Background
Thrombin generation test (TGT) measures kinetics of thrombin activity during coagulation of a blood plasma sample mixed with activators of blood coagulation [
1]. The TGT is widely used in clinical research to measure global hemostatic potential in blood coagulation disorders either for diagnostic purposes [
2,
3] or as means of treatment monitoring [
4,
5]. More recently, TGT became a useful tool in drug development and control of drug potency and quality in drug manufacture [
6‐
9]. Although proposed in the 1950s [
10], the test gained popularity only a decade ago after the technique was revolutionized with the introduction of fluorogenic thrombin substrates and microtiter plate reader format [
2]. However, the convenience of the microtiter plate technology can be deceiving: microtiter plate assays are prone to a special kind of variability caused by the uneven microenvironments in different wells of the plate [
11].
In order to describe the location-based effects for the TGT assay, a recent biotechnology application of TGT was used. We and others found that the procoagulant activity of IVIG correlates with reported myocardial infarction, stroke and other thromboembolic events [
12,
13]. Commercial (CAT® by Stago and Technothrombin® by Technoclone) as well as in house variants of the fluorogenic TGT method were especially helpful in identification of procoagulant IVIG lots by manufacturers and regulatory agencies[
14,
15]. In this report, we describe how both in house and commercial methods are prone to location-based biases which can be addressed with the use of symmetrical strip-plot layout design.
Materials and methods
Materials
Human normal pooled plasma (FACT) was from George King Biomedicals. Immunodepleted Factor XI deficient plasma was from Affinity Biologicals (Ontario, Canada). Human plasma-derived Factor XIa was from Haematologic Technologies Inc (Essex Junction, VT). Recombinant lipidated tissue factor (rTF, Dade Innovin) was from Dade Behring (Marburg, Germany). TF activity was determined using the Actichrome TF chromogenic kit (American Diagnostica). Fluorogenic substrate for thrombin Z-Gly-Gly-Arg-AMC was from Bachem (Torrance, CA). Phospholipid vesicles were from Technoclone (Diapharma, West Chester, OH). The Thrombin Calibrator and TF reagent PPP-low used in the CAT instrument experiment were from Thrombinoscope BV, Maastricht, The Netherlands.
In house thrombin generation test for IVIG procoagulant activity
Thrombin generation was measured as described in [
8,
16,
17]. Thrombin generation profiles in FXI-deficient or normal plasma (75% by reaction volume) supplemented with fluorogenic substrate Z-Gly-Gly-Arg-AMC (800 μM) and phospholipids vesicles (4 μM) were mixed with serially diluted thrombogenic lot of IVIG or FXIa. To reduce procedural errors, all reagents and plates were kept on plate heater (37°C); plasma was mixed with substrate and lipids prior transfer to microtiter wells; IVIG and FXIa samples (20% by reaction volume) were transferred to plasma using a 12- or 8-channel pipette, and reaction was started by rapidly adding 2.54 μL (2.5% by reaction volume) of a mixture containing tissue factor (0.3 pM) and CaCl
2 (20 mM) using another 12-channel pipette. Recording was conducted in two microplate readers, Infinite F500 (Tecan, Durham, NC) and Synergy H4 (Biotek, Winooski, VT) at 37°C.
TG curve processing software
In house assay data processing was performed using an automated software package designed by Dr. Mikhail Ovanesov using OriginPro (OriginLab, Northampton, MA; the package is available from us upon request). The software is capable of applying different processing algorithms during the conversion of raw fluorescence to the processed TG curve, and computes the final TG curve parameters. The software can also apply the thrombin calibration and thrombin-α2macroglobulin (α2MG) correction algorithms [
2]. For quantitative assessment of IVIG procoagulant activity (a bioassay approach [
15]), a calibration curve was prepared from a serially diluted IVIG standard or FXIa. Thrombin peak height (TPH) values were plotted against FXIa concentration and fitted with a cubic polynomial equation. To calculate FXIa activity in the well, the calibration curve was applied to the TPH and multiplied by the pre-dilution of the sample in the well.
Commercial thrombin generation test (CAT)
IVIG samples were mixed with plasma as described in the in house method, but after that the experiment was performed and processed according to the manufacturer’s guidelines using CAT microplate reader and software package (Diagnostica Stago, Inc., Parsippany, NJ).
Discussion
The clinical laboratory version of the TGT is usually intended for comparison of coagulation potentials of patient and healthy donor plasma samples [
2,
3] or patient-specific evaluation of the procoagulant and anti-coagulant effects of existing and investigational treatments [
4,
5]. Consequently, prior research on TGT assay development largely focused on preanalytical conditions and proper calibration of assay in units of thrombin activity. Improved quality of collected blood plasma samples, control of plasma temperature as well as internal thrombin calibration or normalization on standard plasma sample were shown to improve assay performance.
Evaluation of drug efficacy and quality, unlike clinical applications, may be less sensitive to preanalytical conditions because several drug preparations are compared with each other on the same microplate using a single sample of plasma. However, quantitative assessment of the difference between drug preparations usually requires comparison of different drug doses. Ideally, a bioassay is employed in which comparison to a standard drug preparation employs a calibration curve or a parallel line assay. Therefore, biotechnology applications require additional TGT qualification to ensure consistent low limit of detection, linear dose-response range and parallelism of tested and standard drugs.
Randomization of samples is an effective solution to correct location effects in high throughput analytical assays, but complete randomization of sample layout on the microtiter plate is prone to procedural errors [
20]. A more practical approach is a strip-plot design in which samples are assigned to random columns and serial dilutions of each sample to one column [
20]. However, in a typical TGT experiment, more samples are tested than there are columns on a plate, and the number of serial dilutions is smaller than there are rows. Since all coagulation samples are tested at least in duplicates, we propose to use a symmetrical design in which duplicates are positioned symmetrically with respect to the center of the plate. When applied to IVIG-TGT, this approach corrects the row-to-row drift and is easy to implement: (a) prior to addition to plasma, double volumes of IVIG samples are arranged in columns 1 through 6 (half of the plate), each sample in a single well, dilutions arranged vertically, (b) plasma is mixed with other required reagents (e.g., lipids) and transferred to cover all 96 wells of another plate, (c) IVIG samples are transferred to columns 1 through 6 of the plasma plate, (d) the plasma plate is then rotated 180° degrees, and (e) step
c is repeated.
While we encountered location errors during drug screening studies, these findings may be even more important for other TGT applications, e.g., use of TGT for diagnosis and treatment of patients [
5]. Sensitivity of coagulation assays to pre-analytical conditions has been known for decades yet the causes of TGT variability have only recently been investigated [
19], but location artifacts were not discussed in the literature. Location effect can be introduced in various ways. We found that if a coagulation enzyme, e.g., FXIa or FVIIa, is added manually to multiple wells on a microtiter plate, a well-to-well drift appears due to the different durations of contact between the enzyme and plasma inhibitors. A similar effect may be observed if TF or contact activator is added manually before recalicification. In our experience, long exposure of plasma to fluorogenic substrate prior to recalcification also decreases the TGT response. Edge effects can be caused by uneven heating of microplate inside the reader and artifacts of liquid dispenser. It should be noted that internal normalization on the thrombin-α2macroglobulin activity calibrator (e.g., as proposed by Hemker et al. [
21]) is not helpful here because the thrombin calibrator only corrects for fluorogenic signal-related inaccuracies and fails to correct for biological and pipetting variables [
22]. All location effects may be corrected with the use of the strip-plot sample design. Specifically, when edge artifacts are minor, our symmetrical strip-plot design provides the simplest practical solution for microplate-based TGT experiments.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Dr. MVO designed the study, developed thrombin generation analysis software and wrote the paper. Dr. YL and SAW conducted the experiments presented in this paper. Dr. AMS conducted preliminary experiments and assisted with method development. Dr. TKL contributed to the study interpretations and helped writing the paper. All author’s read and approved the final manuscript.