Determination of Antidepressants in Human Plasma by Modified Cloud-Point Extraction Coupled with Mass Spectrometry

Cloud-point extraction (CPE) is rarely combined with liquid chromatography coupled to mass spectrometry (LC–MS) in drug determination due to the matrix effect (ME). However, we have recently shown that ME is not a limiting factor in CPE. Low extraction efficiency may be improved by salt addition, but none of the salts used in CPE are suitable for LC–MS. It is the first time that the influences of a volatile salt—ammonium acetate (AA)—on the CPE extraction efficiency and ME have been studied. Our modification of CPE included also the use of ethanol instead of acetonitrile to reduce the sample viscosity and make the method more environmentally friendly. We developed and validated CPE–LC–MS for the simultaneous determination of 21 antidepressants in plasma that can be useful for clinical and forensic toxicology. The selected parameters included Triton X-114 concentration (1.5 and 6%, w/v), concentration of AA (0, 10, 20 and 30%, w/v), and pH (3.5, 6.8 and 10.2). The addition of 10% of AA increased recovery twice. For 20 and 30% (w/v) of AA, three phases were formed that prolonged the extraction process. The developed CPE method (6% Triton X-114, 10% AA, pH 10.2) was successfully validated through LC–MS/MS simultaneous determination of 21 antidepressants in human plasma. The linearity was in the range of 10–750 ng/mL (r2 > 0.990).


Introduction
Cloud-point extraction (CPE) is a modification of liquid-liquid extraction (LLE) that is more friendly to the environment and users, mainly due to lower consumption of organic solvents according to the global trend of "green chemistry." CPE has also advantages over solid phase extraction (SPE), such as faster and cheaper optimization processes and no need for expensive equipment [1]. In CPE, the surfactant (such as Triton X-114) in a concentration above the critical micelle concentration (CMC) can exist as a homogeneous isotropic liquid, which separates into two isotropic phases with different concentrations of surfactant. Sample ingredients are separated into the surfactant micelle-rich phase (hydrophobic components when nonionic surfactant is used) and the micelle-poor phase (hydrophilic components) [2].
Nowadays, the most reliable technique of pharmaceutical determination is liquid chromatography coupled to mass spectrometry (LC-MS). However, there are only a very few papers reporting LC-MS coupled with CPE in drug determination, i.e., in the determination of memantine in rat plasma [3], The concentration of Triton X-114, used as the surfactant in CPE, was very carefully selected. This was because the overly low concentration of Triton X-114 decreases recovery and prevents  The initial factors of extraction included pH 6.8, temperature 60 • C, and 20% of AA. To reach pH 6.8, only water was used. Thus, pH 6.8 was selected to reduce the time of the process, reduce the cost, and diminish the risk of contamination. AA concentration was selected as a medium value from all tested in the experiment. The level of antidepressants corresponded to a plasma concentration of 100 ng/mL. Many factors affect CPE efficiency. In this paper, the influences of surfactant concentration (Triton X-114), volatile salt (AA-0, 10, 20, and 30% (w/v)), and extraction pH (about 3.5, 6.8, 10.2) was tested to select the optimal conditions for the simultaneous determination of 21 antidepressants. Other parameters, i.e., volume of reagents and equilibrium time, were used as described previously [2,31]. However, in the current study higher temperature was selected (60 • C). The optimal temperature for CPE is 15-20 • C greater than the cloud point of the surfactant, which in the case of Triton X-114 is 23 • C, and was reviewed to be 40-60 • C [32]. With an increase in temperature, the efficacy of the extraction increases, and the volume of the surfactant-rich phase decreases due to the disruption of the hydrogen bonds and the dehydration of the phase [32].

Development of Cloud-Point Extraction
The optimal conditions (concentration of Triton-X-114 and AA, pH sample) were defined as the ones showing the highest mean recovery and the lowest absolute ME (ME A ) (Equation (1)) [33], for simultaneously analyzing all 21 antidepressants. The influences of selected parameters are presented in  and summarized in Figure 6. The chromatography separation was not optimized. We used the gradient elution mode, eluent flow rate, and analytical column commonly used in our analysis [31].
Molecules 2020, 25, x FOR PEER REVIEW 5 of 27 effective separation of the aqueous and micellar phases. On the other hand, too high concentration of surfactant is associated with a high volume of the micelle-rich phase, and the unwanted dilution of the analyte [5,10,32]. The range of tested Triton X-114 concentrations, added to samples in multicompound determination, varied from 1.5 to 9% (w/v) [1,2,4,5,8,32], with the most frequently used concentration being 4% [32]. In this study, two variants of Triton concentration (1.5 and 6%) were studied. The effects of the surfactant concentration on recovery and MEA are shown in Figure 2. sample pH 6.8; ammonium acetate content-20% (w/v). Antidepressants level corresponds to a plasma concentration of 100 ng/mL Higher mean recovery was observed with higher Triton X-114 concentration (the experiment was repeated three times) (p < 0.001). Results are presented as means and standard deviations.
The volume of the micellar rich phase was about 100 μL, and the initial volume of the sample was 1 mL, so the analytes were concentrated about 10 times. Over twofold increases in recovery at higher Triton X-114 concentration were noted for CLO, DOX, FLV, IMI, MAP, MIR, SER, and TRA; over threefold increases for AMI, CIT, DES, FLX, MOC, NOR, OPI, PAR, PRO, and TIA; and the Effects of Triton X-114 concentration on (a) recovery (p < 0.001), (b) absolute matrix effect (ME A ) (p = 0.182). Extraction conditions: equilibrium temperature-60 • C; equilibrium time-20 min; sample pH 6.8; ammonium acetate content-20% (w/v). Antidepressants level corresponds to a plasma concentration of 100 ng/mL Higher mean recovery was observed with higher Triton X-114 concentration (the experiment was repeated three times) (p < 0.001). Results are presented as means and standard deviations. The effect of ammonium acetate concentration (0, 10, 20, and 30% (w/v)) on recovery (p < 0.001). Extraction conditions: equilibrium temperature-60 • C; equilibrium time-20 min; sample pH, 6.8; concentration of Triton X-114-6% (w/v). Antidepressants level corresponds to a plasma concentration of 100 ng/mL. Results are presented as means and standard deviations. Mean recoveries for 0% AA (dashed line), 10% AA (dotted line), 20% AA (dotted-dashed line), and 30% AA (bold dashed) are shown.The highest increases of recovery, as a result of AA addition, were observed for AMI, PAR, PRO, TIA, and TRA, whereas the weakest effect was recorded for both FLV and MIA. No increase in recovery for VEN or decrease for OPI was observed. The increase of recovery (10% vs. 0 % (w/v)) was correlated (r s = 0.465, p = 0.034) with the calculated octanol-water partition coefficient (cLogP).   MEA can be observed as signal suppression (below 100%) or an enhancement (above 100%) in the presence of a matrix. Significant MEA <85% or >115% was observed for 10, 10, 12, and 16 out of 21 studied drugs for 0, 10, 20, and 30% (w/v) of AA, respectively. The differences were not statistically significant (χ 2 test, p = 0.198), nor were they in statistical analysis performed only for the compounds with significant MEA (p = 0.244). In all variants of AA, insignificant MEA was observed for five compounds: FLX, FLV, MAP, PAR, and SER. For 0 and 10% (w/v) of AA, that was additionally observed for AMI, IMI, and TRI, whereas for 0, 10, and 20% (w/v) of AA, that was the case for DES, NOR, and PRO.
The effect of salt adding (sodium chloride) on the recovery of only one antidepressant Extraction conditions: equilibrium temperature-60 • C; equilibrium time-20 min; sample pH-6.8; concentration of Triton X-114-6% (w/v). Antidepressants level corresponds to a plasma concentration of 100 ng/mL. Results are presented as means and standard deviations. ME A between 85% (dotted line) and 115% (dotted line) was considered insignificant.
In alkaline pH, more particles occur in molecular form; thus, higher affinity to nonionic micelles and higher extraction recovery were observed. At acidic pH, the majority of the antidepressants were in ionic form and had their lowest affinity for nonionic micelles. That resulted in the loss of some of the analytes during phase separation and low extraction efficiency. Moreover, the separation of two phases-the aqueous and micelle-rich ones-in the acidic and neutral pHs, was not as pronounced as in alkaline pH. Statistical analysis (p < 0.001) indicated significant differences in recovery of the extraction performed at various pH values. However, no relationship between the pH of the extraction environment and the number of compounds exhibiting significant MEA was observed (test χ 2 ; p = 0.514). Statistical analysis carried out for compounds with significant MEA (ANOVA, p = 0.013) showed statistically significant differences in MEA between pHs. The results of the post hoc test revealed that these differences occurred between the alkaline and acidic environment (p = 0.004). The mean MEA was 6.1% lower for samples with a pH of 3.5 than those of pH 10.2. However, it should be emphasized that in acidic conditions, the basic compounds have less affinity to nonionic surfactant micelles, which resulted in the lower analytical signal and lower recovery. Moreover, the separation into two phases was not distinct in acidic pH or neutral pH, which caused the loss of analytes during phases separation. Thus, pH 10.2 was selected as optimal to use in the final method. Madej and Persona [2] developed a CPE to extract basic compounds (pKa 9.0−9.5) at pH 12, with Triton X-114 at 3.25% (w/v), without salt addition. The recoveries were 26.2, 81.4, 77.3, and 98.5% for paracetamol, amitriptyline, clomipramine, and promazine, respectively [2]. In this paper, the number of isolated compounds from plasma in the basic environment (pH 10.2) was much higher (n = 21 vs. n = 4) and the pKa range was wider (pKa 6.0−10.5). A higher mean recovery (86.2%) was observed than in the reported experiment (70.9%). Compared to the recovery of particular compounds, higher recoveries were achieved for AMI (81.4 vs. 85.4%) and CLO (77.3% vs. 90.1%). OPI was extracted as a neutral compound at pH 6, which resulted in 21.8% recovery, while in this study at pH 10.2, the recovery was 102.5% [2]. Thus, the extraction at pH 10.2 and the addition of AA allow for efficient isolation of antidepressants from biological matrices by the CPE method. Additionally, the extraction environment does not significantly affect the number of compounds exhibiting a matrix effect. Some compounds revealed significant MEA in all tested conditions (CIT, DOX, TRA, VEN). All of them showed a significant surfactant effect (the ratio of peak area of the analyte in a solvent with the surfactant to that without the surfactant) previously [31]. That means that the MEA observed is not related to impurities extracted from plasma, but the presence of Triton in the sample. PCA was  The concentration of Triton X-114, used as the surfactant in CPE, was very carefully selected. This was because the overly low concentration of Triton X-114 decreases recovery and prevents effective separation of the aqueous and micellar phases. On the other hand, too high concentration of surfactant is associated with a high volume of the micelle-rich phase, and the unwanted dilution of the analyte [5,10,32]. The range of tested Triton X-114 concentrations, added to samples in multi-compound determination, varied from 1.5 to 9% (w/v) [1,2,4,5,8,32], with the most frequently used concentration being 4% [32]. In this study, two variants of Triton concentration (1.5 and 6%) were studied. The effects of the surfactant concentration on recovery and ME A are shown in Figure 2.
The volume of the micellar rich phase was about 100 µL, and the initial volume of the sample was 1 mL, so the analytes were concentrated about 10 times. Over twofold increases in recovery at higher Triton X-114 concentration were noted for CLO, DOX, FLV, IMI, MAP, MIR, SER, and TRA; over threefold increases for AMI, CIT, DES, FLX, MOC, NOR, OPI, PAR, PRO, and TIA; and the highest for VEN (4.5 times). A lower than twofold increase was noted only for MIA. Higher recoveries at 6% Triton were expected based on the partition coefficient equation. The partition coefficient is characteristic for the compound and the extraction conditions (e.g., type of surfactant, pH). It is calculated as the ratio of antidepressant concentration in the surfactant to that in the aqueous phase [34]. Thus, with the increase of the surfactant rich phase volume, the mass of the extracted compound increases as well.
Significant ME A values were observed for four (CIT, DOX, TRA, and VEN) and nine antidepressants (CIT, DES, DOX, MAP, MIA, MOC, OPI, TRA, and VEN) out of 21 analyzed at 1.5 and 6% Triton X-114, respectively. In contrast to 6%, at 1.5% Triton X-114 concentration, ME A was not observed for DES, MAP, MIA, MOC, and OPI. Higher surfactant concentration is related to a higher risk of ME A , as reported previously [31]. The possible reason is the interference of surfactant with droplet evaporation or higher recovery not only of analytes, but also impurities of samples that interfere with the analyte ionization. However, in the current study, the differences in the occurrence of ME A for all 21 antidepressants between the two surfactant concentrations were not statistically significant (p = 0.182). The concentration of 6% (w/v) Triton X-114 was selected as optimal considering higher mean recovery for 21 antidepressants and comparable ME A .

The Effects of Salt Addition
Salts affect the extraction efficiency by decreasing or increasing the analyte concentration in the aqueous phase and CMC. Effects of salt addition depend on the concentrations of salts and the type of surfactant used. For the nonionic surfactants, such as Triton X-114, the value of CMC decreases with an increasing concentration of the electrolyte [35][36][37][38][39][40]. The anions (i.e., Cl − and SO 4 2− ) are likely to decrease self-association of water molecules, whereas the cation (i.e., Na + ) may decrease the cloud point by dehydration of the polyoxyethylene chain [39,41]. Electrolytes support demulsification, which is needed to limit the amount of impurities in the final surfactant-rich phase [40]. The other important effects of electrolytes are the decrease in the cloud point temperature and promotion of phase separation by increasing the density of aqueous phase [8,41]. However, overly large amounts of salts can cause the formation of three phases, as observed in our study [8].
The most frequently used concentration of different salts is 4-6% [32], increasing the recovery of the analyte by 10-20%. The opposite effect was observed if the salt concentration ranged from 7 to 10%. Improvement of extraction efficiency associated with salt addition was probably due to the salting-out effect, which reduces the amount of water to dissolve the analyte. While the overly high concentration of salt will competitively carry substances into the protein deposition, it can lower the concentration of drug in the solution and will have an impact on the recovery [41]. Even if the addition of salt increases the recovery, non-volatile salts contaminate the ion source in LC-MS. Thus, we tested the addition of the volatile salt (AA) in three variants: 10, 20, and 30% (w/v) (sample pH 6.8, 6% (w/v) of Triton X-114). The influences of various concentrations of salt added on the recovery and ME A were evaluated (Figures 3 and 4).
There was a statistically significant difference (p < 0.001) in recovery depending on the concentration of AA used in the tested conditions (sample pH 6.8, 6% Triton X-114). The differences occurred between all sample variants. The addition of 10% (w/v) AA increased recovery 1.64 times on average (range 0.87-2.31), while 20% (w/v) AA increased recovery 1.85 times on average (range 0.88-2.76), compared with the sample without the addition of AA.
The addition of 30% (w/v) AA resulted in lower recovery than the addition of 10% or 20% (w/v) of AA. Moreover, for 20 and 30% (w/v) of AA, three phases were observed (from the bottom of the tube: an aqueous phase, a micelle-rich phase, an aqueous phase), unlike for 0 and 10% (w/v) of AA, for which only two phases were formed (lower layer-micelle-rich phase; and aqueous surfactant-lean phase above). To separate the micelle-rich phase, an upper surfactant-lean phase was decanted. In the cases of 20 and 30% (w/v) of AA, the microsyringe should be used to remove the lower aqueous surfactant-lean phase from the bottom of the test tube. Thus, although the mean recovery was higher for 20% (w/v) AA, the addition of 10% (w/v) AA was selected as the most convenient concentration. Similar observations of the influence of electrolytes on the location of the micellar phases in the tube were reported [35]. ME A can be observed as signal suppression (below 100%) or an enhancement (above 100%) in the presence of a matrix. Significant ME A <85% or >115% was observed for 10, 10, 12, and 16 out of 21 studied drugs for 0, 10, 20, and 30% (w/v) of AA, respectively. The differences were not statistically significant (χ 2 test, p = 0.198), nor were they in statistical analysis performed only for the compounds with significant ME A (p = 0.244). In all variants of AA, insignificant ME A was observed for five compounds: FLX, FLV, MAP, PAR, and SER. For 0 and 10% (w/v) of AA, that was additionally observed for AMI, IMI, and TRI, whereas for 0, 10, and 20% (w/v) of AA, that was the case for DES, NOR, and PRO.
The effect of salt adding (sodium chloride) on the recovery of only one antidepressant venlafaxine in the CPE procedure was reported for the concentration range 0.1-0.5 M. The concentration of NaCl 0.3 M was chosen as optimal [8]. In this work, AA was used first time in CPE procedure combined with LC-MS not only for venlafaxine determination but also other 20 antidepressants. As the optimal concentration of salt, 1.3 M AA (10% (w/v) of AA) was selected.

The Effects of pH
Depending on the sample pH, analytes can be either charged or uncharged. That also depends on the chemical properties of said analytes. The interaction of the analyte with the micellar aggregate formed by a nonionic surfactant is weaker for ionic than the neutral form. Thus, the highest recovery can be observed at pH where all analytes are uncharged and partitioned into the rich-micellar phase of Triton X-114 or another nonionic surfactant [37,42].
The effect of sample pH was examined using 6% (w/v) Triton X-114 and 10% (w/v) AA at pH 3.5, 6.8, and 10.2 ( Figure 5). The mean recoveries were 6.4, 64.7, and 86.2% for acidic, neutral, and alkaline conditions, respectively. The majority of antidepressants studied have basic and lipophilic properties. In alkaline pH, more particles occur in molecular form; thus, higher affinity to nonionic micelles and higher extraction recovery were observed. At acidic pH, the majority of the antidepressants were in ionic form and had their lowest affinity for nonionic micelles. That resulted in the loss of some of the analytes during phase separation and low extraction efficiency. Moreover, the separation of two phases-the aqueous and micelle-rich ones-in the acidic and neutral pHs, was not as pronounced as in alkaline pH. Statistical analysis (p < 0.001) indicated significant differences in recovery of the extraction performed at various pH values. However, no relationship between the pH of the extraction environment and the number of compounds exhibiting significant ME A was observed (test χ 2 ; p = 0.514). Statistical analysis carried out for compounds with significant ME A (ANOVA, p = 0.013) showed statistically significant differences in ME A between pHs. The results of the post hoc test revealed that these differences occurred between the alkaline and acidic environment (p = 0.004). The mean ME A was 6.1% lower for samples with a pH of 3.5 than those of pH 10.2. However, it should be emphasized that in acidic conditions, the basic compounds have less affinity to nonionic surfactant micelles, which resulted in the lower analytical signal and lower recovery. Moreover, the separation into two phases was not distinct in acidic pH or neutral pH, which caused the loss of analytes during phases separation. Thus, pH 10.2 was selected as optimal to use in the final method.
Madej and Persona [2] developed a CPE to extract basic compounds (pKa 9.0−9.5) at pH 12, with Triton X-114 at 3.25% (w/v), without salt addition. The recoveries were 26.2, 81.4, 77.3, and 98.5% for paracetamol, amitriptyline, clomipramine, and promazine, respectively [2]. In this paper, the number of isolated compounds from plasma in the basic environment (pH 10.2) was much higher (n = 21 vs. n = 4) and the pKa range was wider (pKa 6.0−10.5). A higher mean recovery (86.2%) was observed than in the reported experiment (70.9%). Compared to the recovery of particular compounds, higher recoveries were achieved for AMI (81.4 vs. 85.4%) and CLO (77.3% vs. 90.1%). OPI was extracted as a neutral compound at pH 6, which resulted in 21.8% recovery, while in this study at pH 10.2, the recovery was 102.5% [2]. Thus, the extraction at pH 10.2 and the addition of AA allow for efficient isolation of antidepressants from biological matrices by the CPE method. Additionally, the extraction environment does not significantly affect the number of compounds exhibiting a matrix effect.

Principal Component Analysis (PCA)
Some compounds revealed significant ME A in all tested conditions (CIT, DOX, TRA, VEN). All of them showed a significant surfactant effect (the ratio of peak area of the analyte in a solvent with the surfactant to that without the surfactant) previously [31]. That means that the ME A observed is not related to impurities extracted from plasma, but the presence of Triton in the sample. PCA was used to determine which molecular descriptors (polar volume, pKa, cLogP, dipole moment, number of hydrogen bond acceptors (HBA), number of hydrogen bond donors (HBD), total volume, molecular mass, total area, polar surface area (PSA)) influenced the ME A in optimal conditions [31]. Retention time and predicted water solubility were also added to the analysis. Three principal components (PC) accounted for 82.4% of the total variation. PC1 (44.7% of the total variation) was correlated mainly with the number of HBA (−0.91), PSA (−0.90), total area (−0.85), molecular mass (−0.85), and total volume (−0.83). In turn, PC2 (27.1% of the total variation) was correlated with the retention time (−0.90), pKa (−0.77) and solubility (0.94), and PC3 (10.6% of the total variation) with the dipole moment (0.78) and HBD (0.58). The best predictors of ME A occurrence were PC1 (PSA, total area, total volume, molecular mass, and the number of HBA) and PC2 (retention time, pKa, and predicted solubility) (Figure 7).
Low polar volume; few HBA; low solubility; and high values of retention time, pKa, and cLogP indicated that the ME A should not be expected. Compounds with lower lipophilicity (low cLogP and high polar volume) and low pKa were prone to significant ME A [31]. retention time (−0.90), pKa (−0.77) and solubility (0.94), and PC3 (10.6% of the total variation) with the dipole moment (0.78) and HBD (0.58). The best predictors of MEA occurrence were PC1 (PSA, total area, total volume, molecular mass, and the number of HBA) and PC2 (retention time, pKa, and p Low polar volume; few HBA; low solubility; and high values of retention time, pKa, and cLogP indicated that the MEA should not be expected. Compounds with lower lipophilicity (low cLogP and high polar volume) and low pKa were prone to significant MEA [31].

Analytical Performance of the Proposed Method
After development, the analytical method's performance was evaluated in terms of linearity, accuracy, precision, the lower limit of quantification (LLOQ = 10 ng/mL), and stability (autosampler, short-term, freeze and thaw). The interference from metabolites was not studied, since it occurs rarely in mass spectrometry. Due to different masses, metabolites were differentiated from parent compounds using LC-MS (Table A1). In some cases, European Medicines Agency (EMA) guidelines recommend the evaluation of the possibility of back-conversion of a metabolite into the parent analyte during the successive steps of the analysis. However, it is performed only when relevant (i.e., in case of potentially unstable metabolites-e.g., acidic metabolites to esters, unstable N-oxides, or glucuronide metabolites, lactone-ring structures) [43]. It is not an issue in the case of antidepressants.

Analytical Performance of the Proposed Method
After development, the analytical method's performance was evaluated in terms of linearity, accuracy, precision, the lower limit of quantification (LLOQ = 10 ng/mL), and stability (autosampler, short-term, freeze and thaw). The interference from metabolites was not studied, since it occurs rarely in mass spectrometry. Due to different masses, metabolites were differentiated from parent compounds using LC-MS (Table A1). In some cases, European Medicines Agency (EMA) guidelines recommend the evaluation of the possibility of back-conversion of a metabolite into the parent analyte during the successive steps of the analysis. However, it is performed only when relevant (i.e., in case of potentially unstable metabolites-e.g., acidic metabolites to esters, unstable N-oxides, or glucuronide metabolites, lactone-ring structures) [43]. It is not an issue in the case of antidepressants.

Linearity
Calibration standards of eight concentrations of all analytes were prepared in blank plasma. The curves (n = 6) were obtained by weighted linear regression analysis with w = 1/x. The regression parameters for all analytes were described by the equation: y = ax + b. The values of a, b, and r 2 for all analytes are presented in Table A2. Good linearities covered the range of 10-750 ng/mL with a coefficient of determination (r 2 ) > 0.990 were obtained for all analytes regarding the peak area ratio of every analyte to internal standard (IS)versus the nominal concentration.
The linearity range of the method corresponds to therapeutic concentrations of determined antidepressants in human plasma and is similar to the reported ranges [20][21][22]28,29].

Limit of Quantification, Precision, and Accuracy
Intra-run and between-run accuracy and precision of the method for LLOQ (10 ng/mL) and QC (quality control; 16, 375, and 625 ng/mL) samples for all analytes were determined. For each LLOQ, a signal-to-noise ratio (S/N) higher than five (from 7.0 to 76.1) was observed. The accuracy was within the acceptance criteria of 85-115 and 80-120% for each QC sample and LLOQ, respectively. The precision was within the acceptance criteria, < 15 and < 20% for QC and LLOQ respectively. Precision was in the range 0.6-14.0%, and accuracy was from 85 to 114% (Table 1). Chromatograms of the analyzed antidepressants extracted from blank plasma and high QC are presented in Figure A1.
The results were repeatable and reproducible and met the criteria of the EMA [43] and the U.S. Food and Drug Administration [44] in the guidelines on bioanalytical method validation. The obtained precision is comparable to the reported values for the other extraction methods of isolating the antidepressants from plasma [20,21].

Stability
All analytes and IS were stable under all tested conditions: autosampler stability (stability in the extract, 48 h at 4 • C); and stability in human plasma, i.e., short-term stability (3 h at room temperature), and freeze-thaw stability (3 cycles at −20 • C). The QC (16 and 625 ng/mL) samples showed no significant changes in comparison to nominal concentrations. In all cases, accuracy and precision were found to be within the acceptable limits of ± 15% in both cases, as shown in Table A3.

Matrix Effect and Recovery
The significant matrix effect was observed for eight and seven out of 21 antidepressants, for 16 and 625 ng/mL, respectively. However, since the effect was compensated by the addition of the internal standard, it did not affect the method's accuracy and precision. It was proven by the variability of relative matrix factor (MF). The calculated coefficient of variation (CV) of the IS-normalized MF did not exceed 15% for all tested compounds (Table A4). The mean CV values were 10.2% (5.7%-15.0%) and 7.1% (2.6%-13.9%) for the concentrations of 16 and 625 ng/mL, respectively. The maximum mean CV equal to 15.0% was obtained for MIA and VEN for low QC (Table A4). The hemolysis and lipemia did not influence the method reliability.

Materials
Four lots of human plasma with EDTA as an anticoagulant were obtained from the Regional Blood Donation and Blood Therapy Centre (Warsaw, Poland). Besides, plasma with visible hemolysis and plasma with visible lipemia were obtained from the Public Central Teaching Hospital (Warsaw, Poland).

Chromatographic and Mass Spectrometric Conditions
Instrumental analysis was performed on an Agilent 1260 Infinity (Agilent Technologies, Santa Clara, CA, US), equipped with an autosampler, degasser, and binary pump coupled to a hybrid triple quadrupole/linear ion trap mass spectrometer QTRAP 4000 (ABSciex, Framingham, MA, USA). The Turbo Ion Spray source was operated in the positive mode. The ion spray voltage and source temperatures were 5000 V and 600 • C, respectively. The curtain gas, ion source gas 1, ion source gas 2, and collision gas were set at 345 kPa, 207 kPa, 276 kPa, and "high" instrument units, respectively [4]. The target compounds were optimized and analyzed in the Multiple Reaction Monitoring (MRM) mode (Table 3). Table 3. The optimized parameters of MRM mode of antidepressant determination in ESI+.

Preparation of Solutions
The surfactant Triton X-114 solutions were prepared at 5 and 20% (w/v) and were used to obtain final surfactant concentrations of 1.5 and 6%, respectively. The standard stock solutions of antidepressants were made by dissolving 10 mg of the reference standard in 10 mL of methanol and were stored at −39 • C. The standard working solutions of 25 µg/mL for the validation and 500 ng/mL for development were prepared by mixing equal volumes of 21 different stock solutions and the appropriate volumes of water.
The internal standard stock solutions SER-d3, FLX-d5 (1 mg/mL), and VEN-d6 (0.1 mg/mL) were prepared in methanol. Each internal standard working solution (500 ng/mL) was prepared by mixing the internal standard stock solution with an appropriate volume of water.

Sample Preparation
The experiment consisted of the method development and validation part. All the experiments were performed in human plasma with EDTA as an anticoagulant. Tested parameters are shown in Table 4 and Figure A3. The method of extraction was the method of Kojro (2019) [31] with our modifications of Triton X-114 concentration (step a), the addition of AA (step b), and pH (step c), which was selected during the development part. Table 4. Variants of samples tested during CPE development and validation-different variants of final concentration of surfactant Triton X-114, concentration of ammonium acetate (AA), and sample pH. Samples were incubated in a water bath for 20 min at 60 • C.

CPE Procedure-Development
The development was carried out in one source of plasma in 3 kinds of sample fortified by antidepressant standard-mix (50 µL of 500 ng/mL aqueous solution without the isotope-labeled standard) for all 21 analytes at a concentration corresponding to 100 ng/mL in plasma in 5 repetitions: sample A (fortified lower phase after diluting with ethanol), sample B (fortified neat ethanol), and sample C (fortified plasma before extraction). Samples A and B were used to calculate the absolute matrix effect ME A [33], also called matrix factor [43] (Equation (1) in Section 3.6). The final concentrations of analytes in sample A and sample B were all the same. Samples A and C were used to assess recovery (Equation (3) in Section 3.6).
In the first step of development, Triton X-114 at concentrations 1.5 and 6% (w/v) was tested. In the second step, the concentrations of extraction salt AA-0, 10, 20, and 30 % (w/v)-were examined; and in the third step-pHs of 3.5, 6.8, and 10.2 were tested. At every step of development, samples A, B, and C were prepared with various modifications of the studied parameters, as described in Table 4 and Figure A3.
Sample A: An aliquot of 250 µL of plasma was placed in a 1.5 mL Eppendorf tube. Then, 50 µL of water was added. The sample was vortexed for 30 s. Afterwards, 300 µL of the surfactant solution, Triton X-114 (5% or 20% to obtain the final concentration of 1.5 or 6%); 400 µL of 100% acetic acid, water, or 25% ammonia solution; and 0, 100, 200, or 300 mg AA (to obtain concentration of 0, 10, 20 or 30% (w/v)) were added. The sample was vortexed at low speed for 5 min and incubated in a water bath for 20 min at 60 • C. After that, the phase separation was accelerated by centrifugation (10,000× g, 5 min, 25 • C). The surfactant-lean aqueous phase was decanted to obtain the surfactant-rich phase, which was diluted with an aliquot of 500 µL of ethanol and incubated for 5 min at -80 • C. Then, the sample was centrifuged (10,000 g, 5 min, 25 • C); 200 µL of supernatant was diluted with 650 µL water and mixed together with 50 µL 500 ng/mL aqueous solution mix of the antidepressants standard without internal standards.
Sample C: Like sample A with modifications: (1) instead of 50 µL water, 50 µL of a 500 ng/mL aqueous solution of antidepressant standard-mix was added to 250 µL plasma; (2) finally, to the vial, 700 µL of water was added instead of 650 µL water and 50 µL of antidepressants standard-mix solution.

CPE Procedure-Validation
An aliquot of 250 µL of human plasma fortified with a solution of mixed standards of 21 antidepressants in appropriate concentrations from QC samples and calibration samples was placed in a 1.5 mL Eppendorf tube. Then, 50 µL of 500 ng/mL aqueous internal standard solution was added. The sample was vortexed for 30 s. Afterward, 300 µL of the surfactant 20% Triton X-114, 400 µL of the 25% ammonia solution, and 100 mg AA (to final concentration of 10% (w/v)) were added. The sample was vortexed at low speed for 5 min and incubated in a water bath for 20 min at 60 • C. The phases were centrifuged and separated (10,000 g, 5 min, 25 • C); the surfactant-lean phase was decanted. The surfactant-rich phase was mixed with a 500 µL aliquot of ethanol and incubated for 5 min at −80 • C. Then, the sample was centrifuged (10,000 g, 5 min, 25 • C). An aliquot of 200 µL of supernatant was diluted with 700 µL of water and analyzed.

Preparation of Calibration Standards and Quality Samples for Validation Optimal CPE Conditions
The calibration standards contained all analytes at eight concentrations ranging from 10 to 750 ng/mL. The quality control samples were prepared at concentrations of 16, 375, and 625 ng/mL. All calibration standards and the quality control samples were prepared via the appropriate dilutions of blank human plasma spiked with all analytes in working solutions and were stored at ≤−39 • C. The precision was calculated as the ratio of standard deviation/mean (%). The accuracy was calculated as the determined value divided by the nominal value and expressed in percent.
Samples A and B for the matrix effect and samples A and C for the recovery test were prepared at two concentrations for all analyte standards and for IS at the working concentration in blank human plasma samples derived from six sources.

Matrix Effect and Recovery Calculation
ME calculated as ME A [33], also called matrix factor (MF) (Equation (1)) [43], was regarded as insignificant if included between 85 and 115%. In development, the samples for matrix effect were prepared from blank human plasma from one source at a level corresponding to a plasma concentration of 100 ng/mL, so the matrix factors were calculated as ME A. The preparation of samples A, B, and C was described in Section 3.5.1. In validation, the MF were calculated as the ratios of the instrument response for substances in sample A and sample B at two concentrations (16 and 625 ng/mL) for all analyte standards and for IS in the working concentration in 6 different sources, including haemolyzed and hyperlipidaemic plasma. The CV of the IS-normalized matrix factor (Equation (2)) should not exceed 15%.
Recovery was calculated by dividing the analytical signal pre-extraction spiked sample plasma (sample C) by post-extraction spiked sample plasma (sample A) (Equation (3)). Sample C concentrations were calculated on account of the volume of the rich-micellar phase. ME A = MF = peak area of the analyte in a matrix (sample A) peak area of the analyte in a solvent (sample B) × 100% (1) During development, recovery was calculated according to Equation (3).
Recovery = peak area of the analyte in variant sample C peak area of the analyte invariant sample A

Statistical Analysis
The statistical analysis of the results was performed with the STATISTICA version 12 for Windows (TIBCO Software Inc., Palo Alto, CA, USA). The normal distribution of the results was evaluated by the Shapiro-Wilk test. For the normal distribution, the t-Student test and ANOVA were used.
The one-way ANOVA followed by the LSD post hoc test for multiple comparisons of the dependent group was applied. Differences between two nominal variables were tested using the chi-square test (χ 2 ). A p-value below 0.05 was considered significant.
PCA was used to find molecular descriptors that divided the studied group into prone and not prone to the ME.

Conclusions
The use of a volatile salt (1.3 M-10% (w/v) of AA) is a promising modification of the CPE procedure for LC-MS applications, since the mean recovery of studied antidepressants increased almost two-fold. The procedure can be applied to other groups of analytes to develop reliable and environmentally friendly screening methods. In the future, other volatile salts such as ammonium formate can be used for comparisons. Non-volatile salts can be applied in CPE-LC/MS when the discharge of the initial part of the chromatographic course is possible. The modified CPE method is an alternative for LLE and SPE for sample preparation prior to LC-MS, giving comparable recovery and matrix effect. Using ethanol instead of methanol or acetonitrile in CPE to reduce the sample viscosity at the last point of sample preparation makes the method more friendly to the environment. The selected conditions of CPE extraction of antidepressants were as follows: 6% (w/v) Triton X-114, 10% (w/v) AA, and pH 10.2. The developed CPE-LC-MS method was proven to be precise, accurate, and reliable, and can be applied for the simultaneous determination of various antidepressants in human plasma in clinical and forensic toxicology.
Limitation of the study: The method was not applied to clinical samples. Not all parameters that can influence the method performance were tested, e.g., extraction temperature, equilibration time, types of volatile salts, and anticoagulants used in blood sample collection.    [49] human plasma    [49] human plasma Figure A3. Variants of samples tested during CPE development.  There is no reported MRM methods for metabolites of other analyzed antidepressants. The uncertainty (standard deviation) was rounded up to two significant figures. The mean was rounded to have the same number of decimal places as the uncertainty.