Increased soluble P-selectin in patients with haematological and breast cancer: a comparison with fibrinogen, plasminogen activator inhibitor and von Willebrand factor

Abstract

Abnormal platelet activation and an increased risk of thrombosis are frequent findings in cancer. As soluble adhesion molecule P-selectin is being increasingly recognized as reflecting increased platelet activation, we hypothesized raised levels in patients with cancer, obtaining plasma from 24 patients with a cross-section of haematological cancers, 41 with breast cancer, and from an equal number of healthy controls for each patient group. Levels of soluble P-selectin were compared with those of von Willebrand factor (vWf), plasminogen activator inhibitor-1 (PAI-1) activity and fibrinogen (markers of endothelial integrity, fibrinolysis and coagulation, respectively). We found raised soluble P-selectin, fibrinogen and vWf in both patient groups compared with their controls (P< 0.01). vWf and soluble P-selectin were higher in the haematological cancers than in breast cancer patients (by 30 and 74%, respectively; bothP< 0.01). There was no significant difference in levels of PAI-1 between any group. There were no differences in soluble P-selectin or vWf when the data from the women with breast cancer were classified according to tumour size, lymph node involvement or presence of vascular invasion. We conclude that the platelet marker soluble P-selectin is raised in both haematological and breast cancer, and is higher in the former, but is unrelated to the type or stage of breast cancer.

Introduction

Cancer is a complex disease characterized by profound changes to physiological systems such as haemostasis and vascular function, and, as a consequence, often confers an increased risk of arterial and venous thrombosis ^[1–3]. Evidence of these changes include abnormalities in the levels of plasma molecules involved in coagulation and fibrinolysis [such as fibrinopeptide A, fibrinogen, plasminogen activator inhibitor-1 (PAI-1) and D-dimer]^[2–5], markers of endothelial cell integrity (soluble E-selectin, von Willebrand factor and soluble thrombomodulin) ^[2–7] and of platelet function (beta-thromboglobulin) ^[8,9]. Some of these changes may be related to, or even partially responsible for, the increased risk of thrombogenesis in these patients.

The adhesion molecule P-selectin is a component of the membrane of the platelet alpha granule and of the membrane of the Weibel–Palade body, the endothelial cell organelle that stores the pro-coagulant molecule von Willebrand factor ^[10,11]. A soluble form of P-selectin has been described in the plasma of normal subjects, with increased levels in patients with atherosclerotic and thrombotic disorders ^[12,13] that seem likely to be related to excess platelet activation ^[14–16]. For example, soluble P-selectin is released from activated platelets in vivo during haemodialysis ^[17]. Increased levels of other soluble adhesion molecules, including soluble E-selectin, have also been described in various cancers ^[7], and upregulation of cell membrane adhesion molecule expression may be important in extravasation and metastasis ^[18,19]. However, although raised soluble P-selectin has been described in patients with melanoma ^[20], this has been disputed ^[21]. Of further interest is the report of the increased expression of P-selectin by endothelial cells within breast cancer tissues ^[22].

As increased levels of the platelet alpha granule matrix product beta-thromboglobulin are present in cancer ^[8,9], it was our primary hypothesis that soluble P-selectin would also be increased in this disease, when compared with healthy controls. However, as already mentioned, cancer is characterized by changes in many plasma proteins. We therefore intended to determine whether such increases in soluble P-selectin would be similar to, or greater than, the expected increases of other plasma markers also relevant to thrombosis and haemostasis. We chose PAI-1, fibrinogen and von Willebrand factor because they are markers of fibrinolysis, coagulation and endothelial function, respectively, are increased in cancer, and may have pathophysiological relevance as contributors to an increased risk of thrombosis ^[1–5,9,11]. We tested our hypothesis in a cross-sectional study of patients with different haematological cancers, and others with breast cancer, each compared with healthy controls. Recruiting patients free of current chemotherapy/blood product support, we sought to avoid the issue of the effects of these interventions.

Subjects and methods

Subjects

We recruited 24 patients (mean ± SD age, 52 ± 11 years; 14 women, 10 men) with haematological cancer (13 with lymphoid neoplasia and 11 with myeloid/granulocytic leukaemia) from outpatient clinics. Exact diagnoses were one acute lymphocytic leukaemia, two chronic lymphocytic leukaemia, eight chronic myeloid leukaemia, two chronic granulocytic leukaemia, five non-Hodgkin's lymphoma, two lymphoma, three malignant myeloma and one Waldenstrom's macroglobulinaemia. These patients had been free of chemotherapy, radiotherapy, or blood transfusion for at least 4 weeks, and had haemoglobin, leukocyte and platelet counts within the normal range and so may be regarded as being in remission. Forty-one women (aged 59 ± 14 years) with early, operable breast cancer were recruited as outpatients prior to surgery. They had also been free of chemotherapy or radiotherapy for 4 weeks. After surgery, the size of the excised tumour (grade T1, < 2 cm diameter; grade T2, 2–5 cm; grade T3, > 5 cm; grade T4, fixed to chest wall), lymph node involvement (presence of tumour in at least one axilliary draining node) and the degree of vascular invasion were recorded according to established criteria. Individual diagnoses were made according to standard protocols, supported by laboratory or imaging services, as appropriate (e.g. immunophenotyping).

As there was clear sex difference between the two patient groups, each had its own different healthy control group. Thus, the 41 women with breast cancer were compared with 41 healthy age-matched (55 ± 8 years) women, and the 24 patients with haematological cancer were compared with 24 control subjects (aged 56 ± 6 years; 10 men, 14 women). All controls were drawn from attenders for endoscopy, hernia repair or for minor operations, their spouses, and from healthy hospital staff. Exclusion criteria for all subjects was diabetes, venous ulceration, serological evidence of hepatitis B virus or HIV infection, atherosclerotic vascular disease, deep vein thrombosis, acute or chronic liver and kidney disease, connective tissue disease, acute infections or inflammatory conditions, or treatment with vasopressin or antibiotics. The approval of the Ethics Committee of West Birmingham Health Authority and informed consent from each subject was obtained.

Methods

Venous blood was obtained following non-traumatic venepuncture into 0.11 mol/l sodium citrate or ethylenediamine tetraacetic acid (EDTA). Citrated whole blood was kept cold (at 4°C), and plasma was obtained within 2 h of venepuncture following centrifugation for 20 min at 1000 ×g and 4°C, and was stored at –70°C until assayed. Soluble P-selectin was measured in citrated plasma by enzyme-linked immunosorbent assay using commercial reagents (R&D Systems, Abingdon, UK). As a component of the membrane of the platelet alpha granule, not its matrix ^[10], a specialized anticoagulant containing substances such as aspirin, theophylline and prostacyclins is not required (as would be the case for beta-thromboglobulin ^[13,23]. von Willebrand factor was measured using an established technique and commercial antisera from Dako (Ely, UK), and reference von Willebrand factor from the NIBSC (Potters Bar, Herts, UK) ^[12]. PAI-1 activity (i.e. that able to bind to immobilized tissue plasminogen activator) was measured using commercial reagents (Actibind PAI-1; Technoclone, Vienna, Austria). Fibrinogen was estimated according to a modified Clauss technique using a Coagulometer and thrombin from Pacific Haemostasis (Huntersville, North Carolina, USA). The intra-assay coefficient of all assays was < 5%, inter-assay variation was < 10%. A haematology screen (white blood cell count, platelet count, haemoglobin) was performed on an Advia 120 (Bayer, Newbury, UK) using the EDTA sample of venous blood.

Data analysis, statistical methods and power

Data was subjected to the Shapiro–Wilks normality test to determine the nature of its distribution. von Willebrand factor and fibrinogen data were distributed normally and are presented as mean and standard deviation. PAI-1 activity and soluble P-selectin data were distributed non-normally and presented as the median and interquartile range. Data between the each of the two groups (controls and patients) was analysed by t test or the Mann–Whitney U test. Correlations were sought using Spearman's ranks method ^[24]. Our recruitment numbers are such that the difference in soluble P-selectin, von Willebrand factor and fibrinogen each provide a power of greater than 0.90 to detect a difference of P < 0.05, or a power of greater than 0.80 to detect P < 0.01 ^[24]. All analyses were performed on a Minitab release 12 package on Windows.

Results

Cross-sectional data

There was no significant difference in the ages of the patients in the two studies (breast cancer study cases/controls, P = 0.097; haematology cancer study cases/controls, P = 0.21), the proportion of smokers (chi-squared, 0.544;P = 0.461; chi-squared 0.444;P = 0.505, respectively) or the proportion of sexes in the haematology study (Table 1). von Willebrand factor (P = 0.001;Fig. 1), soluble P-selectin (P < 0.001;Fig. 2) and fibrinogen (P = 0.005;Fig. 3), but not PAI-1 (P = 0.2482), were all higher in the breast cancer patients than their controls (Table 1). Similarly, von Willebrand factor (P < 0.001;Fig. 1), soluble P-selectin (P < 0.001;Fig. 2) and fibrinogen (P = 0.0072;Fig. 3), but not PAI-1 (P = 0.2482), were all higher in the haematology cancer patients than their controls (Table 1). Levels of both von Willebrand factor (P = 0.0012) and soluble P-selectin (P < 0.001) were higher in the haematology cancer patients than in the breast cancer patients.

Figure 1.:   Levels of soluble P-selectin among the patients with haematological cancers, breast cancer, and their respective controls. The bar represents the mean value.

Figure 2.:   Levels of von Willebrand factor among the patients with haematological cancers, breast cancer, and their respective controls. The bar represents the mean value.

Figure 3.:   Levels of fibrinogen among the patients with haematological cancers, breast cancer, and in their respective controls. The bar represents the median value.

Table 1: Age, sex and study indices in the patients and controls

Correlations

There were no statistically significant correlations between the four indices in either control group, the patients with breast cancer, or the patients with haematological cancer, except that von Willebrand factor correlated weakly with soluble P-selectin in the latter group (r_s= 0.406;P = 0.049). In the entire study group of 130 participants, the correlation was similar but the significance improved (r s = 0.401;P < 0.001). None of the indices correlated with age and there was no difference according to sex or smoking.

Subanalyses within patient groups

The small numbers of patients with diverse haematological malignancy (even in comparing leukaemia versus non-leukaemias or lymphoid malignancy versus myeloid malignancy) does not permit subanalyses: such an analysis was neither hypothesized nor calculated for. However, the number of subjects with breast cancer provides sufficient power for additional analyses. The four markers were analysed according to the size of the excised tumour (T1 versus T2–T4 combined), lymph node involvement (presence/absence), and the presence or absence of vascular invasion (Table 2). Of these, the only significant finding was that fibrinogen was higher in women with larger tumours (the combined group of stages T2, T3 and T4) when compared with women with smaller (T1) tumours grade. Only cases whose classification data was unambiguous were included so that, in some analyses, n < 41.

Table 2: The four plasma indices in 38 cases of breast cancers when classified according to histological criteria

Discussion

Many groups have reported raised levels of various plasma molecules in cancer, including those related to platelets, haemostasis and thrombosis, and vascular biology ^[1–9]. The question, therefore, is which of these markers has most to offer, notably in diagnosis, classification, or in staging prognosis? In the present comparison of two different types of cancer, soluble P-selectin was higher in the haematological cancers than in breast cancer, confirming our previous preliminary finding of raised soluble P-selectin using different reagents and patients with cancer ^[25]. Raised levels of von Willebrand factor, as other endothelial markers, and fibrinogen, in solid cancers (e.g. prostate, bladder), have been reported ^{[3–5,9,25–28]}, as has soluble P-selectin in melanoma ^[20], although this is disputed ^[21]. Worthy of note, patients were not on current treatment so results cannot be due to an artifact of, for example, cytotoxic drugs. Neither soluble P-selectin nor von Willebrand factor predicted histologically defined subgroups of breast cancer.

Membrane-bound P-selectin is found on activated endothelial cells (as a constituent of the Weibel–Palade body membrane) and on the surface of the activated platelet as a component of the alpha granule membrane ^[10,11]. However, lack of correlation with established markers of endothelial damage/dysfunction suggests it is likely to arise from the platelet and reflects platelet activation. For example, von Willebrand factor and soluble P-selectin are influenced by different risk factors and pharmacological interventions, and circulating degranulated platelets lose membrane P-selectin ^{[14–16,29,30]}. However, as breast cancer endothelium expresses increased membrane P-selectin ^[22], we cannot be certain that the plasma levels we have found do not arise from this source instead of, or as well as, from platelets. Nevertheless, the further significance of raised soluble P-selectin is unclear. Membrane-bound P-selectin may be involved in the development of metastases ^[18,19], possibly by binding to mucin-type ligands such as CD24 on tumour cells ^[31]. If so, then soluble P-selectin may interfere with contact between tumour cells and the endothelium or platelet in vivo, as may be possible in interactions between neutrophils and the endothelium in inflammatory disease ^[32]. However, this may not be an issue in certain leukaemias as acute myeloid leukaemia blasts neither express nor release P-selectin ^[33]. Increased soluble P-selectin (and, indeed, von Willebrand factor) in those with a haematological malignancy in apparent remission is notable and implies some degree of platelet perturbation in the absence of overtly active disease. The increased levels of both markers compared with breast cancer probably reflect the pathophysiology of the different diseases.

Failure to find increased levels of PAI-1 were surprising, but this may be because our haematological patients were in remission (with normal haemoglobin, leukocyte and platelet counts), while those with breast cancer generally had early, localized, well-defined disease and were fit for surgery. Blomback et al. ^[3] found raised PAI-1 in prostatic cancer, but these patients may have had more advanced disease. Indeed, Uchiyama et al. ^[28] reported raised PAI-1 in a heterologous groups of patients with various cancers, with higher levels in the presence of metastases. Some of these differences may be technical, as our method measures PAI-1 able to bind to immobilized tissue plasminogen activator. However, we found raised fibrinogen in both cancer groups, indicating some degree of haemostatic disturbance in our patients, in agreement with previous studies ^[9,28]. Nevertheless, with the exception of fibrinogen and the size of the tumour, we are unable to relate any of our markers to three commonly used histological classifications of breast cancer. In this group, the large numbers (comparable with previous work by Gadducci et al. ^[3,4]) provide sufficient power to detect any relevant differences, should they exist ^[24].

Our data has an additional implication. It has been suggested that adhesion molecules may be useful targets for cancer therapy ^[18,34] and in cardiovascular disease ^[35]. If so, then monoclonal antibodies and/or soluble ligands designed to bind to cell surface P-selectin (e.g. on breast cancer endothelium ^[22]) may first have to contend with possibly high levels of the soluble form in the plasma, which may absorb and so neutralize such antibodies, rendering them ineffective. The same principle extends to monoclonal antibodies to other adhesion molecules, many of which have raised levels of soluble forms in the plasma ^[7,20].

In conclusion, in the present pilot cross-sectional study, limited by relatively small numbers, we report that patients with haematological cancer or breast cancer have increased levels of soluble P-selectin that are not due to current therapy. Although we do not have the power for a formal and convincing sensitivity/specificity analysis, raised levels of soluble P-selectin were present in more patients with either type of cancer compared with raised levels of von Willebrand factor. Our study therefore provides preliminary data of an assessment of soluble P-selectin in certain cancers that needs to be confirmed in other common cancers such as prostatic cancer. Furthermore, the ability of soluble P-selectin to predict the presence or development of metastatic disease, arterial or venous thrombosis ^[36–38] and also long-term outcome in these diseases should also be explored.

Acknowledgements

The authors are grateful to the City Hospital Research and Development Programme for their support of this project.