High-throughput platelet spreading analysis: a tool for the diagnosis of platelet-based bleeding disorders

Bleeding disorders are an extremely heterogeneous group of conditions presenting a unique diagnostic challenge. While the gold-standard platelet function assay is light transmission aggregometry, platelet defects are only detected in approximately 50% of patients with a clinical history of bleeding consistent with a platelet disorder. We tested a cohort of patients recruited to the UK-GAPP (Genotyping and Phenotyping of Platelets) study with a suspected platelet function disorder using a high-throughput platelet spreading assay. We detected a platelet spreading defect in 32 out of 55 patients tested (58%), and of these, 16 presented with normal lumiaggregometry results despite a significant Bleeding Assessment Tool (BAT) score. Furthermore, a family identified through this approach was subsequently identified as carrying a rare genetic variant of TUBB1, a gene linked to macrothrombocytopenia. This work suggests that morphological defects detected through a high-content platelet spreading approach can identify platelet dysfunctions not detected by lumiaggregometry.

Platelet function disorders are a heterogeneous group of conditions whose clinical and laboratory diagnosis is complicated by the range of reported bleeding symptoms, as well as functional redundancy within platelet signaling pathways. To address this, patients with suspected inherited bleeding disorders are often recruited for platelet function tests (PFT) following initial assessment.1 The main consideration when designing a panel of PFT is the best use of a limited quantity of sample. Despite the multitude of PFT available, a platelet defect is not always detected in patients with a significant bleeding history. Indeed, in patients with patterns of bleeding consistent with platelet disorders, we and others only find a platelet defect in approximately 50% of patients. To address the limitations of current approaches, the GAPP study performed a range of additional PFT, including impedance aggregometry (Multiplate), 96-well platelet aggregation, aggregation on collagen at arterial shear rates (1000 s) and clot retraction assays, on a subset of patients.32 Unfortunately this use of multiple PFT has not increased the number of patients identified with a platelet function disorder. Here we investigated a high-throughput analysis of platelets spread on fibrinogen to interrogate morphological and platelet defects.

The full methods are provided in the Online Supplement. Briefly, patients with bleeding of unknown cause were recruited to the GAPP study.4 Peripheral blood collected from patients and healthy volunteers was used to generate platelet-rich plasma for lumiaggregometry and washed platelets at a concentration of 2×10/mL for spreading on fibrinogen-coated coverslips. Spreading samples were incubated for 45 min before fixation, immunostaining and fluorescence microscopy. Captured images were subjected to a semi-automated image analysis pipeline using the open source software platforms KNIME and Ilastik. This produced platelet segmentations and corresponding measurements for area and circularity.65

Patients were recruited with a variety of bleeding symptoms consistent with a platelet defect (n=55). The International Society on Thrombosis and Haemostasis (ISTH)-BAT score was used to provide a quantitative evaluation of the patients’ bleeding history; scores were available for 40 individuals with the exception of pediatric patients. The overall mean BAT score was 9.825 (range, 2-23), with scores ≥4 being considered to be indicative of excessive bleeding.7

Lumiaggregometry was carried out as previously described.98 On the basis of this, the platelet defects in the cohort were split into six categories; cyclooxygenase (COX) deficiency (n=4), Gi receptor signaling defect (n=4), secretion defect (n=8), thrombocytopenia (n=8), multiple defects (n=5) and no defects (n=26) (Table 1). Upon recruitment, platelet samples from patients were also subjected to spreading on fibrinogen. Samples were fixed and stained with phalloidin to delineate cell morphology simply and accurately with the high signal/noise ratio required for robust high-throughput imaging and analysis. A large scale dataset was generated by taking six fields of view from a representative area of each slide. To establish a control dataset, platelets from five healthy volunteers were treated identically. Images generated from these assays were subjected to a semi-automated segmentation workflow to measure area and circularity for each platelet, which were plotted with the median of the dataset (Figure 1). Area and circularity measurements from the five representative healthy controls were used to define normal ranges. Patients’ data which lay outside of these ranges were defined as indicating a potential defect.

Table 1.Phenotypes of all patients recruited to the GAPP study who were included in the platelet spreading investigations.

Figure 1.High-throughput measurement of platelet area and circularity. Area and circularity measurements were generated using a semi-automated image analysis workflow and abnormalities were identified by comparison to a control range established by spreading data from healthy volunteers (red box) (n=5). Patients’ samples with medians and confidence intervals outside of the control range were classified as positive hits (blue boxes) for either a defect in platelet area or circularity, while samples which fell outside both normal ranges were classified as double-positives (orange boxes).

Overall, abnormal spreading morphologies were found in 32 of 55 patients tested (58%). This consisted of patients showing defects in platelet area only (n=6), platelet circularity only (n=13), both area and circularity (n=13) or no defect (n=23) (Figure 1). Abnormal area and circularity are consistent with round, unspread platelets, and such double-positive hits were considered robustly indicative of morphological spreading defects. Once identified as displaying a potential platelet spreading defect, samples were correlated to existing PFT and clinical data to establish whether spreading abnormalities predicted platelet dysfunction.

In total, spreading defects were found in nine patients who also had an aggregation defect (Gi=3, secretion=5, COX=1). Interestingly 16 of the 32 patients (50%) who were positive for spreading defects did not demonstrate a functional abnormality through lumiaggregometry, suggesting additional platelet defects have been identified through this spreading assay (Figure 2). Furthermore, these 16 patients had a significant BAT score (median BAT score=11). Representative immunofluorescence and segmentation images for area (patients 2 and 4), circularity (patients 23 and 24.1) or both defects (patients 41.1 and 41.2) are shown in Figure 2C. Of the ten double-positive platelet spreading hits with an available ISTH-BAT score, eight (80%) demonstrated a significant bleeding score (median BAT=11). Together, these data confirm that high-throughput spreading analysis can supplement data generated by lumiaggregometry, by both confirming aggregation defects and highlighting morphological abnormalities not detected by the gold-standard approach in patients with excessive bleeding.

Figure 2.Distribution of platelet spreading defects associated with lumiaggregometry defects. (A) A heat map showing the presence of a spreading or aggregation defect. Each row represents an individual patient. Each column depicts a platelet spreading abnormality (A - platelet area spreading defect, B - platelet circularity spreading defect, C - Platelet circularity and area spreading defect) or D - thrombocytopenia or lumiaggregometry defect, E - platelet secretion defect, F - COX defect, G - Gi signaling defect, H - multiple aggregation defect). In red: double-positive hits. In dark green: single hits for either a spreading or aggregation defect. In black: multiple hits within a category. In light green: no hits for either test. Patients with a double-positive hit and increased Bleeding Assessment Tool score are highlighted in red font. (B) A pie chart indicating the distribution of ‘double-positive’ platelet spreading defects. (C) Representative platelet spreading immunofluorescent and segmentation images of selected patients with clear defects of circularity or area. The top panel shows images for a representative control healthy donor. The lower panels show images for patients 2 and 4, who were found to have only a platelet area spreading defect, for patients 23 and 24.1, who only had a platelet circularity defect and patients 41.1 and 41.2 who had both area and circularity spreading defects.

The GAPP study aims to identify the phenotype and genetic causes of bleeding. A key question is whether platelet spreading can identify patients with a genetic defect in cases in which aggregometry is normal. In this cohort of patients, whole exome sequencing analysis was applied using an established bioinformatic pipeline.10 In one such patient a heterozygous TUBB1 missense variant (p.Arg359Trp) was identified (patient 41.2) but not in the patient’s sibling, 41.1. This genetic defect is also in keeping with the reduced platelet count only observed in patient 41.2 (107×10/L) and not in the sibling (221×10/L). TUBB1 defects are known to cause macrothrombocytopenia1411 and the identification of this variant through platelet spreading analysis supports the efficacy of this tool in the study of suspected platelet function disorders. Interestingly, this finding is consistent with previous reports showing a minimal effect of cytoskeletal inhibition on platelet aggregation. This suggests that the absence of a defect in aggregregometry in TUBB1-defective patients is likely due to a minor effect which overlaps with controls. In this instance, platelet spreading effectively detects a defect classically masked in aggregation experiments.1715

This study systematically correlated high-throughput platelet spreading analysis with clinical and lumiaggregometry data to determine the former’s utility in the investigation of platelet function disorders. We found that patients positive for both circularity and area abnormalities, consistent with the morphology of abnormally small and round platelets, are both positive and negative for aggregation defects. Eighty percent of patients with a spreading defect, but negative for a defect by aggregometry, had a significant bleeding score. Furthermore, within this cohort we identified TUBB1 as the causative genetic variant in an individual with thrombocytopenia.

Of particular interest is that a large number of patients positive for platelet spreading defects and with a history of excessive bleeding are negative for an aggregation defect, suggesting that applying platelet spreading as part of the platelet testing panel could reveal morphological defects which would otherwise be missed by lumiaggregometry alone. Furthermore this platelet spreading assay is not dependent upon platelet count, as is the case with aggregation-based PFT98 and could therefore be a useful tool to test for platelet function defects in patients with thrombocytopenia. Our approach validates the use of high-throughput platelet spreading analysis as an analytical tool to improve and augment existing PFT. As high-content imaging approaches become increasingly accessible and commonplace, applying high-throughput platelet spreading strategies is likely to be an invaluable addition to the current arsenal of PFT.