ACA: Anti-Centromere Antibodies; ANA: Anti- Nuclear Antibodies; dcSSc: Diffuse Cutaneous Scleroderma; ELISA: Enzyme-Linked Immunosorbent Assay; EMP: Endothelial Microparticle; ILD: Interstitial Lung Disease; lcSSc: Limited Cutaneous Scleroderma; PAH: Pulmonary Arterial Hypertension; PAP: Pulmonary Arterial Pressure; PMP: Platelet Microparticle; RVSP: Right Ventricular Systolic Pressure; sAng-1: Soluble Angiopoietin 1; sAng-2: Soluble Angiopoietin 2; sE-selectin: Soluble E-selectin; sICAM-1: Soluble Inter-Cellular Adhesion Molecule-1; sL-selectin: Soluble L-selectin; SSc: Scleroderma (Systemic Sclerosis); sTie-2: Soluble Endothelial Cell Specific Tyrosine Receptor 2; sVCAM-1: Soluble Vascular Cell Adhesion Molecule 1; TTE: Trans-Thoracic Doppler-Echocardiography; VEGF: Vascular Endothelial Growth Factor; VEGF-R1: Receptor 1 for Vascular Endothelial Growth Factor

Insufficient or absent angiogenesis are hallmarks of scleroderma (SSc) [1]. Microvascular change is an early manifestation of SSc followed by intimal proliferation and fibrosis of arterioles resulting in reduced blood flow and tissue ischemia [2]. This ongoing vasculopathy in the lungs presents clinically as pulmonary hypertension and characteristically precedes lung fibrosis [3].

With tissue hypoxia and vascular damage, vascular healing and new blood vessels are required [4]. Angiogenesis is dependent on a tight balance of angiogenic and angiostatic factors [5]. Different proangiogenic stimuli are present in SSc but an appropriate angiogenic response is lacking [6]. In a hypoxic environment, growth factors outweigh inhibitors with angiogenesis the result [7]. Vascular endothelial growth factor (VEGF) is a hierarchical upstream inducer of angiogenic cascade [8-11]. VEGF synthesis is paralleled by increased activation and apoptosis of endothelial cells. This is illustrated on a molecular level by biomarkers inclusive of inter-cellular adhesion molecule 1 (ICAM-1) and the vascular cell adhesion molecule 1(VCAM- 1) [12]. Circulating microparticles regulate inflammation, affect cell proliferation and apoptosis [13]. They were shown in a previous study to be elevated in 10 dcSSc and 27 lcSSc patients combined compared to 15 healthy controls (p<0.05), suggesting that microparticles attenuate fibrotic disease manifestations [14].

A second set of regulatory molecules, the endothelial cell specific receptor tyrosine kinases Tie-1 and Tie-2, and their ligands angiopoietin-1 (Ang-1) and angiopoietin-2 (Ang-2), also regulate both vasculogenesis and angiogenesis [15]. Ang-1 is expressed by many cells including pericytes, smooth muscle cells and fibroblasts [16], while Ang-2 is expressed almost exclusively by endothelial cells. Ang-1 mediated Tie-2 signaling is the adult default pathway to control vascular quiescence. Ang-2 antagonizes Ang-1 and responds to endothelial stimulation by several factors including VEGF, fibroblast growth factor-2, tumor necrosis factors and hypoxia [17]. Ang-2 may be the principal switch controlling the transition from resting to activated endothelium and facilitating inflammatory response [18].

Yu et al. [19] have demonstrated in vivo and ex vivo that E-selectin is required in endostatin-mediated anti-angiogenesis and confers endostatin sensitivity to nonresponsive human endothelial cells in vitro.

Tissue ischemia normally leads to up-regulation of angiogenic factors which, through the proliferation of endothelial cells, aids in vascular remodeling. Despite reduced organ perfusion in SSc, there is paradoxically no increase in angiogenic activity to compensate for this deficiency [20]. The more extensive and severe disease occurs in the diffuse subset. At least one-third of patients have clinically significant fibrosis and lung function impairment.

The purpose of this preliminary study is to elucidate in SSc patients possible interrelationships amongst circulating micro particles and angiogenic and angiostatic factors that are predictive of lung fibrosis and high values of right ventricular systolic pressure (RVSP) by transthoracic Doppler-echocardiography (TTE).

Study population

Thirty cases with SSc and 30 healthy controls gave informed consent. Research was done in compliance with the Declaration of Helsinki and approved by the Research Ethics Boards of the University of British Columbia and the University of Northern British Columbia. All patients fulfilled the American College of Rheumatology Criteria for SSc and were classified as having limited cutaneous (lcSSc) or diffuse cutaneous (dcSSc) disease according to the criteria of LeRoy et al. [21]. The study protocol excluded cases or age- and sex-matched healthy controls with hypertension, known coronary artery disease, known cardiovascular disease, diabetes, lung disease or active infections at the time of phlebotomy. Cases with overlap syndromes or mixed connective tissue disease were excluded from the study. No recruited healthy control was excluded and the sole case excluded was a dcSSc male with hypertension.

Clinical variables

Disease duration was measured from the onset of the first non- Raynaud’s phenomenon. Skin involvement was assessed using the modified Rodnan skin score (mRSS) at 17 different body areas and given a score from 0-3 for each area [22]. Disease severity for each patient was evaluated on a scale of 0-4 for each of nine organ systems using the Medsger scale [23] and then summed.

Pulmonary function was examined but only predicted forced vital capacity (FVC) and the predicted diffusing level of carbon monoxide (DLCO) are reported. Interstitial lung disease (ILD) was defined as the presence of ground glass opacities, fibrosis or honeycombing. It was determined from high resolution computed tomography (HRCT) [24].

Study patients underwent HRCT of the chest with slice thickness of 1.25 mm taken at an interval of 0.8 mm at 100 kV and 360 mA. The scanning was performed through the entire thorax with the subject in a supine position with breath holding in inspiration. The HRCTs were read by an experienced chest radiologist blinded to the clinical and lab data and reviewed by us. Ground glass opacity was defined as a hazy parenchymal opacity with preservation of bronchial and vascular margins. Fibrosis was defined as interlobular septal thickening, intralobular lines, traction bronchiectasis and traction bronchiolectasis. Honeycombing was defined as clustered air filled cysts with well defined walls. Formal scoring was not used in this study.

TTE was used for RVSP and Doppler flow measurement of the TR jet to calculate pulmonary arterial pressure (PAP). Pressures less than 35 mm Hg were considered normal. Echo Doppler studies have been shown to correlate well with right heart catheter studies [25], but tend to underestimate pulmonary artery pressure and cardiac output [26]. All clinical variables were recorded at time of plasma collection or within two months prior.

Microparticle concentrations

Blood samples were collected in buffered sodium citrate and centrifuged at 1 500 g for 15 minutes. The plasma was removed and spun at 12 700 g for 12 minutes. Without disturbing the pellet, the platelet free plasma (PFP) was removed and stored at -80°C. In brief, 100 ml of PFP was incubated (20 minutes) with 20 ml of anti-CD 106- FITC (BD Bioscience, Mississauga, Canada), 7.2 ml of anti-CD 144-PE (eBioscience, San Diego, California, USA), and 20ml of anti-CD41- PECy5 (BD Bioscience). Mouse immunoglobulin (Dako Canada, Burlington, Canada) was used as a negative control. Samples were then incubated with 10 μl of Annexin V-APC (BD Bioscience) and 100 ml of CACl2 (5 mM) binding buffer for 15 minutes. Finally, 250 ml of binding buffer and 100 ml of Flow-Count counting beads (Beckman Coulter Canada, Mississauga, Canada) were added to each tube. Samples were placed on ice and analyzed immediately using a MoFlo high speed cell sorter (Beckman Coulter Canada). Microparticles were defined as having a size of 1 μm or less. With a forward- and side-scatter dot plot, the upper boundary of the analysis region was set using a 1.0 μm fluorescent microsphere (Polysciences, Warrington, Pennsylvania, USA). A 0.5 μm microsphere was used to ensure visualization of events of this size or smaller. Gated events were then interrogated for expression of labeled antibodies. Platelet microparticles (PMPs) were defined as CD41⁺ MPs. Endothelial microparticles (EMPs) were defined as CD41^- /CD144⁺ MPs. To confirm EMPs, samples were further evaluated for the expression of activation (CD106⁺) or apoptosis (Annexin V⁺) markers (data not shown). Sample analysis was stopped at 50 000 MPs. The numbers of EMPs and PMPs per microliter were calculated as per the Flow-Count product insert (Beckman Coulter Canada).

Circulating factors

All plasma samples were aliquoted and stored at -80°C. Plasma levels of soluble Angiopoeitin-1 (sAng-1), Angiopoeitin-2 (sAng-2) and sTie-2 (lower case “s” prefix denotes soluble) were measured by commercially available quantitative colorimetric sandwich enzymelinked immunosorbent assay (ELISA) kits following the manufacturer’s protocol (R&D Systems, Minneapolis, Minnesota, USA). Endostatin was measured by a sandwich ELISA (R&D Systems, Minneapolis, Minnesota, USA). VEGF, soluble intercellular adhesion molecule one (sICAM-1), soluble vascular cell adhesion molecule one (sVCAM-1), sE-selectin and sL-selectin were measured using Searchlight Custom Arrays (Pierce Searchlight Products, Thermo Fisher Scientific Woburn, Massachusetts, USA) and analyzed using a 16-bit Searchlight CCD imaging system (Pierce Biotechnology, Rockford, Illinois, USA).

Statistical analysis

Estimates of location are reported as a mean ± SE. Given small sample sizes and non-normality for most factors, nonparametric methods were used for comparisons. Statistical calculations and graphing were done using release 2.11.1 of the R statistical software system [27]. The stats package in R was used to calculate the exact P values for the Wilcoxon test of no shift in location for matched casecontrol pairs. The coin package in R was used to calculate the exact P values for the Mann-Whitney test for location with two independent random samples [28]. Classification trees for ILD and PAH as an exploratory investigation were produced using the rpart package [29]. Because of small sample sizes, the accuracy of correct classification for each tree was assessed with the delete-one jackknife. Caution in interpretation of the accuracy estimates for classification trees is advisable because of the small sample sizes.

Demographics and clinical features of patients

Blood samples were available for this preliminary study of 19 patients diagnosed with lcSSc and 11 with dcSSc (Table 1). No gender difference in age with the healthy controls was detected for lcSSc patients (p=0.0917) or for dcSSc patients (p=0.6970). Combining genders, patients diagnosed with lcSSc had a mean (±SE) age of 58.53 ± 2.50 years (range: 36-74 years) while patients diagnosed with dcSSc had a mean age of 49.64 ± 3.17 years (range: 30-65 years). The difference in median age between lcSSc and dcSSc is statistically significant (p=0.0399).

Table 1: Demographics.

There was no difference in location of the distribution of maximum disease severity scores between women and the single man with lcSSc (p=0.8889) or amongst the female and male dcSSc patients (p>0.9999). Combining genders, patients diagnosed with lcSSc had a mean (±SE) maximum disease severity index of 7.00 ± 0.78 (range: 2-16). Patients diagnosed with dcSSc had a mean disease severity index of 7.18 ± 0.91 (range: 2-11). There is no statistically significant difference between lcSSc and dcSSc cases with respect to location of the disease severity index (p=0.6638) as a consequence of aggressive clinical management or referral bias associated with a tertiary care clinic, or both.

No gender difference in disease duration was detected for lcSSc (p=0.9474) or for dcSSc (p=0.8485) patients. Combining genders, patients diagnosed with lcSSc had a mean (±SE) disease duration of 10.68 ± 2.68 years (range: 1-43 years) while patients diagnosed with dcSSc had a mean disease duration of 2.91 ± 0.48 years (range: 2-11 years). Disease duration is significantly shorter (p=0.0381) in dcSSc cases compared to lcSSc as expected.

No gender difference in mRSS was detected for either lcSSc (p=0.2632) or dcSSc (p=0.7576) cases. Mean (±SE) mRSS was 9.05 ±1.29 (range: 1-20) in lcSSc cases and 22.82 ± 2.82 (range: 4-36) in dcSSc cases. Skin score is worse in dcSSc cases compared to lcSSc cases (p=0.0002).

With regard to auto antibodies, dcSSc and lcSSc have distinctly different profiles for anti-centromere antibodies (ACA), anti-nuclear antibodies (ANA) and anti-topoisomerase antibodies (Scl-70) (Table 1).

On the basis of the preceding statistical analyses, decisions were made to disregard gender but treat limited and diffuse forms of SSc as distinct.

Comparisons of circulating factors among cases and healthy controls

There are insufficient data to estimate the 78 variances and covariances for the 12 circulating factors. It is reasonable to assume dependencies among the circulating factors, so the Bonferroni correction for 24 independent tests with an overall 0.05 level of significance is conservative in the case of unassessed dependencies. Except for the selectins and VEFG Receptor-1 (VEGF-R1), the distributions of factors and microparticles are not normally distributed (Supplementary Table 1).

The lcSSc and dcSSc cases are comparable with the following exceptions. The serum concentration of sAng-2 in lcSSc is statistically significantly greater than in the matched controls (p<0.0001÷2=0.00005) (Table 2). This is consistent with an activated endothelium facilitating inflammatory response in lcSSc patients. Endostatin levels are higher in both sets of SSc cases compared to their paired controls (Table 2). sEselectin levels are higher in lcSSc patients compared to their controls (p=0.0003÷2=0.0002) but not in dcSSc patients (p=0.0840÷2=0.0420). Amongst the cell adhesion molecules assessed, sE-selectin appears to be the most promising. The only statistically significant result involving microparticles is the lower level of PMPs in lcSSc patients (p=0.0020÷2=0.0010) compared to controls, although the mean for dcSSc patients is even lower.

Table 2: Comparison of mean concentrations of circulating factors among cases and controls.

Overall, the dramatic differences by p-value are higher concentrations of sAng-2, endostatin, sE-selectin, and low PMP in 19 lcSSc patients and higher concentrations of endostatin in 11 dcSSc cases compared to controls.

Comparisons of circulating factors relating to ILD

The prevalence of ILD is 47.4 per cent and 63.6 per cent in the lcSSc and dcSSc cases, respectively (Table 1). In Table 3, the only two circulating factors with concentrations that appear to be different in cases with ILD are sE-selectin which was higher in lcSSc cases (p=0.0535÷2=0.0268) and EMPs which were higher in dcSSc cases (p=0.0121÷2=0.0061).

For a classification tree [30] with the 12 circulating factors for ILD in lcSSc cases, the factors in order of appearance from the top of the tree to the bottom are sE-selectin, sAng-1, and sAng-2 (Figure 1A). The estimate of accuracy of correct classification by the classification tree method is 47.4 per cent. sAng-2 came to our attention in the comparison of lcSSc cases with their controls (Table 2). That sE-selectin has entered the model for the classification tree for lcSSc is not a surprise because of its role in inflammation. The angiopoeitins, especially Ang-2, are important regulators of the inflammatory response.

Figure 1: Classification trees for determining presence or absence of ILD among lcSSc (A) and dcSSc (B) cases. If a statement is true then the branch to the left is taken.

Mean EMP concentrations are more than 4 times higher in dcSSc cases with ILD compared to dcSSc cases without ILD (p=0.0121÷2=0.0061) (Table 3). So it is not a surprise to have EMP as the circulating factor at the top of the classification tree for dcSSc patients (Figure 1B). The estimate of accuracy of correct classification by the classification tree method is 72.7 per cent. Six patients with ILD are picked out at the first branch with a concentration greater than or equal to a count of 95 per μl. The remaining dcSSc patient with ILD is picked out at the second branch with an sAng-1 concentration less than 7.8 ng/ml.

Table 3: Comparison of mean concentrations of circulating factors by ILD and disease type.

The classification trees for lcSSc and dcSSc are distinctly different for ILD. Low concentrations of sE-selectin and sAng-1 with a high concentration of sAng-2 appear to predict an absence of ILD for lcSSc cases, whereas, a low EMP concentration with a high sAng-1 concentration appear to predict an absence of ILD for dcSSc cases. For either lcSSc or dcSSc, the angiopoetins seem to be involved in regulating endothelial activation and inflammation but ILD in lcSSc appears to be a consequence of inflammation associated with E-selectin. In contrast, ILD in dcSSc appears to be characterized by large numbers of circulating endothelial microparticles.

Comparisons of circulating factors relating to pulmonary hypertension

The prevalence of putative PAH is 47.4 per cent and 36.4 per cent in the lcSSc and dcSSc cases, respectively (Table 1). The concordance rates for ILD and high RVSP are 68.4 per cent and 54.5 per cent for lcSSc and dcSSc cases, respectively, so there is not complete overlap of these two clinical conditions (Table 4). No factor appears by the Bonferroni correction to be significantly different with respect to a difference in location whether RVSP ≥ 35 mm Hg, or not, for either lcSSc or dcSSc (Table 5).

Table 4: Distribution of ILD and PAH status.

Table 5: Mean concentrations of circulating factors by disease type and putative PAH status.

The classification tree for putative pulmonary hypertension amongst the 19 lcSSc cases starts with sE-selectin at the top (Figure 2A). The estimate of accuracy of correct classification by the classification tree method 57.9 per cent. Six of the high lcSSc cases have an sE-selectin concentration of at least 37 ng/ml. A sequence of two binary splits at the second branch proceeds through PMP and sICAM-1 to separate the remaining 3 cases with TTE RVSP ≥ 35 mm. PMP first attracted attention in comparison tests with controls (Table 2). Generally, lcSSc cases with low sE-selectin, PMP and sICAM-1 concentrations will have TTE RVSP < 35 mm.

Figure 2: Classification tree for determining putative pulmonary hypertension among lcSSc and dcSSc cases. If a statement is true then the branch to the left is taken.

The classification tree for determining putative pulmonary hypertension amongst the dcSSc cases is simple (Figure 2B): 3 dcSSc cases with putative pulmonary hypertension have a VEGF-R1 concentration of 206 pg/ml or higher. The mean (±SE) VEGF-R1 concentration for dcSSc cases is 178 ± 20 pg/ml. The remaining dcSSc case with putative pulmonary hypertension separates from those without at the second branch with an EMP count less than 56 particles per μl. The estimate of accuracy of correct classification by the classification tree method 63.6 per cent.

Comparing classification trees suggests that the potent inflammatory agent E-selectin is associated with both ILD and high TTE RVSP in lcSSc (Figure 1A and Figure 2A). This is not repeated in the classification trees for dcSSc (Figures 1B and Figure 2B) in which high EMP concentrations are associated with ILD and TTE RVSP lower than 35 mm. So while endothelial microparticles appear responsible for ILD in dcSSc cases they are protective against pulmonary hypertension. The implication is that lcSSc reaches the endpoints of lung fibrosis and pulmonary hypertension through a different biochemical pathway than dcSSc.

With regard to the literature, Blann et al. [31] reported mean concentrations of soluble L-selectin in serum for dcSSc cases (755 ±57 ng/ml), lcSSc cases (961 ± 61 ng/ml) and controls (906 ±35 ng/m). Shimada et al. [32] reported sample means in serum. The estimates reported in this study are for soluble L-selectin in plasma and thus are not comparable with those from either Blann et al. [31] or Shimada et al. [32].

Supplementary Table 2 provides a comparison of estimates of location for VEGF in SSc cases and controls with those of Riccieri et al. [33], Papaioannou et al. [34], Guiducci et al. [14], Hummers et al. [35], Del Papa et al. [36] and Distler et al. [37]. Due to differences in methods it is not possible to perform hypothesis tests between any of these estimates and those of this study. Likewise, we cannot compare our estimate of median VEGF concentration (316 pg/ml, range: 72- 1470 pg/ml) for SSc cases with putative pulmonary hypertension and those without (162 pg/ml, range: 92-382 pg/ml) with estimates of Papaioannou et al. [34]. However, these results and ours are trending to show VEGF concentration to be higher in plasma, or serum, for SSc cases compared to controls and as well higher in SSc patients with a TTE RVSP of 35 mmHg or higher compared to SSc cases with TTE RVSP lower than 35 mmHg.

The literature review for endostatin concentrations revealed no consistency (Supplementary Table 3) [35,37,38]. Statistical tests of comparison are not possible because of the differences in assay methods.

Two articles were found that considered microparticle concentrations in SSc patients. The results of Nomura et al. [39] are not comparable to our results because they used the monoclonal antibody GPIX. The later article by Guiducci et al. [14] counted the number of CD144⁺ microparticles in contrast to the mass of CD41^- /CD144⁺ microparticles reported by us.

There is evidence for greater expression of E-selectin in neutrophil co-cultures from sera of SSc cases compared to controls (p=0.00004) [40]. Our study consistently shows lower ratios for cases to controls compared to the literature for sVCAM-1 and sICAM-1 but not sEselectin (Supplementary Table 4) [36,39,41].

We found plasma levels of sAng-1, sAng-2, and sTie-2 to be higher in SSc cases compared to sex- and age-matched healthy controls. Riccieri et al. [33] examined only sAng-2 in plasma with a difference assay method but also come to the same conclusion (Supplementary Table 5). However, Michalska-Jakubus et al. [42] in considering only sAng-1 and s-Ang-2 in serum found the opposite to be true. Noda et al. [43] found no difference between SSc cases and controls with respect to the concentration of sTie-2 in serum. Nevertheless, ILD in our lcSSc cases is associated with high levels of sAng-1 and sAng-2 (Figure 1). But sAng-1, sAng-2, and sTie-2 are not implicated in putative PAH (Figure 2).

This preliminary study is limited by being cross-sectional on a small number of patients, in measuring only plasma levels of circulating factors and estimating PAP by TTE. Lacking are measurements of profibrotic factors, such as Interleukin 6. Data collection is ongoing.

In attempting to distinguish healthy controls from lcSSc and dcSSc cases, 22 controls can be separated from all but one lcSSc case based upon a cut-off of endostatin concentration below138 ng/ml in a classification tree (Figure 3). In distinguishing the remaining 8 controls from the cases, this is achieved at the last node on the classification tree by an sE-selectin concentration less than 31 ng/ml (Figure 3). This is consistent with the demonstration by Yu et al. [19] in vivo and ex vivo that E-selectin is required in endostatin-mediated antiangiogenesis and confers endostatin sensitivity to nonresponsive human endothelial cells in vitro. The interaction of VEGF and its receptor VEGF-R1 requires further study as does the potential role of L-selectin.

Figure 3: Classification tree for sorting amongst controls, lcSSc and dcSSc cases. If a statement is true then the branch to the left is taken.

In summary, our classification tree analysis in this preliminary report provides support for a role of the sAng-1/sAng-2/sTie-2 system with respect to ILD in both lcSSc and dcSSc, however, more research is needed. For the limited form of disease, ILD and putative pulmonary hypertension are associated with high concentrations of the potent pro-inflammatory agent sE-selectin, whereas, the diffuse form finds endothelial microparticles more strongly connected with ILD but not putative pulmonary hypertension. Higher levels of microparticles have been associated with promoting adhesion in vascular disorders [44].

Thanks for laboratory work to A. Meredith, BS, and E. Whalen, BS (James Hogg Heart and Lung Institute, St. Paul’s Hospital and University of British Columbia, Vancouver, Canada). Thanks also for assistance with statistical analysis to W. Petrcich, BS (Department of Mathematics and Statistics, University of Northern British Columbia, Prince George, Canada).

This work was supported by research grants from the Scleroderma Association of British Columbia and the Heart and Stroke Foundation of Canada. Dr. van Eeden is a Senior Scholar of the Michael Smith Foundation for Health Research, a Career Investigator of the American Lung Association, and the Canadian Institutes for Health Research GlaxoSmithKline Professor in Chronic Obstructive Pulmonary Disease.