Molecularly imprinted polymer for determination of lumefantrine in human plasma through chemometric-assisted solid-phase extraction and liquid chromatography

Author’s Accepted Manuscript Molecularly imprinted polymer for determination of lumefantrine in human plasma through chemometric-assisted solid-phase extraction and liquid cromatography Pedro Henrique Reis da Silva, Melina Luiza Vieira Diniz, Gerson Antônio Pianetti, Isabela da Costa César, Maria Elisa Scarpelli Ribeiro e Silva, Roberto Fernando de Souza Freitas, Ricardo Geraldo de Sousa, Christian Fernandes

PII: DOI: Reference:

www.elsevier.com/locate/talanta

S0039-9140(18)30207-8 https://doi.org/10.1016/j.talanta.2018.02.090 TAL18409

To appear in: Talanta Received date: 18 January 2018 Revised date: 23 February 2018 Accepted date: 24 February 2018 Cite this article as: Pedro Henrique Reis da Silva, Melina Luiza Vieira Diniz, Gerson Antônio Pianetti, Isabela da Costa César, Maria Elisa Scarpelli Ribeiro e Silva, Roberto Fernando de Souza Freitas, Ricardo Geraldo de Sousa and Christian Fernandes, Molecularly imprinted polymer for determination of lumefantrine in human plasma through chemometric-assisted solid-phase extraction and liquid cromatography, Talanta, https://doi.org/10.1016/j.talanta.2018.02.090 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Molecularly imprinted polymer for determination of lumefantrine in human plasma through chemometric-assisted solid-phase extraction and liquid cromatography

Pedro Henrique Reis da Silvaa, Melina Luiza Vieira Diniza, Gerson Antônio Pianettia, Isabela da Costa Césara, Maria Elisa Scarpelli Ribeiro e Silvab, Roberto Fernando de Souza Freitasb, Ricardo Geraldo de Sousab, Christian Fernandesa*

a

Laboratório de Controle de Qualidade de Medicamentos e Cosméticos, Departamento de Produtos

Farmacêuticos, Faculdade de Farmácia, Universidade Federal de Minas Gerais b

Laboratório de Ciência e Tecnologia de Polímeros, Departamento de Engenharia Química, Escola

de Engenharia, Universidade Federal de Minas Gerais Avenida Antônio Carlos, 6627, Pampulha, Belo Horizonte – MG, Brazil, 31270-901

* Corresponding author, Tel: +55 31 34096957; fax: +55 31 34096976. [email protected]

ABSTRACT Lumefantrine is the first-choice treatment of Falciparum uncomplicated malaria. Recent findings of resistance to lumefantrine has brought attention for the importance of therapeutic monitoring, since exposure to subtherapeutic doses of antimalarials after administration is a major cause of selection of resistant parasites. Therefore, this study focused on the development of innovative, selective, less expensive and stable molecularly imprinted polymers (MIPs) for solid-phase extraction (SPE) of lumefantrine from human plasma to be used in drug monitoring. Polymers were synthesized by precipitation polymerization and chemometric tools (Box-Behnken design and surface response

1

methodology) were employed for rational optimization of synthetic parameters. Optimum conditions were achieved with 2-vynilpyridine as monomer, ethylene glycol dimethacrylate as crosslinker and toluene as porogen, at molar ratio of 1:6:30 of template/monomer/crosslinker and azo-bisisobutironitrile as initiator at 65 °C. The MIP obtained was characterized and exhibited high thermal stability, adequate surface morphology and porosity characteristics and high binding properties, with high affinity (adsorption capacity of 977.83 μg g-1) and selectivity (imprinting factor of 2.44; and selectivity factor of 1.48 and selectivity constant of 1.44 compared with halofantrine). Doehlert matrix and fractional designs were satisfactorily used for development and optimization of a MISPE-HPLC-UV method for determination of lumefantrine. The method fulfilled all validation parameters, with recoveries ranging from 83.68 to 85.42%, and was applied for quantitation of the drug in plasma from two healthy volunteers, with results of 1407.89 and 1271.35 ng mL-1, respectively. Therefore, the MISPE-HPLC-UV method optimized through chemometrics provided a rapid, highly selective, less expensive and reproducible approach for lumefantrine drug monitoring. Graphical abstract

Keywords: Malaria; Lumefantrine; Molecularly imprinted polymer; Chemometrics; Molecularly imprinted solid phase extraction 1. Introduction Malaria is an infectious disease caused by protozoan of genus Plasmodium. Although its incidence has declined considerably over recent years, malaria remains as a public health issue in

2

many tropical countries [1,2]. According to World Health Organization (WHO), approximately 212 million cases and more than 400,000 deaths caused by malaria were notified worldwide in 2015, most of them among children under five years old, which is of extreme concern [2]. Plasmodium falciparum was responsible for almost all deaths (more than 99%), which makes treatment even more important. Currently, the first-choice treatment for uncomplicated and severe Falciparum malaria consists in a fixed-dose combination of artemether and lumefantrine [2,3]. Due to the increasing of Falciparum resistance to artemisinin-based combination therapy (ACT), monitoring of lumefantrine levels after administration is crucial to avoid therapeutic failure [4,5]. Biological matrices are generally very complex. Therefore, quantitation of drugs in biological fluids commonly requires a previous sample preparation that must be as efficient, precise, rapid, selective, reproducible, robust and cheap as possible [6,7]. Different sample preparation techniques are available, with particular advantages and drawbacks. Latterly, molecularly imprinted solid-phase extraction (MISPE) has emerged as a highly selective alternative for sample preparation. This approach uses synthetic biomimetic materials, known as molecularly imprinted polymers (MIPs), as sorbent [8-10]. MIPs are synthesized by polymerization of functional monomers and crosslinkers around a template, which can be the analyte of interest itself or an analogue molecule, leading to formation of crosslinked three-dimensional selective cavities. These cavities are complementary to template in size, shape and chemical function, and after removal of template, they are available for selectively binding the analyte from different samples, as biological fluids [8–10]. Apart from its high selectivity, a MIP also presents high thermal, mechanical and physicochemical stabilities. Furthermore, it is adaptable to different extraction techniques, such as solid-phase extraction (SPE) [11–13]. Several studies from different fields, such as food, pharmaceutical and environmental analysis demonstrate the versatility of molecularly imprinted polymers for sample preparation [11,14–16].

3

Nevertheless, ensuring high recovery and selectivity for a MIP can be very complex, because different variables can markedly influence on binding, thermal, mechanical and physicochemical properties of the polymers [10,11]. The optimization of these parameters is mandatory in order to obtain an effective material. Chemometric tools, such as design of experiments and surface response methodology, has gained relevance for multivariate optimization as it is more rational, reducing the number of experiments and reagent consumption [17–19]. Several methods for the determination of lumefantrine in biological matrices have already been published. However, all of them employed conventional approaches, such as liquid-liquid extraction (LLE) [20–23], protein precipitation (PP) [24–31], SPE [32–34] or combination of PP and SPE [35,36]. LLE, PP and SPE with conventional sorbents are extensively used in routine for bioanalytical purposes and can provide suitable results in some circumstances. However, a limited selectivity is a commonplace when these techniques are employed. From our knowledge, until the present moment, there is no study describing the synthesis and application of a MIP to extract lumefantrine from human plasma samples. Therefore, the aim of this study was to synthesize and characterize an innovative molecularly imprinted polymer for lumefantrine, to develop and validate an analytical method by MISPE and high performance liquid chromatography, and to apply the validated method for determination of lumefantrine in human plasma samples from healthy volunteers. In order to rationally optimize the synthesis protocol, the MISPE procedure and the chromatographic conditions, chemometrics were employed.

2. Materials and methods

2.1. Chemicals and reagents Lumefantrine and artemether chemical reference standards (CRS) were purchased from United States Pharmacopeia (Rockville, USA). Halofantrine CRS were granted from

4

GlaxoSmithKline (Rio de Janeiro, Brazil) and lumefantrine active pharmaceutical ingredient (API) was donated by Dafra Pharma (Turnhout, Belgium). Acetaminophen CRS, caffeine CRS, chloroquine diphosphate CRS, primaquine phosphate CRS and sodium artesunate CRS were donated by Farmacopeia Brasileira (Rio de Janeiro, Brazil). All monomers, crosslinkers and dichloromethane were obtained from Sigma-Aldrich (Steinheim, Germany). Acetonitrile and methanol HPLC grade were purchased from JT Baker (Xalostoc, Mexico) and Scharlau (New Jersey, USA). Trifluoroacetic acid (TFA) HPLC grade was purchased from Tedia (Fairfield, USA). Ethyl acetate, chloroform and toluene were purchased from Synth (Diadema, Brazil). Azobisisobutyronitrile (AIBN) was purchased from TCI (Saitama, Japan). Ultrapure water was acquired using a Millipore Gradient Direct Q3 system (Billerica, USA). Gas and liquid nitrogen were purchased from White Martins (Contagem, Brazil) and from Air Products (Belo Horizonte, Brazil), respectively. Human blood samples were collected from 12 healthy volunteers (10 for obtaining blank samples and 2 for obtaining real samples after drug administration) in fasting condition. Plasma samples were separated by centrifugation at 480 x g for 10 minutes and stored at -70 °C prior to use. This project was approved by the Committee on Ethics in Research of Universidade Federal de Minas Gerais (protocol number CAAE 54567716.4.0000.5149).

2.2. Chromatographic apparatus and conditions HPLC analysis was carried out in a Thermo Surveyor Finnegan chromatographic system (Waltham, USA), equipped with a quaternary pump, an autosampler, a diode array detector (DAD) and ChromQuest software. For Sohxlet efficiency evaluation and for static and kinetic adsorption studies, chromatographic separation was performed on a Zorbax SB-CN (250 x 4.6 mm, 5.0 μm) from Agilent Technologies (Santa Clara, USA), maintained at 30 °C. The mobile phase consisted of a mixture of methanol and TFA 0.05%, in a proportion of 75:25, at a flow rate of 1.0 mL min-1. Ultraviolet (UV) detection at 335 nm and injection volume of 20 μL were used. Chromatographic

5

runs for bioanalytical purposes were carried out using a Kinetex core-shell C18 (100 x 4.6 mm, 2.6 μm) column, purchased from Phenomenex (Torrance, USA). Other analytical conditions were previously optimized, as described in section 3.6, except for injection volume, which was established as 20 μL for all chromatographic runs.

2.3. Synthesis of molecularly imprinted polymers (MIP) for lumefantrine Polymers were synthesized by precipitation polymerization. A protocol of synthesis was optimized, since several experimental conditions can influence on physicochemical, mechanical, thermal and binding characteristics of a MIP. The first optimization step consisted of a 3 factors and 3 levels (+1, 0 and -1) Box-Behnken experimental design. The three parameters selected as factors were monomer (X1), crosslinker (X2) and porogen (X3). The experimental design led to a matrix consisting of 13 experiments, as shown in Table 1. The efficiency of copolymerization was evaluated with the combination of 2-vinylpyridine (2-VP) and methacrylic acid (MAA) in three different proportions (25:75, 50:50 and 75:25). Subsequently, two methods of activation (thermochemical and photochemical) of the radicalar initiator were compared. The thermochemically activated polymer was synthesized in a Nova Ética 304D (Vargem Grande do Sul, Brazil) thermostatic bath with orbital agitation maintained at 65 °C, for 24 hours, under mid agitation. The photochemical (UV irradiation) procedure was performed using a mercury lamp at lower temperature (125 W), in a Solab Dubnoff SL157 (Piracicaba, Brazil) thermostatic bath maintained at 4 °C. Thereafter, the efficiency of pre-polymerization was evaluated by storing the mixture of template, monomer and crosslinker at 4 °C for 4 and 8 hours prior to addition of the initiator. Finally, the molar ratio of lumefantrine to monomer was optimized, by testing four different proportions (1:4, 1:6, 1:8 and 1:10). Static adsorption studies were carried out for evaluation of binding properties of the polymers and quantitation of lumefantrine was performed with the HPLC-UV developed and validated method. Adsorption capacity at equilibrium (Q) and imprinting factor (IF) were defined as

6

dependent variables and Statistica 10.0 (Statsoft, Austin, USA) was used for statistical analysis and for obtaining response surfaces. Synthesis was optimized in order to obtain the highest Q and, primarily, the highest IF. Useful formulae and equations are provided in Supplementary data. Polymerization was performed as follows: 80 mg (0.151 mmol) of lumefantrine were weighed in a 25 mL beaker and dissolved in 20 mL of porogen. Then the solution was transferred to a 50 mL threaded test tube. After, the defined amounts of monomer and crosslinker were added to the mixture, which was purged with nitrogen for 5 min and, then, 20 mg (0.122 mmol) of AIBN dissolved in 5 mL of porogen was added. After mixing, the solution was purged again with nitrogen for 10 min and the tube was threaded and sealed with parafilm. Polymerization was performed for 24 hours. Non-imprinted polymers (NIPs) were synthesized as described for MIPs, without addition of lumefantrine. Polymers were then washed in a Soxhlet apparatus, with chloroform, for 48 hours, dried in vacuum oven at 60 °C for 2 hours, sieved to particles of less than 100 μm and stored in desiccators prior to use.

2.4. Physicochemical characterization Particle size determination and particle size distribution were performed, by laser diffraction, in a Beckman Coulter LS 12 320 (Brea, USA) equipment, for suspensions of 2 mg mL-1 in ultrapure water of MIP and NIP. The mean and SPAN values were determined for each polymer. The surface morphologies of polymers were observed by scanning electron microscopy (SEM). The micrograph images were obtained after sputter coating of polymers with gold in aluminum stubs, using a Thermo-Fischer Scientific Quanta 200 FEI (Waltham, USA) microscope at room temperature, voltage of 15 kV, working distance of 15.4 mm and at a magnitude of 50,000 x. Pore size, pore volume and surface area analyses were performed by nitrogen sorption porosimetry with the Brunauer-Emmett-Teller (BET) method (Quantachrome Nova 3200e, Boynton Beach, USA). Differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA) were performed in DSC-60 and TGA-50WS (Shimadzu, Kyoto, Japan), respectively. Temperature

7

program ranged from 25 to 300 °C for DSC. For TGA, temperature was from 25 to 600 °C for polymers and from 25 to 900 °C for lumefantrine. A heating rate of 10 °C min-1 and gas nitrogen at flow rate of 50 mL min-1 were used. Fourier transform infrared spectroscopy (FTIR) was performed in a FTIR Thermo-Fischer Scientific Nicolet 6700 spectrometer (Waltham, USA) at ATR mode, from 4000-400 cm-1, at room temperature and without any previous sample preparation.

2.5. Static and kinetic adsorption studies Static adsorption studies were performed in 25 mL conical flasks undergoing mid shaking in bath maintained at 25 °C by weighing and suspending 50 mg of polymer in 5 mL of a solution of lumefantrine at five different concentrations (12, 18, 24, 30 and 36 µg mL-1). After 48 hours, supernatants were filtered through 0.45 μm membranes and the amount of lumefantrine was determined by HPLC-UV. Then, Q and IF were calculated and fitting of the data to Langmuir and Freundlich adsorptions isotherms were investigated. Kinetic adsorption was carried out as described for static studies, with lumefantrine at 24 µg mL-1. However, at defined time intervals of 0.5, 1, 2, 4, 8, 12, 24 and 48 hours, 200 μL aliquots of supernatant were collected (with reposition) and filtered through 0.45 μm membranes to vials for quantitation of lumefantrine. Plotting and linear regression statistical analyses were performed with GraphPad Prism 5.0 (San Diego, USA).

2.6. Selectivity and reproducibility studies MIP selectivity was evaluated by comparing the static adsorption of lumefantrine and a chemically analog molecule (halofantrine), at concentrations of 24 µg mL-1. Q and IF for both drugs were determined and used for calculating the selectivity constant (α) and the selectivity factor (S). MIP selectivity for artemether, in terms of recovery, was assessed at 300 µg mL-1. Reproducibility was determined through calculations of coefficients of variation (CV) for Q and IF of three independent batches of polymers.

8

2.7. Development and optimization of a molecularly imprinted solid-phase extraction (MISPE) procedure for extraction of lumefantrine Since lumefantrine possesses a high level of protein binding, a previous protein precipitation step was necessary. This step was performed as follows: 500 μL of plasma were mixed with 500 μL of 0.2% perchloric acid (in acetonitrile) and centrifuged at 480 x g for 15 minutes. Thereafter, 500 μL of the supernatant were collected and transferred to MISPE cartridges. Commercially available cartridges were emptied, washed with a 60:40 v/v mixture of methanol and ultrapure water and sonicated for 20 minutes. Then, the cartridges were filled with suspensions of the defined amounts of MIP in 5 mL of pure methanol using a Phenomenex SPE vacuum manifold (Torrance, USA). The conditioning step of MIP cartridges was performed with 2 x 1 mL of methanol followed by 2 x 1 mL of ultrapure water, by centrifugation. The loading step was optimized by a 24-1 fractional factorial design of full resolution, with 4 factors, 2 levels (+1 and -1) and central point (0). The selected factors were pH (X1, ranging from 3.5 to 10.5), polymer weight (X2, ranging from 30 to 90 mg), centrifuge relative force (RCF, X3, ranging from 160 to 320 x g) and sample volume (X4, ranging from 500 to 1000 μL). Tests were performed at the highest level of concentration of the calibration curve (10 µg mL-1 of lumefantrine and 2 µg mL-1 of IS). The experimental design led to a matrix of 8 independent experiments, as demonstrated in Table 2. The central point was evaluated in triplicate (experiments 9 to 11).

Clean up and elution steps were defined by evaluating the desorption of lumefantrine from MISPE cartridges, employing the mobile phase and mixtures of TFA 0.05% and methanol or acetonitrile, in proportions ranging from 10 to 90% of organic solvent.

2.8. HPLC-UV bioanalytical method for determination of lumefantrine in human plasma samples Halofantrine was selected as internal standard (IS) due to its chemical and structural similarity to lumefantrine. The optimization procedure was carried out with a solution of both drugs at 5000 ng mL-1 each, in mobile phase. Three wavelengths of detection were tested: 240, 305 and

9

335 nm. Other analytical parameters were optimized with a 5-variables Doehlert matrix alongside with surface response methodology and desirability function, using Statistica 10.0. The parameters selected as factors were temperature (X1, ranging from 20.0 to 40.0 °C), flow rate (X2, ranging from 0.7 to 1.3 mL min-1), proportion of organic solvent (X3, ranging from 78.0 to 88.0%), TFA concentration (X4, ranging from 0.04 to 0.14%) and type of organic solvent (X5, consisting of pure methanol, a mixture of methanol and acetonitrile 50:50 v/v or pure acetonitrile). The matrix was composed by 31 independent experiments and central point was evaluated in triplicate, totalizing 33 experiments (Table 3). Run time was fixed at 30 min.

The choice of dependent variables (responses) included chromatographic parameters such as lumefantrine peak area, peak height, capacity factor (k), asymmetry (A), peak resolution, peak purity and theoretical plates (N), which were then combined into desirabilities values (D). The classification, target value and lower or upper limits for each dependent variable are presented in Table S.1. The validation was carried out according to European Medicines Agency (EMA), Food and Drug Administration (FDA) and Agência Nacional de Vigilância Sanitária (ANVISA) Guidances for Bioanalytical Method Validation [37–39]. The evaluated parameters were selectivity, carryover effect, matrix effect, linearity, precision, accuracy, limits of detection (LOD) and quantification (LOQ), recovery, and stability in biological matrix (short term, long term, post-processing and freeze-thaw) and in solution. Linearity was additionally assessed as proposed by Souza and Junqueira [40].

2.9. Application to human plasma samples of two healthy volunteers The efficacy of the MISPE-HPLC-UV method was evaluated by determining the concentration of lumefantrine from two healthy volunteers after oral administration of a single dose, consisting of 4 tablets of Coartem (Novartis, Basel, Switzerland), containing 20 mg of artemether

10

and 120 mg of lumefantrine each. Blood collection was performed only once, 6 hours after administration, according to a pharmacokinetic study carried out by our group [26].

3. Results and discussion

3.1. Synthesis of molecularly imprinted polymers (MIP) for lumefantrine Thirteen MIPs and thirteen NIPs for lumefantrine were synthesized with the Box-Behnken design. Q and IF for each experiment are presented in Table 4.

Unexpectedly, experiments 1, 9 and 10 demonstrated IF lower than 1, which indicates that control polymers (NIPs) presented higher values for Q than MIPs. These three experiments used TRIM as crosslinker, which suggests that TRIM was not appropriate. Since both lumefantrine and TRIM are lipophilic, nonspecific adsorption of the drug may have been more prominent. Experiments 2, 5, 6, 11, 12 and 13 demonstrated very similar Q for MIP and NIP, presenting an IF close to unit, showing that there was no molecular imprinting. These polymers were synthesized with MAA or 2-HEMA as monomer. Tarley and coworkers [10] argue that acidic monomers are more indicated for synthetizing a MIP for basic molecules. Therefore, as lumefantrine is a basic compound (with proton-acceptor groups), it would interact stronger with MAA or 2-HEMA instead of 2-VP, through hydrogen-bond formation. However, hydrogen-bonds may not have been formed because lumefantrine is a weak base; its most basic nitrogen consists of a very sterically hindered tertiary amine, surrounded by three large substituents, including 2 butyl groups. Surprisingly, 2-VP (a basic molecule) was the best monomer, since experiments 3, 4, 7 and 8 presented IFs higher than 1, which strongly suggests that specific cavities were formed. Because lumefantrine has an extremely weak acidic group (an alcoholic group), hydrophobic, dipole-dipole, van der Waals and π-π electrons interactions could have prevailed.

11

The statistical model fitted properly for IF, with a determination coefficient (r2) of 0.9341 and regression statistically significant (p < 0.05) at a confidence level of 95%. Thus, we can conclude that changes in factors explain approximately 93% of the variability observed in IF, which demonstrates adequate fit, accuracy and predictive power. However, the model fitted poorly for Q, with a r2 of only 0.3318 and regression statistically insignificant (p > 0.05). F-test and Pareto chart (Figure S.1) demonstrated that for Q, linear and quadratic terms of all factors as well as interactions between linear terms were not statistically significant (p > 0.05), while the linear term of monomer (X1) was shown to be the only significant factor (p < 0.05) for IF, which confirmed that monomer have the highest influence on IF. Surface responses (Figure 1) confirmed that 2-VP was the best monomer for IF, although the opposite was found for Q. However, as IF was considered as priority instead of Q, surface responses were in accordance with the previous discussion. They also showed that all 3 porogens were adequate for synthesis; however, toluene led towards the maximum region of the response. Additionally, toluene has higher boiling point and higher solubility for monomers and crosslinkers. Therefore, toluene was defined as porogen. Regarding to crosslinkers, Box-Behnken design was not conclusive. TRIM has a high swelling capacity, which led to longer extraction times, besides of its negative influence on IF. DVB and EGDMA demonstrated similar effect on binding properties and selectivity of polymers, but even though DVB is more commonly employed with 2-VP, EGDMA was chosen because it produced the best individual MIP (MIP-8) and higher yields.

Copolymerization of 2-VP and MAA increased significantly Q; nonetheless, it decreased even more significantly IF. In general, the higher MAA proportion was, the higher Q was (477.17, 682.74 and 715.27 μg g-1 for 25, 50 and 75% of MAA respectively). Oppositely, the higher MAA proportion was, the lower IF was (1.22, 1.13 and 1.12, respectively). Therefore, copolymerization was discontinued. UV irradiation, although presented better results in related works [41,42], did not lead to any benefit improvement of IF or Q. Thus, thermochemical activation was maintained. The

12

effect of 4 or 8 hours of pre-polymerization at 4 °C on binding properties of MIP was statistically irrelevant. However, as 4 hours improved IF on a small scale (from 1.76 to 1.81), it was included in the protocol. Increasing the ratio to 1:6 led to a clearly improvement (approximately of 60.0%) of Q (972.45 μg g-1) and IF (2.43). This occurred because the increase in the number of 2-VP molecules in the reactional mixture conducted to an increase of interactions between lumefantrine and monomer. Therefore, a more stable template-monomer complex was formed, resulting in a high number of specific binding sites and higher adsorption capacity and selectivity for polymers. Further increases of template/monomer ratios resulted in decreased IF (1.63 and 1.47 for 1:8 and 1:10, respectively), probably because of the increase of the amount of unspecific binding cavities. Thus, a proportion of template, monomer and crosslinker of 1:6:30 was fixed. After optimization, a MIP with Q of 977.83 μg g-1 and IF of 2.44 was obtained.

3.2. Physicochemical characterization The particle sizes (diameter) for MIP and NIP were respectively 38.59 ± 0.88 μm and 27.08 ± 1.14 μm. Thus, the presence of lumefantrine in the mixture contributes for the formation of specific cavities around the template and, consequently, larger particles. These particle diameters are appropriate for SPE cartridges, which generally uses sorbents of particle sizes ranging from 30 to 50 μm [43]. MIP showed a narrower particle size distribution, with a SPAN of 1.39 (NIP had a SPAN of 1.70), which ensures better repeatability for SPE extractions. SEM micrographs at magnitude of 50,000 x are disposed in Figure 2. MIP and NIP showed to have irregular surfaces with complex structure of internal small pores. MIP and NIP showed very similar pore diameters (8.97 and 8.45 Å, respectively) and, therefore, classified as microporous materials. This characteristic is essential for MISPE, since it promotes more accessible specific cavities and more favorable kinetic profile. It also increases reuse and decreases carryover. Surface area and pore volume were approximately 2.5 fold higher for MIP than NIP (139.68 m2 g-1 and 0.13

13

cm3 g-1 and 56.90 m2 g-1 and 0.05 cm3 g-1 for MIP and NIP, respectively), which indicate higher surface area, number of adsorption sites and a extraction efficiency, with fast mass transfer in the adsorption and desorption processes, justifying the higher Q for MIP. Also, these results are in agreement with those for particle diameter and suggest the molecular imprinting.

DSC curves (Figure 3, left) showed a short and broad endothermic peak at 76 °C for MIP and at 78 °C for NIP, which was initially associated to volatilization of non-reacted monomers and crosslinkers or to residual solvents. DSC curve for lumefantrine presented a sharp and pronounced endothermic peak at 135 °C, corresponding to its melting point, as already described in the literature. TGA thermograms showed that polymers were very stable until 330 °C, temperature where their decomposition started. A less important weight loss (less than 5.0%) with onset at 30-35 °C is probably associated to volatilization of non-reacted monomers and crosslinkers. Lumefantrine TGA curve confirmed its high thermal stability. Temperature was increased until 900 °C for the drug, but residual weight of 22% persisted. It can be related to the formation of non-volatile degradation products, as carbonized or inorganic residues. FTIR spectra for polymers (NIP and MIP before and after lumefantrine removal by Soxhlet extraction) and lumefantrine were obtained. MIP and NIP presented very comparable spectra and therefore very similar compositions. Bands at characteristic wavenumbers, such as 3450 cm-1 (O-H stretching of carboxylic acid), 2955 cm-1 (C-H stretching of aliphatic compounds), 1732 cm-1 (C=O stretching of carboxylic acid), 1637 cm-1 (C=N stretching), 1255 cm-1 (C-O stretching of carboxylic acid), 1154 cm-1 (C-O stretching of ester) and 950 cm-1 (O-H bending of carboxylic acid), confirmed the presence of 2-VP and EGDMA in the polymeric matrix. The absence of bands at 1650-1640 cm-1 is indicative of the absence of vinyl groups in polymers and confirms the polymerization of 2-VP. When comparing MIP spectra before and after removal of lumefantrine, no

14

deviation in bands at 3450 cm-1, 1732 cm-1 and 1637 cm-1 was observed, which suggests that the main interaction between the drug and the MIP was not by hydrogen-bond.

3.3. Static and kinetic adsorption studies A HPLC-UV method was developed and properly validated. The method was selective to lumefantrine, which was adequately separated from crosslinkers, monomers, and radicalar initiator used for synthesis of polymers. Furthermore, selectivity of the method was assessed by the absence of products of degradation in lumefantrine samples submitted to thermal (heat, at 70 °C) and photolytic (ultraviolet irradiation) degradation studies. The method was linear in the range of 4.8 to 43.2 µg mL-1 (r = 0.9967, residuals were independent, normal and homoscedastic, and p-value > 0.05, which confirmed that linearity deviation was not significant). Intraday and inter-days precision and accuracy were assessed at 3 different levels (low, medium and high), and coefficients of variation and standard relative errors were lower than 5.0%. Theoretical limits of quantitation (LOQ) and detection (LOD) of 3.2 and 0.3 µg mL-1, respectively, were calculated according to the equation of calibration curve. The method also showed appropriate robustness regarding pH of mobile phase, temperature, proportion of organic modifier, flow rate, methanol brand, time in the ultrasound bath and trifluoroacetic concentration. Static adsorption curves for MIP and NIP are shown in Figure 3 (left). Both polymers had similar adsorption behaviors, with an initial increase of Q, reaching a plateau at 24 µg mL-1. Probably, adsorption sites for lumefantrine have been saturated at this point. MIP showed significantly higher Q (higher affinity) for lumefantrine at all concentrations levels. The highest Q for MIP and NIP were 1003.10 μg g-1 and 404.01 μg g-1, respectively. Figure 3 (right) summarizes the results of kinetic adsorption studies for MIP and NIP. Both curves demonstrated an initial burst, with a rapid increase of Q within the first 30 minutes of experiment. Q increased proportionally with time increasing, but slopes were progressively smaller, until a plateau (equilibrium) was reached when binding sites were saturated (at around 24 hours for

15

MIP). Both curves exhibited similar adsorption profile and, at all evaluated times, MIP showed higher Q (and, therefore, higher affinity). The highest Q obtained were 1010.30 and 414.01 μg g-1 for MIP and NIP, respectively. Fitting of the equilibrium data to Langmuir and Freundlich adsorption isotherms were modeled, using concentration at equilibrium (Ce) and Q. The plots of Ce/Q versus Ce and lnQ versus lnCe were employed to obtain Langmuir and Freundlich isotherms (Figure S.2), respectively. Correlation coefficients (r) for MIP and NIP were 0.9998 and 0.9986, respectively, in the Langmuir model, and 0.9693 and 0.9478, respectively, in the Freundlich model. High correlation coefficients (> 0.9) were obtained, showing that the polymers fit adequately in both isotherm models and the variables are strongly correlated. However, higher correlation coefficients were obtained with Langmuir isotherm and, therefore, the multilayer adsorption model is more suitable to explain the adsorption between drug and MIP.

3.4. Selectivity and reproducibility studies The results of Q for MIP and NIP were respectively 993.63 and 404.78 μg g-1 (for lumefantrine) and 670.33 and 392.40 μg g-1 (for halofantrine), with IF of 2.46 and 1.71, respectively. NIP exhibited similar affinity for both molecules, since interactions are nonspecific. Nonetheless, MIP presented significantly higher affinity for lumefantrine, confirming that very selective binding cavities were formed. Selectivity factor and selectivity constant of 1.48 and 1.44, respectively, were calculated, showing an affinity approximately 50% higher for lumefantrine compared to halofantrine. These results demonstrated that the obtained MIP adequately discriminate lumefantrine, even for molecules structurally related. In every study already published dealing with lumefantrine determination in human plasma, conventional solvents and sorbents were used. Therefore, selectivity can be compromised, since these materials generally do not discriminate molecules with similar physicochemical properties, such as metabolites and drugs used

16

concomitantly. The selectivity of MIP for artemether, a potential interference, was also assessed. The recovery was lower than 15%, showing the poor affinity of the MIP for this drug. For reproducibility, the results obtained for three different batches ranged from 977.83 to 1015.04 μg g-1 for Q and from 2.42 to 2.53 for IF, with means of 995.90 μg g-1 and 2.47 and CV of 1.34% and 1.84%, respectively. The calculated CVs were considerably low, confirming the reproducibility. The obtained MIP can be used for lumefantrine extraction for approximately 25 times without losing its properties.

3.5. Development and optimization of a molecularly imprinted solid-phase extraction (MISPE) procedure for extraction of lumefantrine Extraction of lumefantrine and IS from human plasma applying only MISPE led to low recoveries, preventing direct employment of this technique without previous protein precipitation. These results are in consonance with those found in works of Lindergardh and coworkers [32] and Huang and coworkers [44] and can be explained by the high protein binding rates of lumefantrine (> 99.7%) and IS (between 60.0-70.0%) [45,46]. Therefore, a protein precipitation procedure, adapted from Huang et al. [44], was used. Centrifugation was preferred instead of vacuum system for the extractions because it is less time consuming (3-5 minutes for each step, in contrast of 15-20 minutes for vacuum system) and allows control of flow rate, optimization and more analyses at once. As demonstrated by F-test and Pareto chart (Figure S.3), all factors were statistically significant for lumefantrine and IS extraction (p > 0.05) in the 24-1 fractional factorial design. Polymer weight was the most critical parameter, followed by sample volume. RCF showed only a small influence, whilst pH was the least critical parameter. The interactions between polymer weight (X2) and relative centrifuge force (X3); polymer weight (X2) and sample volume (X4); and relative centrifuge force (X3) and sample volume (X4) were considered confounding effects. A correlation coefficient of 0.8894 confirmed the goodness of fit and high predictability of the model.

17

Cube plots for lumefantrine extraction (Figure S.4) evidenced that the adsorbed amount was considerably increased with increase of polymer weight, but the use of 60 or 90 mg of MIP did not led to significantly different results. So, in order to develop a more economic method, polymer weight was fixed at 60 mg. Oppositely, these plots showed that lower volumes of sample contributed to higher extraction; therefore 500 μL was defined as the ideal sample volume. RCF was also inversely related to lumefantrine and IS extraction. However, minimum tested level for RCF (160 x g) led to extractions that expended about 10 minutes to occur, increasing unnecessarily the time of analysis. Increasing RCF to 240 x g (mid-level) maintained a near identical adsorbed amount, but led to a significant reduction in analysis time. As expected, since hydrogen-bond was proved not to be the main interaction between polymers and drugs, pH was not an important parameter influencing lumefantrine extraction. The best analytical conditions for the loading step were respectively pH of 7.0, polymer weight of 60.0 mg, relative centrifuge force of 800 x g and sample volume of 500 μL. Desorption curves obtained with mixtures of methanol or acetonitrile with TFA 0.05% showed that acetonitrile is a stronger eluent for lumefantrine, since solubility of the drug is higher in acetonitrile. The mixture consisting of 90.0% of acetonitrile and 10.0% of TFA 0.05% desorbed 98.35% and was, therefore, the ideal solution for elution. However, tests with the mobile phase from the bioanalytical chromatographic method led to good similar results (desorption of 98.17%) and was, thus, defined as elution solution, in order to minimize effects of different elution strengths on peak shape. A mixture of 10.0% of methanol and 90.0% of TFA 0.05% desorbed only 2.78% of lumefantrine and was considered ideal for clean up.

3.6. Development and optimization of HPLC-UV bioanalytical method for determination of lumefantrine in human plasma samples With respect to the choice of wavelength for detection, 240 nm was the most sensitive, since lumefantrine and halofantrine have high molar absorptivities at this wavelength (Figure S.5).

18

However, in 240 nm, several interfering peaks from biological matrix appeared. Although selective and widely described in the literature for detection of lumefantrine, 335 nm led to considerably lower absorbance for IS. Thus, 305 nm was chosen because it presented intermediate selectivity and detectability. Doehlert matrix demonstrated that all five factors were statistically significant for desirability function (p > 0.05), as indicated by F-test and Pareto chart (Figure S.6). Temperature, proportion of organic solvent and type of organic solvent appeared, in this order, as the most significant parameters for the response. The results of the chromatographic parameters and the calculated D for each experiment are summarized in Table S.2. Ten surface responses were produced (Figure 4), considering linear and quadratic main effects and linear x linear second order interactions. The experimental design was well delimited, since most of the surface responses presented a maximum.

Visual inspection of surface responses shows that the optimum conditions were found around the central point for all five studied variables, mainly for temperature, proportion of organic modifier and TFA concentration. Optimum conditions were temperature of 30.0 °C, proportion of organic solvent of 83.0%, flow rate of 0.90 mL min-1, TFA concentration of 0.09% and organic phase consisting of a mixture of 60.0% of methanol and 40.0% of acetonitrile. The model was statistically relevant, since a correlation coefficient of 0.8970 was found. Furthermore, lack-of-fit was not significant (p > 0.05). Simultaneously, they confirmed the goodness of fit and the predictability of the model. A chromatographic run with these optimized analytical conditions was performed and retention times of 3.06 and 4.86 minutes for IS and lumefantrine, respectively, were observed. Most of the measured chromatographic parameters were properly optimized; however, resolution (10.58) and asymmetry (1.74) were considered unnecessarily high, even though they fulfilled FDA specifications. Thus, fine adjustments of analytical conditions were rationally proposed for

19

improving these parameters, especially asymmetry. Therefore, TFA concentration (the most influencing parameter for asymmetry according to the Doehlert design, as can be seen in the Pareto chart for asymmetry, in Figure S.7) was initially increased to 0.14% (further increases led to pH < 1.5, incompatible with the stationary phase). Subsequently, temperature (the second most influencing parameter for asymmetry) was increased from 30 °C to 35 °C (further increases to 40 °C did not led to relevant improvement of symmetry and influenced negatively on other chromatographic parameters, such as resolution, capacity factor and column lifetime). As expected, increasing temperature led to an undesirable decrease in capacity factor. Finally, flow rate was progressively adjusted from 0.9 mL min-1 to 0.7 mL min-1. Thereafter, the newly optimized analytical conditions (mobile phase composed of a mixture of methanol, acetonitrile and TFA 0.14% in a proportion of 50:33:17 v/v/v, at a flow rate of 0.7 mL min-1 and temperature of 35 °C) were considered adequate. Halofantrine and lumefantrine eluted at 3.13 and 4.71 minutes, respectively and all chromatographic parameters were fulfilled. The chromatographic run took only six minutes, allowing high throughput in routine analysis. Resolution decreased from 10.58 to 7.48; capacity factor was higher than 2.0 (for lumefantrine) even at a higher temperature and asymmetry was appropriately corrected to less than 1.30. Results for theoretical plates, peak area and height and signal-to-noise ratio for lumefantrine and IS confirmed that adequate detectability was achieved after optimization, since lumefantrine possesses high molar absorptivity in the UV region [47].

3.7. Method validation The chromatograms of lumefantrine at LLOQ and blank plasma samples (Figure S.8) showed that most of the interfering peaks appeared within void volume or between peaks of lumefantrine and halofantrine. Interfering peaks, found in the same retention time of lumefantrine and IS, ranged from 2.86 to 6.15% and 0.56 to 1.38% of the responses obtained for LLOQ, respectively. Also, selectivity was evaluated regarding to other antimalarials (artemether,

20

chloroquine diphosphate, primaquine phosphate and sodium artesunate) and acetaminophen and caffeine. No interfering peaks were found in the retention times of lumefantrine and IS, except for chloroquine, which showed a peak at the retention time of lumefantrine. Nevertheless, this peak presented a response corresponding to only 3.97% of LLOQ for lumefantrine. The method was, then, considered selective. This result was expected, since the obtained MIP was able to properly discriminate lumefantrine from similar molecules (as discussed in the section 3.4), which is not achieved with the conventional solvents and sorbents utilized in the sample preparation methods already reported in the literature. For carryover evaluation, there were no increase or appearance of interfering peaks in the same retention times of lumefantrine and IS. Responses for interfering peaks were not higher than 20.0% and 5.0% of LLOQ responses, which evidenced that no carryover was observed. The CV found for IS-normalized matrix factors (2.64%) were lower than 15.0%, proving that the matrix effect was not significant or was adequately corrected with the use of IS. IS-normalized matrix factors ranged from 0.97 to 1.05, showing that responses obtained in plasma or solution were extremely comparable. The highest differences were found for hemolyzed plasma samples and no statistically differences were found for different concentration levels. The recoveries of the SPE-cartridges of MIP (MISPE) at 3 different levels of concentration of lumefantrine ranged from 84.08-87.12% for the lowest level of concentration, from 82.9485.01% for the mid level of concentration, and from 82.79-84.16% for the highest level of concentration, with means of respectively (85.42 ±1.42%), (83.94 ±1.00%) and (83.68 ±0.64%). These results show high and repeatable recoveries independently of the concentration levels, ranging from 83.68-85.42% and with low CVs. These recoveries are higher than those found by other authors, as Lindergardh and coworkers [32], Blessborn and coworkers [34,35] and Sethi and collaborators [36], and comparable to those found by Annerberg and coworkers [33] and Huang and coworkers [44]. For halofantrine, the mean recovery obtained from three different calibration curves was 78.43 ± 2.48%.

21

The method showed linearity in the range of 50-10000 ng mL-1. Following the procedure proposed by Junqueira and Souza [40], for assessment of linearity, the 3 calibration curves fulfilled requirements of normality, homoscedasticity and independence of residuals. No outliers were found, significance of regression was confirmed by ANOVA and none of the three curves showed a significant linearity deviation, which confirms that data fits adequately the linear regression model without the need of a ponderation factor. Calibration curves were plotted for ratio of peak responses versus concentration of lumefantrine, as presented in Figure S.9. Slopes, intercepts, correlation coefficients (r), determination coefficients (r2) and regression equations obtained for each curve are disposed at Table 5. All curves presented r2 > 0.99, confirming that concentration of lumefantrine and peak responses ratio are variables very correlated. Also, there was no statistical significant difference between the 3 calibration curves (p > 0.05 for slopes and intercepts); so a mean (interdays) calibration curve was obtained. Intra and inter-days precisions and accuracies were shown, since CV and relative errors (Table 5) found were below 15.0% for QC samples and below 20.0% for LLOQ.

The relative errors found for each point of the three curves were below 20.0% for LLOQ and below 15.0% for other concentration levels, confirming linearity of the developed method. Linear regression parameters as slope (0.000803) and standard deviation of intercept (0.001208) were used for determination of LOD and LOQ. The calculated theoretical values were respectively 4.97 ng mL-1 and 15.04 ng mL-1. Thus, the method showed detectability higher [27,30,32–34,48], or, at least, comparable [21,23], than other methods employing ultraviolet detection. Even some mass spectrometry methods showed considerably higher LOQ [22,25,35]. These results highlighted that the rational optimization procedure improved detectability of the developed method. Deviations of all exposed samples from the nominal values ranged from -5.73 to 3.01% and, thus, were lower than the 15.0% allowed; therefore, lumefantrine showed to be sufficiently stable in plasma at all conditions (short term, long term, post-processing and freeze-thaw) and for all the time needed for analyses. In addition, at all tested conditions, lumefantrine and IS showed proper

22

stability in solution, since responses of exposed samples did not differ more than 10.0% from freshly prepared samples (ranging from -1.65 to 1.84%).

3.8. Application to healthy volunteer human plasma samples The developed method was successfully applied for determination of lumefantrine in human plasma samples. The plasmatic concentrations of lumefantrine, determined in triplicate, for volunteers 1 and 2 were 1407.89 ± 55.23 ng mL-1 and 1271.35 ± 47.58 ng mL-1, respectively. Figure 5 shows the chromatograms obtained. These concentrations could be increased if the administration of the tablets were carried out with the intake of milk, which increases the oral bioavailabilty.

4. Conclusions The chemometric tools used in the synthesis, extraction and chromatographic separation allowed to, rationally, choose the better levels for each parameter and to evaluate their possible interactions. The obtained innovative MIP showed high thermal stability (by DSC and TGA) and adequate morphological features (by nitrogen sorption porosimetry) and particle size (by laser ray diffraction) for its employment as sorbent in solid-phase extraction. It also demonstrated adequate binding properties, with adsorption capacity of 977.83 μg g-1, adequate for its purpose, and high selectivity (imprinting factor of 2.44, selectivity factor of 1.48 and selectivity constant of 1.44). The molecularly imprinted solid-phase extraction (MISPE) coupled to HPLC-UV approach was validated according to international requirements, showing suitable selectivity, precision, accuracy and linearity in the range of 50-10000 ng mL-1. No carryover and matrix effects were observed. Recoveries ranging from 83.68-85.42 were found. Both lumefantrine and IS proved to be stable at all tested conditions. Finally, the method was successfully applied to samples obtained from two healthy volunteers. With all this in mind, the MISPE-HPLC-UV method developed emerges as a

23

fast (only 6 minutes run), highly selective (more than methods with conventional sorbents) and sensitive (comparable to LC-MS methods for lumefantrine analysis) and less expensive approach for therapeutic monitoring of patients under lumefantrine therapy.

Conflicts of interest The authors declare having no conflict of interest.

Acknowledgements The authors acknowledge the financial support of the following Brazilian research agencies: Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG). The authors also acknowledge GlaxoSmithKline, Dafra Pharma, Farmacopeia Brasileira and ANVISA for providing the chemicals and reagents. Lastly, the authors would like to thank Centro de Microscopia, Laboratório de Hematologia and Laboratório de Tecnologia Farmacêutica, all from UFMG, for SEM analysis, blood collection and particle size determination and particle size distribution, respectively.

References [1]

P.J. Guerin, P. Olliaro, F. Nosten, P. Druilhe, R. Laxminarayan, F. Binka, W.L. Kilama, N. Ford, N.J. White, Malaria: Current status of control, diagnosis, treatment, and a proposed agenda for research and development, Lancet Infect. Dis. 2 (2002) 564–573.

[2]

World Health Organization, World malaria report 2016, WHO Press, Geneva, 2016.

[3]

World Health Organization, Guidelines for the treatment of malaria, third edition, WHO Press, Geneva, 2015.

[4]

H. Noedl, Y. Se, K. Schaecher, B.L. Smith, D. Socheat, M.M. Fukuda, Evidence of artemisinin-resistant malaria in Western Cambodia, N. Engl. J. Med. 359 (2008) 2619–2620.

24

[5]

A.-C. Uhlemann, M. Ramharter, B. Lell, P.G. Kremsner, S. Krishna, Amplification of Plasmodium falciparum multidrug resistance gene 1 in isolates from Gabon., J. Infect. Dis. 192 (2005) 1830–1835.

[6]

I.M.O. Viana, P.P.R. Lima, C.D.V. Soares, C. Fernandes, Simultaneous determination of oral antidiabetic drugs in human plasma using microextraction by packed sorbent and highperformance liquid chromatography, J. Pharm. Biomed. Anal. 96 (2014) 241–248.

[7]

V.F. Fagundes, C.P. Leite, G.A. Pianetti, C. Fernandes, Rapid and direct analysis of statins in human plasma by column-switching liquid chromatography with restricted-access material, J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 947–948 (2014) 8–16.

[8]

B. Sellergren, Direct drug determination by selective sample enrichment on an imprinted polymer, Anal. Chem. 66 (1994) 1578–1582.

[9]

M. Lasakova, P. Jandera, Molecularly imprinted polymers and their application in solid phase extraction, J.Sep.Sci. 32 (2009) 799–812.

[10]

C.R.T. Tarley, M.D.P.T. Sotomayor, L.T. Kubota, Polímeros biomiméticos em química analítica. Parte 1: preparo e aplicações de MIP (“molecularly imprinted polymers”) em técnicas de extração e separação, Quim. Nova. 28 (2005) 1076–1086.

[11]

F.G. Tamayo, E. Turiel, A. Martín-Esteban, Molecularly imprinted polymers for solid-phase extraction and solid-phase microextraction: recent developments and future trends, J. Chromatogr. A. 1152 (2007) 32–40.

[12]

D.A. Spivak, Optimization, evaluation, and characterization of molecularly imprinted polymers, Adv. Drug Deliv. Rev. 57 (2005) 1779–1794.

[13]

P.A.G. Cormack, A.Z. Elorza, Molecularly imprinted polymers: synthesis and characterisation, J. Chromatogr. B. 804 (2004) 173–182.

[14]

C.. Allender, C. Richardson, B. Woodhouse, C.. Heard, K.. Brain, Pharmaceutical applications for molecularly imprinted polymers, Int. J. Pharm. 195 (2000) 39–43.

[15]

S.A. Piletsky, S. Alcock, A.P.F. Turner, Molecular imprinting: at the edge of the third

25

millennium, Trends Biotechnol. 19 (2001) 9–12. [16]

F. Qiao, H. Sun, H. Yan, K.H. Row, Molecularly imprinted polymers for solid phase extraction, Chromatographia 64 (2006) 625–634.

[17]

B. Barros Neto, I.S. Scarminio, R.E. Bruns, Como fazer experimentos: Pesquisa e desenvolvimento na ciência e na indústria, Artmed, Porto Alegre, 2007.

[18]

M.A. Bezerra, R.E. Santelli, E.P. Oliveira, L.S. Villar, L.A. Escaleira, Response surface methodology (RSM) as a tool for optimization in analytical chemistry, Talanta. 76 (2008) 965–977.

[19]

D.C. Montgomery, Design and analysis of experiments, eighth edition, John Wiley & Sons, New Jersey, 2012.

[20]

Wahajuddin, S.P. Singh, G.K. Jain, Determination of lumefantrine in rat plasma by liquidliquid extraction using LC-MS/MS with electrospray ionization: assay development, validation and application to a pharmacokinetic study, J. Chromatogr. B 877 (2009) 1133– 1139.

[21]

I.F. Khalil, U. Abildrup, L.H. Alifrangis, D. Maiga, M. Alifrangis, L. Hoegberg, L.S. Vestergaard, O.P.E. Persson, N. Nyagonde, M.M. Lemnge, T.G. Theander, I.C. Bygbjerg, Measurement of lumefantrine and its metabolite in plasma by high performance liquid chromatography with ultraviolet detection, J. Pharm. Biomed. Anal. 54 (2011) 168–172.

[22]

L. Huang, X. Li, F. Marzan, P.S. Lizak, F.T. Aweeka, Determination of lumefantrine in small-volume human plasma by LC–MS/MS: using a deuterated lumefantrine to overcome matrix effect and ionization saturation, Bioanalysis 4 (2012) 157–166.

[23]

F. Khuda, Z. Iqbal, Y. Shah, L. Ahmmad, F. Nasir, A.Z. Khan, Amanullah, N. Shahbaz, Method development and validation for simultaneous determination of lumefantrine and its major metabolite, desbutyl lumefantrine in human plasma using RP-HPLC/UV detection, J. Chromatogr. B 944 (2014) 114–122.

[24]

E.M. Hodel, B. Zanolari, T. Mercier, J. Biollaz, J. Keiser, P. Olliaro, B. Genton, L.A.

26

Decosterd, A single LC-tandem mass spectrometry method for the simultaneous determination of 14 antimalarial drugs and their metabolites in human plasma, J. Chromatogr. B 877 (2009) 867–886. [25]

V. Munjal, N. Paliwal, B.K. Chaursia, B. Varshney, T. Ahmed, J. Paliwal, LC-Tandem Mass Spectrometry method for quantification of lumefantrine in human plasma and its application to bioequivalence study, Chromatographia 71 (2010) 505–510.

[26]

I.C. César, J.A. de Aquino Ribeiro, L. de Souza Teixeira, K.B. Bellorio, F.C. de Abreu, J.M. Moreira, P.R. Chellini, G.A. Pianetti, Liquid chromatography-tandem mass spectrometry for the simultaneous quantitation of artemether and lumefantrine in human plasma: application for a pharmacokinetic study, J. Pharm. Biomed. Anal. 54 (2011) 114–120.

[27]

B. Maganda, O. Heudi, A. Cortinovis, F. Picard, O. Kretz, O. Minzi, A fast and reliable reversed phase high performance liquid chromatography method for simultaneous determination of selected anti-retroviral and lumefantrine in human plasma, J. Chromatogr. B 919–920 (2013) 52–60.

[28]

A.V. Silva, N. Mwebaza, M. Ntale, L.L. Gustafsson, A. Pohanka, A fast and sensitive method for quantifying lumefantrine and desbutyl-lumefantrine using LC-MS/MS, J. Chromatogr. B 1004 (2015) 60–66.

[29]

K. Govender, L. Gibhard, L. Du Plessis, L. Wiesner, Development and validation of a LCMS/MS method for the quantitation of lumefantrine in mouse whole blood and plasma, J. Chromatogr. B 985 (2015) 6–13.

[30]

S. Sandhya, P. Shiji Kumar, S. Meena, A sensitive liquid chromatograhic assay for the simulaneous determination of lumefantrine and artemether in human plasma, Indo Am. J. Pharm. Res. 5 (2015) 720–726.

[31]

M. Wahajuddin, S.P. Singh, I. Taneja, K.S. Raju, J.R. Gayen, H.H. Siddiqui, S.K. Singh, Simultaneous quantification of proposed anti-malarial combination comprising of lumefantrine and CDRI 97-78 in rat plasma using the HPLC-ESI-MS/MS method:

27

application to drug interaction study, Malar J. 14 (2015) 172. [32]

N. Lindegårdh, A. Annerberg, D. Blessborn, Y. Bergqvist, N. Day, N.J. White, Development and validation of a bioanalytical method using automated solid-phase extraction and LC-UV for the simultaneous determination of lumefantrine and its desbutyl metabolite in plasma, J. Pharm. Biomed. Anal. 37 (2005) 1081–1088.

[33]

A. Annerberg, T. Singtoroj, P. Tipmanee, N.J. White, N.P.J. Day, N. Lindegårdh, High throughput assay for the determination of lumefantrine in plasma, J. Chromatogr. B 822 (2005) 330–333.

[34]

D. Blessborn, S. Römsing, A. Annerberg, D. Sundquist, A. Björkman, N. Lindegardh, Y. Bergqvist, Development and validation of an automated solid-phase extraction and liquid chromatographic method for determination of lumefantrine in capillary blood on sampling paper, J. Pharm. Biomed. Anal. 45 (2007) 282–287.

[35]

D. Blessborn, S. Romsing, Y. Bergqvist, N. Lindegardh, Assay for screening for six antimalarial drugs and one metabolite using dried blood spot sampling, sequential extraction and ion-trap detection., Bioanalysis 2 (2010) 1839–1847.

[36]

P. Sethi, V.K. Dua, R. Jain, A LC-MS/MS method for the determination of lumefantrine and its metabolite desbutyl-lumefantrine in plasma from patients infected with Plasmodium falciparum malaria, J. Liq. Chromatogr. Relat. Technol. 34 (2011) 2674–2688.

[37]

EMA, Guideline on bioanalytical method validation, London, 2012.

[38]

FDA, Guidance for Industry: bioanalytical method validation, Silver Spring, 2013.

[39]

ANVISA, Resolução RDC no 27, Brasília DF, 2012.

[40]

S.V.C. De Souza, R.G. Junqueira, A procedure to assess linearity by ordinary least squares method, Anal. Chim. Acta. 552 (2005) 23–35.

[41]

Y. Zhang, Q. Na, N. Chen, J. Chen, X. Zhou, L. Bai, Preparation and evaluation of melamine molecularly imprinted polymers by thermal- and photo-initiation methods, J. Chinese Chem. Soc. 60 (2013) 204–211.

28

[42]

H.-H. Pan, W.-C. Lee, Synthesis of molecularly imprinted polymer and its molecular recognition properties of N-acetylneuraminic acid, E-Journal Chem. 4 (2007) 611–619.

[43]

A. Zwir-Ferenc, M. Biziuk, Solid phase extraction technique - trends, opportunities and applications, Polish J. Environ. Stud. 15 (2006) 677–690.

[44]

L. Huang, P.S. Lizak, A.L. Jayewardene, F. Marzan, M.N.T. Lee, F.T. Aweeka, A modified method for determination of lumefantrine in human plasma by HPLC-UV and combination of protein precipitation and solid-phase extraction: application to a pharmacokinetic study, Anal. Chem. Insights. 2010 (2010) 15–23.

[45]

D. Colussi, C. Parisot, F. Legay, G. Lefèvre, Binding of artemether and lumefantrine to plasma proteins and erythrocytes, Eur. J. Pharm. Sci. 9 (1999) 9–16.

[46]

B. Cenni, J. Meyer, R. Brandt, B. Betschart, The antimalarial drug halofantrine is bound mainly to low and high density lipoproteins in human serum., Br. J. Clin. Pharmacol. 39 (1995) 519–526.

[47]

J.K. Verma, H.A. Syed, Spectrophotometric method for determination of lumefantrine in pharmaceutical formulations, J. Pharm. Res. 2 (2009) 1550–1551.

[48]

M. Ntale, J.W. Ogwal-Okeng, M. Mahindi, L.L. Gustafsson, O. Beck, A field-adapted sampling and HPLC quantification method for lumefantrine and its desbutyl metabolite in whole blood spotted on filter paper, J. Chromatogr. B 876 (2008) 261–265.

Figure captions Figure 1 – Surface responses for adsorption capacity (top) and imprinting factor (bottom) for MIPs obtained from the Box-Behnken design.

29

Figure 2 – SEM micrographs for MIP (left) and NIP (right) at 50,000 x of magnitude. Figure 3 – Static (left) and kinetic (right) adsorption curves for MIP and NIP. Figure 4 – Surface responses for desirability obtained from the Doehlert matrix for optimization of a bioanalytical chromatographic method for quantitation of lumefantrine. Figure 5 – Chromatograms of plasma samples from volunteers 1 and 2 obtained with the HPLC-UV validated method.

Table 1 – Box-Behnken design matrix for the synthesis optimization of a MIP for lumefantrine.

Experiment 1 2 3 4 5 6 7 8 9 10 11 12 13 (central point)

Tested values (coded values) X1 X2 X3 2-HEMA (+1) TRIM (+1) TOL (0) 2-HEMA (+1) DVB (-1) TOL (0) 2-VP (-1) TRIM (+1) TOL (0) 2-VP (-1) DVB (-1) TOL (0) 2-HEMA (+1) EGDMA (0) CHL (-1) 2-HEMA (+1) EGDMA (0) ETAC (+1) 2-VP (-1) EGDMA (0) CHL (-1) 2-VP (-1) EGDMA (0) ETAC (+1) MAA (0) TRIM (+1) CHL (-1) MAA (0) TRIM (+1) ETAC (+1) MAA (0) DVB (-1) CHL (-1) MAA (0) DVB (-1) ETAC (+1) MAA (0) EGDMA (0) TOL (0)

2-HEMA: 2-hidroxyethylmethacrylate; 2-VP: 2-vinylpyridine; CHL: chloroform; DVB: divinylbenzene; EGDMA: ethyleneglycol dimethacrylate; ETAC: ethyl acetate; MAA: methacrylic acid; TOL: toluene; TRIM: trimethylolpropane triacrylate.

Table 2 – 24-1 fractional factorial experimental design matrix for the MISPE optimization.

Experiment 1 2 3 4 5 6 7 8 9 10

X1 3.5 (-1) 10.5 (+1) 3.5 (-1) 10.5 (+1) 3.5 (-1) 10.5 (+1) 3.5 (-1) 10.5 (+1) 7.0 (0) 7.0 (0)

Tested values (coded values) X2 X3 X4 30 (-1) 160 (-1) 500 (-1) 30 (-1) 160 (-1) 1000 (+1) 90 (+1) 160 (-1) 1000 (+1) 90 (+1) 160 (-1) 500 (-1) 30 (-1) 320 (+1) 1000 (+1) 30 (-1) 320 (+1) 500 (-1) 90 (+1) 320 (+1) 500 (-1) 90 (+1) 320 (+1) 1000 (+1) 60 (0) 240 (0) 750 (0) 60 (0) 240 (0) 750 (0)

30

11

7.0 (0)

60 (0)

240 (0)

750 (0)

Table 3 - Doehlert matrix for the bioanalytical chromatographic method optimization.

Experiment 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

X1 30.0 (0) 30.0 (0) 30.0 (0) 40.0 (+1) 35.0 (+0.5) 35.0 (+0.5) 35.0 (+0.5) 35.0 (+0.5) 20.0 (-1) 25.0 (-0.5) 25.0 (-0.5) 25.0 (-0.5) 25.0 (-0.5) 35.0 (+0.5) 35.0 (+0.5) 35.0 (+0.5) 35.0 (+0.5) 25.0 (-0.5) 30.0 (0) 30.0 (0) 30.0 (0) 25.0 (-0.5) 30.0 (0) 30.0 (0) 30.0 (0) 25.0 (-0.5) 30.0 (0) 30.0 (0) 30.0 (0) 25.0 (-0.5) 30.0 (0) 30.0 (0) 30.0 (0)

Tested values (coded values) X2 X3 X4 1.0 (0) 83.0 (0) 0.09 (0) 1.0 (0) 83.0 (0) 0.09 (0) 1.0 (0) 83.0 (0) 0.09 (0) 1.0 (0) 83.0 (0) 0.09 (0) 1.3 (+0.866) 83.0 (0) 0.09 (0) 1.2 (+0.289) 88.0 (+0.817) 0.09 (0) 1.2 (+0.289) 84.25 (+0.204) 0.14 (+0.791) 1.2 (+0.289) 84.25 (+0.204) 0.10 (+0.158) 1.0 (0) 83.0 (0) 0.09 (0) 0.7 (-0.866) 83.0 (0) 0.09 (0) 0.8 (-0.289) 78.0 (-0.817) 0.09 (0) 0.8 (-0.289) 81.75 (-0.204) 0.04 (-0.791) 0.7 (-0.866) 81.75 (-0.204) 0.08 (-0.158) 0.8 (-0.289) 83.0 (0) 0.09 (0) 0.8 (-0.289) 78.0 (-0.817) 0.09 (0) 0.8 (-0.289) 81.75 (-0.204) 0.04 (-0.791) 0.8 (-0.289) 81.75 (-0.204) 0.08 (-0.158) 1.3 (+0.866) 83.0 (0) 0.09 (0) 1.1 (+0.577) 78.0 (-0.817) 0.09 (0) 1.1 (+0.577) 81.75 (-0.204) 0.04 (-0.791) 1.1 (+0.577) 81.75 (-0.204) 0.08 (-0.158) 1.2 (+0.289) 88.0 (+0.817) 0.09 (0) 0.9 (-0,577) 88.0 (+0.817) 0.09 (0) 1.0 (0) 86.75 (+0.613) 0.04 (-0.791) 1.0 (0) 86.75 (+0.613) 0.08 (-0.158) 1.2 (+0.289) 84.35 (+0.204) 0.14 (+0.791) 0.9 (-0,577) 84.25 (+0.204) 0.14 (+0.791) 1.0 (0) 79.25 (-0.613) 0.14 (+0.791) 1.0 (0) 83.0 (0) 0.13 (+0.633) 1.2 (+0.289) 84.25 (+0.204) 0.10 (+0.158) 0.9 (-0,577) 84.25 (+0.204) 0.10 (+0.158) 1.0 (0) 79.25 (-0.613) 0.10 (+0.158) 1.0 (0) 83.0 (0) 0.05 (-0.633)

X5* 0 0 0 0 0 0 0 +0.775 0 0 0 0 -0.775 0 0 0 -0.775 0 0 0 -0.775 0 0 0 -0.775 0 0 0 -0.775 +0.775 +0.775 +0.755 +0.775

*For factor X5 (type of organic solvent): -0.775 corresponds to 100.0% of methanol; +0.775 corresponds to 100.0% of acetonitrile; and 0 corresponds to a mixture consisting of 50.0% of methanol and 50.0% of acetonitrile.

Table 4 – Adsorption capacities (Q) and imprinting factors (IF) for MIPs obtained from the BoxBehnken design. Experiment

Adsorption capacity (Q) (μg g-1)

Imprinting factor (IF)

31

1

540.96

0.50

2

558.30

0.99

3

1022.41

1.56

4

583.73

1.68

5

718.79

0.92

6

595.74

1.06

7

649.04

1.21

8

598.87

1.74

9

244.51

0.37

10

566.78

0.46

11

918.14

0.98

12

922.10

0.99

13

773.32

1.10

Table 5 – Figures of merit of linear regression data of intra-days and inter-days for evaluation of linearity, precision and accuracy. Parameter Slope Intercept Correlation coefficient (r) Determination coefficient (r2) Precision (CV, %) Accuracy (RE, %)

Day 1 0.000829 ± 0.000007 0.001879 ± 0.001005 0.999342 0.998685 1.21-4.98 0.38-12.45

Results Day 2 Day 3 0.000808 ± 0.000005 0.000803 ± 0.000006 0.001788 ± 0.000917 0.001848 ± 0.001439 0.999588 0.999849 0.999177 0.999698 0.98-7.21 1.75-6.48 -04.5-4.56 -1.79-1.66

Mean (inter-days) 0.000813 ± 0.000004 0.001802 ± 0.001208 0.999193 0.998387 1.38-7.91 -0.37-5.27

32

33

34

HIGHLIGHTS 

Innovative molecularly imprinted polymer for lumefantrine extraction was synthesized



Characterization of the MIP showed adequate stability, morphology and selectivity



Chemometric approach was used for optimization of a MISPE-HPLC-UV for lumefantrine



A rapid bioanalytical method was validated for determination of lumefantrine in plasma



The MISPE-HPLC-UV method was successfully applied for two healthy volunteers

35