Cystatin C regulates the cytotoxicity of infection‐induced endothelial‐derived β‐amyloid

Infection of pulmonary endothelial cells by Pseudomonas aeruginosa induces the production and release of cytotoxic oligomeric tau and Aβ from the endothelial cells. Cystatin C, which is also secreted by the endothelial cells, can suppress the cytotoxic activity of Aβ. Whether cystatin C directly impacts oligomeric tau remains to be determined. Image produced with BioRender.

Infection of rat pulmonary microvascular endothelial cells with the bacterium Pseudomonas aeruginosa induces the production and release of cytotoxic oligomeric tau and beta amyloid (Ab). Here, we characterized these cytotoxic amyloids. Cytotoxic behavior and oligomeric tau were partially resistant to digestion with proteinase K, but cytotoxicity was abolished by various denaturants including phenol, diethylpyrocarbonate (DEPC), and 1,1,1,3,3,3-hexafluoro-2-isopropanol (HFIP). Ultracentrifugation for 8 h at 150 000 g was required to remove cytotoxic activity from the supernatant. Ultracentrifugation, DEPC treatment, and immunodepletion using antibodies against Ab also demonstrated that cytoprotective protein(s) are released from endothelial cells during P. aeruginosa infection. Mass spectrometry of endothelial cell culture media following P. aeruginosa infection allowed identification of multiple potential secreted modulators of Ab, including cystatin C, gelsolin, and ApoJ/clusterin. Immunodepletion, co-immunoprecipitation, and ultracentrifugation determined that the cytoprotective factor released during infection of endothelial cells by P. aeruginosa is cystatin C, which appears to be in a complex with Ab. Cytoprotective cystatin C may provide a novel therapeutic avenue for protection against the long-term consequences of infection with P. aeruginosa.
Pneumonia is the second most common infection in hospitals and the most common infection in ICU units [1]. Hospital-acquired, or nosocomial, pneumonia causes significant morbidity and is the leading cause of death due to nosocomial infections [2,3]. In addition to these immediate consequences, patients who survive the initial infection and are subsequently discharged from the hospital have elevated death rates in the first year following release from the hospital due to some form of end-stage organ failure, including cardiovascular, renal, Abbreviations ApoJ, apolipoprotein J; Ab, beta amyloid protein; DEPC, diethylpyrocarbonate; HBSS, Hank's balanced salt solution; HFIP, 1,1,1,3,3,3hexafluoro-2-isopropanol; MOI, multiplicity of infection; PA, Pseudomonas aeruginosa; PMSF, phenylmethylsulfonylfluoride; PMVEC, pulmonary microvascular endothelial cells; PrP, prion protein; SEM, standard error of the mean. and pulmonary declines, or stroke [4,5]. Moreover, surviving patients also exhibit decreased cognitive and mental activity [6][7][8][9]. The reasons for these long-term consequences of nosocomial pneumonia have never been defined.
Pseudomonas aeruginosa is one of the major causes of nosocomial pneumonia [10][11][12]. P. aeruginosa is an opportunistic organism that uses a type III secretion system to transfer various exoenzymes into the cytoplasm of lung cells, where those enzymes ultimately target various membrane phospholipids [13], the actin cytoskeleton [14,15], and the cytoskeletal protein tau [16][17][18]. These events lead to endothelial barrier breakdown in the lung resulting in exudative edema, impaired gas exchange, restrictive physiology, and, oftentimes, death.
Recent studies have provided a potential mechanistic explanation for the long-term consequences for patients who survive nosocomial pneumonia caused by organisms like P. aeruginosa. Specifically, it has been demonstrated that infection of pulmonary endothelial cells in culture, in animal models, and in human patients, leads to the production and release of oligomeric tau and Ab [18,19]. These cytotoxic amyloids exhibit characteristics of self-replicating prion molecules [19], and their release into the blood following infection may provide an explanation for the elevated death rates and memory defects in patients who were successfully treated with antibiotics during the initial hospital stay. According to this scheme, the long-lived, self-replicating Ab and tau would be capable of being transported throughout the body where they could subsequently re-infect the surviving patient's cells leading to organ failure sometime following hospital discharge. Support for this possibility was shown in studies in which cell culture supernatants generated following infection of cultured endothelial cells, as well as cerebrospinal fluid from pneumonia patients, were able to disrupt long-term potentiation in rat brain slices [20] and memory in treated animals [21]. Experiments in progress are further investigating the possibility that infection-induced lung-derived Ab and oligomeric tau are transported throughout the circulation leading to poor patient outcomes.
The studies in this paper were designed to allow biochemical characterization of the toxic Ab and oligomeric tau produced following infection of endothelial cells by P. aeruginosa. As will be shown, the cytotoxic amyloids produced following infection are susceptible to various protein denaturing treatments, although pulmonary endothelial cell generated oligomeric tau was relatively resistant to degradation by proteinase K. Moreover, evidence will be presented demonstrating that a cytoprotective regulator of Ab, cystatin C, is also produced during these infections. Regulation of cystatin C may provide a potential clinical avenue for treating infections caused by P. aeruginosa.

Results
Infection of cultured rat pulmonary microvascular endothelial cells (PMVECs) with P. aeruginosa strain PA103 induces the production and release of cytotoxic Ab and oligomeric tau which exhibit prion characteristics [18,19]. The basic assay used for analysis of cytotoxic Ab and tau involves incubating endothelial cells with an infectious strain of P. aeruginosa in Hanks Balanced Salts Solution (HBSS), collecting the supernatant 4-5 h after infection, filter sterilizing to remove bacteria, adding the sterile supernatant to confluent cultures of na€ ıve PMVECs and then assessing cell killing 21-24 h later ( Fig. 1). Cell killing then is quantified directly by use of IMAGEJ software (National Institues of Health, Bethesda, MD, USA) to determine percent area of the culture dish devoid of cells after supernatant treatment (Fig. 1). Supernatant generated following inoculation of PMVECs with DPcrV, a strain that lacks a functional type III secretion system, does not contain secreted cytotoxic Ab and oligomeric tau whereas supernatant produced using strain PA103 does contain the cytotoxic forms ( Fig. 1) [19][20][21].
This basic assay was then used to assess the stability of Ab and oligomeric tau cytotoxins following various treatments. For these experiments, supernatant generated following infection of PMVECs with PA103 bacteria was treated with a reagent and then the treated supernatant was added to na€ ıve PMVECs. The amount of cell killing in the treated supernatant was compared to the level of cytotoxicity in untreated PA103 supernatant. Specific supernatant treatments included Proteinase K, HFIP, DEPC, and phenol. Previously it had been shown that cytotoxic amyloids generated during infection of PMVECs by P. aeruginosa were resistant to digestion by trypsin for 30 min and were sensitive to treatment with HFIP when HFIP was added at a 109 excess of HFIP to supernatant [19]. These responses were investigated in more detail in initial studies.
To assess susceptibility to protease, the broad specificity protease proteinase K was substituted for trypsin and the duration of treatment lasted from 1 to 30 h. As shown in Fig. 2, cytotoxic activity of treated supernatants decreased in a linear fashion until about 4-h treatment duration, after which time only a very slight decrease in cytotoxicity could be detected in the treated samples. To investigate the nature of the protease-resistant cytotoxic activity, immunoblot analysis was performed to establish whether oligomeric tau and Ab were resistant to proteinase K treatment. As shown in Fig. 2, Ab was nondetectable after 1 h of proteinase K treatment. However, oligomeric tau could still be detected after 30 h protease treatment.
Previous studies assaying HFIP inactivation of cytotoxic Ab and oligomeric tau generated by PMVECs used a ratio of 10 parts HFIP to 1 part supernatant [19]. Inactivation of endothelial-derived oligomeric tau and beta amyloid was investigated in a more rigorous manner by performing a dose response using different ratios of HFIP to supernatant. As shown in Fig. 3A, cell killing decreased with treatments of supernatant as low as 0.5 : 1 of HFIP : supernatant, and reached maximal decrease at 1 : 1. No additional decrease in killing was detected at higher ratios of HFIP to supernatant. Previously, we had demonstrated that treatment with a 10-fold excess of HFIP to PA103 supernatant resulted in loss amyloid epitopes [19]. To assess the effects of HFIP on the structure of endothelial-derived toxic amyloids at lower ratios of HFIP to supernatant, immunoblot analysis was performed using A11 anti-amyloid antibody, which we have shown previously recognizes both endothelial-derived Ab and oligomeric tau [20,21]. As shown in Fig. 3B, amyloid characteristics of endothelial-derived Ab and oligomeric tau were nearly completely abolished at treatments as low as 1 part HFIP to 1 part supernatant. Subsequently, the susceptibility of oligomeric tau and Ab to additional protein denaturants was assessed. Specifically, cell extracts were treated with either phenol or DEPC, and then cytotoxic behavior of the residual supernatant was assessed. As shown in Fig. 3B, phenol treatment abolished cytotoxic activity. Immunoblot analysis using A11 anti-amyloid antibody demonstrated that phenol treatment nearly completely disrupted the amyloid structure of oligomeric tau and Ab (Fig. 3B). However, somewhat different results were noted following treatment with DEPC. As shown in Fig. 3C, an increase in cytotoxic activity was noted after treatment of supernatants with 1 mM DEPC. Treatment with higher concentrations of DEPC (10 and 20 mM) completely abolished cytotoxic behavior. Immunoblot analysis using A11 anti-oligomer antibody demonstrated that amyloid characteristics of Ab and oligomeric tau were retained after DEPC treatment. As DEPC disrupts protein function by modification of amino acid side chains, these results demonstrate that both amyloid structure and amino acid side chain characteristics are critical for cytotoxic activity of endothelial-derived Ab and oligomeric tau.
Studies then were performed to investigate whether cytotoxic Ab and oligomeric tau were large enough to be pelleted by ultracentrifugation. For these investigations, supernatant obtained from P. aeruginosa-infected PMVECs was centrifuged at 150 000 g and aliquots were collected at 1, 2, 4, 8, 16, and 24 of centrifugation. The individual supernatants then were applied to cultured PMVECs, and cytotoxicity was assessed. As shown in Fig. 4A, there was an initial increase in cell killing followed by a rapid decrease in cytotoxic activity which reached maximal depletion by 8 h of centrifugation. Immunoblot analyses of the residual fractions following centrifugation was also performed using the A11 anti-amyloid antibody. As shown in Fig. 4B, a significant portion of the A11 reactive Ab and oligomeric tau were removed during the first hour of centrifugation at 150 000 g, with additional residual amounts being gradually lost as the length of centrifugation increased.
The previous experiments suggested that cytoprotective proteins are released from endothelial cells during infection by P. aeruginosa strain PA103, and that these cytoprotective proteins are removed during the first hour of ultracentrifugation at 150 000 g (Fig. 4) and by treatment with 1 mM DEPC (Fig. 3). To investigate whether Ab or oligomeric tau were involved as cytoprotective agents, immunodepletion experiments were performed. Specifically, PA103 supernatants were immunodepleted using either anti-Ab (MOAB) or antioligomeric tau (T22). The depleted supernatants were then added to PMVECs and cell killing was assessed. As shown in both occurred about 10 h earlier in the Ab depleted supernatant than in untreated control PA103 supernatant (Fig. 5). The rapid time of onset of cell killing was specific for depletion of Ab as it was not observed following depletion of oligomeric tau. Recovery of Ab from agarose beads and addition of the immune-isolated Ab to Ab immunodepleted supernatant reconstituted a relative cytoprotection of the PA103 supernatant, with no cytotoxicity apparent in cultures at 10 h following addition of the eluted Ab (Fig. 6). Previous published studies have identified an entire cohort of interacting proteins that modulate Ab toxicity [22]. Mass spectrometry of PA103 supernatant was performed to determine whether any of these reported modulators are released from endothelial cells during infection processes as co-immunoprecipitation of one or more of these proteins may explain why depletion of Ab led to increased cytotoxic activity. As shown in Fig. 7A, mass spectrometry analyses identified three known negative regulators of Ab action, including Apolipoprotein J/clusterin [23][24][25][26][27], cystatin C [28,29], and gelsolin [30,31]. Subsequently, each protein was individually immunodepleted from PA103 supernatants and then the depleted supernatants were added to PMVECs. As shown in Fig. 7B, depletion of neither gelsolin nor Apolipoprotein J/clusterin had any effect on the cytotoxic activity of PA103 supernatant. In contrast, depletion of cystatin C converted PA103 supernatant to a form in which cell killing occurred as early as 7 h after addition of depleted supernatant to PMVECs (Fig. 7B). Direct quantitation of cell killing at the 7 h time point confirmed that depletion of cystatin C increased the cytotoxic activity of supernatant (Fig. 7D). Elution of cystatin C from beads and addition of the eluate to cystatin-depleted supernatant was able to suppress the enhanced cytotoxic activity of immunodepleted PA103 supernatant (Fig. 7C,D).
To establish whether cystatin C and Ab were complexed, co-immunoprecipitation analyses were performed. For this, cystatin C was immunodepleted from PA103 supernatant and from culture supernatant derived from uninfected control cells, and then the beads were probed for co-immunoprecipitation using antibody against Ab. As shown in Fig. 8A, oligomeric Ab species in the range of 10-20 kDa were detected in association with cystatin C in supernatants derived from PA103 infected cells, while lower molecular weight forms of Ab were complexed to cystatin C in supernatants obtained from uninfected cells. Finally, to assess formation of the cystatin C/Ab complex following infection, PMVECS were treated with PA103 strain of P. aeruginosa for either 0, 2, or 4 h, and the individual supernatants were collected. PA103 supernatants were subjected to ultracentrifugation for 1, 2, and 4 h, and the pellets were analyzed for Ab and cystatin C by immunoblotting. As shown in Fig. 8B, neither cystatin C nor Ab were pelleted from supernatants derived from cells in which PA103 was added to PMVECs and then immediately collected for analyses (T = 0) demonstrating that any cystatin C/Ab complexes that may have been present in cell culture medium was successfully removed by rinsing prior to addition of bacteria to cells. Moreover, neither cystatin C nor Ab was present in pellets derived from supernatants collected from cells that were treated with PA103 for 2 h indicating either that Ab and/or cystatin C had not yet been released from PA103 infected cells or that complexes which formed between the two proteins were not large enough to be pelleted (Fig. 8B). However, when supernatants derived from cells that had been treated with PA103 for 4 h were analyzed, both cystatin C and Ab pelleted during the first hour of centrifugation at 150 000 g for 1 h. The size of the pelleted cystatin C was 26 kDa, which is the reported size of cystatin dimers [32][33][34].

Discussion
Various assays were performed to determine the biochemical stability of Ab and oligomeric tau generated by rat PMVECs following infection with P. aeruginosa. Central to the characterization studies was a rapid cell culture cytotoxicity assay in which cell killing could be assessed in 21-24 h. Major findings are that oligomeric tau cytotoxins produced during the infection process are relatively resistant to protease degradation using proteinase K, but both oligomeric tau and Ab are susceptible to neutralization by HFIP, phenol, and DEPC. In addition, ultracentrifugation established that cytotoxic Ab and oligomeric tau were large enough to be pelleted. Immunodepletion, DEPC treatment, and centrifugation provided evidence for the presence of a cytoprotective protein in the PA103 supernatant. This cytoprotective protein was identified as cystatin C, and additional studies determined that cystatin C complexes with Ab. Finally, cytotoxicity was still evident in supernatants depleted of Ab indicating a critical role for oligomeric tau as an endothelial-derived cytotoxin.

Aβ-Depleted Aβ-Add Back
In a previous study [19], we reported that cytotoxic Ab and oligomeric tau, which exhibited prion-like properties, are liberated from rat PMVECs during infection by P. aeruginosa bacteria. Previous preliminary characterization of the endothelial-derived forms of Ab and tau demonstrated that they were resistant to digestion by trypsin and were not denatured by boiling. In addition, the cytotoxins demonstrated prion-like behavior as defined by transmissibility during passage between cells in the absence of bacteria. Experiments reported here expanded these previous observations.
Endothelial-derived cytotoxic Ab and oligomeric tau were sensitive to various protein denaturants including HFIP, phenol, and DEPC (Figs 2 and 3). However, the mechanisms of inactivation appear to be different.
Whereas both phenol and HFIP caused gross rearrangements of oligomeric tau and Ab as evidenced by loss of immunoreactivity when reacted with A11 antioligomer antibody, DEPC did not reduce reactivity toward the A11 antibody. Rather than cause denaturation of proteins, DEPC modifies amino acid side chains. This result indicates that both amyloid structure and side chains are important for the cytotoxic behavior of Ab and oligomeric tau. The importance of amino acid side chains and their modifications for assembly of oligomeric tau complexes has recently been demonstrated [35], and these studies further support a role for the importance of amino acid side chains for oligomeric tau toxicity.
Ab was rapidly digested with proteinase K (Fig. 2) as demonstrated by immunoblot analysis. In contrast, the endothelial oligomeric tau was resistant to digestion by proteinase K with degradation and loss of cytotoxic behavior not being apparent even after incubation for 30 h with 100 µgÁmL À1 proteinase K. This contrasts to the well-characterized PrP sc which can be completely proteolyzed by proteinase K between 2 and 4 h of treatment [36]. The reasons for this difference are not clear at present. Regardless, these results demonstrate that oligomeric tau is cytotoxic in the absence of Ab under the experimental conditions used. The studies reported here present evidence of an endothelial-derived cytoprotective molecule being secreted during P. aeruginosa infection. This conclusion is based on ultracentrifugation studies, in which cytotoxic activity increased after one hour of centrifugation at 150 000 g (Fig. 4), by DEPC treatment, in which cell supernatants treated with low concentrations of DEPC exhibited increased cytotoxicity (Fig. 3), as well as immunodepletion studies (Fig. 5). Collectively, these data indicate that cytoprotective protein(s) are released from endothelial cells. The data presented here indicate that cystatin C is the protein responsible for modifying the cytotoxic behavior of endothelial-derived beta amyloid. At present, it is not possible to state whether cystatin binds to all forms of beta amyloid or only a subset.
Although Ab is classically considered to be a deleterious molecule due to its link to Alzheimer's disease and other neurologic conditions, recent evidence has indicated that Ab may have additional functions. For example, it has been shown that Ab is part of the innate immune response and can act as an anti-microbial peptide [37,38]. Our results demonstrate that different forms of Ab associate with cystatin C, dependent upon whether the Ab was secreted from uninfected cells or from PMVECs that had been infected by P. aeruginosa (Fig. 8), with cystatin C binding to lower molecular weight forms secreted from uninfected cells while being associated with higher molecular weight complexes released from PA103 infected cells. A potential explanation for this varied binding response is provided by our recent studies which demonstrated that infection regulates the type of amyloid secreted from PMVECs [39]. Infection by strains of P. aeruginosa that lack a functional type III secretion system or produce catalytically inactive effector exoenzymes leads to the release of noncytotoxic forms of Ab whereas infection with strains that contain functional type III secretion system and active effector enzymes leads to the conversion of Ab into cytotoxic forms. The behavior of the type III secretion system and exoenzymes may also help to explain why pelleting of cystatin C and Ab was only detected when supernatant collected following 4 h infection was 15-

MOAb IgG MOAb
Infected Uninfected analyzed, and not when supernatant collected after 2 h of infection was subjected to ultracentrifugation (Fig. 8). The absence of complexes large enough to be pelleted at early time points may be a consequence of the amount of time required for the infected PMVECs to respond to the infection [40,41]. Further studies will be required to firmly establish the varied cystatin C/ Ab interactions that occur following P. aeruginosa infection.

15
The cytotoxic forms of Ab and oligomeric tau present in PA103 supernatant were large enough to be pelleted during ultracentrifugation, and cytotoxic activity was lost after 8 h centrifugation at 150 000 g. This is significantly different than the reported sedimentation rate for the PrP sc particle [42] indicating that the endothelial-derived cytotoxic prions are smaller and/or less dense than the PrP sc agent. The cytoprotective factor, however, is much larger than the cytotoxic forms and sedimented after only 1 h of centrifugation at 150 000 g. This finding suggests that the cytoprotective form of Ab is either a large oligomeric form of Ab, that Ab is complexed with other cellular components, or both. As shown in Fig. 8, at least portion of the Ab released following P. aeruginosa infection of PMVECs is complexed to cystatin C. Previous investigations of cystatin C in Alzheimer's disease have demonstrated that it protects neuronal cells from Ab toxicity [29]. Cystatin C appears to be working in a similar fashion in this endothelial system to protect against amyloid cytotoxicity. Additional studies are needed to fully characterize the action of cystatin C in pulmonary infections.
The form of cystatin C that pelleted during ultracentrifugation appeared to be a dimer based on molecular weight (Fig. 8). Cystatin C is a protein that can assemble into oligomers and larger amyloids via the process of domain sharing [32][33][34][43][44][45]. The dimerization process is accelerated by cell stress, and dimers are prevalent in pathological material [32][33][34][43][44][45][46]. Most likely, infection of rat PMVECs by P. aeruginosa strain 103 is inducing a stress that leads to dimerization of cystatin C. Cystatin C is classically considered as a cysteine protease inhibitor, and dimerization is thought to generate a nonfunctional form of cystatin C as crystallization has determined that the enzyme inhibitory domain of the molecule is buried during the dimerization process [33,34]. The data here suggest that dimerized forms of cystatin may have an important function as an innate defense mechanism by binding to Ab thereby inhibiting assembly into polymers. This may have one of three consequences. The data support a model in which the complex of cystatin C and Ab have a direct cytoprotective role and assist cell survival during the infection process by an undefined mechanism. However, in another related scenario, binding to Ab may inhibit assembly of beta amyloid into large toxic species promoting survival. A third possibility is that binding of cystatin C to Ab is an adaptive mechanism of the bacterial cells and assists them with avoiding the innate immune response of the cells. Recent data indicate that beta amyloid oligomers are anti-bacterial in nature [37][38][39]. During evolution, the bacteria may have developed a system where they target cystatin C and induce its dimerization so that it inhibits assembly of beta amyloid into an anti-bacterial species. Studies investigating whether cystatin C protects bacteria from effects of beta amyloid have not been performed. Clearly, additional studies will be needed to establish the significance of the binding of cystatin C to Ab during P. aeruginosa infections.
In summary, cytotoxic Ab and oligomeric tau liberated from cultured endothelial cells during infection by the bacterium P. aeruginosa were analyzed. The cytotoxic activity of the secreted amyloids is abolished by protein denaturants, although the endothelial-derived oligomeric tau is resistant to protease digestion and capable of cell killing in the absence of Ab. Importantly, data are presented identifying cystatin C as a regulator of cytotoxic activity of Ab during infection by P. aeruginosa. As such, cystatin C may be a useful therapeutic target for protection against the consequences of pneumonia caused by P. aeruginosa. Studies in progress are investigating this possibility.

Bacterial strains
Two strains of P. aeruginosa were used in these studies, and both have been described previously [47]. For most experiments, strain PA103, which expresses both exoenzymes ExoT and ExoU and has a functional type III secretion system for transfer of exoenzymes to infected cells, was used. For some experiments, the mutant DPcrV was used. This mutant is incapable of transferring exoenzymes to target cells as the PopB/D channel is not formed [48,49].

Cell culture and production of cytotoxic supernatants
Primary cultures of rat pulmonary microvascular endothelial cells (PMVECs) were maintained as described previously [50,51]. To generate cytotoxic supernatants, PMVECs were grown to confluence and then infected with the appropriate strain of bacteria at a multiplicity of infection (MOI) of 20 : 1 using methods that were described previously [17,19,49]. Bacteria were diluted in Hanks' Balanced Salt Solution (HBSS) prior to addition to cells, and treatment with the bacteria was for 4-5 h. Culture supernatants were collected, centrifuged at 2000 g for 10 min at room temperature to remove debris, and filter sterilized through a 0.22 µm filter to remove bacteria.

Cytotoxicity assay
Cytotoxic supernatant prepared as described above was added to na€ ıve PMVECs, and the cells were placed in a 37°C humidified CO 2 incubator for 21 h. Control cells were treated with HBSS alone. The cells then were photographed and cell killing was quantified by using IMAGEJ software as described previously [19].
For some experiments, the cytotoxic supernatant was pretreated before being applied to PMVECs. For some studies, supernatant was treated with 100 µgÁmL À1 proteinase K at 37°C for 0, 1, 2, 4, 8 24 or 30 h. The protease was then inhibited by addition of a 1 : 1000 dilution of 1 mgÁmL À1 phenylmethylsulfonylfluoride (PMSF; diluted in MeOH), and the treated supernatants were added to PMVECs. For these studies, control cells were treated with HBSS that contained PMSF. For other investigations, cytotoxic supernatant was treated with 1,1,1,3,3,3-hexafluoro-2isopropanol (HFIP) at concentrations ranging from 10 : 1 to 0.5 : 1 HFIP to supernatant for 30 min at 37°C. The treated samples were then dried via centrifugation under vacuum to remove the HFIP. The dried proteins were then resuspended in a volume of HBSS equal to the starting amount of cytotoxic supernatant, and the resuspended material was added to PMVECs. In a separate set of experiments, cytotoxic supernatant was treated with diethylpyrocarbonate (DEPC) at concentrations of 1, 10, and 20 mM for 30 min at room temperature. The DEPC-treated supernatants then were added to PMVECs. The last set of studies to test stability of the cytotoxic amyloids involved treating cytotoxic supernatant with phenol. For these experiments, supernatant and saturated phenol (pH 7.0) were mixed 1 : 1 and then vortexed for 60 s. The sample was centrifuged, and the protein-containing layer was collected, dried as detailed above, resuspended in HBSS and then further desalted back into HBSS by centrifugation using a microcentrifugation device with a 3 kDa cutoff (Millipore product #UFC500324, Burlington, MA, USA). The phenol-treated supernatant then was added to cultured PMVECs to assess killing activity. In each instance, the presence of residual amyloid in the treated supernatant was analyzed by immunoblot analysis using A11 anti-amyloid antibody (see below).

Ultracentrifugation
The relative size of the cytotoxin was determined by subjecting the cytotoxic supernatant to ultracentrifugation and then assaying each centrifuged sample for residual cytotoxic activity. For these studies, equal amounts of cytotoxic supernatant were added to centrifuge tubes and then the samples were centrifuged at 150 000 g at 4°C. Samples were collected at 1, 2, 4, 8, 16, and 24 h, and the samples were then either applied to PMVECs to assess cytotoxic activity or concentrated using microconcentration devices with 3 kDa cutoffs before being analyzed by immunoblotting (see below). For some experiments, the pelleted material was collected after ultracentrifugation and analyzed by immunoblot.

Mass spectrometry
Pulmonary microvascular endothelial cells were rinsed extensively with HBSS and then infected with either PA103 or DPcrV bacteria at a MOI of 20 : 1 in HBSS. Cells were incubated for 4 h and then the supernatants were collected, centrifuged, and filter sterilized as outlined above. The supernatant was concentrated 100-fold using a microcentrifugation filtration apparatus with a 3 kDa cutoff [52]. Proteins were identified using methods outlined previously [53]. Specifically, secreted proteins were separated by SDS/ PAGE (0.8-cm separation; Criterion XT Bis-Tris 12% gel; Bio-Rad, Hercules, CA, USA). After staining with Coomassie Blue, the protein-containing region was excised into six slices and the proteins were reduced and alkylated (iodoacetamide) and digested in situ with trypsin (sequencing grade; Promega, Madison, WI, USA). Each digest was analyzed by capillary HPLC-electrospray ionization tandem mass spectrometry on a Thermo Scientific Orbitrap Fusion Lumos mass spectrometer (Waltham, MA, USA). On-line HPLC separation was accomplished with a Thermo Scientific/Dyonex RSLC NANO HPLC system: column, Pico-Frit TM (New Objective; 75 lm i.d., Woburn, MA, USA) packed to 15 cm with C18 adsorbent (Vydac; 218MS 5 lm, 300 A); mobile phase A, 0.5% acetic acid (HAc)/0.005% trifluoroacetic acid (TFA); mobile phase B, 90% acetonitrile/0.5% HAc/0.005% TFA; gradient 3 to 42% B in 30 min; flow rate, 0.4 lLÁmin À1 . Precursor ions were acquired in the Orbitrap in centroid mode (scan range, m/z 300-1500; resolution, 1-200 000); data-dependent collisioninduced dissociation spectra of ions in the precursor scan were acquired at the same time in the ion trap ('top speed'; threshold to trigger MS2, 50 000; quadrupole isolation, 0.7; charge states, 2+ to 5+; dynamic exclusion, 30 s; normalized collision energy, 30%). MASCOT (v2.6.2; Matrix Science, London, UK) was used to search the spectra against the rat subset of the UniProt database [UniProt_Rat 20170117 (31 383 sequences; 17 231 818 residues)] concatenated with a database of common protein contaminants [contaminants 20120713 (247 sequences, 128 130 residues)]. Cysteine carbamidomethylation was set as a fixed modification and methionine oxidation and deamidation of glutamine and asparagine were considered as variable modifications; trypsin was specified as the proteolytic enzyme, with one missed cleavage allowed. Subset search of the identified proteins by X! Tandem, cross-correlation with the Mascot results and determination of protein and peptide identity probabilities were accomplished by SCAFFOLD (v4.8.7; Proteome Software, Portland, OR, USA). The thresholds for acceptance of peptide and protein assignments in Scaffold were 95% and 99%, respectively.

Immunodepletion and immunoisolation
Cytotoxic supernatant was depleted of either oligomeric tau or Ab using previously reported procedures [19]. Specific antibodies used included T22 (anti-oligomeric tau) and anti-b-amyloid (Invitrogen Product #71-5800, Waltham, MA, USA). For other studies, gelsolin, cystatin C, or apolipoprotein J were depleted using the previously mentioned antibodies. Depleted supernatants were added to cultured cells to assess cytotoxic activity.
For some studies, the antigen bound to the beads was recovered for add-back experiments. To achieve this, Protein A agarose beads containing bound antibody-antigen complexes were washed six times with PBS and then once in PBS containing 0.5 M NaCl. The pelleted beads were suspended in 4 M MgCl 2 in HBSS and the beads were removed by brief centrifugation in a microfuge. The supernatants containing immune-isolated antigens and their respective antibodies were collected. As the cytotoxins are heat stable whereas antibodies are heat labile, the antibodies could be inactivated by placing the eluted material at 100°C for 10 min. The samples then were dialyzed against four changes of HBSS (100-fold excess for 2-4 h per buffer change) and added to PMVECs along with the appropriate antibody-depleted supernatant to assess whether cytotoxic activity could be restored.

Co-immunoprecipitation
To determine whether cystatin C in cytotoxic supernatants was complexed with b-amyloid, cystatin C was immunoprecipitated as described above. The beads were then rinsed extensively as detailed previously, and then, the presence or absence of co-precipitating b-amyloid was assessed by immunoblotting using anti-b amyloid antibody. For the control, the immunoprecipitation was performed using non-immune rabbit IgG (Sigma-Aldrich product #I-5006, St. Louis, MO, USA).