- Email: [email protected]
Analysis of small-molecule interactions using Biacore S51 technology
ANALYTICAL BIOCHEMISTRY Analytical Biochemistry 329 (2004) 316–323 www.elsevier.com/locate/yabio
Analysis of small-molecule interactions using Biacore S51 technology David G. Myszka¤ Center for Biomolecular Interaction Analysis, University of Utah, 50 N. Medical Drive, School of Medicine, Rm 4A417, Salt Lake City, UT 84132, USA Received 29 December 2003 Available online 6 May 2004
Abstract Biacore S51 is a new surface plasmon resonance-based biosensor developed by Biacore AB (Uppsala, Sweden). The instrument was engineered speciWcally to support small-molecule drug discovery and development. The platform includes increased sensitivity, larger sample handling capabilities, and automated data processing to improve throughput. Compared to previously released Biacore instruments, the most signiWcant design change relates to the introduction of the hydrodynamic-addressing Xow cell. This design allows two reaction surfaces and a reference surface to be placed within the same Xow cell, thereby improving data quality and extending the kinetic range of the instrument. Using a set of small-molecule inhibitors of the enzyme carbonic anhydrase II, we tested the reproducibility, sensitivity, and dynamic range of the biosensor. Given the S51’s performance capabilities, it should play an active role in secondary screening by providing high-resolution information for small-molecule/target interactions. 2004 Elsevier Inc. All rights reserved.
Optical biosensors are capable of monitoring the reversible associations of biological molecules in real time. Biacore AB (Uppsala, Sweden) released the Wrst commercial biosensor for biomolecular interaction analysis in 1990. Since that time, the company has developed a Xeet of instruments based on surface plasmon resonance (SPR)1 detection. SPR is sensitive to the refractive index of the buVer, which changes as molecules in solution interact with the target immobilized on the surface. These instruments allow the thermodynamic and kinetic binding properties of biological molecules to be routinely determined using microgram quantities of material and without labeling. A signiWcant advancement in SPR biosensor hardware was made with the release of Biacore S51 from Biacore AB. This platform, which represents an entirely new line of Biacore instruments, displays increased sensitivity, higher throughput, greater automation, and ¤
Fax: 1-801-585-3015. E-mail address: [email protected]. 1 Abbreviations used: SPR, surface plasmon resonance; CAII, carbonic anhydrase II; EDC, N-ethyl-N0-(3-dimethylaminopropyl)carbodiimide; NHS, N-hydroxysuccinimide; PBS, phosphate-buVered saline; Rmax, maximum response; RU, response unit. 0003-2697/$ - see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ab.2004.03.028
enhanced ease of use over previously available platforms. The Xow cell system in Biacore S51 has been completely redesigned and represents the biggest change compared to earlier platforms. In-line referencing is achieved through a proprietary hydrodynamic addressing system, which enables immobilization of two ligands and inclusion of a reference spot within the same Xow cell (Fig. 1). A Y-inlet uses a buVer stream to limit sensor surface coupling chemistry to the two side regions of the Xow cell, while the middle of the sensor surface remains isolated from coupling conditions. In Biacore S51, three detection spots are simultaneously addressed in each Xow cell, thereby eliminating lag time between reference and sample, which improves data quality and permits the resolution of fast kinetics. In addition, simultaneous sample delivery ensures constant analyte concentrations and reduces systematic noise due to temperature and pumping artifacts. Together, these improvements in instrument hardware increase the sensitivity of Biacore S51 relative to Biacore 2000 and 3000 platforms. To evaluate the performance of Biacore S51 and highlight how it may be applied in a drug discovery setting, we analyzed the binding of a number of smallmolecule inhibitors interacting with the enzyme carbonic
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
317
groups were blocked with a 7-min injection of 1.0 M ethanolamine, pH 8.5. DiVerent levels of CAII were immobilized for diVerent experiments as described under Results. Fig. 1. Schematic of Biacore S51 Xow cell. Two separate inlets make it possible to immobilize ligands on diVerent sides in the Xow cell, leaving the center detection spot unmodiWed as a control surface. In this illustration, the ligand for immobilization is being introduced through inlet 1 at a slower Xow rate than the running buVer in inlet 2.
anhydrase II. We have used this enzyme system in the past to compare the binding thermodynamics and kinetics obtained from the biosensor with solution-based methods [1]. And recently, it was used to compare the quality of data obtained from 40 diVerent investigators to gauge experimental variability associated with biosensor analysis [2]. The compounds used in the present study varied in molecular mass from 340 to 95 Da and displayed a »10,000-fold diVerence in aYnity for the enzyme. Binding response data were highly reproducible and changes in response as low as 0.50 RU (0.5 pg/mm2) could be reliably determined. The binding constants for the compound series were compared with regard to aYnity and kinetic components. These detailed mechanistic data with regard to compound binding provide new information to optimize a compound’s activity during drug development.
Materials and methods Materials Biacore 2000, Biacore 3000, Biacore S51, CM5 series sensor chips, and coupling reagents (N-ethyl-N0-(3dimethylaminopropyl)carbodiimide (EDC), N-hydroxysuccinimide (NHS), and ethanolamine-HCl) were purchased from Biacore AB. Carbonic anhydrase isozyme II (CAII) from bovine erythrocytes, sulfonamide inhibitors, and all other general-use reagents were purchased from Sigma Chemical Co. (St. Louis, MO). CAII immobilization CAII was immobilized onto CM5 sensor chips using standard amine coupling [3]. Phosphate-buVered saline (PBS), which consisted of 20 mM Na2HPO4 and 150 mM NaCl, pH 7.4, was used as a running buVer. The carboxymethyl dextran surface within one side of the Xow cell was activated with a 7-min injection of a 1:1 ratio of 0.4 M EDC and 0.1 M NHS. The enzyme was coupled to the surface with a 7-min injection of CAII diluted in 10 mM sodium acetate, pH 5.0. Remaining activated
Sulfonamide binding experiments Biosensor experiments were carried out on the following sulfonamide-based inhibitors: acetazolamide, benzene-disulfonamide, benzene-sulfonamide, carboxybenzenesulfonamide, dansylamide, sulfanilamide, sulpiride, and methylsulfonamide. Each compound was dissolved directly in the PBS running buVer and analyzed using a two- or threefold dilution series. The highest compound concentration varied, but all compounds were tested at 10 diVerent concentrations and each concentration was tested at least Wve times. The ultra-highresolution analysis of carboxybenzenesulfonamide involved the testing of 96 diVerent concentrations using a 4% dilution series and each concentration was tested four times. Within a given compound concentration series, the samples were randomized to minimize systematic errors. All of the bound complexes dissociated back to baseline within a reasonable time frame; therefore, no regeneration was required. Data processing and Wtting All sensorgrams were processed by Wrst subtracting the binding response recorded from the control surface (center reference spot), followed by subtracting an average of the buVer blank injections from the reaction spot [4]. To determine kinetic rate constants, all data sets were Wt to a simple 1:1 interaction model including a term for mass transport using numerical integration and nonlinear curve Wtting [5]. Equilibrium analysis was performed for methylsulfonamide by Wtting the response at the end of the association phase to a single-site binding isotherm.
Results Performance comparison To compare the performance of Biacore S51 with Biacore 2000 and 3000 instruments, identical binding studies using acetazolamide injected over CAII-immobilized surfaces (»3000 RU) were carried out. Acetazolamide was injected at seven diVerent concentrations and each concentration was repeated three times. The data in Fig. 2 shows the binding responses obtained for the small molecule analyte as measured on the diVerent platforms. While similar overall intensities were obtained from each instrument, the quality of the data as judged by the smoothness of the binding response and the lower
318
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 2. Comparison of binding response data from diVerent Biacore platforms. Acetazolamide at concentrations of 1000, 333, 111, 37, 12.3, 4.1, and 1.37 nM was injected over similar densities (»3000 RU) of CAII immobilized within (A) Biacore 2000, (B) Biacore 3000, and (C) Biacore S51. Each concentration was tested in triplicate. The response data were processes using data from a reference surface and buVer injections. To make an easier comparison of the binding responses the data from all instruments are shown at a rate of 1 Hz.
overall random noise is signiWcantly better on Biacore S51 compared to the 2000/3000 series of instruments. Also, having the reference and reaction surface within the same Xow cell in Biacore S51 minimizes the spikes often observed at the beginning and end of sample injections. Small-molecule binding responses To demonstrate the ability of Biacore S51 technology to measure the kinetic interactions of small molecules with immobilized macromolecules, we carried out an analysis of a number of sulfonamide compounds binding to CAII. Fig. 3 shows the sensorgram data collected for eight of the inhibitors interacting with a 5000-RU enzyme surface. The molecular masses of the compounds varied from the highest mass of 341 Da for sulpiride to the lowest of 95 Da for methylsulfonamide. All compounds were assayed over a 1000-fold concentration range using a 2-fold dilution series. Each concentration was also injected Wve times to test the reproducibility of the assay. All of the binding responses were exceptionally reproducible, since there was very little deviation between the repeated analyte concentrations. This indicates that the immobilized CAII did not lose activity over the time course of the assay. Data quality In general, the quality of the binding data itself is very high. The signals for most of the compounds were »100 times greater than the short-term random noise. As expected, even at saturating concentrations, the lowest signals were observed for methylsulfonamide (Fig. 3H), which is consistent with it having the lowest molecular mass (95 Da). However, even these low responses (1– 4 RU) were reproducible and could be analyzed to extract out binding constants.
Kinetic and equilibrium analysis The red lines in Fig. 3 represent a global Wt of the response data to a 1:1 interaction model that included a term for mass transport. All of the data sets for the various inhibitors were well described by this simple model. This illustrates that it is possible to Wt a number of interactions recorded on the biosensor with a simple model. The kinetic rate constants and aYnities determined for each compound are presented alongside the compound structure in Fig. 3. Because the reaction and reference surfaces are within the same Xow cell in Biacore S51 (as opposed be being aligned in a series on Biacore 2000 and 3000 instruments), it is possible to resolve the dissociation kinetics for very rapidly dissociating systems. For example, it is possible to determine a dissociation rate of 0.64 s¡1 for sulpiride (see Fig. 3G). The dissociation phase data for methylsulfonamide (see Fig. 3H) are perhaps just beyond the limit of determination at 2.2 s¡1. In this case, it would be diYcult to determine whether the actual dissociation rate was faster than what we observed. The limit for the dissociation rate constant in Biacore S51 is around 1 s¡1, which is about 10-fold higher than that on the Biacore 2000 and 3000 instruments. For this reason we performed an equilibrium analysis of the methylsulfonamide data. The inset in Fig. 3H illustrates that the equilibrium response data Wt well to a simple 1:1 binding isotherm. Comparison of binding constants A visual comparison of the kinetic binding constants for the sulfonamide inhibitors can be made by plotting the association versus the dissociation rates for each compound (Fig. 4). In this plot, data points to the right on the x axis represent faster association rates while data points further up on the y axis represent slower
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
319
Fig. 3. Sensorgrams for sulfonamide-based inhibitors interacting with a 5000-RU CAII surface. The highest concentration for each compound was as follows: (A) acetazolamide, 1 M; (B) furosemide, 30 M; (C) dansylamide, 6 M; (D) benzenesulfonamide, 100 M; (E) carboxybenzenesulfonamide, 40 M; (F) sulfanilamide, 100 M; (G) sulpiride, 2 mM; (H) methylsulfonamide, 2.5 mM. Each compound was injected over a 1000-fold concentration range using a 2-fold dilution series. Each concentration was injected Wve times and the entire data set was globally Wt to a 1:1 interaction model as shown by red lines. The compound structure, name, molecular mass, and binding constants are provided on each data set. The inset in H represents an equilibrium analysis of the methylsulfonamide data set. The number in parentheses represents the standard error in the last signiWcant digit.
dissociation rates. The diagonal lines represent equilibrium isotherms from 1 mM to 10 nM. Compounds with the highest aYnity appear in the upper right-hand quadrant of the plot and the lowest aYnity will be in the
lower left-hand corner. Comparing data for the nine sulfonamide inhibitors, there is an »10,000-fold diVerence in aYnity between the weakest binder, methylsulfonamide (KD D 270 M), and the tightest binder, acetalzolamide
320
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 4. Kinetic proWle plot of nine sulfonamide-based inhibitors of CAII. Rate constants were determined from Biacore S51. The diagonal lines represent equilibrium binding isotherms.
(KD D 19 nM). Three compounds, benzenesulfonamide, benzenedisulfonamide, and danyslamide, lay in a similar position on the plot, illustrating that they have similar association and dissociation kinetics for CAII. The aYnities for these compounds (»850 nM) are also similar to the aYnity determined for carboxybenzenesulfonamide (893 nM), since they all sit near the 1 M binding isotherm. Therefore, if looking at only the equilibrium dissociation constant, one would conclude that the addition of a carboxyl group to benzenesulfonamide does not eVect the interaction because the aYnities are the same. However, carboxybenzenesulfonamide binds with diVerent kinetics. For example, compared to benzenesulfonamide, the association and dissociation rates for carboxybenzenesulfonamide are each about four times slower. The rate constants changed in equal magnitude so that the overall aYnity remains around 1 M. The aYnity is 6-fold weaker when the carboxyl group on carboxybenzenesulfonamide is converted to an amine in sulfanilamide (Fig. 3F). The association rate for sulfanilamide decreases 2-fold while the dissociation rate increases 3-fold compared to those of carboxybenzenesulfonamide. This detailed kinetic information on the binding interaction is useful when analyzing structure/ function relationships of target/inhibitor interactions, providing more information than equilibrium assays alone. Ultra-low-capacity surfaces An advantage of the Biacore S51 technology is that it is capable of measuring reliable responses even for very low signals. To illustrate this capability, Wrst we analyzed the interaction of a low-molecular-mass analyte, methylsulfonamide (95 Da), as shown in Fig. 3H. To further test the limits of analyte detection, we collected binding data for carboxybenzenesulfonamide over an ultra-low-
capacity surface of CAII. In this case we purposefully immobilized only 100 RU of the enzyme with the intention of measuring a maximum binding response of 01 RU. The sensorgrams in Fig. 4A show the binding responses for a series of carboxybenzenesulfonamide over the ultra-low-capacity CAII surface. The maximum signal appears to be around 0.75 RU for the highest concentrations of the analyte. These data were Wt to a 1:1 interaction model to determine binding constants for the interaction as shown in Fig. 5A. The rate constants determined from the ultra-low-density CAII surface (Fig. 5A) compare very well with the values determined from the higher-density surface (Fig. 3E). However, with all the analyte concentrations overlaid in Fig. 5A it becomes diYcult to discern the concentration-dependent binding responses because of the random noise. To more readily resolve the diVerent binding proWles, the 10 diVerent analyte concentrations are plotted individually in Fig. 5B. From a visual inspection of these data it is possible to observe the expected concentration-dependent binding responses and to assess the quality of the Wt to the reaction model. These results illustrate that it is possible to accurately interpret binding responses that have a maximum signal of less than 1 RU. In this case, the calculated Rmax for the surface was 0.6 RU. High-resolution analysis An added feature of Biacore S51 is the ability to analyze data from a 384-well plate. To test the overall reliability and robustness of the system, we analyzed 96 diVerent concentrations of carboxybenzenesulfonamide, each injected four times over a 4000-RU CAII surface. The binding response from each analyte concentration was highly reproducible, since they overlapped with one another, as shown in Fig. 6A. The data for all 384 sensorgrams (a total of 619,200 data points) were globally Wt to a 1:1 interaction model. Fig. 6B shows a plot of the residuals from the entire data set, which are low and randomly distributed about zero. The standard deviation in the residuals was 0.25 RU, which is similar to the standard deviation of the replicate experiments of 0.24 RU, indicating that the simple model does an excellent job of describing the data. Again, the rate constants determined from this high-resolution analysis (Fig. 6A) are consistent with the values determined from our previous studies in Figs. 3E and 5A. The results of this ultra-high-resolution analysis illustrate that the instrument is capable of collecting exceptionally reproducible data over 384 assays.
Discussion As experimentalists interested in characterizing molecular interactions, we are fortunate to have access to commercial biosensor technology. While the
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
321
Fig. 5. Binding responses for carboxybenzenesulfonamide over an ultra-low-density CAII surface (100 RU). (A) Overlay of binding data for the inhibitor from 40 M to 19.5 nM using a twofold dilution series. (B) Responses for the individual inhibitor concentrations. Red lines represent a global Wt of the binding data to a 1:1 interaction model.
322
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 6. Ultra-high-resolution analysis of carboxybenzenesulfonamide/CAII interaction. (A) Inhibitor was injected using a starting concentration of 100 M in a 4% dilution series, giving a total of 98 diVerent concentrations. Each concentration was repeated four times for a total of 384 samples tested over the same surface. Red lines represent a global Wt for the data to a 1:1 interaction model. (B) Residual plot based on the Wt of the primary data.
biosensor platforms released in early 1990s ushered in a new way of characterizing macromolecular interactions, evolving technology such as Biacore’s S51 platform is making it possible to apply the advantages of label-free and real-time analysis to small-molecule systems in a routine way. All of the major components of the S51 have been reengineered relative to the Biacore 2000 and 3000 technology. The S51 detection unit has been redesigned with state-of-the-art optics, light source, and detector. Automated and integrated loading of sensor chips within the optical detection area permits a closed detection environment for improved thermal stability. The instrument can operate from 4 to 45 °C and the sample racks are protected from light. Higher throughput is made possible by incorporating 384-well microtiter plate compatibility and fast sample loading via a new injection technique. The injection needle is Wxed directly to the Xow cell and the sample racks move to address each sample. The result is decreased sample carry-over, faster sample loading, and the potential for future parallel sample processing possibilities.
Our analysis of a set of sulfonamide inhibitors binding to CAII illustrates the detailed level of information that can be collected for small-molecule interactions. Characterizing the kinetics of an interaction provides more information about a compound’s activity than equilibrium or inhibition studies alone. From a practical standpoint, Biacore S51 is capable of detecting the interaction of exceptionally small ligands (0100 Da), characterizing strong and weak interactions, and interpreting slow and fast kinetic rate constants. High reproducibility makes it possible to interpret very low signals, which could be particularly useful when working with enzyme preparations that are not fully active. Together these advantages make the new sensor technology an ideal tool to support secondary screening in drug discovery. References [1] Y.S.N. Day, C.L. Baird, R.L. Rich, D.G. Myszka, Direct comparison of binding equilibrium, thermodynamic, rate constants determined by surface- and solution-based biophysical methods, Protein Sci. 11 (2002) 1017–1025.
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323 [2] M.J. Cannon, G.A. Papalia, I. Navratilova, R.J. Fisher, L.R. Roberts, K.M. Worthy, A.G. Stephen, G.R. Marchesini, E.J. Collins, D. Casper, H. Qiu, D. Satpaev, S.F. Liparoto, D.A. Rice, I.I. Gorshkova, R.J. Darling, D.B. Bennett, M. Sekar, E. Hommema, A.M. Liang, E.S. Day, J. Inman, S.M. Karlicek, S.J. Ullrich, D. Hodges, T. Chu, E. Sullivan, J. Simpson, A. RaWque, B. Luginbühl, D.G. Myszka, Comparative analyses of a small molecule/enzyme interaction by multiple users of Biacore technology, Anal. Biochem. Submitted.
323
[3] B. Johnsson, S. Löfås, G. Lindquist, Immobilization of proteins to a carboxymethyldextran-modiWed gold surface for biospeciWc interaction analysis in surface plasmon resonance sensors, Anal. Biochem. 198 (1991) 268–277. [4] D.G. Myszka, Improving biosensor analysis, J. Mol. Recognit. 12 (1999) 279–284. [5] D.G. Myszka, T.A. Morton, CLAMP: a biosensor kinetic data analysis program, Trends Biochem. Sci. 23 (1998) 149–150.
Analysis of small-molecule interactions using Biacore S51 technology David G. Myszka¤ Center for Biomolecular Interaction Analysis, University of Utah, 50 N. Medical Drive, School of Medicine, Rm 4A417, Salt Lake City, UT 84132, USA Received 29 December 2003 Available online 6 May 2004
Abstract Biacore S51 is a new surface plasmon resonance-based biosensor developed by Biacore AB (Uppsala, Sweden). The instrument was engineered speciWcally to support small-molecule drug discovery and development. The platform includes increased sensitivity, larger sample handling capabilities, and automated data processing to improve throughput. Compared to previously released Biacore instruments, the most signiWcant design change relates to the introduction of the hydrodynamic-addressing Xow cell. This design allows two reaction surfaces and a reference surface to be placed within the same Xow cell, thereby improving data quality and extending the kinetic range of the instrument. Using a set of small-molecule inhibitors of the enzyme carbonic anhydrase II, we tested the reproducibility, sensitivity, and dynamic range of the biosensor. Given the S51’s performance capabilities, it should play an active role in secondary screening by providing high-resolution information for small-molecule/target interactions. 2004 Elsevier Inc. All rights reserved.
Optical biosensors are capable of monitoring the reversible associations of biological molecules in real time. Biacore AB (Uppsala, Sweden) released the Wrst commercial biosensor for biomolecular interaction analysis in 1990. Since that time, the company has developed a Xeet of instruments based on surface plasmon resonance (SPR)1 detection. SPR is sensitive to the refractive index of the buVer, which changes as molecules in solution interact with the target immobilized on the surface. These instruments allow the thermodynamic and kinetic binding properties of biological molecules to be routinely determined using microgram quantities of material and without labeling. A signiWcant advancement in SPR biosensor hardware was made with the release of Biacore S51 from Biacore AB. This platform, which represents an entirely new line of Biacore instruments, displays increased sensitivity, higher throughput, greater automation, and ¤
Fax: 1-801-585-3015. E-mail address: [email protected]. 1 Abbreviations used: SPR, surface plasmon resonance; CAII, carbonic anhydrase II; EDC, N-ethyl-N0-(3-dimethylaminopropyl)carbodiimide; NHS, N-hydroxysuccinimide; PBS, phosphate-buVered saline; Rmax, maximum response; RU, response unit. 0003-2697/$ - see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ab.2004.03.028
enhanced ease of use over previously available platforms. The Xow cell system in Biacore S51 has been completely redesigned and represents the biggest change compared to earlier platforms. In-line referencing is achieved through a proprietary hydrodynamic addressing system, which enables immobilization of two ligands and inclusion of a reference spot within the same Xow cell (Fig. 1). A Y-inlet uses a buVer stream to limit sensor surface coupling chemistry to the two side regions of the Xow cell, while the middle of the sensor surface remains isolated from coupling conditions. In Biacore S51, three detection spots are simultaneously addressed in each Xow cell, thereby eliminating lag time between reference and sample, which improves data quality and permits the resolution of fast kinetics. In addition, simultaneous sample delivery ensures constant analyte concentrations and reduces systematic noise due to temperature and pumping artifacts. Together, these improvements in instrument hardware increase the sensitivity of Biacore S51 relative to Biacore 2000 and 3000 platforms. To evaluate the performance of Biacore S51 and highlight how it may be applied in a drug discovery setting, we analyzed the binding of a number of smallmolecule inhibitors interacting with the enzyme carbonic
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
317
groups were blocked with a 7-min injection of 1.0 M ethanolamine, pH 8.5. DiVerent levels of CAII were immobilized for diVerent experiments as described under Results. Fig. 1. Schematic of Biacore S51 Xow cell. Two separate inlets make it possible to immobilize ligands on diVerent sides in the Xow cell, leaving the center detection spot unmodiWed as a control surface. In this illustration, the ligand for immobilization is being introduced through inlet 1 at a slower Xow rate than the running buVer in inlet 2.
anhydrase II. We have used this enzyme system in the past to compare the binding thermodynamics and kinetics obtained from the biosensor with solution-based methods [1]. And recently, it was used to compare the quality of data obtained from 40 diVerent investigators to gauge experimental variability associated with biosensor analysis [2]. The compounds used in the present study varied in molecular mass from 340 to 95 Da and displayed a »10,000-fold diVerence in aYnity for the enzyme. Binding response data were highly reproducible and changes in response as low as 0.50 RU (0.5 pg/mm2) could be reliably determined. The binding constants for the compound series were compared with regard to aYnity and kinetic components. These detailed mechanistic data with regard to compound binding provide new information to optimize a compound’s activity during drug development.
Materials and methods Materials Biacore 2000, Biacore 3000, Biacore S51, CM5 series sensor chips, and coupling reagents (N-ethyl-N0-(3dimethylaminopropyl)carbodiimide (EDC), N-hydroxysuccinimide (NHS), and ethanolamine-HCl) were purchased from Biacore AB. Carbonic anhydrase isozyme II (CAII) from bovine erythrocytes, sulfonamide inhibitors, and all other general-use reagents were purchased from Sigma Chemical Co. (St. Louis, MO). CAII immobilization CAII was immobilized onto CM5 sensor chips using standard amine coupling [3]. Phosphate-buVered saline (PBS), which consisted of 20 mM Na2HPO4 and 150 mM NaCl, pH 7.4, was used as a running buVer. The carboxymethyl dextran surface within one side of the Xow cell was activated with a 7-min injection of a 1:1 ratio of 0.4 M EDC and 0.1 M NHS. The enzyme was coupled to the surface with a 7-min injection of CAII diluted in 10 mM sodium acetate, pH 5.0. Remaining activated
Sulfonamide binding experiments Biosensor experiments were carried out on the following sulfonamide-based inhibitors: acetazolamide, benzene-disulfonamide, benzene-sulfonamide, carboxybenzenesulfonamide, dansylamide, sulfanilamide, sulpiride, and methylsulfonamide. Each compound was dissolved directly in the PBS running buVer and analyzed using a two- or threefold dilution series. The highest compound concentration varied, but all compounds were tested at 10 diVerent concentrations and each concentration was tested at least Wve times. The ultra-highresolution analysis of carboxybenzenesulfonamide involved the testing of 96 diVerent concentrations using a 4% dilution series and each concentration was tested four times. Within a given compound concentration series, the samples were randomized to minimize systematic errors. All of the bound complexes dissociated back to baseline within a reasonable time frame; therefore, no regeneration was required. Data processing and Wtting All sensorgrams were processed by Wrst subtracting the binding response recorded from the control surface (center reference spot), followed by subtracting an average of the buVer blank injections from the reaction spot [4]. To determine kinetic rate constants, all data sets were Wt to a simple 1:1 interaction model including a term for mass transport using numerical integration and nonlinear curve Wtting [5]. Equilibrium analysis was performed for methylsulfonamide by Wtting the response at the end of the association phase to a single-site binding isotherm.
Results Performance comparison To compare the performance of Biacore S51 with Biacore 2000 and 3000 instruments, identical binding studies using acetazolamide injected over CAII-immobilized surfaces (»3000 RU) were carried out. Acetazolamide was injected at seven diVerent concentrations and each concentration was repeated three times. The data in Fig. 2 shows the binding responses obtained for the small molecule analyte as measured on the diVerent platforms. While similar overall intensities were obtained from each instrument, the quality of the data as judged by the smoothness of the binding response and the lower
318
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 2. Comparison of binding response data from diVerent Biacore platforms. Acetazolamide at concentrations of 1000, 333, 111, 37, 12.3, 4.1, and 1.37 nM was injected over similar densities (»3000 RU) of CAII immobilized within (A) Biacore 2000, (B) Biacore 3000, and (C) Biacore S51. Each concentration was tested in triplicate. The response data were processes using data from a reference surface and buVer injections. To make an easier comparison of the binding responses the data from all instruments are shown at a rate of 1 Hz.
overall random noise is signiWcantly better on Biacore S51 compared to the 2000/3000 series of instruments. Also, having the reference and reaction surface within the same Xow cell in Biacore S51 minimizes the spikes often observed at the beginning and end of sample injections. Small-molecule binding responses To demonstrate the ability of Biacore S51 technology to measure the kinetic interactions of small molecules with immobilized macromolecules, we carried out an analysis of a number of sulfonamide compounds binding to CAII. Fig. 3 shows the sensorgram data collected for eight of the inhibitors interacting with a 5000-RU enzyme surface. The molecular masses of the compounds varied from the highest mass of 341 Da for sulpiride to the lowest of 95 Da for methylsulfonamide. All compounds were assayed over a 1000-fold concentration range using a 2-fold dilution series. Each concentration was also injected Wve times to test the reproducibility of the assay. All of the binding responses were exceptionally reproducible, since there was very little deviation between the repeated analyte concentrations. This indicates that the immobilized CAII did not lose activity over the time course of the assay. Data quality In general, the quality of the binding data itself is very high. The signals for most of the compounds were »100 times greater than the short-term random noise. As expected, even at saturating concentrations, the lowest signals were observed for methylsulfonamide (Fig. 3H), which is consistent with it having the lowest molecular mass (95 Da). However, even these low responses (1– 4 RU) were reproducible and could be analyzed to extract out binding constants.
Kinetic and equilibrium analysis The red lines in Fig. 3 represent a global Wt of the response data to a 1:1 interaction model that included a term for mass transport. All of the data sets for the various inhibitors were well described by this simple model. This illustrates that it is possible to Wt a number of interactions recorded on the biosensor with a simple model. The kinetic rate constants and aYnities determined for each compound are presented alongside the compound structure in Fig. 3. Because the reaction and reference surfaces are within the same Xow cell in Biacore S51 (as opposed be being aligned in a series on Biacore 2000 and 3000 instruments), it is possible to resolve the dissociation kinetics for very rapidly dissociating systems. For example, it is possible to determine a dissociation rate of 0.64 s¡1 for sulpiride (see Fig. 3G). The dissociation phase data for methylsulfonamide (see Fig. 3H) are perhaps just beyond the limit of determination at 2.2 s¡1. In this case, it would be diYcult to determine whether the actual dissociation rate was faster than what we observed. The limit for the dissociation rate constant in Biacore S51 is around 1 s¡1, which is about 10-fold higher than that on the Biacore 2000 and 3000 instruments. For this reason we performed an equilibrium analysis of the methylsulfonamide data. The inset in Fig. 3H illustrates that the equilibrium response data Wt well to a simple 1:1 binding isotherm. Comparison of binding constants A visual comparison of the kinetic binding constants for the sulfonamide inhibitors can be made by plotting the association versus the dissociation rates for each compound (Fig. 4). In this plot, data points to the right on the x axis represent faster association rates while data points further up on the y axis represent slower
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
319
Fig. 3. Sensorgrams for sulfonamide-based inhibitors interacting with a 5000-RU CAII surface. The highest concentration for each compound was as follows: (A) acetazolamide, 1 M; (B) furosemide, 30 M; (C) dansylamide, 6 M; (D) benzenesulfonamide, 100 M; (E) carboxybenzenesulfonamide, 40 M; (F) sulfanilamide, 100 M; (G) sulpiride, 2 mM; (H) methylsulfonamide, 2.5 mM. Each compound was injected over a 1000-fold concentration range using a 2-fold dilution series. Each concentration was injected Wve times and the entire data set was globally Wt to a 1:1 interaction model as shown by red lines. The compound structure, name, molecular mass, and binding constants are provided on each data set. The inset in H represents an equilibrium analysis of the methylsulfonamide data set. The number in parentheses represents the standard error in the last signiWcant digit.
dissociation rates. The diagonal lines represent equilibrium isotherms from 1 mM to 10 nM. Compounds with the highest aYnity appear in the upper right-hand quadrant of the plot and the lowest aYnity will be in the
lower left-hand corner. Comparing data for the nine sulfonamide inhibitors, there is an »10,000-fold diVerence in aYnity between the weakest binder, methylsulfonamide (KD D 270 M), and the tightest binder, acetalzolamide
320
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 4. Kinetic proWle plot of nine sulfonamide-based inhibitors of CAII. Rate constants were determined from Biacore S51. The diagonal lines represent equilibrium binding isotherms.
(KD D 19 nM). Three compounds, benzenesulfonamide, benzenedisulfonamide, and danyslamide, lay in a similar position on the plot, illustrating that they have similar association and dissociation kinetics for CAII. The aYnities for these compounds (»850 nM) are also similar to the aYnity determined for carboxybenzenesulfonamide (893 nM), since they all sit near the 1 M binding isotherm. Therefore, if looking at only the equilibrium dissociation constant, one would conclude that the addition of a carboxyl group to benzenesulfonamide does not eVect the interaction because the aYnities are the same. However, carboxybenzenesulfonamide binds with diVerent kinetics. For example, compared to benzenesulfonamide, the association and dissociation rates for carboxybenzenesulfonamide are each about four times slower. The rate constants changed in equal magnitude so that the overall aYnity remains around 1 M. The aYnity is 6-fold weaker when the carboxyl group on carboxybenzenesulfonamide is converted to an amine in sulfanilamide (Fig. 3F). The association rate for sulfanilamide decreases 2-fold while the dissociation rate increases 3-fold compared to those of carboxybenzenesulfonamide. This detailed kinetic information on the binding interaction is useful when analyzing structure/ function relationships of target/inhibitor interactions, providing more information than equilibrium assays alone. Ultra-low-capacity surfaces An advantage of the Biacore S51 technology is that it is capable of measuring reliable responses even for very low signals. To illustrate this capability, Wrst we analyzed the interaction of a low-molecular-mass analyte, methylsulfonamide (95 Da), as shown in Fig. 3H. To further test the limits of analyte detection, we collected binding data for carboxybenzenesulfonamide over an ultra-low-
capacity surface of CAII. In this case we purposefully immobilized only 100 RU of the enzyme with the intention of measuring a maximum binding response of 01 RU. The sensorgrams in Fig. 4A show the binding responses for a series of carboxybenzenesulfonamide over the ultra-low-capacity CAII surface. The maximum signal appears to be around 0.75 RU for the highest concentrations of the analyte. These data were Wt to a 1:1 interaction model to determine binding constants for the interaction as shown in Fig. 5A. The rate constants determined from the ultra-low-density CAII surface (Fig. 5A) compare very well with the values determined from the higher-density surface (Fig. 3E). However, with all the analyte concentrations overlaid in Fig. 5A it becomes diYcult to discern the concentration-dependent binding responses because of the random noise. To more readily resolve the diVerent binding proWles, the 10 diVerent analyte concentrations are plotted individually in Fig. 5B. From a visual inspection of these data it is possible to observe the expected concentration-dependent binding responses and to assess the quality of the Wt to the reaction model. These results illustrate that it is possible to accurately interpret binding responses that have a maximum signal of less than 1 RU. In this case, the calculated Rmax for the surface was 0.6 RU. High-resolution analysis An added feature of Biacore S51 is the ability to analyze data from a 384-well plate. To test the overall reliability and robustness of the system, we analyzed 96 diVerent concentrations of carboxybenzenesulfonamide, each injected four times over a 4000-RU CAII surface. The binding response from each analyte concentration was highly reproducible, since they overlapped with one another, as shown in Fig. 6A. The data for all 384 sensorgrams (a total of 619,200 data points) were globally Wt to a 1:1 interaction model. Fig. 6B shows a plot of the residuals from the entire data set, which are low and randomly distributed about zero. The standard deviation in the residuals was 0.25 RU, which is similar to the standard deviation of the replicate experiments of 0.24 RU, indicating that the simple model does an excellent job of describing the data. Again, the rate constants determined from this high-resolution analysis (Fig. 6A) are consistent with the values determined from our previous studies in Figs. 3E and 5A. The results of this ultra-high-resolution analysis illustrate that the instrument is capable of collecting exceptionally reproducible data over 384 assays.
Discussion As experimentalists interested in characterizing molecular interactions, we are fortunate to have access to commercial biosensor technology. While the
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
321
Fig. 5. Binding responses for carboxybenzenesulfonamide over an ultra-low-density CAII surface (100 RU). (A) Overlay of binding data for the inhibitor from 40 M to 19.5 nM using a twofold dilution series. (B) Responses for the individual inhibitor concentrations. Red lines represent a global Wt of the binding data to a 1:1 interaction model.
322
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323
Fig. 6. Ultra-high-resolution analysis of carboxybenzenesulfonamide/CAII interaction. (A) Inhibitor was injected using a starting concentration of 100 M in a 4% dilution series, giving a total of 98 diVerent concentrations. Each concentration was repeated four times for a total of 384 samples tested over the same surface. Red lines represent a global Wt for the data to a 1:1 interaction model. (B) Residual plot based on the Wt of the primary data.
biosensor platforms released in early 1990s ushered in a new way of characterizing macromolecular interactions, evolving technology such as Biacore’s S51 platform is making it possible to apply the advantages of label-free and real-time analysis to small-molecule systems in a routine way. All of the major components of the S51 have been reengineered relative to the Biacore 2000 and 3000 technology. The S51 detection unit has been redesigned with state-of-the-art optics, light source, and detector. Automated and integrated loading of sensor chips within the optical detection area permits a closed detection environment for improved thermal stability. The instrument can operate from 4 to 45 °C and the sample racks are protected from light. Higher throughput is made possible by incorporating 384-well microtiter plate compatibility and fast sample loading via a new injection technique. The injection needle is Wxed directly to the Xow cell and the sample racks move to address each sample. The result is decreased sample carry-over, faster sample loading, and the potential for future parallel sample processing possibilities.
Our analysis of a set of sulfonamide inhibitors binding to CAII illustrates the detailed level of information that can be collected for small-molecule interactions. Characterizing the kinetics of an interaction provides more information about a compound’s activity than equilibrium or inhibition studies alone. From a practical standpoint, Biacore S51 is capable of detecting the interaction of exceptionally small ligands (0100 Da), characterizing strong and weak interactions, and interpreting slow and fast kinetic rate constants. High reproducibility makes it possible to interpret very low signals, which could be particularly useful when working with enzyme preparations that are not fully active. Together these advantages make the new sensor technology an ideal tool to support secondary screening in drug discovery. References [1] Y.S.N. Day, C.L. Baird, R.L. Rich, D.G. Myszka, Direct comparison of binding equilibrium, thermodynamic, rate constants determined by surface- and solution-based biophysical methods, Protein Sci. 11 (2002) 1017–1025.
D.G. Myszka / Analytical Biochemistry 329 (2004) 316–323 [2] M.J. Cannon, G.A. Papalia, I. Navratilova, R.J. Fisher, L.R. Roberts, K.M. Worthy, A.G. Stephen, G.R. Marchesini, E.J. Collins, D. Casper, H. Qiu, D. Satpaev, S.F. Liparoto, D.A. Rice, I.I. Gorshkova, R.J. Darling, D.B. Bennett, M. Sekar, E. Hommema, A.M. Liang, E.S. Day, J. Inman, S.M. Karlicek, S.J. Ullrich, D. Hodges, T. Chu, E. Sullivan, J. Simpson, A. RaWque, B. Luginbühl, D.G. Myszka, Comparative analyses of a small molecule/enzyme interaction by multiple users of Biacore technology, Anal. Biochem. Submitted.
323
[3] B. Johnsson, S. Löfås, G. Lindquist, Immobilization of proteins to a carboxymethyldextran-modiWed gold surface for biospeciWc interaction analysis in surface plasmon resonance sensors, Anal. Biochem. 198 (1991) 268–277. [4] D.G. Myszka, Improving biosensor analysis, J. Mol. Recognit. 12 (1999) 279–284. [5] D.G. Myszka, T.A. Morton, CLAMP: a biosensor kinetic data analysis program, Trends Biochem. Sci. 23 (1998) 149–150.