Introduction Since first being introduced in 1978 by Still et al. 1, Flash chromatography, a medium performance liquid chromatographic ( MPLC) purification technique, has evolved considerably, and has become a common tool for synthetic chemists in their everyday work. This is particularly true in modern industrial environments such as the pharmaceutical industry where tight timelines demand high productivity and quality. Flash- LC is a purification method of choice for large scale preparative separations and is used routinely on a laboratory scale ( g- kg) and pilot plant scale ( 10' s of kg). Advances made by instrument manufacturers have given access to reliable automated Flash- LC purification hardware for the laboratory and dependable pre- packed columns. Whilst most chemists are familiar with this technique, it is appropriate method development that remains the main hurdle in making the best use of it. Developing fit for purpose methods for routine laboratory work can be time consuming. Developing an optimised method to efficiently execute the more demanding applications of Flash- LC within the pharmaceutical industry, such as purification of materials to be used in non-clinical or clinical trials to tight purity requirements, can be particularly testing for those not skilled in the art. Flash- LC employs particles with dimensions on order of 50 µm which give relatively low column efficiencies ( N) compared with other LC techniques such as the ubiquitous high performance liquid chromatography ( HPLC). Column efficiency cannot be relied upon in order to achieve a meaningful preparative Flash- LC separation. Instead a high selectivity ( alpha) is required to achieve the resolution required to translate a Flash chromatographic method into pure product. The optimisation of selectivity is a key factor for successful method development for preparative applications. One very practical and effective means to affect selectivity for normal phase ( NP) Flash- LC is by careful solvent choice - selecting solvent systems composed from a combination of solvents from the eight NP solvent selectivity families ( based on Snyder's et al. 2and Glajch et al. 3 solvent selectivity descriptors). For reversed phase ( RP) Flash- LC the selectivity may be most affected by organic modifier choice or pH. Conventionally the method development process involves screening solvent combinations using TLC to determine the best solvent( s) and relative proportions in order to achieve the required separation. However, in order to achieve the required selectivity it is often necessary to use binary, ternary, or sometimes even quaternary solvent systems composed of a combination of solvents from the eight NP solvent selectivity families. Hence, even a routine method can involve a vast array of solvent combinations and proportions. Once the selectivity effects are known, other factors such as the chromatographic band shape, solubility, stability, solvent cost and environmental issues can be considered. The method development process can be arduous and the consistency and quality of the optimised method is dependant on the skill set of the individual. The key to enabling facile, rapid and successful Flash- LC purifications is to be able to quickly identify the best conditions to maximise selectivity for a given separation problem using minimal effort and achieving consistent results independent of a chemist's experience with the technique. The aim was to deliver method development systems to rapidly and easily develop a method for normal or reverse phase Flash chromatography. The method screening systems would be implemented into intuitive and integrated purification workflows in order to encourage and enable the use of robust and successful purifications. Experimental All solvents used in this work are HPLC grade from Sigma- Aldrich ( Gillingham, UK). The screening instruments were Agilent ( Stockport, UK) 1100 HPLC systems ( quaternary pumps) equipped with a solvent selection valve to enable screening of more solvents on the normal phase system. Detection was performed with a diode array UV/ Vis detector. The evaporative light scattering detector used with the normal phase screening system was an ELS 1000 from PolymerLabs ( Church Stretton, UK). The instruments were controlled using Chemstation Rev. B. 03.01 and Easy- Access Rev. A. 05.01 software. All Flash chromatography experiments were performed on a Biotage ( Uppsala, Sweden) Isolera One equipped with a variable wavelength UV detector and two collections beds. Enabling facile, rapid and successful chromatographic Flash purification Stéphane Dubant and Ben Mathews* Chemical Research & Development, Pfizer Global Research & Development, Sandwich, Kent, CT13 9NJ, UK. * Corresponding Author Flash liquid chromatography ( Flash- LC) is an established technique that can have a very positive impact on productivity in the pharmaceutical research and development chemistry laboratory - provided it is used efficiently. Finding the optimal conditions for a normal phase or reverse phase Flash- LC separation can be a laborious process. The aim of this work was to develop a screening system to enable automated, rapid and reliable method development for preparative Flash chromatography. Implementation of these method screening systems into the chemists workflow has had beneficial productivity and quality impacts at different stages of drug development, and on scales ranging from laboratory to pilot plant. Key words: Flash chromatography, automated method development, normal phase, reverse phase. 10November/ December 2009

11 Sample A: This compound was a mixture of an intermediate from a drug substance synthetic route and contained one major desired product and three minor impurities that required purging. The sample was prepared by dissolving 5 mg in 1 mL of dichloromethane, and then analysed using the normal phase screening system. Normal phase conditions: Column: Luna Silica ( 2) 50 x 3.0 mm, 3 micron Mobile Phase: See Table 1 Flow rate: 4 mL/ min ( 3.5 mL/ min if Toluene is used) Temperature: Ambient Equilibration: see Table 2 Overall screening run time ( 6 methods): 45 min Reverse phase conditions: Column: Biotage C18 250 x 4.6 mm, 60 micron Mobile Phase: See Table 3 Flow rate: 3 mL/ min (~ 1 column volume/ min, based on Biotage data) Equilibration: 5 min equilibration at starting conditions Temperature: Ambient Eluents: A= Water/ TFA ( 0.1 % vol) or Water/ Ammonium Acetate ( 10 mM) B= Acetonitrile or Methanol Overall screening run time ( 4 methods): ~ 1.5 hours Results and Discussion Automated Normal Phase Flash method screening An automated normal phase method screening system was built using standard HPLC instrumentation. Glajch et al. 3demonstrated that solvents can be categorized using three descriptors ( non- localized, base localized and dipole localized). These descriptors led to the classification of solvents for normal phase chromatography into eight selectivity families ( Synder et al2). A comprehensive method development screen should ideally include a solvent from each of the eight selectivity families. However, some solvents are unfavourable in practice for reasons relating to safety, the environment or silica incompatibility. The method screening system includes solvents from several of the selectivity families ( i. e. I, II, V, Via, VIb and VII) and performs a gradient with a choice of either heptane or toluene as the weak eluent ( Tables 1 and 2). For basic analytes, the same methods can be run with basic additive conditions ( using a dedicated column) by selecting the appropriate method in the instrument software. The diverse selectivity that can be obtained using different normal phase solvent systems is illustrated in Figure 1 using a typical reaction product ( Sample A) from our laboratories. Sample A is an intermediate in an exploratory drug substance synthetic route and is comprised of four components, three of which are undesired impurities ( designated 1, 2 and 3) that required purging. Significant selectivity differences can be observed between different solvents systems, leading to beneficial changes in resolution and even to differences in elution order ( Figure 1). This approach enables rapid determination of the best solvent choice for optimal selectivity as well as facile Solvent A Solvent B Snyder Strength ° " weak solvent"" strong solvent" Selectivity group( silica) Heptane0 TolueneVII0.22 EthanolII0.65 Ethyl AcetateVIa0.36 tBMEI0.32 AcetoneVIa0.53 DCMV0.3 AcetonitrileVIb0.52 Table 1: A list of the solvents used in the normal phase Flash screen. For use with basic analytes, additional methods can be selected that incorporate the base diethylamine with the other solvents. Time ( min)% solvent A% solvent B% Ethanol 09550 0.49550 2.301000 2.701000 Table 2: Gradients used for the normal phase Flash screen: on the right, the column conditioning and equilibration gradient; on the left the gradient method used to generate the data. Time ( min)% solvent A% solvent B 00100 0.20100 0.3955 Table 3: Reverse phase scouting gradient on a scaling Flash column. Since one minute approximately equals one column volume on the analytical scale, the same gradient method can easily be scaled to any preparative Flash column based on column volumes. Figure 1: Example chromatograms showing the selectivity obtained using different normal phase solvent systems on silica for Sample A - a four component mixture containing impurities 1, 2 and 3. Time ( min) or % eluent A% eluent B ~ column volume 0955 2955 17595 2055 Figure 2: An example of how to estimate the isocratic Flash- LC solvent composition by relating the gradient retention time of any particular analyte from it chromatogram to the gradient profile. The report is engineered so that the estimated solvent composition that it predicts results in that analyte having a retention factor under those isocratic conditions that is suitable for a preparative purification process.