Fluorescently-Labeled Calmodulin as the Biorecognition Element for

Lexington, Kentucky 40506-0055. Received September 3, 2001; Revised Manuscript Received May 17, 2002. A small-scale, homogeneous, rapid sensing ...
0 downloads 0 Views 151KB Size
1186

Bioconjugate Chem. 2002, 13, 1186−1192

ARTICLES Class-Selective Drug Detection: Fluorescently-Labeled Calmodulin as the Biorecognition Element for Phenothiazines and Tricyclic Antidepressants Phillip M. Douglass, Lyndon L. E. Salins, Emre Dikici, and Sylvia Daunert* Departments of Chemistry and Pharmaceutical Sciences, University of Kentucky, Lexington, Kentucky 40506-0055. Received September 3, 2001; Revised Manuscript Received May 17, 2002

A small-scale, homogeneous, rapid sensing system for phenothiazines and tricyclic antidepressants (TCAs) has been developed by employing fluorescently labeled mutant calmodulin (CaM) as the recognition element. A calmodulin mutant containing a unique cysteine residue at position 109 on the protein was expressed in Escherichia coli. Following purification, the environment-sensitive, thiolspecific fluorophores N-[2-(1-maleimidyl)ethyl]-7-(diethylamino)coumarin-3-carboxamide (MDCC), 6-acryloyl-2-dimethylaminonaphthalene (acrylodan), and 4-[N-(2-(iodoacetoxy)ethyl)-N-methylamino]7-nitrobenz-2-oxa-1,3-diazole (IANBD ester) were coupled to the C109 site of the mutant protein. The response of labeled CaM in the presence of calcium to increasing concentrations of chlorpromazine hydrochloride (CPZ), as well as other phenothiazines and structurally related antipsychotics and antidepressants, was investigated. Fluorescence measurements were performed on benchtop and microtiter plate fluorometers. The responses were characterized as a change in the signal intensity of the labeled protein upon ligand binding, and the stability of the system was monitored over a ninemonth period. The assay showed specificity for the phenothiazine and TCA classes of drugs, with limits of detection in the micromolar range. Selectivity studies indicated negligible response of the biosensing system to structurally unrelated compounds. This work represents a proof-of-concept assay for rapid, homogeneous detection of drugs employing binding proteins as the biorecognition element.

INTRODUCTION

The use of fluorescently labeled proteins has recently resulted in a number of biosensing systems for a variety of biomolecules (1-4). With proteins, one can capitalize on a highly evolved natural sensing system that provides selective and sensitive detection of target ligands and analytes. Such systems have not only been indicated for detection of drugs in a homogeneous-based format as reported herein, but also as a means of studying drug activity in cell-based screens (2, 5). In high-throughput screening (HTPS) applications, assays are not necessarily required to be highly specific. Screens that are designed to detect classes of drugs versus assays that detect single compounds are more practical and cost-effective in the analysis of combinatorial chemistry products (6). “Hits” identified from these preliminary screens need further analysis, and conventional analytical methods are then employed to characterize the nature and level of drug response. Ideal HTPS systems should be homogeneous, facile, and rapid. In addition, they should be amenable for miniaturization and small volume detection of a large number of compounds. Calmodulin is a 148 amino acid, 17-kDa calcium binding protein. Ubiquitous in nature, CaM plays an important role in intracellular signaling pathways (7, 8). * To whom correspondence should be addressed. Phone: (859) 257-7060. Fax: (859) 323-1069. E-mail: [email protected].

In the presence of calcium, the protein undergoes a conformational change resulting in the exposure of two hydrophobic pockets, one each in the N- and C-termini of the protein. These hydrophobic pockets have a high affinity for enzymes such as phosphodiesterase and myosin light chain kinase, which are activated upon binding to CaM (9). The phenothiazines, a family of antipsychotic drugs, bind to the C-terminal hydrophobic pocket of CaM, acting as antagonists of such enzymes (8, 10). In addition, CaM also has affinity for the structurally related tricyclic antidepressants (TCAs). Controversy exists in the literature as to whether the phenothiazines and TCAs also bind to the N-terminal hydrophobic pocket as one might expect, given that the two domains share a 44% residue homology (9). Most, however, tend to support the idea of the drugs only binding to the C-terminal with significant affinity (Kd’s in the micromolar range). Upon binding to the drug, CaM undergoes a hinge-motion conformational change about its central, 8-turn R-helix, resulting in a collapse of the protein as evident in Figure 1. The crystal structure of CaM bound to trifluoperazine (TFP) has been determined (see Figure 1, at right). Cook et al found that 13 residues form contacts with TFP. Ten of these are hydrophobic; the others are glutamates and interact via van der Waals forces. Only one (Glu11) is in the N-terminal domain of the protein (11). The tricyclic ring lies between the side chains of Met124 and Met144; these residues interact directly with the ligand’s aromatic rings (9, 11, 12).

10.1021/bc010080b CCC: $22.00 © 2002 American Chemical Society Published on Web 10/19/2002

Class-Selective Drug Detection

Figure 1. Ribbon diagram of calcium-bound calmodulin in the absence (left) and presence (right) of the phenothiazine trifluoperazine. Calcium is represented by yellow spheres, trifluoperazine is colored blue, and the site of fluorophore conjugation, Cys109, is colored red. Crystal structures were obtained from the Brookhaven Protein Data Bank (files 1CLL and 1CTR) and displayed using Molscript 1.4 (31) and Raster3D (32) software.

In this work, we report a proof-of-concept approach to drug detection utilizing the changes in conformation that binding proteins undergo upon binding to their ligand. Specifically, the study of fluorescently labeled CaM as a sensing element for the phenothiazine class of drugs and the related TCAs is described. The affinity of CaM for phenothiazines has previously been used in our laboratory as the basis of developing an affinity purification method for the protein and other recombinant proteins fused to CaM (13-15). Most recently, site-selective fluorescently labeled CaM has been employed in our group as a biosensing system for the detection of calcium (16). Given that CaM binds not only to calcium but also to phenothiazines, it is hypothesized that the change in conformation and environment that calcium-bound CaM undergoes upon binding to these drugs will allow the protein to act as a highly sensitive sensing element in their detection (16). It has been demonstrated that the location of the fluorophore on the protein highly influences the change in signal observed and thereby the sensitivity and detection limits of the sensing system (16). The CaM employed in our studies lacks any native cysteine residues (17). A mutant of this CaM has been rationally designed to contain a unique cysteine at a position on the protein that appeared to undergo significant environmental and conformational change upon drug binding (16). Utilizing the cysteine’s thiol group, mutant CaM was coupled to three different sulfhydrospecific environment-sensitive fluorophores, namely, MDCC, acrylodan, and IANBD ester. The resulting CaM conjugates were then evaluated as biosensing elements for the detection of phenothiazines and related TCAs. Steady-state fluorescence experiments were conducted to verify a difference in fluorescence signal emitted by the fluorescently labeled protein in the presence or absence of analyte. The class-selectivity of the biosensing system toward compounds both structurally related and unrelated to the phenothiazines and TCAs was also evaluated. EXPERIMENTAL PROCEDURES

Reagents. Luria-Bertani (LB) medium was purchased from GibcoBRL (Gaithersburg, MD). Ethylenebis(oxy-

Bioconjugate Chem., Vol. 13, No. 6, 2002 1187

ethylenenitrilo)tetraacetic acid (EGTA), dithiothreitol (DTT), chlorpromazine hydrochloride (CPZ), trifluoperazine dihydrochloride (TFP), thioridazine hydrochloride, amitryptiline hydrochloride, R-D-glucose, and acetylsalicylic acid were obtained from Aldrich (Milwaukee, WI). Ampicillin, chlorprothixene hydrochloride, and imipramine hydrochloride were obtained from Sigma (St. Louis, MO). Tris(hydroxymethyl)aminomethane (Tris) and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) were purchased from Research Organics (Cleveland, OH). The fluorophores N-[2-(1-maleimidyl)ethyl]7-(diethylamino)coumarin-3-carboxamide (MDCC), 6-acryloyl-2-dimethylaminonaphthalene (acrylodan), and 4-[N(2-(iodoacetoxy)ethyl)-N-methylamino]-7-nitrobenz-2-oxa1,3-diazole (IANBD ester) were obtained from Molecular Probes (Eugene, OR). Isopropyl β-D-thiogalactopyranoside (IPTG) was purchased from GibcoBRL (Gaithersburg, MD). Fluoronunc white C96 Maxisorp microtiter plates were purchased from VWR (South Plainfield, NJ). The expression plasmid pVUC-1 was kindly provided by Dr. D. M. Watterson (17). The phenothiazine 2-(trifluoromethyl)-10H-(3′-aminopropyl) phenothiazine hydrochloride (TAPP) was synthesized in our group as previously described (13, 15, 18). The bicinchoninic acid (BCA) protein microassay reagent kit from Pierce (Rockford, IL) was used to determine the concentration of the purified CaM. Deionized distilled water (Milli-Q water purification system, Millipore, Bedford, MA) was used to prepare all solutions and mobile phases. All chemicals employed were reagent grade or better. Apparatus. Bacterial colonies were grown on agar plates at 37°C in a Fisher Scientific Incubator (Fair Lawn, NJ). Cell cultures were grown in an Orbital Shaker from Forma Scientific (Marietta, OH) and pelleted using a Beckman J2-MI Centrifuge (Palo Alto, CA). Unpurified protein fractions were filtered with a 0.2 µm filter from Nalgene (Rochester, NY). A BIOCAD SPRINT Perfusion Chromatography System by PE Biosystems (Foster City, CA) was used for protein purification. Protein purity was verified by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) using a PhastSystem from Pharmacia Biotech (Uppsala, Sweden). Purified protein was concentrated via lyophilization using a VirTis Bench Top 3 Freeze-Dryer (Gardiner, NY). Protein was dialyzed against the correct buffer using 10000 MWCO Snakeskin pleated dialysis tubing by Pierce (Rockford, IL). Initial fluorescence studies were performed on a Fluorolog-2 fluorometer, Spex Industries (Edison, NJ) equipped with a 450-W xenon lamp and excitation and emission monochromators. Assay development was carried out on a Cytofluor 4000 fluorescence microtiter plate reader by PE Biosystems (Foster City, CA) equipped with the following filters: 420/50-nm excitation and 490/40-nm emission (for MDCC), 380/20nm excitation, 490/40-nm emission (for acrylodan), 485/ 20-nm excitation, 530/25-nm emission (for IANBD ester), 313/55-nm excitation, 460/40-nm emission (for chlorpromazine). Expression of Calmodulin. The plasmid pVUC-1 (17), containing the gene for mutant CaM with a methionine-cysteine mutation at residue 109 (16) was transformed into Escherichia coli JM109 competent cells and then plated on LB plates containing ampicillin (100 µg/mL). After growing overnight at 37 °C, a colony of transformed cells was picked and incubated overnight in 5 mL of LB media containing 75 µg/mL ampicillin. The culture was transferred into 500 mL of LB and incubated for an additional 7-8 h, until an optical density at 600 nm of 0.6 was obtained. Protein expression was induced

1188 Bioconjugate Chem., Vol. 13, No. 6, 2002

by the addition of 1 mM IPTG. Following 5 h of incubation with IPTG, the cells were harvested by centrifugation, washed in 10 mM HEPES buffer containing 0.5 mM CaCl2, pH 7.5 (Buffer A), and lysed via sonication. The crude extract was boiled for 5 min to precipitate out additional undesired proteins. Purification of CaM109. A TAPP affinity column was used to purify the mutant CaM following the method of Hentz et al. (13). Purified protein was concentrated overnight by lyophilization, and the purity of the protein was confirmed by SDS-PAGE employing silver staining development. The lyophilized protein was resuspended in 500 µL of deionized water and dialyzed in Buffer A for 24 h to remove any remaining EGTA present from the purification process. Protein concentration was determined by using the BCA protein assay. Preparation of Fluorescently-Labeled CaM109. A volume of 1 mL of purified CaM109 was dialyzed against Buffer A, containing DTT to eliminate any disulfide linkage formation in preparation for fluorophore conjugation. The protein was then dialyzed to remove any excess DTT. Immediately following DTT removal, a 5-fold molar excess of fluorophore (MDCC, IANBD ester, or acrylodan) was added in the dark. The coupling reactions were allowed to proceed for 4 h with stirring, at which time the samples were loaded onto a Sephadex G-25 sizeexclusion column to separate conjugated protein from excess fluorophore. The conjugates were eluted with Buffer A. A SDS-PAGE was run to verify which elution fractions contained the protein conjugate. The fractions containing the conjugate were pooled together, and the final protein concentration was calculated. The purified CaM109 conjugate was stored at 4 °C in the dark until use. Fluorescence Studies of CaM109-MDCC with CPZ. With the exception of initial optimization studies (where noted), all experiments were performed on the microtiter plate fluorometer with an assay volume of 300 µL, utilizing the CaM109-MDCC conjugate, and CPZ as the analyte. All results reported are the average of a minimum of three replicates, corrected for background, with the blank consisting of Buffer A. Detection limits reported were calculated with S/N ) 3. Chlorpromazine Fluorescence. Solutions of increasing concentrations of CPZ, 1 × 10-7 to 0.1 M, were prepared in Buffer A. Fluorescence measurements were taken on the microtiter plate fluorometer, using the filters previously noted. MDCC/CPZ Interaction. Aliquots of 25 µL of CPZ solutions ranging in concentration from 1 × 10-7 to 0.1 M were added to 1.5 mL samples of MDCC in Buffer A. Response of the biosensing system to MDCC was measured in the fluorescence microtiter plate reader employing the parameters described above. Dilution Curve. To determine the optimum concentration of labeled CaM109 to be used in subsequent experiments, the fluorescence emitted by increasing concentrations of each conjugate was measured on the benchtop fluorometer (CaM109-MDCC) or on the fluorescence microtiter plate reader (CaM109-MDCC, CaM109-acrylodan, CaM109-IANBD). Dilutions of labeled CaM109 were prepared ranging from 1 × 10-10 to 1 × 10-5 M. The protein concentration at which an increase in fluorescence signal was considerably larger than that of the background while still being relatively low was chosen as the concentration to be employed in all subsequent experiments.

Douglass et al.

Time Study of CaM109-MDCC/CPZ Interaction. An experiment was conducted on the benchtop fluorometer to observe the change in fluorescence response of the conjugated CaM109-MDCC to the binding of CPZ at various incubation times. A volume of 1.5 mL of CaM109MDCC was incubated with 25 µL of 1 × 10-2 M CPZ in Buffer A at RT on a 400 rpm shaker for various periods of time. Average fluorescence quenching values were plotted against the incubation times in minutes. The minimum incubation time allowing for maximal fluorescence response was chosen for all subsequent assays. Calibration Curves for CaM109-MDCC with Analyte. Aliquots of 25 µL of various concentrations of analytes (phenothiazines, TCAs, structurally related chlorprothixene, and other analytes) were added to a 1.5 mL volume of the labeled protein. The mixture was then incubated on a 400 rpm shaker at RT for 2 min and analyzed in the benchtop fluorometer (CPZ only) or fluorescence microtiter plate reader as previously described. Average fluorescence quenching was plotted against the log of the analyte concentration in the sample. Experiments were also conducted with the additional fluorophores, acrylodan, and IANBD ester following the same procedure. Stability Studies. The fluorescent response of the CaM109-MDCC conjugate to 1.6 × 10-5 M CPZ was measured on days 3, 14, 21, 99, 151, and 269 of postconjugation in the fluorescence microtiter plate reader following the procedure outlined above. RESULTS AND DISCUSSION

The detection of phenothiazines and TCAs by the fluorescently labeled mutant CaM described here is based on the calcium-dependent change in conformation of the protein. Calmodulin belongs to a family of binding proteins that contain helix-loop-helix motifs referred to as EF hands. Calmodulin has four such motifs; two are located in both the N- and C-terminal regions of the protein (8). Upon binding calcium, CaM undergoes a conformational change about the central R-helix, exposing a hydrophobic region in each domain (11, 12). By capitalizing on this change, we have reported herein the development of a class-selective means of drug detection. In addition, employing a biological recognition element, namely the binding protein CaM, takes advantage of nature’s highly evolved means of detection and demonstrates the potential application of these binding proteins to current needs of the chemical, pharmaceutical, and biotechnology industries. In particular, our approach provides an alternative solution in the demand for rapid, reliable, and reproducible screening methods in drug discovery programs. The synthetic CaM employed in our studies (17) resembles higher plant CaM with the exception of threonine-26, glycine-96, valine-136, and glutamine-143, which are common to vertebrate CaM. As such, it does not contain any cysteine residues in its structure (17). Sitedirected mutagenesis on this gene has been performed previously in our group (16) to study the structureactivity of certain specific mutations on the protein and to develop a biosensing system for the detection of calcium. On the basis of crystallographic analyses of calcium-bound CaM in both the presence and absence of the phenothiazine trifluoperazine (Figure 1) (11, 12), CaM containing a mutation at residue 109 (Met109Cys; CaM109) was chosen as the biosensing element to be employed in the development of the assay for the phenothiazines and TCAs. Located on the surface of the hydrophobic pocket of CaM’s C-terminal domain, residue

Class-Selective Drug Detection

Bioconjugate Chem., Vol. 13, No. 6, 2002 1189

Figure 2. General structure of the substances employed in this study.

109’s location should allow for efficient conjugation of the fluorophore to the protein (16). Studies of these crystal structures have shown that the microenvironment surrounding residue 109 undergoes changes upon the addition of drug, as residue 109 is one of 10 involved in hydrophobic interactions with the phenothiazine class of drugs (11, 12). From the results reported in these studies in addition to previous mutational studies performed with fluorescently labeled CaM (16), it was hypothesized that a fluorophore attached at this position would be sensitive to drug binding. It is important to note, as one can observe from Figure 2, that the phenothiazines and related compounds are naturally fluorescent. Thus, it should be possible to detect them based on their intrinsic fluorescence alone. Having a λex of 340 nm and a λem of 460 nm, CPZ fluorescence could also potentially interfere with the emission wavelength of MDCC, the fluorophore employed in the majority of the studies reported herein. Therefore, detection

of a change in CPZ fluorescence as a result of increasing concentrations was evaluated. Figure 3 demonstrates that although CPZ fluorescence can be used in the direct detection of the drug in a sample solution, the detection limit, 6.3 × 10-4 M, may not be sensitive enough for the HTPS of this class of drugs. The detection of CPZ using CaM109 coupled to three different fluorophores, MDCC, acrylodan, and IANBD ester, was then investigated. The unique cysteine of CaM109 was conjugated with these sulfhydro-specific fluorophores; the concentration of the fluorophore in the reaction mixture was five times that of the protein. Following conjugation, excess fluorophore was separated from the conjugated protein by using size-exclusion chromatography. Fractions of eluant corresponding to CaM109-fluorophore, as determined by SDS-PAGE, were pooled together and employed as the sensing element used in all studies. Initial assay development was performed on a standard benchtop fluorometer, with

1190 Bioconjugate Chem., Vol. 13, No. 6, 2002

Figure 3. Chlorpromazine fluorescence in the absence of CaM109-MDCC. Error bars not visible are concealed by the points on the graph.

Figure 4. Effect of incubation time between CaM109-MDCC and 1 × 10-4 M CPZ. Error bars not visible are concealed by the points on the graph.

assay volumes of 1.5 mL. To determine the CaM109 concentration that would be used in subsequent experiments, dilution curves were constructed. In these experiments, fluorescence measurements were made employing increasing concentrations of the conjugated protein. For CaM109-MDCC, a concentration of 5 × 10-8 M was chosen. This concentration provided a large enough signal for development of the assay while not consuming a significant amount of reagent. To determine how long the biosensing element needed to be incubated with the analyte, a time study was performed. From this study, it was determined that an incubation time of 2 min resulted in the generation of maximal sensitivity of the reagent (CaM) to analyte (CPZ) (Figure 4). Such rapid reaction kinetics can provide an advantage for this type of sensing scheme; rapid detection is one of the desired properties of assays employed in the HTPS of drugs and allows for a greater number of compounds to be screened per time. CaM109 coupled to MDCC has previously been reported by our group as a sensitive biorecognition element

Douglass et al.

Figure 5. Calibration curve for CPZ employing CaM-MDCC and performed in a benchtop fluorometer (b), with a total assay volume of 1.5 mL, and on a microtiter plate fluorometer ([), with a total assay volume of 300 µL.

for calcium detection (16, 19). As such, we decided to initially focus on this conjugate when designing a sensing scheme for phenothiazines due to their calcium-dependent binding to CaM. MDCC is composed of a reactive maleimide connected to a fluorescent coumarin moiety by a flexible aliphatic spacer arm; MDCC is excited at 425 nm and emits fluorescence at 474 nm. Upon coupling CaM109 with MDCC, a blue shift to 460 nm occurs in the fluorescence emission spectra of the fluorophore (data not shown). We postulate that MDCC is residing in the hydrophobic pocket in the C-terminal domain of the protein in the absence of CPZ. Residues 92, 100, 105, 109, 124, 125, 127, 128, 136, 141, and 144 have been shown to compose this hydrophobic pocket (11). As seen in Figure 5, a decrease in the fluorescence quantum yield (quenching) is observed upon binding of the phenothiazine chlorpromazine (CPZ) to the CaM109-MDCC, with a detection limit of 5 × 10-7 M (on the benchtop fluorometer). This quenching could be explained by the change in environment that MDCC experiences upon binding of CPZ to this hydrophobic region. To allow for binding, the flexible spacer arm and the coumarin of the MDCC may be displaced from this region, exposing the fluorophore to a more polar environment. A subsequent red shift in the fluorescence spectra observed upon the addition of increasing concentrations of CPZ supports this idea (data not shown). Red shifts are often, but not always, associated with a decrease in quantum yield (20). Also, it is possible that the diethylamino group of the fluorophore is less constrained in the presence of CPZ. Previous work from other groups has shown that coumarins having the lone pair of the amino substituent constrained and not able to interact optimally with the π electrons of the aromatic ring are highly fluorescent (21). Studies performed by us and others on the mechanism of response of fluorescently labeled proteins support this theory (2123). Although the CPZ native fluorescence spectrum (λex and λem of 340 and 460 nm, respectively) interferes with that of CaM109-MDCC, it does not significantly hinder the sensitivity of our biosensing system. In addition, a steady-state study conducted with MDCC and CPZ indicated that in the absence of CaM the fluorescence signal generated by MDCC is not significantly quenched by the analyte. This observation indicates that the

Class-Selective Drug Detection

quenching observed in the CaM-based assays is a result of the changes occurring in the dynamics of the proteinanalyte interactions, and not simply due to an interaction between the analyte and the fluorophore itself. An alternative explanation, related to the fact that CaM is being employed as a molecular handle, would be that the fluorophore and analyte are brought into close proximity causing a change in fluorescence as a result of the interaction between the two. To elucidate the exact mechanism by which changes in fluorescently labeled CaM conformation affects the fluorescence emission of the fluorophore, studies involving time-resolved anisotropy measurements need to be performed. It is important to note that the detection limit of 1.6 × 10-6 M observed for CPZ when measured on the fluorescence microtiter plate reader was higher than that observed in the larger volume format. A shift in the detection limit to higher concentrations when measuring the CPZ levels on the fluorescence microtiter plate reader can in part be explained to a lower collection efficiency of the instrument, and the total amount of reagent present in the 300 µL samples vs the 1 mL samples. Since this assay has been scaled down 10-fold in volume, for optimal sensitivity the concentration of reagent may need to be increased to compensate for this reduction in sample volume. It should be noted that concentrations in primary high-throughput screens can be as high as 10-5 M, allowing for the identification of compounds with moderate interaction. There is a need, however, for small volume assays, especially when the supply of new compounds from combinatorial libraries may be limited (2427). Reduced consumption as a result of small volume assays allows for synthetic or biological samples to be subjected to a greater number of analyses at these higher concentrations (28). The other two fluorophores, acrylodan and IANBD ester, were similarly evaluated. These CaM109 conjugates were also able to sense increasing levels of CPZ, although with worse detection limits. The acrylodan- and IANBD-labeled CaM each had a shift in detection limits 1 order of magnitude higher (1.6 × 10-5 M), and a reduced maximal fluorescence quenching of 80% at 1 × 10-3 M of CPZ as opposed to 100% for CaM109-MDCC. Nonetheless, CPZ detection using these additional CaM conjugates demonstrates the adaptability of the CaMbased sensing system to fluorophores of differing fluorescence emission profiles. As previously mentioned, this assay was intended to detect the structurally related TCAs as well, which also are used in the treatment of psychoses and depression. These compounds are also CaM inhibitors (29); thus, it was expected the system would elicit a similar response to these drugs. Table 1 includes the response of CaM109MDCC to the TCAs imipramine and amitriptyline and the structurally related chlorprothixene and compares their response to representative phenothiazines. All phenothiazines and related tricyclic drugs tested had detection limits of 7.4 × 10-6 M or lower, with the phenothiazine trifluoperazine having the lowest detection limit of 1.6 × 10-7 M. Also shown in Table 1 is the response to nine unrelated compounds: acetylsalicylic acid, R-D-glucose, and several nontricyclic antidepressants. Acetylsalicylic acid, dopamine, serotonin, and warfarin were selected as models of structurally unrelated drugs that contain an aromatic ring, and serotonin and dopamine act on the central nervous system. Amantadine and glucose were chosen due to their polarity as compared to TCAs. Tianeptine, amoxapine, and fluoxetine were tested, because they are structurally similar

Bioconjugate Chem., Vol. 13, No. 6, 2002 1191 Table 1. Comparison of CaM109-MDCC Response to Phenothiazines, Related Tricyclics, and Other Structurally Unrelated Compoundsc compound

max. fluorescence quenching (%)

detection limit (M)

chlorpromazinea trifluoperazinea thioridazinea chlorprothixeneb amitriptylineb imipramineb amoxapineb tianeptineb fluoxetinec acetylsalicylic acidc glucosec warfarinc dopaminec amantadinec serotoninc

97.2 99.8 100.0 100.0 91.3 100.0 0.5 13.7 30.1 8.7 8.3 8.3 13.7 1.4 6.9

1.6 × 10-6 1.6 × 10-7 1.6 × 10-6 7.4 × 10-6 1.5 × 10-6 7.4 × 10-6 5.9 × 10-4 -

a Note: all studies were performed on fluorescence microtiter plate reader. b Phenothiazines. c Related tricyclics. d Structurally unrelated compounds.

to tricyclic antidepressants but have different activity and mechanism of action. The results showed minimal fluorescence change for all of these compounds, except for fluoxetine. Tianeptine and amoxapine, selective serotonin reuptake inhibitors (SSRI), are structurally similar to tricyclic antidepressants. Their mechanism of action is different, and they do not inhibit calmodulin (30). We postulate that tianeptine’s long hydrophilic chain and amoxapine’s piperazine ring with truncated spacer, prevents them from binding to calmodulin. Fluoxetine, although not a tricyclic antidepressant, showed some response in the assay. This drug has two aromatic rings and contains a hydrophilic side chain with the proper length. We hypothesize that due to its flexibility, fluoxetine can fit into the hydrophobic pocket of calmodulin and the hydrophilic side chain can further stabilize the binding. However, to achieve 30% quenching, as with the TCAs, the concentration required for this was substantially greater. The detection limit for fluoxetine was found to be 5.9 × 10-4 M. This value is above the usual concentration range of drug detection required in highthroughput screening applications. Therefore, CaM provides class-selective detection, more specifically for those drugs belonging to the phenothiazines and TCA families of compounds. A stability study was conducted over a nine-month period to observe the change in fluorescence signal of CaM109-MDCC since initial conjugation (data not shown). The reagent was stored in solution at 4 °C in the dark. These studies were performed to evaluate the long-term storage of the CaM109-MDCC biosensing reagent. Typical calibration curves for CPZ were obtained each time as shown in Figure 5. Results indicated a shift in the detection limit toward higher concentrations; after nine months, the detection limit for CPZ was an order of magnitude higher (1 × 10-5 M) than that observed when fresh CaM109-MDCC was employed. In conclusion, it has been demonstrated that sitespecific fluorescently labeled CaM can be employed as a biosensing reagent in the preliminary screening of certain pharmaceuticals. The CaM-based assay showed to be class-selective toward drugs used in the treatment of psychoses and depression, which contain a phenothiazine/ phenothiazine-like structural moiety. The assay developed is reproducible over long periods of time and analyses volumes in the microliter range have shown adequate sensitivity for HTPS. The work presented

1192 Bioconjugate Chem., Vol. 13, No. 6, 2002

herein illustrates a proof-of-concept, model system employing a binding protein (CaM) as the detection element. Given that the assay is homogeneous in nature, this technology is envisioned to offer a valid alternative approach to rapid initial screening of potential new drug candidates in HTPS formats. In addition, not requiring the addition of reagents to generate a signal (i.e., reagentless) provides potential advantages over more conventional means of pharmaceutical analyses. Specifically, this assay offers minimal sample preparation, minimal analyte consumption, and rapid detection kinetics. It is believed that the strategy employed herein could be easily adapted to other binding proteins specific to other classes of ligand/analytes and/or drugs. ACKNOWLEDGMENT

The authors would like to thank the National Aeronautics and Space Administration for support of this work. P. M. Douglass acknowledges the National Science Foundation IGERT program and the American Foundation for Pharmaceutical Education for predoctoral fellowships. S. Daunert is a Cottrell Scholar and a Lilly Faculty Awardee. LITERATURE CITED (1) D’Auria, S., Gryczynski, Z., Gryczynski, I., Rossi, M., and Lakowicz, J. R. (2000) A protein biosensor for lactate. Anal. Biochem. 283, 83-88. (2) Giuliano, K. A. and Taylor, D. L. (1998) Fluorescent-Protein Biosensors: New Tools for Drug Discovery. Trends Biotechnol. 16, 135-140. (3) D′Auria, S., Herman, P., Rossi, M., and Lakowicz, J. R. (1999) The fluorescence emission of the apo-glucose oxidase from Aspergillus niger as probe to estimate glucose concentrations. Biochem. Biophys. Res. Commun. 263, 550-553. (4) Doi, N., and Yanagawa, H. (1999) Design of generic biosensors based on green fluorescent proteins with allosteric sites by directed evolution. FEBS Lett. 453, 305-307. (5) Giuliano, K. A. and Post, P. L. (1995) Fluorescent Protein Biosensors: Measurement of Molecular Dynamics in Living Cells. Annu. Rev. Biophys. Biomol. Struct. 24, 405-434. (6) Giuliano, K. A., DeBiasio, R. L., Dunlay, T., Gough, A., Volosky, J. M., Zock, J., Pavlakis, G. N., and Taylor, D. L. (1997) High-content screening: a new approack to easing key bottlenecks in the drug discovery process. J. Biomol. Screen. 2, 249-259. (7) Hidaka, H. and Hartshorne, D. J., Eds. (1985) Calmodulin Antagonists and Cellular Physiology, Academic Press, Inc., Orlando. (8) VanEldik, L. and Watterson, D. M., Eds. (1998) Calmodulin and Signal Transduction, Academic Press, Inc., New York. (9) Osawa, M., Swindells, M. B., Tanikawa, J., Tanaka, T., Mase, T., Furuya, T., and Ikura, M. (1998) Solution Structure of Calmodulin-W-7 Complex: The Basis of Diversity in Molecular Recognition. J. Mol. Biol. 276, 165-176. (10) Prozialeck, W. C. and Weiss, B. (1982) Inhibition of Calmodulin by Phenothiazines and Related Drugs: StructureActivity Relationships. J. Pharmacol. Exp. Ther. 222, 509516. (11) Cook, W. J., Walter, L. J., and Walter, M. R. (1994) Drug Binding by Calmodulin: Crystal Structure of a CalmodulinTrifluoperazine Complex. Biochemistry 33, 15259-15265. (12) Chattopadhyaya, R., Meador, W. E., Means, A. R., and Quiocho, F. A. (1992) Calmodulin structure refined at 1.7 A resolution. J. Mol. Biol. 228, 1177-1192. (13) Hentz, N. G., Vukasinovic, V., and Daunert, S. (1996) Affinity Chromatography of Recombinant Peptides/Proteins Based on a Calmodulin Fusion Tail. Anal. Chem. 68, 15501555.

Douglass et al. (14) Schauer-Vukasinovic, V. and Daunert, S. (1999) Purification of recombinant proteins based on the interaction between a phenothiazine-derivatized column and a calmodulin fusion tail. Biotechnol. Prog. 15, 513-516. (15) Patel, J., Daunert, S., Bachas, L. G., and Bhattacharyya, D. Unpublished data. (16) Schauer-Vukasinovic, V., Cullen, L., and Daunert, S. (1997) Rational Design of a Calcium Sensing System Based on Induced Conformational Changes of Calmodulin. J. Am. Chem. Soc. 119, 11102-11103. (17) Roberts, D. M., Crea, R., Malecha, M., Alvarado-Urbina, G., Chiarello, R. H., and Watterson, D. M. (1985) Chemical synthesis and expression of a calmodulin gene designed for site-specific mutagenesis. Biochemistry 24, 5090-5098. (18) Hart, R. C., Bates, M. D., Cormier, M. J., Rosen, G. M., and Conn, P. M. (1983) Synthesis and characterization of calmodulin antagonistic drugs. Methods Enzymol. 102, 195204. (19) Salins, L. L. E., Schauer-Vukasinovic, V., and Daunert, S. (1998) Optical Sensing Systems Based on Biomolecular Recognition of Recombinant Proteins. Proc. SPIE- Int. Soc. Opt. Eng. 3115, 16-24. (20) Lakowicz, J. R. (1983) Principles of Fluorescence Spectroscopy, Plenum Press, New York. (21) Hirshberg, M., Henrick, K., Haire, L. L., Vasisht, N., Brune, M., Corrie, J. E., and Webb, M. R. (1998) Crystal structure of phosphate binding protein labeled with a coumarin fluorophore, a probe for inorganic phosphate. Biochemistry 37, 10381-10385. (22) Lundgren, J. S., Salins, L. L., Kaneva, I., and Daunert, S. (1999) A dynamical investigation of acrylodan-labeled mutant phosphate binding protein. Anal. Chem. 71, 589-595. (23) Watkins, A. N. and Bright, F. V. (1998) Effects of Fluorescent Reporter Group Structure on the Dynamics Surrounding Cysteine-26 in Spinach Calmodulin: A Model Biorecognition Element. Appl. Spectrosc. 52, 1447-1456. (24) Wolcke, J. and Ullmann, D. (2001) Miniaturized HTS technologies - µHTS. Drug Discov. Today 6, 637-646. (25) Masimirembwa, C. M., Thompson, R., and Andersson, T. B. (2001) In vitro high throughput screening of compounds for favorable metabolic properties in drug discovery. Comb. Chem. High Throughput Screen. 4, 245-263. (26) Schullek, J. R., Butler, J. H., Ni, Z. J., Chen, D., and Yuan, Z. (1997) A high-density screening format for encoded combinatorial libraries: assay miniaturization and its application to enzymatic reactions. Anal. Biochem. 246, 20-29. (27) Kenny, B. A., Bushfield, M., Parry-Smith, D. J., Fogarty, S., and Treherne, J. M. (1998) The application of highthroughput screening to novel lead discovery, in Progress in Drug Research (E. Jucker, Ed.) pp 245-269, Birkhauser Verlag, Basel. (28) Ramsey, J. M., Jacobson, S. C., and Knapp, M. R. (1995) Microfabricated chemical measurement systems. Nat. Med. 1, 1093-1096. (29) Roufogalis, B. D., Minocherhomjee, A.-E.-V. M., and AlJobore, A. (1983) Pharmacological Antagonism of Calmodulin. Can. J. Biochem. Cell Biol. 61, 927-933. (30) Montgomery, S. A., 1995. Selective serotonin reuptake inhibitors in the acute treatment of depression, in Psychopharmacology, The Fourth Generation of Progress (F. E. Bloom, and D. J. Kupfer, Eds.) pp 451-461, Raven Press, New York. (31) Kraulis, P. J. (1991) MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946-949. (32) Merrit, E. A. and Murphy, M. E. P. (1994) Acta Crystallogr. D50, 869-873.

BC010080B