Introduction

Schizophrenia is recognizably a severe and chronic psychiatric disorder that affects approximately 0.5% of the world’s general population [1, 2]. Antipsychotics are the first-line evidence-based treatment for schizophrenia. These drugs are usually prescribed to implement rehabilitation programs and to manage the risk of new psychotic episodes [3,4,5]. In routine clinical practice, the antipsychotic drugs can be classified in first-generation antipsychotics (FGAs) (such as haloperidol and chlorpromazine) and second-generation antipsychotics (SGAs) (including clozapine, quetiapine) [6].

In most cases, the effectiveness of treatment and side effects of the antipsychotic drugs are dose-related and the plasma concentrations of these drugs can vary widely between patients due to their interindividual variability [7]. In this context, therapeutic drug monitoring (TDM) of antipsychotic drugs is essential for managing the patient’s medication strategy. TDM constitutes an important resource to evaluate the adherence to therapy, increasing the therapeutic efficiency and minimizing the risk of side effects or toxicity [8, 9]. Therefore, sensitive and selective bioanalytical methods are required to carry out reliable determinations. Liquid chromatography–tandem mass spectrometry (LC–MS/MS) has been used as the reference analytical technique to determine anti-schizophrenic drugs and their metabolites in biological samples for TDM purposes [10].

Sample preparation is a very important step in the bioanalytical process. Biological fluids contain numerous endogenous compounds (such as proteins and salts) that may negatively affect the chromatographic separation, increase the column back pressure, and suppress or intensify the ionization signal in LC–MS/MS analysis (matrix effect) [11].

Microextraction by packed sorbent (MEPS) has emerged as an attractive technique for sample preparation. MEPS is a miniaturized version of the conventional solid-phase extraction (SPE), consisting of a few milligrams of a solid sorbent integrated into the barrel of a gas-tight syringe (100–250 μL) [12,13,14]. The selectivity of the sorbent is related to the sorption capacity of the analytes and adequate sample cleanup. C18, C8, C2, C8-SCX, and silica are examples of commercial sorbents successfully used in MEPS for the determination of a wide range of drugs in different complex biological matrices [15]. However, due to the high complexity of the biological fluids, innovative materials such as molecularly imprinted polymer (MIP) [16,17,18], monoliths [19, 20], and restricted access material (RAM) have been used as MEPS sorbents to provide selectivity and sensitivity to the sample preparation procedure.

RAMs are porous sorbents with a restrictive and hydrophilic outer surface specifically designed for the removal of macromolecules (size-exclusion mechanism), combined with smaller inner pores with hydrophobic surfaces that only molecules with low molecular weight can reach. Two different types of macromolecule exclusion mechanisms are recognized, chemical diffusion barrier (a protein or polymer network is bound covalently to the surface of the particle) or physical barrier made by the pore size (diameter) [21, 22]. In order to combine the advantages of common sorbents (such as selectivity and high sorption capacity) with the ability to exclude macromolecules, new RAMs have been successfully obtained by the insertion of external hydrophilic layers on conventional materials. The most successful examples are silica and polymer-based, and more recently, the carbon nanotubes (CNTs) [23].

To optimize the use of CNTs in bioanalysis, Barbosa et al. [24, 25] developed a new hybrid sorbent called restricted access carbon nanotubes (RACNTs). The synthesis was achieved by chemically covering commercial CNTs with a cross-linked bovine serum albumin (BSA) layer. This new material combines the versatility, high chemical stability, and large surface area of carbon nanotubes with the ability of RAMs to exclude macromolecules. The exclusion mechanism is based on the electrostatic repulsion between the proteins from the sample and from the BSA layer when the pH of the medium is higher than the isoelectric points of both proteins [23].

RACNTs were successfully applied in online SPE for the direct extraction of cadmium and lead in human serum samples [24, 25]. This material was able to exclude about 100% of the total serum proteins. Additionally, RACNTs were successfully used as sorbents in a column-switching liquid chromatography system to determine anticonvulsant [26] and antihypertensive [27] drugs in untreated human plasma and serum samples. These works are related to the use of RACNTS only for online extractions. However, the direct injection of plasma samples in the HPLC (online system) might increase the column back pressure after a few injections, which can damage the analytical column.

RACNTs in this sample preparation scenario allow the utilization of this innovative sorbent in laboratories that do not have a two-dimensional system, demonstrating the versatility of this stationary phase. Compared with column-switching methods, the MEPS procedure is simpler to optimize, is able to use a lower biological sample volume, and has reduced sample preparation time. This manuscript describes the evaluation of RACNT-based material for MEPS (offline extraction) to determine antipsychotic drugs (chlorpromazine, clozapine, quetiapine, and olanzapine) in plasma samples by high-performance liquid chromatography–tandem mass spectrometry for TDM of schizophrenic patients.

Experimental

Materials

Chlorpromazine (CLOR), clozapine (CLOZ), olanzapine (OLA), and quetiapine (QUET) standards were acquired from Cerilliant® (Round Rock, TX, USA). Chlorpromazine-d3 (CLOR-d3) and quetiapine-d8 (QUET-d8) were obtained from Cerilliant® (Round Rock, TX, USA) and used as internal standards (IS). The multi-walled CNTs with outer wall diameters from 6 to 9 nm and lengths of 5 mm were purchased from Sigma-Aldrich® (St. Louis, MO, USA). Bovine serum albumin, glutaraldehyde, and sodium borohydride used for obtaining the RACNTs were acquired from Sigma-Aldrich® (St. Louis, MO, USA). Acetonitrile, ethanol (both of HPLC grade), ammonium acetate, and formic acid were supplied by JT Baker® (Phillipsburg, SXM, USA). The ultrapure water used to prepare the mobile phase had been previously purified in a Milli-Q® system (São Paulo, SP, Brazil).

Synthesis of the RACNTs

First, a polypropylene SPE cartridge containing 500 mg of commercial CNTs was attached to a vacuum manifold system. Subsequently, 20 mL of a 10-g L−1 BSA solution prepared in a phosphate buffer 50 mmol L−1 (pH = 6.0) was percolated through the material at a 1.0-mL min−1 flow rate. Next, 5 mL of a 250-g L−1 glutaraldehyde aqueous solution was added through the sorbent at a flow rate of approximately 0.2 mL min−1. As a result of the cross-linking of amine groups of the BSA with the glutaraldehyde, the protein was fixed in the surface of the material, forming unstable imine groups [24]. After maintaining the system on standby for 5 h, the time required to ensure that all glutaraldehyde reacted with BSA, 10 mL of 10 g L−1 aqueous solution of sodium borohydride was percolated through the cartridge at 0.5 mL min−1 flow rates. In this step, the sodium borohydride acts as a reducing agent, converting the imine groups to amines, which are more stable [24]. The obtained material was exhaustively washed with ultrapure water to remove possible product residues and dried at room temperature for approximately 24 h. The material has been previously characterized by Barbosa et al. [24] and De Faria et al. [27]. A mass of 10 mg of RACNTs was packed into a polyethylene conical tube (1000 μL) using glass wool on the bottom end and a steel filter (2.4 mm diameter) on the top to prevent loss of the adsorbent material. The filter was pushed towards the sorbent until no empty space remained. The tube was properly adapted to a polypropylene Becton Dickinson® syringe (20 mL) for MEPS (Fig. 1).

Fig. 1
figure 1

MEPS apparatus

Full size image

Plasma samples

Blank plasma samples, as well as plasma samples from schizophrenic patients, were supplied by the psychiatric nursing staff of the Hospital das Clínicas de Ribeirão Preto (from the University of São Paulo, Brazil). The plasma samples were collected in agreement with the criteria established by the Ethics Committee of Faculdade de Medicina de Ribeirão Preto from the University of São Paulo, Brazil, and stored at − 80 °C.

The analytical parameters were evaluated with blank plasma samples (100 μL) spiked with analyte standard solutions at different concentrations, considering the drugs’ therapeutic interval.

Before the extraction procedure, the plasma samples were diluted with 400 μL of borate buffer solution (10 mmol mL−1, pH 9).

Optimization of the MEPS extraction procedure

To investigate the best MEPS conditions, 40 μL of methanolic standard solutions of all antipsychotic drugs (1000 ng/mL) were transferred to an Eppendorf® tube (1.5 mL), dried under N2 gas flow, and reconstituted in 100 μL of blank plasma. The spiked samples were diluted with 400 μL of buffer solutions. The extraction variables sample pH, number of draw–eject cycles, and desorption solvent and volume were univariably optimized to attain the best sorption capacity for each analyte. The results were evaluated in terms of absolute recovery. All the experiments were performed in triplicate (n = 3).

MEPS conditions

The diluted plasma solution was manually drawn and ejected through the sorbent three times in the same vial to extract/concentrate the analytes. Then, the sorbent was washed with 150 μL of ultrapure water to remove protein and other interferents. In the elution step, two aliquots of 100 μL of pure acetonitrile were percolated through the cartridge to desorb the analytes in an Eppendorf® tube (1.5 mL). Subsequently, the eluate was dried under N2 gas flow and reconstituted with 50 μL of the initial mobile phase (80:20 ammonium acetate 5 mmol L−1 + 0.1% formic acid/acetonitrile). Finally, for the cleaning and reconditioning step, two acetonitrile aliquots (100 μL) followed by two ultrapure water aliquots (100 μL) were percolated to condition the RACNT sorbent after each extraction. This step aimed to increase the sorbent lifetime and to decrease the carryover effect. The RACNT phase was stored at room temperature in ultrapure water. The solvents have been previously pipetted to an Eppendorf tube and then aspirated by the syringe containing the tip with the RACNT phase. The flow rate was approximately 3 drops per second.

Protein exclusion test

To perform the protein exclusion test, 100 μL of blank plasma was diluted with 400 μL of borate buffer solution (pH 9). Ten microliters of the diluted plasma was directly injected in the HPLC-UV system without any analytical column. The obtained signal area corresponded to 100% of endogenous compounds (protein). The ability of the RACNTs to exclude macromolecules was accomplished by manually drawing and ejecting the diluted plasma three times through the cartridge, using both RACNT and CNT as sorbents. The remaining plasma was analyzed in the HPLC system and the peak areas were compared. The mobile phase consisted of phosphate buffer 0.01 mol L−1 (pH 7) at a 0.4-mL min−1 flow rate with a UV detector operating at 254 nm. All the tests were performed in triplicate.

LC–MS/MS

MEPS/LC–MS/MS analyses were carried out on a Waters ACQUITY UPLC H-Class system coupled to a Xevos TQ-D tandem quadrupole (Waters Corporation, Milford, MA, USA) mass spectrometer equipped with a Z-spray source operating in the ESI positive mode. The samples were kept at 10 °C and 10 μL was injected in the chromatographic system. The antipsychotic drugs were separated on an XSelecta CSH C18 (2.5 μm, 2.1 × 100 mm) column at 40 °C. A binary mobile phase consisting of (A) ammonium acetate aqueous solution 5 mmol L−1 (with 0.1% of formic acid) and (B) acetonitrile at a flow rate of 300 μL min−1 provided an optimal separation. The gradient elution was 20% B, increasing gradually to 85% B (0–3 min) which remained in this configuration up to 4.5 min. Subsequently, the gradient returned to 20% B (4.5–6.0 min), followed by 5 min of re-equilibration. The total analysis time was 11 min. The source and MS parameters were automatically determined by Intellistart® Program as follows: capillary voltage, 3.2 kV; source temperature, 150 °C; desolvation temperature, 400 °C; desolvation gas flow, 700 L h−1 (N2, 99.9% purity); and cone gas flow, 150 L h−1 (N2, 99.9% purity). Argon (99.9999% purity) was used as the collision gas. All the analytes and ISs were analyzed in the multiple reaction monitoring (MRM) mode. The dwell time established for each transition was 0.048 s. The interscan delay was set to the automatic mode. Data were acquired by means of the MassLynx V4.1 software. Table 1 shows the MRM transitions for each determined analyte. The primary transition was used for quantification and the secondary transition for confirmation.

Table 1 Ion transitions, instrument settings, and retention times for the antipsychotics
Full size table

Analytical validation

Analytical validation was based on current international guidelines issued by the European Medicines Agency (EMA) and the Food and Drug Administration (FDA). Linearity, precision, accuracy, limit of quantification (LOQ), selectivity, and matrix and carryover effects were evaluated.

A pool of blank human plasma sample was spiked with CLOZ, QUET, and CLORP from 10 to 700 ng mL−1, as well as from 10 to 200 ng mL−1 for OLA. The deuterated IS QUET-d8 and CLOR-d3 were used at the concentrations of 150 ng mL−1. The linearity was expressed by the determination coefficient (R2) calculated from the intra-assay calibration curves. The intra- and inter-assays related to precision and accuracy were carried out on the same day and on three different days, respectively. The accuracy was evaluated as the relative error (%RE), and the precision by the relative standard deviation (RSD).

These data were evaluated at low-quality control (low QC) (within three times the LLOQ), medium-quality control (medium QC) (around 30–50% of the calibration curve range), and high-quality control (high QC) (at least at 75% of the upper calibration curve range). The lower limit of quantification (LLOQ) was defined as the lowest concentration that could be measured with acceptable precision and accuracy (within ± 20%). Selectivity was carried out comparing the chromatograms (MRM mode) of a blank human plasma sample and a blank human plasma sample spiked with the analytes at the concentrations corresponding to the LLOQ. The carryover effect was evaluated by injecting a blank sample immediately after analysis of the upper limit of quantification (ULOQ) sample. To perform the matrix effect (ME) assay, the RSD of the IS normalized matrix factor (MF) was calculated using the peak areas of the analytes dissolved in water compared with blank plasma spiked with the antipsychotic drugs (n = 3) at two concentration levels (low QC and high QC). The obtained results were compared and the RSD lower than 15% was adopted as the acceptance criterion.

Results and discussion

Optimization of the extraction procedure

Effect of sample pH

Under the investigated buffer solution: ammonium acetate (pH 5.0, 10 mM), ammonium phosphate (pH 7.0, 10 mM), and sodium tetraborate (pH 9.0, 10 mM), the plasma sample dilution with borate buffer solution at pH 9 achieved the highest recovery (Fig. 2a). In this scenario, the majority of the antipsychotics (pKa values from 7.24 to 9.3) (Fig. 3) were partially or totally in the nonionic form, which supported the weak intermolecular forces between the analytes and the inner hydrophobic surface of the RACNT. This dilution also decreased substantially the matrix viscosity, minimizing interference of the plasma in the MEPS process and increasing the diffusion coefficients. Extreme pH conditions were not evaluated, aiming to avoid protein precipitation.

Fig. 2
figure 2

Optimization of the MEPS parameters: (a) effect of pH, (b) number of draw–eject cycles, (c) type of desorption solvent, and (d) desorption volume

Full size image
Fig. 3
figure 3

Chemical structure of the antipsychotics

Full size image

Number of draw–eject cycles

In MEPS, the draw–eject cycles can be performed in the same vial or by drawing and ejecting as waste [28]. In this assay, the extraction cycles were performed in the same vial for the establishment of the sorption equilibrium. The recovery of the antipsychotics did not present a significant difference varying the number of cycles from 1 to 5 (Fig. 2b). These results attest to the high ability of carbon nanotubes to interact with organic compounds by Van der Waals forces and to establish π–π interactions besides the high surface area. Therefore, 3 draw–eject cycles were chosen to ensure reliable and precise extractions with low RSD values.

Eluent nature and volume

The optimization of the desorption conditions was performed aiming to ensure effective elution of the analytes with minimum solvent volume. Acetonitrile, methanol, and ethanol were evaluated as elution solvents. Two aliquots of 100 μL of pure acetonitrile afforded the highest recovery, probably due to the higher hydrophobic character and low viscosity (Fig. 2c, d).

Protein exclusion test

The sample pH may directly affect the sorption of the analytes in the sorbent, as well as interfere in the ability of the material to exclude macromolecules. When a biological sample is percolated through a RACNT sorbent in a pH higher than the isoelectric point of sample proteins and BSA layer, all the proteins are negatively charged. In this case, the macromolecules are excluded by electrostatic repulsion, while low molecular weight compounds penetrate into the inner pores, interacting with the core of the CNTs.

RACNT and CNT were evaluated according to their ability to exclude proteins, as shown in Fig. 4. When the diluted blank plasma sample was extracted by using the RACNT as sorbent, the obtained peak area of the supernatant was practically equal to that obtained by the direct injection of the blank plasma in the HPLC system, attesting the high ability of the RACNT to exclude proteins (about 97%). On the other hand, when commercial CNT was used as the stationary phase, the peak was considerably lower when compared to the direct injection of plasma. In this case, about 50% of the proteins were retained in the sorbent. These results reinforce the ability of RACNT to exclude proteins by the electrostatic repulsion between the plasma proteins and the BSA layer (chemistry diffusion barrier) in the outer surface of the phase.

Fig. 4
figure 4

Analytical signals obtained from the direct injection of 100 μL of plasma diluted with 400 μL of borate buffer (pH 9) at 255 nm (A), after the extraction procedure with the RACNT (B), and with conventional CNT (C). A—peak area, 9,196,219; B—peak area, 8,926,788; and C—peak area, 6176930

Full size image

Analytical validation

The method was linear for all the antipsychotics from the LLOQ (10 ng mL−1) to the ULOQ (200 ng mL−1 for OLA and 700 ng mL−1 for CLOR, CLOZ, and QUET). The coefficients of determination (r2) were greater than 0.99 (Table 2). All linear ranges of the proposed method are in agreement with the preconized therapeutic intervals of the target drugs.

Table 2 Linear ranges, coefficients of determination, LLOQ values, recovery, and internal standard for the antipsychotics
Full size table

The method exhibited suitable precision and accuracy intra- and inter-assays. These data were evaluated at low-quality, medium-quality, and high-quality control. The accuracy was given as RE values that ranged from − 5.86 to 5.79% (intra-assay accuracy) and − 8.01 to 11.53% (inter-assay accuracy). The precision was expressed as RSD values, which varied from 2.98 to 11.71% (intra-assay precision) and 4.29 to 12.85% (inter-assay precision). These data are presented in Table 3.

Table 3 Accuracy and precision of the proposed method from plasma samples spiked with target drugs at different concentrations (n = 5)
Full size table

The matrix effect results proved that there was no significant ionic suppression or enhancement. The normalized factor matrix RSD for all drugs were not greater than 15% (Table 4). No carryover effect was observed when injecting a blank sample immediately after analysis of the ULOQ sample. Signals lower than 20% for the analytes and not greater than 5% in the case of the IS were achieved when comparing the blank chromatograms and the chromatograms in the LLOQ (Fig. 5).

Table 4 Normalized matrix factor (%RSD) of the antipsychotics in plasma samples spiked with analytes at LLOQ and ULOQ concentrations
Full size table
Fig. 5
figure 5

MEPS/LC–MS/MS chromatograms of a drug-free plasma sample and drug-free plasma sample spiked with the antipsychotics at LLOQ concentration

Full size image

Compared to other techniques, the proposed method presented low LLOQ, reduced organic solvent consumption, sample preparation time, and biological sample volume (Table 5). In addition, the tips showed great chemical stability, being re-used over 100 times, which underlines the low cost of the method. It was possible to use the same tip to perform more than four analytical curves in which the angular coefficients and quality controls remained within the acceptable range. Moreover, the RACNTs showed no apparent damage or swelling.

Table 5 Comparison of this MEPS–HPLC–MS/MS method with other counterparts to determine antipsychotics in biological samples
Full size table

Clinical application in plasma samples of schizophrenic patients

To evaluate the proposed method for clinical use, the method was applied to determine the targeted antipsychotics in plasma samples from schizophrenic patients undergoing therapy (Table 6). The measured plasma levels ranged from therapeutic to toxic levels. This variation could be explained by differences in the ability of patients to absorb, distribute, metabolize, and excrete the active compound due to interindividual variability (pre-existing disease, age, gender, or concomitant medication). Therefore, TDM should constitute an important tool to enhance the therapeutic response, design optimal dosing regimens, and monitor a patient’s adherence to treatment.

Table 6 Concentrations of the antipsychotic drugs in the schizophrenic patients’ plasma samples. The omitted values were found lower than the LLOQ
Full size table

Conclusion

A RACNT method was successfully applied for the first time in an offline extraction as a MEPS sorbent. The developed method achieved adequate sensitivity to determine antipsychotics in a reduced volume of plasma samples (100 μL) at sub-therapeutic levels. The MEPS–LC–MS/MS method allowed the direct analyses of human plasma (without pre-treatment) with the exclusion of about 97% of the proteins and other endogenous components. The evaluated validation parameters are in accordance with the EMA (European Medicines Agency) and the FDA (Food and Drug Administration) guidelines. In addition, the chemical stability of the sorbent allowed its reuse over 120 times without significant sensitivity loss. The developed method was successfully applied to determine antipsychotics drugs for TDM of schizophrenic patients.