Structure-function analysis of PrsA reveals roles for the parvulin-like and flanking N- and C-terminal domains in protein folding and secretion in Bacillus subtilis.

The PrsA protein of Bacillus subtilis is an essential membrane-bound lipoprotein that is assumed to assist post-translocational folding of exported proteins and stabilize them in the compartment between the cytoplasmic membrane and cell wall. This folding activity is consistent with the homology of a segment of PrsA with parvulin-type peptidyl-prolyl cis/trans isomerases (PPIase). In this study, molecular modeling showed that the parvulin-like region can adopt a parvulin-type fold with structurally conserved active site residues. PrsA exhibits PPIase activity in a manner dependent on the parvulin-like domain. We constructed deletion, peptide insertion, and amino acid substitution mutations and demonstrated that the parvulin-like domain as well as flanking N- and C-terminal domains are essential for in vivo PrsA function in protein secretion and growth. Surprisingly, none of the predicted active site residues of the parvulin-like domain was essential for growth and protein secretion, although several active site mutations reduced or abolished the PPIase activity or the ability of PrsA to catalyze proline-limited protein folding in vitro. Our results indicate that PrsA is a PPIase, but the essential role in vivo seems to depend on some non-PPIase activity of both the parvulin-like and flanking domains.

The PrsA protein of Bacillus subtilis is an essential membrane-bound lipoprotein that is assumed to assist post-translocational folding of exported proteins and stabilize them in the compartment between the cytoplasmic membrane and cell wall. This folding activity is consistent with the homology of a segment of PrsA with parvulin-type peptidyl-prolyl cis/trans isomerases (PPIase). In this study, molecular modeling showed that the parvulin-like region can adopt a parvulin-type fold with structurally conserved active site residues. PrsA exhibits PPIase activity in a manner dependent on the parvulin-like domain. We constructed deletion, peptide insertion, and amino acid substitution mutations and demonstrated that the parvulin-like domain as well as flanking N-and C-terminal domains are essential for in vivo PrsA function in protein secretion and growth. Surprisingly, none of the predicted active site residues of the parvulin-like domain was essential for growth and protein secretion, although several active site mutations reduced or abolished the PPIase activity or the ability of PrsA to catalyze proline-limited protein folding in vitro. Our results indicate that PrsA is a PPIase, but the essential role in vivo seems to depend on some non-PPIase activity of both the parvulin-like and flanking domains.
The rate-limiting steps in the folding of proteins containing cis-prolines are catalyzed by peptidyl prolyl cis/trans isomerases (PPIases). 1 These ubiquitous proteins can be divided into three families, cyclophilins, FK506-binding proteins (FKBP), and parvulins (1). Because cyclophilins and FKBPs are targets of immunosuppressive drugs, they are called immunophilins. No specific inhibitor is available for parvulins. Although PPIases catalyze the isomerization of prolyl peptide bonds in vitro, the physiological role of many of these proteins is still unclear. The effects of their mutational inactivation vary from no phenotypic effect (2) to severe defects (3)(4)(5)(6). PPIases may also act as chaperones either in a PPIase domain-dependent (7) or -independent (8) manner, and they may have essential overlapping functions with other PPIases and chaperones (4,9,10).
Members of the parvulin family have been found in all types of cells. The prototype member of the family is the parvulin of Escherichia coli (11). It is a small cytoplasmic protein comprising essentially a PPIase domain only. In other members a homologous PPIase domain is flanked by N-and C-terminal domains of various lengths (12). At least three subfamilies can be identified based on structural and functional differences. Pin1-type parvulins consist of an N-terminal WW domain followed by a PPIase domain. The plant members, however, lack the WW domain (12,13). The crystal structure of the human Pin1 protein has been solved, and the PPIase domain has been shown to consist of a four-stranded, anti-parallel, flattened ␤-barrel that is surrounded by four ␣-helices (␤␣ 3 ␤␣␤ 2 ) (13). The Pin1 protein binds specifically to a phosphorylated serine or threonine residue located N-terminal to proline. The WW domain, characteristically a module binding proline-rich sequence, and the PPIase domain are both involved in the binding of phosphorylated residues (14). The phosphate group is recognized by a conserved cluster of three positively charged residues located in the N terminus of the loop preceding ␣1 (13,15). PPIases of this subfamily are involved in the regulation of mitosis (15), and linked to chromatin remodeling and general transcriptional regulation (5). In the second subfamily the PPIase domain is similar to that of hPar14; the ␣1-helix and the loop that precedes it are several residues shorter than in Pin1 (16). Furthermore, these enzymes have an insertion in the loop between ␣4 and ␤3. The biological function of hPar14-type PPIases is unknown. In the third subfamily, the PPIase domain is similar to that of the E. coli parvulin. It lacks the hPar14-type insertion and the Pin1-type cluster of positively charged residues involved in the binding of phosphoserine and -threonine. Many PPIases of this subfamily are bacterial proteins required for the folding and/or maturation of extracytoplasmic proteins (16,17). The periplasmic SurA and PpiD, which have roles in the assembly of outer membrane proteins (4), belong to this subfamily.
We have previously identified in Bacillus subtilis a membrane-bound lipoprotein, PrsA, which according to sequence similarity belongs to the parvulin family of PPIases (18). In PrsA a stretch of 90 amino acids is 45% identical with E. coli parvulin. Furthermore significant homology is observed with parvulin-like or PPIase domains of all members of the parvulin family. This parvulin-like region is flanked by a 115-and 70-residue long N-and C-terminal region, respectively, which are not homologous to any regions of other parvulins except for those of PrsA-like proteins of species related to B. subtilis. So far it is not known whether the parvulin-like region and the flanking regions form separate domains. The first amino acid residue at the N terminus of the mature PrsA is a cysteine, which is covalently linked to a diacylglycerol molecule that anchors the protein to the membrane. It has been suggested that PrsA catalyzes or assists the post-translocational folding of extracytoplasmic proteins at the cytoplasmic membrane-cell wall interface (19,20). The secretion rate of AmyQ ␣-amylase, a model protein used in many of our studies, is linearly dependent on the cellular level of PrsA (6). PrsA is an essential protein in B. subtilis, because decreasing its cellular level below 200 molecules per cell results in cell lysis (6). The lethal effect is partially suppressed by the increased negative charge density of the cell wall matrix (21). A defect in PrsA results in the accumulation of misfolded AmyQ at the membrane-wall interface (19). The accumulation of misfolded protein in the cell envelope activates the CssS/R two-component signal transduction pathway and consequently the synthesis of HtrA-type protein quality control proteases (19).
The mode of action of PrsA has not been studied as yet at the molecular level. We carried out a structure-function analysis of PrsA using deletions, pentapeptide insertion mutagenesis, and substitutions of individual amino acid residues. The structure of the parvulin-like region was modeled and used to mutate amino acid residues of the predicted catalytic site of the PrsA PPIase. We demonstrate that the parvulin-like domain indeed catalyzes peptidyl prolyl isomerization, and the N-and C-terminal domains are not needed for this activity. The parvulin-like domain folds correctly independently of the flanking domains. The PrsA function in protein secretion and viability is dependent on all three domains, but the PPIase activity is non-essential for the PrsA function in vivo. There are determinants in PrsA for the catalysis of proline-limited folding in a non-PPIase manner and for cooperation with the PPIase activity.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The B. subtilis strains used are 168 (trpC2) derivatives. In the viability and exoproteome studies, IH7211 with a chromosomal P spac -prsA gene (6) was used as the parental strain. The complementation of prsA3 was determined with IH7492 (prsA3 trpC2 glyB133 hisA1), which expresses amyQ of Bacillus amyloliquefaciens from pKTH3382. The pKTH3382 plasmid was constructed by inserting the amyQ gene as a SalI fragment into pSR11 (22). IH7492 was also used as the parental strain to produce and purify various lipomodified PrsA proteins. The low protease strain B. subtilis WB600 (trpC2 ⌬nprE ⌬aprE ⌬epr ⌬bpf) (23) was used to produce nonlipomodified PrsA proteins in secreted form and E. coli BL21 (FϪ ompT hsdS (r B Ϫ m B Ϫ ) gal) (24) to express GST-PrsA fusion proteins. Bacteria were typically grown in modified 2ϫ L-broth (2% tryptone, 1% yeast extract, and 1% NaCl).
Molecular Modeling-The PrsA PPIase domain structure was modeled based on the NMR structure of hPar14 (Protein Data Bank (PDB) entry 1EQ3) (16). The model was built using Insight II (Accelrys, Inc.). Compared with hPar14, the PrsA PPIase domain has two one-amino acid insertions and a 6-residue deletion (Fig. 2). The first insertion is located in the loop connecting ␣3 and ␤2. The other insertion is in the loop that participates in proline binding and connects ␤2 and ␣4. The only deletion shortens the loop between ␣4 and ␤3 to two residues. The length of this loop varies within the parvulin family. These were modeled by searching a loop from a data base containing either most or an unbiased selection of PDB structures. The loop data base is a PDBselect data base (1888 PDB structures). Several residues from both ends of the loop were used as anchors. The inserted loop was chosen so that it fits the secondary structures while at the same time having a low root mean square deviation from the anchor points. The model was refined by energy minimization with the Discover program using the Amber force field in a stepwise manner. First, the new loop was minimized while the borders of the new loops and the C␣ atoms of the conserved regions (and finally only the C␣ atoms of the conserved regions) were harmonically constrained. The model was evaluated using the PRO-CHECK program (25).
Mutagenesis Techniques-Site-directed mutagenesis was performed using the QuikChange™ site-directed mutagenesis kit (Stratagene) according to the instructions of the manufacturer. Oligonucleotides designed to mutate the prsA gene in pKTH3327 (6) were used in PCR reactions, followed by digestion of the template DNA with DpnI and transformation of the PCR-amplified mutant derivatives of pKTH3327 into E. coli XL-1 Blue supercompetent cells (Stratagene). The prsA mutations were confirmed by sequencing.
Gene Splicing by the Overlapping Extension (gene SOEing) method (26) was used to construct the prsA NϩC mutant. The N-terminal (the ribosome-binding site, 1-402 and 675-695 base pairs of the encoding region) and C-terminal (675-876 base pairs of the encoding region and the translation termination site) fragments were PCR-amplified and fused using the SOE method. The resulting prsA NϩC fragment was inserted into the unique HindIII site in the pDG148 (27). To construct the prsA derivatives encoding truncated PrsA proteins, DNA fragments containing the ribosome-binding site and 444, 663, 780, 828, or 873 base pairs of the encoding region were PCR-amplified and cloned in pDG148. The constructions were verified by sequencing.
The insertion mutagenesis was carried out as follows. First, the B. subtilis prsA gene was PCR-amplified with oligonucleotides 8874 (5Ј-CACATCGATGGAATGAGGAGTGTT-3Ј) and 8875 (5Ј-CAAGTCG-ACGCAGTTCTCAGCAGCATG-3Ј) using the chromosomal DNA of B. subtilis 168 as a template. The resultant fragment, which contains the ribosome-binding site, the coding region, and the transcription termination site of prsA, was cloned into the pGEM T-easy (Stratagene) vector. The plasmid obtained was mutagenized in vitro with the cat-Mu (NotI) transposon (28) using the Mutation Generation System (Finnzymes) as essentially described (28). A pool of mutated plasmids was transformed into E. coli cells and subjected to a series of manipulations including recloning of the transposon containing prsA genes into pUC19 (24) and elimination of the resistance marker gene (cat) by NotI digestion and religation. The protocol produced a pool of about 35,000 prsA mutants. Next, the prsA genes, which contained 15-bp insertions, were again PCR-amplified using the plasmid pool as a template. The PCR primers contained the same homologous region as 8874 and 8875 oligonucleotides. The PCR fragments were inserted between the Hin-dIII and SalI sites of the pDG148 plasmid, placing the prsA genes under the transcriptional control of the P spac promoter. The recombinant plasmids were transformed into E. coli C600 cells (24). Plasmids from 75 transformants were isolated and subjected to HindIII-NotI and SalI-NotI double digestions to identify plasmids that contain only a single 15-bp in-frame insertion in the prsA gene. The outcome was that 49 plasmids fulfilled this criterion. The exact location of 31 insertions in these plasmids was determined by sequencing.
Complementation of prsA3 and Lethality of PrsA Depletion-Complementation of prsA3 was examined by determining AmyQ secretion with a halo assay as previously described (29). One bacterial colony was suspended in 0.5 ml of water, and 0.5-l samples were pipetted onto L-plates containing 3% starch with or without 1 mM isopropyl-␤-Dthiogalactopyranoside (IPTG). The plates were incubated for 16 h at 37°C followed by staining with iodine. The diameters of the halos around raised bacterial colonies were measured. The complementation of prsA3 with the PrsA variants containing site-directed mutations were also determined in liquid cultures. ␣-Amylase secretion was determined with the Phadebas amylase test (Amersham Biosciences) as described previously (21).
To investigate the complementation of the lethality of PrsA depletion, the B. subtilis IH7211 (P spac -prsA) was transformed with pDG148 derivatives containing mutant prsA genes. In the resultant strains, the chromosomal wild-type prsA gene and a plasmid-carrier of the mutant prsA gene were both under the control of the P spac promoter. However, because the P spac in the plasmids is leaky and that in the chromosome is tight, we were able to determine the complementation. Cells grown to the density of Klett 100 in the presence of 1 mM IPTG were harvested and washed once with L-broth, diluted 1000-fold in 2ϫ L-broth, and then cultivated for 8 -10 h. Growth was measured with a Klett-Summerson colorimeter.
Purification of Lipomodified and Nonlipomodified PrsA Proteins-For the purification of lipomodified PrsA proteins, B. subtilis cells were grown in buffered tryptone broth (1% tryptone, 100 mM potassium phosphate, pH 7.5, 2% glucose, 1 mM MgSO 4 , 3.5 M FeSO 4 , 3.5 M ZnSO 4 , 0.45 mM MnSO 4 , 40 nM CuSO 4 , and 7 nM K 2 Cr 2 O 7 ). In the early stationary phase, the cells were harvested, treated with lysozyme (4 mg/ml) for 30 min, and then sonicated until the viscosity disappeared. The cell membranes were isolated by centrifugation at 44,000 ϫ g for 1 h 15 min and solubilized in 2% n-octyl-␤-D-glucopyranoside (OG) (Sigma-Aldrich). Insoluble material was removed by centrifugation, and the supernatant was passed through a DEAE-Sepharose CL-6B (Amersham Biosciences) column equilibrated with a buffer containing 10 mM Tris-HCl, pH 7.5, and 30 mM NaCl. The flow-through fraction containing PrsA was applied to a CM-Sepharose CL-6B (Amersham Biosciences) column equilibrated with a buffer containing 30 mM MES, pH 6.0, and 30 mM NaCl, and PrsA was eluted with 300 mM NaCl. From the solubilization step onward all the solutions contained 1% OG.
In order to produce PrsA in secreted form, the type II signal peptide anchoring PrsA to the membrane was replaced with the type I signal peptide of AmyQ. The PrsA-secreting strain was grown in Terrific Broth (17 mM KH 2 PO 4 , 72 mM K2HPO4, 24 g liter Ϫ1 yeast extract, 12 g liter Ϫ1 tryptone, and 0.4% glycerol) containing 5% glucose at 37°C to Klett 100 ϩ 2 h. The protease inhibitor phenylmethylsulfonyl fluoride was added every half-hour after Klett 100. The culture supernatant was concentrated 20-fold by a Vivaflow 200 tangential flow concentrator (Vivascience) and applied to a Sepharose XL cation-exchange column (Amersham Biosciences) equilibrated with 30 mM MES, pH 6.0. PrsA bound to the column was eluted with 500 mM NaCl.
Production and Purification of GST-PrsA Fusion Proteins-The GST gene fusion system was used to produce PrsA proteins in E. coli for purification. Cells were grown in 2ϫ TY (1.6% tryptone, 1% yeast extract, 1% NaCl) to Klett 50 -70, and 0.2-1 mM IPTG was added to induce the synthesis of GST-PrsA fusion proteins from pGEX-2T (Amersham Biosciences) derivatives containing prsA genes. After incubating for 3-4 h at 30°C, the cells were harvested by centrifugation, suspended in phosphate-buffered saline, and disrupted with a French Press. Cell debris was removed by centrifugation prior to applying the cell lysate into a glutathione-Sepharose 4 fast flow (Amersham Biosciences) column equilibrated with phosphate-buffered saline. After washing the column with phosphate-buffered saline, thrombin (1 unit/ 100 g of fusion protein) was added to release PrsA from GST, and fractions containing PrsA were collected. Circular dichroism spectroscopy data were collected by a Jasco 720 spectropolarimeter (Tokyo, Japan). PrsA concentrations were determined spectrophotometrically at A 280 and in Coomassie Blue-stained SDS-PAGE.
Ribonuclease T1 Refolding and Protease-coupled PPIase Assays-The proline-limited folding was assayed with ribonuclease T1 (RNase) essentially as has been previously described (30). The RNase was purchased from Sigma-Aldrich. Centricon 3 filters (Amicon) were used to remove the ammonium sulfate of the RNase solution. To unfold the RNase, it was incubated at 20 M concentration in a buffer containing 100 mM Tris-HCl, pH 7.8, 8 M urea, and 1 mM EDTA for 1 h at room temperature. The refolding was initiated by diluting the solution 28.5fold with a buffer containing 100 mM Tris-HCl, pH 7.8, and 1 mM EDTA. The tryptophan fluorescence was measured at 320 nm (5-nm bandwidth) with excitation at 268 nm (3-nm bandwidth) using a Shimadzu RF-5000 or Varian Cary Eclipse spectrofluorophotometer. PrsA was added prior to the dilution at the final concentrations of 1-8 M. When the foldase activity of lipomodified PrsA proteins was determined, 1% octylglucoside was added to the refolding buffer. The RNase refolding activity of nonlipomodified and lipomodified PrsA was measured at 10 and 12°C, respectively. Cyclosporin A (CsA, Sigma-Aldrich) and FK506 (Calbiochem) were used at 5 M concentrations to inhibit PPIase activ-ities of putative contaminating cyclophilins and FKBP-type PPIases in the PrsA preparations, respectively. Juglone (Sigma-Aldrich), which is an inhibitor of numerous parvulins, was used at 7 M concentrations as described (31). Cyclophilin, which was used as a positive control, was from calf thymus and was purchased from Sigma-Aldrich. PPIase activities were expressed as k cat /K m (mM Ϫ1 s Ϫ1 ). A protease-coupled assay with ␣-chymotrypsin was used to determine PPIase activities with a tetrapeptide substrate (succinyl-Ala-Lys-Pro-Phe-4-nitroanilide) (Fischer et al., Ref. 40). The peptide concentration was 37.5 M.
Refolding of Chemically Unfolded Rhodanese-Unfolding and refolding of rhodanese was carried out essentially as previously described (32). Rhodanese (Sigma Aldrich) was incubated at 38 M concentration in a buffer containing 50 mM Tris-HCl, pH 7.8, 6 M GdnHCl, and 5 mM dithiothreitol for 1 h at 25°C. The refolding was initiated by a 60-fold dilution with the refolding buffer (50 mM Tris-HCl, pH 7.8, 10 mM dithiothreitol, and 50 mM sodium thiosulfate) in the absence or presence of folding factors. In experiments with GroELS, 1 mM MgCl 2 , 10 mM KCl, and 1 mM ATP were added to the refolding buffer. Over 60 min of incubation, aliquots were withdrawn at several time points. Rhodanese activity of the refolding reaction samples was measured as described in Ref. 33. PrsA concentration was 5 M. GroELS (2.5 M) and bovine serum albumin (15 M) were used as positive and negative controls, respectively.
Extracellular Proteome Analysis-Cells of the B. subtilis IH7211 (P spac -prsA) were grown with and without IPTG, and its derivative strains harboring plasmids for encoding mutant PrsA proteins were grown without IPTG in L-broth. In the transition to the stationary phase (t 0 ) or an hour later (t 1 ), cells were separated from the medium. Extracellular proteins in the medium were precipitated with ice-cold 10% (w/v) trichloroacetic acid, and 100-g samples were subjected to two-dimensional electrophoresis as described previously (34). The resulting two-dimensional gels were fixed with a solution containing 50% (v/v) methanol and 7% (v/v) acetic acid, and stained with SYPRO Ruby protein gel stain (Molecular Probes). Fluorescence was detected using a Storm860 fluorescence imager (Molecular Dynamics). For quantification, samples from two independent experiments were resolved for each time point investigated, and the quantitative image analysis was performed with DECODON Delta two-dimensional software (www.decodon.com). Protein identification was carried out as previously described (34).

The N-terminal, C-terminal, and Parvulin-like Regions Are
All Essential for PrsA Function in Vivo-Many parvulins consist of functionally and structurally independent domains (8,13,16,35). In order to identify which regions (or domains) of the PrsA protein are required for protein secretion and viability in B. subtilis, a series of deletion mutations of prsA were constructed, and the mutations were characterized for their effects on in vivo PrsA activity (Fig. 1). The constructed deletions of the C-terminal region either removed the whole region (PrsA-(1-202)), or part of it (PrsA-(1-256) and PrsA-(1-242)). We also constructed mutations that deleted the entire parvulin-like region (PrsA-(NϩC)) or both the C-terminal and parvulin-like region (PrsA-(1-128)). Furthermore, PrsA-(33-258) molecules were designed that lacked the 32 N-terminal and 15 C-terminal amino acids. All mutant genes were expressed from an IPTG-inducible promoter in a multicopy plasmid.
All mutant PrsA proteins were detected as abundant proteins by immunoblotting, indicating their in vivo stability. The cellular location of two proteins, PrsA-(1-242) and PrsA-(NϩC), was studied by trypsin accessibility on the surface of protoplasts (36). Trypsin treatment degraded about 95% of PrsA-(1-242) and PrsA-(NϩC), but only about 20% of GroEL, a cytoplasmic protein used as a control for protoplast stability (not shown). In Triton X-100 and ultrasonic-treated protoplasts, all PrsA as well as GroEL were degraded. These results indicate that the mutant PrsA proteins translocate across the cell membrane and become surface-exposed in a manner similar to the wild-type PrsA.
The in vivo activities of the mutant PrsA proteins were determined by two independent complementation assays.
First, we determined their ability to restore AmyQ secretion in the secretion-deficient prsA3 (D249N) mutant. Second, the ability of plasmid-coded mutant PrsA proteins to restore the growth of cells depleted of the chromosomally coded wild-type PrsA was studied. For the PrsA depletion, a strain in which the chromosomal prsA has been placed under the control of the regulatable P spac promoter was used. In the absence of IPTG there is thus an insufficient amount of chromosomally coded PrsA to support growth. The P spac promoter of the plasmid used, however, is leaky, and enough PrsA is produced to rescue growth even in the absence of IPTG if the PrsA encoded by the gene is functional (Fig. 1).
A truncated PrsA (PrsA-(1-256)), lacking the 17 C-terminal amino acid residues including a cluster of serine and threonine residues (the "serine/threonine tail"), was fully active, as evidenced by the complementation of the secretion deficiency of prsA3 as well as the PrsA depletion (viability) (Fig. 1). This indicates that the serine/threonine tail is dispensable. Deleting about half of the C-terminal region (PrsA-(1-242)) inactivated both functions of PrsA ( Fig. 1), implying that the C-terminal region is essential for the PrsA function in vivo. Consistently, the truncation of the whole C-terminal region (PrsA-(1-202)) or both the C-terminal and parvulin-like region (PrsA-(1-128)) resulted in inactive PrsA proteins. A mutant PrsA containing the N-and C-terminal regions but lacking the parvulin-like region (PrsA-(NϩC)) was also unable to restore the growth of PrsA-depleted cells (Fig. 1). This mutant PrsA displayed, however, a weak activity in the AmyQ secretion assay, about 15% of the wild-type level. A short N-terminal deletion (PrsA-(33-258)) also fully inactivated PrsA.
The deletion mutant PrsA proteins were stable in vivo as revealed by immunoblotting, unlike other PrsA mutant proteins, such as PrsA3, which is largely degraded by cell membrane-and wall-associated quality control proteases (19). The finding suggests that they have a native-like fold and folding kinetics. The integrity of the native-like fold of PrsA-(NϩC) was also confirmed by circular dichroism spectroscopy (data not shown).
We also characterized PrsA by insertion mutagenesis. The aim was to obtain a large set of in-frame insertions of five amino acids at diverse positions in PrsA and to analyze the effects of the insertions on AmyQ secretion and cell viability in a manner similar to those outlined above. The mutagenesis was carried out using the in vitro strategy that is based on Mu transposition (28), and properties of 31 insertion mutants were characterized ( Table I). The cellular level of all except one mutant PrsA protein remained unaffected ( Table I), suggesting that the inserted pentapeptides did not significantly alter the PrsA fold. One mutant PrsA was truncated probably because of proteolytic degradation.
Numerous pentapeptide insertions in the N-terminal region (residues 1-128) either inactivated PrsA or deteriorated its activity severely (Table I). The effects of the insertions were comparable for both AmyQ secretion and viability. It is also notable that the insertions close to the N terminus of PrsA more severely affected the PrsA function than those close to the parvulin-like region. Insertion mutation 36 (Table I), which introduced a pentapeptide after Gln 65 in the N-terminal domain, showed a surprising phenotype. The synthesis of PrsA36 in the prsA3 mutant caused a retardation in growth; the cells were aberrant and grew in tiny colonies on agar plates. No growth was observed when cells from the tiny colonies were replated on fresh plates. In contrast, in the wild-type strain, PrsA36 was not harmful to growth. Thus, it seems that the PrsA36 protein is synthetic lethal in the prsA3 mutant. The PrsA36 protein was also unable to rescue the growth of cells depleted of the wild-type PrsA (Table I), indicating that it is an inactive PrsA protein. The correct cellular location of two inactive insertion mutants, PrsA5 and PrsA113, was verified by trypsin accessibility using the method described above (data not shown). Furthermore, the signal peptide of all insertion mutant PrsA variants was cleaved off as revealed by immunoblotting (not shown), also suggesting that these mutant proteins were translocated across the cell membrane, lipomodified, and located at the cell membrane-wall interface in a manner similar to the wild-type PrsA.
In the parvulin-like and C-terminal regions pentapeptide insertions were mostly tolerated. However, two mutations (54 and 113) in the parvulin-like domain also fully disrupted the activity of PrsA in both the AmyQ secretion and viability assays (Table I). Furthermore, it was found that many mutants with an insertion in the parvulin-like or C-terminal region exhibited distinctly different activities in the two assays. None of the pentapeptide insertions apart from 54 and 113 conferred an effect on the complementation of the secretion defect in prsA3, but reduced by about half the activity in the viability assay.
Molecular Modeling of the PPIase Domain of B. subtilis PrsA-The structure-function analysis of PrsA was continued by modeling the three-dimensional structure of the parvulin- like domain of B. subtilis PrsA, taking advantage of the known structure of a human parvulin, hPar14 (16,35). The parvulinlike domain of PrsA, which contains 45% identical residues with hPar14, could be modeled to a typical parvulin-type PPIase structure (Fig. 2). In the Pin1 structure, the peptidyl-prolyl bond isomerization is thought to be catalyzed by a cysteine, two histidines, and a serine (13). Both of the histidine residues are conserved in the PrsA structure, but aspartic acid and tyrosine substitute for the cysteine and serine, respectively. The hydrophobic pocket for the substrate proline is formed in both hPar14 and Pin1 by three amino acids, leucine, methionine, and phenylalanine (13,16,35). According to the model, all the residues forming the proline-binding pocket and the catalytic site are structurally conserved in PrsA and hPar14 ( Fig. 2A). Comparison of the surface charge of PrsA and hPar14 revealed that in PrsA the proline-binding pocket and the catalytic site are clearly more hydrophobic (data not shown).
PrsA Catalyzes the Folding of Ribonuclease T1 and cis/trans Isomerization of the Prolyl Peptide Bond-Because modeling shows that the hPar14/Pin1 fold is conserved in the parvulinlike domain of PrsA, it is likely that PrsA is also capable of catalyzing the isomerization of peptidyl-prolyl bonds. To demonstrate this, we examined the ability of the wild-type PrsA to catalyze in vitro the refolding of ribonuclease T1 (RNase T1) denatured with urea. The rate of the refolding of RNase T1 is limited by the slow kinetics of the cis/trans isomerization of two peptide bonds preceding the cis proline residues (37,38). The PPIase activities of both the lipomodified and nonlipomodified PrsA were examined. The nonlipomodified PrsA does not contain the N-terminal diacylglycerol moiety. Cyclophilin was used as a positive control.
As shown in Fig. 3A, PrsA indeed catalyzed the prolinelimited folding in a concentration-dependent manner. The k cat /K m values were 143 mM Ϫ1 s Ϫ1 and 25 mM Ϫ1 s Ϫ1 for the lipomodified and nonlipomodified PrsA proteins, respectively (Table II). The catalytic activity of the nonlipomodified PrsA was the same whether it was produced in E. coli as a GST fusion protein or a secreted form (see "Experimental Procedures") in B. subtilis (Table II). The activity is dependent on the parvulin-like domain, since PrsA-(NϩC) was inactive. All variants of PrsA either with a complete or partial deletion of the Nor C-terminal domain catalyzed the refolding of denatured RNase T1 in a manner similar to the full-length PrsA (Table  II). Even the mere parvulin-like domain (PrsA-PPI) was fully active. Neither CsA nor FK506 inhibited this activity (Table II), indicating that the PrsA preparations were not contaminated with cyclophilins or FKBP-type PPIases. They were also free of the E. coli parvulin, since juglone did not affect the activity (31). These results suggest that PrsA is a PPIase. Cyclophilin, however, displayed a clearly higher activity (Fig. 3B); k cat /K m was 1300 mM Ϫ1 s Ϫ1 (Table II), about 50-fold higher than that of PrsA. Compared with the previously reported PPIase activity of a cyclophilin (39), our bovine cyclophilin preparation exhib- ited about 20-fold higher activity. The reason for this is unclear.
We confirmed the PPIase activity of PrsA using a tetrapeptide substrate (succinyl-Ala-Lys-Pro-Phe-4-nitroanilide) and the conventional protease-coupled PPIase assay (40). For these measurements, proteins were produced in E. coli as GST fusion proteins. Both the full-length PrsA and parvulin-like domain exhibited a PPIase activity with the tetrapeptide substrate, but again the activity of cyclophilin was significantly higher, by about 60-fold (parvulin-like domain) (Fig. 3C and Table II). PrsA-(NϩC) was consistently inactive. Chymotrypsin, which is used in the PPIase assay to cleave the trans form of Lys-Pro, nicked or partially truncated the full-length PrsA protein during the assay (SDS-PAGE, not shown). However, chymotrypsin at 300 g/ml for 5 min did not affect the level of the PPIase activity (data not shown). The protein consisting of the parvulin-like domain was resistant to chymotrypsin.
The Catalytic Amino Acid Residues of the Predicted Active Site Are Dispensable for the PrsA Function in AmyQ Secretion and Growth-Taking advantage of the modeled structure of the parvulin-like domain, we wanted to determine whether the conserved amino acid residues in the predicted active site are needed for the PrsA function in vivo. The catalytically impor-tant residues, Met 172 and Phe 176 , located in the proline-binding pocket, and His 122 , Asp 154 , Tyr 196 , and His 199 , located in the catalytic site, were replaced either with alanine or tyrosine (Fig. 2B). The residues Gly 197 and Ile 200 were mutated in the same way, because alanine substitutions of the corresponding residues in the E. coli PpiD parvulin are known to disrupt the PPIase activity as well as the in vivo function in the assembly of outer membrane proteins (4,41). Residues Phe 144 , Ser 152 , Gly 161 were also replaced with alanine because of their high conservation. The S156P mutation resulted accidentally from a PCR error. Additionally, three internal deletions in the parvulin-like domain were constructed. The shortest of the deletions, ⌬Met 172 -Lys 182 , removed the whole ␣4-helix (Fig. 2C), which includes residues Met 172 and Phe 176 of the proline-binding pocket. The deletion of ⌬Asp 154 -Gln 171 led to loss of the ␣3helix, ␤2-strand, and almost the whole loop following the ␤2strand, whereas the deletion of ⌬Thr 185 -Lys 202 removed ␤3 and half of ␤4 (Fig. 2C). These mutant PrsA proteins were present in cells at near wild-type levels (immunoblotting, not shown), suggesting that the protein domains folded correctly to the native protease-resistant conformation, apart from three, PrsA-H199A, PrsA-(⌬Asp 154 -Gln 171 ), and PrsA-(⌬Thr 185 -Lys 202 ), which were partially degraded and thus may be at least partially misfolded (Table III). The effects of the constructed mutations on the PrsA function in vivo were analyzed using the complementation assays.
Strikingly, the His 122 , Asp 154 , Met 172 , and Phe 176 residues of the active site, and four other highly conserved residues, Phe 144 , Ser 152 , Gly 161 , and Ile 200 , could be substituted without effect on the PrsA function in vivo (Table III). The S156P mutation also did not affect the PrsA function. These mutant PrsA proteins restored AmyQ secretion in the prsA3 mutant to the wild-type level as well as rescued the growth of cells depleted of wild-type PrsA. Two proteins, PrsA-G197A and PrsA-H199A, also fully restored PrsA activity in the viability assay but displayed only partial activity in the AmyQ secretion assay (Table III). Conversely, the PrsA-Y196A substitution mutant fully restored AmyQ secretion in prsA3 but only partially restored growth in cells depleted of the wild-type PrsA. Despite these minor effects, the results indicate that none of the predicted active site residues are indispensable for the PrsA functions in AmyQ secretion and growth. It was surprising that even PrsA-(⌬Met 172 -Lys 182 ), which lacks most of the likely proline-binding site, exhibited full PrsA activity in the viability assay, and even in the AmyQ secretion assay the activity was only moderately (70% of wild type) reduced (Table III). The longer internal deletions, ⌬Asp 154 -Gln 171 and ⌬Thr 185 -Lys 202 , however, inactivated PrsA (Table III), in a manner similar to the deletion of the whole parvulin-like domain (Fig. 1), consistent with an essential role for this domain.
The Influence of PrsA Depletion and Active Site Mutations of the PrsA PPIase on the Extracellular Proteome-In previous studies the extracellular proteome of B. subtilis wild type has been described (34). Taking advantage of the exoproteome analysis, we now investigated the effects of partial PrsA depletion and four of the active site mutations (H122A, D154A, Y196A, and ⌬Met 172 -Lys 182 ) on the exoproteome. These mutations were chosen either since they were partially defective in the above complementation assays. Y196A and ⌬Met 172 -Lys 182 or the corresponding active site mutations in a Pin1-type eukaryotic parvulin have been shown to cause phenotypic effects (H122A and D154A) (5). IH7211 cells (P spac -prsA) were depleted of the wild-type PrsA protein by cultivating them in the absence of IPTG and thus the P spac -prsA activity was repressed but the initial conditions did not inhibit growth. Strains transformed with plasmids encoding the mutant PrsA proteins were similarly partially depleted of the chromosome-coded wild-type PrsA. The strains were grown to the early stationary growth phase, followed by the sample preparation and two-dimensional gel electrophoresis as described under "Experimental Procedures." For gel comparison we used the false-color image analysis and quantification tool of the DECODON Delta twodimensional program.
The partial PrsA depletion caused prominent alterations in the pattern of the extracellular proteome. First, strongly elevated amounts of cytoplasmic proteins were observed (Fig. 4). The quantification of 17 previously identified cytoplasmic proteins showed 5-20-fold higher relative amounts in the exopro-teome of PrsA-depleted cells (Fig. 6A and data not shown). The finding is consistent with cell lysis in the absence of PrsA (6). Lysis was also indicated by a severalfold increase of the protein content in the culture medium of the PrsA-depleted cells (data not shown). Secondly, the PrsA depletion drastically decreased or removed most secreted proteins (Figs. 4 and 6B), including those expressed and secreted at a high level (34). However, there were also several proteins with a predicted type I signal peptide, like YxaK and YolA, which were present at about the same level in the proteome of both PrsA-depleted and nondepleted cells (Fig. 4).
Plasmid-coded wild-type PrsA reversed the effects of the Values are means of at least two measurements. b Prolyl isomerization of the tetrapeptide succinyl-Ala-Lys-Pro-Phe-4-nitroanilide. c PrsA proteins were produced either as GST fusion proteins in E. coli or as lipoproteins in B. subtilis. PrsA was cleaved from GST, and the product is devoid of the N-terminal cysteine and lipomodification of the native protein.
d BD, below detection limit. e Close to the detection limit and therefore considerable variation from one experiment to another. f ND, not determined. g Preparation similar to GST-PrsA fusion protein but obtained from cells expressing GST not fused to PrsA. h Bovine serum albumin. i The PrsA lipoprotein contains a covalently linked diacylglycerol moiety at the N-terminal cysteine residue. j Produced as a secretory protein and devoid of lipomodification. PrsA depletion on the exoproteome (not shown). The exoproteome analysis of the parvulin domain mutants revealed that PrsA-H122A also restored the wild-type pattern of the exoproteome (Figs. 5A and 6). This is consistent with the complementation in the AmyQ secretion and viability assays above and indicates that PrsA-H122A is a fully functional PrsA protein in vivo. PrsA-Y196A also restored the wild-type pattern of the exoproteome despite the moderate defect in the viability assay (Fig. 6). Interestingly, the exoproteome of the mutant ⌬Met 172 -Lys 182 resembled that of the partial PrsA depletion (Figs. 5B and 6). There was a decrease in secreted proteins (Fig. 6B) and indications of cell lysis as judged by the increased amount of proteins released into the culture medium. However, the contamination of the exoproteome by cytoplasmic proteins was less extensive than for PrsA depletion (Fig. 6A), suggesting partial complementation. A similar partial complementation was also seen with the D154A mutant construct (Fig. 6). PrsA-(PPI-⌬Met 172 -Lys 182 ) constructed in this study were subjected to RNase T1 refolding and protease-coupled PPIase assays. Also the catalytic activities of a PrsA variant with a change of the serine 156 to proline in the parvulin-like domain (PrsA-(PPI-S156P)) were determined. The mutant PrsA proteins were produced either as GST fusion proteins in E. coli or lipoproteins in B. subtilis. The integrity of the PrsA variants purified was confirmed by immunoblotting (not shown).

The PPIase Active Site Is Important for PrsA-catalyzed Folding and Prolyl Isomerization in Vitro-Several
The PrsA-(PPI-S156P) protein catalyzed RNase T1 refolding (Table II) only marginally. A similar reduction in the catalytic activity was also observed when the active site residue His 122 or Asp 154 was mutated. The findings suggest that there are specific structural requirements for the parvulin-like domain in the catalysis of proline-limited protein folding. The PrsA-(PPI-(⌬Met 172 -Lys 182 )) protein was inactive. However, when these mutations were in the full-length, lipomodified PrsA, they had no effect on the catalytic activity (Table II). The activity of these lipomodified PrsA proteins determined in a detergent milieu was significantly higher (about 5-fold) than that of nonlipomodified PrsA in a non-detergent milieu.
The PPIase assay with the succinyl-Ala-Lys-Pro-Phe-4-nitroanilide tetrapeptide substrate revealed that PrsA-(PPI-H122A) is inactive (Table II), indicating that His 122 is indispensable for the prolyl isomerization. The ⌬Met 172 -Lys 182 deletion also inactivated the tetrapeptide isomerization. Although PrsA-(PPI-S156P) was strongly defective in catalyzing RNase T1 refolding, it exhibited a PPIase activity with the tetrapeptide substrate (Table II). The activity was about half of that of the wild-type parvulin-like domain (PrsA-PPI). The PrsA-(PPI-D154A) protein also displayed an activity similar to PrsA-(PPI-S156P).
In the protease-coupled PPIase assay chymotrypsin caused some degradation of PrsA-(PPI-S156P) and PrsA-(PPI-D154A) (not shown), and therefore the actual PPIase activities of these mutant PrsA proteins may be higher than what was observed (Table II). The rate of degradation of PrsA-(PPI-H122A) was similar to that of PrsA-(PPI-S156P) and PrsA-(PPI-D154A) , implying that the degradation does not explain the PPIasenegative phenotype of this PrsA variant. The PrsA-(PPI-(⌬Met 172 -Lys 182 )) protein was more sensitive to proteolytic degradation than the other PrsA variants.
PrsA Does Not Possess a General Chaperone Activity-Since some PPIases are known to have a dual PPIase chaperone role in protein folding (7,8), we determined whether PrsA has the properties of a molecular chaperone. We used chaperone assays in which the ability of PrsA to prevent either unfolded rhodanese or citrate synthase from aggregating during refolding FIG. 4. The extracellular proteome of B. subtilis IH7211 (P spac -prsA) grown in the absence (red image) and presence (green image) of 1 mM IPTG. Cells were harvested 1 h after the entry into the stationary phase, and extracellular proteins in the culture medium were trichloroacetic acid-precipitated, separated by two-dimensional PAGE, and stained with SYPRO Ruby as described under "Experimental Procedures." For two-dimensional image analysis and quantification the DECODON Delta twodimensional software was used. Proteins that appeared as green spots (higher level in the presence of PrsA) are marked in blue and those appeared as red spots (higher level in the absence of PrsA) are marked in red. Proteins that appeared as yellow spots are labeled in white (unchanged according to the quantitation data). This gel is available in the Sub2D proteome data base (microbio2.biologie. uni-greifswald.de:8880/). was determined (see "Experimental Procedures" for details).
In the absence of a chaperone, only about 30% of the initial rhodanese activity was recovered (Fig. 7). GroELS chaperone increased the recovery up to 80%, whereas bovine serum albumin, which was used as a negative control, did not affect the recovery. Unlike GroELS, PrsA did not exhibit a chaperone activity (Fig. 7). The result was similar when the chaperone activity was determined using citrate synthase as a substrate, and the degree of aggregation was measured by light scattering (data not shown).

DISCUSSION
In this study we carried out a structure-function analysis of B. subtilis PrsA. The aim was to identify which regions or domains of PrsA are required for protein secretion and growth with the emphasis on the characterization of the parvulin-like domain, its putative peptidyl-prolyl cis/trans isomerase activity, and the role of the PPIase activity in vivo. PrsA indeed exhibited PPIase activity in a parvulin-like domain-dependent manner. Our results furthermore indicate that in vivo the parvulin-like PPIase domain as well as the flanking N-and C-terminal non-PPIase domains are essential, but the PPIase activity as such is non-essential.
The molecular modeling of PrsA revealed a Par14-or Pin1-type parvulin fold and structural conservation of the proline-binding and catalytic site residues. PrsA in vitro catalyzed the prolinelimited folding of ribonuclease T1. However, the catalytic activity of PrsA was fairly weak. Much higher activities have been reported for the E. coli trigger factor, human Pin1 and E. coli PpiD (4,13,42), whereas the activity of E. coli SurA is comparable to that of PrsA (8). Because PrsA-PPI protein containing only the parvulin-like region catalyzed the proline-limited folding in a similar manner as the full-length PrsA, the parvulin-like region is clearly responsible for the activity. The activities of PrsA-(33-241) and PrsA-(PPIϩC) are consistent with this conclusion. The PPIase activity of PrsA was confirmed with a tetrapeptide substrate. These results indicate that the parvulin-like region indeed forms a separate domain, which is able to fold correctly into the native enzymatically active conformation without any significant influence from the flanking regions. The replacement of the serine 156 in the parvulin-like domain to proline resulted in a mutant PrsA protein (PrsA-(PPI-S156P)) that was unable to catalyze the refolding of denatured RNase T1. Interestingly, however, PrsA-(PPI-S156P) catalyzed the prolyl isomerization of a tetrapeptide substrate. This result indicates that the catalysis of proline-limited folding of a complex protein substrate such as ribonuclease by PrsA also requires some other determinants in the parvulin-like domain than those required for the PPIase activity per se. Similar additional determinants have also been observed in the catalysis of the folding of a reduced and carboxymethylated RNase T1 variant by the parvulin-like domains of SurA (8). These determinants of PrsA are not important for the PrsA function in vivo, since the PrsA-(PPI-S156P) protein was fully active in the complementation assays. PrsA-(NϩC), which does not contain the parvulin-like domain, was unable to catalyze prolyl isomerization, consistent with the conclusion that the PPIase activity resides in the parvulin-like domain.
In the eukaryotic parvulin Pin1, the His 59 and Cys 113 active site residues are highly important for the PPIase activity. The H59A substitution results in almost complete loss of the PPIase activity (15) and a replacement of the Cys 113 decreases the PPIase activity to about 1% of wild-type level (13). Substitutions of the corresponding catalytic amino acids in the parvulin-like domain of PrsA, His 122 and Asp 154 , also affected the PPIase activity as determined with the peptide substrate. The H122A substitution inactivated the PPIase function and  , green image). B, ⌬Met 172 -Lys 182 (red image) versus wild type (green image). Cells were grown in the absence of IPTG, followed by the preparation and analysis of extracellular proteins as described in the legend to Fig. 4. Proteins appearing as green spots (higher level in the wild type) are marked in blue, and those appearing as red spots (higher level in the mutants) are marked in red. Proteins appearing as yellow spots are labeled in white (unchanged according to the quantitation data). These gels are also available in the Sub2D proteome data base.
D154A decreased the activity to about half of the wild-type level. Furthermore, the deletion of the ␣4-helix resulted in a PrsA variant that did not catalyze prolyl isomerization. These mutant variants of the parvulin-like domain were also defective in catalyzing the proline-limited folding of denatured RNase T1. These results suggest that the predicted active site residues are indeed crucially important for the PPIase activity of PrsA. Despite the effect of the H122A substitution on the PPIase activity, this mutation did not affect the PrsA functions in the AmyQ secretion and growth, indicating that the PPIase activity is dispensable for the PrsA functions in vivo. Surprisingly, in the full-length, lipomodified PrsA, the active site mutations did not disturb the catalysis of proline-limited folding in vitro, suggesting that PrsA also facilitates protein folding by some other mechanism than catalyzing the cis/trans isomerization of prolyl peptide bonds. It has recently been shown that the DnaK chaperone catalyzes the cis/trans isomerization of nonprolyl peptide bonds (43). The difference of activity between soluble and lipomodified mutant proteins may reside in the pattern of multimerization or protein-protein association of PrsA monomers.
None of the targeted amino acid replacements of the likely active site residues inactivated PrsA in the AmyQ secretion or growth. These functions also remained unaffected when Gly 197 and Ile 200 were mutated, in contrast to the corresponding mutations in PpiD, which have been shown to inactivate the PpiD function both in the assembly and stability of outer membrane proteins and in vitro-determined PPIase activity (4). However, the parvulin-like domain is most probably essential for the PrsA functions in vivo, as evidenced by the inability of the mutant PrsA proteins lacking the whole PPIase domain (PrsA-(NϩC)) or a major part of it (PrsA-(⌬Asp 154 -Gln 171 )) and PrsA-(⌬Thr 185 -Lys 202 ) to catalyze AmyQ secretion and/or support growth, although the essential activity is not the PPIase activity. An alternative explanation could be that the N-and C-terminal domains in these mutant PrsA proteins are incorrectly folded and they are consequently inactive in vivo rather than the inactivation being due to the absence of the parvulin-like domain. Since PrsA-(⌬Asp 154 -Gln 171 ) and PrsA-(⌬Thr 185 -Lys 202 ) were partially degraded in vivo, their conformation may be incorrect. Based on the structural model, ⌬Asp 154 -Gln 171 and ⌬Thr 185 -Lys 202 deletions are likely to disrupt the conformation of the central ␤-sheet. PrsA-(NϩC), however, was stable in vivo, suggesting that the Nand C-terminal domains in this mutant PrsA protein are in a native-like conformation. PrsA-(NϩC) also exhibited partial PrsA activity in the AmyQ secretion assay, suggesting that the fold is proper. Furthermore, the CD spectroscopic analysis of PrsA-(NϩC) and the full enzymatic activity of PrsA-PPI suggest that the PPIase domain and the flanking domains are able to fold correctly independently.
It is interesting that a similar parvulin domain deletion mutant of E. coli SurA as B. subtilis PrsA-(NϩC) may be functional in the assembly and folding of outer membrane proteins (8), and thus in contrast to PrsA its in vivo function may be independent of the parvulin-like domain. The PrsA-like protein of Lactococcus lactis may also function in a manner independent of the parvulin-like domain (44). However, these results need to be confirmed. In addition to SurA, there is in the E. coli periplasm also another protein functionally to similar PPIase, PpiD, which may interfere with the complementation analysis of functions of SurA mutants (4,8,45). PrsA is the only essential PPIase at the cell membrane-wall interface of B. subtilis and therefore functionally similar PPIases do not disturb its functional analysis. In the parvulin-like domain, most peptide insertions that affected the PrsA function in vivo were found to be located in three spots around the active center (see Fig. 2D). This suggests that despite the high tolerance to substitutions of single amino acid residues and peptide insertions the active center of the parvulin-like domain most probably is important for the PrsA function.
It was peculiar that in the parvulin-like and C-terminal domains many pentapeptide insertions partially disrupted the PrsA activity in the viability assay but did not at all affect the activity in the AmyQ secretion assay. We can put forward three possible explanations for this obvious discrepancy. (i) PrsA may be a multimeric protein and in the prsA3 mutant the insertion mutant PrsA proteins may interact with the PrsA3 protein, form functional heterodimeric or multimeric PrsA3-PrsA peptide complexes and thus restore AmyQ secretion in the prsA3 mutant by intermolecular complementation, but not the function essential for viability. (ii) An interaction of the insertion mutant PrsA proteins with PrsA3 may stabilize this unstable but functional PrsA protein and thereby restore AmyQ secretion in the prsA3 mutant. (iii) The insertion mutant PrsA proteins may interact with the quality-control proteases that degrade PrsA3 and prevent the degradation by competitive inhibition. The synthetic lethal phenotype of the PrsA36 protein in the prsA3 mutant supports the first hypothesis. The most likely explanation for the synthetic lethality is formation of functionally inactive dimeric or multimeric PrsA3-PrsA36 complexes and consequent inactivation of the PrsA3 protein and cell death.
In addition to the complementation of the AmyQ secretion defect in prsA3 and the growth defect of cells depleted of PrsA we also used exoproteome analysis to characterize the function of some of the mutant PrsA proteins. It was observed that when cells were partially depleted of the chromosomally coded wildtype PrsA the level of numerous secreted proteins in the exoproteome decreased, whereas the level of some other exoproteins remained about the same as that of the non-depleted cells. This result suggests that the stability and secretion of several exoproteins is dependent on PrsA. However, it is difficult to assess from our data, how much the PrsA depletion actually decreased the secretion of these PrsA-dependent exoproteins, since the PrsA depletion also caused significant cell lysis and increase of the level of contaminating cytoplasmic proteins in the exoproteome. The active site mutants PrsA-(⌬Met 172 -Lys 182 ) and PrsA-D154A were only partially able to restore the wild-type pattern of the exoproteome even though they fully restored growth in the PrsA-depleted cells, indicating that they are to some extent defective PrsA proteins. The underlying mechanism of the effects on the exoproteome is most probably independent of the PPIase activity, since the PrsA-H122A protein fully restored the wild-type pattern of the exoproteome. It has been shown that the FKBP-type PPIase domain of E. coli FkpA has a PPIase-independent chaperone activity in vitro (7). Also E. coli SurA parvulin (8) exhibits a general chaperone activity. Although our results indicated that PrsA does not have a general chaperone activity, a dedicated chaperone activity is not excluded.
Our results also showed that in addition to the PPIase domain, the N-and C-terminal domains are essential for the PrsA function in vivo. Hence, some non-PPIase activity of the flanking domains is required together with the PPIase domain activity. We propose at least three possible functional roles for the flanking domains. They may be interaction domains that bring the parvulin-like domain into contact with its substrates in the cell wall, in a manner similar to the flanking domains of some other PPIases they may mediate binding to other cellular components or substrates (3,46,47). Alternatively, PrsA might be a modular protein in which the parvulin-like domain and the flanking domains operate in completely unrelated essential functions. The synthetic lethal phenotype of PrsA36 in the prsA3 mutant suggests a third possible role. The flanking re-gions may be domains that are required for the putative formation of PrsA dimers or multimers in a manner similar to the dimerization of the Legionella pneumophila Mip protein, an FKBP-type PPIase (48).