The Pathogenic Role of Staphylococcus epidermidis Capsular Polysaccharide/Adhesin in a Low-Inoculum Rabbit Model of Prosthetic Valve Endocarditis
Background The capsular polysaccharide/adhesin (PS/A) antigen of Staphylococcus epidermidis was required to produce endocarditis in a rabbit model in which infection resulted from hematogenous spread of bacteria from a contaminated catheter in the jugular vein. However, many prosthetic valve endocarditis (PVE) infections probably result from direct contamination of the valve with small numbers of bacteria during surgery. The role of PS/A in this situation was evaluated by modifying a rabbit model of endocarditis to partially mimic PVE.
Methods and Results A Teflon catheter was contaminated with graded inocula of either PS/A-positive S epidermidis strain M187sp11 or the PS/A-negative, isogenic strain M187sn3 and inserted into the left ventricle through the aortic valve. The PS/A-positive strain had a 50% infectious dose of 1.1×102 cfu (95% CI, 3.3 to 3.7×103) compared with 8.5×104 cfu of the PS/A-negative strain (95% CI, 8.6×103 to 8.5×105). The odds for developing endocarditis were estimated to be 42 times higher for any given inoculum level of the PS/A-positive strain (P=.1). When the PS/A-positive strain was adherent to a catheter surface it survived in rabbit blood, whereas under the same conditions the PS/A-negative strain was killed ≈90% in 1 hour.
Conclusions Direct contamination of an intraventricular foreign body by low levels of PS/A-positive S epidermidis results in endocarditis in rabbits, but at suitably high doses PS/A-negative strains have sufficient virulence to infect cardiac vegetations. PS/A enhances but is not absolutely required for bacterial virulence in a rabbit model of PVE.
The incidence of prosthetic valve endocarditis (PVE) is about 2% to 3% in patients undergoing valve replacement,1 with Staphylococcus epidermidis accounting for about 30% overall of these infections.2 Contamination of valves leading to PVE has been reported to occur both directly during the intraoperative period1 3 and indirectly in the postoperative period due to hematogenous spread of bacteria from contaminated intravascular catheters4 or other sites of infection.5 Hematogenous spread of bacteria to the prosthetic valve in the postoperative period probably would require that the bacterial strain possess virulence factors that allow for resistance to host bloodborne defense mechanisms, principally involving phagocytes, antibodies, and complement. However, direct contamination of the valve during the intraoperative period may depend on bacterial factors that promote adherence to a valve surface, survival on that surface, and spread to damaged cardiac tissue. Resistance to host defenses would be a factor here only if the defense mechanisms were operative against surface-bound bacteria. Surface-bound bacteria grow as biofilms, and organisms have been reported to express markedly different external structures in biofilms from those expressed by the suspended or planktonic bacterial cells usually used in laboratory studies. The biofilm state has been reported to alter resistance to both host defense mechanisms and antibiotics.6 7 8 9 10 Therefore, the bacterial factors needed to initiate PVE from a directly contaminated valve may differ from those needed to initiate PVE via hematogenous bacterial spread.
Previous results have demonstrated that the capsular polysaccharide/adhesin (PS/A) antigen of S epidermidis is a critical virulence factor in a rabbit model of endocarditis11 in which a contaminated intrajugular catheter attached to a subcutaneous osmotic pump was used to initiate bacteremia followed by endocarditis in rabbits that also had a second catheter inserted into the ventricle through the aortic valve. A transposon mutant of S epidermidis strain M187 deficient in elaboration of PS/A12 was virtually avirulent in this model, being isolated from only 1 of 98 blood cultures and never initiating endocarditis. In contrast, a PS/A-producing isogenic strain was isolated from 61% of blood cultures and caused endocarditis in 75% of the challenged animals. These differences in virulence were associated with decreased resistance of the PS/A-negative mutant to the phagocytic activity of leukocytes and complement. As a result, the mutant strain could not survive at a sufficient level in the blood to infect the intraventricular catheter and/or damaged cardiac tissue. PS/A also promotes bacterial adherence to biomaterials,12 13 14 which may be important in pathogenesis of PVE caused by this organism.
To evaluate the role of bacterial virulence factors in a setting that resembles direct intraoperative contamination of a prosthetic valve, we modified the rabbit model of endocarditis15 to this situation. We found that the intraventricular catheter used to initiate endocardial infection could be contaminated before surgery by S epidermidis and, when placed in the left ventricle in close proximity to the aortic valve, initiate endocarditis. The model was therefore useful to quantify differences in the ability of two isogenic strains of S epidermidis, differing only in elaboration of PS/A,11 12 to cause endocarditis without regard to the fact that PS/A is required for hematogenous spread of S epidermidis from one infectious focus to the heart. This model also allowed us to monitor bacteremia during the course of the study and determine endocarditis-associated pathologies when the animals were killed and autopsied. Although the major difference between this model and human PVE was the lack of an actual valve being placed into the animal’s heart, the implantation of a contaminated intraventricular Teflon catheter resulted in the development of endocarditis even when low-challenge inocula of S epidermidis were used, mimicking to some degree the clinical situation of contamination of implanted material with small numbers of bacteria during surgery. Therefore, this model provided a unique perspective on the pathogenesis of endocarditis compared with previously published studies and provided a potential explanation for why a significant proportion (approximately one third) of isolates of S epidermidis from cases of PVE fail to elaborate surface factors that inhibit host phagocytic defenses.
Bacterial Strains and Growth Conditions
We used isogenic PS/A-positive and PS/A-negative S epidermidis strains M187sp11 and M187sn3, which have been described previously.11 12 Based on all readily definable phenotypic markers, these strains only differ in elaboration of PS/A.12 These strains were routinely grown on Memphis agar16 containing erythromycin (1 μg/mL) and chloramphenicol (5 μg/mL), a medium that selects for the antibiotic resistances encoded by transposon Tn917LTV1 carried by these strains. This medium was also used to confirm the identity of isolates recovered from infected rabbits. Congo red agar17 was used to differentiate M187sp11 from M187sn3, as these strains have distinctive morphologies on this medium due to their differences in elaboration of PS/A.
Preparation of Bacterial Inoculum for Infection
Strain M187sp11 or M187sn3 was inoculated from Memphis agar into 100 mL of tryptic soy broth (TSB) and incubated overnight at 37°C with end-over-end rotation. The next day, the agglutinated mass of cells of strain M187sp11 was disrupted by repeated passage through a syringe with a 25-gauge needle attached. Cells of strain M187sn3 needed no such treatment. The cultures then were adjusted to an optical density of 1.0 at 650 nm to achieve a concentration of ≈109 cfu/mL. From this starting level the culture was diluted further in TSB to obtain suspensions with various concentrations of bacterial cells. Five to ten 3F Teflon catheters were placed into the suspension and incubated at room temperature for 30 minutes in order to contaminate them. Catheters were rinsed in PBS, and some were individually inserted into rabbit hearts; the remaining catheters were used to determine the level of contamination achieved (cfu per catheter) according to methods we have described previously.11
Endocarditis Model and Determination of Infection
We used many of the same techniques as described previously.11 15 18 Sixty-four New Zealand White rabbits weighing 2.5 to 3.0 kg were anesthetized by a single intramuscular injection of ketamine hydrochloride (Ketalar, 40 mg/kg, Parke Davis) and xylazine (Rompun, 10 mg/kg, Haver). Next the contaminated catheters were inserted into the left ventricles of anesthetized rabbits via a cutdown incision made in the right carotid artery. The catheter was carefully tied in place with ligatures around the outside of the artery, the surgical incision was closed, and the rabbit was returned to its cage. Blood samples were taken from the ear vein of most rabbits over the next 20 to 22 days to culture for the detection of bacteremia, although a few rabbits were studied for shorter (11 to 13 days) or longer (42 to 44 days) periods. Blood was cultured as described11 first by inoculating 5 mL of blood into 50 mL of TSB and then subculturing any growth onto the differential and selective media described above to determine whether S epidermidis strain M187sp11 or strain M187sn3 was present. All of these procedures were approved by the Harvard Medical Area Standing Committee on Animals.
At the end of the experimental period, rabbits were killed and autopsies were performed as described11 18 to determine the following outcomes: presence or absence of an endocardial vegetation and the weight of any vegetations present; level of infection in the vegetations by quantitative bacterial cultures; and the cfu of S epidermidis remaining on the intraventricular catheter. Because we cultured the entire homogenate of the endocardial lesions, endocarditis could be defined as the presence of ≥1 cfu of the appropriate strain of S epidermidis, whereas lack of endocarditis was defined as cultures of the homogenized vegetation remaining sterile after 10 days of incubation at 37°C. Given the relative ease with which S epidermidis grows in laboratory culture media, it is unlikely that we failed to identify infected vegetations because of limitations in culture sensitivity.
To prepare a bacterial antigen for detecting immune responses of infected rabbits, a cell wall extract of S epidermidis strain M187sn3, containing as the major (>80%) component teichoic acid, was prepared by heating bacterial cells for 1 hour at 95°C at a pH of 5.0 followed by dialysis and lyophilization. Although we have previously shown11 18 that teichoic acid is a major antigen of S epidermidis provoking antibody responses during experimental infection in rabbits, we used a less pure preparation here, since preliminary studies indicated that some animals responded to cell wall components other than teichoic acid. Antibody in rabbit serum samples to the extract was measured in an ELISA as described18 19 with the use of serum samples obtained before and at the end of the experimental period. The titer also was determined as described11 by means of regression analysis to find the dilution of serum yielding an optical density of 0.2 after 60 minutes of incubation of the ELISA plate containing enzyme substrate. The highest concentration of rabbit serum tested was a dilution of 1:25; sera at this concentration that failed to reach an optical density of 0.2 after 60 minutes of incubation were given a titer of <25.
Adherence of Bacteria to Platelet-Fibrin Clots
The method of Chugh et al20 was used to compare the adherence of S epidermidis strains M187sp11 and M187sn3 to platelet-fibrin clots.
Survival of Catheter-Associated Bacteria in Rabbit Blood
To determine the level of survival of S epidermidis strains M187sp11 and M187sn3 in rabbit blood when the organisms were adherent to a catheter surface, whole blood was obtained from a rabbit and treated with sodium citrate as an anticoagulant. The blood was aliquoted, and the cellular component was separated from the plasma. The plasma was either heated at 56°C for 30 minutes to inactivate complement components or adsorbed twice for 30 minutes at 4°C with 1 mg of lyophilized cells of strain M187sn3. Untreated plasma samples along with the cells were held at 4°C during this time. The various plasma fractions then were added back to their original cellular fractions, and catheters that had been contaminated in a manner identical to that used for inducing endocarditis were placed into the blood samples. These were incubated at 37°C for 60 minutes, and surviving bacteria were enumerated by standard techniques. In an individual assay, triplicate determinations of surviving bacteria were made and each assay was repeated at least three times. A sample of serum from this rabbit also was analyzed for the presence of antibody to S epidermidis M187sn3 cells with use of the ELISA method described above, and a titer of <25 was determined.
The dose of bacteria needed to infect a specific percentage of animals was determined by logistic regression (Logit) analysis using the systat statistical package (version 5.2.1, SYSTAT, Inc). Spearman rank correlations, Mann-Whitney U tests, Wilcoxon signed rank tests, and t tests were performed with the statview se+graphics software program (Abacus Concepts) on a Macintosh computer. The relationship of inoculum to blood culture results and development of endocarditis was evaluated by logistic regression with the use of an overdispersed binomial model to account for nonindependent observations within animals. The odds ratios for the occurrence of a positive or negative blood culture before or after day 10, when 50% of the blood cultures were obtained, were determined by longitudinal data analysis based on models appropriate for repeated cultures.21 The splus software program (written by Vincent Carey, Channing Laboratory, Boston) was used for this calculation.
Development of Endocarditis and Associated Findings
Tables 1 and 2 present summaries of the principal data obtained during this study with S epidermidis strains M187sp11 and M187sn3, respectively. For clarity, results are grouped together for rabbits that received challenge inocula within 0.5 log10 cfu, except in two cases. The inability to exactly replicate the challenge inocula within a group was due to inherent variability in the experimental protocol used to contaminate the catheters before surgery and limitations on the numbers of animals that could be infected during a single time period. In general, most rabbits were infected for a period of 20 to 22 days, with the following exceptions: among the rabbits challenged with strain M187sp11, 6 were infected for 11 to 13 days, 2 for 28 days, and 3 for 42 to 44 days; among the rabbits infected with strain M187sn3, 4 were infected for 11 to 13 days. Two rabbits died 7 days after surgery, one challenged with 1.8×106 cfu of strain M187sp11 and one challenged with 8.8×105 cfu of strain M187sn3. At necropsy, both of these animals had positive cultures of endocardial lesions. Some rabbits were infected for a shorter period to determine whether endocarditis would develop in <20 days after high-dose challenge, and some rabbits were infected for >22 days after low-dose challenge to determine whether positive blood cultures would result in these animals if a longer study period was used. Rabbits challenged with low doses of strain M187sp11 generally had negative blood cultures within the first 3 weeks (only 12 [15%] of the 82 blood cultures taken in the first 3 weeks from rabbits challenged with ≤2×102 cfu of S epidermidis strain M187sp11 were positive).
Development of Endocarditis
The ID50 for inducing endocarditis by the PS/A-positive strain M187sp11 was 1.1×102 cfu (95% CI, 3.3×100 to 3.7×103 cfu) compared with 8.5×104 cfu (95% CI, 8.6×103 to 8.5×105 cfu; P<.05, logit analysis) for the PS/A-negative strain M187sn3. All five rabbits challenged with doses ≥4.2×105 cfu of strain M187sp11 and killed between 11 and 13 days developed endocarditis. Even more striking was the observation that the infectious dose for 25% of the rabbits challenged with strain M187sp11 was 3 cfu (95% CI, .01 to 81 cfu) compared with 5.7×103 cfu (95% CI, 2.2×102 to 1.5×105 cfu) for animals challenged with strain M187sn3. The large confidence intervals for these determinations probably resulted from the small number of animals studied, but given the obvious differences in pathogenesis between the two bacterial strains, it seemed injudicious to infect more animals for the purpose of achieving narrower confidence intervals. These results also show that regardless of bacterial phenotype for PS/A, endocarditis can be provoked in this model with extremely low doses of S epidermidis, whereas almost all other studies published to date require challenge doses of 108 to 109 cfu of S epidermidis to elicit endocarditis.22 23 24 25 26
We also fit a logistic regression model to analyze the relationship of the inoculum size to development of endocarditis. The odds for developing endocarditis were estimated to be 42 times higher for the PS/A-positive strain at any given inoculum compared with the PS/A-negative strain (P=.1). The relationship was linear over the dose ranges evaluated, and the odds ratios for developing endocarditis for a natural log unit change in inoculum (≈2.7 log10 cfu) were 1.36 (95% CI, 1.07 to 1.74) for strain M187sp11 and 1.57 (95% CI, 1.12 to 2.18) for strain M187sn3. There was no significant difference (P=.51) in the slope estimates for these two strains. Along with the ID50 estimates, these results indicate that PS/A contributes to bacterial virulence by lowering the inocula of S epidermidis needed to induce endocarditis in rabbits from a contaminated intraventricular catheter. However, because there was no difference in the slope estimates relating increases in inoculum to the rate at which endocarditis develops, it appears that once a sufficient dose of S epidermidis is delivered to the heart, increasing the bacterial inoculum results in comparable increases in the incidence of endocarditis whether or not the organism expresses PS/A.
Quantitative Measurement of Infection of Endocardial Vegetations
The development of vegetations and the cfu of S epidermidis per gram of tissue were analyzed next, although in these analyses it was possible that the vegetation weight determined at autopsy was not fully representative of the situation during the earlier infectious stage of the disease, since part of the vegetation may have been lost from blood flow or valve movement. With this caveat, there were significant correlations between the size of the initial inoculum contaminating the intraventricular catheter and the final cfu of S epidermidis per gram of cardiac vegetation determined for PS/A-positive strain M187sp11 (P=.03, Spearman rank correlation) and for PS/A-negative strain M187sn3 (P=.002, Spearman rank correlation). Total vegetation weight also correlated with initial inoculum size in animals challenged with strain M187sn3 (P=.03, Spearman rank correlation) but only modestly so for animals challenged with strain M187sp11 (P=.08, Spearman rank correlation). If the 5 animals challenged with a low dose (≤1.8×102 cfu) of strain M187sp11 and infected for >22 days, which should have produced heavier vegetations, were excluded from the analysis correlating vegetation weight with inoculum, then the significance level equaled .04. These results suggest that once endocarditis is initiated there is an early correlation of inoculum size with vegetation weight and level of infection, but as endocardial infection proceeds, the dependence of cfu per gram of vegetation and vegetation weight on the initial inoculum lessens.
Quantitative Measurement of Bacterial Survival on Intraventricular Catheters
Quantitative measurement of cfu of bacteria surviving on the intracardiac catheter for the most part reflected the level of infection of the cardiac vegetation. At autopsy, 100% of animals that developed endocarditis with strain M187sp11 had a positive catheter culture, and 100% of animals with sterile vegetations also had sterile intraventricular catheters. Six of 8 animals with sterile catheters after challenge with M187sp11 received the lowest challenge doses (<200 cfu/catheter), indicating that intracardiac catheters contaminated with small numbers of PS/A-positive S epidermidis can be sterilized in rabbits. Of the 27 animals challenged with strain M187sn3, 14 developed endocarditis and 13 of these had a positive intraventricular catheter culture; the single exception had only 1 cfu/g in the endocardial vegetation. Of the remaining 13 animals challenged with strain M187sn3 that did not develop endocarditis, 8 developed sterile vegetations in the presence of an intraventricular catheter that remained contaminated at autopsy. The quantitative level of residual bacterial contamination of the catheter determined for these 8 rabbits at autopsy compared with the bacterial level initially determined at infection indicated that >99% of the infecting inoculum of strain M187sn3 was lost from the intracardiac catheter during the experimental period. When compared with strain M187sp11, where endocarditis always occurred as long as the intracardiac catheter remained contaminated, strain M187sn3 appears to have a marked reduction in the ability to survive on the intracardiac catheters in vivo and a lessened ability of the residual bacterial cells to spread to cardiac tissue from a contaminated foreign body and induce endocarditis.
Blood Culture Results
The results from cultures of blood obtained during the study period are given in Table 3⇓. There were more total cultures from animals challenged with strain M187sp11 because a greater number of rabbits were infected with this strain (Tables 1⇓ and 2⇓). There were significantly fewer (P≤.001, longitudinal data analysis) positive cultures of blood obtained over the first 9 days, when half of the blood samples were taken, than over the last 12 days (Table 3⇓) in animals challenged with either strain of S epidermidis, probably because the density of bacteria in the endocardial lesion had to become sufficiently high to allow the establishment of bacteremia at a level that would yield a positive blood culture. This conclusion was supported by the observation that blood cultures were positive only in animals challenged with strain M187sp11 that developed endocarditis (Table 3⇓). Similarly, all except 2 positive blood cultures from animals challenged with strain M187sn3 came from animals that developed endocarditis. In addition, 2 of the 3 rabbits challenged with 1.3 to 2.0×102 cfu of strain M187sp11 that were studied for 42 to 44 days after infection developed endocarditis, yet neither had a positive blood culture before day 21. After this initial 3-week period, 4 of 7 blood cultures from one rabbit and 7 of 7 blood cultures from the second rabbit were positive. The third rabbit observed for the longer period did not develop endocarditis and had sterile blood cultures throughout the observation period.
The logistic regression model fitting the proportion of positive blood cultures as a function of inoculum yielded an odds ratio for a positive culture for a natural log (≈2.7 log10) unit change in inoculum of 1.25 (95% CI, 0.99 to 1.58; P=.065) for PS/A-negative strain M187sn3 and an odds ratio of 1.16 (95% CI, 1.04 to 1.30; P=.008) for PS/A-positive strain M187sp11. There was no significant difference in the slope estimates for these parameters between the two strains, and while the overall level of blood cultures positive for strain M187sp11 was higher than for strain M187sn3 (35% versus 19%, respectively; Table 3⇑), the difference was not significant.
Immune Response to Infection
An extract of surface antigens from strain M187sn3, containing principally (80%) teichoic acid along with other cell wall antigens, was used to measure antibody responses in all of the infected rabbits. Three rabbits had a preinfection antibody titer to this extract of >25; two rabbits had titers of 30 and 33, and one rabbit had a preinfection titer of 155. The remaining animals all had preinfection titers of <25. Geometric mean antibody titers in the cohort of blood samples obtained at the end of the experimental period were significantly greater (P<.0001, Wilcoxon signed rank test) than preinfection titers. There was no significant difference in the final antibody titers between animals challenged with either strain of S epidermidis. The correlation of postinfection titer with inoculum size was not impressive; P values (Spearman rank correlation) were .05 for animals challenged with strain M187sp11 and .28 for animals challenged with strain M187sn3.
The immune response reflected the infectious situation. Geometric mean antibody titers to the cell wall extract antigen were higher in animals that developed endocarditis than in those with sterile vegetations (Fig 1⇓; P<.001 for all animals, P=.001 for animals challenged with strain M187sp11, and P=.02 for animals challenged with strain M187sn3; unpaired two-sided t test). None of the 8 animals challenged with strain M187sp11 that had sterile vegetations had an increase in antibody titer compared with 20 of 27 animals with infected vegetations. The final blood sample from 4 of the 7 infected animals without antibody increases was obtained ≤11 days after surgery, which suggests that this period was too short for immune responses to develop routinely. Only 3 of 13 animals challenged with strain M187sn3 developed antibody titers in the presence of sterile endocardial vegetations, and 2 of these 3 had a colonized intraventricular catheter at autopsy. Ten of 14 animals with vegetations infected with the PS/A-negative strain had antibody titer increases; 3 of the 4 without titer changes were animals whose blood samples were obtained ≤13 days after surgery. There were strong direct correlations between the cfu per gram of vegetation and postinfection titer for animals challenged with strain M187sp11 (P=.0001, Spearman rank correlation) and animals challenged with strain M187sn3 (P=.003, Spearman rank correlation). As expected, a direct correlation was observed between increased postinfection titer and the cfu of bacteria remaining on the intracardiac catheter (P<.001 and P=.005, strains M187sp11 and M187sn3, respectively; Spearman rank correlation).
Mechanisms Accounting for Differential Virulence of S Epidermidis Strains M187sp11 and M187sn3
We postulated that two mechanisms might account for the lower ID50 value of strain M187sp11 compared with that of strain M187sn3: There was either a difference in adherence of these strains to platelet-fibrin clots or a difference in their survival in rabbit blood containing a contaminated catheter. When we added 1 to 2×104 cfu of either strain M187sp11 or M187sn3 to platelet-fibrin clots in vitro using the methods described by Chugh et al20 to form the clots, we found no difference in the ability of these strains to adhere to the clots (n=6 replicates for each strain; mean±SD cfu of strain M187sp11 adhering to clots=1.13×104±2.3×103; mean±SD cfu of strain M187sn3 adhering to clots=1.30×104±2.8×103; P=.3, two-sided t test). However, we found significant (P=.02, Mann-Whitney U test) differences in the survival of strain M187sp11 on a contaminated catheter placed into whole, citrated rabbit blood compared with that of strain M187sn3. The PS/A-positive strain M187sp11 was minimally killed after 60 minutes of exposure to rabbit blood, whereas the majority of strain M187sn3 cells were killed (≈90%, Fig 2⇓). Heating of the separated plasma component of blood resulted in a significant reduction in killing of strain M187sn3 to about 55% (P=.02 compared with killing in intact blood, Mann-Whitney U test), while adsorption of the plasma component still resulted in killing of about 70% of this bacterial strain (not significantly different from killing in intact blood). A combination of heating and adsorption of the plasma component of blood nearly eliminated killing of strain M187sn3 (P<.001, Mann-Whitney U test). These results indicate that heat-labile components of blood, most likely complement, possess the majority of the activity that kills S epidermidis strain M187sn3, but there also appears to be a contribution from antibody and/or another serum component that can both bind to this bacterial strain and activate complement. Since the antibody titer in the rabbit blood to the hot pH 5.0 cellular extract of strain M187sn3 was <25, any antibody in rabbit blood that promoted killing of this strain probably was directed to antigens not detected in our ELISA. Alternately, a nonantibody component such as C-reactive protein could have been responsible for the killing activity that remained after plasma was heated.
In this report we have shown that the rabbit model of bacterial endocarditis can be modified such that endocardial infections are elicited by contamination of the intraventricular catheter with bacteria before implantation. Although there are clearly differences between human PVE and the rabbit model in terms of the surgical procedures and the lack of an actual prosthetic valve in the animal’s heart, the rabbit model was developed in an attempt to mimic PVE as best as possible in a small laboratory animal. We used Teflon catheters because this material is commonly used in portions of prosthetic valves27 and has been implicated as the part of the valve that becomes contaminated and leads to PVE.28 In addition, placing a foreign body in close proximity to valvular and myocardial tissue allowed us to study the effect of contaminating the foreign body with low bacterial inocula, simulating what probably happens when prosthetic material is contaminated during placement into the heart during surgery. This is particularly useful for studies of S epidermidis and other coagulase-negative staphylococci, since others who have used the standard rabbit model of endocarditis, where infection is induced by an intravenous bolus of bacteria, needed doses of 108 to 109 cfu to elicit infection.24 25 26 29 30 Such high-level challenge doses limit the ability to quantify and compare the effects of virulence factors on the development of endocarditis.
We chose to evaluate the role of the PS/A antigen of S epidermidis in virulence by using this model, inasmuch as S epidermidis is the principal pathogen causing PVE,31 32 and we have previously reported11 that the absence of PS/A completely inhibits the induction of endocarditis in a rabbit model in which endocardial infection results from hematogenous spread of bacteria from an infected intrajugular catheter. However, since PS/A protects S epidermidis from circulating host phagocytic defenses,11 18 19 it was not clear that PS/A would be critical for virulence when the infecting bacteria were attached to the surface of a foreign body directly introduced into the heart. While the results presented here provide additional evidence that PS/A promotes S epidermidis endocardial infections, they differ significantly from previously published results11 in that PS/A-deficient S epidermidis retains some capacity to cause endocarditis when introduced directly into the heart attached to a contaminated foreign body.
Similar to previous findings,11 we found PS/A was an important virulence factor for S epidermidis, since its presence significantly lowered the ID50 value for initiating endocarditis. This lower ID50 value was also associated with greater survival of the surface-bound PS/A-positive strain in rabbit blood. Again, similar to previous findings,11 resistance of PS/A-positive strains to host bloodborne killing factors played a role in the model of PVE used in this study. What was entirely different from our previous findings was that a sufficiently high-challenge inoculum of the PS/A-deficient strain could cause endocarditis, and the PS/A-negative strain engendered an increase in the rate of occurrence of endocarditis as a function of increasing bacterial inoculum at a rate comparable to that of the PS/A-positive strain. Therefore, in this model of prosthetic valve infection, the absence of PS/A reduces but does not eliminate the pathogenic potential of S epidermidis.
To compare the capacity of comparable inocula of the PS/A-positive and PS/A-negative strains to induce endocarditis, the protocols were designed to obviate the role of PS/A in promoting adherence of S epidermidis to biomaterials. Thus, the data in Tables 1⇑ and 2⇑ show the inocula of already adherent bacteria implanted into the left ventricle. However, to achieve comparable levels of catheter contamination before surgery, the PS/A-negative strain must be made up to a concentration of 102 to 103 more cfu of bacteria per milliliter than the concentration of the PS/A-positive strain.12 33 Thus, if we based the comparison in ID50 values on the initial bacterial concentration used to contaminate the catheters and not on the actual inoculum, the magnitude of the difference in ID50 would increase by 100- to 1000-fold. Therefore, our conclusion that PS/A promotes the capacity of S epidermidis to cause PVE probably is based on a conservative comparative estimate of the ID50 values for the isogenic PS/A-positive and PS/A-negative strains tested here.
Since the magnitude of an immune response often increases with increasing antigenic dose, we were not surprised by the direct correlations observed between levels of contamination of both vegetations and intraventricular catheters and the resultant antibody titers. Another correlation we observed, which was also predictable, was between the cfu per gram of vegetation and the percentage of positive blood cultures (P<.001, Spearman rank correlation coefficient for both strains). Of greater interest were the poor correlations noted between the initial inoculum and the cfu of S epidermidis per gram of cardiac vegetations, the percentage of positive blood cultures, and the postinfection antibody titers. These results indicate that development of endocarditis itself probably is the principal factor correlating positively with levels of colonization of intraventricular catheters, development of bacteremia, and the magnitude of the antibody response. To the extent that the inoculum is responsible for induction of endocarditis, it affects these other measures, but once endocarditis occurs, it is the primary determinant of other measures of infection.
The ability of strain M187sn3 to cause bacteremia after establishing endocarditis was surprising and completely different from previous findings.11 In the prior studies the lack of PS/A significantly enhanced killing of strain M187sn3 by phagocytes and complement,11 18 19 indicating that it is very difficult for PS/A-negative strains to survive while freely floating in blood. However, in the new model of endocarditis reported here, it appeared that once fairly high concentrations of the PS/A-negative strain were established in the endocardial vegetation, bacteremia developed. Every rabbit except one with ≥2 positive blood cultures of strain M187sn3 had an endocardial lesion with ≥4×104 cfu per gram of vegetation. In addition, all except 2 of 136 positive blood cultures were obtained from animals with endocarditis, a further indication that bacteremia occurred subsequent to endocardial infection.
The finding that a PS/A-negative strain of S epidermidis can cause endocarditis under the conditions described here is consistent with our report that over one third of clinical endocarditis isolates of coagulase-negative staphylococci, of which 90% are S epidermidis, do not produce PS/A.33 These isolates were poorly adherent to biomaterials when evaluated in an in vitro adherence assay.33 Although we have no information regarding whether the PS/A-negative clinical isolates caused endocarditis as a result of intraoperative contamination of the prosthetic valve, the results here suggest that PS/A-negative strains of S epidermidis and other similar bacterial isolates that are readily killed by blood are more likely to cause PVE if they do not have to travel by a hematogenous route to infect damaged cardiac tissue or colonize a prosthetic valve. Endocarditis isolates of S epidermidis and other coagulase-negative staphylococci express PS/A at a rate nearly identical to that of skin isolates (62% and 57%, respectively33 ), consistent with the idea that skin serves as a source for S epidermidis infections.
Our goal was to have an animal model that would be useful for exploring pathogenesis and immunity of bacterial endocarditis resulting from intraoperative contamination of an implanted foreign body like a prosthetic valve. Direct contamination of an intraventricular catheter serves this purpose in that endocarditis readily developed after contamination with very low inocula of PS/A-positive S epidermidis. Even the challenge level of the PS/A-negative S epidermidis needed to induce endocarditis was not overwhelmingly high, indicating an important distinction regarding the role of PS/A in virulence from our previous studies.11 Furthermore, development of interventions that reduce the incidence of PVE will need to address infections that occur from either intraoperative or postoperative contamination. Previous results in a rabbit model of endocarditis mimicking postoperative infection have demonstrated the efficacy of active and passive immunotherapy to prevent endocarditis caused by PS/A-elaborating strains of S epidermidis.18 These studies will now be extended to the model of intraoperative infection leading to endocarditis to evaluate the potential efficacy of PS/A-specific immunotherapy in preventing S epidermidis endocarditis in this situation.
This work was supported by National Institutes of Health grant AI-23335, by a Grant-in-Aid from the American Heart Association, and by Sanofti-Winthrop Pharmaceuticals.
- Received February 28, 1995.
- Revision received May 3, 1995.
- Accepted June 13, 1995.
- Copyright © 1995 by American Heart Association
Braimbridge MV, Eykyn SJ. Prosthetic valve endocarditis. J Antimicrob Chemother. 1987;20(suppl A):173-180.
Threlkeld MG, Cobbs CG. Infectious disorders of prosthetic valves and intravascular devices. In: Mandell GL, Douglas RG, Bennett J, eds. Principles and Practice of Infectious Diseases. New York, NY: Churchill Livingston, Inc; 1994:783-793.
Freeman R, King B. Analysis of results of catheter tip cultures in open-heart surgery patients. Thorax. 1975;30:26-30.
Dismukes WE, Karchmer AW, Buckley MJ, Austen WG, Swartz MN. Prosthetic valve endocarditis: analysis of 38 cases. Circulation. 1973;48:365-377.
Evans DJ, Brown MRW, Allison DG, Gilbert P. Susceptibility of bacterial biofilms to tobramycin: role of specific growth rate and phase in the division cycle. J Antimicrob Chemother. 1990;25:585-591.
Gilbert P, Allison DG, Evans DJ, Handley PS, Brown MRW. Growth rate control of adherent bacterial populations. Appl Environ Microbiol. 1989;55:1308-1311.
Jensen ET, Kharazmi A, Lam K, Costerton JW, Hoiby N. Human polymorphonuclear leukocyte response to Pseudomonas aeruginosa grown in biofilms. Infect Immun. 1990;58:2383-2385.
Shiro H, Muller E, Gutierrez N, Boisot S, Grout M, Tosteson TD, Goldmann D, Pier GB. Transposon mutants of Staphylococcus epidermidis deficient in elaboration of capsular polysaccharide/adhesin and slime are avirulent in a rabbit model of endocarditis. J Infect Dis. 1994;169:1042-1049.
Muller E, Huebner J, Gutierrez N, Takeda S, Goldmann DA, Pier GB. Isolation and characterization of transposon mutants of Staphylococcus epidermidis deficient in capsular polysaccharide/adhesin and slime. Infect Immun. 1993;61:551-558.
Tojo M, Yamashita N, Goldmann DA, Pier GB. Isolation and characterization of a capsular polysaccharide/adhesin from Staphylococcus epidermidis. J Infect Dis. 1988;157:713-722.
Muller E, Takeda S, Goldmann DA, Pier GB. Blood proteins do not promote adherence of coagulase-negative staphylococci to biomaterials. Infect Immun. 1991;59:3323-3326.
Christensen GD, Baddour LM, Madison BM, Parisi JT, Abraham SN, Hasty DL, Lowrance JH, Josephs JA, Simpson WA. Colonial morphology of staphylococci on Memphis agar: phase variation of slime production, resistance to beta-lactam antibiotics, and virulence. J Infect Dis. 1990;161:1153-1169.
Takeda S, Pier GB, Kojima Y, Tojo M, Muller E, Tosteson T, Goldmann DA. Protection against endocarditis due to Staphylococcus epidermidis by immunization with capsular polysaccharide/adhesin. Circulation. 1991;84:2539-2546.
Kojima Y, Tojo M, Goldmann DA, Tosteson TD, Pier GB. Antibody to the capsular polysaccharide/adhesin protects rabbits against catheter related bacteremia due to coagulase-negative staphylococci. J Infect Dis. 1990;162:435-441.
Chugh TD, Burns GJ, Shuhaiber HJ, Bahr GM. Adherence of Staphylococcus epidermidis to fibrin-platelet clots in vitro mediated by lipoteichoic acid. Infect Immun. 1990;58:315-319.
Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13-22.
Baddour LM, Christensen GD, Hester MG, Bisno AL. Production of experimental endocarditis by coagulase-negative staphylococci: variability in species virulence. J Infect Dis. 1984;150:721-727.
Carbon C. Experimental endocarditis: a review of its relevance to human endocarditis. J Antimicrob Chemother. 1993;31:71-85.
Galetto DW, Boscia JA, Kobasa WD, Kaye D. Teicoplanin compared with vancomycin for treatment of experimental endocarditis due to methicillin-resistant Staphylococcus epidermidis. J Infect Dis. 1986;154:69-75.
Kennedy S, Chambers HF. Daptomycin (LY146032) for prevention and treatment of experimental aortic valve endocarditis in rabbits. Antimicrob Agents Chemother. 1989;33:1522-1525.
Archer GL, Vazquez GJ, Johnston JL. Antibiotic prophylaxis of experimental endocarditis due to methicillin-resistant Staphylococcus epidermidis. J Infect Dis. 1980;142:725-731.
West TE, Cantey JR, Apicella MA, Burdash NM. Detection of anti-teichoic acid immunoglobulin G antibodies in experimental Staphylococcus epidermidis endocarditis. Infect Immun. 1983;42:1020-1026.
Ivert TS, Dismukes WE, Cobbs CG, Blackstone EH, Kirklin JW, Bergdahl LA. Prosthetic valve endocarditis. Circulation. 1984;69:223-232.
Muller E, Takeda S, Shiro H, Goldmann D, Pier GB. Occurrence of capsular polysaccharide adhesin among clinical isolates of coagulase-negative staphylococci. J Infect Dis. 1993;168:1211-1218.