Levels of soluble endothelium-derived adhesion molecules in patients with sickle cell disease are associated with pulmonary hypertension, organ dysfunction, and mortality

Summary

Endothelial cell adhesion molecules orchestrate the recruitment and binding of inflammatory cells to vascular endothelium. With endothelial dysfunction and vascular injury, the levels of endothelial bound and soluble adhesion molecules increase. Such expression is modulated by nitric oxide (NO), and in patients with sickle cell disease (SCD), these levels are inversely associated with measures of NO bioavailability. To further evaluate the role of endothelial dysfunction in a population study of SCD, we have measured the levels of soluble endothelium-derived adhesion molecules in the plasma specimens of 160 adult patients with SCD during steady state. Consistent with a link between endothelial dysfunction and end-organ disease, we found that higher levels of soluble vascular cell adhesion molecule-1 (sVCAM-1) were associated with markers indicating renal dysfunction and hepatic impairment. Analysis of soluble intercellular cell adhesion molecule-1 (sICAM-1), sE-selectin and sP-selectin levels indicated partially overlapping associations with sVCAM-1, with an additional association with inflammatory stress and triglyceride levels. Importantly, increased soluble adhesion molecule expression correlated with severity of pulmonary hypertension, a clinical manifestation of endothelial dysfunction. Soluble VCAM-1, ICAM-1, and E-selectin were independently associated with the risk of mortality in this cohort. Our data are consistent with steady state levels of soluble adhesion molecules as markers of pulmonary hypertension and risk of death.

Endothelial cell adhesion molecules play a vital physiological role in the recruitment and binding of inflammatory cells to vascular endothelium, particularly in venules. E-selectin and P-selectin induce specific inflammatory cells to slowly roll across the endothelial surface, until firm adhesive interactions develop on endothelial vascular cell adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1) to specific counter-ligands on the surface of inflammatory cells (Alon & Feigelson, 2002). Recent data from a number of research groups indicate that interactions between sickle reticulocytes, leucocytes, and endothelial cells via these adhesion molecules occur in patients with sickle cell disease (SCD), and may contribute to disease pathology (Gee & Platt, 1995; Setty & Stuart, 1996; Turhan et al, 2002; Hebbel et al, 2004). Vaso-occlusion because of haemoglobin S polymerisation, and the extent of polymerisation itself, may be compounded by adhesion molecule-dependent reticulocyte–monocyte–endothelial interactions in the postcapillary venules (Swerlick et al, 1993; Setty & Stuart, 1996; Spring et al, 2001; Frenette, 2004; Wood et al, 2004a). Expression of vascular VCAM-1, ICAM-1 and P-selectin are associated with sickle retinopathy (Kunz et al, 2002). Adhesion molecules mediate vaso-occlusion in mouse models of SCD (Matsui et al, 2001; Belcher et al, 2003; Embury et al, 2004; Wood et al, 2004b; Belcher et al, 2005). The expression of these cell adhesion molecules is under the influence of inflammatory cytokines, such as tumour necrosis factor-α (TNF-α) and interleukin-1β (IL-1β), higher levels of which have been measured in the plasma of patients with SCD (Vordermeier et al, 1992; Malave et al, 1993; Duits et al, 1996, 2003; Saleh et al, 1998; Conran et al, 2004; Tavakkoli et al, 2004). Sickle cells have also been reported to induce endothelial adhesion molecules (Shiu et al, 2000; Brown et al, 2001). Largely based on in vitro cytokinedependent VCAM-1 and ICAM-1 expression in cell culture systems, it is widely believed that elevated levels of soluble adhesion molecules in the plasma of patients with SCD occurs secondary to increased systemic inflammatory stress (Makis et al, 2000).

However, adhesion molecule expression is also modulated by endothelial nitric oxide (NO) produced from NO synthase. This modulation appears to be mediated by its inhibition of superoxide and secondary downregulation of nuclear factor (NF)-κB (Takahashi et al, 1996; Spiecker et al, 1997, 1998; Space et al, 2000). In fact, inhibition of endothelial NO synthase during TNF-α or IL-1β stimulation significantly augmented VCAM-1 gene expression (De Caterina et al, 1995). Thus, elevated sVCAM-1 levels in plasma reflect impaired endothelial function, a relationship that is observed in the vascular biology of SCD. Our group and others have determined that NO bioavailability is impaired in patients and mice with SCD (Nath et al, 2000; Belhassen et al, 2001; Eberhardt et al, 2003; Gladwin et al, 2003), due in large part to the consumption of NO by cell-free plasma haemoglobin (Reiter et al, 2002) and reactive oxygen species (Aslan et al, 2000, 2001, 2003; Aslan & Freeman, 2004), and compounded by low levels of the substrate for NO synthesis, L-arginine (Enwonwu et al, 1990; Morris et al, 2000, 2005). We and others have observed that increased levels of soluble (s)VCAM-1 and sICAM-1 are associated with decreased NO bioavailability and increased haemolytic rate in patients with SCD (Setty & Stuart, 1996; Nath et al, 2000; Gladwin et al, 2003). In fact, it is known that NO suppresses VCAM-1, ICAM-1 and E-selectin expression and adhesion of sickle erythrocytes to vascular endothelium (De Caterina et al, 1995; Khan et al, 1996; Shin et al, 1996; Spiecker et al, 1998; Space et al, 2000; Lee et al, 2002). Thus, excessive endothelial activation and vaso-constriction because of impaired NO bioavailability may contribute to vascular instability in patients with SCD (Reiter et al, 2002; Reiter & Gladwin, 2003; Nath et al, 2004).

These observations suggest that soluble adhesion molecules are potential markers of endothelial dysfunction, characterised by endothelial activation and insufficient bioavailability of endothelium-derived NO. There are a number of clinical observations in SCD patients that support this thesis: (1) hydroxyurea therapy has been reported to both augment endothelial NO production (Gladwin et al, 2003; Morris et al, 2003) and downregulate sVCAM-1 and sICAM-1 levels (Saleh et al, 1999; Gladwin et al, 2002, 2003; Conran et al, 2004); (2) patients presenting with the acute chest syndrome have plasma sVCAM-1 levels that are inversely proportional to NO metabolite levels (Stuart & Setty, 1999); and (3) vasomotor responsiveness to infusions of nitroprusside (a direct NO donor) is inversely proportional to measured plasma levels of sVCAM-1 (Gladwin et al, 2003). Finally, a polymorphism in the VCAM-1 gene appears to confer risk of stroke in patients with SCD (Taylor et al, 2002). Therefore, in the present study, we hypothesised that plasma levels of sVCAM-1, and other endothelium-derived adhesion molecules, could be used as surrogate biomarkers to characterise vasculopathy and endothelial dysfunction in population studies of patients with SCD. These measurements enabled us to test a number of predictions based on small translational studies, including (1) the relationship between haemolysis and endothelial dysfunction; (2) the modulating effects of hydroxyurea and gender on endothelial function; and (3) the putative role of endothelial dysfunction in the development of pulmonary hypertension, renal insufficiency and death.

Materials and methods

Subjects

The patient population comprised 160 adult (aged 18–74 years) patients with SCD (haemoglobin SC patients were excluded) studied at steady state in an outpatient clinic setting, at a time when hospitalisation was not required. Fifty-two healthy subjects were recruited as normal controls, six of whom were known to have sickle cell trait. For 41 of the healthy control subjects, clinical blood count and chemistry data were available. Informed consent was signed by each subject for an Institutional Review Board-approved protocol to obtain clinical information, echocardiography and blood specimens for research analysis (Gladwin et al, 2004). Clinical and laboratory characteristics of the patients and normal control subjects are described in Table I.

Table I.

Characteristics of the SCD and normal African-American control subject populations.

	Sickle cell patients			Control subjects
Patient characteristic	n	Mean	SD	n	Mean	SD	P-value
Age (years)	160	35	11	41	36	11	ns
Alanine aminotransferase (U/l)	159	28	15	41	22	12	0.02
Albumin (g/l)	158	41	4	41	41	2	ns
Alkaline phosphatase (U/l)	158	126	96	41	77	23	<0.001
Aspartate aminotransferase (U/l)	157	43	22	41	23	8	<0.001
Bilirubin, total (lmol/μl)	157	47.7	20.1	41	10.9	4.6	<0.001
Bilirubin, direct (lmol/μl)	158	8.7	6.7	40	2.4	0.9	<0.001
Blood pressure, systolic (mmHg)	139	121	17	36	133	19	0.002
Blood pressure, diastolic (mmHg)	139	66	11	36	74	17	0.01
C-reactive protein (g/l)	138	7.6	10.1	35	4.2	1.4	<0.001
Creatine kinase (U/l)	155	80	75	41	220	157	<0.001
Creatinine (μmol/l)	159	88.4	141.4	41	78.1	20.0	ns
Erythrocyte sedimentation rate (mm/h)	120	45	31	38	21	18	<0.001
Ferritin (μg/l)	152	1022	1448	39	75	61	<0.001
Gender, female (fraction)	160	0.61		41	0.59		ns
Haematocrit (volume fraction)	158	0.27	0.05	41	0.40	0.04	<0.001
Haemoglobin (g/dl)	158	9.1	1.6	41	13.5	1.6	<0.001
Haemoglobin A (mol/mol)	160	0.010	0.016	40	0.91	0.18	<0.001
Haemoglobin F (mol/mol)	159	0.085	0.066	39	0.004	0.006	<0.001
Iron (μmol/l)	141	18.8	9.7	38	14.1	6.8	0.001
Lactate dehydrogenase (U/l)	142	369	142	41	168	43	<0.001
Mean corpuscular volume (fl)	158	93	12	41	85	7	<0.001
Neutrophil count (109/l)	160	5.6	2.7	41	3.2	1.6	<0.001
Oxygen saturation, transcutaneous (fraction)	94	0.96	0.04	37	0.99	0.01	<0.001
Platelet count (109/l)	156	378	127	41	271	60	<0.001
Protein, total (g/l)	158	77	7	41	76	5	ns
Reticulocyte count (109/l)	150	272	135	37	69	30	<0.001
Transferrin (g/l)	157	2.01	0.52	38	2.71	0.51	<0.001
Transferrin saturation (fraction)	141	0.57	0.33	38	0.30	0.13	<0.001
Triglycerides (mmol/l)	141	1.32	0.82	32	0.89	0.53	<0.001
Urea nitrogen (mmol/l)	159	3.93	3.93	41	4.23	1.23	ns
Uric acid (mmol/l)	158	0.37	0.13	41	0.31	0.08	<0.001
Weight (kg)	124	76	19	36	82	16	<0.001
White blood cell count (109/l)	158	10.8	3.8	41	5.77	2.03	<0.001

ns, not significant; P-value calculated by two-sided t-test.

Echocardiography

Transthoracic echocardiography was performed in all patients with the use of the Acuson Sequoia (Siemens) and Sonos 5500 (Philips, Chicago IL, USA) systems. Cardiac measurements were performed according to the guidelines of the American Society of Echocardiography. Transmitral flow, Doppler determinations of the severity of valvular regurgitation, and left ventricular stroke volume were assessed and graded as previously described (Gladwin et al, 2004). Peak velocities of the E wave and A wave, the ratio of the E wave to the A wave, and the deceleration time were measured in a standard manner. Isovolumic relaxation time was measured as the time from aortic-valve closure to the start of mitral inflow. Tricuspid regurgitation was assessed in the parasternal right ventricular inflow, parasternal short-axis, and apical four-chamber views, and a minimum of five sequential complexes were recorded. Continuous-wave Doppler sampling of the peak regurgitant jet velocity was used to estimate the right-ventricular-to-right-atrial systolic pressure gradient with the use of the modified Bernoulli equation [4 × (tricuspid regurgitant jet velocity)²]. For the purpose of analysis, we prospectively defined pulmonary hypertension as a peak tricuspid regurgitant jet velocity (TRV) of at least 2.5 m/s. Since most patients with clinically significant pulmonary hypertension have measurable tricuspid regurgitation, we assumed that pulmonary-artery pressures were normal in patients with trace or no tricuspid regurgitation. Pulmonary-artery systolic pressure was quantitated by adding the Bernoulli-derived pressure gradient to the estimated mean right atrial pressure. The mean right atrial pressure was calculated according to the degree of collapse of the inferior vena cava with inspiration: 5 mmHg for a collapse of at least 50% and 15 mmHg for a collapse of less than 50%.

Enzyme-liked immunosorbent assay

Plasma levels of sVCAM-1, sICAM-1, sE-selectin, and sP-selectin were measured in duplicate on plasma samples using commercially available kits (R & D Systems, Minneapolis, MN, USA). Assays were performed on plasma specimens diluted as recommended by the manufacturer. Because of limitations of available plasma, not all markers were assessed in all 151 subjects with SCD. The precise numbers of subjects for each variable are indicated in the results section. The performance characteristics of these assays have already been established by the manufacturer, with inter-assay and intra-assay coefficient of variation each <10%. The lower limit of detection for sVCAM-1 was 4.5 ng/ml, sICAM-1 1.86 ng/ml, sE-selectin 0.54 ng/ml, and sP-selectin 1.54 ng/ml.

Measurement of myeloperoxidase levels

Myeloperoxidase levels were measured with use of an enzyme-linked immunosorbent assay utilising monoclonal (capture) and polyclonal (detection) antibodies to human myeloperoxidase (PrognostiX Inc., Cleveland, OH, USA). Each plate included a standard curve with isolated myeloperoxidase (extinction coefficient of 178 000/M/cm) and controls to correct for interplate variability. The inter- and intra-CVs were both <7%. The lower limit of detection was 13 pM.

Clinical laboratory assays

Standard clinical laboratory assays were performed in the Department of Laboratory Medicine in the Clinical Center of the National Institutes of Health (Bethesda, MD, USA). Plasma amino acids were quantified via ion exchange chromatography (Beckman model 6300 amino acid analyzer; Fullerton, CA, USA) at the Mayo Clinic, Rochester MN, USA.

Data analysis

Bivariate associations with the soluble adhesion molecules were assessed using Spearman rank correlation coefficients for continuous variables and Wilcoxon rank-sum tests for dichotomous variables. Associations of adhesion molecules with pulmonary hypertension (defined in three levels) were evaluated by a z-test for linear trend in Kruskal–Wallis analysis of variance (ANOVA). Associations with multiple variables simultaneously were evaluated using multiple linear regression on log₁₀ transformed levels of adhesion molecules, which should reduce the influence of large outlying values compared with regression on untransformed values. All variables with bivariate association with an adhesion molecule at P < 0.15 were considered in the modelling. Regression modelling using ranks for all variables was also done. Associations with mortality were assessed using log10 transformed levels of adhesion molecules in proportional hazards (Cox) regression models and intervals of adhesion molecule levels in logrank tests. P-values ≤ 0.05 were considered statistically significant. No adjustment was made for multiple comparisons. Statistical analysis was done using NCSS 2004 (Number Cruncher Statistical Systems, Kaysville, UT, USA) and Prism 4 (Graph- Pad Software, San Diego, CA, USA).

Results

Patients with SCD versus African-American control subjects

General characteristics of the study populations are shown in Table I. The plasma levels of all four soluble adhesion molecules were compared in SCD patients and African-American control subjects without SCD. Significant elevations are seen in the plasma levels of sVCAM-1 (median 811 vs. 334 ng/mL, P < 0.001, Wilcoxon rank-sum test), sE-selectin (median 74.6 vs. 41.5 ng/mL, P < 0.001), and sP-selectin (median 170 vs. 106 ng/mL, P < 0.001) in patients with SCD (n = 148, 139, and 144, respectively) compared with controls (n = 40) (Fig 1). The difference between patients (n = 143) and controls (n = 41) in the levels of sICAM-1 was not significant (median 189 vs. 228 ng/mL).

Fig 1.

Soluble adhesion molecules in patients with sickle cell disease compared with normal control subjects. Median plasma levels of sVCAM-1, sE-selectin, and sP-selectin are significantly higher in patients with sickle cell disease (SCD) (n = 148, 143, 139, and 144, respectively) compared with normal controls (P < 0.001 for each comparison, Wilcoxon rank-sum test). No significant difference was seen in median levels of sICAM-1 between normal controls and SCD patients, although a wider range of sICAM-1 values was seen in patients with SCD. Horizontal bars indicate median values.

Age and gender

None of the soluble adhesion molecule levels were significantly associated with age (Table II). Similarly, there were no significant associations with gender by Wilcoxon rank-sum tests.

Table II.

Correlations with soluble adhesion molecule expression. Spearman correlation coefficients (ρ) are shown. For each coefficient, n is between 79 and 148.

	Log10 sVCAM-1	Log10 sICAM-1	Log10 sE-selectin	Log10 sP-selectin
Log10 sVCAM-1	–	0.44***	0.26**	0.25**
Log10 sICAM-1	0.44***	–	0.26**	0.11
Log10 sE-selectin	0.26**	0.26**	–	0.33***
Log10 sP-selectin	0.25**	0.11	0.33***	–
White blood count	0.25**	0.10	0.21*	0.24**
Haemoglobin	−0.32***	−0.13	−0.13	−0.09
Platelet count	−0.13	0.01	0.08	0.40***
Reticulocyte count	0.20*	0.07	0.29***	0.15
% HbA	−0.06	−0.05	−0.04	−0.08
% HbS	0.18*	0.03	0.05	0.07
% HbF	−0.27***	−0.02	−0.15	−0.04
ESR	0.04	0.13	−0.09	−0.25**
C-reactive protein	0.12	0.10	0.15	0.07
Myeloperoxidase	0.20*	0.19*	0.15	0.04
Alanine aminotransferase	0.22**	0.25**	0.26**	0.18*
Albumin	−0.04	−0.18*	−0.06	0.08
Alkaline phosphatase	0.23**	0.29***	0.25**	0.07
Aspartate aminotransferase	0.39***	0.28***	0.36***	0.16
Bilirubin, direct	0.36***	0.40***	0.29***	−0.01
Bilirubin, indirect	0.17*	−0.00	0.20*	0.07
Blood urea nitrogen	0.20*	0.05	0.10	0.02
Cholesterol	0.04	0.16	0.13	0.13
Creatinine	0.16*	0.13	0.09	−0.06
Lactate dehydrogenase	0.38***	0.15	0.17	0.04
Triglycerides	0.20*	0.17	0.33***	0.32***
Potassium	0.18*	0.17	0.16	0.04
Transferrin saturation	0.19*	0.11	0.17	−0.06
Uric acid	0.35***	0.11	0.10	−0.03
TRV	0.16	0.12	0.10	0.04
Oxygen saturation	−0.20	−0.07	−0.37***	−0.34**
Age	0.10	0.08	0.13	−0.08

^*P ≤ 0.05;

^**P < 0.01;

^***P < 0.001.

Relationship of soluble adhesion molecules to other clinical variables

We evaluated the relationship of soluble adhesion levels to a variety of laboratory measurements (Table II). A number of interesting patterns are observed. All four adhesion molecule variables showed evidence of statistically significant relationships to each other (except for sICAM-1 versus sP-selectin), although the correlation coefficients were not high. The level of sVCAM-1 was significantly positively associated with markers of haemolysis, iron burden, renal and hepatic function, pulmonary arterial pressure, white blood cell count, and plasma myeloperoxidase, and inversely associated with fetal haemoglobin level (Table II). It appeared that elevation of sVCAM-1 was associated with dysfunction of multiple organs. This potentially represents pathological activation of vascular endothelium in these organs, and as a marker it may integrate the effects of vascular pathology or endothelial dysfunction in each of these organs. Surprisingly, we did not find significant associations of log₁₀ sVCAM-1 with erythrocyte sedimentation rate (ρ = 0.04, P = 0.71) or C-reactive protein (ρ = 0.12, P = 0.20).

In order to understand the most important independent variables for increased levels of sVCAM-1, multiple linear regression modelling was performed for log₁₀ values of sVCAM-1. Dependent variables for which P < 0.05 in the final model were included. The three variables that fit best into a model for log₁₀ sVCAM-1 were lactate dehydrogenase, direct bilirubin, and urea nitrogen (Table III). Although there was evidence on non-normality of the residuals (P = 0.04 by Shapiro–Wilk test), all three of these variables also had P < 0.05 in regression on ranks. These results suggest that the processes of intravascular haemolysis, cholestatic hepatic dysfunction, and renal dysfunction may be independently linked to increased sVCAM-1 levels.

Table III.

Multiple regression models for adhesion molecule expression.

	Log10 sVCAM-1	Log10 sICAM-1	Log10 sE-selectin	Log10 sP-selectin
R2	0.40	0.29	0.28	0.39
n	129	125	112	99
Haemolysis	LDH (<0.001)
Hepatocellular			Log10 ALT (0.031)
Cholestasis	Direct bilirubin (0.003)	Direct bilirubin (0.003)	Log10 direct bilirubin (0.025)
Renal dysfunction	BUN (<0.001)
Bone damage		Alkaline phosphatase (<0.001)
Platelets				Platelet count (<0.001)
Lipid		Log10 triglycerides (0.019)	Log10 triglycerides (<0.001)	Triglycerides (0.003)
Inflammation			CRP (0.038)	ESR (0.020)
Treatment				Hydroxyurea (0.007)

Dependent variables are included for which P < 0.05 in final model. P-values are given in parentheses. All associations are positive, except hydroxyurea treatment, which is negatively associated with log₁₀ sP-selectin.

In regression models using ranks for all variables, the following variables did not have P < 0.05: for sICAM-1, triglycerides (P = 0.08); for sE-selectin, CRP (P = 0.26) and direct bilirubin (P = 0.15); for sP-selectin, ESR (P = 0.0503). BUN, blood urea nitrogen.

Associations with sICAM-1 levels were similar to some extent to those found with sVCAM-1. Log₁₀ sICAM-1 was associated predominantly with multiple markers of liver function, as well as plasma myeloperoxidase. Multiple regression modelling suggested that the levels of direct bilirubin, alkaline phosphatase, and possibly log₁₀ triglycerides were independent variables linked to sICAM-1 (Table III). There was no evidence of departure of residuals from normality (P = 0.68 by Shapiro–Wilk test). The direct bilirubin level appeared to be indicative of a cholestatic hepatic dysfunction similar to the sVCAM-1 linkage, but it is uncertain whether the source of alkaline phosphatase in this case was hepatic or bone, since we did not perform biochemical fractionation. Because alkaline phosphatase was statistically associated with log₁₀ sICAM-1 independently of the association of direct bilirubin to log₁₀ sICAM-1, it seemed likely that the elevation of alkaline phosphatase in patients with higher sICAM-1 levels was because of pathology in the bone, potentially vaso-occlusive damage. Supporting this interpretation, other investigators have found that most of the variation in alkaline phosphatase activity in patients with SCD at steady state was contributed by the bone isoform (Mohammed et al, 1991).

Log₁₀ sE-selectin was positively associated with hepatic dysfunction, reticulocyte count, white blood cell count, and triglyceride level (Table II). In multiple regression modelling, sE-selectin was linked independently to C-reactive protein and to log₁₀ values of alanine transaminase (ALT), direct bilirubin, and triglyceride (Table III). However, there was a suggestion of non-normality of residuals (P = 0.10, Shapiro-Wilk test), and the associations with C-reactive protein and direct bilirubin were not supported by regression on ranks. These results suggested that in these sickle cell patients, sE-selectin was related to an elevated triglyceride level and hepatocellular injury possibly with cholestatic hepatic dysfunction and inflammation.

Finally, sP-selectin is primarily related to the number of platelets, as it serves as a marker of platelet activation. The level of sP-selectin was also positively related to ALT, white blood cell count, and triglyceride level, and negatively associated with erythrocyte sedimentation rate (ESR) and oxygen saturation (Table II). In a multiple regression model, log₁₀ sP-selectin was independently positively associated with platelet count and triglyceride level, and negatively associated with ESR and hydroxyurea treatment (Table III). Although there was evidence of non-normality of residuals (P = 0.05, Shapiro-Wilk test), these associations were also found in regression on ranks. This suggested that sP-selectin, though related primarily to platelet number, may also be related to inflammation and, as for sICAM-1 and sE-selectin, to triglyceride level.

Effects of hydroxyurea

Hydroxyurea (hydroxycarbamide) therapy was significantly associated with lower levels of sVCAM-1 (median 897.5 ng/ml in 78 patients not on therapy vs. 744.5 ng/ml in 54 patients on therapy, P = 0.006 by Wilcoxon rank-sum test), sE-selectin (median 77 vs. 61 ng/ml, P = 0.01), and sP-selectin (median 174 vs. 145.5 ng/ml, P = 0.02). It was also associated with a non-significant reduction in sICAM-1 (median 197.5 vs. 184 ng/ml, P = 0.49). Patients with chronic renal failure had very high levels of sVCAM-1; consistent with common clinical practice, these patients were not treated with hydroxyurea, skewing the association of hydroxyurea treatment with markers of renal function. When patients with serum creatinine levels >90 μmol/l were excluded, the reduction in sVCAM-1 was smaller but still significant (median 866.5 vs. 744.5 ng/ml, P = 0.03, Table IV). The relationship of hydroxyurea treatment to the other soluble adhesion molecules was essentially unchanged by this adjustment for renal function (Table IV). Since it appeared that the sVCAM-1 was associated with clinical severity and since it was standard clinical practice to treat only the more severely affected patients with hydroxyurea, the treatment group probably had higher baseline sVCAM-1 levels than the untreated patients. Therefore, these population data probably underestimate the true magnitude of the treatment effect. Our data confirm and extend previous measurements in a small cohort of SCD patients (Saleh et al, 1999).

Table IV.

Hydroxyurea and soluble adhesion molecule expression. The analysis excludes sickle cell patients with serum creatinine >106.1 μmol/l, above which no patient was prescribed hydroxyurea.

	Off Hydroxyurea		On Hydroxyurea
	n	Median (25th and 75th percentile)	n	Median (25th and 75th percentile)	P-value*
sVCAM-1	78	866.5 (685.5, 1056.5)	54	744.5 (625, 974.75)	0.03
sICAM-1	76	196 (141.75, 259)	51	184 (138, 240)	0.69
sE-selectin	74	77 (60, 93)	49	61 (42, 88)	0.02
sP-selectin	75	174 (133, 241)	54	145.5 (105.75, 202)	0.01

^*From Wilcoxon rank-sum test.

Relationship of soluble adhesion molecule levels to pulmonary hypertension

The levels of each of the soluble adhesion molecules were compared with the TRV, an echocardiographic measure of pulmonary artery pressure (PAP). In this population, right heart catheterisation studies have confirmed that TRV <2.5 m/s corresponded to normal PAP, TRV 2.5–2.9 m/s corresponded to mild pulmonary hypertension, and TRV >2.9 m/s corresponded to moderate to severe pulmonary hypertension (Gladwin et al, 2004). The Spearman rank correlation coefficient with TRV was not significant for any of the adhesion molecules (Table II) but was nearly so for VCAM-1 (P = 0.053). However, there was a positive trend with level of pulmonary hypertension for each of the four molecules (Fig 2). The trends were significant by a test for linearity with Kruskal–Wallis ANOVA for sVCAM-1 (P = 0.009) and sICAM-1 (P = 0.02); P = 0.054 and 0.10 for sE-selectin and P-selectin, respectively. Thus, TRV was associated with multiple markers of endothelial activation in patients with SCD. We did not find evidence that myeloperoxidase was associated with pulmonary hypertension in SCD patients.

Fig 2.

Soluble adhesion molecules in patients with sickle cell disease with and without pulmonary hypertension. The level of sVCAM-1 is higher in patients with TRV 2.5–2.9 m/s (mild pulmonary hypertension, PHT) and in patients with TRV >2.9 m/s (moderate to severe pulmonary hypertension) than in patients with TRV <2.5 m/s (normal range). The difference was significant by test for linearity with Kruskal–Wallis ANOVA (n = 144, P = 0.009). Horizontal bars indicate median values. Similar trends are seen for sICAM-1 (n = 139, P = 0.02), sE-selectin (n = 135, P = 0.054), and sP-selectin (n = 140, P = 0.10).

Relationship of soluble adhesion molecules to mortality rate

Among study patients without haemoglobin SC for whom soluble adhesion molecule data were available, 15 had died by May 2005, with median follow-up of 34 months for all 150 subjects. Values of sVCAM-1 and sP-selectin were available for all 15 subjects who died, sICAM-1 for 14 and sE-selectin for 11. Cox proportional hazard regression analysis showed a significant relationship of mortality to log₁₀ sICAM-1 (P < 0.001) and log₁₀ sE-selectin (P < 0.001) at study entry, and a similar trend for log₁₀ sVCAM-1 (P = 0.07) (Table V). Mortality was negatively and not significantly related to log₁₀ sP-selectin values (P = 0.36). Kaplan–Meier curves suggested differences in mortality for patients in intervals that correspond approximately to the highest, lowest and middle two quartiles of sICAM-1 (χ² = 17.4, d.f. 1, P < 0.0001 by logrank test) and sE-selectin (χ² = 9.2, d.f. 1, P = 0.002 by logrank test) levels (Fig 3). These results suggested that markers of endothelial activation were associated with risk for early mortality in patients with SCD.

Table V.

Relationship of soluble adhesion molecule levels to risk of mortality. Results of separate Cox proportional hazards regression analyses for each type of adhesion molecule are shown.

Independent variable	n	Number of deaths	P-value*	Risk ratio (RR)† (95% confidence interval for RR)
Log10 sVCAM-1	148	15	0.066	1.7 (0.98, 3.1)
Log10 sICAM-1	143	14	<0.001	2.9 (1.7, 4.9)
Log10 sE-selectin	139	11	<0.001	4.2 (2.0, 8.9)
Log10 sP-selectin	144	15	0.36	0.7 (0.4, 1.4)

^*Likelihood ratio test.

^†RR is given for 75th relative to 25th percentile, calculated as e^{coefficient\times (75th percentile-25th percentile)}.

Fig 3.

Survival, sICAM-1 and sE-selectin values. Shown are Kaplan–Meier plots of survival fraction for three subgroups of SCD patients by adhesion molecule level: Low, representing the quartile of patients with the lowest soluble adhesion molecule levels; Mid, representing the second and third quartiles combined; and high, representing the quartile with the highest soluble adhesion molecule levels. (A) sICAM-1 level: Low (<141 ng/ml), Mid (141–256 ng/ml), High (>256 ng/ml) (Logrank test for trend, P < 0.0001). (B) sE-selectin levels: Low (<53 ng/ml), Mid (53–93 ng/ml), High (>93 ng/ml) (Logrank test for trend, P = 0.002). The number of patients at risk is shown below the horizontal axis at 10 month intervals.

Discussion

VCAM-1 and ICAM-1 are normally expressed at low levels on the luminal surface of endothelial cells. Their expression is induced by a variety of biological stimuli, particularly by inflammatory cytokines, such as TNFα. VCAM-1 and ICAM-1 provide an adhesive surface for specific ligands present on the surface of leucocytes physiologically recruited as part of the inflammatory programme. In contrast, as part of the biochemical program that facilitates blood flow and maintains vascular homeostasis, NO normally suppresses expression of VCAM-1, ICAM-1 and E-selectin (De Caterina et al, 1995; Khan et al, 1996; Shin et al, 1996; Spiecker et al, 1998; Lee et al, 2002). The impairment of NO bioavailability, commonly called endothelial dysfunction, results in pathological activation of endothelial cells to express adhesion molecules. Subsequent shedding of soluble adhesion molecules into blood plasma therefore can serve as markers either of endothelial dysfunction or of inflammation with endothelial activation. The measurement of soluble adhesion molecules in the plasma of a well-characterised population of patients with SCD enabled us to evaluate the relationship between endothelial dysfunction and haemolytic rate, hydroxyurea therapy, patient gender, and the development of pulmonary hypertension, end-organ insufficiency and death.

Our results, derived from a large unselected population of adult patients with SCD, indicate that an elevated level of soluble adhesion molecules is a marker related to disease severity, as evidenced by its association with early mortality and organ dysfunction, including pulmonary hypertension, liver disease and renal dysfunction. In particular, pulmonary hypertension is a complication of SCD that is rapidly emerging as an important predictor of early mortality (Gladwin et al, 2004). Consistent with our prior observation that SCD patients with higher than median levels of sVCAM-1 also have impaired vascular reactivity (Gladwin et al, 2003), these data suggest that patients with higher sVCAM-1 levels are also at greater risk for pulmonary hypertension and early mortality. Higher than median sVCAM-1 levels might be used to identify patients at high risk of endothelial dysfunction, or to stratify statistical analyses for this risk. There are few well-documented prognostic indicators in SCD, and the potential utility of plasma levels of sVCAM-1, sICAM-1 and sE-selectin to identify high-risk patients merits further investigation.

The association of endothelium-derived soluble adhesion molecules with pulmonary hypertension in our patients is consistent with our previous evidence suggesting that endothelial activation and dysfunction play a biological role in pulmonary hypertension in patients with SCD. Since NO is known to suppress VCAM-1, ICAM-1 and E-selectin expression, it is attractive to hypothesise that in part, elevated levels of their soluble derivatives indicate impaired NO bioavailability (Khan et al, 1996). We have previously linked elevation of sVCAM-1 to excessive NO consumption and impaired NO-dependent vasodilator response in patients with SCD, and now in our present data to both pulmonary hypertension and early mortality (Gladwin et al, 2003). The linkage of sVCAM-1 levels to markers of haemolysis supports the thesis that plasma cell-free haemoglobin released during intravascular haemolysis is an important factor in NO consumption and endothelial dysfunction in SCD (Reiter et al, 2002). The most powerful association in this regard is between sVCAM-1 and lactate dehydrogenase, the latter widely regarded by clinicians as the most sensitive clinical laboratory marker of the intravascular form of haemolysis that liberates haemoglobin into blood plasma (Neely et al, 1969). This same association has been seen in earlier studies with smaller patient populations (Reiter et al, 2002; Schnog et al, 2003; Setty et al, 2003).

Our association of elevated levels of soluble adhesion molecules to increased haemolytic rate, pulmonary hypertension, low oxygen saturation and endothelial dysfunction in patients with SCD is highly consistent with recent observations by Setty et al 2003. They found a close linkage in children with SCD between low haemoglobin–oxygen saturation and severity of haemolytic anaemia, indicated by low haemoglobin and high reticulocyte count. They also found that low oxygen saturation is related to abnormal endothelial activation, as shown by elevated plasma levels of soluble VCAM-1, L-selectin and P-selectin and by red cell adhesion to cultured endothelial cells. The exact pathophysiological effects of haemolysis and anaemia on lung perfusion and ventilation-to-perfusion matching require further study.

Thus, it appears that severe haemolytic anaemia may underlie a constellation of physiologically linked findings, namely hypoxaemia, pulmonary hypertension, NO resistance and increased plasma levels of soluble adhesion molecules (Rother et al, 2005). The contribution of this vasculopathy to blood flow impairment by haemoglobin S polymerisation and cell adhesion merits further investigation.