Background
Thrombospondin-1 (TSP-1) is a matricellular protein, which is involved in various processes of inflammation, neovascularization, hypoxia and remodeling [
1‐
3]. A major source of TSP-1 are the alpha-granules of platelets [
4]. Nevertheless, it is expressed in numerous cell types like macrophages, endothelial cells, vascular smooth muscle cells and cardiomyocytes [
5‐
7]. Its production increases dramatically under conditions of hypoxia, altered shear stress or exposure to growth factors including platelet derived growth factor and basic fibroblast growth factor [
8]. These conditions are all present in pulmonary hypertension. In vitro and in vivo studies revealed three mechanisms of action of TSP-1. First, binding to CD47 disrupts the nitric oxide pathway and vascular endothelial growth factor pathways based on the inhibition of soluble guanylyl cyclase and protein kinase G in vascular smooth muscle cells [
9]. Second, in higher concentrations TSP-1 binds and activates CD36, thereby decreasing intracellular cGMP-levels due to interaction with cellular uptake of myristic acid and decreased eNOS-dependent synthesis of nitric oxide [
10]. The net result of TSP-1 binding to these receptors is induction of apoptosis and abortion of VEGF and nitric oxide signaling [
11‐
13]. Finally, TSP-1 is able to exert its biological effects by an additional mechanism involving the activation of transforming growth factor ß (TGF-ß) leading to excessive fibrosis [
14]. Therefore, TSP-1 is able to diminish the effects of specific therapeutics of PH and promote disease progression of PH.
In this study we investigated the possible associations of circulating TSP-1 with clinical as well as hemodynamic variables and survival in PH. TSP-1 levels were compared between a control group without PH, patients with pulmonary arterial hypertension (PAH), lung disease associated PH (LD) and chronic thromboembolic PH (CTEPH). Furthermore, humoral cofactors of TSP-1 release were measured and analyzed.
Methods
Participants
Between June 2007 and December 2010, we enrolled prospectively 19 control subjects and 93 patients prospectively from the outpatient clinic for PH at the Department of Pulmonology of the Saarland University Hospital (Homburg / Saar, Germany). The control group consisted of subjects without evidence of any internal disease. In nine out of nineteen controls a right heart catheterization was performed for exclusion of pulmonary hypertension while in nine persons without any medical history of chronic disease or respiratory symptoms, PH was considered absent. Cardiopulmonary function tests including echocardiography and right heart cathererization were within normal range. The patients had incident and prevalent diagnosis of PH and had diagnostic workup according to current guidelines. Cardiovascular diseases aside from PH were excluded prior to inclusion into the study. Diagnosis of PH was based on a mean pulmonary artery pressure above 25 mmHg (PAPm) and a pulmonary artery occlusion pressure (PAOP) below 15 mmHg. Cardiac output was determined by thermodilution method. All patients received pulmonary ventilation/perfusion scintigraphy to diagnose CTEPH. Patients with suspected CTEPH had confirmatory pulmonary angiography. The inclusion into the study was irrespective of operability and plasma samples were taken before planned pulmonary end-arterectomy. Both incident as well as prevalent diagnosis of PH was included. Patients were at least two months clinically and hemodynamically stable before enrollment as indicated by unchanged functional class, lack of hospitalization and stable medication.
The study complies with the declaration of Helsinki and was approved by the local ethics committee (Ethikkommission der Ärztekammer des Saarlandes, Nr. 153/11). Written informed consent was obtained from each patient.
Measurements
Blood was drawn during right heart catheterization as initial work-up or follow-up. For platelet stabilization, blood samples were collected into tubes containing 1 mL volume of citrate 109 mmol/l, theophylline 15 mmol/l, adenosine 3.7 mmol/l, dipyridamole 0.198 mmol/l. Plasma was separated by immediate centrifugation (2700xg at 4 °C, 20 min) and stored at −80 °C for batch analysis. The local clinical laboratory determined routine parameters.
TSP-1 levels were measured by ELISA in duplicates according to the manufacturers instructions (Quantikine human TSP-1 ELISA, R&D, Minneapolis, USA). Optical density was determined using a Tecan Spectra III reader set to 450 nm and 620 nm reference wavelength. The TSP-1 assay had a minimal detection limit of 0.355 ng/mL. The within run repeatability coefficient of variation was 5.1 %. Inter-assay repeatability was calculated with 11.3 % resulting in an overall variability of 8.2 %.
PGDF-β platelet poor plasma levels were measured by the Quantikine Human PDGF-β Immunoassay (R&D Systems, Minneapolis, USA, Catalog Number DBB00). Big-Endothelin ELISA was obtained from the Biomedica Group (Vienna, Austria, Catalog Number BI-20082), SDF1a and PF-4 ELISA from Ray Biotech, Inc. (Catalog Number ELH-SDF1alpha-001, ELH-PF4-001). Measurements were performed as indicated by the manufacturer.
Statistics
Continuous variables were tested for normality and variance homogeneity by Shapiro-Wilk test and Hartley´s Fmax test respectively. Non-normally distributed variables are expressed as median (range), normally distributed variables as mean ± SD unless indicated otherwise. Categorical variables are expressed as numbers (percentage). Outliers were defined outside a three-SD interval from mean of the whole dataset. Differences between groups were tested by ANOVA with post-hoc analysis (Dunnett T3 correction) for normally distributed and Mann–Whitney-U-test for non-normally distributed variables. Comparison of categorical variables was performed by Chi-square-test.
Regression analysis was studied for associations of continuous variables. For bivariate regression a non-parametric model was chosen with locally linear curve fitting of the dataset.
Kaplan-Meyer-Analysis for death was performed with an optimized cut-off of TSP-1 by minimized Matthews coefficient. Hazard ratios were calculated using a cox proportional hazard model (efron method). A p-value of less than 0.05 (two-tailed) was considered statistically significant in all tests.
Data storage and processing, calculations, as well as statistical computations were performed by R-Project version 3.2.0.
Discussion
The presented study showed, that levels of circulating TSP-1 differ significantly between patients with PH compared to subjects without PH. The subgroup analysis demonstrated the importance of the specific subtype of PH for the interpretation of this biomarker. Furthermore, associations of circulating TSP-1-levels and pulmonary hemodynamics were elaborated. And finally, the association of elevated TSP-1 with increased mortality became evident.
As Bauer and Isenberg showed, TSP-1 is abundantly expressed in lung tissue of patients with PH. In pulmonary endothelium, CD47 controls endothelial NO-synthase. Binding of TSP-1 uncouples eNOS via CD47 and promotes progress of hypoxic PH in animal models [
15]. Furthermore, TSP-1 has been shown to deteriorate shear stress dependent vasodilation under hypoxic conditions and is induced by hypoxia in pulmonary vascular endothelium [
16].
In this study we demonstrated a significant increase in plasma levels of TSP-1 in patients with various types of PH. While levels of TSP-1 varied in different subtypes of PH, the common denominator was their hemodynamic state. The univariate non-linear regression analysis showed a significant association of circulating TSP-1 with PVR and CO. This supports the initial hypothesis, that the increase of PVR increases shear stress resulting in a release of TSP-1 from pulmonary endothelial cells. However, hemodynamic parameters to determine shear stress in humans remain scarce. The extent of shear stress in a specific pulmonary vessel is influenced by both right ventricular function and properties of the vascular bed of the lung. PVR and cardiac output remain the best surrogate variables of standard right heart catheterization for estimation of shear stress.
On the other hand it has been shown, that TSP-1 production is minimal at optimal shear rates [
17], explaining nonlinear U-shaped relation of shear stress and TSP-1 release. In this study we showed first clinical evidence for this complex relationship (Fig.
3), but had to omit possibly confounding factors (i.e. age, medication) due to limited data points.
Induction of TSP-1 in endothelial cells has been shown previously [
18] and was confirmed for PAH [
16]. Our data contrasts these findings, as an association of TSP-1 with mixed central venous oxygen saturation, which reflects peripheral oxygen depletion, could not be confirmed. Furthermore, the vascular endothelium might not be the only source of TSP-1 in PH, as increasing right ventricular load might release TSP-1 from cardiomyocytes as hemodynamics deteriorate. The release of TSP-1 from the myocardium especially due to ischemia-reperfusion injury has been shown previously [
19] and might limit specificity of TSP-1 as a marker for diagnosis of PH.
Experimental data supported the hypothesis of TSP-1 interfering with the NO-pathway [
20]. This abrogates soluble guanylate cyclase and protein kinase G as well as adenylate cyclase [
21‐
23], resulting in direct pulmonary arterial vasoconstriction, disease progression [
16] and disruption of key pharmacological targets of specific therapy.
Examination of drug regimes suggested lower TSP-1 levels on treatment with PDEIs, higher levels on PC-treatment and intermediate levels under ERA treatment. This observation might lead to the assumption, that treatment regime was influential on circulating TSP-1. On the other hand, PC therapy is an invasive therapeutic concept and often withheld until treatment goals are not sufficiently achieved by oral medications. Therefore, these observations might reflect the finding, that TSP-1 was elevated with deteriorating hemodynamics and increased functional class. The influence of medical therapy on TSP-1 levels could not be determined in this study and warrants longitudinal trials observing TSP-1 levels before and after initiation of new drug regimes.
As this study was designed as a cross-sectional observation, we included patients in different stages of disease. While some had newly diagnosed PH, others were in the final stage of their illness. Thus a non-matched comparison of TSP-1 levels between groups seems a serious limitation for this study. The comparison of incident and prevalent PH demonstrated slightly higher levels of TSP-1 in incident PH, statistical significance was not reached. While the groups had similar hemodynamics, the influence of medication, the course of disease and adaptive processes might lead to this finding. The reasons and possibly influence on the prognostic value of TSP-1 was beyond the scope of this cross-sectional pilot study.
Based on the data linking TSP-1 to hemodynamics and the possibility of its negative impact on pathophysiological mechanisms of PH, we sought to gain insight of the prognosis attached to TSP-1 as a biomarker. At the end of the study, survivor and non-survivors were compared and in all groups non-survivors had significantly higher levels of TSP-1 when entering the study (Fig.
4). CTEPH patients demonstrated relatively higher levels in both survivors and non-survivors compared to the other subgroups. It is therefore mandatory to exactly classify the type of PH before using TSP-1 as a potential biomarker.
Survival analysis in this study was based on follow-up in regular intervals up to five years. The patients had received standard therapy according to current guidelines. Cardiovascular mortality was chosen as the primary endpoint, but in this dataset it was identical with all cause mortality. We were able to show a significantly increased mortality in patients with high TSP-1 levels. This data is limited by the sample size of 93 patients and inclusion of further confounding factors or multivariate analysis with established predictors of mortality is obsolete. Nevertheless, the optimized cut-off at 2051 ng/mL yielded a highly significant difference between groups. Though, as discussed above, the circulating levels of TSP-1 might vary between subgroups and therefore a different cut-off can be expected especially in patients with CTEPH. Furthermore, serial measurements are needed to evaluate the true impact on survival and various clinical events in more complex frailty models.
Nevertheless, this data provides valuable information on the relation of TSP-1 with pulmonary hemodynamics, a clinical application of in vitro observations considering crucial pathways in PH and indicates an impact on the prognosis of a serious disease. These observations are the clinical implication of previous studies on the role of TSP-1 in experimental PH [
15,
24].
The presented data might form the basis for evaluation of TSP-1 as an indicator for the choice of medication and might facilitate therapeutic decisions. TSP-1 interacts significantly with the NO pathway, thereby leading to uncoupling of specific therapeutic drugs [
22]. Observational studies should evaluate TSP-1 as a marker for possible non-responders before initiation of therapy.
Abbreviations
CI, cardiac index; CTEPH, chronic thrombembolic PH; LD, lung disease associated PH; NO, nitric oxide; PAH, pulmonary arterial hypertension; PAOP, pulmonary artery occlusion pressure; PAP, pulmonary artery pressure; PDGF, platelet derived growth factor; PF-4, platelet factor 4; PH, pulmonary hypertension; PVR/I, pulmonary vascular resistance/index; RAP, right atrial pressure; SAP, systemic artery pressure; SDF-1, stromal derived factor 1; TGF, tissue growth factor; TSP-1, Thrombospondin-1; VEGF, vascular endothelial growth factor