Anal. Chem. 1997, 69, 2159-2164
Pulsed Ultrafiltration Mass Spectrometry: A New Method for Screening Combinatorial Libraries Richard B. van Breemen,* Chao-Ran Huang, Dejan Nikolic, Charles P. Woodbury, Yong-Zhong Zhao, and Duane L. Venton
Department of Medicinal Chemistry and Pharmacognosy, College of Pharmacy, University of Illinois at Chicago, 833 South Wood Street, M/C 781, Chicago, Illinois 60612-7231
In response to the need for rapid screening of combinatorial libraries to identify new lead compounds during drug discovery, we have developed an on-line combination of ultrafiltration and electrospray mass spectrometry, called pulsed ultrafiltration mass spectrometry, which facilitates the identification of solution-phase ligands in library mixtures that bind to solution-phase receptors. After ligands contained in a library mixture were bound to a macromolecular receptor, e.g., human serum albumin or calf intestine adenosine deaminase, the ligandreceptor complexes were purified by ultrafiltration and then dissociated using methanol to elute the ligands into the electrospray mass spectrometer for detection. Ligands with dissociation constants in the micromolar to nanomolar range were successfully bound, released, and detected using this method, including warfarin, salicylate, furosemide, and thyroxine binding to human serum albumin, and erythro-9-(2-hydroxy-3-nonyl)adenine binding to calf intestine adenosine deaminase. Repetitive bind-and-release experiments demonstrated that the receptor could be reused. Thus, pulsed ultrafiltration mass spectrometry was shown to provide a simple and powerful new method for the screening of combinatorial libraries in support of new drug discovery. During classical drug discovery and development, drug candidates are typically synthesized, purified, characterized, and tested for a given biological activity one by one. Although significant increases in efficiency may be gained by simultaneously producing many compounds in so-called combinatorial libraries1-3 and testing them as a group, a difficult challenge has been the screening of these libraries to detect molecules with a desired biological activity or binding profile that is specific for a given receptor. Few methods exist which allow both receptor and ligands to be screened in solution so as to preserve completely their native binding interactions,4,5 and even fewer solution-phase methods exist which permit the receptor to be reused. Presently, * E-mail:
[email protected]. (1) Thompson, L. A.; Ellman, J. A. Chem. Rev. 1996, 96, 555-600. (2) Gordon, E. M.; Gallop, M. A.; Patel, D. V. Acc. Chem. Res. 1996, 29, 144154. (3) Desai, M. C.; Zuckermann, R. N.; Moos, W. H. Drug Dev. Res. 1994, 33, 174-188. (4) Kaur, S.; Huebner, V.; Tang, D.; McGuire, L.; Drummond, R.; Csetjey, J.; Stratton-Thomas, J.; Rosenberg, S.; Figliozzi, G.; Banville, S.; Zuckermann, R.; Dollinger, G. Proceedings of the 43rd ASMS Conference on Mass Spectrometry and Allied Topics, Atlanta, GA, May 21-26, 1995; p 30. (5) Chu, Y. H.; Kirby, D. P.; Karger, B. L. J. Am. Chem. Soc. 1995, 117, 54195420. S0003-2700(97)00132-7 CCC: $14.00
© 1997 American Chemical Society
most screening approaches use classical bioassays, which test compounds singly but which carry out multiple assays in parallel in an approach called high-throughput screening. Other frequently reported library screening methods are affinity-based and use either immobilized ligands or immobilized receptors.6-9 However, these methods suffer from the fact that immobilization may change the affinity characteristics of the bound species from the native, solution-phase form. Based on an extension of our method for pulsed ultrafiltration measurement of affinity constants for ligand-receptor bonding,10 we are developing an on-line combination of ultrafiltration and electrospray mass spectrometry which facilitates the identification of solution-phase ligands that bind with affinity to specific solution-phase receptors and which permits reuse and/or recovery of the receptor molecules. To demonstrate pulsed ultrafiltration mass spectrometry as a method for screening combinatorial libraries, a ligand that binds with high affinity (i.e., Ka ≈ 108 M-1) to a model receptor, adenosine deaminase, was identified in a 20 compound library containing analogs of adenosine. In addition, ligands of moderate affinity (Ka ≈ 105 M-1) for human serum albumin (an important carrier protein for drugs in the blood) were bound from a mixture and released. A key enzyme in purine metabolism, adenosine deaminase, catalyzes the hydrolysis of adenosine or 2′-deoxyadenosine to its corresponding inosine product and ammonia. Adenosine deaminase is present in almost all mammalian tissues, with particularly high activities in lymphoid tissues,11 and its deficiency is associated with severe immunodeficiency disease.12 Adenosine analogs such as ara-A, which are used as anticancer and antiviral agents, are deaminated by adenosine deaminase to less active products.13 Therefore, coadministration of inhibitors of adenosine deaminase can maintain the potency of these drugs. In addition, many adenosine deaminase inhibitors themselves (6) Bunin, B. A.; Plunkett, M. J.; Ellman, J. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 4708-4712. (7) Kelly, M. A.; Liang, H.; Sytwu, I. I.; Vlattas, I.; Lyons, N. L.; Bowen, B. R.; Wennogle, L. P. Biochemistry 1996, 35, 11747-11755. (8) Youngquist, R. S.; Fuentes, G. R.; Lacey, M. P.; Keough, T. J. Am. Chem. Soc. 1995, 117, 3900-3906. (9) Liang, R.; Yan, L.; Loebach, J.; Ge, M.; Uozumi, Y.; Sekanina, K.; Horan, N.; Gildersleeve, J.; Thompson, C.; Smith, A.; Biswas, K.; Still, W. C.; Kahne, D. Science 1996, 274, 1520-1522. (10) Chen, S.; Chen, C. J.; Woodbury, C. P.; Venton, D. L. Abstracts of Papers, 206th National Meeting of the American Chemical Society, Chicago, IL, August 22-27, 1993; ACS: Washington, DC, 1993. (11) Brady, T. G.; O’Donovan, C. I. Comp. Biochem. Physiol. 1965, 14, 101120. (12) Giblett, E. R.; Anderson, J. E.; Cohen, F.; Polara, B.; Meuwissen, H. J. Lancet 1972, 2, 1067-1069. (13) Agarwal, R. P. Pharmacol. Ther. 1982, 17, 399-429.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997 2159
show antitumor and antiviral activities.14 Another promising therapeutic application of adenosine deaminase inhibitors is their use in preventing free radical-mediated ischemic injury to the reperfused heart.15 By preventing deamination of adenosine in the ischemic heart, the formation of xanthine substrates for xanthine oxidase is inhibited, which in turn prevents the burst of damaging free radicals upon reperfusion of the tissue.16 The adenosine deaminase inhibitor, erythro-9-(2-hydroxy-3-nonyl)adenine (EHNA), has shown efficacy in reducing ischemic heart damage.17 An example of a potent reversible inhibitor of adenosine deaminase, the (+) isomer of EHNA, shows a Ki of 1.7 nM and a Kd of 1.9 nM.17-19 For comparison, the Ki and Kd of purine riboside are both 4 µM.19 In this first report of pulsed ultrafiltration mass spectrometry for screening combinatorial libraries, a fully characterized library of adenosine deaminase analogs and known ligands of serum albumin was screened in order to illustrate the use and function of this new method. In addition, a preliminary report on the use of pulsed ultrafiltration mass spectrometry for the measurement of classical equilibrium binding constants, including binding to human serum albumin, has been presented.20 Other mass spectrometry-based methods for measuring binding to serum albumin include the method of Greig et al.21 for measuring dissociation constants for oligonucleotide-serum albumin complexes and the immobilized serum albumin liquid chromatography mass spectrometry approach of Tiller et al.22 EXPERIMENTAL SECTION A liquid chromatograph-electrospray mass spectrometer (LCMS) was used as the screening apparatus, except that an ultrafiltration chamber was substituted for the HPLC column. The ultrafiltration chamber consisted of a 1.0 in. diameter in-line solvent filtration unit (Upchurch Scientific, Oak Harbor, WA; Catalog No. A-333) in which the filter disk had been replaced by a 10 000 molecular weight cutoff ultrafiltration membrane (Amicon; Beverly, MA; Catalog No. YM10) or a comparable home-built chamber that was mechanically stirred. The volume of each ultrafiltration chamber was approximately 100 µL. Mass spectra were obtained using a Hewlett-Packard (Palo Alto, CA) 5989B MS Engine quadrupole mass spectrometer equipped with a 1090L gradient HPLC system. The quadrupole analyzer was maintained at 120 °C, and the range m/z 50-400 was scanned over approximately 2 s at unit resolution. A dwell time of 5 or 8 s/ion was used during selected ion monitoring measurements. Nitrogen gas at 80 psi was used for nebulization of the electrospray, and nitrogen at 300 °C with a flow rate of 10 L/min also served as a bath gas for solvent evaporation. To enhance formation of protonated molecules during positive ion electrospray, the eluate from the (14) Glazer, R. I. Cancer Chemother. Pharmacol. 1980, 4, 227-235. (15) Sandhu, G. S.; Burrier, A. C.; Janero, D. R. Am. J. Physiol. 1993, 265, H1249-H1256. (16) Xia, Y.; Khatchikian, G.; Zweier, J. L. J. Biol. Chem. 1996, 271, 1009610102. (17) Schaeffer, H. J.; Schwender, C. F. J. Med. Chem. 1974, 17, 6-8. (18) Porter, D. J. T.; Abushanab, E. Biochemistry 1992, 31, 8216-8220. (19) Frieden, C.; Kurz, L. C.; Gilbert, H. R. Biochemistry 1980, 19, 5303-5309. (20) van Breemen, R. B.; Huang, C.-R.; Nikolic, D.; Woodbury, C. P.; Zhao, Y.Z.; Venton, D. L. Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, May 12-16, 1996; p 1032. (21) Greig, M. J.; Gaus, H.; Cummins, L. L.; Sasmor, H.; Griffey, R. H. J. Am. Chem. Soc. 1995, 117, 10765-10766. (22) Tiller, P. R.; Mutton, I. M.; Lane, S. J.; Bevan, C. D. Rapid Commun. Mass Spectrom. 1995, 9, 261-263.
2160
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
ultrafiltration chamber was acidified prior to entering the electrospray ion source by addition of methanol/water/acetic acid (49: 49:2 v/v) at a flow rate of 5 µL/min. During negative ion electrospray, postchamber addition of a 5% aqueous ammonia solution in methanol/water (1:1 v/v) at 10 µL/min was used to enhance the formation of deprotonated molecules. Human serum albumin and all library compounds were purchased from Sigma Chemical Co. (St. Louis, MO). EHNA was obtained as a racemic mixture. Calf intestine adenosine deaminase was purchased from Boehringer Mannheim (Indianapolis, IN), and enzyme concentration was determined using the bicinchoninic acid method.23 Positive ion matrix-assisted laser desorption/ionization (MALDI) time-of-flight mass spectrometry (Hewlett-Packard Co.) in a matrix of sinnapinic acid produced a broad, singly charged molecular ion at m/z 41 250 ( 50 (possibly broadened by heterogeneity of the enzyme and combinations of matrix adducts and/or cationization by protons, sodium, or potassium ions). This molecular weight of the native enzyme was in general agreement with that of Kelly et al.,24 who used electrospray mass spectrometry to obtain an average molecular weight of 40 549 for reduced and alkylated bovine adenosine deaminase. An equimolar library of 19 adenosine analogs (17.5 µM each) plus the natural substrate for adenosine deaminase, adenosine, was prepared in 50 mM potassium phosphate buffer, pH 7.5 (see list of compounds in Table 1). The enzyme inhibitor, EHNA, was included in the library at a 10-fold lower concentration than all other compounds in order to bias the screening assay against its detection and to emphasize the sensitivity and specificity of the screening method. The library was mixed with 2.1 µM calf intestinal adenosine deaminase and incubated a minimum of 15 min at room temperature before being injected into the ultrafiltration chamber. For each screening experiment, a mixture containing 420 pmol of adenosine deaminase, 350 pmol of EHNA, and 3.5 nmol of each library compound was injected into the ultrafiltration chamber. After the chamber was flushed with water for 8 min at 50 µL/min (which eluted more than 98% of the unbound compounds from the ultrafiltration chamber and reduced their concentration to background levels in the electrospray mass spectra), the mobile phase was changed to methanol/water (50: 50 v/v) to dissociate the enzyme-ligand complex and thereby release bound ligands for identification by electrospray mass spectrometry. As another screening example, an equimolar (0.4 µM) mixture of ascorbic acid, furosemide, salicylate, thyroxine, tryptophan, and warfarin in 33 mM ammonium acetate, pH 7.5, was infused for 30 min at 100 µL/min (1.2 pmol total of each compound) through the ultrafiltration chamber containing 15 nmol of human serum albumin (150 µM in the ultrafiltration chamber). After a 5 min wash with deionized water at 100 µL/min to remove more than 99% of the unbound compounds from the ultrafiltration chamber, bound ligands were eluted into the mass spectrometer for identification using 60% methanol at 80 µL/min. Deprotonated molecules were detected using selected ion monitoring with negative ion electrospray mass spectrometry. (23) Smith, P. K.; Krohn, R. I.; Hermanson, G. T.; Mallia, A. K.; Gartner, F. H.; Provenzano, M. D.; Fujimoto, E. K.; Goeke, N. M.; Olson, B. J.; Klenk, D. C. Anal. Biochem. 1985, 150, 76-85. (24) Kelly, M. A.; Vestling, M. M.; Murphy, C. M.; Hua, S.; Sumpter, T.; Fenselau, C. J. Pharm. Biomed. Anal. 1996, 14, 1513-1519.
Table 1. Library of Adenosine Analogs for Screening Using Adenosine Deaminase ligand
[M + H]+ m/z
ligand
[M + H]+ m/z
EHNA adenine adenosine adenine-9-R-D-arabinofuranoside 2′-AMP
278 136 268 268
hypoxanthine (-)-inosine 5′-IMP (-)-2′-deoxyinosine
137 269 349 253
348
333
3′-AMP 5′-AMP
348 348
adenosine-5′carboxylic acid ADP cAMP
282
2′-deoxyinosine 5′monophosphate guanosine guanosine 5′-monophosphate purine
121
428 330
purine riboside adenine N1-oxide
253 152
284 364
Another set of experiments was designed to show how ultrafiltration could be used to extract and concentrate a ligand from a dilute solution and then how the receptor might be reused for additional binding studies. The ultrafiltration chamber was loaded with 420 pmol of adenosine deaminase (4.2 µM in the 100 µL chamber), and EHNA at 10.0 nM was continuously infused in 10 mM ammonium acetate buffer, pH 7.5, at 50 µL/min for 40 min until on-line electrospray mass spectrometric detection indicated that the rate of EHNA elution from the chamber was constant. Bound EHNA was released into the mass spectrometer by eluting the chamber with 30% methanol in water. The binding and release process was repeated a total of three times using the same aliquot of adenosine deaminase. As another example, warfarin (300 nM) in 33 mM ammonium acetate, pH 7.5, was infused for 20 min at 100 µL/min (0.6 nmol total) through the ultrafiltration chamber loaded with 25 nmol of human serum albumin. The warfarin-albumin complex was disrupted by introduction of 60% methanol in water, and the deprotonated molecule (m/z 307) of warfarin was continuously monitored by using negative ion electrospray mass spectrometry. The albumin in the ultrafiltration chamber was reused for two additional binding and release experiments. RESULTS AND DISCUSSION After the molecular weight of adenosine deaminase was determined to be 41 250, a 10 000 molecular weight cutoff ultrafiltration membrane was selected for use during all ultrafiltration experiments so that adenosine deaminase and serum albumin would be retained in the binding chamber while allowing lower molecular weight library compounds (i.e., see Tables 1 and 2) to pass through. The principle of pulsed ultrafiltration screening of combinatorial libraries is shown in Figure 1. During pulsed ultrafiltration, ligand-receptor complexes remain in solution in the ultrafiltration chamber while unbound library compounds and buffer are washed away. In this manner, the ultrafiltration chamber functions as a solution-phase extraction device. Then, the ligand-receptor complex is disrupted (e.g., by addition of organic solvent) so that the ligand(s) is released into the mass spectrometer for identification (Figure 1). As an example, the pulsed ultrafiltration mass spectrometric analysis of warfarin binding to human serum albumin is shown in Figure 2. A dilute solution of warfarin was pumped through the ultrafiltration chamber containing serum albumin. Warfarin
Table 2. Screening of Drug Mixture for Binding to Human Serum Albumin Using Pulsed Ultrafiltration Mass Spectrometry peak area (×105)a
ligand
[M - H]m/z
control
HSA
HSA/control
ascorbic acid tryptophan warfarin furosemide salicylate thyroxine
175 203 307 329 137 775
973 350 nd 77 741 678 10 302 863 1 221 337 10 074 161
840 441 nd 386 590 141 42 096 612 8 824 261 15 062 473
0.86 5.0 4.1 7.2 1.5
a Peak areas were calculated from elution profiles in Figure 3. nd, not detected.
Figure 1. Scheme showing the use of pulsed ultrafiltration mass spectrometry for screening a combinatorial library for compounds that bind to a macromolecular receptor. The receptor is trapped in solution by an ultrafiltration membrane, which allows low molecular weight solution-phase compounds in a “pulse” of a combinatorial library to pass through (top). After unbound compounds are washed away, the “hits” in the library are eluted from the chamber by destabilizing the ligand-receptor complex using methanol, pH change, etc. The wash may be discarded to waste (top) or monitored continuously by the mass spectrometer (bottom). The released ligands are identified using mass spectrometry or tandem mass spectrometry.
was extracted out of the mobile phase and retained in the chamber by binding to solution-phase serum albumin. Finally, the warfarin-albumin complex was disrupted by changing the mobile phase to 60% methanol, and the released warfarin was detected as its deprotonated molecule at m/z 307 using negative ion electrospray mass spectrometry. Nonspecific binding of warfarin to the ultrafiltration membrane and the ultrafiltration chamber was determined by carrying out an identical control experiment without serum albumin in the ultrafiltration chamber (Figure 2A). The difference in area between the peaks shown in Figure 2A (control) and B corresponds to the specific binding of warfarin to serum albumin, and the ratio of the two peak areas is 1:4.8 ((5% SD). Using serum albumin as the receptor molecule, a screening experiment was carried out using an equimolar mixture of thyroxine, salicylate, furosemide, warfarin, tryptophan, and ascorbic acid. Ascorbic acid does not bind to serum albumin and thus served as a negative control. Affinity constants (Ka) for albumin binding of thyroxine,25 salicylate,26 furosemide,27 warfarin,26 and (25) Steiner, R. F.; Roth, J.; Robbins, J. J. Biol. Chem. 1966, 241, 560-567. (26) Brown, K. F.; Crooks, M. J. Biochem. Pharmacol. 1976, 25, 1175-1178. (27) Sebille, B.; Thuaud, N. J. Chromatogr. 1978, 167, 159-170.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
2161
Figure 2. Infusion of a dilute solution of warfarin (300 nM) at 100 µL/min through the ultrafiltration chamber containing (A) no receptor protein as a control or (B) 25 nmol of human serum albumin. After 20 min, the warfarin-albumin complex was disrupted by introduction of 60% methanol in water. The deprotonated molecule of warfarin at m/z 307 was continuously monitored by negative ion electrospray mass spectrometry. The difference in peak area between the control and serum albumin experiments reflects specific binding of warfarin to albumin.
tryptophan28 are 16-, 2.2-, 1.7-, 1.5-, and 0.22 × 105 M, respectively. After infusion of a dilute solution of these compounds through the ultrafiltration chamber, bound ligands were released into the mass spectrometer using 60% methanol. Figure 3 shows the selected ion chromatograms of the deprotonated molecules for the ligand mixture as they eluted from the ultrafiltration chamber. The ratio of peak areas (albumin binding/control without albumin) for each ligand was calculated from data in Figure 3 and is summarized in Table 2. A ratio of 1 indicates that there was no binding to albumin, and a value >1 shows that the ligand was concentrated in the ultrafiltration chamber by binding to serum albumin. Warfarin, salicylate, furosemide, and thyroxine showed binding to serum albumin, but as expected, no binding of ascorbic acid was observed (see comparison of peak areas for nonspecific and specific binding in Table 2). In this competitive binding experiment, no signal was observed for tryptophan, since its affinity for serum albumin is approximately 10-fold lower than those of each of the competing ligands, warfarin, salicylate, and furosemide. Although thyroxine has the highest affinity for serum albumin of the ligands tested, only a small signal was detected for this ligand during the screening experiment. Because thyroxine is virtually insoluble in water, it was first dissolved in methanol before being diluted into the aqueous buffer used for the screening experiment. Although it appeared to remain in solution, thyroxine probably precipitated out of the ligand solution or adsorbed to the glassware during sample handling so that little was available for binding to albumin during the screening experiment. This hypothesis is supported by subsequent LC-MS experiments in which aliquots of freshly prepared aqueous and methanolic solutions of thyroxine were compared with each other (28) Sun, S. F.; Wong, F. Chromatographia 1985, 20, 495-499.
2162 Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
Figure 3. Ultrafiltration mass spectrometric screening experiment showing binding and release of drugs interacting with human serum albumin. A 0.4 µM mixture of ascorbic acid, furosemide, salicylate, thyroxine, tryptophan, and warfarin was pumped through the ultrafiltration chamber containing 15 nmol of human serum albumin. After a 5 min wash with deionized water, bound ligands were eluted into the mass spectrometer for identification using 60% methanol. Deprotonated molecules of each ligand were monitored using negative ion electrospray mass spectrometry. See Table 2 for a list of ligands and their corresponding [M - H]- ions.
and with aged solutions. LC-MS analyses showed a higher concentration of thyroxine in a methanol stock solution compared to a nominally identical aqueous solution. In addition, LC-MS indicated that the concentration of thyroxine in the aqueous solution decreased over several hours. Because of the limited aqueous solubility of many compounds contained in combinatorial libraries, an experiment was carried out to demonstrate how ultrafiltration may be used to extract a high-affinity ligand from a dilute solution and concentrate it on a macromolecular receptor. A dilute solution of EHNA (10 nM) was pumped through the ultrafiltration chamber containing adenosine deaminase until a steady state of ligand binding and dissociation was achieved as monitored on-line using mass spectrometry. Then, elution with 30% methanol disrupted ligandreceptor binding, releasing EHNA for mass spectrometric detection (see Figure 4). Multiple cycles of EHNA binding and methanol release using the same enzyme consistently produced essentially the same mass spectrometer detector responses for protonated EHNA, indicating that adenosine deaminase was not irreversibly denatured by treatment with 30% methanol. Carried out immediately before and after the experiments with adenosine deaminase and EHNA, identical control experiments containing no adenosine deaminase showed a significantly lower background signal for EHNA upon methanol elution (Figure 4). The observa-
Figure 4. Repetitive binding and release of EHNA from 420 pmol of adenosine deaminase (ADA) in the ultrafiltration chamber, demonstrated by monitoring the elution of EHNA using selected ion monitoring electrospray mass spectrometry of the protonated molecule of EHNA at m/z 278. EHNA at 10 nM was continuously pumped through the ultrafiltration chamber until a steady-state level was measured eluting from the chamber. Then, bound EHNA was eluted from the chamber using 30% methanol. This cycle of EHNA extraction and release using methanol was repeated twice more without loss of enzyme affinity for EHNA. The first and the last peaks represent control experiments that were carried out without ADA in the ultrafiltration chamber in order to control for nonspecific binding of EHNA to the ultrafiltration membrane and chamber.
tion of signal at m/z 278 in the control experiment was probably the result of a combination of enhanced electrospray ionization during methanol elution compared to pure aqueous buffer and nonspecific binding of EHNA to the ultrafiltration membrane, tubing, and binding chamber. Overall, these experiments demonstrate that a specific ligand may be extracted from solution onto a solution-phase receptor and released for identification, and then the receptor may be reused for subsequent binding experiments. A library of 19 adenosine analogs and adenosine was incubated with adenosine deaminase in phosphate buffer and then injected into the pulsed ultrafiltration chamber for screening of high-affinity ligands. After the ultrafiltration chamber was flushed with water to remove the nonvolatile binding buffer and unbound or weakly binding library compounds, methanol was introduced into the mobile phase to dissociate the enzyme-ligand complex and release bound ligands for identification by electrospray mass spectrometry. During methanol elution, only EHNA, [M + H]+ of m/z 278, was detected (see mass spectrum in Figure 5), presumably because EHNA has the highest affinity for adenosine deaminase among the library compounds. In control experiments using the library without enzyme, no library compounds were detected during methanol elution (Figure 5, control). It should be noted that additional control experiments verified that all 20 library compounds could be detected by using either positive ion (i.e., bases and nucleosides) or negative ion (i.e., nucleotides) electrospray mass spectrometry. Despite being present at a 10-fold lower concentration than the natural substrate adenosine or the other adenosine analogs, EHNA was easily identified in the 20 compound library. This demonstrates the utility of ultrafiltration electrospray mass spectrometry for identifying a high-affinity ligand among a set of analogs that bind to a specific receptor. Since the use of 30% methanol only reversibly disrupted the receptor-ligand complex, binding of EHNA to adenosine deaminase could be fully restored by returning the enzyme to the original binding buffer conditions.
Figure 5. Identification of EHNA as the highest affinity ligand for adenosine deaminase in a combinatorial library of 20 compounds using ultrafiltration electrospray mass spectrometry. After the nonbinding and weakly bound ligands were flushed out of the ultrafiltration chamber with water, EHNA was eluted by disrupting the EHNAadenosine deaminase complex using methanol. The abundance of protonated EHNA at m/z 278 was 16 times higher than any background ions detected in an identical control experiment using the library but no enzyme. (Compare background ions with library [M + H]+ values in Table 1.)
Although other receptors might not be as stable to methanol as adenosine deaminase, alternative methods may be used to more gently disrupt receptor-ligand binding, such as changes in pH, temperature increase, or exposure to other organic solvents or a competing ligand. Since the mobile phase should be completely volatile and of low ionic strength for optimum compatibility with electrospray mass spectrometry, one is limited to the use of volatile buffers such as ammonium acetate or ammonium formate of less than 200 mM (high ionic strength buffers suppress electrospray ionization). Although such mobile phases might not be ideal for maximum ligand-receptor interaction, this limitation may be overcome by the insertion of an off-line aqueous washing step between the injection of ligand or library in nonvolatile, high ionic strength or otherwise mass spectrometer-incompatible buffers and subsequent methanol disruption of the ligand-receptor complex. In the case of adenosine deaminase, deionized water was used to wash away phosphate buffer and unbound ligand prior to methanol elution of bound ligand and mass spectrometric detection. During the wash period, unbound compounds could be detected by electrospray mass spectrometry, but binding buffer and nonbinding compounds were typically washed to waste so as to minimize contamination of the mass spectrometer. If buffer is required in order to maintain association of the ligand-receptor complex, then a volatile buffer should be used instead of water during the washing step. The ability to screen compounds in solution using solutionphase receptors distinguishes pulsed ultrafiltration mass spectrometry from the majority of affinity screening methods, which utilize immobilized ligands or receptors. For example, affinity columns containing immobilized receptor29 or antibody30 have been used to screen ligand mixtures, and other approaches use (29) Kassel, D. B.; Consler, T. G.; Shalaby, M.; Sekhri, P.; Gordon, N.; Nadler, T. In Techniques in Protein Chemistry VI; Crabb, J. W., Ed.; Academic Press: San Diego, CA, 1995; p 39.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
2163
phage-expressed libraries31 or ligands immobilized on beads.9 No matter which scheme is used, immobilization of ligands or receptors can significantly alter their binding characteristics from the native, solution-phase form. Therefore, various approaches are being developed for screening libraries of compounds entirely in solution. In one solution-phase approach, a series of bioassays using progressively smaller libraries is carried out, until the active compound in the library is identified.32 This method is slow and labor intensive. Alternatively, affinity capillary electrophoresis has been applied to the analysis of a peptide library in which peptides that bind to a solution-phase receptor in the electrophoresis buffer have slower mobility than noninteracting library compounds.5 Identification of each ligand is carried out using on-line electrospray mass spectrometry. Finally, another promising solutionphase screening method uses a size exclusion column to separate ligand-receptor complexes from unbound library compounds, followed by a reversed-phase LC-MS analysis that both dissociates the complex and identifies the released ligand.4 Unlike these
other mass spectrometry-based screening methods, pulsed ultrafiltration mass spectrometry allows the solution-phase receptor to be recovered or reused, which is a distinct advantage when the receptor protein is expensive or in short supply. In addition, only pulsed ultrafiltration mass spectrometry allows library compounds to be extracted from a dilute solution and concentrated onto the receptor molecule, which overcomes common library solubility limitations. Finally, pulsed ultrafiltration mass spectrometry is a new method for screening molecular diversity for the “needle in the haystack” or compound that binds with high affinity to a target receptor.
(30) Zuckermann, R. N.; Kerr, J. M.; Siani, M. A.; Banville, S. C.; Santi, D. V. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 4505. (31) Schumacher, T. N. M.; Mayr, L. M.; Minor, D. L., Jr.; Milhollen, M. A.; Burgess, M. W.; Kim, P. S. Science 1996, 271, 1854-1857. (32) Carell, T.; Wintner, E. A.; Sutherland, A. J.; Rebek, J., Jr.; Dunayevshiy, Y.M.; Vouros, P. Chem. Biol. 1995, 2, 171-183.
Received for review February 3, 1997. Accepted March 21, 1997.X
2164
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
ACKNOWLEDGMENT Use of the electrospray and MALDI time-of-flight mass spectrometers was generously provided by the Hewlett-Packard Co. The assistance of Dr. Alexander Schilling with MALDI timeof-flight mass spectrometry is greatly appreciated. Preliminary results were presented at the 44th ASMS Conference on Mass Spectrometry and Allied Topics, May 12-16, 1996, Portland, OR.
AC970132J X
Abstract published in Advance ACS Abstracts, May 1, 1997.
