Novel liquid chromatography method based on linear weighted regression for the fast determination of isoprostane isomers in plasma samples using sensitive tandem mass spectrometry detection

Journal of Chromatography B, 1051 (2017) 17–23

Contents lists available at ScienceDirect

Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb

Novel liquid chromatography method based on linear weighted regression for the fast determination of isoprostane isomers in plasma samples using sensitive tandem mass spectrometry detection c ´ Justyna Aszyk a,∗ , Jacek Kot b , Yurii Tkachenko b , Michał Wozniak , Anna Bogucka-Kocka c , a Agata Kot-Wasik a

Department of Analytical Chemistry, Faculty of Chemistry, Gdansk University of Technology, 11/12, Narutowicza Street, 80-233 Gda´ nsk, Poland National Centre for Hyperbaric Medicine, Institute of Maritime and Tropical Medicine in Gdynia, Medical University of Gdansk, Powstania Styczniowego 9B, 81-519 Gdynia, Poland c Chair and Department of Biology and Genetic Medical University of Lublin, W. Chod´zki 4A, 20-093 Lublin, Poland b

a r t i c l e

a b s t r a c t

i n f o

Article history: Received 29 August 2016 Received in revised form 15 February 2017 Accepted 19 February 2017 Available online 1 March 2017 Keywords: F2-isoprostanes Prostaglandins LC–MS/MS analysis Weighted linear regression Plasma

A simple, fast, sensitive and accurate methodology based on a LLE followed by liquid chromatography–tandem mass spectrometry for simultaneous determination of four regioisomers (8-iso prostaglandin F2␣ , 8-iso-15(R)-prostaglandin F2␣ , 11␤-prostaglandin F2␣ , 15(R)-prostaglandin F2␣ ) in routine analysis of human plasma samples was developed. Isoprostanes are stable products of arachidonic acid peroxidation and are regarded as the most reliable markers of oxidative stress in vivo. Validation of method was performed by evaluation of the key analytical parameters such as: matrix effect, analytical curve, trueness, precision, limits of detection and limits of quantification. As a homoscedasticity was not met for analytical data, weighted linear regression was applied in order to improve the accuracy at the lower end points of calibration curve. The detection limits (LODs) ranged from 1.0 to 2.1 pg/mL. For plasma samples spiked with the isoprostanes at the level of 50 pg/mL, intra-and interday repeatability ranged from 2.1 to 3.5% and 0.1 to 5.1%, respectively. The applicability of the proposed approach has been verified by monitoring of isoprostane isomers level in plasma samples collected from young patients (n = 8) subjected to hyperbaric hyperoxia (100% oxygen at 280 kPa(a) for 30 min) in a multiplace hyperbaric chamber. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Plasma F2 -isoprostanes (F2 -IsoPs) are isomers of prostaglandin F2␣ generated from non-enzymatic free-radical per oxidation of arachidonic acid initiated in vivo, including arachidonic esters in phospholipids. They become biomarker of oxidative stress due to their chemical stability and wide availability in numerous biological samples, such as blood plasma, exhaled breath condensate, urine, cerebrospinal fluid or meconium [1–6]. Theoretically 64 different isomers can be generated during this oxidation among which 8-iso-prostaglandin F2␣ is the most recognized isomer, therefore is used as a marker of oxidative stress. Formation of isoprostanes has been implicated in variety of human disorders including: can-

∗ Corresponding author. E-mail address: [email protected] (J. Aszyk). http://dx.doi.org/10.1016/j.jchromb.2017.02.021 1570-0232/© 2017 Elsevier B.V. All rights reserved.

cer [7], neurodegenerative diseases [8], asthma [9–11], pulmonary sarcoidosis [12] or acute respiratory distress syndrome [13]. Recently, several analytical techniques have been applied for determination of isoprostanes in biological specimens. The isoprostanes are commonly measured in plasma or urine by well-established gas chromatography–mass spectrometry (GC–MS) [14–18] and enzyme-linked immunosorbent assay (ELISA) methods [19–21]. Generally immunoassay methods offer high-throughput analysis and require less expensive instrumentation, however it can be possible to measure only one isomer of isoprostane per immunoassay [22,23]. Despite high sensitivity and specificity of GC–MS based technique, it usually requires expensive and time-consuming sample preparation procedure due to the necessity of use of derivatisation and clean-up steps. Nowadays, a liquid chromatography–mass spectrometry (LC–MS) technique is a method of choice for analysis of isoprostanes in biological specimens and can be used as an alternative approach to GC–MS. This is mainly due to the fact that HPLC–MS can overcome above men-

18

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23

tioned problems. This strategy is often supplemented by the use of multiple stable isotope-labeled standards prior to extraction steps to compensate the ion suppression effects and to control the losses of the analytes and reduce contamination [24,25]. Several scientific groups have already reported the determination of isoprostanes with the use of LC-MS/MS [3,22,23,25,26]. Nowadays in bioanalytical approaches based on MS/MS detection more and more often weighted linear regression models are used. The use of OLSR for calibration of analytical method with negligence of heteroscedasticity (variance is increasing with concentration) can lead to significant impairment of accuracy, especially at the lower concentrations of the calibration curve [27–30]. It should be stressed, that isoprostanes are present in biological specimens at very low concentration levels, typically at pg/mL level [6], therefore precision, especially at the low end points of the calibration curve should be maintained for ensuring the reliability of the results. A simple and effective tool to compensate the strong heteroscedasticity in the data and loss in analytical method precision in chromatographic techniques is the use of weighted least squares linear regression (WLSLR), which is useful if the data random errors are not constant across all levels of the calibration curve [29–31]. The purpose of this study was to develop a sensitive, robust and fast method for the separation and determination of trace levels of four isomers of F2 -isoprostanes in plasma samples. The proposed method consisted of one step isolation and clean-up by simple liquid–liquid extraction (LLE) with ethyl acetate at micro scale followed by high performance liquid chromatography coupled to tandem mass spectrometry (HPLC–MS/MS). To ensure the reliability of the developed method, validation in terms of: matrix effect, linearity, limit of quantification (LOQ), limit of detection (LOD), accuracy, intra- and interday precision was performed. Matrix-induced suppression was avoided thank to the use of matrix-matched calibration curve. As a part of assay validation weighted least squares linear regression model (WLSLR) was used to construct calibration curve for determination of trace levels of isoprostanes in plasma samples. Optimized methodology may be valuable for evaluation of differences in concentration of free isoprostanes in human plasma between various groups of patients with different disease severity. We focused also on stressing the relevance and steps to be taken for application of weighting schemes for linear regression analysis for determination of trace levels of isoprostanes. To the best of our knowledge this is the first paper describing the separation of F2-isoprostanes in biological material within less than 8 min using weighted regression at calibration stage. The developed method was applied to plasma samples collected from patients subjected to hyperbaric hyperoxia in a multiplace hyperbaric chamber for quantification of local (i.e. pulmonary) oxidant stress by measurements of isoprostanes level. Our aim was also to investigate whether hyperbaric hyperoxia increases levels of isoprostanes in plasma of human.

2.2. Standard solutions Standards of four isoprostanes (8-isoP, 8,15-isoP, 11-isoP and 15isoP) and internal standard I.S. were diluted in mixture of methanol and water (80:20, v/v) in order to obtain stock solution at concentration 1 ␮g/mL. They were stored at −20 ◦ C until analysis. The calibration solutions of isoprostanes were prepared by diluting the stock solutions of isoprostanes with mixture of acetonitrile and water (15:85, v/v) to obtain 5, 10, 25, 50, 100, 200 pg/mL of isoprostanes in plasma. The calibration curves were constructed freshly in triplicate by plotting peak area ratio of analytes to internal standard vs concentration of isoprostanes in plasma samples. Calibration samples were prepared by spiking plasma samples with four isoprostanes. To account for endogenous isoprostanes, the ratio of endogenous each isoprostane peak area divided by the IS peak area in unspiked plasma sample was subtracted from area ratios of calibration samples (corrected analyte area/IS area ratio). The concentration of internal standard in final extract was maintained at 100 pg/mL. All solutions were kept at 4 ◦ C until analysis. 2.3. Subjects and specimens collection Blood samples were collected from young (age from 21 to 35 years) and healthy humans (n = 8) subjected to hyperbaric hyperoxia (100% oxygen at 280 kPa(a) for 30 min) in a multiplace hyperbaric chamber. Blood (typically 1 mL) obtained from volunteers was collected immediately before and after the exposure and centrifuged (5 min at 8000 rpm followed by 10 min at 1600 rpm, 4 ◦ C). Separated plasma layers were immediately transferred to Eppendorf tubes and kept at −85 ◦ C and stored at this temperature until LC–MS/MS analysis. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Medical University of Gdansk. 2.4. Extraction of isoprostanes from samples An original, novel sample preparation procedure has been evaluated for four isoprostanes isolated from human plasma samples. Analytes were extracted using liquid–liquid extraction (LLE) with ethyl acetate. Briefly, 400 ␮L of plasma sample was transferred to glass tubes and 4 mL of water acidified with hydrochloric acid (pH = 2) and 4 mL of ethyl acetate was added. Before extraction, deuterated internal standard, 8-isoP-d4 , was added to reach the final concentration of 100 pg/mL. Such mixture was shaken for 10 min using vortex. Subsequently, content of vessel was centrifuged for 4 min at 4400 rpm and the organic layer was collected. Extraction with 4 mL of ethyl acetate has been repeated (totally tree times). Then organic solvent layer was combined and evaporated under a gentle stream of nitrogen. The dry residue was redissolved in 400 ␮L of mobile phase (H2 O/ACN + 0.01% FA, 15/85, v/v). Finally, 50 ␮L was subjected for final LC–MS/MS analysis.

2. Material and methods 2.5. HPLC analysis 2.1. Chemicals HPLC-MS grade acetonitrile was purchased from Sigma–Aldrich (Poland) and formic acid was obtained from P.O.Ch (Gliwice, Poland). Ultrapure water was prepared usingHPLC5 system from Hydrolab (Poland). Standards of isoprostanes: 8-iso prostaglandin F2␣ (8-iso-PGF2␣ , 8-isoP), 8-iso-15(R)-prostaglandin F2␣ , (8iso-15(R)-PGF2␣ , 8,15-isoP), 11␤-prostaglandin F2␣ (11␤-PGF2␣ , 11-isoP), 15(R)-prostaglandin F2␣ (15(R)-PGF2␣ , 15-isoP) and internal standard 8-iso prostaglandin F2␣ – d4 (8-iso-PGF2␣ – d4 ) were obtained from SPI-BIO (Montigny le Bretonneux, France).

Isoprostane analyses were performed using LCMS-8050 system (Shimadzu) that consists of binary pump (NEXERA X2 LC-30 AC LIQUID CHROMATOGRAPHY), thermostat (CTO – 20 AC PROMINENCE COLUMN OVEN) and autosampler (NEXERA X2 SIL – 30 AC AUTOSAMPLER). The separation was achieved using a Kinetex (100 × 2.1 mm, 2.6 ␮m) column (Phenomenex), maintained at 40 ◦ C. The flow rate was kept at 0.8 mL/min and the injection volume was set to 50 ␮L. The selected mobile phase consists of H2 O with 0.01% (v/v) of FA (component A) and ACN with 0.01% (v/v) of FA (component B). At

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23

19

Table 1 MRM transitions and analyte-specific mass spectrometry parameters applied for isoprostane quantitation. Analyte

Transition

MRM transition

Q1 pre bias [V]

Collision energy [V]

Q3 pre bias [V]

Four F2 -isoprostanes

Primary Secondary Primary

353.10 > 193.20a 353.10 > 309.25 357.20 > 197.10a

17 27 27

26 19 26

18 29 18

8-iso prostaglandin F2␣ – d4 a

Quantifier ion transition.

Fig. 1. LC–MS/MS chromatogram of (A) blank and extracted plasma samples spiked with isoprostanes (1. 8,15-isoP; 2. 8-isoP; 3. 11-isoP; 4. 15-isoP and 5. 8-iso-PGF2␣ -d4 ) at LOQ level and internal standard (100 pg/mL) (B) blank and extracted plasma sample spiked with internal standard (100 pg/mL).

the beginning 15% of component B was used, then linear gradient from 15% to 35% of component B in 7 min was applied. A 3 min post-run time back to the initial mobile phase composition was used after each analysis. 2.6. MS/MS detection and data analysis The HPLC-ESI-MS/MS analysis was performed with Shimadzu LCMS-8050 triple quadrupole mass spectrometer (Shimadzu, Japan). The ESI source was operated in the negative ion mode. Source parameters were also optimized: nebulizing gas flow, 3 L/min; heating gas flow, 15 L/min; drying gas flow, 5 L/min; interface temperature, 300 ◦ C; heat block temperature, 350 ◦ C and DL temperature, 250 ◦ C. Quantification was performed using multiple reaction monitoring (MRM) transitions of m/z 353.10 → 193.10 for all isoprostanes, 357.20 → 197.10 for 8-iso prostaglandin F2␣ – d4 (IS). The LabSolutions 5.60 SP1 software was used for data acquisition and analysis. Statistical data analysis was performed with SPSS version 11. Differences between the mean values of each cycle were analyzed by Wilcoxon signed-rank test. The differences with p less than 0.05 were considered statistically not-different. 3. Results and discussion 3.1. Separation of analytes For optimalisation of MS/MS operation parameters flow injection analysis (FIA) was done using 1 ␮g/mL solution of each substance. As to internal standard, the most intense fragment was ion with 357 m/z. Four isoprostanes are deprotonated in negative ion mode to form ions with 353 m/z. The following mass transitions were therefore selected to quantify four isoprostanes: m/z 353.10 → 193.10 and m/z 357.20 → 197.10 for 8-iso prostaglandin F2␣ – d4 . To verify the presence of four isoprostanes in each sample the quantifier/qualifier ion ratio was confirmed. The following acceptance criteria for positive identification of analytes was set: the established ratios ((m/z 353.10 > 193.10):(m/z 353.10 > 309.25)) of the peak areas of four isoprostanes observed for the two tran-

sitions in the investigated plasma samples should not change by ±20% for those employed after analysis of standards. These acceptance criteria was set according to European Commission Council Directive 96/23/EC;SANCO/1805/2000. Samples not conforming to these permitted tolerances were omitted from the data sets. The optimum detection parameters for analytes are presented in Table 1. Isoprostanes from plasma samples were extracted using developed procedure and further analyzed by LC–MS/MS. As it can be seen in Fig. 1(A) all isomers (8-isoP, 8,15-isoP, 11-isoP, 15-isoP) are clearly separated from each other. The retention times of four isoprostanes and IS were 5.43, 5.58, 5.74, 6.13 and 5.56, respectively. HPLC–MS/MS analysis shows that in spiked plasma extracts obtained from human plasma samples which underwent the LLE purification procedure additional peaks of unknown compounds appeared in chromatogram. Unknown peaks can be naturally occurring isomers of isoprostanes and no coelution with determined compounds was observed. 3.2. Method validation 3.2.1. Evaluation of matrix effect An estimation of matrix effect was performed by comparing the slopes of the calibration curves performed in plasma extracts and in solvent according to strategy described by [32,33]. Therefore, two standards lines were prepared: (1) in plasma extracts spiked after extraction with analytes and (2) in mobile phase. The concentration of internal standards in both cases was kept at 100 pg/mL. Matrix effect (ME) was calculated according to equation: ME [%] =

matrix-matched curve slope × 100% solvent curve slope

(1)

The value of ME below 100% indicates suppression of signal. As it can be seen additionally from data presented in Table 2, the calculated matrix effects for each analyte were relevant and showed suppression of the signal. Additionally, visual analysis of the curves (Fig. 2) clearly indicates an exhibition of matrix-induced decreased responses by four isoprostanes.

20

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23

Table 2 Regression parameters and matrix effect for calibration curve with standard solutions and calibration curve with matrix matched standards. Analyte

Calibration curve equation for plasma

Calibration curve equation for solvent

The ratio of the slopes for matrix/standard solution

Matrix effect [%]

8,15-isoP 8-isoP 11-isoP 15-isoP

y = 0.0139x + 0.1428 y = 0.0141x − 0.1213 y = 0.0069x + 0.0419 y = 0.0082x − 0.0076

y = 0.0175x + 0.1480 y = 0.0178x + 0.1960 y = 0.0080x + 0.0850 y = 0.0138x + 0.1118

0.79 0.79 0.85 0.59

79 79 85 59

Fig. 2. Analytical curves (LC–MS/MS) in solvent (red quadrate) and in the matrix extract (blue circle) for the studied isomers of isoprostanes. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Therefore matrix-matched approach instead of external calibration was applied in order to obtain reliable measurements. For compensation of all matrix effects on the analytes and monitoring the recovery during sample analysis, the internal standard calibration was applied.

3.2.2. Linearity and calibration range Matrix matched-calibration curves were prepared in plasma samples at six concentration levels ranging from 5 to 200 pg/mL for each analyte. Calibration curves for all analytes exhibited coefficient of determinations with the value over 0.99. However, the value of coefficient of determination (r2 ) being greater than 0.99 is often easily misinterpreted in evaluation of regression mode and linearity. The variances are often increasing with concentration (data sets are heteroscedastic). The negligence in the heteroscedasticity of the data, can lead to significant loss of accuracy despite the acceptable r2 values, especially at the lower end of the calibration range of analytical curve [34].

Test of homoscedasticity of the data was conducted by performing F-test in accordance with following equation: F exp =

s22 s12

(2)

where s2 and s1 is a variance obtained at upper limit of quantitation (200 pg/mL) and lower limit of quantitation (5 pg/mL), respectively. When experimental F value is bigger than the limiting tabled F value for appropriate degrees of freedom and confidence level (Ftab < Fexp ), it indicates that data are heteroscedastic [17,28,29,31]. The limiting tabled F-value was taken from the table with 2 degrees of freedom and 95% confidence level (df1, df2 = n − 1) and is equal to 19. Results obtained in F-test for 8-isoP, 8,15-isoP, 11-isoP and 15isoP were 150, 1258, 58,95, respectively. Experimental F-value for each analyte was greater than tabled F-value (Ftab < Fexp ). Assumption of homoscedasticity was not met for obtained results. In order to circumvent this problem, weighted least squares linear regression model was applied. To decide which weighting factor should

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23 Table 3  %RE) for isoprostanes. Respective sum of the relative errors ( wi

1 1/x 1/x2 1/x0.5 1/y 1/y2 1/y0.5



Table 5 Recovery data obtained for isoprostanes extracted from human plasma spiked at three different concentration levels: 10, 50, 200 pg/mL.

RE [%]

Recovery (CV) [%] (n = 3)

8,15-isoP

8-isoP

11-isoP

15-isoP

284 212 176 252 223 189 253

541 135 106 199 201 138 249

391 236 228 235 233 215 230

1429 345 445 425 259 555 402

The underlined values represents the smallest sum of the relative errors.

be applied, values of relative errors were calculated according to equation: cmeasured − cnominal × 100% cnominal

Analyte

10 pg/mL

50 pg/mL

200 pg/mL

8,15-isoP 8-isoP 11-isoP 15-isoP

98.9 (4.5) 97.8 (2.0) 97.9 (3.5) 99.2 (4.6)

101.9 (1.2) 102.4 (4.5 99.5 (3.5) 101.2 (2.3)

99.4 (6.0) 99.6 (6.4) 100.8 (2.5) 100.5 (3.0)

3.2.3. Limit of detection (LOD) and limit of quantification (LOQ) The limit of detection (LOD) was calculated based on formula: LOD = 3.3*Sb/a (a – slope of the calibration curve, Sb – standard deviation of the intercept), while value of LOQ is adequate to 3 times LOD. The results are presented in Table 4. These results showed that the method was very sensitive and enabled the detection and quantification of all of the substances investigated in the study.

Fig. 3. Percentage of relative errors (%RE) vs concentration obtained for unweighted model and models: 1/x2 , 1/y, 1/y2 for 8-iso-PGF2␣ as an example.

RE [%] =

21

(3)

where cmeasured is concentration obtained from the weighted equation, and cnominal is nominal standard concentration. According to appropriate formulas [29,32], weighted regression equation parameters: a (slope), b (intercept) and R2 (correlation coefficient) values were calculated. The weighting factor, which presents the smallest sum of the relative errors calculated according to Eq. (3) and exhibits the best distribution scatter around the axis of concentrations. In this study, the evaluation of six empirical weights such as 1/x0.5 , 1/x, 1/x2 , 1/y0.5 , 1/y and 1/y2 have been studied. In Table 3 the sums of the relative errors for each of the seven models for each isoprostane have been summarized. As an example, %RE plots for unweighted model and three weighted regressions of the 8-isoP across the whole concentration range are shown in Fig. 3. As can be seen, the weighting factor 1/x2 exhibited the best %RE distribution scatter among three presented weighted models and therefore was chosen. The same approach was applied to other analytes. For the quantification of isoprostanes in plasma samples new equations of the calibration curves with weighted parameters a and b was applied.

3.2.4. Trueness and repeatability Trueness of developed method (expressed in terms of recoveries and coefficient of variation – CV) was evaluated by spiking plasma samples with isoprostanes at three different concentrations levels: 10, 50, 200 pg/mL. The recoveries were evaluated by comparing the peak area ratios of each analyte to internal standard for spiked and extracted blood samples with analogous ratios for extracts spiked with standards post – extraction. Each analysis was performed in three replicates. The results of recovery experiments are presented in Table 5. The values of recoveries ranging from 97.8 to 101.9%), with relative standard deviations (RSDs) below 7% indicate good accuracy and precision of developed method. The repeatability (intra-day precision of the method) expressed as RSD was determined by analysis of plasma extract at three spiking levels (n = 6) during one day. Inter-day precision of the method was evaluated by analysis of samples at one spiking level (50 pg/mL) over three days. The intra-day repeatability ranged from 2.1 to 3.5%, while inter-day reproducibility ranged from 0.1 to 5.1%. Developed method provided to be applicable in determination of isoprostanes in human plasma due to satisfactory repeatability and reproducibility.

3.2.5. Analysis of real samples To demonstrate applicability of the validated method series of human plasma samples collected from patients exposured to hyperbaric hyperoxia exposure (100% oxygen at 280 kPa(a) for 30 min) in a multiplace hyperbaric chamber were analyzed. The results shown in Fig. 4 revealed elevated concentrations in plasma of only 8-iso-PGF2␣ , because concentration of other three isoprostanes was below LOQ. The post-exposure median for 8iso-PGF2␣ [25–75th percentiles] was 41 [21–72] pg/mL, and the pre-exposure median was 38 [14–41] pg/mL. Concentration of plasma free 8-iso-PGF2␣ didn’t increase significantly after hyperbaric hyperoxia (27.5 ± 13.1 vs 46.2 ± 25.6 pg/mL, respectively; p = 0.26).

Table 4 Regression parameters for weighted calibration curves for four isoprostanes. Analyte 8.15-isoP 8-isoP 11-isoP 15-isoP

Weighted factor 2

1/x 1/x2 1/y2 1/y

a

b

Sb

R2

LOD [pg/mL]

LOQ [pg/mL]

0.0181 0.0269 0.0069 0.0051

0.0317 0.6302 0.0511 −0.0247

0.0072 0.0075 0.0045 0.0032

0.991 0.998 0.992 0.990

1.3 1.0 2.1 2.1

3.9 3.0 6.3 6.3

22

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23

Conflict of interest The authors declare no conflict of interest. Acknowledgment This study was partially supported by the European project of the PHYPODE Marie Curie Initial Training Networks (FP7-PEOPLE2010-ITN). References

Fig. 4. Box plot of plasma 8-iso-PGF2␣ levels established from patients (n = 8) preand post-hyperbaric hyperoxia exposure. Data shown are plasma isoprostane concentrations in pg/mL. The median, 25–75th percentile (box) and 5–95th percentiles (whiskers) for 8 subjects are shown. Mean isoprostanes concentrations and relative standard deviations (RSD) were calculated for each plasma sample analyzed in triplicate. The plasma 8-iso-PGF2␣ values did not change significantly after hyperbaric hyperoxia exposure (p = 0.26) compared to controls.

4. Conclusions A LC–ESI-MS/MS method for the detection and quantitation of 8iso-PGF2␣ , 8-iso-15(R)-PGF2␣ , 11-PGF2␣ and 15(R)-PGF2␣ based on the use of weighted linear regression in plasma samples was developed. All target compounds can be measured simultaneously in multiple reaction monitoring mode (MRM) with the transitions m/z 353 → 193 (regioisomers) and m/z 357 → 197 (IS) within a 7 min separation. Our study describes an approach utilizing the weighted least squares linear regression (WLSLR) in order to counteract observed heteroscedasticity of the data and significantly improve the accuracy, particularly at the lower end of the calibration curve. Isoprostanes are present at low concentrations in human biological fluids, therefore, maintaining precision at the low end of the curve is herein especially important. Our study stressed also the steps to be taken to apply the weighting schemes for quantification of trace levels of isoprostanes in human plasma samples. The combination of WLSLR approach together with matrix-matched calibrations standards allows for accurate, reproducible and reliable validation and quantification of isoprostanes over the whole calibration range as well as for obtaining satisfying limits of detection, which were comparable and sometimes even better than those described in other studies. Our method stands out good reproducibility and high sensitivity and it was considered efficient, precise and accurate for all tested analytes. A deuterated internal standard was used to obtain a high quantitation precision by compensation of detector’s response. The analytical validation parameters were satisfied in terms of analytical curve, precision, trueness, limit of detection and limit of quantification. Matrix effect has been also established showing clear suppression in MS signal during electrospray ionization. To avoid and compensate remaining matrix effects, matrix matched-calibration was applied. The calculated limits of quantification ranging from 3.9 to 6.3 pg/mL, with clear evidence that the lower values were quantified based on the lowest concentration of the calibration curve within the range of values validated with acceptable trueness and precision. The presented accurate, reproducible and high-throughput method combines simplicity and specificity during biological sample preparation using easy LLE together with selectivity of fast HPLC separation followed by sensitive ESI-MS/MS detection and it able to be used in large clinical studies and standard clinical laboratories. Despite the fact, that weighted linear regression is more sophisticated than ordinary regression, it should be performed in order to obtain results with greater trueness.

[1] K.C. Bohnstedt, B. Karlberg, L.O. Wahlund, M.E. Jönhagen, H. Basun, S. Schmidt, Determination of isoprostanes in urine samples from Alzheimer patients using porous graphitic carbon liquid chromatography–tandem mass spectrometry, J. Chromatogr. B Biomed. Sci. Appl. 796 (2003) 11–19. [2] A.W. Taylor, R.S. Bruno, B. Frei, M.G. Traber, Benefits of prolonged gradient separation for high-performance liquid chromatography–tandem mass spectrometry quantitation of plasma total 15-series F 2-isoprostanes, Anal. Biochem. 350 (2006) 41–51. [3] M. Janicka, P. Kubica, A. Kot-Wasik, J. Kot, J. Namie´snik, Sensitive determination of isoprostanes in exhaled breath condensate samples with use of liquid chromatography–tandem mass spectrometry, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 893 (2012) 144–149. [4] T.J. Montine, W.R. Markesbery, J.D. Morrow, L.J. Roberts, Cerebrospinal fluid F2-isoprostane levels are increased in Alzheimer’s disease, Ann. Neurol. 44 (1998) 410–413. [5] M.P. Lu, L.Z. Du, W.Z. Gu, Z.Z. Yu, X.X. Chen, Z.S. Yu, Anti-inflammation and anti-oxidation effects of recombinant human superoxide dismutase on acute lung injury induced by meconium aspiration in infant rats, Zhejiang Da Xue Xue Bao Yi Xue Ban 34 (2005) 55–59. [6] G.L. Milne, H. Yin, K.D. Hardy, S.S. Davies, L.J. Roberts 2nd, Isoprostane generation and function, Chem. Rev. 111 (2011) 5973–5996. [7] D.A. Barocas, S. Motley, M.S. Cookson, S.S. Chang, D.F. Penson, Q. Dai, G. Milne, L.J. Roberts 2nd, J. Morrow, R.S. Concepcion, J.A.J.A. Smith Jr., J.H. Fowke, Oxidative stress measured by urine F2-isoprostane level is associated with prostate cancer, J. Urol. 185 (2011) 2102–2107. [8] K.M. Kim, B.H. Jung, K.J. Paeng, I. Kim, B.C. Chung, Increased urinary F(2)-isoprostanes levels in the patients with Alzheimer’s disease, Brain Res. Bull. 64 (2004) 47–51. [9] O. Keskin, S. Balaban, M. Keskin, E. Kucukosmanoglu, B. Gogebakan, M.Y. Ozkars, S. Kul, H. Bayram, y. Coskun, Relationship between exhaled leukotriene and 8-isoprostane levels and asthma severity, asthma control level, and asthma control test score, Allergol. Immunopathol. (Madr.) 42 (2014) 191–197. [10] S. Carraro, P.E. Cogo, I. Isak, M. Simonato, M. Corradi, V.P. Carnielli, E. Baraldi, EIA and GC/MS analysis of 8-isoprostane in EBC of children with problematic asthma, Eur. Respir. J. 35 (2010) 1364–1369. [11] S. Zanconato, S. Carraro, M. Corradi, R. Alinovi, M.F. Pasquale, G. Piacentini, F. Zacchello, E. Baraldi, Leukotrienes and 8-isoprostane in exhaled breath condensate of children with stable and unstable asthma, J. Allergy Clin. Immunol. 113 (2004) 257–263. [12] K. Psathakis, G. Papatheodorou, M. Plataki, P. Panagou, S. Loukides, N.M. Siafakas, D. Bouros, 8-Isoprostane, a marker of oxidative stress, is increased in the expired breath condensate of patients with pulmonary sarcoidosis, Chest 125 (2004) 1005–1011. [13] P.V. Price, B.W. Christman, Exhaled breath condensate isoprostanes are elevated in patients with acute lung injury or ARDS, Chest 114 (1998) 1653–1659. [14] W. Liu, J.D. Morrow, H. Yin, Quantification of F 2-isoprostanes as a reliable index of oxidative stress in vivo using gas chromatography-mass spectrometry (GC–MS) method, Free Radic. Biol. Med. 47 (2009) 1101–1107. [15] G.L. Milne, B. Gao, E.S. Terry, W.E. Zackert, S.C. Sanchez, Measurement of F 2-isoprostanes and isofurans using gas chromatography–mass spectrometry, Free Radic. Biol. Med. 59 (2013) 36–44. [16] D. Tsikas, M.T. Suchy, Protocols for the measurement of the F 2-isoprostane, 15 (S)-8-iso-prostaglandin F 2␣, in biological samples by GC–MS or GC–MS/MS coupled with immunoaffinity column chromatography, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 1019 (2016) 191–201. [17] E. Mas, F. Michel, A. Guy, V. Bultel, Y. Falquet, P. Chardon, J.C. Rossi, J.P. Cristol, T. Durand, Quantification of urinary F 2-isoprostanes with 4 (RS)-F 4t-neuroprostane as an internal standard using gas chromatography–mass spectrometry: application to polytraumatized patients, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 872 (2008) 133–140. [18] D.R. Briskey, G.R. Wilson, R.G. Fassett, J.S. Coombes, Optimized method for quantification of total F2-isoprostanes using gas chromatography–tandem mass spectrometry, J. Pharm. Biomed. Anal. 90 (2014) 161–166. [19] D. Il’yasova, J.D. Morrow, A. Ivanova, L.E. Wagenknecht, Epidemiological marker for oxidant status: comparison of the ELISA and the gas chromatography/mass spectrometry assay for urine 2,3-dinor-5,6-dihydro-15-F 2t-isoprostane, Ann. Epidemiol. 14 (2004) 793–797.

J. Aszyk et al. / J. Chromatogr. B 1051 (2017) 17–23 [20] K. Samitas, D. Chorianopoulos, S. Vittorakis, E. Zervas, E. Economidou, G. Papatheodorou, S. Loukides, M. Gaga, Exhaled cysteinyl-leukotrienes and 8-isoprostane in patients with asthma and their relation to clinical severity, Respir. Med. 103 (2009) 750–756. [21] Y. Li, V. Chongsuvivatwong, A. Geater, A. Liu, Exhaled breath condensate cytokine level as a diagnostic tool for obstructive sleep apnea syndrome, Sleep Med. 10 (2009) 95–103. [22] T. Sicilia, A. Mally, U. Schauer, A. Pähler, W. Völkel, LC–MS/MS methods for the detection of isoprostanes (iPF2alpha-III and 8,12-iso-iPF2alpha-VI) as biomarkers of CCl4 -induced oxidative damage to hepatic tissue, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 861 (2008) 48–55. [23] T. Petrosino, M. Serafini, Matrix effect in F2-isoprostanes quantification by HPLC–MS/MS: a validated method for analysis of iPF2␣-III and Ipf2␣-VI in human urine, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 96 (2014) 100–106. [24] J.R. Bain, R.D. Stevens, B.R. Wenner, O. Ilkayeva, D.M. Muoio, C.B. Newgard, Metabolomics applied to diabetes research: moving from information to knowledge, Diabetes 58 (2009) 2429–2443. [25] W. Yan, G.D. Byrd, M.W. Ogden, Quantitation of isoprostane isomers in human urine from smokers and nonsmokers by LC–MS/MS, J. Lipid Res. 48 (2007) 1607–1617. ´ [26] M. Janicka, A. Kot-Wasik, J. Paradziej-Łukowicz, G. Sularz-Peszynska, A. Bartoszek, J. Namie´snik, LC–MS/MS determination of isoprostanes in plasma samples collected from mice exposed to doxorubicin or tert-butyl hydroperoxide, Int. J. Mol. Sci. 14 (2013) 6157–6169.

23

[27] E. Rozet, R.D. Marini, E. Ziemons, B. Boulanger, P. Hubert, Advances in validation, risk and uncertainty assessment of bioanalytical methods, J. Pharm. Biomed. Anal. 55 (2011) 848–858. [28] T. Singtoroj, J. Tarning, A. Annerberg, M. Ashton, Y. Bergqvist, N.J. White, N. Lindegardh, N.P. Day, A new approach to evaluate regression models during validation of bioanalytical assays, J. Pharm. Biomed. Anal. 41 (2006) 219–227. [29] A.M. Almeida, M.M. Castel-Branco, A.C. Falcão, Linear regression for calibration lines revisited: weighting schemes for bioanalytical methods, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 774 (2002) 215–222. [30] J. Tellinghuisen, Weighted least squares in calibration: the problem with using “quality coefficients” to select weighting formulas, J. Chromatogr. B Biomed. Sci. Appl. 872 (2008) 162–166. [31] B. Matuszewski, M. Constanzer, C. Chavez-Eng, Strategies for the assessment of matrix effect in quantitative bioanalytical methods based on HPLC–MS/MS, Anal. Chem. 75 (2003) 3019–3030. [32] C.P. Da Silva, E.S. Em’idio, M.R.R. de Marchi, Method validation using weighted linear regression models for quantification of UV filters in water samples, Talanta 131 (2015) 221–227. ´ [33] M. Wiergowski, M.K. Wozniak, M. Kata, M. Biziuk, Determination of MDPBP in postmortem blood samples by gas chromatography coupled with mass spectrometry, Monatsh. Chem. 147 (2016) 1415–1421. [34] F. Raposo, Evaluation of analytical calibration based on least-squares linear regression for instrumental techniques: a tutorial review, Trends Analyt. Chem. 77 (2016) 167–185.