hydrogen bonding inter- actions. Interaction with a quinone molecule in the cytosol appears quite unlikely; therefore, this observa- tion may give a hint to a putative binding parter of the ISP in the cytosol, probably playing some chaper- one role. The chromatographic purification step may have removed this binding partner. Further experi- ments are needed to examine this aspect in detail. A potential binding partner could be involved in cluster assembly; likely candidates are components of the Nif or Isc machinery responsible for iron–sulfur cluster assembly in bacteria and mitochondria [46,47]. Prelim- inary data from our laboratory indicate that overex- pression of the isc or nif operon can indeed promote [2Fe)2S] cluster assembly to the P. denitrificans ISP in the heterologous host E. coli [48]. In a BLAST search on the draft version of the P. denitrificans genome, putative genes homologous to those of the isc operon J. Bachmann et al. Rieske protein from Paracoccus denitrificans FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4825 from E. coli and the nif operon from A. vinelandii were identified (data not shown). Alternatively, ‘proof-reading’ chaperones may inter- act with the ISP in the cytosol. Recent evidence points to such specific chaperones acting on Tat sub- strates and preventing premature translocation [16,49]. However, such specific proof-reading chaperones were typically acknowledged as accessory genes in the operon context of the respective Tat substrate protein [40]. No such ORF of yet unknown function is pre- sent in the fbc operon coding for the cytochrome bc 1 complex of P. denitrificans. Also, chaperone binding is assumed to be associated with conserved sequence ele- ments in the signal peptide n-region [12]. We did not find such sequence conservation in case of bacterial ISPs; however, comparison of h-regions shows signifi- cant similarities among the different ISPs and might therefore be specifically recognized by a putative chaperone. An additional level of control for export competence is designated ‘quality control’ [17]. Here, the folding status of the protein is examined, presumably by the Tat translocon itself. From our data it is not clear whether cofactor ‘proof-reading’ or general ‘quality- control’ keeps the apo-ISP from being efficiently trans- located and inserted into the membrane; a third explanation for the lower membrane levels, as men- tioned above, is that the apo-ISP is normally targeted but subsequently degraded by periplasmic proteases. From the crystal structure of the homologous bovine soluble Rieske protein fragment (ISF), it seems plaus- ible that the apo-protein may adopt its almost terminal tertiary structure, as the cluster is bound only by a minor subdomain on top of the b-sandwich fold [50]. Furthermore, a CD spectrum of the heterologously expressed and refolded apo-ISF shows secondary structure features similar to the native holo-protein. However, native PAGE indicates a partially mobile or disordered structure for the refolded apo-ISF [48]. In addition, the Rieske protein carries a cystine bridge, and it was shown that various disulfide-containing pro- teins may be exported by the Tat pathway only under conditions in which a mutant strain provides an oxid- izing cytosolic environment [17]. However, it is reason- able to assume that disulfide bonds are formed in the periplasm, catalysed by homologues of the DsbA ⁄ B machinery. Therefore, we argue that the cystine bridge is not essential for an export-competent structure of the ISP. In summary, it seems that the apo-ISP can adopt its tertiary structure to a large extent and might well be accepted by the Tat translocon; however, disor- dered elements (the cluster binding subdomain) may hamper this process. Genetic inactivation of Tat machinery components in P. denitrificans will be an interesting goal for future experiments. Respiration is obligatory for this bac- terium [29]; however, a Tat-inactivated strain should by viable under oxic conditions, given the bioenergetic flexibility of P. denitrificans. The expected defect of the cytochrome bc 1 complex can be bypassed by the ba 3 quinol oxidase. If such a mutant can be obtained, it will certainly provide valuable information about the assembly of redox proteins in this important model system for the study of respiratory chains. Another interesting outlook is the identification of a putative cytosolic binding partner of the ISP, e.g. by using chemical cross-linking approaches and MS, giving interesting insights into the assembly of Rieske proteins. Experimental procedures Bioinformatic tools Protein sequences were obtained from Swiss-Prot pro- tein database (http://www.expasy.org/sprot); (a) bacterial Rieske proteins: R. rubrum P23136, R. capsulatus P08500, R. sphaeroides Q02762, R viridis P81380, B. japonicum P51130, P. denitrificans P05417; (b) [NiFe] hydrogenase small subunits: A. chroococcum P18190, A. hydrogenophilus P33375, B. japonicum P12635, R. capsulatus P15283; (c) MauM ferredoxins: M. extorquens Q49130, M. flagellatum Q50423, M. methylotrophus Q50235, P. denitrificans Q51659; (d) NapA periplasmic nitrate reductases: A. eutro- phus P39185, D. desulfuricans P81186, R. sphaeroides Q53176, P. pantotrophus Q56350; (e) NosZ nitrous oxide reductases: A. eutrophus Q59105, P. aeruginosa Q01710, P. denitrificans Q51705, P. stutzeri P19573, R. meliloti Q59746; (f) TorA trimethylamine-N-oxide reductases: E. coli P33225, R. capsulatus Q52675, R. sphaeroides Q57366. Multiple sequence alignments were performed using clustal x v. 1.81 [27]. For secondary structure prediction based on multiple alignments, the web server JPRED [51] (http://www.compbio.dundee.ac.uk) was used. Kyte–Doolittle hydropathy plots [25] were generated using an online tool from the ExPASy molecular bio- logy server (http://www.expasy.org/tools/protscale.html); the window size was set to 19 residues for comparison of ISP sequences and the comparison group of Tat- translocated proteins. Differences in hydropathy were statistically assessed with a two-tailed Mann–Whitney U-test (http://eatworms.swmed.edu/leon/stats/utest.html). To estimate the hydropathy of TM helices, a limited collection of P. denitrificans TM proteins was examined: cytochrome c 552 , cytochrome c 1 , cytochrome b and cyto- chrome c oxidase SU II. A set of 20 predicted Sec- exported proteins was obtained from the SPDb server Rieske protein from Paracoccus denitrificans J. Bachmann et al. 4826 FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS [...]... Iwasaki T, Crofts AR & Dikanov SA (2001) The interaction of the Rieske iron–sulfur protein with occupants of the Qo-site of the bc1 complex, probed by 1D and 2D electron spin echo envelope modulation J Biol Chem 277, 4605–4608 33 de Vries S & Cherepanov A (1998) Spectroscopic investigations on the water-soluble fragment of the Rieske [2Fe)2S] protein from Paracoccus denitrificans Inorgan Chim Acta 275–276,... c reductase from a bc1 subcomplex and the Rieske iron–sulfur protein isolated by a new method Eur J Biochem 132, 395–407 Breyton C, de Vitry C & Popot JL (1994) Membrane association of cytochrome b6f subunits The Rieske iron–sulfur protein from Chlamydomonas reinhardtii is an extrinsic protein J Biol Chem 269, 7597–7602 Gonzalez-Halphen D, Vazquez-Acevedo M & GarciaPonce B (1991) On the interaction... protein from Paracoccus denitrificans J Bachmann et al 50 mm Mops pH 7.5, 100 mm NaCl, 1 mm EDTA, 1 mm KCN, and 0.04% lauryl maltoside In order to assess the stability of the cytochrome bc1 complex of the Y20K mutant strain, membranes from the Y20K mutant and the wild-type (as control) were subjected to a sodium carbonate treatment prior to the enzymatic assay: membrane aliquots in 20 mm KPi pH 8 (protein. .. FPLC system (Pharmacia) Proteins were eluted with a 0–300 mm NaCl gradient; the eluate from 150 to 250 mm salt was pooled and concentrated by ultrafiltration (cut-off 5 kDa) to a final protein concentration of  200 mgÆmL)1 To obtain a reference Rieske [2Fe-2S] cluster spectrum, a soluble fragment of the Rieske iron–sulfur -protein was produced by limited proteolysis of the purified P denitrificans cytochrome... critical role for FEBS Journal 273 (2006) 4817–4830 ª 2006 The Authors Journal compilation ª 2006 FEBS 4829 Rieske protein from Paracoccus denitrificans 40 41 42 43 44 45 46 47 48 J Bachmann et al the second arginine within the twin-arginine motif Arch Microbiol 177, 107–112 Palmer T & Berks BC (2003) Moving folded proteins across the bacterial cell membrane Microbiology 149, 547–556 Bogsch E, Brink S &... Nucleic Acids Res 22, 4673–4680 Rieske protein from Paracoccus denitrificans 28 Korn M (1994) Doppeldeletion der Cytochrom c Oxidase und Reduktase (cta und fbc Operon) in Paracoccus denitrificans (Diploma thesis) Johann Wolfgang Goethe-Universitat, Frankfurt am Main ¨ 29 Baker SC, Ferguson SJ, Ludwig B, Page MD, Richter OMH & van Spanning RJM (1998) Molecular genetics of the genus Paracoccus: metabolically... periplasmic supernatant by centrifugation at 10 000 g Spheroplasts were resuspended in 50 mm KPi pH 7, 10 mm EDTA, 100 lm Pefabloc SC, and 0.1 mgÆmL)1 lysozyme and lysed by sonication Intact cells and cellular debris was removed by centrifugation (10 min, 10 000 g) The membrane fraction was separated from the cytoplasm by two successive ultracentrifugation steps (1 h, 125 000 g) The membrane pellet was... Yonetani A, Baron A, Bush DR, Cline K & Martienssen R (1997) Sec-independent protein translocation by the maize Hcf106 protein Science 278, 1467–1470 Molik S, Karnauchov I, Weidlich C, Herrmann RG & Klosgen RB (2001) The Rieske Fe ⁄ S protein of the cytochrome b6 ⁄ f complex in chloroplasts: missing link in the evolution of protein transport pathways in chloroplasts? J Biol Chem 276, 42761–42766 Finazzi... ice; it was checked that the phosphate buffer of the membrane samples did not interfere with the desired alkaline pH For the following enzymatic measurement, the pretreated sample was typically diluted 1 : 200 in assay buffer Optical spectroscopy c-Type cytochromes served as a periplasmic marker, as they exhibit characteristic redox difference spectra in the visible range For the oxidized spectra, 1... Paracoccus denitrificans PhD thesis, Johann Wolfgang GoetheUniversitat, Frankfurt am Main ¨ 4830 49 Hatzixanthis K, Clarke TA, Oubrie A, Richardson DJ, Turner RJ & Sargent F (2005) Signal peptide–chaperone interactions on the twin-arginine protein transport pathway Proc Natl Acad Sci USA 102, 8460–8465 50 Iwata S, Saynovits M, Link TA & Michel H (1996) Structure of a water soluble fragment of the Rieske . The Rieske protein from Paracoccus denitrificans is inserted into the cytoplasmic membrane by the twin-arginine translocase Julie Bachmann 1 ,. was argued that the ISP of Paracoccus denitrificans, a Gram-negative soil bacterium, is inserted into the cytoplasmic membrane via the twin-arginine translocation