Conformation of a plasmid replication initiator protein affects its proteolysis by ClpXP system

Abstract

Proteins from the Rep family of DNA replication initiators exist mainly as dimers, but only monomers can initiate DNA replication by interaction with the replication origin (ori). In this study, we investigated both the activation (monomerization) and the degradation of the broad-host-range plasmid RK2 replication initiation protein TrfA, which we found to be a member of a class of DNA replication initiators containing winged helix (WH) domains. Our in vivo and in vitro experiments demonstrated that the ClpX-dependent activation of TrfA leading to replicationally active protein monomers and mutations affecting TrfA dimer formation, result in the inhibition of TrfA protein degradation by the ClpXP proteolytic system. These data revealed that the TrfA monomers and dimers are degraded at substantially different rates. Our data also show that the plasmid replication initiator activity and stability in E. coli cells are affected by ClpXP system only when the protein sustains dimeric form.

Introduction

Proteins belonging to the Rep family of DNA replication initiators exist mainly as dimers, however, only Rep monomers can initiate DNA replication by the interaction with direct repeat sequences (iterons) of replication origin (ori).^1,2 The replication initiation protein TrfA of broad-host-range plasmid RK2 exists in two translational variants, 44 and 33 kDa.^3,4 Both proteins were found in the cell largely in the dimeric form and both bind iterons located in RK2 oriV as monomers.^5–7 The binding of TrfA monomers to oriV results in DNA helix destabilization (opening).⁸ The replication activity of TrfA protein could be detected in vitro by the formation of a substantially unwound form of the supercoiled plasmid DNA (FI*), which requires active TrfA monomers and the combined activity of bacterial replication proteins, including a helicase.^9,10 Many TrfA mutants with well characterized phenotypes have been described, however, despite attempts to obtain TrfA crystals, the structure of the protein has not yet been determined.

The activation of Rep proteins by converting dimers to replicationally proficient monomers is mediated by E. coli Hsp70 chaperone protein DnaK together with DnaJ, GrpE and the Clp/Hsp100 ATPases ClpA and ClpX.^11–15 Dimers of TrfA are monomerized by the DnaK/ClpB system¹⁰ or by ClpX.¹⁶ The ClpX-dependent activation of TrfA results in TrfA monomers with a high ability to bind iterons and initiate RK2 replication.

Beside chaperone-like activity, ClpX has been shown to be involved in ATP-dependent protein degradation through a transient association with the catalytic protease ClpP.¹⁷ Structural studies have demonstrated that one or two hexamers of ClpX bind to a barrel-like structure comprised of two heptameric rings of ClpP to form the ClpXP protease.¹⁸ ClpP also combines with hexamers of the other member of the Clp/Hsp100 protein family, namely the E. coli protein ClpA.¹⁹ The ClpP proteases alone do not degrade native proteins or even unfolded peptides.^20,21 The ATPase subunit of the proteolytic complex provides substrate specificity. ClpX or ClpA binds the specific substrate, unfolds it in an ATP-dependent reaction, and translocates the polypeptide into the protease degradation chamber.^22,23 ClpXP and ClpAP proteases degrade substrates processively starting from one end of the substrate protein. A specific sequence, comprising a few amino acids, called the tag or proteolytic signal, is required for substrate identification by the ATPase subunit.²⁴ Proteomic discovery of more than 50 cellular substrates of the ClpXP system revealed five classes of ClpX recognition signals in E. coli.²⁵

Because of Hsp100 protease activity, Rep proteins are unstable in E. coli cells, but chaperone components of the Hsp100 system are required for Rep protein activation.^26–31 The overall cellular activity of Rep proteins is therefore affected by both molecular chaperones and proteases. The interplay of activation and proteolysis of DNA replication initiators by Clp/Hsp100 systems remains ambiguous. It is not known for example if the activated Rep protein monomers are subjected to degradation or if conformation of replication initiators affects its proteolysis and therefore intracellular stability.

In this study we investigated both chaperone-dependent activation and proteolysis of the plasmid RK2 replication initiation protein TrfA. The in vitro tests and the analysis of intracellular TrfA stability revealed that the protein is degraded by the ClpXP system. We also demonstrate that monomerization of TrfA by ClpX inhibits its subsequent degradation by the ClpXP protease. Moreover, mutations within the TrfA dimerization interface affecting dimer formation result in a different efficiency of TrfA proteolysis by ClpXP and thus its stability in E. coli cells.

Results

TrfA is proteolysed by ClpXP

We previously demonstrated that the TrfA protein is activated by ClpX, which results in TrfA monomers proficient in plasmid RK2 replication initiation.¹⁶ It was not known, however, whether the intracellular level of TrfA protein is affected by proteolysis by ClpXP or other proteolytic systems. To address this question we analyzed the stability of the wild-type TrfA in E. coli cells in the presence of a protein synthesis inhibitor, a technique previously used for testing stability of protease substrates. [See Ref.25, and Materials and Methods] Because of low expression of trfA gene from it's natural promoter we had to use an inducible promoter and chemiluminescence detection for the protein stability tests (see Materials and Methods). The level of TrfA protein after induction was not high and comparable to the natural level of DnaB helicase [Fig. 1(A,B)]. Samples taken at the various time points were analyzed using SDS-PAGE and Western blotting with TrfA specific antibodies [Fig. 1(A)]. A substantial decrease in the amount of TrfA protein in the course of the experiment and its complete disappearance after 60 min indicate the TrfA degradation in E. coli cells [Fig. 1(A)]. A similar intracellular instability of TrfA was observed in pulse-chase experiment (data not shown) indicating that some protease system(s) is (are) involved in TrfA degradation.

Figure 1.

TrfA stability in E. coli cells (panel A). The stability of TrfA protein was analyzed after inhibition of translation as described under “Materials and Methods.” The intracellular stability of His6-TrfA was tested in E. coli wild-type and clpX, clpP, clpXP mutant strains. Densitometry analysis of Western blots enabled an estimation of TrfA half-life time in the analyzed strains, displayed in table. The stability of DnaB protein was analyzed in E. coli wild-type cells similarly as for TrfA protein (panel B). The dnaB gene was expressed constitutively from it's natural promoter. Anti-DnaB and anti-TrfA sera giving similar detection level, SuperSignal West Pico and chemiluminescence substrate kit was used for TrfA and DnaB analysis.

To test whether ClpX together with the ClpP proteolytic subunit control the intracellular concentration of TrfA, we compared TrfA stability in an E. coli wild-type strain with TrfA protein stability in E. coli clpX, clpP and clpXP mutants. In comparison to the wild-type strain, TrfA was more stable in the clpX mutant [Fig. 1(A)]. The protein half-life in the clpX mutant was 25 min in comparison to 10 min in the wild-type strain [Fig. 1(A)], indicating that ClpX is indeed involved in TrfA degradation. Moreover, we observed stabilization of TrfA in clpXP and also in clpP mutant strains, with protein half-life of approximately 60 min in either strain [Fig. 1(A)]. The higher stability of TrfA in the clpP strain versus the clpX strain indicates that, in addition to ClpXP, ClpAP could be involved in TrfA degradation.

To further analyze TrfA proteolysis by ClpXP we conducted in vitro experiments with purified proteins. The protein concentrations used during our in vitro assays were in the micromolar range, typically 1.8 μM. In this concentration, the wild-type TrfA remains as a dimer. It has been demonstrated by crosslinking experiments that once TrfA dimer is formed, there is little dissociation of the protein subunits. TrfA remains as dimer to the final protein concentration of 0.125 μM.³² To perform proteolytic tests in vitro the incubation of purified wild-type TrfA with ClpX and ClpP was followed by electrophoretic analysis. A substantial decrease in the wild-type TrfA protein concentration in the samples taken during the course of the experiment was observed, regardless of whether native TrfA protein [Fig. 2(A)] or N-terminal hexahistidine-tagged TrfA variants [Fig. 6(A)] were subjected to ClpXP-dependent degradation. When ClpX or ClpP were omitted no TrfA degradation was detected indicating requirement for both ClpX and ClpP in the reaction [Fig. 2(A) lane 6 and (B) lanes 1–4]. Interestingly, even 2 h incubation of TrfA with ClpXP did not result in complete proteolysis [Fig. 2(A) lane 6]. Activity tests showed that the remaining protein is replicationally active indicating the presence of TrfA monomers that must have been generated during incubation (Fig. 9 supporting data). Also, TrfA incubation with ClpXP resulted in the appearance of a specific electrophoretic band corresponding to a lower molecular mass protein of approximately 27 kDa [Fig. 2(A) lanes 3–5 and (B) lane 2]. In a control reaction, where TrfA was incubated without protease, no such additional band was found [Fig. 2(A), lane 6]. This suggested that the observed electrophoretic band was a result of TrfA proteolysis. This accumulating protein fragment, designated TrfA* [Fig. 2(A,B)], was extracted from the gel and subjected to N-terminal sequencing. TrfA* consisted of a 198 aa long C-terminal TrfA fragment starting at position 185 of the wild-type protein. These results indicated that TrfA is most likely proteolyzed from the N terminus.

Figure 2.

TrfA is proteolyzed by ClpXP. A proteolytic assay containing native 33-kDa TrfA as the subtrate was performed as described in the “Materials and Methods.” Time course (panel A) and control experiment where TrfA was incubated with ClpX or ClpP alone (panel B) are presented. Reactions were analyzed by SDS-PAGE and Coomassie staining. White arrow indicates the position of TrfA protein, black arrow indicates the position of the TrfA proteolysis product, TrfA*.

Figure 6.

Mutations within the TrfA dimer interface inhibit the protein proteolysis by ClpXP. Proteolytic reactions containing His6-TrfA (A), His6-TrfA(S257F) (B), and His6-TrfA(G254D/S267L) (C) were assembled as described under “Materials and Methods.” Incubation was performed for indicated time and products analyzed by SDS-PAGE and Coomassie staining.

ClpX-dependent activation of TrfA inhibits its subsequent proteolysis

Because TrfA could be degraded by the ClpXP protease complex, but it was also activated by ClpX,¹⁶ the question arose if TrfA activation affects its subsequent proteolysis. To address this, a two-step experiment was designed [Fig. 3(A)]. In the first step TrfA was activated by ClpX, whereas in the second step TrfA proteolysis by ClpXP was monitored. After incubation of TrfA with ClpX (first step) an aliquot of the sample was analyzed for TrfA replication activity as measured by the formation of FI*. The remaining sample after addition of ClpP was analyzed for TrfA degradation (second step).

Figure 3.

Activation by ClpX inhibits TrfA proteolysis. The experimental design is presented in panel (A). For activation (first step), His6-TrfA protein was incubated with or without (control reaction) ClpX. After activation, a portion of the reaction mixture was analyzed for TrfA activity in the FI* assay (see Materials and Methods) (panel B). The appearance of the FI* band indicates that the TrfA protein has been activated. The TrfA activation reaction was followed by the proteolytic assay (second step), which was initiated by the addition of ClpP or ClpXP. After incubation for the indicated time, samples were analyzed by SDS-PAGE and Coomassie staining (panel C) as described under “Materials and Methods.” The quantity of TrfA protein was estimated by densitometry and plotted (panel C).

When the wild-type TrfA was incubated in the presence of ClpX and ATP under conditions which result in TrfA monomerization,¹⁶ the formation of the FI* indicated that the protein was indeed activated by ClpX [Fig. 3(B) lane 3]. In the control reaction with no ClpX added to the incubation mixture, no FI* formation was detected [Fig. 3(B) lane 2]. Pre-incubation of TrfA with ClpX also consistently resulted in a substantial inhibition of the proteolysis by ClpXP analyzed during the second experimental step [Fig. 3(C)]. An approximately twofold higher degradation of TrfA, measured by densitometry scans, was observed in the control reaction when ClpX was omitted from the first step incubation mixture [Fig. 3(C)]. This result was repeatable with 10% variation. The observed inhibition of TrfA degradation was not due to loss of ClpX activity during first incubation because addition of fresh ClpX with ClpP did not result in increasing of TrfA degradation (data not shown). Light scattering analysis excluded the possibility that TrfA activation by ClpX results in substantial aggregation of TrfA (Fig. 10 supporting data). Thus we concluded that the ClpX-dependent activation of TrfA resulted in decreased TrfA proteolysis. The TrfA monomer degradation by ClpXP must have been much less effective comparing to the TrfA dimer degradation. Because during the proteolytic tests [Fig. 2(A)] the remaining protein contains a replicationally active TrfA monomer (Fig. 9 supporting data), this together explains the incomplete TrfA degradation.

Mutations within the TrfA dimerization interface altering dimer formation also affect the protein proteolysis by ClpXP

Our experiments indicated that the activation of TrfA leading to replicationally proficient TrfA monomers decreases TrfA proteolysis by ClpXP. We then asked if mutations altering TrfA dimer stability and therefore replication activity could affect TrfA proteolysis. The previously studied TrfA(S257F) and TrfA(G254D/S267L) mutant proteins,^16,32–34 were chosen for our experiments. In glycerol gradient centrifugation, a technique which allows the separation of the protein dimeric and monomeric forms, TrfA(S257F), was detected in fractions corresponding to the protein dimer, similarly as the wild-type TrfA [Fig. 4(A)]. The shapes of the curves obtained after densitometry analysis indicated that both proteins preparations contained homogenous populations with no substantial contamination with monomeric protein form [Fig. 4(A)]. Although both proteins are dimeric, we found that the S257F substitution results in the protein with a higher α-helical stability measured as chemical denaturation profile determined by the changes in ellipticity in the increasing concentration of guanidine HCl [Fig. 4(B)]. Also, FI* analysis demonstrated that in comparison to wild-type TrfA, TrfA(S257F) is less efficiently activated by ClpX [Fig. 4(C) compare lanes 4 and 8]. Consequently, this dimeric mutant protein is phenotypically defective in DNA binding and unable to support RK2 replication.³⁴ The phenotypic effects of the G254D/S267L substitutions are opposite. During glycerol gradient centrifugation, the TrfA(G254D/S267L) mutant is present largely in the form of a monomer [Fig. 4(A)]. The TrfA(G254D/S267L) profile obtained by densitometry scan of fractions after gradient centrifugation indicates no substantial contamination with the dimer protein form [Fig. 4(A)]. Moreover, in activity tests in the absence of ClpX in the reaction mixture, we observed a complete conversion of the plasmid template to the FI* form indicating that the protein is constitutively active and does not require ClpX for its activation [Fig. 4(C) lane 9]. Furthermore, the TrfA(G254D/S267L) mutant protein exhibits the lowest α-helical stability in comparison to the wild-type TrfA or the TrfA(S257F) mutant [Fig. 4(B)]. The analysis of isothermal circular dichroism spectra [Fig. 4(D)] did not indicate substantial differences of secondary structure content of the analyzed TrfA variants. The wild-type TrfA and the TrfA(G254D/S267L) mutant exhibited a 4% difference in β-sheet content, whereas the wild-type TrfA and the TrfA(S257F) secondary structure contents were very similar. The agreement of the CD spectra suggests that the wild-type and the TrfA mutant proteins maintain similar secondary structure, which strongly suggests that no major structural rearrangements occurred owing to mutations.

Figure 4.

Mutations altering the TrfA dimer formation affect replication activity and requirement for the ClpX-dependent activation. The His6-TrfA(S257F) and His6-TrfA(G254D/S267L) were analyzed along with His6-TrfA in glycerol gradients for dimer and monomer formation (A). Densitometry analysis of Western blots enabling an estimation of protein position in gradient is plotted in panel A. The chemical denaturation with guanidine HCl profiles of the His6-TrfA(S257F), the His6-TrfA(G254D/S267L), and the His6-TrfA were determined by the changes in ellipticity at 222 nm (B). Panel C shows comparison of the ClpX-dependent activation of the TrfA(S257F), the His6-TrfA(G254D/S267L), and the His6-TrfA. Panel D shows isothermal circular dichroism spectra of the analyzed TrfA variants in a range from 195 to 260 nm.

To further analyze the significance of the S257F, G254D, and S267L mutations, and in the absence of any experimental evidence as to the structure of TrfA, we performed the protein secondary and tertiary structure predictions. The well established method for tertiary structure prediction is by detection of homology to proteins with already known structure. Although this approach works best for proteins with high sequence identity, it also allows for construction of structural models based on structural templates with very little sequence similarity, provided that homology between the target and the template is established. In such cases, it is usually accomplished by fold-recognition methods and utilization of so-called metaservers that collect results from a number of these methods and use post-processing software to infer a consensus prediction. The validity of this approach has been demonstrated in a series of acclaimed benchmarking experiments called Critical Assessment of techniques for protein Structure Prediction (CASP).³⁵

Because significant sequence similarity between TrfA and other Rep proteins has not been identified,³⁶ to predict the structure of TrfA, we used the GeneSilico metaserver, [See Ref.37 and see Methods for details] which groups together most of the fold-recognition methods that have been very successful in CASP. According to structure prediction, the N-terminal region of TrfA (residues 1–160) is predicted to be disordered in solution (e.g., to lack a unique three-dimensional structure in the absence of other stabilizing factors). It has been shown that many natively disordered proteins actually retain significant fraction of secondary structure (often helical),^38–40 hence we might expect that the N-terminal part of TrfA actually does show some content of secondary structure. For the major part of the TrfA C-terminal region (residues 195–382), the fold-recognition methods implemented in the GeneSilico metaserver confidently predicted the presence of two copies of the WH domain. Most importantly, the closest homolog of TrfA among proteins with known structure appears to be the RepE initiator protein from plasmid F⁴¹ (1rep in the PDB; top match with high scores, e.g., 2.27 according to the PCONS consensus method, with scores >1 indicating statistical significance). The second best match (2nra in the PDB; PCONS score 1.81) is to the π initiator protein structure.⁴² The third best match (1hkq in the PDB; PCONS score 0.37) corresponds to the alignment between the central domain of TrfA and the N-terminal domain (NTD) of RepA from plasmid pPS10.⁴³ It must be noted that PCONS assigns scores based not on sequence similarity but on the detection of homology indicated by several individual fold recognition methods used by GeneSilico metaserver (see also Methods). Although the RepE and π exhibit very similar structure, RepA covers roughly their N-terminal half. The alignments returned by the fold-recognition methods indicate relatively low sequence identity between TrfA and the corresponding parts of the templates (top match 14% to RepE, 12% to π initiator). However, the secondary structure predicted for TrfA matches perfectly the secondary structure observed in the crystal structures of the templates. In addition, the pattern of residues predicted to be buried in the protein core or exposed to the solvent in TrfA matches very well e.g. the amphipathic character of helices of the WH domains in the templates. This, together with significant scores returned by fold-recognition methods, indicates high likelihood of remote homology between TrfA and the WH proteins, in particular with RepE predicted as the potentially closest homolog with known structure.

Based on the alignments returned by fold-recognition methods, using the structure of the best template (RepE), we constructed a comparative model of the structured region of TrfA, predicted to comprise a tandem repeat of WH domains (residues 160–382) [Fig. 5(A)]. Although the final determination of the model quality will be possible only by comparison to the structure solved by high-resolution experimental methods (e.g., X-ray crystallography), there are a number of bioinformatics methods called Model Quality Assessment Programs (MQAPs) that can predict the approximate deviation of the model from the real structure, without the knowledge of the real structure and instead relying on statistical evaluation of various structural parameters of the model (for review see Ref.44). Our model of TrfA has been evaluated as potentially ‘very good’ by the PROQ model quality assessment program (score 3.9, i.e., quite close to 4, which would indicate an ‘extremely good’ model) and is predicted to exhibit root mean square deviation of ∼3.5 Å to the native (currently unknown) structure, according to MetaMQAP (see Methods). It must be emphasized that these methods only predict the actual error in the modeled structure. Nonetheless, the values of their scores suggest that the model is likely to be approximately correct, hence supporting the prediction of a tandem WH fold in TrfA.

Figure 5.

TrfA shares homology with members of the Rep initiators family. A model of the TrfA dimer (aa 190-382; coordinates are available from ftp://genesilico.pl/iamb/models/), with individual protomers in blue and green, showing residues that interfere with dimerization (A). A combination of two substitutions G254D/S267L (orange) results in protein that is monomeric. The exchange S257F (shown in gray) produces a protein with a more stable dimer interface. Panel B shows TrfA monomer bound to DNA. Substitutions P314S and D198N result in TrfA mutants defective in DNA binding. The substitution E361K results in a mutant protein that binds DNA more efficiently). Table in C shows properties of TrfA mutants: G254D/S267L, S257F, P314S, D198N, E361K, and CΔ305. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

In the model the dimerization interface (residues 234–290 within the WH1 domain) consists of an extended anti-parallel β-sheet forming two β-hairpins [Fig. 5(A)]. We found that mutations G254D, S267L, and S257F affecting dimer stability are located within the predicted dimer interface [Fig. 5(A)].

The secondary structure content inferred using CD for 44 or 33 kDa TrfA variants could not be used to directly validate the predicted structure of TrfA, as the model corresponded to a different range of sequence. We have obtained genetic constructs to express TrfA mutants with N-terminal truncations corresponding to the modeled globular region, but the corresponding protein variants were impossible to purify in such concentration and buffer conditions that would be appropriate for CD analysis. Similar difficulties with purification of the TrfA N-terminal mutants were also described previously.⁶ Therefore, additionally to the model quality prediction in silico, to test the correctness of structure prediction we analyzed if the TrfA mutants with characterized phenotypes [Fig. 5(C)] fit the model of TrfA. As it was shown for RepA and RepE⁴⁵ we predict that only one, the N-terminal WH domain, is involved in dimerization, whereas both WH1 and WH2 domains of TrfA are involved in interactions with DNA. Consistent with this prediction, the deletion mutant TrfA(CΔ305) that lacks the predicted WH2 domain, forms a dimer structure but does not bind DNA. [See Ref.6, Fig. 5(C)] Substitutions P314S and D198N that result in TrfA mutants defective in DNA binding⁶ occur in residues that are close to the binding interface and may perturb protein-DNA interactions [Fig. 5(B,C)]. Finally, the substitution E361K generates a mutant protein that binds DNA more efficiently than the wild-type protein [See Ref.6, Fig. 5(C)]. According to our model, the positively charged Lys residue may form a contact with the DNA phosphate backbone, thereby stabilizing the protein-DNA complex [Fig. 5(B)].

The data presented above suggest that mutations S257F, G254D, and S267L are located in the predicted dimerization interface and consistent with their observed effect on TrfA dimer formation, replication activity and the ClpX-dependent activation. To answer the question if these mutations can also affect TrfA proteolysis by ClpXP, we conducted the analysis of TrfA stability. During in vitro tests, in control experiments after 2 h of incubation more than 70% of the wild-type TrfA was degraded by ClpXP [Fig. 6(A)]. In contrast, a significantly limited degradation or no degradation was detected when, respectively, TrfA(S257F) or the monomeric TrfA(G254D/S267L) mutant proteins were incubated with the ClpXP proteolytic complex [Fig. 6(B,C)]. These results were consistently repeatable. The inhibition of ClpXP-dependent degradation of TrfA(S257F) and TrfA(G254D/S267L) is not due to the failure of TrfA mutant's recognition by ClpX but rather due to failure of substrate processing. In ELISA experiments those TrfA mutant proteins interact with ClpX equally well as the wild-type TrfA (see Fig. 8).

Figure 8.

Binding of ClpX to TrfA conformational variants. Protein interactions were tested using ELISA reaction according to procedure described under “Materials and Methods.” Immobilized proteins: His6-TrfA (filled circles), His6-TrfA(G254D/S267L) (open rectangles), His6-TrfA(S257F) (filled rectangles), λO (opened circle), and BSA (filled triangles) were incubated with 62.5, 125, 250, 500, or 750 ng of ClpX in the presence of ATP. Plate wells contained 0.5 μg of the relevant protein.

Consistent with the in vitro observations, when we tested the TrfA mutants' stability in vivo, almost no degradation of TrfA(S257F) was detected after 60 min in the wild-type E. coli (see Fig. 7). Also, the monomeric TrfA(G254D/S267L) protein, with a 37-min half-life, was more stable than the wild-type TrfA (see Fig. 7). We conclude that the mutations in the TrfA dimerization interface affect not only protein dimer formation, replication activity, and activation by ClpX, but also TrfA proteolysis by ClpXP.

Figure 7.

TrfA mutant proteins with S257F and G254D/S267L substitutions exhibit increased intracellular stability in E. coli. The stability of His6-TrfA, His6-TrfA(S257F), and His6-TrfA(G254D/S267L) was tested in E. coli wild-type strain as described under “Materials and Methods.” The results of densitometry analysis of Western blots enabling an estimation of TrfA mutant proteins half-life times are displayed in the table.

Discussion

Our observations demonstrate that TrfA appears to be a natural substrate for the E. coli ClpXP proteolytic complex and that the oligomeric state of the TrfA protein influences its degradation and therefore intracellular stability. Moreover, we provide evidence that the TrfA protein of broad-host-range plasmid RK2 belongs to the family of replication initiation proteins consisting of WH domains.

Our in vivo and in vitro experiments clearly demonstrated that TrfA is a substrate for the ClpXP-dependent proteolysis, however, we must admit that the TrfA proteolysis was not as efficient as demonstrated for other ClpXP substrates for example lambda O protein.²⁹ Substrate recognition and proteolysis by the ClpXP system relies on the presence of a specific sequence (tag).^46,47 Based on the proteomic analysis described by Flynn et al.,²⁵ using the ScanProSite Web tool (http://www.expasy.org/tools/scanprosite/) we identified 13 putative ClpX recognition sequences in the proximity of both N- and C-terminal ends and in the interior of the TrfA protein primary sequence (Table 1 supporting data). The analysis of TrfA truncated variants encompassing these 13 putative recognition sequences (MP and IK results unpublished) did not reveal a single sequence required for TrfA proteolysis suggesting the possibility that multiple signals within TrfA sequence are required.

The prediction of WH domains in the TrfA structure indicates that this protein is homologous to RepE, Pi, and RepA replication initiators of plasmids F, R6K, and pPS10, respectively. Formerly, the lack of evident sequence similarity between TrfA and other Rep proteins resulted in the exclusion of TrfA from comparative analyses of replication initiation proteins consisting of WH domains.^1,36 However, the lack of high sequence identity is not an evidence for lack of homology, as indicated by numerous homologous proteins with very similar structures, yet little sequence similarity (many such cases for WH domains are documented e.g., in the SCOP database http://scop.mrc-lmb.cam.ac.uk/scop/data/scop.b.b.i.e.html). Recently, and consistent with the model presented here, the presence of a WH motif within the TrfA protein has been suggested.⁴⁸ The TrfA model presented here represents only the 190–382aa residues, as attempts to predict the structure of the N-terminal region revealed that it exhibits high propensity to be intrinsically unstructured. The model of the globular part of TrfA predicts a tandem repeat of WH domains beginning approximately at amino acid 190. The appearance of a TrfA* degradation product comprising the C-terminal TrfA fragment starting at position 185 (see Fig. 2) indicates that ClpXP most probably is hindered by structural distortions caused by the WH domains. Thus, multiple ATPase cycles and/or multiple attempts may be required for complete TrfA unfolding and proteolysis. It has been proposed that the stability of the three-dimensional structure of the substrate affects its processing by chaperone/proteolytic systems.^23,49 The release of stable substrates can occur even after translocation and degradation of a substrate have commenced.⁵⁰ Because under different reaction conditions the TrfA* was always observed, most likely during the TrfA processing by ClpXP, it is released as a stable protein domain comprising WH motifs.

Two TrfA mutants, TrfA(S257F) and TrfA(G254D/S267L), tested in the course of this work appeared resistant to ClpXP degradation in vitro and, in comparison with the wild-type protein, are much more stable in E. coli cells. In contrast, truncated variants of TrfA protein are prone to ClpXP degradation (MP and IK unpublished results). Intriguingly, both TrfA(S257F) and TrfA(G254D/S267L) have substitutions located within the predicted dimer interface [Fig. 5(A)]. ELISA tests showed that the lack of degradation of those mutants is not due to the failure of recognition by ClpX. ClpX activates TrfA(S257F), however, this activation is much less efficient than that observed for the wild-type TrfA [Fig. 4(C)]. Also, previously it has been shown that TrfA(S257F) could be activated to only a limited extent by ClpX or guanidine HCl treatment, as determined by in vitro DNA binding.³⁴ The TrfA structure prediction presented here places S257 at the end of the β2–β3 hairpin structure involved in protein dimerization [Fig. 5(A)]. The end of a hairpin loop is not directly involved in the interaction between monomers, however the S257F substitution generates a hydrophobic cluster which, together with V274 and L275, could stabilize TrfA dimer formation. The CD analysis of TrfA GuHCl denaturation profiles demonstrates that TrfA(S257F) has a higher α-helical stability than the wild-type TrfA. It is very likely that tight dimer interface of TrfA(S257F) can stabilize the protein structure in such a way that it can be bound by ClpX, but cannot be efficiently unfolded and degraded by ClpXP. In consequence, the protein remains in dimeric, replicationally inactive form.

The other TrfA mutant, which appeared resistant to ClpXP degradation, is TrfA(G254D/S267L). The two substitutions result in a monomeric, hyperactive protein that shows runaway replication and a plasmid copy-up phenotype.^{8,9,16,33,34,51,52} The TrfA model presented here predicts that the 254 and 267 residues are very close to each other and are located in the middle parts of the opposite β strands of the β2–β3 hairpin structure in the dimer interface. The G254D and S267L substitutions most probably cause a conformational change leading to the perturbation of the dimer interface.

Unlike wild-type TrfA, the TrfA(G254D/S267L) monomeric mutant does not exhibit requirement for chaperone-dependent activation [Fig. 4(C)]. This indicates that this mutant protein is not only truly monomeric, but that it is also a fully active form of the plasmid replication initiator. Similar Rep protein variants have been characterized for plasmids P1, F, R6K, and pPS10.^36,41,53,54 Structural analysis of the plasmid F RepE protein and the plasmid pPS10 RepA protein, to which our model for TrfA predicts homology, demonstrated that the dissociation of Rep into monomers, as a result of specific mutations or chaperone activity, alters the Rep monomer conformation by introducing changes within the WH domains.^43,54,55 In this work, we observed that the G254D/S267L substitutions in WH domains prevent TrfA dimer formation and result with lower α-helical stability in comparison to the dimeric wild-type TrfA or the TrfA(S257F) mutant. These substitutions also result in a substantial inhibition of TrfA proteolysis by ClpXP [Fig. 6(C)]. Moreover, when wild-type TrfA was activated by ClpX leading to formation of the replication proficient monomeric form, similar inhibition of TrfA proteolysis by ClpXP was observed (see Fig. 3). It is likely that steric distortions, as proposed for other Rep proteins, or/and distance changes between recognition motifs generated upon monomerization, could make the TrfA monomers unable to be processed by ClpXP. The analysis of isothermal CD spectra [Fig. 4(D)] revealed minor differences in secondary structure content of the wild-type TrfA and the monomeric TrfA(G254D/S267L) mutant. It has been recently demonstrated that induced high local flexibility activates a dormant ClpX recognition motif of bacteriophage Mu repressor.⁵⁶

The involvement of molecular chaperones and proteolytic systems in Rep processing has been studied in vitro for many replicons (for review see Ref.24). Because of the redundant, overlapping activity of cellular chaperone/protease systems those in vitro results do not necessarily translate into phenotypic effects in a relevant genetic background. We observed that TrfA dimers and monomers differ in susceptibility to ClpXP-dependent proteolysis. This result suggests a novel view of the stability of Rep proteins in bacterial cells. Further experiments with different chaperones and proteases as well as other replication initiators are needed to explore this issue.

Materials and Methods

Bacterial strains, proteins, plasmids, and reagents

The E. coli strains used in this study were: SG20080 (clpX::kan), SG22098 (clpP::cm), SG22097 (clpX::kan, clpP::cm) derivatives of SG20250.⁵⁷ Purification of E. coli proteins ClpX, DnaC, DNA gyrase and histidine-tagged DnaB and DnaA was performed following established protocols described previously.^16,51,58 E. coli HU was purified using an over-expression vector and a purification protocol kindly provided by Dr. Roger McMacken. The wild-type TrfA was purified as previously described.^5,6 Published protocols have been used for the purification of the previously described N-terminal hexahistidine-tagged TrfA variants including His6-TrfA, His6-TrfA(G254D/S267L) and His6-TrfA(S257F).^16,32,34,52 The E. coli ClpP protein was kindly provided by Dr. Zylicz from IIMCB Warsaw. Commercially available proteins were SSB from Promega, creatine kinase and bovine serum albumin (BSA) (Fraction V) from Sigma. Plasmid pKD19L1 containing the minimal RK2 origin region has been previously described.⁵⁹ Plasmid pAT30 derivatives are vectors for inducible expression of the TrfA proteins.³² Immunodetection assays were conducted with the use of anti-TrfA or anti-DnaB rabbit sera, goat anti-rabbit HRP-IgG from BioRad, SuperSignal West Pico and chemiluminescence substrate kit from PIERCE.

Proteolytic assay

The reaction mixture (25 μL) containing ClpX (1.5 μg), ClpP (1.5 μg) and TrfA (1.5 μg) in 20 mM Tris-HCL pH 7.6, 40 mM Hepes-KOH pH 7.6, 4% (w/v) sucrose, 4 mM dithiothreitol, 11 mM magnesium acetate, 80 μg/mL bovine serum albumin (BSA) and 4 mM ATP was incubated for 2 h at 32°C. After incubation the reactions were stopped by the addition of loading buffer (0.1M Tris-HCl pH 6.8, glycerol 5%, sodium dodecyl sulphate 2%, 0.1% bromophenol blue, 0.75M β-Mercaptoethanol) followed by 5 min. incubation at 95°C. One third of the reaction volume was loaded on a 12.5% SDS-polyacrylamide gel and Coomassie Brilliant Blue staining or immunodetection was applied after electrophoresis. The densitometry analysis was performed using Gel-Doc 2000 Imaging System (BioRad) and Scion Image software (Scion Corporation).

ClpX-dependent activation of TrfA

TrfA (500 ng) was incubated with ClpX (2 μg) for 15 min at 32°C in the same buffer as used for proteolysis. After incubation, the activation assay was coupled with proteolytic assay and/or TrfA activity test. When activation was coupled with proteolytic assay reaction buffer containing ClpP (2 μg) was added giving a total reaction volume of 25 μL. After the subsequent incubation period samples were analyzed as described for the proteolytic assay. When the TrfA activation reaction was coupled with the TrfA activity test the activation step was performed as described above. After incubation the portion of the activation reaction mixture was diluted in the buffer containing components for the protein activity test.

In the experiments that compared the activation of His6-TrfA, His6-TrfA(S257F) and His6-TrfA(G254D/S267L), 500 ng of proteins was used along with the increasing amounts of ClpX (1, 1.5, and 2 μg). After activation reaction in vitro TrfA activity test was performed as described below.

In vitro TrfA activity test

The helicase unwinding assay (FI*) was used for the analysis of the TrfA conformational activation. This reaction requires the active monomeric form of the TrfA protein for establishing the helicase complex on the plasmid RK2 replication origin. The assay was performed essentially as previously described.¹⁶ Reactions (25 μL) contained 40 mM HEPES-KOH, pH 7.6, 25 mM Tris-HCl, pH 7.6, 4% (w/v) sucrose, 4 mM dithiothreitol, 80 μg/mL BSA, 11 mM magnesium acetate, 2 mM ATP, 500 mM (each) CTP, GTP, UTP, 20 mM creatine phosphate, 20 μg/mL creatine kinase, 160 ng of SSB, 160 ng of DNA gyrase, 125 ng of DnaB, 625 ng of DnaA, 24 ng of DnaC, 300 ng of pKD19L1 template, 500 ng of His6-TrfA(G254D/S267L) or a portion of activation mixture (see above) containing 500 ng of His6-TrfA or His6-TrfA(S257F). After incubation, the reactions were analyzed on agarose gel.

Glycerol gradient centrifugation

To analyze the monomeric and dimeric forms of TrfA protein, glycerol gradient centrifugation was performed. Analysed proteins (3 μg) were diluted in 100 μL of gradient buffer without glycerol. Samples were then applied to a 3.2-mL linear 15–35% (v/v) glycerol gradient in buffer (20 mM Tris, pH 7.4; 50 mM KCl; 5 mM MgCl₂; 1 mM DTT; 0.1% Nonidet P-40; and 0.5M NaCl) and centrifuged at 40,000 rpm for 24 h at 2°C in a Beckman SW60 rotor. Fractions were collected from the top of the tube and analyzed using SDS/PAGE followed by Western blotting with anti-TrfA serum. DnaK (70 kDa) and carbonic anhydrase (30 kDa) were used as molecular weight markers.

In vivo stability assay

Bacterial cells used for TrfA stability analysis were transformed with pAT30 derivatives enabling inducible expression of His6-TrfA, His6-TrfA(G254D/S267L) and His6-TrfA(S257F) proteins. Overnight cultures were diluted 1:25 in fresh LB broth, and allowed to grow to OD₆₀₀ = 0.5. The expression of TrfA was induced by the addition of IPTG to a final concentration of 1 mM. After 30 min., protein translation was stopped by the addition of tetracycline to a final concentration of 20 μg/mL. Samples were removed at specific time points and analyzed using SDS-PAGE followed by Western blotting.

Enzyme-linked immunosorbent assay

A modified ELISA assay with polyclonal ClpX antibodies was used for detection of TrfA-ClpX interactions. A conjugate of horseradish peroxidase coupled with goat anti-rabbit IgG (Bio-Rad) and colorimetric detection was performed with TMB Peroxidase EIA Substrate Kit.

Circular dichroism

Measurements were performed on a Jasco J-810 Circular Dichroism spectrometer equipped with PFD 350S automatic Peltier accessory. Spectra were collected at 25°C in a range from 195 to 260 nm with 2-nm spectral bandwidth. Scan speed was 50 nm/min with a 2s response time. Each spectrum was averaged from five independent measurements. The protein concentration was 0.3 mg/mL in 25 mM KPi pH = 7.5, 100 mM NaCl, in a quartz cuvette with 1-mm path length. Spectra were measured in millidegrees, corrected for buffer effects, and converted to mean residue molar ellipticity [Θ] MRW. Calculation of secondary structure content was performed with the use of CDPro software package.⁶⁰ Guanidine hydrochloride-induced TrfA denaturation was monitored by changes of ellipticity at 222 nm in the increasing denaturant concentration. Proteins were diluted to a concentration of 0.1 mg/mL in a phosphate buffer containing appropriate GuHCl concentration. Spectra were collected as described above and corrected by subtraction of the spectra measured from a blank containing the same concentration of GuHCl.

Bioinformatic analyses of TrfA

For prediction of the TrfA structure, we used comparative modeling by fold-recognition which is an established bioinformatic approach, allowing for detection of very remote relationships between proteins, regardless of pairwise sequence similarity. We used GeneSilico metaserver³⁷ which uses a number of different fold-recognition methods including PDBBLAST, HHSEARCH, mGenTHREADER, and it contains PCONS⁶¹ as the consensus-inferring program for calculation of a consensus prediction. This approach is currently the recommended procedure, established in the course of the acclaimed CASP benchmarking experinment. GeneSilico metaserver, uses a variety of methods, not only for protein fold-recognition, but also for secondary structure prediction, order/disorder, solvent accessibility of individual amino acid residues etc. (links to individual methods are provided at the website https://genesilico.pl/meta2/). In particular, for prediction of regions of potential intrinsic disorder we used PONDR⁶² and DISOPRED⁶³, and for the tertiary structure prediction we used the protein fold-recognition approach, which allows for detection of remote relationships between proteins, regardless of pairwise sequence similarity. The consensus fold-recognition prediction (i.e., calculation of alignments between the query sequences to proteins of known structure) was performed using PCONS, program that uses results from fold recognition methods applied by GeneSilico metaserver. The structural model of TrfA was constructed basing on the alignments returned by the fold-recognition methods and refined using the ‘FRankenstein's monster’ approach (a variation of comparative modeling) successfully applied in the CASP6 competition.⁶⁴ Finally, the quality of the model (estimated deviation from the true structure, without the knowledge of the true structure) was predicted using PROQ⁴⁴ and MetaMQAP.⁶⁵