Anal. Chem. 1998, 70, 865-872
Multistage Accurate Mass Spectrometry: A “Basket in a Basket” Approach for Structure Elucidation and Its Application to a Compound from Combinatorial Synthesis Qinyuan (Quincey) Wu*
Selectide, a Subsidiary of Hoechst Marion Roussel, Inc., Tucson, Arizona 85737
A “basket in a basket” method based on a multistage accurate mass spectrometric (MAMS) technique was developed and demonstrated by obtaining a unique elemental composition of a compound (with a molecular weight of 517) from combinatorial synthesis. The accurate masses for the parent and the fragment ions were obtained with up to five stages of MAMS using Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS). This approach requires only input of elements used in the synthetic processes and some constraints about unusual light elements, such as fluorine, while the compositions of the parent ions and their fragments are obtained for structure elucidation. Conversely, accuracy of better than 0.02 ppm (assuming elements C, H, N, O, S, and F are involved) would be required in order to define a unique composition for the same mass using a direct accurate mass measurement because the number of possible elemental compositions increases sharply as the mass increases. Similarly, due to the uncertainty in determining elemental compositions of fragments and complexity of possible internal fragmentation, tandem mass spectrometry may not provide enough information for structure elucidation of unknown compounds, especially of the organic molecules in the mass range of 300-1000 Da, typically encountered in combinatorial lead generation. The application of MAMS to combinatorial drug discovery is particularly advantageous since the built-in chemical information from the synthesis can be used as constraints. The implementation of a nanoelectrospray ionization technique makes this approach practical for characterization of small quantities of compounds typically available from lead generation processes. Structure elucidation constitutes an important aspect of analytical chemistry. In the pharmaceutical industry, structure identification of biologically potent compounds either from natural products or from combinatorial libraries1,2 is directly related to the throughput of lead generation and drug discovery processes. Extracts from natural products often generate compounds in * For correspondence: Selectide/HMR, 1580 E. Hanley Blvd., Tucson, AZ 85737. S0003-2700(97)01132-3 CCC: $15.00 Published on Web 01/28/1998
© 1998 American Chemical Society
limited quantities, while an active compound from combinatorial synthesis may be a side-product of the chemical reactions or may involve a large number of structural possibilities if a deconvolution method is not included in the synthesis. Mass spectrometry has been recognized as an efficient technique for these applications due to its superior speed, high sensitivity, fragmentation capability in the gas phases, and compatibility with separation techniques in solution. Although the structure of compounds from combinatorial synthesis can often be confirmed by measuring its molecular weight if chemical reactions proceed as expected, structure elucidation using mass spectrometry is often the first choice, if not the only one, in cases where reaction side-products are involved. Mass spectrometry also plays a key role in characterizing unknown lead compounds extracted from natural products. It is obvious that the more accurate the structural assignment is, the less work will be needed in the usually more time-consuming chemical synthesis of potential leads or reextraction of the natural products. Molecular structures can be difficult to determine directly by using mass spectrometry. Instead, the masses of the corresponding ions and their fragments are measured, which can often be used to confirm the structures from a set of known choices, such as the proposed structures from chemical synthesis. In addition, fragmenting ions in the gas phases using tandem mass spectrometry3-5 usually reveals sequence information of the corresponding molecules, which can be used to elucidate their structures. Up to 10 stages of fragmentation and chemical reaction in the gas phase have been demonstrated for small molecules using an ion trap mass spectrometer.6 MS4 experiments have also been demonstrated for proteins7 and for small peptides (using a (1) Geysen, M. H.; Rodda, H. M.; Mason, T. J. In Synthetic Peptides as Antigens; Porter R., Wheelan, J., Eds.; Ciba Foundations Symposium 119; Wiley: New York, 1986. (2) Lam, K. S.; Lebl, M.; Krchnak, V. Chem. Rev. (Washington, D.C.) 1997, 97, 411-448. (3) Tandem Mass Spectrometry; McLafferty, F. W., Ed.; Wiley: New York, 1983. (4) Cooks, R. G. In Collision Spectroscopy; Cooks, R. G., Ed.; Plenum: New York, 1978. (5) Busch, K. L.; Glish, G. L.; McLuckey, S. A. Mass Spectrometry/Mass Spectrometry: Techniques and Application of Tandem Mass Spectrometry; VCH: New York, 1988. (6) Nourse, B. D.; Cox, K. A.; Morand, K. L.; Cooks, R. G. J. Am. Chem. Soc. 1992, 114, 2010-2016. (7) Wu, Q.; Van Orden, S.; Bakhtiar, R.; Cheng, X.; Smith, R. D. Anal. Chem. 1995, 67, 2498-2509.
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998 865
quadrupolar axialization technique)8 using Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS). As an alternative to tandem mass spectrometry, the masses of ions can be measured to such an accuracy that a unique elemental composition can be found.3,9-11 One of the obvious prerequisites of obtaining such a high-accuracy measurement of an unknown sample is that high mass resolution be achieved to avoid interference from other species. Accurate mass measurements of combinatorial peptide libraries and proteins have been demonstrated recently with high resolution using FTICR-MS.12-14 Ultrahigh-resolution data have been obtained at various magnetic field strengths,15-18 and high mass accuracy has been shown for both parent and their fragment ions using either internal or external calibration methods.7,12,19 High-sensitivity FTICR results have also been reported by coupling to nanoelectrospray ionization (ESI), matrix-assisted laser desorption/ionization (MALDI), and capillary electrophoresis.20-22 By using current techniques with a mass accuracy of better than 2 ppm, unique empirical formulas can often be found for small molecules (C2OF3 (TFA)d
C7H8N C4H9NOF C12H12N7O2 C16H16NO4 C9H13N7O3F C14H15NO2F3 C11H16NO3F4
C7H8N C4H9NOF C12H12N7O2 C16H16NO4 C9H13N7O3F C14H15NO2F3 C11H16NO3F4
C17H17NFS C14H18NOF2S
C17H17NFS C14H18NOF2S
>C8H6Fe C7H8N C12H12N7O2 C16H16NO4 C14H15NO2F3 C11H16NO3F4 C17H17NFS
a The elemental compositions of the previous stage are used as a constraint. b The differences between columns 2 and 3. c Compositions without impossible elemental ratios in column 4 (see text). d Those compositions in column 5 that also contain this combination in their fragments (column 4) if the number of F is higher than 3. e Those compositions in column 6 that also include this combination in their fragments (column 4) if it contains F.
For the measured mass of 106.0658 Da at MS5, the maximum numbers of C, H, N, O, F, and S atoms in the formula were calculated to be 9, 15, 6, 6, 4, and 3, respectively, from the corresponding hypothetical structures, i.e., (-CtC-CtC-), (-CH2-CH2-), (-NH-NH-), (-O-O-), (-CF2-CF2-), and (-S-S-). Based on these and a mass tolerance of 15 ppm, a computer search found only two matching compositions: C7H8N and C4H9NOF. As shown in Table 1, the second composition is actually in conflict with the constraint that a composition should at least include C8H6F, if it were to involve F. Therefore, the composition should be C7H8N, which is within 6 ppm of the measured value. It must be noted that a mass tolerance of about twice the experimental uncertainty (8.5 ppm) was used here (i.e., at 95% confidence level), and for the rest of the example, in searching for the matching compositions. In routine operation, where sample and time limitations may not allow optimization of experimental conditions for the best accuracy, it is obviously advantageous to consider a larger group of possible compositions with the “correct” compound rather than to consider only a smaller one which might exclude it. Similarly, a computer search for the mass 286.1061 Da (MS4) with a 5 ppm tolerance resulted in 12 possibilities, consisting of C7H8 from the MS5. Table 1 also listed the differences between these compositions and C7H8N. An examination of these composition differences quickly ruled out several compositions consisting of impossible elemental ratios:24 for example, composition CH9O4F9 included more monovalent atoms (H and F) than the maximum number of covalent bonds for a “CO4” group. As a result, seven compositions were left for further consideration, of which five remained possible when the constraints for fluorine were applied. The same procedure was applied for the MS3 ions using the five unresolved compositions from the MS4 stage (see Table 2). Four elemental compositions were found within a 5 ppm mass tolerance, of which only one had a reasonable elemental ratio when the composition differences were considered. Thus, the ions detected at MS3 were determined to be C20H25N2O4 with 1.1 ppm accuracy, and the corresponding C16H16NO4 composition was 870 Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
the only choice for MS4. This also led to an unambiguous assignment of C21H25N2O5 for the MS2 ions. Finally, although there were 11 possible elemental compositions for the parent ions based on the known composition for MS2, most of them were composed of impossible elemental ratios, as shown in Table 2. Therefore, the composition of the parent ions was determined to be C26H37N4O7, with an accuracy of 2.9 ppm. Since the elemental compositions for each of the fragments were also resolved (Table 3) and were found to match the building blocks used for the randomization steps and the linker, the structure of this compound was readily elucidated. The fragment at the MS2 stage resulted from cleavage of the amide bond between the linker and the “R1” group (see above), which lost a “CO” during MS3. At the fourth and the fifth stages, the second amide bond and the ether bond were cleaved, respectively. This result is in significant contrast to direct accurate measurements, which, even with unrealistic 0.02 ppm mass accuracy, would still generate two compositions: C26H37N4O7 (517.265 677 Da) and C17H47N6OF2S4 (517.265 671 Da) for the same mass and the same kinds of elements considered. Moreover, the constraints on the elemental ratios and on the fluorine-containing compounds cannot be used to rule out either one of these two possibilities because of the large number of possible structural isomers at higher masses. These constraints generally work more effectively for lower masses, which is similar to the effects of mass tolerance on the number of elemental compositions shown in Figure 1. As the result, there would be 115 unresolved elemental compositions within 1 ppm for the same parent ions considering all the elements C, H, N, O, F, and S. Although mass accuracy of better than 2 ppm can be achieved with optimization of experimental conditions12,25 and with a better understanding of the parameters that govern the mass accuracy measurement, it would require an accuracy of better than 0.02 ppm in order to define a unique composition using a direct accurate mass measurement. Conversely, MAMS is a “basket in a basket” approach, where multiple constraints can be applied toward obtaining elemental compositions at each stage of mass spectrometry. As illustrated in Table 3, not only the composition of the smaller fragments (the
Table 2. Elemental Compositions for the Third to the First Stages of MS MSn 3
2 1
possible compositionsa
previous stage
C20H25N2O4 C13H22N8O3F C13H25N8O2S C15H28N2O2F3S none none C21H25N2O5 C22H33N10O5 C26H37N4O7 C21H37N6O9 C23H38N4O8F C21H35N8O5F2 C24H36N4O5F3 C21H37N4O6F4 C23H41N4O7S C22H38N6O5FS C21H40N4O5F3S C21H45N2O8S2
C16H16NO4 C12H12N7O2 C12H12N7O2 C14H15NO2F3 C17H17NFS C11H16NO3F4 C20H25N2O4 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5 C21H25N2O5
elemental ratioc
>C2OF3 (TFA)
>C8H6Fe
C4H9N CH10NOF CH13NS CH13NS
C20H25N2O4
C20H25N2O4d
C20H25N2O4
CO CH8N8 C5H12N2O2 H12N4O4 C2H13N2O3F H10N6F2 C3H11N2F3 H12N2OF4 C2H16N2O2S CH13N4FS H15N2F3S H20O3S2
C21H25N2O5
C21H25N2O5
C21H25N2O5
C26H37N4O7
C26H37N4O7
C26H37N4O7
composition differenceb
a The elemental compositions of the previous stage are used as a constraint. b The differences between columns 2 and 3. c Compositions without impossible elemental ratios in column 4 (see text). d Those compositions in column 5 that also contain this combination in their fragments (column 4) if the number of F is higher than 3. e Those compositions in column 6 that also include this combination in their fragments (column 4) if it contains F.
Table 3. Results of a Multistage Accurate Mass Measurement
MS
measured mass (Da)a
elemental compositionb
calcd mass (Da)
accuracy of the measurement (ppm)
1 2 3 4 5
517.2642 ( 0.0021 (4.0 ppm) 385.1721 ( 0.0005 (1.3 ppm) 357.1805 ( 0.0006 (1.7 ppm) 286.1061 ( 0.0005 (1.7 ppm) 106.0658 ( 0.0009 (8.5 ppm)
C26H37N4O7 C21H25N2O5 C20H25N2O4 C16H16NO4 C7H8N
517.2657 385.1758 357.1809 286.1074 106.0651
2.9 9.6 1.1 4.5 6.4
a The average monoisotopic mass from 9, 5, 5, 5, and 3 experiments, respectively, for the first to the fifth stages of MS. Also indicated are the standard deviations for each stage. b See text and the preceding tables for details.
smaller rings in the inserted figure) served to determine those for the larger fragments, but also the composition of these large fragments (larger rings) set the upper limit for the next generation of ions. In this example, the determination of the ions at the MS3 left only one choice for the ions at the MS2, which would otherwise have 30 possible compositions at 5 ppm accuracy, even when fluorine was not involved (see Figure 1). On the other hand, the measured mass at MS3 served to exclude four other compositions at the MS4 stage, just like a distorted small basket would not fit into the normal larger basket. Depending on the relative sizes of these “baskets”, multiple elimination of impossible compositions may also be available so that even if one composition may be
overlooked in one analysis, it will still be excluded by the next step. For example, if the composition of C9H21N3O2FS2 for the MS4 (see Table 1) were to be carried over for the MS3 analysis, an impossible composition difference of CH10NOF would be found, resulting in the dismissal of the related elemental combinations. Such a “basket in a basket” approach also greatly reduces mass accuracy requirements since the original larger ions were dissociated into smaller fragments, which could be determined using only a moderate mass accuracy (see Figure 1). In this example, a mass accuracy of 2-5 ppm was shown to be good enough for composition determination, although relaxing mass accuracy requirements should obviously be dependent on the availability Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
871
of other constraints, such as more stages of MS. Interestingly, this example included the effectiveness of MAMS in an extreme case: the ions at the MS2 stage were measured to a precision of 1.3 ppm with poor accuracy of 9.6 ppm, which may occur under routine operating conditions when special optimization for the best accuracy may not be performed due to time or sample limitations. The high mass error at MS2 could be attributed to the fact that its total ion population was about 50% higher than in the other stages of MS (vide supra). In a separate control experiment, the mass error increased from 2 to 9 ppm when the total ion intensity was increased by 50%, indicating that space charging effects may dominate such measurements at high ion density. On the other hand, significant mass errors were also observed for those ions with a signal-to-noise ratio (S/N) below 5, in which case the peak centroid may not be statistically well defined. In these experiments, lower S/N might be one of the reasons for the decreased mass accuracy for the higher order of MS. Additional uncertainty may also result from the fact that the m/z range of the reference ions used was not particularly optimized for ions at lower m/z. Product ions of the higher order of MS are generated by multiple cycles of collisions with gases, during which significant magnetron motion may build up and ion clouds may be dephased for both the analytes and the references. The loss in mass accuracy for those ions could probably be recovered by the application of the quadrupolar axialization techniques.30,31 Reaxialization and remeasurement31 may also reduce the mass errors for ions with low S/N. A potential limitation of the “basket in a basket” method is that certain information from the starting materials needs to be used as a constraint, if light elements other than C, H, N, and O are involved. As can be seen from Tables 1 and 2, the number of possible elemental compositions would decrease to 1, 3, 1, 1, and 5 for the fifth to the first stages of MS, respectively, if fluorine were not involved. In fact, the 115 possible elemental compositions (vide supra) would be reduced to 16 without fluorine, indicating that this is a common situation for accurate mass measurements. Nevertheless, one of the advantages of MAMS over direct accurate mass measurement is that the constraints for these uncommon light elements can be applied to dramatically reduce the number of possible elemental compositions. Other information potentially useful as constraints include unusual isotopic patterns, such as those from Cl or Br, and information about a particular element, such as the number of protons or carbons from NMR. Multichannel fragmentation may also serve as a constraint, although only one channel of fragmentation was available in this example. The coupling of a reliable nanoelectrospray ionization technique with MAMS is important for its practical usage. In general, the amount of sample available for analytical work is limited, at least in the lead generation stage. There have been many cases (30) Schweikhard, L.; Guan, S. H.; Marshall, A. G. Int. J. Mass Spectrom. Ion Processes 1992, 120, 71. (31) Speir, J. P.; Gorman, G. S.; Pitsenberger, C. C.; Turner, C. A.; Wang, P. P.; Amster, I. J. Anal. Chem. 1993, 65, 1746. (32) Edmondson, R. D.; Gadda, G.; Fitzpatrick, P. F.; Roussel, D. H. Anal. Chem. 1997, 69, 2862-2865.
872
Analytical Chemistry, Vol. 70, No. 5, March 1, 1998
where biological activities came from unexpected products of the combinatorial synthesis. It is important to notice that, given a common threshold of biological activity, the actual potency from a minor component may be several times higher than that if a major component is responsible for the activity. In such cases, structural determination becomes extremely important, while the amount of sample available for analysis is usually small. In the above experiments from MS2 to MS5, a total of 3 µL of the solution was loaded into the flow injection source, i.e., the amount of the analyte compound injected was ∼30 pmol. Such experiments would be impossible without the nanoelectrospray technique. CONCLUSION A “basket in a basket” method based on a multistage accurate mass spectrometric (MAMS) technique was developed and demonstrated by obtaining unique elemental compositions for a compound (MH+ ) 517 Da) and its fragments from four stages of dissociation. The structure of this compound was readily elucidated on the basis of these results and the built-in chemical information from the combinatorial synthesis. This approach is useful for structure analysis of unknown compounds since it only requires input of the elements used (in combinatorial synthesis) and some constraints about unusual light elements. Obviously, this method can also be extended to structure elucidation of other unknown compounds, such as extracts from natural products. In these cases, the elements involved may be determined using techniques such as atomic absorption spectroscopy (AA) and inductively coupled plasma mass spectrometry (ICP-MS). An extension of such a “basket in a basket” approach is identification of compound modifications from a known structure, as reported recently.32 However, MAMS can be expected to yield more structural information, such as the site(s) of the modification, especially when multiple variations are involved. Due to the availability of complementary information from different stages of MS in this method, structural assignment is less dependent on the mass accuracy, which should tolerate less optimized experimental conditions and allow higher throughput analysis. This is particularly amenable to routine applications where optimization toward the best accuracy may not be practical due to sample availability, time constraints, or even instrumental capability. Such an approach can also be implemented with other types of instruments, such as an ion trap mass spectrometer, which is known to have multistage fragmentation capabilities but generally shows moderate mass accuracy. With the continued improvement in FTICR-MS, an external calibration method may also be used for multistage accurate mass spectrometry. ACKNOWLEDGMENT I would like to thank Dr. Nikolai Sepetov for helpful suggestions. I also thank Luwei Zhao for synthesizing the compound, and Nina Ma, Olga Issakova, Magda Stankova, and Shelly Wade for useful discussions. Received for review December 8, 1997. AC971132M
October
13,
1997.
Accepted
