INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sepsis occurs in approximately 1% of all hospital inpatients and accounts for between 20% and 30% of intensive care unit (ICU) admissions (1, 2). Despite modern techniques of resuscitation and organ support, septic shock continues to have a mortality rate in the region of 50%. Septic shock develops during an episode of infection and is characterized by the appearance of hypotension that is resistant to volume replacement alone, and by organ hypoperfusion (3).

Microorganisms have a clear initiating role in septic shock syndrome, but the circulatory disturbances that follow are a result of the release of various toxins and inflammatory mediators (1, 4). Some of these substances are direct bacterial products, but many are released by the host in response to infection. They include both pro- and antiinflammatory cytokines, complement components, platelet-activating factor, prostaglandins, and nitric oxide (NO). These inflammatory mediators cause complex alterations in microvascular flow (5). Vascular dilation and constriction, the presence of arteriovenous shunts, and changes in tissue blood flow as a result of vascular occlusion have all been reported in humans and animals during sepsis (6).

In the past decade, a unifying hypothesis has been developed to explain the vascular changes in septic shock (7) on the basis of the effect of inflammatory mediators on the vascular endothelium (4, 8, 9). The endothelium is involved in the control of vascular tone, vascular permeability, and coagulation (10, 11), and a number of changes in it follow exposure of endothelial cells (EC) and isolated blood vessels to relevant proinflammatory mediators (12). It has therefore been proposed that widespread endothelial damage occurs during human sepsis and leads to organ dysfunction and failure. However, there is little direct evidence for endothelial damage or dysfunction in humans during septic shock.

The presence of increased numbers of circulating EC has previously been reported after various forms of severe vascular disturbance, including sickle cell crisis and myocardial infarction (13). We hypothesized that the severe vascular damage associated with septic shock would also result in endothelial shedding. We therefore measured the number of circulating EC in critically ill patients with sepsis and septic shock.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Our study had full ethical approval from the Joint Ethics Committee of Newcastle Health Authority and the University of Newcastle-upon-Tyne. Patients meeting the American College of Chest Physicians/Society for Critical Care Medicine criteria for sepsis and septic shock (21) were identified in the ICUs of the Royal Victoria Infirmary and Newcastle General Hospital, Newcastle-upon-Tyne. The criteria for sepsis included the development of two or more of the following conditions during an episode of infection: a temperature of > 38° C or < 36° C; a heart rate of > 90 beats/min; a respiratory rate of > 20 breaths/min or an arterial carbon dioxide tension (Pa_CO2) < 4.25 kPa; a white blood cell count > 12 × 10⁹/L; or < 4 × 10⁹/L or the presence of more than 10% immature neutrophils (21). The criteria for septic shock was sepsis-induced hypotension (systolic blood pressure < 90 mm Hg or a reduction of systolic blood pressure of 40 mm Hg or more from baseline) despite adequate fluid resuscitation, along with the presence of perfusion abnormalities (21). ICU controls were recruited from among patients who did not meet the sepsis, septic shock, or systemic inflammatory response criteria of our study (21) but who were sufficiently ill as to require mechanical ventilation. Healthy controls were recruited from among staff members of the University of Newcastle-upon-Tyne and Newcastle-upon-Tyne Hospital. Clinical information on each patient was recorded, including diagnosis, the Acute Physiology and Chronic Health Evaluation (APACHE) II data set (22) on ICU admission, and maximum Sepsis-related Organ Failure Assessment (SOFA) score (23). The following groups were recruited: 15 patients with septic shock, eight patients with sepsis, but without septic shock, nine patients without sepsis from ICUs, and 11 healthy volunteers.

Blood Sampling

In pilot studies it was found that the trauma associated with venipuncture increased the yield of EC in a blood sample. Blood samples were therefore taken from freshly placed, flushed venous cannulae. In healthy controls, a venous cannula (18-gauge) was inserted in a hand or arm vein. In all groups the initial 5-ml blood sample drawn was discarded to minimize EC contamination from the puncture wound of the vascular wall. Following this, 18 ml of venous blood was drawn into tubes containing sodium citrate (0.105 M) as anticoagulant. Samples were collected from patients within 24 h of the initial diagnosis of sepsis or septic shock. Blood samples were analyzed without access to clinical information.

EC Isolation by Isopycnic Centrifugation

Isopycnic centrifugation separates cells on the basis of differences in specific gravity. We have previously described our method of isolating EC with this principle (24). A Percoll (Sigma, Ltd., Poole, UK) suspension was prepared as follows: a solution of 1.54 M sodium chloride (NaCl) with 1 M sodium dihydrogen phosphate (NaH₂PO₄ · H₂O) in deionized water was made up and filter-sterilized with 0.22-µm Flowpore. Percoll of specific gravity 1.130 g/ml was used to prepare a Percoll stock suspension by adding 7 ml of theNaCl/NaH₂PO₄ · H₂O solution to 93 ml of sterile Percoll solution. A volume of 100 ml of Percoll suspension of specific gravity 1.060 g/ml was made up by using 46 ml of Percoll stock suspension, 10 ml of 5 mg/ml human albumin fraction V in water (Sigma), 34 ml of Dulbecco's modified phosphate-buffered saline (PBS), and 10 ml of 3.8% (wt/vol, in water) sodium citrate (prepared from Na₃C₆H₅O₇ · 2 H₂O. Preparation of the Percoll solution was done under aseptic conditions in a Class II cabinet, and all solutions (except Percoll) were filtered through 0.22-µm Flowpore. The specific gravity of the final Percoll suspension was determined with an electronic laboratory weighing scale (Sartorius), with Percoll of specific gravity 1.130 g/ml as a standard for calibration.

Isolation of Circulating EC

Blood was layered onto an equal volume of Percoll suspension (specific gravity = 1.060 g/ml) and centrifuged at 1,000 × g for 10 min at room temperature. The top cell layer was collected and further centrifuged at 400 × g for 10 minutes to pellet the cells. The cells were then pooled into one tube and resuspended in medium at a final volume of 2 ml. Human umbilical vein endothelial cells (HUVEC), mononuclear cells, and platelets were used as controls. HUVEC were isolated from fresh cords by using collagenase, and were grown in culture. Cells were detached at confluence by using trypsin. Mononuclear cells were prepared by layering blood onto Lymphoprep (Cardinal Laboratories), a separation gradient specifically designed for isolation of lymphocytes from blood, and were centrifuged at 400 × g for 25 min. The top cell layer was transferred into fresh tubes and further centrifuged at 400 × g for 8 min to pellet the cells, which were then resuspended in medium. For preparation of platelet-rich suspension, blood was collected into tubes containing sodium citrate as anticoagulant, and was centrifuged at 1,000 × g for 10 min. The platelet-rich plasma was transferred into a fresh tube and further centrifuged at 400 × g for 10 min to pellet the platelets, which were then resuspended in PBS. Cytospin smears were prepared for each of the cell preparations, and were stained with antibodies.

Identification and Quantitation of Circulating EC

Immunocytochemical methods were used to identify circulating EC. Two antibodies with specificity for EC, a polyclonal antihuman von Willebrand factor (vWf) and a monoclonal anti-vascular endothial growth factor receptor (VEGFr) antibody for the KDR receptor (Sigma), were used to identify and count circulating EC. HUVEC were used as positive controls and mononuclear cells were used as negative controls. Optimal antibody dilutions were determined before the clinical studies were done. Cytospin preparations were made from a standard concentration of HUVEC in PBS, and were stained with antibody to determine the sensitivity of cell detection in vitro. Serial dilutions of HUVEC were seeded into a standard volume of whole blood, and the recovery rate was determined after isolation. Final concentrations of 1,000, 500, 100, 75, 50, 30, 20, 15, and 10 EC per milliliter of whole blood were achieved. Five separate series of seeding experiments were performed.

Indirect Immunofluorescence Technique

One hundred microliters of the cell suspension was used to prepare cytospin smears, which were then stained with the antibodies at a dilution of 1:200 and incubated at 37° C for 30 min. After washing with PBS, the slides were further incubated with the secondary antibodies; goat antirabbit and goat antimouse (for the anti-vWf and anti-VEGFr antibodies, respectively) at 37° C for 30 min. Each test was done in duplicate at a minimum. Nonspecific binding was assessed with an isotype control antibody. Final cell counts were adjusted by subtraction of nonspecific binding. The slides were viewed with a Diaphot EPI fluorescence microscope (490-nm excitation, 520-nm emission) (Nikon, Tokyo, Japan). A conservative approach to cell identification was adopted. Only cells with at least a complete rim of immunofluoresecent staining were counted. In addition, stained cells were differentiated from possible clumps of debris or platelets by changing to phase-contrast light microscopy and assessing the size and morphology of the stained cells. Of these stained cells, those that were between 20 µm and 50 µm in diameter and which were nucleated were counted. All positively stained cells on the slide were directly counted, with the result and expressed as the number of circulating EC per milliliter of whole blood. Staining of cells with both antibodies (anti-vWf and anti-VEGFr), with different fluorochromes (TRITC and fluorescein isothiocyanate, respectively), was done on six samples, to assess the proportion of dual-staining cells.

Cell Culture

EC culture was attempted from all samples obtained from patients with septic shock and all healthy volunteers. The cells were resuspended in a complete medium (Medium 199) supplemented with 20% heat-inactivated fetal calf serum, 2 mM/100 ml L-glutamine, 100 IU/ml penicillin G and 100 g/ml streptomycin, and 300 µg/ml EC growth supplement (all from Sigma). The cells were grown in multiwell plates and incubated at 37° C in 5% CO₂. Unattached cells were removed after 24 h by washing the flasks twice with warm medium. The adherent cells were then grown in complete EC medium, with changes of medium made every 7 d. The cells were identified as EC by using monoclonal antibodies to EC markers, including vWf, VEGFr, and CD 105 (endoglin).

Statistical Analysis

Results are presented as the mean ± SEM unless otherwise stated. Analysis of variance (ANOVA) with Bonferroni's correction was used when comparing the four clinical groups. Significance was taken as a value of p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cell Staining and Retrieval

Mononuclear cell suspensions did not stain with either anti-vWf or anti-KDR antibody. Platelet clumps did stain with anti-vWf antibody, but these were easily distinguished from nucleated cells through phase-contrast microscopy. Platelets and platelet clumps did not stain with anti-KDR antibody. Ninety percent of HUVEC suspended in PBS were identified with the anti-vWf antibody. In whole blood, the mean recovery of vWf-positive cells was 86% (range: 73% to 92%) at seeding concentrations of 20 to 1,000 HUVEC per milliliter. At seeding concentrations below 20 cells per milliliter, recovery fell to less than 40%. Seventy-six percent of HUVEC suspended in PBS were identified with the anti-KDR antibody. In whole blood, the mean recovery of KDR-positive cells was 65% (range: 60% to 85%) at seeding concentrations between 20 and 1,000 HUVEC per milliliter of whole blood. Below this level, recovery was 30% or less.

Patient Data

The mean ages of the four study groups were similar (control group: 58 yr; ICU control group: 48 yr; sepsis group: 62 yr; septic shock group: 58 yr). The sex distributions of the groups were: healthy controls: eight males and three females; ICU controls: seven males and two females; sepsis group: five males and three females; and septic shock group: 11 males and four females. The percentage in each group who were current cigarette smokers were: healthy control group, 45%; ICU control group, 71%; sepsis group, 38%; and septic shock group, 47%.

Respiratory, urinary, and postsurgical intraabdominal infections and primary blood-borne sepsis accounted for the majority of cases of sepsis and septic shock in our series (Tables 1 and 2). APACHE II scores were high in the septic shock group, reflecting the severity of illness in this group of patients, and their elevated SOFA scores showed a high incidence of multiorgan failure. Severity of illness and the degree of multiorgan failure were significantly lower in the ICU control group (Table 3). ICU mortality was 67% in the septic shock group and 22% in the ICU control group. Mortality in the group with sepsis was 13%.

Circulating EC Numbers

Very few circulating EC were identified either in healthy controls or ICU controls. The numbers of vWf-positive cells were 1.9 ± 0.5/ml (mean ± SEM) in healthy controls and 2.6 ± 0.6/ml in ICU controls. Mean KDR-positive cells were 0.7 ± 0.3/ml in healthy controls and 0.5 ± 0.2/ml in ICU controls. There was a significant increase in circulating EC (between-group ANOVA: f = 44.7, p < 0.0001) both in patients with sepsis and those with septic shock (Figures 1 and 2). In patients with sepsis, vWf-positive cells were 16.1 ± 2.7/ml and KDR-positive cells were 4.2 ± 1.1/ml. In patients with septic shock, vWf-positive cells were 30.1 ± 3.3/ml (Figure 3) and KDR-positive cells were 10.4 ± 1.2/ml (between-group ANOVA: f = 26.6, p < 0.0001; Figure 4). The differences between the numbers of circulating vWf-positive and KDR-positive cells in patients with sepsis and septic shock were also significant (p < 0.001 for both groups). Absolute cell counts were higher with the anti-vWf antibody than with the anti-KDR antibody. The correlation between the two counts was high (r² = 0.93; Figure 5), and a series of dual-staining studies showed that 84% of the KDR-positive cells were also vWf-positive. There was no significant correlation between the total circulating mononuclear cell count and the number of vWf-positive or KDR-positive cells detected in patients with septic shock.

View larger version (24K): [in this window] [in a new window]   Figure 1.   High-power view of an individual EC from a patient with septic shock. The cells were separated by isopycnic centrifugation. Cytospin preparations were then made and stained with fluorescein isothiocyanate-conjugated anti-vWF antibody (original magnification: ×1,000).

View larger version (27K): [in this window] [in a new window]   Figure 2.   High-power view of an individual EC from a patient with septic shock. The cells were separated by isopycnic centrifugation. Cytospin preparations were then made and stained with fluorescein isothiocyanate-conjugated antibody to anti-VEGF-R (KDR) antibody (original magnification: ×1,000).

View larger version (10K): [in this window] [in a new window]   Figure 3.   Plot of the number of circulating vWf-positive EC/ml blood recovered from healthy controls, ICU controls, and patients with septic shock. Each data point represents one patient or control. Mean circulating EC numbers were significantly higher in sepsis (16.1 ± 2.7/ml) and septic shock patients (30.1 ± 3.3/ml) than in healthy controls (1.9 ± 0.5/ml) and ICU controls (2.6 ± 0.6/ml).

View larger version (10K): [in this window] [in a new window]   Figure 4.   Plot of the number of circulating KDR-positive EC/ml blood recovered from healthy controls, ICU controls, and patients with septic shock. Each data point represents one patient or control. Mean circulating EC numbers were significantly higher in sepsis (4.2 ± 1.1/ml) and septic shock (10.4 ± 1.2/ml) than in healthy controls (0.7 ± 0.3/ ml) and ICU controls (0.5 ± 0.2/ml).

View larger version (11K): [in this window] [in a new window]   Figure 5.   Plot of the number of vWf-positive circulating EC versus the number of KDR-positive circulating EC. The correlation between the two counts was high (r2 = 0.93).

The number of circulating KDR-positive EC was significantly higher in patients who died of septic shock than in survivors (12.0 ± 1.6/ml versus 7.1 ± 1.2/ml, p = 0.026). The difference between numbers of vWf-positive cells in the survivors and nonsurvivors of septic shock was not significant (34.0 ± 4.2/ml, versus 23.4 ± 3.8/ml, p = 0.09).

Cell Culture

EC were successfully cultured from six of the 15 patients with septic shock, but from none of the controls. Successful cultures were obtained from patients with the highest circulating cell counts. The cells attained between 50% and 80% confluence in 4 to 6 wk. The cells were confirmed as endothelial by the appearance of characteristic "cobblestoning" and with fluorescent antibodies to vWF, KDR, and endoglin. From 70 to 90% of cultured cells were positive for these markers.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We found an increase in circulating EC in patients with sepsis and septic shock as compared with both healthy volunteers and ICU patients who did not have sepsis. These results can be explained by the hypothesis that the vascular endothelium is the principal target of the inflammatory response in sepsis, and that widespread vascular damage and disruption occur to this endothelium (2, 4, 8). Our findings of an increase in circulating EC in patients with sepsis but without shock suggest that endothelial damage precedes the clinical development of organ failure. We also found that the mean number of circulating KDR-positive EC was significantly higher in patients who died of septic shock. These observations support the view that the magnitude of increase in circulating cell numbers in sepsis is related to the severity of vascular injury.

The endothelium modulates vascular tone, has important interactions with the coagulation system, and plays a major role in the movement of circulating leukocytes into various tissue beds (10, 11, 25). There is considerable indirect evidence for changes in vascular endothelial function in septic shock. A large number of in vitro studies have demonstrated that exposure of EC to lipopolysaccharide and proinflammatory cytokines initiates a series of molecular programs that alter cell function and may lead to cell dysfunction and damage (12). Stimulated EC upregulate surface adhesion molecules, express and release procoagulant factors, and secrete vasoactive compounds that include endothelins, NO, and prostaglandins. All of these substances and molecules have been implicated in the pathogenesis of septic shock (1).

Evidence for involvement of the vascular endothelium in human septic shock is less clear. There are surprisingly few studies of endothelial ultrastructure in human sepsis, but postmortem studies of patients who died of sepsis-related acute respiratory distress syndrome (ARDS) showed patchy EC swelling and injury (26). Skin biopsy specimens from patients with sepsis resulting from peritonitis showed upregulation of intercellular adhesion molecule-1 on vascular EC (27), and adhesion-molecule upregulation has also been reported on cultured pulmonary microvascular cells taken from patients who died of ARDS (28). Widespread endothelial damage has been reported in a number of studies of animals injected with lethal doses of endotoxin. These include endothelial separation from the basement membrane in primates (29) and vascular EC detachment and loss in primates (30), dogs (31), and rats (32).

Circulating EC have previously been detected in humans through immunofluorescence, immunohistochemical, and flow cytometric techniques (14). George and coworkers, using a flow cytometric method, detected a significant number of circulating EC after venous trauma (16). Sbabati and associates, using an immunofluorescence method, reported a rise in circulating EC following cardiac catheterization (19), and Mutin and colleagues reported increased circulating EC during myocardial infarction in humans (13). Solovey and coworkers, using an immunohistochemical method, found raised circulating numbers of EC during sickle cell crisis (20). Increased numbers of circulating EC have also been reported in humans during cytomegalovirus and rickettsial infections (14, 15, 17, 18). In addition, increased numbers of circulating EC have been found during experimental endotoxemia in the dog (31). In all of these studies, few circulating EC were detected in control subjects, and the order of magnitude of increase in cell numbers in patients was similar to that in the current study (10 to 100 cells/ml blood).

The detection of rarely circulating EC requires specific antibodies to EC receptors. We used two different antibodies to EC receptors, and also used morphologic criteria, to identify EC. vWf is synthesized by EC and subsequently released onto the cell surface (33). It is not completely specific for EC, since it is also synthesized by platelets. However, additional size and morphologic criteria were used to identify the EC in our study. Vascular endothelial growth factor (VEGF) is a cytokine with a major role in angiogenesis (34). Vascular EC express two VEGF receptors. VEGF-R1, or flt-1, is also expressed on monocytes, whereas VEGF-R2, or KDR, is expressed only on EC, pancreatic duct cells, and human testicular tissue. The anti-KDR antibody used in our study should be a highly specific marker for circulating EC, since it is absent on other circulating hemopoetic cells.

In the present study, we detected a consistently higher level of EC with the anti-vWf antibody than with antibody to KDR. The majority of KDR-positive cells were also vWF-positive, and the correlation between the numbers of these two kinds of EC, on a subject-by-subject basis, was high (r² = 0.93). This suggests that the KDR-positive EC were a subset of the vWf-positive cells. The difference in cell numbers detected with the two antibodies probably indicates a difference in antibody-receptor binding or a difference in the number of receptors expressed. We found that in vitro, the sensitivity of the polyclonal anti-vWF antibody was higher than that of the monoclonal anti-KDR antibody, with 90% of HUVEC staining with the anti-vWF antibody as compared with 75% that were stained with the anti-KDR antibody. This is likely to reflect the higher number of epitopes recognized by the polyclonal anti-vWF antibody than by the monoclonal anti-KDR antibody.

In summary, we have shown that patients with sepsis and septic shock have increased numbers of circulating EC. These findings support the hypothesis that widespread endothelial damage occurs in human sepsis.