The Inner Filter Effect

To test whether the IFE non-linearity can disrupt the TG assay, plasma samples were spiked with increasing concentrations of fluorophore AMC (0 to 200 µM) prior to triggering TG (Fig. 2). The resulting fluorescence curves showed a baseline fluorescence intensity proportional to AMC concentration in both hemophilia and FVIII supplemented plasma samples (Fig. 2A,B). An unexpected artifact of fluorescence signal drop was observed within the first 5 min on each fluorescence curve, which corresponded to a negative substrate consumption rate. It is possible that these momentary changes in fluorescence signal are indicative of the settling of plasma-calcium mixtures following the injection of calcium chloride into AMC-containing plasma. This artifact was also observed in experiments without added substrate (compare supplemental Fig. S1A and Fig. S1B), suggesting that the drop is unrelated to the IFE under investigation in this experiment, but is likely caused by the optical changes in plasma absorbance after addition of DMSO (AMC diluent).

Fig. 2

IFE of AMC fluorophore and its correction via calibration or normalization. FVIII-DP was supplemented with 1 IU/mL FVIII to normalize hemophilia plasma, or not, and was subsequently premixed with the indicated concentrations of AMC prior to initiating coagulation with Ca2+ and substrate. Raw fluorescent data were produced by the CAT microplate reader and software and analyzed in several different ways: (A, B) raw AMC fluorescence in relative fluorescent units (RFU), (C, D) internally calibrated TG curves via a thrombin calibration coefficient (see Materials and Methods), (E, F) normalized-uncalibrated curves, (G, H) calibrated TG curves (via TS software), (I, J) calibrated TG curves (via OR software), and (L, M) calibrated TG curves (via SH software). Uncalibrated curve data were produced by differentiating the AMC curves observed in (A, B). Calibrated curves were produced using TS software, our in-house OR software, which uses published algorithms similar to CAT calibration, or SH software, our second in-house app based on CAT algorithm. An asterisk (*) next to the indicated concentrations in panels G & H denotes high AMC concentrations in which commercial TS software did not report TG curves, possibly due to their noisy appearance as suggested by the TG curves reported by OR and SH software apps at these high concentrations. Normalized-uncalibrated curves were produced by normalizing each uncalibrated curve pairing of hemophilic sample and normalized hemophilia sample (hemophilic plasma supplemented with FVIII) at each pre-spiked AMC concentration against the TPH value of the normalized plasma sample in each pairing. TG was recorded for 40–60 min. Assay conditions: 63 µL of FVIII-DP, 1 µL of FVIII (1 IU/mL), 16 µL of AMC at indicated concentrations, 20 µL of PPP trigger, and 20 µL of FluCa.

Without correction for IFE and substrate consumption (i.e. only differentiation of the fluorescence curve and internal linear thrombin calibration, see Materials and Methods), the resulting TG curves showed an inverse, concentration-dependent effect of pre-spiked fluorophore on the general size of the curve (Fig. 2C,D). We normalized the uncalibrated TG curves at each pre-spiked AMC concentration against the TPH value of the corresponding sample of hemophilia plasma supplemented with FVIII (Fig. 2E,F). This method produced overlapping normalized TG curves for all TG conditions, confirming that AMC did not change the TG curves but only induced the IFE.

The TG curves were reanalyzed to test the algorithmic correction of IFE. TS software produced overlapping curves at AMC concentrations below 115 µM (Fig. 2G,H; and Fig. 3); no TG curve data (curve = 0) were reported for experiments with AMC above 115 µM or 75 µM in samples with either 1 IU/mL FVIII or no added FVIII, respectively (Fig. 3A-D). When OR software was used for corrections analysis, the resulting TG curves appeared to overlap except at the highest concentrations of spiked AMC fluorophore, where they were erratic (Fig. 2I,J). Correction via SH software produced overlapping TG curves, with small overestimation at the highest AMC concentrations (Fig. 2L,M).

To investigate the reasons behind the overestimation of TG curves by OR and SH apps at high AMC concentrations, we plotted thrombin calibrator curves before (red lines in Fig. 2C-D) and after CAT algorithm corrections (red lines Fig. 2I-M) by OR and SH software. TS software app data were excluded from this analysis because TS does not report thrombin calibrator data. As expected, a sharp decline in thrombin activity was observed immediately after the beginning of the experiment (Fig. 2C-D), demonstrating the effects of IFE and substrate consumption in the TG assay. This decline was accurately resolved by CAT algorithm in the OR software, as evidenced by the same average thrombin calibrator activity at a horizontal line position of red calibrator curves in Fig. 2I-J. However, an increasing level of distorting noise was seen after 30 min, reflecting the amplification of noise in the uncalibrated curves in Fig. 2C-D. In contrast, SH software maintained accurate linearization for a period of 20 min only, followed by an overestimation of the thrombin calibrator activity from 20 to 30 min and progressive underestimation after 40 min (red lines in Fig. 2L-M). These differences in accuracy of corrective linearization of the thrombin calibrator curve correlated with the degree of overestimation of TG curves in AMC-supplemented experiments, i.e., underestimation of thrombin calibrator activity by SH in the area of extreme non-linearity of the calibrator curve (above 40 min, see red lines in Fig. 2A-B) resulted in less extreme overestimation of TG curves at the highest AMC spiked concentrations.

Figure 3 shows the TG assay parameters TPH and ETP as a function of AMC concentration. Since AMC fluorophore reagent does not interfere with the reaction of substrate cleavage by thrombin, the algorithms were expected to report comparable thrombin activities for either AMC concentration. Under- or over-estimation of reported activity indicates that the algorithm, as implemented in a particular software package, will over- or undershoot its correction, respectively. The OR software was able to accurately recover both TPH and ETP up to AMC fluorophore concentrations of ~ 170 µM, overestimating TPH (Fig. 3A,B) and ETP (Fig. 3C,D blue lines) at higher AMC. The TS software produced similar values for both parameters, except when the software did not compute the TG parameters for FVIII-supplemented plasma above 115 µM of spiked AMC (Fig. 3A,C) and hemophilic plasma samples above 75 µM of spiked AMC (Fig. 3B,D, red lines). For both plasma samples tested, SH software was able to recover both TPH and ETP at all AMC concentrations, however, a trend to overestimation was apparent above 115 µM of AMC (Fig. 3A-D). The mean TPH values for the indicated correction algorithms were significantly different when compared to values produced by internal linear calibration in both plasma samples (Fig. 3E,F). A correlation between the three studied apps is shown in Fig. 3G. The side-by-side comparisons between the three software outputs (Fig. 3A-D) indicated similar curve shapes. Additionally, the curves did not fully overlap. Further investigation suggested that these differences can be due to the smoothing applied by the TS software to the calibrated TG curves prior to parameter acquisition (refer to Supplemental Fig. S2 for additional evidence of smoothing).

Fig. 3

CAT calibration algorithm distorts TG curves when the fluorescence signal extends into the non-linear range of the calibrator. Thrombin peak height (TPH) and endogenous thrombin potential (ETP; measured as area under the curve) values were obtained from data in Fig. 1. TPH values from indicated software apps in (A) FVIII-DP supplemented with 1 IU/mL FVIII and (B) FVIII-DP with added buffer. ETP values from indicated software apps in (C) FVIII-DP supplemented with 1 IU/mL FVIII and (D) FVIII-DP with added buffer. Internal linear calibration is shown as black curves, CAT calibration is shown as blue curves (OR software), red curves (TS software), and green curves (SH software). CAT calibration correction did not return values at AMC concentrations > 115 µM. (E) TPH values from FVIII-DP supplemented with 1 IU/mL FVIII and (F) FVIII-DP with added buffer in indicated software apps, where p < 0.05 was deemed significantly different via a paired T-test (denoted as *) as compared to the Internal Linear group. Correlations of TPH from (G) TS software vs. OR Software, (H) SH software vs. OR Software, and (I) SH software vs. TS software in plasma with 0 IU/mL FVIII (gray squares) and with 1 IU/mL FVIIII (black squares). Assay conditions: 63 µL of FVIII-DP, 1 µL of FVIII (1 IU/mL), 16 µL of AMC at indicated concentrations, 20 µL of PPP trigger, and 20 µL of FluCa.

Overall, calibration by the CAT method, as implemented in the TS, OR and SH software apps, demonstrated adequate correction of suppressed rate of substrate consumption due to IFE that was induced by added baseline fluorescence. This correction was accurate in recovering correct thrombin activity when the fluorescence values were within the linear range of the thrombin calibrator wells. At higher fluorescence (above ~ 115 µM AMC, Table 1), calibration by TS or OR software apps eventually reached the edge of failure, either failing to produce TG results (TS software) or overestimating the TG parameters (OR software). SH software recovered TG parameters at all AMC concentrations. Interestingly, normalization to TPH in FVIII-supplemented plasma successfully produced overlapping TG curves, that where much less erratic than the curves produced by TS or OR software, suggesting another unintended consequence of correction algorithm, i.e. amplification of noise. This amplification was more apparent at higher concentrations AMC, consistent with the stronger IFE induced by AMC and stronger corrections applied by the algorithm.

Table 1 Edge of failure set points for fluorescent artifacts on TG analysis. The significance the IFE and substrate consumption on TG as analyzed by three software packages: TS, OR, and SH Effect of reduced substrate concentration

We next sought to investigate the corrective capabilities of algorithmic calibration in conditions of substrate depletion. Indeed, complete substrate consumption would be characterized by a fluorescence increase rate reaching zero value at a certain time point, leading to no TG activity recorded regardless of the CAT artifact correction algorithm. To facilitate substrate depletion, plasma samples were supplemented with decreasing concentrations of fluorogenic substrate (Fig. 4). The fluorescent curves showed a concentration-dependent decrease in overall fluorescent intensity as substrate concentration decreased (Fig. 4A,B). As expected, internal linear calibration without CAT correction resulted in TG curves that were decreased with lower substrate concentrations (Fig. 4C,D).

Fig. 4

Substrate consumption and its correction via calibration or normalization. FVIII-DP was supplemented with 1 IU/mL FVIII to normalize hemophilic plasma, or not, and was subsequently premixed with the indicated concentrations of substrate, ZGGR-AMC, prior to initiating coagulation with Ca2+ and substrate. Raw data was produced by the CAT microplate reader and analyzed in several different ways: (A, B) raw AMC fluorescence in relative fluorescent units (RFU), (C, D) internally calibrated TG curves via a thrombin calibration coefficient (see Materials and Methods), (E, F) normalized-uncalibrated curves, (G, H) calibrated TG curves (via TS software), (I, J) calibrated TG curves (via OR software), and (L, M) calibrated TG curves (via SH software). Uncalibrated curves data were produced by differentiating the AMC curves observed in (A, B). Calibrated curves were produced using TS, OR or SH apps all of which employed the same CAT correction algorithm. Normalized-uncalibrated curves were produced by normalizing each uncalibrated curve pairing of hemophilic and normalized sample (hemophilic plasma supplemented with FVIII) at each pre-spiked AMC concentration against the TPH value of the normalized plasma sample in each pairing. TG was recorded for 40 min. Assay conditions: 63 µL of FVIII-DP, 1 µL of FVIII (1 IU/mL), 16 µL of substrate ZGGR-AMC (at indicated concentrations), 20 µL of PPP trigger, and 20 µL of Ca2+ (calcium chloride buffer)

The CAT algorithm (in all three software packages) was able to reconstruct TG curves at all tested substrate concentrations without reaching the point of failure, suggesting that complete substrate consumption was not observed under any tested condition in Fig. 4. Another indication of incomplete substrate consumption was that normalization to TPH (in FVIII-supplemented plasma) appeared to correct for the substrate concentration effects with great consistency for all samples, producing overlapped curves, suggesting that the shape of TG curve was not affected by substrate concentrations (Fig. 4E,F). Faster onset of substrate consumption would have changed the shape of TG curve. However, normalized TG curves appeared distorted by noise at low substrate concentrations, likely because reduced signal (in samples with reduced substrate concentration) to noise ratio resulted in more erratic curves.

Interference of substrate with TG reactions

Interestingly, in substrate titration experiments, all three software packages produced an overestimation of TG curves at the highest substrate concentration (Fig. 4I-M). Previous studies suggested that fluorogenic substrate can interfere with TG kinetics. Therefore, decreasing the initial substrate concentration, which is used at a concentration of 416 µM in commercial assays and 800 µM in our laboratory, may have impacted the TG kinetics independently of, and in addition to, the artifact of substrate depletion. To address the effect of substrate on TG reactions, we used another ZGGR substrate based on an AFC fluorophore rather than AMC. A mixture of AFC and AMC substrates can maintain the total concentration of ZGGR peptide allowing us to investigate the effect of AMC-based substrate consumption without adverse manipulation of TG kinetics. We assumed that the specificity of ZGRR substrate to thrombin and FXa and other enzymes has not changed substantially as a result of replacing the AMC fluorophore with the AFC fluorophore.

We confirmed that AFC has peak fluorescent emission shifted to the right of that of AMC (~ 490 nm vs. ~450 nm; Supplemental Fig. S3A,B). Unfortunately, in the CAT instrument the AFC fluorophore alone produced a much higher signal than that of AMC, because the CAT microplate reader uses a filter set favored by the AFC, i.e., the 390 nm excitation is closer to preferred AFC excitation than that of AMC and the 460 nm emission is close enough to AFC emission peak (see Supplemental Fig. S4). Therefore, we used a monochromator microplate reader (Biotek) in a narrow-band mode of 380 nm excitation and 430 nm emission (shifted lower than CAT instrument’s wavelengths). A control experiment on this microplate reader using an AMC-only substrate titration under the same conditions and reagents as used for the CAT experiment demonstrated consistent results between CAT and Biotek readers (compare Fig. 4 to Supplemental Fig. S5). TS software was not able to use fluorescent data produced by Biotek reader, and therefore only OR and SH software were compared below.

Because the AMC/AFC substrate mixture experiments contained the same combined concentration of ZGGR peptide at the beginning of the experiment, the samples within each hemophilic plasma pair (with and without 1 IU/mL FVIII) were expected to have the same TG reactions regardless of the concentration of added ZGGR-AMC substrate. In contrast, recording of the TG curves were predicted to be affected by the consumption of AMC substrate. As expected, a decrease in the substrate consumption rate (Fig. 5A,B) was observed with decreasing concentrations of AMC substrate.

Similarly to AMC substrate titration experiments above (Fig. 4), the AMC/AFC substrate mixture experiments showed that the AMC substrate was not completely consumed in either plasma sample group, as evidenced by the continuous sloped increase at the tail end of the fluorescent curves (Fig. 5A,B), which suggests that residual substrate cleaving activity (likely that of the thrombin-α2 macroglobulin complex formed at early stages of the TG curve) was continuing to cleave substrate. Therefore, an edge of failure caused by complete substrate depletion did not occur in hemophilic plasma with and without added FVIII regardless of the substrate concentration (Table 1). However, CAT correction algorithm failed to alleviate the underestimation of the TG in samples with lower than typical substrate concentrations, see Fig. 5G,H (OR software) and Fig. 5I,J (SH software). Normalization of the TG curves to the TPH values produced curves that were slighlty more overlapped and less erratic than the calibration method (Fig. 5E,F). It should be noted, however, that both the normalization and CAT methods produced TG curves with higher lag times as the concentration of ZGGR-AMC increased, preventing a full overlap of TG curves.

Fig. 5

An attempt to study substrate consumption in plasma samples supplemented with two substrates, ZGGR-AMC and ZGGR-AFC. FVIII-DP was supplemented with 1 IU/mL FVIII to normalize hemophilic plasma, or not, and was subsequently premixed with the indicated concentrations of two substrates, ZGGR-AMC and ZGGR-AFC, such that the ratio of AMC:AFC equaled to a concentration of 800 µM, prior to initiating coagulation with Ca2+. Raw data was produced by the Biotek microplate reader and analyzed in several different ways: (A, B) raw AMC fluorescence in relative fluorescent units (RFU), (C, D) internally calibrated TG curves via a thrombin calibration coefficient (see Materials and Methods), (E, F) Normalized-Uncalibrated curves, (G, H) Calibrated TG curves (via OR software), and (I, J) calibrated TG curves (via SH software). Uncalibrated curve data were produced by differentiating the AMC curves observed in (A, B). CAT calibrated curves were produced using our in-house OR and SH software apps, which use published algorithms similar to CAT calibration. Normalized-uncalibrated curves were produced by normalizing each uncalibrated curve pairing of hemophilic and normalized sample (hemophilic plasma supplemented with FVIII) at each pre-spiked AMC concentration against the TPH value of the normalized plasma sample in each pairing. TG was recorded for 40–60 min. An artifact resembling TG signal in early minutes in Fig. 5 is only seen with ZGGR-AFC experiments, suggesting that it is caused by either the AFC fluorophore itself (similar to Fig. 2 and Fig. S1 discussed above) or background fluorescence signal of un-cleaved ZGG-AFC substrate. Assay conditions: 78 µL of FVIII-DP, 2 µL of FVIII (1 IU/mL), 20 µL of PPP trigger, and 20 µL of custom FluCa mixtures (substrates ZGGR-AMC and ZGGR-AFC at indicated concentrations with calcium chloride buffer)

Interestingly, similar to experiments with spiked AMC, the baseline fluorescence intensity of each sample increased as AFC substrate concentration increased, suggesting that ZGGR-AFC substrate produced a baseline fluorescent signal even before it was cleaved, despite employing a wavelength set by the Biotek plate reader (Fig. 5A,B). This bleed-through of the ZGGR-AFC and possible AFC fluorescence into the AMC recording channel is the likely reason for the failure of CAT algorithm to correct for AMC substrate consumption in the AMC/AFC substrate mixing experiments.

Complete substrate consumption by Procoagulant samples

Previous TG investigations suggested that complete substrate depletion can be achieved in procoagulant conditions caused by deficiency of antithrombin [5, 7]. In another attempt to produce forced controlled substrate depletion, we used moderately antithrombin deficient plasma (ATIII-DP, 5% deficiency) with or without a dose of heparin that partially compensated for antithrombin deficiency (by heparin-mediated increase in antithrombin activity). Further, to determine the edge of failure for the variable substrate consumption conditions, we produced dose-dependent elevation of TG curves with increasing concentrations of TF trigger.

As expected, AMC fluorescence curves were dose dependent on TF concentration (Fig. 6A,B). Internal linear calibration without CAT corrections demonstrated dose dependent TG curves, with higher and quicker curves generated at high TF concentrations (Fig. 6C,D). CAT algorithm with the OR software produced highly distorted TG curves comprised of the initial elevation and peak reminiscent of the beginning portion of the uncalibrated TG curved, yet with no decline in thrombin activity following the peak (Fig. 6G,H). Instead, TG peaks were followed by a brief plateau of thrombin concentration and another sharp elevation into infinite values. The second TG peak in OR software was caused by an artifact of thrombin activity overestimation, suggesting algorithm failure following complete depletion of the substrate, which was seen previously in our study of TG at elevated prothrombin levels [5, 7]. In contrast, SH software produced curves that were much less erratic but had plateau-like thrombin peaks between 20 and 4.3 pM of TF (Fig. 6I,J). Again, less overstimation by SH than OR software was consistent with its underestimation of linear calibrator concentrations closer to the end of the calibration curve (similar to red lines in Fig. 2L-M).

Fig. 6

Effect of CAT calibration on TG curves in procoagulant plasma samples. Antithrombin deficient plasma (ATIII-DP) was treated with or without heparin and the indicated TF concentration to assess the effect of calibration via different software apps in procoagulant samples. Raw data was produced by the CAT assay microplate reader and analyzed in several different ways: (A, B) raw AMC fluorescence in relative fluorescent units (RFU), (C, D) internally calibrated TG curves via a thrombin calibration coefficient (see Materials and Methods), (E, F) normalized-Uncalibrated curves, (G, H) calibrated TG curves (via CBER algorithm), and (I, J) calibrated TG curves (via SH software). Uncalibrated curve data were produced by differentiating the AMC curves observed in (A, B). CAT calibrated curves were produced by our in-house OR and SH software apps. Normalized-uncalibrated curves were produced by normalizing each uncalibrated curve pairing of hemophilic and normalized sample (hemophilic plasma supplemented with FVIII) at each pre-spiked AMC concentration against the TPH value of the normalized plasma sample in each pairing. TG was recorded for 60 min. Assay conditions: ATIII-DP with 0.2 U/mL of normal pooled plasma with or without heparin (0.03 USP/mL), TF (0.12–20 pM), tPA (0.13 µg/mL), thrombomodulin (12.5 nM), PC:PS vesicles (4 µM) and custom FluCa mixture (800 µM ZGGR-AMC and calcium chloride)

Consistent with previous observations in procoagulant samples with complete substrate depletion [5, 7], OR and SH software apps produced TG curves with a highly erractic tails (portion of the curve after the TG peak) in ATIII-DP samples, with higher noise as TF concentration increased. Erratic increases in TG curve tails are not likely to represent sudden spikes in thrombin activity, but rather an effect of amplification of fluorescence curve noise which is caused by an overshooting of the CAT algorithm corrections at small bumps in uncalibrated TG curve tails. This overshooting represents a failure of CAT algorithm to accurately reconstruct TG curves from miniscule growth of the fluorescence signal when fluorescence growth is hindered by substrate depletion [5, 7].

To further evaluate if exaggerated TG tails are an artifact of CAT algorithm, we studied clot formation and lysis through observation of clot turbidity in parallel with the TG experiments described above (see Supplemental Fig. S6). As expected, times to clot formation aligned well with the beginning of the TG curves (Fig. S6) and were inversely correlated with TPH. Further, the peak clot density was inversely correlated with TPH, and clot density peaked at a TF concentration of ~ 2.6 pM in the presence of heparin, whereas in the samples without heparin, clot density was highest at very low TF concentrations (~ 0.1 pM). Since clot formation is limited by the depletion of fibrinogen, clot density may not accurately reflect the TG curves. Yet, increased TG should translate into protection from tPA-induced fibrinolysis as we reported previously in this experimental system [12]. Nonetheless, time to clot lysis appeared to (inversely) correlate with the TPH calculated from the first clear peak on the TG curve (see Fig. S6). This is likely explained by the fact that no reactions of fibrinolysis begin before fibrin formation. When the lysis time is corrected for the clot time, the duration of clot lysis does correlate positively with the TF dose. Overall, monotonous changes in clotting and lysis parameters corresponded to monotonous changes in TPH parameter rather than apparently erratic changes of TG curve tail ends.