Anal. Chem. 2005, 77, 1345-1353
High-Throughput Solution-Based Medicinal Library Screening against Human Serum Albumin Jimmy Flarakos,† Kenneth L. Morand,‡ and Paul Vouros*,†
Barnett Institute and Department of Chemistry and Chemical Biology, Northeastern University, Boston, Massachusetts 02115, and Procter & Gamble Pharmaceuticals, Cincinnati, Ohio 45202
High-throughput screening of combinatorial libraries has evolved from studying large diverse libraries to analyzing small, structurally similar, focused libraries. This paradigm shift has generated a need for rapid screening technologies to screen both diverse and focused libraries in a simple, efficient, and inexpensive manner. We have proactively addressed these needs by developing a highthroughput, solution-based method combining size exclusion (SEC), two-dimensional liquid chromatography (2-D LC), and mass spectrometry (MS) for determining the relative binding of drug candidates in small, focused medicinal libraries against human serum albumin (HSA). Two types of libraries were used to evaluate the performance of the system. The first consisted of five diverse ligands with a wide range of hydrophobicities and whose association constants to HSA cover 3 orders of magnitude. A β-lactam library composed of structurally similar compounds was used to further confirm the validity of the methodology. The ability to distinguish site-specific interactions of drugs competing for individual domains of the HSA receptor is also demonstrated. Comparison of chromatographic profiles of the library components before and after incubation with the receptor using multiple reaction monitoring allowed a ranking of the ligands according to their relative binding affinities. The observed rankings correlate closely with literature values of the association constants between the respective ligands and HSA. This simple, rugged methodology can screen a wide spectrum of chemical entities from combinatorial mixtures in less than 6 min. Protein-drug binding greatly influences absorption, distribution, metabolism, and excretion (ADME) properties of typical drugs. Human serum albumin (HSA) and acid glycoproteins are the most abundant proteins in the systemic circulation, with HSA comprising 60% in plasma.1 An appropriate free-drug concentration (therapeutic level) is necessary to achieve a desired effect. Thus, extensive HSA-drug binding inhibits a drug’s ability to reach therapeutic levels and reduces its value as a viable therapeutic.2 * Corresponding author address: Department of Chemistry and Chemical Biology, 360 Huntington Ave., 102 Hurtig Hall, Boston, MA 02115. † Northeastern University. ‡ Procter & Gamble Pharmaceuticals. (1) Peters, T. All about Albumin: Biochemistry, Genetics and Medical Applications; Academic Press, Inc.: San Diego, 1996. 10.1021/ac048685z CCC: $30.25 Published on Web 02/03/2005
© 2005 American Chemical Society
In addition, a drug with a relatively high HSA binding affinity will have a long half-life, which may increase its toxicity. Conversely, a drug with a low HSA binding affinity is limited in its ability to perfuse tissues and reach the site of action. This is especially relevant to hydrophobic drugs, in which protein binding enhances their solubility and subsequent tissue distribution.3 Therefore, understanding the extent of drug binding will assist in screening out drugs with unfavorable binding characteristics. To address these issues, a variety of affinity selection techniques have been developed for screening ligands against protein receptors. They include equilibrium dialysis,3-5 ultrafiltration,6,7 pulsed ultrafiltration8-13 combined with LC/MS,7,11,14-18 or affinity selection CE/MS.14,19-21 Equilibrium dialysis has served as the benchmark method for determining receptor-ligand association constants. Although it is considered a comprehensive approach, equilibrium dialysis is time-consuming, requiring long incubation times,3 and therefore, it cannot be accepted as a viable, “quick” (2) Lenz, G. R.; Nash, H. M.; Jindal, S. Drug Discovery Today 2000, 5, 145156. (3) Kariv, I.; Cao, H.; Oldenburg, K. R. J. Pharm. Sci. 2001, 90, 580-587. (4) Zini, R.; Morin, D.; Jouenne, P.; Tillement, J. P. Life Sci. 1988, 43, 21032115. (5) Banker, M. J.; Clark, T. H.; Williams, J. A. J. Pharm. Sci. 2003, 92, 967974. (6) Wieboldt, R.; Zweigenbaum, J.; Henion, J. Anal. Chem. 1997, 69, 16831691. (7) Whitlam, J. B.; Brown, K. F. J. Pharm. Sci. 1981, 70, 146-150. (8) Gu, C.; Nikolic, D.; Lai, J.; Xu, X.; Van Breemen, R. B. Comb. Chem. High Throughput Screening 1999, 2, 353-359. (9) Johnson, B. M.; Nikolic, D.; van Breemen, R. B. Mass Spectrom. Rev. 2002, 21, 76-86. (10) Shin, Y. G.; Bolton, J. L.; van Breemen, R. B. Comb. Chem. High Throughput Screening 2002, 5, 59-64. (11) van Breemen, R. B.; Huang, C. R.; Nikolic, D.; Woodbury, C. P.; Zhao, Y. Z.; Venton, D. L. Anal. Chem. 1997, 69, 2159-2164. (12) Zhao, Y. Z.; van Breemen, R. B.; Nikolic, D.; Huang, C. R.; Woodbury, C. P.; Schilling, A.; Venton, D. L. J. Med. Chem. 1997, 40, 4006-4012. (13) Chen, C. J.; Chen, S.; Woodbury, C. P., Jr.; Venton, D. L. Anal. Biochem. 1998, 261, 164-182. (14) Dunayevskiy, Y. M.; Lai, J.-J.; Quinn, C.; Vouros, P. Rapid Commun. Mass Spectrom. 1997, 11, 1178-1184. (15) Jenkins, B. G.; Lauffer, R. B. Mol. Pharmacol. 1990, 37, 111-118. (16) Wabnitz, P. A.; Loo, J. A. Rapid Commun. Mass Spectrom. 2002, 16, 8591. (17) Dunayevskiy, Y.; Vouros, P.; Carell, T.; Wintner, E. A.; Rebek, J., Jr. Anal. Chem. 1995, 67, 2906-2915. (18) Blom, K. F.; Larsen, B. S.; McEwen, C. N. J. Comb. Chem. 1999, 1, 82-90. (19) Chu, Y.-H.; Dunayevskiy, Y. M.; Kirby, D. P.; Vouros, P.; Karger, B. L. J. Am. Chem. Soc. 1996, 118, 7827. (20) Kaur, S.; McGuire, L.; Tang, D.; Dollinger, G.; Huebner, V. J. Protein Chem. 1997, 16, 505-511. (21) Liu, J.; Abid, S.; Hail, M. E.; Lee, M. S.; Hangeland, J.; Zein, N. Analyst 1998, 123, 1455-1459.
Analytical Chemistry, Vol. 77, No. 5, March 1, 2005 1345
screening methodology. Ultrafiltration and its associated pulsed ultrafiltration technique are commonly used, but may suffer from nonspecific binding.5 Although they are still effective for determining binding constants, these methods are still relatively timeconsuming for quick screening purposes. In addition, a number of studies have been conducted to predict binding affinity using computational approaches22-24 or immobilized receptors.25,26 Immobilized receptors have been used with some success to predict binding constants;25,26 however, the immobilized HSA columns are expensive. In addition, the restrictive nature of the bound receptor diminishes the HSA’s ability to fold and reduces its activity.1 Computational techniques are increasingly commonplace for predicting binding; however, these techniques have not matured enough to allow accurate and consistent determination of solutionbased drug-protein association constants under different solution conditions. Off-line solution-based screening methods using size exclusion have been investigated and found capable of accurately predicting relative binding between ligands and receptors.14,16,17 Their success and versatility is related to their use of Sephadex G-25 as the size exclusion medium. Sephadex, a cross-linked dextran, was developed over 40 years ago for the purpose of solute separation based on size.27-29 Sephadex-based size exclusion separations still serve as the benchmark for protein purification due primarily to their minimal nonspecific interactions with the sample components. More recently, automated approaches have been developed to improve throughput.18 These automated approaches typically use silica-based size exclusion media, which tend to show nonspecific interactions with highly basic analytes. A widely applicable, robust, reproducible, and affordable protein separation system requires Sephadex as the backbone. Although off-line techniques use Sephadex, they could still be automated to reduce sample preparation steps and increase throughput. With this in mind, we developed an online, solution-based, affinity selection methodology to rapidly identify relative binding affinity of combinatorial library components against HSA. Our online methodology required minimal sample preparation, building on work previously developed in our laboratory using G-25 microspin cartridges,14,16,17 and uses the Sephadex online with a column-switching configuration. The methodology was tested in three ways. First, five compoundssindomethacin, warfarin, imipramine, acetylprocainamide, and quinidinespossessing a wide range of binding constants, structural characteristics, and hydrophobicities (see Figure 1) were selected to validate the methodology. Their association constant values ranged from 1.4 × 106 M-1 to 1.6 × 103 M-1. The methodology was next applied to a five-component β-lactam library. β-Lactams constitute an important class of antibiotics used extensively to treat bacterial infections.30-33 These antibiotics are structurally similar, possessing a fused five-membered ring or a (22) Hall, L. M.; Hall, L. H.; Kier, L. B. J. Comput.-Aided Mol. Des. 2003, 17, 103-118. (23) Kratochwil, N. A.; Huber, W.; Muller, F.; Kansy, M.; Gerber, P. R. Biochem. Pharmacol. 2002, 64, 1355-1374. (24) Takamatsu, Y.; Itai, A. Proteins 1998, 33, 62-73. (25) Woodbury, C. P., Jr.; Venton, D. L. J. Chromatogr., B 1999, 725, 113-137. (26) Buchholz, L.; Cai, C. H.; Andress, L.; Cleton, A.; Brodfuehrer, J.; Cohen, L. Eur. J. Pharm. Sci. 2002, 15, 209-215. (27) Winzor, D. J. J. Biochem. Biophys. Methods 2003, 56, 15-52. (28) Porath, J. Clin. Chim. Acta 1959, 4, 776-778. (29) Porath, J.; Flodin, P. Nature 1959, 183, 1657-1659. (30) Chow, B. F.; McKee, C. M. Science 1945, 101.
1346
Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
fused six-membered dihydrothiazine ring.34 Our methodology was further extended to explore its capability for determining sitespecific interactions using as model targets Sites I and II, two of the three most characterized binding domains of HSA.1 EXPERIMENTAL DETAILS HSA, acetylprocainamide, warfarin, indomethacin, quinidine, imipramine, ammonium acetate, trifluoroacetic acid, and all other reagents were purchased from Sigma Chemical Company (St. Louis, MO) unless otherwise specified. Deionized water was generated by a Milli-Q-Plus water system from Millipore (Waltham, MA). HPLC grade acetonitrile was purchased from Fisher Scientific (Pittsburgh, PA). Samples were centrifuged in 1.5-mL siliconized microcentrifuge tubes in a Marathon 21000 (R) refrigerated multipurpose centrifuge. Both were purchased from Fisher (Pittsburgh, PA). Incubations of HSA-ligand complexes were isolated from excess ligand on a custom-packed 1-mL HiTrap desalting column (Amersham Biosciences), all in buffer B (10 mM of Tris/HCl, pH ) 7.4; 10 mM MgCl; 10 mM KCl). The trap cartridge was a protein macrotrap 3 × 8 mm from Michrom BioResources (Auburn, CA) or a Waters Oasis HLB 3 × 20 mm (Milford, MA). The analytical column was a Luna C18 50 × 2.1 mm, 5 mm from Phenomenex (Torrance, CA). LC/MS. The compounds were separated with an HP1100 HPLC fitted with a 10-µL flow cell and a photodiode array detector. The eluting compounds were introduced into a PE-Sciex API-III + triple quadrupole tandem mass spectrometer (Thornhill, Ontario) in either APCI+ or ESI+ mode. The electropray conditions were turboionspray temperature, 480 °C; needle voltage, 4200 V; curtain gas, 1.0 L/min; orifice, 50 V; nebulizer pressure, 40 psi; auxiliary gas, 4000 cm3/min; and the collision gas set to 1.5 × 10-14 molecules/cm2. In APCI, some parameters were changed: the heated nebulizer temperature was 480 °C, nebulizer pressure was 100 psi, and nebulizer flow was1.0 L/min; the rest were kept the same as those used for ESI. Specific mobile phase compositions and associated components are listed in Table 1. The compounds are listed in order of increasing affinity for HSA on the basis of binding constants or percent binding. Table 1 summarizes the conditions used for the appropriate ligands. Off-Line Gel Filtration Isolation. CentriSpin G-20 columns (Princeton Separations) were used off-line to establish a benchmark method for comparison with our on-line technique. The CentriSpin columns can separate proteins of g25 kDa from salts, unwanted low-molecular-weight impurities, and ligands. The columns were hydrated with 650 µL of buffer B for 1 h, followed by equilibration at 750g. On-Line Gel Filtration and Isolation. A 75-µL solution of 100 µM HSA was incubated in buffer B with 75 µL of 100 µM each of the five component ligand library for 1 h before injecting the incubate into the system. On-line determination was performed using two Valco six-port switching valves (Valco Instruments Co. Inc. Houston, TX) with a two-column, ternary pumping configuration (see Figure 2). The L and E in the center of the each sixport valve designate “load” or “elute” positions for the flow path. (31) Horimoto, S.; Mayumi, T.; Aoe, K.; Nishimura, N.; Sato, T. J. Pharm. Biomed. Anal. 2002, 30, 1093-1102. (32) Niessen, W. M. J. Chromatogr., A 1998, 812, 53-75. (33) Parker, C. E.; Perkins, J. R.; Tomer, K. B.; Shida, Y.; O’Hara, K. J. Chromatogr. 1993, 616, 45-57. (34) Riediker, S.; Stadler, R. H. Anal. Chem. 2001, 73, 1614-1621.
Figure 1. Compounds tested: (a) diverse library compounds and (b) β lactams.
Incubations of HSA-ligand complexes (10 µL) were analyzed at a flow rate of 0.4 mL/min by SEC for mass selective separation. HSA and the HSA-ligand complexes eluted in the void and traveled directly to the trap column (step (a)). At 1.5 min, (b) valve 1 is switched, and the smaller, late-eluting ligands from the SEC column are sent to waste. This also places mobile phase 2 in line with the trap cartridge at a flow of 0.5 mL/min. The acid pH of mobile phase 2 denatures the HSA disrupting the complex, washing off the HSA and leaving the previously complexed ligands to bind to the trap cartridge. At 3.5 min, valve 2 is switched, allowing mobile phase 3 to backflush the bound ligands onto the analytical column for further separation and detection. β-Lactams were studied using a modified method (see Table 1). Specifically, a Waters Oasis HLB 30 × 2 mm, 5-µm was used
as the trap column to permit multiple analyses. The ionization mode was positive ESI with a cycle time of 6 min. Competitive binding experiments were performed similarly with indomethacin, warfarin, dansylsarcosine, and tryptophan as the competing ligands. All sample injections were done in triplicate to ensure validity of data and methodology. RESULTS AND DISCUSSION Evaluation of Column Switching System. A fundamental requirement for the effective operation of the on-line system shown in Figure 2 is the clean separation of the protein or protein/ligand complex from the free ligands. It is also assumed that the off rates for the dissociation of the complexes in their equilibrium process were slow. The efficacy of the experimental approach was Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
1347
Table 1. LC/MS/MS Conditions condition
generic method
solvent A solvent B buffer B mobile phase 1 mobile phase 2 mobile phase 3 SEC column trap column reversed-phase (RP) column
β-lactam method
10 mM ammonium acetate in water 10 mM ammonium acetate in acetonitrile 10 mM Tris/HCl, pH ) 7.4; 10 mM MgCl2; 10mM KCl 95/5 buffer B/methanol 0.1% TFA in water 5-80% B over 3.5 min 50% B/A HiTrap 1 mL Protein Trap Oasis HLB Luna C18 APCI+
ionization mode increasing binding affinity V
ESI+
Compound
SRM
Compound
SRM
quinidine acetylprocainamide imipramine warfarin indomethacin
325.2 > 160.2 278.1 > 205.2 281.2 > 208.3 309.1 > 250.9 358.1 > 139.0
ampicillin piperacillin cefotaxime cloxacillin dicloxacillin
349.9 > 106.2 518.0 > 143.3 456.3 > 156.2 436.1 > 277.2 470.2 > 311.2
Figure 2. Column switching configuration.
evaluated first using the diverse library of compounds of known binding constants and is demonstrated in the results summarized in Figures 3 and 4. Figure 3 shows the UV trace of a 10-µL injection of a 100 µM HSA solution eluting from the SEC. The sample was subsequently loaded for 1.5 min onto the trap column, followed by reversed-phase LC/MS/MS analysis. As shown in the top panel, the HSA protein elutes in the void volume at 0.765 min, and the MS/MS extracted ion chromatograms (XICs) show no potential ion interferences from the ligands at their elution time (2-4min). This experiment confirmed that the system does not respond to any components that may reside in the HSA at the retention time and the m/z values of our library compounds. A reverse experiment was then conducted whereby 10 µL of the diverse library solution was injected onto the SEC column. As seen from the SEC-UV profile (Figure 4, upper panel), the ligands elute after the void volume (after 2.3 min), and the specificity of the methodology is verified by the MS/MS XICs, which show no 1348
Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
responses related to the ligands in the time window before 1.5 min. Identical results were obtained with the β-lactam library. Thus, these experiments served to validate the specificity of the method and establish that any response in the MS/MS XICs before the 1.5-min switch would have to be due to ligands eluting from the SEC column bound to the HSA receptor. Two model libraries were tested to demonstrate the general effectiveness of the on-line system toward screening for ligandreceptor interactions. It was of particular interest to determine whether the proposed experimental approach could rank the relative affinities of the ligand HSA interactions in accordance with their known association values given in the literature. For the first library, the strength of the interaction is represented by the association constant, Ka, values, which ranged from 1.4 × 106 to 1.6 × 103 M-1, where the greater the Ka, the stronger the binding. Literature values of the relative affinities for members of the β-lactam library are reported in terms of percent binding to the
Figure 3. SEC-UV-LC/MS/MS of 100 µM HSA, 10 µL.
circulating physiological concentrations of HSA, which is essentially related to the Ka values. Two variations of the same methodology have been described in Table 1. This was done to test the wider applicability of the methodology should changes need to be made in the trapping column to retain different types of compounds. Two ion source modes (APCI and ESI) were used to show that the methodology is effective regardless of the type of ionization mode used. Binding Selectivity of Diverse Library. An equimolar solution of the five-member diverse library (100 µM each) was incubated with an equal volume of 100 µM HSA for 1 h and allowed to reach equilibrium. A 5-µL aliquot of the mixture was injected into the SEC column, and its UV elution profile is shown in Figure 5a. Two bands are observed, of which the first (elution maximum at ∼0.8 min.) contains the free HSA as well as HSA-ligand complexes. The second band eluting after ∼1.5 min (maximum at 2.18 min), represents the elution of free ligands. The first band was diverted to the trap at 1.5 min, and after denaturation of the complex with mobile phase 2 (0.1% TFA in water) the released ligands were analyzed by reversed phase LC/MS/MS (Figure 6). The SRM chromatograms shown in
Figure 6b may be compared to the chromatograms of the same library analyzed using the same experimental setup but prior to incubation with HSA (Figure 6a). The ratio of the peak areas for any given compound may be used to ascertain the propensity of the drug to bind to the receptor. For example, if all the drug present in the library solution is fully bound to the receptor, the area counts in 6a and 6b would be equal (ratio of 1.00). On the other hand, for weakly bound ligands, the peak area ratio will be progressively larger. Accordingly, a ranking system may be established to categorize the components of the library in terms of their relative affinities to the receptor, with the larger ratio indicating the stronger binder. The results for the diverse library are summarized in Table 2, where the peak area ratios are also compared to the reported association constants of the ligands with HSA.1,23 As noted, quinidine and acetylprocainamide are the weakest binders, whereas the highest ratio is calculated for indomethacin. It is conceivable that residual HSA, adsorbed on the trap column, may interfere with the peak area ratios of postincubation samples. However, the consistency of our results leads us to believe that any ion suppression effects are not a factor in these measurements. Nonetheless, this issue would be imporAnalytical Chemistry, Vol. 77, No. 5, March 1, 2005
1349
Figure 4. SEC-UV-LC/MS/MS of 100 µM β-lactam library, 10 µL.
tant to investigate further, especially as it pertains to ESI. Significantly, the overall trend observed for binding of the library components with HSA is fully in line with their respective association constants, and the reliability of the measurements is further supported by their high reproducibility, with relative standard deviation values of 10% or less. Analysis of β-Lactam Library. The results of the above experiments demonstrate that our methodology is capable of screening diverse libraries to provide a direct correlation of association constants between ligands and receptors. Alternatively, drug-receptor interactions are often expressed in terms of the percent drug bound to the protein using as a guideline the circulating molar concentrations of the protein and physiologically active levels of the drug.35 A β-lactam library consisting of structurally similar components (Figure 1), for which the percent binding information has been well documented,22 was selected for further evaluation of our on-line screening methodology. To accommodate the distinct structural features of the β-lactams, some minor modifications to the chromatographic separations were incorporated, as outlined in Table 1. More1350
Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
over, since the analysis of the lactams is often complicated because of the facile cleavage of the lactam ring, a compatible ionization method was necessary. Since APCI analysis of β-lactams would require bromine modifiers and other nonconventional operating parameters,31 we opted to use positive ESI for their characterization. As with the previous example, an equimolar solution of the five-member β-lactam library (100 µM each) was incubated with an equal volume of 100 µM HSA for 1 h and allowed to reach equilibrium. The mixture was analyzed as described above, and the results are presented in Figures 5b and 7. It is important to note the reproducible separation of the bands related to the protein and protein-drug complex from the unbound ligands eluting from the SEC column (Figure 5b). In addition, distinct differences are observed in the SRM chromatograms (Figure 7a and b), the results of which are summarized in Table 3. The trend observed for the library components is in line with their percent binding values,23 further validating the capability (35) Gilman, A. G. Goodman and Gilman’s The Pharmacological Basis of Therapeutics; 7th ed; Macmillan: New York, 1985.
Figure 5. SEC-UV-LC/MS/MS of 100 µM incubation of HSA and (a) diverse ligand library or (b) β-lactam library.
Figure 6. SEC-UV-LC/MS/MS peak areas of (a) reference 10 µM of diverse library and (b) incubation of 100 µM of HSA and diverse library.
Table 2. Relative Binding Ratio of Diverse Library ligand
incubation (100 µM) peak area
ref (10 µM) peak area
association constant Ka (M-1)
av peak area ratio (incubation/ref)
quinidine acetylprocainamide imipramine warfarin indomethacin
3.20 × 102 0.00 2.84 × 103 7.56 × 104 9.48 × 103
9.29 × 104 8.96 × 104 4.80 × 104 4.67 × 105 2.32 × 104
1.6 × 103 3.3 × 103 2.5 × 104 3.3 × 105 1.4 × 106
3.80 × 10-3 (% RSD ) 12.7, n ) 5) 0.00 (% RSD ) N/AP, n ) 5) 5.18 × 10-2 (% RSD ) 10.8, n ) 5) 1.54 × 10-1 (% RSD ) 5.70, n ) 5) 3.92 × 10-1 (% RSD ) 7.28, n ) 5)
of the methodology for screening even closely related library components and providing a reliable ranking based on percent relative binding affinities.
Multiple Site Binding. There are at least two known binding sites on HSA, and it was of interest to examine further the specificity of our method under conditions that one site might Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
1351
Table 3. Relative Binding Ratio of (β-lactam library ligand
incubation (100 µM) peak area
ref (10 µM) peak area
% binding
av peak area ratio (incubation/ref)
ampicillin piperacillin cefotaxime cloxacillin dicloxacillin
8.60 × 1.00 × 104 2.30 × 103 5.00 × 103 5.55 × 103
1.03 × 8.93 × 104 1.80 × 104 2.10 × 104 1.64 × 104
18 30 38 89 97
8.07 × 10-2 (% RSD ) 5.68, n ) 5) 1.26 × 10-1 (% RSD ) 8.82, n ) 5) 1.35 × 10-1 (% RSD ) 11.8, n ) 5) 2.54 × 10-1 (% RSD ) 6.44, n ) 5) 3.38 × 10-1 (% RSD ) 11.3, n ) 5)
103
105
Figure 7. SEC-UV-LC/MS/MS peak areas of (a) reference 10 µM of β-lactams and (b) incubation of 100 µM of HSA and β-lactams.
Figure 8. SEC-UV-LC/MS/MS peak areas of (a) reference 10 µM of site I and site II ligands and (b) incubation of 100 µM of HSA and ligands.
interfere with binding at an adjacent site. Four compounds, each pair exhibiting binding characteristics to either site I or II of HSA were incubated with the protein. Site I is specific to indomethacin 1352 Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
and warfarin; Site II is specific to dansylsarcosine and tryptophan. The results of the experiment are shown in Figure 8, which compares the LC/MS/MS profiles of the solution of the two pairs
Table 4. Relative Binding Ratio of Site I and II Ligands ligand
incubation (100 µM) peak area
ref (10 µM) peak area
association constants (M-1)
peak area ratio (incubation/ref)
warfarin (site I) indomethacin (site I) tryptophan (site II) dansylsarcosine (site II)
4.56 × 105 1.30 × 105 1.93 × 103 2.34 × 105
1.06 × 106 1.88 × 105 2.99 × 104 4.16 × 105
3.3 × 105 1.4 × 106 1.4 × 104 1.8 × 105
4.77 × 10-1 (% RSD ) 7.48, n ) 5) 7.24 × 10-1 (% RSD ) 6.09, n ) 5) 7.21 × 10-2 (% RSD ) 10.8, n ) 5) 5.56 × 10-1 (% RSD ) 2.23, n ) 5)
of compounds before (left panel) and after (right panel) incubation with the protein, and the data are summarized in Table 4. Indeed, we see that the relative ratios in Table 4 for the site I binding ligands, indomethacin and warfarin, were 0.72 and 0.48, respectively. Similarly, site II binding ligands, dansylsarcosine and tryptophan, gave relative ratios of 0.56 and 0.07. On the basis of their association constants, binding affinity to HSA would favor a higher peak area ratio for warfarin than dansylsarcosine if both ligands competed for the same binding site. The opposite is seen in Table 4, reflecting no competition between site I and II ligands. It may be noted that the binding ratio for warfarin and indomethacin in the site binding experiment (0.47 and 0.72; see Table 4) were not the same as the binding ratios for warfarin and indomethacin with the diverse library (0.15 and 0.39; see Table 2). Considering the same two ligands (warfarin and indomethacin) were in different competitive environments in the diverse ligand and multiple site binding experiments, we expect slightly different ratios but no significant change in the rank order of binding. This experiment demonstrated that the binding was site-specific, and ligands bound to one site did not interfere with binding to an adjacent site. Although some HSA conformational changes are likely to occur during site I and II binding, they do not seem to interfere with their respective binding affinity to particular ligands. Thus, these results indicate that it should be possible to screen site I and site II ligands simultaneously and potentially double the capacity of our binding assay. CONCLUSION In summary, a simple and rapid method has been developed for determining the relative binding affinity of combinatorial library members to a protein receptor using an on-line SEC column switching system. Previously, microspin cartridges have been used successfully in an off-line configuration to determine relative protein-ligand binding affinities;14,16 however, for on-line purposes, the microspin method had significant sample handling problems and required dedicated equipment, and the price per cartridge proved to be relatively high. Alternative approaches based on 96well plate methods are of higher throughput, but they require a centrifuge and autosampler specifically designed to handle these plates, and the price per plate is still high. The methodology
presented here offers a balanced approach of high throughput using conventional laboratory resources and is designed to handle multiple injections of libraries on one cartridge in a continuous flow format. Initially, conventionally packed SEC columns were tested, but they tended to be expensive and came in impractical column dimensions, and some displayed nonideal SEC behavior with more basic compounds. These secondary interactions also caused differences in protein-ligand binding affinities by denaturing the protein. To address this deficiency, we custom-packed sephadex G-25 columns, mimicking the microspin characteristics. This allowed us to automate our methodology, reducing our cost per injection to one-third, based on current price of consumables, i.e., that of the microspin columns. Moreover, addition of 5% organic to the mobile phase in order to reduce nonspecific binding and protein aggregate formation further increased the life of the cartridges, enabling us to routinely run 25 or more injections per cartridge without loss of efficiency. In addition, these cartridges can be regenerated to prolong their yet undetermined maximum lifetime. Future work will require the miniaturization of the methodology. Relatively large amounts of protein were used in this methodology to show proof of principle. The Sephadex G-25 media can be packed into capillaries, and nanovalves can be used to reduce solvent and sample requirements. Since the system is solution-based and does not denature the protein, with some modifications, this methodology can be used to automate any offline, solution- and Sephadex-based protein-ligand bioassay. ACKNOWLEDGMENT The authors thank MDS Pharma Services for their generous donation of the HP 1090, technical support group at Amersham biosciences for custom packing our cartridges, Caroline Ceailles and John Williams for editing and useful manuscript criticism, and P&G and National Institutes of Health (Grant no. RO1CA693006) for their generous support. This is publication number 850 from the Barnett Institute. Received for review December 13, 2004.
September
3,
2004.
Accepted
AC048685Z
Analytical Chemistry, Vol. 77, No. 5, March 1, 2005
1353
