《天然药物化学》课程参考文献（天然药物研究与开发）Natural Products in Drug Discovery - Concepts and Approaches for Tracking Bioactivity.pdf_大学文库

Current Organic Chemistry, 2006, 10, 899-920 899 1385-2728/06 $50.00+.00 © 2006 Bentham Science Publishers Ltd. Natural Products in Drug Discovery - Concepts and Approaches for Tracking Bioactivity O. Potterat and M. Hamburger* Institut für Pharmazeutische Biologie, Universität Basel, Klingelbergstrasse 50, CH-4053 Basel, Switzerland Abstract: Effective methods for localization and characterization of bioactivity are a cornerstone for natural product based drug discovery efforts. Over the last decade, a wealth of new technologies and conceptual approaches for bioactivity screening have emerged. These developments are reviewed under the perspective of their applicability in the field of natural products discovery. The methods discussed here include bioautography, HPLC-based activity profiling, HPLC-based on-flow bioassays, assays based on capillary electrophoresis, molecular imprinted polymers, various MSand NMR-based methods, biosensors, and chip-based technologies for affinity separation and expression profiling. Selected examples illustrate the potential and limitations of the different approaches for contemporary natural products lead discovery. INTRODUCTION Natural products have traditionally played a major role in drug discovery and still constitute a prolific source of novel lead compounds or pharmacophores for medicinal chemistry. About 40% of current ethical drugs are directly or indirectly derived from compounds of biogenic origin [1, 2]. Molecules originating from natural products leads are well represented in the worldwide 35 top selling prescription drugs with percentages ranging approximately from 20 to 40 % over the last five years. Substances like taxol, cyclosporine and the statins are cornerstones of modern pharmacotherapy. Despite increasing competition from combinatorial and classical compound libraries, there has been a steady introduction of natural product-derived drugs in the last years. According to a recent review on the role of natural product research in drug discovery [3], a total of 15 natural products or natural product derived drugs have been launched in either the United States, Europe or Japan in the years 2000-2003. They include essential medicines such as the anti-malaria drug arteeether (ArtemotilÒ), galanthamine (ReminylÒ) for the treatment of Alzheimer’s disease, and pimecrolimus (ElidelÒ) for atopic dermatitis. The impact of natural products on the development pipelines of the pharmaceutical industry is unabated. As per December 2003, some 15 compounds were in phase III clinical trials or registration. Among these are representatives of novel compound classes such as the anticancer drug ixabepilone, a semisynthetic derivative of epothilone B produced by the myxobacterium Sorangium cellulosum, or ziconotide, a compound for the treatment of chronic pain which is the synthetic equivalent of w-conotoxin found in the dart venom of the cone shell Conus magus [3]. The Potential of Natural Products A striking feature of natural products accounting for their lasting importance in drug discovery is their unmet and still *Address correspondence to this author at the Institut für Pharmazeutische Biologie, Universität Basel, Klingelbergstrasse 50, CH-4053 Basel, Switzerland; Tel: 0041 61 267 14 25; Fax: 0041 61 267 14 74; E-mail: matthias.hamburger@unibas.ch largely untapped structural diversity. In the postgenomic era with its increasing number of druggable targets, chemical diversity of screening pools is a key factor in the tremendous competition between pharmaceutical industries. In this respect, natural products remain a rather indispensable complement to synthetic compound collections. Natural products are sterically more complex and differ from synthetic compounds with respect to the statistical distribution of functionalities [4]. They cover a much larger volume of the chemical space and display a broader dispersion of structural and physicochemical properties than compounds issued from combinatorial synthesis [5]. Even though some academic groups and companies nowadays synthesise increasingly complex structures to match the chemical space occupied by natural products, a strategy referred to by Schreiber as “diversity oriented synthesis” (DOS) [6], about 40% of the chemical scaffolds found in natural products are still absent in today’s medicinal chemistry [7]. Compared to synthetic molecules, a large proportion of natural products exhibit also more favourable ADME/T properties, despite the fact that they often do not satisfy “drug-likeness” criteria, such as Lipinski’s Rule of Five [8]. Besides their potential as lead structures, natural products also provide attractive scaffolds for combinatorial synthesis and remain essential tools for the validation of new drug targets [9]. Challenges Faced by Industrial Natural Product Research Despite the proven track record of natural products in drug discovery and their uncontested unique structural diversity, pharmaceutical companies have drastically scaled down or even terminated their activities in this field during the last decade [3]. Remaining operations have been quite often outsourced to biotech companies offering natural product related services such as libraries of pure compounds or semi-purified fractions. In industries where natural products remain in the screening program, they face increasing competition from other technologies such as combinatorial chemistry, drug modeling and virtual screening, for the allocation of drug discovery resources. The

900 Current Organic Chemistry, 2006, Vol. 10, No. 8 Potterat and Hamburger recent decline of interest observed in the pharmaceutical industry is, in part, due to the accelerated transformation of the drug discovery process during the last decade. Natural products have been facing major obstacles to fit into this new drug research environment. The advent of combinatorial chemistry and high throughput screening [10] in the 1990s enabled the testing of hundreds of thousands samples within a few weeks. This paradigm shift had major implications for natural product research. The classical and historically successful approach of screening crude or pre-purified extracts followed by several iterative steps of activity guided fractionation could not match with the short target cycle times in HTS, where testing capacities are only provided for a limited time window. A further complication is that natural product hits must go from mixture to pure compounds with enough time left for hit-to-lead assessment. Full structural information and accurate IC50 data are indispensable to compete with synthetic compounds in the lead selection phase. Hit clustering is gaining importance for establishing structure-activity relationship early in the lead selection process. Natural product hits are often observed as singletons, a fact that puts them at a disadvantage in comparison to compound families typically encountered in synthetic libraries. A further issue with extract screening is the comparatively high number of false positives which are due to the common presence of compounds which display unspecific activities or interfere with the assay format. Tannins, for example, form tight complexes with metal ions and with a wide array of proteins and polysaccharides. This leads to false-positive result in most assays involving a purified protein. Detergent-like compounds have a tendency to disrupt membranes and produce misleading results in cellbased assays. Examples include widely occuring plant metabolites such as saponins, and fatty acids and panosialins, which are common in streptomycetes. Compounds such as polyenes and polyethers often display general cytotoxicity and may produce false results in cell-based assays. Strong metal chelators are susceptible to react with assay components, e.g. when nickel beads are used as linkers. UV quenchers such as the chlorophyll breakdown product phaeophorbide A, and autofluorescent compounds are prone to interfere with the readout in assays based on light measurement [11]. Confronted to unrealistic hit rates in HTS and a slow and labour-intensive deconvolution process, many pharmaceutical companies have switched from extract screening to prefractionated extracts or pure compounds libraries [12]. Large collections of compounds and semipurified fractions have been generated using parallel fractionation and purification technology [12-15]. While these methods have the undeniable advantage of considerably reducing or even eliminating the timeconsuming follow-up process, they also have some intrinsic drawbacks. Pure compound libraries will never be a full substitute for the huge structural diversity found in extracts. Trace components which are, in principle, as promising as major constituents, are likely not found in such collections. The splitting of an extract into a large number of fractions, on the other hand, leads to a considerable increase of the number of samples to be screened. This can be an issue when working with expensive assay formats or costly targets, such as recombinant proteins. A well-balanced combination of pure compounds, fractions and extracts, and a differential use thereof depending on the target and the screening format, appears to be the most promising approach. The Discovery Process of Bioactive Natural Products In recent years, natural product research has become a technology-driven process. The impact of HPLC-coupled spectroscopy has been tremendous. The concerted use of HPLC-DAD, -MS and -NMR has opened entirely new possibilities for the characterization of secondary metabolites in biological extracts. These techniques provide a wealth of structural information on-line with minute amounts of sample [16, 17]. Even absolute configuration of a molecule can be established using HPLC-CD [18] or HPLC-NMR after Mosher’s ester derivatization [17]. With the more recent emergence of mass spectrometry-controlled preparative HPLC [19], compound purification has also become straightforward, provided the compounds exhibit sufficient chemical stability. Recent developments in NMR probe technology and higher magnetic fields [20], and miniaturization in X-ray crystallography make structural elucidation with sub-mg amounts becoming increasingly routine. While analysis, purification and structure elucidation of natural products have experienced a technological breakthrough over the last decade, tracking bioactivity in complex matrices remains a highly challenging task. Extracts are complex mixtures. There is a continuing need for faster and more reliable methods to identify compounds that interact with therapeutic targets, with minimal interference from the multitude of chemicals present in the matrix. The classical process leading from a bioactive extract to a pharmacologically active pure constituents has always been a long and tedious process requiring substantial material amount and financial resources [21]. It consists of several consecutive steps of preparative chromatographic separation, whereby each fraction has to be submitted to suitable bioassays to track the activity ultimately to a defined pure compound. While this procedure has led to the successful isolation of many bioactive molecules, its weaknesses cannot be overlooked. Besides being slow and costly, the separation performance is poor, at least in the initial fractionation steps which are typically by open column chromatography. The loss of bioactivity in the course of the purification process is not uncommon, and there is little means for early dereplication of known or otherwise uninteresting compounds. The approach described above obviously no more matches the timelines and the workflow of modern drug discovery. There is a compelling need for faster and more effective strategies, susceptible to be implemented in a high throughput environment. Probably the greatest challenge in this context is the judicious interfacing of chemistry and biology to correlate chemical analysis with biochemical data. The development of highly sensitive and miniaturized assays provides the technological basis for that purpose. Various innovative methodologies for the analysis of macromoleculeligand interactions, and the on-line integration of

Natural Products in Drug Discovery - Concepts Current Organic Chemistry, 2006, Vol. 10, No.8 901 immunochemical and enzymatic methods with chemoanalytical systems have been recently implemented. New off-line strategies such as HPLC-based profiling, directly applicable to a broad range of mechanism-based and cellular assays, have become increasingly popular in the context of industrial natural product screening. The most significant developments in this field are presented below, together with a selection of representative examples. BIOAUTOGRAPHY Bioautography was the first and quite successful attempt to combine chemo-analytical and biological assay principles in a seamless manner. Early bioactivity-related detection methods for TLC were established in 1970s already [22]. Bioautography offered the possibility of directly tracking bioactive compounds in complex mixtures with minute amounts of material at hand. The principle of bioautographic assays is straightforward (Fig. 1): A microorganism enzyme or biomolecule is applied onto a developed TLC plate. After suitable incubation, zones of inhibition may either become visible or may be visualized after a detection step involving enzymatic conversion of a chromogenic substrate [23]. Thanks to the relative simplicity of most assays and the possibility of parallel detection with various TLC staining reagents, bioautographic assays are of continuing popularity among natural product researchers. At first, direct bioautographic assays were established for antifungal activity using spore forming fungi [22]. Due to their innocuous and simple handling in a standard phytochemical laboratory, plant pathogenic fungi, such as species of the genus Cladosporium, have been particularly popular. Later developments included assays for antibacterial [24] and anti-Candida [25] activities, whereby microbial reduction of tetrazolium salts and agar overlays were needed for visualization and maintenance of microbial viability. There is meanwhile a wealth of antimicrobial compounds, which have been detected and isolated upon TLCbioautography (for a selection of recent examples see [26- 29]). At the same time, new assay setups, such as 2D-TLC [30], continue to be reported and the list of organisms which have been used in combination with TLC is further expanding. With the growing interest into natural antioxidants, assays using stable radicals such as DPPH as a detection reagent for radical scavengers were established. Although this test cannot be considered as a bioautographic assay in a strict sense, the link of chemical and biological properties is obvious and the assay predictive in the search of antioxidants. The DPPH assay was first proposed for the screening of antioxidants in marine bacteria from fish and shellfish [31]. It has become increasingly popular among phytochemists and is widely used for the detection of radical scavengers in plant [29, 32-36] and fungal [37, 38] extracts. The assay consists of spraying a methanolic solution of the purple coloured DPPH radical onto a developed TLC plate. Upon reduction, the compound turns yellow, and active compounds appear as yellows spots against a purple background. Another simple approach which has been used for the detection of antioxidants on TLC plates relies on the decolouration of b-carotene, a compound undergoing bleaching in the presence of air and light [33, 36, 39]. In this assay, substances with antioxidative properties appear as orange spots against a colourless background. After compound purification, the activity can be quantified in a corresponding solution assay using the water soluble carotenoid crocin, readily available from saffron [40]. Recently, attempts have been made to visualize natural products interactions with biomolecules, such as enzymes and nucleic acids, on TLC. Representative examples of such assays are TLC screens for new acetylcholinesterase inhibitors, as potential drugs in the treatment of Alzheimer’s disease. Two assay protocols involving different detection methods have been reported. In one of them [41], the TLC plate is sprayed with the enzyme, acetylthiocholine as a Fig. (1). Principle of bioautographic assays on TLC (reprinted from [23], with kind permission of Wissenschaftliche Verlagsgesellschaft mbH, Stuttgart)

902 Current Organic Chemistry, 2006, Vol. 10, No. 8 Potterat and Hamburger substrate, and Ellman’s reagent. The enzyme hydrolyses acetylthiocholine resulting in thiocholine which reacts with the Ellman’s reagent. The plate stains yellows and active spots appear as white spots. In an alternative method which has been applied to both actetylcholinesterase and butylcholinesterase, naphthyl acetate is used as a substrate and fast blue salt B as a detection reagent. Inhibitors of cholinesterases produce white spots on the background which is stained purple by the diazonium dye [42]. An interesting variant of bioautography has been developed for visualizing the binding properties of secondary metabolites to biomacromolecules [43, 44]. Binding can be detected via the differential chromatographic mobility of a compound with and without the presence of a target macromolecule. The method has been used in particular to detect interaction with DNA revealed by a significant decrease of the Rf value on the TLC plate. While bioautography is inexpensive and can be set up in almost every laboratory, its applicability remains limited by mainly two factors: The restricted number of relevant biological targets, which can be developed into an assay of this format and the lack of quantitative data. As a further drawback, there is no simple correlation with structural information provided by modern LC-coupled spectroscopic techniques. Dereplication of active compounds is thus not straightforward. HPLC-BASED ACTIVITY PROFILING Over the last decade, the focus in natural product analysis has shifted towards HPLC. The on-line coupling of HPLC with powerful spectroscopic methods provides a wealth of structural information from minute sample amounts without the need for tedious preparative isolation. Preparative HPLCMS is nowadays ubiquitous as a purification method in the pharmaceutical industry. The advent of high throughput purification platforms combining UV and MS triggering modes has let to an unprecedented level of automation in natural products isolation. With the development of microtitre based biossays, the great potential of HPLC micro-fractionation became apparent. Bioactivity can be tracked in complex mixtures without isolation of compounds and correlated with spectroscopic information available on-line. The principle of the approach is shown in Fig. 2 [45]: An extract is separated by analytical gradient HPLC. Via a T-split, a portion of the effluent is fractionated in the 96-well format, while the other part serves for the on-line spectroscopic characterization of the eluted peaks. After drying, the fractions are redissolved in a small amount of a suitable solvent, typically DMSO, and assayed for bioactivity. The activity profile is then matched with the HPLC chromatogram and the spectroscopic data. A targeted preparative isolation is carried out if the active principles are deemed of sufficient interest. While the potential of this approach has been first realized in the pharmaceutical industry, similar procedures have subsequently been implemented in academic research [46 - 49]. HPLC-Based Profiling in Medicinal Plant Research The identification of the cyclooxygenase-2 (COX-2) inhibitory principles in Isatis tinctoria L. (Brassicaceae) illustrates the potential of this approach in medicinal plant research [47]. In the search for the anti-inflammatory principles in lipophilic leaf extracts of this traditional dye and medicinal plant, a pronounced COX-2 inhibitory activity had been detected in a cell-based assay with Mono Mac 6 cells. Subsequently, the extract was submitted to activity profiling for rapid identification of the active constituents. The HPLC profiles, fractionation steps and COX-2 inhibition of individual fractions are shown in Fig. 3. In a first step, 11 fractions were taken at 8 min intervals. The inhibition profile revealed that virtually all activity of the extract was located in fraction 4. In a second round, the time window of 24-32 min was assayed at higher resolution. The COX-2 inhibitory principle was concentrated in fraction Tf 25-26 min. Finally, the active compound could be located in the shoulder at 25.0 min preceding the peak at 25.5 min. On the basis of ESI-MS and UV-vis data recorded on-line, the major peak was identified as indirubin (1), whereas the active compound was the indoloquinazoline alkaloid tryptanthrin (2) (Fig. 3). The compound was subsequently characterized as a potent dual inhibitor of COX-2 (IC50 in Mono Mac 6 cells 0.037 µM) Fig. (2). Principle of HPLC-based activity profiling (reprinted from [23] with kind permission of Wissenschaftliche Verlagsgesellschaft mbH, Stuttgart). N H O NH O indirubin (1)

904 Current Organic Chemistry, 2006, Vol. 10, No. 8 Potterat and Hamburger the overall effect. A typical example is shown in Fig. 4 with the tracking of the inhibitory activity of a lipophilic extract of Salvia miltiorrhiza Bunge (Lamiaceae) on recombinant monoamine oxidase (MAO) A and inducible NO synthase (iNOS) [49]. S. miltiorrhiza, known as “Danshen”, is a renowned Chinese medicinal plant chiefly used in the treatment of various cardiovascular disorders and in some infectious and inflammatory diseases. MAO A activity was assayed with a kinetic measurement of the conversion of kynuramine to 4-hydroxyquinoline [50]. Correlation of the activity profiles with the HPLC fingerprint strongly pointed towards the tanshinone-type diterpenoids as compounds with dual inhibitory properties for MAO A and iNOS induction. Targeted preparative purification afforded tanshinone IIA (3), 15,16-dihydrotanshinone I (4), cryptotanshinone (5) and tanshinone I (6). Dihydrotanshinone I was in both assays the most potent compound (MAO A: IC50 23 µM; iNOS: IC50 2.4 µM) while tanshinone IIA showed only marginal activity. The robustness of the kinetic assay for the detection of MAO A inhibition in the presence of potentially interfering compounds was established by spiking crude plant extracts, selected to contain diverse compounds exhibiting strong UV/Vis absorption and/or fluorescence, with small amounts of known inhibitors [50]. HPLC-Based Profiling in Industrial Environment HPLC-based activity profiling has been widely used in natural product HTS programs to prioritize hits and guide isolation work. Prioritization is of utmost importance in a HTS environment, since screens not seldom deliver a large number of hits exceeding by far the isolation capacities. A complete work-up of all active extracts is impossible within the timeframe allocated to follow-up activities in the lead discovery phase, which is usually restricted to as less as a few weeks. As a major asset, the microfractionation step can be performed with the bioactive sample stored in the library, without need of macroscopic isolation nor time-consuming refermentation or recollection activities and the associated issues of unsatisfactory reproducibility. With the activity profile at hand, isolation capacities can be dedicated to extracts where bioactivity correlates with chromatographic peaks. Preparative purification can be performed using a straightforward peak-guided strategy. Low priorities are assigned to extracts when activity cannot be recovered after fractionation or appears dispersed over a broad time window. The scale-up of the separation is straightforward, since the chromatographic system can be easily transposed to preparative HPLC. Examples of HPLC-based activity profiling in the context of industrial HTS programs have been recently reported. The application of this approach in a screening program for novel glucagon receptor antagonists demonstrates its potential for early dereplication. In this project, an extract of Streptomyces sp. strongly inhibited glucagon induced cAMP elevation. Semipreparative gradient HPLC separation of 30 µl of extract into 30 one-minute fractions and subsequent testing of each fraction enabled the activity to be attributed to a main peak at 16.1 min. (Fig. 5). HPLC-UV-MS analysis revealed a UV-spectrum similar to that of tryptophan and a MW of 2036 amu. Since no record corresponding to these data could be found in the literature, purification of the compound was undertaken on a preparative scale. The compound, a new bicyclic peptide (7), exhibited potent and selective antagonist activity towards the human glucagon receptor in a functional assay (IC50 0.44 µM) [51]. At the same time, micro-scale fractionation reliably filtered out extracts in which the inhibitory activity could be assigned to known compounds of little interest, such as cephalochromin (8) or resistomycin (9). The same strategy was successfully applied to the detection and isolation of a series of terphenylquinones as part of a HTS-supported lead finding process where more than 80’000 natural product extracts of plant and microbial origin were tested. Among other hits, an extract of a culture of the fungus Stillbella sp. showed a remarkable inhibition of the tyrosine kinase src, the prototype member of the src family of kinases involved in several signaling pathways. Comparative analysis of the LC-MS-UV dataset and the bioactivity plot of the fractionated extract indicated that the activity of this extract was based on a group of hydroxyquinone derivatives (10-13). Targeted purification of the active compounds was subsequently achieved by mass spectrometry-controlled preparative HPLC using ESIMS. The most potent compound, the trihydroxyphenyl derivative 12, inhibits the tyrosine kinase activity of human src with an IC50 of 3.9 µM [52]. A similar approach was also applied in the course of a screening program for new CDK5 inhibitors. CDK5 is a cyclin dependent kinase playing a crucial role in the development of the central nervous system and the regulation of neuronal signal transduction. A lipophilic extract of the stems and leaves of the shrub Clausena excavata Burm. (Rutaceae) was selected upon HPLC-based activity profiling. Activity profile and spectroscopic data strongly pointed towards a compound with a UV spectrum characteristic of a 9H-carbazole chromophor. Peak-guided isolation afforded a new alkaloid (14) with potent antagonist activity (IC50 0.51 µM) [53]. 15,16-dihydrotanshinone I (4) cryptotanshinone (5) tanshinone I (6) O O O O O O O O O O O O tanshinone IIA (3)

Natural Products in Drug Discovery - Concepts Current Organic Chemistry, 2006, Vol. 10, No.8 907 natural product extracts are separated using HPLC. A postcolumn biochemical assay determines the bioactivity of the effluent, while parallel mass spectrometry and diode array detection provide chemical characterization. Through additional coupling of evaporative light scattering detection (ELSD), quantitative information about the active components can be obtained. The greatest challenge of this technology is to find conditions which are compatible with both the HPLC separation and the biochemical assay. This on-line approach requires fast assays (typically 60-120 s) which match the timescale of HPLC chromatography. To increase the throughput, the LC-separation step can be omitted for primary screening and the assay performed in the flow injection mode. Active samples are subsequently reanalyzed with HPLC separation in the so-called highresolution screening (HRS) mode. A wide range of assay formats applicable to a large variety of protein targets have been developed. Basically, after separation by HPLC, a solution of the target is added and reacts with the analytes in a first reaction coil. A reporter ligand or substrate is subsequently added and interacts with the target in a second coil. Afterwards, the reaction products are analyzed using a UV, MS or fluorescence detector (Fig. 6). In one of the first on-flow applications, the Ellman’s procedure was adapted for the detection of acetylcholine esterase inhibitors in natural product extracts [57]. The enzyme reaction product was detected by a UV-Vis detector at 405 nm. Presence of an inhibitor produced a negative peak. The potential of this method for screening purposes was demonstrated by the identification of ungiminorine (17), as a new acetylcholinesterase inhibitor in an extract of a Narcissus cultivar [58]. Fluorescence-Based Detection Systems Current on-flow systems mostly use fluorescence-based readouts. In a common assay format, competition between binding compounds in the sample and a fluorescence labeled ligand is detected by means of a temporary fluorescence decrease in the effluent [56]. This approach has been used in particular for the detection of estrogen receptor a and b binding activity in plant extracts [59]. Coumestrol (18), a high affinity ligand for the oestrogen receptor was used as a reporter molecule. The binding of coumestrol to the receptor induces a blue shift and four times increase in fluorescence intensity. Luteolin (19), quercetin (20) and kaempferol (21) could be identified as active components of a pomegranate (Punica granatum L.; Punicaceae) pericarp extract [56]. An alternative procedure well suited for screening enzyme inhibitors is based on the measurement of substrate conversion. This assay format is applicable to various classes of enzymes, such as kinases, phosphatases, phosphodiesterases, and caspases. Presence of an inhibitor in the effluent produces a change in fluorescence intensity due to reduced substrate conversion. In the majority of cases, the substrate used does not exhibit high fluorescence prior to its conversion so that a negative response is obtained. An increase of fluorescence can be observed when the fluorescence of the substrate is decreased by the conversion. This assay format has been used for the screening of Fig. (6). Principle of an HPLC on-flow receptor assay (adapted from [56]; © 2003 Kluwer Academic Publishers, with kind permission of Springer Science and Business Media). O O N O H HO OH H unginimorine (17) O OH O OH HO O coumestrol (18) O O HO OH OH OH R R = H luteolin (19) R = OH quercetin (20)

908 Current Organic Chemistry, 2006, Vol. 10, No. 8 Potterat and Hamburger phosphodiesterase (PDE) inhibitors [60]. This enzyme family hydrolyses cGMP and cAMP to GMP and AMP, respectively, and has attracted much attention recently since selective inhibitors have a therapeutic potential in diseases such as hypertension, asthma, heart failure and erectile dysfunction. The test system was based on the conversion of the fluorescent reporter substrate mant-cGMP into the less fluorescent mant-GMP. Active compounds were detected by means of an increase in fluorescence. The assay was validated by spiking extracts with the known PDE inhibitors papaverine (22) and theophylline (23). In spite of their intrinsic selectivity, fluorescence-based assays suffer from background interferences which may decrease their sensitivity and lead to the assignment of false positives. Therefore, attempts have been made to switch to detection methods with increased selectivity, such as fluorescence resonance energy transfer detection (FRET). The suitability of FRET for on-flow systems has been demonstrated using the protease subtilisin as a model enzyme [61]. A further selectivity increase could be achieved with time-resolved FRET, which entirely suppresses the interferences from short lived fluorescence. A flow injection kinase assay using TR-FRET has been recently reported. The assay measures the phosphorylation of poly(GT)-biotin by the receptor tyrosine kinase EGFR [62]. While FRET fluorescence has been applied so far only to flow injection assays, work aiming at coupling this method to liquid chromatography is ongoing. MS-Based Detection Systems Mass spectrometry is an attractive alternative to fluorescence detection in on-flow assays. The principle consists of continuously monitoring a MS reporter molecule at its specific m/z. A major advantage over a fluorescence based measurement is that it does not require the synthesis and purification of labeled ligands [56]. As a further asset, this format is insensitive to natively fluorescent compounds which often cause problems in the screening of natural product extracts. The feasibility of MS based detection was demonstrated using streptavidin/biotin and antidigoxigenin/digoxin competitition model assays. Subsequently, a substrate conversion based assay involving the detection of the reaction products by ESI-MS was described [63]. The assay was validated with the cysteine protease cathepsin B. It is based on the decrease of the MS signal corresponding to the reaction product in response to the presence of an inhibitor in the effluent. The reduction of the product signal can be readily detected using the single ion trace locked on the product mass. Operating the MS in the scan mode provides structural information about the compounds responsible for the inhibition. Besides the research work carried out at the University of Amsterdam, there have been a few attempts to establish online detection systems for biologically relevant properties. Of particular interest is an on-line HPLC procedure for the detection of radical scavenging compounds in plant extracts. The method measures the reduction of the DPPH radical and can be run with mobile phase compositions ranging from 10 to 90% organic solvent in water or buffer. Reduction of the radical is detected as a negative absorbance peak at 517 nm [64]. Full integration of the bioassay into the HPLC flow, providing both chemical and biological information in a single run, can be viewed as the ultimate achievement in interfacing bioactivity and chromatography. However, there are some limitations for on-line systems: Separation and bioassay conditions must be compatible, or rendered to be so, with the aid of make-up buffers. The considerable consumption of immunochemicals, receptors or enzymes may represent a major cost factor when operating such online assays. The resolution is typically lower than in an offline configuration, since band broadening is induced by the post-column reaction. A further drawback in comparison to the off-line HPLC-based activity profiling is that on-line assays require substantial development and optimization for each target to be screened. CAPILLARY ELECTROPHORESIS-BASED ASSAYS Capillary electrophoresis is a well-established technique for high resolution separation of compounds based on their charge-to-mass ratio and shape. Affinity capillary electrophoresis (ACE) detects the binding of small molecules to a labeled macromolecule (enzyme, receptor, nucleic acids) through a shift in electrophoretic mobility due to changes in surface charges or target conformation. The ACE technology has been recently optimized for the detection of bioactive compounds in complex mixtures such as natural product extracts [65]. An ACE screening platform has been commercialized by the biotech company Cetek Corporation (Marlborough, MA), which claims to have established validated assays for more than 150 pharmaceutical targets (CE Assay™). The basic principle is the detection of a shift in the capillary electrophoretic mobility of a fluorescentlylabeled target in response to the binding of an active compound. The technology is well suited for the screening of natural product crude extracts, combinatorial libraries and N O O O O papaverine (22) O O OH HO OH OH kaempferol (21) N N N H N O O theophylline (23)