Background. The ability of Staphylococcus aureus to adhere to endothelial cells is a major prerequisite for the tissue-invasive stage of bacterial infection.
Methods. To develop a model for the study of endothelial attachment and detachment kinetics of S. aureus within the host's microvasculature in vivo, we labeled inactivated staphylococci with fluorescein isothiocyanate and investigated their interaction with the vascular endothelium of arterioles, capillaries, and venules in the dorsal skin-fold chamber of untreated and tumor necrosis factor (TNF)-α-treated hamsters by use of intravital fluorescence microscopy.
Results. During the first 20 min after injection, >99% of the bacteria were removed from the microvascular bloodstream. In parallel, single bacteria and bacterial clusters adhered to the endothelial lining of postcapillary venules and to nutritive capillaries. Bacterial adherence to the endothelium of arterioles was only rarely observed. TNF-α treatment significantly accelerated bacterial clearance and resulted in a significant increase of venular, but not arteriolar and capillary, bacterial adherence, indicating the venular endothelium to be the target structure for bacterial recruitment.
Conclusion. The insights into host-pathogen interaction gained with this new in vivo model offer highly promising novel aspects of the understanding of infections caused by S. aureus.
Staphylococcus aureus, a highly virulent gram-positivemicroorganism, is a predominantly extracellular pathogen that causes a variety of community-acquired and nosocomial infections that range from harmless superficial skin infections to serious invasive diseases, such as endocarditis , osteomyelitis , and pyarthritis . The incidence of select S. aureus infections—such as endovascular diseases or device-associated infections—is on the rise, paralleled by an increasing resistance to practically all antistaphylococcal compounds due to the acquisition of multiple resistance traits . This trend has spurred interest in the biology and virulence properties of this microorganism and its interaction with the host organism, to aid in the development of new therapeutic strategies to prevent and combat disease.
Besides the secretion of extracellular toxins—for example, toxic shock syndrome toxin 1 and staphylococcal enterotoxins —the pathogenicity of S. aureus depends on the expression of cell wall-associated proteins, which promote adherence to platelets , endothelial cells , extracellular matrix components , and soluble plasma proteins . In fact, bacterial adherence is likely to play a central role in allowing the microorganism to transition from the bacteremic stage to the tissue-invasive stage of disease, because this is a major prerequisite for bacterial transmigration through the endothelial barrier . Accordingly, a number of studies have focused on identifying different surface receptors of S. aureus, which bind in a specific manner to host ligands. These adhesins include the collagen-binding protein Cna , the fibrinogen-binding proteins clumping factor A and B [12, 13], fibronectin-binding protein (FnBP) A and FnBPB [14, 15], and protein A . In addition, a family of secreted adhesive proteins that recognize various extracellular matrix molecules (Eap, Emp, Efb, and coagulase) [9, 17, 18] has been identified. As has been shown elsewhere [19, 20], the control of their expression is a complex dynamic process that is controlled by a regulatory system that consists of the global regulators SigB, SarA, and agr/RNA III.
Although it is widely recognized that the ability of S. aureus to adhere to endothelial cells is a critical step in the development of bacterial infection, little is known about the attachment kinetics and mechanisms of S. aureus within the microvascular system—that is, arterioles, capillaries, and venules—under in vivo conditions. This is because current experimental models that evaluate pathogenic mechanisms of staphylococcal diseases— for example, endocarditis models [21, 22]—do not allow the time-dependent determination of the number of resident bacteria at the vessel wall. Therefore, the aim of the present study was to develop a new model for the in vivo imaging of S. aureus in the host's microcirculation. For this purpose, we labeled inactivated staphylococci with fluorescein isothiocyanate (FITC) and investigated their interaction with the vascular endothelium of arterioles, capillaries, and venules in the dorsal skin-fold chamber of Syrian golden hamsters, by use of intravital fluorescence microscopy.
Materials and Methods
Animals. The experiments were conducted in accordance with German legislation for the protection of animals and with the US National Institutes of Health's Guide for the Care and Use of Laboratory Animals and were approved by the local governmental animal care committee. Eight- to 10-week-old male Syrian golden hamsters with a body weight of 60–80 g were used for the study. The animals were housed 1 hamster/cage and had free access to tap water and standard pellet food (Altromin) throughout the experiment.
Preparation of the dorsal skin-fold chamber. The chamber contains 1 layer of striated muscle and skin and allows for the intravital microscopic observation of the microcirculation in awake animals over a prolonged period of time (figure 1). In previous studies , we have shown that the preparation of the skin-fold chamber induces some vasoconstriction and a leukocytic inflammatory response during the first 24 h after surgery. Therefore, we guaranteed a 48-h period of recovery after the surgical preparation before the start of the experiments. After this period, microcirculation has normalized and shows no significant changes in arteriolar perfusion, functional capillary density, venular leukocyte-endothelial cell interaction, and macromolecular leakage over a period of several days [23–25]. This chamber preparation can be studied for a 3-week period and has been successfully used in multiple studies of vascular biology, ischemia and reperfusion, inflammation, transplantation, and tumor growth pathology [26, 27].
The chamber technique and its implantation procedure have been described in detail elsewhere . Briefly, under sodium pentobarbital anesthesia (50 mg/kg body weight, administered intraperitoneally), 2 symmetrical titanium frames were implanted on the extended dorsal skin fold of the hamsters, such that they sandwiched the double layer of skin. One layer of skin was then completely removed in a circular area ~15 mmin diameter, and the remaining layers (consisting of striated skin muscle and subcutaneous tissue) were covered with a removable cover slip incorporated into 1 of the titanium frames. In addition, a permanent polyethylene catheter (0.28-mm internal diameter; PE10) was passed from the dorsal to the ventral side of the neck and inserted into the right carotid artery for the application of bacteria.
Preparation of bacteria. To study the in vivo interaction of staphylococci with the vascular endothelium, bacteria were stained with FITC (20 mg/mL ethanol stock solution; Sigma). For this purpose, S. aureus strain Cowan 1 was cultured in 5 mL of Muller-Hinton broth medium (Becton Dickinson) under constant rotation at 37°C overnight. Then, 1 mL of the culture was washed in 1 mL of sterile 0.9% NaCl, and the pellet was resuspended, under constant rotation for 2 h, in 250 µL of sterile NaHCO3 (0.1 mol/L, pH 9.0) that contained FITC (100 mg/µL). FITC-labeled bacteria were washed 3 times in 1 mL of sterile PBS (Biochrom). To avoid uncontrolled effects caused by the secretion of bacterial pyrogenic superantigens or staphylococcal a-toxin on the activation status or integrity of the endothelium, bacteria were inactivated in 500 µL of PBS/1% formaldehyde under constant rotation for 2 h at room temperature. In previous studies [7, 29], we demonstrated that inactivation does not interfere with the interaction of S. aureus Cowan 1 with eukaryotic cells. Inactivated, FITC-labeled staphylococci were again washed 3 times in PBS, and the final pellet was resuspended in 1 mL of sterile PBS/1% bovine serum albumin. The sample could be stored for a period of 2–3 days at 4°C before injection.
Intravital fluorescence microscopy. For in vivo microscopic observation, awake hamsters were immobilized in a Plexiglas tube, and the dorsal skin-fold preparation was attached to the microscopic stage. After the intra-arterial injection of ~108 inactivated, FITC-labeled bacteria, intravital fluorescence microscopy was performed by use of a modified Leitz Orthoplan microscope with a 100-W mercury lamp attached to a Ploemo-Pak illuminator with blue, green, and UV filter blocks (Leitz) for epi-illumination. The microscopic images were recorded by a charge-coupled device video camera (CF8/1 FMC; Kappa) and transferred to a video system for off-line evaluation. A 10-fold long-distance objective (Leitz), resulting in a total magnification of ×216 on a 14-inch monitor screen (PVM 1444; Sony), was used to select 10 optical fields of interest per chamber that contained arterioles, capillaries, and venules.
Microcirculatory analysis. Quantitative off-line analysis of the videotapes was performed by use of the computer-assisted image analysis system CapImage (version 6.02; Zeintl) and included the determination of microvascular bacterial clearance and adherence of microorganisms to the endothelium of arterioles, capillaries, and venules. Single bacteria and bacterial clusters were analyzed separately, because we assumed differential attachment kinetics of single bacteria, compared with clusters, given the small diameters of the microvessels, particularly of nutrient capillaries (see below). For that purpose, a single fluorescence signal corresponding to 1 pixel on the monitor screen was defined to be a single bacterium, whereas a signal of size >1 pixel was interpreted as a bacterial cluster. Microvascular bacterial clearance was determined by repeated scanning of an identical venular segment during the experiment and counting the number of all nonadherent (free flowing) bacteria within the vessel segment (given as the number of bacteria per cubic millimeter, under the assumption of a cylindrical vessel geometry). In accordance with studies that have investigated the interaction of leukocytes and endothelial cells [30, 31] and to studies in which bacterial-flow cell systems have been used with adhesive substrates to demonstrate the definitive attachment of microorganisms transported to the surface , we defined adherent staphylococci as bacteria that did not move or detach from the endothelial lining of arterioles or venules during an observation period of 20 s (given as the number of adherent bacteria per square millimeter of endothelial surface). In each hamster, analysis was performed in 8–10 postcapillary or collecting venules (diameter, ~30–90 µm) and 3–4 precapillary arterioles (diameter, ~30–70 mm). Because capillaries were too small in diameter to delineate them from the surrounding tissue without contrast enhancement (by staining plasma with FITC), bacterial adherence in capillaries was determined indirectly by counting the number of arrested single bacteria and bacterial clusters that were seen outside of venules and arterioles in 10 optical fields of interest (area, 0.55 mm2) per dorsal skinfold chamber (given as the number of bacteria per square millimeter of tissue area). To allow a comparison of bacteria–endothelial cell interaction between arterioles, capillaries, and venules, capillary bacterial adherence was also calculated as the number of adherent bacteria per square millimeter of endothelial surface. Therefore, capillary lengths and diameters per tissue area were measured in some additional hamsters that received FITC-dextran (molecular weight, 150,000 Da) for contrast enhancement. This allowed us to clearly visualize the capillary structures for the quantitative analysis of their geometry (figure 1). An analysis of capillary geometry by FITC-dextran could not be performed in the identical hamsters in which bacterial adherence was studied, because the fluorescent staining of intravascular plasma by FITC-dextran would have made impossible the analysis of the adherence kinetics of the FITClabeled bacteria.
To study whether differences in bacterial adherence are due to changes in local microhemodynamics caused by tumor necrosis factor (TNF)-α, we additionally analyzed the diameters, center-line velocity, volumetric blood flow, and wall-shear rate in those venules in which bacterial clearance had been determined. Diameters (d) were measured, in microns, perpendicularly to the vessel path. The center-line velocity (v) was analyzed by the computer-assisted image analysis system by use of the line-shift method. Volumetric blood flow was calculated as Q = π × (d/2)2 × ν/1.6 (pL/s), where 1.6 represents the Baker-Wayland factor  to correct for the parabolic velocity profile in microvessels with diameters >20 µm.
Experimental protocol. Seven Syrian golden hamsters were prepared by implantation of a dorsal skin-fold chamber and of a permanent catheter into the right carotid artery. The hamsters were allowed to recover from anesthesia and surgery for 48 h before the experiment. To investigate bacterial clearance and attachment kinetics of staphylococci within the microvascular system of the dorsal skin-fold chamber, intravital fluorescence microscopy was performed 1, 3, 5, 10, 20, 30, and 60 min after the intra-arterial application of ~108 FITC-labeled bacteria (untreated group). Vessel diameters, center-line velocity, volumetric blood flow, and wall-shear rate were additionally determined at the 1-min time point.
After 24 h of recovery, when no more bacteria were detectable within microcirculation, the tissue of the dorsal skin-fold chamber was exposed to TNF-α (topical application of 2000 U dissolved in 100 µL PBS for 30 min; Boehringer Mannheim). Topical application was chosen to deliver the cytokine locally without systemic alterations, in particular to avoid any deterioration of macrohemodynamic conditions. The hamsters were then challenged again with ~108 bacteria (TNF-α-treated group; n = 7). Intravital microscopy was performed at times similar to those of the untreated group.
Statistics. Differences between treatment (before and after the application of TNF-α) were tested separately at each time point by paired Student's t test. To test for time effects separately for each experimental intervention, multivariate analysis of variance for repeated measures was applied. This was followed by paired Student's t test, including a correction of the a-error according to Bonferroni probabilities for repeated measurements (SigmaStat version 2.03; Jandel). All values are expressed as mean ± SE. Statistical significance was accepted for a value of P < .05.
Using the technique of intravital fluorescence microscopy, we detected FITC-labeled single bacteria and bacterial clusters in the bloodstream of the microvessels of the dorsal skin-fold chamber of hamsters shortly after an intra-arterial injection of S. aureus Cowan 1. In untreated hamsters, the analysis of microvascular clearance exhibited the highest concentrations of free-flowing bacteria 1 min after injection. This was followed by a progressive decrease during the first 10 min of the experiment (figure 2). After 20 min, >99% of the bacteria were found to have been removed from the microvascular bloodstream. This was associated with the adherence of single bacteria and bacterial clusters to the endothelial lining of the microvasculature. Treatment with TNF-α accelerated microvascular clearance of bacteria, as indicated by a significantly lower concentration of bacteria still moving intravascularly 1 min (3765 ± 1292 vs. 7621 ± 1802 bacteria/mm3, respectively, for treated and untreated hamsters; P < .05) and 3 min (1086±327 vs. 2620±529 bacteria/mm3, respectively, for treated and untreated hamsters; P < .05) after the injection (figure 2).
In untreated hamsters, the adherence of single bacteria in venules peaked 5 min after injection (11.5±2.4 bacteria/mm2 endothelial surface). This was followed by a progressive decline during the course of the experiment, resulting in 1.7±0.5 bacteria/mm2 of endothelial surface at 60 min (figure 3A). Treatment with TNF-α resulted in a significant enhancement of the initial bacterial adherence to the venular wall (figure 4), as indicated by a 6-fold increase (64.5±13.4 bacteria/mm2 of endothelial surface at 5 min; P < .05) over that of untreated control hamsters. Although the number of adherent bacteria also progressively declined during the subsequent observation period (figures 3B and 5), the frequency of adherence remained significantly higher (31.8±8.2 bacteria/mm2 of endothelial surface at 60 min), compared with that of untreated control hamsters (figure 3A and 3B).
In contrast, single bacteria adhering to the endothelial lining of arterioles were only rarely observed throughout the experiment. This lack of adherence was found in both untreated and TNF-α-treated chambers (figures 3 and 5).
An analysis of the adherence of single bacteria in capillaries was performed in individual tissue areas of 0.55 mm2, which encompassed nutritive capillaries with a mean ± SE diameter of 5.5±0.1 µm and a total length of 8.5±0.5 mm. These data indicate a capillary length of 15.5 mm/mm2 of tissue, and, accordingly, a capillary endothelial surface of 0.27 mm2/mm2 of tissue. In untreated hamsters, the capillary adherence of single bacteria was most pronounced 5 min after injection—3.9±0.7 bacteria/mm2 of tissue (table 1), which reflects ~15 bacteria/mm2 of capillary endothelial surface. As the experiment progressed, the adherence of single bacteria in these capillaries decreased only slightly, to 2.2±0.4 adherent bacteria/mm2 of tissue at 60 min (table 1), which indicates ~8 bacteria/mm2 of capillary endothelial surface. Thus, 60 min after injection, the adherence of single bacteria in capillaries of nonstimulated tissue was 5-fold more pronounced than that in postcapillary venules (1.7±0.5/mm2 of endothelial surface; figure 3). Treatment with TNF-α resulted in only a minor (<2-fold) enhancement of the capillary adherence of single bacteria, compared with that in untreated control hamsters, and this did not prove to be significant over the course of the entire 60-min observation period (table 1).
The number of bacterial clusters that were adherent in the postcapillary venules of untreated hamsters ranged between 6.8±2.3/mm2 of endothelial surface at 5 min and 4.5±2.2/mm2 of endothelial surface at 60 min and did not significantly differ over time (figure 6A). After treatment with TNF-α, the analysis of the venular adherence of bacterial clusters yielded significantly (P < .05) higher numbers, compared with those of untreated control hamsters, throughout the entire observation period (figure 6B). However, cluster adherence decreased from 60.4±14.5/mm2 of endothelial surface 5 min after bacteria injection to 24.0±7.2/mm2 of endothelial surface 60 min after bacteria injection (figure 6B).
Comparable to the adherence of single bacteria, clusters that adhered to the arteriolar endothelial lining were only rarely observed during the entire observation period. This included both untreated control and TNF-α-treated chambers (figure 6).
In untreated hamsters, the analysis of capillary adherence of bacterial clusters revealed values of 5.7±1.2/mm2 of tissue, and this did not change during the 60-min observation period (table 1). This indicates ~20 clusters/mm2 of capillary endothelial surface, which represents a 3-fold higher value than that observed in postcapillary venules (6.8±2.3 clusters/mm2 of endothelial surface; figure 6). Most interestingly, during the entire 60-min period after the injection of bacteria, local TNF-α exposure was not effective in significantly increasing the adherence of bacterial clusters in these capillary network structures (table 1).
The analysis of microhemodynamics revealed an increased center-line velocity, volumetric blood flow, and wall-shear rate in the TNF-α-treated hamsters, compared with untreated hamsters. These differences, however, were not statistically significant (table 2).
Intravital fluorescence microscopy is a novel tool for the close examination of the interactions of pathogenic microorganisms with components of the vessel wall. In fact, this technique permits a precise spatial and kinetic resolution of the endothelial interaction with bacteria. In our study, the evaluation of microbial attachment to the vessel wall in vivo was realized with inactivated S. aureus Cowan 1 for a proof of principle and to provide a technical platform for future studies examining staphylococcal cell-wall-molecule interactions (MSCRAMMs and protein A) with the host endothelium.
The major findings of the present study were that 1×108 intra-arterially applied S. aureus rapidly adhere as single bacteria and as bacterial clusters to the endothelial lining of nutritive capillaries and postcapillary venules but not of precapillary arterioles. Strikingly, a major fraction of the adherent single bacteria detached during the 60-min observation period, whereas bacterial clusters remained permanently adherent, particularly within the capillary network. Local treatment with TNF-α significantly increased bacterial recruitment in venules but not in arterioles and capillaries. Because there was no evidence of a decrease in volumetric blood flow and wall-shear rate after the application of TNF-α, this increased bacterial adherence was not caused by deteriorated microcirculation but, rather, by a specific inflammatory activation of venular endothelial cells that was probably caused by the overexpression of specific surface ligands.
A limited number of in vitro studies have addressed the interaction of S. aureus with host ligands under physiological flow conditions [32, 34–37]. By use of bacteria displaying culture-dependent variations in adhesin densities  or mutants defective in defined adhesins , it has been shown that S. aureus cell-wall components contribute to reversible and irreversible attachment even under conditions of elevated shear rates. However, although these studies analyzed the adherence of S. aureus to plasma, collagen, and fibrinogen, an in vitro analysis of the interaction between S. aureus and endothelial cells indicated that the FnBP-mediated adherence to resting endothelial cells observed under static conditions is completely abolished under flow conditions, including shear rates in the range of 10–200 cells/s . These latter studies, however, were performed in flow chambers seeded with bovine aortic endothelial cells, which may markedly differ from arteriolar, capillary, and venular endothelium in their expression characteristics of adherence molecules.
The present study allowed us (1) to study the interaction of S. aureus and endothelial cells within the microcirculation, which probably represents the primary target for intravascular bacterial recruitment, compared with the aorta and large arteries; and (2) to distinguish among arteriolar, capillary, and venular adherence events. In the hamster skin-fold preparation used in the present study, shear rates in precapillary arterioles have to be considered to be >400 cells/s, whereas capillaries and postcapillary venules present with significantly lower shear rates of ~170 and ~125 cells/s, respectively . Thus, our results show that, within the unstimulated microcirculation in vivo, S. aureus adheres to capillary and venular endothelium at low shear rates, whereas shear rates >400 cells/s in arterioles prevent bacterial adherence. Although these results may indicate a strong dependency of endothelial adherence on the wallshear rate, the endothelial specificity of the different vascular compartments may also contribute to bacterial recruitment. This is underlined by the aforementioned results obtained from in vitro analyses, which demonstrated that S. aureus adherence to aortic endothelial cells is completely abolished at a shear rate of 10 cells/s .
One particular advantage of our new model is that it makes possible the study of the consequences of different treatments of the animal—for example, with proinflammatory substances—in a crossover design. Herein, we used TNF-α, because of its extensive characterization as a major inducer of the inflammatory response. TNF-α induces the activation of leukocytes and endothelial cells and increases the adherence of leukocytes and platelets to the endothelium, as well as the level of fibrinogen in plasma . It may also augment the binding of fibronectin by endothelial cells . More recently, it has been shown that the specific increase of fibrinogen release and cell attachment was principally due to the induction of the procoagulant and permeability effects orchestrated by TNF-α [41, 42]. The increased permeability induced by TNF-α is due to the surface expression of tissue factors on endothelial cells. TNF-α disrupts the vasculature and the anchoring endothelial tissue, exposing the basement membrane; it has similar effects on arterioles and venules and alters the permeation properties of the glycocalyx on the luminal surface independently of leukocyte adherence . S. aureus is known to interact with major basementmembrane components—fibrinogen, fibronectin [6, 43, 44], collagen, and laminin. TNF-α may induce bacterial attachment by exposing these ligands. Accordingly, in our study, treatment with TNF-α yielded a significant increase in the adherence of single bacteria and bacterial clusters. This effect, however, was observed only in venules and not in arterioles and capillaries. Because TNF-α-mediated endothelial damage to the endothelial lining has to be considered to be similar in arterioles, capillaries, and venules , and because wall-shear rates in capillaries are comparable to those in venules, as reported in a model used elsewhere , the selective adherence of S. aureus in venules after treatment with TNF-α indicates that the postcapillary venular endothelium is the target structure for bacterial recruitment in TNF-α-mediated inflammation. Our study design did not allow us to determine the role of platelets in the observed attachment events. Because platelets may provide adhesive surfaces to both staphylococci and the activated endothelium , and because they may release ligands that are adhesive for staphylococci [8, 46, 47], they can support a bridge between the vessel wall and the bacteria. However, at this stage, additional experiments are required to address this specific role.
In conclusion, we successfully established a reproducible in vivo model to study the interaction of S. aureus with the vascular endothelium. To our knowledge, ours is the first successful application of this technique for studying the pathogenic events of microorganisms at the endothelial site, and it could be demonstrated that our method allows the determination of realtime bacterial interaction kinetics with the endothelium. In vivo treatment with TNF-α enhanced bacterial attachment and reduced bacterial detachment and, thus, could be used as a model intervention to augment bacterial adherence to the microvascular endothelium of venules. Yet, because systemic treatment with TNF-α elicits a plethora of effects on various endovascular and blood cells, it is difficult to attribute this observation to a defined mechanism of interaction. Thus, the present results might prompt additional approaches that use, for example, defined staphylococcal deletion mutants to compare in closer detail the interaction kinetics of bacteria with vessel walls. The insight into the host-pathogen interaction gained with this powerful new in vivo technique offers highly promising novel aspects that could enhance the understanding of the mechanisms of endovascular diseases and of the development of novel strategies to combat endovascular infections caused by S. aureus.