INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

E-selectin or endothelial-leukocyte adhesion molecule 1 (ELAM-1) is a 115 kDa endothelial transmembrane glycoprotein, which, along with other members of the selectin family, is an early mediator of endothelial-leukocyte adhesion in inflammatory states. The selectins, E-selectin and P-selectin on endothelium and L-selectin on leukocytes, bind specific ligands to promote rolling of leukocytes along the endothelial surface prior to vascular cell adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1)-mediated tight binding and extravasation of leukocytes through endothelial cell junctions (1). There is minimal constitutive E-selectin expression (2), but the molecule is inducible in response to various cytokines (TNF-, interferon-gamma, IL-1, thrombin) as well as endotoxin, with significant expression on the luminal surfaces of activated endothelial cells (3, 4). Enhanced expression is also well documented in vivo. Monkeys infused with lipopolysaccharide show pronounced E-selectin expression in lung and skin vasculature (5). Baboons with lethal septic shock have widespread de novo E-selectin expression far in excess of that found in traumatic/hypovolemic shock (6, 7). Substantial E-selectin expression was found in skin vasculature at biopsy of 75% of patients with peritonitis due to perforation of the gastrointestinal tract (8).

A shed, soluble form of E-selectin of reduced molecular weight is present in the supernatant of activated endothelial cells within 24 h, with further increases for up to 72 h (9). Low levels of E-selectin are detected in serum of normal individuals (11) and several investigators have recently documented increased levels in patients with sepsis and other critical illnesses (11). There is interest in examining serum levels of E-selectin and other adhesion molecules in sepsis and other inflammatory conditions relating their measurement to outcome from critical illness (11).

In this study, we sought to examine more closely the relationship of E-selectin serum levels, as a putative marker for endothelial activation, to systemic inflammation, infection, hemodynamic compromise, organ dysfunction, and outcome in a large group of critically ill medical patients. Of the various adhesion molecules, E-selectin is uniquely displayed only on endothelium and is almost exclusively inducible, and thus may prove to be an excellent marker for endothelial activation (1). Further, E-selectin is expressed rapidly following cellular stimulation and therefore might be expected to be an early marker of systemic inflammation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Clinical Protocol

The protocol was approved by the Medical College of Virginia /Virginia Commonwealth University Committee on the Conduct of Human Research (CCHR) and adhered to guidelines issued by National Institutes of Health for protection of confidentiality of participating subjects. Informed written consent was waived by the CCHR except for patients in whom multiple blood sampling was performed. All patients were prospectively identified and clinical data and blood specimens were obtained within 24 h of presentation. In selected patients, as many as three additional blood samples were obtained on: (1) ICU day 2; (2) ICU day 3, 4, or 5; and (3) ICU day 6, 7, or 8. Clinical data were collected over the course of ICU hospitalization. Patients were followed until death or hospital discharge.

Definitions

Criteria proposed in the American College of Chest Physicians-Society of Critical Care Medicine (ACCP/SCCM) (19) consensus statement were utilized to define systemic inflammatory response syndrome (SIRS), sepsis, hypoperfusion, hypotension, severe sepsis (or severe SIRS), and septic shock (or SIRS with shock). Patients were categorized using data from the first 24 h of ICU hospitalization.

Systemic inflammation was indicated by the presence of at least two of the following: (1) body temperature > 38° C or < 36° C; (2) heart rate > 90 beats/min; (3) respiratory rate > 20 breaths/min or Pa_CO2 <  32 mm Hg; (4) WBC count > 12,000 cells/mm³ or < 4,000 cells /mm³ or > 10% immature neutrophils (19). Patients were considered to have sepsis if they had: (1) systemic inflammation, as defined above, plus (2) clinical and/or microbiological evidence of infection as the likely cause of inflammation (19). Clinical manifestations of intraabdominal infection included abdominal guarding, pain, and/or rigidity, increased number of neutrophils in the peritoneal fluid, and/or imaging evidence of focal infection or abscess. Clinical manifestations of pneumonia included purulent sputum, a new infiltrate seen on chest radiograph, and gram stain of sputum demonstrating many neutrophils, bacteria, and a paucity of epithelial cells. Evidence for central nervous system infection included nuchal rigidity and cerebrospinal fluid pleocytosis, elevated protein, and decreased glucose. Soft tissue swelling and tenderness with overlying skin erythema, and/or purulent drainage was consistent with wound infection or necrotizing fasciitis. Purulent drainage at the catheter insertion site was consistent with catheter infection. Microbiological evidence included positive cultures of blood and/or relevant body fluid or drainage. Based on inflammation and infection criteria, each patient was categorized as having (1) no SIRS; (2) non-infectious SIRS (no clinical or microbiological evidence of infection); (3) culture-negative sepsis (systemic inflammation plus clinical evidence of infection but no confirmatory microbiological results); or (4) culture-positive sepsis (systemic inflammation plus clinical evidence of infection as well as confirmatory microbiological results).

Hemodynamic categories were established based upon ACCP/ SCCM consensus criteria (19). Hypotension was defined as systolic BP < 90 mm Hg in the absence of other causes for hypotension, or use of > 5 µg /kg /min of dopamine, or any dose of norepinephrine (19). Organ hypoperfusion was manifested by one or more of the following: (1) lactic acidosis, or unexplained increased anion gap acidosis; (2) oliguria (< 30 cc/h for at least one hour); or (3) acute alteration in mental status unexplained by other conditions (19). Based on these criteria, patients were classified into three hemodynamic categories: shock, severe, and uncomplicated. Patients who had both hypotension and organ hypoperfusion were considered to have "shock" (19). Patients who had either hypotension or organ hypoperfusion, but not both were considered to have "severe" sepsis or SIRS (19). Patients with no hypotension or hypoperfusion were categorized as "uncomplicated."

Severity of illness was estimated from day 1 data using the APACHE II severity of illness classification system (20). Organ failure was determined using the method of Tran and coworkers (21). This scoring system was selected because of its development in the clinical setting of a Medical ICU. The Day 1 Multiple Organ Failure (MOF) score was determined by summing the number of organ systems failed (maximum of seven systems: cardiovascular, pulmonary, renal, neurologic, hematologic, hepatic, gastrointestinal) on day 1 of ICU hospitalization. The peak MOF score reflected the greatest number of organ systems failed on any one day over the course of ICU hospitalization. Death during hospitalization or survival to hospital discharge was determined.

Collection and Processing Of Blood Specimens

Blood specimens were collected from indwelling arterial or venous catheters or by venipuncture into nonheparinized glass tubes and placed at 4° C. Specimens were permitted to clot and centrifuged within 30 min at 4,000 g for 10 min. Serum specimens were then aliquoted and stored at 70° C until the time of assay.

Soluble E-selectin Assay

Samples were assayed by an ELISA specific for E-selectin by an investigator (D.B.) blinded to clinical parameters. Samples underwent four sets of three-fold dilution and each set run in triplicate for comparison to the standard curve. The assay is sensitive to 60 pg /ml for E-selectin and there is no cross reactivity with ICAM-1, VCAM-1, or P-selectin. The intra-assay and inter-assay coefficient of variation are <  5%. Assay results are expressed as ng/ml. Levels of soluble E-selectin for normal volunteers in this laboratory average 1 ng/ml (0.92 ± 0.66 [SD] ng/ml) (11).

Statistical Analysis

Continuous descriptive data (age, etc.) are displayed as mean ± SD and compared using unpaired t-tests. E-selectin values are not normally distributed (Shapiro-Wilk test: p < 0.0001). The values were skewed with a number of small values and a few very large values (skewness = 2.41). As a result, all analyses were performed on the log transformed E-selectin values. Corresponding nonparametric analyses were also performed and demonstrated identical results for all analyses. All summary values are reported on the original scale (after a reverse transformation from log values was performed) as mean and 95% confidence intervals. Comparison of groups was performed using ANOVA on log transformed data, followed by a post-hoc multiple comparison procedure (Tukey's HSD) as necessary. Tests of Correlation were also performed on log transformed E-selectin values. A p value of less than 0.05 was considered significant for all statistical tests.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied 119 adults with critical illness admitted to the Medical College of Virginia Hospitals 12-bed Medical Respiratory Intensive Care Unit (MRICU) (Table 1). We initially studied 24 patients with severe sepsis, then expanded the series to include additional septic patients as well as patients with other forms of critical illness by evaluating 95 consecutive, HIV-negative patients admitted to the MRICU over a 6-wk period. The initial group of septic patients were similar (p > 0.1) to the subset of the 95 consecutive ICU patients who met criteria for sepsis (n = 43) for age, sex, APACHE II score, day 1 MOF score, E-selectin level, and survival. Accordingly, data from the groups were combined for analysis. The most common admission diagnoses are displayed in Table 1. Of note, many patients who satisfied ACCP/SCCM criteria for sepsis (19) had other primary admission diagnoses. Patients were moderately ill with an average APACHE II score of near 20, two organs failed at onset, and 22% hospital mortality.

There were eight patients without SIRS, 44 with non-infectious SIRS, 24 with culture-negative sepsis and 43 with culture-positive sepsis on ICU day 1 (Table 2). The respiratory tract was the most common site of infection for sepsis, regardless of whether microbiological confirmation of infection was present, followed by the urinary tract and abdomen (Table 2). Among microbiologically-confirmed sepsis cases, there were similar numbers of gram positive and gram negative infections. There were two systemic fungal infections and one case each of Rocky Mountain spotted fever (previously reported in references 22, 23) and malaria. Thirteen patients had documented bacteremia.

Log transformed E-selectin levels obtained within 24 h of ICU admission averaged 5.28 ng/ml (95% CI = 3.81, 7.32) with a median of 7.3 and a range from 0.03 to 123 ng/ml. This far exceeds the average level of 0.92 ± 66 ng/ml for normal volunteers in the Otsuka America laboratory. Serial E-selectin levels measured on as many as four occasions over the course of the first week of ICU hospitalization in 25 patients demonstrated no appreciable change over time (Figure 1). E-selectin levels were similar for men and women (p = 0.2) and among all ages (r = 0.12, p = 0.19). Day 1 E-selectin correlated modestly with APACHE II score (r = 0.23, p = 0.012) and Day 1 MOF score (r = 0.18, p = 0.053).

View larger version (23K): [in this window] [in a new window]   Figure 1.   Serial measurement of E-selectin in 25 randomly selected patients demonstrated stable levels for up to 1 wk. Open figures denote survivors and closed figures nonsurvivors. Circles represent septic patients; squares represent nonseptic patients.

Patients from the four inflammation/infection categories had different E-selectin levels (one way ANOVA, F[3, 115] = 5.46, p < 0.0001) (Figure 2, Table 2). Patients with culture-positive sepsis had higher E-selectin levels than all other inflammation/infection categories (p < 0.05, Tukey's HSD), whereas other categories did not differ from each other. Patients from the three hemodynamic categories had significantly different E-selectin levels (one way ANOVA, F[2, 116] = 12.45, p < 0.0001) (Figure 3, Table 2) and each of the three groups were significantly different from the others (each p < 0.05, Tukey's HSD). In order to evaluate possible interactions between the inflammation/infection and hemodynamic categories, a two way ANOVA on the log of E-selectin was performed, demonstrating no interaction (F[5, 108] < 1, p > 0.7). This analysis demonstrated significant, and independent, differences for log E-selectin levels among the three hemodynamic categories (F[2, 113] = 6.80, p < 0.0016) as well as for culture-positive sepsis versus other inflammation/infection categories (F[3, 113] = 7.01, p < 0.0002). Thus the association of higher E-selectin levels with culture positive sepsis is independent of the severity of hemodynamic compromise. The log-transformed mean E-selectin levels for all subsets of inflammation/infection category and hemodynamic category, as well as numbers of patients, are displayed in Figure 4. The above analyses were also performed on the 95 consecutive ICU patients, with significant differences again found for infection/inflammation categories (one way ANOVA, F[3, 91] = 4.18, p < 0.008) and for hemodynamic categories (one way ANOVA, F[2, 92] = 8.54, p  < 0.0004).

View larger version (18K): [in this window] [in a new window]   Figure 2.   Individual (open symbols) and log-transformed mean (bar) E-selectin levels are displayed for four patient groups based upon inflammation and infection status. Open triangles represent the initial group of 24 patients with severe sepsis and open circles represent the 95 consecutive ICU patients. E-selectin levels were different among all groups (p < 0.0001) and were higher (p < 0.05) for culture-positive sepsis versus all other groups. Data are displayed in Table 2.

View larger version (19K): [in this window] [in a new window]   Figure 3.   Individual (open symbols) and log-transformed mean (bars) E-selectin levels are displayed for groups based upon hemodynamic status. Open triangles represent the initial group of 24 patients with severe sepsis and open circles represent the 95 consecutive ICU patients. The three groups were significantly different (p < 0.0001) and each group was significantly different from the others (p < 0.05). Data are displayed in Table 2.

View larger version (47K): [in this window] [in a new window]   Figure 4.   Log-transformed mean E-selectin levels are displayed for subgroups categorized by hemodynamic severity (uncomplicated, severe, shock) and inflammation/infection (no SIRS, noninfectious SIRS, culture-negative sepsis, and culture-positive sepsis). The number of patients in each group is displayed on the corresponding column.

There was no significant difference in E-selectin levels by site of infection (p = 0.14) or by class of organism (p = 0.9) (Table 2). The 13 patients who had positive blood cultures and clinical evidence of infection had higher (p = 0.003) E-selectin levels than 106 patients with negative blood cultures or for whom no blood cultures were obtained.

We found a modest correlation (r = 0.30, p = 0.001) between day 1 E-selectin level and peak MOF score (2.3 ± 1.9) during ICU hospitalization for all patients. E-selectin levels were higher (p < 0.05) for nonsurviving critically ill patients (Table 2), but the difference was not significant for surviving and nonsurviving septic patients (p = 0.57).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

As expected, our critically ill patients had higher E-selectin levels than was found in normal volunteers. Within this group of ICU patients, we observed striking, and independent, relationships between serum E-selectin levels and the presence of well-documented infection as the cause of systemic inflammation and the severity of hemodynamic compromise. As has been documented elsewhere (24) the vast majority of our ICU patients (93%) had systemic inflammation, as judged by conventional criteria (19). Patients with systemic inflammation had higher E-selectin levels than non-SIRS patients, however when examined more closely, patients with microbiologically documented infection as the cause of systemic inflammation (culture-positive sepsis) accounted for this difference. In fact, patients with noninfectious SIRS had E-selectin levels which were virtually identical to levels for patients without SIRS and were statistically similar to E-selectin levels for patients with clinical evidence of infection, but no microbiological confirmation (culture-negative sepsis). Patients with culture-positive sepsis had E-selectin levels three- to eightfold higher than other groups. Statistical testing confirmed that higher E-selectin levels in culture-positive sepsis was independent of the severity of hemodynamic compromise. The subset of patients with documented bacteremia had the highest E-selectin levels.

There are a number of interesting implications from these findings. First, the intensity of endothelial activation, considered to be a pivotal feature of systemic inflammation (20, 25, 26) may differ for sepsis as compared with other causes of critical illness. In support of this theory, Redl and coworkers (7) observed that baboons with shock due to infusion of live bacteria had greatly enhanced vascular expression of E-selectin, whereas baboons with hemorrhagic shock did not. Cowley and colleagues (12) documented higher E-selectin levels for patients with sepsis and organ dysfunction than for "miscellaneous severely ill patients" who were comparably ill (similar APACHE II scores and mortality rates). Recently, Moss and colleagues (16) documented significantly higher E-selectin and ICAM-1 levels and von Willebrand factor antigen activity, as measures of endothelial activation, for patients with sepsis compared with similarly ill trauma patients at risk for ARDS. Sessler and colleagues (18) reported a trend (p = 0.1) for higher circulating ICAM-1 levels for patients with sepsis than for critically ill noninfectious SIRS patients.

Similar observations are reported for blood cytokines. Pinsky and colleagues (27) found septic shock patients to have significantly higher peak TNF- and IL-6 blood levels than patients with shock from other causes. Casey and colleagues (28) reported higher TNF-, IL-1, and IL-6 plasma levels at onset of illness for septic patients as compared to critically ill medical ICU patients.

Current thinking emphasizes the similarities of fulminant widespread inflammation caused by sepsis and systemic inflammation associated with noninfectious conditions such as acute pancreatitis, ischemia, multiple trauma, hemorrhagic shock, and immune-mediated organ injury within the "systemic inflammatory response syndrome" or SIRS concept (19). Our findings and those of other investigators (7, 12, 16, 18, 27, 28) strongly suggest that significant quantitative differences in endothelial activation as well as for circulating cytokines exist for sepsis versus noninfectious systemic inflammation.

We were intrigued by the finding that our patients who had culture-negative sepsis had significantly lower E-selectin levels than patients with culture-positive sepsis. This observation was consistent among the principle sites of infection (respiratory and intraabdominal). The absence of microbiological confirmation of infection despite strong clinical suspicion is common. Thirty-six percent of our septic patients had culture-negative sepsis, similar to recent multicenter prospective clinical trials and large prospective series of patients with severe sepsis (24, 28) in which as many as one-half of all septic patients were culture-negative. Potential explanations for lack of microbiological confirmation include the inability to obtain material for culture, concomitant antimicrobial use at the time material was obtained for culture, and infection due to fastidious bacteria or other pathogens; however, it is also possible that infection is not actually present in some cases. We observed similar survival rates for culture-positive and culture-negative sepsis (69% and 67%, respectively), as have others (29). However, in a large prospective survey Rangel-Frausto and associates (24) found patients with culture-negative sepsis to have a lower mortality rate as well as lower attack rates for ARDS, disseminated intravascular coagulation, acute renal failure, and shock than culture-positive sepsis. No other investigators have compared E-selectin or other indicators of endothelial activation for culture-negative and culture-positive sepsis, although plasma cytokine levels were similar for the two groups in one series (28). Our results suggest that the "intensity" of endothelial activation, as judged by expression and subsequent shedding of E-selectin, may be greater for septic patients who have microbiologically confirmed infection compared to culture-negative sepsis. The explanation for our observations, as well as the potential implications remain to be defined.

We documented a striking positive relationship between E-selectin level and concomitant hemodynamic dysfunction as reflected by the presence of ACCP/SCCM consensus-defined hypotension and/or organ hypoperfusion (19). Patients with the combination of hypotension and organ hypoperfusion (i.e., shock) had the highest E-selectin levels. This relationship persisted after adjustment for infection/inflammation category. We also documented a significant correlation between E-selectin on day 1 of sepsis and peak multiple organ failure score over the ICU hospitalization. Other investigators have also documented a relationship between soluble adhesion molecule levels and hemodynamic and/or dysfunction parameters. Newman and colleagues (11) found much higher E-selectin levels among bacteremia patients with hypotension versus those with normal blood pressure. Cowley and coworkers (12) documented higher initial and peak E-selectin levels for septic patients with organ dysfunction than those without. Similarly, higher cICAM-1 and sVCAM-1 levels have been found in patients with more severe shock and greater organ dysfunction or failure (12, 18).

Is it possible that increased amounts of soluble E-selectin in blood result from endothelial injury rather than mere activation? Numerous studies have demonstrated that E-selectin is shed from cultured endothelial cells which are activated but have no evidence of injury or detachment (9). Further, shed E-selectin differs from that recovered from the lysates of detergent-fragmented endothelial cells by being of lower molecular weight and by having no detectable cytoplasmic domains (11). E-selectin detected in the blood of four normal subjects had no detectable cytoplasmic domain suggesting shedding rather than injury accounts for the low level expression in vivo (11); however, such studies were not reported for patients with septic shock who had markedly elevated E-selectin blood levels. Thus, while it is possible that microvascular disruption and injury is in part responsible for E-selectin release, this remains speculative and requires further study.

For all critically ill patients we documented higher E-selectin levels for nonsurvivors compared to survivors. The difference, however, was not statistically significant for septic patients. Other investigators have reported varying results regarding outcome. Boldt and coworkers (14), studying patients with severe trauma or postoperative complications, found nonsurvivors to have higher blood levels of E-selectin, ICAM-1, and VCAM-1 at study entry. Further, they observed blood levels to be stable or decline over 5 days in survivors whereas blood levels increased further in nonsurvivors. Cowley and coworkers (12) found no difference in E-selectin levels for survivors versus nonsurvivors; however, all patients with initial or peak E-selectin levels above 36 U/ml died. We previously demonstrated significantly higher levels of cICAM-1 on ICU day 1 for nonsurvivors than survivors (18). Others (14) have made similar observations for circulating ICAM-1 and VCAM levels in specific populations. We feel these conflicting data reinforce the need for extensive validation of the reliability and clinical value before these and other putative biochemical markers of outcome are applied clinically.

Although we can successfully detect and quantify E-selectin and other soluble adhesion molecules in circulating blood and other biological fluids, the actual role these shed molecules play in the inflammatory process is far from clear. Initial cellular expression mediates the sequential process of placing effector inflammatory cells such as neutrophils in the region of the inflammatory stimulus. There is evidence that soluble E-selectin, as well as other adhesion molecules remain active once they are shed and can influence subsequent adhesion in several ways. E-selectin upregulates neutrophil CD11b, the ligand for ICAM-1 enhancing firm leukocyte attachment and transendothelial migration (31). Since recombinant E-selectin can inhibit leukocyte adhesion (32) to endothelium in vitro, it is likely that circulating E-selectin limits E-selectin mediated rolling of activated leukocytes via competitive binding of cell-bound ligands, thus "downregulating" the inflammatory response. In rats, intravenous injection of soluble E-selectin 2 and 4 h after intratracheal injection of LPS inhibits subsequent neutrophil emigration into airspaces (33). The magnitude of protection against LPS was similar to that seen after intravenous injection of anti-E-selectin monoclonal antibody in LPS treated rats.

Since the intensity of E-selectin expression and shedding appears to correlate with organ hypoperfusion and dysfunction, inactivation of cell-bound E-selectin might be expected to ameliorate these sequelae. Anti-E-selectin monoclonal antibody protected rats against LPs-induced lung injury (33). Pre-treatment of Pseudomonas-infused swine with a dual anti-E- and anti-L-selectin monoclonal antibody significantly attenuated lung injury (34). In a preliminary study, administration of a murine monoclonal antibody to E-selectin to patients with septic shock was well tolerated and associated with dose-related improvement in oxygenation and trends for faster resolution of organ failure (35).

In summary, we found E-selectin to be detectable in increased quantities in patients with SIRS and sepsis, but particularly in patients with bacteremia or other microbiological confirmation of sepsis. Patients with organ hypoperfusion and/ or systemic hypotension had higher serum E-selectin levels than patients with uncomplicated illness. We found a modest correlation between day 1 E-selectin levels and organ dysfunction as well as survival.