In vitro assays of Staphylococcus epidermidis chara cteristics a n d outcome in a n endocarditis model

In vitro assays of Staphylococcus epidennidis characteristics and outcome in an endocarditis model. OBJECTIVE: Staphylococcus epidennidis adherence to indwelling polymers is important in prosthetic valve endocarditis. Earlier studies have related streptococcal endocarditis to isolates with high levels of cell-associated hexoses. The objective of the present study was to determine if a relationship exists between an S epidennidis isolate assay score and production/ severity of experimental endocarditis. DESIGN: Groups of patient S epidennidis isolates were screened for surface hexoses and an animal model of endocarditis with isolates testing highest and lowest on the screen was produced. Disease severity produced by 'high hexose' versus 'low hexose' organisms was evaluated. Endocarditis responding variables were bacterial vegetation weight and log10 colony forming units (cfu) and in survival tests, comparative time to death with different isolates. Bacterial characteristics were not measured. Baseline data showed a vegetation weight difference so that with a ~ error of 0.20 and a two-tailed a. error of 0.05, a significant difference would be noted using 30 animals. A total of 64 animals was used. POPULATION STUDIED: Bacterial isolates from two patient groups (n=42 and n=68) on which in vitro assays were run. An animal model of endocarditis (n=64) was used to evaluate four selected isolates for vegetation size, Iog10 cfu/ g, and survival time. MAIN REsULTs: In a group of S epidennidis endocarditis animals evaluated for time of death, a significantly more rapid death time resulted in the group dosed with the high hexose-scoring organism (P<0.025). Vegetations and logw cfu produced by test high hexose isolates averaged larger but were not significantly different. CoNCLUSIONs: A significantly more rapid death rate occurs in untreated endocarditis using a high hexose isolate than with S epidennidis with low surface hexoses. Using bacterial vegetation and cfu as endpoints, however, experimental endocarditis using patient isolates of S epidennidis does not show the same strong correlation to bacterial surface hexoses as does streptococcal endocarditis.


S TAPHYLOCOCCUS EPIDERMIDIS. BECAUSE OF ITS HIGH
tendency to colonize polymeric implants, is a major factor in human prosthetic valve endocarditis (25 to 30% of cases). Occurrence of native valve endocarditis caused by these organisms in humans is infrequent; it is reported in 1 to 3% of cases (1). S epidermidis and other coagulase-negative staphylococci have been used by us (2) and by others (3,4) in the animal model of endocarditis. S epidermidis, in our tests, virtually always produced cardiac vegetations in the presence of an indwelling polymeric arterial catheter and 10 7 intravenous bacteria. Our earlier work has shown that the propensity towards endocarditis production in humans by the viridans streptococci -high native valve endocarditis producers -was related to high scores by isolates on a colorimetric test for cell-associated hexoses (5)(6)(7). It was of interest to determine if this assay for cell-adherent bacterial polysaccharides would also reflect behaviour of different isolates of S epidermidis in the animal endocarditis model. Coagulasenegative polysaccharides have been evaluated as adhesion or virulence-promoting substances both in vivo and in vitro (8)(9)(10)(11)(12)(13). Since the tryptophan assay for cell-adherent hexoses (t0 assay) is rapid and can be used for screening large numbers of isolates, we were interested in the possible relationship between virulence in the animal endocarditis model and quantity of bacterial surface hexose as determined by this assay. This study used S epidermidis isolates selected from two clinical populations because of their extreme values in the t0 assay.

MATERIALS AND METHODS
Organisms: S epidermidis samples used for these studies included a group of 42 from adult patients at Truman Medical Center West and 68 from infants at Children's Mercy Hospital, both in Kansas City, Missouri. Patient swabs were implanted on mannitol salt agar and were tested for catalase, hemolysis patterns, protein A, and 204 clumping factor. Coagulase-negative staphylococcal isolates were biochemically identified using either the API Staph-Ident system (Analytab Products, Plainview, New York) or the Microscan Gram-positive combo plate (American Scientific Products, McGaw Park, Illinois). Isolates were frozen in tryptic soy broth 15% glycerol (TSB) (Remel Labs, Lenexa, Kansas) 15% glycerol and stored at -70°C. Each was streaked on blood agar and grown 24 h before being assayed, cultured in either TSB or pooled normal rabbit serum, 76 mg/ elL blood glucose. Patie nt characteristics: The adult isolates comprised one unselected box from a group of 1676 specimens collected and identified by the Truman Medical Center microbiology laboratory. A large percentage of the adult specimens were obtained from an obstetrics-gynecology clinic, and were isolated from nasopharynx, skin, blood and cervix. The adult population would be considered to be a low-infection group; few had polymeric implants or catheters ( 14). The pediatric bacterial samples came from: noses of healthy infants (n=21); and blood cultures (n=47), 34% of which had central venous or umbilical catheters. Nine of the 4 7 were considered probable contaminants, 10 true sepsis, and 28 were from infants with one or more clinical signs of sepsis. Tryptophan (t(/)) assay: This colorimetric assay (15) was used to measure alcohol-precipitated cell-adherent polysaccharide, stripped by sonication from salinesuspended coagulase-negative staphylococci which had been grown in either rabbit serum or TSB for 48 h. The precipitated carbohydrate was treated with 66% aqueous sulphuric acid and 1% tryptophan, heated at 1 00°C for 20 mins, and absorbance was read at 500 nm against a standard of 500,000 molecular weight dextran. Values reported were actual peak absorbance Asoo. The two clinical bacterial groups (n=42 and n =68) were measured by t0 assay using four to 12 replications per isolate, and the high score isolate and low score isolate were selected from each group. These four isolates, used in endocarditis studies. were all S epider-midis, although many low scoring isolates were Staphylococcus hominis.
Animal model: The University Animal Care and Use Committee approved the methods (4, 16) used to produce endocarditis in 280 g S pragu e-Dawley rats (SASCO, Omaha, Nebraska). Th irty-two animals were used to test the adult staphylococcal is ola tes and 16 to test the pediatric isolates. Briefly, polyethylene sleeves (lntramedic PE-10, Clay Adams, Parsippany, New Jersey) over sterile 32 gauge hypodermic tubing were inserted into the left ventricular lumen via the right common carotid artery. The stainless steel insert was removed upon correct placement. and the polyethylene sleeve was sutured in place. About 20 h after cannula implantation, 0.5 rnL containing 1 to 4x10 7 organisms (16 h growth, saline suspend ed) were dosed via tail vein. Animals were killed three days after b acterial challenge, and 1.5 mL blood from the dorsal a orta were drawn and cultured to confirm bacteremia. Hearts with cannulas in place were removed, placed in tissue culture cellwells and necropsied in sequence under a 12.5x microsurgical scope. Vegetations adhering to valve leaflets and to cannulas across the valves were aseptically excised and added to pre-weighed sterile bags; wet weights were taken. Each vegetation was then macerated in the bag, serial dilutions were made and plated to blood agar. Bacterial vegetation cou nts were made at 24 h, and Iog10 bacteria per gram of vegetation calculated for each animal. A greater than 90% average infection rate was obtained.
In survival tests, rodents with implanted cannulas were dosed with bacteria as above. Animals were checked twice daily and the time of death recorded. Blood samples were cultured to confirm bacteremia as above. In vitro bacterial analyses: A qu antitative slime assay (1 7) and a bacterial hydrophobicity assay were also performed on the pediatric bacterial p opulation (n=68). The slime assay evaluated material remaining on a glass tube after 24 h culture in TSB with 10% glucose. Decanted tubes were fixed, stained with 0.1% safranin, and heated (85°C for I h in 0 .2 M sodium hydroxide). Colour was read at 530 run. Hydrophobicity analysis (18) evaluated the absorbance of a 10 8 / mL bacterial suspension in phosphate b u ffered saline before and after vortexing 2 mins w ith p-xylene (Merck, Darmstadt) . Hydrophobicity was expressed as absorbance (600 nm) of extracted versus initial suspensions times 100. Neutrophil phagocytosis/bactericidal assays: The adult isolates used in endocarditis s u rvival studies were analyzed further for mechanisms of bacterial virulence. Human blood neutrophils (polymorphonuclear leukocytes [PMN)) were isolated as reported (19). Overnight cultures of the two test S epidennidis isolates were washed and suspended at 7.
.<::: 0.5% gelatin with 10%i autologous serum. The organisms were cultured in duplicate tubes with and without 1.5x 10 6 PMN with shaking at 3 7°C. Ten microlitres from each tube were sampled at 0 , 30, and 60 mins, placed in 10 mL iced sterile water, and aliquots were plated on b lood agar and counted after a 24 h incubation. Percentage kill was calculated as follows: (No-PMN tube cfu/ml-PMN tube cfu / ml) X 100 zero time No-PMN tube cfu / ml An additional phagocytosis/bactericidal assay was performed on these organisms using an acridine orange crystal violet microassay as described previously (20). In this assay, S epidennidis isolates used in endocarditis survival tests were opsonized with 10% rabbit serum in HBSS, then incubated with rabbit PMN on glass slides 45 mins before staining. The living bacteria fluoresced green and the dead fluoresced red. Crystal violet quenched the fluorescence of extracellular /membrane adherent organisms so these were not visualized. Counts were made of total and nonviable S epidennidis in 100 PMN. Analysis: Differences in mean vegetation weights and mean vegetation bacterial concentrations were evaluated by Student's t test for unpaired data. Probability of survival (untreated endocarditis) was plotted with Kaplan Meier curves and with linear regression analysis (21). Significance of in vitro tests was determined using the Asystant statistical software package.

RESULTS
Coagulase-negative staphylococcal blood culture isolates from two groups of patients, 42 adult and 68 pediatric, were quantified by the t0 assay. The data  spread was wider using serum-grown isolates compared with TSB culture, and the endocarditis test strains were selected from serum-cultu red assay runs. Figure  1 illustrates the t0 assay data compared with slime production by the same isolate from pediatric bacterial samples. Using averaged data, isolates from each population with the highest and lowest t0 assay values were selected for further testing. Table 1 gives data on in vitro assays performed on the fourS epidermidis isolates. Tested by the qualitative slime screen, three of the four S epidermidis strains used to produce endocarditis were slime-negative, although the groups tested 60% s lime-positive. By the quantitative slime assay, there was no pattern; one high t0 isolate was much below the median, whereas others were average. Table 2, showing data separated by the source of the infective bacterial strains, presents the mean vegetation weights and bacterial counts of cfu/ g of vegetation, taken from valve leaflets and from valve-contacting catheters in the rodent model of endocarditis. There was a marginal difference in vegetation weight and S epidermidis cfu / g between the two isolates from the adult patient sources (P<0.1). with the higher t0 assay S epidermidis isolate producing a larger vegetation and greater growth.

206
The right-hand side of Table 2 details animal endocarditis models using t0 high and low isolates from the pediatric popu lation. Again t h e higher t0 scoring S epidermidis produ ced a somewhat larger cfu/ g than the isolate scoring at the bottom of the t0 series. Figure 2 shows mortality (time of survival) in untreated S epidermidis endocarditis in a group of 16 animals, eight of which received the high t0 adult isolate and eight the low t0 isolate at identical infective doses. Using the S epidermidis isolates from the adult population group, endocarditis was established using 2.2x10 7 intravenous organisms. At the end of seven days, one death had occu rred in the low t0 assay bacterial group, and eight of eight had died in the h igh t0 assay bacteria group. This difference in mortality, even with the small populations, is significant ([P<0.025) x 2 test). Table 3 gives results of phagocytosis and neutrophil bactericidal assays performed on the two isolates showing the significant difference in endocarditis mortality. Ability to escape from phagocytosis in vivo is a classical bacterial virulence factor that has been related to surface glycocalyx. Using a bacteria to PMN ratio of 5:1, there was a 4. 7% kill of the high virulence isolate and a 6% kill of the low virulence isolate. These data were not significant at 60 mins in vitro, but offered enough  A simple capsule evaluation (carbon plus safranin staining of a thin smear) was performed on several saline-washed isolates from the pediatric bacterial group, including the two t0 assay 'extreme value' strains. There was a rather h igh agreement between this test and the t0 a s say (80%), with the high t0 isolates receiving a two or greater capsule score, while the lowest t0 isolates were all negative.
The four endocarditis isolates showed hydrophobic differences. The high t0 assay isolates were strongly hydrophobic, whereas the low t0 isolates scored around the median of the large test series.

DISCUSSION
These results demonstrate that the cell-adherent hexose sugars of coagulase-negative staphylococci can be quantified by the assay which described and correlated with virulence (endocarditis production) in streptococci. The use of this t0 assay to select staphylococcal isolates that produce a m ore virulent endocarditis, however, is not so clear cut. The isolates selected represented well-tested group extremes, all typed as S epidennidis. In no series was the difference in vegetation weight or log10/ g bacterial growth statistically different between groups, although bacterial growth and vegetation weight were consistently greater in the high t0 S epidennidis-dosed groups. It was of interest in necropsy that the quantity of actual valve vegetation (as opposed to the catheter-associated vegetation with valve contact) was strikingly less than the valve-adherent growth observed in streptococcal endocarditis rats which were being necropsied at similar times for unrelated studies. Local effect of the catheter C AN J INFECT DIS VOL 4 No 4 JULY/ A UGUST 1993 in in vivo experiments has been reported ( 11,15) but the definite bacterial differences in valve versus catheter vegetation noted with our necropsy procedure again suggests how very important the catheter is to vegetation size.
In a test of endocarditis time of survival ( Figure 2), a significantly more rapid death rate occurred (eight of eight versus one of eight at one week) when the adult group t0 test high isolate was compared with the t0 low test isolate. Although in vitro growth curves with quantitative plating were not performed for all isolates, early tests showed that bacterial growth in vitro was faster in some high t0 scoring coagulase-negative staphylococci. It is possible that, given the same intravenous challenge dose of bacteria, more rapid growth also occurred in vivo with the high t0 isolate, producing more deaths. Additional experiments were performed to evaluate phagocytosis resistance of those isolates, using two types of assay. In neither in vitro test was PMN internalization and killing significantly different between strains, but PMN killing rates of 4 . 7% versus 6% in 1 h may indicate a difference that could, in vivo , show up as a host survival difference. Phenotypic variants of bacteria producing endocardial infections have also been noted (3) and the resultant bacteria exhibit differences in virulence.
To investigate further the differences in these endocarditis strains of S epidennidis, hydrophobicity and slime measurements were made. Three of the four S epidennidis test isolates were average or below average slime producers. The tests on washed cells of a few isolates, staining for capsule with carbon and safranin, showed that capsule score was higher more often than was slime score in the high t0 assay isolates (Figure 1) .
The time-of-survival endocarditis tests, using the two adult population S epidennidis isolates, show significant differences in the death rates of the groups. The isolates ( Table 2) differ widely in both hydrophobicity and in the t0 assay for cell-associated polysaccharides. Wadstrom (22) has shown that staphylococci with high cell surface hydrophobicity bind to plastic polymers in higher numbers than do cells with low hydrophobicity. Since the animal endocarditis model uses a polyethyl-ene catheter across the aortic valves, a higher bacterial adherence to the catheter in vivo could contribute to the higher death rate with the hydrophobic bacterial isolate. Similarly, Hazen (23) in ex vivo experiments showed that hydrophobic cells were more likely than hydrophilic cells to seed and colonize various organs of an infected host.
Bacterial polysaccharides themselves have occasionally been considered to be more or less hydrophobic