Ubiquitin and ubiquitin-modified proteins activate the Pseudomonas aeruginosa T3SS cytotoxin, ExoU


E-mail frankd@mcw.edu; Tel. (+1) 414 955 8766; Fax (+1) 414 955 6535.


Pseudomonas aeruginosa is an opportunistic Gram-negative pathogen that possesses a type III secretion system (T3SS) critical for evading innate immunity and establishing acute infections in compromised patients. Our research has focused on the structure–activity relationships of ExoU, the most toxic and destructive type III effector produced by P. aeruginosa. ExoU possesses phospholipase activity, which is detectable in vitro only when a eukaryotic cofactor is provided with membrane substrates. We report here that a subpopulation of ubiquitylated yeast SOD1 and other ubiquitylated mammalian proteins activate ExoU. Phospholipase activity was detected using purified ubiquitin of various chain lengths and linkage types; however, free monoubiquitin is sufficient in a genetically engineered dual expression system. The use of ubiquitin by a bacterial enzyme as an activator is unprecedented and represents a new aspect in the manipulation of the eukaryotic ubiquitin system to facilitate bacterial replication and dissemination.


Pathogens have evolved a variety of mechanisms to overcome host barriers to parasitism. Physical barriers include an intact epithelium, the outward flow of mucosal fluids, skin keratinization and the synthesis and secretion of antimicrobial peptides (Collins and Brown, 2010). Other barriers encode specific systems to recognize and traffic intracellular invaders through the lysosomal or autophagy pathways for destruction (Dupont et al., 2010). Pathogens that have co-evolved with their hosts encode gene products that manipulate the mammalian environment to enhance replication and spread to the next host.

Microbial interference with the ubiquitylation/deubiquitylation pathway is now recognized as a major evasive mechanism (reviewed in Angot et al., 2007; Rytkonen et al., 2007; Shames et al., 2009; Collins and Brown, 2010; Dupont et al., 2010). A variety of bacterial effector proteins possess deubiquitylation (Orth et al., 2000; Zhou et al., 2005; Misaghi et al., 2006; Catic et al., 2007; Rytkonen et al., 2007; Sweet et al., 2007; Ye et al., 2007; Le Negrate et al., 2008) or E3 ligase activities (Haraga and Miller, 2006; Janjusevic et al., 2006; Zhang et al., 2006; Rohde et al., 2007; Kubori et al., 2008). Upon intracellular delivery, bacterial effectors also serve as ubiquitylation targets (Kubori and Galan, 2003; Schnupf et al., 2006; Patel et al., 2009) or as scaffolding proteins for modification complexes (Kim et al., 2005; Nomura et al., 2006; Angot et al., 2007; Jubelin et al., 2010). Collectively, these incursions benefit the bacterium by modulating host signalling pathways, particularly those involved in inflammation, as well as altering host or bacterial protein trafficking and stability.

Pseudomonas aeruginosa is a soil bacterium and significant opportunistic pathogen (Pier and Ramphal, 2005). This organism takes advantage of tissue that is damaged or hosts with compromised immune systems to establish a nidus of infection that can spread systemically. It is particularly problematic in people who suffer neutropenia, mechanically ventilated patients and individuals with severe burns. Treatment is difficult because of intrinsic and acquired resistance to antimicrobial agents (Giamarellou, 2000). In contrast to acute infections, P. aeruginosa can also establish long-term, localized, chronic infections in cystic fibrosis patients (Hauser et al., 2011). Damage in this case is not only due to bacterial replication but also to the host inflammatory response to an invader that cannot be eliminated.

Pseudomonas aeruginosa is notable for its expression of many tissue degradative enzymes and toxins that alter eukaryotic cell physiology. The organism uses a type III secretion system (T3SS) to deliver or inject at least four toxins into the cytoplasm of infected cells (Yahr et al., 1997). Injection of ExoS or ExoT perturbs cellular signalling and cytoskeletal components (Frithz-Lindsten et al., 1997; Ganesan et al., 1999). Each enzyme is bifunctional and contains Rho GAP and ADP-ribosyltransferase domains (Goehring et al., 1999; Sun and Barbieri, 2003). The introduction of junctional gaps in endothelial tissue is mediated by the injection of ExoY (Sayner et al., 2004), an adenylyl cyclase (Yahr et al., 1998). The most cytotoxic component injected into cells, ExoU, possesses phospholipase A2 activity (Sato et al., 2003). ADP-ribosyltransferase, adenylyl cyclase and phospholipase activities are generally not detectable from purified proteins in vitro unless an eukaryotic activator/cofactor is present. For ExoS and ExoT the activators are members of the 14-3-3 family of scaffolding proteins (Fu et al., 1993). ExoU is activated in the presence of certain preparations of superoxide dismutase or SOD1 (Sato et al., 2006). The cofactor for ExoY is currently unknown (Yahr et al., 1998). The interaction of cofactor proteins with each enzyme is poorly understood and the development of defined inhibitors will depend on characterizing the mechanisms of activation at the molecular level.

Our studies have focused on defining the mechanism of activation of ExoU by SOD1. This protein was originally identified as an activator of ExoU by using biochemical enrichments of yeast soluble fractions and proteomic approaches (Sato et al., 2006). Similar to the 14-3-3 activators of ExoS and ExoT, SOD1 is ubiquitous in eukaryotes, exists in high concentration in the cytoplasm and the sequence differences between prokaryotic and eukaryotic forms potentially accounted for the specificity of ExoU's toxicity for eukaryotes (Sato et al., 2006). Commercially prepared bovine SOD1 (bSOD1) is also an activator, but the activation activity appears to depend on the tissue of origin and commercial supplier (Benson et al., 2010). The amount of bSOD1 or recombinant yeast SOD1 (ySOD1) required for activation of ExoU is relatively high and is not saturable in kinetic analyses (Benson et al., 2010). These data suggested that a minor population of SOD1 molecules are responsible for the activation of ExoU.

The aim of this study was to determine the properties of bSOD1 and ySOD1 mediating the activation of ExoU. Our analyses indicate that post-translational ubiquitylation of ySOD1 and monoubiquitin, contained within bSOD1 commercial preparations, is responsible for the activation of ExoU in vitro. Of the multiple mechanisms that prokaryotic pathogens use to interact with or modulate the eukaryotic ubiquitylation system, this is the first report of a T3SS toxin utilizing ubiquitin as an activator of enzymatic activity that mediates host cell death.


Activation of ExoU phospholipase activity by peptides

Our first aim was to characterize the minimal region of superoxide dismutase required to activate ExoU in an in vitro assay. To test relatively large fragments with potential overlaps, we used a partial protease digestion approach and controlled the time of digestion (Koth et al., 2003). Initial experiments utilized thrombin (2 peptide fragments), glutamyl endopeptidase (10 fragments), endoproteinase Lys-C (12 fragments), and trypsin (15 fragments) either alone or in combinations (data not shown). Bovine liver SOD1 was denatured to facilitate digestion and the resulting peptide fragments were precipitated for testing in a fluorescence-based in vitro activity assay measuring cleavage of the phospholipid mimic, PED6 (Benson et al., 2010). Interestingly, peptide activation of ExoU comparable with non-treated bSOD1 was detectable in all instances, even samples sequentially digested to completion with three different proteases (Fig. 1A and B). The most prevalent products appeared in the 4–8 kDa range, suggesting that a relatively small fragment possessed the capability to activate ExoU (approximately 36–72 amino acids, Fig. 1A).

Figure 1.

Peptides from a liver bSOD1 preparation activate ExoU in vitro. A. 10–20% Tris-tricine Coomassie-stained SDS-PAGE gel showing peptides generated using alkylated liver bSOD1. Lanes 1, 6 and 12 are molecular weight standards; Lane 2, undigested bSOD1; Lane 3, endoproteinase Lys-C (Lys-C) only; Lane 4, Lys-C digestion; Lane 5, Lys-C and thrombin only; Lane 7, Lys-C + thrombin digestion; Lane 8, Lys-C and glutamyl endopeptidase (Glu-C) only; Lane 9, Lys-C + Glu-C digestion; Lane 10, Lys-C, Glu-C and thrombin only; Lane 11, Lys-C + Glu-C + thrombin digested bSOD1. Arrows indicate a peptide that is present in undigested SOD1 and retained in SOD1 digested to completion with endoproteinase Lys-C, thrombin and glutamyl endopeptidase as outlined. B. Activity assay using either untreated bSOD1, sequentially and completely digested protein or a protease only control as cofactor material. ‘Triple Digestion’ corresponds to lane 11 in (A). C. Western blot analysis of bSOD1 from a 1:1000 proteinase K: bSOD1 (w/w) digestion over a 120 min time-course. Anti-SOD1 (1:5000) was used as the primary antibody in this analysis. D. Specific phospholipase activity (nmol PED6 cleaved min−1 mg−1 rExoU) when proteinase K digested products from denatured bSOD1 are used to activate rExoU over a digestion time-course (error bars ± SEM, n = 3). Untreated refers to bSOD1 that is not denatured nor protease treated and undigested refers to bSOD1 that is denatured but not treated with proteinase K.

To more completely digest bSOD1, we used proteinase K, which possesses 58 predicted cleavage sites within the protein. Similar to site-specific digestion, full-length bSOD1 was not detectable on silver stained gels (data not shown) or by Western blot analysis after the 60 min time point (Fig. 1C), yet specific activity remained relatively constant until 75–90 min digestion (Fig. 1D). We postulated that the accumulation of a peptide activator over time might be demonstrated by an increase in specific activity but this result was not observed. In fact, digestion beyond 75 min showed a decrease in specific activity, indicating that the cofactor was at least partially proteinaceous (Sato et al., 2006). Together, these findings suggested that an SOD1 derived peptide or a protease resistant molecule or both contributed to ExoU activation.

We noted that a peptide corresponding to 6.5 kDa was not only present in the undigested bSOD1 preparation, but was also partially retained in the preparation of triply digested SOD1 (Fig. 1A, lane 11). This peptide was extracted from the gel and identified by mass spectrometry as ubiquitin (Table S1). To determine if the presence of ubiquitin was the common factor determining whether or not different SOD1 preparations activated ExoU, we examined bovine liver, kidney and red blood cell derived proteins by Western blot analysis with a monoclonal antibody specific to ubiquitin (Fig. S1). Using similar amounts of SOD1, we detected ubiquitin in the only preparation capable of activating ExoU, bovine liver SOD1 (Fig. S1 and data not shown). These data suggest that ubiquitin is a cofactor for ExoU activation.

Ubiquitylated proteins from yeast and mammalian cells activate ExoU

Commercial preparations of bSOD1 purified from SDS gels activate ExoU (Sato et al., 2006) suggesting that a subpopulation of bSOD1 molecules might be modified by ubiquitin. Western blot analysis of the bSOD1 indicated that ubiquitin was not covalently associated with liver SOD1 (Fig. S1). To determine whether the activation activity associated with recombinant ySOD1 (Sato et al., 2006) was due to ubiquitylated protein or released monoubiquitin, ySOD1 was purified by cobalt chromatography and subjected to Western blot analysis using equivalent amounts (4 µg) of monoclonal antibodies to the hexa-histidine tag or ubiquitin as probes. Anti-histidine tag antibody recognized a 34 kDa protein when the antigen was loaded at nanogram levels (Fig. 2A). Anti-ubiquitin antibodies recognized protein within the same molecular weight range at levels of 1 µg or higher (Fig. 2A). Detection of signal with anti-ubiquitin required an exposure time of 1 h as compared with a few seconds for blots probed with anti-histidine tag antibody. Monoubiquitin was not detectable (data not shown). These data support the conclusion that a small subpopulation of ubiquitlyated recombinant ySOD1 serves as an ExoU cofactor in vitro.

Figure 2.

Ubiquitylated proteins activate ExoU. A. Histidine tagged recombinant ySOD1 was purified and subjected to a twofold dilution series followed by SDS-polyacrylamide gel electrophoresis and Western blot analysis. The beginning protein concentration for each blot was 4 µg of purified ySOD1. Left panel, Western blot of ySOD1 titration probed with antibody to the histidine tag. Right panel, Western blot analysis identical to the left panel except probed with monoclonal antibody specific for ubiquitin. Equivalent amounts of primary antibody were used for each blot. Twofold more secondary antibody was used to detect the anti-ubiquitin probe. Exposure times: anti-His, 3 sec; anti-ubiquitin, 1 h. B. Western blot analysis of ubiquitylated proteins from U2OS cells. Left panel, cultures were subjected to an 8 h treatment with MG132 (10 µM, + lanes) or medium only (− lanes) before harvest. Five percent of each cell lysate (∼ 5.0 × 105 cells) was probed with anti-ubiquitin or antibody to GAPDH (loading control) after electrophoresis (SDS-PAGE, 15%) and Western blot transfer. Right panel, the remaining portion of each corresponding lysate was subjected to immunoprecipitation with an isotype control or anti-ubiquitin monoclonal antibody and probed in Western blot analysis of an SDS-PAGE (8% gel) with anti-ubiquitin. Bottom right panel, SDS-PAGE gel (15%) of isotype control or anti-ubiquitin precipitated material probed with anti-ubiquitin. Free or monoubiquitin is not detectable. A twofold monoubiquitin titration (100–3.1 ng) was loaded to lanes on the same gel to establish the detection range of the antibody. C. In vitro assay measuring cleavage of a phospholipid mimic, PED6, using identical amounts of protein shown in each lane for material immunoprecipitated with anti-ubiquitin or isotype control antibodies. Assay measurements from equivalent amounts of each antibody bound to Protein G beads exposed to lysis buffer only were subtracted as background. Data are expressed as means ± SEM, n = 3.

To determine whether ubiquitylated mammalian proteins activate ExoU, an osteosarcoma cell line, U2OS, was exposed or not exposed to the proteasome inhibitor MG132. Exposure to MG132 shifts the pool of free ubiquitin to ubiquitin conjugated to proteins. Western blot analysis of total cell lysates revealed an increase in the detectable levels of high molecular weight species, consistent with a shift in the accumulation of ubiquitylated proteins upon treatment with MG132 (Fig. 2B). Immunoprecipitated material from isotype controls revealed neither ubiquitylated proteins nor free ubiquitin as expected. In the absence of MG132 few high molecular weight ubiquitin conjugates were detectable. Conjugated protein material increased substantially upon treatment with MG132 (Fig. 2B). Importantly, in immunoprecipitation assays ubiquitylated proteins appeared to be enriched while free ubiquitin was not detectable (Fig. 2B). ExoU activity assays were performed utilizing the bound materials as a source of the activator. Beads from the anti-ubiquitin immunoprecipitated MG132-treated lysates possessed significant activity (Fig. 2C). Minor activity was detectable from MG132 untreated cells (Fig. 2C). Overall, these data support the hypothesis that in addition to free ubiquitin, ubiquitylated proteins from either yeast or mammalian cells serve as cofactors for ExoU.

Various species of ubiquitin but not ubiquitin-like proteins serve as activators

To address the mechanism of activation of ExoU, it is important to determine whether different forms of ubiquitin affect ExoU activity. A PED6 titration was initially conducted in the presence of monoubiquitin or polyubiquitin to quantify the concentration of substrate required to be near saturation. For PED6, this concentration was determined to be 100 µM (data not shown). To determine how well ubiquitin activated ExoU, each species was titrated into the assay and phospholipase A2 activity was measured as an increase in cleaved PED6 (fluorescence) over time (Fig. S2A and B). For most of the ubiquitin species tested, a lag time in ExoU activation was observed in the progression curves. This lag in activation was not dependent on preincubation time and could be reduced or eliminated as the concentration of activator was increased. These two points suggest that the structural ordering of ExoU and the binding to ubiquitin reaches equilibrium fairly rapidly when sufficient amounts of the activator are present.

Progression curves were analysed for the rates of enzymatic activity after steady-state conditions were reached and the concentration of activator (nM) as a function of rate (nmoles PED6 cleaved min−1) was plotted (Fig. 3A). With the exception of monoubiquitin, the enzymatic rate of ExoU in the presence of all ubiquitin species reached saturation at concentrations of activator less than 4 µM, about 2 µM less than the intracellular concentration of ubiquitin (Haas and Bright, 1987).

Figure 3.

Multiple isoforms of ubiquitin activate rExoU. A. Michaelis–Menten plots representing in vitro phospholipase activity measured from titrations of several ubiquitin species. Nanomoles PED6 cleaved were calculated from steady-state rates of RFU accumulation. Dashed lines represent non-linear regression fit of data. The inset graph for each ubiquitin species is the double reciprocal plot with linear regression analysis designated as the dashed line. B. The bar graph shows the affinity of ExoU for different ubiquitin isoforms plotted in terms of 1/Kact and chain length (the mean ± SEM, *P < 0.01).

Non-linear regression analysis of the Michaelis–Menten plots (as represented by dashed lines) fit well with calculated correlation coefficients of 0.89 or greater (Fig. 3A and Table 1) making it possible to derive kinetic constants for each ubiquitin species. As the number of ubiquitin moieties increased for each isoform, the activation constant (Kact) corresponding to the Kd for ExoU binding to the ubiquitin, decreased, which is indicative of tighter binding of rExoU for ubiquitin (Fig. 3B and Table 1). Although rates of maximal catalysis (Vmax) were similar, we noted a disparity in the catalytic efficiency (Vmax/Kact) between the different forms of ubiquitin. Compared with monoubiquitin, activation efficiency of rExoU increases by approximately 10-fold when K48-linked diubiquitin is used as the activator, and 250- to 500-fold when K48-linked tetraubiquitin, linear tetraubiquitin or K63-linked octaubiquitin are used (Fig. 3B and Table 1).

Table 1.  Activation of ExoU by different isoforms of ubiquitin.
  • a. 

    µM, the affinity of rExoU for ubiquitin.

  • b. 

    nmoles of PED6 cleaved min−1.

  • c. 

    nmoles of PED6 cleaved min−1 nmoles ExoU−1µM ubiquitin−1.

  • d. 

    r2 is the correlation coefficient listed for the Michaelis–Menten fit for each ubiquitin isoform.

Monoubiquitin17.18 ± 7.580.13 ±
K48-linked Diubiquitin1.19 ± 0.330.08 ± 0.01400.988
K48-linked Tetraubiquitin0.06 ± 0.020.10 ± 0.011.1 × 1030.916
Linear Tetraubiquitin0.08 ± 0.020.21 ± 0.011.5 × 1030.967
K63-linked Octaubiquitin0.04 ± 0.010.15 ± 0.012.0 × 1030.973

Importantly, ubiquitin-like proteins including ISG15, FAT10, SUMO-1 and NEDD8 do not activate rExoU at the highest concentrations tested (10 µM, Fig. S2C and data not shown). These data suggest that ubiquitin specifically activates ExoU in vitro. ExoU activation, however, is not restricted to a specific ubiquitin chain linkage but occurs with multiple forms. Better activation is observed as the number of ubiquitin moieties increases up to n = 4 (Fig. 3B).

Recombinant ExoU binds several immobilized ubiquitin species

To determine if ExoU is able to bind ubiquitin in vitro, a solid phase binding assay previously established for ExoU binding to bSOD1 was utilized (Schmalzer et al., 2010). Recombinant histidine-tagged ExoU was able to bind both monoubiquitin (Fig. 4A) and K48-linked diubiquitin (data not shown) with similar affinities (Kd ∼ 1.4 nM). The affinity of rExoU for K48-linked polyubiquitin (Kd ∼ 0.4 nM, Fig. 4B) is approximately 3.5-fold greater than monoubiquitin or diubiquitin (Fig. 4C and Table 2). These data suggest that the association of ExoU with ubiquitin increases as the number of ubiquitin moieties increases. Double reciprocal plots (insets for each ubiquitin species) demonstrate that the concentration range for rExoU is sufficient and the linear regression analysis suggests that there is good correlation between recombinant protein added and binding (Fig. 4A and B). Histidine-tagged rPcrV, a protein present at the tip of the type III needle apparatus, does not bind ubiquitin under these conditions. These experiments provide evidence that ExoU associates with ubiquitin and that this association is enhanced with polyubiquitin species (Fig. 4C).

Figure 4.

Recombinant ExoU binds immobilized ubiquitin. A. The binding of rExoU to immobilized monoubiquitin was determined using a solid-phase binding assay (Schmalzer et al., 2010 and Experimental procedures). rExoU binding was detected using monoclonal antibody and a fluorescent horse radish peroxidase conjugated secondary reagent. The data are plotted and analysed by non-linear regression (dashed line). Each point represents three independent experiments. The closed circles (●) represent histidine-tagged rExoU and the closed squares (inline image) represent the negative control protein, histidine-tagged rPcrV. The inset graph is the double reciprocal plot with linear regression analysis designated as the dashed line. B. Solid phase binding assay using immobilized K48-linked polyubiquitin. C. Bar graph shows comparison in affinity (1/Kd) of monoubiquitin versus polyubiquitin chains. Data are expressed as the mean ± SEM, n = 3. *P < 0.01.

Table 2.  Binding constants for the association of ExoU and different ubiquitin isoforms.
Immobilized activatorKdaBmaxbr2c
  • a. 

    nM units.

  • b. 

    relative fluorescence units (RFU).

  • c. 

    r2 is the correlation coefficient listed for the non-linear regression fit for each immobilized ubiquitin isoform.

Monoubiquitin1.45 ± 0.4727 430 ± 3 8440.891
K48-linked Diubiquitin1.41 ± 0.5334 025 ± 5 4260.864
K48-linked Polyubiquitin0.38 ± 0.0815 885 ± 1 0550.959

Monoubiquitin is sufficient for ExoU activity in a genetically engineered E. coli system

A dual expression bacterial system was constructed in which both ExoU and ubiquitin were under the control of tightly repressible promoters (pBAD and T7 polymerase, respectively) in E. coli. Western blotting experiments of bacterial cells demonstrated that ubiquitin, ExoU or a control protein, PcrV, could be induced independently and simultaneously, depending on the medium composition (Fig. 5A and B). A time-course of cell survival after dual induction (derepression) showed a significant drop in viability after 1 h in the ExoU-ubiquitin expressing cells, correlating with the timing of ExoU expression (Fig. 5B and C). PcrV-ubiquitin expressing cells did not suffer any loss in viability; in fact, cell numbers, after a 3 h induction, increased almost sevenfold compared with a nearly 50-fold decrease in survival of ExoU-ubiquitin induced cells (Fig. 5C and D). At the final time point, bacterial cells were harvested by centrifugation, lysed and PED6 substrate added to the lysate. Only ExoU/ubiquitin-induced lysates actively cleaved PED6 in vitro without addition of exogenous enzyme or activator (Fig. 5E). From these experiments, we concluded that in the presence of monoubiquitin and ExoU, the bacterial inner membrane serves as an in vivo substrate for ExoU phospholipase activity.

Figure 5.

Coexpression of ExoU and ubiquitin in E. coli is lethal. A. Western blot analysis of the independent or simultaneous expression of PcrV and ubiquitin within E. coli cells transformed with two expression plasmids. The expression of monoubiquitin is induced by the addition of IPTG and the expression of PcrV is induced by the addition of arabinose to the medium. Each lane is labelled by the number of hours post induction that the sample was harvested for analysis. B. Western blot analysis of the independent or simultaneous expression of ExoU and monoubiquitin within E. coli cells transformed with similar expression plasmids as described in A. C. Cells harbouring inducible ExoU or PcrV and ubiquitin vectors were measured for cell survival after dual expression of both constructs over a 3 h time-course. Error bars are means ± SEM (n = 3). D. Relative survival of each strain at each time point over the induction period in relation to cell number at time zero induction. Error bars are means ± SEM (n = 3), *P < 0.0001. E. Lysates were made using cells harvested from the 3 h time point and one OD600 unit was analysed for PED6 cleavage in vitro. Error bars are for the mean ± SEM (n = 3).

Permeabilization of bacterial membranes by ExoU in the presence of monoubiquitin

We next used a microscopic approach to characterize ExoU phospholipase activity on bacterial membranes. Cell morphology was first studied using a lipophilic styryl dye, FM 4-64, for membrane staining. ExoU and monoubiquitin coexpression resulted in changes in membrane staining patterns from predominantly rod shapes to punctate, irregular spheres, which accumulated with time (Fig. 6A). Differential interference contrast (DIC) microscopic images show what appears to be the formation of coccus-shaped cells and debris after 1 h induction with the eventual breakdown of any discernable cell structure at 90 min. PcrV-ubiquitin induced cells, in contrast, retained a defined rod shape throughout the time-course (Fig. 6A).

Figure 6.

Morphological changes in E. coli cells after induction of ExoU and monoubiquitin. A. Bacterial cells were induced for expression of ExoU or PcrV with ubiquitin for the indicated times. Cells were stained with a lipophilic dye FM4-64, washed, and detected by using a filter for FM4-64 or differential interference contrast (DIC). Arrows indicate morphology changes. B. The coexpression cells were induced for 90 min and stained with nucleic acid stains SYTOX green (membrane impermeable) and permeable Hoechst 33342 (blue) in the presence of FM4-64 (red). Cells were incubated on poly-lysine coated glass coverslips and washed to remove unbound cells before image acquisition. Scale bars in overlay images indicate 2 µm.

Live cell microscopy was subsequently performed on immobilized E. coli cells using three fluorescent reagents: FM 4-64 to stain the membranes, Hoechst stain and SYTOX green to mark DNA. Intact cells exclude SYTOX green, but when the membrane barrier is compromised, the dye flows into the cell, binds to nucleic acids and green fluorescent signals are amplified. The inclusion of SYTOX green in the medium confirmed the permeabilization of the bacterial membrane as an intense green fluorescence can be seen originating near the septum and flowing into induced cells (Fig. 6B). A green haze appears around some cells suggesting that nucleic acids are being released (Fig. 6B).

Phospholipase-mediated membrane destruction is rapid

Further characterization of ExoU activity aimed to examine the kinetics of ExoU-mediated destruction of bacterial membranes. The membrane and DNA staining system described for Fig. 6 was used to observe SYTOX green infiltration (after a 1 h induction period) into ExoU-permeabilized cells over a time-course of 30 min. In agreement with the timing of ExoU expression in previous experiments, significant SYTOX fluorescence signal was detectable after 1 h of induction. Once initial punctate green signals were visible at a pole or septum, approximately 15 min was required to progress to an intense green fluorescence throughout the bacterium and then to loss of nucleic acid and membrane signals in ExoU-induced cells expressing ubiquitin (Fig. 7). The loss of fluorescent signals may be related to a release of turgor pressure and the reduction of intact bacteria to cellular debris as shown by DIC at 90 min (Fig. 6A). Motion visualization of the occurrence displays a sweeping green signal moving through doubly induced cells (ExoU and ubiquitin) as the reporter dye appears to enter through a predominant site of the membrane (Movie S1). Conversely, PcrV and ubiquitin expressing strains retained membrane staining and excluded SYTOX green over the entire time-course (Fig. 7 and Movie S2).

Figure 7.

Time-lapse imaging of the coexpression E. coli strains during induction. Bacterial cells were induced for the expression of ExoU or PcrV with monoubiquitin and stained with SYTOX green, Hoechst 33342 (blue) and FM4-64 (red) in the presence of inducers in a glass-bottom dish. After 1 h post induction, cells were analysed at 30°C by time-lapse microscopy. Images were acquired with 10 steps of a 3 min interval as indicated.


ExoU is the most potent type III-secreted toxin synthesized by P. aeruginosa (Lee et al., 2005), but its mechanism of action remained unclear due to limited homology with other proteins and its potent cytotoxicity in mammalian cells (Sato and Frank, 2004). The use of yeast as a model system revealed a phenotype of vacuolar fragmentation suggesting membrane destruction or remodelling (Sato et al., 2003). These data, combined with the inhibition of cytotoxicity in both mammalian and yeast cells by phospholipase A2 inhibitors, allowed alignment with phospholipases and the identification of a catalytic dyad, S142 and D344 with similarity to the patatin protein family (Phillips et al., 2003; Sato et al., 2003). In vitro phospholipase activity was detectable only in the presence of other cellular materials indicating the requirement for a eukaryotic activator (Sato et al., 2003). Affinity methods for identifying this activator generally failed as ExoU appeared to interact with multiple proteins and the inclusion of detergents or other agents that decrease non-specific protein–protein interactions inhibited enzyme activity (Sato et al., 2005; 2006). It was unclear whether this inhibition was due to the inability of ExoU to associate with phospholipid substrates or activator or both. Proteomic approaches and the use of enzymatic activity as a screen resulted in the discovery of eukaryotic SOD1 as an activator (Sato et al., 2006). Caveats pertaining to SOD1, however, were that it had poor specific activity as a cofactor, saturable kinetics for ExoU could not be obtained in vitro and only particular commercial preparations of the protein displayed activation capabilities (Sato et al., 2006; Benson et al., 2010). The goal of this study was to determine the specific properties of yeast and bovine SOD1 that mediated the activation of ExoU. Our results indicate that ExoU is a unique toxin in that it specifically associates with ubiquitin and/or ubiquitylated proteins to activate the enzyme. The mechanism of activation coupled with type III delivery ensures that eukaryotic cells are specifically and potently targeted and that the bacterium is protected from its own enzyme.

In addition to using ubiquitin as an activator, ExoU is itself ubiquitylated (Stirling et al., 2006). Ubiquitylation of proteins delivered by type III and type IV secretion systems has been shown to play critical roles in effector stability and trafficking (Kubori and Galan, 2003; Schnupf et al., 2006; Angot et al., 2007; Patel et al., 2009). For ExoU, two monoubiquitin molecules are added to K178 predominantly via a K63 linkage (Stirling et al., 2006). Modification appears to have no significant impact on the half-life of the toxin. In terms of intracellular localization, ubiquitylated, catalytically inactive ExoU (ExoU-S142A) as well as ExoUS142A that cannot be ubiquitylated at K178 (K178R) traffic to the plasma membrane suggesting that ubiquitin modification is not involved with plasma membrane localization (Stirling et al., 2006). The observation of cytotoxicity in the prokaryotic dual expression system also suggests that eukaryotic proteins, other than monoubiquitin, are not required for ExoU to traffic to membrane substrates, or compromise membrane integrity.

The discovery of ubiquitin as an activator and the fact that ExoU is modified by ubiquitin in cells suggests the hypothesis that ubiquitylated ExoU may self-activate. In this model, the injection of ExoU would lead to ubiquitlyation at K178, followed by a conformational change of the molecule. This conformational change might be mediated by the intramolecular recognition of attached diubiquitin by another domain within ExoU. Structure–function analyses of ExoU have implicated the importance of C-terminal residues for phospholipase activity (Finck-Barbançon and Frank, 2001; Sato et al., 2003; 2005; 2006; Benson et al., 2010; 2011; Schmalzer et al., 2010). It is also clear from EPR analyses that ExoU N- and C-terminal residues change conformation in the presence of ubiquitin (bSOD1, Benson et al., 2011; data not shown). To account for these observations, we considered whether the C-terminus of ExoU might encode an ubiquitin-binding domain. Using a variety of bioinformatic approaches, no known ubiquitin binding motifs or domains were identified (data not shown) in the C-terminus or other regions of ExoU. Importantly, Stirling et al. showed that ExoU K178R retained full toxicity, implying that ubiquitylation of ExoU is not required for phospholipase activity. Finally, rExoU is produced in bacteria and the in vitro enzyme activity assay used to measure phospholipase activity lacks ATP and other enzymes required for ubiquitylation. Whether intracellular ubiquitylation of ExoU serves to accelerate activation is unclear and will require direct testing of ubiquitin-modified ExoU derivatives.

The activation of ExoU is highly specific to ubiquitin as ubiquitin-like proteins SUMO-1, ISG15, FAT10 and NEDD8 (10 µM) do not activate ExoU (data not shown and Fig. S2C). Although the specificity of activation relates to ubiquitin, different chain lengths, types of linkages or conjugation to other proteins all function to activate ExoU. Longer chains of ubiquitin, however, have a greater ability to activate phospholipase activity in vitro. We postulate that the interaction of ExoU with ubiquitin may involve multiple sites accounting for the apparent high affinities measured in the solid phase binding assays and the absence of an identifiable motif. Polyubiquitin may act as the best scaffold on which ExoU folds to produce an active enzyme. Alternatively, cofactor interaction may serve a bifunctional role in this toxin's activation, facilitating both a global conformational change (Benson et al., 2011) as well as contributing to catalysis. Our kinetic data suggest that a single ubiquitin molecule may not be able to efficiently accomplish both tasks, as high concentrations of monoubiquitin are required to reach saturable kinetics relative to chain-linked counterparts (Fig. 3 and Table S2). These data indicate that the ExoU–ubiquitin interaction may define a novel type of binding or unique motifs.

In summary, we have demonstrated that the P. aeruginosa phospholipase toxin ExoU is activated by several ubiquitin isoforms, as well as by ubiquitylated proteins. This is, to our knowledge, the first report of ubiquitin serving as an activator for a bacterial toxin. The exact role of ubiquitin in the activation process is unknown but is postulated to facilitate a conformational change in ExoU to allow catalysis. Polyubiquitin molecules associate with and activate ExoU with the greatest efficiency in vitro suggesting that either the size of the cofactor or multiple interaction sites within ExoU are important for phospholipase activity. The association of ExoU with ubiquitin apparently involves novel structural contacts, as no recognizable ubiquitin binding motifs were identified within the ExoU sequence. The toxicity and membrane degradation observed in an E. coli dual expression system for ExoU and ubiquitin reinforces the importance and absolute requirement of a eukaryotic cofactor in regulating the activity of this potent type III effector.

Experimental procedures

Bacterial strains and media

Bacterial strains E. coli DH5α and BL21 (DE3) pJY2 (Enzo Life Sciences) JN105-ExoU or PcrV pET15b-ubiquitin were grown in Luria–Bertani (LB) medium at 30°C or 37°C supplemented with appropriate antibiotics (ampicillin 100 µg ml−1; gentamicin 10 µg ml−1; chloramphenicol 30 µg ml−1). Saccharomyces cerevisiae Y258 YJR104C (Open Biosystems) was maintained in standard defined-ura medium and grown as per manufacturer's recommendation (Thermo scientific). U2OS cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum.


Antibodies used were as follows: MAb166 mouse anti-PcrV (Frank et al., 2002), U29F8 mouse anti-ExoU, rabbit anti-SOD1 (RDI-Fitzgerald), mouse anti-ubiquitin (Santa Cruz, sc-271289), mouse anti-His (GE healthcare), goat anti-rabbit HRP (Sigma), goat anti-mouse HRP (Invitrogen) and mouse IgG (Santa Cruz, sc-2025). SuperSignal West Pico chemiluminescent substrate (Thermo Scientific) was used for Western blot detection. SUMO-1, FAT10 and all ubiquitin except K48-linked polyubiquitin (Enzo Life Sciences, BML-UW0670-0100) were purchased from Boston Biochem. Bovine ubiquitin was obtained from Sigma and purified to apparent homogeneity (> 99.9%, Baboshina and Haas, 1996). Recombinant NEDD8 and ISG15 were expressed and purified as active molecules to apparent homogeneity (> 99.9%) as described previously (Bohnsack and Haas, 2003 and Narasimhan et al., 2005). Proteases used for digestion experiments include: endoproteinase Lys-C (Promega), glutamyl endopeptidase (Glu-C; Sigma), proteinase K (Sigma), thrombin (Sigma) and trypsin (Gibco).

Protein purification

Recombinant ExoU was produced and purified as described in Schmalzer et al. 2010, with the exception that cells were lysed by passage through a French pressure cell. Recombinant ySOD1 was produced by an overnight galactose induction (2% final concentration) of cells (final OD600 = 2.5–3.0). Washed cells were suspended in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM β–mercaptoethanol, 1 mM PMSF, protease inhibitors, DNase and RNase and lysed by bead beating. Metal affinity chromatography was used for recombinant protein purification.

Partial proteolysis

For sequential site-specific protease digestions, bSOD1 was first denatured and then alkylated in the dark for 30 min at 37°C with 10 mM iodoacetamide. The solution was desalted and the flowthrough was acetone precipitated. Pellets were washed with 90% acetone, suspended in buffer and digested with Endoproteinase Lys-C. The reaction was stopped with 5 mM PMSF, 1% SDS and acetone precipitated. Precipitated proteins collected by centrifugation were suspended in the appropriate buffer and digested with glutamyl endopeptidase following the same procedure. The reaction was stopped as described, acetone precipitated, and the protein pellet suspended in buffer and digested with thrombin. This reaction was stopped, acetone precipitated and suspended in 10 mM KPO4, pH 6.3. Single digestions followed similar procedures. For Proteinase K digestions, bSOD1 was diluted in buffer before heating to 95°C for 20 min. Aliquots were removed and placed in stop tubes containing 15 mM PMSF in 10 mM KPO4 buffer (controls). Proteinase K was added to the reaction tube and aliquots were removed at their respective time points and placed into the stop solution and kept on ice until the last time point. All solutions were acetone precipitated, washed and suspended in 10 mM KPO4 buffer, pH 6.3 for use in activity assays as described. Supplementary experimental procedures provide additional detail.

Cell culture and immunoprecipitation

U2OS cells were seeded (7.5 × 105) into 10 cm dishes to grow 72 h, 37°C, 5% CO2. At this time (∼ 90% confluency) they were treated with either medium alone or 10 µM MG132 (Sigma) for 8 h. The cells were scraped into PBS, harvested by centrifugation for 5 min, 2000 g and suspended in NP-40 buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA) with protease inhibitors (EDTA-free tablet, Roche) and 5 mM iodoacetamide. Lysates were sonicated three times with 5 s pulses (Branson Sonifier 150) and placed on ice 30 min with occasional vortexing. The supernatant was collected after a 10 min, 16 000 g, 4°C centrifugation step and placed in a new tube with 5 µg antibody for 12 h, 4°C. Protein G dynabeads (50 ml, Invitrogen) were then washed in NP-40 buffer and incubated with the lysate for 1 h at 4°C. The beads were washed three times with 200 ml PBS and suspended in 50 ml 50 mM MOPS, pH 6.3, 50 mM NaCl. Protein bound beads were added directly to the in vitro PED6 assay (30 mM PED6 and 135 nM rExoU) or boiled in SDS loading buffer for Western blot analysis.

In vitro phospholipase activity assay

The phospholipase A2 activity of rExoU was measured using the established ExoU activity assay previously described (Benson et al., 2010). Briefly, the assay conditions were optimized such that the reaction included 50 mM MES (pH 6.3), 750 mM monosodium glutamate (MSG; pH 6.3), 100 µM PED6 and 33.8 nM untagged rExoU in a final volume of 50 µl. The addition of ubiquitin (in 10 mM KPO4, pH 6.3) was required for activation of the phospholipase activity of ExoU and the concentration was dependent on the species used. Background fluorescence was measured from reactions containing all components except ubiquitin. Fluorescence was measured every 5 min for no more than 90 min at an excitation wavelength of 488 nm and emission wavelength of 511 nm (495 nm cut-off filter; Spectramax M5 microplate reader; Molecular Devices). The conversion equation used was y = 63 654x, where y is RFU and x is nmoles PED6 cleaved (Benson et al., 2010). The data were fit to the Michaelis–Menten equation and analysed by non-linear regression using Prism 5.0 (GraphPad Software).

Solid phase binding assay

The interaction of rExoU with several ubiquitin species was analysed using the previously established solid phase binding assay with modifications (Schmalzer et al., 2010) related to the use of ubiquitin isoforms rather than bSOD1. Approximately 250 ng of the appropriate ubiquitin species was immobilized to the surface of a polyvinyl plate in 10 mM sodium carbonate pH 9.6 and 0.5% gelatin was used in the blocking and diluent buffers. The non-specific binding of proteins and antibodies was evaluated and the data plotted as the mean ± standard error of three independent experiments. The data were fit using non-linear regression and binding constants were determined using the one-site binding model in Prism 5.0.

Dual expression studies

Bacterial cells were harvested from plates in LB broth, incubated for 30–45 min at 30°C and adjusted in LB broth with antibiotics with or without inducers to a starting OD600 of 0.25. Aliquots from cultures (30°C) were removed at the indicated time points for plating. Colonies visible after overnight growth (37°C) were counted. After a 3 h induction, cells were harvested by centrifugation and suspended to OD600 of 0.1 in assay buffer with DNase, RNase, protease inhibitors and 0.5 mg ml−1 lysozyme. Further lysis was carried out via sonication and lysates were analysed for phospholipase activity with PED6 as a substrate. For Western blotting, cells were grown and induced as described, with addition of 0.5% glucose medium to control for the addition of arabinose. The load per lane on SDS-polyacrylamide gels was normalized to 2 × 107 cfu/lane.

Live cell imaging of coexpression strains

Bacterial cells (initial OD600 of 0.4) were grown in the inducing medium (LB containing 0.1 mM IPTG, 0.5% arabinose, and appropriate antibiotics) at 30°C for indicated periods. Cells were collected from 800 µl of culture by centrifugation at 4300 g for 2 min and suspended in 100 µl staining solution for 5–10 min. The staining solution consisted of 1.5 µg ml−1 FM4-64 lipophilic styryl dye in Hank's balanced salt solution (HBSS) containing calcium and magnesium in the presence or absence of 75 nM SYTOX green nucleic acid stain (membrane impermeable) and 20 µg ml−1 Hoechst 33342 nucleic acid stain (membrane permeable) as indicated (all reagents from Invitrogen). Stained cells were washed with HBSS twice before mounting on a glass slide with a coverslip or on a poly-lysine-coated glass at the bottom of a culture dish for microscopy. Cells were analysed with a Nikon Eclipse Ti-U inverted microscope with Chroma Sedat filters using a 100 × oil immersion objective lens (Plan Apo VC with NA 1.40, Nikon). NIS-Elements software (Nikon) was used for image acquisition.

For time-lapse imaging, cells were stained for 10 min in 100 µl of the staining solution in the presence of inducers 0.1 mM IPTG and 0.5% arabinose on a poly-lysine-coated glass bottom dish. Unbound cells were removed by washing with HBSS twice. At the 1 h induction time point, bound cells were covered with 200 µl staining solution containing inducers for time-lapse microscopy at 30°C. Images were acquired with 10 steps of a 3 min interval for a total of 30 min.

Statistical analysis

Two-tailed P-values were generated from unpaired t-tests using Prism 5.0 (GraphPad Software).


This work was supported by the National Institutes of Health (AI49577 to D.W.F.), the Center for Infectious Disease Research and the Advancing a Healthier Wisconsin Foundation at the Medical College of Wisconsin.