Miniaturized Chip Integrated Ecological Sensor for the Quantitation of Milnacipran Hydrochloride in the Presence of Its Impurities in Dosage Form and Human Plasma

The expanded graphite (EXG) was synthesized after modifications on the following reported scheme. ²⁵ Briefly, 6 ml of concentrated H₂SO₄ (98%) was mixed with 10 gm of (NH₄)₂S₂O₈ in a sealed 50 ml beaker and was subjected to low power sonication for 5 min till a homogenous slurry was formed. A 2 gm of G was added to the slurry and mixed gently under an overhead agitator for 5 min at room temperature till reaching the desired consistency and the formation of intercalation graphite (IG). The resulting IG compound was partially decomposed by exposure to 2.5 MHz microwave irradiation for 60 s till the worm-like flakes appeared, and the EXG was formed. The EXG product was then left to cool down for an additional 60 s and was stored in a rubber-sealed glass vial.

The PANI particles were synthesized utilizing the interfacial polymerization technique after optimizing the synthesis conditions based on previous computational simulations.

Computational design of PANI imprinted polymer

First, Aniline and MLN molecules were modeled in MOE using universal smiles codes from the PubChem database. ²⁶ Structure preparations, protonation, partial charges calculation, and energy minimizations were also performed based on the EHT: Amber 10 force field. Molecule files in.sdf format were saved in a combined file for the next step in DFT calculations.

DFT geometry optimizations and binding energy calculations under different solvents were conducted using the ORCA 5.0.4 software. ²⁷ Geometry optimizations were performed utilizing the Bp86 functional besides the split valence polarization function (def2-SVP), Grimme's dispersion correction method (D3BJ), and the integral equation formalism variant (IEFPCM) solvation mode to simulate the effect of each different solvents (CCl₄, CHCl₃, ether, dichloromethane) on the binding between MLN and aniline. By the end of the convergence, HOMO-LUMO energies, binding energies, solvation energies, and gap energies were calculated for the MLN-aniline complex under each solvent.

PANI molecules were modeled on MOE by the polymer building tool, then the PANI structure was prepared for protonation and error corrections using the structure preparation tool and finally was stored as a PDB file. A 0.7 ns molecular dynamic simulation (MD) in water as solvent at pH = 5 was performed for both MLN and PANI using Nose-Poincare Andersen equations of motion, 0.5 ps sampling window, 0.002 ps time step. The simulation time was segregated as follows; 0.03 ns heating from 0 to 300 K followed by an equilibration phase of 0.1 ns, then a production period of 0.54 ns followed by a cooling period of 0.03 ns. By the end of the simulation, the trajectory files were collected and sent for analysis by origin pro software.

Synthesis of non-imprinted and imprinted Ag-PANI polymer

Ag-PANI particles were prepared by interfacial polymerization utilizing two immiscible phases. To the first aqueous phase (20 ml deionized LC/MS grade water), we added 0.25 gm of AgNO₃ and sonicated for 5 min till complete dissolution and formation of solution A. Solution A was stirred on a magnetic stirrer, and 1.07 ml of H₂SO₄ was added to it. Afterward, we added 0.28 gm of standard MLN to 40 ml of diethyl ether/CCl₄ (50:50% v/v) under continuous stirring for 5 min, then 0.75 ml of aniline was added to the mixture to prepare solution B. Finally, into a 100 ml beaker, solution B was added to solution A; the beaker was sealed and continued stirring for 24 h till the completeness of polymerization as indicated by the formation of deep green color of the Ag-PANI (Emraldin). The resulting particles were washed several times using methanol: acetic acid (90:10% v/v) till complete clearance of the MLN from the particles. Finally, the particles were washed with deionized water several times to remove residual washing solvent and dried overnight at room temperature. As shown above, the non-imprinted Ag-PANI polymer was prepared without adding the MLN into the organic phase.

The Nano-Ag ink was prepared following the procedure reported by ²⁸ after slight modifications in the technique. Briefly, 1.02 gm of AgNO₃ was dissolved into 60 ml of EG by sonication for 10 min and left on a magnetic stirrer for 10 min; then, the resulting solution was held into a burette (solution A) for the next step. Another solution of 10 gm of PVP dissolved into 60 ml of EG and heated on a hot plate at 100 °C for 5 min until the solution turned yellow (solution B). At continuous heating at 100 °C, add solution A from the burette at a slow rate (0.05 ml s⁻¹) to the hot solution B with constant stirring at 700 rpm using an overhead agitator. After the complete addition of solution A, the whole mixture was kept stirred at 100 °C for 3 h to permit the capping of the reduced Ag particles with PVP. Excess PVP was removed from the mixture using 100 ml of ethanol added to it under sonication for 3 h. Eventually, the particles were collected by cool centrifugation at 4 °C and 10000 rpm for 15 min After that, the supernatant was decanted and replaced by ethanol/acetone (50:50%v/v) solution, sonicated for 5 min, to wash the particles for several cycles with the aid of centrifugation at 10000 rpm and 15 min for each cycle. Finally, when the resulting supernatant was clear, the particles were collected into a 5 cm petri dish and were flash dried in a microwave oven for 90 s to evaporate residual ethanol/acetone and prevent excessive oxidation of Ag particles. The dried particles were dispersed into 10 ml of methanol and stored in a tightly closed glass vial under refrigeration at 4 °C till needed.

A diagrammatic illustration summing up the assembly process of the proposed chip sensor was supplied in Fig. 1a.

Zoom In Zoom Out Reset image size

Preparation of the quasi-reference (Ag/AgCl) electrode

A PVC cocktail was prepared by weighing 190 mg of PVC powder and transfer into a 50 ml sealed vial containing 5 ml of THF and stirred by magnetic stirrer for 2 min, then 0.34 ml of NOPE was added with continuous stirring for an additional 5 min till the complete dissolution of all components and formation of a homogenous cocktail solution. A 300 μL of the above cocktail was transferred to a 2 ml Eppendorf containing 6 mg of EXG, 3 mg of the non-imprinted Ag-PANI fine powder, and 20 μL of the Nano-Ag ink, and then the Eppendorf was tightly sealed and subjected to vortex followed by a 30 min sonication into an ice bath. A 50 μL of the above mixture was printed to the unprotected half (the position designated for reference electrode) of a polyethylene tetra phthalate (PET) chip. The printed layer was dried with a hot air gun (60 °C) for 15 min; then, the protective film was carefully removed. Finally, the electrode was immersed into a (40 mg ml⁻¹) NaOCl solution for 30 s to form the AgCl layer, then washed with deionized water and dried in a hot air gun (60 °C) for 5 min.

Preparation of the MLN-sensitive electrode

A similar PVC cocktail was prepared as described above but with the addition of 8 mg of the ion associate (MLN-TPB) and 0.34 ml of NOPE. A 300 μL of the above cocktail was transferred to a 2 ml Eppendorf containing 3.9 mg of the imprinted Ag-PANI fine powder and 6 mg of EXG, and the Eppendorf was tightly sealed and subjected to vortex followed by 30 min sonication into an ice bath. A 50 μL of the above mixture was printed to the un-protected half (the position designated for the sensor electrode) of the same PET chip to construct sensor III (D₁). The printed layer was dried by a hot air gun (60 °C) for 15 min. Then, the protective film was carefully removed. Finally, the crocodile connectors were applied, and the electrodes were insulated by paraffin film. The sensor I comprised 0.35 ml of NOPE and 10 mg of the MLN-TPB without Ag-PANI particles. While sensor II was of a similar composition to sensor I with the addition of 5 mg of Ag-PANI particles, and both sensors were fabricated in a similar fashion to sensor III.

Another design of the proposed sensors (D₂), implying that each electrode was printed on a separate PET chip, was executed according to the above procedures without the need for protective films during their printing.

Each sensor was calibrated by immersing the non-insulated part into a 5 ml measuring cell containing varying concentrations of MLN standard solution in the range of (1 × 10⁻⁷ − 1 × 10⁻² M). The obtained potential due to each concentration vs each concentration was plotted, and the slope of each sensor was recorded.

The stability of the reference electrode, either integrated with the MLN sensitive electrode on the same chip or printed on a separate PET chip, was tested by investigating the resulting potential from Cl⁻ solution (KCl) of various concentrations in the range of (1 × 10⁻⁵ − 1 × 10⁻¹ M) against a commercial double junction Ag/AgCl electrode. Finally, the relation between the - Log of the solution's molar concentration vs the resulting potential (mV) was plotted, and the slope of the reference electrode was computed accordingly.

Commercially available dosage form averomilan^®, each tablet was claimed to contain 50 mg of MLN, was analyzed utilizing sensor III (D₁). Ten tablets from each product were pulverized in a clean, dry mortar into a fine powder, then a suitable aliquot representing the average weight of one tablet was quantitatively transferred to a 100 ml volumetric flask containing 70 ml BR buffer (pH 5). The flask was subjected to vortex for 10 min, then sonicated for 30 min, and the volume was completed to the mark with the same solvent. Finally, a suitable amount was transferred into a 15 ml falcon tube for cool centrifugation at 10000 rpm for 10 min. An appropriate aliquot (10 ml) was withdrawn from the clear supernatant and transferred quantitatively into another 100 ml volumetric flask. The volume was completed using the same solvent to achieve a final concentration of 1.77 × 10⁻⁴ M. A suitable volume was transferred to the 5 ml measurement cell, and the resulting potential was recorded by sensor III (D₁) accordingly.

The human plasma was devoid of drugs and was procured from the tumor marker research unit situated at the Faculty of Pharmacy, Al-Azhar University, Cairo, Egypt, before the commencement of the actual investigations. Various aliquots ranging from 0.88 to 1.3 ml were precisely withdrawn from standard solutions of MLN with a concentration of 1 × 10⁻⁵ M and transferred into 100 ml volumetric flasks. Subsequently, 1 ml of human plasma was introduced and eventually diluted to the designated level with BR solution, which had been adjusted to a pH of 5, resulting in a final concentration range of 0.88 × 10⁻⁶ to 1.3 × 10⁻⁶ M. Prior to potentiometric measurement using the suggested sensors, the solution was subjected to sonication for the duration of 1 min. This concentration range was selected to correspond to the highest concentration of MLN that has been seen after a single dose (MLN's Cmax is 250 ng ml⁻¹). Following that, the suggested method's protocols were used to analyze MLN directly in plasma. The corresponding regression equation was used to get the recovery percentage and standard addition once the DVN concentration was measured.

The stability of the integrated quasi-reference electrode was essential to be tested for overall stable sensor response. The reference electrode was investigated regarding its stability and was tested three consecutive times (n = 3) on three different newly fabricated sensors. The quasi-reference electrode was responsive and sensitive to changes in concentrations of various Cl⁻ solutions in the range of (1 × 10⁻⁴ − 1 × 10⁻¹ M) with an average slope value of (56.3 ± 0.625 mV decade⁻¹). Pertaining to the Nernestian slope values obtained from the reference electrodes, the proposed Ag/AgCl electrode was qualified to act as a pseudo reference in the proposed sensor.

This study aimed to analyze the impact of pH levels using BR buffer, which ranges from 2 to 12, on the electrode potential response. The experiment utilized 1 × 10⁻⁴ and 1 × 10⁻³ M solutions of MLN. The data was presented as a graph showing a relation between recorded potential values for each mentioned concentration obtained at different pH levels against pH levels (Figs. 3a–3c). It was noticed that the potential stability of all sensors (II and III) was achieved in a pH range of (4–8) while sensor I was stable in the range of (5–8). Below pH 4, a significant increase in the potential was noticed, most probably due to the interference of the elevated proton concentration with the measurements. Above pH 8, there was a noticeable decrease in the potential, most probably due to the deprotonation of the MLN ions in the high pH medium or the interference from the elevated hydroxyl concentration in the medium. From these data, we concluded that preserving the pH value at 5 was critical for the stable response of the proposed sensors. Although sensor III had a wide pH range (4–8), the commitment to pH 5 was reasoned by the requirement of the PANI particles for sufficient concentration of protons in the surrounding media to perform its ultimate conductivity function enhancing the overall sensor sensitivity. ²⁹ The Prepared Ag-PANI nanoparticles were in the salt form (-NH₂-) and still if the surrounding medium was acidic enough. Since the conductivity of PANI is dependent on the allocation of the imine protons onto its polymeric chain of the salt form so, if the medium pH was above 5 a gradual decrease in the PANI conductivity is noticed due to the deprotonation of the PANI salt into the base (-NH-) form. ³⁰

Zoom In Zoom Out Reset image size

The suggested sensors' selectivity in the existence of interfering ions was assessed using the separate solutions approach. For this work, six different inorganic ions in the form of chloride salts were chosen. The suggested sensor responded to the inorganic ions in a non-Nernstian pattern, as seen in Fig. S3a. Additionally, the method's selectivity was checked by applying matched potential method (MPM). This method was evaluated using different organic components, related impurities (PTD and BCN), and other co-formulated drugs (mirtazapine) with MLN. The results were illustrated in Fig. S3b, proving the selectivity of the fabricated sensor towards MLN ions.

Dynamic response time

The dynamic response time is a crucial factor in the analytical utilization of ion-selective sensors. This study indicated that the modified sensor (sensor 3) had a faster response time than the other two fabricated sensors (sensor I and II) within the concentration range of 10^–7 to 10^–2 M, which took five seconds to achieve a stable potential.

The proposed sensor's electrochemical performance was evaluated per the IUPAC guidelines data. The calibration curves section was followed to test the sensors and the PVC membrane electrodes; the detection limit, linear range, and slope were monitored at specific time periods. After conducting several weeks of evaluation, it was determined that the PVC polymeric membrane is capable of withstanding the soaking effect for a duration of up to 7 weeks.

The MD simulations aimed to scan the effect of each organic solvent on the non-bonding interactions between MLN (target) and the aniline (monomer) in the Ag-PANI synthesized particles. The trajectory files from the 0.7 ns simulations of the MLN-aniline complex in four different solvents (CCl₄, CH₃Cl, CH₂Cl₂, and ether) were analyzed regarding the radius of gyration (R_g), root mean square error (RMSD), hydrogen bonding energies (HB) and the non-bonded interactions energies (NB). The involvement criterion for solvents was limited by the solubility of both the monomer and MLN facilitating the optimal interaction between them.

The initial HB energies analysis of the four simulations revealed a relative abundance of the hydrogen bonding in MLN-aniline complex in ether according to the following order ether > CHCl₃ > CCl₄ > CH₂Cl₂ as seen in Fig. S4a, which implied the choice of ether as the progenic solvent for polymerization that support HB between MLN and aniline. The NB interaction energies analysis results aligned with those from HB energies, as seen in Fig. S4b. Due to the closeness of the HB and NB energies' values for the MLN—aniline complex in both CHCl₃ and CCl₄ making the comparison between them impractical, we chose to compare the R_g values of the investigated complex in these two solvents as shown in Fig. S4c. The complex of lower Rg values reflected the compactness of its structure and its superior stability which was achieved under CCl₄ rather than CH₃Cl clarifying the ambiguity in the HB and NB obtained results.

By examining the RMSD data of the MLN-aniline complex in all four solvents, both ether and CHCl₃ showed a significant un-stability due to its elevated and fluctuating RMSD profile (Fig. S4d). Again we needed to decide the second most suitable solvent, which achieves abundant HB interaction between MLN and aniline in their complex and shows acceptable stability. This criterion made us conclude that CCl₄ was the second most suitable choice regarding its relatively stable RMSD profile.

Finally, the PANI polymer was designed utilizing the MOE polymer building tool then the whole polymer was equilibrated with MLN ion in CCl₄ for 0.7 ns under the same above MD simulation criteria. The obtained equilibrium MLN-PANI complex structure was saved in the.pdb format after constraining the PANI molecule at the obtained equilibrium conformation to simulate the induced imprinting effect by the presence of MLN ion, as seen in Fig. 4a. Also, the selectivity of the proposed PANI equilibrium conformation towards MLN ion in the presence of its related impurities PTD and BCN was tested in an aqueous medium (pH 5). The RMSD analysis of this simulation (0.63 ns) revealed that the MLN-PANI complex had the highest stability among other complexes indicating the selectivity of the proposed PANI equilibrium structure towards MLN ions Fig. 4b.

Zoom In Zoom Out Reset image size

The BP86 function was selected here for some specific characters, making it the best option for screening purposes, as the situation implies here. These characters can be summarized as; being pure DFT functional. It is computationally cheaper and faster than hybrid functional making it more useful for small systems that require higher accuracy and efficiency. Also, BP86 can give reliable results about the geometries and bond energies in smaller systems. On the other hand, unlike the hybrid functional, the BP86 functional has many drawbacks while encountering hydrogen bonding systems. This work compensated for these pitfalls by involving both the def2-SVP basis set and the D3BJ method.

The def2-SVP basis set was utilized because it can be extended to optimizations of the hydrogen bonding systems' energy and structure. ³¹ This method compensated for the shortness of the BP86 in dealing with hydrogen bonding as it effectively accounts for the anisotropy and electronic distortions around the hydrogen atoms.

Also, the involvement of the D3BJ method was essential to correct the BP86 calculations for the dispersion effect, compensating for the shorthand abilities of the BP86 in simulating hydrogen bonding and enhancing the accuracy of the utilized method. The net result was a faster, computationally cheaper, and more accurate method capable of simulating and optimizing the geometry of a small hydrogen bonding system with greater reliability.

The DFT output files resulting from each solvent were analyzed regarding the energies of the optimized molecules in single and complex form, and the binding energy of the MLN-aniline was calculated according to the following equation:

$$\begin{eqnarray*}&&{{\rm{E}}}_{{\rm{Binding}}}={{\rm{E}}}_{{\rm{opt}}\left({\rm{MLN}}-{\rm{aniline}}\right)}\unicode{x02014}\left[{{\rm{E}}}_{{\rm{opt}}\left({\rm{aniline}}\right)}+{{\rm{E}}}_{{\rm{opt}}\left({\rm{MLN}}\right)}\right]\end{eqnarray*}$$

By comparing the resulting binding energies in the investigated solvents, ³² CCl₄ and ether both showed the highest binding energies, as shown in Fig. S5a. Also, after calculating the dipole moment (μ) of the MLN-aniline complex in each solvent, it was noticed that there was a linear increase in the complex dipole moment with the rise in the dielectric constant (ε) of each different solvent which implies the charge transfer establishment in the complex stabilized by the increment in the dielectric constant of the surrounding solvent molecules (Fig. S5b). Besides, the molden files were generated in Orca software to investigate the HUMO-LUMO orbitals and calculate the MLN-aniline complex's gap energies (ΔE_H-L) in different solvents. After plotting the data of gap energy vs the dipole moment of the MLN-aniline complexes in other solvents, it was noticed that CCl₄ and ether provided the lowest gap energies for the complex, as in Fig. S5c. The lower gap energy of the MLN-aniline complex is related to the enhancement in its electroconductive properties as the lower ΔE_H-L facilitates the electron transfer between HUMO and LUMO orbitals and the current flow through the complex or through the transducers the complex is composing. ³³ Also, the lower ΔE_H-L suggests that MLN would be tightly bound to the aniline upon approaching its polymeric network of the Ag-PANI particles.

From the above data and the data pertained from MD simulations, we concluded that CCl₄ was the most suitable choice as a non-aqueous layer during the interfacial polymerization and imprinting of Ag-PANI particles. Due to the low solubility of the MLN in CCl₄, we decided to use a co-solvent system composed of (1:1 v/v) CCL₄: ether to aid the MLN solubility into the organic layer without affecting the established interaction between MLN and aniline during polymerization.

Molecular docking was performed to test the selectivity of the imprinted polymer towards MLN ions in the presence of its related impurities BCN and PTD. A semi-flexible docking protocol was followed utilizing the triangle matcher placement method with a rigid receptor, London ΔG scoring function, and the generalized Born/solvent accessible surface area (GBVI/WSA ΔG) rescoring function. A database composed of the conformational search results for all potential ligands (MLN, BCN, and PTD) was utilized to perform docking against the equilibrium structure of PANI as a receptor. The resulting docking poses were obtained, as shown in Fig. S6a, revealing that MLN had the highest score indicating the selectivity of the proposed PANI MIP towards the MLN ions. The highest score and lowest RMSD poses were selected and isolated for the MD 0.63 ns simulation in a buffered aqueous medium, as previously detailed under the MD simulation section.

Five different MLN concentrations were examined in triplicate to determine linearity. The observed potential responses and their associated concentrations had an established linear relationship, as tabulated in Table IV.

Both repeatability and intermediate precision of the proposed methods were evaluated, where the repeatability was done by determining the average of three concentrations repeated three times within a single day. Additionally, the intermediate precision was assessed by calculating the average of three concentrations repeated three times over three consecutive days. The results of both types of precision were interpreted in the form of percent relative standard deviation (% RSD), as summarized in Table IV.

The accuracy of an analytical methodology is determined by the degree of conformity between the measured value and the actual value. It shows the % recovery of three MLN measurements, as presented in Table IV.

The LOD is the minimal concentration that the technique can detect, albeit it is not always quantified as an exact and precise amount. LOD was calculated by extending the highest Nernstian concentration and the lowest non-responsive concentration lines and subsequently determining the main ion concentration where these lines intersect. The LOD results of the proposed methods are shown in Table IV.

Modifications were made to the experimental variables, specifically the pH with a tolerance of ±0.2, to evaluate their impact on the robustness of the methods. The results indicate that the minor adjustments made during the experimental procedure did not significantly affect the measured potential, as shown in Table IV.

The built-in sensors were used for MLN analysis in pharmaceutical formulation (Averomilan tablet). Based on the percentage of recoveries and standard deviation (SD) obtained, it was shown that the proposed sensors could detect the specified drug. As shown in Table V, the percentage recovery and SD were reliable and exact, confirming the sensor's applicability. Furthermore, a standard addition method was used to guarantee that tablet excipients didn't interfere, as shown in Table V.

Additionally, the validity of the proposed sensor was then studied by comparing the results to those of the published approach, but no statistically significant difference was found where the estimated t and F values were less than the tabulated ones (Table VI).

The results obtained from utilizing the recommended sensor for measuring DVN in human plasma are presented in Table VII. The outcomes obtained from individual sensors demonstrated the assay's accuracy and precision.