Multiple Roles for the Twin Arginine Leader Sequence of Dimethyl Sulfoxide Reductase of Escherichia coli*

Dimethyl sulfoxide (Me2SO) reductase of Escherichia coli is a terminal electron transport chain enzyme that is expressed under anaerobic growth conditions and is required for anaerobic growth with Me2SO as the terminal electron acceptor. The trimeric enzyme is composed of a membrane extrinsic catalytic dimer (DmsAB) and a membrane intrinsic anchor (DmsC). The amino terminus of DmsA has a leader sequence with a twin arginine motif that targets DmsAB to the membrane via a novel Sec-independent mechanism termed MTT for membrane targeting and translocation. We demonstrate that the Met-1 present upstream of the twin arginine motif serves as the correct translational start site. The leader is essential for the expression of DmsA, stability of the DmsAB dimer, and membrane targeting of the reductase holoenzyme. Mutation of arginine 17 to aspartate abolished membrane targeting. The reductase was labile in the leader sequence mutants. These mutants failed to support growth on glycerol-Me2SO minimal medium. Replacing the DmsA leader with the TorA leader of trimethylamineN-oxide reductase produced a membrane-bound DmsABC with greatly reduced enzyme activity and inefficient anaerobic respiration indicating that the twin arginine leaders may play specific roles in the assembly of redox enzymes.

Recently, we and others have reported the discovery of a novel system, which targets and translocates folded, cofactorcontaining proteins to and across the cytoplasmic membrane of bacteria (1)(2)(3)(4). A similar system was previously identified in chloroplasts and some mitochondria (5). This system, named MTT (for membrane targeting and translocation) or TAT (for twin arginine translocation), 1 recognizes an amino-terminal leader sequence that is distinct from the one recognized by the well-characterized Sec system (1)(2)(3)(4). The MTT leader sequence is typically 30 -60 amino acids long, contains a conserved twin arginine motif, (S/T)RRXF(X/L)K, and has been found to be associated with a large number of periplasmic redox proteins (6,7). Although the role of this signal sequence in protein export was originally proposed for the periplasmic hydrogenase of Desulfovibrio vulgaris (8), understanding of this unique leader had to await the discovery of the MTT genes.
One of the twin Arg leader proteins identified in a data base search was the catalytic subunit of Me 2 SO reductase (6). This enzyme is a heterotrimer, composed of a molybdenum molybdopterin guanine dinucleotide containing a catalytic subunit (DmsA, 85.8 kDa), an iron-sulfur subunit (DmsB, 23.1 kDa), and a membrane intrinsic subunit (DmsC, 30.8 kDa) (9,10). The DmsA and -B subunits function as a catalytic dimer, whereas DmsC is the membrane anchor subunit and binds menaquinone (11). It was argued by Berks (6) that DmsA together with DmsB could be targeted for export via the MTT export pathway. However, extensive biochemical, immunological, electron microscopic, and electron paramagnetic resonance studies have been carried out on the topological organization of the Me 2 SO reductase (12,13). The two membrane extrinsic subunits, DmsA and -B were shown to face the cytoplasmic side of the membrane, contrary to the periplasmic localization of most other twin Arg leader-containing proteins.
The twin Arg motif of DmsA was not recognized in the initial cloning and sequencing of the dms operon (9). N-terminal analysis of the purified DmsA subunit revealed that the mature, purified protein began with the sequence Val-Asp-Ser . . . (see Fig. 1) and that the first upstream Met (16 residues upstream, M30 in Fig. 1) was proposed as the initiating Met (9). However, it now appears that DmsA may actually initiate at a Met, 46 residues upstream of the Val-46 (Fig. 1). This 45-amino acid leader encodes a twin Arg motif and has Ala-His-Ala immediately upstream of the cleavage site, in accord with other proteins containing this type of leader (6,14).
The electron transfer properties of the Me 2 SO reductase in Escherichia coli are well documented (15,16). However, very little is known about the role of the leader peptide. In the present study, we confirm that DmsA does indeed initiate at the more upstream Met (which we have now designated as M1, Fig. 1) We also examined the influence of the DmsA leader sequence on the assembly and stability of the enzyme, together with the membrane-targeting functions.
sham Pharmacia Biotech. All other reagents were of analytical grade.
Bacterial Strains and Plasmids-The following bacterial strains were used from our laboratory collection. TG1 (⌬(lac-pro) supE thi hsdD5/FЈ traD36 proAB lacI q lacZ DM15), DSS301 (as TG1; Km r ⌬dmsABC), and K38 (HfrC, Ϫ ). The plasmids carrying mutations in the DmsA leader sequence were generated using the Sculptor sitedirected mutagenesis kit (15). The following primers were used to construct the plasmids containing the native anaerobic promoter: pDMSM1 Stop (GTGAGCCATTtaaAAAACGAAAATCC), pDMSR17D (GGTGAGTCGCgaTGGTTTGGTAAAAAC), pDMSA43N:A 45N (TTA-GTCGGATTaatCACaaTGTCGATAGCG), and pDMSA43N (TTAGTC-GGATTaatCACAATGTCGATAGCG). The mutated residues are indicated here in lowercase letters. Plasmid pDMS160 (derived from pBR322 vector) served as the template DNA. The wild-type and mutant dms plasmids under the control of the T7 promoter were constructed as described earlier (17)(18)(19). Briefly, the DNA fragments carrying the dms operon (EcoRI/SalI digest) from pDMS160, pDMSR17D, and pDMSM1 Stop were cloned to pTZ18R as EcoRI/SalI inserts to yield the T7 promoter constructs pDMS223, pT7R17D, and pT7M1 Stop, respectively. E. coli strain K38 carrying the plasmid pGP1-2 (20) was transformed with the T7 promoter-driven dms constructs to study the expression of Dms polypeptides. Molecular biological procedures, such as restriction enzyme digestion, gel elution of DNA, ligation, filling in of the recessed ends of DNA after restriction digestion, and transformation of DNA into bacterial strains, were carried out as described (21).
Plasmid pTDMS10 has a deletion of the dms leader sequence up to and including the putative cleavage site at position 45 of the leader (Fig.  1). This corresponds to a 133-base pair (bp) deletion (nucleotide 808 to 940 of the dms operon map (9,18)) and was constructed using polymerase chain reaction (PCR). Two PCR products (A, 868 bp; B, 640 bp) were generated using the plasmid pDMS160 as the template and the primer pairs, BR322-R1 (CACGAGGCCCTTTCG)/TDMS3 (cccggggcta-gccagctgtctagacatatgccccatggATAATGGCTCACTCAAGC) for product A and TDMS2 (ccatggggcatatgtctagacagctggctagccccgggGTCGATAGC-GCCATTCC)/TDMS1 (CGCGTTTCGCCAGGG) for product B, respectively. The primer pairs TDMS3/TDMS2 encompass a 38-bp complimentary overhang (shown in lowercase letters) coding for the multicloning site: NcoI, NdeI, XbaI, PvuII, NheI, and SmaI. Except for PvuII, all other sites are unique. Introduction of the multicloning site codes for the following amino acid sequence and includes the initiating methionine: MGHMSRQLASPG. Following these 11 engineered amino acids is Val-46 of the mature protein (see Fig. 1).
PCR products A and B were gel purified, mixed, and subjected to a second round of PCR using primers BR322-R1 and TDMS1. The resulting composite PCR product (1471 bp) has the multicloning site replacing the leader sequence. The DNA fragment was gel-purified and subjected to EcoRI and EcoRV restriction digestion to yield a 1394-bp fragment. This DNA now lacking the entire leader was used to replace the native DNA sequence between the nucleotide positions 1 and 1491 of the plasmid pDMS160, at the EcoRI and EcoRV restriction sites, respectively, to generate pTDMS10 ( Fig. 1). Plasmids pDMS160 and pTDMS10 harbor the native anaerobic promoter and the rest of the dms operon.
Construction of Plasmids Carrying the dms Genes under Control of the lac Promoter-Plasmid pTDMS10 was digested with NcoI and EcoRI, and the fragments were separated on a 1% agarose gel in a TAE (40 mM Tris acetate and 2 mM EDTA) buffer system. The larger fragment (7762 bp) containing the entire dms operon (less the leader) was gel-eluted. The smaller NcoI/EcoRI fragment (805 bp) containing the native anaerobic promoter was discarded. The 205-bp fragment containing the lac promoter was obtained by a PCR reaction using the DS-18 (atatgaattcGAATTAATTCTCACTCATTAGG)/DS-19 (CTAGAA-GAAGCTTGGGATC) primer pair and pYZ4 DNA (pBR322-derived vector) as a template (22). The PCR product was digested with EcoRI and NcoI and ligated to the gel-eluted 7762-bp fragment to generate pDMS189.
Plasmids pDMS190 and pDMS191 were derived from pDMS189 and harbor the DmsA and TorA leader sequences, respectively, upstream of mature DmsA. The DNA sequence for the dms and torA leaders were obtained by PCR, using the primers DS-11 (atatccatggtgAAAAC-GAAAATCCCTGATG)/DS-12 (atggatccagctgGACAGCGTGCGCAATC-CG) and SL-S2 (atatccatggtgAACAATAACGATCTCTTTC)/SL-A2(atg-gatcccagctgCGAGATGACAGCGTCAGT), respectively. The plasmid pDMS160 served as a template DNA for DS-11/DS-12 PCR, and chromosomal DNA was used as a template for amplifying the torA leader. The PCR DNA products have NcoI/PvuII flanking ends and were cloned in to NcoI/SmaI-cut pDMS189 to generate plasmids pDMS190 and pDMS191.
All constructs were verified by restriction digestion analysis of the mini DNA preparations and DNA sequencing reactions to confirm the mutations or deletions in the dms operon.
Media and Growth Conditions-Aerobic overnight bacterial cultures were grown at 37°C unless indicated otherwise, in Luria-Bertani (LB) medium supplemented with the required antibiotics (11). A 1% inoculum was used from these overnight cultures for all experiments, unless indicated otherwise. Anaerobic growth experiments in glycerol minimal media were carried out in 160-ml screw capped conical flasks with a Klett tube attachment (Klett flasks) for direct monitoring of the bacterial growth as a function of time. The minimal medium was supplemented with casamino acids, vitamin B1, terminal electron acceptor, ammonium molybdate, and antibiotics, as described (11). Me 2 SO served as the terminal electron acceptor for the growth experiments in the Klett flasks.
For the measurement of Me 2 SO reductase activity, the cells were grown anaerobically in 1000-ml conical flasks at 37°C for 48 h on glycerol-fumarate minimal medium (G-F medium), for 24 h in peptonefumarate medium, or for 8 h in Terrific broth (TB) medium (21). Me 2 SO reductase was expressed constitutively under the above anaerobic growth conditions. Cells were harvested and treated as described earlier for the preparation of the membrane and supernatant (periplasm plus cytoplasm) fractions (11).
Gel Electrophoresis and Western Blotting-Proteins were fractionated on a 7.5% or a 12% resolving polyacrylamide gel using a Bio-Rad mini slab gel apparatus with an SDS buffer system (23). Proteins from SDS-polyacrylamide gel electrophoresis (PAGE) gels were transferred to nitrocellulose filters and probed with the anti-DmsA and anti-DmsB antibodies as described (11). Western blot analysis of whole cells was performed essentially as above, except the cell pellets were directly lysed in the SDS solubilization buffer and heated in a boiling water bath for 2 min before loading on SDS-PAGE gels.
T7 Expression Studies-Strain K38/pGP1-2, carrying pDMS223-encoding dmsABC or various mutants, was grown overnight in LB medium at 30°C. A 5% inoculum of this culture was transferred into fresh LB medium (6.0 ml) and grown for 2 h, and a 1.0-ml aliquot was drawn for the zero time sample. To the remaining 5.0 ml of bacterial culture, was added fresh LB medium (5.0 ml) at 54°C, to facilitate rapid temperature equilibration of the culture to 42°C. The incubation was continued for another 15 min in a water bath set at 42°C to induce T7 RNA polymerase. Rifampicin (at 250 g/ml of culture), an inhibitor of chromosomally encoded RNA polymerase, was added, and the incubation was continued at 42°C for 15 min. All subsequent incubations were at 37°C (20). Aliquots (1.0 ml) were drawn at various time points and spun in an Eppendorf centrifuge, and the resulting bacterial pellets were resuspended in 50 l of the SDS solubilization buffer. The solubilized pellets were heated in a boiling water bath for 2 min and spun to separate the insoluble material, and a 5-l aliquot of the clear supernatant was loaded per lane onto an SDS-PAGE gel.
Analytical Techniques-DNA sequencing was carried out using an Applied Biosystems model 373A DNA sequencer at the DNA core facility in the Department of Biochemistry, University of Alberta, Edmonton, Canada.
All the data reported here represent at least two independent experiments, and the results from a representative experiment are presented.

Investigation of the Translational Start Site of the DmsA Leader Sequence and Its Role in Expression of DmsAB
Polypeptides-When the dmsABC sequence was reported (9, 10), we proposed that DmsA initiated at a methionine, 16 residues upstream of a valine, the first amino acid of the purified DmsA subunit (M30 in Fig. 1). Subsequent bioinformatic analysis of the dms sequence by Berks (6) suggested that DmsA actually initiated at a methionine, which is 45 residues upstream of the valine (M1, Fig. 1) and contained a consensus twin Arginine motif. To determine if M1 was the initiating Met, we mutated this start codon to the stop codon TAA, and expressed the construct (pT7M1 Stop) using a T7 promoter system. Following expression, the proteins were probed by Western blotting with anti-DmsA and anti-DmsB antibodies. As shown in Fig. 2, vector pTZ18R did not express DmsAB, whereas pDMS223 (wild-type dmsABC) showed good expression of the DmsAB subunits. The DmsAB polypeptides were observed as early as 30 min and peaked at 2 h. The expression from the pT7M1 Stop plasmid showed that the intensity of the DmsA polypeptide was drastically reduced with limited expression at 30 and 60 min followed by rapid degradation. Expression of the DmsB polypeptide was relatively unaffected, even up to 21 h. These results indicated that DmsA had a long 45-amino acid leader, which contained a consensus twin arginine motif, and that an internal Met-30 could not serve as an efficient initiating methionine ( Fig. 1). Under these experimental conditions, DmsB could not protect the truncated form of DmsA originating at the putative second initiating Met-30 in the pT7M1 Stop construct (Fig. 1)

from degradation.
Role of the Conserved Arginine in the DmsA Leader Sequence-The two arginines of the twin Arg consensus sequence have been shown to be highly conserved within the family of polypeptides exhibiting the motif (Fig. 1, R16 and R17). These residues have been shown to be important for the expression and/or the function of the polypeptides that rely on the MTT system for protein targeting and transport (14,26,27). We mutated Arg-17 in the DmsA leader to an Asp (R17D) and examined the effect on the expression and stability of DmsA in the T7 system as described above (Fig. 2). Western blot analysis revealed that DmsAB were expressed to levels comparable to the wild-type, and the amount of DmsA was only slightly reduced at 21 h. compared with the wild-type ( Fig. 2;  pT7R17D). These results indicate that the R17D mutation has no adverse affects on the expression or stability of the DmsAB polypeptides under the conditions of the T7 expression and analysis of the whole cell lysates by Western blotting. The expression of the DmsC polypeptide could not be analyzed in these experiments due to the lack of the corresponding antibody (12).

The DmsA Leader Sequence Is Essential for Anaerobic
Growth on Me 2 SO-Anaerobic growth with glycerol as carbon and energy source and Me 2 SO as the terminal electron acceptor (G-D medium) provides a simple means of monitoring the physiological function of DmsABC (11). E. coli DSS301, deleted for the chromosomal dms operon, was unable to support growth on G-D medium (11). The wild-type plasmid, pDMS160, carrying the entire dms operon, restored the anaerobic growth of this strain (Fig. 3A). As expected, pDMSM1 Stop could not complement growth. Similarly, the construct lacking the entire leader (pTDMS10) could not complement growth. Although pDMSR17D produced the DmsAB (Fig. 2, and presumably the DmsC) polypeptides, this construct was unable to complement growth of DSS301. A mutant at the putative cleavage site in which alanine 43 was changed to asparagine (pDMSA43N) and a double mutant in which both the alanines at 43 and 45 were changed to asparagine (pDMSA43N:A45N) showed normal growth profiles (Figs. 1 and 3A). We investigated replacement of the dms leader with the TorA leader of inducible TMAO reductase. We used lac promoter constructs, pDMS189, pDMS190, and pDMS191 to facilitate the interpretation of the leader substitution (or deletions) while keeping the promoter background constant. The TorA leader has a twin Arg motif and translocation of TMAO reductase to the periplasm is dependent on the MTT system (2, 3). Plasmid pDMS191, with a torA leader in place of the dmsA leader, showed poor growth (Fig. 3B). These studies indicate that an intact dmsA leader was essential for the production of functional enzyme. Examination of the control lac promoter constructs (Fig. 3B) indicated that pDMS189, lacking a dmsA leader, could not support growth, similar to its parent pT-DMS10 (Fig. 3A). The wild-type lac promoter-dmsA leader construct (pDMS190) supported good growth even in the absence of added inducer.
The Role of the DmsA Leader on the Targeting, Assembly, and Activity of Me 2 SO Reductase-To probe the enzymatic activity of the leader mutants described above, we measured the Me 2 SO-dependent oxidation of benzyl viologen (BV) as a measure of DmsAB function and lapachol-catalyzed reduction of Me 2 SO as a measure of the DmsABC in the membrane and soluble fractions from E. coli DSS301 cells grown on minimal glycerol-fumarate medium (Table I). Me 2 SO reductase activity was localized predominantly to the membrane fraction in cells harboring pDMS160, pDMS190, and the cleavage site mutants pDMSA43N and pDMSA43N:A45N. The activity data correlated well with the anaerobic growth experiments (Fig. 3).
Membrane vesicles from E. coli strain DSS301/pDMSM1 Stop expressed very low levels of BV and lapachol activities. The BV activity observed was predominantly in the soluble fraction (Table I), in agreement with the growth measurements. Similarly, there was no detectable BV or lapachol enzyme activity when the leader was deleted (pTDMS10). In this mutant, the leader was replaced with a multicloning site that encodes 11 amino acids, including an initiating Met at the amino terminus of DmsA unrelated to the DmsA leader (Fig. 1). These studies indicate that the leader sequence is an absolute requirement to sustain the anaerobic growth, enzyme activity, stability, and localization of the Me 2 SO reductase holoenzyme ( Fig. 3 and Table I).
Strain DSS301/pDMSR17D with a mutation in the second Arg of the twin Arg motif expressed DmsA (Fig. 4B), but this enzyme did not associate with the membrane and only very minimal levels of enzyme activity were detected in the soluble fraction. Although nearly normal levels of DmsA appear to accumulate following T7 expression (Fig. 2), we have found that under the standard growth conditions (peptone-fumarate medium for 24 h or on glycerol-fumarate medium for 48 h; data not shown), the precursor form of DmsA is degraded. As will be shown below (Fig. 4B), mutant precursor enzyme from pDMSR17D was poorly processed to the mature form.
We have examined the expression of dmsABC using the lac promoter construct (pDMS190). In minimal medium pDMS190 expressed similar total activity units in the absence of an inducer (isopropyl-1-thio-␤-D-galactopyranoside) compared with pDMS160. The membrane distribution of the BV and lapachol activities was also similar to pDMS160 (Table I). We also compared the activity profiles of the various constructs grown on rich medium (TB) for 8 h (Table II). For the wild-type pDMS190, 97% of the activity was found in the membrane fraction.
E. coli/pDMS189, lacking a leader, had only 6% of the wildtype activity, and 76% of this activity was soluble. DSS301/ pDMS191 with a TorA leader in place of the DmsA leader, had about 20% of the wild-type activity, and 90% was membranebound, corroborating the slow anaerobic growth seen on glycerol-Me 2 SO medium. These results indicate that the twin arginine leader is required for stability and for the association of DmsAB with the membrane.
Western Blot Analysis of the DmsA Subunit from the Wildtype and Signal Sequence Mutants-To confirm the role of the leader in membrane targeting and stability of DmsA, we used Western blot analysis of membrane and soluble fractions from cells harboring various plasmids encoding Me 2 SO reductase. The distribution of DmsA with the wild-type and mutant leaders was examined from cells grown anaerobically on G-F medium for 48 h to corroborate the activity measurements summarized in Table I. DmsA expressed by DSS301/pDMS160, pDMS190, pDMSA43N, and the double mutant pDMSA43N: A45N was observed predominantly in the membrane fraction (Fig. 4A). The distribution of DmsA reflected the distribution of specific activity (Table I). Cells harboring plasmids pDMSM1 Stop (Fig. 4A) and pTDMS10 (data not shown) showed no immunoreactive material in either the membrane or soluble fractions.
Analysis of the DmsA from DSS301/pDMSR17D mutant was carried out from cells grown on TB medium for 8 h (Fig. 4B). Cells grown on G-F or peptone-fumarate media did not show any immunoreactive DmsA presumably due to degradation (data not shown). However, analysis of the cell fractions derived from cells grown in rich medium revealed three immuno- These results are in agreement with the hypothesis that the DmsA leader is required for the correct targeting and stability of the holoenzyme.
In Fig. 4B we also analyzed the membrane and soluble fractions for lac promoter-controlled expression of DmsABC. The DmsA immunoreactive bands were observed predominantly in the membrane fraction for E. coli DSS301/pDMS190 (wild-type leader) and pDMS191 (TorA leader). A very small amount of the D-form of DmsA was observed in the soluble fraction of pDMS191. There was no detectable immunoreactive material in the membrane or the soluble fraction from the leaderless construct, pDMS189, consistent with the distribution of the enzyme activity data (Table II). DISCUSSION The MTT system plays an essential role in the translocation of severalfolded, cofactor-containing, redox proteins to the periplasm (1-3). These proteins have long amino-terminal leader sequences with a twin Arg motif and are usually soluble or bound to the periplasmic side of the membrane. The DmsABC and FdoGHI enzymes are membrane-bound, multisubunit proteins, and the DmsA and FdoG subunits have similar twin Arg signal peptides. However, a great deal of experimental evidence has shown that DmsAB and FdoGH subunits face the cytoplasmic side of the membrane and are not translocated across the membrane (12,28). We have proposed that the MTT system is needed to associate the DmsAB subunits with the membrane anchor subunit, DmsC (1,11).
In this study we provide experimental evidence for the presence of a 45-residue twin Arg leader on the DmsA subunit. We also show that the leader is essential for the stability and function of the holoenzyme. This contrasts with the periplasmic TMAO and Me 2 SO reductases from E. coli and Rhodobacter sphaeroides (2,29). In these enzymes, deletion of the twin Arg leader resulted in cytoplasmic localization of the enzyme, consistent with the predicted role of the leader peptide (2,3,29); however, the leaderless enzymes were stable and incorporated the molybdopterin cofactor.
The DmsA leader has two methionine residues, Met-1 and Met-30 (Fig. 1). Met-30 was found to be a poor initiation site and led to minimal synthesis of the reductase (Fig. 2). This Met is not conserved in DmsA homologues (6). The confirmation of a 45-residue leader necessitates renumbering of amino acid residues in DmsA. Previous publications have assumed that DmsA initiated at Met-30, based on the original sequence study (9), and the first residue of the mature protein, Val46, was previously numbered Val-16.
Mutagenesis studies of the conserved Arg, within the twin Arg motif, were performed on the [Ni-Fe] hydrogenase from D.   vulgaris, the nitrous oxide reductase from the Pseudomonas stutzeri, and the glucose-fructose oxidoreductase from Zymomonas (14,26,30). In all cases the mutant enzymes were not translocated to the periplasm but were stable. The R17D mutant of Me 2 SO reductase was highly unstable, was not targeted to the membrane, and did not support anaerobic growth (Table  I and Fig. 3). These studies suggest that the twin Arg leader could have two distinct functions: targeting and stability. The twin Arg motif in DmsA differs from the consensus twin Arg leaders by having a Leu in place of a Phe in the fourth position, however, it is unlikely that the targeting versus translocation role is related to this variation. When we replaced the DmsA leader with the TorA leader, which has a Phe in place of Leu, DmsA was still targeted to the membrane. The TorA leader-DmsA construct appeared to be far less stable than the DmsA leader-DmsA construct, suggesting that there are unique interactions between the leader and mature protein.
The twin Arg leader is cleaved by a peptidase (leader peptidase I?), and we have examined the processing of DmsA by mutating the conserved amino acids adjacent to the putative cleavage site. Nearly all of the twin arginine leaders have a conserved AXA motif amino-terminal to the cleavage site, and cleavage at this site was confirmed for a number of enzymes (6,14). The DmsA leader also has the AXA motif (Fig. 1), and the N-terminal analysis of the purified DmsA polypeptide showed cleavage at this site (9). Both the cleavage site mutants (pDMSA43N and pDMSA43N:A45N) exhibited enzyme activity, were processed to the M-form, supported anaerobic growth on G-D medium, and had the expected membrane localization of activity. These mutations were chosen based on algorithms of von Heijne (31), which suggested that Asn residues were least preferred by leader peptidase I. Data in Fig. 4A show that the DmsA43N:A45N protein migrates as mature DmsA, suggesting that cleavage is catalyzed by a protease that is not inhibited by Asn or a nearby alternate cleavage site is being utilized. These results lead to a conundrum. If DmsA is not translocated to the periplasm, when and where is it cleaved? Although in earlier papers we suggested that cleavage may be an artifact of cell lysis and protein purification (9,10), it now appears that cleavage is an essential step to maturation of functional the enzyme. 2 The twin Arg leader is able to export both homologous and heterologous proteins. Reporter proteins such as dihydrofolate reductase, ␤-lactamase, alkaline phosphatase, and the 23-kDa protein of the plant photosystem II oxygen-evolving complex were exported and processed to the mature forms when expressed with a TorA signal peptide (3,14,32,33). We have replaced the DmsA leader with the TorA leader (pDMS191) and investigated the membrane assembly and function of the TorA-Me 2 SO reductase holoenzyme (Table II, Figs. 3 and 4). Although the TorA leader behaved similarly to the DmsA leader in its ability to target the holoenzyme to the membrane, nearly a 5-fold decrease in specific activity was noted for the mutant enzyme, compared with the wild-type enzyme. The mutant enzyme did not support efficient anaerobic growth on glycerol-Me 2 SO minimal medium. The incompatibility of the TorA leader supports our contention that the DmsA leader has important stabilizing properties during maturation of the DmsABC holoenzyme. Interestingly, when the TorA leader was fused to the soluble Me 2 SO reductase from R. sphaeroides, the enzyme also failed to show optimal enzyme activities, implying that the leader has a role in the overall folding and function of the enzyme complex (29). These studies show that twin Arg leaders may play specific roles in the assembly of redox enzymes and may not be interchangeable.