HIGH THROUGHPUT SCREENING

At least four considerations are required for an academic laboratory to use high throughput screening as part of their research arsenal. The first is to engineer an experimental situation in which the ‘system’ itself tells us the answer to a difficult problem. For example, establishing a method of identifying novel proteins involved in host cell invasion by a parasite (Carey et al. 2004). The second relates to the availability of the necessary technology. There has been a significant drop in the cost of the required robotic equipment and it is now possible to buy an automated liquid handler and multimode plate reader with a minimal budget. Third, compound collections are now commercially available and a myriad of companies cater to the increasing demand (Table 1). Fourth, academic researchers have realised that they can not only validate a novel protein target but can also identify ‘proof of concept’ compounds that modulate the protein's activity (Smith et al. 2004).

TARGETED HTS

Providing an exhaustive literature review of high throughput screening in parasitology is a daunting task. We have therefore chosen to review only a sub-section of the literature (for a more complete review see Werbovetz, 2000). All the articles we will discuss in detail have a common link: the use of unbiased compound collections. This decision means that a large number of targeted HTS reports in parasitology will not be discussed. Our view is that in the majority of these articles a lead compound against the selected parasite target already exists and that the main focus of the papers is usually not on the discovery aspect but on lead optimisation. For example, the recent report of the screening of kinase inhibitors against the Leishmania mexicana CRK3 Cyclin-Dependent Kinase, although interesting, is not discussed in detail here as this research is inspired by previous knowledge of likely kinase inhibitor structures (Grant et al. 2004).

As a result of intellectual property issues, financial constraints and the many technical barriers inherent in establishing a miniaturised functional protein assay, there are very few cases where purified parasite proteins have been screened against unbiased compound collections (e.g. Samson et al. 1995; Asai et al. 2002). An excellent example of this approach is the work by Asai et al. (2002). Toxoplasma gondii expresses the protein nucleoside triphosphate hydrolase (NTPase) in two isoforms, NTPase-I and NTPase-II. These enzymes are released into the parasite-containing vacuole and are essential for tachyzoite replication within the host cell. Approximately 150000 compounds were assayed against NTPase in T. gondii in an automated 96-well format. The activity of the compounds was monitored by the presence of orthophosphate, formed by hydrolysis of ATP. Five compounds were shown to inhibit both NTPase-I and NTPase-II. One of the compounds was a selectivite inhibitor of NTPase-I. On further investigation all five compounds were found to inhibit parasite replication. With the availability of genome sequences and improvements in our ability to classify and express recombinant proteins, it seems likely that the number of examples of this HTS mode will increase dramatically. The use of small molecule chips to aid high throughput binding assays, followed by low throughput functional assays, will also enhance this discovery mode (Koehler, Shamji & Schreiber, 2003).

A second class of HTS projects that use an unbiased compound set does exist but this set is a virtual one. Table 2 summarises reports of projects that have started with an in silico screen against a specific protein. In the majority of cases, these studies are aided by the availability of structural information on the chosen protein, although homology models are also used. For example Bond et al. (1999) used this approach to study the protein target trypanothione reductase (TR) in the protozoan parasite Trypanosoma cruzi. This protein was selected due to its potential importance in the survival of the parasite. TR reduces trypanothione disulfide (T[S]₂) to dihydrotrypanothione, which can in turn reduces glutathione disulfide (GSSH) to glutathione (GSH). GSH protects cells by removing potential oxidising agents. Bond et al. (1999) obtained a crystal structure of TR with its T[S]₂ substrate bound. An area of the active site of TR which is the most structurally dissimilar to the human equivalent glutathione reductase (GR) was choosen to screen the Cambridge Structural Database (CSD) for potential inhibitors. Two natural products, cadabacine and lunarine, were identified from this screen and considered suitable for further investigation. Lunarine, which is commercially available, showed inhibition kinetics indicating non-covalent binding to TR followed by covalent binding to the reduced form of TR.

PHENOTYPE HTS

The workhorses of HTS in parasitology are the cell-based live/dead assays. This class of assays includes the tritiated hypoxanthine incorporation assay used to determine the activity of a compound against P. falciparum (Harmse et al. 2001). By carrying out these assays in conjunction with mammalian cell cytotoxicity assays, it is possible to make an initial assessment of whether or not a compound has an acceptable therapeutic window. The protein targets of compounds cannot be assigned from these assays, but the assays do provide a clear indication of relevant biological activity in a cellular environment. Recent technological advances enable the use of more focused cell-based assays to assess the effect of compound collections on specific aspects of the parasite lifecycle. An example that we have been involved in is the use of high throughput imaging technology to identify small molecules that perturb host cell invasion by T. gondii. Fig. 1 summarises the approach used to screen 12160 commercially available compounds in this study.

Fig. 1. A phenotype-based high throughput invasion assay. Each well of the 384-well plate contained either a unique compound or an equivalent volume of DMSO (control). Intracellular and extracellular parasites were distinguished by fluorescence microscopy; extracellular parasites were red and green appearing yellow in merged images; intracellular parasites were green only. Compounds that reduced the invasion levels to <20% compared to control wells were considered inhibitors. This figure is reproduced from Carey et al. (2004) copyright, National Academy of Sciences, USA. (Bar=10 μm.)

In addition to the supply of parasites, compounds and the miniaturisation of the assay format, two other key factors made this screen possible. Firstly, just short of 100000 images had to be captured. This was done at the Institute of Chemistry and Cell Biology (ICCB) at Harvard Medical School using an inverted fluorescence microscope fitted with an x/y stage and piezoelectric-motorized objective holder. Metamorph® software was used to control auto-focusing, all aspects of image acquisition and movement between the 4 different fields sampled in each well and between wells. The same technology has also been used to identify compounds that modulate the early stages of the wound healing response (a process that is analogous to parasite invasion in that it involves the cell motility machinery). This work has recently been published (Yarrow et al. 2004) and provides further details on the automated microscopy. Whilst acquiring, storing and accessing the images proved challenging, sorting through them to identify hits proved impossible by hand/eye. An algorithm to enable automated image analysis was therefore established. The goal of this algorithm was to identify and count fluorescent objects (parasites) present in each of the 100000 images. The number of parasites per field was then used to calculate the average number of invaded parasites per field (by subtraction of parasite numbers in paired images, see Fig. 1). This number was compared to that in the control wells present on each plate and inhibitors of invasion were defined as compounds that gave invasion levels of less than 20% of the control levels. A parasite was defined by input of numerical values for minimum and maximum object area. This enabled fragments or clumps of parasites to be excluded. The overall protocol was sufficiently reliable to identify 65 inhibitors from the primary screen. This number was eventually reduced to 24 using additional selection criteria (Carey et al. 2004).

THE NEXT CHALLENGE: TARGET IDENTIFICATION

Phenotype-based approaches, such as the one discussed above, minimally deliver compounds that are active in cells. Unfortunately, unless the detailed phenotype provides additional clues (Mayer et al. 1999), such approaches do not provide information about the protein target(s) of the compound. For this, a series of other techniques and some good fortune is required. Numerous articles summarising biochemical approaches to protein target identification have been published recently, reflecting the fact that this is becoming one of the major challenges in post-genomic research (King, 1999; Gray, 2001; Ward, Carey Westwood, 2002; Lokey, 2003). Here we focus on the input that the biologically-aware chemist can, and arguably, must have. To date, our laboratory has encountered six scenarios in which chemical input is required.

Affinity chromatography

These studies typically involve the immobilisation of the active compound onto a water-compatible matrix. The resulting affinity matrix is then used to capture protein targets from crude or prefractionated cell lysates. In order for the compound to be optimally displayed for protein binding, it is necessary to modify its structure by the addition of an inert linker unit. For example, in order to identify parasite-derived binding partners of olomoucine, the agarose 95 matrix was prepared (Fig. 2) (Knockaert et al. 2000). Deciding where the linker unit should be attached so that it has a minimal effect on biological activity frequently requires rounds of analogue synthesis and retesting. As shown in Fig. 2, studies of this type also require the preparation of a control matrix (agarose 95 M matrix). This involves the use of an inactive analogue of the compound that is as close in chemical structure to the original as possible. The use of this matrix enables the identification of proteins that bind non-specifically to the matrices. Analogous studies using other kinase inhibitors known to be of relevance to parasite systems (Taddei and Westwood, unpublished results) and several of the 24 inhibitors that were identified in the T. gondii screen (Morgan, Ward and Westwood, unpublished results; Carey et al. 2004) are ongoing in our laboratory using these techniques. Affinity chromatography can also be used in a second mode. ATP or cGMP binding subsets of the crude lysate are isolated using ATP or cGMP immobilised resins. Comparison of proteins bound to the resin in control lysates versus lysates that have been pretreated with the compound of interest leads to the identification of protein binding partners (Graves et al. 2002). This was elegantly used (Gurnett et al. 2002) to provide additional evidence in support of the assignment of a cGMP-dependent protein kinase (PKG) as the target of the pyrrole containing compound 1 shown in Fig. 4.

Fig. 2. Successful target identification in parasitology – affinity chromatography. Affinity chromatography studies were carried out using crude protein extracts from four different protozoan parasites Plasmodium falciparum, Leishmania mexicana, Trypanosoma cruzi and Toxoplasma gondii and resins loaded with either 95 or 95 M, see B. The bound proteins were eluted and analysed by SDS-PAGE/silver stain, see A. The gels clearly show the isolation of one protein CK1 for P. falciparum, two proteins are identified with L. mexicana, CK1 and an unknown protein. CK1 was also pulled down from the T. gondii and T. cruzi extracts (with two other unidentified proteins in the latter case, Knockaert et al. 2000)*Reprinted from Knockaert, M., Gray, N., Damiens, E., Chang, Y-T, Grellier, P., Grant, K., Fergusson, D., Mattram, J., Soete, M., Dubremetz, J-F., Le Roch, K., Doerig, C., Schultz, P. G. & Meijer, L. (2000). Intracellular targets of cyclin-dependent kinase inhibitors: identification by affinity chromatography using immobilised inhibitors, Chemistry and Biology, 7, 411–422. Copyright (2000), with permission from Elsevier..

High throughput synthesis technology to increase compound potency and aid rapid generation of structure activity relationship data (SAR)

It is currently a very rare occurrence for a commercial compound collection to yield a hit that enables direct target identification. One reason for this is that the initially identified compounds are often not very potent. It is therefore necessary for the activity of a compound to be improved before techniques such as affinity chromatography can be used successfully. Fig. 3 summarises the type of technology that is available to chemists as they try to increase the potency of an analogue series as rapidly as possible.

Fig. 3. Chemical technology that aids protein target identification. The main goal of this technology is to increase the rate of analogue synthesis enabling the rapid acquisition of structure activity relationship (SAR) data (Hird, 1999). Images A–D show examples of the technology used in our research group. A and B: semi- and fully-automated reaction blocks often supplied with filtration and separation devices, enabling both solid and solution phase chemistry. Maximum use of these systems requires automated purification systems (up to 10 compounds can be purified in parallel, equipment not shown); C: focused microwave apparatus. This technology enables reactions that have been traditionally carried out on the time scale of hours to be carried out in minutes. For a recent application from our laboratory see Patterson et al. (2004); D: automated analytical systems. The liquid chromatography-mass spectrometry system shown here enables the assessment of compound purity and structure 24 hours a day. Additional analytical techniques include evaporative light scattering and both fully automated and flow NMR.[Dagger]A web based search identified the following companies as manufactures of technology required for library synthesis. The letters in brackets correspond to the type of instrumentation they supply. A=manual reaction blocks, B=automated systems, C=Microwaves, D=LCMS systems, E=HPLC systems, F=ELSD's, G=NMR systems. Accelab (B), Advanced ChemTech (B), Agilent Technologies (D, E), Alltech Associates (F), Applied Biosystems/MDS SCIEX (D), Argonaut Technologies (B, E), Biotage (E), Brucker Biospin (G), Buchi (A, E), CEM Corporation (C), Charbdis Technologies (A, B, E), Chemglass (A), Chemspeed (A, B), CSPS Pharmaceuticals (A), FlexChem (A), GBC Scientific Equipment PTY Ltd (E), Genevac (A), Gilson Inc. (B, E), H&P (A), Hitachi (E), Heidolph (A), HEL Inc (B), Innolabtech (B), Isco Inc (E), J-Kem Scientific (A), Mettler Toledo (Bohdan) (A, B), Milestone Inc. (C), Orochem Technologies (A), Personal Chemistry/Biotage (C), Polymer Laboratories (F), Prolab Instruments (B, E), Radleys (A, E), Rapp Polymere (A, B), Robosynthon (A), Sedere (F), Shimadzu (D, E), Stem corporation (A), Spyder Instruments (A), Tecan (B), Torviq (A), Varian Inc. (D, E, G), Waters (D, E) and Zinsser Analytic (A, B). Solid phase technology has also improved facilitating split and pool synthesis (e.g. IRORI and Mimotopes, not shown (Ley & Baxendale, 2000)). In the majority of cases, the advancements in technology described here shift the rate determining step in analogue synthesis away from practical issues and back to fundamental questions of chemical design and feasibility.

The use of parallel synthesis equipment coupled with automated purification and analysis technology speeds up this process. For example, we recently prepared 96 analogues of an invasion inhibitor of T. gondii, resulting in the identification of compounds with increased potency and a clearer view of the substituent-dependence of the biological activity (Catti, Ward and Westwood, unpublished data). However, there is often a lag period while chemists become acquainted with the chemistry associated with a particular hit. When the compound has been selected using HTS from a collection previously prepared by the chemist this delay is reduced.

Radiolabeled analogue synthesis

One of the most inspiring reports of successful target identification in parasitology comes from the work of the Merck Research Laboratories (Gurnett et al. 2002; Fig. 4). They showed that the pyrrole containing compound 1 has broad spectrum activity against Eimeria in poultry and T. gondii in a murine toxoplasmosis model (Gurnett et al. 2002; Nare et al. 2002). The target of 1 was identified using a tritiated version of 1 prepared in house at Merck. As Fig. 4A and B show, very few proteins other than cGMP-dependent protein kinase (PKG, 120 kD band), specifically associate with radiolabeled-1 in a ligand-binding assay using soluble extracts of unsporulated E. tenella oocysts. Further studies showed that 1 inhibits the activity of PKG with an IC₅₀ of 0·8 nM. The evidence in support of PKG being the cellular target of 1 was further strengthened when mutant parasites that are resistant to 1 in vitro and in the mouse in vivo model were produced by incorporating mutations into the catalytic site of PKG (Donald et al. 2002). More recently, 1 has been used to study the role of PKG in a series of parasite related processes (Wiersma et al. 2004).

Fig. 4. Successful target identification in parasitology – radiolabeled analogue synthesis. A Chromatogram indicating the two fractions (34 and 35) bound by radiolabled compound 1. B Silver stained gels indicate two bands at 150 kD and 120 kD which are present in both fractions 34 and 35 (Gunnett et al. 2002)**Reprinted from, The Journal of Biological Chemistry, (2002) 277, 18. Gurnett A. M., Liberator, P. A., Dulski, P. M., Salowe, S. P., Donald, R. G. K., Anderson, J. W., Wiltsie, J., Diaz, C. A., Blum, P. S., Misura, A. S., Tamas, T., Sardana, M. K., Yuan, J., Biftu, T. & Schmatz, D. M. Purification and Molecular Characterization of cGMP-dependent Protein Kinase from Apicomplexan Parasites. A Novel Chemothreapeutic Target, 15913–15922. Copyright (2002), with permission from The American Society for Biochemistry and Molecular Biology. http://www.jbc.org. C Structure of the radiolabeled compound used in these target identification stuidies.

This research emphasizes the dramatic effect that compounds can have on biological research and in part justifies the time and resources required. The synthesis of sufficient quantities of radiolabeled analogues can be difficult in an academic laboratory, but our experience has shown that with the ‘cold’ synthetic route in hand there are several contract synthesis companies that can help to progress this approach to target identification (Patterson, Westwood, unpublished data).

CONCLUSION

This review has focused on the use of unbiased compound collections as a means of linking parasite biology and synthetic chemistry. In particular, it has described targeted and phenotype-based high throughput screening approaches. The ultimate driving force behind the marriage of chemistry and parasitology is, of course, drug discovery. In this context, the majority of the research discussed here only impacts on the early phases of this complex process (particularly when it is carried out in academia). However, the impact is still a crucial one and, in our opinion, is set to increase as the technologies discussed here become more common-place in academic environments (Verkman, 2004). A recent review (Burdine & Kodadek, 2004) argues the major challenge arising from phenotype-based assays is identifying the protein target(s) of hit compounds. There appears to be a renewed optimism (or is it determination) about our ability to succeed in target ID programmes. Whether this is misplaced or not, it is clear that this venture is a multidisciplinary one that requires the input of biologists, chemists, informatics experts and cutting edge technology.