Biochemical Journal

Research article

Activation of human pro-urokinase by unrelated proteases secreted by Pseudomonas aeruginosa

Nathalie Beaufort, Paulina Seweryn, Sophie de Bentzmann, Aihua Tang, Josef Kellermann, Nicolai Grebenchtchikov, Manfred Schmitt, Christian P. Sommerhoff, Dominique Pidard, Viktor Magdolen

Abstract

Pathogenic bacteria, including Pseudomonas aeruginosa, interact with and engage the host plasminogen (Plg) activation system, which encompasses the urokinase (uPA)-type Plg activator, and is involved in extracellular proteolysis, including matrilysis and fibrinolysis. We hypothesized that secreted bacterial proteases might contribute to the activation of this major extracellular proteolytic system, thereby participating in bacterial dissemination. We report that LasB, a thermolysin-like metalloprotease secreted by Ps. aeruginosa, converts the human uPA zymogen into its active form (kcat=4.9 s−1, Km=8.9 μM). Accordingly, whereas the extracellular secretome from the LasB-expressing pseudomonal strain PAO1 efficiently activates pro-uPA, the secretome from the isogenic LasB-deficient strain PDO240 is markedly less potent in pro-uPA activation. Still, both secretomes induce some metalloprotease-independent activation of the human zymogen. The latter involves a serine protease, which we identified via both recombinant protein expression in Escherichia coli and purification from pseudomonal cultures as protease IV (PIV; kcat=0.73 s−1, Km=6.2 μM). In contrast, neither secretomes nor the pure proteases activate Plg. Along with this, LasB converts Plg into mini-Plg and angiostatin, whereas, as reported previously, it processes the uPA receptor, inactivates the plasminogen activator inhibitor 1, and activates pro-matrix metalloproteinase 2. PIV does not target these factors at all. To conclude, LasB and PIV, although belonging to different protease families and displaying quite different substrate specificities, both activate the urokinase-type precursor of the Plg activation cascade. Direct pro-uPA activation, as also reported for other bacterial proteases, might be a frequent phenomenon that contributes to bacterial virulence.

  • bacterial invasion
  • human urokinase
  • LasB
  • plasminogen activation system
  • protease IV
  • Pseudomonas aeruginosa

INTRODUCTION

Proteases are present in all living organisms. In mammals, these enzymes are involved in a multitude of (patho)physiological processes, both intracellularly, such as protein quality control or apoptotic cell death, and extracellularly, such as complement system activation, coagulation, fibrinolysis and inflammation, as well as normal cell migration or malignant tumour cell invasion [1]. Micro-organisms, including pathogenic bacteria, also produce proteases, which primarily provide free amino acids and/or participate in housekeeping functions [24]. Interestingly, some of these proteases also regulate the transition from an adhesive to an invasive bacterial phenotype, dampen the host innate and acquired defences, and/or damage host tissues [36]. Accordingly, a number of microbial proteases have been reported and, in some cases, demonstrated as significant virulence factors [24,6].

One of the host multicomponent systems that is affected by bacterial proteases is the Plg (plasminogen) activation system [79]. This major fibrinolytic/matrilytic and inflammatory system is composed of the serine protease Pm (plasmin), its precursor Plg, the urokinase-type (uPA) and tissue-type (tPA) proteolytic activators of Plg, as well as the serpin-type protease inhibitors α2-antiplasmin and PAI-1 (plasminogen activator inhibitor type-1) [1012]. Numerous pathogenic bacteria actually interact with this host system through (i) binding of Plg via surface-exposed lysine-rich residues, and/or (ii) production of Plg-activating factors, either displaying proteolytic activity (e.g. the outer membrane aspartic protease Pla from Yersinia pestis), or functioning as non-proteolytic activators (e.g. the well-characterized staphylo- or strepto-kinase), and, furthermore, (iii) expression of proteases which degrade host serpins [2,79]. Altogether, these host–pathogen interactions induce dysregulated engagement of the Plg activation system and are thus thought to promote host tissue destruction and bacterial dissemination and invasion, which led to the concept of ‘bacterial metastasis’, by analogy to the uPA/Pm-driven progression of invasive cancer cells [2,79]. In line with this, we demonstrated recently that aureolysin, a thermolysin-like metalloprotease secreted by the Gram-positive bacterium Staphylococcus aureus, does not only disrupt the host protease inhibitor network by degrading α2-antiplasmin and PAI-1, but also activates the host protease zymogen pro-uPA [13].

Considering that other pathogenic prokaryotes might hijack this major mammalian proteolytic system via direct pro-uPA activation as well, we investigated the capacity of proteases secreted by Pseudomonas aeruginosa to do so. Ps. aeruginosa is a widespread opportunistic human pathogen, which can provoke local (e.g. from the skin and wounds, as well as respiratory and urinary tracts) to systemic, benign to life-threatening infections [14]. In addition, this Gram-negative bacterium, which produces biofilms and displays a high capacity to develop resistance to antibiotics, is a frequent cause of nosocomial infections [14]. Among the secreted pseudomonal factors, there is an array of proteases, including the metalloproteases LasA, LasB (pseudolysin) and AprA (alkaline protease), as well as the serine protease PIV (protease IV)/PrpL [4,15,16]. In addition to these proteases, a high- and a low-molecular-mass serine protease [LepA, ~100 kDa, and PASP (Ps. aeruginosa small protease), ~18.5 kDa] have been identified in the pseudomonal secretome [1719]. LasB, which is one of the major secreted pseudomonal proteins, is expressed to high levels both in vitro and in vivo, and is an acknowledged virulence factor in lung and skin infection models [16]. This 33 kDa protein belongs to the thermolysin/M4 family of metalloproteases and, accordingly, it shares sequence and structure homologies with staphylococcal aureolysin [20,21]. LasB thus appears to be a good candidate to display a direct pro-uPA-activating capacity.

Using purified or recombinant bacterial proteases, as well as pseudomonal extracellular secretomes, we demonstrate that LasB and PIV, although being unrelated proteases, both target and activate human pro-uPA, while leaving Plg proteolytically inactive. This pathway is likely to contribute to bacterial invasion [22]. Moreover, whereas LasB processes a number of components of the Plg system including Plg itself, the uPAR (uPA receptor), its inhibitor PAI-1, as well as one major Pm target, the pro-MMP (matrix metalloproteinase)-2, PIV does not target any of these proteins.

EXPERIMENTAL

Purified and recombinant proteins

Purified LasB (0.5 mg/ml, ~15 μM) isolated from a pathogenic strain of Ps. aeruginosa was from Elastin Products Company with a specific activity of 260 units/mg of protein. Native PIV (0.085 mg/ml, ~3.2 μM) was purified from the extracellular medium of pseudomonal cultures, as reported by Engel and co-workers [23]. Recombinant human pro-uPA (1 mg/ml, ~20 μM) expressed in Escherichia coli, was a gift from A.G. Saunders (Grünenthal, Aachen, Germany). Human Glu-Plg (2 mg/ml, ~20 μM) and Pm (1 mg/ml, ~10 μM, 4.6 units/mg), purified from plasma, were from Sigma–Aldrich. Human active uPA (1 mg/ml, ~20 μM, 100000 IU/mg), purified from urine, was from ProSpec-Tany TechnoGene. Human recombinant uPAR (0.5 mg/ml, ~10 μM), encompassing residues 1–281 fused to a C-terminal His6 tag, was from R&D Systems. Recombinant human PAI-1 (10 μg/ml, ~0.2 μM) was from Hyphen Biomed.

Antibodies

A chicken polyclonal Ab (antibody) directed to human uPA and strongly reacting with the A-chain of the protease has been produced in-house [24], a mouse monoclonal Ab directed to uPAR domain 2 (#3932, reacting with amino acids 125–132 [25]) was from American Diagnostica, a mouse monoclonal Ab directed to PAI-1 (C-9, reacting with amino acids 24–158) was from Santa Cruz Biotechnology, and a mouse monoclonal Ab directed to polyhistidine was from Qiagen (Penta-His). HRP (horseradish peroxidase)-conjugated Abs against chicken or mouse Ig were from Sigma–Aldrich and from Jackson ImmunoResearch respectively.

Bacterial strains and preparation of bacterial extracellular secretomes

The Ps. aeruginosa LasB-producing strain PAO1 [26] and the strain PAO1lasB9 (referred to as PDO240), in which the lasB gene sequence has been deleted and replaced with a spectinomycin-resistance cassette [27], were used in the present study. Bacteria were grown overnight at 37 °C with shaking in Luria–Bertani broth liquid medium, until cells had reached the stationary growth phase. Bacterium-free culture supernatants were then obtained by a first centrifugation at 6000 g, followed by a second one at 12000 g, both for 10 min. Supernatants were filtered further through 0.22 μm membranes to provide the conditioned growth medium, which contains the array of bacterial factors secreted under these experimental conditions, including proteases, and are thereafter referred to as the bacterial secretomes PAO1-Sec and PDO240-Sec. These milieus were immediately frozen in aliquots of 500 μl and stored at −80 °C.

Cloning, expression and purification of recombinant PIV

The coding sequence of mature PIV (amino acids 212–462; Swiss-Prot accession number Q9HWK6) was amplified by nested PCR, and 5′ and 3′ extensions harbouring appropriate restriction sites (BamHI, HindIII) for cloning were added by additional PCRs. The PCR fragment was then inserted into the bacterial expression vector pQE-30 (Qiagen). The resulting fusion gene encodes mature PIV preceded by an N-terminal extension of 17 amino acids including the His6 tag.

Overnight cultures of transformed XL1 Blue cells (Stratagene) were diluted and grown at 37 °C with vigorous shaking in 2× TY medium (16 g/l tryptone, pH 7, 10 g/l yeast extract and 5 g/l NaCl) to an attenuance (D600) of 0.7, then induced with 1.5 mM IPTG (isopropyl β-D-thiogalactoside) and cultured for an additional 3 h. After centrifugation at 10000 g for 5 min, the bacterial pellet was frozen at −20 °C. Cells were lysed in 6 M guanidinium chloride, 10 mM Tris/HCl, 100 mM NaH2PO4 and 8 mM 2-mercaptoethanol (pH 8.0) on a rotating wheel for 2 h at 4 °C. The lysate was then centrifuged at 10000 g for 10 min, the supernatant was collected, and the pH was adjusted to 8.0.

Thereafter His6-tagged PIV was purified by affinity chromatography using an Ni-NTA (Ni2+-nitrilotriacetate) Superflow resin (Qiagen) under denaturing conditions in two steps. First, proteins were washed and eluted by a pH gradient: starting off with 6 M guanidinium chloride, 10 mM Tris/HCl, 100 mM NaH2PO4 and 8 mM 2-mercaptoethanol (pH 8.0), the column was washed further with 8 M urea, 10 mM Tris/HCl, 100 mM NaH2PO4 and 8 mM 2-mercaptoethanol (pH 8.0, 7.0, 6.0 and 5.3 respectively). Most of the bound recombinant PIV was finally eluted with the same buffer at pH 4.0. In the second purification step, an imidazole concentration gradient was used. For binding of PIV, the pH of the eluate of the first column was adjusted to 8.0 and the column was loaded in the presence of 5 mM imidazole. Subsequently, the column was washed with 8 M urea, 10 mM Tris/HCl, 100 mM NaH2PO4 and 8 mM 2-mercaptoethanol (pH 8.0) containing increasing concentrations of imidazole: 5, 10, 20, 30 and 40 mM. Finally, PIV was eluted with the same buffer containing 200 mM imidazole.

Before refolding, purified PIV was incubated overnight at 20 °C in the elution buffer containing 10 mM DTT (dithiothreitol) and then dialysed overnight in a Spectra/Por membrane (molecular-mass cut-off 10 kDa) against an 100-fold volume of 4 M urea, 50 mM Tris/HCl, 100 mM NaCl and 0.005% Tween 20 (pH 8.0). Refolding of PIV was performed in 2 M urea, 50 mM Tris/HCl, 100 mM NaCl, 2.2 mM MgCl2, 50 mM glycine, 5 mM GSH, 0.5 mM GSSG (GSH/GSSG ratio 10:1) and 0.005% Tween 20 (pH 8.0), in an 100-fold volume of the sample at 4 °C for 10 h and subsequently in an 100-fold sample volume of the same buffer containing 1 M urea at 4 °C for 72 h. The refolding buffer was then exchanged with 100 mM NaCl, 50 mM Tris/HCl and 0.005% Tween 20 (pH 8.0), incubated for 12 h, and changed twice. Finally, the protein solution was cleared by centrifugation at 12000 g and the purified protein was stored in this buffer at −20 °C. The protein concentration of recombinant PIV was estimated to be 10 μg/ml (~0.4 μM), as determined on Coomassie Blue-stained gels calibrated with the Page-Ruler unstained protein ladder from Fermentas Life Sciences (0.2 mg/ml of each protein).

As a control, an irrelevant His6-tagged protein (the uPAR domain 1, uPAR1-94, hereafter referred to as uPAR-D1 [25]) was expressed, purified and refolded using a similar procedure as for recombinant PIV.

Exposure of purified pro-uPA, Plg, uPAR and PAI-1 to pseudomonal proteases

Pro-uPA (200 ng for immunoblot analysis and activity measurement, 5 μg for N-terminal microsequencing), Plg (200 ng for activity measurement, 1 μg for Coomassie Blue staining and 5 μg for N-terminal microsequencing), uPAR (20 ng) or PAI-1 (200 ng) were adjusted to a 10–20 μl final volume in TBS (Tris-buffered saline: 100 mM Tris/HCl and 100 mM NaCl, pH 7.5). Proteins were either left untreated for 1 h at 37 °C, or exposed (i) to the respective activators of pro-uPA and Plg, Pm (45 nM) and uPA (90 nM), (ii) to purified LasB in the range 1–100 nM, (iii) to recombinant PIV in the range 7.5–200 nM, or to 200 nM recombinant uPAR-D1, as negative control, (iv) to 100 nM native PIV, or (v) to PAO1-Sec or PDO240-Sec diluted in the range 0.04–1% (v/v) in TBS. Some assays were performed in the presence of various combinations of the following selective protease inhibitors (Sigma–Aldrich), at the indicated final concentrations: OPA (o-phenanthroline), 10 mM; benzamidine, 5 mM; E-64 [trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane], 10 μM; and pepstatin A, 5 μM. After incubation, samples were cooled on ice, and frozen before further analysis (see below).

Exposure of fibroblast-derived human pro-MMP-2 to pseudomonal proteases

Human myofibroblasts were isolated from surgically resected stenotic calcified cardiac valves as reported previously [28]. Patient informed consent was given for collection of these samples, and the work was carried out according to appropriate ethical regulations. Cell cultures were routinely grown using Smooth Muscle Cell Growth Medium 2 (Promocell Bioscience) containing the following growth additives, all from Promocell Bioscience: 5% (v/v) FBS (fetal bovine serum); 2 ng/ml basic fibroblast growth factor; 0.5 ng/ml epithelial growth factor; and 5 μg/ml insulin. For experimental purposes, cells between passages 1 and 5 were detached with trypsin, then seeded in six-well plates (Tissue Culture Products) and cultured up to high density in the same medium. Then, they were washed with PBS, and maintained in culture for 24 h in 1 ml of FBS-free medium. Cell culture medium, which contains the array of secreted fibroblast factors, including proteases, was then collected and centrifuged at 300 g for 10 min, then at 18000 g for 30 min, in order to eliminate detached cells and debris. In such medium, pro-MMP-2 has been reported to be the major secreted gelatinolytic enzyme [28]. For zymography assay 20 μl of cell culture medium was left untreated as control, or incubated with purified or recombinant pseudomonal proteases, or with bacterial secretomes, as described above. For proteolytic activity measurement, 100 μl of cell culture medium was left untreated as control, or exposed to 2 nM LasB or to 10 nM native PIV, overnight at 37 °C. After incubation, samples were cooled on ice, and frozen before further analysis (see below).

Immunoblot analysis

Proteins were denatured in the presence of 2% (w/v) SDS and 5% (v/v) 2-mercaptoethanol for 5 min at 95 °C, followed by SDS/PAGE on 10 or 14% polyacrylamide gels in a Mini-PROTEAN 3 apparatus (Bio-Rad Laboratories). For apparent molecular mass determination, the Page-Ruler Plus pre-stained protein ladder was used as a standard (range 10000–250000 Da, Fermentas Life Sciences). Separated proteins were then transferred on to PVDF membranes (Pall) in a semi-dry transfer device (Biometra). Membranes were incubated for 60 min in PBS-T (PBS, pH 7.4, containing 0.1% Tween 20) and 5% (w/v) dried skimmed milk, and then overnight at 4 °C with the primary Ab diluted in the same medium. After washings in PBS-T, membranes were incubated for 60 min with the HRP-coupled secondary Ab (dilution 1:10000), then washed. Peroxidase activity was finally developed using the ECL (enhanced chemiluminescence) or ECL+ reagent kits (GE Healthcare) and the corresponding luminescence was revealed by exposure of membranes to X-ray-sensitive blue films (CEA). Films were scanned with a GS-800 calibrated densitometer and pictures were prepared with the PDQuest 7.1.1 software (both Bio-Rad Laboratories).

N-terminal microsequencing

After electrophoresis under reducing conditions, proteins were transferred on to PVDF membranes, as described above. Protein bands, visualized by Coomassie Blue staining (0.1%), were then excised and subjected to automated Edman degradation in a pulse liquid-phase sequencer (Procise 492, Applied Biosystems).

Gelatin zymography

For gelatinolytic activity assays, proteins (200 ng of recombinant PIV or uPAR-D1, or 20 μl of protease-exposed fibroblast culture medium) were added with 2% (w/v) SDS and separated by SDS/PAGE on 10% polyacrylamide gels, including 0.2% (w/v) gelatin (Gibco-Invitrogen). Gels were then washed twice for 30 min in a 2.5% (v/v) Triton X-100 solution, developed overnight at 37 °C in 0.2 M Tris/HCl and 5 mM CaCl2 (pH 7.8), and finally stained with Coomassie Blue (0.1%). Gels were scanned as described above.

Proteolytic activity measurement towards synthetic substrates

For uPA and Pm activity measurement, samples were placed in a 96-well plate and adjusted to a 160 μl final volume in 50 mM Tris/HCl and 150 mM NaCl (pH 8.0), in the presence of 0.13 mM benzoyl-β-Ala-Gly-Arg-pNA (p-nitroanilide) (Pefachrome uPA, Loxo) or in 100 mM Tris/HCl, 0.05% Tween 20 and 0.01% BSA (pH 7.5), in the presence of 0.19 mM D-Val-Leu-Lys-pNA (DVLK-902, Molecular Innovations) respectively. The release of pNA at 37 °C was then monitored at 405 nm in a spectrophotometer (SLT-Labinstruments) over a 30 min time period. The initial rate of substrate hydrolysis derived from the data was then expressed as ΔA, in m-units/min.

For PIV activity measurement, 250–500 ng of the recombinant PIV, or 500 ng of control recombinant uPAR-D1, were placed in a 96-well plate and completed to a 250 μl final volume in 50 mM Tris/HCl, 100 mM NaCl and 0.005% Tween 20 (pH 8), in the presence of 0.3 mM of tosyl-glycyl-prolyl-lysine-4-nitroanilide acetate (Roche), before substrate hydrolysis was recorded as described above.

For MMP-2 activity measurement, samples were placed in a 96-well plate and adjusted to a 200 μl final volume in 50 mM Tris/HCl, 50 mM NaCl and 5 mM CaCl2 (pH 7.5), in the presence of 10 μM (7-methoxycoumarin-4-yl)acetyl-Pro-Leu-Gly-Leu-[3-(2,4-dinitrophenyl)-L-2,3-diaminopropionyl]-Ala-Arg (M-1895, Bachem). Substrate hydrolysis was then recorded in a spectrofluorimeter (SpectraMax Gemini EM, Molecular Devices) over a 30 min time period, with excitation and emission wavelengths of 355 and 460 nm respectively.

Determination of the kinetic parameters of pro-uPA activation

LasB (0.01 nM), native PIV (0.1 nM) or Pm (0.01 nM, standardized by active-site titration using 4-methylumbelliferyl p-guanidinobenzoate [29]), were incubated with various concentrations of pro-uPA (0–10 μM) in 50 mM Hepes, 150 mM NaCl, 0.01% sodium azide and 0.01% Triton X-100 (pH 7.4). Zymogen activation was followed by recording the release of AMC (7-amino-4-methylcoumarin) generated by cleavage of the reporter substrate Boc (t-butoxycarbonyl)-Leu-Gly-Arg-7-amino-4-methylcoumarin (20 μM) by activated uPA. In each experiment, the specific activity of uPA towards the reporter substrate was determined using active-site-titrated uPA (0–10 nM) under the same conditions. Substrate cleavage by pro-uPA, LasB, PIV or Pm was negligible. Data were recorded every 30 s for 30 min using a HTS 7000 Bio Assay Reader (PerkinElmer), with excitation and emission wavelengths of 360 and 465 nm respectively.

The time-dependent increase in AMC was fitted to a modification of an integrated rate equation describing a coupled zymogen activation and reporter reaction initially used to analyse the activation of Plg [30,31]: Embedded Image where AMC(t) and AMC0 are the concentrations of AMC measured at time t and baseline respectively, t0 is the delay between the start of the reaction and the first measurement, Vz is the velocity of zymogen activation, and ars is the specific activity of uPA towards the reporter substrate. The zymogen activation velocities Vz at different pro-uPA concentrations calculated from at least three different experiments were subsequently fitted to the Michaelis–Menten equation to obtain the kinetic parameters Km, kcat and kcat/Km. Both curve-fitting steps were performed with proFit (Quantum Soft) using a Levenberg–Marquardt algorithm.

Data presentation

Results are expressed as means±S.E.M. for the indicated number of experiments performed independently.

RESULTS

Activation of human pro-uPA by purified LasB

To investigate the capacity of pseudomonal LasB to activate pro-uPA, recombinant human pro-uPA (200 ng) was first exposed to increasing concentrations (1–100 nM) of the purified bacterial protease for 60 min at 37 °C. It is of note that such LasB concentrations are comparable with those produced by bacteria in vitro (up to 500 nM [32]) and actually detected in vivo, such as in the airways of patients with cystic fibrosis (up to 300 nM [33]). Immunoblot analysis performed under reducing conditions revealed that LasB progressively converts single-chain pro-uPA (~47000 Da) into two molecular species of ~29000 and ~17000 Da (Figure 1A). These molecular masses are similar to those of the B- and A-chains of uPA, i.e. the catalytic and binding domains of the protease respectively, as they are generated via exposure of pro-uPA to its canonical activator Pm (Figure 1A). Accordingly, N-terminal microsequencing indicated that these LasB-generated species possess N-terminal sequences Ile-Ile-Gly-Gly and Ser-Asn-Glu-Leu respectively, corresponding to amino acids 159–163 and 1–4 of human pro-uPA. Both fragments are thus generated upon hydrolysis of the Lys158–Ile159 peptide bond within pro-uPA (Table 1), i.e. the activation cleavage site. Proteolytic activity measurement using a selective uPA chromogenic substrate confirmed that pro-uPA can be readily activated by LasB, with a bacterial protease concentration as low as 1 nM and an enzyme/substrate molar ratio of 1:450 (Figure 1B). On one hand, after exposure of pro-uPA (200 ng) to LasB in the range 10–100 nM, the activity of LasB-generated uPA reached up to 75% of the activity displayed by 200 ng of purified human active uPA, used as a positive control (Figure 1B). On the other hand, LasB-driven pro-uPA activation is actually comparable with the activation of pro-uPA resulting from exposure to Pm in an enzyme/substrate molar ratio of 1:10 (Figure 1B). It is of note that, for a high LasB concentration (i.e. 100 nM) (Figure 1A), the A-chain is further converted into an ~13000 Da molecular species, bearing a Phe-Ser-Asn-Ile N-terminus and thus resulting from the hydrolysis of the Tyr24–Phe25 peptide bond within the growth factor domain that makes part of the uPA A-chain (Table 1). In contrast, the B-chain remains remarkably stable. Complementary experiments showed an efficient time-dependent activation of pro-uPA by LasB, detectable as soon as 15 min following exposure to 10 nM LasB, whereas a further processing of the A-chain is again observed for the longest incubation time (i.e. 5 h) (results not shown). Finally, to characterize the efficiency of pro-uPA activation by LasB, we determined the kinetic parameters of the reaction, and found kcat, Km and kcat/Km values of 4.9±0.34 s−1, 8.9±1.1 μM and 0.55±0.08 μM−1·s−1 respectively, compared with values obtained with active-site-titrated Pm of 4.2±0.13 s−1, 3.2±0.24 μM and 1.3±0.11 μM−1·s−1 respectively.

Figure 1 Activation of human pro-urokinase by purified LasB, or by the secretomes of Ps. aeruginosa LasB-expressing or LasB-deficient strains

Purified human pro-uPA (200 ng, ~450 nM) was left untreated (NT), exposed to its canonical activator Pm (45 nM), to increasing concentrations of purified LasB (A and B), or to increasing quantities (expressed as a percentage of the final volume of the reaction mixture) of the secretome either from the LasB-expressing pseudomonal strain PAO1 (PAO1-Sec; C and D) or from the isogenic LasB-deficient strain PDO240 (PDO240-Sec; E and F), for 60 min at 37 °C. (A, C and E) Proteins were analysed by immunoblotting using a chicken anti-uPA polyclonal Ab (dilution 1:2000), as described in the Experimental section. Portions of the films corresponding to the location of the relevant molecular species, located by an arrow on the left-hand side, are shown and their apparent molecular mass is indicated on the right-hand side. One out of three analyses showing similar results is illustrated. (B, D and F) In parallel, uPA activity was assayed towards a selective chromogenic substrate (Pefachrome uPA). The proteolytic activity of 200 ng of purified uPA was used as an external positive control and corresponds to 100% activation of the zymogen present in the assay. Results are means+S.E.M. for three independent experiments. Note that pro-uPA, LasB, PAO1-Sec and PDO240-Sec have no or minimal activity on Pefachrome uPA. ΔOD, change in absorbance (ΔA).

View this table:
Table 1 Cleavage sites for LasB and PIV within components of the human Plg activation system

None, no cleavage observed. Preferred amino acids in the P1 (for PIV) and P1′ (for LasB) position of the cleavage site are in bold.

Activation of human pro-uPA by bacterial secretomes

We next explored whether the secretome from a pseudomonal LasB-expressing strain (i.e. PAO1) could trigger activation of pro-uPA. Within such a secretome, LasB is both the major secreted bacterial protein and one major protease, reaching a concentration up to 500 nM (see the gelatinolytic activity of PAO1-Sec in Figures 3E and 5D, right-hand panels, and [32]). Exposure of recombinant human pro-uPA to this secretome actually results in its activation, as determined by both immunoblot analysis (Figure 1C) and proteolytic activity measurement (Figure 1D). This processing is already detectable with a 1:2500 dilution (i.e. 0.04%) of PAO1-Sec, whereas a 1:100 dilution (i.e. 1%) results, as observed with purified LasB, in a further processing of the A-chain (Figure 1C). In order to evaluate the role of LasB as a major pro-uPA-activating enzyme within the pseudomonal secretome, we analysed the capacity of the secretome from an isogenic LasB-deficient strain (i.e. PDO240) to activate the human zymogen. Using PDO240-Sec, a significant pro-uPA activation is observed (Figures 1E and 1F), although limited compared with that produced with PAO1-Sec.

Altogether, these results suggest that LasB is indeed a major pseudomonal pro-uPA activator, although Ps. aeruginosa also secretes one or several other pro-uPA activator(s).

Identification of PIV as a potent pro-uPA activator

In order to identify the unknown pro-uPA activator(s) expressed within the pseudomonal secretome, pro-uPA was exposed to PDO240-Sec in the presence of selective protease inhibitors, before analysing pro-uPA activation by immunoblot. Pro-uPA activation by the LasB-deficient PDO240-Sec is not affected by metallo-, cysteine or aspartic protease inhibitors (i.e. OPA, E-64 and pepstatin A respectively), whereas it is clearly reduced, although not abrogated, in the presence of the wide-spectrum serine protease inhibitor benzamidine (Figure 2A). Consistently, and although the cation chelator OPA totally inhibits pro-uPA activation by 100 nM purified LasB (Figure 2B, right-hand panel), it is not sufficient to fully repress activation of the zymogen by PAO1-Sec, although activation is drastically impaired (Figure 2B, left-hand panel). This reflects both a major LasB-dependent and a minor LasB- and metalloprotease-independent pro-uPA activation by PAO1-Sec. As observed with PDO240-Sec, the metalloprotease-independent pathway of activation is actually sensitive to a serine protease inhibitor, as indicated by the almost total abrogation of activation provided by the combination of OPA and benzamidine, whereas none of the other inhibitor combinations (OPA/E-64, OPA/pepstatin A) adds any further inhibition over that provided by OPA alone (Figure 2B, left-hand panel). Taken together, these observations confirm that, besides LasB, activation of human pro-uPA by the pseudomonal secretome also involves a serine protease activity.

Figure 2 Activation of human pro-urokinase by a benzamidine-sensitive secreted pseudomonal protease

Pro-uPA was left untreated (NT), or exposed to 1% (v/v) PDO240-Sec (A), 1% (v/v) PAO1-Sec (B, left-hand panel), or 100 nM LasB (B, right-hand panel), for 60 min at 37 °C, in the presence or absence of the following protease inhibitors: OPA (10 mM), benzamidine (Benz., 5 mM), E-64 (10 μM) or pepstatin A (Pep., 5 μM), which were either added alone, or in combination (Mix, combination of all four inhibitors). Pro-uPA activation was then evaluated by immunoblotting (n=3), as described in the legend to Figure 1.

Among the pseudomonal secreted serine proteases is PIV, also referred to as PrpL or Ps-1 [3436]. This enzyme, for which activity was reported to be affected, but not completely blocked, by benzamidine [36], targets substrates bearing a lysine residue at the P1 position of the sessile peptide bond [23,36]. Such inhibitor and substrate specificities thus share similarities with the metalloprotease-independent pro-uPA activation that is observed upon its exposure to pseudomonal secretomes. In support of the assumption that PIV may be a pro-uPA activator, it must be noted that PIV is expressed to significant and similar levels in both the PAO1 and PDO240 secretomes, as demonstrated by evaluation of the global gelatinolytic activity of the secretomes using zymography (Figure 3E). Indeed, the gelatinolytic profile of PAO1-Sec identifies, under the particular electrophoretic conditions used, three major components of ~52000, ~167000 and ≥250000 Da, which, based on previous reports [15,35], correspond to AprA, LasB and PIV respectively. The gelatinolytic profile of PDO240-Sec shows the AprA and PIV components, with activities similar to those observed in PAO1-Sec, although lacking, as expected, the LasB component. In order to ascertain that PIV can be a pro-uPA activator, the DNA encoding mature PIV was cloned from the PAO1 genome [37] and then inserted into a prokaryotic expression plasmid encoding for a translation product bearing an N-terminal His6 tag, as depicted schematically in Figure 3(A), following a procedure described previously for a number of eukaryotic serine proteases [38,39]. As expected, plasmid transformation in E. coli, followed by protein expression, extraction under denaturating conditions and purification, yields a major ~30000 Da fusion protein (Figure 3B). It is of note that the two-step purification procedure (i.e. pH-dependent affinity chromatography on Ni-NTA followed by a second Ni-NTA chromatography using an imidazole concentration gradient) results in a rather pure preparation of recombinant PIV, which, however, remains slightly contaminated by a ~27000 Da protein (Figure 3B). This latter contaminant is also present in the final preparation of an E. coli-expressed His6-tagged protein generated through an identical procedure and taken as a negative control, in this case the ~14500 Da D1 of the uPA receptor, uPAR (uPAR-D1, Figure 3B). SDS/PAGE performed under reducing conditions indicates that, upon dialysis-based protein refolding, recombinant PIV readily proceeds itself into smaller species with ~20000 and ~17000 Da (Figure 3B), a feature which is also observed when proteins are analysed under non-reducing conditions (results not shown). These truncated PIV species remain reactive with an Ab directed against the His6 tag (Figure 3C), demonstrating that they result from cleavage of the enzyme within its C-terminal region. Remarkably, such PIV autoprocessing has already been reported for naturally produced, as well as for recombinantly expressed, PIV, and was found not to affect the enzyme's proteolytic activity [23,35,37,40,41]. Accordingly, our recombinant PIV preparation is active towards a known PIV chromogenic susbstrate, chromozyme PL [34] (Figure 3D), and appears, upon gelatin zymography, as a high-molecular-mass doublet species (Figure 3E). In contrast, the preparation of recombinant uPAR-D1 is devoid of any peptidase/proteinase activity in both assays (Figures 3D and 3E).

Figure 3 Production of recombinant active pseudomonal PIV

(A) Schematic structure comparison of the natural and E. coli-expressed PIV proteins (upper and lower panels respectively). aa, amino acids. (B and C) Preparations of E. coli-expressed PIV, or uPAR-D1 (used as a control non-protease recombinant protein preparation), either purified or purified and refolded, were analysed by SDS/PAGE under reducing conditions followed by transfer on to a PVDF membrane and Coomassie Blue staining (B), or immunoblotting using an anti-His6 monoclonal Ab (Penta-His, 1 μg/ml, C). In (B), # is an E. coli contaminant and * and ** are truncated PIV species. (D and E) Refolded recombinant PIV and uPAR-D1 preparations were evaluated for peptidase (D) and proteinase (E) activity, using the chromogenic substrate Chromozym PL or gelatin zymography respectively, as detailed in the Experimental section. For zymography, a reference gelatinolytic profile was provided by the bacterial secretomes originating from the pseudomonal PAO1 and PDO240 strains. Results obtained for PIV and uPAR-D1 batches representative of two or three independent preparations are illustrated. ΔOD, change in absorbance (ΔA).

Exposure of recombinant human pro-uPA to increasing concentrations of recombinant PIV (7.5–200 nM) results in a progressive conversion of single-chain pro-uPA into two-chain active uPA, with uPA activity reaching up to 80% of the activity of an equivalent amount of purified uPA, whereas, in contrast, recombinant uPAR-D1 leaves pro-uPA unchanged and inactive (Figure 4). These data were validated by the observation that native PIV (100 nM), purified directly from the extracellular medium of pseudomonal cultures, also targets and activates the zymogen (Figure 4). It is of note that, for a high PIV concentration (i.e. over 100 nM PIV), a further processing of the uPA A-chain is observed, leading to the generation of an ~13000 Da species which results from hydrolysis of the Lys23–Tyr24 peptide bond (Figure 4A and Table 1). In addition, exposure of pro-uPA to recombinant PIV in the presence of selective protease inhibitors revealed that, as observed with bacterial secretomes, PIV-induced pro-uPA activation is drastically sensitive to benzamidine, whereas it is not affected by the other inhibitors tested (results not shown). Finally, kinetic parameters of pro-uPA activation by native PIV were found to be kcat=0.73±0.04 s−1, Km=6.2±0.65 μM and kcat/Km 0.12±0.01 μM−1·s−1.

Figure 4 Activation of human pro-urokinase by recombinant pseudomonal PIV

Pro-uPA was left untreated (NT), or exposed to Pm, to increasing concentrations of recombinant PIV (rPIV, 7.5–200 nM), to recombinant uPAR-D1 (200 nM), or to native PIV (nPIV, 100 nM), for 60 min at 37 °C. Immunoblot analysis for pro-uPA processing (A, representative of two experiments) and measurement of uPA activity (B, means+S.E.M., n=4) were then performed exactly as described in the legend to Figure 1. Note that pro-uPA, rPIV, uPAR-D1 or nPIV have no appreciable direct activity on the uPA substrate. ΔOD, change in absorbance (ΔA).

Altogether, our results thus demonstrate that, within the pseudomonal secretome, not only LasB, but also PIV targets and activates pro-uPA.

Identification of other substrates for pseudomonal proteases within the host Plg activation system

We analysed further whether PIV cleaves other known LasB targets within the human Plg activation system (i.e. uPAR, PAI-1 and pro-MMP-2), and, more generally, whether pseudomonal proteases can process Plg directly [32,4244].

Exposure of Plg to either purified LasB (10 nM) or to 1% of the LasB-expressing PAO1-Sec results in the generation of identical new molecular species, including a doublet of ~72000/65000 Da, and a singlet of ~40000 Da (Figure 5A). These species are distinct from the 88/84 kDa A-chain, and from the 28 kDa B-chain produced upon Plg activation by uPA (results not shown). In line with this, LasB-processed Plg does not show any proteolytic activity towards a synthetic Pm chromogenic substrate (results not shown). N-terminal microsequencing allowed us to establish that these Plg-derived species result from hydrolysis of the Ser441–Val442 and Val442–Val443 peptide bonds within human Plg (Table 1). They thus correspond to two glycosylation variants of angiostatin, encompassing the Plg kringle domains 1–4 (~72000/65000 Da), and to mini-Plg, composed of the inactive catalytic domain and of the kringle domain 5 (~40000 Da). For higher LasB concentrations (i.e. 100 nM) or with 20% PAO1-Sec, these Plg fragments are processed further into ~43000/39000 and 29000 Da forms, which were found to result from further C-terminal processing of both angiostatin and mini-Plg. In contrast, neither the LasB-deficient PDO240-Sec nor recombinant PIV had any detectable effect on the structure and activity of human Plg (Figure 5A and results not shown respectively).

Figure 5 Processing of components of the human Plg activation system by pseudomonal proteases

Human Plg (~1 μM), recombinant human uPAR (~20 nM), recombinant human PAI-1 (~200 nM), or fibroblast-derived culture medium containing pro-MMP-2 were either left untreated (NT) or exposed to pseudomonal proteases, either pure, or as they are expressed within PAO1-Sec or PDO240-Sec, as indicated in the legends to Figures 1 and 4. (A) The proteolytic processing of Plg was analysed by SDS/PAGE (1 μg of Plg per lane) followed by Coomassie Blue staining. (B and C) uPAR and PAI-1 processing were evaluated by immunoblot (10 ng of protein per lane) using an anti-uPAR monoclonal Ab directed against D2 (#3932, 0.05 μg/ml) (B) or an anti-PAI-1 monoclonal Ab recognizing residues 24–158 of human PAI-1 (C-9, 0.4 μg/ml) (C). (D) (Pro-)MMP-2 activity was detected by gelatin zymography (20 μl of reaction mixture per lane), with the gelatinolytic profiles of the bacterial secretomes depicted in the right-hand panel. Note that the AprA activity is higher in PDO240-Sec compared with PAO1-Sec. Results are representative of two or three independent experiments.

As we reported previously [32], it was confirmed that purified LasB (100 nM), as well as the PAO1-derived secretome (20%), convert recombinant soluble uPAR (~55000 Da) into a ~45000 Da D2 (domain 2)–D3 (domain 3)-truncated species lacking the shed N-terminal D1 (Figure 5B), which reflects hydrolysis within the uPAR D1–D2 linker region. Exposure of uPAR to either PDO240-Sec or recombinant PIV shows no evidence of a proteolytic processing of this important cell receptor (Figure 5B).

Furthermore, LasB was reported to target and inactivate PAI-1 [42]. Accordingly, LasB at 100 nM, as well as PAO1-Sec (20%), process recombinant PAI-1 (~57000 Da) into a slightly shorter molecular species of ~54000 Da (Figure 5C). This feature has also been reported for other bacterial thermolysin-related proteases [45], in which case it corresponds to cleavage within the C-terminal domain of the serpin, in the vicinity of the reactive-centre loop. In contrast, PAI-1 was not processed at all using the LasB-deficient secretome or recombinant PIV (Figure 5C).

Finally, the present study confirms that LasB at 100 nM, as well as PAO1-Sec in the range 1–20%, transforms pro-MMP-2, as it is spontaneously released by human vascular fibroblasts in their culture medium with a molecular mass of 63000 Da [28], into the ~57000 Da active form of the enzyme, as determined by gelatin zymography (Figure 5D) [43,44]. Activation of fibroblast-derived pro-MMP-2 by purified LasB was also confirmed by protease activity measurement towards an MMP-selective fluorigenic substrate (results not shown). In contrast, under conditions which trigger efficient pro-uPA activation, recombinant PIV or PIV naturally expressed by the PDO240 strain, leave pro-MMP-2 unaffected (Figure 5D).

Similar results regarding the absence of Plg, uPAR, PAI-1 or pro-MMP-2 processing by PIV were also obtained using the purified native protease (results not shown).

DISCUSSION

A number of pathogenic bacteria take profit on the host Plg system to promote their own spread and invasion through host tissues and physiological barriers, in a way that mimics that observed for malignant tumour cells [79]. In the case of Ps. aeruginosa, Plg/Pm binding at the bacterial surface has been reported for a number of reference and clinical isolates [22,46,47]. Moreover, bound Pm promotes invasion of Ps. aeruginosa within fibrin and reconstituted basement membrane matrices [22]. However, in contrast with S. aureus and Y. pestis [7,9,13], Ps. aeruginosa had not been reported so far to display any activator for pro-uPA, tPA or Plg.

In the present study, we have established that Ps. aeruginosa secretes proteases that directly activate the zymogen form of human uPA, including the thermolysin-like metallopeptidase LasB and the serine protease PIV. With a kcat/Km of 0.55 μM−1·s−1, LasB appears to be a potent pro-uPA activator, with a specificity constant in the range of that observed for Pm (kcat/Km of Pm=1.3 μM−1·s−1), whereas PIV is approx. 5-fold less efficient than LasB (kcat/Km=0.12 μM−1·s−1). Still, both proteases activate pro-uPA much more efficiently than the S. aureus-expressed thermolysin-like protease aureolysin (kcat/Km of aureolysin=0.003 μM−1·s−1) [13]. Remarkably, besides Ps. aeruginosa, several pathogenic bacteria that express thermolysin-like peptidases, including S. aureus [13], Burkholderia cepacia and Serratia marcescens (N. Beaufort, P. Wojciechowski, J. Potempa and V. Magdolen, unpublished work), also display pro-uPA-activating capacities. It is thus tempting to propose that pro-uPA activation might be a feature common to bacterial M4-related peptidases and thus a property shared by other pathogenic bacterial species.

Within human tissues, (pro-)uPA is locally expressed at low levels by epithelial, endothelial and stromal cells [11,12]. Upon inflammation or infection, its expression by these cells is increased, whereas it is also massively released from neutrophils recruited at the site of inflammation/infection, where plasma proteins, including Plg, also exudate [11,12]. As pro-uPA is positioned upstream in the auto-amplified uPA/Pm/MMP proteolytic loop, its activation by bacterial enzymes certainly results in an efficient triggering and maintenance of this host proteolytic system. Concerning Plg, the zymogen was not processed into active Pm by any of the pseudomonal proteases. Plg was, however, cleaved by LasB, but not by PIV, into (i) mini-Plg, which can be converted further into mini-Pm [13], and (ii) angiostatin, which displays the interesting capacity to dampen leucocyte migration, and might thus provide bacteria with immunoevasive capacities [48]. Inactivation of host serpins and/or activation of pro-MMPs are other known features of bacterial proteases that may allow a pathogen to use one of the major extracellular proteolytic systems of the host for its own dissemination [7,13,43,44]. In this respect, we confirmed that LasB neutralizes PAI-1 [42], whereas it activates pro-MMP-2 [43,44], thus participating further in the bacterial-driven deregulation of the otherwise tightly regulated Plg activation system. By contrast, PIV leaves both substrates unaffected. Of note is the fact that the proteolytic processing of the other principal Plg activator tPA by pseudomonal proteases was not investigated in the present study, since its activation is promoted by its binding to fibrin rather than by proteolysis [11].

Although the effects of bacteria and bacterial compounds on the host Plg activation system have been analysed extensively [7], the reciprocal activity of the Plg system on bacteria has been poorly investigated so far, with the exception of (pro-)uPA, which promotes pseudomonal growth and enhances the bacterial proteolytic activity [49,50]. Interestingly, we found that both LasB and PIV process the A-chain of (pro-)uPA, a feature which prevents uPA binding to its cellular receptor uPAR [51]. Along with this, LasB targets the D1–D2 linker region within uPAR, which abrogates the uPA–uPAR interaction [32]. Altogether, we speculate that the Pseudomonas-dependent disruption of the (pro-)uPA–uPAR complex might increase the bioavailability of the otherwise membrane-anchored (pro-)uPA, thereby promoting bacterial growth.

In contrast with the rather well-characterized LasB, little information is currently available concerning PIV. PIV is widely expressed among reference and clinical isolates [37,52], and is a potent virulence factor in eye and lung infection models [35,37,41]. Interestingly, this serine peptidase appears to be clearly distinct from eukaryotic serine proteases. On one hand, PIV readily processes its own mature catalytic domain (~30 kDa) into smaller 20 and 17 kDa fragments, a cleavage that does not impair its proteolytic activity [40,41]. On the other hand, it displays a bulky 186-amino-acid pro-domain, which has been proposed to prevent intracellular autoprocessing, or to act as a chaperone protein [41]. We now establish that replacement of the PIV pro-domain by a short, 17-residue, sequence, actually leads to an active form of the enzyme, although it is still N-terminally extended. This suggests that PIV activation does not result from an insertion of the N-terminus of the mature catalytic domain into a cleft near the active site, triggering a conformational rearrangement of the latter as observed with eukaryotic serine proteases [53], but rather from uncovering of the pre-existing active site, as described, e.g., for subtilisin-like peptidases [54]. Concerning the PIV substrate specificity, this protease targets peptide bonds bearing a basic residue in the P1 position [23], whereas LasB targets peptide bonds bearing a bulky or apolar residue in the P1′ position [4]. As a consequence, and although these two proteases belong to different protease families, they both hydrolyse the unique activation cleavage site within human pro-uPA, i.e. Lys158–Ile159. In contrast, a number of LasB substrates (e.g. Plg, uPAR, PAI-1 and pro-MMP-2; Table 1) are not affected at all by PIV, or are processed at different peptide bonds (e.g. the uPA A-chain; Table 1).

In conclusion, the ubiquitous bacterium Ps. aeruginosa expresses proteases, including the well-known thermolysin-like metallopeptidase LasB, and the more recently described PIV, which both combine to hijack the host Plg system, primarily via pro-uPA activation, and, in the case of LasB, via PAI-1 inactivation and pro-MMP-2 activation. Such processes might constitute an important virulence mechanism, which may well utilize other pathogenic bacteria expressing similar extracellular proteases.

AUTHOR CONTRIBUTION

Nathalie Beaufort, Paulina Seweryn, Sophie de Bentzmann and Aihua Tang designed, performed and analysed experiments. Josef Kellermann, Christian Sommerhoff, Dominique Pidard and Viktor Magdolen designed and analysed experiments. Nicolai Grebenchtchikov and Manfred Schmitt contributed vital new reagents. Nathalie Beaufort, Christian Summerhoff, Dominique Pidard and Viktor Magdolen wrote the paper. All authors read and approved the final paper.

FUNDING

This work was supported by fellowships from the Alexander von Humboldt Stiftung, Bonn, Germany, and from the Fondation Lefoulon-Delalande/Institut de France, Paris, France (to N. B.), by an Erasmus student mobility grant (to P. S.), and by a collaboration grant from the Centre de Coopération Universitaire Franco-Bavarois/Bayerisch-Französisches Hochschul-zentrum (CCUFB/BFHZ), Munich, Germany (to N. B. and V. M.). D. P. is Chargé de Recherche at the Institut National des Sciences du Vivant within the Centre National de la Recherche Scientifique, Paris, France.

Acknowledgments

We express gratitude to Sabine Streicher and Antoine Freyss for excellent technical assistance, to Reinhard Mentele for help with N-terminal microsequencing, to Michel Chignard for providing purified LasB, to Christine Choqueux for providing the procedure for culturing human myofibroblasts, and to Wolfram Bode for helpful discussions.

Abbreviations: Ab, antibody; AMC, 7-amino-4-methylcoumarin; AprA, alkaline protease; D1, domain 1; D2, domain 2; FBS, fetal bovine serum; HRP, horseradish peroxidase; MMP, matrix metalloproteinase; Ni-NTA, Ni2+-nitrilotriacetate; OPA, o-phenanthroline; PAI-1, plasminogen activator inhibitor type-1; PBS-T, PBS (pH 7.4) containing 0.1% Tween 20; PIV, protease IV; Plg, plasminogen; Pm, plasmin; pNA, p-nitroanilide; TBS, Tris-buffered saline; tPA, tissue-type plasminogen activator; uPA, urokinase-type plasminogen activator; uPAR, uPA receptor

References

View Abstract