In-vitro and in-vivo metabolism of different aspirin formulations studied by a validated liquid chromatography tandem mass spectrometry method

Low-dose aspirin (ASA) is used to prevent cardiovascular events. The most commonly used formulation is enteric-coated ASA (EC-ASA) that may be absorbed more slowly and less efficiently in some patients. To uncover these “non-responders” patients, the availability of proper analytical methods is pivotal in order to study the pharmacodynamics, the pharmacokinetics and the metabolic fate of ASA. We validated a high-throughput, isocratic reversed-phase, negative MRM, LC–MS/MS method useful for measuring circulating ASA and salicylic acid (SA) in blood and plasma. ASA-d4 and SA-d4 were used as internal standards. The method was applied to evaluate: (a) the "in vitro" ASA degradation by esterases in whole blood and plasma, as a function of time and concentration; (b) the "in vivo" kinetics of ASA and SA after 7 days of oral administration of EC-ASA or plain-ASA (100 mg) in healthy volunteers (three men and three women, 37–63 years). Parameters of esterases activity were Vmax 6.5 ± 1.9 and Km 147.5 ± 64.4 in plasma, and Vmax 108.1 ± 20.8 and Km 803.2 ± 170.7 in whole blood. After oral administration of the two formulations, tmax varied between 3 and 6 h for EC-ASA and between 0.5 and 1.0 h for plain-ASA. Higher between-subjects variability was seen after EC-ASA, and one subject had a delayed absorption over eight hours. Plasma AUC was 725.5 (89.8–1222) for EC-ASA, and 823.1(624–1196) ng h/mL (median, 25–75% CI) for plain ASA. After the weekly treatment, serum levels of TxB2 were very low (< 10 ng/mL at 24 h from the drug intake) in all the studied subjects, regardless of the formulation or the tmax. This method proved to be suitable for studies on aspirin responsiveness.

www.nature.com/scientificreports/ formation of the platelet agonist thromboxane A 2 (TxA 2 ) 2 . Measurement in serum of TxB 2 , the stable TxA 2 metabolite, is used to evaluate the pharmacologic efficacy of ASA 3 which is able to induce a decrease of TxB2 levels from 200 to 300 ng/mL to ≤ 10 ng/mL 4,5 when administered in healthy subjects. Enteric-coated aspirin (EC-ASA) is commonly used in clinical practice because it was initially hypothesized that it would cause less gastrointestinal discomfort and bleeding [6][7][8] , compared to plain ASA. This surmise was not confirmed by controlled studies 9 , but EC-ASA is still the most commonly used ASA formulation to prevent arterial thrombosis. In recent years, it has been shown that, despite its extensive use, a significant proportion of patients taking the therapeutic dose of EC-ASA (100 mg o.d) displays inadequate pharmacodynamics (PD) response, with partial or null inhibition of TxA 2 biosynthesis or of arachidonic acid-induced platelet aggregation, thus increasing the probability of atherothrombotic and ischaemic events. Although "resistance" to ASA in patients with coronary artery or cerebrovascular disease is relatively rare 10 , conversely it is rather frequent in some categories of patients at high risk of thrombotic events, such as those undergoing coronary artery bypass or those with essential thrombocytemia (TE) 1,11 . A recent meta-analysis reported that the overall prevalence of "laboratory defined" aspirin resistance in CVD patients is 24.7% 12 , and although nowadays a standardized test that could identify "resistant patients" is lacking, it's clear that subjects on ASA therapy show a great variability in response to ASA. These patients are classified as "non-responders", or having an absent or incomplete pharmacological response to therapy 3 , for reasons not yet well clarified. The potential mechanisms for "ASA resistance" has been largely investigated [13][14][15] and may be caused by several factors: (a) poor adherence; (b) decreased efficiency of some formulations of ASA to inhibit TxA 2 production [16][17][18][19] ; (c) inadequacy of the standard low doses of daily aspirin (81-100 mg) to inhibit completely COX-1 activity 5,20,21 (to these patients, less protected from thrombotic events, treatment with 100 mg b.i.d. regimen or with a different ASA formulation has been proposed [16][17][18][19] ); (d) competition of ASA with other NSAIDs (for example ibuprofene, indometacine) which could block ASA access at binding site (Ser-530) in COX-1 22 ; (e) unknown interaction with proton pump inhibitors (PPIs) routinely co-prescribed to patients on ASA chronic treatment at high risk of bleeding; (f) esterase-mediated metabolism of ASA: in vivo ASA may undergo hydrolysis to salicylate prior to absorption because of esterases in the gastrointestinal tract 23 . Variations in blood-borne esterase activity have been documented in healthy subjects (HS) 24 , but attempts to correlate activity with pathological states have yielded inconsistent results; (g) platelet multidrug resistance protein-4 (MRP4) overexpression 25,26 ; (h) increased production of platelets with a new COX-1 not inhibited by ASA and able to synthetized TxA 2 10 ; (i) genetic polymorfirsm: presence of COX-1 variants that may be less responsive to aspirin inhibition 27,28 (l) biosynthesis of TxA 2 by pathways that are not blocked by ASA; (m) interventions of coronary revascularization with coronary artery bypass surgery or coronary angioplasty that may induce temporary aspirin resistance 10 ; (n) loss of antiplatelet effect of ASA with prolonged administration: tachyphylaxis.
Among all these different assumptions, the most plausible are (a) the lack of a pharmacodynamics effect; (b) the presence of excessive esterase activity; (c) the presence of an impaired intestinal absorption. To verify these assumptions, it is necessary to evaluate «in vitro» the esterases activity in plasma and blood, and to study «in vivo» the kinetics of ASA and salicylic acid (SA), and TxB2 inhibition.
While different commercial methods are available for TxB2 evaluation, ASA and SA detection needs a fully validated method based on isotope dilution liquid chromatography-mass spectrometry (LC-MS/MS). Consequently, the availability of a suitable analytical procedure is pivotal for any clinical study on aspirin responsiveness.
Here we describe the set-up and optimization of the LC-MS/MS method, and we demonstrate its suitability for further clinical studies by assessing: (1) the "in vitro" ASA-degrading esterase activity in whole blood and plasma of a small cohort of HS; (2) the "in vivo" ASA and SA plasma pharmacokinetics after oral administration of two different ASA formulations (EC-ASA and plain ASA) to HS. To complete our pilot study with pharmacodynamics information, serum TxB 2 was also measured.

LC-MS/MS method for aspirin quantification in plasma.
Optimization of mass spectrometry and liquid chromatography conditions, fragment ion spectra of ASA and SA, and the Compounds Parameter for each analyte are fully described in "Supplemental Data" (Note S1, Figs. S1, S2, and Table S1). It is worthy to note that more than 50% of ASA undergoes in-source fragmentation and forms SA, likewise ASA-d4 forms SA-d4. Liquid chromatography plays an important role in the method development of ASA and SA: in pharmacokinetics (PK) studies, chromatographic separation is pivotal to distinguish between the SA fragment peak generated into the source, and the SA generated in vivo during ASA metabolism. Figure 1 reports an example of chromatograms of all components: retention times were 2.6 min for ASA and ASA-d4; 3.5 min for SA and SA-d4.
Method validation. The developed method was validated according to the FDA and EMA guidance for bioanalytical method validation 29,30 and performed using MultiQuant software 2.1 and GraphPad Prism v. 7.0. The parameters determined were: selectivity, specificity, linearity, precision, accuracy, recovery and stability.
Selectivity was determined by analysing six blank plasma samples, obtained from six different sources, spiked with analytes and internal standards at the respective lowest limit of quantification (LLOQ) concentration.
As reported in FDA guidelines 29 , the lowest standard on the calibration curve should be accepted as LLOQ if the analyte response is at least five times the response compared to blank (solvent) response and analyte peak should be identifiable, discrete and reproducible with a precision of ± 20% and accuracy of 80-120% 29 . Using this analytical method LLOQ for ASA and SA was 20 ng/mL. The limit of detection (LOD) was determined considering the lowest concentration at which signal to noise ratios was > 3 30 . For ASA and SA LOD was 10 ng/ mL. Figure  Intra-and inter-day accuracy and precision were evaluated by the analysis of six replicates containing ASA and SA at different QC levels (20, 60, 400 and 1250 ng/mL for ASA; 20, 60, 200, 500 and 5000 ng/mL for SA) prepared on the same day and in different days. The accuracy was expressed as % bias: (observed concentration)/(nominal concentration) × 100 and precision by % of the CV. The acceptable criteria of the data included accuracy within ± 15% deviation from the nominal value and precision within ± 15% CV except for LLOQ, which was set at ± 20%. Results of intra-day and inter-day accuracy and precision are reported in "Supplemental Data", Tables S2-S3. Intra-day and inter-day precision were within ± 15% for each QC at low medium and high levels and within ± 20% at LLOQ levels. The intra-day and inter-day accuracies were all within 100 ± 15% of the nominal value and were within 100 ± 20% at LLOQ levels.
Proteins precipitation with acetonitrile containing 0.1% of formic acid was trustworthy and provided clean samples. The recoveries of analytes were good and reproducible: for ASA at concentrations of 20, 100, 500 and 2000 ng/mL, recoveries were 85.60 ± 4.74, 68.46 ± 0.67, 57.12 ± 4.15 and 56.02 ± 5.05; for SA at concentrations of 20, 100, 500, 5000 and 8000, they were 72.93 ± 5.0470, 71 ± 0.80, 77.76 ± 4.16, 77.85 ± 0.61 and 71.39 ± 2.06. Internal Standard normalized matrix factor was calculated as reported by Sirok et al. 31  Stability experiments of ASA and SA in plasma samples were carried out by analysing QC samples at two different concentrations for ASA (60, 1250 ng/mL) and three different concentrations for SA (60, 200 and 5000 ng/ mL), under three different conditions: after three freeze-thaw cycles (− 20 °C; 5 °C), after short term storage (6 h) in ice-bath, after long-term stability (2 months at − 20 °C). Short term stability of post-extracted plasma was also evaluated in autosampler at 5 °C for 72 h. No significant degradation of ASA and SA was observed under the conditions studied (Tables S4, S5).
Esterase activity in plasma and blood. The LC-MS/MS method was applied to study the "in vitro" plasma esterase activity in the study subjects. Figure 2 shows the plasma enzyme activity as a function of time. Maximal activity was observed after 120 min of incubation, then declined.
Esterase activity, as a function of substrate concentration, was studied both in whole blood and in plasma (Fig. 3). Using the Michaelis-Menten model, V max and K m were 6.54 ± 1.87 nmol/mL/min and 147.5 ± 64.4 nmol/10 µL in plasma, and 109 ± 20.8 nmol/ mL/ min and 803 ± 171 nmol/10 µL (mean ± SEM, n = 10), in whole blood. Figure 4 shows the plasma concentration-time profile of ASA and SA in each HS studied (n = 6) after EC-ASA and plain-ASA intake. At each time point, ASA and SA were recorded simultaneously in plasma. www.nature.com/scientificreports/ After EC-ASA administration, the drug reached the plasma compartment quite variably among subjects, and never before 2 h from its administration. No ASA or SA signal was observed in plasma from one subject across the 8 h of observation (Fig. 4). The mean of time-concentration pharmacokinetics curves for ASA and SA are reported in Fig. 5. The median maximal ASA plasma concentration (C max ), which was observed between 3 and 6 h, was 571.7 (IQR, 40.28-827.1) ng/mL, and the median AUC was 725.5 (89.8-1222) ng h/mL. SA C max and AUC were 2089 (206-4518) ng/mL and 5920 (600-13,883) ng h/mL, respectively (Fig. 6).

Pharmacokinetics of ASA formulations in healthy subjects.
After administration of plain ASA, drug absorption occurred in all subjects between 0.5 and 1.5 h after intake. ASA C max and AUC were 814 (516-1300) ng/mL and 823 (624-1196) ng h/mL; SA C max and SA AUC were 3520 (2748-5805) ng/mL and 11,663 (8648-18,194) ng h/mL (median, IQR) (Figs. 4 and 5). By comparing the PK parameters between the two ASA formulations, only t max was statistically different both for ASA (p < 0.0001) and SA (p = 0.0002).
The pharmacodynamics of the two tested ASA formulations was evaluated by measuring the serum levels of TxB 2 . Its concentrations were very low (< 10 ng/mL at 24 h from drug intake) in all the subjects after 7 days of daily treatment with 100 mg ASA, regardless of the formulation.

Discussion
Availability of a robust and validated analytical procedure is essential for developing clinical protocols aimed at studying in vivo ASA metabolism.
Several published methods are currently available for the simultaneous quantification of ASA and its metabolite. Sirok et al. 31 recently published a sensitive method for determining ASA and SA using ASA-d4 and SA-d4 as internal standards, a liquid-liquid extraction followed by an evaporation step and chromatographic gradient separation. The method allowed to obtain LLOQs of 1 ng/mL for ASA and 80 ng/mL for SA. Xu et al. 32 extracted ASA and SA from human plasma using protein precipitation with 6-methoxysalicylic acid as internal standard and an isocratic chromatographic separation. Barathi et al. 33 get a LLOQ of 1.09 ng/mL and a relatively rapid chromatographic separation but included an SPE extraction procedure. Chhonker et al. 34 proposed the simultaneous quantitation of ASA and clopidogrel along with their metabolites. The pre-analytical purification included  With the perspective of using the analytical method to study PK of ASA in large clinical trials on patients under chronic treatment, and/or to compare different drug formulations, none of the described methods met our needs, so we elected to develop a rapid, sensitive, robust and cheap method. Despite the desired cheapness and robustness of analysis, aspirin quantification for clinical studies also needs great accuracy. Due to the known in-source aspirin degradation, this can be ensured by a proper chromatographic resolution of ASA and SA peaks  www.nature.com/scientificreports/ and by the use of stable-labelled internal standards, highly recommended for each analytical method based on LC-MS/MS. We propose an isocratic mobile phase that allows high throughput analysis and avoids the tedious long re-equilibration time that is necessary in gradient analysis and, in the meanwhile, we get the baseline separation (~ 1 min) of ASA and SA within 6 min with good S/N ratio (Fig. S4). The method was linear over a large concentrations range as necessary in PK studies, and LLOQ was 20 ng/mL for both analytes. We chose simple protein precipitation with organic solvent avoiding longer and more time-consuming procedures. Of course, we are aware that this would result in a sensibility lower than the ones reported by others, but, if necessary, those lower limits can also be yielded with our method by merely drying the sample and dissolving with a smaller volume of the solvent prior LC-MS/MS analysis. In vivo, ASA undergoes spontaneous and enzymatic hydrolysis by erythrocyte and plasma esterase. In plasma, enzymatic activity is mainly corresponding to that of plasma cholinesterase, while a minor activity is due to albumin acting as an esterase. In whole blood esterase activity is greater than in plasma, and an arylesterase with specificity for aspirin has been isolated from human erythrocytes 35 . Zhou et al. 36 showed that ASA is firstly hydrolysed within erythrocytes by PAF acetylhydrolase; however, this enzyme activity has been verified to be very variable among individuals. In a subsequent study, Zhou et al. 37 showed a statistically significant higher plasma esterase activity in 1928 coronary artery disease patients (16.5 ± 4.4 nmol/mL/min) compared with 298 control subjects (15.1 ± 3.7 nmol/mL/min) (p = 3.4 × 10 -8 ). Excessive esterase metabolism could cause the loss of PD effect of ASA treatment observed in some patients. Here, using plasma of HS and LC-MS/MS, we set-up a precise protocol to study esterase activity in both whole blood and plasma. This was necessary to rule out the contribution of the plasma activity from that of leucocytes and erythrocyte. This method is therefore suitable to investigate cohorts of patients at risk of inadequate response to ASA.
The LC-MS/MS method was also applied to study PK of two different low-dose ASA formulations in human plasma. ASA is rapidly and completely absorbed from the gastrointestinal tract. During and after absorption, ASA is converted to its main metabolite SA. Plain-ASA is rapidly absorbed in the stomach, while the enteric-coating of EC-ASA is resistant to the acid environment and, therefore, it releases ASA for absorption in the alkaline environment of the intestine, rather than in the stomach. From our results, it appears clear that in subjects on treatment with EC-ASA, ASA and SA in plasma peak from 2 to 6 h, with significant inter-individual variability. After witnessed intake of EC-ASA, one subject showed a flat time-concentration serum line for both ASA and SA during all 8 h of observation. This was probably due to an impaired dissolution of the enteric coating in the alkaline environment of the intestine, preventing the release of the active ingredient, or to an impaired intestinal function of the subject, causing a delayed intestinal absorption across the cells membrane. This hypothesis may be plausible as the EC-ASA pharmacodynamics effect was also demonstrated in this patient, with TxB2 concentrations very low (< 10 ng/mL) at 24 h from the drug intake.
On the contrary, plain-ASA was rapidly absorbed with a t max between 0.5 and 1.5 h, with minor inter-individual variability. Despite the reported differences in the absorption time, and excluding the subjects with impaired Figure 6. PK parameters of ASA (above) and SA (below) in 6 HS after EC-ASA (blue) and plain-ASA (red) administration. Only T max of ASA and SA were significantly different between the two treatments. To note that t max after EC-ASA administration could be calculated only on n = 4 patients. www.nature.com/scientificreports/ absorption, the two treatments were comparable in plasma AUCs. Notably, all the patients had inhibition of TxB 2 formation also after 24 h from the drug intake (< 10 ng/mL) 36 .

Conclusion
Our procedure allows studying the analytical fate of ASA "in vivo" following plasma ASA and SA levels. We compared two different ASA formulations: the enteric-coated formulation showed an irregular behaviour and absorption, which was impaired in one subjects. On the contrary, plain-ASA absorption variability was lower, as all tested subjects absorbed the drug efficiently. Our procedures could be used for studying ASA metabolism in patients who display inadequate response to the drug.

LC-MS/MS instrumentation.
The mass spectrometry measurements were performed with the same instrument previously described 38 . Briefly, the liquid chromatography system was an UltiMate 3000 LC Systems Preparation of stocks, calibrators, and quality controls. Two independent calibration curves were prepared in human plasma (250 µL, containing KF 1 mg/mL) by spiking 50 µL of the appropriate working solution giving a final concentration of 20, 50, 100, 200, 500, 1000, 2000 ng/mL for ASA and 20, 50, 100, 200, 500, 1000, 2000, 5000, 8000 ng/mL for SA. A specific SA calibration curve between 0.1 μM and 0.2 mM was constructed to study esterases activity. Quality control (QC) samples were prepared by spiking control human plasma in bulk with ASA and SA at appropriate concentrations in the low, medium and high range: for ASA 20, 60, 500 and 1500 ng/mL; for SA 20, 100, 500, 5000 and 8000 ng/mL. The developed method was validated according to the FDA and EMA guidance for bioanalytical method validation 29,30 . Sample preparation for mass spectrometry. ASA and SA were purified from human plasma using protein precipitation: 250 µL of human plasma were added 25 µL of ASA-d4 (4 µg/mL), 25 µL of SA-d4 (10 µg/ mL) and 700 µL of 0.1% formic acid in acetonitrile. Samples were processed in an ice-bath in order to prevent ASA hydrolysis by esterase. The mixture was vortexed for 1 min, then centrifuged at 14,000g, 4 °C for 10 min. The supernatant was transferred into an analytical vial, and 10 µL were injected into the LC-MS/MS system. Study population. Six HS were recruited among clinicians and students at the Department of Health Sciences. They included three women and three men subjects aged 53 (37-63) yrs and with BMI 22.7 (19.4-27.2); median (CI 25-75%). For haematological data 3 mL blood were collected into K-EDTA tubes and analysed by Coulter analyser. Blood count values were WBC (× 10 9 /L) 6.1 (5.4-6. All subjects, who voluntarily accepted to participate in the study, were informed and authorization was obtained by signing a letter of consent. These subjects were chosen among those who participated to a larger clinical study 1 approved by the institutional local ethical committee (Comitato Etico, Ospedale San Paolo, Milano, Italy). None of the volunteers was under pharmacological treatment. The exclusion criteria were: pregnancy, lactation and nonsteroidal anti-inflammatory drug assumption. A diary containing information about drugs assumption, weight, height, breakfast, age, and withdrawal times was written off for each enrolled subject.