Investigation of isomerization of dexibuprofen in a ball mill using chiral capillary electrophoresis

Besides the racemate, the S‐enantiomer of ibuprofen (Ibu) is used for the treatment of inflammation and pain. Since the configurational stability of S‐Ibu in solid state is of interest, it was studied by means of ball milling experiments. For the evaluation of the enantiomeric composition, a chiral CE method was developed and validated according to the ICH guideline Q2(R1). The addition of Mg2+, Ca2+, or Zn2+ ions to the background electrolyte (BGE) was found to improve Ibu enantioresolution. Chiral separation of Ibu enantiomers was achieved on a 60.2 cm (50.0 cm effective length) x 75 μm fused‐silica capillary using a background electrolyte (BGE) composed of 50 mM sodium acetate, 10 mM magnesium acetate tetrahydrate, and 35 mM heptakis‐(2,3,6‐tri‐O‐methyl)‐β‐cyclodextrin (TM‐β‐CD) as chiral selector. The quantification of R‐Ibu in the mixture was performed using the normalization procedure. Linearity was evaluated in the range of 0.68–5.49% R‐Ibu (R2 = 0.999), recovery was found to range between 97 and 103%, the RSD of intra‐ and interday precision below 2.5%, and the limit of quantification for R‐ in S‐Ibu was calculated to be 0.21% (extrapolated) and 0.15% (dilution of racemic ibuprofen), respectively. Isomerization of S‐Ibu was observed under basic conditions by applying long milling times and high milling frequencies.


Introduction
There are numerous chiral drugs on the market, which are administered as racemic mixtures, although pharmacological activity often resides in one enantiomer only. Its chiral antipode may show reduced activity, no activity at all, or even toxicity. The probably most widely known example of a toxic distomer is the S-isomer of thalidomid showing pronounced teratogenic effects [1]. Most of the β-blockers like metoprolol and bisoprolol salts are applied as racemic mixtures, although it is known that pharmacological activity is almost exclusively ascribed to one enantiomer only [2]. The same holds true for nonsteroidal anti-inflammatory drugs (NSAIDs) such as flurbiprofen and ibuprofen [3].
Ibuprofen is an NSAID with one stereogenic center at the α carbon atom. It inhibits the cyclooxygenase providing analgesic, antipyretic, and antiphlogistic effects, because it prevents the production of inflammatory prostaglandins [4,5]. Hence, ibuprofen is frequently used for the treatment of slight and moderate pain, and rheumatic diseases. The anti-inflammatory effects of Ibu can be attributed to the (+)-S-enantiomer (dexibuprofen), whereas (-)-R-Ibu is inactive (Fig. 1). Since there is a unidirectional enzymatic conversion of R-into S-Ibu in vivo [6][7][8], the racemic mixture is usually applied. However, interindividual differences in pharmacokinetics of the enantiomers might justify the use of enantiomeric pure S-Ibu [9,10], which was thus launched for patient's treatment also. Certainly, a compendial monograph for S-Ibu does not exist.
Isomerization of S-Ibu was already observed in vitro under acidic and basic conditions, where a kinetic model of base-catalyzed isomerization was evaluated [11][12][13]. The experiments were carried out in aqueous solution or in the melt applying high temperatures and long reaction times, including the usage of strong organic bases as isomerization additives [13]. However, the ibuprofen enantiomers were found to be highly stable under acidic conditions. Furthermore, isomerization processes in solid state have been described in the scope of asymmetric synthesis or using thermal or irradiating stress [14][15][16][17][18][19].
Mechanochemistry can offer attractive alternatives allowing a control of the reaction conditions by performing the processes in automated devices such as ball mills [20][21][22]. The grinding, shearing, pulling, kneading, and milling in such instruments transmits mechanical energy to reactive media inducing chemical reactions including bond-breaking and -forming processes. Many of these transformations proceed in the absence of solvent resulting in exceptionally high reactivities and leading to unusual product compositions. Recently, ball milling has been proven to be predictive for degradation processes of a drug [23]. To the best of our knowledge, mechanochemistry has not been used to investigate the isomerization of Ibu so far. Hence, the isomerization of S-Ibu, which was stressed in a ball mill under solid-state conditions using basic, acidic, and/or liquid additives, was studied here.
Up to now there is no validated CE method in literature applied on the determination of the enantiomeric purity of S-Ibu; reported CE methods dealing with the quantification of Ibu enantiomers are either validated [28,29,33], or tested on enantiomeric purity [37,39]. Hence, for quantitative evaluation of the R-Ibu content in the S-enantiomer after ball milling, a simple CE separation using TM-β-CD as a chiral selector and a magnesium salt as cationic additive was developed and validated as an addition to other proposed CE methods and HPLC [42].

Instrumentation
Stressing experiments (20-30 Hz milling frequencies) were performed using a Retsch MM400 ball mill from Retsch GmbH (Haan, Germany) with a milling jar (10 mL) and milling ball (10 mm diameter) made of ZrO 2 -Y (zirconia dioxide stabilized with yttria) or stainless steel. An IST 636 ball mill from InSolido Technologies (Zagreb, Croatia) was used for the milling frequency of 35 Hz. For evaporation of solvents, a Heidolph Laborota 4000 rotary evaporator from Heidolph Instruments GmbH & Co.KG (Schwabach, Germany) was used.
CE separations were performed using a P/ACE MDQ capillary electrophoresis system from Beckman Coulter (Krefeld, Germany), equipped with a DAD. The fused silica capillaries purchased from BGB Analytik (Rheinfelden, Germany) were of 60.2 and 50.2 cm total length (50.0 or 40.0 cm effective length) with an i.d. of 75 and 50 μm, respectively.

CE procedure
The final method was carried out on a 60.2 cm (50.0 cm effective length) x 75 μm fused silica capillary. The optimized BGE (background electrolyte) consisted of 35 mM TM-β-CD, 50 mM sodium acetate buffer (pH 5.0), and 10 mM magnesium acetate tetrahydrate. The cartridge temperature was maintained at 27°C and a voltage of + 26 kV was applied (E = 432 V/cm, current approximately 75 μA). The detection was performed using a DAD at λ = 202 nm (high sensitivity filter). The samples were injected hydrodynamically by applying a pressure of 3.45 kPa for 5 s on the anode side. Sample concentration was 0.6 mg/mL in 0.005 M NaOH.

Stressing of S-Ibu
Isomerization tests were performed using S-Ibu in form of its free acid according to the milling processes shown in Tables S1 and S2 (n = 1, Supporting Information). For each experiment, the milling jar and milling ball were of the same material: ZrO 2 -Y for schemes a, b,d-g or stainless steel for scheme c. For each milling process, the jar was loaded with 103.1 mg of S-Ibu (0.5 mmol) and optionally additives before milling. KOH was used as a basic additive at 2.0 molar equivalents (56 mg), and FeCl 3 as a Lewis acid at 0.1 molar equivalents (8.0 mg), respectively. As grinding auxiliaries, 103 mg of SiO 2 (1 weight equivalent) or 500 mg Al 2 O 3 were applied.
In case of using KOH as an additive, after the milling procedure, the mixture was neutralized with an aqueous solution of 1 M HCl. The pH was checked to be 6-7, using pH paper (schemes b, f, g). Afterward, the mixture was extracted with ethyl acetate and evaporated. If grinding auxiliaries such as SiO 2 or Al 2 O 3 were applied, the extracted mixture was filtered before evaporation. The resulting powder was used for CE analysis.

BGE solutions
1 M sodium acetate buffer was prepared by dissolving 1.43 mL of acetic acid ≥ 99.8% in water, adjusting to the desired pH with 10 M sodium hydroxide solution and subsequent dilution to 25.0 mL with water. The 0.1 M magnesium acetate solution was prepared by dissolving 536 mg of magnesium acetate tetrahydrate in water and diluted to 25.0 mL with water. BGE stock solutions were stored at 4°C, protected from light and used within 7 days. The final BGE was always freshly prepared by mixing the appropriate volumes of acetate buffer and magnesium acetate stock solutions, adding the required amount of the respective CD. The mixture was diluted with water to achieve the desired concentration.

Ibu solutions
Stock solutions of S-and racemic Ibu were prepared by dissolving 50.0 mg of the respective compound in 10.0 mL aqueous 0.05 M NaOH.
Test solutions for validation were prepared by mixing 600 μL of S-Ibu stock solution with varying amounts of rac-Ibu stock solution, diluting with water to 5.0 mL. For initial method development, rac-Ibu solutions were prepared by diluting a 5 mg/mL stock solution of rac-Ibu in methanol to a concentration of 0.4 mM. For the method optimization on the larger i.d. capillary, a 0.6 mg/mL S-Ibu solution in 0.005 M NaOH was used. All solutions were stored at 4°C, protected from light.
The sample solutions were prepared by dissolving the respective sample in 0.005 M NaOH, resulting in a concentration of 0.6 mg/mL. Samples were vortexed and sonicated until complete dissolution. Sample measurement was conducted in randomized order in triplicates. The solutions were passed through a 0.20 μm PVDF filter and degassed prior to injection.

Method development
According to Rawjee et al. [24], only the uncharged forms of the Ibu enantiomers are interacting with the CD as chiral selector. However, when using a neutral CD, differences in electrophoretic mobility and hence resolution are only possible if Ibu (pK S ≈ 4.5 [43]) is partially charged. Consequently, the buffer pH plays an important role for the resolution of the Ibu enantiomers. The abovementioned concept was adopted for the CE method development approach, using an untreated fused silica capillary and applying the normal polarity mode (anode at the inlet). TM-β-CD was used as chiral selector because it has been proven to be well suitable for chiral recognition of the Ibu enantiomers [26]. An acetate buffer was chosen because it is suitable for ensuring a stable pH near the pK S of Ibu. The separation was initially developed on a 50.2 cm x 50 μm fused silica capillary, investigating the effects of the pH, the concentration of BGE components (buffering agent, cationic additives) and TM-β-CD on the migration time and resolution R S of the Ibu enantiomers. R S was calculated according to the equation where t m is the migration time and w 0.5 is the peak width at half height. To enhance sensitivity, the method was transferred to a 75 μm i.d. capillary, considering the adjustment of the separation parameters.  least 125 mM. However, the resulting high currents and Joule heating led to massive band broadening, drifting, and noisy baselines.
Next, it was aimed to combine low separation currents and sufficient enantioresolution. The experiments were carried out using a BGE composed of 75 mM acetate buffer (pH 5.5) and 30 mM TM-β-CD, applying a voltage of 20 kV (E = 398 V/cm) and adding varying amounts of Zn 2+ ions. Cationic additives are known to modulate the EOF by interacting with the negatively charged capillary wall. To decrease the conductivity of the BGE, zinc sulfate heptahydrate was considered as polycationic buffer additive. The Zn 2+ concentration was varied from 1-10 mM and the effects on migration time, resolution, and current were investigated (Table 2). Migration times and resolution of the Ibu enantiomers consistently increased with rising Zn 2+ concentration, while the resulting current did not change considerably. Thus, it can be concluded that adding Zn 2+ ions to the BGE is more efficient than high acetate concentrations regarding resolution and separation current. However, the Zn 2+ cations were reduced to Zn 0 on the cathode during separation, leading to arbitrary current distortions. Hence, cations with a smaller redox potential but similar size were considered, such as Mg 2+ . 10 mM zinc sulfate heptahydrate could be replaced by the same molar amount of magnesium sulfate heptahydrate, achieving similar separation results. The separation conditions with a BGE composed of 100 mM acetate buffer pH 5.25, 15 mM MgSO 4 × 7H 2 O, applying a voltage of 29 kV resulted in migration times of less than 10 min for the Ibu isomers and an R s value of 2.1.
The TM-β-CD concentration was varied between 10 and 80 mM to study the impact on migration time, resolution, and current. The respective results are depicted in Fig. 2. Higher TM-β-CD concentrations increased the migration times and the resolution of the Ibu enantiomers ( Fig. 2A). The current during separation decreased due to the lowering conductivity of the BGE (Fig. 2B). In fact, satisfying resolution (R S ≥ 2) and migration times could be achieved with TM-β-CD concentrations of 30-40 mM.
A detection wavelength of λ = 202 nm was chosen, even though the Ph.Eur. used λ = 214 nm in the HPLC method for the test on related substances [44], our CE method showed the highest sensitivity for detection of the R-isomer at λ = 202 nm.
To sum up the method development results obtained on the 50.2 cm x 50 μm capillary, sufficient separation (R S = 2.1) could be achieved in less than 10 min using the abovementioned separation conditions. Current during separation was   [25,26,45,46]. As part of our method development, this was studied by adding 1 and 5% (V/V) methanol and acetonitrile to the BGE, respectively. As shown in Table 3, increasing amounts of the organic modifier resulted in prolonged migration times, which was more pronounced in the case of methanol compared to acetonitrile. The current during separation was slightly decreased. However, severe and irreproducible baseline shifts were observed at methanol concentrations of 5% V/V (Fig. 3). Hence, methanol and ace- tonitrile were withdrawn for further method development. Moreover, organic solvents were avoided for the preparation of Ibu samples because they led to noisy baselines and sample overloading effects. 0.005 M NaOH was used instead.
The cationic additives magnesium, sodium and potassium salts were tested at a concentration of 12 mM and compared with respect to separation performance. Therefore, 12 mM magnesium sulfate, magnesium chloride, magnesium acetate, calcium chloride, sodium chloride or potassium chloride were added each to the BGE. The enantiomers were not resolved using NaCl or KCl. Best results regarding resolution and migration time were achieved using magnesium acetate and magnesium sulfate. Since magnesium sulfate led to noisy baselines after the S-Ibu peak, magnesium acetate was applied (Table 4).
Finally, a sufficient separation (migration time about 21 min, R S = 2.6) including a smooth baseline was achieved on a 60. The capillary temperature was studied as described in Table 5. Elevated temperatures provided faster migration times and still sufficient resolution. Best results were achieved at a temperature of 27°C.
To ensure constant migration times, the rinsing procedure between each run was optimized. Rinsing schemes were tested as suggested by Wahl and Holzgrabe [47,48], where acidic, basic and organic liquids were used in varying orders upon flushing of a 60.2 cm (50.0 cm effective length) × 50 μm capillary. For our tests, the rinsing durations were adjusted accordingly. If no 0.1 M NaOH was applied as a rinsing solvent, the migration times dramatically increased, which was also observed when successively water, 0.1 M NaOH and BGE were applied. Flushing with organic solvents such as methanol followed by basic rinsing led to decreasing migration times. The optimized rinsing sequence was water (0.5 min), 0.1 M NaOH (6 min), water (1 min) and BGE (3 min). After each separation, a vial with fresh BGE was used, coping with changes such as degradation and/or electrolysis in BGE and, consequently, in separation conditions. Injection parameters including injection duration and sample concentration were optimized to achieve the maximum sensitivity, which was determined by evaluation of the R-Ibu peak height. The injected S-Ibu concentration was tested from 0.1 to 0.8 mg/mL, applying a pressure of 3.45 kPa for 4 s at the anode side. As shown in Fig. 4, the peak height got more intense with increasing S-Ibu concentration, reaching a maximum at about 0.65 mg/mL. At higher concentrations, the peak height dropped because of peak broadening due to overloading effects. The resolution decreased because of the broadening Ibu peaks, but was sufficient throughout the tested concentration range (R S > 2). Along these lines, long injection times also resulted in a decrease of resolution and increase of the peak height (Fig. 5 ).
The resolution decreased because of the broadening Ibu peaks, but was sufficient throughout the tested concentration range (R S > 2) The S-Ibu peak area increased linear with the tested concentration until 0.65 mg/mL, which may be indicative for beginning sample overloading at higher concentra-  tions. Thus, preserving a safety margin, a slightly decreased S-Ibu concentration of 0.6 mg/mL was used. The injection duration was varied from 3 to 10 s at a pressure of 3.45 kPa. Figure 5 shows a linear increase of the R-Ibu peak height from 4 to 10 s, but also a severe decrease of resolution at injection durations above 5 s. As a compromise the following injection parameters were chosen: a S-Ibu concentration of 0.6 mg/mL, which were injected for 5 sec at a pressure of 3.45 kPa.
An electropherogram of S-Ibu (0.6 mg/mL) containing 0.68% R-Ibu is shown in Fig. 6. Good separation of the enantiomers (R S ≥ 2.3) could be achieved within about 18 min with the minor enantiomer (R) migrating in front of (S). Quantification of R-Ibu was performed by area normalization using the migration time corrected area A corr (area divided through migration time of the peak). The percentage content of R-Ibu was determined using The sample of S-Ibu provided by Pen Tsao contained a small amount of the R-isomer (0.68%), which can be regarded as an advantage, because it is not necessary to achieve an LOQ below 0.1% for R-Ibu as generally recommended by the Ph.Eur. for impurities.

Method validation
The validation was performed according to the ICH Guideline Q2(R1) [49]. All measurements were conducted in randomized order.
The linearity was determined at five concentration levels ranging from 0.68 to 5.49% R-in S-Ibu (y = 160.09x + 0.0705). Each concentration level was measured in triplicate. The coefficient of determination was found to be (R 2 = 0.9991). Accuracy was evaluated by determining the recovery ranging from 97 to 103% (Table S3). The small amount of R-in S-Ibu (0.68%) was taken into account by subtracting the area of the R-Ibu peak from the area of the S-Ibu peak.
The LOQ of R-Ibu in the presence of S-Ibu was determined based on an S/N of 10 by plotting the percentage R-Ibu against the corresponding S/N. As already mentioned, the pure S-Ibu sample contained 0.68% R-Ibu; thus the LOQ was determined by extrapolation of the calibration curve (y = 108.59x -12.625; R 2 = 0.9984). The extrapolated LOQ was found to be 0.21% for R-Ibu, referring to a concentration of 1.25 μg/mL. A second approach for evaluating the LOQ of R-Ibu was performed by dilution of a solution of rac Ibu until an S/N of 10 was reached and determination of the corresponding concentration of R-Ibu. The results are summarized in Table S4. The slight difference in determined LOQs pointed out in percentage R-Ibu (0.21 and 0.15%, respectively) may be explained by the presence of massive excess of S-Ibu. The comparison of the LOQs of other CE methods for the evaluation of the enantiomeric purity of S-Ibu (1% [37] and 0.5% [39]), revealed our method to be superior with regard to sensitivity (0.21%).
Evaluation of precision comprises repeatability and intermediate precision. The repeatability was determined by the measurement of a solution of S-Ibu (0.6 mg/mL) containing 0.68% R-Ibu (n = 6): an RSD of 1.52% for the R-Ibu content was found (Table S5). The RSD of the migration times was 1.31% for the R-Ibu and 1.46% for the S-Ibu peak. For intermediate precision (consecutive measurements on 2 days), with freshly prepared sample and BGE stock solutions, the RSD of the percentage R-Ibu content was 1.97% and of the migration times for the R-Ibu and the S-Ibu peak 2.65% and 2.78%, respectively, indicating a sufficient precision of R-Ibu determination. Figures 7A and 7B show the CE results of isomerization after applying the milling schemes a-g. As can be seen, without the addition of KOH neither increasing the milling speed nor the reaction time leads to an isomerization of S-Ibu (Fig. 7A). However, a partial isomerization was observed when KOH was used. R-Ibu increased significantly when a combination of high milling frequency and extended milling time was applied. In particular, after milling for at least 30 h at 20 Hz milling frequency, a significant isomerization was observed (samples 43, 45, and 46, Fig. 7A). Only at a milling frequency of 25 Hz or higher an isomerization was already detected after 2 h of milling. To observe isomerization at a milling time of 20 min, a milling frequency of 35 Hz had to be applied (sample 39). Certain samples from milling schemes a and b were stored in solution at 4°C for 7 days to study further isomerization (samples 21,23,39,43,44,46). No increase in  Tables S1 and S2 percentage R-Ibu could be detected, indicating an isomeric stability of S-Ibu in solution for at least 7 days.

Measurement of stressed S-Ibu samples
Under acidic conditions, a slight isomerization was detected for sample 50 (0.80% R-Ibu, Fig. 7A). In contrast, sample 48 did not show any increase in the content of R-Ibu (0.65% R-Ibu) which indicates that the milling speed might be a crucial factor for isomerization for this milling scheme. No isomerization was observed in the case SiO 2 was used as the only grinding auxiliary (scheme d, Fig. 7A). For milling scheme e, water was applied as a liquid assisted grinding (LAG) agent [50] without further additives. As for schemes a and d, no isomerization could be ascertained, except for sample 57 (0.77% R-Ibu, Fig. 7B), where 0.1% molar equivalents of water were used. However, for sample 60, where the five-fold amount of water was employed, no isomerization was detected (0.68% R-Ibu). Process scheme f combines the usage of KOH and LAG agent. Here, only sample 72 shows minor isomerization of S-Ibu (0.72% R-Ibu, Fig. 7B). Sample 33 reveals a higher percentage of R-Ibu (0.80%) where no water was added. Consequently, the effect of water on isomerization of S-Ibu could be considered negligible. The conditions used for scheme g did not lead to isomerization (Fig. 7B).

Concluding remarks
The CE method development revealed the addition of small cations such as Mg 2+ or Zn 2+ to highly increase the separation of Ibu enantiomers. The developed and validated CE method is proven to be capable to reliably determine the enantiomeric purity of S-Ibu.
To our knowledge, the isomerization of S-Ibu stressed in a ball mill has not been described before. The results extend the knowledge of S-Ibu isomerization under mechanical stress and several comparatively mild stressing conditions. Measurement results indicate that significant isomerization of S-Ibu only occurs on milling in presence of KOH, which gets more intense when milling time and milling speed exceed particular values. However, the extent of isomerization is limited as shown by the small increase of percentage R-Ibu, which emphasizes the stereochemical stability of S-Ibu under the conditions tested. Since a slight isomerization was only observed under basic conditions, which do not occur in tablets, capsules, and ointments, formulations are most likely stable upon storage.