MUC1 ectodomain is a flagellin-targeting decoy receptor and biomarker operative during Pseudomonas aeruginosa lung infection

We previously reported that flagellin-expressing Pseudomonas aeruginosa (Pa) provokes NEU1 sialidase-mediated MUC1 ectodomain (MUC1-ED) desialylation and MUC1-ED shedding from murine lungs in vivo. Here, we asked whether Pa in the lungs of patients with ventilator-associated pneumonia might also increase MUC1-ED shedding. The levels of MUC1-ED and Pa-expressed flagellin were dramatically elevated in bronchoalveolar lavage fluid (BALF) harvested from Pa-infected patients, and each flagellin level, in turn, predicted MUC1-ED shedding in the same patient. Desialylated MUC1-ED was only detected in BALF of Pa-infected patients. Clinical Pa strains increased MUC1-ED shedding from cultured human alveolar epithelia, and FlaA and FlaB flagellin-expressing strains provoked comparable levels of MUC1-ED shedding. A flagellin-deficient isogenic mutant generated dramatically reduced MUC1-ED shedding compared with the flagellin-expressing wild-type strain, and purified FlaA and FlaB recapitulated the effect of intact bacteria. Pa:MUC1-ED complexes were detected in the supernatants of alveolar epithelia exposed to wild-type Pa, but not to the flagellin-deficient Pa strain. Finally, human recombinant MUC1-ED dose-dependently disrupted multiple flagellin-driven processes, including Pa motility, Pa biofilm formation, and Pa adhesion to human alveolar epithelia, while enhancing human neutrophil-mediated Pa phagocytosis. Therefore, shed desialylated MUC1-ED functions as a novel flagellin-targeting, Pa-responsive decoy receptor that participates in the host response to Pa at the airway epithelial surface.

. Pa-expressed flagellin provokes NEU1-mediated MUC1-ED desialylation to generate a flagellintargeting MUC1-ED decoy receptor. The Pa flagellin subunit engages cell-associated MUC1-ED (step 1), leading to recruitment of a preformed pool of intracellular NEU1, together with its chaperone protein, PPCA, to MUC1-ED (step 2). NEU1 desialylates the MUC1-ED (step 3), to increase its adhesiveness for Pa (step 4) and unmask its glycine-serine (Gly-Ser) protease recognition site (step 5), permitting sheddase-mediated MUC1-ED release from the airway EC surface (step 6). Shed, desialylated MUC1-ED in the airway lumen acts as a soluble decoy receptor that reduces Pa motility, competitively inhibits Pa adhesion to cell-associated MUC1-ED and subsequent invasion of airway epithelia, protects against Pa biofilm formation, and enhances Pa phagocytosis by host PMNs (step 7). sialidase, NEU1, together with its chaperone protein, protective protein/cathepsin A (PPCA), to surfaceexpressed MUC1 21,22 . Further, we found that NEU1 desialylates the MUC1-ED to unmask cryptic binding sites for flagellin-expressing Pa 21,23 (Fig. 1). In contrast, NEU1-mediated MUC1-ED desialylation did not increase its adhesiveness for a flagellin-deficient, isogenic Pa mutant 21 . While NEU1 overexpression in MUC1-expressing human embryonic kidney (HEK)293 cells enhanced their adhesiveness for wild-type Pa, NEU1 overexpression in Toll-like receptor (TLR)5-expressing HEK293 cells did not 21 . Although NEUs are well established virulence factors for multiple bacterial pathogens 24 , this is the only case, to our knowledge, where a prokaryote hijacks a host sialidase to influence its pathogenicity. At the same time, we also found that NEU1-mediated MUC1-ED desialylation uncovers a juxtramembranous protease recognition site, permitting MUC1-ED cleavage by endogenous sheddases and release from the EC surface 21,[25][26][27] (Fig. 1). More specifically, NEU1 overexpression in HEK293 cells expressing a MUC1-ED S317A mutant, containing a serine-to-alanine substitution within the glycine-serine protease recognition site, displayed no MUC1-ED shedding 22 . Shed, desialylated MUC1-ED in the airway lumen retains its ability to bind to flagellin-expressing Pa, thereby acting as a soluble decoy receptor that reduces Pa motility and competitively inhibits Pa adhesion to cell-associated MUC1-ED 21,22 , both reducing Pa invasion of underlying tissues 22 . In fact, in a mouse model of Pa pneumonia, co-administration of the MUC1-ED decoy receptor with a lethal challenge of Pa diminished lung bacterial burden, lung cytokine production, and pulmonary leukostasis, and increased 5-day survival from 0 to 77%, indicating that the shed MUC1-ED constitutes a protective component of the host response to Pa 22 (Fig. 1). In the current studies, we sought to extend our prior findings in the mouse model to an in vivo human setting. We asked whether the shed MUC1-ED levels might be elevated in BALF of Pa-infected patients, as an indirect but specific measure of NEU1 catalytic activity 21,23 . Further, we tested human recombinant MUC1-ED for its ability to influence Pa adhesion to alveolar epithelia, Pa motility, Pa biofilm production, Pa-stimulated interleukin-8 (IL-8) production, and neutrophil-mediated Pa phagocytosis.

MUC1-ED levels in bronchoalveolar lavage fluid (BALF) from Pa-infected patients.
We previously reported that Pa increases MUC1-ED shedding using an in vivo mouse model of Pa pneumonia 22 . We asked whether MUC1-ED levels also might be increased in BALF from patients with Pa airway infection. Between 12 June 2013 and 20 February 2020, BALF was collected from 61 patients undergoing mechanical ventilation and clinically diagnosed with VAP in three separate units at the University of Maryland Medical Center (Baltimore, MD). Of these 61 patients, the BALF of 13 (21.3%) were culture-positive for Pa, 32 (52.5%) culture-positive for other microorganisms, 6 of which were polymicrobial, and 16 (26.2%) culture-negative (Table 1). Power analysis indicated that this study is 100% powered with these sample sizes to detect different BALF MUC1-ED levels in Pa-positive (n = 13) vs. Pa-negative (n = 48) patients. Pa-infected VAP patients were more likely to have been admitted to the long-term Comprehensive Pulmonary Rehabilitation Unit (CPRU) (61.5% vs. 12.5%), and experienced fewer ventilator-free days (8.2 vs. 18.7 days), compared with Pa-negative patients (Table 1). Interestingly, Pa-infected VAP patients were less likely to have been admitted to the acute care Medical Intensive Care   Fig. 2A,B). Therefore, the mean normalized MUC1-ED level in Pa-infected VAP patients was 5.6-fold greater than that measured in patients infected with other microorganisms, and 7.7-fold greater compared with culture-negative patients. These combined data suggest that elevated BALF levels of MUC1-ED might provide a diagnostic biomarker for Pa lung infection. We asked whether MUC1-ED levels in BALF of Pa-infected VAP patients might correlate with Acute Physiology and Chronic Health Evaluation (APACHE) II scores as a predictor of disease severity. In 11 patients where APACHE II scores were available, MUC1-ED levels did not correlate with APACHE II values (Supplemental Fig. 1).
Since NEU1-mediated MUC1-ED desialylation is required for its proteolytic release and shedding 22 , we next asked whether MUC1-ED in BALF from Pa-infected patients was desialylated. MUC1-ED immunoblotting of BALF proteins bound to the lectin, peanut agglutinin (PNA), which binds to subterminal galactose residues only after removal of terminal sialic acid, was used to address this question. First, we validated the specificity of the PNA lectin. As anticipated, PNA recognized asialofetuin, but not sialylated fetuin (Fig. 2C). MUC1-ED in the BALF of Pa-infected VAP patients was desialylated, while desialylated MUC1-ED in the BALF from Pa-negative patients was not detected (Fig. 2D). Similarly, total shed MUC1-ED was detected in the BALF from Pa-infected VAP patients, but not in the BALF from Pa-negative patients (Fig. 2E). The total MUC1-ED concentrations in the four selected samples from the Pa-positive group were comparable, as were the MUC1-ED levels in the four selected Pa-negative samples (Fig. 2E). Of note, although low levels of MUC1-ED were seen by ELISA in BALF harvested from the two Pa-negative patient groups ( Fig. 2A,B), desialylated MUC1-ED could not be detected in Pa-negative patients either by the combined PNA enrichment/MUC1-ED immunoblotting protocol (Fig. 2D, lanes 1-4) or the immunoblotting-only protocol (Fig. 2E, lanes 1-4). It is conceivable that the relatively higher sensitivity of the ELISA procedure, compared with the PNA/immunoblotting or immunoblotting-only protocols, allows detection of very low levels of MUC1-ED by the former and not the latter.
MUC1-ED levels were also quantified in tracheal aspirate (TA)s and BALFs simultaneously collected from a subset of VAP patients (n = 15), including 2 patients infected with Pa, 6 patients infected with other microorganisms, and 7 culture-negative patients. No correlation was observed between MUC1-ED levels in the paired BALF and TA samples (Supplemental Fig. 2). Therefore, the MUC1-ED levels in the more easily obtained TA samples did not reflect BALF levels. However, due to the small sample size for TAs obtained from patients who cultured positive for Pa, whether a correlation exists could not be definitively excluded. MUC1-ED shedding by Pa-derived flagellin. MUC1-ED levels are increased in the BALF of mice challenged with either flagellin-expressing Pa or purified Pa-derived flagellin 22 and in BALF from VAP patients infected with Pa ( Fig. 2A,B). We asked whether Pa-derived flagellin might be responsible for MUC1-ED shedding, in vitro. First, we used an in vitro bacterial motility assay to establish flagella expression in selected microorganisms, including wild-type (WT)-PAK, Stenotrophomonas maltophilia, Legionella pneumophila, and Salmonella enterica serovar Typhimurium. A nonmotile PAK/fliC¯ flagellin-deficient, isogenic mutant strain was used as a negative control. With the exception of the PAK/fliC¯ strain, all microorganisms were motile (Fig. 3A). We next established the time requirements for WT-PAK-provoked increases in MUC1-ED shedding from in vitro cultures of A549 human alveolar epithelial cells (ECs) (Fig. 3B). At times ≥ 2 h, shed MUC1-ED levels were increased over time, with maximal levels at 24 h. Over the same time period, the PAK/fliC¯ mutant strain only induced minimal MUC1-ED shedding at later time points that did not achieve statistical significance compared with the phosphate-buffered saline (PBS) vehicle control. Stimulation of in vitro cultures of A549 ECs with either of two Pa laboratory strains, PA01 or PAK, dramatically increased MUC1-ED levels in cell culture supernatants compared with the PBS control (Fig. 3C). A549 EC stimulation with the PAK/fliC¯ flagellin-deficient strain was associated with dramatically reduced MUC1-ED shedding compared with that provoked by flagellin-expressing

Formation of complexes comprised of flagellin-expressing Pa and the shed MUC1-ED decoy receptor.
Since NEU1-mediated MUC1-ED desialylation both renders the ED hyperadhesive for Pa 21 and unmasks its protease recognition site, permitting its proteolytic release 21,22 , we asked whether shed MUC1-ED might bind to flagellin-expressing Pa. WT-PAK and the PAK/fliC¯ strain each were co-cultured for 24 h with A549 cells, the supernatants collected and centrifuged to collect the bacteria. The collected Pa were lysed and the lysates processed for MUC1-ED immunoblotting (Fig. 3E). The WT-PAK was detected in Pa:MUC1-ED complexes, whereas the PAK/fliC¯ strain was not. Whether these Pa:MUC1-ED complexes form with cell-tethered MUC1-ED at the airway epithelial cell surface and are then released, and/or form with already shed MUC1-ED unoccupied by Pa, is unclear.

Pa-derived flagellin is present in the BALF of Pa-infected VAP patients. Flagellin-expressing Pa
stimulates MUC1-ED shedding in vitro (Fig. 3), and MUC1-ED levels are increased in the BALF of mice with Pa lung infection 22 and Pa-infected patients with VAP ( Fig. 2A,B). We asked whether Pa flagellin could be detected in the BALF of Pa-infected VAP patients. FlaA and FlaB flagellin levels were quantified by ELISA and normalized to total BALF protein. The mean ± S.E. normalized FlaA level was 131.5 ± 9.1 ng/mg BALF protein (range, 91.6 to 154.9 ng/mg) in 7 of 13 VAP patients infected with FlaA-expressing Pa (Fig. 4A,B). The mean ± S.E. normalized FlaB level was 139.7 ± 6.8 ng/mg (range, 112.0 to 163.4 ng/mg) in 6 of 13 VAP patients infected with FlaB-expressing Pa (Fig. 4A,B). Only one flagellin type, either FlaA or FlaB, was present in each of the 13 patient BALFs, suggesting that only one Pa strain was responsible for each patient's infection, but not absolutely   www.nature.com/scientificreports/ excluding coinfection with two Pa strains expressing the same flagellin type. To establish how many Pa CFUs a given flagellin level reflects, PAK and PA01 bacteria were cultured in either bacterial medium or eukaryotic tissue culture medium, in the presence or absence of A549 cells, to simulate Pa lung infection more closely, in vivo. Increasing inocula of PAK and PA01 bacteria were lysed and the lysates processed for FlaA or FlaB ELISA (Fig. 4C). Using these FlaA and FlaB standard curves, and given the comparable levels of flagellin found in laboratory and clinical Pa strains 28 , a crude estimate of the Pa lung burden for each patient was calculated.

Human recombinant MUC1-ED as a flagellin-targeting decoy receptor and promoter of phagocytosis. Flagellin contributes to Pa pathogenesis through increased motility, biofilm formation, adhesion
to and invasion of airway epithelia, and inflammation through activation of EC-expressed TLR5, downstream interleukin (IL)-8 production, and pulmonary leukostasis, predominantly with polymorphonuclear leukocytes (PMNs) [29][30][31][32][33][34] . In a mouse model of Pa pneumonia, endogenous murine MUC1-ED in BALF diminishes Pa motility and adhesion to airway ECs 22 . Further, using metabolic inhibitors of protein glycosylation and a nonglycosylated, E. coli-derived human rMUC1-ED, we demonstrated that the decoy receptor function of MUC1-ED resides within its protein backbone and not its N-or O-linked carbohydrates 22 . To further validate the utility of human rMUC1-ED as a Pa flagellin-targeting intervention, we used immunogold transmission electron microscopy (EM) to establish its binding to the intact Pa flagellum. Gold-labeled anti-mouse IgG secondary antibody immunolocalized to the Pa flagellar filament following incubation of the bacteria with human rMUC1-ED and mouse anti-MUC1-ED antibody (Fig. 5A, panel i), but not to bacteria incubated with rMUC1-ED and a nonimmune mouse IgG control (Fig. 5A, panel ii). Next, we tested human rMUC1-ED for its ability to inhibit flagellindependent Pa motility. Human rMUC1-ED dose-dependently inhibited Pa motility by up to 81.4%, compared with the simultaneous PBS vehicle control. (Fig. 5B). In a Pa biofilm assay where the bacteria were incubated with increasing amounts of rMUC1-ED in plastic microwells, rMUC1-ED reduced Pa biofilm formation by up to 90.5% (Fig. 5C). In a bacterial adhesion assay where Pa were cultured with A549 cells in the presence of increasing amounts of rMUC1-ED, Pa adhesion to the ECs was dose-dependently diminished by up to 78.4% (Fig. 5D). When these same ECs were incubated with Pa in the presence of the same escalating concentrations of rMUC1-ED, IL-8 levels in the cell culture supernatants were reduced by up to 80.5% (Fig. 5E). Finally, in a PMN opsonophagocytosis assay where green fluorescent protein (GFP)-expressing Pa were incubated with primary human PMNs in the presence of increasing amounts of rMUC1-ED, intracellular levels of GFP-Pa were dosedependently elevated up to 3.87-fold compared with the simultaneous PBS control (Fig. 5F). However, human rMUC1-ED at 100 µM displayed no effect on the growth of flagellin-expressing Pa, compared with the PBS control (Fig. 5G). Thus, human rMUC1-ED, at concentrations comparable to those of MUC1-ED found in the BALF of Pa-infected patients ( Fig. 2A,B) exhibits robust, flagellin-targeting decoy receptor function and enhances the PMN phagocytic response to Pa, without inhibiting bacterial replication.

Discussion
In the current study, 21.3% of mechanically ventilated patients were infected with Pa. Pa is the second most common microorganism isolated in U.S. hospital-acquired lung infections with 90,000 cases per year 35 . In a prospective study of mechanically ventilated ICU patients across 56 sites in 11 countries, the incidence of Pa pneumonia ranged from 13.5% to 19.4% 36 . In a review of 24 studies, Pa was the most frequent isolate from ICU VAP patients, accounting for 24.4% of all microorganisms isolated 12 . Of the 13 Pa-positive patients in our study, 8 (61.5%) were identified in a chronic ventilator Comprehensive Pulmonary Rehabilitation Unit and 5 (38.5%) in an acute tertiary care medical ICU ( Table 1). The increased incidence of Pa-positive patients in the chronic vs. acute ICU setting likely represents longer duration of intubation in the chronic ICU 7 . In fact, Pa lung infection was associated with a reduction in the number of ventilator-free days (8.2 ± 11.3 days) compared with Pa-negative patients (18.7 ± 10.0 days). The current study supports Pa-derived flagellin as a ligand for the MUC1-ED, and suggests that flagellin engagement of the cell-associated MUC1-ED provokes MUC1-ED shedding in the airways of VAP patients. Using an in vivo experimental murine model to explore lung responsiveness to Pa, we have established flagellin-driven MUC1-ED shedding from the airway EC surface as a novel, Pa-selective host response to lung infection 22  www.nature.com/scientificreports/ specifically, purified Pa-derived flagellin dose-dependently increased MUC1-ED shedding from the murine airway epithelium into the brochoalveolar compartment. In the current study, BALF levels of the two Pa flagellin types, FlaA and FlaB, each could be tightly correlated with BALF levels of shed MUC1-ED in the same patient. Pa flagellin levels predicted MUC1-ED levels in a concentration-dependent manner. In these same patients, we found a greater than 6.0-fold increase in the mean ± S.E. MUC1-ED level in the BALF of Pa-infected VAP patients (4.01 ± 0.19 µg/mg BALF protein) compared with that observed in Pa-negative patients (0.66 ± 0.05 µg/ mg) ( Fig. 2A,B). No differences in BALF MUC1-ED levels were seen between patients infected with microorganisms other than Pa (0.71 ± 0.05 µg/mg) and those who were culture-negative (0.52 ± 0.08 µg/mg), indicating that the elevation of MUC1-ED levels is specific for Pa lung infection. Although S. maltophilia increased MUC1-ED shedding in vitro (Fig. 3C), whether BALF MUC1-ED levels are increased in patients with S. maltophilia lung infections is unknown. Of note, 100% of VAP patients who were Pa-positive had MUC1-ED levels ≥ 2.87 µg/mg BALF protein (range, 2.87 to 5.34 µg/mg), whereas all patients who were infected with other microorganisms or culture-negative had MUC1-ED levels ≤ 1.55 µg/mg (range, 0.05 to 1.55 µg/mg). Based on these data, 2.0 µg MUC1-ED/mg BALF protein provides a reliable diagnostic cut-off value to discriminate between Pa-infected and non-Pa-infected patients. Elevated BALF MUC1-ED levels may offer a predictive biomarker for Pa lung infection in VAP patients. The MUC1-ED ELISA is more rapid, less labor intensive, and less expensive than conventional bacterial cultures. However, the MUC1-ED ELISA identifies only Pa, and unlike bacterial cultures, does not provide antibiotic sensitivities. Further studies are required to establish whether elevated BALF MUC1-ED levels might also result from S. maltophilia lung infections.
In the mouse, Pa-stimulated MUC1-ED desialylation and shedding is a NEU1 sialidase-dependent process 22 . While NEU1 has historically been studied in context of its indispensable role in glycan catabolism 37 , far less is known about its role in the host response to septic processes, including Pa lung infection. We previously established NEU1 as the predominant sialidase expressed in human airway epithelial cells 23 . Silencing of NEU1 expression reduces total human lung airway EC sialidase activity by more than 70%. We detected NEU1 immunostaining in the superficial epithelium along the entire human airway, including the brush border 23 . This NEU1 expression pattern closely correlates with that known for MUC1 in these same tissues 38,39 . The MUC1-ED isolated from BALF of Pa-infected patients was desialylated (Fig. 2D). We have previously demonstrated that increased levels of soluble, desialylated MUC1-ED offer an indirect but specific measure of NEU1 catalytic activity 21,22 . We found that after siRNA-induced NEU1 silencing, MUC1-ED desialylation and shedding from Pa flagellin-exposed airway ECs is nonexistent 21 . Similarly, in a previous study, we found that NEU1-selective sialidase pharmacologic inhibition dramatically reduces MUC1-ED shedding in Pa-challenged mice 22 . These combined data indicate that NEU1 catalytic activity is increased in the lungs of Pa-infected patients. In previous coimmunoprecipitation and PNA lectin blotting studies, flagellin stimulation increased NEU1 association with and desialylation of the MUC1-ED 22 . NEU1-driven MUC1-ED desialylation unmasked not only cryptic binding sites for Pa flagellin, dramatically increasing Pa adhesion to lung ECs, but also the MUC1-ED glycine-serine protease recognition site, permitting its proteolytic cleavage and release from the airway EC surface 22 . With these results in mind, it is not surprising that desialylated MUC1-ED could not be detected in the BALF of patients who were Pa culturenegative or infected with microorganisms other than Pa (Fig. 2D, lanes 1-4).
Pa flagellin constitutes an important virulence factor through bacterial motility, adhesion to and invasion of host epithelia, and inflammatory gene expression via engagement of host TLR5 [29][30][31][32] . Flagellin-dependent motility and adhesion to ECs are both prerequisites for Pa biofilm formation 33 and flagellin is a target of opsonophagocytosis by host PMNs 34 . Host factor(s) disrupting one or more flagellin-driven processes would be predicted to protect against Pa pathogenesis. At MUC1-ED concentrations similar to those found in the BALF of Pa-infected patients, the Pa flagellin-targeting, E. coli-expressed human rMUC1-ED decoy receptor reduced Pa motility, adhesion to alveolar ECs and biofilm formation, as well as IL-8 production, and enhanced neutrophil-mediated Pa phagocytosis. Whether MUC1-ED bound to Pa flagellin functions as an authentic opsonin is unclear. Human rMUC1-ED exerted no measurable bacteriostatic or bactericidal activity. Through its ability to impair flagellin-driven Pa motility, MUC1-ED may interfere with Pa penetration of the mucus blanket 40 , invasion of the epithelium 23 , movement through the paracellular pathway into subepithelial tissues 41 , and/or Pa resistance to www.nature.com/scientificreports/ mucociliary clearance 42 . Inhibition of flagellin-dependent Pa adhesion to host airway epithelia likely decreases Pa colonization and invasion of subepithelial tissues. Disruption of the MUC1-ED-Pa flagellin receptor-ligand interaction blocks MUC1-driven downstream intracellular signaling 43 , biosynthesis of proinflammatory cytokines/ chemokines 44,45 , pulmonary leukostasis 22 , and neutrophil recruitment to the bronchoalveolar compartment 46 .
On the other hand, NEU1-mediated release of sialic acid residues from the highly sialylated MUC1-ED substrate may facilitate their incorporation into sialic acid-rich Pa biofilms 33 and/or dampen complement activation 47 , thereby contributing to Pa pathogenesis.
In conclusion, MUC1-ED and Pa-derived flagellin were detected in the BALF of Pa-infected VAP patients at dramatically elevated levels compared with those found in Pa-negative patient BALF. Extending the ex vivo studies to an in vitro model of Pa infection, stimulation of human alveolar ECs with either Pa or S. maltophilia, but not other recognized airway pathogens, increased MUC1-ED levels in cell culture supernatants and purified rFlaA and rFlaB flagellins recapitulated the effect of intact Pa bacteria. E. coli-expressed, human rMUC1-ED dose-dependently inhibited Pa motility, adhesion to alveolar epithelia, biofilm formation, and IL-8 production, while at the same time, enhanced neutrophil-mediated Pa phagocytosis, without influencing bacterial growth. The current studies expand our previous observation of a flagellin-provoked, NEU1-mediated generation of a MUC1-ED decoy receptor in a mouse model of Pa pneumonia 22 to human in vivo pathophysiology. Measurement of MUC1-ED and flagellin BALF levels offer a rapid, reliable means to identify ventilated patients with Pa pneumonia that might serve as a guide for empiric antibiotic therapy. Of note, the mechanism of action of most conventional antibiotics requires bacterial replication and cell wall remodeling 13 . It is conceivable that human rMUC1-ED, which itself does not alter bacterial replication, might be combined with antibiotics for synergistic therapy against MDR Pa.

Methods
Ethics statement. This study was conducted in accordance with the Declaration of Helsinki and other local statutes or regulations protecting subjects in biomedical research, and was approved by the Institutional Review Boards of the University of Maryland Baltimore (protocol numbers HP-00059183 and HP-00083805). Informed consent was obtained from all participants.
Study setting and design. This study was conducted using BALF collected from VAP patients at the University of Maryland Medical Center (Baltimore, MD). VAP patients were admitted either to the 29-bed Medical Intensive Care Unit (MICU) or the 9-bed Critical Care Resuscitation Unit (CCRU) of the University Teaching Hospital, or received prolonged mechanical ventilation in the 14-bed Comprehensive Pulmonary Rehabilitation Unit (CPRU) of the long-term acute care University Specialty Hospital. Patient data were collected as part of routine hospital standard of care, including age, race, gender, body mass index (BMI), ICU location, ventilatorfree days, antibiotics administered, chest x-ray, APACHE II score, and microbiology culture results. Bacterial identification of surveillance cultures was done at the University of Maryland Medical Center Clinical Microbiology Laboratory following standard laboratory procedures using the Vitek MS for bacterial identification and Vitek-2 system for antimicrobial susceptibilities (Biomerieux, Marcy l'Etoile, France) and interpreted according to Clinical and Laboratory Standards Institute (CLSI) criteria 48 . Clinical diagnosis of VAP was based on findings of a new pulmonary infiltrate by chest x-ray in the setting of fever, purulent sputum, leukocytosis, and diminished oxygenation 1 . Bronchoalveolar lavage and tracheal aspiration. BAL was performed on VAP patients undergoing standard of care diagnostic bronchoscopy. Briefly, after conscious sedation with fentanyl and midazolam, and local anesthesia with 2% lidocaine, the bronchoscope was wedged in a 3rd or 4th order bronchus, after which 125 ml of sterile pyrogen-free 0.9% NaCl was injected in 25-ml aliquots. The BALF was retrieved, pooled, filtered through sterile gauze, centrifuged at 450 xg to remove cells, and the supernatant concentrated 25-fold by ultrafiltration and stored until use at − 80 °C. Tracheal aspiration was performed using a 12 French siliconized polyvinyl chloride tracheal aspiration probe (Embramed, Brazil) as described 49 . Bacteria. Pa strains PA01, GFP-PA01, PAK, and PAK/fliC¯, Pa clinical strains isolated from pneumonia patients, L. pneumophila, S. pneumoniae, H. influenzae, S. aureus, and K. pneumoniae were previously described 22,23,[50][51][52] . S. maltophilia strain 810-2 was obtained from the American Type Culture Collection (Manassas, VA). All bacteria were cultured at 37 °C in LB medium (10 mg/ml tryptone, 5.0 mg/ml yeast extract, 10 mg/ ml NaCl) (Thermo Fisher Scientific, Waltham, MA), and quantified spectrophotometrically at A 600 . In selected experiments, Pa strains PA01 and PAK were cultured at 37 °C in DMEM containing 10% FBS (Hyclone Laboratories, Logan, UT) in the presence or absence of confluent monolayers of A549 human alveolar ECs (American Type Culture Collection).
Purification of Pa and STm flagellins. An overnight culture of PA01 was centrifuged at 5000×g for 30 min, resuspended in Krebs-Ringer buffer, and incubated for 1 h at 37 °C. The bacteria were collected by centrifugation and the supernatant was filtered through a 0.22 μm pore membrane and the filtrate boiled for 20 min. The filtrate was concentrated by centrifugal ultrafiltration, adjusted to pH 6.0, and flagellin purified by sequential ion exchange chromatography using Macro-Prep High S and Macro-Prep High Q resins (Bio-Rad). Flagellin was purified from STm strain CVD1925 as described 53,54 . Pa and STm flagellins were incubated with polymyxin B-agarose (Thermo Fisher Scientific) to remove LPS, after which less than 0.1 endotoxin units/µg of protein was detected by the Limulus amebocyte lysate assay.  MUC1-ED immunoblotting. Equal protein aliquots of BALFs were resolved by SDS-PAGE. In order to adequately visualize the > 250 kDa MUC1-ED protein, discontinuous gel electrophoresis was performed under reducing conditions in 3% stacking/5% separating acrylamide gels containing 25 mM Tris-HCl, 190 mM glycine, 0.1% SDS, pH 8.3 until the 150 kDa prestained protein molecular weight marker reached the bottom of the gel. The resolved proteins were transferred to polyvinylidene fluoride (PVDF) membranes (Bio-Rad, Hercules, CA) and the membranes probed with rabbit anti-mouse MUC1-ED antibody, followed by HRP-conjugated goat anti-rabbit IgG secondary antibody and enhanced chemiluminescence reagents (Thermo Fisher Scientific), as described 22 .
PNA lectin blotting of desialylated MUC1-ED. Equal protein aliquots of BALF were incubated with PNA immobilized on agarose beads (Vector Laboratories, Burlingame, CA) to selectively enrich for PNA-binding proteins and the PNA-bound proteins resolved by SDS-PAGE as described above. The resolved proteins were transferred to PVDF membranes, and the membranes processed for MUC1-ED immunoblotting as above. Asialofetuin (1.0 µg) was used as a positive control for PNA-binding proteins whereas fetuin (1.0 µg) was used as a negative control.
Bacterial motility assay. Bacteria in mid-log phase (A 600 = 0.5), were resuspended in LB medium, stabinoculated into 0.3% LB agar plates, incubated overnight, and colony diameters measured as an indicator of bacterial motility, as described 22 . In other experiments, PAK were preincubated with increasing concentrations rMUC1-ED, washed, resuspended in LB medium, and analyzed for motility.
Detection of Pa:MUC1-ED complexes. A549 cells (2.0 × 10 5 cells/well) were cultured in DMEM containing 10% FBS in 24-well plates. The cells were incubated for 24 h at 37 °C with 1.0 × 10 8 CFUs/well of WT-PAK or the PAK/fliC¯ flagellin-deficient strain. The bacteria were collected by centrifugation of culture supernatants at 5,000 xg for 10 min, washed 3 times with PBS, pH 7.4, and lysed. Equal protein aliquots (10 µg) of the lysates were processed for MUC1-ED immunoblotting as above.

Expression and purification of human rMUC1-ED. Human MUC1-ED incorporating flanking NcoI
and EcoRI restriction sites was amplified by PCR from a full-length MUC1-pcDNA3 plasmid 55 and the amplicons subcloned into the pBAD/6X-His plasmid (Thermo Fisher Scientific). The plasmid was transformed into E. coli NEB 5-alpha competent cells (New England Biolabs), the cells were cultured to mid-log phase (A 600 = 0.5), induced for 3 h with 0.01% arabinose, and lysed by sonication. The 6X-His epitope-tagged human rMUC1-ED was purified on a nickel-nitrilotriacetic acid affinity column, its identity verified by immunoblotting with antihuman MUC1-ED antibody, and purity confirmed by a single Coomassie blue-stained protein band following SDS-PAGE.
Immunogold EM. Immunogold electron microscopy using a Tecnai T12 transmission electron microscope (Thermo Fisher Scientific) to detect rMUC1-ED binding to Pa flagellin was performed as previously described 56 .