Generic Capillary Electrochromatographic Screening and Optimization Strategies for Chiral Method Development

Drug molecules are often chiral, which means that they possess a chiral centre that causes two or more non-superimposable geometric forms that are mirror images of each other to occur. These forms are called enantiomers and different pharmacological activities are often displayed by the enantiomers in chiral environments, such as living systems.^1,2 Because of these different actions, several regulations are followed by the pharmaceutical industry when developing a chiral drug substance.^3,4 We will not go into detail on these regulations, but briefly it can be stated that methods have to be developed that can separate and quantify all enantiomers of a chiral molecule.

In an early stage of drug development, racemates, (i.e., mixtures of enantiomers), are usually synthesized because of the high costs of enantiopure reagents to synthesize enantiopure drugs. Afterwards, these mixtures are separated into their individual enantiomers to investigate the possible activities.

Thus chiral method development is needed in early drug development, as well as in the registration procedures. In both situations, generic separation strategies, applicable on large sets of structurally diverse substances, can be very useful to speed up method development.

In this article, capillary electrochromatography (CEC), a hybrid between high performance liquid chromatography (HPLC) and capillary electrophoresis (CE), is used as an analytical separation technique. CEC is characterized by highly efficient separations, a low sample and solvent consumption during analysis, and it allows similar sample concentration ranges in HPLC to be used.^5,6 To perform separations, an electrical field is applied over a capillary filled with stationary phase, resulting in an electro-osmotic flow (EOF), which is the driving force of the mobile phase. In analogy with earlier defined strategies in HPLC,^7–10 CE^11–13 and SFC,¹⁴ CEC was considered to develop generic separation strategies and methods for chiral method development. The strategies have to be generic, meaning applicable on large sets of structurally diverse molecules. They should also provide information concerning the enantioselectivity of the system towards a given compound in a relatively limited number of experiments. Finally, baseline separation should be achieved for most compounds employing the strategy.

Initial Evaluation of Chiral Separation Possibilities in CEC

To perform chiral separations, two approaches can be used; direct and indirect. The latter is mostly avoided in industry because a derivatization is required, which implies that the enantiomers are modified on the one hand and that the procedure takes time on the other. Therefore, only the direct methods, where a chiral selector is used for enantioselective recognition, were considered for the development of strategies. These latter methods can be sub-divided into two types: one where the chiral selector is added to the mobile phase and one where it is bonded onto a chromatographic support.

In earlier experiments in CE,^11,12 it was seen that the addition of highly-sulphated cyclodextrins (HSCDs) as a chiral selector to the background electrolyte allowed defining generic strategies. This approach was also potentially applicable in CEC, using an achiral stationary phase. To reduce costs, because HSCDs are quite expensive, we started with a derivative of β-cyclodextrin, hydroxypropyl-β-cyclodextrin, as chiral selector for the chiral separation of chlorthalidone,¹⁵ in analogy with reference 16. The separation was optimized with an experimental design approach resulting in a considerably reduced analysis time compared with reference 16 while the baseline separation was maintained.

However, this system appeared to be highly selective, as it was not successfully used for other molecules. Highly-sulphated cyclodextrins were then briefly tested to obtain a broader enantiorecognition range but without satisfying results. It was also seen that the selector consumption was higher than expected because of the renewal of vials after each analysis caused by buffer depletion and the necessity to rinse the columns with mobile phase containing cyclodextrin to load the selectors onto the column. These three criteria made us conclude that this approach was not attractive for our goals. The use of chiral stationary phases (CSP) where the selector is coated or bonded onto the chromatographic support, appeared more promising.

Preliminary Studies on Polysaccharide CSP

The finally selected CSP were all from the polysaccharide type because of their broad enantiorecognition range and because of earlier experience in HPLC and SFC.^7–10,14 Our first CEC study involved Chiralcel OD-H and Chiralpak AD-H stationary phases, both normal-phase HPLC materials but used in reversed-phase (RP) CEC mode (since RP-phases were not available at that time).¹⁷ An EOF could be generated and bad separations of basic compounds could largely be improved by the addition of small amounts of hexylamine to the mobile phase. Later on, it was found that the addition of amines positively influenced the separation caused by the increase in pH, allowing the substances to be uncharged. Thus, the pH of the mobile phase was an important parameter to obtain chiral separations on these stationary phases. This was to be expected, as neutral chiral selectors are present and no or little interaction with charged species was seen in earlier experiments (Chiral Technologies, unpublished results).

Neither elution nor separation of the acidic compounds was observed because of the high pH used. Using a low pH on Chiralcel OD-H did not allow eluting thiourea within 100 minutes (see further). It became obvious that to define separation strategies a distinction had to be made between acidic and basic compounds. The influence of other parameters was evaluated as well. Generally, it was seen that acetonitrile (ACN) performed best as organic modifier and that an increased fraction reduced analysis times but also resolution. An increased voltage during analysis resulted in a similar influence. The addition of an ion pairing reagent, such as sodium perchlorate, as in RPLC,⁸ was not possible in concentrations above 100 mM because of the high organic modifier content needed to have acceptable analysis times.

Therefore, this reagent was probably not at its optimal concentration to form ion pairs with positively charged basic compounds and to allow their separation at lower pH values. As a consequence, it was decided that for basic compounds a pH of 11.5 would be used. A preliminary screening mobile phase consisting of a 5 mM phosphate buffer pH 11.5 mixed with 70% of ACN and containing 0.15% (v/v) hexylamine was defined. Using this mobile phase, applying 15 kV at a temperature of 20 °C, 21 out of 29 compounds showed enantioselectivity on at least one stationary phase, while 15 of them could be baseline separated. When the voltage was reduced to 5 kV, four extra compounds were partially separated. It was also seen that AD-H exhibited broader enantioselectivity than OD-H for the 24 basic compounds. For the three bifunctional species, OD-H showed a better recognition. For neutral compounds, no conclusions were drawn at this stage, as only two compounds were considered.

In a next study, the intention was to study the separation of acidic compounds at low pH. However, it was seen that at these conditions on Chiralcel OD-H a reversal of EOF occurred, which was the reason why thiourea was not detected at low pH in reference 17. This reversal was not seen with the reversed-phase material, Chiralcel OD-RH. Therefore, differences between normal- and reversed-phase materials were investigated first. Both materials were tested for the phases Chiralcel OD, Chiralcel OJ, Chiralpak AD and Chiralpak AS by means of a small test set of six compounds. Simultaneously, different electrolytes were tested for their suitability in later to be defined strategies. Two low pH (pH 3) electrolytes and three high pH (pH 11 or 11.5) electrolytes were tested. For acidic compounds, an ammoniumformate electrolyte performed best, as current drops were frequently observed with a low pH phosphate buffer. For basic compounds, no difference was seen using either an ammonium bicarbonate-, a borate- or a phosphate buffer, but the latter was chosen given earlier experience in reference 17.

Differences between normal- and reversed-phase materials mostly occurred at low pH. The largest difference was seen between OD-RH and OD-H, because of the reversal of EOF. This difference is probably a result of the use of different types of silica for these two phases, which for confidentiality reasons is not confirmed nor denied by the manufacturer. No other phases showed a reversal of EOF at low pH. Small differences were also seen on the level of enantioselectivity. For example, acenocoumarol could be partially separated on an OJ-RH phase, while no enantioselectivity was seen on the normal-phase counterpart. Faster elution was generally observed at low pH on Chiralpak AS-RH with respect to AS-H. Because it is best to select only one type of stationary phase to develop generic strategies, the reversed-phase material was preferred to the normal-phase one in reversed-phase CEC mode. Another conclusion from this study is that bifunctional and neutral compounds are best treated as the basic compounds: no advantages were seen when analysed at low pH, only a longer analysis time was observed. Finally, the addition of amine at high pH mobile phases was not found necessary to obtain good separations and, therefore, omitted in further studies.

Definition of the Separation Strategy

The use of experimental designs allowed the effect of several parameters on the separation to be studied. The studied parameters for acidic compounds were the content of ACN, pH, buffer concentration, operating temperature and applied voltage.¹⁹ For the non-acidic compounds (basic, bifunctional and neutral), most parameters were already studied in reference 17 and it was assumed that their effects also occur for the reversed-phase material and for the non-investigated stationary phases; Chiralcel OJ-RH and Chiralpak AS-RH. Additionally, the effects of buffer concentration and temperature were studied for the latter compounds in reference 20.

It was also seen that most observations in HPLC could be extrapolated to the CEC separations.^8,9 The content of organic modifier determines the elution strength of the mobile phase and the generated EOF. The applied voltage and temperature allow the analysis to be slowed down or speeded up, though the effect of the temperature was not always obvious for non-acidic compounds. The pH is best as low as possible for acids because they are more charged at higher pH, resulting in less interaction with the chiral selectors; for non-acidic compounds the pH was set at 11.5. For the buffer concentration, it was seen for acids that the highest resolutions were obtained at high concentrations, without a significant decrease of the EOF because of double layer overlap. For non-acidic compounds, no general conclusions could be drawn regarding the effect of the buffer concentration. Based on the results from references 17, 19 and 20, screening conditions for acids were defined as a 45 mM ammoniumformate pH 2.9/ACN (35/65) mobile phase, an applied voltage of 15 kV and a temperature of 25 °C. For non-acidic compounds these conditions were 5 mM phosphate buffer pH 11.5/ACN (30/70) with the same applied voltage and temperature as for acids. It must be noted that the applied voltages used in the strategy were determined on a 20 cm packed column with a total length of 31.2 cm. For shorter or longer columns, the applied voltage must be adapted accordingly.

The column testing order in the screening experiments was determined by analysing all compounds of a test set (n=63) on the four columns. For acids, it was seen that OJ-RH and AD-RH exhibited most enantioselectivity and gave complementary results. Therefore, it is recommended to screen them simultaneously in the first instance, occasionally followed by an experiment on AS-RH and to test OD-RH in the last instance. For basic, bifunctional and neutral compounds, most and complementary enantioselectivity was seen on AD-RH and OD-RH so they are to be screened first and simultaneously. The next phase to screen is OJ-RH, followed by AS-RH.

Depending on the results at screening conditions different actions are proposed. The general structure of the strategy is given in Figure 1, including the steps taken when no separation is seen in the screening phase. In the instance of a partial separation at screening conditions (0 < Rs < 1.5), the resolution can be enhanced in optimization level 1. When a baseline separation is obtained, the analysis time can be reduced in optimization level 2. The next paragraphs describe these two optimization levels. Where no separation is seen (Rs = 0), the modifier content and occasionally voltage are reduced, prior to switching to another technique.

Optimization Level 1

This section explains the experiments proposed when a partial separation occurred at screening conditions or after application of the adapted conditions when initially the Rs was zero (Figure 1). For more details, we refer to references 19 and 20.

Figure 1: General structure of the strategy.

For acids, a two-level full factorial design for two factors is performed (Table 1). Acetonitrile content and temperature are decreased compared to screening conditions. Then, the best result is selected. If still no baseline separation is obtained but the analysis time is less than 40 minutes, one can reduce the applied voltage. If then again no baseline separation is achieved, it is recommended to switch to another technique since similar strategies are defined in CE,^11,12 RPLC,^8,9 NPLC,^7,9 SFC,⁴ and POSC.¹⁰ When a Rs > 1.5 is obtained, the analyst can optimize the analysis time in optimization level 2, of course with the risk of losing the baseline separation.

Table 1: Experiments to perform according to the 22 full factorial designs in optimization levels 1 and 2.

For non-acidic compounds, again ACN content and applied voltage are varied (Table 1). After selection of the best result and if no baseline resolution is observed, the temperature can be decreased to 15 °C, though the outcome of a temperature decrease will not always be positive on the Rs because of increased peak broadening at longer analysis times.²⁰ When a baseline separation is achieved, the same criteria and remarks apply as for the acids.

Figure 2: Separation of fenoprofen on Chiralcel OJ-RH at (a) screening conditions, and (b) after application of optimization 1. Conditions: (a) see Figure 1, (b) with 60% ACN at 20 °C.

Some electrochromatograms comparing the separations seen at screening conditions and after application of optimization level 1 are given in Figures 2 and 3.

Figure 3: Separation of toliprolol on a Chiralpak AD-RH stationary phase. (a) screening experiment, (b) after application of optimization 1. Conditions: (a) see Figure 1, (b) with 65% ACN at 10 kV.

Optimization Level 2

The goal of this level is to reduce the analysis time, but still maintaining the baseline resolution earlier obtained. It is therefore not recommended to perform this step when a resolution close to 1.5 is obtained. This level is equal for both acidic and non-acidic molecules.^19,20 First, four experiments are performed according to a full factorial design in which the factors voltage and temperature are varied (Table 1). If the best result is still not satisfying, the ACN content of the mobile phase can be increased. Separations obtained at screening conditions and after application of this optimization level can be found in Figures 4 and 5.

Figure 4: Separation of coumachlor on Chiralpak AD-RH at (a) screening conditions and (b) after optimization of the analysis time using optimization 2. Conditions: (a) see Figure 1, (b) at 35 °C and 25 kV.

Strategy Evaluation

The applicability of the strategy was evaluated by means of test sets. Fifteen acidic and 48 non-acidic compounds were selected.^19,20 The difference in numbers comes from the fact that most drug molecules are in the latter category.

Figure 5: Separation of meberevine on Chiralcel OD-RH. (a) result of the screening experiment, (b) best result of optimization level 2. Conditions: (a) see Figure 1, (b) at 35 °C and 25 kV.

For the acids, five compounds were baseline-, six partially and four not separated after application of the screening step. This means that enantioselectivity is seen for most compounds. After application of optimization level 1, an extra five compounds were brought to a baseline separation, bringing the total of baseline separated compounds to ten. One compound remained partially separated, and for the four initially co-eluting compounds, no additional selectivity was seen. Optimization level 2 allowed a significant reduction in analysis time for all compounds where a baseline separation was achieved at the screening step.

For the non-acidic compounds, 17 compounds were not separated at the screening experiment. This number could be reduced to seven after application of the Rs = 0 step, but no baseline separation in a reasonable time could be achieved using optimization level 1 for these latter compounds. At screening conditions, 15 compounds partially separated, of which five achieved a baseline resolution after optimization level 1, and ten remained partially separated. Thus, the number of baseline separated compounds increased from 16 to 21 after optimization 1.

The number of baseline resolutions achieved in optimization 1 is not spectacularly high because in the optimization level no profound optimization is performed but only a fast evaluation on separation improvement. The strategy aims at evaluating several compounds for enantioselectivity rather than maximally optimizing the separation of individual compounds. Therefore, the strategy suggests using another separation technique if after optimization 1 still no baseline separation is achieved. For optimization 2, the same observations as for acids apply (i.e., an acceptable reduction of the analysis time was observed for all compounds that were baseline separated at screening conditions.)

The best results for the test compounds after application of the proposed strategy are given in Table 2. In the entire test set, enantioselectivity was seen for 82.5% of the compounds, and for 49.5% a baseline separation was obtained.

Table 2: Best results for the test compounds. Results extracted from references 19 and 20.

Conclusions

The definition of generic separation strategies for chiral method development appears to be feasible using CEC as a separation technique. Two strategies were defined, which showed their generic character towards a large set of compounds, as 82.5% was partially or baseline separated. Fast information on the enantioselectivity was obtained because 66.7% of the compounds were already separated at screening conditions. For the final requirement of a generic strategy, baseline resolution for the majority of compounds, the approach performs less, because only 49.2% of the substances could be baseline resolved. However, with some extra optimization on individual compounds, this number can certainly be increased, but this approach was outside the scope of our study.

Table 2: Continued.

For the future perspectives, it is worthwhile mentioning that an instrument is commercially available which allows CE, CEC, pressurized CEC and capillary liquid chromatographic separations. This should allow differences in enantioselectivity between those techniques to be evaluated rather quickly. Loop injections and high pressurization during analysis to work with submicronsized particles are also possible. The use of monolithic columns also seems to be a challenge in the field of chiral CEC separations, as they eliminate the need for frits and their consequent problems such as bubble formation and column fragility.^21,22 The presented work and the future perspectives are an indication that CEC is a technique that can give an extra dimension to those currently available.

Acknowledgements

The authors thank Peter J. Schoenmakers, Wim Th. Kok, Sebastiaan Eeltink and Christina Suteu for their contributions.

Debby Mangelings is a PhD student at the Department of Analytical Chemistry and Pharmaceutical Technology. Her main research area focuses on the evaluation of capillary electrochromatography as a separation technique and the development/evaluation of new stationary phases to be used with this technique. CEC columns can be packed upon request.

Mohamed Maftouh is currently head of the Biology Analytics Activity at the Discovery Analytics Department of Sanofi-Aventis Research. His main research area focuses on small molecule biomarkers method development (profiling, identification and quantification), biophysical methods for identification of protein-ligand interactions and micro-nanobiotechnology. Formerly, he managed a separation sciences group for chemical synthesis follow up.

Desire Luc Massart is an emeritus professor at the Vrije Universiteit Brussel, Belgium and performs research on chemometrics in process analysis and its use in the detection of counterfeiting products or illegal manufacturing processes.

Yvan Vander Heyden is a professor at the same university and heads a research group on chemometrics and separation science.

References

1. D.J. Cordato, L.E. Mather and G.K. Herkes, Journal of Clinical Neuroscience, 10, 649–654 (2003).

2. K.M. Rentsch, J. Biochem. Biophys. Methods, 54, 1–9 (2002).

3. http://www.ich.org

4. FDA's policy statement for the development of new stereisomeric drugs, US Food and Drug Administration (1992) (on-line): http://www.fda.gov/cder/guidance/stereo.htm

5. I.S. Krull et al., Capillary Electrochromatography and Pressurized Flow Capillary Electrochromatography: An Introduction, HNB Publishing, New York (2000).

6. K.D. Bartle and P. Myers, J. Chromatogr. A, 916, 3–23 (2001).

7. C. Perrin et al., J. Chromatogr. A, 947, 69–83 (2002).

8. C. Perrin et al., J. Chromatogr. A, 966, 119–134 (2002).

9. N. Matthijs et al., J. Chromatogr. A, 1041, 119–133 (2004).

10. N. Matthijs, M. Maftouh and Y. Vander Heyden, Screening approach for chiral separation of pharmaceuticals, Part IV, Polar Organic Solvent Chromatography, submitted.

11. N. Matthijs et al., J. Pharm. Biomed. An., 27, 515–529 (2002).

12. N. Matthijs et al., Anal. Chim. Acta, 525, 247–263 (2004).

13. N. Matthijs, M. Maftouh and Y. Vander Heyden, Enantiomeric impurity determination using a highly-sulfated cyclodextrins based screening method in capillary electrophoresis, in press (Biomedical Chromatography).

14. M. Maftouh et al., J. Chromatogr. A, 1088, 67–81 (2005).

15. D. Mangelings et al., An. Chim. Acta, 509, 11–19 (2004).

16. F. Lelièvre et al., J. Chromatogr. A, 723, 145–156 (1996).

17. D. Mangelings et al., Electrophoresis, 24, 2567–2576 (2003).

18. D. Mangelings et al., Electrophoresis, 25, 2808–2816 (2004).

19. D. Mangelings et al., Electrophoresis, 26, 818–832 (2005).

20. D. Mangelings et al., Electrophoresis, 26, 3930–3941 (2005).

21. D. Mangelings, M. Maftouh and Y. Vander Heyden, J. Sep. Sci., 28, 691–709 (2005).

22. F. Svec, J. Sep. Sci., 28, 729–745 (2005).