Pseudomonas aeruginosa infection liberates transmissible, cytotoxic prion amyloids

Abstract

Patients who recover from pneumonia subsequently have elevated rates of death after hospital discharge as a result of secondary organ damage, the causes of which are unknown. We used the bacterium Pseudomonas aeruginosa, a common cause of hospital-acquired pneumonia, as a model for investigating this phenomenon. We show that infection of pulmonary endothelial cells by P. aeruginosa induces production and release of a cytotoxic amyloid molecule with prion characteristics, including resistance to various nucleases and proteases. This cytotoxin was self-propagating, was neutralized by anti-amyloid Abs, and induced death of endothelial cells and neurons. Moreover, the cytotoxin induced edema in isolated lungs. Endothelial cells and isolated lungs were protected from cytotoxin-induced death by stimulation of signal transduction pathways that are linked to prion protein. Analysis of bronchoalveolar lavage fluid collected from human patients with P. aeruginosa pneumonia demonstrated cytotoxic activity, and lavage fluid contained amyloid molecules, including oligomeric τ and Aβ. Demonstration of long-lived cytotoxic agents after Pseudomonas infection may establish a molecular link to the high rates of death as a result of end-organ damage in the months after recovery from pneumonia, and modulation of signal transduction pathways that have been linked to prion protein may provide a mechanism for intervention.—Balczon, R., Morrow, K. A., Zhou, C., Edmonds, B., Alexeyev, M., Pittet, J.-F., Wagener, B. M., Moser, S. A., Leavesley, S., Zha, X., Frank, D. W., Stevens, T. Pseudomonas aeruginosa infection liberates transmissible, cytotoxic prion amyloids.

Pneumonia is a serious pulmonary infection that is responsible for upwards of 50,000 deaths per year in the United States (1). The infection is caused either by bacteria, viruses, or fungi and is generally divided into 2 broad classes: community-acquired pneumonia and hospital-acquired (nosocomial) pneumonia. Although rarely a cause of community-acquired pneumonia, Pseudomonas aeruginosa is one of the most common causes of nosocomial pneumonia in mechanically ventilated, critically ill patients (2–5). Nosocomial infection by P. aeruginosa is associated with high in-hospital mortality rates and extended lengths of hospital stay (6–10). Sequencing of the genome of P. aeruginosa has shown that it encodes various antibiotic resistance factors and drug efflux systems that make antibiotic treatment difficult and that contributes to the high mortality rates associated with infection (11).

During infection, P. aeruginosa uses a type III secretion system to transfer bacterial toxins into the cytoplasm of target cells. Principal among these bacterial toxins are enzymes referred to as ExoS, -T, -U, and -Y. ExoS and ExoT are dual-functioning enzymes with both Rho GTPase and ADP-ribosyltransferase activities that impact cell signaling (12–15), whereas ExoU is a phospholipase A2 that targets host cell membranes, which leads to cell lysis and modulation of signal transduction pathways (13, 16). ExoY is a multiaction nucleotide cyclase (17–20), and production of cyclic nucleotides by ExoY in pulmonary microvascular endothelial cells targets the microtubule-associated protein τ, which leads to loss of cellular microtubules and breakdown of the endothelial barrier (18, 21). Infection of the lungs by P. aeruginosa leads to transfer of the previously described exoenzymes into pulmonary cells, which results in a loss of barrier integrity in the lung, leading to edema, flooding of the alveolar airways, decreased pulmonary function, and, oftentimes, death (22, 23).

It has been established that patients with pneumonia who are successfully treated and who survive the initial infection subsequently have elevated rates of death as a result of secondary end-organ injury in the months after hospital discharge. Several groups have analyzed long-term effects of pneumonia on patient survival and quality of life (24–33). Major findings from these studies have included increased mortality, particularly among elderly patients, with major causes of death that include cardiovascular disease, stroke, renal failure, respiratory insufficiency, and additional infections (32, 33). Two recent studies have also reported not only decreased quality of life but also increased costs of long-term care of patients after pneumonia (34, 35). Clearly, understanding the reasons for long-term end-organ effects after pulmonary infection by P. aeruginosa as well as developing effective treatments to alleviate those conditions have important clinical and economic consequences.

The reasons for long-term elevated rates of death after treatment for pneumonia have never been determined. In this study, we investigated the hypothesis that infection by P. aeruginosa causes production and release of a long-acting host-derived toxin that can lead to cytotoxicity and hyperpermeability, which may cause secondary organ failure in the absence of living bacteria. Support for this hypothesis comes from 2 sources. First, previous work has demonstrated that infection of pulmonary endothelial cells by P. aeruginosa induced long-term effects on endothelial cell proliferation (36). Specifically, infection of cultured pulmonary endothelial cells by P. aeruginosa inhibited growth of treated endothelial cells for up to 1 wk after removal of the bacteria from the cell culture environment by antibiotic treatment. This result suggests either that infection of cells modified them in some way to inhibit their growth or that something was retained in the medium that repressed cell proliferation even after bacteria were killed. Second, transmissible cellular components, such as prions and prion-like molecules, have been implicated in various human diseases, including Creutzfeldt-Jakob disease (37), Alzheimer’s disease (38), Parkinson’s disease (39), and amyotrophic lateral sclerosis (40). In these diseases, transfer of modified proteins between cells has been implicated in the pathogenesis of disease. Production of a modified protein after P. aeruginosa infection of the lung could explain the long-term effects that have been reported to occur in various organs after pneumonia caused by P. aeruginosa, as release of such a substance into the blood could target endothelium throughout the body, leading to organ failure at sites distinct from the original site of infection. Our analyses demonstrate that infection of pulmonary endothelium by P. aeruginosa results in the liberation of a cytotoxic amyloid-like substance from endothelial cells. Production of such a substance could explain the elevated levels of mortality and various types of organ failure that occur after recovery from P. aeruginosa infection.

MATERIALS AND METHODS

Cell culture

Rat primary microvascular endothelial cells (PMVECs) and pulmonary artery endothelial cells (PAECs) were maintained as previously described (41, 42). Mouse primary cortical neurons were isolated as previously described (43).

Bacterial strains

Two strains of P. aeruginosa were used in these studies and both have been previously described (44). For most experiments, we used strain PA103, which expresses both exoenzymes ExoT and ExoU and has a functional type III secretion system for transfer of exoenzymes to infected cells. For some experiments, we used strain ΔPcrV. This strain is incapable of transferring exotoxins to target cells as the PopB/D channel is not formed (45, 46).

Cytotoxicity assay

PMVECs were grown to confluence and then infected with the appropriate strain of bacteria at a multiplicity of infection of 20:1 by using previously described methods (21). Bacteria were diluted in HBSS before addition to cells, and treatment with bacteria was for either 2.5 or 4 h. Culture supernatants were collected, centrifuged at 2000 g for 10 min at room temperature to remove debris, filter sterilized through a 0.22-µm filter to remove bacteria, and then added to either naïve PAECs, PMVECs, or cortical neurons. Filtered supernatant sterility was independently confirmed by testing for bacterial growth (data not shown). Cells were maintained for 21–30 h before being fixed with 2% paraformaldehyde, and then photographed by using a Nikon Eclipse TS100 inverted microscope (Nikon, Tokyo, Japan). For some experiments, uninfected PMVECs were scraped from a plate, suspended in HBSS, homogenized by multiple passes through a 27-g needle, and filter sterilized, then cell homogenate was added to cells and cells were incubated, fixed, and photographed as described. To quantify killing, cells were analyzed for lactate dehydrogenase release by using manufacturer procedures (04744926001; Roche Diagnostics, Mannheim, Germany). To control for potential effects of HBSS, cells were cultured in HBSS alone for the same duration as cells that were treated with supernatants.

For some studies, supernatants obtained from PA103-infected PMVECs were treated with either RNase A (10 U for 1 h at 37°C), DNase I (5 U for 30 min at 37°C), or trypsin (0.05% for 20 min at 37°C), or were boiled (100°C for 20 min) before being added to cells. To inactivate trypsin before adding trypsin-digested supernatants to cells, the supernatant was either boiled for 10 min or fetal bovine serum was added to 2% final concentration. For other experiments, PAECs were pretreated with anti–prion protein (PrP) Ab (1:1000 dilution) for 60 min (ab136919; Abcam, Cambridge, MA, USA). Anti-PrP Ab was dialyzed twice for 4 h each against a ×100 excess of HBSS to remove preservatives before addition to cells. In another set of experiments, supernatants were pretreated with anti-amyloid oligomer Ab (SPC-506D; StressMarq Biosciences, Victoria, BC, Canada) for 2 h at 4°C with shaking, and Ab/antigen complexes were depleted by addition of protein A agarose beads. After centrifugation, the depleted supernatant was dialyzed against 2 changes of HBSS for 4 h each to remove preservatives included in Abs and protein A agarose beads before addition to the cells. To perform immunoblot analysis of supernatants by using anti-amyloid Ab, supernatants were first concentrated at least 10-fold by using Centricon devices (EMD Millipore, Darmstadt, Germany) before electrophoresis. Immunoblot studies were performed by using standard procedures (21, 47). Finally, in some experiments, cytotoxic supernatant was treated with a 10-fold excess of 1,1,1,3,3,3-hexafluor-2-propanol (HFIP; 003409; Oakwood Chemical, Estill, SC, USA) for either 5, 15, or 30 mins. HFIP was removed by lyophilization and residual protein was either resuspended in HBSS and added to cells to assay effects on cytotoxicity or resuspended directly in sample buffer and assayed by immunoblot analysis by using anti-amyloid oligomer Ab.

Thioflavin T assay for amyloid fibril formation

Formation of amyloid fibrils was assessed by using thioflavin T (ThT) and spectroscopic analysis. ThT (2390-54-7; Sigma-Aldrich, St. Louis, MO, USA) salt was dissolved into PBS and filtered by using a 0.2-µm syringe filter to achieve a stock concentration of 25 mM. Before each experiment, ThT stock was diluted in PBS to a working concentration of 0.49 mM. For spectrophotometric assay, 1 ml of diluted ThT was placed in a disposable 1-cm path-length cuvette (Thermo Fisher Scientific, Waltham, MA, USA), and fluorescence emission spectrum of the ThT solution was acquired by using a QuantaMaster 40 spectrofluorimeter (Photon Technology International, Princeton, NJ, USA) with excitation wavelength set at 425 nm. Fluorescence emission then was collected at 450–575 nm at 2-nm increments. An integration time of 0.1 s/wavelength was used and 5 sequential scans were averaged for each reading. After acquiring the baseline emission spectrum, a time-lapse scan was performed by using an excitation wavelength of 425 nm and an emission wavelength of 482 nm, with data acquired every 0.2 s for 60 s. To measure amyloid, scans were stopped at the 20-s time point, and 10 µl sample solution was added (either PA103 or ΔPcrV supernatant). After time-lapse scan, a second emission spectral scan was performed by using settings that were identical to those previously described. For fluorescence emission spectral scans, spectra were plotted for visual inspection. For time-lapse scans, baseline (ThT alone) intensity was calculated as the average of intensities between time points 0 and 19 s, and each time point was then expressed as the ratio of total intensity divided by baseline intensity. Time-lapse intensity ratios were also plotted for visual inspection. Average intensity for time points 21–60 s was also calculated as well as the corresponding ratio of 21–60 s intensity divided by baseline intensity, which was used as an estimate of fibril formation.

Analysis of filtration coefficient in whole rat lungs treated with cytotoxic supernatant

All experimental procedures were performed in accordance with current provisions of the U.S. Animal Welfare Act and were approved by the Institutional Animal Care and Use Committee of the University of South Alabama. Permeability was assessed by filtration coefficient (K_f) in standard isolated lung experiments (48). Male Fischer 344 (F344) rats, ranging from age 10 to 15 wk, were purchased from Harlan Laboratories (Madison, WI, USA). Animals were anesthetized using Nembutal (65 mg/kg body weight). Once a surgical plane was achieved—as defined by the absence of a withdrawal reflex after toe and tail pinch—animals were intubated and ventilated, a sternotomy was performed, and pulmonary artery and left ventricular catheters were placed. Blood was taken by heart puncture from the right ventricle. Heart and lungs were removed en bloc and suspended in a humidified chamber where mechanical ventilation and blood flow was established. Rat lungs were perfused at constant flow (40 ml/min/kg body weight) with buffer (mM: 119.0 NaCl, 4.7 KCl, 1.17 MgSO₄, 1.0 Na₂HPO₄, 1.18 KH₂PO₄, 2.2 NaHCO₃, 5.5 glucose) that contained 4% bovine serum albumin/6% blood in 50 ml total volume and physiologic (2.2 mM) CaCl₂ at pH 7.35 at 38°C. After lungs were perfused for 15 min to reach an equilibrated status, baseline K_f was measured as previously described (48). K_f, the product of specific endothelial permeability and surface area for exchange, is a sensitive measure of lung endothelial permeability when surface area is fully recruited. K_f was calculated as the rate of lung weight gain obtained 13–15 min after a 10-cm H₂O increase in pulmonary venous pressure, normalized per 100 g predicted wet lung weight. After K_f measurement, venous pressure was set back to the previous level, and 1 ml concentrated supernatant—from either PA103- or ΔPcrV-infected PMVECs—was slowly infused into the lung circulation by injection into the inflowing perfusate and circulated for a total of 4 h, during which period K_f was measured again as previously described at 2- and 4-h time points after supernatant injection. In experiments that involved pretreatment with anti-PrP Ab, the initial K_f measurement was made with venous pressure kept at the 10-cm-higher level, and then perfusion was interrupted by stopping the peristaltic pump. For Ab pretreatment studies, 25 µl anti-PrP A dissolved in 1 ml perfusate was quickly infused into the inflowing perfusate, flow was resumed for 10 s to allow the Ab to advance into the lung vasculature, the peristaltic pump was stopped, and both the inflow and outflow were clamped for 5 min to allow Ab to bind to the surface of endothelial cells. After 5 min of Ab incubation in the lung, venous pressure was set to the initial lower limit and the flow of perfusate was resumed and allowed to continue for 25 min, which allowed the Ab to continue to circulate through the pulmonary vasculature. Perfusate that contained any unbound Ab was then washed out by using 50 ml fresh perfusate. Supernatant from PA103-infected PMVECs was then injected and K_f measurements were recorded at 2 and 4 h after addition as reported above. A time control was performed after the same procedures, except that neither Ab, nor culture supernatant were added. Data are reported as means ± se. Two-way ANOVA followed by Tukey’s multiple comparison analysis was used to evaluate the differences between 2 groups. Significance was considered at P < 0.05.

Generation and analysis of PrP depleted cells

To knock out the gene that encodes PrP (PrnP) in PMVECs using CRISPR (49, 50), a gRNA directed against sequence CTGCAAAAAGCGGCCAAAG was inserted into the custom-made vector and cotransfected into PMVECs with plasmids CAS9_GFP (44719; Addgene, Cambridge, MA, USA), pX330-U6-Chimeric_BB-CBh-hSpCas9 (42230; Addgene), pExodus CMV.Trex2 (40210; Addgene), and pEF1aEGFP. Forty-eight hours after transfection, cells were subjected to fluorescence-activated cell sorting, and EGFP-positive cells were plated at 300 and 600 cells/150-mm dish. One hundred ninety-two resulting colonies were picked, expanded in 24-well plates, and screened by PCR with primers rPrnPmultiF (CTGTCCTAAGAGGATGGGAATG), rPrnPmultiR (CAGGAAGATGAGGAAGGAGATG), and rPrnPcrispr1F (TACTGATGTTGGCCTCTGC). The targeted region of PrnP in clones in which a 706-bp PCR band failed to amplify—indicating that a deletion was introduced—was amplified with primers rPrnPmultiF and rPrnPmultiR and cloned into pCR 4-TOPO TA vector (Thermo Fisher Scientific). Twelve clones from each ligation were directly amplified with primers m13SPORTf (CCCAGTCACGACGTTGTAAAACG) and m13SPORTr (AGCGGATAACAATTTCACACAGG) and PCR products were treated with Escherichia coli exonuclease I and shrimp alkaline phosphatase (both New England Biolabs, Ipswich, MA, USA) and sequenced by using BigDye Terminator (v3.1) Cycle Sequencing Kit (Thermo Fisher Scientific). Sequencing products were precipitated with ethanol and subjected to capillary run at Functional Biosciences (Madison, WI, USA). Clones in which all PrnP alleles contained frameshifts were further subjected to Western blotting with Abs against PrP (PA1795; Bosterbio, Pleasanton, CA, USA). Once generated, various PrP^−/− clones were analyzed for sensitivity to supernatant that was collected from P. aeruginosa strain PA103-infected PMVECs by using procedures previously detailed. In addition, wild-type (WT) and PrP^−/− cells were either untreated or pretreated with anti-PrP Ab as previously described to test whether survival pathways were activated by Ab treatment as has been demonstrated in neurons (51, 52). Cells were collected and subjected to immunoblot analyses by using Abs directed against either ERK 1/2 (M5670; Sigma-Aldrich), P-Thr202/Tyr204 ERK 1/2 (SAB43012082; Sigma-Aldrich), Fyn kinase (SAB2108139; Sigma-Aldrich), or P-Tyr530 Fyn kinase (SAB4301504; Sigma-Aldrich). Blots were developed by using chemiluminescence procedures and quantified as previously described (21).

Collection and analysis of human bronchoalveolar lavage fluids

This preliminary clinical study was approved by the institutional review board of the University of Alabama at Birmingham with a waiver of patient consent. Five-milliliter samples of bronchoalveolar lavage (BAL) fluid were collected from patients who were mechanically ventilated. An aliquot was used to diagnose that ventilator-acquired pneumonia was caused by P. aeruginosa (which took 24–48 h). After diagnostic verification, the remainder of each BAL fluid was rapidly frozen and samples were sent to the Stevens laboratory at the University of South Alabama. Upon arrival, samples were thawed and BAL fluids were initially centrifuged at 5000 g for 10 min at 4°C to remove dead cells and particulates and were then sterilized by passage through a 0.22-µm filter. BAL fluids were diluted 1:5 in HBSS and added to PMVECs for overnight culture. Cytotoxicity was then determined by direct observation as previously detailed. BAL fluids were then assayed by immunoblot analysis using previously detailed procedures (21, 47). The protein concentration in each sample was determined by using previously described methods (47), and equal amounts of protein were loaded for each BAL fluid sample. Blots were probed with anti-amyloid Ab as described previously, anti-Aβ Ab (NBP2-13075; Novus Biologicals, Littleton, CO, USA), and 2 anti-τ oligomer Abs, T22 and TOC1. T22 was purchased from EMD Millipore (ABN454), whereas TOC1 Ab was provided by Dr. Nick Kanaan (Michigan State University, East Lansing, MI, USA).

Statistical methods

Two-way ANOVA was used to compare sample groups in various studies. For lung permeability analyses, Tukey’s multiple comparison analysis was used for comparison of groups. Significance in all experiments was considered when P < 0.05.

Study approval

All experiments that involved animals and humans were approved by the appropriate university regulatory panels. All experimental procedures that used rats were performed in accordance with current provisions of the U.S. Animal Welfare Act and were approved by the Institutional Animal Care and Use Committee of the University of South Alabama. Clinical studies that involved collection of BAL fluids were approved by the institutional review board of the University of Alabama at Birmingham with a waiver of patient consent.

RESULTS

Demonstration of an amyloid cytotoxin produced during P. aeruginosa infection

To test whether P. aeruginosa infection caused the liberation of a cytotoxic substance from endothelial cells, an in vitro bioassay was developed that involved infecting rat PMVECs with P. aeruginosa strain PA103, incubating for 4 h, collecting the supernatant, filter sterilizing to remove bacteria, and then applying the filtered supernatant to naive cells. Cells were cultured and observed. Inoculation of PAECs (Fig. 1A), PMVECs, and cortical neurons (Supplemental Fig. 1) with culture supernatant resulted in cytotoxicity and death within 21–24 h. Lactate dehydrogenase release from PAECs that were treated with supernatant collected from PA103-infected cells began between 12–18 h after supernatant addition and increased in a linear manner during the course of treatment (Fig. 1B).

Figure 1.

Infection of pulmonary endothelial cells by P. aeruginosa strain PA103 results in the release of a cytotoxin. A) Bioassay demonstrating cytotoxin. Rat PMVECs were infected with P. aeruginosa strain PA103. After 4 h, supernatant was collected, filtered, and added to naive PAECs. Cells were maintained for 22 h, then fixed and photographed (PA103). Growth in HBSS had no effect (HBSS). Release of the cytotoxic substance depends on infection of cells by bacteria, and the cytotoxic substance cannot be removed by dialysis. PMVECs were inoculated with either P. aeruginosa infectious strain PA103 or noninfectious strain ΔPcrV, and supernatants were collected, filtered, and then added to naive PAECs. Supernatant from ΔPcrV-inoculated PMVECs lacked cytotoxic activity (ΔPcrV). Supernatant from PA103-infected cells was also dialyzed before addition to PAECs, and dialysis did not reduce cytotoxicity (dialyzed). Whole-cell lysate prepared from uninfected PMVECs had no effect (lysate). B) Time course of cell killing. Supernatants collected from treated PAECs were collected and lactate dehydrogenase (LDH) release was measured. LDH release was initially detected at 18 h postaddition of supernatant and increased in a linear manner during the course of the experiment. Controls included cells treated with either HBSS alone or supernatant collected from cells inoculated with ΔPcrV strain of bacteria; n = 4. **P ≤ 0.05; ***P ≤ 0.01.

To verify that the observed effects were not a result of the release of a bacterial metabolite and depended on infection of the PMVEC monolayer by bacteria, we performed 3 experiments. First, supernatant was dialyzed against HBSS to remove any potential low MW metabolites before addition of the supernatant to cells, and, second, supernatant was collected from PMVECs that were inoculated with the ΔPcrV strain of P. aeruginosa. ΔPcrV lacks the functional type III secretion system and is incapable of injecting exoenzymes into host cells. As shown in Fig. 1A, dialyzing supernatant had no effect on the cytotoxic activity of the filtered material. Moreover, supernatant that was collected from PMVECs that had been inoculated with ΔPcrV bacteria lacked cytotoxic activity (Fig. 1A). Finally, to verify that cytotoxin was not a molecule that was already present in cells that was simply liberated as cells died, uninfected PMVECs were homogenized and centrifuged to remove nuclei and unlysed cells, and the homogenate was filter sterilized and applied to PMVECs for 24 h. No cell death was observed (Fig. 1A).

Characterization of the cytotoxin (Supplemental Fig. 2) determined that it was completely resistant to digestion with DNase, RNase, and trypsin (20–30 min treatment). In addition, the cytotoxin was resistant to boiling (20 min) when the supernatant that contained it was collected from cells that had been treated with bacteria for >4 h, although it was present in a form that was susceptible to boiling when collected from cells that were treated with P. aeruginosa for shorter time periods (Supplemental Fig. 2). In addition, ammonium sulfate precipitation determined that the toxin precipitated from the cytotoxic supernatant in the 30–50% ammonium sulfate fraction with no activity detected in other fractions under the assay conditions used (Supplemental Fig. 3).

The characteristics that were identified in the previous experiments are suggestive of an amyloid-type molecule; therefore, ThT spectrofluorometric assays were performed using PA103 and ΔPcrV supernatants. As shown in Fig. 2A, B, supernatant collected from PA103-infected cells contained amyloids, whereas supernatant from ΔPcrV-inoculated cells was devoid of detectable amyloid under the conditions used. Subsequently, immunoblot analysis was performed by using anti-amyloid Ab (53–55). A major reactive band of 110 kDa and a minor band at 75 kDa were detected in supernatant collected from cells that were inoculated with infectious P. aeruginosa strain PA103, whereas no reactive bands were detected in supernatant collected from PMVECs that were inoculated with ΔPcrV strain (Fig. 2C). As amyloid proteins have been shown to be disrupted by treatment with HFIP (56), we performed studies to determine whether HFIP would affect cytotoxic activity. As shown in Fig. 2D, HFIP treatment abolished cytotoxic activity and immunoblot analysis determined that HFIP converted the endothelial-derived cytotoxin to a form that migrated near the gel front on immunoblots (Fig. 2C). To assess whether amyloid molecules could be involved in the cytotoxic response, supernatant was immunodepleted with anti-amyloid Ab and the depleted supernatant was then added to cultured PAECs and PMVECs. Immunodepletion of amyloid proteins partially removed cytotoxic activity from the supernatant (Fig 2E, F). Immunoblot analysis was performed to begin to characterize the nature of cytotoxic amyloids. In agreement with our previous studies (47), oligomeric τ was present in the cytotoxic supernatant (Fig. 2G). In addition, oligomeric Aβ complexes were also present in the cytotoxic supernatant (Fig. 2G).

Figure 2.

The cytotoxin contains amyloid-like substances. A, B) Supernatant from PA103-infected endothelial cells possesses ThT-detectable amyloids. A) ThT emission spectra (Ex: 425 nm; Em: 450–575 nm] were obtained in the absence and presence of supernatant obtained from PA103-infected endothelial cells. B) Supernatant addition increased fluorescence intensity. A time-scan of samples (Ex: 425 nm; Em: 482 nm) illustrates that supernatants obtained from PA103-infected but not ΔPcrV-infected endothelial cells possess detectable amyloids. C) Immunoblot analysis using anti-amyloid Ab. Immunoblot analysis (using A11 anti-amyloid Ab) of supernatant collected from PMVECs that had been inoculated with either P. aeruginosa strain PA103 (lane A) or with noninfectious strain ΔPcrV (lane B). Treatment of cytotoxic supernatant with HFIP for 15 (lane C) or 30 min (lane D) converted the cytotoxic amyloid to a lower-MW form. Values at left represent M_r (kDa). D) Treatment with HFIP abolishes cytotoxic activity. PMVECs were treated with either HBSS (1), supernatant collected from cells that were inoculated with strain PA103 of P. aeruginosa (2), or supernatant that was collected from P. aeruginosa–infected PMVECs and then subsequently treated with HFIP (3). E) Immunodepletion of the toxin. Treatment with anti-amyloid (A11) Ab could deplete the killing activity of the supernatant, whereas treatment of the supernatant with nonimmune rabbit IgG had minimal effect on cytotoxic activity. F) Quantitation of cell killing in cells that were treated with supernatant depleted of amyloid proteins using anti-amyloid Ab. Cells were treated with either complete supernatant collected from PA103-infected PMVECs (black) or with supernatant that was preadsorbed with either nonimmune rabbit IgG (white) or anti-amyloid Ab (gray). Lactate dehydrogenase (LDH) release was measured 26 h after addition of the appropriate supernatant. n = 3, mean ± se; *P ≤ 0.05; **P ≤ 0.01. G) Identification of amyloid species in the cytotoxic supernatant. Supernatant from PA103-infected PMVECs was immunoblotted using Abs specific for oligomeric τ (T22; lane A) and Aβ oligomers (lane B). Values at left represent M_r (kDa).

As self-induction and propagation are traits of many amyloid proteins, we tested whether cytotoxin could self-reproduce. For these experiments, cytotoxic supernatant was added to PMVECs for a brief period (4 h), cells were then rinsed extensively, and fresh HBSS was added for 16 h. This second supernatant was then collected and added to naïve PMVECs. As shown in Fig. 3A, this second supernatant, which was collected from cells that never encountered P. aeruginosa, also contained cytotoxic activity, which demonstrated self-propagating prion-like behavior. Moreover, immunoblot analysis of this second supernatant using anti-amyloid Ab demonstrated that the 110- and 75-kDa reactive proteins were released from cells that were treated with supernatant alone (Fig. 3B).

Figure 3.

Amyloids are regenerated by cells that never encountered bacteria. The cytotoxin exhibits prion-like behavior and is self-propagating. A) Filter-sterilized supernatant obtained from PA103-infected cells was added to cells for 4 h, cells were rinsed 4 times with HBSS, and then fresh HBSS was added for 16 h. This second supernatant (2° supernatant) was then collected and added to PMVECs, and cells were maintained for 20 h. As a negative control, PMVECs were also treated with HBSS for 4 h, then HBSS was removed and replaced with fresh HBSS for 16 h. This HBSS was then collected and added to cells for 20 h (HBSS control). Positive control cells (PA103 1° supernatant) were treated with the initial supernatant. B) Anti-amyloid immunoblot analysis of supernatant that was collected from cells that were only treated with supernatant from cells that had been initially treated with P. aeruginosa (lane 2°). Positive control supernatant from cells that were treated with bacterial strain PA103 is also included (lane 1°). Values at right represent M_r.

Activation of PrP protects against effects of the cytotoxin

To determine whether PrP was involved directly either as the cytotoxin or as the receptor for an unknown amyloid (57) in the supernatant, we performed 2 assays. In the first experiment, PrP was depleted from the supernatant by using anti-PrP Ab and the depleted supernatant was then added to PAECs. In the second assay, PAECs were pretreated with anti-PrP Ab and complete supernatant was then added to cells. As shown in Fig. 4, immunodepletion of PrP from the supernatant had no effect on cytotoxic activity. In contrast, pretreatment of PAECs with anti-PrP Ab protected cells against the cytotoxic activity of the supernatant (Fig. 4).

Figure 4.

Pretreatment of PAECs with anti-PrP Ab protects against the effects of the cytotoxin. Anti-PrP immunodepletion of supernatant collected from PA103-infected PMVECs had no effect on killing activity (immunodepletion). In contrast, when PAECs were pretreated with anti-PrP Ab before addition of supernatant, cells were protected from killing (cell pretreatment). Control cells were treated with supernatant that was collected from PA103-infected PMVECs (PA103).

The previous result has 2 possible interpretations. First, PrP may be the receptor for the cytotoxin, as has been demonstrated for other infectious amyloid proteins (57). Alternatively, pretreatment of endothelial cells with anti-PrP Ab may be activating a survival pathway in cells, as has been shown to occur in neurons (51, 52). To discriminate between these 2 possibilities, PrP^−/− PMVECs were generated by using Crispr/Cas 9 gene editing (49, 50), and knockout cells were then analyzed. Individual clones were selected, and PrP deletion was verified by direct gene sequencing, RT-PCR, and immunoblot analysis (Supplemental Fig. 4). When cytotoxin-containing supernatant was added to PrP ^−/− cells, they died much more quickly than did WT PMVECs, which suggested that a role for PrP in microvascular endothelial cells is to contribute to survival (Fig. 5A). In neurons, it has been shown that anti-PrP Ab pretreatment cross-links PrP surface proteins and initiates a survival pathway via activation of ERK1/2 and Fyn kinases (51, 52). To determine whether ERK1/2 and Fyn were activated in PMVECs by Ab pretreatment, immunoblot analysis was performed using control and PrP Ab-pretreated PMVECs. Pretreatment of wt PMVECs with anti-PrP Ab resulted in increased phosphorylation of both ERK1/2 and Fyn without changing overall levels of those proteins in treated cells, whereas pretreatment of PrP^−/− cells had no effect on phosphorylation status of those proteins (Fig. 5B).

Figure 5.

PrP regulates a survival pathway in PMVECs. A) WT PMVECs (control) and 2 PrP^−/− clones (4D5 and 1A2) were treated for 10 h, a time point that is before induction of cell death in WT PMVECs after supernatant treatment, with supernatant collected from P. aeruginosa–infected PMVECs. B) WT and PrP^−/− PMVECs were analyzed for levels of phosphor-ERK1/2 and phosphor-Fyn. Both untreated control (C) and anti-PrP pretreated (Ab) cells were analyzed. Control PMVECs showed an increase in both ERK1/2 phosphorylation (P = 0.001; n = 7) and Fyn phosphorylation (P = 0.018; n = 6) after Ab treatment, whereas PrP ^−/− cells exhibited no increase. Blots of WT cells were reprobed by using Ab against total ERK1/2 and Fyn to verify that Ab treatment did not affect protein levels.

Activation of PrP protects intact lungs from effects of the cytotoxin

To determine whether similar processes are at work in the intact lung, both the cytotoxicity and anti-PrP Ab protection assays were repeated using isolated rat lungs. In initial studies, supernatant from P. aeruginosa–infected PMVECs was added to isolated rat lungs for 2–4 h using a recirculating pump, and K_f was measured. Treatment of lungs with supernatant led to increased K_f, which was indicative of the loss of endothelial barrier function (Fig. 6). In contrast, addition of supernatant collected from PMVECs that were inoculated with ΔPcrV did not significantly increase K_f of isolated lungs (Fig. 6). To test whether PrP activation could protect intact lungs from cytotoxin, K_f was measured in isolated lungs that were pretreated with anti-PrP Ab before addition of supernatant collected from P. aeruginosa–infected PMVECs. As shown in Fig. 6, activation of survival signals in lungs by PrP Ab cross-linking protected against the effects of cytotoxin-containing supernatant.

Figure 6.

Cytotoxic supernatant collected from P. aeruginosa disrupts barrier integrity in intact lungs in a PrP-dependent manner. To document the transferability of cytotoxicity of supernatant that was collected from strain PA103-infected PMVECs, K_f measurements were made in the isolated rat lung at baseline (BL) and at 2 and 4 h after addition of ΔPcrV (n = 8) or PA103 (n = 9) supernatant. In a third group of rats (n = 4), lung vasculature was incubated with anti-prion protein PrP Ab for 30 min before PA103 supernatant to document the involvement of prion protein in the cytotoxicity. A time control (n = 3) was performed in parallel. At 2 h, K_f increased 2.5-fold in PA103 group, which is significantly different from K_f of BL, ΔPcrV, and time control at 2 and 4 h. At 4 h, K_f increased ∼3.5-fold in PA103 group, which is significantly different from BL, ΔPcrV, time control, and PrP Ab treatment at this time point. Results are shown as means ± se. The respective baseline K_f value for each group was as follows: PA103 infected, 0.11 ± 0.02; ΔPcrV, 0.15 ± 0.02; PA103 + Ab, 0.20 ± 0.03; PA103 without Ab, 0.17 ± 0.04; ΔPcrV without Ab, 0.18 ± 0.03; and time control, 0.19 ± 0.05. **P < 0.01 vs. other groups as indicated above.

Human patients with pneumonia produce amyloid cytotoxin

To determine whether amyloid molecules are produced during pneumonia in humans, BAL fluid was obtained from mechanically ventilated patients with a primary P. aeruginosa lung infection and BAL fluids were then analyzed. As shown in Fig. 7, 5 of 6 patient samples induced cytotoxicity when added to cultured PMVECs. Immunoblot analysis of the 6 patient samples using anti-amyloid Ab determined that the 5 samples that exhibited cytotoxicity also contained amyloid proteins, with major reactive bands that ranged from 25 to 125 kDa (Fig. 7). The sixth sample (patient 4), which lacked cytotoxic activity when added to cultured cells, did not contain any anti-amyloid reactive proteins (Fig. 7).

Figure 7.

BAL fluid obtained from human patients with pneumonia contains cytotoxic activity and amyloid proteins. Six patient BAL samples were added to PMVECs and then analyzed for cytotoxic activity. Images obtained using BAL from patients 1–5 were collected 19 h after addition of BAL fluid, whereas the image obtained using BAL fluid from patient 6 was obtained after only 6 h of treatment. (Note: at 19 h, cells treated with this BAL fluid had died and lifted off the plate into the medium and could not be photographed.) Immunoblot analysis of the 6 patient samples using anti-amyloid Ab (A11) revealed the presence of amyloid species in BAL fluids. To characterize the amyloid proteins present in BAL fluids, immunoblot analyses were repeated using T22, an anti-τ oligomer-specific Ab (T22), TOC1, another anti-τ oligomer-specific Ab (TOC1), and Ab specific for Aβ (Aβ). Values at left of blots indicate MW (kDa).

Additional immunoblot studies were performed to begin to characterize amyloid species generated during human P. aeruginosa infections. Previously, we had demonstrated that τ and Aβ oligomers were released from endothelial cells after infection by P. aeruginosa (Fig. 2) (49). To test whether τ oligomers were released into human BAL fluid during pneumonia, BAL fluid was analyzed by immunoblotting using 2 Abs (T22 and TOC1) that recognize τ oligomers exclusively. As shown in Fig. 7, both Abs reacted with proteins present in BAL fluids that exhibited cytotoxic activity, with major bands detected near 25 and 50 kDa. Multiple minor M_r species were also recognized in the 5 BAL samples that exhibited cytotoxic activity. τ Abs failed to recognize any proteins in the sample that lacked cytotoxicity (patient 4). Subsequently, the same samples were probed with Ab that was specific for Aβ amyloid protein, and 50- and 25-kDa proteins were recognized in the 5 BAL samples that exhibited cytotoxic activity, whereas the sample that did not demonstrate cytotoxicity was negative by immunoblotting with Aβ Ab (Fig. 7).

DISCUSSION

The results of the studies reported here demonstrate that: 1) infection of pulmonary endothelial cells by the bacterium P. aeruginosa results in production and liberation of a cytotoxin from the infected cells; 2) the cytotoxin can damage endothelial and nonendothelial cells in culture as well as in intact lungs; 3) the cytotoxin is amyloid in nature and exhibits characteristics of a prion-like agent; 4) endothelium can be rendered resistant to the effects of the cytotoxin by activation of cell-surface PrP; 5) human patients that develop pneumonia after P. aeruginosa infection produce various cytotoxic amyloids, including oligomeric τ and Aβ; and 6) a known infectious agent, P. aeruginosa, can be the initiator of a self-replicating prion-like molecule.

It is well established that patients who recover from pneumonia have elevated rates of death as a result of secondary end-organ injury in the months after hospital discharge (24–33). The widespread nature of this end-organ effect was highlighted by Corrales-Medina et al. (33), who demonstrated that pneumonia is as significant a risk factor for subsequent cardiovascular death as such well-established risk factors as diabetes, hypertension, and smoking. Reasons for susceptibility to secondary organ damage in these patients have never been defined. In this study, we investigated the hypothesis that pulmonary infection leads to production of an element that persists after the pneumonia-causing infectious agent has been cleared. The studies demonstrate that infection of pulmonary endothelial cells by the bacterium P. aeruginosa causes production and liberation of a cytotoxic agent from endothelial cells that has the capacity to damage naïve, unexposed endothelial cells and neurons. Liberation of such an agent during infection may explain 2 observations that are related to P. aeruginosa pneumonia. First, this cytotoxic agent could be transported in the blood to distant regions of the body where it may damage endothelium and, perhaps, the parenchyma of different organs that are far from the sight of the initial insult. Second, the presence of this long-lived cytotoxic agent in the blood may explain why patients succumb from secondary organ failure after hospital discharge even after the infectious pathogen has been treated successfully by antibiotics (32, 33). We are in the process of testing these possibilities.

Several additional questions remain to be addressed. First, identification of the infectious particle has not been completely elucidated. The studies reported here demonstrate that it is amyloid in nature. In this and a previous study, we demonstrated that certain strains of P. aeruginosa can lead to release of oligomeric τ species from infected endothelial cells and, furthermore, that relatively high MW τ species display cytotoxicity (47). Oligomeric τ has been implicated as a prion in Alzheimer’s disease (38), and its production during certain pneumonias may be critical for long-term outcomes; however, attempts to immunodeplete τ from cytotoxic supernatants have failed to abrogate the cytotoxicity (47), which suggests either that additional factors are present or that available Abs are not sufficient to neutralize the injurious species (see ref. 47 for discussion). Our finding that both oligomeric τ and Aβ are present in cytotoxic supernatant and BAL samples supports the possibility that multiple amyloid cytotoxins are liberated during infection (Figs. 2 and 7). Unfortunately, the resistance of these amyloid materials to trypsin degradation has rendered total characterization of the cytotoxins futile by current mass spectroscopy approaches. Additional biochemical studies will be required to establish the detailed identity of cytotoxins.

A second question that must be addressed is whether the phenomenon that we have described—the liberation of amyloid cytotoxins during pneumonia—is unique to Pseudomonas infections or whether it is an event that occurs during infection by other agents. It is worth pointing out that other bacteria share exoenzymes that have characteristics similar to those of exoenzymes produced by P. aeruginosa. For example, Bacillus anthracis and Bordetella pertussis both produce exoenzymes similar to P. aeruginosa ExoY (44), and it is possible that these bacteria may mimic the activity of P. aeruginosa and produce amyloid cytotoxins during infection. Likewise, additional bacterial species produce proteins that are homologous to P. aeruginosa ExoU (58). Studies are underway to address whether other bacterial agents that are responsible for pneumonia can also produce cytotoxic amyloid molecules.

An additional question that should be addressed is whether the cytotoxic activity that is described in these in vitro studies is generated in vivo in patients with bacterial infections. For our hypothetical model that explains the long-term deleterious effects of pneumonia to be correct, it is essential that amyloid agents be generated during human infections. In support of our model, Kumar et al. (59) recently demonstrated that amyloid is released during bacterial infections of the brain in animal models of Alzheimer’s disease, which supports the possibility that infection may be a link to the generation of amyloid molecules during dementia onset. Our observations also show that amyloids are released during bacterial infections in human patients, although in an entirely different organ system. Our results further demonstrate the cytotoxic nature of pulmonary-derived amyloids. Our studies of 6 patient samples (Fig. 7) indicate that in 5 of 6 tested, cytotoxic amyloids were produced. Moreover, we clearly demonstrated that oligomeric τ and Aβ are 2 amyloid species that are present in these patient samples. That oligomeric species are present in samples that exhibit cytotoxicity while being absent from the sample that did not harbor cytotoxic activity is of importance; however, we recognize that the sample size is small and further studies will be necessary to verify this observation.

That pretreatment of endothelial cells with anti-PrP Abs protected against the effects of the amyloid substance may have important clinical implications. Specifically, either direct targeting of the prion protein on the cell surface or the intracellular signal transduction pathways activated after PrP stimulation may provide a mechanism for intervention to protect patients from long-term effects of pneumonia. This could be achieved either by identification of ligands that activate the PrP protein or by isolation of compounds that activate either ERK1/2 or Fyn signaling in endothelial cells. Proof of concept for this strategy is provided in our whole lung experiments (Fig. 6), which demonstrated that PrP Ab pretreatment of endothelial cells protected against the increase in permeability (K_f) that was observed in untreated lungs when cytotoxic supernatant was added. The possibility that PrP generates survival signals for cells was initially proposed for studies in neurons (51, 52), and the studies reported here suggest a similar role for PrP in endothelium.

In summary, the main result of our studies is the demonstration that an infectious agent, such as P. aeruginosa, causes production of cytotoxic amyloid prions. Prions have been implicated as transmissible infectious agents (37) in several rare human and animal (scrapie, mad cow) diseases (1, 60–62), whereas prion-like molecules, such as oligomeric τ, Aβ, and other amyloids are suspected of playing roles in the progression of additional human diseases (38–40). However, with the exception of instances that involve the consumption of contaminated material or inoculation with tainted human blood products (61, 62), the initiating agent that leads to the generation of prion-like substances, such as aggregated τ in tauopathies, has never been established. Here, we demonstrate that a common infectious bacterium is a trigger for the formation of a transmissible cytotoxic amyloid agent. Moreover, we further demonstrate that it is the exoenzymes injected into target cells by P. aeruginosa whose activities are critical for this process, as the cytotoxins was not produced by inoculation of PMVECs by the ΔPcrV strain, which lacks a functional type III secretion system because of its inability to properly assemble the PopB/D pore (45, 46). To address the long-term consequences of bacterial pneumonia in humans, such as end-stage organ failures that are characteristic of these infections, it is essential to understand the biochemical functions of each of the exoenzymes and to determine how they modify cellular behavior to lead to the production of an infectious amyloid.

Glossary

BAL

bronchoalveolar lavage

HFIP

1,1,1,3,3,3-hexafluoro-2-propanol

K_f

filtration coefficient

PAEC

pulmonary artery endothelial cell

PMVEC

pulmonary microvascular endothelial cell

PrP

prion protein

ThT

thioflavin T

WT

wild-type

AUTHOR CONTRIBUTIONS

R. Balczon, K. A. Morrow, C. Zhou, M. Alexeyev, S. Leavesley, X. Zha, and T. Stevens designed and performed experiments; B. Edmonds performed experiments; J.-F. Pettit, B. M. Wagener, S. A. Moser, X. Zha, and D. W. Frank provided patient materials and cell lines; R. Balczon, K. A. Morrow, C. Zhou, M. Alexeyev, and T. Stevens analyzed data and wrote the paper; and all authors provided editorial input to the final manuscript.