|Home | About | Journals | Submit | Contact Us | Français|
Synthesis of Escherichia coli LpxL, which transfers a secondary laurate chain to the 2′ position of lipid A, in Yersinia pestis produced bisphosphoryl hexa-acylated lipid A at 37°C, leading to significant attenuation of virulence. Our previous observations also indicated that strain χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) stimulated a strong inflammatory reaction but sickened mice before recovery and retained virulence via intranasal (i.n.) infection. The development of live, attenuated Y. pestis vaccines may be facilitated by detoxification of its lipopolysaccharide (LPS). Heterologous expression of the lipid A 1-phosphatase, LpxE, from Francisella tularensis in Y. pestis yields predominantly 1-dephosphorylated lipid A, as confirmed by mass spectrometry. Results indicated that expression of LpxE on top of LpxL provided no significant reduction in virulence of Y. pestis in mice when it was administered i.n. but actually reduced the 50% lethal dose (LD50) by 3 orders of magnitude when the strain was administered subcutaneously (s.c.). Additionally, LpxE synthesis in wild-type Y. pestis KIM6+(pCD1Ap) led to slight attenuation by s.c. inoculation but no virulence change by i.n. inoculation in mice. In contrast to Salmonella enterica, expression of LpxE does not attenuate the virulence of Y. pestis.
Yersinia pestis has evolved several strategies to evade phagocytosis and to inhibit the inflammatory response. One strategy depends on a large plasmid called pCD1 (~70 kb) (1), which encodes a type III secretion apparatus necessary for the translocation of effector proteins (Yops) into eukaryotic target cells (2) and is the core of the Yersinia pathogenicity machinery that targets cells of the immune system (3). The injected Yops disturb the dynamics of the cytoskeleton, disrupt phagocytosis, and block the production of proinflammatory cytokines, thus favoring the survival of the invading Yersinia (2, 3). In addition, Y. pestis also evolved the ability to evade host immune recognition by altering its lipid A structure (4, 5).
Lipid A, the hydrophobic anchor of lipopolysaccharide (LPS), is a glucosamine-based saccharolipid that makes up the outer monolayers of the outer membranes of Gram-negative bacteria. It is also known as endotoxin because of its ability to induce toxic inflammatory responses (6, 7). Many of the immune-activating abilities of LPS can be attributed to lipid A. It is a very potent activator of the immune system, stimulating cells (such as monocytes or macrophages) at picogram-per-milliliter quantities (8). Sensing of lipid A by the human immune system is critical for the onset of immune responses and clearance of Gram-negative bacterial infections. LPS activates cells via Toll-like receptor 4 (TLR4) and MD-2 on the cell surface (9–12), but the process of activation is dependent upon the structure of lipid A (13).
In Escherichia coli and Salmonella enterica serovar Typhimurium, the final steps of lipid A synthesis occur in the inner membrane, where two acyl groups are added to the tetra-acylated Kdo2 [(3-deoxy-d-manno-octulosonic acid)2]-lipid IVA before the mature hexa-acylated lipid A is exported to the outer membrane. At normal growth temperatures, the late acyltransferases LpxL (HtrB) and LpxM (MsbB) consecutively add lauroyl (C12) and myristoyl (C14) groups, respectively, to the tetra-acylated intermediate (14–16). At 12°C, however, the cold-temperature-specific late acyltransferase LpxP acts instead of LpxL to add palmitoleate (C16:1) to form hexa-acylated lipid A (17). Y. pestis carries the lpxP and lpxM genes, encoding two late acyltransferases, LpxP and LpxM, respectively, but lacks lpxL, accounting for the absence of hexa-acylated lipid A at 37°C (18, 19). The hexa-acylated lipid A predominates at 21°C to 27°C (consistent with the flea host body temperature), while mainly tetra-acylated lipid A predominates at 37°C (consistent with the mammalian host temperature) (4, 20, 21). The lpxM deletion in the Y. pestis EV strain, which mainly produces tetra-acylated lipid A, does not change its virulence significantly (22, 23). However, the Y. pestis KIM1001 strain expressing LpxL, which produces hexa-acylated structures at both 26°C and 37°C, was attenuated and could induce potent protective immunity against plague (5). Our work also confirmed previous observations that Y. pestis KIM6+(pCD1Ap) strains expressing E. coli lpxL are highly attenuated through subcutaneous (s.c.) administration (5, 24). Nevertheless, our observations indicated that the LpxL-expressing Y. pestis strain χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) induced strong inflammatory reactions causing mice to become very sick before recovery by s.c. immunization and also retained virulence via intranasal (i.n.) infection (25). The histopathological analysis at 48 h postinoculation showed that mice infected i.n. developed more severe lung lesions (unpublished data). These results suggested that hexa-acylated lipid A synthesized by χ10015(pCD1Ap) might have strong toxicity.
High-resolution X-ray crystallographic models demonstrated that the C-1 and C-4 phosphate groups on lipid A bind to basic amino acid residues on TLR4 and MD-2 and play a crucial role in dimerization of the TLR4–MD-2–LPS complex, which is necessary for induction of endotoxic or septic shock in susceptible hosts (26). Removal of a phosphate group has been shown to substantially reduce lipid A toxicity (27, 28). Monophosphoryl lipid A (phosphate in the 4′ position only) and derivatives have been used as high-potency adjuvants, retaining the immunostimulatory properties of lipid A with significantly reduced toxicity (7, 29, 30). Large amounts of highly purified 1-dephospho-Kdo2-lipid A, which has been proved useful as a novel adjuvant, can be obtained from E. coli expressing the lpxE gene of Francisella novicida (lpxE) (31). Introduction of the F. novicida lpxE gene into live S. Typhimurium leads to lipid A 1-dephosphorylation and low endotoxic activity while retaining immunogenicity (32). Therefore, we hypothesized that altering the lipid A structure of χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) (25) by introducing lpxE of F. novicida into its chromosome might reduce its toxicity and still retain its immunogenicity. Our results, however, showed that 1-dephosphorylated lipid A isolated from strain χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) exhibited a reduced capacity to activate cells in vitro. However, the mutant did not significantly reduce virulence by i.n. inoculation relative to its parental strain χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) and was actually more virulent in mice by 3 orders of magnitude through s.c. administration. In addition, expressing LpxE in wild-type Y. pestis KIM6(pCD1Ap) caused only slight attenuation by s.c. inoculation but no virulence change by i.n. inoculation in mice.
Tryptose blood agar (TBA) and heart infusion broth (HIB) were from Difco. HIB Congo red agar plates were used to confirm the pigmentation (Pgm) phenotype of Y. pestis strains (33)). Ampicillin, chloramphenicol, and l-arabinose were from Sigma (St. Louis, MO). Oligonucleotides were from IDT (Coralville, IA). Restriction endonucleases were from New England BioLabs (Ipswich, MA) unless indicated otherwise. Taq DNA polymerase (New England BioLabs) was used in all PCR tests. Vent DNA polymerase (New England BioLabs) was used to amplify fragments for cloning. Qiagen products (Hilden, Germany) were used to isolate plasmid DNA, gel purify fragments, or purify PCR products. T4 ligase, T4 DNA polymerase, and shrimp alkaline phosphatase (SAP) were from Promega (San Luis Obispo, CA).
All bacterial strains and plasmids used in this study are listed in Table 1. All strains were stored at −70°C in phosphate-buffered glycerol. Y. pestis KIM6+ (Pgm+) cells were grown routinely at 28°C on Congo red agar from glycerol stocks and then grown in HIB or on TBA (34). All the Y. pestis constructions used in this study were based on KIM6+. All E. coli strains were grown routinely at 37°C in LB broth (35) or on LB solidified with 1.2% Bacto agar (Difco). Secreted virulence factors were prepared by using a modification of a previously described method (36). Y. pestis was grown in HIB medium overnight at 26°C. The cells were then harvested and washed three times with PMH2 (33), inoculated into 40 ml of fresh PMH2 medium, and grown to an optical density at 600 nm (OD600) of 0.05 by shaking overnight at 26°C. Cultures were shifted to 37°C for 6 h with shaking to provide mild aeration.
All primers used in this paper are listed in Table 2. Primer sets lacI-1/lacI-2 and lacI-3/lacI-4 were used for amplifying the ′y3233 fragment (part of y3233, downstream of lacI) and the lamB′ fragment (part of lamB, upstream of lacI), respectively. Complementarity between primers lacI-2 and lacI-3 is indicated by bold lettering in Table 2. The ′y3233 and lamB′ fragments were fused by overlapping PCR using primers lacI-1 and lacI-4. The resulting PCR product was digested with EcoRI and HindIII and ligated into pYA4454, digested with the same enzymes, to construct the plasmid pYA4734. Plasmid pYA4734 was digested using PstI, blunted by T4 DNA polymerase, and dephosphorylated by SAP. The Plpp lpxE fragment was purified from plasmid pYA4295 using the SbfI enzyme and blunted by T4 DNA polymerase. Then, the Plpp lpxE fragment was ligated into the blunted site of pYA4734 to form plasmid pYA4735. Plasmid pYA4735 was SacI digested, blunted by T4 DNA polymerase, and dephosphorylated with SAP. The cat-sacB fragment was cut from pYA4373 using PstI and SacI restriction endonucleases and blunted by T4 DNA polymerase. The two fragments were ligated to form plasmid pYA4736. The DNA sequence was confirmed through sequencing.
Y. pestis strains with ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE and ΔlacI23::Plpp lpxE deletion-insertion mutations were constructed using the two-step recombination method as described previously (37). Perry's group suggested that lacI of Y. pestis is not functional or that there is no LacI-binding site in the promoter region of YplacZ (38), so we rationally replaced the Y. pestis lacI with Plpp lpxE in this work. Briefly, linear ′y3233-sacB-cat-Plpp lpxE-lamB′ fragments cut and gel purified from plasmid pYA4736 using EcoRI and HindIII were transformed into Y. pestis KIM6+ (33) and χ10015 (ΔlpxP32::PlpxL lpxL) containing pKD46 (25) by electroporation. Selected colonies all had the expected insertion by PCR. Then, electrocompetent cells were prepared from a sucrose-sensitive isolate and electroporated with approximately 1 μg of linear ′y3233-Plpp lpxE-lamB′ fragments purified from pYA4735 to replace the cat-sacB cassette. Plasmid pKD46 was eliminated from Y. pestis strains with ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE and ΔlacI23::Plpp lpxE deletion-insertion mutations by growth at 37°C to yield χ10027 and χ10039, respectively (Fig. 1). The mutant strains were confirmed by DNA sequence analysis. Under biosafety level 3 (BSL-3) containment, plasmid pCD1Ap was then introduced by electroporation into each strain, yielding χ10027(pCD1Ap) and χ10039(pCD1Ap), respectively.
Lipid A isolation and structure analysis were performed as described previously (25). Peaks corresponding to 1-dephosphorylated lipid A species detected from strains are likely artifacts of the acid hydrolysis as seen previously (39). The peaks corresponding to 1-dephosphorylated lipid A in the doubly charged region of the spectra include those derived from deprotonation of the phosphate as well as adducts with negatively charged ions from the solvent such as chloride, acetate, or piperidine carboxylate. The LPS was purified by extraction with 45% phenol plus triethylamine and sodium deoxycholate (40) and was used for the cell stimulation assay. The LPS was subjected to two rounds of phenol reextraction to remove contaminating TLR2-stimulating lipoproteins (41).
HEK293 cells stably expressing human TLR4/human MD-2 were obtained from InvivoGen. The HEK293-TLR4/MD-2 cells were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal calf serum, 100 IU/ml penicillin, and 100 μg/ml streptomycin, with addition of Blasticidin plus hygromycin (50 μg/ml) for HEK293-TLR4/MD-2 cells. Cells were seeded at 3 × 104 per well in 96-well tissue culture plates (Costar) and stimulated in triplicate with LPS (10 ng/ml) from E. coli strain O111:B4 (Sigma) or Y. pestis grown at 26°C or 37°C. Culture supernatants were collected after 18 h of incubation and analyzed with the human interleukin-8 (IL-8) Ready-Set-Go kits (eBioscience).
BALB/c murine macrophage cells (RAW 264.7) were cultured in DMEM supplemented with 10% fetal calf serum, 100 IU/ml penicillin, and 100 μg/ml streptomycin and incubated at 37°C in 5% CO2. RAW 264.7 cells were cultured at a density of 5 × 105 cells/ml and stimulated in triplicate with LPS (10 ng/ml) from E. coli strain O111:B4 (Sigma) or Y. pestis grown at 26°C or 37°C. Culture supernatants were collected after 18 h of incubation and analyzed with the mouse tumor necrosis factor alpha (TNF-α) Ready-Set-Go kit (eBioscience). Human THP-1 cells (ATCC TIB-202) were cultured in RPMI 1640 supplemented with 5 μM β-2-mercaptoethanol and 10% fetal calf serum and then incubated at 37°C in 5% CO2. THP-1 cells were cultured at a density of 5 × 105 cells/ml and stimulated in triplicate with LPS (10 ng/ml) from E. coli strain O111:B4 (Sigma) or Y. pestis grown at 26°C or 37°C. Culture supernatants were collected after 18 h of incubation and analyzed with the human TNF-α Ready-Set-Go kit (eBioscience).
MIC assays were performed with samples in Corning microtiter trays. Susceptibility to polymyxin B (Sigma) was determined by bactericidal assays according to a previous method with slight modification (42). For each experiment, bacteria grown overnight in HIB at 21°C or 37°C were diluted in HIB to an OD620 of 0.1. An aliquot of 10 μl of diluted culture (~5 × 105 bacteria) was added to each well of a polypropylene 96-well plate containing 100 μl of a 2-fold serial dilution of polymyxin B in HIB. The wells were examined for growth, and the MIC was determined to be the lowest concentration at which no visible growth occurred after 20 h at 26°C or 37°C. All assays were performed in triplicate.
Conversion of plasminogen to plasmin was performed as described previously (43). Briefly, strains with Pla were grown at 37°C on HIB liquid medium supplemented with 0.25 mM CaCl2. Bacteria were pelleted and resuspended in 0.1 M Tris with 0.1% Tween 20. The assays were completed in a 96-well microtiter plate (Nalge Nunc International, Rochester, NY). Each reaction was performed in triplicate, and each reaction mixture contained 5 μl of bacterial solution (approximately 2.5 × 107 bacteria), 10 μl of 4 mM chromogenic plasmin substrate S2251 (Val-Leu-Lys-p-nitroanilide) (DiaPhamar, Rochester, NY), 10 μl of 160 nM human plasminogen, and 80 μl Tris-Tween buffer. The reaction was monitored at OD405. The outer membrane fraction was extracted from Y. pestis as previously described (44).
Single colonies of each strain were used to inoculate HIB cultures and grown overnight at 26°C. To select for plasmid pCD1Ap, ampicillin was added into the medium at a concentration of 25 μg/ml. Bacteria were diluted into 10 ml of fresh HIB enriched with 0.2% xylose and 2.5 mM CaCl2 to obtain an OD620 of 0.1 and incubated at 26°C for s.c. infections (bubonic plague) or at 37°C for i.n. infections (pneumonic plague). Both cultures were grown to an OD620 of 0.6. The cells were then harvested, and the pellet was resuspended in 1 ml of isotonic phosphate-buffered saline (PBS).
All animal procedures were approved by the Arizona State University Animal Care and Use Committee. Female 7-week-old Swiss Webster mice from Charles River Laboratories (Wilmington, MA) were inoculated by s.c. injection under the skin on the back of the neck with 100 μl of bacterial suspension or by the i.n. route with 20 μl of bacterial suspension (around 10 μl/naris). Actual numbers of CFU inoculated were determined by plating serial dilutions onto TBA. To determine the 50% lethal dose (LD50), groups of six mice were infected with serial dilutions of bacteria. Mice were monitored twice daily for 21 days, and the LD50 and 95% confidence limits were calculated by probit analysis (45).
For colonization/dissemination analysis, groups of mice were inoculated either s.c. or i.n. At the indicated times after infection, 3 mice for each time point were euthanized, and samples of lungs, spleen, and liver were collected and weighed. The bacterial load for each organ was determined by plating dilutions of the homogenized tissues onto TBA plates containing 25 μg/ml ampicillin and reported as CFU per gram of tissue or CFU per ml blood. Infections were performed in at least two independent experiments.
Data are expressed as means ± standard deviations (SD). The log rank test was used for analysis of the survival curves. The nonparametric Mann-Whitney test (one-tailed) was used for colonization data. A P value of <0.05 was considered significant.
To remove the 1-phosphate group from lipid A, we inserted F. tularensis lpxE under the constitutive transcriptional control of the strong E. coli promoter Plpp into the lacI site of the Y. pestis chromosome. We constructed mutant Y. pestis strains χ10027 (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) and χ10039 (ΔlacI23::Plpp lpxE) (Fig. 1). Then, we used electrospray ionization (ESI) mass spectrometry (MS) to characterize the lipid A species made by χ10027 and its parent strain χ10015 (ΔlpxP32::PlpxL lpxL) and χ10039 and its parent strain Y. pestis KIM6+ grown at 26°C and 37°C, respectively.
The structures of bisphosphoryl hexa-acylated lipid A (I) and 1-dephosphorylated hexa-acylated lipid A (II) present in LpxL-expressing Y. pestis are shown in Fig. 2A. The presence of LpxL in χ10015 (ΔlpxP32::PlpxL lpxL) resulted in predominantly bisphosphoryl hexa-acylated lipid A decorated with two 4-amino-4-deoxy-l-arabinose (l-Ara4N) modifications, one l-Ara4N, or no l-Ara4N at m/z 1,014, m/z 949, and m/z 884, respectively (Fig. 2B and andD),D), at 26°C or 37°C as previously reported (25). Lipid A isolated from χ10027 was predominantly 1-dephosphorylated hexa-acylated lipid A with one l-Ara4N (various peaks) or no l-Ara4N (various peaks) at 26°C or 37°C, respectively. The spectra of the doubly negatively charged lipid A region contain several peaks consistent with 1-dephosphorylated lipid A ions, including the doubly deprotonated ion [M − 2H]2− as well as noncovalent salt adducts [M − H + acetate]2−, [M − H + piperidine-carboxylate]2−, and [M − H + Cl]2−, as previously reported for 1-dephosphorylated lipid A from Salmonella (32). The assignment of the peaks was confirmed with ESI-tandem MS (MS/MS) (data not shown). The m/z values for the lipid A and adduct ions are summarized in Table S1 in the supplemental material. The predominance of 1-dephosphorylated species of lipid A is consistent with the action of LpxE due to insertion of Plpp lpxE into the chromosome (Fig. 2C and andEE).
Because of the presence of an intact Y. pestis lpxP gene, the mass spectra of the lipid A isolated from KIM6+ grown at 26°C contained peaks corresponding to hexa-acylated lipid A decorated with two, one, or zero l-Ara4N modifications at m/z 1,041, m/z 976, and m/z 910, respectively, in addition to significant peaks consistent with tetra-acylated (m/z 832) and penta-acylated (m/z 924) lipid A (Fig. 3A). At 37°C, lipid A isolated from Y. pestis KIM6+ was predominantly tetra-acylated lipid A (lipid IVA) as previous reported (20, 25) (Fig. 3C). For χ10039, the expression of LpxE shifted the lipid A profile to produce peaks consistent with predominantly 1-dephosphorylated tetra-acylated lipid A species (m/z 726.9). The level of l-Ara4N modification is higher for both strains at 26°C than at 37°C, consistent with Fig. 2 and previous reports (25).
The lpxM mutation in the live Y. pestis vaccine strain EV NIIE and Y. pestis strain 231 had pleiotropic effects on expression and immunoreactivity of numerous major protein and carbohydrate antigens, including F1, Pla, Ymt, V antigen, LPS, and enterobacterial common antigen (ECA) (46). This information suggested that variation of LPS structure in Y. pestis might affect expression of some virulence factors. In order to evaluate this possibility, we compared LcrV and YopM synthesis and secretion in the wild-type strain Y. pestis KIM6+(pCD1Ap) and the different LPS modification strains, χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL), χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE), and χ10039(pCD1Ap) (ΔlacI23::Plpp lpxE). Two additional LPS-related mutants, χ10013(pCD1Ap) (ΔlpxP32) and χ10031(pCD1Ap) (ΔrfaC725), were also included in these assays. The lipid A structures of χ10013 (ΔlpxP32) and χ10031 (ΔrfaC725) (see Fig. S3 and S4 in the supplemental material) were consistent with Y. pestis 231 with an lpxP deletion and a waaC (rfaC) deletion as described by Anisimov et al. (47, 48). Results showed that there were no significant differences in LcrV levels in whole-cell lysates or in YopM secretion in each of the strains tested, except that the LcrV secretion level was significantly reduced in χ10031(pCD1Ap) (see Fig. S1).
Previous work indicated that some LPS modifications affected the sensitivity of Gram-negative bacteria to cationic antimicrobial peptides (19, 49). Here, we compared the MICs of polymyxin B for Y. pestis KIM6+, χ10013, χ10015, χ10027, χ10031, and χ10039 grown at 26°C and 37°C in HIB medium to determine the susceptibility of those strains to polymyxin B. The results showed that all strains grown at 26°C were less susceptible to polymyxin B than were strains grown at 37°C (Table 3). The MICs of polymyxin B were equivalent in Y. pestis KIM6+, χ10013 (ΔlpxP32), and χ10015 (ΔlpxP32::PlpxL lpxL). In contrast, strains χ10027 (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) and χ10039 (ΔlacI23::Plpp lpxE) with LpxE expression were more sensitive to polymyxin B than were their parent strains χ10015 and KIM6+, respectively. Strain χ10031, as a deep-rough mutant, lacks the outer core of lipid A and was, as expected, highly sensitive to polymyxin B (Table 3).
Research demonstrated that the temperature-induced changes in LPS potentiated Pla-mediated proteolysis in Y. pestis (50). Plasminogen activation and Pla synthesis were dramatically higher in Y. pestis grown at 37°C than in that grown at 20°C (51–53). Therefore, Pla activities were compared among different mutants only at 37°C. Strain χ10023 (Δpla-525) was used as the negative control. Our results showed that Pla activities of χ10013 (ΔlpxP32) and χ10015 (lpxL insertion) did not vary significantly relative to that of the parent strain (KIM6+). In contrast, χ10027 and χ10039 (both synthesizing LpxE) had significantly decreased Pla activity relative to their parent strains, with χ10027 having a higher level of Pla activity than χ10039 (Fig. 4A). Compared with KIM6+, the Pla activity in χ10031 (ΔrfaC725) was very low (Fig. 4B). Dentovskaya et al. reported that Pla activity was totally abolished in the Y. pestis 231 waaC (rfaC) mutant (48). Pla is an outer membrane protein and plasminogen activator, and its level of synthesis would be expected to affect its activity. Therefore, we compared Pla syntheses in the whole-cell lysate and outer membrane fractions among those strains. Relative to the wild-type strain KIM6+, Pla synthesis was not significantly different in χ10013 (ΔlpxP32) or χ10015 (ΔlpxP32::PlpxL lpxL) in the whole-cell lysate and outer membrane fractions. Pla synthesis in χ10027 (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) did not show very significant changes in the Pla levels from the whole-cell lysate but was slightly reduced in the outer membrane fractions. Pla synthesis in χ10039 (ΔlacI23::Plpp lpxE) was significantly lower (Fig. 4B). In addition, the level of Pla synthesis was near the limit of detection in χ10031 (ΔrfaC725) (Fig. 4B).
In agreement with the temperature-dependent shift from primarily hexa-acylated lipid A at 21°C to tetra-acylated lipid A structures at 37°C, Kawahara et al. (20) reported that lipid A from Y. pestis grown at 37°C stimulated less TNF-α secretion in both murine and human macrophage cell lines than did lipid A from Y. pestis grown at 27°C. Recently, Montminy et al. (5) and Telepnev et al. (54) further demonstrated that tetra-acylated lipid A produced in Y. pestis at 37°C had poor TLR4-stimulating activity and stimulated less TNF-α secretion in human peripheral blood mononuclear cells and dendritic cells than did lipid A produced at 26°C.
Here, we evaluated immunostimulatory properties of LPS isolated from different Y. pestis strains grown at 26°C or 37°C. Purified LPS was incubated with transfected HEK293 cell lines expressing TLR4 and MD-2. The results indicated that high levels of human IL-8 were observed in cultures in which cells were exposed to LPS isolated from E. coli (positive control) and Y. pestis KIM6+ grown at 26°C and χ10015 (ΔlpxP32::PlpxL lpxL) grown at both 26°C and 37°C. Relative to χ10015, LPS isolated from χ10027 (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) grown at both 26°C and 37°C stimulated significantly lower levels of human IL-8 (Fig. 5A). Low levels of IL-8 were present in cultures in which cells were exposed to LPS isolated from Y. pestis KIM6+ grown at 37°C and χ10013 and χ10039 grown at both 26°C and 37°C (Fig. 5A). No IL-8 was detected in cells exposed to PBS as the negative control.
The profiles of TNF-α in murine macrophage cells (RAW 264.7) and human THP-1 cells exposed to LPS isolated from those strains were very similar to the profiles in transfected HEK293 cell lines expressing TLR4 and MD-2 (Fig. 5B and andC).C). LPS from χ10015 and χ10027 led to potent stimulation of TNF-α levels at both 26°C and 37°C, while LPS from χ10013 and χ10039 led to weak stimulation of TNF-α levels at both 26°C and 37°C, consistent with previous reports for wild-type Y. pestis (4, 5). Together, these results confirmed that 1-dephosphorylation of hexa-acylated lipid A muted the inflammatory response of mammalian cells to Y. pestis LPS.
To investigate the contribution of modified lipid A to Y. pestis virulence, we infected Swiss Webster mice with Y. pestis KIM6+(pCD1Ap), χ10013(pCD1Ap) (ΔlpxP32), χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL), χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacZ123::Plpp lpxE), and χ10039(pCD1Ap) (ΔlacI23::Plpp lpxE) by s.c. and i.n. administration. Table 4 lists the LD50 for each strain. The s.c. and i.n. LD50s of Y. pestis KIM6+(pCD1Ap) (s.c., <10 CFU, and i.n., ~102 CFU) were consistent with our previous results (25). The ΔlpxP32 single mutant χ10013(pCD1Ap) had the same s.c. and i.n. LD50s as did Y. pestis KIM6+(pCD1Ap). The s.c. and i.n. LD50s of χ10015(pCD1Ap) (Δlpxp32::PlpxL lpxL) (s.c., ~107 CFU, and i.n., 2.7 × 104 CFU) were the same as our previous report (25). Compared with χ10015(pCD1Ap), the i.n. LD50 of χ10027(pCD1Ap) was not significantly different but the s.c. LD50 of χ10027(pCD1Ap) decreased ~103-fold. These results suggest that 1-dephosphorylated hexa-acylated lipid A from χ10027(pCD1Ap) lacks sufficient immunostimulatory properties in s.c. infection for efficient clearance of bacteria.
Our results did not rule out the possibility that LpxE expression in χ10027(pCD1Ap) might increase virulence. To address this possibility, the virulence of LpxE synthesis in Y. pestis KIM6+(pCD1Ap) was determined. Our results demonstrated that the s.c. LD50 of χ10039(pCD1Ap) (ΔlacI23::Plpp lpxE) increased ~50-fold compared with wild-type Y. pestis KIM6+(pCD1Ap), and the i.n. LD50 of χ10039(pCD1Ap) increased less, 10-fold compared to its parental strain. These results suggest that LpxE expression alone does not alter the virulence of Y. pestis but rather that lipid A 1-dephosphorylation significantly reverses the reduction in virulence of the Y. pestis strain with constitutive hexa-acylated lipid A due to LpxL expression.
To gain further insight into the differential virulence of the LpxE-expressing strains, we evaluated the ability of those mutant bacteria to disseminate and colonize target organs. We infected mice via the s.c. route with χ10013(pCD1Ap), χ10015(pCD1Ap), χ10027(pCD1Ap), χ10039(pCD1Ap), or the wild-type strain Y. pestis KIM6+(pCD1Ap). The bacterial burdens in lung, spleen, and liver were determined at days 2, 4, and 6 after s.c. infection with ~103 CFU of each strain (Fig. 6). Both wild-type KIM6+(pCD1Ap) and χ10013(pCD1Ap) (ΔlpxP32) reached similar titers of 1 × 104 to 1 × 105 CFU in the lung, spleen, and liver at 2 days postinfection. The numbers of KIM6+(pCD1Ap) and χ10013(pCD1Ap) organisms steadily increased at 4 and 6 days postinfection. The titers of χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) in the lung, spleen, and liver were very low at 2 days postinfection, and almost no bacteria were recovered from the organs at 4 and 6 days postinfection. Strain χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) was found in the lung, spleen, and liver at 2 and 4 days postinfection, and its colonization efficiency in different organs was around 10- to 100-fold lower than that of KIM6+(pCD1Ap) but higher than that of its parent strain χ10015(pCD1Ap). The titers of χ10027(pCD1Ap) recovered in the lung, spleen, and liver at 6 days postinfection were ~103- to 104-fold lower than those for KIM6+(pCD1Ap). A modest decrease in bacterial burden in the lung, spleen, and liver for χ10039(pCD1Ap) (ΔlacI23::Plpp lpxE) was seen by 2, 4, and 6 days after inoculation compared with those for wild-type strain KIM6+(pCD1Ap) and χ10013(pCD1Ap) (ΔlpxP32).
We also determined bacterial burdens in the different organs after i.n. administration. Mice infected with roughly the same dose of each strain (1.5 × 104 CFU) were euthanized at 12, 24, and 48 h postinfection to determine the bacterial loads in lung, spleen, and liver (Fig. 7). KIM6+(pCD1Ap) and χ10013(pCD1Ap) reached very similar bacterial loads in the lung, spleen, and liver at all the times postinfection. At 12 h postinfection, there was no major difference in the ability of χ10039(pCD1Ap) (ΔlacI23::Plpp lpxE) to establish infection within the lungs compared with those of KIM6+(pCD1Ap) and χ10013(pCD1Ap) (ΔlpxP32). While χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) and χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacI23::Plpp lpxE) displayed a modest but significant reduction in bacterial counts in the lung compared with KIM6+(pCD1Ap), χ10013(pCD1Ap), and χ10039(pCD1Ap), this decreased bacterial burden was constant in the lungs throughout the course of the experiment (Fig. 7A). By 24 and 48 h postinfection, χ10015(pCD1Ap) was present in the lungs, but the numbers were significantly lower than those of KIM6+(pCD1Ap), χ10013(pCD1Ap), and χ10039(pCD1Ap) (Fig. 7A). At 24 and 48 h postinfection, there were significant differences between bacterial counts of χ10027(pCD1Ap) and those of KIM6+(pCD1Ap). The numbers of χ10027(pCD1Ap) bacteria in the lungs of infected mice at 48 h postinfection were approximately 1,000-fold lower than those in wild-type-infected animals (Fig. 7A). Compared with KIM6+(pCD1Ap), the χ10039(pCD1Ap) burden in the lung was only modestly lower by 48 h postinfection (Fig. 7A).
The bacterial burden of both χ10015(pCD1Ap) and χ10027(pCD1Ap) in the spleens and livers was lower than those of KIM6+(pCD1Ap) and χ10013(pCD1Ap) (ΔlpxP32) (Fig. 7B and andC).C). By 48 h postinfection, the bacterial numbers in the spleens and livers of mice infected with the mutant χ10015(pCD1Ap) or χ10027(pCD1Ap) were about 10- to 100-fold lower than those in KIM6+(pCD1Ap)-infected mice (Fig. 7B and andC).C). At 12 h postinfection, the bacterial burden of χ10039(pCD1Ap) (ΔlacZ123::Plpp lpxE) in the spleens and livers was slightly lower than that of KIM6+(pCD1Ap), but the numbers of χ10039(pCD1Ap) in the spleens and livers were similar to those of KIM6+(pCD1Ap) and χ10013(pCD1Ap) at late infection stages (Fig. 7B and andCC).
The lipid A extraction from KIM6+ and χ10015 in this work is a different batch from our previous paper (25), so that lipid A MS of KIM6+ and χ10015 has some differences from our previous work (25). However, our results still verified that in Y. pestis KIM6+, lipid A secondary acylations of C16:1 (LpxP) occurred only at 26°C but not at 37°C (4, 19, 25). The lpxP deletion strain χ10013 (ΔlpxP32) lacks the palmitoleoyl (C16:1) transferase, leaving tetra-acylated lipid IVA as the predominant lipid A species (see Fig. S3 in the supplemental material). The lpxL insertion into the lpxP locus in strain χ10015 results in the production of hexa-acylated lipid A containing a 2′ secondary C12:0 chain, as previously reported (5, 25). LpxE, an inner membrane phosphatase from Francisella novicida, can selectively remove the 1-phosphate group of lipid A in E. coli and Salmonella (32, 55). Our results also indicated that lpxE introduced into the chromosome of Y. pestis can lead to marked lipid A 1-dephosphorylation. The result is consistent with a recent report examining the lipid A from Y. pestis containing a plasmid expressing lpxE (56).
Resistance of Y. pestis to cationic antimicrobial peptides depends on the content of l-Ara4N, whose temperature-dependent incorporation into lipid A is regulated by the two-component PhoP-PhoQ regulatory system (4, 57, 58). Our results, which are consistent with a previous report (20), showed that the LPS was more extensively modified with l-Ara4N at 26°C than at 37°C, consistent with the susceptibility to polymyxin B (Table 3). Normally, in Y. pestis, the lipid A backbone is modified with one or two l-Ara4Ns on the phosphate groups (59, 60). In E. coli (31) and Helicobacter pylori (61), the removal of the 1-phosphate decreases the susceptibility to polymyxin B. The presence of LpxE leads to the removal of the 1-phosphate group from lipid A in χ10027 and χ10039, but unlike E. coli and H. pylori, this leads to increased susceptibility to polymyxin B (Table 3). One difference with Y. pestis is that the lipid A seems to be at least partially modified by l-Ara4N at both 26°C and 37°C, in contrast to E. coli, which normally lacks l-Ara4N modification at all temperatures. Perhaps the preexisting l-Ara4N modification would cause the effects of lipid A 1-dephosphorylation to be less dramatic in Y. pestis than in E. coli. Anisimov et al. showed that the efficiency of l-Ara4N transfer to lipid A in LPS with an incomplete inner core was lower than that in LPS with a complete inner core, as observed in the waaA, waaC, or waaE mutants that were highly susceptible to polymyxin B (47). Our results also indicated that the rfaC (waaC) deletion strain χ10031 was highly sensitive to polymyxin B (Table 3). This may be in part due to the loss of stabilizing LPS cross-links mediated through the core, leading to a defect in the permeability barrier (62).
Pla is a surface protease encoded by the Y. pestis-specific plasmid pPCP1 (63, 64). Its activity is related to the LPS structure (50, 65). Low Pla activity was observed in χ10031 (ΔrfaC) (Fig. 4A), which is a deep-rough LPS mutant with one Kdo and one Ko (d-glycero-d-talo-oct-2-ulosonic acid) residue attached to lipid A. Our results verify that the presence of a complete core is critical for Pla activity, which is consistent with previous reports (47, 48, 50). Pla synthesis in χ10031 is much lower than that in other strains (Fig. 4B). This may be the major reason for low Pla activity in χ10031, but the mechanism of rfaC mutation affecting Pla synthesis is unclear. The synthesis levels of Pla in Y. pestis strains KIM6+, χ10013, and χ10015 are similar (Fig. 4B), which suggested that acylation in LPS did not significantly affect the synthesis level or activity of Pla. A previous report also showed that LpxL synthesis in Y. pestis did not affect Pla activity (5). Additionally, Pla belongs to a unique family of integral outer membrane proteases known as omptins, which are widely distributed within the family Enterobacteriaceae (66). A conserved feature in omptins is the presence of a three-dimensional protein motif for binding to lipid A phosphates (66, 67). Pla also contains the three-dimensional motif for protein binding to lipid A phosphates (65, 68). Incorporation of LpxE (lipid A 1-phosphatase) into Y. pestis to construct strains χ10027 (ΔlpxP32::PlpxL lpxL ΔlacZ123::Plpp lpxE) and χ10039 (ΔlacI23::Plpp lpxE) significantly decreased Pla activity (Fig. 4A) and slightly affected Pla synthesis (Fig. 4B). However, comparison of different strains cultured at 26°C or 37°C indicated that those genetic manipulations did not significantly affect their growth rate (see Fig. S2 in the supplemental material). Our results suggested that the bacterial growth was not a major factor that affected Pla activity and Pla synthesis but that the phosphate group of lipid A (a possible binding motif for Pla) affected Pla activity.
LPS purified from KIM6+ grown at 37°C is a predominantly tetra-acylated LPS, perhaps with minor amounts of penta-acylated LPS, which has very poor stimulatory activity toward HEK293 cells expressing TLR4 and MD-2 or other cell lines (6, 54). Our results shown in Fig. 5 were not completely consistent with these reports. The LPS from KIM6+ grown at 37°C still had significant immune stimulation in such cell lines, which might be caused by some dubious factors such as LPS extraction, LPS purity, or others. The Fig. 5 results showed that activation by LPS from the strain synthesizing LpxL was significantly different than that by LPS from KIM6+ grown at 37°C, while LPS from strains synthesizing LpxE, with or without LpxL, was not. LPS from χ10013 (ΔlpxP32) and χ10039 (ΔlacI23::Plpp lpxE) had low levels of immunostimulatory properties in vitro (Fig. 5). The likely reason for this observation is that the LPS from χ10013 (ΔlpxP32) is tetra-acylated at all temperatures and that the LPS from χ10039 (ΔlacI23::Plpp lpxE) is 1-dephosphorylated at all temperatures.
Strain χ10027(pCD1Ap), which produces 1-dephosphoryl hexa-acylated lipid A (Fig. 2), had a significantly reduced LD50 compared with its parent strain χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL), with a 1,000-fold-lower LD50 by the s.c. administration route (Table 4). Consistent with the virulence, colonization data also showed that χ10027(pCD1Ap) was better able to colonize infected mice than was the parent χ10015(pCD1Ap) (Fig. 6). It seems likely that just as Y. pestis with tetra-acylated lipid A is virulent because it does not provoke a significant innate immune response, so too Y. pestis with 1-dephosphorylated hexa-acylated lipid A may also fail to provoke the full innate immune response required to clear the bacteria. This is in contrast to what is seen in Salmonella, where strains expressing 1-dephosphorylated hexa-acylated lipid A are dramatically less virulent. One of the key differences is that Y. pestis seems to have multiple mechanisms to attenuate the innate immune response. Y. pestis has a virulence plasmid which encodes a type III secretion apparatus for the translocation of Yops (suppressing the innate immune response) into target immune cells (2, 3). Therefore, the modest immune response induced by 1-dephosphorylated hexa-acylated lipid A from χ10027(pCD1Ap) might not be enough to overcome the suppressed immune response triggered by Y. pestis virulence factors. In addition, research indicates that Pla can modulate the susceptibility of Y. pestis to pulmonary antimicrobial peptides (69). Therefore, reduced Pla synthesis in strain χ10027(pCD1Ap) (ΔlpxP32::PlpxL lpxL ΔlacZ123::Plpp lpxE) may impair its colonization in the lung relative to that of its parent strain χ10015(pCD1Ap) (ΔlpxP32::PlpxL lpxL) (Fig. 7A).
Researches have indicated that Pla promotes the invasion of Y. pestis from subcutaneous sites of inoculation into the lymphatic system and deeper tissues in models of bubonic plague (43, 70) and that inactivation of Pla increased the LD50 of Y. pestis for mice a millionfold (43, 71), but Y. pestis lacking Pla was reported to be nearly equivalent in virulence to the wild type by aerosol infection (72, 73). The LpxE synthesis in χ10039(pCD1Ap) (ΔlacZ123::Plpp lpxE) significantly reduced Pla activity (Fig. 4A), which might be the reason for the moderate LD50 increase for s.c. infection (Table 4).
In summary, our findings build upon the observation that Y. pestis expressing LpxL is dramatically less virulent than wild-type Y. pestis. The hexa-acylated lipid A produced by that strain was still sufficiently toxigenic that we sought to further reduce its endotoxicity through lipid A 1-dephosphorylation. While expression of LpxE based on the diphosphoryl hexa-acylated strain was confirmed to lead to lipid A 1-dephosphorylation, it reduced stimulatory properties of LPS and also the bacterial LD50. Therefore, our work to understand the complex relationship of the LPS structure and pathogenicity of Y. pestis will be essential for rationally designing safe and effective vaccines against this pathogen.
We thank Susan Straley for providing anti-YopM antibodies, Jon D. Goguen for providing anti-Pla antibodies, and Praveen Alamuri for assisting with animal experiments.
This work was supported by National Institutes of Health grant 5R01 AI057885 to R.C. and by grants GM51310 and GM-069338 to C.R.H.R. The mass spectrometry facility in the Department of Biochemistry of the Duke University Medical Center is supported by the LIPID MAPS Large Scale Collaborative Grant number GM-069338 from NIH.
All authors declare no conflicts of interest.
Published ahead of print 28 January 2013
Supplemental material for this article may be found at http://dx.doi.org/10.1128/IAI.01403-12.