Structure-based drug discovery approaches

In recent years, the structure-based drug discovery (SBDD) shows promise as a good way of developing highly specific drugs with high binding affinity (Blundell et al., 2002; von Itzstein, 2008b). The high-resolution X-ray crystal structures always provide valuable information in SBDD, especially the complex structures of the targeted enzymes with potential leading compounds. Based on the structural features, modification of a candidate compound framework for a more efficient pattern of protein-ligand interactions is feasible. Remarkable examples are the SBDD efforts targeting the influenza virus sialidases, which successfully yielded two marketed antiviral drugs, Zanamivir (von Itzstein et al., 1993) and Oseltamivir (Kim et al., 1997). Both drugs were developed on the basis of transition-state analogue Neu5Ac2en skeleton (previously shown in Section 1.2) with the crystal structures of influenza virus N2 and N9 enzymes, and both show nanomolar-range binding affinity towards the viral sialidases (von Itzstein, 2007).

Figure 1.15 The structures of Zanamivir and Oseltamivir. (a) Zanamivir (Relenza), formula: C12H20N4O7; MW: 332.13 g/mol; (b) Oseltamivir (Tamiflu), formula: C14H24N2O4; MW: 284.35 g/mol.

However, it is not simple to obtain an ideal starting framework for rational drug design. One strategy is to use the currently known substrate of the targeted enzyme as the initial template. Based on the catalytic mechanism, it is aimed to design new compounds with higher binding affinity and lower metabolic speed that mimicks the transition state of catalysis, as seen in the case of anti-influenza virus development (von Itzstein, 2007, 2008b).

Another promising strategy is a combinatorial screening of the selected target against libraries of small molecules with MWs 100-300 Da (termed as fragments) exhibiting relatively high ligand efficiency (Bembenek et al., 2008). The well-recognized criteria for the fragment selection is “Rule of Three”: the MW of the fragment should be below 300Da; the number of hydrogen bond acceptors should not exceed 3 and the cLogP is

≤ 3 (Congreve et al., 2003). The major advantages of the fragment-based drug

discovery (FBDD) approach include the fact that the small molecules have high structural diversity and significantly reduced structural complexity, and no previous knowledge of the catalysis of the targeted enzyme is necessary (Fattori et al., 2008). During the past decade, such a fragment-based screening strategy has evolved to a working procedure that consists of a variety of techniques used at different stages of fragment screening, leads optimisation and validation. These techniques now include the fluorescence-based thermal shift assay, saturation transfer difference-nuclear magnetic resonance (STD-NMR), isothermal titration calorimetry (ITC), fragment cocktail crystallography, enzyme kinetics and some virtual docking techniques (Ciulli and Abell, 2007), good combination of these techniques would make the FBDD screening more efficient.

The fluorescence-based thermal shift assay is normally selected for the first-line high- throughput screening against large libraries, as only about 1-10 µM protein and 1-10

mM compounds are used in each 50 µl reaction in a 96-well or 384-well plate. In this

assay, an environment sensitive fluorescent dye Sypro Orange is used, which binds to the gradually exposed hydrophobic region of the protein in a thermal unfolding process and shows fluorescence at 490/530 nm. The thermal unfolding curve of fluorescence intensity against temperature would give the Tm (midpoint temperature of thermal

unfolding ramp), a key parameter reflecting the thermostability of the protein. Compounds that stabilize the protein for a ΔTm of over 0.5 °C could be considered as

hits (Lo et al., 2004).

The hits from the thermal shift assay then could be verified by STD-NMR spectroscopy, which could detect the fragments with binding affinities in the 10-3 to 10-

8 _{M range, even from a substance mixture (Klages et al., 2007). In STD-NMR, a}

selective irradiation is applied to the protein, where efficient saturation could be transferred within the protein and to the bound ligand via spin diffusion, resulting a detectable difference signal compared to a reference spectrum without on-resonance irradiation. This technique also enables the mapping of binding epitopes between the ligand and protein (Meyer and Peters, 2003). Being a powerful method in FBDD approaches, STD-NMR is widely used in both hits screening and leads validation. Once high quality protein crystals were obtained, X-ray crystallography could be the most powerful tool for obtaining direct evidence of how the fragments or leading compounds bind to the protein. In practice, crystals can be soaked with individual hits separately or with cocktails of about 10 fragments from the library, where the final concentration of the fragments are normally 50-100 mM, about ten times over the binding affinity observed in kinetics (Blundell and Patel, 2004; Jhoti et al., 2007). In addition, virtual screening approaches also play a key role in SBDD, enjoying major advantages in many aspects like high speed, easy operation, low cost and good

predictions. Moreover, the computational docking method is particularly important in validating the various designs of novel compounds based on previously leads (Villar and Hansen, 2007). Some quantitative methods, such as ITC and enzyme kinetics, are also essential in lead validation. A combination of these techniques makes SBDD no doubt a reliable way for novel antimicrobial development.

Figure 1.16 Schematic view of FBDD strategy. Modified from previously published scheme (Ciulli and Abell, 2007; Orita et al., 2008).

From the fragments lead-like hits, some different approaches are available for the development of drug-like molecules. For instance, the fragment evolution approach often refers to building a larger compound with more functional groups based on leads. In fact, some fragments could be sub-structure components of a starting skeleton, with the linking of two or more fragments at suitable placement in the enzyme active site. A third approach is fragment self-assembly, which uses the reactive fragments to form an active inhibitor on the protein template. The fragment optimisation is also a common way in lead progression, which involves engineering the ADME properties (absorption, distribution, metabolism and elimination) of only a particular part rather than just binding potency of the leading hit (Alex and Flocco, 2007; Rees et al., 2004). Once novel drug-like molecules were obtained, further evaluation and validation using the various techniques mentioned above will be carried out to confirm if potent inhibitors have been achieved.

The current projects are aiming to determine the crystal structures of sialidases from S. pneumoniae and P. aeruginosa. With the crystals of these targets, FBDD from our fragment libraries and SBDD from sialic acid analogues would be feasible. Preliminary work of our SBDD efforts is present in Chapter 4.