Báo cáo khoa học: The periplasmic peptidyl prolylcis–transisomerases PpiD and SurA have partially overlapping substrate specificities doc

The periplasmic peptidyl prolyl cis–trans isomerases PpiD and SurA have partially overlapping substrate specificities Krista H. Stymest and Peter Klappa Department of Biosciences, University of Kent, Canterbury, UK During the process of protein folding, several poten- tially rate-limiting steps can occur, one of which is the cis–trans isomerization of proline peptide bonds, catal- ysed by a ubiquitously expressed group of peptidyl prolyl cis–trans isomerases (PPIases) [1,2]. Three major classes of PPIases have been identified in both prok- aryotes and eukaryotes: the cyclophilins, i.e. cyclospo- rin A-binding proteins, the FK506-binding proteins, and PPIases that have sequence similarity with the catalytic domain of the small PPIase parvulin [3,4]. The periplasmic space of Escherichia coli contains all three classes of PPIases, namely PpiA, a cyclophilin- type PPIase, FkpA, a FK506-binding PPIase, and two members of the parvulin-family, SurA and PpiD. SurA was originally isolated as a protein product of stationary-phase survival genes of E. coli [5–7]. Null mutants or mutants with reduced levels of SurA in the periplasm lead to a significant decrease in folding of the maltose-inducible porin LamB and the trimeric porins OmpC and OmpF [5,6]. It has been shown that SurA is important in maintaining outer membrane integrity by facilitating maturation of integral outer membrane proteins [8–10]. SurA contains two domains with high sequence similarity to parvulin, in addition Keywords peptidyl prolyl cis–trans isomerase; periplasmic space; protein folding; protein– protein interaction; substrate specificities Correspondence P. Klappa, Department of Biosciences, University of Kent, Canterbury CT2 7NJ, UK Fax: +44 1227 763912 Tel: +44 1227 823515 E-mail: p.klappa@kent.ac.uk (Received 19 February 2008, revised 14 April 2008, accepted 6 May 2008) doi:10.1111/j.1742-4658.2008.06493.x One of the rate-limiting steps in protein folding has been shown to be the cis–trans isomerization of proline residues, catalysed by a range of peptidyl prolyl cis–trans isomerases (PPIases). In the periplasmic space of Escherichia coli and other Gram-negative bacteria, two PPIases, SurA and PpiD, have been identified, which show high sequence similarity to the catalytic domain of the small PPIase parvulin. This observation raises a question regarding the biological significance of two apparently similar enzymes present in the same cellular compartment: do they interact with different substrates or do they catalyse different reactions? The substrate-binding motif of PpiD has not been characterized so far, and no biochemical data were available on how this folding catalyst recognizes and interacts with substrates. To char- acterize the interaction between model peptides and the periplasmic PPIase PpiD from E. coli , we employed a chemical crosslinking strategy that has been used previously to elucidate the interaction of substrates with SurA. We found that PpiD interacted with a range of model peptides indepen- dently of whether they contained proline residues or not. We further dem- onstrate here that PpiD and SurA interact with similar model peptides, and therefore must have partially overlapping substrate specificities. However, the binding motif of PpiD appears to be less specific than that of SurA, indicating that the two PPIases might interact with different substrates. We therefore propose that, although PpiD and SurA have partially overlapping substrate specificities, they fulfil different functions in the cell. Abbreviations DSG, disuccinimidyl glutarate; PDI, protein disulfide isomerase; PPIase, peptidyl prolyl cis–trans isomerase; scRNase, ‘scrambled’ ribonuclease A. 3470 FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS to an N-terminal peptide-binding domain (Fig. 1). In vitro, this N-terminal domain has the properties of a molecular chaperone, i.e. it prevents heat-denatured citrate synthase from aggregation [8], and is essential and sufficient for interaction with model peptides [11]. Recently, it was reported that SurA has a bi-partite substrate-binding area: while the N-terminal peptide- binding site appears to interact with unfolded proteins, the first parvulin domain confers substrate specificity by interacting specifically with aromatic residues in the tested peptides [12]. Studies with a recombinant frag- ment of SurA revealed that only the second, C-termi- nal, parvulin domain has catalytic activity [8]. PpiD was discovered in a genetic screen as a multi- copy suppressor of a surA deletion strain [13]. PpiD is anchored to the inner membrane via a single N-termi- nal transmembrane segment, with the catalytic parvu- lin-like domain (Fig. 1) facing the periplasmic space [13]. Deletion of the ppiD gene leads to an overall reduction in the level and folding of outer membrane proteins. The simultaneous deletion of both ppiD and surA genes was found to confer synthetic lethality, and therefore it was suggested that these PPIases have an overlapping substrate specificity [14]. However, this observation was recently disputed by Justice et al. [15]: in their experiments, a ppiD ⁄ surA double mutant did not show any loss in viability and only a quadruple mutant lacking all four periplasmic PPIases (SurA, PpiD, PpiA and FkpA) showed a significant decrease in the growth rate. The presence of two enzymes with apparently similar catalytic properties in the same compartment raises a question regarding their biological function. Do they catalyse different reactions or do they interact with dif- ferent substrates? The substrate-binding motif of SurA was recently identified, and the enzyme was shown to bind to proteins with the motif Ar-X-Ar, or modifica- tions of it, where Ar is an aromatic residue [16,17]. However, the substrate-binding motif of PpiD has not been characterized so far, and there are no biochemical data available on how this folding catalyst recognizes and interacts with its substrates. To address the question of how PpiD interacts with its substrates, we used chemical crosslinkers, which are powerful tools for the study of interactions between proteins, and can be applied to proteins that are avail- able in small amounts even in crude cell extracts [18–20]. Results To investigate how PpiD interacts with its substrates, and to compare its substrate specificity with that of SurA, we overexpressed PpiD as an N-terminally hexahistidine-tagged protein without leader sequence or transmembrane domain (Fig. 1). SurA was expressed as full-length mature protein in which a hexahistidine tag replaced the leader sequence, as pre- viously described [11]. Expression of both PPIases resulted in high yields of the respective proteins. The molecular masses for SurA and PpiD were approxi- mately 45 and 62 kDa, respectively. Sedimentation analysis showed that both proteins were found pre- dominantly in the soluble fraction, with < 5% in the pellet (data not shown). Using purified recombinant proteins for crosslinking experiments has occasionally been observed to produce false-positive results, as model substrates with very weak binding affinities can also be crosslinked to the purified protein (P. Klappa, unpublished results). To avoid such crosslinking artefacts, all crosslinking experiments were routinely carried out with whole E. coli cell lysates expressing the respective PPIases unless otherwise stated. We observed that addition of the chemical cross- linker disuccinimidyl glutarate (DSG) to a lysate exp- ressing recombinant PpiD resulted in crosslinking products [collectively indicated as (b)] with an appar- ent molecular mass of approximately 80 kDa (Fig. 2, lane 4), in addition to unmodified PpiD (a). Identical crosslinking products were also detected with purified PpiD (Fig. 2, lane 2), indicating that these additional crosslinking products were not proteins from the E. coli lysate interacting with PpiD. Both (a) and (b) were recognized by an antibody directed against PpiD (data not shown), whereas anti-hexahistidine serum only recognized band (a) (compare with Fig. 3, lanes 6 and 7). Electrospray mass spectroscopy analysis indi- Fig. 1. Domain constructs of recombinant SurA and PpiD. The numbering is based on the full-length sequences of SurA and PpiD. S, signal sequence; Pep, peptide-binding domain; TM, transmem- brane portion; Parv, parvulin-like domain; ?, segments of unknown function. The substrate-binding domain of SurA has been reported to be located in the peptide-binding domain and the first parvulin domain [12]. K. H. Stymest and P. Klappa Substrate interactions of periplasmic PPIases FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 3471 cated that addition of DSG generates a heterogeneous mixture of DSG–PpiD adducts. However, as DSG modifies lysine residues and thus changes the m ⁄ z ratio, a detailed analysis of the crosslink products (b) was not possible. From the observation that all bands were recognized by anti-PpiD serum, but only band (a) was detected using anti-hexahistidine serum, we pro- pose that, in conformation (b), the N-terminal hexahis- tidine tag is no longer exposed and is therefore not recognized by the hexahistidine antibody. No such effect was observed for chemical crosslinking of SurA (data not shown), indicating that the reduced mobility of conformation (b) did not merely result from the attachment of the chemical crosslinker to a hexahisti- dine tag. We believe that the bands (b) are most likely cross- linking artefacts that change the mobility of PpiD in SDS–PAGE, but do not affect substrate binding. PpiD, like SurA, interacts with a misfolded protein To investigate whether PpiD interacts with a misfolded protein, we used ‘scrambled’ ribonuclease A (scRNase) as a model substrate. We previously demonstrated that biotinylated scRNase could be chemically crosslinked to purified SurA [11]. Here we extended our studies to determine whether this technique could also be employed to study the binding properties of PpiD. To avoid potential artefacts through interaction with the biotin moiety, we used unmodified scRNase, which was incubated with crude E. coli lysates expressing recombinant SurA or PpiD. Samples without scRNase and DSG served as controls. Analysis was carried out by Western blotting of the gels, with subsequent detec- tion using anti-hexahistidine serum (Fig. 3). In the absence of DSG or scRNase, no crosslinking products could be detected (lanes 1–3 for SurA and 6–8 for PpiD, respectively). In the presence of DSG and scRN- ase, a single specific crosslinking product was observed for PpiD, with an approximate molecular mass of 75 kDa (lane 9). Crosslinking of scRNase to an E. coli lysate expressing SurA resulted in a crosslinking prod- uct with an approximate molecular mass of 60 kDa (lane 4). This result clearly showed that PpiD, like SurA, interacts with misfolded proteins. That this interaction was dependent on the native conformation of the PPIases was demonstrated by heat inactivation of the lysates. After heat treatment of the lysates (5 min at 95 °C) prior to the addition of scRNase and chemical crosslinking, the interaction between scRNase and the PPIases was strongly reduced (lanes 5 and 10, respectively). PpiD, like SurA, interacts with a radiolabelled model peptide without proline residues To facilitate analysis of the recognition motif of PpiD, we used a radiolabelled peptide, D-somatostatin, with the amino acid sequence AGSKNFFWKTFTSS, as a model substrate. [ 125 I]-Bolton–Hunter-labelled D-somatostatin was chemically crosslinked to recombi- Fig. 2. Chemical crosslinking of PpiD. Purified PpiD (5 lM, Pur) or E. coli lysate (lys) expressing recombinant PpiD in a total volume of 10 lL were incubated with DSG (final concentration 0.5 m M) in buf- fer B for 60 min at 0 °C or were left untreated. The samples were then analysed on 10% polyacrylamide gels with subsequent Coomassie Brilliant Blue staining. M, molecular mass marker. Fig. 3. Interaction of PPIases with scRNase. E. coli lysates expressing recombinant SurA or PpiD, respectively, were heat-inac- tivated at 95 °C for 5 min (lanes 5 and 10) or left untreated. The samples were then incubated with 100 l M scRNase or buffer B prior to crosslinking with DSG. After crosslinking, the samples were analysed on 10% polyacrylamide gels with subsequent electro- transfer onto poly(vinylidene) fluoride (PVDF) membranes. The sam- ples were probed with primary anti-hexahistidine serum, raised in mouse, and secondary anti-mouse serum conjugated to horseradish peroxidise, and visualized by enhance chemiluminescence. Cross- linking products are indicated with an asterisk. The positions of the molecular mass markers are indicated. Substrate interactions of periplasmic PPIases K. H. Stymest and P. Klappa 3472 FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS nant PpiD overexpressed in E. coli cell lysate (Fig. 4A). Crosslinking of the radiolabelled peptide to a cell lysate expressing SurA served as a positive con- trol, while a cell lysate without recombinant SurA or PpiD, served as a negative control (lanes 1 and 2). Cell lysates that contained recombinant PPIases showed crosslinking products, indicated by arrows (lanes 4 and 7, respectively). The lysate expressing SurA showed only one specific crosslinking product, with an approx- imate molecular mass of 47 kDa, but the lysate expressing PpiD showed two specific crosslinking prod- ucts with molecular masses of 64 and 82 kDa. As the addition of chemical crosslinker to PpiD resulted in two bands [(a) and (b), see Fig. 2], we believe that these crosslinking products are most likely due to interaction of the two PpiD bands with the radio- labelled peptide. In the absence of DSG, no crosslink- ing products were detected (Fig. 4A, lanes 3 and 6). Heat inactivation of the lysates (5 min at 95 °C) prior to addition of the peptide and chemical crosslinking inhibited the interaction between radiolabelled peptide and the PPIases (lanes 5 and 8), indicating that the interaction was specific for native proteins. We observed that all cell lysates, irrespective of whether they expressed recombinant PPIases or not, showed an unidentified crosslinking product of approximately 25 kDa, marked x, which has been observed previously [20]. We also noted that the interaction between PpiD and radiolabelled D-somatostatin was much weaker A B C Fig. 4. Interaction of PPIases with radiolabelled D-somatostatin. (A) [ 125 I]-Bolton–Hunter-labelled D-somatostatin (33 lM) was incu- bated with E. coli lysates expressing recombinant SurA, PpiD or nei- ther of the PPIases (Lys) in buffer B for 10 min at 0 °C in a total volume of 10 lL. As controls, E. coli lysates expressing recombinant SurA or PpiD, respectively, were heat-inactivated at 95 °C for 5 min (lanes 5 and 8) prior to addition of the radiolabelled peptide. Samples were subsequently incubated with DSG (final concentration 0.5 m M) for 60 min at 0 °C or were kept untreated. After crosslinking, the samples were analysed on 10% polyacrylamide gels with subse- quent autoradiography. The positions of the molecular mass markers are indicated. (B) [ 125 I]-Bolton–Hunter-labelled D-somatostatin (33 l M) was incubated with E. coli lysates expressing recombinant SurA or PpiD in the presence of 0.2% w ⁄ v Triton X-100 prior to crosslinking with DSG. After cross-linking, the samples were analy- sed on 10% polyacrylamide gels with subsequent autoradiography. The positions of the molecular mass markers are indicated. (C) [ 125 I]- Bolton–Hunter-labelled D-somatostatin (D-som, 10 l M) was incu- bated with an E. coli lysate expressing recombinant PpiD in the absence or presence of 50 l M (+) or 100 lM (++) scRNase. After crosslinking with DSG, the samples were analysed on 10% poly- acrylamide gels with subsequent electrotransfer onto poly(viny- lidene) fluoride (PVDF) membranes. The membranes were subjected to autoradiography (upper panel) and subsequently probed with primary anti-hexahistidine serum, raised in mouse, and second- ary anti-mouse serum conjugated to horseradish peroxidise, and visualized by enhance chemiluminescence (Western blot, lower panel). Crosslinking products are indicated by arrows. Exposure times for samples containing PpiD were on average 20 times longer than for samples containing SurA. ‘a’, crosslinking to PpiD band (a); ‘b’, crosslinking to PpiD band (b); x, unspecified crosslinking product. K. H. Stymest and P. Klappa Substrate interactions of periplasmic PPIases FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 3473 than that between SurA and the peptide, as exposure of the PpiD crosslinking autoradiographs for 20 times longer was required to obtain similar intensities of the crosslinking products. As the intensities of the scRN- ase crosslinking products were comparable between SurA and PpiD (compare with Fig. 3), we speculate that PpiD might bind preferentially to misfolded pro- teins rather than short peptides (see below). Qualitatively and quantitatively identical results were obtained when we substituted the E. coli lysates by purified SurA or PpiD, with the exception that the unspecific crosslinking product, denoted x, was no longer detectable (data not shown). Using an alterna- tive chemical crosslinker, bis(sulfosuccinimidyl) suber- ate, gave quantitatively and qualitatively identical results (data not shown). It is important to point out that D-somatostatin does not contain any proline residues, and therefore it is clear that the presence of a proline residue in the model peptide is not essential for interaction with PpiD. We recently demonstrated that the binding of pep- tides to purified SurA required hydrophobic interac- tions [11], and we therefore wished to determine whether this is also the case for the interaction of pep- tides with PpiD. Radiolabelled D-somatostatin was chemically crosslinked to a cell lysate overexpressing recombinant PpiD in the absence or presence of Tri- ton X-100. A cell lysate expressing SurA served as a positive control. As with SurA, binding of radiola- belled D-somatostatin to PpiD was strongly inhibited in the presence of Triton X-100 (Fig. 4B). This experi- ment demonstrated that the interaction of PpiD with peptides, like that of SurA, is detergent-sensitive and presumably requires hydrophobic interactions. A simi- lar observation was made for interaction between the radiolabelled peptide and the unidentified protein giving rise to the crosslinking product indicated x. To demonstrate that the peptide-binding site in PpiD is identical to the site that interacts with a mis- folded protein, we carried out a competition experi- ment (Fig. 4C). Chemical crosslinking of 10 lm radiolabelled D-somatostatin to a cell lysate expressing PpiD was reduced in the presence of excess scRNase (100 or 50 lm) (Fig. 4C, upper panel, lane 1 versus lanes 2 and 3). To show that the competition is not due to a crosslinking artefact, e.g. quenching of the crosslinker by the excess of scRNase, we also per- formed a Western blot of the same gel, and detected a crosslinking product between scRNase and PpiD using anti-hexahistidine serum (lower panel, lanes 2 and 3). From this experiment, we concluded that excess of scRNase could replace the radiolabelled peptide. We therefore suggest that peptides and misfolded proteins interact with the same substrate binding site in PpiD. Interestingly, when we carried out a competition experiment using an excess of unlabelled D-somato- statin (100 lm) to inhibit the binding of scRNase (20 lm), we could not detect any competitive effects (data not shown). The most likely interpretation of this result is that PpiD has a higher affinity for the mis- folded protein than for a small peptide (see below). Taken together, our results indicated that there is no principal difference between the two PPIases SurA and PpiD with respect to interaction with a misfolded pro- tein or a radiolabelled model peptide. PpiD interacted with a misfolded protein and a radiolabelled model peptide that did not contain any proline residues. The interactions were detergent-sensitive and required native PpiD. PpiD and SurA interact with different peptides To address the question of whether PpiD showed simi- lar substrate specificity to that of SurA, we chemically crosslinked various radiolabelled peptides to SurA or PpiD (Fig. 5A). A lysate expressing human archetypal protein disulfide isomerase (PDI) served as a control to demonstrate that absence of a crosslinking product is not due to low efficiency of radiolabelling of the peptide. PDI interacted with all the peptides tested to a similar extent, indicating that each peptide could be crosslinked with similar efficacy. We found that SurA showed strong interactions with D-somatostatin and somatostatin, while other peptides gave only weak crosslinking signals. In contrast, PpiD showed a strong interaction with most peptides tested. However, we noted that the interaction with radiolabelled D-somato- statin was rather weak compared to that with other peptides. For all the peptides tested, we observed crosslinking products related to the PpiD bands (a) and (b). For efficient crosslinking of a peptide to a protein, correct spatial orientation of the crosslinked residues in the target molecules is essential. It is therefore con- ceivable that failure to detect crosslinking products in the case of SurA and peptides other than D-somato- statin and somatostatin was due to unfavourable ori- entation of crosslinkable residues. We therefore carried out competition experiments, using cell lysates express- ing recombinant SurA or PpiD, and radiolabelled D-somatostatin (10 lm) in the presence of an excess of unlabelled peptides (100 lm). Although D-somatostatin did not show very strong binding to PpiD, we used it in our experiments to enable direct comparison with SurA. Competition between radiolabelled D-somato- Substrate interactions of periplasmic PPIases K. H. Stymest and P. Klappa 3474 FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS statin (10 lm) and unlabelled D-somatostatin (100 lm) served as a positive control. Figure 5B shows a typical example of these experiments. Unlabelled peptide M did not compete with radiola- belled D-somatostatin for binding to SurA, but showed competition for binding to PpiD. This result is in excellent agreement with the results in Fig. 5A, in which radiolabelled peptide M interacts with PpiD, but not SurA. Unlabelled peptide A, however, inhib- ited the interaction between radiolabelled D-somato- statin and SurA, but was far less efficient with PpiD. This result is in line with experiments using radio- labelled peptide A: while radiolabelled peptide A bound to SurA, no interaction with PpiD was detected (data not shown). Table 1 summarizes the competition experiments. We used the competition between radiolabelled D-somatostatin and unlabelled D-somatostatin as the reference point: strong competition between radiola- belled D-somatostatin and unlabelled peptide is indi- cated by +++, no competition by ). A B Fig. 5. Interaction of PPIases with model peptides. (A) The indi- cated [ 125 I]-Bolton–Hunter-labelled peptides (30 lM) were incubated with E. coli lysates expressing recombinant SurA, PpiD or human PDI in buffer B for 10 min at 0 °C in a total volume of 10 lL. After crosslinking with DSG, the samples were analysed on 10% poly- acrylamide gels with subsequent autoradiography. The positions of the molecular mass markers are indicated. (B) [ 125 I]-Bolton–Hunter- labelled D-somatostatin (D-som, 10 l M) was incubated with E. coli lysates expressing recombinant SurA or PpiD in the presence of 100 l M of the indicated unlabelled peptides. A sample without unla- belled peptides served as controls. After crosslinking with DSG, the samples were analysed on 10% polyacrylamide gels with sub- sequent autoradiography. ‘a’, crosslinking to PpiD band (a); ‘b’, crosslinking to PpiD band (b); x, unspecified crosslinking product. Table 1. Competition between peptides and radiolabelled D-somatostatin for binding to SurA and PpiD. E. coli cell lysates expressing recombinant SurA or PpiD were incubated with 10 l M radiolabelled D-somatostatin in the presence of a 10-fold excess of the indicated unlabelled peptides prior to crosslinking. Samples were subsequently incubated with DSG (final concentration 0.5 m M) for 60 min at 0 °C. After crosslinking, the samples were analysed on 10% polyacrylamide gels with subsequent autoradiog- raphy. Quantification was performed with a Bio-Rad (Hemel Hemp- stead, UK) phosphoimager. A sample without competing peptides served as a control. Competition was expressed as reduction in the intensity of the respective crosslinking product relative to the con- trol without unlabelled peptides. +++, strong competition (intensity of crosslinking product < 10% of control); ++, moderate competi- tion (intensity of crosslinking product between 10% and 30% of control); +, weak competition (intensity of crosslinking product between 50% and 80% of control); ), no competition (intensity of crosslinking product > 80% of control). Peptide SurA PpiD AGSKNFFWKTFTSS (D-som) +++ +++ AGCKNFFWKTFTSC (somatostatin) +++ +++ AASKNFFWKTFTSS +++ +++ AGAKNFFWKTFTSS +++ ++ AGSKAFFWKTFTSS ++ ++ AGSKNAFWKTFTSS (F6A) + +++ AGSKNFAWKTFTSS (F7A) + + AGSKNFFAKTFTSS (W8A) ) + AGSKNFFWKAFTSS +++ +++ AGSKNYFWKTFTSS (F6Y) + +++ AGSKNYFAKTFTSS (F7Y) ) + AGSKNWFWKTFTSS +++ +++ AGSKNFFWKTFT +++ ) AGSKNFFWKT +++ ) AGSKNYFWKSAS (peptide S) )) AGSKNFFWKS (peptide A) +++ ) AASKAFFWKS +++ ) AGSKNFFWAT +++ ) TKWFFNKSGA +++ ) - -SKNFFWKTFT +++ ) - -SKNFFWKT )) - - - -NFFWKT )) INLKALAALAKKIL (peptide M) ) +++ WEYIPNV + ) WEYIP )) PTIKFFNGDTASPK (peptide P) ) +++ K. H. Stymest and P. Klappa Substrate interactions of periplasmic PPIases FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 3475 Our results demonstrate that changing the phenylal- anine at position 6 to alanine (F6A) in the D-somato- statin sequence prevented this peptide from competing with the binding of radiolabelled D-somatostatin to SurA. Interestingly, the same modified peptide showed strong competition for the interaction of radiolabelled D-somatostatin with PpiD. Changing the phenylalanine at position 7 to alanine (F7A) reduced the competitive effect of the peptide for the interaction with both SurA and PpiD. A similar result was observed for the tryp- tophan to alanine (W8A) modification. Changing the phenylalanine at position 6 to a tyrosine (F6Y) pre- vented the peptide from competing with the binding of radiolabelled D-somatostatin to SurA, but not to PpiD. We noticed that peptides shorter than 14 amino acids did not efficiently compete for the binding of radiolabelled D-somatostatin to PpiD. Interestingly, peptide S did not compete for the binding to PpiD, although radiolabelled peptide S interacted with PpiD (Fig. 5A, lane 13). This result is most likely due to the increase in the size of peptide S during direct radio- labelling: treatment of peptides with [ 125 I] Bolton– Hunter labelling reagent results in an increase in the length of the peptide chain by one residue. We there- fore suggest that the interaction of PpiD with model peptides is dependent on their size. Taken together, our results show that the interaction between peptides and SurA is strongly dependant on the motif FFW in the peptide, whereas the substrate specificity for PpiD appears to be less specific. Discussion In native polypeptides, about 5–7% of the peptidyl prolyl bonds are in the cis configuration, and almost half of the 1453 non-redundant protein structures in the protein database contain at least one cis peptidyl prolyl bond [21]. Conversion of the trans peptidyl prolyl bond into the cis conformation, which is cataly- sed by peptidyl prolyl cis–trans isomerases, has been reported to be essential for the correct folding of many proteins, e.g. the folding of outer membrane proteins [5,13] and periplasmic proteins as well as certain toxins [22] in prokaryotes. In the periplasmic space of E. coli, two PPIases, SurA and PpiD, with sequence similarity to the cata- lytic domain of parvulin have been identified. This observation raises a question regarding the biological significance of these two PPIases in the same cellular compartment: do they interact with different sub- strates or do they catalyse different reactions? The simultaneous deletion of both ppiD and surA genes was reported to confer synthetic lethality, and hence it was suggested that the PPIases have overlapping substrate specificity [14]. However, Justice et al. [15] challenged this observation, showing that a double deletion of the ppiD and surA genes did not result in loss of viability. To investigate how these PPIases interact with their substrates and whether they have potentially overlap- ping substrate specificities, we employed a crosslinking approach, which we had used previously to determine the interaction between peptides and other folding catalysts [18,20,23]. PpiD interacts with model peptides and a misfolded protein We recently showed that model peptides and a mis- folded protein, ‘scrambled’ RNase A, interacted specif- ically with purified recombinant SurA from E. coli [11]. Using a similar crosslinking approach, we demon- strate here that a recombinant fragment of PpiD, lack- ing the leader sequence and the transmembrane segment, also interacted specifically with radiolabelled model peptides and scRNase. As with SurA, we found that the interaction between PpiD and model peptides was independent of the presence of a proline residue within the peptide. This result is in line with the results for other protein isomerases, such as PDI [18] and the PPIases trigger factor [24] and FkpA [25]. As with PDI [18] and SurA [11], the interaction of PpiD with model peptides is sensitive to the presence of Triton X-100, which indi- cates that hydrophobic interactions play a role in the initial binding of peptides to PpiD. Triton X-100 is widely used for the functional recovery of intracellular soluble and membrane-bound proteins, and therefore it is unlikely that the detergent interferes with the structure of the binding site. This was confirmed by our observation that addition of Triton X-100 did not interfere with the interaction between scRNase and PpiD or SurA (data not shown). We speculate that the interaction between PPIases and a misfolded pro- tein is stronger than that with a short peptide, proba- bly due to a more extensive array of interactions, which include interactions other than hydrophobic interactions, e.g. electrostatic interactions and hyd- rogen bonds. Interaction of PPIases with model peptides The substrate-binding motif of SurA was recently identified, and the enzyme was shown to bind to peptides with the motif Ar-X-Ar, or modifications of Substrate interactions of periplasmic PPIases K. H. Stymest and P. Klappa 3476 FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS it, where Ar is an aromatic residue [10,16,17]. Recent reports also showed that SurA has affinity for pep- tides enriched in aromatic residues with positively charged residues in the vicinity [26]. Our results con- firmed these observations; however, based on our crosslinking competition data, we speculate that the substrate specificity for SurA is more extensive and comprises more amino acids. Furthermore, our data support previous reports indicating that the conforma- tion of the peptide plays an important role in binding to SurA [8–10]. For example, peptide S contained the required binding motif Ar-X-Ar, but was not found to interact with SurA. It is unlikely that the lack of the phenylalanine residue in position 11 (F11) was the sole reason for this result, as peptide A, which also lacked F11, showed efficient binding to SurA. We therefore propose that the alanine and serine residues at the C-terminus of peptide S induce a structural change such that the Ar-X-Ar motif is no longer accessible. Replacement of phenylalanine 6 (F6) with alanine in D-somatostatin strongly reduced the inter- action with SurA, but did not affect the interaction with PpiD. Interestingly, scRNase does not contain an Ar-X-Ar motif; however, the efficient interaction between this misfolded protein and SurA indicates that SurA might recognize hydrophobic patches in a tertiary structure, which places two or more aromatic amino acids in close proximity. Alternatively, scRNase might bind predominantly to the general substrate-binding site as identified by Xu et al. [12]. Based on their extensive study of interactions between model pep- tides and SurA by X-ray crystallography, it was sug- gested that the interaction between peptides and SurA is facilitated by the N-terminal binding domain, as well as the first parvulin domain. The N-terminal binding domain appears to act as a general interac- tion site for misfolded proteins, whereas the first parvulin domain confers substrate specificity through very specific interactions between the aromatic resi- dues in substrate molecules and the binding site of SurA [12]. PpiD, like SurA, can interact with aromatic residues in the model peptide D-somatostatin. However, the substrate specificity for the interaction with PpiD does not appear to be restricted to aromatic residues, as demonstrated by the binding of peptide M, which does not contain aromatic residues. We therefore propose that the substrate specificity of PpiD is less specific than that for SurA. We speculate that the substrate specificity of PpiD is determined more by the hydro- phobicity of residues in the model peptides than the presence of aromatic residues. Biological significance In our experiments, we confirmed the finding by Bitto & McKay that the Ar-X-Ar binding motif is essential for the interaction between SurA and model peptides [16,17]. It was demonstrated that this is the signature motif for specific outer membrane proteins, which have been proposed to be the predominant substrates of SurA [9,10]. Our results show that even short peptides (< 11 amino acids) interacted with SurA, as long as they contained this signature motif and probably exhibited some other structural requirements [9]. The interaction with PpiD, however, required a longer pep- tide chain, with at least 13 amino acids. This difference might reflect the different biological functions of the two PPIases. We propose that the main biological function of SurA is to facilitate the folding of predom- inantly outer membrane proteins, which might contain a simple signature motif for the interaction with SurA. Although the precise mechanisms of targeting of outer membrane proteins to the outer membrane has not yet been established [27], it is likely that translocation of most outer membrane proteins from the inner to the outer membrane occurs via the periplasmic space, thus allowing soluble SurA to efficiently interact with incor- rectly folded substrates. PpiD, however, is anchored to the inner membrane, with the catalytically active site facing the lumen of the periplasmic space. This particular localization makes it likely that PpiD is pre- dominantly involved in the folding of inner membrane- associated proteins rather than soluble proteins. PpiD might therefore interact with a variety of slowly fold- ing proteins, for which a simple and specific recogni- tion motif does not exist. We therefore propose that PpiD has a broader substrate specificity than that of SurA: while PpiD acts as a general folding catalyst rec- ognizing various binding motifs, SurA requires only a short and characteristic recognition motif specific for outer membrane proteins. Experimental procedures ‘Scrambled’ RNase A, the homobifunctional crosslinking reagent disuccinimidyl glutarate (DSG), and all other chem- icals were obtained from Sigma (Poole, UK). The [ 125 I] Bol- ton–Hunter labelling reagent, anti-hexahistidine serum, enhanced chemiluminescence reagent and X-ray films were purchased from Amersham GE Healthcare (Little Chalfont, UK). Peptides were synthesized as described previously for other peptides [18], and peptide sequences are given in Table 1. Antibodies against recombinant SurA and PpiD were raised by injection of the purified proteins (see below) into New Zealand White rabbits as described previously K. H. Stymest and P. Klappa Substrate interactions of periplasmic PPIases FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 3477 [23]. Secondary antibodies coupled to horseradish peroxi- dase were purchased from Dako (Glostrup, Denmark). Cloning of PpiD and purification of proteins The cloning of the surA gene from E. coli has been described previously [11]. The gene encoding mature PpiD without the transmembrane portion (positions 37–623) was amplified from the E. coli genome by PCR using the fol- lowing primers: forward primer, 5’-TTTTTTTTCATAT GGGAGGCAATAACTACGCCGCAAAAG-3’; reverse primer, 5’-TTTTTTTTCTCGAGCTATTATTGCTGTTC CAGCGCATCGC-3’. These primers allowed the insertion of an NdeI site at the N-terminus and an Xho I site at the C-terminus. The primers complementary to the 3’ end included a stop codon. The inserts were cloned between the NdeI and XhoI sites of pLWRP51, a modified pET23d vec- tor, which contained an insert coding for an initiating methionine residue followed by a hexahistidine tag, as described previously [11]. The construct was verified by DNA sequencing. Protein expression and purification of PpiD were per- formed as described for SurA [11]. Cloning and protein expression of human PDI were performed as described previ- ously [11]. [ 125 I]-Bolton–Hunter labelling of peptides was per- formed as recommended by the manufacturer of the reagent. Binding of peptides and scRNase After precipitation with trichloroacetic acid, the radio- labelled peptides were dissolved in buffer B (100 mm NaCl, 25 mm KCl, 25 mm sodium phosphate buffer pH 7.5). Labelled peptides or scRNase were added to E. coli lysates expressing recombinant SurA, PpiD or human PDI in buf- fer B. The samples (10 lL) were incubated for 10 min on ice prior to crosslinking [11]. Chemical crosslinking Crosslinking was performed using the homobifunctional crosslinking reagent disuccinimidyl glutarate (DSG) [11]. Crosslinking solution (2.5 mm DSG in buffer B) was added to the samples at one-fifth of the sample volume. The reaction was carried out for 60 min at 0 °C. Crosslink- ing was stopped by the addition of SDS–PAGE sample buffer [11]. The samples were subjected to electrophoresis in 10% SDS polyacrylamide gels. Acknowledgements We wish to thank Kevin Howland and Judy Hardy (Department of Biosciences, University of Kent, Can- terbury, UK) for synthesis of the peptides, and the Wellcome Trust for establishment of a Protein Science Facility. This work was supported by a Biotechnology and Biological Sciences Research Council PhD studentship to KHS. References 1 Fischer G, Tradler T & Zarnt T (1998) The mode of action of peptidyl prolyl cis ⁄ trans isomerases in vivo: binding vs. catalysis. FEBS Lett 426, 17–20. 2 Bang H, Pecht A, Raddatz G, Scior T, Solbach W, Brune K & Pahl A (2000) Prolyl isomerases in a minimal cell. Catalysis of protein folding by trigger factor from Myco- plasma genitalium. Eur J Biochem 267, 3270–3280. 3 Schmid FX, Mayr LM, Mucke M & Schonbrunner ER (1993) Prolyl isomerases: role in protein folding. Adv Protein Chem 44, 25–66. 4 Gothel SF & Marahiel MA (1999) Peptidyl-prolyl cis– trans isomerases, a superfamily of ubiquitous folding catalysts. Cell Mol Life Sci 55, 423–436. 5 Lazar SW & Kolter R (1996) SurA assists the folding of Escherichia coli outer membrane proteins. J Bacteriol 178, 1770–1773. 6 Lazar SW, Almiron M, Tormo A & Kolter R (1998) Role of the Escherichia coli SurA protein in stationary- phase survival. J Bacteriol 180, 5704–5711. 7 Tormo A, Almiron M & Kolter R (1990) surA, an Escherichia coli gene essential for survival in stationary phase. J Bacteriol 172, 4339–4347. 8 Behrens S, Maier R, de Cock H, Schmid FX & Gross CA (2001) The SurA periplasmic PPIase lacking its parvulin domains functions in vivo and has chaperone activity. EMBO J 20, 285–294. 9 Behrens S (2002) Periplasmic chaperones – new structural and functional insights. Structure 10, 1469– 1471. 10 Hennecke G, Nolte J, Volkmer-Engert R, Schneider- Mergener J & Behrens S (2005) The periplasmic chaper- one SurA exploits two features characteristic of integral outer membrane proteins for selective substrate recogni- tion. J Biol Chem 280, 23540–23548. 11 Webb HM, Ruddock LW, Marchant RJ, Jonas K & Klappa P (2001) Interaction of the periplasmic pepti- dylprolyl cis–trans isomerase SurA with model peptides. The N-terminal region of SurA is essential and suffi- cient for peptide binding. J Biol Chem 276, 45622– 45627. 12 Xu X, Wang S, Hu YX & McKay DB (2007) The peri- plasmic bacterial molecular chaperone SurA adapts its structure to bind peptides in different conformations to assert a sequence preference for aromatic residues. J Mol Biol 373, 367–381. 13 Dartigalongue C & Raina S (1998) A new heat-shock gene, ppiD, encodes a peptidyl-prolyl isomerase required for folding of outer membrane proteins in Escherichia coli. EMBO J 17, 3968–3980. Substrate interactions of periplasmic PPIases K. H. Stymest and P. Klappa 3478 FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 14 Rizzitello AE, Harper JR & Silhavy TJ (2001) Genetic evidence for parallel pathways of chaperone activity in the periplasm of Escherichia coli. J Bacteriol 183, 6794– 6800. 15 Justice SS, Hunstad DA, Harper JR, Duguay AR, Pinkner JS, Bann J, Frieden C, Silhavy TJ & Hultgren SJ (2005) Periplasmic peptidyl prolyl cis–trans isomerases are not essential for viability, but SurA is required for pilus biogenesis in Escherichia coli. J Bacteriol 187, 7680– 7686. 16 Bitto E & McKay DB (2003) The periplasmic molecular chaperone protein SurA binds a peptide motif that is characteristic of integral outer membrane proteins. J Biol Chem 278, 49316–49322. 17 Bitto E & McKay DB (2004) Binding of phage-display- selected peptides to the periplasmic chaperone protein SurA mimics binding of unfolded outer membrane proteins. FEBS Lett 568, 94–98. 18 Klappa P, Hawkins HC & Freedman RB (1997) Inter- actions between protein disulphide isomerase and pep- tides. Eur J Biochem 248, 37–42. 19 Freedman RB & Klappa P (1998) Protein disulphide- isomerase: a catalyst of thiol:disulphide interchange and associated protein folding. In Molecular Biology of Chaperones and Folding Catalysts (Bukau B, ed), pp. 437–459. Harwood Academic Press, London. 20 Klappa P, Ruddock LW, Darby NJ & Freedman RB (1998) The b’ domain provides the principal peptide- binding site of protein disulfide isomerase but all domains contribute to binding of misfolded proteins. EMBO J 17, 927–935. 21 Reimer U, Scherer G, Drewello M, Kruber S, Schut- kowski M & Fischer G (1998) Side-chain effects on peptidyl-prolyl cis ⁄ trans isomerisation. J Mol Biol 279, 449–460. 22 Ruddock LW, Coen JJ, Cheesman C, Freedman RB & Hirst TR (1996) Assembly of the B subunit pentamer of Escherichia coli heat-labile enterotoxin. Kinetics and molecular basis of rate-limiting steps in vitro. J Biol Chem 271, 19118–19123. 23 Klappa P, Stromer T, Zimmermann R, Ruddock LW & Freedman RB (1998) A pancreas-specific glycosylated protein disulphide-isomerase binds to misfolded proteins and peptides with an interaction inhibited by oestro- gens. Eur J Biochem 254, 63–69. 24 Scholz C, Mucke M, Rape M, Pecht A, Pahl A, Bang H & Schmid FX (1998) Recognition of protein sub- strates by the prolyl isomerase trigger factor is indepen- dent of proline residues. J Mol Biol 277, 723–732. 25 Bothmann H & Pluckthun A (2000) The periplasmic Escherichia coli peptidylprolyl cis,trans-isomerase FkpA. I. Increased functional expression of antibody fragments with and without cis-prolines. J Biol Chem 275, 17100– 17105. 26 Alcock FH, Grossmann JG, Gentle IE, Likic VA, Lith- gow T & Tokatlidis K (2008) Conserved substrate bind- ing by chaperones in the bacterial periplasm and the mitochondrial intermembrane space. Biochem J 409, 377–387. 27 Danese PN & Silhavy TJ (1998) Targeting and assembly of periplasmic and outer-membrane proteins in Escheri- chia coli. Annu Rev Genet 32, 59–94. K. H. Stymest and P. Klappa Substrate interactions of periplasmic PPIases FEBS Journal 275 (2008) 3470–3479 ª 2008 The Authors Journal compilation ª 2008 FEBS 3479 . The periplasmic peptidyl prolyl cis–trans isomerases PpiD and SurA have partially overlapping substrate specificities Krista H. Stymest and Peter. double deletion of the ppiD and surA genes did not result in loss of viability. To investigate how these PPIases interact with their substrates and whether they have