Introduction
Pharmaceutical drug discovery depends increasingly on speed and efficiency. Currently there is an ever increasing drive to decrease turnaround times for samples. Over the past six years, SFC has been adopted as the instrumentation of choice for the purification of chiral pharmaceutical compounds based on savings in production time and cost [1]. Many researchers have worked on developing an in-house mass directed SFC for achiral purifications [2]. The commercial availability of a mass directed supercritical fluid instrument has now made the purification of achiral pharmaceutical compounds a real possibility. The use of the MS-directed collection of the peak of interest eliminates some of the other steps previously needed to insure collection of the peak of interest using UV detection [3]. The AstraZeneca drug discovery laboratory has now turned to SFC for the purification of both chiral and achiral compounds.
Traditionally, mass directed HPLC has been the method of choice for the purification of many achiral pharmaceutical compounds. However, SFC/MS offers several advantages over the use of LC/MS. SFC/MS eliminates the need for extended dry-down times. LC/MS fractions typically require an overnight dry down, however SFC/MS fractions of methanol can be dry in two hours. The elimination of a wait time for the overnight dry down of fractions is compelling. The crisis with the availability and price of acetonitrile during the past year is another reason for using SFC/MS for achiral purification. SFC/MS fractions are usually smaller as the main component, CO2, is no longer present, adding to the decreased dry down time. The advent of more selective achiral columns has precluded the use of multicolumn approaches to separating achiral compounds [4]. For some compounds, the different selectivity of the columns used in SFC/MS can eliminate the tailing or fronting of a peak on LC/MS and be replaced by sharp peaks on SFC/MS [5]. In testing the system we selected several achiral columns for screening and then used several for the purification of test compounds. These test compounds are typical of the basic compounds we routinely see.
SFC/MS Purification System Specifications
Compared to preparative LC/MS, the hardware components to enable preparative SFC/MS are initially more daunting. However, with use and familiarity, the primary maintenance activity is filling of methanol for the pumps. The preparative SFC/MS hardware specifications for the purification of samples examined here were as follows:
• Pumps: CO2 pump, modifier pump, make-up pump for collection, dilution pump for MS which can help with ionization in ES mode, pump for diluting the stream going to the splitter
• Detectors: UV and MS
• Flow splitter
• Gas/liquid separator
• Column oven
• Open bed collection system
• Two heat exchangers
• A back pressure regulator
• A degasser
The system has the capability of having five different split ratios; very low, low, medium, high, and very high. Since the current focus is on developing a system for routine use over a wide range of compounds, the medium splitter setting is being used. As we gain familiarity with the range of compounds that require purification, the different splitter settings may prove useful for both those compounds that ionize very well and those that do not.
Currently the system is optimized for a flow rate of 100 ml/min. It would be an improvement to have the system optimized for a variety of flow rates, particularly lower flow rates. The flow rate of 100 ml/min is not optimal for either 21mm or 30 mm id. columns. Currently we are only using 30 mm id. columns, however the ability to use smaller id. columns would save significant money.
For the separations shown in this paper, we used the SFC/MS system with some modifications as follows:
• The flow rate for the system was 100ml/min.
• The columns were 30 x 150 mm with a variety of stationary phases.
• The system initially arrived with an ESI source; however it was switched to an APCI source to match the analytical source currently in use for the screening of samples.
• The initial setup for the instrument was with a 1ml syringe and 1 ml sample loop which was changed to a 2.5 ml syringe and a 2ml sample loop.
• The DAD wavelength was 220 nm.
• The oven temperature was 25 degrees C.
• The splitter was set to medium.
Methanol was used as the modifier with 0.5% dimethyl ethylamine as the additive. A gradient from 5 to 40% modifier over 8 minutes was used. There were two compounds in each sample with a total weight of approximately 50 mgs. The samples were diluted with methanol or a mixture of methanol and DMSO to dissolve the sample. The amount of each compound in the mixture varied, so at times the setting of the threshold for collection required modification. Each mixture was purified in triplicate.
Screening Samples
Column selection: There are several types of achiral stationary phases currently available for the separation of achiral compounds. Since our compounds are primarily basic, we chose to prescreen using the following stationary phases: pyridine, DEAP, amino, nitro and cyano. From the results we selected three columns, DEAP, pyridine and amino for the purification of the test compounds. These columns showed the best resolution of the two-component mixture. None of the columns worked for all of the mixtures. The use of two compounds in each sample highlighted which column gave the best separation for each set and the differences in selectivity for the various columns.
As the SFC/MS system is incorporated into our purification process, this series of five columns will be used to screen achiral compounds. Over time and as applicable, we anticipate a pattern will emerge to define which column works best with specific compounds and projects. This knowledge base would decrease the ongoing need to screen using all five columns.
Retention times: The gradient for both the analytical prescreen and the purification runs were the same and the flow rate was set to 2.35ml/minute to scale to 100ml/minute flow rate on the preparative instrument. As shown in figure 1, the prescreen chromatogram, the bottom three traces, and the purification chromatogram, the top three traces, have fairly similar retention times, generally between +/- 0.32 to 0.59 minutes. There were differences up to 1.19 minutes between the analytical retention time and the preparative retention time. The order of elution for both the analytical and the preparative chromatograms was always the same. The analytical and the preparative phases are not from the same lots and the analytical phases were in use for a longer period of time. This may account for some of the retention time differences and this is subject to further investigation.
Testing the System: Comparison of SFC/MS and LC/MS:
These compounds were separated using both SFC/MS and LC/MS. With SFC/MS the APCI source was used, while the LC/MS used the ESI source. Using the APCI source in the SFC/MS eliminated the formation of adducts that are typically seen with the ESI source.
The LC/MS column for all of the test compounds was a reverse phase column using a gradient under basic conditions and ethanol for the separation. On the LC/MS system we had switched from acetonitrile to ethanol with the difficulties of availability and cost for acetonitrile. The SFC/MS column for the test compound to be discussed below was a pyridine phase using a methanol/CO2 gradient. For two of the test compounds, there was significant fronting of the peak in the preparative LC/MS. None of the test compounds showed fronting or tailing in the preparative SFC/MS. An example of this is shown in Figure 2, where the LC/MS preparative chromatogram shows the first peak fronting and multiple fractions are collected. The SFC/MS preparative chromatogram is shown in Figure 3. The two compounds form sharp peaks without fronting. Two of the mixtures had reversal of peak elution order using SFC/MS compared to the elution order on LC/ MS. This would be expected given the differences in the mechanisms of separation between SFC and LC.
The reproducibility of the retention times for the peaks was very good for preparative SFC/MS, with a typical variation in retention time of +/- 0.03minutes. The LC/MS also had good reproducibility of retention times. Generally the compounds were soluble in methanol; however two of the five samples required the addition of a small amount of dimethyl sulfoxide (DMSO) to completely dissolve the sample. For separation of the test compounds on the LC/MS, DMSO is the solvent of choice for dissolving the sample. In one case, the first peak was poorly retained on the SFC/MS and eluted with the DMSO peak. The SFC/MS collected the peak of interest; however it was a broad peak containing DMSO. This was not an issue for this compound on the LC/MS. The test mixtures for both the LC/MS and the SFC/MS contained various concentrations of the two compounds. The varying concentrations and differences in ionization of the compounds created challenges for collection in both the LC/MS and the SFC/MS. The use of peak collection within a time window helped for compounds with low ionization when also collecting a good ionizing compound. This capability is available on both SFC/MS and LC/MS. Running an initial test injection on both SFC/MS and LC/MS of 100ul helped to sort out the threshold for collection and the need for using time windows for collection.
Once the SFC/MS and LC/MS fractions were collected, the fractions were dried down. The dried fractions were pooled and placed into tared vials. The vials were dried again and the results compared with the amount originally placed in the vial. The SFC/MS recovery results were consistently between 75-91%, with two exceptions. The two low recovery peaks that had poor ionization and were collected using threshold, had recoveries around 70%. Thus, proving the need to used time windows for collection. For one compound in particular, there was high recovery, between 89-99%. The recovery data for LC/MS was comparable with recovery results between 72-99%, with similar issues with the poorly ionizing compounds and the high recovery on one particular compound.
Integration into Purification Process
Once the testing was completed with all the data supporting good chromatography with good recovery, the SFC/MS system was ready for integration into the work flow. Two similar samples which were difficult to purify by LC/MS were chosen for the first test. These samples were quite polar. Both compounds had retention times on the SFC/MS analytical screen of approximately 7.5 minutes on an 8 minute gradient with the impurities eluting earlier and well separated. The preparative analysis showed similar pattern in chromatography, resulting in an easy removal of the impurities from both compounds. The initial LC/ MS screen of these two compounds indicated that the separation of the peak of interest from the impurities would be difficult. This guided the selection of SFC/MS for the purification. We have determined that the SFC/MS system can be used for chiral purifications in cases where there is very little UV signal. Such a test of the system was successfully made with a sample that needed to be collected using MS signal since the other SFC systems for chiral purifications relied on the UV signal for collection of the compound. Previously, we would have required the chemist to use derivatization of the compound in order to obtain an adequate UV signal for purification.
Conclusions
The SFC/MS was successfully tested and compared with the LC/ MS system. The results showed comparable recoveries of the test compounds. The SFC/MS showed some chromatographic advantages for some compounds with less peak fronting compared with LC/MS. As the preparative SFC/MS is incorporated into the work flow, we will look for the preparative system that allows for good chromatographic separation of the peak of interest from the impurities.
The SFC/MS had considerable time savings for obtaining final compound in a vial when compared to LC/MS. The LC/MS had larger fractions containing water which required overnight dry down. The SFC/MS fractions were smaller and required only a two hour dry down. The use of SFC/MS for the purification of achiral compounds will help us to achieve our goal of faster turnaround times.
References
1.Taylor, L.T.”Supercritical Fluid Chromatography” (1. 2008) Anal. Chem. 80 pp.4285-4294.
2. Zhang, X., Towle, M.H., Felice, C.E., Flament, J.H., and Goetzinger, W.K., “Development of a Mass-Directed Preparative Supercritical Fluid Chromatography Purification System” (2006) J. Comb.Chem. 8, 705-714.
3. White, C and Burnett, J., “Integration of Supercritical Fluid Chromatography into Drug Discovery as a Routine Support Tool: II. Investigation and Evaluation of Supercritical Fluid Chromatography for achiral Batch Purification” (2005) Journal Chrom. A, 1074, pp.175-185.
4. Phinney, K.W., Sander, L.C., and Wise, S.A. “Coupled Achiral/Chiral Column Techniques in Subcritical Fluid Chromatography for the Separation of Chiral and Nonchiral Compounds” (1998) Anal.Chem,70 pp.2331-2335.
5. Yan, T.Q., Bradow, J., Chang, S.P., Depoanta, R., and Phillippe, L. “Approaches to Singleton Achiral; Purification of Difficult Samples for Discovery Research Support” (2009) LCGC North America April.
Author Biography
Jennifer Van Anda is a Senior Scientist at AstraZeneca Pharmaceuticals in Wilmington, DE where she leads a purification group in the Chemistry Department. Previously, she was a Marketing Chemist with Mettler-Toledo, where she was responsible for the analysis of customer samples and the demonstration of the Supercritical Fluid Chromatographic product portfolio. She has also worked at Agilent Technologies, in a variety of positions. Jennifer Van Anda received her doctorate degree from the University of North Carolina, Chapel Hill in Pharmacology.