d-Alanine Modification of a Protease-Susceptible Outer Membrane Component by the Bordetella pertussis dra Locus Promotes Resistance to Antimicrobial Peptides and Polymorphonuclear Leukocyte-Mediated Killing

Abstract

Bordetella pertussis is the causative agent of pertussis, a highly contagious disease of the human respiratory tract. Despite very high vaccine coverage, pertussis has reemerged as a serious threat in the United States and many developing countries. Thus, it is important to pursue research to discover unknown pathogenic mechanisms of B. pertussis. We have investigated a previously uncharacterized locus in B. pertussis, the dra locus, which is homologous to the dlt operons of Gram-positive bacteria. The absence of the dra locus resulted in increased sensitivity to the killing action of antimicrobial peptides (AMPs) and human phagocytes. Compared to the wild-type cells, the mutant cells bound higher levels of cationic proteins and peptides, suggesting that dra contributes to AMP resistance by decreasing the electronegativity of the cell surface. The presence of dra led to the incorporation of d-alanine into an outer membrane component that is susceptible to proteinase K cleavage. We conclude that dra encodes a virulence-associated determinant and contributes to the immune resistance of B. pertussis. With these findings, we have identified a new mechanism of surface modification in B. pertussis which may also be relevant in other Gram-negative pathogens.

INTRODUCTION

Whooping cough, or pertussis, a highly communicable infection, is caused by the Gram-negative bacterium Bordetella pertussis. The incidence of pertussis is increasing steadily in the United States, leading the CDC to classify pertussis as a reemerging disease (1–4). In 2012, pertussis epidemics were declared in Wisconsin, Vermont, and Colorado, and multiple states witnessed an increased incidence of pertussis higher than the national incidence. Overall, in the United States, more than 41,000 cases and 18 pertussis-related deaths have been reported for 2012 (3; www.cdc.gov). Globally, in 2008, 16 million cases of pertussis and 195,000 deaths were estimated to have occurred by the World Health Organization. Although infants are the primary targets of pertussis, adolescents and adults constitute 60% of the reported cases in the United States (5). It is now broadly accepted that these individuals serve as sources of transmission to infants and young children (6–9). The reemergence of pertussis calls for intensified research efforts to discover new pathogenic mechanisms of B. pertussis. Specifically, identification and comprehension of additional immune resistance determinants are essential for a detailed understanding of its virulence and for the development of novel vaccines and other therapeutic agents.

Antimicrobial peptides (AMPs) are one group of innate immune effectors produced by myeloid-derived host defense cells, such as macrophages and neutrophils, and by the skin and mucosal epithelia (10, 11). AMPs are broad-spectrum antimicrobials and display potent microbicidal activities against bacterial pathogens (11–14). AMPs largely exert their antimicrobial activities by damaging bacterial membranes and forming pores, thereby resulting in the efflux of essential ions and nutrients and disruption of membrane potential (15).

B. pertussis displays differential susceptibilities to AMPs isolated from different organisms (16–18). While cecropins and porcine AMPs (pBD-1 and PG-1) are highly effective in killing of B. pertussis, human AMPs HNP-1 and hBD-2 are relatively less effective. BrkA and BapC, two autotransporter proteins of B. pertussis, have been implicated in providing resistance to killing by a single AMP, cecropin P1 (17, 19). In both these studies, neither the brkA nor the bapC mutant strain was found to be more susceptible to any other AMPs tested. A B. pertussis factor that promotes resistance to human AMPs is not known. Additionally, the mechanisms by which BrkA or BapC confer resistance to cecropin P1-mediated killing are also not known.

Of the multiple ways employed by bacteria to resist the action of AMPs, surface alteration by enzymatic chemical modifications is a common theme in both Gram-positive and Gram-negative bacteria (12, 20). In many Gram-positive bacteria, d-alanyl esterification of teichoic acid by the genes of the dlt operons leads to a decreased negative charge on the bacterial cell surface (12, 20). This change in surface charge results in diminution of the interaction of innate immune effectors like cationic AMPs.

In this article, we report the investigation of a previously uncharacterized B. pertussis locus, the dra locus, which is homologous to the dlt loci of Gram-positive bacteria. We constructed a dra-deficient mutant of B. pertussis and compared the mutant and the wild-type (WT) strains with respect to cell morphology, growth characteristics, and susceptibility to innate immune components and cells. The mutant strain was more susceptible to killing by several human AMPs and by human polymorphonuclear lymphocytes (PMNs). We show that the dra locus is involved in the incorporation of d-alanine into an outer membrane (OM) component that is susceptible to proteinase K cleavage. Our findings uncover a unique mechanism of surface modification in B. pertussis which is distinct from the amino acid modification of polysaccharides by the homologous dlt loci.

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions.

Bacterial strains and plasmids used in this study are listed in Table 1. B. pertussis strains were maintained on Bordet-Gengou agar (BG) supplemented with 10% defibrinated sheep blood. Liquid cultures were grown in Stainer-Scholte (SS) broth with heptakis (2,6-di-O-methyl-β-cyclodextrin) (21, 22). Escherichia coli strains were grown in Luria-Bertani medium. As necessary, the various growth media were supplemented with the appropriate antibiotics: chloramphenicol (Cm; 10 μg ml⁻¹), kanamycin (Km; 25 μg ml⁻¹), and streptomycin (Sm; 50 μg ml⁻¹).

Table 1.

Bacterial strains and plasmids

Strain or plasmid	Characteristics	Reference or source
Strains
B. pertussis
Bp536	WT reference strain	Laboratory stock
Δdra	Bp536 derivative containing an in-frame deletion in the draABUDC locus	This study
Δdravec	Δdra strain containing vector plasmid pBBR1MCS	This study
Δdracomp	Δdra strain containing pNKT7, the draABUDC complementation plasmid	This study
E. coli DH5αλpir	F− ϕ80lacZΔM15 Δ(lacZYA-argF)U169 deoR recA1 endA1 hsdR17(rK− mK+) phoA supE44 thi-1 gyrA96 relA1 λ pir	62
SM10λpir	Conjugation strain	24
Plasmids
pBBR1MCS	Broad-host-range plasmid; Cmr	26
pSS4245	Allelic exchange vector; Kmr Ampr	25
pNKT4	draABUDC locus of Bp536 cloned in the pSS4245 vector	This study
pNKT7	draABUDC locus cloned in pBBR1MCS	This study

Molecular biology and bioinformatics.

Standard procedures were used for plasmid isolation, restriction digestion, cloning, and transformation. Conjugal transfer of plasmids to B. pertussis strains was performed with the E. coli SM10λpir strain (23, 24). Bordetella transconjugants were selected on BG agar containing the appropriate antibiotics. Sequence analysis was performed and homology was investigated by using the BLAST (available from the NCBI website and PATRIC) and Clustal W (available at the Biology Workbench at the San Diego Supercomputing Center) programs, and pairwise alignments were conducted at uniprot.org.

Deletion of the dra locus.

An in-frame nonpolar deletion in the dra locus in the B. pertussis Bp536 WT strain was constructed using a previously published allelic exchange method (25). A 378-bp MfeI-HindIII fragment spanning regions 5′ to and including the first 25 codons of draD was amplified from Bp536 genomic DNA using the primers dra5AMfeI and dra3AHindIII (Table 2). A 509-bp fragment containing regions 3′ to and including the last 17 codons of draA was similarly amplified using primers dra5BHindIII and dra3BBglII. A three-way ligation containing the MfeI-HindIII- and HindIII-BglII-digested PCR fragments along with the EcoRI-BamHI-digested allelic exchange vector pSS4245 (25) (a generous gift from Scott Stibitz, Center for Biologics Evaluation and Research, FDA) was carried out, resulting in plasmid pNKT4. This plasmid was then introduced into the Bp536 chromosome by mating with SM10λpir cells on BG agar containing 50 mM MgSO₄ (Bvg⁻-phase conditions) for 6 to 8 h. Cointegrants were selected and colony purified on BG-Sm-Km agar containing 50 mM MgSO₄. To allow secondary crossover events to occur, cointegrants were streaked on BG plates without MgSO₄ (Bvg⁺-phase conditions) and grown for 2 to 3 days at 37°C. Double-crossover recombinants were then restreaked on BG-Km and BG-Sm. Colonies that were streptomycin resistant but kanamycin sensitive contained putative deletions. The identity of the deletions was verified by PCR, followed by DNA sequencing of the PCR products.

Table 2.

Primers

Primer	Sequence (5′ to 3′)
dra5AMfe I	CCCGCAATTGTATATGCCTGACGAGGCAA
dra3AHindIII	CCCAAGCTTCGGGTGCACCAACGCACTGAG
dra5BHindIII	CCCAAGCTTGACCGCAAGAAGCTGCTGGAA
dra3BBglII	CGGGGTACCGCTGGCACGCGCCATCTACCA
cdltAFKpnI	CGGGGTACCGTATATGCCTGACGAGGCAA
cdltARXbaI	TGCTCTAGAGCTGGCACGCGCCATCTACCA

Genetic complementation of the dra locus.

A 4,611-bp fragment containing the entire dra locus plus 38 bp upstream of the putative translational start site of draC and 458 bp downstream of the termination codon of draA was amplified from Bp536 chromosomal DNA with primers cdltAFKpn1 and cdltARXba1 by PCR utilizing Pfu DNA polymerase. The resulting PCR fragment containing flanking KpnI and XbaI restriction sites was cloned into the corresponding sites of plasmid pBBR1MCS (26), resulting in the complementation plasmid pNKT7. This plasmid was transformed into DH5αλpir and subsequently mobilized into the Δdra strain as described above.

Antimicrobial peptide killing assays.

B. pertussis strains were grown to logarithmic phase (optical density at 600 nm [OD₆₀₀], ∼1.0) in SS medium at 37°C under shaking conditions. The bacterial cells were then harvested by centrifugation (5 min, 5,000 rpm), washed with 10 mM sodium phosphate buffer (pH 7.0), and resuspended in the same buffer. For AMP killing assays under physiological conditions, SS medium and Dulbecco modified Eagle medium (DMEM) were used instead of the sodium phosphate buffer. Serial dilutions of antimicrobial peptides were prepared in 10 mM sodium phosphate buffer. Bacteria (10⁶) were incubated with the concentrations of AMPs indicated below for 2 h at 37°C on a rotator, and appropriate dilutions were plated on BG agar plates containing Sm for colony counting. Percent survival was determined by dividing the number of CFU recovered after AMP treatment by the number of CFU recovered from nontreated controls. Each experiment was performed in triplicate.

Binding of cationic proteins and peptides.

The ability of B. pertussis strains to bind positively charged proteins and peptides was determined on the basis of the method previously reported (26). Briefly, stationary-phase bacteria were harvested by centrifugation, washed twice with phosphate-buffered saline (PBS), and resuspended in 0.1 M HEPES buffer, pH 7.0, to an OD₆₀₀ of 1.0. Cytochrome c (Sigma) was added to a final concentration of 500 μg/ml, whereas fluorescein-labeled LL-37 and fluorescein isothiocyanate (FITC)-labeled poly-l-lysine (Sigma) were added at a concentration of 20 μg/ml. After incubation for 10 min at room temperature, bacteria were centrifuged at 13,000 rpm for 5 min, washed twice with PBS, and then resuspended in PBS. To assess the relative amount of cytochrome c bound to each B. pertussis strain, the absorbance at 530 nm of the cell suspension was measured. The fraction of fluorescein-labeled LL-37 and FITC-labeled poly-l-lysine associated with the bacterial suspension was determined by measuring the fluorescence (excitation at 480 nm and emission at 520 nm for LL-37 and excitation at 500 nm and emission at 530 nm for poly-l-lysine).

To obtain LL-37 peptide with a single fluorescein at the N terminus, LL-37 (AnaSpec) was dissolved in distilled H₂O at 1 mg/ml, and 100 μl was incubated with 0.5 mM 5 (and 6)-carboxyfluorescein succinimidyl ester (Molecular Probes) for 1 h at room temperature in 50 mM potassium phosphate, pH 7.0. Fresh 5 (and 6)-carboxyfluorescein succinimidyl ester was then added to a final concentration of 1 mM and the reaction was allowed to proceed for another hour. Fluorescein-labeled LL-37 precipitated out of solution and was collected by centrifugation for 5 min at 14,000 rpm. The peptide was washed 3 to 4 times with 25 mM potassium phosphate, pH 7, and resolubilized in 150 μl 50% ethanol-water. The peptide molecular weight was determined using a Bruker Autoflex matrix-assisted laser desorption ionization–time of flight mass spectrometer using sinapinic acid as the matrix and indicated that 50 to 80% of the LL-37 peptide was singly labeled with fluorescein.

Preparation of total membranes for d-alanine incorporation assay.

Stationary-phase cultures of B. pertussis strains were centrifuged at 6,000 rpm for 20 min, and the cell pellets were resuspended in cell disruption buffer (10 mM Tris-HCl [pH 8.0], 20% sucrose, 1 mM EDTA, 0.1 mg/ml lysozyme, 1 mM phenylmethylsulfonyl fluoride). After incubation on ice for 10 min, the samples were frozen in dry ice and then thawed in cold water. The bacterial suspension was sonicated on ice and centrifuged at 12,000 rpm for 10 min to pellet unlysed cells. The clarified suspension obtained after the first centrifugation step was centrifuged at 25,000 rpm for 1 h, and the pellet (total membrane fraction) was resuspended in cold 10 mM Tris-HCl (pH 8.0). The protein content was estimated using the standard Bradford assay, and 100 μg of total protein was used for each sample of the d-alanine incorporation assay.

d-Alanine incorporation assay.

Incorporation of d-alanine was performed as described previously (27). The reaction mixture contained 25 mM MgCl₂, 5 mM ATP, 0.07 mM d-[¹⁴C]alanine (36 mCi/mmol), 30 mM Tris-HCl (pH 7.5), and 100 μg of the total membrane fraction in a final volume of 50 μl. The reaction mixtures were incubated at 37°C for 30 min, after which the reaction was terminated by addition of ice-cold buffer (5 mM Tris-Cl, pH 7.8, 10 mM MgCl₂). Samples were centrifuged at 30,000 rpm for 1 h at 4°C. The pellet was washed 3 to 4 times with 10 volumes of ice-cold termination buffer. The final washed membranes were resuspended in the termination buffer and filtered through a 0.45-μm-pore-size membrane filter, followed by washing of the filter one time with termination buffer. The filter was dissolved in the scintillation fluid, and radioactivity was measured in a Wallac 1209 Rack beta scintillation counter.

Cellular fractionation and isolation of cellular components.

For separation of inner and outer membrane components, the radiolabeled membranes were incubated with 2% Triton X-100 in the presence of 10 mM MgCl₂ for 30 min on ice and then centrifuged at 17,000 rpm for 1 h, as described by us earlier (28). The pellet (detergent insoluble) represented the outer membrane fractions, while the supernatant represented the inner membrane fractions. The relative radioactivity in each fraction was measured in a scintillation counter.

For extraction of polysaccharides, including lipooligosaccharide (LOS), the radiolabeled membranes were resuspended in 100 μl of 0.5 M EDTA, pH 8.0, and 2% SDS and boiled for 5 min at 100°C, followed by overnight treatment with proteinase K (1 mg/ml) at 37°C. Proteinase K was heat inactivated by incubating for 30 min at 90°C. To the aqueous solution, 2.5 volumes of ethanol and 0.3 M of sodium acetate were added. The mixture was incubated at −80°C for 2 h, followed by centrifugation at 16,000 rpm for 30 min. The supernatant was collected. The pellet was washed with 70% ethanol and air dried. The radioactivity associated with the pellet and that associated with the supernatant were separately measured in the scintillation counter.

Murein was extracted from the radiolabeled total membranes as the SDS-insoluble material (29). Radiolabeled membranes were resuspended in 8% SDS, boiled at 100°C for 30 min, and centrifuged at 130,000 rpm for 1 h. The insoluble residue was reextracted twice by boiling in 4% SDS. The murein sacculus prepared in this way was washed twice with water and once with 2 M NaCl. The final insoluble residue represented the SDS-resistant murine sacculus.

To demonstrate the integrity of the different extraction procedures, assays for specific components were performed. To confirm that pure outer membranes were obtained, the outer membrane pellet was separated by PAGE, followed by Western blotting to probe for BcfA (an outer membrane protein) and for BvgA (a cytoplasmic protein) (28). While BcfA was detected in the pellet fractions, BvgA was not detected (data not shown).

For determination of the LOS content in the pellet obtained after boiling in SDS-EDTA, the Limulus amoebocyte lysate (LAL) assay was conducted. The endotoxin concentration was found to be 0.35 μg/ml of culture at an OD₆₀₀ of the WT and Δdra strains.

For quantification of peptidoglycan, we measured the levels of diaminopimelic acid (DAP), an amino acid present in the cell wall of B. pertussis (30, 31). Briefly, the pellet containing the peptidoglycan fraction was acid hydrolyzed for 16 h at 95°C, followed by the addition of HCOOH and ninhydrin reagent to the hydrolyzed sample and incubation at 37°C for 1.5 h. The A₄₄₀ was measured and compared with standards to determine the DAP content. The pellet fraction contained 80 ng/ml of culture at the OD₆₀₀ of DAP. DAP was undetectable in the supernatant fraction.

PMN killing assay.

B. pertussis strains were grown to logarithmic phase (OD₆₀₀ = 1) and then harvested by centrifugation (8,000 rpm, 5 min at room temperature). Human PMNs were purified from the peripheral blood of healthy human donors by discontinuous plasma-Percoll centrifugation, in accordance with a Wake Forest School of Medicine Institutional Review Board-approved protocol. Bacteria (5 × 10⁶) resuspended in PBS were incubated with PMNs (at multiplicities of infection of 1:1 and 1:5) at 37°C for 1.5 h in a 24-well plate. After incubation, PMNs were lysed for 30 min on ice by the addition of 0.01% saponin. Bacterial cells were harvested from the wells by vigorous pipetting in 0.01% saponin and then plated as serial dilutions on BG agar plates containing streptomycin. The inoculum unexposed to PMNs served as the control (untreated bacteria). Percent killing was calculated using the following formula: 100 × [(c_u − c_t)/c_u], where c_u represents the number of CFU of untreated bacteria and c_t represents the number of CFU of bacteria treated with PMNs.

Microscopy.

For determining differences in morphology, the WT and Δdra strains were grown to stationary phase (OD₆₀₀ = 4). An aliquot of the bacterial suspension was spotted on a glass coverslip, heat fixed, and visualized with either a light or an electron microscope. For light microscopy, a ×100 oil immersion lens was used. Electron microscopic analysis was carried out by adsorbing bacteria onto carbon-coated gold grids (Electron Microscopy Sciences, PA) in a humidified chamber for 1 h, followed by fixing with 2.5% glutaraldehyde and negative staining with 2% phosphotungstic acid (pH 6.6). The images were analyzed with a Technai transmission electron microscope at a magnification of ×30,000.

RESULTS

The Bordetella dra locus.

Scanning of the genome sequence of B. pertussis strain Tohama I revealed the presence of homologs (BP2987 to BP2991) of the dlt loci of Gram-positive bacteria (Fig. 1). In addition to Tohama I, dlt homologs were also identified in the genomes of other classical Bordetella spp., B. bronchiseptica strain RB50 (BB4380 to BB4384) and B. parapertussis strain 12822 (BPP3907 to BPP3911) (32) (Fig. 1). In Gram-positive bacteria, the dltABCD loci catalyze the incorporation of d-alanine esters into teichoic acid (32–35). Gram-negative bacteria lack teichoic acid. Thus, the Bordetella loci must modify a different component of B. pertussis. Based on the phenotypic characterization presented below and to distinguish these loci from the dlt loci of Gram-positive bacteria, we have named these loci dra for their role in d-alanine incorporation and resistance to AMPs.

Fig 1.

Comparison of dra and dlt loci in Bordetella and other Gram-positive and Gram-negative bacteria. Shown are the organizations of the dra and related operons in different bacterial species. The species shown in the figure (in order of relatedness to the B. pertussis dra locus) and used in sequence comparisons are B. pertussis strain Tohama I, B. bronchiseptica strain RB50, B. parapertussis strain 12822, Achromobacter xylosoxidans A8, B. avium 197N, Acidovorax avenae subsp. avenae ATCC 19860, Delftia acidovorans SPH-1, Dickeya dadantii Ech586, Enterobacter cloacae SCF1, Pectobacterium wasabiae WPP163, Bacillus cereus ATCC 14579, Clostridium difficile 630, and Vibrio cholerae ATCC 39315. Homologous genes use the same patterns across species. The black arrow for V. cholerae represents a lipid A transacylase. The region of the dra operon deleted in the Δdra strain is shown at the top.

Sequence analysis of the gene products of the Bordetella dra locus.

The Bordetella dra loci consist of five open reading frames (ORFs) and have an identical gene organization. The five dra genes from the three species shown at the top of Fig. 1 encode proteins with 98 to 100% amino acid sequence identity between homologs. The order of the Bordetella dra genes (draCDUBA) differs from that of the dlt genes (dltABCD) generally found in Gram-positive bacteria, except in the case of Clostridium difficile, where the dlt genes are arranged in the order dltDABC (Fig. 1) (36).

On the basis of the activities of some of the proteins from the dlt loci of Gram-positive organisms, putative biochemical functions can be attributed to several of the Dra proteins. For example, the Bordetella draA genes are homologous (34 to 37% identity at the amino acid level for these 477- to 504-residue proteins) to the dltA genes, which code for DltA, an ATP-dependent d-alanine transferase and ligase. DltA both adenylates d-alanine (activating it with an AMP group) and transfers the alanyl group to the d-alanyl carrier protein, DltC (37).

draC, corresponding to dltC, is relatively more divergent (with ∼15% amino acid identity between DraC and DltC from Bacillus cereus) but is always 73 to 79 amino acids in length and aligns well with acyl carrier proteins in multiple-sequence alignments (including conservation around the phosphopantetheinylated Ser residue at position 37, within a DS motif including hydrophobic residues on each side). Thus, DraA and DraC are expected to play roles analogous to those of DltA and DltC in preparing the d-alanyl group for transfer.

draB encodes a 375-amino-acid protein homologous to the dltB gene products (with 28% amino acid identity between DraB and B. cereus DltB and with lengths ranging from 372 to 395 residues across species). On the basis of the proposed function of DltB, DraB may similarly act as a transmembrane protein supporting the transport of the activated d-alanyl group across the inner membrane (37). DltB/DraB proteins are in some cases annotated as membrane-bound O-acyltransferases.

draD, encoding a 375-amino-acid protein, is homologous to the dltD gene products (ranging from 375 to 395 residues across species). DraD aligns convincingly with DltD proteins (including a conserved GSSEXXXXD motif near the N terminus) but is more divergent from its Gram-positive counterparts than DraA and DraB (∼19% identity between DraD and DltD proteins). This may reflect the difference in the identity of the membrane-associated alanyl acceptor, which in Gram-positive organisms is lipoteichoic acid.

A distinctive feature of the Bordetella dra operons is the presence of a small, conserved ORF (BP2989/BB4382/Bpp3909) encoding a hypothetical 41-amino-acid protein. Some Gram-positive bacteria harbor small ORFs (e.g., dltX in C. difficile) near the dlt genes (Fig. 1) (32, 36, 38). The small ORF present in Bordetella spp. did not display any significant similarity to dltX or to any other bacterial ORFs. To distinguish it from the dltX genes and because of its unknown function, we have designated this ORF draU.

Deletion of the dra locus and growth characteristics.

To investigate the function of the dra locus in B. pertussis, we generated an in-frame deletion by utilizing allelic exchange, which resulted in the deletion of a 4.6-kb region of the dra locus (Fig. 1). Following construction of the mutant strain, we examined the effect of dra deletion on the phenotypic characteristics of B. pertussis. There were no notable differences in the growth rate between the WT and Δdra mutant when they were grown in SS broth, as the strains reached similar optical densities (data not shown). Compared to the WT strain, the Δdra mutant did not display any significant alterations in colony size or appearance on BG agar plates (data not shown). Aberrant cell shapes were also not observed either by light microscopy or by electron microscopy (Fig. 2A and B).

Fig 2.

Deletion of the dra locus has no effect on cell morphology of B. pertussis. Phase-contrast (A) or electron (B) micrographs of bacterial cells after 72 h of growth are shown.

The dra locus is required for resistance to human AMPs, proteins, and polymyxin B.

The microbicidal activities of human defense proteins, peptides, and the peptide antibiotic polymyxin B against the Bp536 WT strain of B. pertussis and the Δdra mutant were tested using viable count analysis. HNP-1 and HNP-2 are human proteins that belong to the α-defensin family of antimicrobial peptides. These are abundant in human neutrophils and are also expressed in the respiratory tract (39, 40). The SPLUNC1 (short palate, lung, nasal epithelium clone 1) protein is a crucial component of innate immunity and is highly expressed in the oral cavity and respiratory tract of humans and many other mammals (41–43). Human CAP18/LL-37 (hCAP18/LL-37) is a prototype member of the human cathelicidin family and is produced by neutrophils, macrophages, and various epithelial cells (44).

As shown in Fig. 3, Bp536 was markedly resistant to HNP-1, HNP-2, and human SPLUNC1 (hSPLUNC1). Concentrations as high as 50 μg/ml of HNP-1 and HNP-2 and 10 μg/ml of hSPLUNC1 resulted in little or no killing of Bp536. While higher concentrations of hSPLUNC1 (25 μg/ml) were effective in killing of Bp536, ∼34% of the input bacteria still survived (Fig. 3C). Compared to the results for these AMPs, lower relative amounts of LL-37 and polymyxin B were required to kill Bp536 (Fig. 3D and E). Taken together, these results confirm and expand the previous observations of B. pertussis displaying various levels of resistance to structurally different antimicrobial peptides (16–18).

Fig 3.

The dra locus promotes resistance to antimicrobial peptides. The susceptibilities of the WT, Δdra, Δdra^vec, and Δdra^comp strains to various antimicrobial peptides were assessed. Bacteria (1 × 10⁶ CFU) were incubated in 10 mM sodium phosphate buffer with the indicated concentrations of AMPs for 2 h. Bacterial numbers were determined by plate counts on BG agar. Each data point represents the mean and standard deviation of triplicates from one of three independent experiments. *, P < 0.05 compared with the untreated control based on Student's t test. The number 0 on top of the bars in some panels indicates the lack of any detectable bacterial growth. PmB, polymyxin B.

Compared to the WT strain, the Δdra mutant was killed in greater numbers at relatively lower concentrations of all the AMPs tested (Fig. 3A to E). Complementation of the dra locus in the mutant strain harboring plasmid pNKT4 containing the dra locus (Δdra^comp) increased the level of resistance of the mutant strain to LL-37 (Fig. 3D) and polymyxin B (Fig. 3E) toward that observed for the WT strain. As expected, the presence of the vector plasmid (pBBR1MCS) in the Δdra strain (Δdra^vec) did not have a significant effect on the susceptibility of the mutant strain. As expected from the growth experiments described above, there were no differences in the growth rates between the Δdra^vec and Δdra^comp strains (data not shown).

The AMP susceptibility assays described above were conducted using low salt concentrations (10 mM sodium phosphate), conditions typically used for AMP assays. Although several AMPs, including LL-37, are highly active in phosphate buffer, these display significant reductions in activity or are inactive in the presence of biological concentrations of monovalent and divalent cations and in tissue culture medium (45, 46). In an attempt to mimic these conditions, we tested the activity of LL-37 in DMEM and SS medium. DMEM is used as a standard tissue culture medium and contains physiological concentrations of Na⁺ ion (greater than 100 mM), Ca²⁺, and Mg²⁺. SS medium is used as an optimal growth medium for B. pertussis and has concentrations of Ca²⁺, Mg²⁺, and monovalent ions similar to those found in human lung secretions (16). Compared to 10 mM phosphate buffer (Fig. 3D), where low concentrations of LL-37 were effective in significant killing of the WT strain, concentrations as high as 5 μg/ml of LL-37 did not result in any observable killing in DMEM (Fig. 4A). In contrast, only 39%, 21%, and 19% of the mutant strain survived at 0.05, 1, and 5 μg/ml of LL-37, respectively (Fig. 4A). Similarly, larger amounts of LL-37 were needed for significant killing of the WT strain when incubated in the SS medium. In comparison, the mutant strain was more sensitive to killing by LL-37 than the WT strain in SS medium (Fig. 4B). Taken together, these results demonstrate that the dra locus promotes the AMP resistance of B. pertussis under low-salt conditions and under in vitro conditions which mimic those found in human lungs.

Fig 4.

Sensitivities of the WT and the Δdra strains to LL-37 under physiologically relevant conditions. Bacteria (1 × 10⁶ CFU) were incubated in either DMEM (A) or SS medium (B) with the indicated concentrations of LL-37 for 2 h. Bacterial numbers were determined by plate counts on BG agar. Data represent the mean and standard deviation of triplicates from one of three independent experiments. *, P < 0.05.

Absence of the dra locus results in increased binding to cationic protein and peptides.

One of the mechanisms by which bacteria resist cationic AMPs is by reducing the net negative charge of the bacterial cell surface, thereby repelling cationic AMPs (12, 47). To determine if the increased susceptibility of the Δdra mutant to AMPs was due to an altered surface charge, the abilities of the WT, Δdra, Δdra^vec, and Δdra^comp strains to bind cationic protein cytochrome c (47) and peptides (poly-l-lysine and LL-37) were compared. As shown in Fig. 5, the Δdra mutant bound significantly more of all three cationic molecules than the WT strain. As expected, the binding of these molecules to the Δdra^comp strain was lower than that of the Δdra and Δdra^vec strains. These results suggest that the dra locus contributes to the AMP resistance of B. pertussis by decreasing the net negative surface charge.

Fig 5.

Interaction of B. pertussis with positively charged protein and peptides. The B. pertussis strains were treated with cytochrome c (A), fluorescein-labeled LL-37 (B), or FITC-labeled poly-l-lysine (C) and washed twice with PBS, and then the cells were resuspended in PBS. The amount of peptide or protein associated with each B. pertussis cell suspension was measured by determination of the absorbance or fluorescence, as described in Materials and Methods. Data are expressed as the percentage of the signal from the B. pertussis suspension compared to the signal from the input protein or peptide. **, P < 0.01.

The dra locus is important for resistance to killing of B. pertussis by human PMNs.

The antimicrobial potential of PMNs depends in part on the microbicidal action of AMPs (48). Based on the in vitro hypersusceptibility of the Δdra strain to AMPs, we hypothesized that the dra locus promotes resistance to PMN-mediated killing. To test this hypothesis, we evaluated the survival of the WT and Δdra strains exposed to human PMNs. Compared to the WT strain, the Δdra strain was killed in greater numbers by PMNs (Fig. 6). Complementation of the dra locus in the mutant strain (Δdra^comp) resulted in enhanced survival, whereas the presence of the vector plasmid in the Δdra strain (Δdra^vec) did not have a significant effect on the survival of the mutant strain. These results suggest that dra promotes the resistance of B. pertussis to PMN-mediated killing.

Fig 6.

The dra locus promotes the resistance of B. pertussis to killing by human PMNs. B. pertussis WT, Δdra, Δdra^vec, and Δdra^comp strains were incubated separately with PMNs at multiplicities of infection of 1:1 and 1:5 for 1.5 h. Bacterial survival was calculated by the number of CFU recovered after incubation with PMNs divided by the number of CFU recovered from untreated controls. Each data point represents the mean and standard deviation of triplicate samples from one of three independent experiments. **, P < 0.01.

The dra locus is involved in the incorporation of d-alanine into B. pertussis.

Given the homology among gene products of the dra and dlt loci, we asked if dra is involved in the incorporation of d-alanine into B. pertussis. To answer this question, we utilized a previously described assay (27) that measures the incorporation of d-[¹⁴C]alanine into purified bacterial membrane fragments. Membrane fragments prepared from the Δdra mutant incorporated smaller amounts of d-alanine than membrane fragments prepared from the WT strain (Table 3). The Dra-dependent incorporation of d-alanine into B. pertussis membranes accounted for 54% of the total d-alanine incorporation (comparing the WT strain with the Δdra mutant). Boiled membranes from either the WT or the Δdra strain incorporated very small amounts of d-[¹⁴C]alanine, and the counts per minute incorporated were similar to the background levels of radioactivity (Table 3).

Table 3.

Incorporation of d[¹⁴C]alanine into various cellular fractions of B. pertussis WT and Δdra strains

Fraction or treatment	cpm (100 μg protein)a		P value
	WT	Δdra mutant
Total membranes	1,056 ± 7	484 ± 36	<0.005
Boiled membranes	48 ± 6	41 ± 7	NS
2% Triton X-100
Initial counts (total membranes)	1,208 ± 144	502 ± 62	<0.05
Supernatant	330 ± 42	247 ± 88	NS
Pellet	812 ± 66	384 ± 129	<0.05
8% SDS
Initial counts (total membranes)	2,137 ± 30	781 ± 45	<0.001
Supernatant	1,664 ± 24	632 ± 60	<0.001
Pellet	328 ± 11	135 ± 28	<0.001
SDS-EDTA boiling
Total membranes	1,056 ± 71	484 ± 36	<0.005
Supernatant	1,110 ± 62	482 ± 219	<0.05
Pellet	108 ± 66	85 ± 66	NS
Proteinase K
Initial counts (total membranes)	1,056 ± 71	484 ± 37	<0.005
Supernatant	587 ± 132	182 ± 50	<0.05
Pellet	184 ± 26	103 ± 22	<0.05

^aRepresentative data from one of at least two to three independent experiments are shown.

NS, not significant.

Membrane fractionation demonstrates the dra-dependent d-alanylation of a protease-susceptible outer membrane component of B. pertussis.

Four kinds of fractionation of the purified membranes were conducted to clarify the type of molecule that serves as the d-alanine acceptor in the dra system, as described below.

(i) Triton X-100 (2%) extraction.

The total radiolabeled membrane fragments were extracted with 2% Triton X-100, a reagent that leads to solubilization of the inner membrane components, leaving the remainder of the outer membrane (OM) components in the pellet. Comparison of the WT strain with the dra mutant revealed that essentially all of the dra-dependent d-alanine incorporation was found in the pellet, suggesting that an OM component of B. pertussis is modified by Dra-dependent d-alanylation (Table 3).

(ii) Boiling with 8% SDS.

d-Alanine is one of the components of the B. pertussis peptidoglycan (30, 31). To determine if the dra locus led to the incorporation of d-alanine into B. pertussis peptidoglycan, murein was extracted from radiolabeled membrane fragments as the 8% SDS-insoluble fraction after boiling (29). The majority of the d-alanine incorporated into membranes in the absence of dra (Δdra strain) was found in the supernatant after SDS extraction. In the WT strain containing the dra locus, while there was a moderate increase of approximately 200 cpm in the peptidoglycan-containing pellet, there was an increase of about 1,000 cpm in the supernatant. This result thus suggests that dra does not catalyze the incorporation of d-alanine into B. pertussis peptidoglycan.

(iii) SDS-EDTA boiling.

Boiling the membrane-containing pellet in the presence of SDS and EDTA followed by proteinase K treatment and precipitation with ethanol is a frequently utilized method to purify polysaccharides, including lipopolysaccharides (LPSs) from the pellet fraction (49). Very little of the radioactivity, if any, was incorporated in the pellet fractions obtained from the SDS-EDTA-extracted membrane preparations of either the WT or the dra mutant. This result indicates that d-alanine is not incorporated into either the LPS or other SDS-EDTA-extractable polysaccharides of B. pertussis.

(iv) Proteinase K treatment.

Incubation of the radiolabeled total membrane fragments with proteinase K, which cleaves peptide bonds, revealed that most of the increase in d-alanine incorporation in the presence of the dra locus (i.e., in the WT strain) was in the supernatant after protease treatment, suggesting that the acceptor molecule(s) modified by the dra system includes proteinase K-susceptible peptide bonds and may be either a protein or another peptide-containing complex macromolecule.

Taken together, these results suggest that the dra locus is mainly involved in the incorporation of d-alanine into a proteinase K-susceptible OM component that is not part of the B. pertussis peptidoglycan.

DISCUSSION

Antimicrobial peptides constitute a crucial first line of human innate host defenses. AMPs work in concert with other clearance and barrier mechanisms of epithelial surfaces to help maintain the lower respiratory tract and other mucosal surfaces free from infection. Despite these antibacterial strategies, upon entry into the host, pathogens easily circumvent these defenses and rapidly multiply, resulting in a full-blown infection. This is true also for B. pertussis, which, upon natural infection of humans and experimental infection of animals, rapidly replicates in the respiratory tract. In mice, B. pertussis is able to maintain infections in the mouse respiratory tract for longer than a month (50). One explanation for this is that like other bacterial pathogens, B. pertussis is able to counter the action of AMPs by having developed efficient resistance mechanisms.

We investigated in this work a previously uncharacterized B. pertussis locus that is similar in amino acid sequence to the sequences of dlt operons. Inactivation of the B. pertussis dra operon resulted in enhanced sensitivity to several host defense peptides, including (i) those with a β-sheet structure and disulfide bridges, such as HNP-1 and HNP-2; (ii) the α-helical peptide LL-37; (iii) the cyclic amphipathic peptide antibiotic polymyxin B; and (iv) hSPLUNC1, a human defense protein with structural similarities to the BPI family of proteins. The dra mutant was also strongly impaired in its ability to survive in human PMNs. Since LL-37, HNP-1, and HNP-2 are representative antibacterial peptides from neutrophil-specific granules, these results imply that the B. pertussis dra locus protects against neutrophil-mediated killing by conferring AMP resistance. The increased sensitivity of the dra mutant toward structurally diverse AMPs led us to carry out experiments which demonstrated that the absence of the dra locus reduced the binding of three positively charged proteins and peptides to the B. pertussis cell surface. Since the common structural feature of many of the AMPs examined in this study is a net positive charge, we propose that the mechanistic basis for the increased AMP sensitivity of the dra mutant is altered electrostatic AMP-cell interactions.

The results obtained in this study showed that the dra locus is not of major importance for the basic cell physiology of B. pertussis. Unlike dlt mutants of some Gram-positive bacteria (32, 38, 51, 52), morphological alterations like the presence of aberrant shapes were not observed. The dra mutant also did not show any significant differences in growth characteristics in standard B. pertussis growth medium. Taken together, these data suggest that growth and morphological differences are not responsible for the observed defect in the survival of the mutant strain in PMNs.

Earlier work by Abi Khattar et al. with the dlt system from Bacillus cereus included the identification of similar genes from Gram-negative organisms, including the three Bordetella species described here that infect mammals, as well as Erwinia carotovora subsp. atroseptica SCRI1043 and Photorhabdus luminescens TT01 (32). Our sequence searches have identified dra homologs in several additional Gram-negative pathogens with a wide host range comprised of plants, animals, birds, and humans (Fig. 1). In addition to the three Bordetella spp. that infect mammalian hosts, a dra homolog was also identified in another Bordetella spp., B. avium, which is the causative agent of bordetellosis in birds (53) (Fig. 1). Dra homologs were also identified in the betaproteobacterium Achromobacter xylosoxidans, an opportunistic human pathogen that has been linked to a variety of human diseases (54); the lambdaproteobacterium Dickeya dadantii, a phytopathogen (55, 56); and Enterobacter cloacae, a component of the normal flora of the gastrointestinal tract that causes opportunistic infections (57). Two additional betaproteobacteria (Acidovorax avenae and Delftia acidovorans) were found to include in their genomes only two of the five genes, which were similar to draA and draC (Fig. 1). Interestingly, a partially homologous system was recently identified in the Gram-negative pathogen Vibrio cholerae and included DltA and DltC-like proteins (designated AlmE and AlmF, respectively), yet no DltB or DltD-like components were encoded within the operons (58). Instead, an additional gene in this operon encodes AlmG, a lipid A transacylase that incorporates a glycyl (or glycyl-glycine) group into one of the fatty acid chains of lipid A in that organism. It is striking that this Alm system in V. cholerae is more distantly related to the Dra systems of other Gram-negative bacteria (∼20% amino acid identity between DraA and AlmE) than the Gram-positive Dlt system is to the Dra system (∼35% amino acid identity between DraA and DltA from B. cereus) (Fig. 7).

Fig 7.

Dendrogram illustrating the relatedness of DraA homologues from various bacterial species. A rooted dendrogram of homologous DltA/DraA/AlmE protein sequences from the various species described in the legend to Fig. 1 was generated by the Clustal W program and Phylip's Drawtree program using the Biology Workbench at the San Diego Supercomputing Center. Shown in parentheses are the amino acid identities after pairwise alignments (at uniprot.org) of B. pertussis DraA and the corresponding homologue from each species.

The discovery of these homologs in Bordetella and other Gram-negative bacteria begs the question about the identity of the surface molecule(s) that is being modified. While teichoic acids are common components of the cell wall membrane of a large number of Gram-positive bacteria, these are not found in Gram-negative bacteria.

We show herein that membranes from the Δdra mutant incorporated smaller amounts of d-[¹⁴C]alanine than membranes from the WT strain of B. pertussis did and that a large proportion of d-alanine was incorporated into the B. pertussis outer membrane. The major d-alanine-containing polymer in B. pertussis is peptidoglycan. However, this is expected to rely on a metabolic pathway that is distinct from the dra locus (30, 31, 59). In agreement with this, the fractionation of d-[¹⁴C]alanine-labeled membranes by boiling in 8% SDS revealed that little of the dra-specific radioactivity was contained in the SDS-insoluble peptidoglycan fraction. As is true for other Gram-negative bacteria, LPS is another major cell surface glycopolymer of Bordetella spp. d-Alanine was not identified as a component of the Bordetella LPS (60), and data obtained here demonstrate the lack of any significant incorporation of d-[¹⁴C]alanine into B. pertussis LPS and other polysaccharides extractable by SDS-EDTA. We found that a large portion of the membrane-incorporated radioactivity could be digested and released into solution by the serine protease proteinase K from pelletable membranes. Taken together, these results suggest that dra is involved in the incorporation of d-alanine into an outer membrane component that is susceptible to cleavage by proteinase K. In the Gram-positive organism B. subtilis, the presence of d-alanine in a covalent linkage to two unidentified cellular proteins was detected, but further characterization of these d-alanylated proteins was not pursued (61). Precise identification of the B. pertussis surface component modified by the dra locus awaits further experimentation.

In summary, our results constitute the identification of a previously unknown mechanism of immune resistance in a Gram-negative pathogen, B. pertussis. Continued research will enhance our understanding of infections caused not only by B. pertussis and other members of the Bordetella spp. but also by several other Gram-negative bacterial pathogens with a wide host range which may utilize similar modifications of their surface component as a means to develop immune resistance.