Biochemical Journal

Research article

Varying dependency of periplasmic peptidylprolyl cistrans isomerases in promoting Yersinia pseudotuberculosis stress tolerance and pathogenicity

Ikenna R. Obi, Roland Nordfelth, Matthew S. Francis

Abstract

Periplasmic PPIases (peptidylprolyl cistrans isomerases) catalyse the cistrans isomerization of peptidyl-prolyl bonds, which is a rate-limiting step during protein folding. We demonstrate that the surA, ppiA, ppiD, fkpA and fklB alleles each encode a periplasmic PPIase in the bacterial pathogen Yersinia pseudotuberculosis. Of these, four were purified to homogeneity. Purified SurA, FkpA and FklB, but not PpiD, displayed detectable PPIase activity in vitro. Significantly, only Y. pseudotuberculosis lacking surA caused drastic alterations to the outer membrane protein profile and FA (fatty acid) composition. They also exhibited aberrant cellular morphology, leaking LPS (lipopolysaccharide) into the extracellular environment. The SurA PPIase is therefore most critical for maintaining Y. pseudotuberculosis envelope integrity during routine culturing. On the other hand, bacteria lacking either surA or all of the genes ppiA, ppiD, fkpA and fklB were sensitive to hydrogen peroxide and were attenuated in mice infections. Thus Y. pseudotuberculosis exhibits both SurA-dependent and -independent requirements for periplasmic PPIase activity to ensure in vivo survival and a full virulence effect in a mammalian host.

  • chaperone
  • immunosuppressant
  • infection
  • membrane biogenesis
  • periplasmic peptidylprolyl cistrans isomerase (periplasmic PPIase)
  • protein folding
  • survival

INTRODUCTION

PPIases (peptidylprolyl cistrans isomerases) catalyse the cistrans isomerization of peptidyl-prolyl bonds in peptide and protein substrates [1,2]. Isomerization of proline residues is a rate-limiting step in protein folding, underlying the ubiquitous presence of PPIases in all kingdoms of life and in different cellular compartments [2]. Three major classes of PPIases exist: the cyclophilins, which bind the immunosuppressive drug CsA (cyclosporin A); the FKBPs (FK506-binding proteins), which bind the FK506 immunosuppressant; and the PPIases that have sequence similarity with parvulin from the Escherichia coli cytoplasm [1,2].

The bacterial periplasm contains all three PPIase classes. In E. coli, four definitive periplasmic PPIases have been described: SurA, PpiA, PpiD and FkpA [3,4]. Another newly identified PPIase is FklB, which might constitute the newest member of the periplasmic PPIase family [5]. SurA and PpiD belong to the parvulin class of PPIases; PpiA is a cyclophilin, whereas FkpA and FklB are related to the FKBP family. SurA is well studied; most bacteria lacking surA are more susceptible to detergents, antibiotics and hydrophobic dyes suggestive of a compromised outer membrane barrier function (reviewed in [6]). On the other hand, in vitro phenotypic defects in bacteria lacking the other periplasmic PPIases are seldom reported.

Yersinia pseudotuberculosis is an environmental bacterium associated with sporadic enteric disease outbreaks known as yersiniosis caused by the ingestion of contaminated food or fluids [7]. Although this disease causes gastrointestinal discomfort, it is usually self-limiting and rarely associated with systemic disease. It is also a close relative of another enteric pathogen Y. enterocolitica [7] as well as the infamous plague-causing pathogen Y. pestis [8].

In the present study, in silico analysis revealed all five of these putative periplasmic PPIases in the Y. pseudotuberculosis YPIII genome sequence (GenBank® accession number CP000950) based on their amino acid sequence identity with their homologues in E. coli K12 [9]. We purified SurA, PpiD, FkpA and FklB and performed a subsequent biochemical analysis of their PPIase activity. Furthermore, we created cumulative in cis deletions of the five periplasmic PPIase genes in Y. pseudotuberculosis and examined their contribution to envelope integrity and survival during exposure to extracytoplasmic stress, ROIs (reactive oxygen intermediates), low pH and in vivo mouse infections.

EXPERIMENTAL

Bacterial strains and plasmids

The bacterial strains and plasmids used in the present study are shown in Table 1.

View this table:
Table 1 Bacterial strains and plasmids used in this study

Growth conditions

Unless stated otherwise, bacteria were cultivated routinely in LB (Luria–Bertani) agar or broth at either 26°C (Y. pseudotuberculosis) or 37°C (E. coli) with aeration. When required, antibiotics were used at the following final concentration: 50 μg/ml kanamycin, 25 μg/ml chloramphenicol and 100 μg/ml carbenicillin.

To assay for the viability of PPIase mutants in the presence of bile salts or detergents, overnight cultures of Y. pseudotuberculosis grown in LB broth at 26°C were sub-cultured (0.1 volume) into 3 ml of fresh medium. The bacteria were allowed to grow for a further 2 h before being serially diluted and then plated on LB agar, Yersinia-selective agar rich in bile salts (Becton Dickinson) or LB agar supplemented with 0.0125% SDS or 0.25% sodium deoxycholate. CFUs (colony forming units) were counted after 48 h of growth.

An antibiotic-sensitivity assay involved mixing 100 μl from overnight cultures of Y. pseudotuberculosis with 10 ml of molten 0.6% soft-agar that was then poured over LB agar plates. Vancomycin- and bacitracin-impregnated discs (Oxoid) were then overlaid on the agar surface. After 24 h incubation at 26°C, the diameter of the inhibitory zone around the discs was measured.

To assay for stationary phase survival, Y. pseudotuberculosis overnight cultures were sub-cultured (0.1 volume) into fresh LB broth and grown at 26°C to mid-exponential phase. These bacteria were again back-diluted (0.001 volume) into 50 ml of LB broth. During 6 days of incubation, samples were taken at daily intervals, serially diluted and then plated on LB agar. CFUs were determined after 48 h.

To test sensitivity to menadione and hydrogen peroxide, Y. pseudotuberculosis overnight cultures were sub-cultured (0.005 volume) into fresh LB broth and grown at 26°C for 1 h. Samples were serially diluted and then plated on LB agar supplemented with 200 μM menadione or 1 mM hydrogen peroxide. CFUs were counted after 48 h of growth.

For the pH testing, overnight cultures of Yersinia were sub-cultured in LB broth (0.1 volume) buffered with 100 mM citrate pH 3 or pH 5. Bacteria were grown at 26°C for 6 h. Samples were taken at 0, 3 and 6 h. After serial dilution, samples were plated on LB agar. CFUs were counted after 48 h. LB broth buffered with 20 mM potassium phosphate (pH 7) was used as a control.

Generation of mutants

Overlap PCR with template DNA from Y. pseudotuberculosis YPIII/pIB102 was used to generate mutated alleles for constructing the full-length in-frame deletion mutants. The primer pairs used to generate each mutation are listed in Supplementary Table S1 (at http://www.BiochemJ.org/bj/439/bj4390321add.htm). Amplified fragments were cloned into the suicide mutagenesis vector pDM4. E. coli S17-1λpir harbouring each specific mutagenesis vector was used in conjugal mating experiments with Y. pseudotuberculosis. Selection for the appropriate allelic exchange events giving rise to the desired Y. pseudotuberculosis mutants used our conventional sucrose-sensitivity methodology [10].

Overexpression and purification of recombinant PPIases

PPIase genes were PCR amplified from the Y. pseudotuberculosis YPIII/pIB102 genome using the primer pairs listed in Supplementary Table S1. Codons 2–105 of ppiD were omitted to avoid sequences encoding for a putative transmembrane region. The DNA fragments were cloned in pET30a(+) generating PPIase gene sequences in frame with the 5′ His tag resident in the plasmid. Recombinant PPIases were over-expressed in E. coli BL21 cells grown to exponential phase (D600 of 0.5) in 250 ml of LB broth. Upon addition of 0.4 mM IPTG (isopropyl β-D-thiogalactoside) bacteria were cultured for a further 3 h. Bacteria were washed with 50 ml of lysis buffer [20 mM Tris/HCl (pH 7.5), 500 mM NaCl and 30 mM Imidazole] and concentrated 10-fold in this buffer containing lysozyme (1 mg/ml). Afterwards, bacteria were lysed by sonication and soluble lysates were clarified by centrifugation (32000 g for 1 h at 4°C) and passaged through a 0.2 μm filter prior to protein purification by affinity chromatography on Ni2+–agarose columns using the AKTA purifier (GE Healthcare). Columns were washed with 50 ml of lysis buffer and the proteins were eluted with 20 mM Tris/HCl (pH 7.5), 500 mM NaCl and 1 M imidazole using a linear gradient. Pooled eluted fractions were dialysed overnight against 20 mM Tris/HCl (pH 8) and 50 mM NaCl. Protein purity was assessed by SDS/PAGE (12% gel) and Coomassie Blue staining. Protein concentration was determined using a BCA (bicinchoninic acid) protein assay kit (Pierce).

PPIase assay

The PPIase activity was performed in 50 mM Tris/HCl (pH 8) at 10°C using the protease-coupled assay [11]. The peptide substrate Suc-Ala-Leu-Pro-Phe-pNa (pNa is p-nitroanilide) was purchased from Bachem. Nitroaniline appearance due to α-chymotrypsin cleavage of the peptide trans-isomer was followed at 390 nm for 3 min with a Beckman Coulter DU 730 Life Science UV–visible spectrophotometer. First-order rate constants (kobs) were derived by fitting reaction progress curves to a first-order rate equation [y390=y0+a(1−ekt), where k is the rate constant, y0 is the absorbance at zero time]. kcat/Km was calculated from the plot of kobs against PPIase concentration.

The PPIase inhibition assay used stock solutions of each immunosuppressant CsA, FK506 or rampamycin (LC laboratories) prepared in ethanol/water (1:1, v/v). PPIases (0.1 μM) were incubated with various concentrations of the drug on ice for 7 min prior to PPIase activity measurements. The respective rate constants in the presence of the inhibitors were derived by fitting reaction progress curves to the first-order rate equation.

Whole bacterial lysates and cellular fractionation

To prepare whole bacterial cell lysates, overnight cultures of Y. pseudotuberculosis were grown at 26°C and 37°C. After standardization of the attenuance at 600 nm and centrifugation (5000 g for 10 min at 22°C), cells were lysed in SDS sample buffer [50 mM Tris/HCl (pH 6.8), 2% SDS, 10% glycerol, 5% 2-mercaptoethanol and 0.1% Bromophenol Blue) and heat-inactivated.

For fractionation experiments, bacteria grown to exponential phase in 200 ml of LB broth at 26°C were resuspended in 30 mM Tris/HCl (pH 8) and 20% sucrose at 80 ml per g of wet weight. After addition of 500 mM EDTA dropwise to an end concentration of 1 mM and a 10 min incubation on ice with gentle agitation, cells were harvested (8000 g for 20 min at 4°C) and the pellet was resuspended in 500 μl of ice-cold 5 mM MgSO4. Following agitation for 1 h at 4°C, the periplasmic fraction (supernatant) was collected (8000 g for 20 min at 4°C) and subsequently mixed with SDS sample buffer. The pellet was resuspended in 500 μl of 100 mM Tris/HCl (pH 8) and 10 mM EDTA, and the osmotic-shocked cells were lysed by three freeze–thaw cycles and a 1 min sonication. The cytoplasmic fraction (supernatant) was recovered (22000 g for 30 min at 22°C) and mixed with SDS sample buffer. The remaining pellet (crude membrane fraction) was washed with 500 μl of 20 mM potassium phosphate buffer (pH 7) and recovered by microcentrifugation (22000 g). To separate inner membrane from outer membrane, the pellet was resuspended in 100 μl of 0.5% sarkosyl in 20 mM potassium phosphate buffer (pH 7). After incubation at room temperature (22°C) for 30 min, the inner membrane (soluble fraction) was collected by microcentrifugation (22000 g) and treated with SDS sample buffer. The outer membrane (pellet) was washed once with 100 μl of 0.5% sarkosyl in 20 mM potassium phosphate buffer (pH 7) and resuspended in SDS sample buffer. Following separation by SDS/PAGE (12% gel), proteins were transferred on to a PVDF membrane (Millipore) and identified with rabbit polyclonal antibodies raised against individually purified recombinant PPIase proteins (Agrisera AB) as approved by the Animal Ethics Committee of Umeå University (ethics number A102/09).

Mass spectrometry and protein identification

Sample preparation, MS and protein identification was based on procedures described previously [12]. A protein ID was considered valid if it was top-most on the putative hit list generated by the MASCOT search engine (http://www.matrixscience.com) and with a probability based Mowse score above the significance threshold (P<0.05; see below).

Isolation of LPS (lipopolysaccharide)

To isolate LPS liberated from Y. pseudotuberculosis grown overnight in 200 ml of LB broth at 26°C, we adapted a method for outer membrane vesicle purification [13] by including a terminal proteinase K digestion step. Briefly, to pellets resuspended in 20 mM Tris/HCl (pH 8), SDS was added to a final concentration of 0.1% and samples were heated for 3 min at 95°C to denature the proteins. Final concentrations of 5 mM CaCl2 and 7 μg of proteinase K were added and the samples were incubated overnight at 37°C. Following treatment with sample buffer, isolated LPS were analysed by SDS/PAGE (13.5% gel) and silver staining (Fermentas Life Sciences).

Crude total cellular LPS was prepared from stationary-phase Y. pseudotuberculosis grown in LB broth at 26°C. Bacterial pellets were collected by brief centrifugation, resuspended in sample buffer and lysed at 95°C for 10 min. Final concentrations of 5 mM CaCl2 and 7 μg of proteinase K were added and the samples were incubated overnight at 37°C. LPS analysis was performed by SDS/PAGE (13.5% gel) and silver staining.

Isolation and characterization of cellular FAs (fatty acids) and phospholipids

The analysis of FA and phospholipid composition of Y. pseudotuberculosis YPIII was performed commercially within the quality assured laboratories at Mylnefield Lipid Analysis using essentially the method of Folch et al. [14].

TEM (transmission electron microscopy)

Y. pseudotuberculosis was harvested from LB plates after 48 h. Bacteria were resuspended in 10 mM Tris/HCl buffer (pH 7.4) and 10 mM MgCl, and allowed to adhere to formvar-coated grids for 3 min at room temperature. Following negative staining with 1% sodium silicotungstate, images of bacteria were acquired using a Jeol JEM 1230 electron microscope at ×40 000 magnification.

Mouse infections

Infections were performed in accordance with the guidelines of the Animal Ethics Committee of Umeå University (ethics number A81-08). Groups of five female BALB/c mice (Scanbur) of similar age (8–9 weeks) and weight (~18 g), were injected intraperitoneally with 0.1 ml of different dilutions of Y. pseudotuberculosis parent and mutant bacteria that had been grown overnight in LB broth at 26°C. Pelleted bacteria were serially diluted to 107, 106 and 105 CFUs/ml in PBS. Infection was monitored for 14 days post-injection. The ID50 (50% infectious dose) was determined using the Reed–Muench method [15], with positivity of infection reflected by symptoms of ruffled fur, weight loss, listlessness and diarrhoea. Infected mice showing symptoms of a terminal infection were immediately killed.

RESULTS

Identification of Y. pseudotuberculosis periplasmic PPIases

Amino acid sequence alignments with their homologues in E. coli K12 revealed five potential periplasmic PPIases in the Y. pseudotuberculosis YPIII genome: surA (NCBI locus tag YPK_3571; UniProt identifier B1JKY1), ppiD (NCBI locus tag YPK_3230; UniProt identifier B1JHR7), ppiA (NCBI locus tag YPK_0242; UniProt identifier B1JIS0), fkpA (NCBI locus tag YPK_0270; UniProt identifier B1JIU8) and fklB (NCBI locus tag YPK_3778; UniProt identifier B1JML8) (Figure 1A). Their amino acid sequences were at least 99% identical with their homologues in Y. pestis, between 85 and 98% in Y. enterocolitica and 58 and 79% in E. coli (Figure 1B). Based upon BLAST motif searches (Figure 1C), SurA (434 amino acids) is composed of four domains: a unique N-terminal domain, two parvulin domains and a C-terminal domain possessing an unknown fold. PpiD (628 amino acids) contains three domains: a transmembrane domain, an N-terminal domain that is similar to that of SurA and a parvulin domain. In addition, the protein contains a long C-terminal region of unknown fold. PpiA (189 amino acids) possesses a cyclophilin domain, whereas FkpA (266 amino acids) and FklB (206 amino acids) consist of two domains: an N- and a C-terminal domain both displaying (FKBP) consensus.

Figure 1 Periplasmic PPIases of Y. pseudotuberculosis YPIII

(A) Schematic representation of the genetic organization of gene loci of Y. pseudotuberculosis periplasmic PPIases. surA (NCBI locus annotation YPK_3571), ppiA (NCBI locus annotation YPK_0242) and fkpA (NCBI locus annotation YPK_0270) are all coded on the sense strand at genome positions 3935985–3937289, 284007–284576 and 309956–310756, whereas ppiD (NCBI locus annotation YPK_3230) and fklB (NCBI locus annotation YPK_3778) are coded on the complementary strand at genome positions 3531083–3532969 and 4168658—4169278 respectively. Symbols denote genes encoding helix–hairpin–helix repeat-containing competence protein ComEA (§), transcriptional regulator subunit β (*), glycoside hydrolase family protein (**), putative NPT hydrolase (***), SlyX (+), YheO domain-containing protein (++), iron–sulfur cluster repair di-iron protein (#) and opacity-associated protein A (##). Predicted gene names are given in parentheses or denoted by a symbol to reflect function. (B) Amino acid sequence identity of periplasmic PPIases from Y. enterocolitica 8081, Y. pestis CO92, E. coli K12 MG1655 and Y. pseudotuberculosis YPIII. Amino acid sequences were aligned using the BLAST algorithm. NCBI annotation is provided in square brackets and the numbers in round brackets represents the length of each protein in amino acids. (C) Domain organization of periplasmic PPIases from Y. pseudotuberculosis. The predicted molecular mass for each protein is provided in parentheses.

Production and purification of Y. pseudotuberculosis periplasmic PPIases

The surA, ppiA, ppiD, fkpA and fklB genes were PCR-amplified from the Y. pseudotuberculosis YPIII/pIB102 genome and cloned into the pET-30a(+) vector. IPTG induction of protein expression in E. coli BL21 cells resulted in soluble forms of recombinant SurA, PpiD, FkpA and FklB, whereas PpiA was produced as an insoluble protein (results not shown). Varying the IPTG concentration or induction temperature did not improve the solubility of recombinant PpiA (results not shown). Soluble forms of SurA, PpiD, FkpA and FklB were purified from E. coli cell lysates using nickel-affinity chromatography and protein purity was analysed by SDS/PAGE (Figure 2). This one-step purification process provided very pure protein that migrated to approximately 52 kDa, 74 kDa, 33 kDa and 27 kDa by SDS/PAGE, corresponding to the apparent molecular masses of recombinant SurA, PpiD, FkpA and FklB respectively. Protein yield of approximately 3 mg was obtained from a 250 ml culture for each protein.

Figure 2 Purification of Y. pseudotuberculosis periplasmic PPIases

The purified recombinant proteins over-produced in E. coli BL21 were boiled for 5 min in SDS sample buffer and separated by SDS/PAGE (12% gels). Proteins were stained with Coomassie Blue. Approximate molecular masses of the individual protein bands in the PageRuler Prestained Protein ladder (Fermentas Life Sciences) are provided to the left of the panel in kDa. Adjacent lanes are purified recombinant SurA, PpiD, FkpA and FklB respectively.

SurA, FkpA and FklB exhibit detectable PPIase activity

Having purified recombinant SurA, PpiD, FkpA and FklB, we performed a rarely executed comparative assessment of their individual PPIase activity using a α-chymotrypsin-coupled spectrophotometric assay that monitors the cistrans isomerization of the peptidyl-prolyl bond in the tetrapeptide Suc-Ala-Leu-Pro-Phe-pNa. This assay is based on the fact that α-chymotrypsin can hydrolyse the 4-nitroanilide amide bond in the tetrapeptide only when the leucine–proline bond is in the trans conformation. SurA, FkpA and FklB displayed PPIase activity (Figure 3A). No detectable activity was observed with PpiD, consistent with an earlier study concerning E. coli PpiD [16] or for the two-component response regulator CpxR, which was used as a negative control (Figure 3A). SurA PPIase activity was low in comparison with the robust activity of FkpA and FklB. The lower intrinsic SurA PPIase activity has also been observed for E. coli SurA [17,18]. Furthermore, the PPIase activity of SurA, FkpA and FklB was concentration-dependent (Figure 3B). Under our assay conditions, the catalytic efficiencies (kcat/Km) of FklB (1.1×106 M−1 · s−1) was highest, followed by that of FkpA (2.6×105 M−1 · s−1), whereas SurA (4.7×104 M−1 · s−1) has the least specificity for the tetrapeptide substrate. FklB and FkpA catalytic efficiencies are comparable with other biochemically verified FKBPs [19,20].

Figure 3 PPIase activity of recombinant SurA, PpiD, FkpA and FklB

(A) Progress curves for the cistrans isomerization of the chromogenic peptide Suc-Ala-Leu-Pro-Phe-pNa. α-Chymotrypsin-mediated hydrolysis of Suc-Ala-Leu-Pro-Phe-pNa was measured in the absence or presence of 0.1 μM recombinant SurA, PpiD, FkpA, FklB or CpxR (protein negative control) at 10°C for 3 min. (B) PPIase activity of SurA, FkpA and FklB is concentration-dependent. Various concentrations of the recombinant proteins were assayed for PPIase activity and the rate constants were calculated by fitting the absorbance data to a first-order exponential rise function. Results are means±S.D. of three independent measurements and were highly reproducible.

To confirm the presence of an intrinsic PPIase activity in the Y. pseudotuberculosis periplasm, we isolated periplasmic extracts from the parental bacteria and a quintuple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) lacking the five periplasmic PPIase genes. These extracts were assayed for PPIase activity using the tetrapeptide Suc-Ala-Leu-Pro-Phe-pNa. Periplasmic extract from the parental bacteria was active in cistrans isomerization of the peptidy-prolyl bond of the tetrapeptide, whereas no PPIase activity was detected in the periplasmic extract of the quintuple mutant (Supplementary Figure S1 at http://www.BiochemJ.org/bj/439/bj4390321add.htm). We therefore considered that a lack of PPIase activity in the periplasmic fraction of the ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA quintuple mutant was indicative of an important in vivo role for these periplasmic PPIases in maintaining optimal Y. pseudotuberculosis fitness.

PPIase inhibition by immunosuppressive drugs

PPIases were first discovered because of their high affinity to bind immunosuppressive drugs. To investigate the effects of CsA, FK506 and rampamycin immunosuppressants on SurA, FkpA and FklB, 0.1 μM of protein was incubated with various concentrations of the drugs and then assayed for remaining PPIase activity. CsA, FK506 and rampamycin showed no detectable effect on SurA PPIase activity, even at excessive concentrations (10 μM for CsA and 0.5 μM for FK506 and rampamycin) (Figure 4A and results not shown). This is consistent with SurA being classified as a parvulin-like PPIase. However, FkpA (Figure 4B) and FklB (Figure 4C) were inhibited strongly by FK506 and rampamycin in a dose-dependent manner. Rampamycin was more effective at causing complete inhibition of FkpA and FklB activity with an IC50 of 19 nM and 51 nM respectively; this compared with an IC50 of 43 nM and 119 nM needed for FK506 to inhibit FkpA and FklB activity respectively. These results reflect the higher affinity of rampamycin to FKBPs [21]. As anticipated, CsA failed to diminish the PPIase activity of FkpA or FklB (Figures 4B and 4C).

Figure 4 Inhibition of PPIase activity with immunosuppressive drugs

Inhibition assay was measured with 0.1 μM SurA (A), FkpA (B) or FklB (C) in 50mM Tris/HCl (pH 8) at 10°C using Suc-Ala-Leu-Pro-Phe-pNa as substrate in the presence of different concentrations of the immunosuppressants CsA, FK506 and rampamycin prepared in ethanol/water (1:1, v/v). Remaining activity was calculated using the rate constants in the presence of the inhibitors. Note that the final concentration of ethanol in the assay mixture was 1%, but this had negligible effect on PPIase activity as determined by control assays in the presence of ethanol alone (results not shown). Results are means±S.D. of three independent measurements and were highly reproducible.

Subcellular localization of SurA, PpiD, FkpA and FklB in Y. pseudotuberculosis

Protein prediction algorithms indicate the presence of signal peptides in SurA and FkpA as well as a transmembrane domain in PpiD (results not shown). This implies that these proteins are likely to be localized in the bacterial envelope. However, no related information is available for FklB. We therefore utilized each recombinant purified PPIase as an antigen to generate specific polyclonal rabbit antibodies. These were then used to experimentally analyse the subcellular localization of SurA, PpiD, FkpA and FklB in the same Y. pseudotuberculosis whole-cell lysate that was separated into outer membrane, periplasmic, inner membrane and cytoplasmic fractions. To control fractionation purity, TolC, MBP (maltose-binding protein), FtsH and H-NS (histone-like nucleoid structuring protein) were used as outer membrane, periplasmic, inner membrane and cytoplasmic markers respectively. All of the PPIases could be detected in the whole-cell lysate (Figure 5). SurA and FkpA were enriched in the periplasmic fraction, whereas PpiD was enriched in the inner membrane fraction. These results are consistent with their respective localizations in E. coli [22,23]. However, contrary to a previous study that identified FklB in the periplasm using indirect detection with the Mip antibody [5], our anti-FklB antibodies identified FklB exclusively in the inner membrane.

Figure 5 Subcellular localization of SurA, PpiD, FkpA and FklB in Y. pseudotuberculosis

An overnight culture of Y. pseudotuberculosis was sub-cultured in fresh LB broth. After 2 h, the bacteria were harvested and whole-cell lysate was separated into outer membrane, periplasmic, inner membrane and cytoplasmic fractions. TolC, MBP, FtsH and H-NS were used as outer membrane, periplasmic, inner membrane and cytoplasmic markers respectively to control the purity of the fractionation. Each cellular fraction was then probed with anti-SurA, anti-PpiD, anti-FkpA and anti-FklB antibodies.

Pal (peptidoglycan-associated lipoprotein) and OmpA are among several outer membrane proteins reduced in surA mutants

In E. coli, SurA functions as a periplasmic folding factor of various outer membrane proteins (reviewed in [6]). To study effects of loss of surA on outer membrane proteins in Y. pseudotuberculosis, outer membrane protein profiles of parental, surA single (ΔsurA), quadruple (ΔppiA ΔppiD ΔfkpA ΔfklB) and quintuple (ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) deletion mutants were analysed by SDS/PAGE and Coomassie Blue staining. Indeed, the surA single and quintuple mutants showed decreased amounts of proteins in their outer membrane in comparison with the parental bacteria (Supplementary Figure S2A at http://www.BiochemJ.org/bj/439/bj4390321add.htm). This defect could be complemented by transforming these surA mutants with a low-copy-number plasmid expressing a wild-type copy of surA from its native promoter. Strikingly, the SurA+ quadruple mutant lacking the other four periplasmic PPIase genes exhibited the same outer membrane protein levels as the parental bacteria (Supplementary Figure S2A). To identify some of the reduced outer membrane proteins caused by surA deletion, protein bands of interest from parental bacteria were excised from the Coomassie Blue-stained gel. Preliminary MALDI–TOF (matrix-assisted laser-desorption ionization–time-of-flight) MS analysis of in-gel-digested tryptic fragments identified a 18 kDa protein as Pal (Supplementary Figure S2B) and a 36 kDa protein as OmpA (Supplementary Figure S2C).

Y. pseudotuberculosis lacking surA are leaky

An alteration in outer membrane protein profiles suggests a compromised outer membrane. To identify whether this induced morphological changes in the affected bacteria, we used TEM to visualize bacterial morphology. Parental bacteria and the quadruple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB) possessed a uniform peripheral exterior suggestive of an intact bacterial envelope (Figures 6A and 6D). In contrast, the surA mutants (ΔsurA or ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) displayed an irregular surface and leaked material into the surrounding milieu (Figures 6B and 6E). These morphological effects were due to the loss of functional surA because a normal morphology could be restored by trans-complementation with a wild-type copy of surA (Figures 6C and 6F).

Figure 6 TEM of PPIase mutants of Y. pseudotuberculosis

Cultures were grown on LB agar at 26°C prior to the TEM analysis of the parental strain (A), surA single mutant (ΔsurA) (B), surA single mutant (ΔsurA) trans-complemented with surA (C), quadruple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB) (D), quintuple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) (E) and quintuple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) trans-complemented with surA (F). Arrows indicate the abundance of material released into the extracellular milieu by Yersinia mutants lacking surA.

Despite many bacteria being known to release outer membrane blebs or vesicles, this released material is apparently not vesicular in nature because the structures lack any obvious membranous form (results not shown). To identify the nature of this liberated cellular material, cell-free bacterial supernatants were subjected to ultracentrifugation and the pelleted material collected. No obvious differences in protein content in these samples were observed (results not shown). However, following proteinase K digestion to remove protein material, SDS/PAGE and silver staining analysis revealed an enrichment of released LPS from the quintuple ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA mutant, and to a lesser extent the single ΔsurA mutant (Supplementary Figure S3A at http://www.BiochemJ.org/bj/439/bj4390321add.htm). This was not observed from all other strains examined, including the trans-complemented surA mutants (Supplementary Figure S3A). Moreover, LPS discharge from the surA mutants seems to reflect a defect in LPS assembly control because total cellular LPS levels isolated from whole bacteria were indistinguishable irrespective of the strain background (Supplementary Figure S3B). It is currently unknown whether this putative LPS assembly defect is a direct or indirect consequence of SurA depletion.

Deletion of surA confers altered FA and phospholipid composition

Given that Y. pseudotuberculosis lacking surA have reduced outer membrane protein content and possesses defects in LPS assembly, we undertook a comparative survey of the membrane FA content of parental and mutant Y. pseudotuberculosis grown at ambient (26°C) and elevated temperature (37°C). The raw results can be seen in Supplementary Table S2 (at http://www.BiochemJ.org/bj/439/bj4390321add.htm). At 26°C, UFAs (unsaturated fatty acids) were reproducibly produced in abundance by all bacteria (62–67.8% of total FA), followed by SFAs (saturated fatty acids) (28.8–32.6%) and with minor amounts of CFAs (cyclic fatty acids) (3.3–5.5%) (Figure 7A, upper panel). At 37°C, production of UFAs was dramatically curtailed to between 36.7 and 48% of total FAs. Interestingly, both surA mutants reproducibly produced more UFAs regardless of growth temperature. Consistent with earlier reports [24,25], CFA production increased dramatically at 37°C (between 12.8 and 23%; Figure 7A, lower panel). However, this effect was noticeably diminished in the surA mutants. The ratio of CFA produced by Y. pseudotuberculosis at 37°C compared with 26°C was therefore higher for SurA+ bacteria than for SurA bacteria (Figure 7B). On the other hand, the SFA ratios and the UFA ratios remained steady in all bacteria independent of growth temperature (Figure 7B).

Figure 7 Comparative analysis of the FA levels in isogenic strains of Y. pseudotuberculosis lacking periplasmic PPIases

Relative FA levels were calculated from the mg of free FA/100 g of total lipid samples (see Supplementary Table S2 at http://www.BiochemJ.org/bj/439/bj4390321add.htm) extracted from exponentially grown parent, surA single mutant (ΔsurA), quadruple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB) and quintuple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA) bacteria. The raw results are represented as the percentages of SFA, UFA and CFA from bacteria grown at 26°C and 37°C (A) and the ratios of SFA, UFA and CFA content at 37°C compared with 26°C (B). Results are means±S.D. of three independent measurements and were highly reproducible.

A partial analysis of individual phospholipid content within whole bacteria was also determined (Supplementary Table S3 at http://www.BiochemJ.org/bj/439/bj4390321add.htm). At 26°C, PE (phosphatidylethanolamine) appeared to dominate total phospholipid content (accounting for 79.2–81.5%), with PG (phosphatidylglycerol; 8.7–11.9%) and CL [cardiolipin (diphosphatidylglycerol) 7.6–10.4%] far less prolific (Supplementary Figure S4A at http://www.BiochemJ.org/bj/439/bj4390321add.htm). Elevated temperature did not have a dramatic impact on the relative levels of PE or PG being produced by any of the bacteria (Supplementary Figure S4A). On the other hand, bacteria lacking surA or SurA+ bacteria deficient in all four other periplasmic PPIases increased the proportion of CL and other minor phospholipids at elevated temperature (Supplementary Figure S4B). As anticipated from the total cellular FA profiles (see Supplementary Table S2 and Figure 7), the make-up of individual phospholipids at ambient temperature was dominated by UFAs, but the relative levels of UFAs and SFAs equilibrated in bacteria grown at 37°C (Supplementary Figure S4C). In addition, the surA mutants routinely produced phospholipids composed of modestly higher levels of UFAs irrespective of growth temperature (Supplementary Figure S4C). Thus subtle alterations in FA and lipid content occur in Y. pseudotuberculosis lacking surA, but generally not when lacking all four other periplasmic PPIases.

Yersinia lacking surA are susceptible to diverse antimicrobials

Given the obvious perturbations in the bacterial envelope of Yersinia surA mutants, we investigated whether the envelope barrier function was compromised in these strains. All single mutants, the quadruple mutant and the quintuple mutant were subjected to growth in the presence of antibiotics (vancomycin and bacitracin), bile salts [a major component contained within YSA (Yersinia selective agar)] and detergents (SDS and sodium deoxycholate). Quite expectedly, the two Y. pseudotuberculosis mutants lacking surA were more susceptible to these antimicrobials compared with any other strain (Supplementary Figures S5A and S5B at http://www.BiochemJ.org/bj/439/bj4390321add.htm). Envelope barrier function was restored to surA mutants by ectopic expression of a wild-type copy of surA. Not surprisingly, the surA mutants were modestly less able to maintain long-term viability during extended culturing (Supplementary Figure S5C). Remarkably, none of these conditions impeded growth of the SurA+ quadruple mutant. Overall, this reflects a strict requirement of SurA for maintenance of envelope integrity in pathogenic Y. pseudotuberculosis.

PPIases other than SurA are important for Yersinia survival during exposure to ROIs

A confounding aspect of the present study has been the lack of discernable phenotypes for the quadruple ΔppiA ΔppiD ΔfkpA ΔfklB mutant that still encodes an intact surA allele. Since it is unlikely that Y. pseudotuberculosis would retain proteins in the absence of any important function, we assessed the sensitivity of our mutants to hydrogen peroxide, the superoxide-generating compound menadione and acid stress. These are relevant stresses that must be counteracted by Y. pseudotuberculosis in order to survive during an intracellular life cycle inside an infected host [26]. Decisively, exposure to hydrogen peroxide resulted in significant (Supplementary Figure S6 at http://www.BiochemJ.org/bj/439/bj4390321add.htm, *P<0.05) growth defects of the quadruple mutant, although to a lesser degree than Y. pseudotuberculosis lacking surA. This result agrees with the recent finding that a Par10-negative E. coli strain was also more sensitive towards hydrogen peroxide than was the wild-type strain [27]. Significantly, the presence of hydrogen peroxide is the first in vitro growth condition in which the quadruple mutant displayed a reduced tolerance to stress. Interestingly however, this mutant did grow to the same level as parental bacteria in the presence of 200 μM menadione, whereas surA mutants still displayed a significant growth defect (Supplementary Figure S6). These differential sensitivities of bacteria to hydrogen peroxide and menadione is not uncommon [28], possibly because the oxidative stress pathways for the superoxide and hydrogen peroxide sensing are independently regulated by SoxRS (superoxide stress regulator) and OxyR respectively [29].

We also tested the sensitivity of these bacteria to increasingly acidic conditions. Despite pH 5.0 being bacteriostatic and pH 3.0 being bacteriocidal for Y. pseudotuberculosis, we could not detect any differences between parental and mutant bacteria under either growth condition (results not shown). Thus, unlike their protective effect against the action of ROIs, no periplasmic PPIase appeared to confer resistance on Y. pseudotuberculosis during growth in low pH.

PPIases and Y. pseudotuberculosis virulence

The quadruple mutant (ΔppiA ΔppiD ΔfkpA ΔfklB) exhibited an in vitro growth defect in the presence of hydrogen peroxide, an antimicrobial compound commonly associated with activation of host immune cells during microbial infection. In the light of this, we were keen to address whether this mutant was attenuated during in vivo infections of female BALB/c mice. We scored attenuation on the basis of ID50 as determined by the extent of visible symptoms of infection, such as ruffled fur, weight loss, listlessness and diarrhoea. Mice infected with high doses (106 and 105) of parental Y. pseudotuberculosis were terminally infected within 4 and 8 days respectively (Figure 8A). With these trends, an ID50 of 2.0×104 could be calculated for the parental strain. On the other hand, mice infected with the single surA mutant, the quadruple PPIase mutant or the quintuple PPIase mutant were mostly symptom free (Figure 8A). This suggested that ID50 measurements for these mutant bacteria were in excess of 3.0×106, 3.2×106 and 2.8×106 respectively; a 100-fold attenuation compared with that of the parent. Interestingly, we did observe a small degree of weight loss (Figure 8B) and ruffled fur (results not shown) in those mice infected with higher doses of the quadruple PPIase mutant. This was not evident in infections with surA mutant bacteria. We therefore conclude from these in vivo mouse infection studies that full Y. pseudotuberculosis pathogenicity does depend on a functioning collection of several periplasmic PPIases, and not solely on active SurA. Hence, one or more of the other periplasmic PPIases possess an important role in Y. pseudotuberculosis survival in a particular niche inside the infected host. We therefore interpret the two-log reduction in ID50 observed for SurA+ and SurA bacteria to reflect SurA-dependent and -independent modes of virulence attenuation.

Figure 8 PPIases are required for virulence in mice

Dilutions of Y. pseudotuberculosis parent and mutant bacteria were prepared from washed pellets that were serially diluted to 107, 106 and 105 CFUs/ml in PBS. Mice were then injected intraperitoneally with 0.1 ml to give final inocula of approximately 106, 105 and 104 bacteria per mouse (five mice for each strain and dilution). Mice were observed for 14 days post-injection for survival (A) and alterations in their body weight (B). Results are means±S.D. for five independent measurements. In (A), only mice infected with dilutions of parental bacteria show any visible signs of infection. Hence much of the data reflect healthy symptom-free mice that are concealed by the bold continuous grey line stylized with a filled square symbol.

DISCUSSION

In comparison with SurA [6], it has proven much more difficult to establish the physiological roles of the other periplasmic PPIases. FkpA is a PPIase with chaperone activity [23,30] that might co-operate with SurA in the starvation stress response [31]. PpiD was originally described as a multicopy suppressor of surA [22] with previous studies also providing evidence of a periplasmic chaperone role [32,33]. On the other hand, FklB and PpiA functions still remain obscure. We used two types of functional assay to study a quadruple ΔppiA ΔppiD ΔfkpA ΔfklB deletion mutant still containing functional SurA, a single ΔsurA mutant, and a quintuple ΔppiA ΔppiD ΔfkpA ΔfklB ΔsurA deletion mutant lacking all five extracytoplasmic PPIases; those that broadly examined bacterial envelope integrity and those that scrutinize survival in the host. Only in the absence of SurA could we detect obvious defects in Yersinia envelope integrity. Yet, like the surA mutants, SurA+ bacteria lacking all other periplasmic PPIases were sensitive to conditions that mimicked the vacuolar environment inside an activated host immune cell (such as elevated hydrogen peroxide). Indeed, PPIases could have an important role during oxidative stress [27]. To this end, all three mutant bacteria were also highly attenuated in a mouse infection model. These results imply that SurA cannot compensate for the collective loss of PpiA, PpiD, FkpA and FklB when Y. pseudotuberculosis enters a mammalian host, but it can during routine in vitro culturing. SurA is therefore generally required for bacterial survival under both in vitro and in vivo conditions, whereas all other periplasmic PPIases are required for growth in a particular ecological niche, most notably during mammalian infections. The mechanistic basis for these attenuating effects remain obscure; to define the in vivo roles for the PpiA, PpiD, FkpA and FklB PPIases during Y. pseudotuberculosis pathogenesis is a goal for the future. We anticipate that their roles in promoting pathogenicity are probably indirect. Perhaps in the absence of PpiA, PpiD, FkpA or FklB, one or more of their native substrates that are also important for Yersinia survival in vivo fail to assemble in the outer membrane.

Having purified four of the five PPIases from Y. pseudotuberculosis, specific antibodies were generated to enable a definitive investigation into the subcellular localization of SurA, PpiD, FkpA and FklB. Although the localizations of SurA and FkpA in the periplasmic fraction and PpiD in the inner membrane are consistent with observations from E. coli [22,23], we identified FklB to be predominately associated with the inner membrane and not the periplasm as previously reported for E. coli [5]. These two assays were performed differently, which may account for this discrepancy. On the other hand, FklB was previously seen to interact with inner membrane-associated thioredoxin [34]. Perhaps this is a transient interaction that could occasionally cause FklB to locate to the inner membrane. Further work is needed to resolve this inconsistency. Nonetheless, the extracytoplasmic compartmentalization of these PPIases in Yersinia ideally positions them for interactions with early protein folding intermediates in the periplasm.

SurA is the subject of intensive research (reviewed in [6]). It is critical for periplasmic quality control, particularly when the bacteria are faced with optimizing growth and survival in adverse environmental conditions. Although long regarded as a PPIase enzyme, evidence now supports a primary role of SurA as a chaperone assisting in the transport and assembly of a subset of outer membrane proteins [35,36]. In fact, the requirement for SurA PPIase-independent chaperone activity supersedes the need for its PPIase activity [35]. This could explain why we observed such low in vitro PPIase activity from SurA derived from Y. pseudotuberculosis, since the two parvulin domains and the associated PPIase activity may only play a secondary role in SurA function. However, a role for the SurA parvulin domains in protein folding in vivo during growth of the organism in some defined ecological niche should not be discounted. We intend to generate site-directed surA mutants to pinpoint whether the physiological defects in our existing full-length surA mutants are due to loss of SurA PPIase activity or chaperone activity.

Target substrates of SurA are thought to possess a W-X-W motif (where W is an aromatic amino acid) that is a characteristic of β-barrel proteins [3739]. In our preliminary search for potential SurA targets in Yersinia, we identified a significant reduction of OmpA and Pal in outer membrane fractions. OmpA, containing a singular W-X-W motif, constitutes a major component of the bacterial outer membrane and therefore possesses important pleiotropic functions [40]. OmpA levels are noticeably lessened in E. coli mutants lacking surA [17,41] and SurA can interact with OmpA peptides [38]. However, it is unclear whether this interaction is a requirement for OmpA assembly in the outer membrane because ompA transcription is also diminished in surA mutants [42] and another periplasmic chaperone, Skp, is already implicated in OmpA assembly [43]. The outer membrane Pal lipoprotein also performs essential functions in the bacterial envelope [44]. As an integral component of the Tol-Pal system, it bridges the inner and outer membranes with the intervening peptidoglycan layer. Formation of these molecular bridges is dependent on the proton-motive force and is necessary for cell division, solute transport across the inner membrane and the preservation of outer membrane integrity. Pal transport to the outer membrane is dependent on the Lol pathway [45]. Since Pal is not considered a true substrate of SurA [42], loss of surA may decrease Lol-dependent transport efficiency, perhaps as a result of alterations in membrane lipid content (discussed below) [46]. To define the true molecular targets of SurA in Y. pseudotuberculosis might well identify novel virulence factors essential for Yersinia survival during host infections.

To optimize growth in varied environments, bacteria actively regulate FA composition to modulate lipid content and distribution in order to maintain membrane structure and function [4749]. Clinically important bacteria, such as the Yersiniae, also alter lipid content as a means to evade the hostile host innate immune response [50,51]. Significantly, alterations in FA and lipid content specifically occurred in Y. pseudotuberculosis lacking surA. Presumably, this limits the adaptiveness of surA mutants to extracytoplasmic stress and the ability to instigate outer membrane repair, especially at elevated temperatures; a poorer survival outcome would result. SurA influence on FA and lipid content could be direct or indirect, affecting de novo fatty acid or lipid biosynthesis, transport and assembly and/or modification of existing FAs or lipid structures. A simple explanation is that surA depletion indirectly stimulates an increased production of key FAs or lipids to compensate for the compromised envelope. This precedent exists given that chronic extracytoplasmic stress responsiveness affects LPS biogenesis [52] and deletion of surA constitutes a strong extracytoplasmic stress stimulant in E. coli [17]. It is also in balance with our observations that Yersinia surA mutants release significant amounts of LPS into the surrounding milieu. In view of this, it would be interesting to examine σE or Cpx pathway activity in Yersinia lacking surA. Knowing this may benefit our future understanding of the molecular connection between SurA and lipid biosynthesis/assembly in the Yersinia envelope.

AUTHOR CONTRIBUTION

Ikenna Obi, Roland Nordfelth and Matthew Francis designed, performed and analysed the experiments. Ikenna Obi and Matthew Francis wrote the paper. All authors read and approved the final paper.

FUNDING

Performed within the virtual framework of the Umeå Center for Microbial Research Linnaeus Program (UCMR-LP), this work was supported by the Swedish Research Council [grant numbers 2006-3869 and 2009-3660], the Carl Tryggers Foundation for Scientific Research [grant number 06:141] and the Foundation for Medical Research at Umeå University. I.R.O. is supported by a UCMR-LP postdoctoral fellowship.

Abbreviations: CsA, cyclosporin A; CFU, colony-forming unit; CL, cardiolipin; FA, fatty acid; CFA, cyclic FA; FKBP, FK506-binding protein; H-NS, histone-like nucleoid structuring protein; ID50, 50% infectious dose; IPTG, isopropyl β-D-thiogalactoside; LB, Luria–Bertani; LPS, lipopolysaccharide; MALDI–TOF, matrix-assisted laser-desorption ionization–time-of-flight; MBP, maltose-binding protein; Pal, peptidoglycan-associated lipoprotein; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; pNa, p-nitroanilide; PPIase, peptidylprolyl cis–trans isomerase; ROI, reactive oxygen intermediate; SFA, saturated FA; TEM, transmission electron microscopy; UFA, unsaturated FA

References

View Abstract