Identification of active-site cysteines in the conserved domain of PilD, the bifunctional type IV pilin leader peptidase/N-methyltransferase of Pseudomonas aeruginosa.

PilD is a bifunctional enzyme responsible for cleavage of the leader peptides from the precursors of the type IV pilin and four proteins with type IV pilin-like amino termini that are required for extracellular protein secretion in Pseudomonas aeruginosa. Following cleavage, PilD also catalyzes the second major posttranslational modification of these proteins, namely the N-methylation of the amino-terminal phenylalanine residues of the mature polypeptides. In this report, we demonstrate that the enzymatic activities of PilD involve cysteine residues that lie within a cytoplasmic domain that shows a high degree of similarity to other proteins postulated to perform the same function in other bacterial species. Both activities are reduced in the presence of sulfhydryl-reactive reagents such as N-ethylmaleimide and p-chloromercuribenzoate. Mutagenesis of pilD resulting in specific amino acid substitutions in all of the Cys residues in PilD show that the 4 conserved cysteines in the cytoplasmic domain are required for full peptidase activity in vivo and for complete peptidase and methyltransferase activities in vitro. Conversely, substitution for a Cys residue in a membrane spanning domain had no effect on PilD activities in vivo or in vitro. Evidence suggests that the peptidase and methyltransferase sites of PilD are adjacent, with the Cys residues in the cytoplasmic domain important for methyl donor binding, as prior reaction of PilD with the S-adenosyl-L-methionine analogue sinefungin afforded complete protection of peptidase activity from inactivation with N-ethylmaleimide.

by similar, short leader peptides and by extensive sequence conservation within the first 10-16 amino acids of the mature protein. The consensus sequence that forms the PilD recognition site is -Gly-'1-Phe-Thr-Leu/Ile-Glu-, where ' 1 is the site of leader peptide cleavage. Following removal of the leader peptide, the amino-terminal Phe residue in type IV pilins and the Pdd proteins is methylated at the new amino terminus. N-Methylation is an unusual posttranslational modification of prokaryotic proteins demonstrated previously for only a handful of polypeptides in Escherichia coli, including CheZ (chemotaxis chemoreceptor), IF3 (initiation factor in protein synthesis), and the ribosomal proteins L16, S11, and L11 (Stock, 1988). This posttranslational modification appears to be essential for polymerization of the pilin monomers into pili, as mutations in the precursor which allow proteolytic processing but block this modification are not assembled into the mature organelle (Pasloske and Paranchych, 1988;. Recently, we have demonstrated that PilD is a bifunctional enzyme, and as such, it not only catalyzes the removal of the leader peptides of its target substrates, it is also responsible for the subsequent N-methylation of the amino-terminal phenylalanine residue in both pilin and PddD precursor substrates (Strom et al., 1993). Type IV pili are found in a wide variety of Gram-negative pathogens, including those of Neisseriagonorrhoeae (Meyer et al., 1984), Moraxella bouis (Marrs et al., 1985), Moraxella nonliquifaciem (Tonjum et al., 1991), Dichelobacter nodosus (McKern et a i , 1983), and Vibrio cholerae (Shaw and Taylor, 1990). The pilin subunits in this class all have N-methylated residues at the amino terminus of the mature protein. We have also shown previously that PilD both efficiently cleaves the N. gonorrhoeae pilin precursor and methylates the mature protein (Strom and Lory, 1992;Strom et al., 1993), and it is likely that the type IV peptidases in these organisms have a bifunctional role as well.
Homologues of PilD have been isolated in a number of other genera. These include TcpJ in V. cholerae, in which mutations prevent the proteolytic cleavage of the leader peptide of the toxin-coregulated pilin (Kaufman et al., 1991), and PulO, which is necessary for extracellular secretion of puilulanase in Klebsiella onytoca (Pugsley and Reyss, 1990). For the latter, PulO has been shown to be involved in the proteolytic processing of PulG, one of four proteins (PulGHIJ) in the pullulanase secretion pathway with amino-terminal sequences that contain the type IV peptidase consensus cleavage site (Pugsley and Dupuy, 1992). These four proteins are homologous counterparts to the P. aeruginosa secretion proteins PddABCD (Nunn and Lory, 1992). PulO has also been shown to cleave correctly the leader peptide from N . gonorrhoeae prepilin in vivo (Dupuy et al., 1992). ComC from Bacillus subtilis is also a PilD homologue and is a necessary This alignment was done using MACAW from the National Center for Biotechnology Information (Schuler et aL, 1991), available by anonymous ftp from ncbi.nlm.nih.gov over the Internet). The three blocks of highest functional homology are shown as white capital letters on black. The locations of the conserved Cys residues are marked with *, and the location of the fusion junction between PilD and LacZ (see Fig. 2) is shown by V.
component for the competence state during B. subtilis sporulation (Mohan et al., 1989). The amino acid homologies of PilD, TcpJ, PulO, and ComC are shown in Fig. 1. Recently, the PilD homologue from N. gonorrhoern has been cloned and sequenced and shown to have this same high level of homology (Lauer et al., 1993).
During the initial characterization of the peptidase activity of PilD, Nunn and Lory (1991) noted that the enzyme was sensitive to sulfhydryl-blocking reagents and that the cleavage activity of purified PilD was stimulated by dithiothreitol. Moreover, a cluster of cysteines is found within a conserved region shared with other type IV peptidases (Fig. 1). This suggested that the enzyme may have cysteines in or near its active site.
To determine the involvement of the Cys residues in the enzymatic activities catalyzed by PilD, the inhibitory effects of several sulfhydryl-reactive reagents on peptidase and methyltransferase activity were tested and compared. The cloned pilD gene was also subjected to oligonucleotide-specific sitedirected mutagenesis to replace individually the Cys residues with Ser and/or Gly residues. The mutated forms of PilD were able to process pilin that was subsequently assembled into pili. However, the mutations in the Cys residues at positions 72, 75, 97, and 100 caused a substantial decrease in peptidase specific activity, with a concomitant loss of methyltransferase activity.

MATERIALS AND METHODS
Bacterial Strains, Plasmids, and Reagents-All bacterial strains and plasmids used in this study are described in Table I. All restriction enzymes and T4 DNA ligase were purchased from Life Technologies, Inc.
PilD Enzyme and Pilin Substrate Preparation-P. aeruginosa PAK-DR containing either the wild-typepiZD plasmid pRBS-L or the mutated clones were grown in 500 ml of minimal A medium (Davis and Mingioli, 1950) supplemented with 50 mM monosodium glutamate, 1% (v/v) glycerol and containing 100 pg/ml carbenicillin and 200 pg/ml IPTG and total cell membranes extracted as described previously (Strom and Lory, 1992). The concentrated membranes were initially resuspended to a total protein concentration of about 25 mg/ml in 25 mM triethanolamine, pH 7.5, 10% glycerol; PilD in these preparations represented approximately 1-5% of the total protein as judged by Western immunoblots. The concentration of PilD in the membrane preparations was then determined by comparing dilutions of each with purified PilD of known concentration on Western immunoblots. These results were then used to normalize the concentration of PilD in each preparation to equal that of purified PilD, 1 pglpl. Approximately 5 ng of PilD could be detected by our antisera in these assays.
Prepilin for the peptidase assays was obtained by overexpressing the piol gene, from the tac promoter in pMStac27PD, in PAK-2B18, followed by purification as described previously (Strom and Lory, 1992). Purified prepilin was resuspended and stored in water to a concentration of 10 mg/ml.
Previously, we have shown that the rate of methylation catalyzed by PilD is comparable whether the substrate is prepilin or already processed pilin with the leader sequence previously removed by in vitro cleavage with PilD (Strom et al., 1993). Therefore, to measure the differences in methylation ability between wild-type and mutated PilD separate from the differences seen in cleavage rates, prepilin was first precleaved to completion with purified PilD. This preparation was then treated with SDS to 2%, which irreversibly inactivates PilD , and precipitated with 4 volumes of acetone. After centrifugation, the pilin was resuspended to the original 10 mg/ml concentration prior to use in the methyltransferase assays.
Enzyme Assays-Enzyme activity assays were performed using total membrane extracts because previous work has shown that in this form, PilD catalyzed purified prepilin cleavage at a 50-fold higher rate than did the immunoaffinity-purified enzyme (Strom and Lory, 1992). All of the assays were carried out in triplicate for each enzyme preparation and dilution.
Peptidase activity of PilD was assayed as described, with typical 10-pl reactions containing enzyme, substrate, and 0.05% cardiolipin in a solubilization or "start" buffer with a final concentration of 25 mM triethanolamine, pH 7.5, 0.5% Triton X-100. The latter is added from a 5 X stock after all other components are mixed. For determination of specific cleavage activities of wild-type and mutagenized PilD, serial 2-fold dilutions of membranes containing PilD were made and used to cleave 10 pg of purified prepilin in 10-pl reactions carried out at 37 "C for 15 min. Negative controls contained membranes isolated from the pilD mutant P. aeruginosa PAK-DR carrying the cloning vector pMMB66EH but with all other reactants. Reactions were stopped with the addition of 20 p1 of electrophoresis buffer containing 0.125 M Tris-HC1, pH 6.8, Pilin in each reaction was then electrophoresed on SDS-Tricine-15% polyacrylamide gels to separate precursor and processed forms (Schaegger and von Jagow, 1987), and after staining with Coomassie Blue, the percentage of prepilin converted to the processed form was determined by densitometry. One unit of PilD activity was calculated as the amount necessary to convert 50% of 1 pg of prepilin to the processed form in 1 h. From these experiments, relative specific cleavage activities of wild-type and mutated PilD, as measured in total membrane extracts, were obtained. The limit of detection of prepilin cleavage as measured by densitometry was approximately 0.5%.
For determination of relative methyltransferase activities, a dilution of membranes calculated to contain about 1 pg of PilD (31.2 pmol) was incubated with 10 pg of precleaved pilin and 5 pCi of [3H] AdoMet in a 10-pl reaction containing 5 mM dithiothreitol, 0.5% Triton X-100 (v/v), and 100 mM triethanolamine, pH 7.5. These were incubated at 37 "C for 1 h, after which an aliquot was removed and the protein precipitated with the addition of trichloroacetic acid to 10%. The precipitated protein was collected on 0.22-pm cellulose filters (Millipore GS), followed by a wash with 5% trichloroacetic acid. The filters were then dissolved in 5 ml of scintillation fluid (Liquiscint, Du Pont-New England Nuclear) and counted for 3H in a scintillation counter. Relative specific activities of the mutants were lac2 fusion vector, in pMMB67HE pilD-lac2 fusion (at Alayz) pilD-phA fusion (at Alayz) pilD, in pMMB66EH pikt promoter-lacZ transcriptional fusion bla-phoA fusion pikt, in pMMB66EH Cys17 "* Gly Cyd7 "* Ser Cysi' + Gly Cys7' + Ser -+ Gly Cysy7 + Ser Cyss7 -+ Gly Cysl@' .--* Ser Cys'@' + Gly determined as a fraction of the total cpm of the mutated PilD enzyme divided by the cpm obtained with wild-type enzyme after subtraction of background counts obtained from negative controls. These counts were typically less than 5% of those obtained with the wild-type enzyme and were comparable to that obtained by trichloroacetic acidprecipitating an equivalent amount of prepilin and [3H]AdoMet alone.
Inhibition of P i D Activity with Sulfhydryl-reactive Reagents-To determine the sensitivity of PilD to the effects of sulfhydryl reagents, 1 pl of diluted wild-type PilD (about 75 ng/pl) contained in total membranes extracted from P. aeruginosa PAK-DO was incubated in 7-p1 reactions with the inhibitors at concentrations ranging from 0.5 to 2.0 mM for 30 min at 25 "C. Substrate was then added with the start buffer to bring the final volume to 10 pl, followed by incubation at 37 "C for 15 min (peptidase activity) and 60 min (methylase activity). The latter reactions included 5 pCi of [3H]AdoMet and were performed in the presence and absence of 5 mM dithiothreitol. The reactions were stopped with the addition of 2 volumes of electrophoresis buffer, and aliquots were analyzed either by SDS-Tricine-polyacrylamide gel electrophoresis or trichloroacetic acid precipitation and liquid scintillation counting.
Construction of PilD-LacZ and PilD-PhoA Hybrid Proteins-To isolate pilD-lacZ fusions, E. coli CC118 containing plasmid pRBS-L was infected with X-TnlacZ followed by selection on L-agar with 100 pg/ml ampicillin and 300 pg/ml kanamycin' (Manoil, 1990) at 30 "C. After incubation for 2 days, plasmid DNA was isolated from pooled colonies by alkaline-SDS extraction and used to transform E. coli CC118. Selection for growth was done on L-agar containing kanamycin and the 8-galactosidase-chromogenic substrate 5-bromo-4chloro-3-indolyl 8-D-galactoside. Colonies that grew contained plasmids carrying T n k Z insertions, whereas blue colonies were indicative of in-frame fusions between an open reading frame in a plasmid gene and lacZ. Plasmids from these blue colonies were then mapped by restriction endonuclease digestion to identify insertions in pilD. These were then sequenced (see below) to identify the pilD-lac2 fusion junction. The DNA sequence of one such fusion, pMS506,  (1992) Strom and Lory (1987)  This study  This study  This study  This study  This study  This study  This study  This study  This study encoded a PilD-LacZ hybrid protein containing the first 92 residues of PilD (at Alagz) fused to LacZ, under tac promoter control. A negative control plasmid, pMS505, was constructed by ligating the 3.0-kilobase BamHI-EcoRI-containing TnlacZ fragment from pCM320 into the same sites in pMMB67HE.
To construct the complementary pilD-phoA fusion of the pilD-lac2 fusion in pMS506, it was first necessary to create a corresponding tac promoter signal sequence-deficient phoA vector. Plasmid pMS502 was constructed by inserting the 3.0-kilobase BamHI-Hind111 fragment containing TnphoA from pCM335 into the same sites in pMMB67EH. Both TnphoA and TnlacZ have BamHI sites located just upstream of the respective reporter genes, and an in-frame gene fusion in one can be excised at this site with BamHI and ligated inframe into the other. In addition, the pilD gene in pRBS-L has a single BamHI site located just upstream of the ribosome binding site. Therefore, the complementary pilD-phoA hybrid was constructed by subcloning the 0.3-kilobase BamHI fragment from pMS506 into the same site in pMS502, and designated pMS507. Both the orientation and maintenance of the correct reading frame were verified by DNA sequencing.
Levels of @-galactosidase and alkaline phosphatase activity produced by the various gene fusions were determined as described elsewhere (Brickman and Beckwith, 1975;Miller, 1972).
Substitution of Cys Residues in PilD by Oligonucleotide-directed Site-specific Mutagenesis-Mutations in pilD were constructed by oligonucleotide-directed site-specific mutagenesis of the M13mp19 derivative of pRBS-L (M13mpRBS-L) using the method of Kunkel et al. (1987) as described previously . The primers used to introduce changes in the codons specifying the Cys residues in PilD were as follows. The underlined bases indicate the position of the original Cys codon with the first position (in brackets) changed to encode Gly (GGC) or Ser (AGC) residues. After the mutagenesis reaction, individual plaques isolated from infected E. coli GW5180 were picked, and the single-stranded template was prepared and sequenced by the dideoxy method of Sanger et al. (1977) using the Sequenase version 2.0 kit (U. S. Biochemical Corp.). The replicative forms of derivatives of M13mpRBS-L containing the desired mutations were then used to subclone the region of DNA containing the pilD gene into the broad host range vector, pMMB67EH (Furste et al., 1986), as SstI-Hind111 fragments. This places the pilD gene under control of the tac promoter on a vector that also carries the laclq repressor gene, allowing induction of expression with IPTG. These clones were then transferred into the P. aeruginosa PAK pilD mutant DQ by conjugation.
Phage Sensitivity Assays-For determination of phage PO4 sensitivity of P. aeruginosa pilD mutants containing the cloned wild-type and mutated pilD genes, cells were grown in L-broth with IPTG induction to an A m of -1.0 and then spread on a L-agar plate with a loop. After the culture dried, 5 pl of bacteriophage PO4 (about 1 X lo8 plaque-forming units/ml) was spotted in the center of the streak.
Phage sensitivity was scored as a zone of clearing where the phage was spotted and is indicative of the presence of pili.

Effect of Sulfhydryl Reagents on PilD-catalyzed Cleavage
and Methylation-To assess the role of cysteine residues in PilD enzymatic activity initially, a determination of the sensitivity of PilD peptidase and methyltransferase activity to the sulfhydryl-reactive reagents NEM, p-chloromercuribenzoate, p-chloromercuriphenylsulfonate, and iodoacetamide was carried out. Incubation of PilD with any of these inhibitors substantially reduced both the peptidase and methyltransferase activity of the enzyme. The effects of p-chloromercuribenzoate and p-chloromercuriphenylsulfonate were readily reversible with the addition of dithiothreitol, further demonstrating that the sulfhydryl-blocking reagents are specifically acting on cysteine residues in PilD. It was not possible to determine whether some of the cysteines are more reactive than others because of a lack of sufficient amounts of purified PilD.
Site-directed Mutagenesis and Expression of PilD-In addition to the results demonstrating the sensitivity of PilD to sulfhydryl-blocking reagents, another factor pointing toward the involvement of the conserved Cys residues in the activities of the enzyme is the probable orientation of the domain containing these amino acids toward the cytoplasmic side of the inner membrane. There are several facts that support this topological model. We have isolated a pilD-lac2 fusion that encodes a hybrid protein consisting of the first 92 amino acids of PilD fused to j3-galactosidase (Fig. 2). This fusion junction is in between the two Cys pairs at positions 72, 75, 97, and 100. As shown in Fig. 2, expression of this hybrid in P.
aeruginosa PAK results in high levels of j3-galactosidase activity, whereas expression of the complementary pilD-phA fusion shows little alkaline phosphatase activity. This is strong evidence that the domain containing the Cys residues extends into the cytoplasm (Manoil, 1990). Proteolytic cleavage of leader peptides and subsequent methylation of PilD substrates is also likely to take place on the cytoplasmic face of the membrane, based on models developed from N. gonorrheae and P. aeruginosa pilin-PhoA hybrid data (Strom and Lory, 1987;Dupuy et al., 1991) and on the presence of the methyl donor AdoMet in the cytoplasm only.
Therefore, to assess the role of the Cys residues in PilD activity directly, oligonucleotide-directed site-specific mutagenesis of pilD was used to replace the Cys residues with Ser and/or Gly residues at positions 17, 72,75,97, and 100 of the 290 amino acid protein ( Fig. 3 and Table I).
To assess the effect of cysteine mutations in PilD on in vivo processing of prepilin, PAK-DO derivatives containing the different mutated pilD clones were tested for sensitivity to killing by the pilus-specific phage P04. After IPTG induction of the alteredpilD genes, the bacteria became completely phage-sensitive, whereas in the absence of IPTG they remained resistant to phage killing. This indicated that all of the mutants in PilD were processing prepilin to some extent, which allowed subsequent assembly of the subunits into pili.
To determine if the mutations in PilD altered the extent of processing of wild-type prepilin in vivo, the extracts from IPTG-induced cells grown overnight were subjected to immunoblot analysis using anti-pilin antisera (Fig. 4). In the PAK-DR strain containing the cloning vector pMMB66EH, only uncleaved pilin is seen, whereas complete processing of prepilin is seen in cells containing the wild-type parental clone pRBS-L. Prepilin was also fully processed in bacteria with the mutatedpilD clones encoding the Cys17 + Gly, Cys17 + Ser, Cys7' + Gly, and Cys7' + Ser substitutions. However, expression of the remaining mutated pilD clones in PAK-DQ resulted in varying amounts of processed and unprocessed pilin. Cysg7 + Ser and Cys1Oo + Ser each showed approximately equal amounts of prepilin and pilin, whereas expression of PilD with Cys75 + Gly, Cysg7 + Gly, and Cys1O0 + Gly mutations each show approximately 80% uncleaved prepilin.
The differences seen in pilin processing with the mutated PilD forms could not be attributed to decreased levels of expression. Estimation of the amount of PilD made from various clones by quantitative immunodot blot analysis showed that the levels of PilD expressed from the mutated genes are comparable or slightly exceed that obtained by expression of the wild-type gene (data not shown). The one exception was the Cys7' + Ser-mutated PilD, where no PilD antigen was detectable on immunoblots. Furthermore no peptidase or methylase activity was observed in in vitro using membrane extracts, even though all endogenous pilin was processed in vivo (Fig. 4). It is likely that this mutation resulted in an unstable form of PilD which is still able to cleave prepilin rapidly in growing cells for pilus assembly leading to phage PO4 sensitivity. However, PilD fails to accumulate in cell extracts to levels detectable by the immunoblot assay and is therefore not present in sufficient concentration to demonstrate peptidase and methylase activity above the limits of the detection by these assays.
Specific Endopeptidase and N-Methyltransferase Activities of Mutated PiZD-To determine the effects of cysteine substitutions on PilD enzymatic activity, membranes were prepared, and peptidase and methylase activities were measured in vitro. Relative specific activities were determined by comparison with values obtained using membranes from PAK-DQ (pRBS-L) and PAK-DO (pMMB66EH) as positive and negative controls, respectively. As shown in Table 11, the CysI7 -+ Gly and Cys17 + Ser mutations had no effect on either activity. However, all single mutations in any of the remaining Cys residues resulted in a dramatic reduction in the ability of the enzyme to cleave prepilin and methylate pilin. The specific peptidase activity of the mutants as a whole was less attenuated than the decrease seen in methyltransferase activity. Substitutions of either of the Cys residues at positions 72 and 75 appear to have a lesser effect on peptidase activity than the Cys residues at 97 and 100. No such differences were seen in the decrease in methyltransferase activity of the mutants. Substitutions for any of the conserved Cys residues resulted in a 98-99% reduction in methylase activity. The limit of detection for either activity was approximately 0.5%.
Protection of PilD Peptidase Activity from Alkylation by Sinefungin-We have shown previously that the methyltransferase activity of PilD is extremely sensitive to inhibition by the AdoMet analogue sinefungin, whereas the ability of the enzyme to bind and cleave substrate is unaffected (Strom et al., 1993). This suggests that the sites for prepilin/Pdd protein substrate binding to PilD may not necessarily overlap the methyl donor binding site. The decrease of both cleavage and methylase activities caused by sulfhydryl-reactive reagents and mutations in the cysteine residues, however, suggests that the active sites for both functions may be adjacent to each other. To determine whether the conserved cysteines are in a region of PilD which is shared by the peptidase and methylase sites, we examined the ability of sinefungin, which binds to the methylation site, to interfere with inactivation of the 0.9 Cys'" + Ser 3.6 1.7 Cys'" + Gly 6.3 0.5 The percentage of prepilin converted to the processed form was determined by densitometry of Coomassie-Blue stained SDS-Tricine gels, as described under "Materials and Methods." One unit of PilD activity was calculated as the amount necessary to convert 50% of 1 pg of prepilin to the processed form in 1 h. For determination of relative specific cleavage activities, these values were compared with those obtained for wild-type PilD. Processing of prepilin was not detected with negative controls (in vitro peptidase assays with membranes from P. oeruginosa PAK-DR containing the cloning vector pMMB66EH).
For determination of relative methyltransferase activities, a dilution of membranes calculated to contain about 1 pg of PilD (31.2 pmol) was incubated with 10 pg of precleaved pilin and 5 pCi of [3H] AdoMet in a IO-pl reaction as described under "Materials and Methods.'' Aliquots were precipitated with trichloroacetic acid, collected on cellulose filters, and counts determined by liquid scintillation. Relative specific activities of the mutants were determined as a fraction of the total cpm of the mutated PilD enzyme divided by the cpm obtained with wild-type enzyme, after subtraction of background counts obtained with the negative controls. incubated in 10 pl with sinefungin at a concentration of 1 mM, for 10 min at 22 "C. NEM was then added to 1 mM, followed by a second 10-min incubation at 22 "C, prior to addition of prepilin or precleaved pilin for peptidase and methylase assays as described under "Materials and Methods." peptidase activity by NEM. Alkylation of PilD by NEM was carried out in the presence and absence of sinefungin, and the enzymatic activities were compared with untreated controls. The results of both peptidase and methylase assays on untreated, sinefungin alone, NEM alone, and sinefungin followed by NEM treatments are shown in Fig. 5. As can be seen, pretreatment with sinefungin afforded complete protection of PilD peptidase activity from inactivation with NEM. It is not possible to determine if this protection from NEM alkylation also extends to the methyltransferase activity, since sinefungin appears to bind PilD very tightly and very likely cannot be removed without irreversibly inactivating the enzyme (Strom et al., 1993).

DISCUSSION
The product of the P. aeruginosa pilD gene is a bifunctional enzyme and is responsible for the maturation and posttranslational modification of type IV pili and four proteins which comprise the machinery for extracellular protein secretion. The PilD protein catalyzes the endoproteolytic cleavage of the leader peptide from the pilin subunit precursor and performs the same function for a group of four export determinants, PddABCD.3 In addition, PilD catalyzes N-methylation of the amino-terminal phenylalanine residue after cleavage of the leader peptides of prepilins.
The 290-amino acid long PilD protein is largely hydrophobic, with four to six potential transmembrane spanning helices (Nunn et al., 1990) and is tightly associated with the cytoplasmic membrane . The protein contains a large, 80-amino acid, relatively hydrophilic domain that is notable for the presence of 4 cysteine residues (Nunn et al., 1990). The arrangement of these 4 cysteines is in a pairwise fashion at positions 72 and 75, and 97 and 100, each pair separated by 2 amino acids (Fig. 1). The location of this domain in the cytoplasm was confirmed by engineering fusions to the cytoplasmic and periplasmic localization markers 8-galactosidase and alkaline phosphatase, respectively (Fig.  2).
The environment in the bacterial cytoplasm is strongly reducing because of the high content of soluble thiols, primarily in the form of reduced glutathione (Fahey et al., 1978). It is therefore likely that the cysteines in the hydrophilic domain of PilD are not involved in disulfide bonds. Although added thiols are required for maximal activity of the Nmethyltransferase, this requirement may reflect artificial formation of inter-or intrachain disulfide bonds during preparation of membranes and subsequent extraction of the enzyme.
Although the protease activity is modestly stimulated in the presence of thiols, the cysteine-containing region of PilD lacks the characteristic sequence surrounding the active-site cysteine of cellular serine proteases, QXXX(G/E)XCW (Dufour, 1988), or viralproteases, G(Q,W,Y)CG(S/G) (Bazan and Fletterick, 1988). The enzyme therefore, may represent a new class of thiol proteases.
Alternatively, the cysteines may be involved in binding of metal cofactors. However, although the arrangement of cysteines resembles a zinc finger motif such as seen in the yeast transcriptional factor GAL4 (Pan and Coleman, 1989), there is no evidence that PilD requires any metals for activity and is not stimulated by the addition of divalent cations .
Although PilD appears to be unrelated to any known proteases, a weak homology has been detected with two known methyltransferases. A search for homologies to thiol methyltransferases using Multiple Alignment Construction and Analysis Workbench (MACAW) (Schuler et al., 1991) found a conserved 6-amino acid region in PilD containing Cysa7 (LGGKCS) and similar domains in the E. coli EcoRI (INGKCP) (Greene et al., 1981) and EcoRII (INGKCS) (Som et al., 1987) methyltransferases. Interestingly, an essential cysteine was identified in the CheR methyltransferase of Salmonella typhimurium by site-directed mutagenesis (Subbaramaiah et al., 1991) and photocross-linking of AdoMet D. N. Nunn and S. Lory, submitted for publication. (Subbaramaiah and Simms, 1992). However, the region surrounding this cysteine shows no similarity with the region flanking any 1 of the 4 conserved cysteines in PilD.
Substitutions for the cysteines at positions 72, 75, 97, and 100 resulted in a striking reduction in both peptidase and methylase activities, The methylase activity of all of the mutated PilD proteins was less than 2% of wild-type, whereas the peptidase activity was reduced to between 20% and less than 0.6%. The substitution of glycine for the Cys residue at 72 or 75 resulted in the highest levels of peptidase activity at 20 and 11% respectively, whereas Cysg7 + Ser, Cys'"" .--, Ser, and Cys'"" + Gly were comparable at approximately 5% of wild-type. Cysg7 .--, Gly gave the lowest of the in uitro measured peptidase activities, possibly because of a deleterious conformational change around the glycine residue.
Replacement of the cysteine residues in the conserved cytoplasmic domain of PilD with serine or glycine results in a large decrease in peptidase activity in uitro, and in many cases a corresponding decrease in the levels of processed prepilin in viuo, without affecting piliation (Fig. 4). One possible reason is that the measured turnover rate or Kcat of the wild-type enzyme at 180 min" (Strom and Lory, 1992) is sufficiently high that as much as a 100-fold decrease in activity still results in enough processed pilin for assembly of functional pili to serve as receptors for PO4 phage.
Based on the observations that both the leader peptidase and N-methyltransferase activities are reduced by exposure to sulfhydryl-active reagents and by mutagenesis of any 1 of the Cys residues in the cytoplasmic domain of PilD, we postulate a model in which the catalytic sites for proteolysis and methylation are adjacent and share the region containing the 4 cysteine residues in the cytoplasmic domain. This hypothesis is supported by the experiment in which sinefungin, a competitive inhibitor of N-methyltransferase, protected 1 or several of the cysteines from alkylation, whereas binding of sinefungin did not interfere with leader peptide cleavage. The methyl donor in PilD, therefore, directly interacts with 1 or several cysteines. Peptidase activity is reduced in the mutants as a result of conformational changes in the methyl donor binding site, possibly affecting the substrate binding or the peptidase active site directly. Modification of cysteines may likewise inhibit PilD peptidase activity because of steric hindrance by the introduction of a bulky molecule bound to Cys residues at an adjacent methyl donor binding site. This result allows a prediction of the location of the enzyme active site to residues within a few amino acids upstream of the Cys72,75 pair or downstream of the Cysg7.'"" pair. Alternatively, the peptidase active site may lie within the two pairs of cysteines. However, it is also possible that the peptidase active site may be topologically close to this region because of protein folding but distant on the linear polypeptide. It is clear from the results that no single amino acid residue is involved in both activities of PilD. Isolation of mutants in PilD, which exclusively affect the leader peptidase or the methyltransferase activity, but not both, may allow a more precise localization of the catalytic site within this bifunctional enzyme.