Label-Free Parallel Screening of Combinatorial

neous plate imaging by a CCD camera for the parallel detection of specific ... library synthesis.3-5 An interesting alternative to the on-bead screeni...
0 downloads 0 Views 180KB Size
Download PDF
Anal. Chem. 2002, 74, 834-840

Label-Free Parallel Screening of Combinatorial Triazine Libraries Using Reflectometric Interference Spectroscopy Oliver Birkert,*,† Rolf Tu 1 nnemann,‡ Gu 1 nther Jung,‡ and Gu 1 nter Gauglitz†

Institute of Physical Chemistry, University of Tu¨bingen, Auf der Morgenstelle 8, D-72076 Tu¨bingen, Germany, and Institute of Organic Chemistry, University of Tu¨bingen, Auf der Morgenstelle 18, D-72076 Tu¨bingen, Germany

The parallel reflectometric interference spectroscopy is presented as a label-free optical detection method. A new setup was adapted to accommodate sample carriers in a 96-well microplate. It allows for the first time simultaneous plate imaging by a CCD camera for the parallel detection of specific biomolecular interaction in the microplate wells at heterogeneous phase using direct optical monitoring. The detection of binding events with time resolution enables a highly parallel functional biomolecular interaction analysis (BIA). The combination of this new screening setup with combinatorial solid-phase synthesis is performed in the wells of glass-bottom microplates to accomplish the synthesis and the screening platform within one device. As a model system for a solid-phase substance library, synthesis of a triazine library and the subsequent BIA with four different antibodies were carried out. The presented setup enables a time resolution of 18 s with a total screening time of less than 35 min including baseline adjustment, BIA, and regeneration of the screening device for 96 samples in parallel. The binding studies reveal a fast classification of the different monoclonal and polyclonal antibodies and enable the detection of triazines with high binding affinity. The presented prototype is the first parallelized optical labelfree detection system for biomolecular interaction analysis that is suitable for a high-throughput screening based on the 96-well microplate format. Combinatorial chemistry has emerged as an important technique in the search and the optimization of target lead structures in the drug discovery process because of the large number of compounds that can be synthesized in parallel with a low expenditure of human labor.1 Additionally, recent improvements in screening-instrumentation hardware and data processing enabled the examination of biomolecular interaction behavior of compound libraries. The primary screening for lead compounds requires a large number of miniaturized syntheses. With regard to an efficient * Corresponding author: (phone) 49 7071 294667; (fax) 49 7071 295490; (e-mail) [email protected]. † Institute of Physical Chemistry. ‡ Institute of Organic Chemistry. (1) Jung, G., Ed. Combinatorial Chemistry; Wiley-VCH: Weinheim, 1999.

834 Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

solution of this challenge, combinatorial synthesis should be linked closely to the drug discovery process.2 The closest link between synthesis and screening can be achieved by conducting both on the same support. The easiest way is the performance of the screening using radiolabel- or fluorescence-based detection techniques directly on the resin beads after a “split and combine” library synthesis.3-5 An interesting alternative to the on-bead screening is the upcoming screening of microarrays on functionalized surfaces such as cellulose membranes6 or poly(ethylene glycol) membranes.7 If compounds are screened against different targets, regeneration of the screening array is necessary to run all the screening experiments on the same array. Otherwise, the arrays must be produced in duplicate, which is combined with an additional consumption of time and material. Besides optical detection methods based on labeling strategies, label-free optical detection for biomolecular interaction analysis, especially based on surface plasmon resonance, became more and more important because native protein-ligand interaction can be examined without modifying one of the binding partners and because of the high information content originated by these methods. However, until now it was not yet possible to adapt direct optical detection methods to highly parallel screening platforms providing high amounts of information on, for example, kinetic parameters and affinity constants. Screening of substance libraries was carried out sequentially or with a low level of parallelization even though their application in high-throughput screening is strongly propagated.8 Here, a first screening application of reflectometric interference spectroscopy (RIfS)910 is presented. RIfS is a label-free direct (2) Ganesan, A., Angew. Chem. 1998, 110, 2989-2992. (3) Lam, K. S.; Salmon, S. E.; Hersh, E. M.; Hruby, V. J.; Kazmierski, W. M.; Knapp, R. J. Nature 1991, 354, 82-84. (4) Furka, A. In Combinatorial Peptide and Nonpeptide Libraries; Jung, G., Ed.; Wiley-VCH: Weinheim, 1996; pp 111-137. (5) Rademann, J.; Groetli, M.; Meldal, M.; Bock, K. J. Am. Chem. Soc. 1999, 121, 5459-5466. (6) Frank, R.;. Overwin, H. Methods. Mol. Biol. 1996, 66, 149-169. (7) Scharn, D.; Wenschuh, H.; Reineke, U.; Schneider-Mergener, J.; Germeroth, L. J. Comb. Chem. 2000, 2, 361-369. (8) Myszka, D. G.; Rich, L. R., Pharm. Sci. Technol. Today 2000, 9, 310-317. (9) Abbreviations used: Ahx, aminohexanoic acid; CCD, charge coupled device; DAPEG, diaminopoly(ethylene glycol); DCM, dichloromethane; DIC, N, N′diisopropylcarbodiimide; DIEA, N,N-diisopropylethylamine; DMF, N,Ndimethylformamide; Fmoc, fluorenylmethoxycarbonyl; Glu, glutamic acid; HOBt, 1-hydroxybenzotriazole; OtBu, tert-butyl ester; PBS, phosphatebuffered saline; Phe, phenylalanine; RIfS, reflectometric interference spectroscopy. 10.1021/ac0106952 CCC: $22.00

© 2002 American Chemical Society Published on Web 01/18/2002

Figure 1. Scheme for the synthesis of the triazine library in the wells of the RIfS microplates: silanization, immobilization of the poly(ethylene glycol), and three-step synthesis of the triazine derivatives.

optical detection method which was recently adapted to a parallel microplate platform.11 Simultaneous plate imaging by a CCD camera allows the time-resolved detection of specific biomolecular interactions. We show both solid-phase synthesis of a compound library and subsequent screening for biomolecular interaction. Both procedures are performed on special microplates that were adapted for the RIfS screening setup. The upgraded sample carrier has the dimensions of a standard 96-well microplate and represents a system completely compatible with commercially available screening systems. With the parallel RIfS, the organic compound library is tested on the solid glass carrier. Therefore, all screening hits must be verified with the free compound in solution to determine the exact interaction parameters. This confirmation is necessary because the conjugation of a compound to a surface as well as to a carrier protein may result in a change of affinity due to spacer recognition.12 As a model system, we used the well-known synthesis of triazine libraries 7,13-15 via successive chloride substitution of trichlorotriazine by amines (Figure 1). The subsequent screening of the activity toward triazine antibodies was used as a model highthroughput screening assay. In contrast to typical screening targets, antibodies are bivalent. The speciality of the antibody system that the avidity may influence the binding parameters16 was not taken in account. EXPERIMENTAL PROCEDURES Equipment and Reagents. Common chemicals and biochemicals were purchased from Sigma (Deisenhofen, Germany) (10) Schmitt, H.-M.; Brecht, A.; Piehler, J.; Gauglitz, G. Biosens. Bioelectron. 1997, 12, 809-816. (11) Rothmund, M.; Schu ¨ tz, A.; Brecht, A.; Gauglitz, G.; Berthel, G.; Gra¨fe, D. Fresenius J. Anal. Chem. 1997, 359, 15-22. (12) Weller, M. G., Niessner, R., SPIE Proc. 1997, 3105, 341-352. (13) Stankova, M.; Lebl, M. Mol. Div. 1996, 2, 75-80. (14) Falorni, G.; Giacomelli, L.; Mameli, A.; Porcheddu, A. Tetrahedron Lett. 1998, 39, 7607-7610. (15) Gustafson, G. R.; Baldino, C. M.; O'Donnell, E. O.; Sheldon, A.; Tarsa, R. T.; Verni, C. J.; Coffen, D. L. Tetrahedron 1998, 54, 4051-4065.

and Fluka (Neu-Ulm, Germany). Diaminopoly(ethylene glycol) (DAPEG, mean molar mass of 2000 g/mol) was from Rapp Polymere (Tu¨bingen, Germany). Fmoc-protected amino acids were purchased from NovaBiochem (Bad Soden, Germany). The polyclonal anti-triazine antibodies were supplied by Dr. Ram Abuknesha, Kings College (London, England). Monoclonal antibodies K4E7 and K1F4 were supplied by Prof. Bertold Hock, TU Mu¨nchen (Mu¨nchen, Germany). Interference transducers were produced by Schott (Mainz, Germany) in an ion plating process. The optical detection setup as well as the gluing of the RIfS microplates was carried out by Carl Zeiss Jena GmbH (Jena, Germany). The parallel sample handling device was realized by CyBio (Jena, Germany). Principle of Reflectometry. The RIfS allows the optical online detection of specific biomolecular interaction without using labeling techniques. It is based on the spectral distribution of reflectance from thin transparent layers on a glass substrate. Monochromatic light beams reflected at both sides of an interference layer (330-nm thickness) lead to a interference pattern from which the optical thickness of a thin layer may be deduced. Changes in the thickness of the layer surface caused by binding of biomolecules at this modified surface lead to a shift of the interference pattern which is evaluated in real time. Recently, we parallelized our RIfS experimental setup enabling high-throughput screening in 96-well microplates. This was accomplished by using a halogen light source, a turnable filter wheel with seven interference filters to address light of different wavelengths in the range of 516-707 nm, and a 12-bit CCD camera (EEV, type CCD05-20) as detection unit. To achieve compatibility with common test formats for combinatorial compound libraries, the transducer was made by gluing the bottomless scaffold of a 96-well plastic microplate onto the transducer slide. (16) Strachan, G.; Grant, S. D.; Learmonth, D.; Longstaff, M.; Porter, A. J.; Harris, W. J. Biosnens. Bioelectron. 1998, 13, 665-673.

Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

835

Table 1. Amino Acids and Amines Used for the Synthesis of the Triazine Collectiona

Figure 2. Scheme of the RIfS setup for microplate format screening. From the white light source (a) seven wavelengths are filtered sequentially and illuminate the RIfS microplate from beneath. The light passes the bulk carrier glass slide (c) and is reflected partially at the bottom and the top of the thin interference layer (d) consisting of a 330-nm silicium dioxide layer and the functionalized poly(ethylene glycol) layer. The reflected light beam (I1 + I2) results in a λ-dependent interference pattern. The intensities of the reflected beams are registrated by the CCD camera (f) and allow the calculation of the optical thickness of layer d. The plastic grid (e) has the dimensions of a standard 96-well microplate. Liquid handling is accomplished by a 96-pipet robot (g).

For suppression of nonspecific binding of biomolecules to the sensor surface, the transducers were coated with DAPEG.17 Furthermore, this biocompatible coating provides anchor groups for the solid-phase synthesis. A scheme of the transducer plate is shown in Figure 2. With this setup, all wells of a microplate can be addressed and read out on-line and in parallel with a time resolution of 18 s. As is typically the case, if diffusion and binding of antibodies to the transducer surface occurs at a slower time than 18 s, binding events at the surface can be monitored on-line. Preparation of DAPEG Transducers. DAPEG was covalently attached onto glass transducers using a modified protocol of Birkert et al.18 The interference transducers were cleaned in freshly prepared, hot H2SO4/H2O2 (6:4), rinsed with water, and dried in a nitrogen stream. The activated transducers were silanized with glycidoxypropyltrimethoxysilane (400 µL) for 1 h. DAPEG (2 mg) was dissolved in dichloromethane (5 mL), and the solution was dropped on the cleaned transducers to obtain an uniform coating. After evaporation of the solvents, the transducer was kept at 70 °C for 8 h. Thereafter, the transducer was rinsed with water and dried in a nitrogen stream. After the transducer was glued to the microplate grid with a silicone glue,19 the triazine derivatives and the reference compounds were synthesized by a combinatorial chemistry approach. Solid-Phase Synthesis of the 1,3,5-Triazine Library. Solutions (100 µL) of the respective Fmoc-amino acid derivative (0.2 M in DMF; see Table 1), HOBt (0.2 M in DMF), and DIC (0.2 M in DMF) were pipetted into the wells of the DAPEG-coated RIfS microplate. After a coupling time of 16 h at room temperature, the wells were washed with a continuous flow of DMF. The Fmoc (17) Piehler, J.; Brecht, A.; Valiokas, R.; Liedberg, B.; Gauglitz, G. Biosens. Bioelectron. 2000, 15, 473-481. (18) Birkert, O.; Haake, H.-M.; Schu ¨ tz, A.; Mack, J.; Brecht, A.; Jung, G.; Gauglitz, G. Anal. Biochem. 2000, 282, 200-208. (19) Nawracala, B.; Berndt, M.; Gauglitz, G.; Elender, G.; Graefe, D.; Berthel, G. United States Patent 6,018,388, 2000.

836 Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

a The surface-bound triazine derivatives are indicated in the text via the related numbers and letter code. *Each compound was synthesized in two neighboring wells.

group was removed by treatment with piperidine (20 % in DMF, 200 µL/well, 20 min). After successive washing with DMF, methanol, and DCM in continuous flow, 200 µL of a 1:1 (v/v) solution of trichlorotriazine (0.1 M in DCM) and DIEA (0.2 M in DCM) was added into each well. After 2 h, the wells were washed thoroughly with DCM. The amines used for the second randomization are listed in Table 1. Solutions (200 µL) of the respective amine (0.2 M in DMF) were added. After 2.5 h, the solutions were removed and the microplate was washed with DMF, methanol, and DCM. Assay Format. Specific binding of antibodies to the triazinecoated surfaces was examined using a direct assay format. The wells of the RIfS microplates were simultaneously filled by the

Figure 3. Binding signal in one well of a 96-well RIfS microplate. The inset shows the events in the microplate cavity. Data registration takes place every 18 s. After confirmation of the baseline stability, buffer was replaced by antibody solution and binding of the antibody to the surfacebound triazine derivative occurred. To regenerate the sensor surface, pepsin solution was pipetted into the microplate wells. Finally, we checked for complete regeneration by measuring a baseline with buffer solution.

programmed pipetting robot. First, PBS buffer (80 µL) was filled into each well and the baseline was recorded in order to determine signal stability. Thereafter, buffer was removed and the sample solution (80 µL, PBS), containing antibody (20 mg/L; 133.3 µM) and ovalbumin (400 mg/L) was filled into the wells. Ovalbumin was used in high concentration as a background protein to prevent nonspecific binding of the antibody to the plastic grid of the microplate. After the binding event was recorded, the solutions were removed and the microplate was regenerated by digestion of the antibody with pepsin solution (80 µL, 9.000 units/mL, pH 1.9). RESULTS AND DISCUSSION To show the potential of the new parallel RIfS setup in combination with combinatorial solid-phase synthesis on the transducer plates in the field of the high-throughput screening, we performed an exemplary screening test by using antibody binding studies. This feasibility study was carried out with antibodies against triazines (polyclonal antibody I, monoclonal antibodies K4E7 and K1F4). As a reference we used an antiisoproturone antibody (polyclonal antibody II) for which no binding affinity to the class of triazines can be expected. Characterization of the Transducers. Typical binding curves for antibodies in two wells of a 96-well plate are shown in Figure 3. After confirmation of the baseline stability, the buffer was replaced by antibody solution and binding of the antibody to the surface-bound triazine derivative occured (solid line). To

regenerate the sensor surface, pepsin solution was pipetted into the microplate wells. The large increase in the binding signal during the regeneration is caused by the difference in the refractive index of plain buffer and regeneration solution (optical effect) and by conformational changes in the polymer surface due to the change in pH (chemical effect). To ensure the regeneration efficiency, the pepsin solution was substituted by buffer solution and the baseline was recorded. After the complete cycle, the next binding study can be carried out with the fully regenerated transducer. For screening experiments with 96 samples, the assay time per plate was reduced to 35 min with an antibody binding time of 900 s. The strength of the affinity binding is given by the binding signal at the end of the measurement, just before regeneration. For data reduction, the evaluated registration time was minimized to 2 times 18 s at the beginning and at the end of the binding event. The degree of binding was determined by measuring the total increase of the optical thickness. Moreover, this procedure reduces the demands on the evaluation software and is suitable for primary screening with high sample throughput. Because nonspecific binding is a common problem in labelfree optical detection, a proof of the specificity of the interaction of the antibodies with the triazine surface was carried out by an inhibition assay prior to carrying out the screening studies. The signal of the free polyclonal anti-triazine antibody (solid line, Figure 3) was compared with the binding signal of the same Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

837

Figure 4. Binding curves of four antibodies: K4E7 (4.1), polyclonal anti-simazine (4.2), K1F4 (4.3), and polyclonal anti-isoproturone (4.4).

antibody preincubated with its parent molecule 4-chloro-6-(isopropylamino)-1,3,5-triazine-2-(6′-amino)hexanoic acid (atrazinecaproic acid, 25 mg/L) (dotted line, Figure 3). The finding that the preincubated antibody is not able to interact with the surfacebound triazine 2e gives proof of the specifity of the binding as well as of the quality of the transducer surface. The triazine-coated transducer is reusable. The regeneration with pepsin completely detached the antibody from the transducer but did not affect the surface-bound triazines. Hence, the synthesized compound library can be screened against a variety of different targets in subsequent screening runs. The binding capacity of the plate decreased only slightly after each assay cycle. A total decrease of less than 20% after 20 measurements was found. Antibody Binding Studies. Figure 4 summarizes the binding signals successively recorded for all four antibodies in the 96 wells of one RIfS microplate and provides an overview of the antibody binding behavior and specifity to the different triazine derivatives. In contrast to most of the binding studies carried out with triazines in the field of environmental analytics, we used triazines with bulkier residues. As the size of the triazine has an influence on affinity and cross-reactivity,20 the screening results differ from studies with atrazine-like triazines. As expected, the polyclonal antibody II, which has no activity toward triazine derivatives, showed no significant binding to the 838

Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

transducer surface (Figure 4.4). Moreover, this confirms that nonspecific binding to the surface is negligible. Different binding patterns were obtained for the reaction of three antibodies with the triazine derivatives. Clearly, evaluable signals were obtained within a measurement time of less than 750 s and susperseded the detection of the equilibrium binding, which after 750 s was still not reached. If equilibrium binding is not important, the total assay time can be reduced to time frames that are acceptable for a fast primary screening. The increase of the optical thickness ranged from 0.7 to 1.0 nm within 750 s. The results show binding curves for all of the synthesized triazine derivatives. Even though the limited sensitivity of the screening setup demands high antibody concentrations, only negligible binding was detected in the 24 reference wells with the polyclonale anti-simazine antibody and with K4E7. This demonstrates the discrimination between active and inactive compounds in a high-throughput screening. Furthermore, the binding signals gave supplementary information about the crossreactivity of the antibodies toward surface-bound triazines. K1F4 shows higher nonspecific binding which decreases the signal range for hit identification. (20) Bruun, L., Koch, C., Jakobsen, M. H., Pedersen, B., Christiansen, M., Aamand, J. Anal. Chim. Acta 2001, 436, 87-101.

Figure 5. Three-dimensional plot of the binding signals obtained after 750 s with a polyclonal anti-simazine antibody. White without pattern: nonactive reference samples. Compounds that differ only in residue R1 are plotted in the same pattern. Two neighboring columns with the same pattern correspond to the same compound synthesized in duplicate.

Binding of the monoclonal antibody K1F4 results in high binding signals for all triazine derivatives. This very strong crossreactivity makes it less interesting for structure-binding relationship studies but useful for tests where a whole class of substances has to be detected in sum. This shows the ability of our RIfS instrument to classify a number of targets due to their use in addition to the use of the setup in standard high-throughput screening applications. The antibodies I and K4E7 show small signals in the reference wells and a wide difference in response in binding to the different triazine derivatives. A study of this kind would allow research to deduce relationships between structure and affinity.

The three-dimensional plot in Figure 5 shows the binding signal after 750 s in each well of the RIfS microplate for the polyclonal anti-triazine antibody I as a typical screening result. Triazine derivatives with the same building block R1 but different building block R2 are plotted with the same pattern. Two neighboring wells contain the same triazine derivative. For a fast hit identification in a high-throughput screening, such graphic displays are very common. The exact binding data for all 36 triazines with the polyclonal antibody I and K4E7 are listed in Table 2. Generally, the amine residue R2 of the triazine derivatives has a greater influence on the affinity of both the polyclonal antitriazine antibody I and K4E7 than the amino acid residue R1. This can be explained by the better accessibility of the free R2 residue. Especially the triazines 1e, 2e, 3e, 1h, 2h, and 3h and also 1f, 2f, and 3f with small residues R2 show a high affinity to both antibodies. In contrast, for the substances 1b, 2b, and 3b containing the bulkiest amino substituent R2, binding of both antibodies was hindered. Hence, the RIfS measurement reveals that the size of R2 is the most critical factor for the investigated antibodies and bulky residues lead to a weaker interaction for both antibodies. Even though the antibody affinity toward the triazine derivatives is mainly influenced by the residue R2 as shown in the screening plot (Figure 5), a closer look at the binding signals in Table 2 points to the effect of residue R1. Binding data of six selected compounds (1a, 2a, 3a, 1e, 2e, 3e) shall be discussed in detail. The monoclonal antibody K4E7 shows a binding behavior identical to that of the triazines 1a, 2a, and 3a, which contain different residues R1 but the same bulky phenylalanine methylester residue R2. This proves that the binding behavior is mostly influenced by the more accessible residue R2 and that residue R1 has no significant influence on the biomolecular interaction. This effect was observed for most of the triazines examined in this studie. However, for 1e, 2e and 3e containing different residues R1 but the same small isopropylamine residue R2, the binding behavior to the monoclonal antibody K4E7 depends strongly on

Table 2. Binding Signals of the Reaction of Two Antibodies with the Triazine Collection

a

triazine

polyclonal antibody signal (nm)

K4E7 signal (nm)

triazine

polyclonal antibody signal (nm)

K4E7 signal (nm)

1a 1b 1c 2a 2b 2c 3a 3b 3c 1d 1e 1f 2d 2e 2f 3d 3e 3f

0.46 ( 0.001 0.22 ( 0.014 0.62 ( 0.115 0.30 ( 0.036 0.16 ( 0.009 0.65 ( 0.020 0.365 ( 0.035 0.16 ( 0.018 0.70 ( 0.1140 0.45 ( 0.011 0.91 ( 0.017 0.79 ( 0.043 0.37 ( 0.009 0.62 ( 0.154 0.69 (0.059 0.34 ( 0.002 0.94 ( 0.014 0.83 ( 0.017

0.49 ( 0.042 0.33 ( 0.059 0.53 ( 0.049 0.45 ( 0.040 0.39 ( 0.033 0.50 ( 0.013 0.45 ( 0.004 0.34 (0.011 0.61 ( 0.042 0.49a 0.65 ( 0.005 0.55 ( 0.038 0.44 ( 0.007 0.48 ( 0.045 0.55 ( 0.018 0.42 ( 0.023 0.80 ( 0.018 0.61 ( 0.017

1g 1h 1i 2g 2h 2i 3g 3h 3i 1j 1k 1l 2j 2k 2l 3j 3k 3l

0.50 ( 0.011 1.06 ( 0.049 0.58 ( 0.034 0.48 ( 0.016 0.93 ( 0.071 0.51 ( 0.051 0.45 ( 0.086 0.86 ( 0.046 0.54 ( 0.027 0.65 ( 0.025 0.60 ( 0.038 0.62 ( 0.056 0.50 ( 0.016 0.62 ( 0.008 0.74 ( 0.049 0.71 ( 0.020 0.60 ( 0.081 0.81 ( 0.027

0.56 ( 0.030 0.67 ( 0.020 0.33 ( 0.080 0.50 ( 0.027 0.55 ( 0.033 0.31 ( 0.017 0.52 ( 0.023 0.73 ( 0.015 0.22 ( 0.056 0.55 ( 0.057 0.46 ( 0.106 0.38 ( 0.042 0.42 ( 0.013 0.37 ( 0.043 0.43 ( 0.012 0.62 ( 0.035 0.50 ( 0.073 0.39 ( 0.107

Single measurement.

Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

839

the residue R1. The strongest interaction was detected with the Phe residue R1; the weakest interaction was found with the Ahx residue R1. Since biomolecular interaction depends on both binding partners, for the antibody I, a different performance can be expected. In fact, we found a noticeable difference in binding activity for 1a, 2a, and 3a in their binding to the polyclonal antibody I whereas with this antibody the differences found for 1e 2e, and 3e are less distinct. Even though the RIfS signal pattern seems to be nearly identical (Figure 5), a detailed study of the binding curves for the different antibodies reveals remarkable differences in binding behavior. CONCLUSION Exemplified by the synthesis and screening of a library of substituted triazines, we determined the scope and limitations of the parallel label-free screening in combination with a combinatorial solid-phase synthesis in a high-throughput screening. The main advantages of the presented approach are the combination of synthesis and screening, the short time for one screening run, the possibility for further miniaturization, and the label-free analysis of the samples. (21) Piehler, J.; Brecht, A.; Gauglitz, G.; Maul, C.; Grabley, S.; Zerlin, M. Biosens. Bioelectron. 1997, 12, 531-538. (22) Haake, H. M.; Tu ¨ nnemann, R.; Brecht, A.; Austel, V.; Jung, G.; Gauglitz, G., submitted to Angew. Chem.

840

Analytical Chemistry, Vol. 74, No. 4, February 15, 2002

The main limitation of this new approach is the low sensitivity. Only high molecular mass molecules such as antibodies or receptors can be used for detection. We are working on further developments of the optical setup that will allow the detection of the binding of small molecules. Piehler et al.21 and Haake et al.22 have already shown that organic reactions may be monitored directly using a single-channel setup. With such an upgraded system, not only the screening but also the on-line monitoring of the combinatorial library syntheses will be possible within one setup. Furthermore, on-line recording of binding curves would offer information about the affinity constant as well as the kinetic parameters of the binding events. Evaluation of the whole binding curve is of interest for secondary screening and hit validation where a high information content for a smaller number of samples is necessary. ACKNOWLEDGMENT Part of this work was funded by BMBF project LIBRARIAN II/0310838. O.B. and R.T. were supported by the DFG-Graduate colleage “Quantitative Analysis and Characterisation of Pharmaceutically and Biochemically relevant Substances” at the University of Tu¨bingen and by the Fonds der Chemischen Industrie.

Received for review June 20, 2001. Accepted October 30, 2001. AC0106952