Anal. Chem. 2001, 73, 247-252
Techniques for Increasing the Throughput of Flow Injection Mass Spectrometry Kenneth L. Morand,* Thomas M. Burt, and Brian T. Regg
Health Care Research Center, Procter & Gamble Pharmaceuticals, 8700 Mason-Montgomery Road, Mason, Ohio 45040 Thomas L. Chester
Miami Valley Labs, Procter & Gamble, 11810 East Miami River Road, Ross, Ohio 45061
Improvements to the design and operation of a Gilson 215 multiprobe liquid-handling system have resulted in a significant increase in the throughput for flow injection molecular weight characterization of combinatorial chemistry libraries. The rapid injection sequence, and subsequent increased sample throughput, is effected by directing the entire mobile-phase flow through each of the injection loops sequentially while isolating or “deadending” the remaining nonactive loops. This mode of operation was accomplished by incorporating columnswitching valves prior to and following the set of eight parallel injectors. Analysis rates are achieved without sacrificing the integrity of the flow injection peak profile as baseline resolution is maintained for all samples. Using this system, the total analysis time for a 96-well microtiter plate has been reduced to ∼5 min. The use of high-throughput organic synthesis (HTOS) and biological screening (HTS) has provided a powerful tool to more broadly investigate the range of molecular diversity available for pharmaceutical development. Mass spectrometry has played an important role in the support of both HTOS and HTS as evident in the number and varied approaches that may be cited from the literature.1,2 As examples, mass spectrometry (MS) characterization is now routinely required for the following: (a) support-bound and solution-phase libraries; (b) single-component, medium-sized (5-1000 compounds) and large (.1000 compounds) mixtures; (c) analogues from a variety of chemical scaffolds; and (d) chemistries of widely varying reliability and yield. In addition, the use of automated/robotic synthesis and screening systems underscores the need for very high-throughput MS characterization capabilities as format changes rapidly evolve to accommodate the desire for larger chemical arrays. With these developments, the analytical community has been challenged to develop technologies that are compatible with the analysis of increasingly larger numbers of compounds. A significant portion of the literature, which references the support of HTOS, has focused on the use of either high* To whom correspondence should be directed: (e-mail) Morand.kl@ pg.com; (tel) 513-622-2493; (fax) 513-622-1196. (1) Su ¨ ssmuth, R. D.; Jung, G. J. Chromatogr., B 1999, 725, 49-65. (2) Fitch, W. L. Mol. Diversity 1999, 4, 39-45. 10.1021/ac0003201 CCC: $20.00 Published on Web 12/14/2000
© 2001 American Chemical Society
throughput (HT) flow injection (FIA) or high-performance liquid chromatography (HPLC) MS for rapid compound profiling. For example, Kassel et al.,3 first reported on interfacing a Gilson 215 multiprobe liquid-handling system to a quadrupole mass spectrometer to effect rapid FIA-MS characterization. Using this system, the authors were able to realize nearly a 4-fold increase in throughput over conventional single-probe systems. For example, the total analysis time for a 96-well microtiter plate was reduced from ∼48 min for a single-probe autosampler to 12 min for the multiprobe autosampler.3 Additionally, there have been a number of initiatives to address high-throughput HPLC-MS profiling of combinatorial chemistry arrays. Typically these techniques have fallen in to one of two categories, i.e., (1) HT serial HPLC-MS analysis4 or (2) parallel HPLC-MS analysis.5 Kyranos et al.,4 have been instrumental in the development of rapid serial HPLC-MS methods, i.e., 1 min/ sample. Their approach utilizes short chromatographic columns, e.g., 4.6 × 30 mm, in combination with high solvent flow rates (4 mL/min) and steep chromatographic gradients (15-95% organic in 0.7 min). Compound characterization is focused toward the rapid determination of sample purity using MS in combination with on-line UV and evaporative light scattering (ELS) detection. Other, more recent, developments have focused on parallel chromatographic systems that are compatible with serial-based mass spectrometric detection. These systems incorporate mechanical segmentation and sampling of multiplexed chromatographic column flow paths in order to increase sample throughput for HPLC-MS. At present, we discuss efforts to address the development of HT FIA-MS protocols to meet the many needs and requirements of HTOS and HTS programs. Specifically, we will discuss the development of applications to support very high-throughput molecular weight identification, i.e., ∼ 5 min/96-well plate. Highthroughput analysis is achieved via improvements to the design and operation of a Gilson 215 multiprobe liquid-handling system. (3) Wang, T.; Zeng, L.; Strader, T.; Burton, L.; Kassel, D. Rapid Commun. Mass Spectrom. 1998, 12, 1123-1129. (4) Lee, H.; Li, L.; Kyranos, J. Improvement of Ultrafast HPLC/MS for HighThroughput Analysis of Combinatorial Libraries. Proceedings of the 1999 ASMS Conference. (5) De Biasil, V.; Haskins, N.; Organ, A.; Bateman, R.; Giles, K.; Jarvis, S. Rapid Commun. Mass Spectrom. 1999, 13 (12), 1165-1168.
Analytical Chemistry, Vol. 73, No. 2, January 15, 2001 247
Figure 1. High-throughput flow injection mass spectrometry system. System configuration includes the following: HP 1100 series HPLC system, Gilson 215 multiprobe liquid-handling system modified to include a Valco Cheminert model C5 eight-position column selection valve, and a Micromass LCT electrospray ionization time-of-flight mass spectrometer. Inset: flow diagram for the Valco Cheminert model C5 eightposition column selection valve. The solid line represents the “selected” flow path, while the dotted lines represent the remaining seven “inactive” flow paths.
EXPERIMENTAL SECTION Flow Injection Mass Spectrometry System. Figure 1 shows a general diagram of the high-throughput FIA-MS instrumentation. The system is based on a modified Gilson 215 multiprobe liquidhandling system (LHS) (Gilson, Inc., Madison, WI) with an 889 multiport injection module and integrated to a Micromass LCT electrospray ionization time-of-flight (ESI-TOF) mass spectrometer (Micromass UK Ltd.). The Gilson 215 and 889 modules were controlled via an off-line computer operating the Gilson 709 Sample Manager software. Initiation of the injection sequence was affected via a contact closure signal at the start of the sample analysis run on the mass spectrometer. Modifications to the Gilson 215 system included the replacement of the splitter and combiner static splitters (I and II in Figure 1) with a Valco Cheminert model C5 eight-position column selection valve (Valco Instruments Co., Inc., Houston, TX) with a 0.25-mm (0.010 in.) bore and zero dead volume (ZDV) fittings. The valve assembly consists of two selection valves mounted on a single EMT high-torque microelectronic actuator. Actuation of the selection valves is controlled by a remote logic level signal from the 889 injector control program. The flow path for the HT FIA-MS system is shown in the inset of Figure 1. The solid line represents the single “active” flow path while the dashed lines represent the remaining seven “nonactive” and dead-ended flow paths. The 889 injectors and the Valco valves were programmed to switch at interval times between approximately 1 and 6 s. HPLC Conditions. Mobile phase, 100% methanol with 0.1% formic acid and 2 mM ammonium acetate, was pumped at a flow 248
Analytical Chemistry, Vol. 73, No. 2, January 15, 2001
rate of 1-3 mL/min with a Hewlett-Packard 1100 series (HewlettPackard, Santa Clarita, CA) HPLC system. Solvents and additives were of HPLC grade (J. T. Baker). Unless noted otherwise, the mobile-phase flow was directed into the mass spectrometer without a split. Samples were prepared as neat solutions in 100% methanol at a concentration of 10-50 µg/mL and injected without further dilution at 5-µL injection volumes. The total system back pressure was measured at 30 bar. Mass Spectrometry Conditions. All experiments were performed on a Micromass LCT ESI-TOF mass spectrometer. The LCT was operated in the positive ion mode across a range of 101000 Da/e and a cone voltage of 30 V. Due to the high solvent flow rate into the mass spectrometer, the source block and desolvation temperatures were set at 150 and 350 °C, respectively, and the nitrogen desolvation gas flow was set at 750 L/h. Flow injection profiles for an entire 96-well microtiter plate were collected into a single data file. Flow Injection Modeling. Modeling the FIA-MS peak widths provided insight to the influence of a number of operational parameters. Five contributors to peak broadening were considered: (1) the initial (temporal) width of the peak in the sample loop, (2) transport of the sample from the loop, and (3-5) transport through the tubes connecting the injector to the combiner, the combiner to the splitter, and the splitter to the mass spectrometer. Peaks were assumed to be distributed temporally, and the treatment of Attwood and Golay was applied to estimate
the broadening occurring in these individual tubes.6 We ignored broadening occurring within the combiner, splitter, and the various fittings and assumed that rapid radial mixing occurs at each of these locations so that the variances from the remaining components will be additive.6 Initially, the sample is uniformly distributed over the filled sample loop with a standard deviation, σs
σs )
1 Vs x12 F
(1)
where Vs is the loop volume and F is the flow rate through the loop. To estimate the broadening due to transport, it is necessary to express the length of each tube in terms of theoretical plates. In FIA, longitudinal diffusion may be safely ignored6 and the plate number, N, for a tube is
N)
L 96DL ) 2 ) 24πDL/F H d u
(2)
c
where D is the diffusion coefficient for the sample in the liquid phase and L is the length of the tube.6 We then calculated the standard deviation from a particular tube, j, as
σj ) Vj[Nj-1/2(1 + 3/Nj)-1/4], Nj g 0.01
(3)
where Vj is the volume of the tube.6 To transport the sample out of the sample loop, the peak center must be displaced through only half the loop volume. Thus, for this source only, N is calculated using L/2 and σ is calculated using this value of N and V/2. The total standard deviation observed by the mass spectrometer, σt, is then estimated from the square root of the sum of the squares of the individual standard deviations. The predicted peak width at the peak base was then taken as 4σt. These equations were coded into a spreadsheet (Microsoft Excel, version 7.0a, Microsoft Corp., Redmond, WA). Pump rate, tube dimensions, and splitting and recombination details were specified as input. The output included the standard deviations for the individual broadening sources, the total peak width, and the contribution of each broadening source to the total variance. The experimental values for data obtained using the standard operating conditions for the Gilson 215 multiprobe were used to validate the modeled predictions for the peak profiles. DISCUSSION The FIA-MS total ion chromatogram in Figure 2 provides an example of the type of data that may be acquired using a standard Gilson 215 multiprobe configuration. In acquiring these results, the HPLC system was operated at a flow rate of 5 mL/min with a 17:1 split ratio, waste:MS, and the injection valves were triggered at intervals of 6 s, or 48 s/column on the microtiter plate, i.e., one set of eight injections.7 Evident from the figure, peak widths under these conditions were measured to be approximately 2.5(6) Attwood, J. G.; Golay, M. J. E. J. Chromatogr. 1981, 218, 97-122. (7) The standard microtiter plate naming conventions will be used throughout this paper. Such that, the coordinates of a 96-well microtiter plate contain 12 columns, labeled 1-12, and eight rows, labeled A-H. The Gilson LHS is configured to sample individual columns, e.g. A1-H1.
Figure 2. FIA-MS total ion chromatogram of terfenadine (MW ) 471.3, C32H41NO2), for single set of eight injections obtained with a standard 215 multiprobe configuration at a flow rate of 5 mL/min and a split ratio of 17:1. The peak widths were measured as 2.5-3.0 s at fwhm and 5.5-6.0 s at the base.
3.0 s fwhm and 5.5-6.0 s at the base. Baseline separation of the flow injection profiles was maintained at greater than 90% valley for all peak profiles with the exception of peaks 6 and 7, which may be attributed to extra dead volume associated with injector 6. The total analysis time for a 96-well microtiter plate was ∼20 min. Notable from this and later experiments was the presence of large dead-time zones. The dead time, which accounts for greater than 9 min of the total time for the present experiment, can be attributed to a number of steps in the injection sequence for the Gilson 215 system. This includes the probe and injector wash cycles, probe arm movements, sample loading, etc. For the purpose of this discussion, the dead time will be referred to as the “intraplate” cycling time. Some effort was made to optimize the injection and wash sequences in order to reduce the intraplate cycle time, and for subsequent experiments the time was reduced to approximately 4.5-5.5 min/plate. Nevertheless, on the basis of the results of these early experiments, it was recognized that limitations of the standard system configuration would significantly limit our analysis speed and, subsequently, the overall sample throughput. Reducing the FIA-MS peak profile is directly related to how fast the sample plug can be pushed through the tubing leading up to the mass spectrometer. A major drawback in the design of the standard 215 multiprobe is the necessity to split the total solvent flow among the eight separate sample injectors. Hence, when the flow is split, the rate through the each of the injectors is not defined as the total system flow rate, but as one-eighth of this rate; e.g., for a 5 mL/min flow, the rate through each injector is 625 µL/min. Therefore, while optimizing the tubing diameter and length between the autosampler and the mass spectrometer will certainly lead to some reduction of the peak width and the overall analysis time, the linear velocity of the sample plug will be limited by the flow rate through the region of the injector and the combining valve. Therefore, to effect a significant reduction of the peak width, it is necessary to look at modes of operation that dramatically increase the linear velocity of the sample plug Analytical Chemistry, Vol. 73, No. 2, January 15, 2001
249
Table 1. High-Throughput Flow Injection-Mass Spectrometry Optimization Programa pump rate diffusion coeff injector dinj(in) flow rate sample vol divisor inj-to-combiner d(in) L(cm) combiner-splitter-MS d(in) L(cm) splitter-MS d(in) L(cm) split ratio square width of initial peak
initial peak inj loop inj-to-container combiner-to-splitter splitter-to-MS total
2 mL/min 1.50 × 10-5 cm2/s 0.007 2 mL/min 5 mL 1
dinj(cm) N Linj(cm)
0.02 0.68 20.14
0.005 50
d(cm) N
0.01 1.70
0.005 10
d(cm) N
0.01 0.34
0.005 0.00000001 1 winit
d(cm) N
0.01 0.00
0.0025 min or 0.15 s
σ(min)
4σ
% of variance
0.0007 0.0020 0.0019 0.0006 0.0000
0.0029 0.0079 0.0075 0.0025 0.0000
6.21 46.95 42.35 4.49 0.00 100.00
total predicted baseline peak width 4σ(min) ) 0.01 or 4σ(s) ) 0.70 a Modeled results for configuration of Gilson 215 multiprobe with sequential selection of the sample injector flow paths.
through this region, specifically, a system in which the flow rate through the injector is equal to the total system flow rate. This may be accomplished by directing the entire system flow through each injector sequentially. A schematic diagram of the system with this configuration is shown in Figure 1. Table 1 outlines the parameters and modeling results for such modifications to the 215 multiprobe. Although the total flow rate was reduced from 5 to 2 mL/min, this still represented a 3-fold increase in the flow rate through each injector when compared with the standard system configuration and results in a reduction of the modeled peak widths from 6 to