Importance of gls24 in Virulence and Stress Response of Enterococcus faecalis and Use of the Gls24 Protein as a Possible Immunotherapy Targe
Division of Infectious Diseases, Department of Medicine, Center for the Study of Emerging and Reemerging Pathogens
Department of Microbiology and Molecular Genetics, University of Texas Houston Medical School, Houston
Gls24 was previously identified as a general stress protein of Enterococcus faecalis. In the present study, we found that a gls24 disruption mutant (TX10100) of E. faecalis strain OG1RF showed a considerably increased 50% lethal dose in a mouse peritonitis model, and, at high inocula, TX10100 was either not lethal or was much less so than wild-type OG1RF (P < .001). TX10100 was also more sensitive to bile salts (mean ± SD survival rate relative to wild-type OG1RF, 9.7% ± 2.0%) at late stationary phase, as previously found by Giard et al. with another strain. Inactivation of glsB, downstream of and cotranscribed with gls24, had no effect on E. faecalis virulence but resulted in reduced bile-salts resistance (to a mean ± SD survival rate of 28.0% ± 5.1%) relative to wild-type OG1RF. Results of complementation of TX10100 with different combinations of gls24-glsB and 2 promotersa remote promoter (P1) and an adjacent promoter (P2)suggested that both genes and both promoters, especially P1, are important for bile-salts resistance. Anti-Gls24 immune rabbit serum, which showed some Gls24 on the cell surface, protected mice against a lethal challenge of OG1RF in the peritonitis model (e.g., survival of 12/18 mice vs. 1/18 mice with preimmune rabbit serum; P = .008). In conclusion, the E. faecalis gls24 gene is important for virulence as well as stress response, and anti-Gls24 immune rabbit serum shows protection against E. faecalis infection in a mouse peritonitis model.
Enterococci have been well established as antibiotic-resistant opportunistic organisms that are commonly recovered from hospitalized patients, with Enterococcus faecalis accounting for the majority of the clinically important enterococcal isolates. These organisms cause 5%15% of endocarditis, as well as other diseases, such as bacteremia and urinary tract infections [1]. Several factors that are important for E. faecalis virulence have been identified, including aggregation substance, cytolysin/hemolysin, EfaA, gelatinase (GelE), serine protease (SprE), the fsr system controlling expression of GelE and SprE, and the 2-component system EtaRS, as well as 2 polysaccharide biosynthesis gene (epa and cps) clusters [24]. Some other factors that have been suggested as potential virulence factors include Ace (an adhesin for collagen) and Esp (biofilm related protein) [2].
Bacterial stress response has been linked to virulence, and the link appears to involve bacterial adaptation to different stress environments in the host. For example, glutathione peroxidase and MtsABC of Streptococcus pyogenes have been shown to be important for resistance to oxidative stress and, at the same time, were found to be important for virulence of S. pyogenes in several animal models [5, 6]. The 2-component system LisRK of Listeria monocytogenes has also been shown to be involved in both stress response and virulence [3, 7], as has EtaRS, a 2-component system we previously identified that is important for virulence of E. faecalis in a mouse peritonitis model [3]; EtaR shows 77% similarity to LisR and 70% similarity to CsrR, a negative regulator of several virulence factors of S. pyogenes, such as ska, sagA, speMF, and hasA [8]. As part of our subsequent studies of EtaR, we identified the gls24 gene, shown previously by Giard et al. to encode a general stress protein involved in bile-salts resistance, the transcription of which was induced under glucose starvation and other stress conditions, such as the presence of bile salts and CdCl2 [9, 10]. In the present study, we found that a gls24 mutant was attenuated in a mouse peritonitis model, which led us to further investigate this system. In addition to analyzing the OG1RF gls24 operon and the effect of its different components on bile-salts resistance, we also found that gls24 is important for E. faecalis virulence and that anti-Gls24 immune rabbit serum protects mice against E. faecalis infection.
MATERIALS AND METHODS
Bacterial strains and media.
E. faecalis strain OG1RF has been described elsewhere [11]. E. faecalis and Escherichia coli were grown in brain-heart infusion (BHI) and Luria-Bertani medium (Difco Laboratories), respectively, with appropriate antibiotics at 37°C, unless otherwise described.
DNA and RNA techniques.
Standard polymerase chain reaction (PCR) amplification, DNA cloning, and DNA sequencing were performed as described elsewhere [12]. Long-range PCR was performed with OG1RF genomic DNA and the primers con1d and con1b (table 1), by use of the XL PCR Kit (Applied Biosystems) and the protocol of the supplier. Total RNA was isolated from E. faecalis by use of the RNeasy Mini Kit (Qiagen) and the protocol of the supplier, with slight modifications. E. faecalis was grown in BHI broth (BHIB) for 24 h at 37°C, and 10 mg/mL lysozyme solution was used for the lysis step. After this, the total RNA (3040 g) was treated with 5 U of RQ1 DNase (Promega) for 20 min at 37°C, extracted with chloroform-phenol-isoamyl alcohol (25 : 24 : 1), and precipitated with ethanol; the RNA was treated 3 times as above before use in reverse-transcriptase (RT) PCR. RT-PCR was performed using the SuperScript One-Step RT-PCR Kit (Invitrogen Life Technologies); the different primers used in the RT-PCR are listed in table 1. Twenty nanograms of E. faecalis RNA was used in a 25-L reaction, and PCR was performed for 30 cycles. Primers designed for 16S-rRNA were used in RT-PCRs as a positive control, and reactions without RT were used as controls to detect DNA contamination.
Construction of disruption mutants and complementation.
The oligonucleotides used for construction of the gls24 and glsB mutants and complementation derivatives are listed in table 1. Gene disruption or inactivation (by a polar effect) in OG1RF was achieved by the method described in a previous article [3], by use of an intragenic fragment (in the case of gls24 disruption) or a fragment containing the whole length of gls24 but not glsB (in the case of glsB inactivation) in the suicide vector, which results in a single crossover; the gls24 and glsB mutants were named TX10100 and TX10200, respectively. Complementation constructs (figure 4) were made with the shuttle vector pAT18 [13]. Briefly, DNA fragments containing P2+gls24 and P2+gls24+glsB were amplified with primers gls24up/gls24down and gls24up/EglsBdown (table 1), respectively, from OG1RF genomic DNA and were cloned into the TA cloning vector (Invitrogen). The fragments were then released from the vector by EcoRI and inserted into the EcoRI site of pAT18. The constructs were sequenced with M13 forward and reverse primers to verify the sequence and to determine the orientation of the fragments. To make constructs with both P1 and P2, the P1 fragment (200 bp) was amplified from OG1RF genomic DNA with primers orf1p-up and orf1p-down (table 1) and inserted into XbaI/BamHI sites of pAT18 containing P2+gls24 or P2+gls24+glsB. To make the complementation constructs with P1 but not P2, the construct containing P1+P2+gls24 was digested with EcoRI and relegated to produce pAT18 containing P1, and the amplified fragment with gls24+glsB or glsB (primers gls24rbs/Eglsdown or glsBrbs/Eglsdown; table 1) was inserted into the BamHI/KpnI sites to produce pAT18 containing P1+gls24+glsB or P1+glsB. These constructs were individually electroporated into TX10100 to produce complementation derivatives, which were confirmed by PCR and Southern-blot analysis. Growth curves of OG1RF, mutants, and complementation strains and stability of single-crossover and complementation derivatives were determined as described elsewhere [14].
Bile-salts resistance and other assays.
After growing in BHIB for 24 h at 37°C, 1 mL of E. faecalis culture was aliquoted, harvested, and resuspended in 1 mL of fresh BHIB, from which 20 L was inoculated into 1 mL of fresh BHIB containing 0.3% bile salts. After incubation for 30 min in a 37°C water bath, the bacterial cells were counted and the percentage survival was calculated by comparing colony-forming units at 30 min to colony-forming units at time (t) 0, by use of the equation
Other potentially stressful conditions examined included growth in BHIB with 40% horse serum, at high temperature (55°C), low pH (pH 3.4), and in an anaerobic chamber. Biofilm and gelatinase production assays were performed as described elsewhere [12, 15].
Production of recombinant proteins and antiserum.
A 561-bp fragment containing the full-length gls24 gene was amplified from OG1RF genomic DNA with primers Egls24up and Egls24down (table 1) and was cloned in-frame into BamHI/KpnIdigested pQE30 (QIAexpressionist; QIAGEN). The His-tagged Gls24 protein was expressed in E. coli and purified under native conditions, as described by the manufacturer. The purified His-tagged Gls24 protein was sent to Bethyl Laboratories (Montgomery, TX), where immune rabbit serum (IRS) was generated by intravenous immunization with His-tagged Gls24 protein in PBS.
Protein analysis.
To localize the Gls24 protein, surface proteins and cell lysates were prepared by Zwittergent (Calbiochem) and sodium dodecyl sulfate treatment, as described elsewhere [16], and supernatant proteins were prepared by concentration to 1/30 volume with a centrifugal filter device (10-kDa cutoff; Millipore Corporation). SDS-polyacrylamide gel electrophoresis, Western blots, and whole-cell ELISA were performed as described elsewhere [16, 17].
Mouse peritonitis and antibody protection model.
The mouse peritonitis model, calculation of LD50 values and Kaplan-Meier survival curves, and log-rank analysis were performed as described elsewhere [14, 18], except that 12.5% sterile rat fecal extract (SRFE) was used. The inocula were adjusted using an optical density at 600 nm (OD 600), and the number of colony-forming units in each inoculum was determined.
Prior to their use, anti-Gls24 IRS and preimmune rabbit serum (PRS) were absorbed with 1 × 109 cells of TX10100 for 1 h at 4°C (IRS-abs and PRS-abs), to remove preexisting antibody. To determine the effect of IRS-abs on colony-forming units of OG1RF in spleens, 8 mice/group were injected intraperitoneally (ip) with 200 L of either anti-Gls24 IRS-abs (titer 1 : 600) or PRS-abs 1 h before ip injection of OG1RF (1.7 × 107 cfu in 1 mL of saline containing 12.5% SRFE). Mice were killed at 72 h, and spleens were removed, weighed, homogenized, and serially diluted in saline for colony count. The experiment was performed twice, and comparison between the 2 groups (IRS-abs vs. PRS-abs) was performed by use of nonparametric Mann-Whitney U tests. For determination of antibody effect on mortality of OG1RF, 9 mice/group were injected ip with 300 L of either IRS-abs or PRS-abs 1 h before and 4 h after OG1RF challenge (1 × 109 cfu in 1 mL of saline containing 12.5% SRFE). The experiment was performed twice; in each experiment, an additional 34 mice were challenged with OG1RF without serum (infection control), and 23 mice received IRS-abs or PRS-abs without bacterial challenge (serum control). Comparison of the number of surviving mice or the survival curves between the 2 groups (IRS-abs vs. PRS-abs) was performed by use of Fisher's exact test or log-rank analysis.
RESULTS
To test whether disruption of gls24 had an effect on E. faecalis virulence, TX10100 was examined in a mouse peritonitis model, along with wild-type OG1RF. Initial experiments using inocula of 1.6 × 109 cfu and 6 mice/group showed that the gls24 disruption mutant was not lethal (100% survival at 96 h with inocula of 1.6 × 109 cfu), whereas wild-type OG1RF showed 83.3%, 66.7%, and 16.7% lethality at 96 h with inocula of 1.2 × 109, 6 × 108, and 3 × 108 cfu, respectively (P < .001 by log rank for 1.6 × 109 cfu of TX10100 vs. 1.2 × 109 cfu of wild-type OG1RF). A larger-scale experiment was then performed for TX10100 and wild-type OG1RF with inocula up to 1.6 × 1010 cfu (for TX10100) and 12 mice/group (figure 1B). The results confirmed that TX10100 was highly attenuated in the mouse peritonitis model, compared with wild-type OG1RF (91.7% survival after an inoculum of 1.6 × 109 cfu [TX10100] vs. 8.3% survival after an inoculum of 1.3 × 109 cfu [wild-type OG1RF] at 120 h; P < .001). The LD50 values for TX10100 and wild-type OG1RF were found to be 8.2 × 109 and 3.8 × 108 cfu, respectively. The colony-forming units recovered from spleens of dead mice receiving TX10100 were comparable on BHI agar and on BHI-kanamycin agar plates, indicating that the disruption was stable in vivo.
RT-PCR with RNA extracted from wild-type OG1RF and TX10100 showed no obvious difference in orf4 levels (figure 3A), suggesting that gls24 is not involved in autoregulation of its upstream genes under our testing conditions. Transcription of glsB was not detectable in TX10100 (figure 3A), consistent with gls24 and glsB being cotranscribed. To investigate whether glsB contributes to E. faecalis virulence, a glsB inactivation mutant of OG1RF was constructed and confirmed by RT-PCR, which showed transcription of gls24 but not glsB (figure 3B). The glsB mutant showed results similar to those for wild-type OG1RF with multiple inocula in the mouse peritonitis model (data not shown), indicating that glsB does not contribute to E. faecalis virulence in this model.
Characterization of the Gls24 protein in E. faecalis.
The gene products of gls24 and glsB have predicted molecular masses of 20.2 and 7.85 kDa and PIs of 5.21 and 11.26, respectively. Gls24 was predicted by PSORT to be a cytoplasmic protein, and GlsB was predicted to be a membrane protein. Recombinant Gls24 protein was successfully expressed in E. coli and used to generate antiserum, whereas attempts to express recombinant GlsB in E. coli were not successful. When anti-Gls24 IRS was used, Gls24 (which ran at 28 kDa on an SDS-PAGE gel) was found in both log- and stationary-phase OG1RF cells (figure 5), whereas PRS did not detect the Gls24 protein band (data not shown). Western blots with different fractions of OG1RF cells detected the Gls24 protein in the Zwittergent preparations (figure 5) and in cell lysates after Zwittergent preparation (data not shown) but not in concentrated culture supernatants (figure 5), suggesting that some Gls24 is present on the cell surface. Although the Gls24 protein was absent from the gls24 disruption mutant TX10100, transcomplementation with P2+gls24 or P2+gls24+glsB restored Gls24 expression, and complementation strains with P1+P2+gls24+glsB overexpressed Gls24, which showed a smear on Western blot, likely as a result of degradation of the protein (figure 5). Reproducible results were obtained with multiple extracts, and whole-cell ELISA also detected Gls24 on the surface (data not shown). Western blots with whole-cell lysates of wild-type OG1RF and TX10100 showed that, after absorption, anti-Gls24 IRS-abs specifically reacted with Gls24, whereas the anti-Gls24 antibody was not present in the PRS-abs (figure 5C).
DISCUSSION
Identification of factors important for enterococcal virulence may help in the design of treatment (such as active or passive immunotherapy) against these organisms, which are a leading cause of nosocomial infections. In the present study, we identified gls24, which was previously shown by Giard et al. to be a general stress gene [9]. Our interest in gls24 was enhanced when we found that the gls24 disruption mutant (TX10100) was highly attenuated in vivo. For example, even after inoculation with 1 × 109 cfu, TX10100 was not lethal in the mouse peritonitis model, whereas the wild-type OG1RF caused 90% death at the same inoculum. The LD50 of TX10100 was 21.6-fold higher than that of wild-type OG1RF, making it, to our knowledge, the most attenuated of all published E. faecalis OG1RF mutants. Our previously published attenuated E. faecalis mutants generally showed prolonged survival, compared with wild-type OG1RF, with similar or slightly higher LD50 values, in the mouse peritonitis model; for example, the fsr and sprE mutants had LD50 values similar to wild-type OG1RF [12], and the LD50 values of etaRS, orfde4, orfde6, gelE, and efaA mutants were 4.6-, 2.5-, 2.1-, 1.6-, 1.4-fold higher, respectively, than that of wild-type OG1RF [3, 14, 18, 19]. When the hemolysin-encoding plasmid pAM714 was introduced into OG1RF, an 35-fold lower LD50 was observed [18, 20]. However, hemolysin is not present in wild-type OG1RF; thus, an isogenic mutant was not available to be compared with the gls24 mutant of OG1RF. We have to point out that these results were obtained at different times with different batches of SRFE; thus, the results may not be directly comparable. Nevertheless, we recently repeated the experiment with an sprE mutant [12] with the same batch of SRFE used for the gls24 mutant and obtained results similar to those previously published. Although it was observed previously by Giard et al. that the gls24 mutant had a longer generation time (58 vs. 45 min) in a semisynthetic medium (Bacto Folic AOAC medium) and shorter chain length, compared with its parental strain JH2-2 (a nongelatinase-producing strain) [9], we did not detect a growth rate or morphological difference between the gls24 mutant and its parental strain OG1RF (gelatinase and serine protease producer) when they were grown in BHIB, suggesting that, for wild-type OG1RF versus TX10100, the different phenotypes are not explained by a growth effectat least, not one observed in BHIB.
To further study the role of gls24 in E. faecalis virulence, antiserum against recombinant Gls24 was generated, and antibody protection assays showed that mice injected with anti-Gls24 antiserum had lower bacterial counts in their spleens (23-fold lower than in mice given PRS; P = .02) after sublethal OG1RF challenge, as well as significantly decreased mortality after lethal OG1RF challenge. These results suggest that the Gls24 protein is a potential candidate for immunotherapy for E. faecalis infection. As far as we could determine from a literature search, the only other antiserum that has been shown to be protective in E. faecalis infection was against a polysaccharide [21]. In that study, the investigators showed that injection of their antiserum into mice reduced E. faecalis counts in mouse organs, compared with counts after injection of preimmune serum or sterile saline; however, the effect of their antiserum on mouse mortality was not studied [21]. Antibody to aggregation substance has also been examined for a protective effect (in a rabbit endocarditis model), but the results were negative [22].
The gls24 locus previously identified by Giard et al. contains 4 ORFs (orf14) upstream of gls24 and 1 ORF (glsB) downstream [9]. In that study, Northern blots from strain JH2-2 showed transcripts with the length of orf1glsB and gls24glsB, and 2 transcriptional start sites were identified by primer extension, 1 upstream of orf1 and 1 upstream of gls24. Our RT-PCR results with strain OG1RF also indicate that there is cotranscription of orf1glsB, and we have demonstrated that the transcript includes orf1 and glsB but not the genes upstream or downstream. Since our RNA analysis showed that disruption of gls24 also disrupted the transcription of glsB, we constructed a glsB mutant (TX10200) and found that inactivation of glsB alone caused mortality in mice that was similar to that caused by wild-type OG1RF, suggesting that gls24, but not glsB, contributes to E. faecalis virulence.
The increased sensitivity of TX10100 to bile salts at the late stationary phase is consistent with the results of the study by Giard et al. showing that gls24 is important for bile-salts resistance after starvation [9]. In their article, Giard et al. indicated that complementation with gls24 and glsB only partially restored bile-salts resistance of the gls24 mutant, but it is not clear exactly what region was used for complementation. Since there are 2 possible promoters (P1 and P2) that may drive transcription of gls24 and glsB (on the basis of our results and those of Giard et al.), in the present study we constructed different complementation derivatives of TX10100 and examined their resistance to bile salts. The results showed that the 2 promoters P1 and P2 (but especially the remote promoter, P1) and both gls24 and glsB are important for bile-salts resistance under our testing conditions. The function of the gls24 gene and the operon in vivo remains unclear, but it may be involved in adaptation to the in vivo environment(s), suggested by their involvement in stress response in vitro, and is an area of interest for future studies.
In summary, gls24 was identified in E. faecalis strain OG1RF and was shown to be important for virulence as well as stress response. Antiserum against the Gls24 protein was able to protect against E. faecalis infection in a mouse peritonitis model, illustrating the potential usefulness of this protein as an immunotarget.
Acknowledgments
We thank Suresh R. Pai and Kavindra V. Singh for performing some of the animal experiments, Ernesto Baca for his assistance with consensus search of the Enterococcus faecalis V583 genome database, and Sandor E. Karpathy for providing some of the primers used in reverse-transcriptase polymerase chain reactions.
References
1. Murray BE. The life and times of the Enterococcus. Clin Microbiol Rev 1990; 3:4665. First citation in article
2. Gilmore MS, Coburn PS, Nallapareddy SR, Murray BE. Enterococcal virulence. In: Gilmore MS, Clewell D, Courvalin P, Dunny GM, Murray BE, Rice L, eds. The enterococci: pathogenesis, molecular biology and antibiotic resistance. Washington, DC: ASM Press, 2002. First citation in article
3. Teng F, Wang L, Singh KV, Murray BE, Weinstock GM. Involvement of PhoP-PhoS homologs in Enterococcus faecalis virulence. Infect Immun 2002; 70:19916. First citation in article
4. Teng F, Jacques-Palaz KD, Weinstock GM, Murray BE. Evidence that the enterococcal polysaccharide antigen gene (epa) cluster is widespread in Enterococcus faecalis and influences resistance to phagocytic killing of E. faecalis. Infect Immun 2002; 70:20105. First citation in article
5. Brenot A, King KY, Janowiak B, Griffith O, Caparon MG. Contribution of glutathione peroxidase to the virulence of Streptococcus pyogenes. Infect Immun 2004; 72:40813. First citation in article
6. Janulczyk R, Ricci S, Bjorck L. MtsABC is important for manganese and iron transport, oxidative stress resistance, and virulence of Streptococcus pyogenes. Infect Immun 2003; 71:265664. First citation in article
7. Cotter PD, Emerson N, Gahan CG, Hill C. Identification and disruption of lisRK, a genetic locus encoding a two-component signal transduction system involved in stress tolerance and virulence in Listeria monocytogenes. J Bacteriol 1999; 181:68403. First citation in article
8. Miller AA, Engleberg NC, DiRita VJ. Repression of virulence genes by phosphorylation-dependent oligomerization of CsrR at target promoters in S. pyogenes. Mol Microbiol 2001; 40:97690. First citation in article
9. Giard JC, Rince A, Capiaux H, Auffray Y, Hartke A. Inactivation of the stress- and starvation-inducible gls24 operon has a pleiotrophic effect on cell morphology, stress sensitivity, and gene expression in Enterococcus faecalis. J Bacteriol 2000; 182:451220. First citation in article
10. Giard JC, Hartke A, Flahaut S, Boutibonnes P, Auffray Y. Glucose starvation response in Enterococcus faecalis JH22: survival and protein analysis. Res Microbiol 1997; 148:2735. First citation in article
11. Murray BE, Singh KV, Ross RP, Heath JD, Dunny GM, Weinstock GM. Generation of restriction map of Enterococcus faecalis OG1 and investigation of growth requirements and regions encoding biosynthetic function. J Bacteriol 1993; 175:521623. First citation in article
12. Qin X, Singh KV, Weinstock GM, Murray BE. Effects of Enterococcus faecalis fsr genes on production of gelatinase and a serine protease and virulence. Infect Immun 2000; 68:257986. First citation in article
13. Trieu-Cuot P, Carlier C, Poyart-Salmeron C, Courvalin P. Shuttle vectors containing a multiple cloning site and a lacZ alpha gene for conjugal transfer of DNA from Escherichia coli to gram-positive bacteria. Gene 1991; 102:99104. First citation in article
14. Xu Y, Singh KV, Qin X, Murray BE, Weinstock GM. Analysis of a gene cluster of Enterococcus faecalis involved in polysaccharide biosynthesis. Infect Immun 2000; 68:81523. First citation in article
15. Mohamed JA, Huang W, Nallapareddy SR, Teng F, Murray BE. Influence of clinical origin, especially endocarditis, and of various genes on biofilm formation by Enterococcus faecalis. Infect Immun 2004; 72:365863. First citation in article
16. Xu Y, Jiang L, Murray BE, Weinstock GM. Enterococcus faecalis antigens in human infections. Infect Immun 1997; 65:420715. First citation in article
17. Nallapareddy SR, Weinstock GM, Murray BE. Clinical isolates of Enterococcus faecium exhibit strain-specific collagen binding mediated by Acm, a new member of the MSCRAMM family. Mol Microbiol 2003; 47:173347. First citation in article
18. Singh KV, Qin X, Weinstock GM, Murray BE. Generation and testing of mutants of Enterococcus faecalis in a mouse peritonitis model. J Infect Dis 1998; 178:141620. First citation in article
19. Singh KV, Coque TM, Weinstock GM, Murray BE. In vivo testing of an Enterococcus faecalis efaA mutant and use of efaA homologs for species identification. FEMS Immunol Med Microbiol 1998; 21:32331. First citation in article
20. Ike Y, Hashimoto H, Clewell DB. Hemolysin of Streptococcus faecalis subspecies zymogenes contributes to virulence in mice. Infect Immun 1984; 45:52830. First citation in article
21. Huebner J, Quaas A, Krueger WA, Goldmann DA, Pier GB. Prophylactic and therapeutic efficacy of antibodies to a capsular polysaccharide shared among vancomycin-sensitive and -resistant enterococci. Infect Immun 2000; 68:46316. First citation in article
22. McCormick JK, Hirt H, Waters CM, Tripp TJ, Dunny GM, Schlievert PM. Antibodies to a surface-exposed, N-terminal domain of aggregation substance are not protective in the rabbit model of Enterococcus faecalis infective endocarditis. Infect Immun 2001; 69:330514. First citation in article, 百拇医药(Fang Teng, Esteban C. Nan)
Department of Microbiology and Molecular Genetics, University of Texas Houston Medical School, Houston
Gls24 was previously identified as a general stress protein of Enterococcus faecalis. In the present study, we found that a gls24 disruption mutant (TX10100) of E. faecalis strain OG1RF showed a considerably increased 50% lethal dose in a mouse peritonitis model, and, at high inocula, TX10100 was either not lethal or was much less so than wild-type OG1RF (P < .001). TX10100 was also more sensitive to bile salts (mean ± SD survival rate relative to wild-type OG1RF, 9.7% ± 2.0%) at late stationary phase, as previously found by Giard et al. with another strain. Inactivation of glsB, downstream of and cotranscribed with gls24, had no effect on E. faecalis virulence but resulted in reduced bile-salts resistance (to a mean ± SD survival rate of 28.0% ± 5.1%) relative to wild-type OG1RF. Results of complementation of TX10100 with different combinations of gls24-glsB and 2 promotersa remote promoter (P1) and an adjacent promoter (P2)suggested that both genes and both promoters, especially P1, are important for bile-salts resistance. Anti-Gls24 immune rabbit serum, which showed some Gls24 on the cell surface, protected mice against a lethal challenge of OG1RF in the peritonitis model (e.g., survival of 12/18 mice vs. 1/18 mice with preimmune rabbit serum; P = .008). In conclusion, the E. faecalis gls24 gene is important for virulence as well as stress response, and anti-Gls24 immune rabbit serum shows protection against E. faecalis infection in a mouse peritonitis model.
Enterococci have been well established as antibiotic-resistant opportunistic organisms that are commonly recovered from hospitalized patients, with Enterococcus faecalis accounting for the majority of the clinically important enterococcal isolates. These organisms cause 5%15% of endocarditis, as well as other diseases, such as bacteremia and urinary tract infections [1]. Several factors that are important for E. faecalis virulence have been identified, including aggregation substance, cytolysin/hemolysin, EfaA, gelatinase (GelE), serine protease (SprE), the fsr system controlling expression of GelE and SprE, and the 2-component system EtaRS, as well as 2 polysaccharide biosynthesis gene (epa and cps) clusters [24]. Some other factors that have been suggested as potential virulence factors include Ace (an adhesin for collagen) and Esp (biofilm related protein) [2].
Bacterial stress response has been linked to virulence, and the link appears to involve bacterial adaptation to different stress environments in the host. For example, glutathione peroxidase and MtsABC of Streptococcus pyogenes have been shown to be important for resistance to oxidative stress and, at the same time, were found to be important for virulence of S. pyogenes in several animal models [5, 6]. The 2-component system LisRK of Listeria monocytogenes has also been shown to be involved in both stress response and virulence [3, 7], as has EtaRS, a 2-component system we previously identified that is important for virulence of E. faecalis in a mouse peritonitis model [3]; EtaR shows 77% similarity to LisR and 70% similarity to CsrR, a negative regulator of several virulence factors of S. pyogenes, such as ska, sagA, speMF, and hasA [8]. As part of our subsequent studies of EtaR, we identified the gls24 gene, shown previously by Giard et al. to encode a general stress protein involved in bile-salts resistance, the transcription of which was induced under glucose starvation and other stress conditions, such as the presence of bile salts and CdCl2 [9, 10]. In the present study, we found that a gls24 mutant was attenuated in a mouse peritonitis model, which led us to further investigate this system. In addition to analyzing the OG1RF gls24 operon and the effect of its different components on bile-salts resistance, we also found that gls24 is important for E. faecalis virulence and that anti-Gls24 immune rabbit serum protects mice against E. faecalis infection.
MATERIALS AND METHODS
Bacterial strains and media.
E. faecalis strain OG1RF has been described elsewhere [11]. E. faecalis and Escherichia coli were grown in brain-heart infusion (BHI) and Luria-Bertani medium (Difco Laboratories), respectively, with appropriate antibiotics at 37°C, unless otherwise described.
DNA and RNA techniques.
Standard polymerase chain reaction (PCR) amplification, DNA cloning, and DNA sequencing were performed as described elsewhere [12]. Long-range PCR was performed with OG1RF genomic DNA and the primers con1d and con1b (table 1), by use of the XL PCR Kit (Applied Biosystems) and the protocol of the supplier. Total RNA was isolated from E. faecalis by use of the RNeasy Mini Kit (Qiagen) and the protocol of the supplier, with slight modifications. E. faecalis was grown in BHI broth (BHIB) for 24 h at 37°C, and 10 mg/mL lysozyme solution was used for the lysis step. After this, the total RNA (3040 g) was treated with 5 U of RQ1 DNase (Promega) for 20 min at 37°C, extracted with chloroform-phenol-isoamyl alcohol (25 : 24 : 1), and precipitated with ethanol; the RNA was treated 3 times as above before use in reverse-transcriptase (RT) PCR. RT-PCR was performed using the SuperScript One-Step RT-PCR Kit (Invitrogen Life Technologies); the different primers used in the RT-PCR are listed in table 1. Twenty nanograms of E. faecalis RNA was used in a 25-L reaction, and PCR was performed for 30 cycles. Primers designed for 16S-rRNA were used in RT-PCRs as a positive control, and reactions without RT were used as controls to detect DNA contamination.
Construction of disruption mutants and complementation.
The oligonucleotides used for construction of the gls24 and glsB mutants and complementation derivatives are listed in table 1. Gene disruption or inactivation (by a polar effect) in OG1RF was achieved by the method described in a previous article [3], by use of an intragenic fragment (in the case of gls24 disruption) or a fragment containing the whole length of gls24 but not glsB (in the case of glsB inactivation) in the suicide vector, which results in a single crossover; the gls24 and glsB mutants were named TX10100 and TX10200, respectively. Complementation constructs (figure 4) were made with the shuttle vector pAT18 [13]. Briefly, DNA fragments containing P2+gls24 and P2+gls24+glsB were amplified with primers gls24up/gls24down and gls24up/EglsBdown (table 1), respectively, from OG1RF genomic DNA and were cloned into the TA cloning vector (Invitrogen). The fragments were then released from the vector by EcoRI and inserted into the EcoRI site of pAT18. The constructs were sequenced with M13 forward and reverse primers to verify the sequence and to determine the orientation of the fragments. To make constructs with both P1 and P2, the P1 fragment (200 bp) was amplified from OG1RF genomic DNA with primers orf1p-up and orf1p-down (table 1) and inserted into XbaI/BamHI sites of pAT18 containing P2+gls24 or P2+gls24+glsB. To make the complementation constructs with P1 but not P2, the construct containing P1+P2+gls24 was digested with EcoRI and relegated to produce pAT18 containing P1, and the amplified fragment with gls24+glsB or glsB (primers gls24rbs/Eglsdown or glsBrbs/Eglsdown; table 1) was inserted into the BamHI/KpnI sites to produce pAT18 containing P1+gls24+glsB or P1+glsB. These constructs were individually electroporated into TX10100 to produce complementation derivatives, which were confirmed by PCR and Southern-blot analysis. Growth curves of OG1RF, mutants, and complementation strains and stability of single-crossover and complementation derivatives were determined as described elsewhere [14].
Bile-salts resistance and other assays.
After growing in BHIB for 24 h at 37°C, 1 mL of E. faecalis culture was aliquoted, harvested, and resuspended in 1 mL of fresh BHIB, from which 20 L was inoculated into 1 mL of fresh BHIB containing 0.3% bile salts. After incubation for 30 min in a 37°C water bath, the bacterial cells were counted and the percentage survival was calculated by comparing colony-forming units at 30 min to colony-forming units at time (t) 0, by use of the equation
Other potentially stressful conditions examined included growth in BHIB with 40% horse serum, at high temperature (55°C), low pH (pH 3.4), and in an anaerobic chamber. Biofilm and gelatinase production assays were performed as described elsewhere [12, 15].
Production of recombinant proteins and antiserum.
A 561-bp fragment containing the full-length gls24 gene was amplified from OG1RF genomic DNA with primers Egls24up and Egls24down (table 1) and was cloned in-frame into BamHI/KpnIdigested pQE30 (QIAexpressionist; QIAGEN). The His-tagged Gls24 protein was expressed in E. coli and purified under native conditions, as described by the manufacturer. The purified His-tagged Gls24 protein was sent to Bethyl Laboratories (Montgomery, TX), where immune rabbit serum (IRS) was generated by intravenous immunization with His-tagged Gls24 protein in PBS.
Protein analysis.
To localize the Gls24 protein, surface proteins and cell lysates were prepared by Zwittergent (Calbiochem) and sodium dodecyl sulfate treatment, as described elsewhere [16], and supernatant proteins were prepared by concentration to 1/30 volume with a centrifugal filter device (10-kDa cutoff; Millipore Corporation). SDS-polyacrylamide gel electrophoresis, Western blots, and whole-cell ELISA were performed as described elsewhere [16, 17].
Mouse peritonitis and antibody protection model.
The mouse peritonitis model, calculation of LD50 values and Kaplan-Meier survival curves, and log-rank analysis were performed as described elsewhere [14, 18], except that 12.5% sterile rat fecal extract (SRFE) was used. The inocula were adjusted using an optical density at 600 nm (OD 600), and the number of colony-forming units in each inoculum was determined.
Prior to their use, anti-Gls24 IRS and preimmune rabbit serum (PRS) were absorbed with 1 × 109 cells of TX10100 for 1 h at 4°C (IRS-abs and PRS-abs), to remove preexisting antibody. To determine the effect of IRS-abs on colony-forming units of OG1RF in spleens, 8 mice/group were injected intraperitoneally (ip) with 200 L of either anti-Gls24 IRS-abs (titer 1 : 600) or PRS-abs 1 h before ip injection of OG1RF (1.7 × 107 cfu in 1 mL of saline containing 12.5% SRFE). Mice were killed at 72 h, and spleens were removed, weighed, homogenized, and serially diluted in saline for colony count. The experiment was performed twice, and comparison between the 2 groups (IRS-abs vs. PRS-abs) was performed by use of nonparametric Mann-Whitney U tests. For determination of antibody effect on mortality of OG1RF, 9 mice/group were injected ip with 300 L of either IRS-abs or PRS-abs 1 h before and 4 h after OG1RF challenge (1 × 109 cfu in 1 mL of saline containing 12.5% SRFE). The experiment was performed twice; in each experiment, an additional 34 mice were challenged with OG1RF without serum (infection control), and 23 mice received IRS-abs or PRS-abs without bacterial challenge (serum control). Comparison of the number of surviving mice or the survival curves between the 2 groups (IRS-abs vs. PRS-abs) was performed by use of Fisher's exact test or log-rank analysis.
RESULTS
To test whether disruption of gls24 had an effect on E. faecalis virulence, TX10100 was examined in a mouse peritonitis model, along with wild-type OG1RF. Initial experiments using inocula of 1.6 × 109 cfu and 6 mice/group showed that the gls24 disruption mutant was not lethal (100% survival at 96 h with inocula of 1.6 × 109 cfu), whereas wild-type OG1RF showed 83.3%, 66.7%, and 16.7% lethality at 96 h with inocula of 1.2 × 109, 6 × 108, and 3 × 108 cfu, respectively (P < .001 by log rank for 1.6 × 109 cfu of TX10100 vs. 1.2 × 109 cfu of wild-type OG1RF). A larger-scale experiment was then performed for TX10100 and wild-type OG1RF with inocula up to 1.6 × 1010 cfu (for TX10100) and 12 mice/group (figure 1B). The results confirmed that TX10100 was highly attenuated in the mouse peritonitis model, compared with wild-type OG1RF (91.7% survival after an inoculum of 1.6 × 109 cfu [TX10100] vs. 8.3% survival after an inoculum of 1.3 × 109 cfu [wild-type OG1RF] at 120 h; P < .001). The LD50 values for TX10100 and wild-type OG1RF were found to be 8.2 × 109 and 3.8 × 108 cfu, respectively. The colony-forming units recovered from spleens of dead mice receiving TX10100 were comparable on BHI agar and on BHI-kanamycin agar plates, indicating that the disruption was stable in vivo.
RT-PCR with RNA extracted from wild-type OG1RF and TX10100 showed no obvious difference in orf4 levels (figure 3A), suggesting that gls24 is not involved in autoregulation of its upstream genes under our testing conditions. Transcription of glsB was not detectable in TX10100 (figure 3A), consistent with gls24 and glsB being cotranscribed. To investigate whether glsB contributes to E. faecalis virulence, a glsB inactivation mutant of OG1RF was constructed and confirmed by RT-PCR, which showed transcription of gls24 but not glsB (figure 3B). The glsB mutant showed results similar to those for wild-type OG1RF with multiple inocula in the mouse peritonitis model (data not shown), indicating that glsB does not contribute to E. faecalis virulence in this model.
Characterization of the Gls24 protein in E. faecalis.
The gene products of gls24 and glsB have predicted molecular masses of 20.2 and 7.85 kDa and PIs of 5.21 and 11.26, respectively. Gls24 was predicted by PSORT to be a cytoplasmic protein, and GlsB was predicted to be a membrane protein. Recombinant Gls24 protein was successfully expressed in E. coli and used to generate antiserum, whereas attempts to express recombinant GlsB in E. coli were not successful. When anti-Gls24 IRS was used, Gls24 (which ran at 28 kDa on an SDS-PAGE gel) was found in both log- and stationary-phase OG1RF cells (figure 5), whereas PRS did not detect the Gls24 protein band (data not shown). Western blots with different fractions of OG1RF cells detected the Gls24 protein in the Zwittergent preparations (figure 5) and in cell lysates after Zwittergent preparation (data not shown) but not in concentrated culture supernatants (figure 5), suggesting that some Gls24 is present on the cell surface. Although the Gls24 protein was absent from the gls24 disruption mutant TX10100, transcomplementation with P2+gls24 or P2+gls24+glsB restored Gls24 expression, and complementation strains with P1+P2+gls24+glsB overexpressed Gls24, which showed a smear on Western blot, likely as a result of degradation of the protein (figure 5). Reproducible results were obtained with multiple extracts, and whole-cell ELISA also detected Gls24 on the surface (data not shown). Western blots with whole-cell lysates of wild-type OG1RF and TX10100 showed that, after absorption, anti-Gls24 IRS-abs specifically reacted with Gls24, whereas the anti-Gls24 antibody was not present in the PRS-abs (figure 5C).
DISCUSSION
Identification of factors important for enterococcal virulence may help in the design of treatment (such as active or passive immunotherapy) against these organisms, which are a leading cause of nosocomial infections. In the present study, we identified gls24, which was previously shown by Giard et al. to be a general stress gene [9]. Our interest in gls24 was enhanced when we found that the gls24 disruption mutant (TX10100) was highly attenuated in vivo. For example, even after inoculation with 1 × 109 cfu, TX10100 was not lethal in the mouse peritonitis model, whereas the wild-type OG1RF caused 90% death at the same inoculum. The LD50 of TX10100 was 21.6-fold higher than that of wild-type OG1RF, making it, to our knowledge, the most attenuated of all published E. faecalis OG1RF mutants. Our previously published attenuated E. faecalis mutants generally showed prolonged survival, compared with wild-type OG1RF, with similar or slightly higher LD50 values, in the mouse peritonitis model; for example, the fsr and sprE mutants had LD50 values similar to wild-type OG1RF [12], and the LD50 values of etaRS, orfde4, orfde6, gelE, and efaA mutants were 4.6-, 2.5-, 2.1-, 1.6-, 1.4-fold higher, respectively, than that of wild-type OG1RF [3, 14, 18, 19]. When the hemolysin-encoding plasmid pAM714 was introduced into OG1RF, an 35-fold lower LD50 was observed [18, 20]. However, hemolysin is not present in wild-type OG1RF; thus, an isogenic mutant was not available to be compared with the gls24 mutant of OG1RF. We have to point out that these results were obtained at different times with different batches of SRFE; thus, the results may not be directly comparable. Nevertheless, we recently repeated the experiment with an sprE mutant [12] with the same batch of SRFE used for the gls24 mutant and obtained results similar to those previously published. Although it was observed previously by Giard et al. that the gls24 mutant had a longer generation time (58 vs. 45 min) in a semisynthetic medium (Bacto Folic AOAC medium) and shorter chain length, compared with its parental strain JH2-2 (a nongelatinase-producing strain) [9], we did not detect a growth rate or morphological difference between the gls24 mutant and its parental strain OG1RF (gelatinase and serine protease producer) when they were grown in BHIB, suggesting that, for wild-type OG1RF versus TX10100, the different phenotypes are not explained by a growth effectat least, not one observed in BHIB.
To further study the role of gls24 in E. faecalis virulence, antiserum against recombinant Gls24 was generated, and antibody protection assays showed that mice injected with anti-Gls24 antiserum had lower bacterial counts in their spleens (23-fold lower than in mice given PRS; P = .02) after sublethal OG1RF challenge, as well as significantly decreased mortality after lethal OG1RF challenge. These results suggest that the Gls24 protein is a potential candidate for immunotherapy for E. faecalis infection. As far as we could determine from a literature search, the only other antiserum that has been shown to be protective in E. faecalis infection was against a polysaccharide [21]. In that study, the investigators showed that injection of their antiserum into mice reduced E. faecalis counts in mouse organs, compared with counts after injection of preimmune serum or sterile saline; however, the effect of their antiserum on mouse mortality was not studied [21]. Antibody to aggregation substance has also been examined for a protective effect (in a rabbit endocarditis model), but the results were negative [22].
The gls24 locus previously identified by Giard et al. contains 4 ORFs (orf14) upstream of gls24 and 1 ORF (glsB) downstream [9]. In that study, Northern blots from strain JH2-2 showed transcripts with the length of orf1glsB and gls24glsB, and 2 transcriptional start sites were identified by primer extension, 1 upstream of orf1 and 1 upstream of gls24. Our RT-PCR results with strain OG1RF also indicate that there is cotranscription of orf1glsB, and we have demonstrated that the transcript includes orf1 and glsB but not the genes upstream or downstream. Since our RNA analysis showed that disruption of gls24 also disrupted the transcription of glsB, we constructed a glsB mutant (TX10200) and found that inactivation of glsB alone caused mortality in mice that was similar to that caused by wild-type OG1RF, suggesting that gls24, but not glsB, contributes to E. faecalis virulence.
The increased sensitivity of TX10100 to bile salts at the late stationary phase is consistent with the results of the study by Giard et al. showing that gls24 is important for bile-salts resistance after starvation [9]. In their article, Giard et al. indicated that complementation with gls24 and glsB only partially restored bile-salts resistance of the gls24 mutant, but it is not clear exactly what region was used for complementation. Since there are 2 possible promoters (P1 and P2) that may drive transcription of gls24 and glsB (on the basis of our results and those of Giard et al.), in the present study we constructed different complementation derivatives of TX10100 and examined their resistance to bile salts. The results showed that the 2 promoters P1 and P2 (but especially the remote promoter, P1) and both gls24 and glsB are important for bile-salts resistance under our testing conditions. The function of the gls24 gene and the operon in vivo remains unclear, but it may be involved in adaptation to the in vivo environment(s), suggested by their involvement in stress response in vitro, and is an area of interest for future studies.
In summary, gls24 was identified in E. faecalis strain OG1RF and was shown to be important for virulence as well as stress response. Antiserum against the Gls24 protein was able to protect against E. faecalis infection in a mouse peritonitis model, illustrating the potential usefulness of this protein as an immunotarget.
Acknowledgments
We thank Suresh R. Pai and Kavindra V. Singh for performing some of the animal experiments, Ernesto Baca for his assistance with consensus search of the Enterococcus faecalis V583 genome database, and Sandor E. Karpathy for providing some of the primers used in reverse-transcriptase polymerase chain reactions.
References
1. Murray BE. The life and times of the Enterococcus. Clin Microbiol Rev 1990; 3:4665. First citation in article
2. Gilmore MS, Coburn PS, Nallapareddy SR, Murray BE. Enterococcal virulence. In: Gilmore MS, Clewell D, Courvalin P, Dunny GM, Murray BE, Rice L, eds. The enterococci: pathogenesis, molecular biology and antibiotic resistance. Washington, DC: ASM Press, 2002. First citation in article
3. Teng F, Wang L, Singh KV, Murray BE, Weinstock GM. Involvement of PhoP-PhoS homologs in Enterococcus faecalis virulence. Infect Immun 2002; 70:19916. First citation in article
4. Teng F, Jacques-Palaz KD, Weinstock GM, Murray BE. Evidence that the enterococcal polysaccharide antigen gene (epa) cluster is widespread in Enterococcus faecalis and influences resistance to phagocytic killing of E. faecalis. Infect Immun 2002; 70:20105. First citation in article
5. Brenot A, King KY, Janowiak B, Griffith O, Caparon MG. Contribution of glutathione peroxidase to the virulence of Streptococcus pyogenes. Infect Immun 2004; 72:40813. First citation in article
6. Janulczyk R, Ricci S, Bjorck L. MtsABC is important for manganese and iron transport, oxidative stress resistance, and virulence of Streptococcus pyogenes. Infect Immun 2003; 71:265664. First citation in article
7. Cotter PD, Emerson N, Gahan CG, Hill C. Identification and disruption of lisRK, a genetic locus encoding a two-component signal transduction system involved in stress tolerance and virulence in Listeria monocytogenes. J Bacteriol 1999; 181:68403. First citation in article
8. Miller AA, Engleberg NC, DiRita VJ. Repression of virulence genes by phosphorylation-dependent oligomerization of CsrR at target promoters in S. pyogenes. Mol Microbiol 2001; 40:97690. First citation in article
9. Giard JC, Rince A, Capiaux H, Auffray Y, Hartke A. Inactivation of the stress- and starvation-inducible gls24 operon has a pleiotrophic effect on cell morphology, stress sensitivity, and gene expression in Enterococcus faecalis. J Bacteriol 2000; 182:451220. First citation in article
10. Giard JC, Hartke A, Flahaut S, Boutibonnes P, Auffray Y. Glucose starvation response in Enterococcus faecalis JH22: survival and protein analysis. Res Microbiol 1997; 148:2735. First citation in article
11. Murray BE, Singh KV, Ross RP, Heath JD, Dunny GM, Weinstock GM. Generation of restriction map of Enterococcus faecalis OG1 and investigation of growth requirements and regions encoding biosynthetic function. J Bacteriol 1993; 175:521623. First citation in article
12. Qin X, Singh KV, Weinstock GM, Murray BE. Effects of Enterococcus faecalis fsr genes on production of gelatinase and a serine protease and virulence. Infect Immun 2000; 68:257986. First citation in article
13. Trieu-Cuot P, Carlier C, Poyart-Salmeron C, Courvalin P. Shuttle vectors containing a multiple cloning site and a lacZ alpha gene for conjugal transfer of DNA from Escherichia coli to gram-positive bacteria. Gene 1991; 102:99104. First citation in article
14. Xu Y, Singh KV, Qin X, Murray BE, Weinstock GM. Analysis of a gene cluster of Enterococcus faecalis involved in polysaccharide biosynthesis. Infect Immun 2000; 68:81523. First citation in article
15. Mohamed JA, Huang W, Nallapareddy SR, Teng F, Murray BE. Influence of clinical origin, especially endocarditis, and of various genes on biofilm formation by Enterococcus faecalis. Infect Immun 2004; 72:365863. First citation in article
16. Xu Y, Jiang L, Murray BE, Weinstock GM. Enterococcus faecalis antigens in human infections. Infect Immun 1997; 65:420715. First citation in article
17. Nallapareddy SR, Weinstock GM, Murray BE. Clinical isolates of Enterococcus faecium exhibit strain-specific collagen binding mediated by Acm, a new member of the MSCRAMM family. Mol Microbiol 2003; 47:173347. First citation in article
18. Singh KV, Qin X, Weinstock GM, Murray BE. Generation and testing of mutants of Enterococcus faecalis in a mouse peritonitis model. J Infect Dis 1998; 178:141620. First citation in article
19. Singh KV, Coque TM, Weinstock GM, Murray BE. In vivo testing of an Enterococcus faecalis efaA mutant and use of efaA homologs for species identification. FEMS Immunol Med Microbiol 1998; 21:32331. First citation in article
20. Ike Y, Hashimoto H, Clewell DB. Hemolysin of Streptococcus faecalis subspecies zymogenes contributes to virulence in mice. Infect Immun 1984; 45:52830. First citation in article
21. Huebner J, Quaas A, Krueger WA, Goldmann DA, Pier GB. Prophylactic and therapeutic efficacy of antibodies to a capsular polysaccharide shared among vancomycin-sensitive and -resistant enterococci. Infect Immun 2000; 68:46316. First citation in article
22. McCormick JK, Hirt H, Waters CM, Tripp TJ, Dunny GM, Schlievert PM. Antibodies to a surface-exposed, N-terminal domain of aggregation substance are not protective in the rabbit model of Enterococcus faecalis infective endocarditis. Infect Immun 2001; 69:330514. First citation in article, 百拇医药(Fang Teng, Esteban C. Nan)