Bps polysaccharide of Bordetella pertussis resists antimicrobial peptides by functioning as a dual surface shield and decoy and converts Escherichia coli into a respiratory pathogen

Infections and disease caused by the obligate human pathogen Bordetella pertussis (Bp) are increasing, despite widespread vaccinations. The current acellular pertussis vaccines remain ineffective against nasopharyngeal colonization, carriage, and transmission. In this work, we tested the hypothesis that Bordetella polysaccharide (Bps), a member of the poly-β-1,6-N-acetyl-D-glucosamine (PNAG/PGA) family of polysaccharides promotes respiratory tract colonization of Bp by resisting killing by antimicrobial peptides (AMPs). Genetic deletion of the bpsA-D locus, as well as treatment with the specific glycoside hydrolase Dispersin B, increased susceptibility to AMP-mediated killing. Bps was found to be both cell surface-associated and released during laboratory growth and mouse infections. Addition of bacterial supernatants containing Bps and purified Bps increased B. pertussis resistance to AMPs. By utilizing ELISA, immunoblot and flow cytometry assays, we show that Bps functions as a dual surface shield and decoy. Co-inoculation of C57BL/6J mice with a Bps-proficient strain enhanced respiratory tract survival of the Bps-deficient strain. In combination, the presented results highlight the critical role of Bps as a central driver of B. pertussis pathogenesis. Heterologous production of Bps in a non-pathogenic E. coli K12 strain increased AMP resistance in vitro, and augmented bacterial survival and pathology in the mouse respiratory tract. These studies can serve as a foundation for other PNAG/PGA polysaccharides and for the development of an effective Bp vaccine that includes Bps.


Introduction
Bordetella pertussis (Bp) is a strict human-adapted Gram-negative respiratory tract pathogen, which causes whooping cough or pertussis, a highly contagious disease. Despite high and widespread vaccine coverage, the incidence of pertussis has increased in many countries. Pertussis is traditionally described as a childhood disease, which results in severe and sometimes fatal infections in newborns and infants. There has also been an increase in infections of vaccinated adolescents and adults, in whom pertussis manifests as a persistent cough with milder symptoms [1][2][3][4].
The adaptation of Bp to humans as an exclusive host was primarily associated with extensive loss of genomic content and gene inactivation [4,5]. Thus, Bp is likely evolving to retain only genes required for its survival in human hosts. The bpsA-D locus, which encodes the Bordetella polysaccharide (Bps), is conserved in all sequenced and annotated strains of Bp. Therefore, investigating pathogenic roles of Bps will provide insights into how Bp survives in its only niche, the human respiratory tract. Bps belongs to the poly-β-1,6-N-acetyl-D-glucosamine (PNAG/PGA) family of polysaccharides. These polysaccharides are produced by numerous Gram-positive and Gram-negative bacteria, fungal and eukaryotic organisms including Plasmodia spp., the causative agents of malaria [6][7][8][9]. For Bp, Bps was the first identified factor to promote colonization of the mouse nose. In addition, Bps was also critical for the colonization of the mouse trachea and lungs [6]. However, the mechanism(s) by which Bps promotes respiratory tract survival is unclear.
After exposure to inhaled microorganisms, the innate immune system of the respiratory tract functions as a highly effective defense against infectious agents. Antimicrobial peptides (AMPs) are critical innate immune components within the respiratory tract which exhibit broad-spectrum and potent microbicidal activities against both Gram-negative and Gram-positive bacteria [10][11][12]. However, bacteria employ diverse strategies to resist killing by AMPs. Alterations of net charge and permeability of the cell surface are some commonly employed strategies for AMP resistance. Common bacterial factors in Gram-negative and Gram-positive bacteria that contribute to AMP resistance are LPS and teichoic acids (TA), respectively [13][14][15]. There are few reported mechanisms that detail how Bp resists human AMPs. Modification of the lipo-oligosaccharide lipid A region with glucosamine increases resistance against LL-37 [16]. Similar to that observed for TA of Gram-positive bacteria and LPS of V. cholerae [15,[17][18][19], we described that the addition of the amino acid D-alanine to a yet unidentified outer membrane component enhances resistance to several human AMPs [20].
In the current manuscript, we tested the hypothesis that Bps promotes resistance of Bp to AMPs. We show that Bps binds AMPs and protects bacteria from AMP-mediated killing when present on the bacterial surface and as a secreted factor. We also investigated the pathogenic consequences of Bps production in a non-pathogenic commensal E. coli K12 strain. Production of Bps as the sole Bp factor was sufficient to confer on E. coli the ability to survive in the mouse respiratory tract and resulted in aggravated lung pathology.

The bpsA-D polysaccharide locus of B. pertussis contributes to AMP resistance
We showed previously that a Bp strain harboring an in-frame deletion in the genes of the bpsA-D locus (ΔbpsA-D), which encodes the Bps polysaccharide, colonized the mouse respiratory tract less efficiently than the wildtype (WT) strain as early as six hours post-challenge [6]. Since AMPs act early after infection to protect the respiratory tract from bacterial pathogens, we hypothesized that the ΔbpsA-D strain will be more sensitive to these antibacterial compounds than the WT strain. We first compared the sensitivities of these two strains to polymyxin B (PmB), a cationic antibiotic peptide that is an excellent model for bactericidal actions of AMPs [21]. Compared to the WT strain, the ΔbpsA-D mutant was significantly more sensitive to PmB-mediated killing (Fig 1A). Complementation of the ΔbpsA-D mutant with the plasmid pMM11 containing the cloned bpsA-D locus (ΔbpsA-D comp ) considerably increased bacterial survival in the presence of PmB when compared to that of the mutant strain harboring the vector plasmid only (ΔbpsA-D vec ) ( Fig 1A).
Next, we tested the sensitivities of the WT and ΔbpsA-D strains to human AMPs. We used LL-37 (a human cathelicidin produced by phagocytes and epithelial cells in the respiratory tract) [22] along with HNP-1 and HNP-2 (members of α-defensin family found in the human respiratory tract) [23,24]. Compared to the WT strain, the ΔbpsA-D mutant was killed significantly better in the presence of all these host AMPs (Fig 1B-1D). Furthermore, ΔbpsA-D comp strain showed significantly more resistance compared to the ΔbpsA-D vec strain when exposed to each AMP (Fig 1B-1D). Interestingly, compared to the WT bacteria, complementation of the ΔbpsA-D strain with pMM11 resulted in increased resistance to PmB and LL-37 ( Fig 1A-1B), which could largely be because the ΔbpsA-D comp strain produces higher amounts of Bps than WT bacteria (S1 Fig). Taken together, these results demonstrate that the bpsA-D locus promotes resistance to several different AMPs.

Dispersin B increases the susceptibility of B. pertussis to PmB and LL-37
Dispersin B (DspB) is a glycoside hydrolase that specifically hydrolyzes poly-β-1,6-N-acetyl-Dglucosamines [25][26][27]. While the exact structure of Bps is unknown, based on its immune reactivity and susceptibility to DspB, Bps appears to be a poly-β-1-6-N-acetyl-D-glucosamine polysaccharide [6,27,28]. Thus, DspB offers a useful biochemical tool to assess the contribution of Bps to AMP resistance independent of the ΔbpsA-D strain. The WT and ΔbpsA-D strains were treated with various concentrations of DspB before incubation with PmB or LL-37. Based on the results in Fig 2A, 50 μg/ml of DspB was chosen for further experiments, since it reduced the levels of detectable Bps on WT bacteria to that detectable on ΔbpsA-D bacteria. The residual detectable material in the ΔbpsA-D strain probably represents some other material that is weakly cross-reactive with WGA-HRP. Compared to incubation with PmB or LL-37 alone, treatment of the WT strain with either DspB + PmB (Fig 2B) or DspB + LL-37 ( Fig 2C) resulted in a 3.5-and 3.0-log-fold increase in bacterial killing, respectively. Compared to treatment with PmB or LL-37 alone, treatment of the ΔbpsA-D strain with DspB + PmB ( Fig 2B) or DspB + LL-37 ( Fig 2C) did not result in any further increase in bacterial killing, suggesting that the activity of DspB is specific to Bps. Treatment with DspB alone did not have any significant effect on the survival of either the WT or ΔbpsA-D strains (S2A Fig), suggesting that DspB does not have any toxic effect on Bp. Furthermore, treatment with heat-inactivated DspB before exposure to LL-37 had no significant reduction in survival compared to bacteria treated with LL-37 alone (S2B Fig). These data show that enzymatic degradation of Bps increases the susceptibility of Bp to killing by PmB and LL-37.

The presence of Bps on the B. pertussis cell surface inhibits LL-37 and PmB binding
Since the results presented thus far suggest that Bps protects Bp by limiting the killing activity of PmB and LL-37, we hypothesized that Bps inhibits the binding of AMP to the bacterial cell surface. To test this, we used flow cytometry to quantify the binding of FITC-labelled LL-37 to fixed bacteria. Fig 3A shows overlaid histograms comparing median fluorescence intensities of FITC, which are quantified in Fig 3B. Addition of FITC-labeled LL-37 to WT bacteria resulted in an increase in FITC fluorescence compared to control (WT bacteria alone), indicating that LL-37 binds to WT bacteria. Addition of FITC-labeled LL-37 to ΔbpsA-D bacteria led to a

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide considerable increase in FITC fluorescence, suggesting that higher amounts of LL-37 bind to the ΔbpsA-D strain than to the WT strain. This phenotype was also observed with PmB, using an ELISA-based method with an α-PmB antibody. As shown in S3 Fig, PmB bound to WT bacteria, indicated by an increase in signal compared to the negative control (blank). However,

Bps is released during laboratory growth of B. pertussis
In addition to producing polysaccharides on the cell surface, many bacteria also release them in the growth medium [8,29]. It is not known if Bps is naturally released during laboratory growth of Bp. To determine this, the amounts of Bps present on the bacterial cell surface (cellassociated) and released into the growth medium (supernatant; collected after centrifugation and filtration of the spent medium) were quantitated by ELISA using the lectin wheat germ agglutinin (WGA) conjugated to HRP. In WT cells, Bps was detected on both the cell surface and in the supernatant ( Fig 4A). As expected, negligible amounts of Bps were detected on the cell-surface or in supernatant obtained from the ΔbpsA-D strain. These results demonstrate that Bp releases Bps during laboratory growth.

Cell-free Bps increases the resistance of B. pertussis to AMPs
Next, we determined if Bps released into the culture medium would increase AMP resistance. First, WT bacteria were incubated with PmB in the presence of SS media or filtered supernatants from either the WT or ΔbpsA-D bacteria. Compared to the addition of either SS media or supernatant from the ΔbpsA-D strain (ΔbpsA-D sup), addition of supernatant from the WT

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide strain (WT sup) resulted in enhanced survival of the WT strain by PmB ( Fig 4B). We also tested if Bps purified from bacterial cells would increase bacterial survival. Addition of purified Bps (Bps prep) increased survival of the ΔbpsA-D bacteria in the presence of LL-37. In contrast, addition of a mock-purified preparation from the ΔbpsA-D strain did not have any significant effect on the survival of the ΔbpsA-D strain ( Fig 4C). Collectively, these results suggest that the addition of cell-free Bps increases AMP resistance of Bp.'

Cell-free Bps binds to PmB and LL-37
We used immunoblotting and ELISA to test whether cell-free Bps bound PmB and LL-37, respectively. For immunoblots, different amounts of Bps prep or mock prep were spotted on a nitrocellulose membrane, followed by incubation with PmB. After extensive washing, the bound PmB was detected by incubation with a mouse α-PmB primary antibody followed by a goat α-mouse HRP-conjugated secondary antibody. PmB bound to the Bps prep but not to the mock prep ( Fig 4D). Similarly, LL-37 bound to Bps in a dose-dependent manner as detected by ELISA, whereas the mock prep showed very weak binding which did not increase with increasing amounts of LL-37 ( Fig 4E). Collectively, these results suggest that cell-free Bps limits the killing activity of AMPs by binding to them.

Presence of the WT strain increases resistance to LL-37-mediated killing in vitro and promotes respiratory tract survival of the ΔbpsA-D strain
Based on the finding that cell-free Bps protects against AMP-mediated killing, we hypothesized that the WT strain will protect the ΔbpsA-D strain from AMP-mediated killing. To test this in vitro, we incubated monocultures or co-cultures (1:1) of the two strains with LL-37 and enumerated CFUs. When incubated with the WT strain, the susceptibility of the ΔbpsA-D strain to LL-37 was nearly identical to that of the WT strain ( Fig 5A).

Fig 5. The presence of the WT strain increases ΔbpsA-D resistance to LL-37 in vitro and enhances its survival in the mouse respiratory tract and Bps is released in mouse lungs during infection. (a) Survival of WT and
ΔbpsA-D strains either in monoculture or in a 1:1 co-culture of both strains against LL-37. Each data point represents the mean and s.e.m. of triplicates from one of five independent experiments. Statistical differences were assessed by two-way ANOVA. �� , p<0.005; ��� , p<0.0005. n.s. not significant. (b) Bacterial CFUs recovered from the nasal septum and lungs of C57BL/6J mice four days after aerosol infection with WT or ΔbpsA-D in monoculture or in a 1:1 co-culture. Bars indicate the mean and s.e.m. of ten mice. Data from two independent experiments with groups of five mice each are shown. Statistical differences were assessed by two-way ANOVA. � , p<0.05; �� , p<0.005; ��� , p<0.0005; n.s., not significant. Dotted line represents the lower limit of detection for nasal septum, and dashed line represents the lower limit of detection for lungs. (c) Amounts of Bps in supernatants of lung lysates obtained from mice either instilled with PBS (blank circles) or infected with the indicated strains was quantified by ELISA using WGA conjugated to HRP. For infected mice, bars indicate the mean and s.e.m. of two independent experiments consisting of groups of five mice each (from Fig 5B). For PBS-instilled mice, bars indicate the mean and s.e.m. of one experiment consisting of four mice. Statistical differences were assessed by one-way ANOVA. �� , p<0.005; ��� , p<0.0005. https://doi.org/10.1371/journal.ppat.1010764.g005

Innate immunity and Bordetella pertussis Bps polysaccharide
We then tested if a similar survival advantage will exist for the ΔbpsA-D mutant in the mouse respiratory tract when co-infected with the WT strain. C57BL/6J mice were infected by aerosol exposure with the WT or the ΔbpsA-D strain either as single strains or when combined in a 1:1 ratio. Bacterial burden was determined by enumeration of CFUs from the nasal septum and lungs at four days post-challenge. Enumeration of CFUs approximately 30 minutes after aerosol challenge confirmed that similar numbers of the two strains were delivered into the nose and the lungs (S4 Fig). Mice infected with only the ΔbpsA-D strain (gray circles) harbored significantly lower bacterial burden on the nasal septum and in the lungs compared to mice infected with the WT strain alone (black circles) (Fig 5B). This result is consistent with our previously published results using the intranasal challenge route [6]. Strikingly, when coinfected with the WT strain, the bacterial burden of the ΔbpsA-D strain (grey squares) on the nasal septum and in the lungs was similar to that of the WT strain (black squares; Fig 5B).
These results indicate that the presence of the WT strain supports the colonization of the ΔbpsA-D strain in the mouse nose and lungs.

B. pertussis releases Bps in the mouse lungs
Lung lysates collected from mice were centrifuged and filtered, followed by quantification of Bps in the supernatants by ELISA. Bps was detected in lung supernatants of mice infected with the WT strain (black circles; Fig 5C). In comparison, Bps levels in the lung lysates of mice infected with the ΔbpsA-D strain (gray circles) were similar to those of mice instilled only with PBS (clear circles). We speculate that this background level of Bps reactivity is due to the crossreactivity of the WGA with host carbohydrates [30]. These results suggest that Bps released from the WT strain contributes to the increased bacterial burden of the ΔbpsA-D strain in the mouse respiratory tract.

Production of Bps in E. coli confers resistance to PmB and LL-37, enhances bacterial survival in the mouse respiratory tract, and augments pathology in the lungs
We then tested whether Bps as the sole Bordetella factor was sufficient to provide resistance to AMP-mediated killing in E. coli. Since E. coli harbors the pga locus (a polysaccharide locus with high sequence homology to the bpsA-D locus [28], we utilized a derivative of MG1655 (E. coli K12 strain) that lacks this locus. This strain was named ARF001. The pMM11 plasmid which encodes Bps was transformed into ARF001 [31][32][33][34]. Compared to the strain containing the empty vector (ARF001 vec ), the strain containing pMM11 (ARF001 bpsA-D ) survived better in the presence of PmB (Fig 6A) and LL-37 ( Fig 6B). We also confirmed using FITC-labeled LL-37 that production of Bps in E. coli inhibited AMP binding (Fig 6C-6D). These results suggest that Bps confers resistance to PmB and LL-37 in E. coli independent of other Bp factors by inhibiting their binding. C57BL/6 mice were infected utilizing both intranasal and aerosol routes [35] with either the ARF001 vec or ARF001 bpsA-D strain and CFUs were enumerated from the nasal septum and lungs four days post-challenge. Compared to mice infected with the ARF001 vec strain, mice intranasally infected with the ARF001 bpsA-D strain had approximately 100-and 400-fold higher bacterial burden in the nasal septum and lungs, respectively (Fig 6E). Upon infection of mice by the aerosol route, the ARF001 vec bacteria were recovered from the nasal septum of only one mouse at the lower limit of detection (3 CFUs) (Fig 6F). No bacteria were recovered from the lungs of any of the mice infected by aerosol. This suggests that ARF001 vec strain is unable to survive in the mouse nose and lungs when infected by the aerosol route. In comparison, the ARF001 bpsA-D bacteria were recovered from the nasal septum when infected by the aerosol

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide route ( Fig 6F). The observed decrease in colonization in the nasal septum and lungs after aerosol infection was not due to the inability of bacteria to reach these sites as enumeration of CFUs approximately 30 minutes after aerosol infection resulted in the recovery of both the ARF001 vec and ARF001 bpsA-D strains from the nasal septum and lungs (S5A Fig). However, the number of bacteria recovered 30 minutes after aerosol infection was lower than that recovered 30 minutes after intranasal infection (S5B Fig), explaining the differences in bacterial burden between intranasal and aerosol infections.
Evaluation and semi-quantitative lesion scoring performed on lungs from mice intranasally infected with ARF001 vec (Fig 7A-7D) strain revealed mild neutrophilic interstitial pneumonia, mild to moderate thickening of the pulmonary interstitium, and moderate numbers of neutrophils infiltrating the interstitium and bronchioles. In contrast, lungs from mice infected ARF001 bpsA-D (Fig 7E-7H) were characterized by a marked neutrophilic and macrophagic pneumonia with significant regional consolidation and thickening of the interstitium. In central areas of consolidation, very large numbers of neutrophils and macrophages obscured the distinction between alveoli and interstitium, and multifocally, alveolar walls were lined by hyperplastic type II pneumocytes, a common pulmonary response to injury. BALT was similarly expanded in this group of lungs adjacent to bronchioles. Degeneration and necrosis, edema, and hemorrhage were similar in all examined lung specimens. Total histopathology scores from mice infected with ARF001 bpsA-D were significantly higher than total scores from

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide mice infected with ARF001 vec (Fig 7I). Taken together, these results suggest that the production of Bps in E. coli is sufficient to impart the ability to colonize the mouse respiratory tract and induce pathology in the lungs.

Discussion
Successful prevention of microbial infection and disease necessitates a deeper understanding of the mechanisms by which microorganisms avoid host immunity. AMPs constitute a major component of the first line of immune defense in the mammalian respiratory tract. However, the respiratory pathogen Bp multiplies rapidly upon natural human infection and experimental infection of laboratory animals and is thus capable of overcoming these defenses [36,37]. In this manuscript, we investigated the role of the Bps polysaccharide in protecting Bp against AMP-mediated killing. We demonstrated that compared to the WT strain, the isogenic ΔbpsA-D mutant strain was more sensitive to PmB and several structurally diverse human AMPs. Treatment of the WT strain by DspB, a glycoside hydrolase that specifically cleaves Bps, also enhanced AMP sensitivity. Consistent with a positive role for Bps in AMP resistance, overproduction of Bps in the ΔbpsA-D strain by utilizing multiple copy plasmid complementation enhanced resistance to both PmB and LL-37. The pathogenic impact of elevated production of Bps in Bp is currently under investigation.
We have discovered two different strategies by which Bps confers resistance to AMPs. First, we demonstrated that while the WT strain bound PmB and LL-37, the ΔbpsA-D mutant strain

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide bound relatively higher amounts of these AMPs. Because of high affinity for LPS, interaction of AMPs with the gram-negative bacterial surface results in the displacement of the divalent cations (Ca 2 + and Mg 2 + ) that bind to the phosphate groups of the inner core of LPS. Because of their bulky nature, AMP binding then results in destabilization of the bacterial membrane [38]. In B. bronchiseptica and other bacteria, the LPS O-antigen promotes AMP resistance by limiting the interaction of AMPs with the bacterial surface [39,40]. Unlike B. bronchiseptica, Bp does not produce O-antigen [41]. We propose that in the absence of the O-antigen, Bps functions to inhibit AMP interaction with LPS and blocks their bactericidal effects. PNAG/ PGA polysaccharides contain variable amounts of de-N-acetylated glucosamine residues, rendering these polymers positively charged [8,42,43]. The deacetylation status of Bps is not known. Previously, the PIA polysaccharide of Staphylococcus epidermidis was shown to protect against killing by LL-37, and it was suggested that electrostatic repulsion is a likely mechanism of this protection [42]. A similar mechanism can also be hypothesized for cell-associated Bps.
Second, we report that Bp releases Bps both in the medium during laboratory growth and in mouse lungs during infection. Incubation with both naturally released and purified Bps increased the resistance of bacterial cells to PmB and LL-37. Cell-free Bps bound to both PmB and LL-37. This suggests that cell-free Bps can sequester/trap AMPs by functioning as a sink and thereby neutralize their bactericidal activity. This property of Bps resembles that of decoy receptors found in mammalian systems, which bind specific growth factors or cytokines [44,45]. Rather than signaling or activating the receptor complex this binding resulted in inhibition because the signaling molecules were trapped and rendered inactive. While many PNAG polysaccharides are released during laboratory growth [8,42,43], this is the first time a naturally released member of the PNAG polysaccharide family has been found to be released during infection and to mediate protection against a host immune component by functioning as a decoy.
The dual shielding and decoy effect of Bps reported in this manuscript appears somewhat paradoxical. How can a polysaccharide simultaneously inhibit AMP binding in one form and promote binding in another form? A previous study investigated the conformations of a series of linear and cyclic oligosaccharides related to PNAG. Linear oligosaccharides were not rigid and adopted several different conformations. The cyclic di-, tri, and tetra-saccharides adopted a symmetrical flattened ring conformation whereas the larger cyclic oligosaccharides were characterized by complicated shapes, resembling twisted rings [46]. It is possible that conformational differences exist between cell-associated and cell-free Bps resulting in variable effects on AMP binding. We propose that the binding of AMPs to Bps and other PNAG polysaccharides is dependent on its cellular location and conformation and involves various molecular forces including ionic forces.
Previously when inoculated intranasally with single strains and compared to the WT strain, the ΔbpsA-D strain survived at considerably lower numbers in the mouse respiratory tract [6]. In the current report, we obtained similar results when mice were infected with single strains using the aerosol route. Further, we observed that when mice were infected by the aerosol route with a 1:1 mixture of the WT and ΔbpsA-D strains, the numbers of the mutant strain harvested from the nasal septum and lungs were similar to that of the WT strain. We also observed increased survival of the mutant strain when incubated with LL-37 in the presence of the WT strain. We propose that Bps released by the WT strain in the mouse respiratory tract can sequester AMPs and is a major contributor to the enhanced survival of the mutant strain in the presence of the WT strain. Additionally, as evident from our AMP binding data, a proportion of AMPs could also be sequestered by binding to Bps on the cell surface of WT cells.
Production of Bps as a single Bp virulence factor in a laboratory-derived non-pathogenic E. coli strain conferred the ability to resist killing by PmB and LL-37. Like that observed in Bp,

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide production of Bps inhibited AMP binding to E. coli. Production of Bps also increased the survival of E. coli in the mouse nose when mice were infected by the aerosol route, and in both the nose and lungs when infected by the intranasal route. Histological analyses of the lungs showed marked enhancement of pathology when mice were intranasally infected with the Bps-producing E. coli strain. We consider these findings to be quite striking, since continuous laboratory passage and culturing of E. coli has led to several genetic mutations, deletions, and loss of surface structures resulting in a non-pathogenic organism which is not suited to a life inside the host or outside a laboratory [47,48]. Furthermore, E. coli is rarely considered to be a respiratory bacterium, as most pulmonary infections by E. coli are a result of dissemination from the gastrointestinal or urinary tract and from oropharyngeal aspirations [49].
In conclusion, Bps promotes resistance against AMP-mediated killing of Bp by functioning as a dual surface shield and decoy. The current findings further corroborate Bps as a principal determinant of Bp virulence and should serve as a model for investigating the pathogenic roles of other PNAG polysaccharides, which are conserved in many microbial species. Since Bps is not a component of current acellular pertussis vaccines, a vaccine containing Bps could control the resurgence of pertussis. Finally, Bps is a rare bacterial factor in that it confers virulence by itself in animal models of infection.

Ethics statement
Housing, husbandry, and experiments with animals were carried out in accordance with the guidelines approved by the Institutional Animal Care and Use Committee of The Ohio State University. C57BL/6J mice (Jackson; male and female, 6 to 12 weeks old) were bred in-house. All experiments were reviewed and approved by The Ohio State University Institutional Animal Care and Use Committee (Protocols #2017A00000090 and #2021A00000069).

Antimicrobial peptide killing assays
Bp and E. coli strains were grown to an OD 600 = 1.0 in SS or LB broth, respectively. Bacterial cells were harvested by centrifugation (13,500 rpm, 5 minutes), washed twice with sterile PBS, and resuspended in 10mM sodium phosphate buffer (pH 7.0). Dilutions of AMPs were prepared in 10mM sodium phosphate buffer, and bacterial cells were incubated with indicated concentrations of AMPs rotating at 37˚C for 2 hours. CFUs were enumerated by spotting 10 μl of 10-fold serial dilutions onto BG agar plates with Sm (Bp) or LB agar plates (E. coli) with Cm. Percent survival was determined by dividing the number of CFUs recovered after AMP treatment by the number of CFUs recovered from only buffer added controls.
For co-culture killing assays, WT and ΔbpsA-D strains were mixed in a ratio of 1:1 in 10mM sodium phosphate buffer (pH 7.0) and incubated with LL-37 at 37˚C rotating for 2

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide hours. We recently discovered that while the WT strain is resistant to both streptomycin and nalidixic acid, absence of the bpsA-D locus results in sensitivity to nalidixic acid (S6 Fig). Therefore, we took advantage of the differential sensitivities of the two strains to nalidixic acid to track their survival in co-culture experiments. Bacterial suspensions were plated on both BG agar with Sm (to enumerate both WT and ΔbpsA-D strains) and BG agar with Nal (to enumerate only the WT strain). CFUs of ΔbpsA-D strain were calculated by subtracting the counts obtained on BG agar with Nal from the counts obtained on BG agar with Sm.
To determine the role of spent culture supernatant in AMP protection, bacteria grown to an OD 600 = 1.0 in SS medium were pelleted by centrifugation (13,500 rpm, 5 minutes). After adjustment of pH to 7.4, the culture supernatants were filtered through a .22 μm filter (Millipore, catalog no. SCGP0052) and stored at -20˚C. Bacteria were resuspended in either culture supernatants or SS broth. Bps prep or mock prep were purified as previously described [6]. AMP killing assays with E. coli strains were done under the same conditions as with Bp, and bacteria were enumerated on LB agar with Cm. For all AMP killing assays, at least two independent experiments were performed with n = 3.

Quantification of Bps by ELISA
Bp strains were grown to OD 600 = 1.0, harvested by centrifugation (13,500 rpm, 15 minutes), washed with sterile 1X PBS twice and resuspended in sterile 1X PBS. To quantify cell-associated Bps, 1 x 10 8 CFUs in 100 μl of PBS were incubated overnight at 4˚C in 96-well plates (Corning, NY), washed three times with PBST (Tween 20) followed by blocking with 5% milk at 37˚C for 1 hour. After washing three more times with PBST, the plates were incubated with wheat germ agglutinin (WGA) conjugated to horseradish peroxidase (HRP) (Biotium Inc. Hayward, CA, USA) at a dilution of 1:1000. After incubation at 37˚C for 1 hour followed by five washes with PBST, 100 μl of 3,3',5,5'-tetramethylbenzidine (TMB, Sigma Aldrich, St. Louis, MO, USA) were added to each well. The reaction was stopped using 100μl of 1M H 2 SO 4 .
To quantify released Bps, 1 ml of bacterial culture corresponding to OD 600 = 1.0 was centrifuged (13,500 rpm, 15 minutes). Then, the supernatant was carefully aliquoted, filtered through a .22 μm filter (Millipore, catalog no. SCGP00525), and stored at -20˚C for later use. 100 μl of supernatants were added to each well of 96-well plates and the plates incubated overnight at 4˚C. ELISA assays were performed as described above. Six independent experiments were performed with n = 8.
To quantitate Bps from mouse lungs, homogenized lung lysates (for preparation of mouse lung lysates, please see below) from PBS-inoculated or Bp-infected mice were centrifuged (5000 rpm, 5 minutes), the supernatant carefully aliquoted, filtered through a .22 μm filter, and stored at -20˚C for later use. 300 μl of supernatants were added to each well of 96-well plates and Bps was quantitated by ELISA as described above. Two independent experiments were performed with n = 5.

Dispersin B treatment
Dispersin B (DspB) was purified as previously described [27]. Bacterial cells corresponding to OD 600 = 1.0 were harvested by centrifugation (13,500 rpm 5 minutes), washed twice with 1X PBS, and resuspended in DspB buffer (20mM Tris base, pH 8.0; 500mM NaCl). DspB treatments were carried out with 1 x 10 8 CFUs at indicated concentrations by incubating at 37˚C rotating for 2 hours. DspB was heat-inactivated via incubation at 56˚C for 30 minutes. The effect of DspB on cell-associated Bps was quantified by ELISA as described above. To perform AMP killing assays after treatment with DspB, bacteria were washed twice with 1X PBS, then resuspended in 10mM sodium phosphate buffer and incubated with indicated concentrations of AMPs. Two independent AMP killing assays were performed in technical triplicate.

Binding of LL-37 to bacteria by flow cytometry
Bacteria were harvested by centrifugation (13,500, 5 minutes), washed and fixed in 4% paraformaldehyde (PFA) with overnight rotation at 4˚C. Cells were then washed with PBS, blocked with 1% BSA and probed with 1mM FITC-labelled LL-37 overnight rotating at 4˚C. After two washes with PBS, approximately 10,000 events were collected per sample by using a Cytek Aurora Flow Cytometer (Cytek, Fremont, CA). FloJo software was used for data analysis. Two independent experiments were performed with n = 3.

Binding of PmB to bacteria by ELISA
Bacteria were harvested by centrifugation (13,500, 5 minutes), washed, fixed in 4% paraformaldehyde (PFA), then coated on high-binding 96-well plates (Corning, NY) and incubated at 4˚C overnight. Next, plates were washed three times with PBST, blocked with 5% milk, and probed with 10 μg/ml polymyxin B at 4˚C overnight. Next, mouse monoclonal anti-polymyxin B antibody (ab40014) (Abcam) was added at a dilution of 1:1000 for one hour followed by three washes with PBST. The HRP-conjugated secondary antibody was added at a dilution of 1:1000. After incubation at 37˚C for 1 hour followed by five washes with PBST, 100 μl of 3,3',5,5'-tetramethylbenzidine (TMB, Sigma Aldrich, St. Louis, MO, USA) were added to each well. The reaction was stopped using 100μl of 1M H 2 SO 4 . Two independent experiments were performed with n = 3.

Binding of AMPs to purified Bps
Serial dilutions of LL-37 were coated on high-binding 96-well plates (Corning, NY) and incubated at 4˚C overnight. Next, plates were washed three times, blocked with 5% milk, and probed with 50 μg/ml Bps and a mock preparation as described [6]. WGA conjugated to HRP (Biotium) was used to quantify the amount of Bps bound to LL-37 by ELISA as described above. Two independent experiments were performed with n = 6.
For PmB binding, 10 μl of 1:2, 1:5 and 1:10 dilutions of 10 mg/ml stock of purified Bps preparations or mock preparations strain were spotted on a nitrocellulose membrane and allowed to dry overnight. The membrane was blocked in 5% milk for 30 minutes followed by incubation with 25 μg/ml of PmB for one hour. Next, mouse monoclonal anti-polymyxin B antibody (ab40014) (Abcam) was added at a dilution of 1:1000 for one hour followed by three washes with TBST. The HRP-conjugated secondary antibody was added at a dilution of 1:5000 for one hour, followed by three washes with TBST and detection by the ECL system.

Mouse models of infection
Male and female C57BL/6J (Jackson) mice were used for all experiments. For all mouse experiments, two independent experiments were performed with groups of five mice. For aerosol infection, mice were infected with 10 8 CFU/ml of Bp WT or ΔbpsA-D either in monoculture or mixed in a 1:1 ratio, E. coli ARF001 vec or ARF001 bpsA-D in monoculture, or PBS as a negative control in an Allied Schuco S5000 Nebulizer for 30 minutes. A cohort of infected mice was sacrificed within 30 minutes after aerosol infection to determine the initial bacterial burden. After designated times post-challenge, the nasal septum and lungs were harvested, mechanically disrupted in PBS + 1% casein using hand-held homogenizers [52], and Bp counts were enumerated on BG agar plates with Sm for single strain inoculum or by plating separately on BG agar

PLOS PATHOGENS
Innate immunity and Bordetella pertussis Bps polysaccharide with Sm and BG agar with Nal as described above. For mice infected with E. coli strains, CFUs were enumerated by plating on LB agar with Cm. No bacterial colonies were recovered from mice that were administered PBS.
For intranasal infection with E. coli, 5x10 7 CFUs in 50 μl (1X PBS) were intranasally administered to each mouse. A cohort of infected mice was sacrificed within 30 minutes after aerosol infection to determine the initial bacterial burden. After 3 days, the nasal septum and lungs were harvested from each mouse and bacterial counts were enumerated by plating LB agar with Cm.
A board-certified comparative pathologist (Dr. Corps) performed blinded semi-quantitative lesion scoring on n = 5 lungs infected with E. coli strains and routinely stained with hematoxylin and eosin. Semi-quantitative lesion scores addressed the following parameters (S2 Table) and were devised based on numerous previously published methodologies and in reflection of lesions present in the experimental cohort: degree of cellularity and consolidation of lung tissue as a percent of total tissue; thickness of alveolar walls; degeneration or necrosis in any portion of the examined lung; presence of edema; presence of hemorrhage; percent of examined alveolar and interstitial tissue infiltrated by neutrophils; percent of bronchioles infiltrated by neutrophils; percent of examined alveoli distant to the lesion center containing alveolar macrophages; and perivascular or peribronchiolar expansion of lymphoid populations +/plasma cells (bronchioalveolar lymphoid tissue, BALT).

Statistics
Statistical analyses of results were performed by unpaired two tailed t-test, one-way ANOVA, two-way ANOVA and Bonferroni posttest. All statistical analyses were performed using GraphPad Prism 7.05.