Thrombocytopenia model with minimal manipulation of blood cells allowing whole blood assessment of platelet function.

Abstract In vitro models of thrombocytopenia are useful research tools. Previously published models have shortcomings altering properties of platelets and other blood components. The aim of the present study was to develop a whole blood method to induce thrombocytopenia with minimal manipulation, and to describe platelet function in induced thrombocytopenia in individuals with healthy platelets. Hirudin anticoagulated blood was obtained from 20 healthy volunteers. One part of the blood was gently centrifuged at 130g for 15 minutes. The platelet-rich plasma was replaced with phosphate-buffered saline to establish thrombocytopenia. Various levels of thrombocytopenia were achieved by combining different volumes of baseline whole blood and thrombocytopenic blood. Platelet counts were measured by flow cytometry (Navios, Beckman Coulter) and routine haematological analyser (Sysmex XE-5000). Platelet function was analysed by impedance aggregometry (Multiplate® Analyzer, Roche) and by flow cytometry (Navios, Beckman Coulter) using collagen, adenosine diphosphate, thrombin receptor activating peptide-6 and ristocetin as agonists. Median baseline platelet count was 227×109/l. The in vitro model yielded median platelet counts at 51×109/l (range 26–93×109/l). We observed minor, yet significant, changes in platelet size and maturity from baseline to modelled thrombocytopenia. In the thrombocytopenic samples, significant and positive linear associations were found between platelet count and platelet aggregation across all agonists (all p-values<0.001). Platelet function assessed by flow cytometry showed minimal alterations in the thrombocytopenic samples. A new whole blood-based model of thrombocytopenia was established and validated. This new model serves as a useful future tool, particularly to explore platelet function in patients with thrombocytopenia.


Introduction
Platelet function tests are used for diagnostic purposes in patients with increased bleeding tendency [1][2][3] and to evaluate the effect of antiplatelet agents [4][5][6]. There are multiple ways to evaluate platelet function in vitro, which all have inherent disadvantages [7,8].
Several laboratory models have been developed to induce thrombocytopenia for research purposes [9][10][11][12][13], but all these methods change the properties of platelets and other blood components, challenging translation of the findings to clinical purposes. These changes are due to excessive physical manipulation by additional centrifugation [10,13] and dilution with plateletpoor plasma [9], thereby adding potential platelet activating substances [14] in addition to reducing red and white blood cell counts [9]. Finally, some models are very time-consuming and labour-intensive [11]. Therefore, a new feasible model is warranted to induce thrombocytopenia focusing to preserve platelet properties.
The aim of the present study was to develop an in vitro model of thrombocytopenia with minimal manipulation of platelets and other blood cells. We validated this new model by quantifying changes in the properties of platelets and other blood components.
Finally, we investigated platelet function in the induced thrombocytopenia in individuals with healthy platelets in order to facilitate future diagnostic investigation of patients with thrombocytopenia by whole blood impedance aggregometry.

Methods
Twenty healthy adult volunteers were included after informed consent. No intake of medications affecting platelet function was allowed. The study was approved by The Central Denmark Region Committees on Health Research Ethics (Reference no.: 1-10-72-571-12) and by the Danish Data Protection Agency. The study was carried out in accordance with the Declaration of Helsinki.
Red blood cell concentration used for platelet counting as well as assessment of haematocrit, white blood cells, mean platelet volume and immature platelet fraction was performed in hirudin using a routine haematological analyser (Sysmex XE-5000, Kobe, Japan). Platelet count, red and white blood cell count, haematocrit, mean platelet volume and immature platelet fraction were measured in all samples, including baseline as well as thrombocytopenic samples. Baseline assessment of C-reactive protein (CRP) was analysed in lithium-heparin (Becton Dickinson Bioscience) using a Cobas 6000 (Roche, Basel, Switzerland).

Inducing thrombocytopenia
One part of the blood was centrifuged at 130g for 15 minutes at 20°C, without brake. The platelet-rich plasma was carefully removed and replaced with an equivalent amount of phosphatebuffered saline (PBS) 0.01M to establish severe thrombocytopenia (level 4) ( Figure 1). Intermediate levels of thrombocytopenia (levels 1, 2, 3) were achieved by combining fixed volumes of baseline whole blood and thrombocytopenic blood (level 4): level 1: 1080 µl baseline whole blood and 2400 µl thrombocytopenic blood, level 2: 600 µl and 2400 µl, level 3: 360 µl and 3000 µl ( Figure 1). This resulted in five samples from each individual (baseline and thrombocytopenia levels [1][2][3][4]. A preactivation sub-study was performed in five healthy volunteers to investigate the effect of physical manipulation on the platelets. Expression of P-selectin was used as a measure of preactivation. In the sub-study expression of P-selection, platelet aggregation was analysed in centrifuged and handled blood and compared with resting blood samples. The manipulation consisted of centrifugation at 130g for 15 min at 20°C followed by transfer of the blood to a new tube. This was done to simulate the blood handling in the thrombocytopenia model without platelet removal. Preactivation of platelets was determined as the proportion of platelets expressing P-selectin, using flow cytometry as described below. Platelet aggregation was measured by Multiplate® Analyzer (Roche, Basel, Switzerland) before and after manipulation using collagen 3.2 µg/ml and adenosine diphosphate (ADP) 6.5 µM as agonists (final concentrations).

Platelet function analyses
All blood samples rested for at least 30 minutes after manipulation and were analysed within two hours of sampling using the Multiplate® Analyzer. Hirudinized whole blood was diluted 1:1 with 37°C isotonic saline followed by incubation at 37°C for three minutes. The following agonists (final concentrations) were used: collagen 3.2 µg/mL, ADP 6.5 µM, thrombin receptor activating peptide-6 (TRAP-6) 32 µM and ristocetin 0.77 mg/mL (all from Roche, Basel, Switzerland). Platelet aggregation was quantified as arbitrary aggregation units (AU) and as area under the aggregation curve (AUC, AU × min), which integrates platelet aggregation and aggregation velocity [16].
Platelet function was expressed as median fluorescence intensity (MFI) of all platelets. The fluorescence intensity is registered in the photomultiplier tube as the fluorescence emitted by the antibodies on the platelet surface. Expression of P-selectin after the addition of HEPES-buffer instead of an agonist was used to assess platelet preactivation. The gating strategy is described and visualised in Appendix I in the supplemental data. Additional information on the flow cytometric investigations required Figure 1. Induction of thrombocytopenia to various platelet counts. Induction of thrombocytopenia was performed as follows: Hirudin anticoagulated whole blood (A) was centrifuged at 130g for 15 minutes separating the red blood cells from platelet-rich plasma (B). The plateletrich plasma was removed (C) and an equivalent amount of phosphate-buffered saline (D) was added. Following gentle mixture severely thrombocytopenic whole blood was achieved (E). Mixture of baseline whole blood (A) with fixed amounts of severely thrombocytopenic whole blood (E) provided different levels of thrombocytopenia (F). according to Lee et al. [18] includes quality control measures, type of flow cell with full optical path and characteristics of the signal and type of compensation. This information is available in a recent publication by Rubak et al. [17].

Statistics
Statistics were performed in Stata IC 13.1 (StataCorp LP, College Station, TX). Data distribution was visually evaluated by Q-Q plots. Normally distributed data are shown as mean ± standard deviation (SD) and non-normally distributed data are displayed as median and range. Normally distributed data were analysed using the paired t-tests or repeated measurements analysis of variance (ANOVA) as appropriate. Data not following normal distribution were evaluated by the Wilcoxon Signed-Rank test, and comparisons between more than two groups were performed using the Friedman test modified to accept missing values. To allow for multiple measurements for each participating person, we used a mixed-effects linear regression analysis to evaluate the association between platelet count and platelet aggregation, specifying random effects at the person level.

Results
The 20 healthy volunteers were distributed as 9 males and 11 females with a median age of 32 years (range 24-56 years). Baseline biochemical characteristics including fibrinogen are shown in Table I. All participants displayed platelet aggregation within the reference intervals at baseline.
Mean white blood cell count decreased significantly from 5.5 to 2.8×10 9 /l and red blood cells showed a small, yet significant, decrease from 4.8 to 4.6×10 12 /l (ANOVA, all p-values<0.001). The median proportion of platelets expressing P-selectin increased significantly from 21% to 29 % (p<0.005), without agonist addition. Furthermore, with no agonist addition, no change were observed in the median proportion of platelets expressing CD63 (1.8-1.8%, p=0.26) and minor increases were observed in the median proportion of platelets expressing bound fibrinogen (1.8-2.6%, p <0.03).
The preactivation sub-study compared centrifuged and handled whole blood with resting blood samples without platelet removal. The sub-study showed that the physical sample manipulation increased the median expression of P-selectin with 15% (from 16% to 31%, p =0.04), but platelet aggregation applying ADP and collagen was unaffected (p-values >0.23).
A significant and positive linear association was found between platelet count and platelet aggregation in the thrombocytopenic range determined by the Multiplate® Analyzer using collagen, ADP, TRAP and ristocetin as agonists (Figure 3) (all p-values <0.001).
Platelet function measured by flow cytometry showed minor changes. The collagen and TRAP induced expression of P-selectin, bound fibrinogen and CD63 was significantly reduced in the thrombocytopenic samples. The ADP-induced expression of P-selectin and CD63 was also significantly reduced, whereas the expression of bound fibrinogen showed no change. Ristocetin induced expression of P-selectin and CD63 was increased whereas bound fibrinogen showed no change ( Figure 4).

Discussion
In the present study, a new in vitro model of thrombocytopenia was established. The model is characterized by minimal influence on platelet properties. This method enables further exploration of patients with thrombocytopenia using impedance aggregometry in hirudinized whole blood. This implies the perspective to evaluate platelet function by impedance aggregometry in thrombocytopenic patients. Several initiatives were taken to ensure optimal preservation of platelet properties and platelet aggregation. We only observed minor changes in platelet properties including mean platelet volume, immature platelet fraction and platelet preactivation.  [15]. †The reference intervals of the utilised lab [16], except TRAP which was supplied by the manufacturer (Roche, Basel, Switzerland). Platelet function at various reduced platelet counts has previously been explored using light transmission aggregometry in platelet-rich plasma [19,20]. Thrombocytopenia has only been sparsely investigated in whole blood by impedance aggregometry [10] probably because previous studies have shown that whole blood platelet aggregation is highly dependent on platelet count [10,12,20,21]. However, investigations addressing only specific levels of low platelet counts (such as 25, 50, 75, 100 × 10 9 /l) are of limited value in a clinical setting, where patients present all levels of thrombocytopenia [9]. Using the present new model of thrombocytopenia, the exploration was taken further on a continuous scale and confirmed a significant and positive association between platelet count and platelet aggregation.
In previously published models of thrombocytopenia the properties of platelets and quantity of blood components were altered [9][10][11][12][13]20], limiting their value in exploration of platelet function. On the basis of the reported findings in previous studies, several modifications were conducted in the present thrombocytopenia model. Cattaneo et al. showed that the addition of platelet-poor plasma may inhibit platelet aggregation [14]. Therefore, PBS was used to replace platelet-rich plasma instead of using platelet-poor plasma. Thereby, the interference of platelet activators that may be present in a sample after high-speed centrifugation was minimized [22]. However, the comparison between these two methods was not performed in this study. Harrison et al. reported that physical manipulation activates platelets [7]. Therefore, the centrifugation velocity and duration were below the well-established practice applied in light transmission aggregometry [23]. Another important limitation of previous models was sample dilution resulting in extensive disturbances of red and white blood cell counts [9]. Accordingly, reduced haematocrit and white blood cell count have been shown to affect platelet aggregation [16,21]. By using PBS in the study, only minor yet statistically significant, changes were induced in red blood cell counts. The white blood cell count was reduced by the procedure, but only weak correlations to Multiplate® Aggregometry were found in previous studies [16,21]. Hence, in this model the white blood cell count was  sufficiently preserved not to have implications for interpretation of the aggregation analyses.
In order to further ensure that the manipulation of blood did not affect platelet function, we performed a preactivation substudy comparing centrifuged and handled blood with resting blood samples without platelet removal. An increase in platelet preactivation was found, but this did not change platelet aggregation. This implies that the manipulation used in the present in vitro thrombocytopenia model did not change platelet function considerably when applying the Multiplate® Analyzer. The preactivation was evaluated by flow cytometry without addition of agonist. The platelet marker of choice was P-selectin, which is known to be the gold standard for in vivo platelet activation [24]. Comparing P-selectin with CD63 and bound fibrinogen, P-selectin showed a substantial greater rise in activation after induction of thrombocytopenia compared to CD63 and bound fibrinogen. This confirms that P-selectin is the most sensitive marker for preactivation of platelets.
Flow cytometry was applied to further explore potential changes in platelet function, showing only minor changes. However, significant reductions in expression of platelet surface markers were observed. According to Panzer et al. [25], this may reflect a diminished haemostatic potential of the platelets after manipulation. Panzer et al. argued that increased preactivation would lead to a reduced response to agonists, which corroborates with our observations. The reduction in platelet function was limited despite the reduction in platelet count, which supports that flow cytometric evaluation of platelets is independent of platelet count [24].
Flow cytometry has only been used to evaluate platelet function in recent years. It is becoming more widely used and is considered very sensitive as regards evaluation of platelet function [26,27] and carries a future potential for this purpose [28][29][30]. However, flow cytometry is expensive, time consuming and only available in specialized laboratories. Therefore, optimization of other methods to evaluate platelet function in thrombocytopenia continues to be relevant.
Fibrinogen is an important cofactor for platelet aggregation. The presumably decrease in fibrinogen could potentially add to the reduction in aggregation, which would be the primary concern in replacing platelet-rich plasma (PRP) with PBS. However, according to the work by Bennett and colleagues [31,32], a reduction of fibrinogen to lower than 10% of the normal concentration is sufficient to saturate the fibrinogen binding sites on activated platelets. Furthermore, it is also regular practice only to add 0.4 mg/ml fibrinogen to washed platelets to measure platelet aggregation by light transmussion aggregometry [14]. In the present study, the fibrinogen level of the healthy volunteers was well within the reference range, and the reduction induced by this method is not expected to influence platelet function.
The present study has some limitations to be considered. Extremely low platelet counts were not obtained. This may be feasible, but it would most likely tend to further reduce white blood cell counts and cause further platelet preactivation. Extremely low platelet counts are seen in a clinical setting, but this model is not validated beneath the obtained platelet count. The fibrinogen level was not measured in the thrombocytopenic samples. However, applying the present method, fibrinogen level will be well above the critical level needed to support platelet aggregation. Although extensive measures were made to minimize potential changes in platelet properties, delicate changes of physiological importance cannot be completely ruled out.

Conclusion
A new in vitro model of thrombocytopenia with minimal manipulation of platelets and other blood cells was developed and validated. Platelet function was investigated in induced thrombocytopenia in individuals with healthy platelets, and a significant and positive association of platelet count and platelet aggregation was confirmed. The present study represents an optimized model of thrombocytopenia allowing assessment of platelet function in whole blood and facilitates future evaluation of platelet aggregation in thrombocytopenic patients.