Anal. Chem. 1999, 71, 2410-2416
Packed Column Supercritical Fluid Chromatography/Mass Spectrometry for High-Throughput Analysis Manuel C. Ventura, William P. Farrell, Christine M. Aurigemma, and Michael J. Greig*
Alanex Corporation, 3550 General Atomics Court, San Diego, California 92121-1194
A supercritical fluid chromatograph was interfaced to a mass spectrometer, and the system was evaluated for applications requiring high sample throughput. Experiments presented demonstrate the high-speed separation capability of supercritical fluid chromatography (SFC) and the effectiveness of supercritical fluid chromatography/ mass spectrometry (SFC/MS) for fast, accurate determinations of multicomponent mixtures. A high-throughput liquid chromatography/mass spectrometry (LC/MS) analysis cycle time is reduced 3-fold using our general SFC/ MS high-throughput method, resulting in substantial time saving for large numbers of samples. Unknown mixture characterization is improved due to the increased selectivity of SFC/MS compared to LC/MS. This was demonstrated with sample mixtures from a 96-well combinatorial library plate. In this paper, we report a negative mode atmospheric pressure chemical ionization (APCI) method for SFC/MS suitable for most of the components in library production mixtures. Flow injection analysis (FIA) also benefits from this SFC/MS system. A broader range of solvents is amenable to the SFC mobile phase compared with standard LC/MS solvents, and solutes elute more rapidly from the SFC/MS system, reducing sample carryover and cycle time. Finally, our instrumental setup allows for facile conversion between LC/MS and SFC/MS modes of operation. In recent years, SFC has been exploited as an alternative to high-performance liquid chromatography (HPLC) because of its superior selectivity and speed. Capillary SFC has been employed in agricultural/environmental applications requiring herbicide or pesticide analyses.1,2 Packed column SFC is employed in the petroleum industry for quantitative analysis of diesel aromatics, replacing methods employing fluorescent indicators.3 The pharmaceutical industry has begun to rely on SFC over liquid chromatography for assays of chiral drugs.4 (1) Berger, T. A. Packed Column SFC, 1st ed.; The Royal Society of Chemistry: London, 1995; Chapter 1. (2) Wright, B. W.; Smith, R. D. J. High Resolut. Chromatogr. 1985, 8, 8-11. (3) Klee, M. S.; Wang, M. Z. Hewlett-Packard Application Note 228-226; HewlettPackard: Palo Alto, CA, 1993. (4) Blum, A. M.; Lynam, K. G.; Grasso, C. C. Chirality 1994, 6, 302-13.
2410 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
The use of SFC should expand as knowledge of its advantages accumulates, especially in high-throughput applications.1,5 Since it is often of interest to separate multicomponent mixtures in a minimal amount of time, the technique is finding increased use in industry for applications in which liquid chromatography separations are considered too slow. Similarly, when mass analysis of the analytes is necessary, SFC/MS will begin to replace LC/ MS.6-20 Most of the advantages of SFC relative to HPLC arise from the lowered intramolecular energy of interaction between mobile phase molecules in SFC.1 The molecular interaction energies of supercritical fluids are generally lower than those of liquids, which translates to lower viscosity, higher diffusivity, and lower column pressure drops relative to HPLC.21 Van Deemter plots show the range of optimal linear velocity (µ0) is increased up to 5-fold with SFC relative to HPLC.22 Additionally, lower supercritical fluid viscosities relative to those of liquids facilitate the use of longer columns in SFC while maintaining optimal conditions for highspeed separations.22-24 Further, since diffusion coefficients are an order of magnitude higher in supercritical fluids than in liquids, (5) Combs, M. T.; Ashraf-Khorassani, M.; Taylor, L. T. J. Chromatogr., A 1997, 785, 85-100. (6) Huang, E.; Henion, J. D.; Covey, T. R. J. Chromatogr. 1990, 511, 257-70. (7) Matsumoto, K.; Nugata, S.; Hattori, H.; Tsuge, S. J. Chromatogr. 1992, 605, 87-94. (8) Tyrefors, L. N.; Moulder, R. X.; Markides, K. E. Anal. Chem. 1993, 65, 2835-40. (9) Thomas, D.; Sim, P. G.; Benoit, F. Rapid Commun. Mass Spectrom. 1994, 8, 105-10. (10) Broadbent, J. K.; Martincigh, B. S.; Raynor, M. W.; Salter, L. F.; Moulder, R.; Sjo ¨berg, P.; Markides, K. E. J. Chromatogr., A 1996, 732, 101-10. (11) Scalia, S.; Games, D. E. Org. Mass Spectrom. 1992, 27, 1266-70. (12) Sadoun, F.; Virelizier, H.; Arpino, P. J. J. Chromatogr. 1993, 647, 351-9. (13) Via, J.; Taylor, L. T. Anal. Chem. 1994, 66, 1385-95. (14) Jedrzejewski, P. T.; Taylor, L. T. J. Chromatogr., A 1994, 677, 365-76. (15) Jedrzejewski, P. T.; Taylor, L. T. J. Chromatogr., A 1995, 703, 489-501. (16) Pinkston, J. D.; Chester, T. L. Anal. Chem. 1995, 67, 650A-656A. (17) Arpino, P. J.; Haas, P. J. Chromatogr., A 1995, 703, 479-88. (18) Pinkston, J. D.; Baker, T. R. Rapid Commun. Mass Spectrom. 1995, 9, 108794. (19) Morgan, D. G.; Islam, S.; Kitrinos, N. P., Jr. Proceedings of the 46th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 1998; p 545. (20) Pinkston, J. D.; Baker, T. R. Proceedings of the 46th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 1998; p 1157. (21) Berger, T. A. Packed Column Supercritical Fluid Chromatography Course Manual; Lake Tahoe, NV, 1998; p A-28. (22) Berger, T. A.; Wilson, W. H. Anal. Chem. 1993, 65, 1451-5. (23) Schleimer, M.; Schurig, V. In Analysis with Supercritical Fluids: Extraction and Chromatography; Wenclawiak, B., Ed.; Springer-Verlag: Berlin, 1992; pp 134-50. 10.1021/ac981372h CCC: $18.00
© 1999 American Chemical Society Published on Web 05/07/1999
the number of theoretical plates generated per unit time increases approximately 3-fold, contributing to SFC’s increased resolving power relative to HPLC.1,23 Neither HPLC nor SFC achieves plate counts per unit time approaching those possible with gas chromatography (GC) due to the extremely high diffusion coefficients of gases. Nevertheless, GC remains impractical for many highthroughput applications since most compounds of biological interest are incompatible with GC. The initial development of SFC/MS techniques was made cumbersome by the choice of established vacuum ionization techniques. Complex interfaces were designed to eliminate the gas load on the mass spectrometer source from the SFC effluent. The first SFC/MS system reported involved multiple pumping stages in the interface to an electron ionization (EI) source.25 Other applications employed chemical ionization (CI) with either a postcolumn split26 or a high-flow rate interface.27-29 In the former case, sensitivity was compromised, while in the latter, generally poor chromatography resulted with packed columns and flow rates were still limited to impractical levels for many applications. A moving belt interface for LC/MS capable of tolerating high flow rates was then applied to SFC/MS using EI and CI.30 Problems associated with this interface include losses of thermally labile solutes, carryover, and poor recoveries due to inefficient desorption from the belt. Modified thermospray interfaces have been employed by several researchers but have recently declined in prominence due to low sensitivity and poor reproducibility in general.11,13,31-35 Particle beam interfaces originally designed for LC/MS have been applied to SFC/MS, but their usage has decreased of late due to limited sensitivity and nonlinear response factors for major classes of compounds.14,15,36 A large factor in the recent decline in use of the aforementioned interfaces is due to the recent success of SFC/MS using atmospheric pressure ionization (API) interfaces. Very little interface design or modification is required to adapt an SFC system to the API source of a commercially available LC/MS system. Providing sufficient restriction at the API source is the critical point in maintaining a supercritical fluid prior to the source. Both ESI and APCI have been employed in numerous SFC/MS experiments.7-10,12,13,17,37-39 High flow rates typical of packed (24) Berger, T. A. Packed Column SFC, 1st ed.; The Royal Society of Chemistry: London, 1995; Chapter 4. (25) Randall, L. G.; Wahrahaftig, A. L. Anal. Chem. 1978, 50, 1703-5. (26) Crowther, J. B.; Henion, J. D. Anal. Chem. 1985, 57, 2711-6. (27) Smith, R. D.; Udseth, H. R. Anal. Chem. 1987, 59, 13-22. (28) Kalinoski, H. T.; Write, B. W.; Smith, R. D. Biomed. Environ. Mass Spectrom. 1988, 15, 239-42. (29) Kalinoski, H. T.; Smith, R. D. Anal. Chem. 1988, 60, 529-35. (30) Berry, A. J.; Games, D. E.; Perkins, J. R. J. Chromatogr. 1986, 363, 14758. (31) Chapman, J. R. Rapid Commun. Mass Spectrom. 1988, 2, 6-7. (32) Berry, A. J.; Games, D. E.; Mylchreest, I. C.; Perkins, J. R.; Pleasance, S. Biomed. Environ. Mass Spectrom. 1988, 15, 105-9. (33) Baskevich, J.; Hogge, L. R.; Berry, A. J.; Games, D. E.; Mylchreest, I. C. J. Nat. Prod. 1988, 51, 1173-7. (34) Morgan, E. D.; Murphy, S. J.; Games, D. E.; Mylchreest, I. C. J. Chromatogr. 1988, 441, 165-9. (35) Musser, S. M.; Callery, P. S. Biomed. Environ. Mass Spectrom. 1990, 19, 348-52. (36) Beaman, J.; Games, D. E.; Jackson, P. J.; Thomas, D. Adv. Mass Spectrom. 1992, 12. (37) Schmeer, K.; Nicholson, G.; Zhang, S.; Bayer, E.; Bohning-Gaese, K. J. Chromatogr., A 1996, 727, 139. (38) Morgan, D. G.; Harbol, K. L.; Kitrinos, N. P., Jr. J. Chromatogr., A 1998, 800, 39-49.
column SFC are amenable to ESI-MS even when appreciable levels of organic modifiers are used.12 However, response factors are dependent on the concentration of modifier in solution as ESI depends on proton transfer in solution. APCI also supports high flow rates as the nebulizer operates at atmospheric pressure. The solutes and mobile phase are completely vaporized in the heated nebulizer with the coaxial addition of a nebulizing gas, usually N2, and solutes are ionized by proton addition or abstraction in the corona discharge plasma consisting of protons and electrons. Since APCI generally results in M + H or M - H ions, mass spectral interpretation is often trivial. This benefits open-access mass spectrometry applications which are becoming increasingly common in modern high-throughput laboratories. Of particular importance for this work, SFC-APCI/MS can be exploited for the demands of high-throughput combinatorial library analysis. Shorter SFC/MS run times coupled with the greater resolving power of SFC relative to HPLC make this technique well suited to our application. We describe a flexible system, capable of rapid change between SFC/MS and LC/MS modes of operation and convertible between autosamplers injecting from individual vials or multiple 96-well plates. Rapid mass spectrometric flow injection analyses (FIA) are also described using this interface. This is accomplished by eliminating the chromatographic column to further reduce turnaround time for screening extremely large numbers of samples.40 The SFC mobile phase system transports solutes more efficiently than HPLC; therefore, a wide variety of sample solvents may be introduced to the SFC/MS system without adverse effect. This makes SFC-FIA/MS a “user-friendly” open-access method, requiring less preparation time for analysts seeking rapid screenings or confirmation of compound purity. The results presented here show that separations and mass identification of components in library mixtures generated by parallel synthesis can be achieved more rapidly with SFC/MS than with LC/MS. Samples consisting of products and byproducts of reaction mixtures were examined using SFC/MS and LC/MS methods. The data presented show the SFC separation of these compounds is achieved with better selectivity and speed than HPLC while maintaining mass spectrometric integrity. A sample for which our high-throughput LC/MS method failed to achieve separation of components is examined using SFC/MS. The utility of SFC/MS for fast flow injection screening is also discussed. EXPERIMENTAL SECTION Instrumentation. The systems used in this experiment are an HP 1100 LC/MSD (Hewlett-Packard, Palo Alto, CA), a Berger supercritical fluid chromatograph (Berger Instruments, Newark, DE), and a 215 liquid handler (Gilson, Middleton, WI). LC/MS data were acquired using Chemstation software (Hewlett-Packard, Palo Alto, CA) on the HP LC/MSD, and both the Chemstation and Berger software packages contributed to SFC/MS operation. Ultraviolet (UV) absorption and mass spectral data were also acquired and processed using Chemstation software for the LC/ MSD system. Berger software on a slave computer was used for (39) Pinkston, J. D.; Baker, T. R. J. Am. Soc. Mass Spectrom. 1998, 9, 498-509. (40) Hayward, M. J.; Snodgrass, J. T.; Alluri, M. R.; Thomson, M. L.; Carter, G. T. Proceedings of the 42nd ASMS Conference on Mass Spectrometry and Allied Topics, Chicago, IL, 1994; p 488.
Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
2411
Figure 1. Schematic of the SFC/MS system with mobile phase flow path.
mobile phase control as well as Berger autosampler control for initiation of individual vial injections. Samples were injected from the vials into the system with a Berger ALS 3100 autosampler equipped with a closed-loop injector. The Gilson 215 liquid handler was used to inject samples from 96-well plates and was controlled by CombiChrome software (Integ, Rotterdam, The Netherlands) incorporated in the HP Chemstation. Automated data analysis of multiplate runs was completed using vendor-supplied software designed by Alanex and Hewlett-Packard. A general schematic illustration showing the instrumental components of the SFC/MS system is shown in Figure 1. The sample is introduced into the mobile phase stream by the autosampler and then travels through 0.010 in. i.d. stainless steel tubing into one of six available column ports in the SFC oven. It continues through the column and into the Berger diode array detector (DAD), equipped with a high-pressure flow cell capable of operating at mobile phase pressures up to 400 bar. Effluent from the DAD is split upstream of the SFC back-pressure regulator using a Valco (Houston, TX) 1/16 in. tee to direct a fraction of the flow into the mass spectrometer. The back-pressure regulator is able to maintain high mobile phase pressures because sufficient restriction is provided by 1 m of 0.0025 in. i.d. PEEK tubing between the tee and the MSD APCI source. This i.d. conforms to the optimal restrictor diameter range 50-75 µm (0.0020-0.0030 in.) determined by Morgan and co-workers for SFC/APCI-MS.38 The split ratio between the fraction of effluent sent to waste and that sent to the APCI source was measured to be approximately a 3:2 ratio. To maintain stable solvent temperature between the UV detector and the APCI source, Thermolyne heat tape (Barnstead Thermolyne, Dubuque, IA) was wrapped around the stainless steel tubing and maintained at 60 °C by a Thermolyne constant-wattage controller. The design of the system with direct introduction of the chromatographic effluent to the source allows for facile conversion between SFC/MS and LC/MS operation modes as well as between 96-well plate and single vial formats. To achieve this, the 2412 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
HP 1100 autosampler or the Gilson 215 liquid handler replaces the Berger autosampler. The stream from the LC column replaces the SFC stream through the Berger DAD to the APCI source. LC/MS analyses are initiated through connection of the start cable for the HP autosampler to the MSD after disconnecting the external autosampler start cable. The HPIB cable from the data acquisition computer remains connected to the Berger DAD, configured on Chemstation where data are collected and processed. The HP 1100 MSD also allows for rapid conversion between APCI and ESI modes, although no ESI data are presented in this paper. For the analysis of samples in individual vials, the Berger autosampler was configured on the SFC slave computer, and each sample injection preceding an experiment was initiated with a start command from the SFC software. Upon injection, the Berger 3100 autosampler sends a start signal to the SFC pump, oven, and DAD, as well as to the MSD and HP 1100 modules. Finally, the Chemstation software receives a signal to start recording spectral data from the HP 1100 and MSD module and data acquisition begins. When samples from multiple 96-well plates were analyzed, the SFC pump was programmed on the slave computer. The CombiChrome software contains the method and sample list and sends an inject signal to the Gilson 215 via a serial cable from the Chemstation computer. The Gilson 215 sends a start signal to the other components, beginning data acquisition. Automated data analysis is completed off-line by the custom software created by HP and Alanex. Analysis Conditions. Because our libraries contain highly diverse chemical entities and thus a wide range of polarities, we developed a single LC/MS method for the entire library. The general high-throughput SFC/MS method developed for this work is compared with our general LC/MS method. A 2.1 × 30 mm Zorbax rapid resolution XDB-C18 cartridge column with 3.5 µm StableBond packing was used for LC/MS (Hewlett-Packard, Palo Alto, CA). The LC column temperature
was held at 35 °C. The HPLC mobile phase gradient consists of a 3-min initial ramp from 15% methanol/85% water to 35% methanol/65% water with the flow rate increasing linearly from 0.50 mL/min to 0.55 mL/min. This ramp is followed by a 1-min ramp to 95% methanol/5% water at 0.55 mL/min. The methanol contains the additive 1,2-dichloroethane at 3% in solution, while the water is saturated with dichloroethane (at 0.15 vol %). Dichloroethane assists in electron transfer for the proton loss process of negative APCI and thus greatly enhances the signalto-noise ratio.41 After a 2-min hold at 95% methanol, the column is returned to its initial conditions over 30 s. At this stage, the gradient program consumes 6.5 min. The system’s 0.75-mL dead volume results in a 1.5 min delay in the gradient. A 2.5-min column reequilibration time at the initial conditions is required to maintain consistent baselines for successive sample runs, resulting in an LC run time of 9 min. Data acquisition is stopped at 7.5 min. The final method conditions for the SFC data presented here included use of a methanol mobile phase modifier, with 0.3% isopropylamine and 1% 1,2-dichloroethane additives. Isopropylamine is believed to change the polarity of the stationary phase relative to that of the mobile phase and adsorb to active sites on the column to decrease chromatographic tailing. It may be used in conjunction with negative APCI but not positive APCI because it competes with less basic analytes for gas phase protons, suppressing the efficiency of ionization. The stationary phase was a 4.6 × 50 mm, 5 µm/60 Å ZyroSil Diol column (Zymor Inc., Wayne, NJ) held at a temperature of 45 °C by the SFC column oven. A constant flow rate of 3.5 mL/ min was used. The mobile phase gradient began immediately after sample injection with 5% modifier and increased to 32.5% at 10%/ min before dropping back down to 5%, post-data acquisition, to restore the starting conditions. The full gradient time was thus 2.75 min. The back-pressure regulator was set to maintain column outlet pressures of 150 bar for chromatographic data presented. Complete elution of analytes from all samples observed was achieved in 2.5 min, but the total cycle time including injection and column reequilibration time was approximately 3.5 min. Further reduction of the total run time is possible and is currently under investigation. The HP MSD was used for mass spectrometric analysis with the standard APCI source installed. Direct fluid introduction with no presource sample split was used for the LC/MS analysis, whereas a tee was used to split the SFC effluent downstream of the UV cell between waste and flow to the APCI source. APCI settings for the nebulizer, drying gas, corona potential difference, and discharge current were maintained for both LC/MS and SFC/ MS experiments reported here. Alteration of temperature and pressure for the nebulizer and drying gas from their optimum setpoints for LC/MS did not reveal any significant improvement in signal intensities for SFC/MS. APCI conditions used included a corona needle voltage of 3 kV with 10-µA discharge current. The mass spectrometer scanned the mass range 105-1000 Da throughout each run. For the flow injection analysis, isocratic conditions were maintained for both LC/MS and SFC/MS systems. The HPLC mobile phase was 95% methanol/5% water flowing at 0.50 mL/ (41) Fischer, S. M.; Miller, C. A.; Kuhlmann, F. Proceedings of the 46th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 1998; p 1401.
Figure 2. General reaction diagram displaying the final step of thiohydantoin synthesis used in the preparation of each well analyzed.
min, whereas the SFC mobile phase was 20% methanol in CO2 with a flow rate of 3.5 mL/min. The SFC outlet pressure was set at a constant 250 bar for the analysis performed. The same tee and APCI inlet restrictor assembly described above was used. Negative mode APCI was employed for the examples shown with the same source conditions used in chromatographic experiments. Sample Preparation and Analysis. The experiments presented resulted from comparative LC/MS and SFC/MS analyses of a 96-well plate consisting of structures obtained in the synthesis of thiohydantoins as displayed in Figure 2.42 This general reaction was used in the experiment since its reactants, products, and byproducts span a range of polarities somewhat representative of those historically investigated by LC/MS in our laboratory. The library plates were sampled and an aliquot was diluted with sufficient methanol/water (95:5) to produce a final theoretical concentration of 250 µM, assuming 100% conversion of reactants to product. The injection volumes were 25 µL using a fixed sample loop on the Gilson 215 for LC/MS and 10 µL using the fixed loop for SFC/MS. Flow injection results were generated using a 10 ng/µL solution of reserpine (Sigma Chemical Co., St. Louis, MO) in methanol, again with negative mode APCI. RESULTS AND DISCUSSION A typical total ion current (TIC) chromatogram from analysis of one well (D6) in the plate is shown in Figure 3. Both the UV and negative APCI mass spectral data are shown for the data acquired in the SFC mode (Figure 3a,b) and in the LC mode (Figure 3c,d). The analysis time required for complete elution of all components in the well is reduced from 6 min with the LC/ MS method to less than 1.7 min with SFC/MS. The delay between the corresponding UV and TIC peaks is due to the time required for the fluid stream to travel from the DAD cell to the mass spectrometer. The average time delay between the LC/MS UV and TIC signals is 2.79 s, which is reduced with SFC/MS to 2.26 s. In addition to this delay time, some band broadening is apparent in TICs relative to UV chromatograms due to diffusion in the tubing, which is longer in the case of SFC/MS, between the UV cell and APCI source. Still, because it is possible to use higher flow rates for SFC/MS, chromatographic peaks are better resolved in a shorter time than is achieved with LC/MS. It should be noted that the examples given reflect the more challenging separations from the plates examinedsthese wells provide more interesting comparisons. In the SFC TIC chromatogram from Figure 3b, neither the order nor relative intensities of the peaks match results from the LC analysis (Figure 3d). The elution order has changed is because (42) Sim, M. M.; Ganesan, A. J. Org. Chem. 1997, 62, 3230-5.
Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
2413
Figure 3. Well D6 chromatograms from (a) UV detection of solutes eluted from the SFC, (b) SFC/MS total ion current, (c) UV detection of solutes eluted from the HPLC, and (d) LC/MS total ion current.
the analysis has switched from a reverse-phase technique (HPLC) to normal phase (SFC). An explanation for the relative signal intensity differences between TIC peaks from the same sample using LC/MS and SFC/MS may be that the relative basicity of the gas phase analytes changes with the different plasmas each mobile phase produces in the APCI source. The background gas in the source changes in composition with the solvent gradient as well. In negative APCI mode, the SFC/MS response increases with modifier concentration in the mobile phase as the dichloroethane concentration in the source increases with the gradient. Therefore, later eluters in SFC/MS tend to have higher response factors and earlier eluters tend to have lower responses relative to corresponding LC/MS signals. The response can be made linear across the TIC profile by introducing a postcolumn auxiliary flow of dichloroethane to the stream entering the APCI source. The negative APCI/MS TIC traces for well B5 displayed in Figure 4 also demonstrate some advantages of SFC/MS over LC/ MS. A 4-fold reduction in retention time for the final peak is achieved using the SFC/MS analysis. This comparison also demonstrates the ability of the SFC/MS analysis to resolve moderately nonpolar compounds that tend to coelute by highspeed LC/MS methods. The TIC chromatogram for well B5, shown in Figure 4a demonstrates the capability of the SFC/MS method to achieve separation of two components with parent ion signals at 317 and 333 Da, respectively. The 317-Da signal peaks at about 0.6 min, well displaced from that for 333 Da, which peaks at 1.1 min. In Figure 4b, the LC/MS TIC on this sample shows the separation is incomplete, as suggested by the shoulder on the tail of the peak at 5.7 min. Since it is known that the expected parent ion signal is at m/z 317 using negative APCI, the abundant presence of m/z 333 under the same integrated TIC peak masks 2414 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
the contribution of the expected compound. Because the signal from m/z 317 is not the principal ion signal under the 5.7 min TIC peak, the product compound might be missed altogether using automated high-throughput data analysis. At minimum, quantitation of purity may be compromised. Since the SFC/MS extracted ion chromatograms (EICs) exhibit complete separation of these two species, a more accurate purity determination for the reaction mixture can be obtained. The SFC/MS TIC shown in Figure 5a exhibits higher chromatographic speed and resolving performance exceeding that observed in the LC/MS TIC trace in Figure 5b for the compounds in well A3. In addition, separation of the expected product compound from reaction byproducts in the mixture is achieved in the SFC/MS analysis where it is not by LC/MS. A clear distinction can be made between the TIC peaks corresponding to the species at 346 Da (product) and 398 Da (byproduct) revealed by EICs. Using these EICs, the m/z ratio of the predominant species detected under each chromatographic peak is labeled above each TIC peak shown in Figure 5. EICs from the LC/MS chromatogram exhibit complete coelution of the two species corresponding to parent ions of m/z 346 and 398. Integration of the two peaks generated by the SFC/MS analysis with automated data processing will yield a more accurate determination of the product purity in this mixture than that obtainable from the LC/MS TIC. Additional mass spectral analysis of these TICs revealed incomplete separation of two major byproducts by LC/MS whereas complete separation was achieved by SFC/MS. Mass spectra corresponding to chromatographic peak maxima are displayed in Figure 6. The spectra shown in Figure 6a,b correspond to the SFC/MS chromatographic peak maxima indicated
Figure 4. Well B5 TIC chromatograms with overlaid extracted ion chromatograms from (a) SFC/MS analysis yielding complete separation of m/z 317 and 333 with elution of all solutes in under 1.5 min and (b) LC/MS analysis revealing only partial resolution of m/z 317 and 333, with elution of all solutes requiring nearly six minutes.
Figure 5. Well A3 TIC chromatograms for (a) SFC/MS analysis indicating four distinct peaks with predominant parent ion species at 346, 398, 304, and 302 Da and (b) LC/MS analysis exhibiting coelution between compounds yielding parent ions at 304 and 302 Da, as well as between species of 346 and 398 Da.
by the labels “304” and “302”, respectively, in Figure 5a. The signal from m/z 302 shown in Figure 6a is present at only about 15% the abundance of the m/z 304 signal. In Figure 6b, the signal at m/z 304 appears at a level less 5% of the primary m/z 302 signal. The LC/MS mass spectrum shown in Figure 6c corresponds to
the maximum of the peak labeled “302” in Figure 5b. Due to inadequate chromatographic resolution between the compounds generating primary ions of m/z 302 and 304, the mass spectrum exhibits these two signals at nearly the same abundance. These results demonstrate superior mass spectral purity due to the Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
2415
Figure 6. Mass spectra corresponding to SFC/MS chromatographic peak maxima for well A3 SFC/MS analysis with the primary ion signal at (a) m/z 304 and (b) m/z 302. (c) Mass spectrum corresponding to the LC/MS chromatogram peak maximum for the primary ion signal at m/z 302.
improved separation capability of high-throughput SFC/MS. A 10 ng/µL reserpine (MW 608) standard was analyzed by high-throughput FIA screening methods using both the HPLC and the SFC fluid systems with negative APCI. For both systems, no chromatographic column is used; the sample travels from the injector to the DAD cell, and the effluent is sent to the mass spectrometer source through the same direct connection used in experiments described above. The reduced viscosity of the SFC mobile phase allows the sample to be detected by UV with almost no delay after injection. Comparison of HPLC signal traces from flow injection analyses to the SFC traces revealed that a much higher elution speed can be achieved with the SFC system and accurate mass spectral information can be obtained. We observed that the maxima of HPLC UV absorbance signals occurred in the range of 0.10 min as compared to 0.037 min with the SFC system. The SFC-UV signal returns fully to baseline level in about 4.2 s compared with 12 s for the HPLC-UV signal. These time differences are significant, considering the prospect of daily screening of thousands of samples in high-throughput laboratories. A 30-s cycle time reduced by 7.8 s (injection to injection) amounts to nearly 6 h of analytical time saved, or more than 1000 additional sample analyses, per 24-h period. Mass spectral characterization by SFC-FIA is possible within 6 s after injection with this system, leaving the autosampler as the rate-limiting step in flow injection analysis. CONCLUSIONS The experiments outlined reveal the significant time-saving and analytical advantages available using SFC/MS for routine high2416 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999
throughput analyses. We have shown that SFC/MS is advantageous over LC/MS not only because it increases sample throughput due to shorter analysis times but also because chromatographic separation of the well components is enhanced. Combining a multiplate autosampler and the SFC/MS system allows unattended UV and mass spectral analyses of over 400 samples per 24-h period. Many compounds that coelute using LC/MS are separated by SFC/MS, improving the accuracy of automated quantitative analysis based on UV and ion chromatograms. While the sensitivity of SFC/APCI-MS may be modestly reduced relative to LC/ APCI-MS, this issue is usually not a concern for combinatorial applications where sample quantities are well above mass spectral sensitivity limits. When mass spectral sensitivity is an issue, it may be improved by introducing an appropriate auxiliary fluid to the system before the APCI source or by optimization of the SFC modifier/additive mixture. Further studies of these enhancements are currently being explored in our laboratory. ACKNOWLEDGMENT The authors thank Izabella Khanoukova for preparation and repreparation of the plates containing sample mixtures used in this presentation.
Received for review December 11, 1998. Accepted March 16, 1999. AC981372H
