Staphylococcus aureus Strains Lacking d-Alanine Modifications of Teichoic Acids Are Highly Susceptible to Human Neutrophil Killing and Are Virulence Attenuated in Mice

Abstract

Staphylococcus aureus is resistant to α-defensins, antimicrobial peptides that play an important role in oxygen-independent killing of human neutrophils. The dlt operon mediates d-alanine incorporation into teichoic acids in the staphylococcal cell envelope and is a determinant of defensin resistance. By using S. aureus wild-type (WT) and Dlt⁻ bacteria, the relative contributions of oxygen-dependent and -independent antimicrobial phagocyte components were analyzed. The Dlt⁻ strain was efficiently killed by human neutrophils even in the absence of a functional respiratory burst, whereas the killing of the WT organism was strongly diminished when the respiratory burst was inhibited. Human monocytes, which do not produce defensins, inactivated the WT and Dlt⁻ bacteria with similar efficiencies. In addition, mice injected with the Dlt⁻ strain had significantly lower rates of sepsis and septic arthritis and fewer bacteria in the kidneys, compared with mice infected with the WT strain

Mucosal membranes and skin are efficient barriers against infections, in part because of the production of antimicrobial peptides, such as human β-defensin–2 (HBD-2), that contribute to protection against microbial pathogens (reviewed in [1–3]). Staphylococcus aureus a major cause of community- and hospital-acquired infections, has evolved means to circumvent host defenses [4, 5]. S. aureus exhibits resistance to HBD-2 [6] and can invade epithelial and endothelial cells [7, 8], thereby allowing the pathogen to persist by “hiding” from phagocytes and antibiotics

Once bacteria infect the subepithelial tissues, polymorphonuclear leukocytes (PMNL) are recruited by the host to prevent further spread of the infection. Oxygen-dependent and -independent mechanisms are used to eliminate ingested bacteria [9]. The former strategy is mediated by toxic oxygen intermediates generated by NADPH oxidase and myeloperoxidase (MPO) [10], the latter by antimicrobial proteins and peptides, such as defensins and cathelicidins, stored inside the PMNL granules. The importance of these peptides was recently underscored by the profound susceptibility of a cathelicidin-deficient mouse strain to bacterial infections [11] and by the increasing number of reports on bacterial mechanisms conferring resistance to this human antimicrobial peptide class [12]

The ability of S. aureus to resist neutrophil killing has a profound influence on its virulence [13]. The PMNL of patients with inherited oxidative burst deficiency (chronic granulomatous disease [CGD]) are inefficient at killing S. aureus [14], since these bacteria are insensitive to many neutrophilic bactericidal components, such as lysozyme and α-defensins [15]. The molecular basis for the defensin resistance of S. aureus however, has remained elusive. We previously described 2 loci on the S. aureus chromosome, dltABCD [16] and mprF [17], whose disruption abolished the resistance of S. aureus to positively charged antimicrobial peptides, such as defensins and other host defense peptides. DltA, DltB, DltC, and DltD catalyze the introduction of d-alanine into teichoic acids, staphylococcal cell wall polymers, whereas MprF is involved in modification of membrane phosphatidylglycerol with l-lysine. Esterification of cell envelope components with amino acids leads to a decrease in the net negative surface charge of the bacteria and, consequently, to the repulsion of positively charged antimicrobial peptides [16, 17]. The lack of d-alanine esters had further remarkable consequences for the Dlt⁻ mutant, such as increased susceptibility to vancomycin and other glycopeptide antibiotics [18] and a dramatic reduction in its ability to colonize plastic and glass surfaces, an important factor in catheter-associated staphylococcal infections [19]

In an attempt to better understand the relevance of DltABCD-mediated evasion of host defenses for S. aureus infections, we compared the killing of wild-type (WT) and Dlt⁻ bacteria by isolated neutrophil components and by human neutrophils and monocytes. We also assessed the virulence of the 2 strains in a sepsis/septic arthritis mouse model

Materials and Methods

Growth conditions and bacterial strainsAll bacterial strains were grown in Luria broth (LB; 1% tryptone, 0.5% yeast extract, and 0.5% NaCl) or BM broth (LB supplemented with 0.1% K₂HPO₄ and 0.1% glucose) at 37°C, unless otherwise noted. Disruption of the dlt operon in S. aureus Newman was accomplished by homologous recombination, using the temperature-sensitive recombination vector pBTDdlt1. The plasmid construction protocol and the procedures used for recombination and phenotypic and genotypic characterizations of the Dlt⁻ mutant have been described in detail for S. aureus Sa113 [16]

Bacterial killing by MPO and defensin HNP1-3To prepare bacteria for killing experiments, Mueller-Hinton broth was inoculated with 0.01 volumes of an overnight bacterial culture and shaken vigorously at 37°C until the midlogarithmic phase was reached. The bacteria were washed twice with PBS containing 0.0005% human serum albumin (HSA) for incubation with MPO or 3 times with potassium phosphate buffer (10 mM pH 7.5) containing 0.0005% HSA for incubation with HNP1-3. Bacteria at a concentration of 3×10⁵ cfu/mL were incubated with 0.05 U/mL of MPO (Calbiochem) plus 10 μM H₂O₂ or with 100 mg of HNP1-3/mL (gift of H. Kalbacher, University of Tübingen) in the same buffers, after preheating all compounds at 37°C. We found 154 mM NaCl from the PBS present in MPO-containing samples, which permitted generation of toxic chlorinating compounds [20]. Samples (10 mL each) were shaken at 37°C, and killing was stopped after 45 min (MPO) or 60 min (HNP1-3) by 25-fold dilution in ice-cold PBS or potassium phosphate buffer, respectively. Viable bacteria were counted 24 h after plating appropriate dilutions on Luria agar (L. agar)

Teichoic acid analysisStaphylococcal teichoic acids were isolated and analyzed, as described elsewhere [16]. In brief, bacteria were grown overnight in BM broth containing 0.3% glucose, resuspended in sodium acetate buffer (20 mM pH 4.6), and disrupted by use of glass beads. Cell walls were prepared by extraction of crude cell lysates with 2% SDS, and cell wall teichoic acids were isolated by use of 5% trichloroacetic acid. The amounts of phosphorus and hexosamines in teichoic acid samples were determined by colorimetric assays, as described elsewhere [21, 22]. We propagated phages 3A and f11 in S. aureus Sa113, following standard procedures [23], and studied their activity by dropping phage suspensions on lawns of S. aureus strains

Killing of S. aureus by human phagocytesBlood drawn from healthy human volunteers was heparinized. Neutrophils were isolated from peripheral blood by use of ficoll-histopaque gradients, as described elsewhere [24]. Monocytes, defined as CD14-positive cells, were purified by using a monocyte isolation kit (Dynal Biotech) and were determined by flow cytometric analysis (FACScan; Becton Dickinson) to be >90% pure. Cells were resuspended to a concentration of 5×10⁶ cells/mL in Hanks’ balanced salt solution containing 0.05% HSA (HBSS-HSA). Bacteria were grown as described above for in vitro killing assays, were washed twice in HBSS-HSA, and were adjusted to a density of 8.5×10⁶ bacteria/mL. Normal human serum was added to a final concentration of 4%, and bacteria were opsonized for 10 min at 37°C. Prewarmed bacterial and neutrophil or monocyte suspensions were mixed to final concentrations of 8.5×10⁴ cfu/mL and 2.5×10⁶ leukocytes/mL. To inhibit the neutrophil NADPH oxidase, 20 μM diphenyleniodonium (DPI) was added immediately before the bacteria and neutrophils were mixed. Samples (50 mL) were shaken at 37°C, and incubation was stopped at various time points by the addition of 2 mL of ice-cold distilled water to disrupt the leukocytes. Appropriate sample volumes were spread on L. agar plates, and colonies were counted after incubation for 24 h at 37°C. When bacteria were incubated for 4 h under the same conditions without neutrophils, we observed no significant changes in colony numbers, compared with the initial counts

Inhibition of the oxidative burst was confirmed by monitoring the luminol-enhanced chemiluminescence of neutrophils in the presence or absence of 20 μM DPI, as described elsewhere [24]. Neutrophils were stimulated with S. aureus Sa113 cells that had been prepared and opsonized as described above

Phagocytosis studiesFor phagocytosis studies, bacteria were grown as described above, washed, resuspended in PBS, and inactivated by heating for 25 min at 70°C. Subsequently, 0.1 mg/mL of fluorescein isothiocyanate (FITC) was added, and the bacteria were labeled at 37°C for 5 h. After a wash with PBS, the bacteria were resuspended in HBSS-HSA, adjusted to the same density, and opsonized, as described above. We mixed and shook 50-mL aliquots of the prewarmed bacterial and leukocyte suspensions at 37°C. The final concentrations were 8.5×10⁷ cfu/mL or 8.5×10⁶ cfu/mL and 2.5×10⁶ leukocytes/mL for incubation with neutrophils or 10⁷ cfu/mL and 10⁶ leukocytes/mL for incubation with monocytes. Incubation was stopped after 8 min by the addition of 100 mL of ice-cold 1% paraformaldehyde. The percentage of leukocytes bearing FITC-labeled bacteria was determined by flow cytometric analysis of 5000 cells (FACScan) or by fluorescence microscopy (Zeiss). We quenched the extracellular fluorescence by adding 0.1 mg/mL of trypan blue, to subtract the noningested leukocyte-associated bacteria. The curves obtained in the presence and absence of trypan blue were superimposable, which demonstrated that the number of adherent, noningested bacteria was negligible

Animal studiesFemale NMRI mice aged 5–7 weeks were injected in the tail vein with a bacterial suspension containing 1.8×10⁷ cfu of either S. aureus Newman WT or the isogenic Dlt⁻ mutant and were evaluated for weight loss, arthritis, and sepsis over 7 days, as described elsewhere [17, 25]. Blood samples were taken from the tail vein 24 h after infection, diluted, and plated for viable bacterial counts on blood agar plates. Mice were killed at postinoculation day 7, and their kidneys were removed to assay bacteria loads by plating serial dilutions of the organs, homogenized in PBS, on blood agar plates. The differences between the means of the values for weight loss and bacteria loads in blood and kidneys were tested for significance, using the 2-tailed version of Student’s t test. The differences in mortality rate and frequency of arthritis between the groups were analyzed by Fisher’s exact test. P⩽.05 was considered to be statistically significant

Results

In vitro killing of the Dlt⁻ mutant by α-defensins and MPO-dependent mechanismsWe have shown that human α-defensin HNP1-3 inhibited the growth of an S. aureus Sa113 Dlt⁻ mutant but had no effect on WT bacteria [16]. To investigate the potential host defense mechanism involved in this phenomenon, we analyzed the sensitivity of the bacteria to neutrophil-derived antimicrobial components. The mutant was effectively inactivated by HNP1-3, whereas WT bacteria were not affected (figure 1), which demonstrates that the lack of d-alanine esters leads not only to inhibition but also to efficient killing of staphylococci

Figure 1

Inactivation of Staphylococcus aureus strains by isolated neutrophil defensins and by myeloperoxidase (MPO). Nos. (in cfu) of viable wild-type (black bars) and Dlt⁻(gray bars) bacteria after incubation with defensin HNP1-3 or with MPO are expressed as percentages of the initial counts (means+SE of 4 counts of a representative experiment)

Because normal neutrophils use oxygen-dependent killing mechanisms in addition to defensins, we compared the rate of inactivation of S. aureus Sa113 WT and Dlt⁻ strains by human neutrophil MPO. Bacteria were incubated with MPO in the presence of H₂O₂ and chloride ions to permit the generation of toxic oxidizing and halogenizing products [20]. The Dlt⁻ mutant was killed to a greater extent than the WT bacteria (figure 1), indicating that increased susceptibility to antimicrobial phagocyte substances due to defective d-alanine substitution of teichoic acids is not restricted to defensins

Teichoic acid analysisAlthough earlier studies have demonstrated the lack of teichoic acid d-alanine in the Dlt⁻ strain [16], possible differences between parent and mutant strains in the amount of N-acetylglucosamine (GlcNAc) substituents have not, to our knowledge, been analyzed. Since d-alanine and GlcNAc are both attached to the hydroxyl groups of ribitol in cell wall teichoic acids [26], the lack of d-alanine might affect the level of GlcNAc and, hence, influence interactions with host factors. However, the ratios of cell wall teichoic acid phosphate and GlcNAc in the 2 strains were found to be equal (table 1), which indicates that the loss of alanylation does not affect the level of substitution with GlcNAc. Since the cell walls of the 2 strains contain very similar amounts of phosphate, the teichoic acid content of the Dlt⁻ strain is most probably unaltered. Previous studies have indicated that cell wall teichoic acids and their GlcNAc substituents are the receptor for most S. aureus phage types [27]. Both, the WT and Dlt⁻ strains were susceptible to phages 3A and f11 from the serologic groups A and B, respectively (table 1), which indicates that the d-alanine residues are not involved in phage recognition

Table 1

Analysis of teichoic acids from Staphylococcus aureus Sa113 wild-type (WT) and Dlt− bacteria

Increased killing of the Dlt⁻ mutant by human neutrophils.Since human neutrophils use both α-defensins and MPO to kill ingested microorganisms, the susceptibilities of the mutant and WT strains to neutrophil killing were compared. Bacteria in log-phase growth were opsonized with normal human serum and incubated with human neutrophils for various time intervals, after which the numbers of surviving bacteria were determined. The Dlt⁻ mutant was killed considerably faster than WT organisms (figure 2A). The number of applied WT or Dlt⁻ bacteria was reduced by 50% after 100 or 12 min, respectively. The 2 strains appear to be ingested by neutrophils with similar efficiencies (figure 2B). Therefore, the difference in killing kinetics most probably reflects increased susceptibility of Dlt⁻ bacteria to neutrophil antimicrobial activities rather than increased phagocytosis

Figure 2

Killing (A) and phagocytosis (B) of Staphylococcus aureus strains by human phagocytes. A Nos. (in cfu) of viable opsonized wild-type (black circles) and Dlt⁻(gray circles) bacteria after incubation with human neutrophils or monocytes are expressed as percentages of the initial counts of a representative experiment. B No. of neutrophils containing wild-type (black bars) or Dlt⁻(gray bars) bacteria after incubation with different nos. of bacteria are shown as percentages of total polymorphonuclear leukocytes (PMNL) from a representative experiment

Inhibition of oxygen-dependent killing mechanisms disrupts human neutrophil killing of WT but not Dlt⁻ bacteriaThe respiratory burst of neutrophils was inhibited by DPI, which inactivates the neutrophil NADPH oxidase and thereby prevents the formation of the MPO substrate H₂O₂ [28, 29]. The addition of 20 μM DPI completely abolished respiratory burst generation, as monitored by neutrophil chemiluminescence responses, whereas bacterial viability was unaffected (data not shown). Whereas treatment of neutrophils with DPI led to strongly reduced WT killing, it had no significant influence on the inactivation of the Dlt⁻ mutant (figure 3). Taken together, these results demonstrate that S. aureus WT bacteria are predominantly killed by oxygen-dependent mechanisms in human neutrophils, whereas the enhanced killing of the Dlt⁻ mutant results from a greater susceptibility to oxygen-independent mechanisms

Figure 3

Killing of Staphylococcus aureus by neutrophils, with or without a functional respiratory burst. Nos. (in cfu) of viable opsonized wild-type (WT) and Dlt⁻ bacteria after a 30-, 60-, or 120-min incubation with human neutrophils in the presence (black bars) or absence (gray bars) of the NADPH oxidase inhibitor diphenyleniodonium (20 μM) are expressed as percentages of the initial counts. Data are mean+SE of ⩾3 independent experiments

Defensin susceptibility is not a factor in bacterial resistance to killing by human monocytesNeutrophils and monocytes use the same respiratory burst enzymes, but human monocytes do not produce defensins [30]. Thus, it was intriguing to compare the killing of opsonized WT and Dlt⁻ bacteria by monocytes. As before, no differences in the uptake kinetics of the 2 bacterial strains were observed (data not shown). In contrast to neutrophils, the monocytes inactivated the mutant and WT bacteria with similar efficiency (figure 2A), which supports the notion that increased killing of Dlt⁻ bacteria by neutrophils results from susceptibility to defensins

Virulence characteristics of WT and Dlt⁻ bacteria in a mouse model of sepsis and arthritisThe virulences of the Dlt⁻ mutant and WT bacteria were assessed in a mouse model of sepsis and septic arthritis [25] by using S. aureus Newman, which is more virulent than laboratory strain Sa113. This model uses mice injected intravenously (iv) with staphylococci and permits the analysis of parameters such as mortality due to sepsis, frequency and severity of septic arthritis, general health (weight loss measurements), and bacteria loads in the blood and kidneys [31]. The Newman WT strain was highly virulent in this model; 7 days after the iv injection of 1.8×10⁷ cfu of the WT strain, the mortality rate was 33%, and the level of arthritis in surviving mice was 94%. In contrast, there were no deaths among mice injected iv with the Dlt⁻ strain, and the frequency of arthritis 7 days after iv infection was 21% (table 2). With respect to general condition, mice injected with Dlt⁻ lost significantly less weight after 3 and 7 days than the WT-infected mice. Although the number of bacteria per milliliter of blood measured 24 h after infection was similar for WT and Dlt⁻ organisms (mean±SD, 1.3±1.3×10³ and 4.5±11.4×10³, respectively), the number of bacteria isolated from the kidneys at day 7 after infection was significantly higher for WT-infected than for Dlt⁻-infected mice (P<.0001; table 2)

Table 2

Comparative virulence of Staphylococcus aureus Newman wild-type (WT) and Dlt− strains following intravenous inoculation of NMRI mice

Discussion

Microbial pathogens have variable susceptibility to the oxygen-dependent and -independent killing mechanisms of phagocytes. Even within a specific bacterial species, clinical isolates may be killed with different efficiencies, as demonstrated elsewhere for S. aureus [32]. The ability to resist neutrophil killing contributes to the virulence of S. aureus as recently demonstrated elsewhere [13]. Inherited defects in phagocyte function lead to increased susceptibilities to certain infections. For instance, patients with CGD who have a mutation in NADPH oxidase, the first enzyme of the respiratory burst cascade, are much more susceptible to S. aureus and Burkholderia cepacia infection than to infection with Pseudomonas aeruginosa or Candida albicans [33, 34]. The mechanisms conferring resistance or susceptibility to a particular antimicrobial phagocyte component are complex and have largely remained elusive; elucidating their molecular pathways may be of great importance in finding new ways to prevent and combat microbial infections

The α-defensins are key components of the oxygen-independent antimicrobial system in human neutrophils. The existence of redundant defensin genes and differences in defensin expression in rodents and primates are major obstacles to analyzing the role of defensins by use of knockout mice. Therefore, we approached this question by generating defensin-susceptible bacterial mutants in a defensin-resistant bacterial WT strain. We recently identified 2 independent loci (dltABCD and mprF) whose inactivation resulted in susceptibility to defensins and related cationic pore-forming peptides [16, 17]. Both systems are based on particular modifications of the net charge of the cell envelope. We have demonstrated here that the elimination of d-alanine ester substitution of teichoic acids by disruption of the dltABCD operon renders S. aureus Dlt⁻ mutants susceptible to killing by defensin peptides and prone to faster and more efficient in vitro inactivation by neutrophils containing large amounts of α-defensins. Accordingly, WT and Dlt⁻ bacteria were equally susceptible to killing by monocytes that do not produce defensins. Efficient neutrophil killing of the WT strain was dependent on a functional respiratory burst cascade, which confirms that oxygen-independent pathways do not play a major role in the inactivation of WT S. aureus in normal neutrophils

The inhibition of the respiratory burst with DPI did not affect the efficiency of the neutrophils’ killing of the Dlt⁻ mutant; however, these bacteria were more susceptible to MPO-mediated killing in vitro. We hypothesize that the cationic enzyme MPO binds more efficiently to the highly negatively charged surface of the Dlt⁻ bacteria, as recently demonstrated with other cationic proteins [16], and this may explain the increased in vitro susceptibility to MPO-mediated killing. We can only speculate why this difference did not translate into decreased killing of the mutant by DPI-treated neutrophils. It may be that oxygen-independent killing of Dlt⁻ bacteria in neutrophils is so efficient that inhibition of the oxidative burst impacts minimally on overall bactericidal activity. Human neutrophils contain further oxygen-independent antimicrobial compounds, such as cathelicidins and group IIA phospholipase A₂ [35]. Since these molecules share cationic properties with defensins, they may contribute to the increased neutrophil killing of Dlt⁻

There was a clear correlation between the mutation in dlt and reduced virulence in mice. Both the mortality and the arthritis frequencies were dramatically reduced in NMRI mice infected with Dlt⁻, compared with those infected with WT. Thus, it seems that the Dlt⁻ mutant is not able to home to or colonize joints in sufficient numbers to induce arthritis. Of interest, there were no significant differences in the numbers of bacteria in the blood 24 h after infection in the 2 groups, which suggests that the clearance of the Dlt⁻ strain by blood neutrophils and monocytes early in infection is not greater than that of the WT strain. However, there were significantly fewer Dlt⁻ than WT bacteria in the kidneys of mice 7 days after inoculation. Taken together, these results suggest that defensin sensitivity does not affect S. aureus viability in the peripheral blood early in infection but renders the bacteria susceptible to killing in tissues, such as the kidney

Cationic peptides in tissues presumably cooperate with the host immune system to dramatically reduce the toxicity and arthritogenicity of Dlt⁻ bacteria. In line with this idea, cationic peptide production has been observed in various mouse tissues [36–38]. Although it is a generally held belief that mouse neutrophils do not contain defensins [39], antimicrobial cathelin-related peptides have been discovered within mouse neutrophils [36]. Moreover, it is possible that, in addition to increased susceptibility to antimicrobial peptides, altered capacities of Dlt⁻ to interact with host cells or matrix proteins contribute to the attenuated virulence in mice

The ability to abrogate the inherent resistance of S. aureus to cationic peptides by changing alanine or lysine residues, and thus the net charge, on the bacterial surface offers exciting possibilities for the development of therapies against previously drug-resistant bacteria. It should be noted that modifications of the lipopolysaccharide mediate protection of gram-negative bacteria against cationic antimicrobial peptides [40, 41]. The modulation of cell surface properties can thus be considered to be a general bacterial strategy to escape from host defense peptides

Acknowledgments

We thank Hubert Kalbacher for providing α-defensins; Jiri Doskar for help with phages; and Erik Heezius, Maartje Ruyken, Miriam Poppelier, Lena Svensson, and Shahin Hajizadeh for technical assistance

Informed consent was obtained for blood draws. Animal experimentation guidelines were followed in animal studies

Financial support: European Molecular Biology Organization (fellowship ASTF 9521 to A.P.); German Research Council (priority program grant 1047 and research group 449/T2 to A.P.); Swedish Inflammation Network and Medical Research Council; Swedish Association against Rheumatism; the Nana Svartz and King Gustaf V Foundations (to L.V.C.); German Ministry of Education and Research (Fö, 01KS9602); Interdisciplinary Center of Clinical Research Tübingen (to F.G. and B.N.)

L.V.C. and S.A.K. contributed equally to this work