Anal. Chem. 1998, 70, 3227-3234
High-Throughput Flow Injection Analysis Mass Spectroscopy with Networked Delivery of Color-Rendered Results. 2. Three-Dimensional Spectral Mapping of 96-Well Combinatorial Chemistry Racks Ekkehard Go 1 rlach,*,† Ramsay Richmond,†,‡ and Ian Lewis§
Core Technology Area Analytics Unit and Transplantation Department, Novartis Pharma AG, CH-4002 Basel, Switzerland
For the last two years, the mass spectroscopy section of the Novartis Pharma Research Core Technology group has analyzed tens of thousands of multiple parallel synthesis samples from the Novartis Pharma Combinatorial Chemistry program, using an in-house developed automated high-throughput flow injection analysis electrospray ionization mass spectroscopy system. The electrospray spectra of these samples reflect the many structures present after the cleavage step from the solid support. The overall success of the sequential synthesis is mirrored in the purity of the expected end product, but the partial success of individual synthesis steps is evident in the impurities in the mass spectrum. However this latter reaction information, which is of considerable utility to the combinatorial chemist, is effectively hidden from view by the very large number of analyzed samples. This information is now revealed at the workbench of the combinatorial chemist by a novel three-dimensional display of each rack’s complete mass spectral ion current using the in-house RackViewer Visual Basic application. Colorization of “forbidden loss” and “forbidden gasadduct” zones, normalization to expected monoisotopic molecular weight, colorization of ionization intensity, and sorting by row or column were used in combination to highlight systematic patterns in the mass spectroscopy data. The introduction of combinatorial chemistry methodology to accelerate drug discovery has had a large impact on the analytical techniques within pharmaceutical companies, especially on mass spectroscopy (MS). Traditional mass spectroscopy facilities faced with the demands of their combinatorial chemistry groups have been obliged to drastically increase their sample processing rate. This factor has had multiple consequences and has forced the delegation of traditional MS operator actions, e.g., results reporting, to automatic systems.1 * Corresponding author. † Core Technology Area Analytics Unit. ‡ Corresponding author on analytics. § Transplantation Department. S0003-2700(98)00026-2 CCC: $15.00 Published on Web 07/03/1998
© 1998 American Chemical Society
Within combinatorial chemistry, two strategies are available to synthesize a library of diverse compounds: parallel synthesis, generally producing single compounds, and split and mix synthesis, producing mixtures. The technique presented here is a consequence of our facing the chemistry of the former rather than the latter.2,3 The chief analytical characteristics of these samples within Novartis Pharma are a relatively high concentration, a general expectation of only one structure in each rack well, and the building-block-dependent intelligent arraying of the 96-well rack’s rows and columns by Novartis Pharma’s Synthman software.4 A flow injection analysis mass spectroscopy system (FIA-MS) able to give purity estimates5 on industrial scale combinatorial chemistry was designed in-house; it uses the 96-well Micronic rack as the basic sample handling unit. The system has some parallels with an earlier report.6 It also has some similarities with automated open-access mass spectrometers analyzing conventional synthetic chemistry samples.7 However, the dominant ionization mode for open-access MS is atmospheric pressure chemical ionization (APcI) rather than electrospray ionization.8 This is because APcI is a harder ionization mode than electrospray ionization (ESI), with less gas-adduction and clustering,9 and produces less multiple charging. Consequently, APcI is more likely to generate an obvious [M + H]+ pseudomolecular ion, (1) Bauer, B. E. Molecular Diversity and Combinatorial Chemistry: Libraries and Drug Discovery; Chaiken, I. M., Janda, K. D., Eds.; American Chemical Society: Washington, DC, 1996; pp 233-243. (2) Bray, A. M.; Chiefari, D. S.; Valerio, R. M.; Maeji, N. J. Tetrahedron Lett. 1995, 36, 5081-5084. (3) Bray, A. M.; Valerio, R. M.; Dipasquale, A. J.; Greig, J.; Maeji, N. J. J. Pept. Sci. 1995, 1, 80-87. (4) Parrott, N. Henry Stewart Conference Studies: Latest advances in data management for combinatorial chemistry and high throughput screening; London, 1996; pp 18-36. (5) Hegy, G.; Go¨rlach, E.; Richmond, R.; Bitsch, F. Rapid Commun. Mass Spectrom. 1996, 10, 1894-1900. (6) Smart, S. S.; Mason, T. J.; Bennell, P. S.; Maeji, N. J.; Geysen, H. M. Int. J. Pept. Protein Res. 1996, 47, 47-55. (7) Pullen, F. S.; Richards, D. S. Rapid Commun. Mass Spectrom. 1995, 9, 188190. (8) Taylor, L. C. E.; Johnson, R. L.; Raso, R. J. Am. Soc. Mass Spectrom. 1995, 6, 387-393. (9) Garcia, D. M.; Huang, S. K.; Stansbury, W. F. J. Am. Soc. Mass Spectrom. 1996, 7, 59-65.
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998 3227
Figure 1. RackViewer route map, showing the interrelationships of the various RackViewer displays.
which facilitates spectral interpretation for the synthetic chemists.10 However, for the purity screening of combinatorial chemistry samples, the gas clustering of electrospray ionization is preferable to the risk of unanticipated fragmentation induced by APcI, especially as the very large sample numbers effectively prevent comprehensive manual inspection of the spectra. Therefore our method of choice is ESI, without any up-front or in-source collisioninduced dissociation declustering. The entire analytical process encompasses automated input of sample information, setup of mass spectrometer analysis parameters, purity estimation versus an expected compound, and subsequent reporting to the customers. This information is used by the combinatorial chemists, first, to improve syntheses during the optimization phase, and second, to characterize the quality of compound collections subject to biological screening, providing a sound foundation for the design of optimized libraries.2 The paramount task of the Novartis Pharma MS scientists was to deliver these purity estimates in a quick and clear way to the geographically spread out combinatorial chemistry customers (via their networked PCs). Mass spectroscopy analysis of the customer’s 96-well Micronic rack samples generates a wealth of chemical information. However, to avoid data interpretation fatigue, color rendering of the data within a Visual Basic interface program is employed,5 with numerical purities relegated to the (accessible) background. Speed of delivery is also critical due to the emphasis on a decreased analytical cycle time concomitant with very large sample numbers. Approximately 65 000 samples have now been measured using our high-throughput FIA-MS (10) Pullen, F. S.; Perkins, G. L.; Burton, K. I.; Ware, R. S.; Teague, M. S.; Kiplinger, J. P. J. Am. Soc. Mass Spectrom. 1995, 6, 394-399.
3228 Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
system,5 and we report here recent improvements stemming from experience of this sample load. EXPERIMENTAL SECTION Analytics. Two SSQ 7000 mass spectrometers from Finnigan MAT (San Jose, CA) were used in these experiments. Each is controlled by a DEC AXP 3000/300 workstation under the OSF/1 V3.2D operating system running the Finnigan Interactive Chemical Information System (ICIS) V8.2.2 application software. Two Ethernet interfaces connected each workstation to a SSQ 7000 and the Novartis local area network, respectively. Both workstations shared a common disk area as a network file system. A direct access from the chemists’ PCs running Microsoft Windows to the workstations was not possible because they were only connected to the services of a DEC PATHWORKS V4.1 network. Thus, a dedicated file service was set up on this network. The workstations in the MS laboratory exchanged files with that area through TCP/IP (File Transfer Protocol). The Finnigan ICIS package controls a Hewlett-Packard 1090 HPLC through a HP-IB interface and the autosampler via the RS-232C interface. The single 96-well rack capacity CTC A200S autosampler was replaced by the new six-rack capacity CTC HTS PAL model (CTC Analytics, CH-4222 Zwingen, Switzerland). The HTS PAL autosampler is run by the same Finnigan software driver as for the CTC A200S. The initial CTC HTS PAL installation software (v1.3.108) was replaced by a subsequent CTC version (v1.3.135) enabling the autosampler six ports Valco valve inject/load orientation to be verified prior to every injection. The 96-well Micronic Blueline racks were from Micronic BV, P.O. Box 604, 8211 Ap Lelystad, The Netherlands.
Figure 2. Overview showing % RICsum purity colorization. This is the default depiction on entering RackViewer. The cursor-tracking data box corresponding to well B7 is illustrated. A Micronics rack has an 8 rows by 12 column format; but an internal Novartis Pharma convention is that the first row is left free.
The mass spectrometer operating conditions in ESI positive and negative modes were the following: heated capillary temperature of 220 °C; conversion dynode at 15 kV; electron multiplier at 1.3 kV, a collision-induced dissociation offset of 0.0 V and a spray voltage of 4.5 kV. The flow injection analysis mobile phase was acetonitrile/water 70:30 (v/v), with the gradient grade acetonitrile and the water for chromatography from E. Merck (D64271, Darmstadt, Germany). The flow rate was 50 µL/min. Normally, 2 µL of sample was injected. The 40 cm long electrospray fused silica capillary was changed from the original 50 to 100 µm i.d. to minimize clogging (Polymicro Technologies, Phoenix, AZ 85023-1200, part numbers TSP050192 and TSP100200. respectively). The RackViewer application was developed using MS Visual Basic V5.0 for the 32-bit environment of Microsoft Windows NT. All the x-y graphics in RackViewer, e.g., spectra and 3-D maps, are built from peak lists using a commercial graphics control (ProEssentials from Gigasoft Inc., Keller, TX 76248) instead of importing Hewlett-Packard Graphics Language (HPGL) graphics.5 Combinatorial Chemistry. Syntheses of tripeptoids were carried out on 7 µmol of hexamethylenediamine methacrylic acid dimethylacrylamide Macro Crowns of Chiron Technologies functionalized with the fluorenylmethyloxycarbonyl (FMOC) protected Rink amide linker. The FMOC protecting group was removed with 20% piperidine in dimethylformamide (DMF). Acylation reactions were performed with stirring for 18 h by adding a
premixed DMF solution containing 0.5 M iodoacetic acid and 0.5 M N,N′-diisopropylcarbodiimide. Displacement reactions were performed with stirring for 4 h by addition of a primary or secondary amine as 0.5 M solutions in dimethyl sulfoxide. After twice repeating these reaction cycles, side-chain deprotection and cleavage of the oligomer from the solid support was performed using 50% trifluoracetic acid in methylene chloride.11 Solvent was removed in vacuo using a Genevac Atlas HT SpeedVac. Samples were then redissolved in acetonitrile/water 70:30 (v/v) prior to MS analysis. RESULTS AND DISCUSSION General Rackviewer Design Changes. In the previous version of the system, spectra were thresholded at 5% relative abundance (RA) for both generating the HPGL files and the purity calculations on the UNIX system. Now a peak list of the automatically background subtracted spectrum i.e., m/z, % RA, thresholded at 1% is used for the purity calculations, then transferred to the file server for display of the spectrum via RackViewer. Thresholding is a critical parameter and if set too low, e.g., 0.1%, could cause slowness in graphical display of data over the network, and if set too high can cause inaccuracies in the purity calculations. (11) Zuckermann, R. N, Kerr, J. M.; Kent, S. H. B.; Moos, W. H. J. Am. Chem. Soc. 1992, 114, 10646-10647.
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
3229
Figure 3. 3-D map with no normalization to expected monoisotopic molecular weight, sorted by row and colorized to mass, e.g., red ) forbidden zones. It can be switched from non-normalized to normalized by selecting the check box “normalized to mw”. At a threshold ) 15% RA, one ion at m/z ) 224 runs in all wells. At threshold values of