Anal. Chem. 1997, 69, 4878-4884
Biochemical Detection for Direct Bead Surface Analysis E. S. M. Lutz, H. Irth,* U. R. Tjaden, and J. van der Greef
Division of Analytical Chemistry, Leiden/Amsterdam Center for Drug Research, Leiden University, P.O. Box 9502, 2300 RA Leiden, The Netherlands
A continuous-flow biochemical detection system is presented which recognizes biologically active compounds immobilized to solid phases. This approach can be used to screen, for example, solid-phase combinatorial libraries for lead compounds. Biochemical detection is performed by mixing a plug of a solid-phase suspension with labeled affinity protein. During a short reaction time, the labeled affinity protein will only bind to ligands, i.e., compounds with biological activity. Hereafter, the free and bound labels are separated by means of a hollow fiber module. Quantitation of the free label is performed with a conventional flow-through fluorescence detector. Total assay time amounts to less than 3 min. Biochemical detection for direct bead surface analysis was developed for two model systems. The first model system used fluorescencelabeled avidin as affinity protein and its ligands biotin and iminobiotin immobilized to agarose as analytes. The second model system used fluorescence-labeled antisheep (Fab)2 fragments as affinity protein and different IgGs immobilized to agarose as analytes. The feasibility of this approach for recognition of solid-phase immobilized ligands was documented by screening 50 samples with a 100% hit rate. Combinatorial chemistry is a recent approach for building highdiversity libraries for generation and optimization of drug candidates. The synthetic strategy in combinatorial chemistry is based on systematic, repetitive, covalent connection of a number of “building blocks” (e.g., amino acids, nucleic acids) of varying structures to each other, thus yielding diverse molecular entities.1 It often makes use of solid-phase chemistry, since excess reagents and byproducts not bound to the solid support can be removed simply by washing.2 Typically, large polymeric beads are used, synthesizing the drug candidate to a selectively cleavable linker group, such as dicyclocarbodiimide.3,4 The developing importance of combinatorial libraries requires recognition of active ligands within the vast amount of chemical entities. This calls for rapid, automated, and straightforward methods to determine biologically active compounds. (1) Gallop, M. A.; Barrett, R. W.; Dower, W. J.; Fodor, S. P. A.; Gordon, E. M. J. Med. Chem. 1994, 37, 1232-51. (2) Gordon, E. M.; Barrett, R. W.; Dower, W. J.; Fodor, S. P. A.; Gallop, M. A. J. Med. Chem. 1994, 37, 1385-1401. (3) Haskins, N. J.; Hunter, D. J.; Organ, A. J.; Rahman, S. S.; Thom, C. Rapid Commun. Mass Spectrom. 1995, 9, 1437-40. (4) Brummel, C. L.; Vickerman, J. C.; Carr, S. A.; Hemling, M. E.; Roberts, G. D.; Johnson, W.; Weinstock, J.; Gaitanopoulos, D.; Benkovic, S. J.; Winograd, N. Anal. Chem. 1996, 68, 237-42.
4878 Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
Biological activity is determined by testing the chemical entities of the library against a molecular or cellular target, e.g., an enzyme or a receptor.5 Recognition is based on affinity interactions between the target and the compounds of interest. Several strategies have been described for screening compounds originating from bead-bound libraries. When the synthesized compound is cleaved off, established assay formats can be employed, such as competition binding assays, enzyme inhibition assays, or cell-based bioassays. In these assays, affinity reactions are monitored by labeling the target or by using a labeled ligand that competes for the same binding site as the compounds of interest. Recognition based on affinity interactions has also been coupled to other analytical systems. For example, affinity chromatography can be used for isolation of biologically active compounds.6 By coupling the affinity chromatography with CEMS,7 ES-MS,8 or Edman peptide sequencing,9 recognition of biological activity is combined with structural analysis. For these methods, it is necessary to establish that the immobilized target has retained its biological activity.2 However, as soon as the compound is cleaved off of the solid phase, it cannot be used for further steps such as identification or continued synthesis. In an approach to overcome this limitation, solid phases were developed with linkers with varying cleaving optimums, thus allowing one to cleave only a part of the compound from the solid phase.10 Cleavage is fully circumvented when the assay procedure employs a soluble target for binding to the immobilized ligands. The beads can be stained with labeled target for selecting the biologically active compounds. Stained beads are then isolated either by visual inspection and manual removal,9,11 or by flow cytometric sorting.12 Though flow cytometric sorting allows a high sample throughput, these methods require long incubation times (5) Ashton, M. J.; Jaye, M. C.; Mason, J. S. Drug Discovery Today 1996, 1, 11-5. (6) Evans, D. M.; Williams, K. P.; McGuinness, B.; Tarr, G.; Regnier, F. E.; Afeyan, N.; Jindal, S. Nature Biotechnol. 1996, 14, 504-7. (7) Chu, Y.-H.; Dunayevskiy, Y.; Kirby, D. P.; Vouros, P.; Karger, B. L. J. Am. Chem. Soc. 1996, 118, 7827-35. (8) Kelly, M. A.; Liang, H.; Sytwu, I.-I.; Vlattas, I.; Lyons, N. L.; Bowen, B. R.; Wennogle, L. Biochemistry 1996, 35, 11747-55. (9) Lam, K. S.; Salmon, S. E.; Hersh, E. M.; Hruby, V. J.; Kazmierski, W. M.; Knapp, R. J. Nature 1991, 354, 82-4. (10) Salmon, S. E.; Lam, K. S.; Lebl, M.; Kandola, A.; Khattri, S. P.; Wade, S.; Pa´tek, M.; Kocis, P.; Krchna´k, V.; Thorpe, D.; Felder, S. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 11708-12. (11) Needels, M. C.; Jones, D. G.; Tate, E. H.; Heinkel, G. L.; Kochersperger, L. M.; Dower, W. J.; Barrett, R. W.; Gallop, M. A. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 10700-4. (12) Mu ¨ ller, K.; Gombert, F. O.; Manning, U.; Grossmu ¨ ller, F.; Graff, P.; Zaegel, H.; Zuber, J. F.; Freuler, F.; Tschopp, C.; Baumann, G. J. Biol. Chem. 1996, 271, 16500-5. S0003-2700(97)00485-X CCC: $14.00
© 1997 American Chemical Society
and sophisticated instrumentation. When immobilized ligands are screened by direct target binding, the covalent attachment of the ligand strongly influences whether or not the target is able to interact. It is therefore crucial to design the ligand in such a way as to limit any adverse effects, e.g., by using an appropriate spacer. In order to enhance the application of combinatorial technologies to drug discovery, synthesis and screening should be integrated.2 However, the methods developed so far are difficult to link to the synthesis procedures. In addition, most assays still require long incubation times, thus not allowing a fast feedback. The purpose of the present paper was to develop an on-line procedure for recognition of active ligands that are immobilized to a solid phase. Recently, a biochemical solid phase assay was successfully translated into a flow assay which offers short reaction times and allows automation of the assay with conventional LChardware.13 The same approach is used in this paper. However, in contrast to the previous assay where the solid-phase bound reagent is added continuously, the assay format described in this paper introduced solid-phase suspensions as the sample plug injected into the flow system. Labelled affinity proteins were used for recognition and quantitation of biological activity of solid-phase bound compounds. A key element in the setup is a previously described hollow fiber module,14 which separates free and bound label and simultaneously ensures compatibility of the solid phases with conventional LC hardware. In this way, direct analysis of compounds attached to bead surfaces is achieved within a time frame of 3 min. MATERIALS AND METHODS Chemicals. Immobilized d-biotin, immobilized iminobiotin, and immobilized avidin were purchased from Pierce (Rockford, IL). CNBr-activated Sepharose was obtained from Pharmacia (Uppsala, Sweden), and streptavidin immobilized on macroporous metacrylate polymer beads (streptavidin beads) was obtained from Boehringer Mannheim (Mannheim, Germany). Goat-, sheep-, and mouse-IgG and fluorescein isothiocyanate-labeled avidin (FITCavidin) were obtained from Sigma (St. Louis, MO). Labeled antisheep (Fab)2 fragments [DTAF-(Fab)2] were obtained from Chemicon (Temecula, CA). Tween 20, sodium chloride, sodium phosphate, potassium chloride, and potassium phosphate were purchased from Merck (Darmstadt, Germany). All chemicals used were of analytical grade. HPLC-grade water was produced in a Milli-Q system (Millipore, Bedford, MA) and used throughout this work. Instrumentation. The flow injection (FI) system consisted of a Gilson (Villiers-le-Bel, France) XL autosampler equipped with a Rheodyne (Cotati, CA) six-port injection valve (injection loop 20 µL), a Pharmacia LKB (Uppsala, Sweden) 2248 LC pump, which delivered the FI carrier at a flow rate of 0.4 mL/min, and a Gilson Minipuls peristaltic pump, which delivered the affinity protein solution [5 nmol/L FITC-avidin or 4 nmol/L DTAF-(Fab)2 unless specified otherwise] at a flow rate of 0.2 mL/min. A Jasco (Tokyo, Japan) fluorescence detector FP 920 (λexc ) 486 nm, λem ) 516 nm; digital filter with 10 s delay time) with a 16 µL flow cell was used for detection. Data acquisition occurred with Gilson 715 software with a data collection rate of 100/min. The polarity of (13) Lutz, E. S. M.; Irth, H.; Tjaden, U. R.; van der Greef, J. J. Chromatogr., A 1997, 776, 169-78. (14) Lutz, E. S. M.; Irth, H.; Tjaden, U. R.; van der Greef, J. J. Chromatogr., A 1996, 755, 179.
Table 1. Characteristics of the Hollow Fiber Module internal fiber diameter fiber wall thickness nominal MWCO fiber length membrane surface area retentate volume
0.85 mm 0.25 mm 150 000-200 000 23 mm 61 mm2 13 µL
the signal output was reversed in order to obtain positive signals. The analytical system was controlled by Gilson 719 Turbo Pascal software. Phosphate-buffered saline (136.9 mmol/L NaCL, 2.7 mmol/L KCl, 9.2 mmol/L Na2HPO4, 1.8 mmol/L KH2PO4) pH 7.4 containing 0.5% Tween 20 (PBST) or 0.2 mol/L carbonate buffer pH 8.7 containing 0.5% Tween 20 (CBT) were used as FI carrier. The affinity protein solution was prepared in the same buffer, which served as FI carrier and was added to the FI carrier via an inverted Y-type mixing union (Upchurch, Oak Harbor, WA). A knitted 0.5 mm i.d. PTFE reaction coil with a volume of 858 µL was used. The system was operated at room temperature. Hollow Fiber Module. Separation of free and bound label was performed with a custom-made hollow-fiber module which has been described previously14 and consisted of one capillary cross-flow membrane (X-Flow BV, Almelo, The Netherlands), which was fitted into an exactly fitting cylindrical channel (1.35 mm i.d.). The characteristics of the hollow-fiber module are summarized in Table 1. Back pressure, necessary to create the driving force for membrane separation, was created by pumping the retentate flow with the peristaltic pump as described earlier,13 causing a flux of 0.05 mL/min. Preparation of Samples. Solid phases (biotin-, iminobiotin-, and avidin-agarose) with for the model system using FITC-avidin as target were purchased from Pierce as 50% aqueous slurries. According to the manufacturer, the binding capacity of biotinagarose and iminobiotin-agarose was 30 and 15 nmol/mL, respectively. For the model system using DTAF-(Fab)2 as target, ligands (none, mouse-, goat- and sheep-IgG) were immobilized in-house to CNBr-Sepharose according to the manufacturer’s procedure15 and kept as 50% aqueous slurries in PBST at 4 °C. In the first batch, the amount of goat-IgG immobilized was 130 nmol/mL agarose. For investigating the influence of the ligand surface density, goat-IgG was immobilized in concentrations of 0, 0.007, 0.0013, 0.07, 0.13, 0.7, 1.3, 6.7, and 13.3 nmol/mL agarose. For the experiment comparing mouse-, goat-, and sheep-IgG as ligands for DTAF-(Fab)2 as target, a concentration of 16 nmol/mL was used. Prior to use, the 400 µL of the solid-phase stock suspensions was washed in an emptied SPE cartridge with 10 mL of the FI carrier in order to eliminate signals as a consequence of different buffer compositions. After washing, the solid-phase material was suspended in 400 µL of the FI carrier and diluted to the required density. For the screening experiment, samples were made from the three different batches of blank- and IgG-agarose, containing none, 130 nmol/mL goat-IgG, 4 mg/mL goat-IgG, and 2 mg/mL mouse-, goat-, and sheep-IgG. The screening experiment was carried out in a double-blind procedure. (15) Affinity Chromatography: Principles and Methods; Pharmacia, L. B., Uppsala, Sweden.
Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
4879
Figure 1. Biochemical detection system based on labeled affinity protein: (a) labeled affinity protein; (b) solid-phase bound analyte.
Prior to injection, the sample suspensions were aspirated and dispensed three times by the autosampler at a flow rate of 20 mL/ min in order to ensure homogeneity of the suspensions. RESULTS AND DISCUSSION Model Systems. Screening for compounds with biological activity is performed with a biological target, such as an enzyme or a receptor, which is linked directly to the disease under consideration.5 The screening is based on the affinity recognition between the target and the biologically active compounds. Here, we used avidin and anti-sheep (Fab)2 fragments as model targets. In this way, the biochemical detection setup for direct analysis of surface-bound compounds is investigated for both low and high molecular mass ligands. Solid-phase bound ligands for avidin are commercially available in the form of biotin- and iminobiotinagarose, whereas solid-phase bound ligands for antisheep (Fab)2 fragments were obtained through the immobilization of sheepor cross-reactive goat-IgG antibodies to agarose. Therefore, agarose was used as model solid-phase during this investigation. With a particle size in the range of 45-165 µm, agarose represents challenges with respect to the hardware similar to the polymeric particles of ∼50 µm which are typically used in combinatorial chemistry.3 Design of the Biochemical Detection System. Assay Format. For the detection of active ligands on the bead surface, affinity proteins labeled with fluorophores are used as reporter molecules (scheme, see Figure 1). The sample suspension containing solid-phase bound ligand is injected into the FI carrier and the labeled affinity protein is added via a mixing union. During the following reaction, the affinity protein binds to the active ligand. After a reaction time of 85 s, the free and beadbound fractions of the label are separated on the basis of their large size difference using a hollow fiber module. The free label fraction in the permeate stream is directed to a fluorescence detector. Accordingly, maximum response is obtained in the absence of analyte. When analyte is present, the concentration of free label and, thus, the signal is decreased in analogy with other indirect detection methods. In order to obtain positive signals, the polarity of the signal output was reversed. Hardware Considerations. Recently, we developed a biochemical detection system employing solid phases as reagents.13 In the present paper, we demonstrate that a similar approach can also be used for the analysis of solid-phase surfaces. When solid phases are introduced into an FI system, it is important to avoid 4880
Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
clogging. The most sensitive parts of the present system were the injection needle and the mixing union. When clogging occurred in the injection needle, mixing of the sample was not sufficient to obtain a homogeneous suspension prior to injection. When clogging occurred in the mixing union, back pressure in the system increased, leading ultimately to a breakdown of the system. However, unless the injected suspension was too thick (>100 µL of agarose/mL), clogging was avoided, which is in concurrence with the compatibility of bead suspensions in the previously described flow system.13 Compatibility of the bead suspensions with conventional flowthrough detectors was ensured by introducing a hollow fiber module into the flow system prior to detection. In the hollow fiber module, cross-flow filtration takes place yielding two streams, namely the retentate and the permeate (see Figure 1). The beads remain in the retentate when a hollow fiber with a cutoff of 150200 kDa is employed, whereas compounds with a molecular mass below 150 kDa, such as FITC-avidin (∼66 kDa) and DTAF-(Fab)2 (∼100 kDa), pass the membrane. In this way, a particle-free permeate stream was obtained which allowed detection with a conventional flow-through detector. Reproducible injection of the sample suspension requires a homogeneous suspension. However, the large beads in the sample suspension sediment rapidly. As a consequence, the samples need to be mixed vigorously immediately before injection, which was achieved by aspirating and dispensing the suspension three times at a flow rate of 20 mL/min. Optimization of Reaction Conditions. Nonspecific Binding. Nonspecific binding (NSB) of proteins is a general problem in most types of biological assays.16 It commonly is reduced by adding nonionic surfactants, high salt concentrations, other proteins or small amounts of organic modifier to the buffers. The optimal conditions for reducing NSB depend on the affinity proteins employed. In the biochemical detection systems developed previously,17 NSB of the affinity protein occurred to the capillary walls of the reaction coils, resulting in an unstable baseline. Though unwanted, this effect generally levels off after a certain time when equilibrium has been reached. However, in the assay format described in this paper, NSB of the affinity protein to the beads of the sample suspension also needs to be considered. In order to evaluate NSB, the signals obtained for a number of materials (see Table 2) that do not allow specific binding, and accordingly are referred to as “blank materials”, were investigated. For avidin, one of the model targets, NSB is known to be pHdependent, with the lowest NSB observed at high pH.18 In the present system, high signals were initially obtained for injections of blank materials when using PBST as FI carrier. In order to overcome NSB, the FI carrier was changed to carbonate buffer, pH 8.7, containing 0.5% Tween 20 (CBT), which is recommended by the manufacturer as the binding buffer in affinity chromatography.19 As a consequence, the signals for the blank materials were reduced to levels below 200 arbitrary units. Blank- and IgGagarose lead to higher signals than avidin-agarose or streptavidin beaded agarose, which is due either to different NSB properties or to different injection peaks. Differences in NSB may occur (16) Gooding, K. M.; Regnier, F. E. HPLC of biological macromolecules; Marcel Dekker, Inc.: New York, 1990. (17) Lutz, E. S. M.; Oosterkamp, A. J.; Irth, H. Chim. Oggi 1997, 15, 11-5. (18) Dutramel, R. C.; Whitehead, J. S. In Avidin-Biotin Technology; Wilchek, M., Bayer, E. A., Eds.; Academic Press, Inc.: San Diego, 1990. (19) Pierce Catalogue, Rockford, IL, 1996.
Table 2. Nonspecific Binding of the Target with Two Different Carriersa signal in PBST (arb units)
RSD (n ) 3)
signal in CBT (arb units)
RSD (n ) 3)
FITC-avidinb 8.0 193 8.6 169 30.4 80 32.8 76
3.7 6.3 21.3 45.9
(B) Model Target: DTAF-(Fab)2c blank-sepharose 56 12.5 105 avidin-agarose nd 67 streptavidin-beads nd 59
0.7 14.9 2.4
(A) Model Target: blank-sepharose 865 IgG-sepharose 768 avidin-agarose 431 streptavidin-beads 481
a Signals for 20 µL injections of 50 µL/mL. The carrier (CBT and PBST, (A) and (B), respectively) was delivered at 0.4 mL/min, the labeled affinity protein solution at 0.2 mL/min. The time for the affinity reaction was 85 s. Fluorescence detection occurred at λexc ) 486 nm and λem ) 516 nm. b The affinity protein solution was 5 nmol/L FITCavidin. c The affinity protein solution was 4 nmol/L DTAF-(Fab)2.
due to differences in specific surface area and in surface functionalities. However, the fact that blank-agarose gives a signal similar to IgG-agarose indicates that a proteinaceous surface does not influence the signal. The observed differences are therefore most probably a consequence of different injection peaks due to differences in size and size distribution. In order to reduce the differences in size and size distribution between the ligand materials and the blank material, avidin-agarose was chosen as blank material for the first model system, since it was obtained from the same manufacturer as biotin- and iminobiotin-agarose. Under these circumstances, a significant difference was observed between the signals with the blank material and with biotin- and iminobiotin-agarose (see Figure 2A). According to the manufacturer, the same quantity of iminobiotin-agarose binds 50% less avidin than biotin-agarose. This was reflected in the lower signal with iminobiotin-agarose in comparison to the signal with biotin-agarose. Flow injection peaks obtained with this model system are shown in Figure 3A. In another experiment, we investigated NSB to the beads in the sample suspension by preincubating FITC-avidin with a 250fold excess of biotin prior to using it in the FI system. In this way, specific binding of FITC-avidin to the immobilized ligand is eliminated during the short reaction time available. As shown in Figure 2A, the signals for injections of the different solid phases were identical, regardless of the presence of immobilized ligand, demonstrating that the difference in signal for injections of biotin, iminobiotin, and the blank material when untreated FITC-avidin was used as target originate from specific binding. The higher signals for the blank when FITC-avidin that had been incubated with an excess of biotin was used are due to an increased quantum yield of FITC-avidin when occupying the binding sites,20 leading to increased background levels and accordingly to increased injection peaks. When the second model target, DTAF-(Fab)2 was used for ligand recognition, the signals obtained for injections of the blank materials were of an acceptable level (
