Expression of the Lipopolysaccharide-Modifying Enzymes PagP and PagL Modulates the Endotoxic Activity of Bordetella pertussis

Abstract

Lipopolysaccharide (LPS) is one of the major constituents of the gram-negative bacterial cell envelope. Its endotoxic activity causes the relatively high reactogenicity of whole-cell vaccines. Several bacteria harbor LPS-modifying enzymes that modulate the endotoxic activity of the LPS. Here we evaluated whether two such enzymes, i.e., PagP and PagL, could be useful tools for the development of an improved and less reactogenic whole-cell pertussis vaccine. We showed that expression of PagP and PagL in Bordetella pertussis leads to increased and decreased endotoxic activity of the LPS, respectively. As expected, PagP activity also resulted in increased endotoxic activity of whole bacterial cells. However, more unexpectedly, this was also the case for PagL. This paradoxical result may be explained, in part, by an increased release of LPS, which we observed in the PagL-expressing cells.

The genus Bordetella includes several well-known pathogens that are capable of colonizing the upper respiratory tracts of mammals. Bordetella pertussis and the majority of Bordetella parapertussis strains are obligate human pathogens that usually cause acute respiratory tract diseases. Bordetella bronchiseptica has a much broader host range and is capable of colonizing rabbits, rats, guinea pigs, dogs, and piglets (18). Normally, B. bronchiseptica does not infect humans; however, some cases of infection in elderly or immunocompromised individuals have been described (3, 19, 45).

The majority of virulence factors in the bordetellae are regulated by the two-component system BvgAS (34, 52). In response to environmental stimuli, the BvgAS system regulates the transcription of several target genes (13). Bacteria in the Bvg⁺ phase express virulence factors, such as filamentous hemagglutinin, pertactin, and pertussis toxin. During the Bvg⁻ phase, the majority of virulence factors are down regulated, whereas other genes (for example, those that are required for motility) are up regulated (1). Some virulence genes, including those encoding filamentous hemagglutinin and fimbriae, are expressed also in a Bvg-intermediate phase, known as Bvgⁱ (14).

Lipopolysaccharide (LPS) is one of the major components of the gram-negative bacterial outer membrane. It is usually composed of a highly variable O antigen, a less variable core oligosaccharide, and a highly conserved lipid moiety designated lipid A. The structure of lipid A is well conserved among gram-negative bacteria and consists of a glucosamine disaccharide substituted with one or two phosphate groups and a variable number of acyl chains (Fig. 1). LPS is also known as endotoxin, due to its endotoxic activity when administered to higher organisms. This endotoxic activity is based on recognition of the LPS by the innate immune system. LPS is first bound by the LPS-binding protein (LBP), after which it is transferred, with the help of CD14, to the Toll-like receptor 4 (TLR4)/MD-2 complex. Activation of the TLR4/MD-2 complex leads to a downstream signaling cascade and eventually to increased production of proinflammatory cytokines, such as interleukin-6 (IL-6) (reviewed in reference 39). The endotoxic activity of LPS is determined entirely by the composition of the lipid A moiety (23). Previous studies have shown that the phosphate groups as well as the number and length of the acyl chains are critical determinants of the endotoxic activity of lipid A (31, 44, 54). Besides its endotoxic activity, LPS also has a powerful adjuvant activity and in that respect is potentially an interesting vaccine component.

FIG. 1.

B. pertussis lipid A architecture and modification by PagP and PagL. B. pertussis lipid A consists of a bisphosphorylated glucosamine disaccharide substituted with three R-3-hydroxytetradecanoic acid moieties and one R-3-hydroxydecanoic acid moiety, of which the 2′ fatty acyl chain is esterified with myristate. B. pertussis PagP [PagP_(Bp)] and E. coli PagP [PagP_(Ec)] add a secondary palmitoyl chain at the 3′ and 2 positions, respectively. Removal of the 3-hydroxydecanoic acid moiety at the 3 position is mediated by PagL.

The role of LPS in Bordetella pathogenesis and immunity is unclear. It has been shown that Bordetella LPS is highly immunogenic, acts as an immunological adjuvant, and exhibits the properties that are expected of an endotoxin (4, 12, 61). Furthermore, Bordetella LPS and its recognition by the TLR4/MD-2 complex have been shown to be important determinants for the course of infection in mice (6, 32, 33). B. pertussis can express two forms of LPS, i.e., band A and band B (40). Band B LPS is composed of lipid A and a core oligosaccharide. Addition of a terminal trisaccharide to band B LPS forms the LPS referred to as band A. In contrast to B. parapertussis and some strains of B. bronchiseptica, B. pertussis lacks a repetitive O-antigen structure. In B. pertussis, lipid A is substituted via ester or amide linkage with R-3-hydroxytetradecanoic acid (3OH-C₁₄) residues at positions 2, 2′, and 3′ and with an R-3-hydroxydecanoic acid (3OH-C₁₀) residue at the 3 position via ester linkage. A secondary myristoyl group substitutes the hydroxyl group of 3OH-C₁₄ at the 2′ position (Fig. 1) (10). Limited information on the genetics of B. pertussis LPS biosynthesis is currently available. Detailed studies have been performed only on the wlb locus, which is required for addition of the band A trisaccharide (2).

In recent years, several enzymes capable of modifying mature lipid A in gram-negative bacteria have been identified. One of them is the palmitoyl transferase PagP, which adds a secondary palmitoyl chain (C₁₆) at the 2 position of lipid A (7). It was recently shown that treating Escherichia coli with EDTA leads to activation of PagP via a membrane perturbation mechanism (25). Palmitoylation of lipid A leads to an increased resistance of the bacterium to cationic antimicrobial peptides (20). Furthermore, palmitoylated LPS is attenuated for signaling through TLR4/MD-2 (24). A PagP homolog, which, in contrast to E. coli and Salmonella PagP, adds a secondary palmitoyl chain at the 3′ position of lipid A (Fig. 1), has been identified in B. bronchiseptica (43). In B. pertussis, however, expression of PagP seems to be abolished through insertion of an insertion sequence element in the promoter region (43). Another lipid A-modifying enzyme, PagL, was discovered in Salmonella enterica serovar Typhimurium and shown to hydrolyze the ester bond at the 3 position of lipid A, thereby releasing the primary 3-hydroxymyristoyl moiety (58). In S. enterica serovar Typhimurium, PagL activity can be inhibited by aminoarabinose modification of lipid A (28). Deacylation by PagL modulates the recognition of lipid A by the TLR4/MD-2 complex (27). We recently reported the identification of PagL homologs in a wide range of gram-negative bacteria, including B. pertussis, B. parapertussis, and B. bronchiseptica (17). Interestingly, in B. pertussis, the pagL open reading frame was disrupted by a frameshift, whereas in B. bronchiseptica and B. parapertussis the open reading frame was intact.

Until the second half of the previous century, pertussis was one of the major causes of early childhood mortality and morbidity. The introduction of whole-cell pertussis vaccines in the 1950s and 60s changed this and led to a dramatic decrease in the number of pertussis cases. However, a problem was that the whole-cell pertussis vaccines contained several toxins, such as LPS and pertussis toxin, and therefore displayed considerable reactogenicity. This led to the development of several acellular pertussis vaccines in the 1980s and 1990s. These vaccines are clearly less reactogenic and have an efficacy of between 58% and 97%, which is comparable to that of whole-cell pertussis vaccines (51). However, causes of concern are that these acellular vaccines are rather expensive, can be expected to select (potentially) for escape mutants more rapidly, and were shown to skew towards a Th2 immune response (5, 49). The development of a new but less reactogenic whole-cell pertussis vaccine seems to be a feasible solution.

Here, we studied the consequence of PagP and PagL expression for the endotoxic activity of B. pertussis LPS and whole bacterial cells. The goal of this study was to investigate whether the PagP and PagL enzymes might be useful tools for decreasing the LPS-mediated reactogenicity of whole-cell pertussis vaccines. In addition, this study provides the first description of B. pertussis mutants with altered lipid A.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

All bacterial strains used are described in Table 1. Typically, the E. coli strains were grown at 37°C in a modified Luria-Bertani broth, designated LB (57), with shaking at 200 rpm. The medium was supplemented with 0.2% glucose. When appropriate, bacteria were grown in the presence of 100 μg/ml ampicillin, 50 μg/ml kanamycin, 10 μg/ml chloramphenicol, 50 μg/ml nalidixic acid, or 300 μg/ml streptomycin, for plasmid maintenance or strain selection. B. pertussis was grown at 35°C on Bordet-Gengou agar (Difco) supplemented with 15% defibrinated sheep blood (Biotrading). To induce the expression of the plasmid-carried pagL and pagP genes in B. pertussis, the bacteria were grown in synthetic THIJS medium (56) supplemented with 1 mM isopropyl-1-thio-β-d-galactopyranoside (IPTG) at 35°C with shaking (175 rpm). To grow B. pertussis in the Bvg⁻ phase, 50 mM magnesium sulfate (final concentration) was added to the growth medium. A Bvg⁺ phase-locked mutant of B. pertussis strain Tohama was created by using suicide plasmid pJM503 (Table 1) for allelic exchange mutagenesis. The presence of the correct mutation in the bvgS gene on the chromosome was confirmed by nucleotide sequencing in both directions after obtaining the gene by PCR. The correct phenotype was confirmed by testing colonies for hemolytic activity on Bordet-Gengou agar supplemented with 15% defibrinated sheep blood and 50 mM MgSO₄. Neisseria meningitidis was grown at 37°C on GC agar plates (Oxoid) supplemented with Vitox (Oxoid). To induce the expression of pagL in N. meningitidis, bacteria containing plasmid pEN11-PagL_(Bb) were grown on GC agar plates supplemented with Vitox and 1 mM IPTG at 37°C.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Genotype or description	Source or reference
B. pertussis
Tohama	Wild-type strain	NVIa
Tohama Bvg+ phase lock	Bvg+ phase-locked mutant of B. pertussis Tohama strain	This study
E. coli
TOP10F′	F′[lacIq Tn10 (Tetr)] mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 deoR recA1 araD139 Δ(ara-leu)7697 galU galK rpsL endA1 nupG	Invitrogen
DH5α	F− Δ(lacZYA-algF)U169 thi-1 hsdR17 gyrA96 recA1 endA1 supE44 relA1 phoA φ80 ΔlacZΔM15	22
SM10λpir	thi thr leu tonA lacY supE recA::RP4-2-Tc::Mu λ pir R6K Kanr	NVI
N. meningitidis
H44/67	Wild-type strain	NVI
H44/67-lpxA	LPS-deficient derivative of H44/76	53
Plasmids
pCRII-TOPO	E. coli cloning vector; Ampr Kanr	Invitrogen
pET-11a	E. coli high-copy expression vector, Ampr, T7 promoter	Novagen
pMMB67EH	Broad-host-range expression vector	15
pPagL(Pa)	pET-11a derivative harboring P. aeruginosa pagL	17
pPagL(Bb)	pET-11a derivative harboring B. bronchiseptica pagL	17
pPagL(Pa)(S151A)	pPagL(Pa) encoding PagL(Pa) with S151A substitution	17
pPagP(Bp)	pET-11a derivative harboring B. pertussis pagP	This study
pMMB67EH-PagL(Pa)	pMMB67-EH derivative harboring P. aeruginosa pagL	This study
pMMB67EH-PagL(Bb)	pMMB67-EH derivative harboring B. bronchiseptica pagL	This study
pMMB67EH-PagL(Pa)(S151A)	pMMB67EH-PagL(Pa) encoding PagL(Pa) with S151A substitution	This study
pMMB67EH-PagP(Bp)	pMMB67-EH derivative harboring B. pertussis pagP	This study
pMMB67EH-PagP(Ec)	pMMB67-EH derivative harboring E. coli pagP	55
pEN11-PagL	pEN11 derivative harboring B. bronchiseptica pagL	8
pJM503	Suicide vector harboring bvgS region with a G2527A substitution	35

^aNVI, Netherlands Vaccine Institute, Bithoven, The Netherlands.

Recombinant DNA techniques.

All plasmids used are described in Table 1. Plasmid DNA was isolated using the Promega WizardPlus SV Minipreps system. Calf intestine alkaline phosphatase and restriction endonucleases were used according to the instructions of the manufacturer (Fermentas). DNA fragments were isolated from agarose gels by using the QIAGEN quick gel extraction kit. Ligations were performed using the rapid DNA ligation kit (Roche).

The pagP gene from B. pertussis strain Tohama (pagP_(Bp)) was obtained by PCR. The chromosomal template DNA was prepared by resuspending ∼10⁹ bacteria in 50 μl of distilled water, after which the suspension was heated for 15 min at 95°C. The suspension was then centrifuged for 1 min at 16,100 × g, after which the supernatant was used as template DNA. The sequence of the forward primer, which contained an NdeI site (underlined) and included an ATG start codon, was 5′-AACATATGACCCAGTATTTCCGGTCCC-3′. The sequence of the reverse primer, which contained a BamHI site (underlined) and included a stop codon, was 5′-AAGGATCCTTAGAACTCCCAGCGGCCA-3′. The PCR was done under the following condition: 50-μl total reaction volume, 25 pmol of each primer, 0.2 mM deoxynucleoside triphosphates, 3 μl of template DNA solution, 1.5% dimethyl sulfoxide, and 1.75 units of Expand High Fidelity enzyme mix with buffer supplied by the manufacturer (Roche). The temperature program was as follows: 95°C for 3 min; a cycle of 1 min at 95°C, 1 min at 60°C, and 1 min 30 s at 72°C repeated 30 times; followed by 10 min at 72°C and subsequent cooling to 4°C. The PCR product was purified from agarose gel and subsequently cloned into pCRII-TOPO. Plasmid DNA from a correct clone was digested with NdeI and BamHI, and the PagP-encoding fragment was ligated into NdeI- and BamHI-digested pET-11a. The ligation mixture was used to transform E. coli DH5α by the CaCl₂ method (50). A plasmid containing the correct insert was designated pPagP_(Bp). The nucleotide sequence of the cloned pagP gene was confirmed by nucleotide sequencing in both directions.

To allow for expression in B. pertussis, pagP_(Bp) and the pagL genes of B. bronchiseptica [pagL_(Bb)] and Pseudomonas aeruginosa [pagL_(Pa)] were subcloned into the broad-host-range, low-copy vector pMMB67EH. pPagP_(Bp), pPagL_(Bb), pPagL_(Pa), and pPagL_(Pa)(S151A) plasmid DNAs (Table 1) were digested with XbaI and HindIII, and the relevant fragments were ligated into XbaI- and HindIII-digested pMMB67EH. The ligation mixture was used to transform E. coli DH5α. Plasmids with the correct inserts were designated pMMB67EH-PagP_(Bp), pMMB67EH-PagL_(Bb), pMMB67EH-PagL_(Pa), and pMMB67EH-PagL_(Pa)(S151A), respectively (Table 1). The pMMB67EH-based plasmids were used to transform E. coli SM10(λpir), which allowed for subsequent transfer of the plasmids to B. pertussis by conjugation.

Isolation and analysis of LPS.

LPS was isolated using the hot-phenol/water extraction method (63) with slight modifications. In short, bacteria were grown in THIJS medium supplemented with 1 mM IPTG to stationary phase. Cells were harvested by centrifugation and resuspended in 40 mM sodium phosphate buffer (pH 7.0) containing 5 mM EDTA. The cells were treated overnight with lysozyme at 4°C, after which an equal volume of phenol was added. The suspension was heated to 70°C, incubated for 30 min with shaking, and subsequently cooled to 10°C, after which phases were separated by centrifugation for 10 min at 8,000 × g. The upper phase was collected, and the extraction was repeated after addition of an equal volume of distilled water to the lower phase. The two upper phases were combined, dialyzed against tap water until the phenol odor disappeared, freeze-dried, and subsequently taken up in distilled water. The LPS was pelleted by centrifugation for 3 h at 150,000 × g and dissolved in distilled water, after which the LPS concentration was determined by analyzing the 3OH-C₁₄ content, using an Agilent 6890 gas chromatograph, as described previously (62). Approximately 250 pmol purified LPS was applied to a Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel (30). The electrophoresis was run for 17 h at a constant current of 20 mA, after which the gel was fixed with water-ethanol-acetic acid (11:8:1, vol/vol/vol) and subsequently stained with silver as described previously (60). For electrospray ionization-mass spectrometry (ESI-MS), a 200-μl aliquot of isolated LPS (50 nmol/ml) was freeze-dried and taken up in 0.1 ml of 2% acetic acid (pH 2.8). The mixture was heated for 2 h at 95°C to hydrolyze the LPS and release the lipid A moiety. Subsequently, the mixture was cooled to room temperature and centrifuged for 10 min at 16,100 × g. The pellet was washed twice in 0.1 ml of double-distilled water and taken up in 0.1 ml of double-distilled water, and 0.3 ml of chloroform-methanol (2:1, vol/vol) was added. After vigorous vortexing, phases were separated by centrifugation for 10 min at 16,100 × g. The upper phase was then used for structural analysis of the lipid A by nanoelectrospray tandem MS on a Finnigan LCQ instrument in the negative-ion mode (64).

Determination of LPS release by B. pertussis cells.

To determine the total LPS content, stationary-phase cultures grown in THIJS medium in the presence of 1 mM IPTG were heat inactivated (1 h at 56°C), freeze-dried, and stored at 4°C. To determine the amount of LPS released into the culture medium, the cells were removed from cultures by centrifugation (10 min at 16,100 × g), after which remaining cells were removed from the supernatant by filtration through a 0.22-μm filter (Millipore). The filtrate was subsequently freeze-dried and also stored at 4°C. Whole-cell suspensions, used for endotoxic activity assays (see below), were prepared by collecting the cells from cultures by centrifugation (for B. pertussis) or from GC agar plates (for N. meningitidis). The cells were washed once with phosphate-buffered saline (PBS) and then resuspended in PBS at an A₅₉₀ of 1.0, heat inactivated, and stored at 4°C. To determine whether LPS was released from the B. pertussis cells during storage, whole-cell B. pertussis suspensions that had been stored at 4°C for 48 h were centrifuged and remaining cells were removed from the supernatant by filtration. The filtrate was freeze-dried and stored at 4°C. The LPS content of the cells and cell-free supernatants was determined by measuring the 3OH-C₁₄ content in the different preparations. The freeze-dried materials were resuspended in PBS, after which the samples were analyzed using an Agilent 6890 gas chromatograph as described previously (62). The results shown represent the data from four individual experiments and were statistically analyzed using Student's t test (two-tailed, two-sample unequal variance). P values lower than 0.05 were considered to be statistically significant.

Endotoxic activity assays.

Stimulation of IL-6 production was tested with the human macrophage cell line MM6 (66), the TLR4-deficient mouse macrophage cell line CTC3H/HeJ, or CTC3H/HeJ cells expressing human TLR4 (hTLR4) (41). The macrophages were seeded in 24-wells plates (2.5 × 10⁵ cells/well) in 400 μl of IMDM medium (Gibco BRL) supplemented with 100 units/ml penicillin, 100 μg/ml streptomycin, 292 μg/ml l-glutamine (Gibco BRL), and 10% fetal calf serum (Gibco BRL). The cells were stimulated with 100 μl of serial dilutions of whole-cell suspensions. (prepared as described above, starting at an A₅₉₀ of 0.2 in the total stimulation volume), cell-free supernatants of whole-cell suspensions (starting with 0.1 ml supernatant in the total stimulation volume), or purified B. pertussis LPS (starting at 10 nmol/ml in the total stimulation volume). The stimulation was for 16 to 18 h at 37°C in a humid atmosphere containing 5% or 10% CO₂ for the MM6 cells and CTC3H/HeJ cells, respectively. For the determination of the effect of EDTA on the endotoxic activity of a whole-cell suspension, MM6 cells were stimulated with a B. pertussis whole-cell suspension, at an A₅₉₀ of 0.01 in the total stimulation volume, in the presence of serial dilutions of EDTA, starting at 10 mM. As an internal standard for the stimulation potency of the cells, CTC3H/HeJ cells were stimulated with 100 μl of serial dilutions of a whole-cell suspension of an LPS-deficient N. meningitidis strain (53) or the TLR2/6 agonist lipoprotein FSL (InvivoGen) (38). Following stimulation, IL-6 concentrations in the culture supernatants were quantified with an enzyme-linked immunosorbent assay (ELISA) against mouse or human IL-6 according to the manufacturers' instructions (eBioscience and PeliKine Compact, respectively).

RESULTS

Contribution of LPS to the total endotoxic activity of whole B. pertussis cells.

Since our goal was to investigate whether the PagP and PagL enzymes might be useful tools for the development of a less reactogenic whole-cell pertussis vaccine, we first investigated the contribution of LPS to the total endotoxic activity of whole B. pertussis cells. Serial dilutions of a whole-cell suspension of B. pertussis strain Tohama were used to stimulate the production of the proinflammatory cytokine IL-6 by TLR4-deficient C3H/HeJ cells or by C3H/HeJ cells expressing hTLR4. The hTLR4-expressing cells were stimulated much more strongly (Fig. 2A), indicating a major contribution of the LPS to the total endotoxic activity of the B. pertussis cells. However, we also tested whether the stimulation potencies of the two cell lines were comparable. Therefore, the macrophages were stimulated with a whole-cell suspension of an LPS-deficient N. meningitidis strain and with the TLR2/6 agonist lipoprotein FSL. In both cases, the TLR4-deficient C3H/HeJ cells exhibited a clearly reduced potency to be stimulated (Fig. 2A and B). These results demonstrate that the TLR4-transfected C3H/HeJ cells are much more sensitive to stimulation by non-TLR4 agonists than the parent cells, emphasizing the importance of proper controls when using these cells in determining the relative contributions of various TLR agonists to cytokine induction. Nevertheless, since the macrophages expressing hTLR4 produced more IL-6 after stimulation with B. pertussis cells than after stimulation with LPS-deficient N. meningitidis, whereas this was just the opposite for the TLR4-deficient C3H/HeJ cells, B. pertussis LPS still seems to contribute to the endotoxic activity of whole B. pertussis cells. Thus, reducing the endotoxic activity of B. pertussis LPS seems to be a valid strategy for obtaining a strain that can be used as the basis for a less reactogenic whole-cell pertussis vaccine.

FIG. 2.

Contribution of LPS to the endotoxic activity of whole B. pertussis cells. Stimulation of IL-6 production by (A) wild-type, whole B. pertussis cells (B.P.) or LPS-deficient N. meningitidis cells (N.M.) or by (B) the TLR2/6 ligand FSL was tested with the TLR4-deficient mouse macrophage cell line C3H/HeJ (TLR4−) and C3H/HeJ cells expressing hTLR4. C3H/HeJ cells were seeded in microtiter plates and stimulated with serial dilutions of a whole-cell suspension. The IL-6 concentration in the culture supernatant was quantified with an ELISA against mouse IL-6. The data represent the averages and standard deviations from three individual experiments.

Expression of PagP and PagL in B. pertussis.

To produce PagP and PagL in B. pertussis strain Tohama, the pagL gene of B. bronchiseptica [pagL_(Bb)] and the pagP genes of B. pertussis [pagP_(Bp)] and of E. coli [pagP_(Ec)] were expressed from the broad-host-range, low-copy-number vector pMMB67EH. LPS isolated from the PagL_(Bb)-expressing strain appeared to be unaffected upon Tricine-SDS-PAGE analysis, whereas that from the PagP-expressing strains appeared to be modified, since a band with a lower electrophoretic mobility than that of the wild-type LPS was detected in case of the PagP-expressing strains (Fig. 3). The modification efficiency appeared to be higher in the PagP_(Ec)-expressing strain than in the PagP_(Bp)-expressing strain. To evaluate the possible LPS modifications in further detail, the lipid A moieties of the strains were analyzed by ESI-MS in the negative-ion mode. This analysis revealed the presence of four major lipid A species in wild-type LPS (Fig. 4A). The peak at m/z 1557 represents the characteristic penta-acylated bisphosphate species that is typically found in B. pertussis (10), whereas the peak at m/z 1477 corresponds to a penta-acylated monophosphate species. The two remaining peaks at m/z 1307 and 1251 represent deacylated lipid A species of the molecular ion at m/z 1477, which lack the primary 3OH-C₁₀ residue at the 3 position and a primary 3OH-C₁₄ residue (at either the 2 or the 3′ position), respectively. Interestingly, calculation of the relative amounts of the individual lipid A species from the corresponding peaks revealed that in wild-type B. pertussis LPS, a large quantity of lipid A species (∼50%) consists of tetra-acylated forms. Furthermore, the large majority (∼80%) of lipid A species are monophosphate forms. These results indicate a high heterogeneity among the lipid A species in wild-type B. pertussis, which was apparently not resolved in the gel analysis (Fig. 3). This observed heterogeneity could be either the consequence of fragment ionization during mass spectrometry analysis, although this is an unlikely option with respect to the heterogeneity in acyl chain substitution, or the consequence of active modification by lipid A-modifying enzymes. However, to exclude the possibility that the observed heterogeneity was a consequence of overhydrolysis prior to the analysis, we tested whether shorter or longer periods of hydrolysis (varying between 1 and 4 h) influenced the ratio of the various lipid A species, and this was not the case (data not shown).

FIG. 3.

Analysis by Tricine-SDS-PAGE of in vivo LPS modification. LPS was isolated from wild-type (WT) and PagP- or PagL-expressing B. pertussis strain Tohama and analyzed by Tricine-SDS-PAGE.

FIG. 4.

Structural analysis by ESI-MS of purified B. pertussis LPS. Lipid A species from wild-type B. pertussis strain Tohama (A) and its derivatives expressing PagL_(Bb) (B), PagP_(Ec) (C), and PagP_(Bp) (D) were analyzed by ESI-MS. Major peaks at m/z 1557, 1477, 1387, 1307, 1251, and 1081 were interpreted as the characteristic penta-acylated bisphosphate species that is typically found in B. pertussis, the corresponding penta-acylated monophosphate species, the deacylated lipid A species of the molecular ion at m/z 1557 missing the primary 3OH-C₁₀ residue at the 3 position, the deacylated lipid A species of the molecular ion at m/z 1477 missing the primary 3OH-C₁₀ residue at the 3 position, the deacylated lipid A species of the molecular ion at m/z 1477 missing a primary 3OH-C₁₄ residue, and the deacylated lipid A species of the molecular ion at m/z 1477 missing both the primary 3OH-C₁₀ residue at the 3 position and a primary 3OH-C₁₄ residue, respectively. The peaks at m/z 1320, 1490, 1545, 1625, 1715, and 1796 correspond to the PagP-mediated palmitoylation of the molecular ions present at m/z 1081, 1251, 1307, 1387, 1477, and 1557, respectively.

Upon expression of PagL_(Bb) (Fig. 4B), three lipid A species, at m/z 1081, 1307, and 1387, were detected. The major peak at m/z 1307 corresponds to the monophosphate deacylated form missing the 3OH-C₁₀ residue at the 3 position, whereas the peak at m/z 1387 corresponds to the bisphosphorylated form of the molecular ion at m/z 1307. The peak at m/z 1081 corresponds to a monophosphate form missing both the 3OH-C₁₀ residue and a 3OH-C₁₄ residue. The relative content of lipid A species that lack the 3OH-C₁₀ residue at the 3 position was increased from ∼37% in wild-type B. pertussis LPS to more than 92% in the strain expressing PagL_(Bb). Thus, even though the electrophoretic mobility of the LPS was not affected (Fig. 3, lane 2), the pagL_(Bb)-encoded lipid A 3-O-deacylase was active in B. pertussis.

Upon expression of PagP_(Ec) (Fig. 4C) and PagP_(Bp) (Fig. 4D), several new lipid A species were detected. The peaks at m/z 1320, 1490, 1545, 1625, 1715, and 1796 correspond to the expected PagP-mediated palmitoylation of the molecular ions present at m/z 1081, 1251, 1307, 1387, 1477, and 1557, respectively. The difference in modification efficiency between E. coli PagP and B. pertussis PagP, which was seen after analysis by Tricine-SDS-PAGE (Fig. 3), was also revealed in the mass spectrometry analysis. In the strain expressing E. coli PagP, approximately 47% of the total lipid A pool was palmitoylated, in contrast to only ∼9% in the strain expressing PagP_(Bp). Remarkably, palmitoylated lipid A species missing a 3OH-C₁₄ residue were found only in the strain expressing PagP_(Ec) and not in that expressing PagP_(Bp). E. coli PagP and B. pertussis PagP were previously shown to have different specificities (7, 43). Whereas E. coli PagP adds a secondary C₁₆ to the primary acyl chain at the 2 position of lipid A, B. pertussis PagP adds the C₁₆ to the acyl chain at the 3′ position (Fig. 1). Thus, if the B. pertussis lipid A species lacking a 3OH-C₁₄ residue (Fig. 4A) lack it specifically at their 3′ positions, the different modification efficiencies of the two PagP enzymes could be explained, as the substrate pool for B. pertussis PagP would then be smaller than that for E. coli PagP. Such an explanation would also be consistent with the presence of hypoacylated lipid A species in vivo. An overview of the relative abundances of all lipid A molecular ions as determined by ESI-MS is presented in Table 2.

TABLE 2.

Relative abundances of lipid A molecular ions as determined by ESI-MS

Protein	Relative abundance of molecular ion at m/z:
	1081 (−C14-3OH, −C10-3OH, −PO4)	1161 (−C14-3OH, −C10-3OH)	1251 (−C14-3OH, −PO4)	1307 (−C10-3OH, −PO4)	1320 (−C14-3OH, −C10-3OH, −PO4, +C16)	1331 (−C14-3OH)	1387 (−C10-3OH)	1477 (−PO4)	1490 (−C14-3OH, −PO4, +C16)	1545 (−C10-3OH, −PO4, +C16)	1557	1625 (−C10-3OH, +C16)	1715 (−PO4, +C16)	1796 (+C16)
Wild type	3.0	0.0	15.6	29.9	0.0	4.5	3.9	29.0	0.0	0.0	14.1	0.0	0.0	0.0
PagL(Bb)	8.5	0.0	2.1	70.9	0.0	0.0	12.8	3.5	0.0	0.0	2.1	0.0	0.0	0.0
PagP(Ec)	4.5	1.3	2.3	25.0	5.0	2.3	10.3	2.0	3.0	14.8	6.0	5.3	6.3	12.3
PagP(Bp)	6.3	2.5	5.0	27.6	0.0	6.1	21.3	3.0	0.0	3.9	17.1	2.5	0.0	2.9

Endotoxic activity of LPS.

To assess the endotoxic activity of wild-type and PagP- or PagL-modified B. pertussis LPSs, their potency to stimulate the production of IL-6 by the human macrophage cell line MM6 was tested. Compared with wild-type LPS, the purified LPS from the strains expressing PagP_(Ec) or PagP_(Bp) had a strongly increased ability to stimulate the macrophages (Fig. 5A). In contrast, modification by PagL_(Bb) reduced the ability to stimulate IL-6 production. Thus, it can be concluded that in vivo palmitoylation results in an increase in endotoxic activity of B. pertussis LPS, whereas PagL_(Bb)-mediated deacylation results in LPS that is, in comparison to wild-type B. pertussis LPS, biologically less active. Comparable results were also obtained for other cytokines, such as IL-10 and IL-8 (data not shown).

FIG. 5.

IL-6 induction by purified B. pertussis LPS and whole cells. The production of IL-6 by the human macrophage cell line MM6 was stimulated with serial dilutions of (A) a stock solution of purified wild-type (WT) or PagP- or PagL-modified LPS from B. pertussis strain Tohama, (B) a whole-cell suspension of B. pertussis cells expressing PagP or PagL or containing the empty vector pMMB67EH (WT), or (C) a whole-cell suspension of whole B. pertussis cells containing the empty pMMB67EH vector (WT) or expressing PagL_(Bb), wild-type P. aeruginosa PagL [PagL_(Pa)], or an active-site mutant derivative of PagL_(Pa) [PagL_(Pa)(s151A)]. The IL-6 concentration in the culture supernatant was quantified in an ELISA against human IL-6. The data represent the averages and standard deviations from three individual experiments.

Endotoxic activity of whole B. pertussis cells.

To determine whether expression of PagP and PagL affects the endotoxic activity of whole bacterial cells similarly to that of purified LPS, serial dilutions of whole-cell suspensions were used to stimulate IL-6 production by the macrophage cell line MM6. As expected, B. pertussis cells expressing PagP_(Ec) or PagP_(Bp) showed, compared to B. pertussis cells containing the empty vector, an increased potency to stimulate the macrophages (Fig. 5B). However, unexpectedly, the cells expressing PagL_(Bb) also showed this effect (Fig. 5B), while the purified LPS of this strain was biologically less active (Fig. 5A). This paradoxical result suggests that some other mechanism counterbalances the reduced endotoxic activity of PagL_(Bb)-modified LPS in the whole cells. The majority of virulence factors in the bordetellae are regulated by the two-component system BvgAS (34). It is known that some of these virulence factors, e.g., adenylate cyclase, are capable of modulating the immune response and suppress the production of proinflammatory cytokines by macrophages (9, 47). Therefore, we hypothesized that the increased endotoxic activity of the whole cells expressing PagL_(Bb) might be a consequence of an altered expression of virulence factors through an effect on the Bvg system. To test this hypothesis, we expressed PagL_(Bb) in a Bvg⁺ phase-locked B. pertussis strain and compared the whole-cell endotoxic activity of this strain with that of the same strain containing the empty vector. Expression of PagL_(Bb) in the Bvg⁺ phase-locked strain resulted in an increase in endotoxic activity comparable to that in the wild-type strain, indicating that bvg-mediated alterations in the expression levels of virulence factors are not responsible for the increased endotoxic activity of whole PagL-expressing B. pertussis cells (data not shown). In addition, we also observed a similar increase in IL-6 induction with PagL-expressing cells grown under conditions known to induce the Bvg⁻ phase (results not shown). To investigate whether the enzymatic activity as opposed to merely the expression of PagL is responsible for the observed increase in endotoxic activity, we constructed two other B. pertussis strains, one expressing P. aeruginosa PagL [PagL_(Pa)] and the other expressing an inactive variant thereof [PagL_(Pa)(S151A)] in which one of the catalytic-site residues, Ser151, has been replaced by alanine (17). Like that of B. bronchiseptica PagL, expression of P. aeruginosa PagL led to increased endotoxic activity of B. pertussis whole cells (Fig. 5C). Furthermore, enzymatic activity of PagL was needed for the increased endotoxic activity, since stimulation of the MM6 cells with bacteria expressing the mutant PagL did not lead to increased IL-6 production (Fig. 5C).

Deacylation of LPS influences its release from B. pertussis membranes.

LPS is normally anchored into the outer membranes of gram-negative bacteria via its lipid A part. However, by disturbing LPS-LPS interactions, e.g., by EDTA treatment (29), LPS can be released from the outer membrane into the external environment. In previous studies, it was shown that membrane-bound LPS has a lower proinflammatory cytokine-inducing activity than soluble LPS aggregates (36, 42), probably due to the better accessibility of LPS in soluble aggregates for extraction and subsequent transfer by LBP/CD14 to the TLR4/MD-2 complex. Deacylation of LPS could potentially weaken the LPS-LPS interactions, leading to increased release from the cells and counterbalancing the effect of the reduced endotoxic activity of the purified, deacylated LPS.

To test the hypothesis that deacylated B. pertussis LPS is more readily released, we compared the amounts of released LPS in whole-cell suspensions of wild-type and PagL_(Bb)-expressing cells. The fraction of released LPS was assessed by determining the total 3OH-C₁₄ content in the whole-cell suspensions and cell-free supernatants obtained from the suspensions. The proportion of LPS released in the whole-cell suspensions of the PagL_(Bb)-expressing strain was significantly higher than that from the wild-type strain, i.e., 17% (±1.8%) versus 11% (±0.7%), showing that deacylation indeed led to an increased release of LPS. For reasons of safety, B. pertussis cells were incubated for 1 h at 56°C during the preparation of whole-cell suspensions. To test whether the higher LPS release from the PagL_(Bb)-expressing cells was due to this heat inactivation step, we also determined the fraction of LPS released during growth into the culture medium. A significantly larger proportion of LPS was detected in the cell-free culture medium from the PagL-expressing strains than in that of the wild-type strain, i.e., 12% (±4.6%) for the wild-type strain and 20% (±1.6%) for the PagL_(Bb)-expressing strain. Interestingly, this analysis also revealed that the total amount of LPS produced, corrected for optical density, by wild-type cells was significantly higher, i.e., 21% (±8.3%), than that produced by the bacteria expressing PagL_(Bb). Overall, these data indicate that deacylation of B. pertussis LPS indeed leads to an increased LPS release from the cells into the external environment and that this is independent of heat inactivation at 56°C. However, we cannot exclude the possibility that this increased release is an indirect consequence of the removal of the acyl chain from the 3 position. For instance, PagL-mediated deacylation of lipid A could induce secondary changes in the LPS core region, which, in turn, could be primarily responsible for the observed release. Yet, the observation that deacylated LPS migrates at the same position as wild-type LPS on a Tricine-SDS-PAGE gel (Fig. 3, lanes 1 and 2) indicates that major changes in LPS composition, other than deacylation at the 3 position, are absent. Hence, the increased release of LPS after modification by PagL is probably directly related to the removal of the acyl chain from the 3 position.

EDTA treatment increases the endotoxic activity of a B. pertussis whole-cell suspension.

To test the hypothesis that increased LPS release enhances endotoxic activity, we stimulated MM6 cells with a whole-cell suspension of wild-type B. pertussis cells in the presence of various amounts of EDTA (Fig. 6). The addition of EDTA indeed led to a concentration-dependent increase in the endotoxic activity of the whole-cell suspension, which is consistent with the hypothesis that destabilization of the outer membrane and increased LPS release could potentially counterbalance the reduced endotoxic activity of deacylated LPS.

FIG. 6.

Stimulation of IL-6 production by a B. pertussis whole-cell suspension in the presence of EDTA. The production of IL-6 by the human macrophage cell line MM6 was stimulated with a whole-cell suspension in the presence of various amounts of EDTA. The IL-6 concentration in the culture supernatant was quantified in an ELISA against human IL-6. The data represent the averages and standard deviations from three individual experiments.

Released LPS is the main determinant of whole-cell suspension endotoxicity.

It was shown above that LPS release increases the endotoxic properties of a whole-cell suspension. To determine the contribution of released LPS to the endotoxic activity of whole-cell suspensions, MM6 cells were stimulated with either a whole-cell suspension of wild-type B. pertussis or with the cell-free supernatant of such a suspension. As shown in Fig. 7, ∼90% of the endotoxic activity appeared to be present within the supernatant. Since only ∼11% of the total LPS was released from the cells (see above), these data suggest that the majority of LPS, which remained associated with the bacterial cells, only marginally contributed to the endotoxic activity of the suspensions.

FIG. 7.

Released LPS determines endotoxicity of B. pertussis whole-cell suspensions. Production of IL-6 by the human macrophage cell line MM6 was stimulated with serial dilutions of a whole-cell suspension of wild-type B. pertussis (WC) or with the cell-free supernatant of such a suspension (SUP). The data represent the averages and standard deviations from three individual experiments.

PagL expression lowers the endotoxic activity of N. meningitidis cells.

To test whether PagL expression would also lead to an increased endotoxic activity in another gram-negative bacterium, we stimulated MM6 cells with serial dilutions of a whole-cell suspension of wild-type or PagL_(Bb)-expressing N. meningitidis cells. Previously, we have shown that PagL(Bb) expression in this host leads to efficient 3-O deacylation of the LPS (8). As a control, the macrophages were also stimulated with LPS-deficient N. meningitidis cells. Compared to the wild-type cells, both the PagL_(Bb)-expressing cells and the LPS-deficient cells displayed a reduced endotoxic activity (Fig. 8). This reduction was most pronounced for the LPS-deficient cells, implying that the LPS in the PagL-expressing strain still had residual endotoxic activity. Most importantly, this result shows that, in contrast to B. pertussis cells, N. meningitidis cells expressing PagL do show a lowered biological activity compared to the wild-type cells. Thus, the paradoxical results obtained with PagL-expressing B. pertussis cells cannot directly be extrapolated to other gram-negative bacteria.

FIG. 8.

PagL modification reduces endotoxic activity of N. meningitidis cells. The production of IL-6 by the human macrophage cell line MM6 was stimulated with serial dilutions of a whole-cell suspension of wild-type N. meningitidis cells (WT H44/76), N. meningitidis cells expressing PagL_(Bb) [PagL(Bb)], or LPS-deficient N. meningitidis cells (LPS−). The IL-6 concentration in the culture supernatant was quantified in an ELISA against human IL-6. The data represent the averages and standard deviations from three individual experiments.

DISCUSSION

The lipid A part is mainly responsible for the endotoxic activity of LPS, and it is known that changes in its physicochemical properties influence its biological activity. In recent years, several enzymes that are capable of modifying lipid A have been discovered in various gram-negative bacteria. Two of these enzymes, PagP and PagL, are located in the outer membrane and function as a lipid A acylase and deacylase, respectively. Although genes for these enzymes have been identified in B. pertussis, they are actually pseudogenes (17, 43). Here, we investigated the consequence of PagP and PagL expression in B. pertussis on the endotoxic activity of the LPS and whole bacterial cells in order to determine whether the PagP and PagL enzymes might be useful tools for the development of a less reactogenic whole-cell pertussis vaccine.

In initial experiments, we wanted to determine the contribution of the LPS to the endotoxic activity of whole B. pertussis cells. Importantly, we observed that TLR4-deficient C3H/HeJ macrophages, compared to C3H/HeJ macrophages expressing hTLR4, exhibit a strongly reduced potency to be stimulated by LPS-deficient N. meningitidis cells as well as by the TLR2/6 agonist lipoprotein FSL. These observations suggest possible cross talk between different TLRs or, alternatively, a reduced sensitivity of the C3H/HeJ macrophage cell line to TLR agonists in general. Therefore, the use of these cell lines to identify LPS responses is less straightforward than anticipated. Nevertheless, the observation that the B. pertussis cells stimulated the TLR4-expressing cells more strongly than did the LPS-deficient N. meningitidis cells, whereas the opposite was the case for the TLR4-deficient macrophages, suggests that B. pertussis LPS, in spite of its relatively low biological activity, might still contribute significantly to the endotoxic activity of the whole bacterial cells.

Interestingly, we found that B. pertussis LPS shows a high heterogeneity. This observation is consistent with the high LPS heterogeneity that was previously reported for other Bordetella strains (10, 65) and other gram-negative bacteria (reviewed in reference 11). Since the enzymes that cooperate in the lipid A biosynthesis route are thought to be interdependent and to work in a preferred order (reviewed in reference 59), it is hard to imagine how significant lipid A heterogeneity could arise during biosynthesis. It is therefore more likely that the heterogeneity observed is the consequence of active modification by lipid A-modifying enzymes. Mass spectrometry analysis showed that ∼37% of the total LPS in the wild-type strain consists of tetra-acylated forms that miss the 3OH-C₁₀ residue at the 3 position, i.e., the position where PagL removes the acyl chain. Since B. pertussis pagL is a pseudogene, it cannot be responsible for removal of this fatty acyl chain. This finding thus suggests the presence of another enzyme in B. pertussis, with an activity comparable to that of PagL. However, no other PagL homolog could be identified in the genome sequence (data not shown). We recently solved the crystal structure of the P. aeruginosa PagL homolog (48). That study revealed a close resemblance of the active site of PagL to that of the outer membrane phospholipase A. It is therefore tempting to speculate that B. pertussis outer membrane phospholipase A has relaxed substrate specificity and might be responsible for removal of the 3OH-C₁₀ residue at the 3 position. Furthermore, the data suggest the existence of at least one additional lipid A deacylase with a specificity different from that of PagL, i.e., one that removes the 3OH-C₁₄ acyl chain from either the 2 or the 3′ position. Recently, an outer membrane-located lipid A deacylase, i.e., LpxR, with specificity for the 3′ position of lipid A has been identified in S. enterica serovar Typhimurium (46). During the analysis, we also observed a high abundance of monophosphate lipid A species. Although this hypophosphorylation could well be the consequence of fragment ionization during mass spectrometry analysis as observed previously (17), it should be noted also that a lipid A-1-phosphatase, i.e., LpxE, was recently identified in Rhizobium leguminosarum (26). Hence, we screened the B. pertussis genome sequence for possible homologs of the LpxE enzyme. BLAST searches resulted in the identification of open reading frame BP0835 (GenBank accession number CAE41138), which encodes a protein with a size comparable to that of LpxE and which also shares the conserved tripartite active-site motif of the lipid phosphatase family that was identified in LpxE (26).

Previous studies have shown that both the phosphate groups and the number and length of the acyl chains are important determinants of the endotoxic properties of lipid A (31, 44, 54). Recognition of LPS is based upon its binding to LBP and subsequent transfer to CD14, followed by interaction with the TLR4/MD2 complex. Earlier studies with purified LPS have shown that the strongest CD14-dependent response requires lipid A to be substituted with six acyl chains and two phosphates and the presence of an O antigen (16). Moreover, it was shown that wild-type B. pertussis LPS already is a poor cytokine inducer in human monocytes, as can be expected from its penta-acylated structure (21). Indeed, in our IL-6 induction assay, wild-type B. pertussis LPS was found to be at least 1,000-fold less active, on a weight basis, than hexa-acylated N. meningitidis LPS (data not shown). Here we showed that further deacylation by PagL reduces the endotoxic activity of B. pertussis LPS even more, thus indicating that the number of acyl chains significantly contributes to the overall endotoxic activity of B. pertussis LPS. This notion is supported by the observation that the PagP-mediated addition of an extra acyl chain to B. pertussis LPS led to increased endotoxic activity. Notably, only a small increase in palmitoylated species, i.e., 9%, in the PagP_(Bp)-expressing strain led to a strongly increased potency to stimulate macrophages. This result can be explained by the notion that hexa-acylated LPS, compared to penta-acylated LPS, is biologically far more active (44). Hence, small increases in hexa-acylated species content will give large effects on the displayed endotoxic activity. Although modification of hexa-acylated Salmonella LPS by PagP was earlier reported to result in LPS with reduced endotoxic activity (24, 27), the increase in endotoxic activity of B. pertussis LPS after modification by PagP is not surprising, since B. pertussis LPS is a mixture of tetra- and penta-acylated forms, which, after modification by PagP, will be converted into penta- and hexa-acylated forms.

Whereas purified PagL-modified LPS showed reduced endotoxic activity, we unexpectedly found that whole bacterial cells expressing PagL from either B. bronchiseptica or P. aeruginosa had increased potency to induce the production of IL-6. This increase in stimulatory potency was not caused by an altered expression of Bvg⁻-regulated virulence factors and was shown to be dependent on the enzymatic activity of PagL. It is well established that purified LPS, on a molar basis, is biologically more active than LPS incorporated into membrane structures such as in bacterial cells, outer membrane vesicles, or liposomes (36, 42). This can probably be explained by an increased accessibility of LPS in soluble aggregates for extraction and subsequent transfer by LBP/CD14 to the TLR4/MD-2 complex. As deacylation of LPS may diminish LPS-LPS interactions, we hypothesized that the increased endotoxic activity of the PagL-expressing cells could be due to an increased accessibility of the deacylated LPS to LBP/CD14. Such a mechanism could then counterbalance the effect of the reduced endotoxic activity of purified, deacylated LPS. Two observations support this interpretation. First, we found that PagL expression leads to an increased release of LPS from the bacterial cells into the medium. Second, a treatment known to result in LPS release, i.e., incubation in the presence of EDTA, results in an increased IL-6 induction by a B. pertussis cell suspension (Fig. 6). Therefore, increased release of deacylated LPS could eventually lead to a net increase in endotoxic activity of PagL-expressing cells. However, we cannot exclude the possibility that other factors also contribute to the observed increase in endotoxic activity of B. pertussis cells expressing PagL, since it is conceivable that after deacylation of LPS, other factors also can be more easily released from the B. pertussis cells.

It is tempting to speculate about a possible physiological significance of LPS release after PagL-mediated deacylation. By regulating PagL expression and/or activity, bacteria could, for instance, control the amount of LPS present within the outer membrane and thereby the membrane fluidity. Furthermore, released LPS could potentially function as a decoy and prevent recognition of the bacterium by the immune system. Interestingly, and consistent with this speculation, S. enterica serovar Typhimurium, one of the gram-negative bacteria that does contain a PagL enzyme, was previously shown, in contrast to other gram-negative bacteria, to be capable of releasing LPS, more or less selectively, from its outer membrane (37). In addition to the increased LPS release, we also observed that the cells expressing PagL produced ∼21% less LPS than the control strain. The reason for this reduced LPS content remains elusive, but it suggests the presence of a feedback mechanism on LPS biogenesis, following deacylation of the LPS by PagL.

The endotoxic activity of LPS forms a serious problem for the development of new whole-cell vaccines. We studied the consequences of the expression of two LPS-modifying enzymes, PagP and PagL, for the endotoxic activity of B. pertussis LPS and whole bacterial cells to see whether they might be useful tools for decreasing the LPS-mediated reactogenicity of whole-cell pertussis vaccines. We showed that these enzymes indeed modulate the endotoxic activity of B. pertussis LPS. Our results demonstrate how LPS modification influences biological activity at different levels, i.e., directly through its effect on LPS per se and indirectly through its effect on LPS accessibility. In the case of PagL, these effects work in opposite directions. Determination of whether the PagP and/or PagL modifications can be used in improved pertussis vaccines will require further in vivo studies on immunogenicity and reactogenicity. We did, however, show that PagL could readily be used for lowering the endotoxic activity of N. meningitidis cells. We think that the discrepancy between the effects on B. pertussis and N. meningitidis cells can be explained by differences in the physical properties of their LPS species, i.e., a mixture of tetra- and penta-acylated species versus primarily hexa-acylated species. Nevertheless, our data show that PagP, PagL, and probably also other lipid A-modifying enzymes, such as LpxE and LpxR, can be used for altering the endotoxic activity of gram-negative bacterial cells and that they thus potentially form useful tools for the development of future vaccines.