Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Biophysical screening in fragment-based drug design: a brief overview

Biophysical screening in fragment-based drug design: a brief overview Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 BioscienceHorizons Volume 11 2018 10.1093/biohorizons/hzy015 ............................................................................................ ..................................................................... Review article Biophysical screening in fragment-based drug design: a brief overview Jacob Robson-Tull Department of Life Sciences, Imperial College London, London SW7 2AZ, UK *Corresponding author: Jacob Robson-Tull, Department of Life Sciences, Imperial College London, London SW7 2AZ, UK. Email: jacob.robson-tull16@imperial.ac.uk Supervisor: Dr Ernesto Cota, Department of Life Sciences, Imperial College London, London SW7 2AZ, UK ............................................................................................ ..................................................................... High-throughput screening (HTS) has been firmly rooted at the heart of many drug discovery programs over the past few decades, having provided a starting point for the development of many of the drugs on the market. However, this technique is often accompanied by high-attrition rates and has proven to be unsuccessful at developing therapeutics against more non-conventional targets such as protein–protein interactions (PPIs). Fragment-based drug discovery (FBDD) offers an alter- native approach and is steadily being taken up by the industry to tackle the high-attrition rate associated with many chem- ical leads developed by HTS. FBDD takes a structure-guided approach and uses a small chemical library of fragments during the initial screening process, to identify weakly binding ligands with the potential for therapeutic development. This review aims to summarise the challenges with screening in FBDD and the key biophysical screening techniques used for identifica- tion of weakly interacting ligands. Key words: fragment-based drug discovery, biophysical, protein–ligand interactions, fragment screening, fragment-based lead discovery, ligand screening Submitted on 27 September 2018; editorial decision on 27 November 2008 ............................................................................................ ..................................................................... 30% of attrition during clinical trials (Segall and Barber, Introduction 2014), some of which is likely due to off-target interactions. Different approaches have since been developed that take a Drug discovery programs have seen huge changes over the structure-guided approach to drug design, to target specific past 20 years, with the advent of new technologies constantly proteins and reduce toxicity. changing the approach to drug discovery and design. High- throughput screening (HTS) has been the technique of choice Fragment-based drug design (FBDD) is a recently devel- for large pharmaceutical companies due to its ability to screen oped workflow that addresses some of the shortfalls of other large numbers of molecules to discover chemical leads techniques and has been successful in developing numerous (Macarron et al. 2011). The development of higher quality clinical-stage drugs, as shown in Fig. 1. FBDD typically chemical libraries has enabled HTS to continue to identify involves screening chemical libraries composed of up to a few hits for conventional drug targets such as protein kinases thousand small molecules, as opposed to millions in HTS (Barker et al., 2013). However, HTS has proven unsuccessful libraries, through a variety of biophysical techniques such as for screening compounds against more difficult targets such NMR spectroscopy (Erlanson et al., 2016). These small mol- as protein–protein interactions (PPIs; Hubbard, 2016), which ecule libraries, termed fragment libraries, are attractive as are of increasing therapeutic importance. What is more con- they address some of the current causes of attrition of lead cerning is that recent estimates place toxicity as a cause of compounds. Fragments are low-MW molecules (<300 kDa) ............................................................................................... .................................................................. © The Author(s) 2019. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. and form weaker interactions with protein targets, normally challenges that must be overcome due to the weakly interact- in the mM–μM range (Murray and Rees, 2009). Although the ing nature of fragments. This review will summarise some of interactions are weaker than those observed in HTS, the inter- the challenges of fragment screening and recent developments actions tend to be higher quality, with each non-hydrogen in biophysical techniques that have helped overcome them. atom (heavy atom) involved in the binding interaction. A low-MW and high-interaction quality give more freedom in Fragment screening the fine-tuning of the physiochemical properties later in the drug development process (Hubbard, 2016), making them HTS typically involves the use of a robust, cell activity-based attractive starting points for lead compounds. Fragment assay such as the luciferase reporter screen (Macarron et al. libraries can also sample a large range of chemical space due 2011). Since fragments are generally smaller and simpler than to their inherent size (Hann, Leach, Harper, 2001), which compounds used in HTS, cell activity assays are difficult to allows them to better fit binding sites and pockets. Together, use as the binding event is usually too weak to cause a detect- these properties contribute to the higher hit rate observed for able change (Murray and Rees, 2009). Therefore, more sensi- fragment libraries than that of HTS libraries. tive biophysical techniques must be employed to detect these interactions, but these present their own challenges. The FBDD is now commonplace in drug discovery programs throughput of these techniques is generally lower and they and has even enabled academic participation due to smaller often require larger quantities of protein. The fragments also chemical libraries and availability of biophysical screening need to be highly pure, and soluble at high concentrations to instruments (Rees et al., 2004). Although the quality of hits in detect their binding. However, this can be an advantage as FBDD is generally higher than HTS, there are certain approved drugs tend to have a low lipophilicity (Lipinski et al., 2001). Nevertheless, biophysical techniques can offer insight into the binding mode of fragments, enabling fragment hits to be developed into lead-like molecules in a structure- guided approach, as summarised in Fig. 2. Several key techni- ques will be focused on, including NMR spectroscopy, sur- face plasmon resonance and X-ray crystallography. It should be noted that there are other techniques used and in develop- ment, although these are less commonly practised at present. NMR spectroscopy FBDD was first made possible 20 years ago by the develop- ment of NMR spectroscopy as a tool to screen fragments by Shuker et al. (1996), in what was called a structure-activity relationship by NMR. Since then, numerous NMR-based techniques have been developed. NMR spectroscopy is well suited to studying protein-fragment interactions as it is pos- Figure 1. Fragment-derived drugs at various stages of clinical trial as sible to determine the binding affinity and binding mode of of 2016. Data from Erlanson (2016). fragments, allowing for dissociation constants (K ) in the Figure 2. Summary of an FBBD program. (a) Selection of an appropriate target compatible with the chosen biophysical screening technique. (b) Target protein is produced and purified. (c) The fragment library designed for the target protein. (d) Biophysical screening of the fragment library. (e) Validation of hits from screening to identify the fragment binding mode. (f) Development of the fragment(s) into a lead molecule. ............................................................................................... .................................................................. 2 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. mM range to be identified (Carr et al., 2005). Additionally, targets. To the contrary, WaterLOGSY can be used to calcu- NMR spectroscopy observes the direct binding event between late dissociation constants through ligand titration, although the protein and fragment, reducing the chance of false posi- special care must be taken to ensure that the dissociation con- tives from assay artefacts as seen in HTS (Wu et al. 2013). stant measured is not dependent on experimental conditions NMR spectroscopy can be used in two different approaches, (Huang et al., 2017). Other ligand-observed techniques exist, either ligand-observed or protein-observed NMR, and these such as CPMG and relaxation experiments (Jhoti et al., are sometimes used in combination to optimise throughput. 2007), but these are not as commonly used as STD and waterLOGSY. Before fragment screening can take place, the fragment Protein-observed NMR is a more laborious process, but it libraries must be optimised as NMR is particularly sensitive is the gold-standard in NMR analysis of protein–ligand inter- to false positives from certain types of fragments. These types actions, allowing for determination of the binding affinity and of interfering fragments often form non-specific interactions binding mode of fragments. It can be used to follow up on with proteins, contain reactive covalent modifiers, are chela- hits from an initial screen by ligand-observed NMR. This tors or are aggregators, and so libraries must be designed to technique requires large quantities of protein (30 nmol) and exclude these (Harner, Frank, Fesik, 2013). Unfortunately, the protein must be isotopically labelled (Renaud et al. 2016). these types of fragments can be difficult to identify due to 1 15 For proteins up to ~40 kDa, a series of H– N HSQC spectra unexpected reactivity with proteins (Baell, 2010), thus caus- can be recorded whilst titrating fragment. This allows the ing false positives. Since high fragment concentrations are binding affinity to be determined and if the spectrum has pre- required for screening, to saturate low-affinity binding sites, viously been assigned, the binding site can be approximated. these effects are likely to become more pronounced in NMR The sensitivity of this technique is increased by using higher screening. A 1D spectra of the fragments in solution may be magnetic field strengths (Kiraly et al. 2015) to identify more measured at 1 mM (Ma et al., 2016) and from this, an estima- 1 15 distinct peaks. For proteins whose H– N HSQC spectrum tion of aqueous solubility can be obtained by comparing the has not been assigned, FBDD campaigns can still occur, dem- spectra to a reference. However, it is not known how the frag- onstrating the power of FBDD. The binding affinity can still ments will behave until they are screened. Even with optimisa- be determined by fragment titration. A technique termed tion, Baell (2010) highlights that some false positives will still ‘NOE matching’ can model the binding mode of ligands by not be identified in early screens. analysing 3D C NOESY spectra and comparing it to a com- Ligand-observed NMR is suitable for screening as it mea- putational model (Constantine et al., 2006). However, this sures the change in intensity, sign or relaxation of fragment technique is not popular at present, perhaps because it protons in the presence of the target protein (Ma et al. 2016). requires several protein–ligand NOEs to be observed and This requires no isotopic labelling of the protein and smaller assigned; this is difficult for fragments as they are small and quantities, thus increasing throughput when compared to interact weakly with target proteins and so its applications to protein-observed NMR. There is no theoretical protein size FBDD may be limited. For proteins larger than 40 kDa, 1 15 1 13 limitation but ligand-observed NMR tends to be used for pro- H– N TROSY HSQC or H– C HSQC can be recorded teins between 15 and 100 kDa (Harner, Frank, Fesik, 2013). for specific isotopically labelled residues such as valine, leu- Saturation transfer difference (STD; Mayer and Meyer, 1999) cine and isoleucine (Hajduk et al., 2000). Binding sites typic- and waterLOGSY (Dalvit et al., 2000) are used most often in ally contain hydrophobic residues, especially in pockets, and FBDD screening and these rely on nuclear Overhauser effect so these residues can be used to report on ligand binding. (NOE) transfer of magnetisation, from the protein to the frag- Although large quantities of protein are required, the protein ment, as shown in Fig. 3a and b. These two techniques are can be recycled for different experiments, with yields of up to used in combination to cross-validate any hits as many false 80% (Harner, Frank, Fesik, 2013). A series of other protein- positives can occur, particularly with STD where fragment observed experiments exist, such as F amino acid labelling methyl protons may be saturated. (Gee et al. 2016). Unfortunately, STD does not give information on the bind- Inhibitors developed against myeloid cell leukaemia 1 ing mode and cannot directly distinguish between specific and (Mcl-1) protein are just one example of where NMR has had non-specific binding interactions, so follow up on hits is great success in screening fragment libraries (Friberg et al. always required. A reporter screen (Jahnke et al. 2002) can 2012). Mcl-1 dysregulation prevents cancerous cells from make use of known ligands to identify binding specificity undergoing apoptosis and has been a promising cancer target (Fig. 3c), but this will lead to a decreased throughput. for many years (Adams and Cory, 2007). However, Mcl-1 is Reporter screens can also enable detection of higher affinity difficult to target as it exerts its effects through PPIs and as of interactions and F atoms can be incorporated to increase 2012, there were no inhibitors in clinical trials. Friberg et al. sensitivity and throughput (Dalvit, Flocco, Marina Veronesi, (2012) overcame this challenge using FBDD. Fragments were 1 15 2002). Alternatively, quantitative STD analysis can be used to screened via SOFAST H– N HMQC, which is similar to a identify binding specificity and does not require a known lig- HSQC-type experiment, and a series of fragments, shown in and of the target protein (Cala and Krimm, 2015). This is Fig. 4, were identified for progression. Since the hit rate for more attractive as it can be applied to a wider range of protein proteins is typically below 1% (Hajduk, Huth, Fesik, 2005), ............................................................................................... .................................................................. 3 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Figure 3. NOE-based screening techniques. (a) In STD, the protein methyl groups are saturated and eventual cross-saturation of the bound ligand causes a signal intensity decrease. RF is the radio frequency pulse applied to the sample. The H spectra before and after protein addition are used to calculate a difference spectrum and identify if a fragment has bound. (b) In waterLOGSY, fragments that bind the target protein are often identified by positive resonances and those that do not have negative resonances. Bound ligands in waterLOGSY are therefore identified by peak inversions. (c) Ligand-observed reporter screening. (ci) The known ligand has sharp peaks in a H spectrum when isolated. (cii) Upon addition of protein, the peaks are broadened due to reduced tumbling when the ligand is bound to the protein. (ciii) Sharp peaks of the known ligand are restored if a fragment competes for binding. it is unlikely for 2 hits to occur in the same fragment cocktail. success in fragment screening due to its sensitivity and incred- Hence, Friberg et al. (2012) increased throughput by screen- ibly high throughput (Neumann et al., 2007). The technique ing 12 fragment cocktails. Since then, more potent inhibitors can characterise fragment binding kinetics and has been used have been developed using FBDD (Pelz et al., 2016). to screen compounds against a variety of protein targets, including PPIs (Pfaff et al., 2015). NMR is a powerful tool for fragment screening and can be tailored to suit the protein target due to the availability of many Generally, SPR works by measuring the change in refract- different techniques. This tool is particularly useful in initial ive index (RI) at a surface where a target is immobilised library screening however, identification of the binding mode of (Fig. 5). There are two distinct ways to prepare the sensor fragments will be limited by the protein target and whether chip and this determines screening properties; either the target peaks have been assigned in the aforementioned 2D spectra. protein or the fragment library can be immobilised to the chip (Chavanieu and Pugnière, 2016). Both methods require small Surface plasmon resonance quantities of protein and a complete screening campaign against a target may require <1 mg of protein (Giannetti, Surface plasmon resonance (SPR) is a spectroscopic technique 2011). This is because the protein can be washed and reused. that, through recent improvements in the technology, has had ............................................................................................... .................................................................. 4 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. Figure 4. Fragment to lead development for Mcl-1. Low-MW fragments were identified that bound to different regions of the same binding pocket of myeloid cell leukaemia 1 (Mcl-1) and were subsequently linked to create a chemical lead with a higher association constant and relatively low MW. Immobilisation of the target protein in microfluidic sys- However, the addition of an affinity tag may cause structural tems is perhaps the most common method as it requires less changes that could interfere with screening and inactive the preparation and can screen 100 s of fragments per day protein. Immobilisation of membrane proteins is particularly (Renaud et al. 2016). But, immobilisation can be the limiting challenging as most require their native lipid environment for step in SPR due to protein inactivation, especially for low sta- activity. There has however been success in this area and a bility proteins (Locatelli-Hoops et al., 2013). A protein can be variety of systems for membrane proteins exist (Patching, immobilised through a covalent or non-covalent linkage. 2014). A steady flow of buffer over the chip allows fragments Immobilisation to the chip is usually through a flexible linker to be introduced and the binding kinetics analysed. Attention molecule such as dextran, and a covalent linkage with NH , must be paid to buffer conditions in non-covalent systems to SH, CHO, OH or COOH functional groups on the target ensure that fragment addition does not cause changes that protein (Chavanieu and Pugnière, 2016). This method poten- may uncouple the protein from the chip. Microfluidic technol- tially requires no modification of the protein prior to immo- ogy allows for the determination of the equilibrium K and bilisation, however the protein orientation can be important. stoichiometry (Bulfer, Jean‐Francois, Arkin, 2016). However, Optimisation to ensure an active protein conformation will the K can only be determined when the solubility of the frag- reduce the quantity of protein needed to generate a spectro- ment is larger than the K , as binding sites need to be satu- scopic signal. Non-covalent linkages are usually formed from rated to obtain a measurable signal. either a His-tagged or biotin-tagged protein, exploiting nickel Microarray-based immobilisation of a fragment library is and streptavidin affinities, respectively (Chavanieu and the most sensitive and highest throughout method, with the Pugnière, 2016). The advantage of this method is that protein ability to screen up to 10 000 molecules in a single day purification can be coupled to linkage to the sensor chip, elim- (Renaud et al. 2016). This method requires the fragment inating the need to remove the purification tag before screen- library to be chemically modified for immobilisation (Singh ing. Kim et al. (2006) have taken this further and developed a et al., 2015), so that multiple fragment orientations are pre- system for on-chip cell culture, protein expression and purifi- sented to the target protein. More protein is also required to cation, and detection of protein–ligand interactions. ............................................................................................... .................................................................. 5 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Figure 5. SPR schematic. Reflected light has a minimum intensity at a critical angle of reflection. A change in mass at the gold surface, due to ligand binding, causes a change in the refractive index. This alters the resonant angle of the intensity minimum and this change in intensity is used to calculate the kinetics of the protein–ligand interaction. detect the binding event, but still only nanomolar quantities the binding mode of the fragment is determined from the crys- (Renaud et al. 2016). This technology is used in combination tal structure. Although this technique has a much lower with a charged-coupled device (CCD) camera and can con- throughput, recent advancements have begun to address this. firm whether a fragment has bound or not (Neumann et al. Crystallography requires the largest quantity of homoge- 2007); binding specificity can then be determined through neous protein (milligrams) out of all the techniques mentioned competition screens. Since the binding of the protein to the so far. This can be a challenge but recent developments in chip is being detected, and not the fragment, this technique is yeast, insect, and even mammalian expression systems have more sensitive as larger changes in signal are produced. made expression of complex eukaryotic proteins easier To discriminate between specific and non-specific binders, (Olivier et al., 2012; Contreras‐Gómez et al., 2013; Khan, competition experiments with known ligands or mutagenesis 2013). Crystallisation presents its own problems and is nor- experiments can be performed. The development of multi- mally the most time-consuming aspect. Depending on the tar- channel/surface sensor chips has allowed for multiple experi- get, proteins must be soluble at concentrations of at least ments to be carried out at once, allowing for false positives 2 mg/ml to form crystals (McPherson and Gavira, 2013). and false negatives to be identified during the screening pro- Membrane proteins are challenging targets due to the pres- cess (Navratilova and Hopkins, 2010). For example, competi- ence of hydrophobic transmembrane regions (TMRs). tion experiments may be carried out on several channels These hydrophobic regions hinder crystal formation but whilst background noise is measured on another. Active site any TMRs are often engineered out to facilitate crystal mutagenesis can also be used to identify false positives but formation. Recent developments in detergents to solubilise this may identify allosteric ligands as false positives, so membrane proteins have made crystallisation easier but the knowledge of any allosteric sites is necessary. crystals are sometimes too small to be useful (Loll, 2014). However, mini beam technology is being developed that SPR is well-suited for initial screening due to its sensitivity, allows protein crystals as small as 10 μmtobeused(Evans throughput, and ability to identify problems with a fragment et al., 2011; Smith, Fischetti, Yamamoto, 2012). A novel library such as non-specific binding. However, expensive approach for difficult targets involves crystallisation within equipment costs (Perkel, 2011) may limit use to large pharma- 3D nanotemplates, where only the surface chemistry of the ceutical companies. nanotemplate is optimised to encourage crystallisation (Shah et al., 2015). This has since been taken further and X-ray crystallography can be coupled with protein purification (Shah et al., X-ray crystallography is a technique that offers the ability for 2017), potentially increasing throughput massively. Target direct validation of any hits during the screening process, as proteins may also be engineered to increase their likelihood ............................................................................................... .................................................................. 6 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. of crystallisation, such as combination with fusion proteins takes time, and generally only one fragment is crystallised in (Derewenda, 2010). complex with the target protein (Hassell et al. 2006). Crystallisation kits exist that may help to identify optimal condi- There are two main techniques employed for crystallisation tions quicker (Molecular Dimensions, 2018). Crystal packing is in fragment screening: crystal soaking and co-crystallisation. In unlikely to interfere with fragment binding as fragments are crystal soaking, protein crystals are transferred to a solution low-MW compounds, but false negatives generated by containing high concentrations of fragment, allowing the frag- obstructed binding sites can only be identified by alterna- ments to diffuse through the crystal solvent channels and bind tive screening techniques. On the other hand, false positives the protein. Co-crystallisation involves crystallisation of the pro- are an issue. Davis and Erlanson (2013) note that the high tein with the fragment. Crystal soaking is the first choice as it is concentrations of fragment may generate interactions higher throughput and does not require optimisation of crystal- unidentifiable by other techniques. These fragments are lisation conditions (Hassell et al. 2006). Crystals can even be likely binding very weakly (Lamoree and Hubbard, 2017) soaked with multiple fragments (up to 10), allowing one to iden- and so this raises the question of the usefulness of such tify the binding mode of multiple fragments (Patel, Bauman, fragments, given that they will need to be developed into Arnold, 2014). Interestingly, it is likely that fragments would lead-like compounds with nanomolar affinities. compete for binding with the target protein and those observed in the crystal structure are likely to form the highest affinity Crystal formation, data collection and analysis are bottle- interactions – whether these are the highest quality interactions necks for this process but automation has since improved would require further investigation. High concentrations of frag- throughput, such as the automation at Diamond Light Source ments are typically required (>10 mM) and this may limit the Synchrotron (2018). Robotics are now able to automate crys- fragment library (Patel, Bauman, Arnold, 2014); this is to ensure tallisation, crystal soaking, crystal selection and orientation saturation of binding sites. But this is also advantageous as this of crystals at the beamline. Moreover, these robots can accur- may identify fragments that bind too weakly to be identified by ately work with smaller quantities of protein, thus reducing other screening methods. If the fragment is not soluble enough, the amount required. A newer generation of detectors, termed co-crystallisation can be used at the cost of lower throughput. pixel array detectors (PADs), are replacing CCDs and these The crystallisation conditions may have to be altered which offer faster data collection (Philipp et al., 2016). Software Figure 6. Development of Hsp90 inhibitors. Red protein surface denotes a negative charge whilst blue denotes a positive charge. Hydrogen bonds are shown by dashed lines. (a) This fragment occupies the binding pocket similarly to the pyrimidine ring of ADP (PDB: 2XDK). (b) Evolution of the fragment to occupy the adjacent hydrophobic pocket increased the activity of the compound (PDB: 2XHR). ............................................................................................... .................................................................. 7 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. packages like AutoSolve can automatically identify bound ligands from an electron density map (Carolan and Lamzin, 2014). Fragment binding is unlikely to change protein con- formation due to their weakly interacting nature and low MW, and so molecular replacement should satisfy for solv- Figure 7. LE metrics. (i) Calculation of LE where N = number of heavy ing the phases. In any case, efforts have been made to auto- atoms. R = molar gas constant; T = temperature in kelvin; K = mate anomalous dispersion at the beamline (de Sanctis equilibrium dissociation constant. (ii) Calculation of lipophilic ligand efficiency (LLE) which accounts for the activity (pIC ) of the et al. 2012). 50 compound and its lipophilicity (c log P). A crystal structure will identify the binding mode of a frag- ment and can help guide fragment evolution. However, no Once true hits have been identified, they are then assessed kinetic information is directly obtained and so it is difficult to for their quality so that a comparison can be made and frag- compare hits by crystallography alone. Molecular docking could ments selected for further progression. Biophysical techniques be performed to estimate some kinetic parameters (Meng et al., are very sensitive and can measure incredibly weak protein– 2011) although it is more typical to use a different biophysical fragment interactions. At the screening stage, fragments are technique such as differential scanning fluorimetry, which is fast often compared by calculating their ligand efficiencies (LE) and low-cost (Scott, Christina, Chris, 2016). (Fig. 7i). This property was introduced to drug discovery as a Crystallography was particularly useful during early screen- way of classifying the quality of contacts made by a fragment ing of inhibitors for Heat shock protein 90 (Hsp90). Hsp90 is (Hopkins, Groom, Alex, 2004). Several groups have found responsible for assisting a series of proteins in folding, notably modified versions of the LE equation, that account for proper- proteins involved in signal transduction (Mahalingam et al., ties like lipophilicity (Fig. 7ii), to be useful (Hopkins et al., 2009). In cancerous cells, upregulated Hsp90 may help tumours 2014), especially during the development of chemical leads survive in more acidic or hypoxic environments, and helps cells where a low lipophilicity correlates with lower attrition tolerate genetic mutations that would otherwise be lethal. (Murray, Verdonk, Rees, 2012). Murray et al. (2010) initially screened a larger library of frag- ments using NMR and identified 125 fragments for crystallo- graphic screening. Only 26 crystal structures were obtained, one The future of FBDD fragment of which made a series of hydrogen bonds in the active FBDD is still evolving and there have been recent develop- site (Fig. 6a), mimicking the way in which the pyrimidine ring of ments that might play a larger role in drug discovery pro- the natural ADP ligand binds. Knowledge of the crystal structure grams of the future. guided growth of the fragment to fill the adjacent lipophilic pocket (Fig. 6b), producing a compound with an improved Covalently-binding fragments have been of interest lately, affinity for Hsp90. Mapping of the binding site was essential for with various groups developing assays to assess fragment evolution of this fragment. affinity. Solvent-exposed cysteine residues on proteins were As crystallography begins to play a larger role in fragment originally targeted, with fragments capable of reversibly screening, it will be interesting to see how data is handled. forming covalent bonds (Erlanson, Wells, Braisted, 2004). Fragment screening produces huge volumes of crystallo- Ostrem et al. (2013) have used this approach to target a graphic data that must be analysed and stored. This will con- mutant of K-Ras (G12C) seen in some tumours and developed tinue to become more of an issue as throughput is increased. inhibitors that target this. These types of fragments are prob- Nevertheless, X-ray crystallography offers fine mapping of lematic though, as it is difficult to make an affinity compari- the fragment binding site and is essential for structure-guided son between them due to the unpredictable nature of evolution of the fragment. electrophile-thiol reactivity. Craven et al. (2018) have since addressed this by developing an assay to analyse the kinetics of covalently-binding fragments, allowing them to ranked by What makes a good hit? their respective affinities. This assay will likely accelerate dis- covery and development of potent covalent inhibitors. Screening techniques can be used to identify hits, but what is Any improvements to the sensitivity of screening techni- often difficult is determining the true nature of a hit. A single ques will reduce the need for high fragment solubility, allow- screening technique will likely produce a series of false posi- ing more chemical space to be explored. Increased cryo-EM tives and/or false negatives. A combination of screening resolution could have a huge impact on drug discovery pro- techniques will allow for cross-validation of any hits, but grams, with the eventual possibility of in-cell screening. The occasionally false negatives may still be missed. However, applications of in-cell NMR are probably limited due to tech- hit correlation between different biophysical techniques can nical challenges. Background noise from non-specific labelling be quite low, most likely because of differences in experi- can be overcome via alternative labelling strategies (Luchinat mental conditions and detection (Jhoti et al., 2013). and Banci, 2016) although delivery of compounds to cells for Therefore, any correlation should be taken as a guide and screening will be difficult. The lipophilicity of fragments has not be absolute. ............................................................................................... .................................................................. 8 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. not been optimised at this stage and the fragments are likely PPI Protein–Protein Interaction to bind with off-target proteins, interfering with the assay. NMR Nuclear Magnetic Resonance Remarkably, groups have been successful in delivering labelled STD Saturation Transfer Difference proteins to mammalian cells and measuring protein–ligand WaterLOGSY Water-Ligand Observed via Gradient interactions. This could therefore be useful during the opti- SpectroscopY misation stage of lead compounds. Development of low-cost NOESY Nuclear Overhauser Effect SpectroscopY SPR spectrophotometers is also being explored (Chen et al. HSQC Heteronuclear Single Quantum Correlation 2015) and this will allow for wider academic participation. HMQC Heteronuclear Multiple Quantum Correlation TROSY Transverse Relaxation Optimized SpectroscopY It is also important to mention the increasing role of docking CPMG Carr-Purcell-Meiboom-Gill Pulse Sequence simulations in FBDD screening. Docking will probably become SPR Surface Plasmon Resonance standard practise in initial fragment screening as it is fast and is not resource intensive. This enables different types of fragment libraries to be screened and an appropriate one selected before References biophysical screening, potentially saving a lot of time. Adams, J. M. and Cory, S. (2007) The Bcl-2 apoptotic switch in cancer development and therapy, Oncogene, 26 (9), 1324–1337. Final remarks Baell, J. B. (2010) Observations on screening-based research and some FBDD has seen many changes since its inception and has been concerning trends in the literature, Future Medicinal Chemistry,2 successful in developing inhibitors against a variety of protein (10), 1529–1546. targets. The ability to screen large chemical space whilst pro- Barker, A., Kettle, J. G., Nowak, T. et al. (2013) Expanding medicinal viding a flexible starting point for the medicinal chemistry is chemistry space, Drug Discovery Today, 18 (5), 298–304. what makes FBDD so attractive. The detection of weakly binding fragments and subsequent development into lead Bulfer, S. L., Jean‐Francois, F. L. and Arkin, M. R. (2016) Making FBDD molecules will continue to be the biggest challenges for work in Academia, in Jahnke W. and Erlanson D. A. (eds), Fragment- FBDD. Additionally, there are still some limitations with frag- based Drug Discovery: Lessons and Outlook, Wiley, Wiley-VCH, pp. ment screening when it comes to large macromolecular com- 223–240. plexes and membrane proteins. Due to the therapeutic Cala, O. and Krimm, I. (2015) Ligand-orientation based fragment selec- importance of these targets, there will likely be developments tion in STD NMR screening, Journal of Medicinal Chemistry, 58 (21), in biophysical techniques to aid screening of these targets. 8739–8742. We are already starting to see FBDD ideas being incorpo- rated into pharmaceutical companies and being used together Carolan, C. G. and Lamzin, V. S. (2014) Automated identification of with HTS (Rüdisser, Vangrevelinghe, Maibaum, 2016). crystallographic ligands using sparse-density representations, Academia are also participating in FBDD as access to bio- Acta Crystallographica Section D: Biological Crystallography,70, physical techniques needed for screening are available at 1844–1853. many institutions. In summary, FBDD is a rapidly evolving Carr, R. A. E., Congreve, M., Murray, C. W. et al. (2005) Fragment-based field and it will undoubtedly play a larger role in future drug lead discovery: leads by design, Drug Discovery Today, 10 (14), development. 987–992. Chavanieu, A. and Pugnière, M. (2016) Developments in SPR fragment Author’s biography screening, Expert Opinion on Drug Discovery, 11 (5), 489–499. Jacob Robson-Tull is a final year student studying BSc Chen, R., Wang, M., Wang, S. et al. (2015) A low cost surface plasmon Biochemistry (Hons) at Imperial College London. Through resonance biosensor using a laser line generator, Optics his studies and lab placements, he has developed an interest in Communications, 349, 83–88. structural biology and its role in identifying therapeutic tar- Constantine, K. L., Davis, M. E., Metzler, W. J. et al. (2006) Protein-ligand gets. Jacob plans to pursue a PhD in structural biology with NOE matching: a high-throughput method for binding pose evalu- the aim of carrying out research that could impact society. ation that does not require protein NMR resonance assignments, Journal of the American Chemical Society, 128 (22), 7252–7263. Funding Contreras‐Gómez, A., Sánchez‐Mirón, A., García‐Camacho, F. et al. (2013) Protein production using the baculovirus‐insect cell expres- The author received no specific funding for this work. sion system, Biotechnology Progress, 30 (1), 1–18. Craven, G. B., Affron, D. P., Allen, C. E. et al. (2018) High-throughput List of Abbreviations kinetic analysis for target-directed covalent ligand discovery, Angewandte Chemie International Edition, 57 (19), 5257–5261. FBDD Fragment-Based Drug Discovery ............................................................................................... .................................................................. 9 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Dalvit, C., Flocco, M. and Marina Veronesi, B. J. (2002) Fluorine-NMR Hann, M. M., Leach, A. R. and Harper, G. (2001) Molecular complexity competition binding experiments for high-throughput screening of and its impact on the probability of finding leads for drug discov- large compound mixtures, Combinatorial Chemistry & High ery, Journal of Chemical Information and Computer Sciences, 41 (3), Throughput Screening, 5 (8), 605–611. 856–864. Dalvit, C., Pevarello, P., Tatò, M. et al. (2000) Identification of com- Harner, M. J., Frank, A. O. and Fesik, S. W. (2013) Fragment-based drug pounds with binding affinity to proteins via magnetization transfer discovery using NMR spectroscopy, Journal of Biomolecular NMR,56 from bulk water*, Journal of Biomolecular NMR, 18 (1), 65–68. (2), 65–75. Davis, B. J. and Erlanson, D. A. (2013) Learning from our mistakes: the Hassell, A. M., An, G., Bledsoe, R. K. et al. (2006) Crystallization of ‘unknown knowns’ in fragment screening, Bioorganic & Medicinal protein-ligand complexes, Acta Crystallographica Section D: Chemistry Letters, 23 (10), 2844–2852. Biological Crystallography, 63, 72–79. de Sanctis, D., Beteva, A., Caserotto, H. et al. (2012) ID29: a high‐inten- Hopkins, A. L., Groom, C. R. and Alex, A. (2004) Ligand efficiency: a sity highly automated ESRF beamline for macromolecular crystal- useful metric for lead selection, Drug Discovery Today,9(10), lography experiments exploiting anomalous scattering, Journal of 430–431. Synchrotron Radiation, 19 (3), 455–461. Hopkins, A. L., Keserü, G. M., Leeson, P. D. et al. (2014) The role of ligand Derewenda, Z. S. (2010) Application of protein engineering to enhance efficiency metrics in drug discovery, Nature Reviews. Drug Discovery, crystallizability and improve crystal properties, Acta Crystallographica 13, 105. Section D: Biological Crystallography, 66, 604–615. Huang, R., Bonnichon, A., Claridge, T. D. W. et al. (2017) Protein–ligand Diamond Light Source. (2018) Fragment Screening - XChem. Diamond. binding affinity determination by the water LOGSY method: an Available from: https://www.diamond.ac.uk/Instruments/Mx/Fragment- optimised approach considering ligand rebinding, Scientific Screening.html [Accessed 12/12/2018]. Reports, 7, 43727. Erlanson, D. A. (2016) Fragments in the clinic: 2016 edition. Practical Hubbard,R.E.(2016) Theroleoffragment‐based discovery in lead Fragments. Available from: http://practicalfragments.blogspot.co.uk/ finding, in Jahnke W. and Erlanson D. A. (eds), Fragment‐based 2016/07/fragments-in-clinic-2016-edition.html [Accessed 09/09/2018]. Drug Discovery Lessons and Outlook,Wiley,Wiley-VCH,pp. 3–29. Erlanson, D. A., Fesik, S. W., Hubbard, R. E. et al. (2016) Twenty years on: the impact of fragments on drug discovery, Nature Reviews. Drug Jahnke, W., Floersheim, P., Ostermeier, C. et al. (2002) NMR reporter Discovery, 15, 605. screening for the detection of high‐affinity ligands, Angewandte Chemie International Edition, 41 (18), 3420–3423. Erlanson, D. A., Wells, J. A. and Braisted, A. C. (2004) Tethering: fragment-based drug discovery, Annual Review of Biophysics and Jhoti, H., Cleasby, A., Verdonk, M. et al. (2007) Fragment-based screen- Biomolecular Structure, 33 (1), 199–223. ing using X-ray crystallography and NMR spectroscopy, Current Opinion in Chemical Biology, 11 (5), 485–493. Evans, G., Axford, D., Waterman, D. et al. (2011) Macromolecular micro- crystallography, Crystallography Reviews, 17 (2), 105–142. Jhoti, H., Williams, G., Rees, D. C. et al. (2013) The ‘rule of three’ for fragment-based drug discovery: where are we now? Nature Friberg, A., Vigil, D., Zhao, B. et al. (2012) Discovery of potent myeloid cell Reviews. Drug Discovery, 12, 644. leukemia 1 (Mcl 1) inhibitors using fragment based methods and structure based design, Journal of Medicinal Chemistry,56 (1), 15–30. Khan, K. H. (2013) Gene expression in mammalian cells and its applica- tions, Advanced Pharmaceutical Bulletin, 3 (2), 257–263. Gee, C. T., Arntson, K. E., Urick, A. K. et al. (2016) Protein-observed 19F- NMR for fragment screening, affinity quantification and druggabil- Kim, M., Lee, S., Choi, H. et al. (2006) On-chip Escherichia coli culture, ity assessment, Nature Protocols, 11, 1414. purification, and detection of expressed proteins, European Biophysics Journal, 35 (8), 655–662. Giannetti, A. M. (2011) Chapter eight - from experimental design to val- idated hits: a comprehensive walk-through of fragment lead identi- Kiraly, P., Adams, R. W., Paudel, L. et al. (2015) Real-time pure shift (15)N fication using surface plasmon resonance, in Kuo L. C. (ed), HSQC of proteins: a real improvement in resolution and sensitivity, Fragment-Based Drug Design: Tools, Practical Approaches, and Journal of Biomolecular NMR, 62 (1), 43–52. Examples, Academic Press Elsevier, pp. 169–218. Lamoree, B. and Hubbard, R. (2017) Current perspectives in Hajduk, P. J., Augeri, D. J., Mack, J. et al. (2000) NMR-based screening of fragment-based lead discovery (FBLD), Essays in Biochemistry, proteins containing 13C-labeled methyl groups, Journal of the 61 (5), 453. American Chemical Society, 122 (33), 7898–7904. Lipinski, C. A., Lombardo, F., Dominy, B. W. et al. (2001) Experimental Hajduk, P. J., Huth, J. R. and Fesik, S. W. (2005) Druggability indices for and computational approaches to estimate solubility and perme- protein targets derived from NMR-based screening data, Journal of ability in drug discovery and development settings, Advanced Drug Medicinal Chemistry, 48 (7), 2518–2525. Delivery Reviews, 46 (1), 3–26. ............................................................................................... .................................................................. 10 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. Locatelli-Hoops, S., Yeliseev, A. A., Gawrisch, K. et al. (2013) Surface plas- Patching, S. G. (2014) Surface plasmon resonance spectroscopy for mon resonance applied to G protein-coupled receptors, Biomedical characterisation of membrane protein–ligand interactions and its Spectroscopy and Imaging, 2 (3), 155–181. potential for drug discovery, Biochimica Et Biophysica Acta (BBA) - Biomembranes, 1838 (1, Part A), 43–55. Loll, P. J. (2014) Membrane proteins, detergents and crystals: what is the state of the art? Acta Crystallographica Section F, Structural Patel, D., Bauman, J. D. and Arnold, E. (2014) Advantages of crystallo- Biology Communications, 70, 1576–1583. graphic fragment screening: functional and mechanistic insights from a powerful platform for efficient drug discovery, Progress in Luchinat, E. and Banci, L. (2016) In-cell NMR: a topical review, IUCRJ,4, Biophysics and Molecular Biology,116 (0), 92–100. 108–118. Pelz, N. F., Bian, Z., Zhao, B. et al. (2016) Discovery of 2-Indole- Ma,R., Wang,P., Wu, J.etal. (2016) Processof fragment-basedleaddiscov- acylsulfonamide Myeloid Cell Leukemia 1 (Mcl-1) inhibitors using ery - a perspective from NMR, Molecules (Basel, Switzerland), 21 (7), E854. fragment-based methods, Journal of Medicinal Chemistry, 59 (5), 2054–2066. Macarron, R., Banks, M. N., Bojanic, D. et al. (2011) Impact of high- throughput screening in biomedical research, Nature Reviews. Drug Perkel, J. M. (2011) Surface Plasmon Resonance (SPR) and Other Discovery, 10, 188. Options for Label-free Detection. Biocompare. Available from: https://www.biocompare.com/Editorial-Articles/41803-Surface-Plasmon- Mahalingam, D., Swords, R., Carew, J. S. et al. (2009) Targeting Resonance-SPR-and-Other-Options-for-Label-free-Detection/ [Accessed HSP90 for cancer therapy, British Journal of Cancer, 100 (10), 09/09/2018]. 1523–1529. Pfaff, S. J., Chimenti, M. S., Kelly, M. J. S. et al. (2015) Biophysical meth- Mayer, M. and Meyer, B. (1999) Characterization of ligand binding by ods for identifying fragment-based inhibitors of protein–protein saturation transfer difference NMR spectroscopy, Angewandte interactions, in Meyerkord C. L. and Fu H. (eds), Protein–Protein Chemie International Edition, 38 (12), 1784–1788. Interactions: Methods and Applications, Springer New York, New McPherson, A. and Gavira, J. A. (2013) Introduction to protein crystal- York, NY, pp. 587–613. lization, Acta Crystallographica Section F, Structural Biology Philipp, H. T., Tate, M. W., Purohit, P. et al. (2016) High-speed X-ray Communications, 70, 2–20. imaging pixel array detector for synchrotron bunch isolation, Meng, X., Zhang, H., Mezei, M. et al. (2011) Molecular docking: a power- Journal of Synchrotron Radiation, 23 (2), 395–403. ful approach for structure-based drug discovery, Current Computer- Rees, D. C., Congreve, M., Murray, C. W. et al. (2004) Fragment-based Aided Drug Design, 7 (2), 146–157. lead discovery, Nature Reviews. Drug Discovery, 3, 660. Molecular Dimensions. (2018) Crystallisation Screens. Molecular Renaud, J., Chung, C., Danielson, U. H. et al. (2016) Biophysics in drug Dimensions. Available from: https://www.moleculardimensions. discovery: impact, challenges and opportunities, Nature Reviews. com/products/c255-Crystallization-Screens/ [Accessed 09/09/2018]. Drug Discovery, 15, 679. Murray, C. W., Carr, M. G., Callaghan, O. et al. (2010) Fragment-based drug Rüdisser, S., Vangrevelinghe, E. and Maibaum, J. (2016) An integrated discovery applied to Hsp90. Discovery of two lead series with high lig- approach for fragment‐based lead discovery: virtual, NMR, and and efficiency, Journal of Medicinal Chemistry, 53 (16), 5942–5955. high‐throughput screening combined with structure‐guided Murray, C. W. and Rees, D. C. (2009) The rise of fragment-based drug design. Application to the aspartyl protease renin, in Jahnke W. and discovery, Nature Chemistry, 1, 187. Erlanson D. A. (eds), Fragment‐based Drug Discovery: Lessons and Outlook, Wiley, Wiley-VCH, pp. 447–480. Murray, C. W., Verdonk, M. L. and Rees, D. C. (2012) Experiences in fragment-based drug discovery, Trends in Pharmacological Sciences, Scott, D. E., Christina, S. and Chris, A. (2016) Differential scanning fluori- 33 (5), 224–232. metry as part of a biophysical screening cascade, in Jahnke W. and Erlanson D. A. (eds), Fragment-based Drug Discovery: Lessons and Navratilova, I. and Hopkins, A. L. (2010) Fragment screening by surface Outlook, Wiley, Wiley-VCH, pp. 139–170. plasmon resonance, ACS Medicinal Chemistry Letters, 1 (1), 44–48. Segall, M. D. and Barber, C. (2014) Addressing toxicity risk when design- Neumann, T., Junker, H.-D., Schmidt, K. et al. (2007) SPR-based frag- ing and selecting compounds in early drug discovery, Drug ment screening: advantages and applications, Current Topics in Discovery Today, 19 (5), 688–693. Medicinal Chemistry, 7 (16), 1630–1642. Shah, U. V., Jahn, N. H., Huang, S. et al. (2017) Crystallisation via novel Olivier, B., Fatima, A., Christel, L. et al. (2012) Overexpression of mem- 3D nanotemplates as a tool for protein purification and bio- brane proteins using Pichia pastoris, Current Protocols in Protein separation, Journal of Crystal Growth, 469, 42–47. Science, 29 (1), 29.2.1–29.2.24. Shah, U. V., Parambil, J. V., Williams, D. R. et al. (2015) Preparation and Ostrem, J. M., Peters, U., Sos, M. L. et al. (2013) K-Ras(G12C) inhibitors characterisation of 3D nanotemplates for protein crystallisation, allosterically control GTP affinity and effector interactions, Nature, Powder Technology, 282, 10–18. 503 (7477), 548–551. ............................................................................................... .................................................................. 11 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Shuker, S. B., Hajduk, P. J., Meadows, R. P. et al. (1996) Discovering high- Smith, J. L., Fischetti, R. F. and Yamamoto, M. (2012) Micro- affinity ligands for proteins: SAR by NMR, Science (New York, N.Y.), crystallography comes of age, Current Opinion in Structural Biology, 274 (5292), 1531. 22 (5), 602–612. Singh, V., Singh, K., Nand, A. et al. (2015) Small molecule microarray screen- Wu, B., Zhang, Z., Noberini, R. et al. (2013) HTS by NMR of combinatorial ing methodology based on surface plasmon resonance imaging, libraries: a fragment-based approach to ligand discovery, Chemistry Arabian Journal of Chemistry. https://doi.org/10.1016/j.arabjc.2014.12.020 & Biology, 20 (1), 19–33. ............................................................................................... .................................................................. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BioScience Horizons Oxford University Press

Biophysical screening in fragment-based drug design: a brief overview

BioScience Horizons , Volume 11 – Jan 1, 2018

Loading next page...
 
/lp/oxford-university-press/biophysical-screening-in-fragment-based-drug-design-a-brief-overview-O91l9wRmUj
Publisher
Oxford University Press
Copyright
Copyright © 2022 Oxford University Press
eISSN
1754-7431
DOI
10.1093/biohorizons/hzy015
Publisher site
See Article on Publisher Site

Abstract

Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 BioscienceHorizons Volume 11 2018 10.1093/biohorizons/hzy015 ............................................................................................ ..................................................................... Review article Biophysical screening in fragment-based drug design: a brief overview Jacob Robson-Tull Department of Life Sciences, Imperial College London, London SW7 2AZ, UK *Corresponding author: Jacob Robson-Tull, Department of Life Sciences, Imperial College London, London SW7 2AZ, UK. Email: jacob.robson-tull16@imperial.ac.uk Supervisor: Dr Ernesto Cota, Department of Life Sciences, Imperial College London, London SW7 2AZ, UK ............................................................................................ ..................................................................... High-throughput screening (HTS) has been firmly rooted at the heart of many drug discovery programs over the past few decades, having provided a starting point for the development of many of the drugs on the market. However, this technique is often accompanied by high-attrition rates and has proven to be unsuccessful at developing therapeutics against more non-conventional targets such as protein–protein interactions (PPIs). Fragment-based drug discovery (FBDD) offers an alter- native approach and is steadily being taken up by the industry to tackle the high-attrition rate associated with many chem- ical leads developed by HTS. FBDD takes a structure-guided approach and uses a small chemical library of fragments during the initial screening process, to identify weakly binding ligands with the potential for therapeutic development. This review aims to summarise the challenges with screening in FBDD and the key biophysical screening techniques used for identifica- tion of weakly interacting ligands. Key words: fragment-based drug discovery, biophysical, protein–ligand interactions, fragment screening, fragment-based lead discovery, ligand screening Submitted on 27 September 2018; editorial decision on 27 November 2008 ............................................................................................ ..................................................................... 30% of attrition during clinical trials (Segall and Barber, Introduction 2014), some of which is likely due to off-target interactions. Different approaches have since been developed that take a Drug discovery programs have seen huge changes over the structure-guided approach to drug design, to target specific past 20 years, with the advent of new technologies constantly proteins and reduce toxicity. changing the approach to drug discovery and design. High- throughput screening (HTS) has been the technique of choice Fragment-based drug design (FBDD) is a recently devel- for large pharmaceutical companies due to its ability to screen oped workflow that addresses some of the shortfalls of other large numbers of molecules to discover chemical leads techniques and has been successful in developing numerous (Macarron et al. 2011). The development of higher quality clinical-stage drugs, as shown in Fig. 1. FBDD typically chemical libraries has enabled HTS to continue to identify involves screening chemical libraries composed of up to a few hits for conventional drug targets such as protein kinases thousand small molecules, as opposed to millions in HTS (Barker et al., 2013). However, HTS has proven unsuccessful libraries, through a variety of biophysical techniques such as for screening compounds against more difficult targets such NMR spectroscopy (Erlanson et al., 2016). These small mol- as protein–protein interactions (PPIs; Hubbard, 2016), which ecule libraries, termed fragment libraries, are attractive as are of increasing therapeutic importance. What is more con- they address some of the current causes of attrition of lead cerning is that recent estimates place toxicity as a cause of compounds. Fragments are low-MW molecules (<300 kDa) ............................................................................................... .................................................................. © The Author(s) 2019. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. and form weaker interactions with protein targets, normally challenges that must be overcome due to the weakly interact- in the mM–μM range (Murray and Rees, 2009). Although the ing nature of fragments. This review will summarise some of interactions are weaker than those observed in HTS, the inter- the challenges of fragment screening and recent developments actions tend to be higher quality, with each non-hydrogen in biophysical techniques that have helped overcome them. atom (heavy atom) involved in the binding interaction. A low-MW and high-interaction quality give more freedom in Fragment screening the fine-tuning of the physiochemical properties later in the drug development process (Hubbard, 2016), making them HTS typically involves the use of a robust, cell activity-based attractive starting points for lead compounds. Fragment assay such as the luciferase reporter screen (Macarron et al. libraries can also sample a large range of chemical space due 2011). Since fragments are generally smaller and simpler than to their inherent size (Hann, Leach, Harper, 2001), which compounds used in HTS, cell activity assays are difficult to allows them to better fit binding sites and pockets. Together, use as the binding event is usually too weak to cause a detect- these properties contribute to the higher hit rate observed for able change (Murray and Rees, 2009). Therefore, more sensi- fragment libraries than that of HTS libraries. tive biophysical techniques must be employed to detect these interactions, but these present their own challenges. The FBDD is now commonplace in drug discovery programs throughput of these techniques is generally lower and they and has even enabled academic participation due to smaller often require larger quantities of protein. The fragments also chemical libraries and availability of biophysical screening need to be highly pure, and soluble at high concentrations to instruments (Rees et al., 2004). Although the quality of hits in detect their binding. However, this can be an advantage as FBDD is generally higher than HTS, there are certain approved drugs tend to have a low lipophilicity (Lipinski et al., 2001). Nevertheless, biophysical techniques can offer insight into the binding mode of fragments, enabling fragment hits to be developed into lead-like molecules in a structure- guided approach, as summarised in Fig. 2. Several key techni- ques will be focused on, including NMR spectroscopy, sur- face plasmon resonance and X-ray crystallography. It should be noted that there are other techniques used and in develop- ment, although these are less commonly practised at present. NMR spectroscopy FBDD was first made possible 20 years ago by the develop- ment of NMR spectroscopy as a tool to screen fragments by Shuker et al. (1996), in what was called a structure-activity relationship by NMR. Since then, numerous NMR-based techniques have been developed. NMR spectroscopy is well suited to studying protein-fragment interactions as it is pos- Figure 1. Fragment-derived drugs at various stages of clinical trial as sible to determine the binding affinity and binding mode of of 2016. Data from Erlanson (2016). fragments, allowing for dissociation constants (K ) in the Figure 2. Summary of an FBBD program. (a) Selection of an appropriate target compatible with the chosen biophysical screening technique. (b) Target protein is produced and purified. (c) The fragment library designed for the target protein. (d) Biophysical screening of the fragment library. (e) Validation of hits from screening to identify the fragment binding mode. (f) Development of the fragment(s) into a lead molecule. ............................................................................................... .................................................................. 2 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. mM range to be identified (Carr et al., 2005). Additionally, targets. To the contrary, WaterLOGSY can be used to calcu- NMR spectroscopy observes the direct binding event between late dissociation constants through ligand titration, although the protein and fragment, reducing the chance of false posi- special care must be taken to ensure that the dissociation con- tives from assay artefacts as seen in HTS (Wu et al. 2013). stant measured is not dependent on experimental conditions NMR spectroscopy can be used in two different approaches, (Huang et al., 2017). Other ligand-observed techniques exist, either ligand-observed or protein-observed NMR, and these such as CPMG and relaxation experiments (Jhoti et al., are sometimes used in combination to optimise throughput. 2007), but these are not as commonly used as STD and waterLOGSY. Before fragment screening can take place, the fragment Protein-observed NMR is a more laborious process, but it libraries must be optimised as NMR is particularly sensitive is the gold-standard in NMR analysis of protein–ligand inter- to false positives from certain types of fragments. These types actions, allowing for determination of the binding affinity and of interfering fragments often form non-specific interactions binding mode of fragments. It can be used to follow up on with proteins, contain reactive covalent modifiers, are chela- hits from an initial screen by ligand-observed NMR. This tors or are aggregators, and so libraries must be designed to technique requires large quantities of protein (30 nmol) and exclude these (Harner, Frank, Fesik, 2013). Unfortunately, the protein must be isotopically labelled (Renaud et al. 2016). these types of fragments can be difficult to identify due to 1 15 For proteins up to ~40 kDa, a series of H– N HSQC spectra unexpected reactivity with proteins (Baell, 2010), thus caus- can be recorded whilst titrating fragment. This allows the ing false positives. Since high fragment concentrations are binding affinity to be determined and if the spectrum has pre- required for screening, to saturate low-affinity binding sites, viously been assigned, the binding site can be approximated. these effects are likely to become more pronounced in NMR The sensitivity of this technique is increased by using higher screening. A 1D spectra of the fragments in solution may be magnetic field strengths (Kiraly et al. 2015) to identify more measured at 1 mM (Ma et al., 2016) and from this, an estima- 1 15 distinct peaks. For proteins whose H– N HSQC spectrum tion of aqueous solubility can be obtained by comparing the has not been assigned, FBDD campaigns can still occur, dem- spectra to a reference. However, it is not known how the frag- onstrating the power of FBDD. The binding affinity can still ments will behave until they are screened. Even with optimisa- be determined by fragment titration. A technique termed tion, Baell (2010) highlights that some false positives will still ‘NOE matching’ can model the binding mode of ligands by not be identified in early screens. analysing 3D C NOESY spectra and comparing it to a com- Ligand-observed NMR is suitable for screening as it mea- putational model (Constantine et al., 2006). However, this sures the change in intensity, sign or relaxation of fragment technique is not popular at present, perhaps because it protons in the presence of the target protein (Ma et al. 2016). requires several protein–ligand NOEs to be observed and This requires no isotopic labelling of the protein and smaller assigned; this is difficult for fragments as they are small and quantities, thus increasing throughput when compared to interact weakly with target proteins and so its applications to protein-observed NMR. There is no theoretical protein size FBDD may be limited. For proteins larger than 40 kDa, 1 15 1 13 limitation but ligand-observed NMR tends to be used for pro- H– N TROSY HSQC or H– C HSQC can be recorded teins between 15 and 100 kDa (Harner, Frank, Fesik, 2013). for specific isotopically labelled residues such as valine, leu- Saturation transfer difference (STD; Mayer and Meyer, 1999) cine and isoleucine (Hajduk et al., 2000). Binding sites typic- and waterLOGSY (Dalvit et al., 2000) are used most often in ally contain hydrophobic residues, especially in pockets, and FBDD screening and these rely on nuclear Overhauser effect so these residues can be used to report on ligand binding. (NOE) transfer of magnetisation, from the protein to the frag- Although large quantities of protein are required, the protein ment, as shown in Fig. 3a and b. These two techniques are can be recycled for different experiments, with yields of up to used in combination to cross-validate any hits as many false 80% (Harner, Frank, Fesik, 2013). A series of other protein- positives can occur, particularly with STD where fragment observed experiments exist, such as F amino acid labelling methyl protons may be saturated. (Gee et al. 2016). Unfortunately, STD does not give information on the bind- Inhibitors developed against myeloid cell leukaemia 1 ing mode and cannot directly distinguish between specific and (Mcl-1) protein are just one example of where NMR has had non-specific binding interactions, so follow up on hits is great success in screening fragment libraries (Friberg et al. always required. A reporter screen (Jahnke et al. 2002) can 2012). Mcl-1 dysregulation prevents cancerous cells from make use of known ligands to identify binding specificity undergoing apoptosis and has been a promising cancer target (Fig. 3c), but this will lead to a decreased throughput. for many years (Adams and Cory, 2007). However, Mcl-1 is Reporter screens can also enable detection of higher affinity difficult to target as it exerts its effects through PPIs and as of interactions and F atoms can be incorporated to increase 2012, there were no inhibitors in clinical trials. Friberg et al. sensitivity and throughput (Dalvit, Flocco, Marina Veronesi, (2012) overcame this challenge using FBDD. Fragments were 1 15 2002). Alternatively, quantitative STD analysis can be used to screened via SOFAST H– N HMQC, which is similar to a identify binding specificity and does not require a known lig- HSQC-type experiment, and a series of fragments, shown in and of the target protein (Cala and Krimm, 2015). This is Fig. 4, were identified for progression. Since the hit rate for more attractive as it can be applied to a wider range of protein proteins is typically below 1% (Hajduk, Huth, Fesik, 2005), ............................................................................................... .................................................................. 3 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Figure 3. NOE-based screening techniques. (a) In STD, the protein methyl groups are saturated and eventual cross-saturation of the bound ligand causes a signal intensity decrease. RF is the radio frequency pulse applied to the sample. The H spectra before and after protein addition are used to calculate a difference spectrum and identify if a fragment has bound. (b) In waterLOGSY, fragments that bind the target protein are often identified by positive resonances and those that do not have negative resonances. Bound ligands in waterLOGSY are therefore identified by peak inversions. (c) Ligand-observed reporter screening. (ci) The known ligand has sharp peaks in a H spectrum when isolated. (cii) Upon addition of protein, the peaks are broadened due to reduced tumbling when the ligand is bound to the protein. (ciii) Sharp peaks of the known ligand are restored if a fragment competes for binding. it is unlikely for 2 hits to occur in the same fragment cocktail. success in fragment screening due to its sensitivity and incred- Hence, Friberg et al. (2012) increased throughput by screen- ibly high throughput (Neumann et al., 2007). The technique ing 12 fragment cocktails. Since then, more potent inhibitors can characterise fragment binding kinetics and has been used have been developed using FBDD (Pelz et al., 2016). to screen compounds against a variety of protein targets, including PPIs (Pfaff et al., 2015). NMR is a powerful tool for fragment screening and can be tailored to suit the protein target due to the availability of many Generally, SPR works by measuring the change in refract- different techniques. This tool is particularly useful in initial ive index (RI) at a surface where a target is immobilised library screening however, identification of the binding mode of (Fig. 5). There are two distinct ways to prepare the sensor fragments will be limited by the protein target and whether chip and this determines screening properties; either the target peaks have been assigned in the aforementioned 2D spectra. protein or the fragment library can be immobilised to the chip (Chavanieu and Pugnière, 2016). Both methods require small Surface plasmon resonance quantities of protein and a complete screening campaign against a target may require <1 mg of protein (Giannetti, Surface plasmon resonance (SPR) is a spectroscopic technique 2011). This is because the protein can be washed and reused. that, through recent improvements in the technology, has had ............................................................................................... .................................................................. 4 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. Figure 4. Fragment to lead development for Mcl-1. Low-MW fragments were identified that bound to different regions of the same binding pocket of myeloid cell leukaemia 1 (Mcl-1) and were subsequently linked to create a chemical lead with a higher association constant and relatively low MW. Immobilisation of the target protein in microfluidic sys- However, the addition of an affinity tag may cause structural tems is perhaps the most common method as it requires less changes that could interfere with screening and inactive the preparation and can screen 100 s of fragments per day protein. Immobilisation of membrane proteins is particularly (Renaud et al. 2016). But, immobilisation can be the limiting challenging as most require their native lipid environment for step in SPR due to protein inactivation, especially for low sta- activity. There has however been success in this area and a bility proteins (Locatelli-Hoops et al., 2013). A protein can be variety of systems for membrane proteins exist (Patching, immobilised through a covalent or non-covalent linkage. 2014). A steady flow of buffer over the chip allows fragments Immobilisation to the chip is usually through a flexible linker to be introduced and the binding kinetics analysed. Attention molecule such as dextran, and a covalent linkage with NH , must be paid to buffer conditions in non-covalent systems to SH, CHO, OH or COOH functional groups on the target ensure that fragment addition does not cause changes that protein (Chavanieu and Pugnière, 2016). This method poten- may uncouple the protein from the chip. Microfluidic technol- tially requires no modification of the protein prior to immo- ogy allows for the determination of the equilibrium K and bilisation, however the protein orientation can be important. stoichiometry (Bulfer, Jean‐Francois, Arkin, 2016). However, Optimisation to ensure an active protein conformation will the K can only be determined when the solubility of the frag- reduce the quantity of protein needed to generate a spectro- ment is larger than the K , as binding sites need to be satu- scopic signal. Non-covalent linkages are usually formed from rated to obtain a measurable signal. either a His-tagged or biotin-tagged protein, exploiting nickel Microarray-based immobilisation of a fragment library is and streptavidin affinities, respectively (Chavanieu and the most sensitive and highest throughout method, with the Pugnière, 2016). The advantage of this method is that protein ability to screen up to 10 000 molecules in a single day purification can be coupled to linkage to the sensor chip, elim- (Renaud et al. 2016). This method requires the fragment inating the need to remove the purification tag before screen- library to be chemically modified for immobilisation (Singh ing. Kim et al. (2006) have taken this further and developed a et al., 2015), so that multiple fragment orientations are pre- system for on-chip cell culture, protein expression and purifi- sented to the target protein. More protein is also required to cation, and detection of protein–ligand interactions. ............................................................................................... .................................................................. 5 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Figure 5. SPR schematic. Reflected light has a minimum intensity at a critical angle of reflection. A change in mass at the gold surface, due to ligand binding, causes a change in the refractive index. This alters the resonant angle of the intensity minimum and this change in intensity is used to calculate the kinetics of the protein–ligand interaction. detect the binding event, but still only nanomolar quantities the binding mode of the fragment is determined from the crys- (Renaud et al. 2016). This technology is used in combination tal structure. Although this technique has a much lower with a charged-coupled device (CCD) camera and can con- throughput, recent advancements have begun to address this. firm whether a fragment has bound or not (Neumann et al. Crystallography requires the largest quantity of homoge- 2007); binding specificity can then be determined through neous protein (milligrams) out of all the techniques mentioned competition screens. Since the binding of the protein to the so far. This can be a challenge but recent developments in chip is being detected, and not the fragment, this technique is yeast, insect, and even mammalian expression systems have more sensitive as larger changes in signal are produced. made expression of complex eukaryotic proteins easier To discriminate between specific and non-specific binders, (Olivier et al., 2012; Contreras‐Gómez et al., 2013; Khan, competition experiments with known ligands or mutagenesis 2013). Crystallisation presents its own problems and is nor- experiments can be performed. The development of multi- mally the most time-consuming aspect. Depending on the tar- channel/surface sensor chips has allowed for multiple experi- get, proteins must be soluble at concentrations of at least ments to be carried out at once, allowing for false positives 2 mg/ml to form crystals (McPherson and Gavira, 2013). and false negatives to be identified during the screening pro- Membrane proteins are challenging targets due to the pres- cess (Navratilova and Hopkins, 2010). For example, competi- ence of hydrophobic transmembrane regions (TMRs). tion experiments may be carried out on several channels These hydrophobic regions hinder crystal formation but whilst background noise is measured on another. Active site any TMRs are often engineered out to facilitate crystal mutagenesis can also be used to identify false positives but formation. Recent developments in detergents to solubilise this may identify allosteric ligands as false positives, so membrane proteins have made crystallisation easier but the knowledge of any allosteric sites is necessary. crystals are sometimes too small to be useful (Loll, 2014). However, mini beam technology is being developed that SPR is well-suited for initial screening due to its sensitivity, allows protein crystals as small as 10 μmtobeused(Evans throughput, and ability to identify problems with a fragment et al., 2011; Smith, Fischetti, Yamamoto, 2012). A novel library such as non-specific binding. However, expensive approach for difficult targets involves crystallisation within equipment costs (Perkel, 2011) may limit use to large pharma- 3D nanotemplates, where only the surface chemistry of the ceutical companies. nanotemplate is optimised to encourage crystallisation (Shah et al., 2015). This has since been taken further and X-ray crystallography can be coupled with protein purification (Shah et al., X-ray crystallography is a technique that offers the ability for 2017), potentially increasing throughput massively. Target direct validation of any hits during the screening process, as proteins may also be engineered to increase their likelihood ............................................................................................... .................................................................. 6 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. of crystallisation, such as combination with fusion proteins takes time, and generally only one fragment is crystallised in (Derewenda, 2010). complex with the target protein (Hassell et al. 2006). Crystallisation kits exist that may help to identify optimal condi- There are two main techniques employed for crystallisation tions quicker (Molecular Dimensions, 2018). Crystal packing is in fragment screening: crystal soaking and co-crystallisation. In unlikely to interfere with fragment binding as fragments are crystal soaking, protein crystals are transferred to a solution low-MW compounds, but false negatives generated by containing high concentrations of fragment, allowing the frag- obstructed binding sites can only be identified by alterna- ments to diffuse through the crystal solvent channels and bind tive screening techniques. On the other hand, false positives the protein. Co-crystallisation involves crystallisation of the pro- are an issue. Davis and Erlanson (2013) note that the high tein with the fragment. Crystal soaking is the first choice as it is concentrations of fragment may generate interactions higher throughput and does not require optimisation of crystal- unidentifiable by other techniques. These fragments are lisation conditions (Hassell et al. 2006). Crystals can even be likely binding very weakly (Lamoree and Hubbard, 2017) soaked with multiple fragments (up to 10), allowing one to iden- and so this raises the question of the usefulness of such tify the binding mode of multiple fragments (Patel, Bauman, fragments, given that they will need to be developed into Arnold, 2014). Interestingly, it is likely that fragments would lead-like compounds with nanomolar affinities. compete for binding with the target protein and those observed in the crystal structure are likely to form the highest affinity Crystal formation, data collection and analysis are bottle- interactions – whether these are the highest quality interactions necks for this process but automation has since improved would require further investigation. High concentrations of frag- throughput, such as the automation at Diamond Light Source ments are typically required (>10 mM) and this may limit the Synchrotron (2018). Robotics are now able to automate crys- fragment library (Patel, Bauman, Arnold, 2014); this is to ensure tallisation, crystal soaking, crystal selection and orientation saturation of binding sites. But this is also advantageous as this of crystals at the beamline. Moreover, these robots can accur- may identify fragments that bind too weakly to be identified by ately work with smaller quantities of protein, thus reducing other screening methods. If the fragment is not soluble enough, the amount required. A newer generation of detectors, termed co-crystallisation can be used at the cost of lower throughput. pixel array detectors (PADs), are replacing CCDs and these The crystallisation conditions may have to be altered which offer faster data collection (Philipp et al., 2016). Software Figure 6. Development of Hsp90 inhibitors. Red protein surface denotes a negative charge whilst blue denotes a positive charge. Hydrogen bonds are shown by dashed lines. (a) This fragment occupies the binding pocket similarly to the pyrimidine ring of ADP (PDB: 2XDK). (b) Evolution of the fragment to occupy the adjacent hydrophobic pocket increased the activity of the compound (PDB: 2XHR). ............................................................................................... .................................................................. 7 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. packages like AutoSolve can automatically identify bound ligands from an electron density map (Carolan and Lamzin, 2014). Fragment binding is unlikely to change protein con- formation due to their weakly interacting nature and low MW, and so molecular replacement should satisfy for solv- Figure 7. LE metrics. (i) Calculation of LE where N = number of heavy ing the phases. In any case, efforts have been made to auto- atoms. R = molar gas constant; T = temperature in kelvin; K = mate anomalous dispersion at the beamline (de Sanctis equilibrium dissociation constant. (ii) Calculation of lipophilic ligand efficiency (LLE) which accounts for the activity (pIC ) of the et al. 2012). 50 compound and its lipophilicity (c log P). A crystal structure will identify the binding mode of a frag- ment and can help guide fragment evolution. However, no Once true hits have been identified, they are then assessed kinetic information is directly obtained and so it is difficult to for their quality so that a comparison can be made and frag- compare hits by crystallography alone. Molecular docking could ments selected for further progression. Biophysical techniques be performed to estimate some kinetic parameters (Meng et al., are very sensitive and can measure incredibly weak protein– 2011) although it is more typical to use a different biophysical fragment interactions. At the screening stage, fragments are technique such as differential scanning fluorimetry, which is fast often compared by calculating their ligand efficiencies (LE) and low-cost (Scott, Christina, Chris, 2016). (Fig. 7i). This property was introduced to drug discovery as a Crystallography was particularly useful during early screen- way of classifying the quality of contacts made by a fragment ing of inhibitors for Heat shock protein 90 (Hsp90). Hsp90 is (Hopkins, Groom, Alex, 2004). Several groups have found responsible for assisting a series of proteins in folding, notably modified versions of the LE equation, that account for proper- proteins involved in signal transduction (Mahalingam et al., ties like lipophilicity (Fig. 7ii), to be useful (Hopkins et al., 2009). In cancerous cells, upregulated Hsp90 may help tumours 2014), especially during the development of chemical leads survive in more acidic or hypoxic environments, and helps cells where a low lipophilicity correlates with lower attrition tolerate genetic mutations that would otherwise be lethal. (Murray, Verdonk, Rees, 2012). Murray et al. (2010) initially screened a larger library of frag- ments using NMR and identified 125 fragments for crystallo- graphic screening. Only 26 crystal structures were obtained, one The future of FBDD fragment of which made a series of hydrogen bonds in the active FBDD is still evolving and there have been recent develop- site (Fig. 6a), mimicking the way in which the pyrimidine ring of ments that might play a larger role in drug discovery pro- the natural ADP ligand binds. Knowledge of the crystal structure grams of the future. guided growth of the fragment to fill the adjacent lipophilic pocket (Fig. 6b), producing a compound with an improved Covalently-binding fragments have been of interest lately, affinity for Hsp90. Mapping of the binding site was essential for with various groups developing assays to assess fragment evolution of this fragment. affinity. Solvent-exposed cysteine residues on proteins were As crystallography begins to play a larger role in fragment originally targeted, with fragments capable of reversibly screening, it will be interesting to see how data is handled. forming covalent bonds (Erlanson, Wells, Braisted, 2004). Fragment screening produces huge volumes of crystallo- Ostrem et al. (2013) have used this approach to target a graphic data that must be analysed and stored. This will con- mutant of K-Ras (G12C) seen in some tumours and developed tinue to become more of an issue as throughput is increased. inhibitors that target this. These types of fragments are prob- Nevertheless, X-ray crystallography offers fine mapping of lematic though, as it is difficult to make an affinity compari- the fragment binding site and is essential for structure-guided son between them due to the unpredictable nature of evolution of the fragment. electrophile-thiol reactivity. Craven et al. (2018) have since addressed this by developing an assay to analyse the kinetics of covalently-binding fragments, allowing them to ranked by What makes a good hit? their respective affinities. This assay will likely accelerate dis- covery and development of potent covalent inhibitors. Screening techniques can be used to identify hits, but what is Any improvements to the sensitivity of screening techni- often difficult is determining the true nature of a hit. A single ques will reduce the need for high fragment solubility, allow- screening technique will likely produce a series of false posi- ing more chemical space to be explored. Increased cryo-EM tives and/or false negatives. A combination of screening resolution could have a huge impact on drug discovery pro- techniques will allow for cross-validation of any hits, but grams, with the eventual possibility of in-cell screening. The occasionally false negatives may still be missed. However, applications of in-cell NMR are probably limited due to tech- hit correlation between different biophysical techniques can nical challenges. Background noise from non-specific labelling be quite low, most likely because of differences in experi- can be overcome via alternative labelling strategies (Luchinat mental conditions and detection (Jhoti et al., 2013). and Banci, 2016) although delivery of compounds to cells for Therefore, any correlation should be taken as a guide and screening will be difficult. The lipophilicity of fragments has not be absolute. ............................................................................................... .................................................................. 8 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. not been optimised at this stage and the fragments are likely PPI Protein–Protein Interaction to bind with off-target proteins, interfering with the assay. NMR Nuclear Magnetic Resonance Remarkably, groups have been successful in delivering labelled STD Saturation Transfer Difference proteins to mammalian cells and measuring protein–ligand WaterLOGSY Water-Ligand Observed via Gradient interactions. This could therefore be useful during the opti- SpectroscopY misation stage of lead compounds. Development of low-cost NOESY Nuclear Overhauser Effect SpectroscopY SPR spectrophotometers is also being explored (Chen et al. HSQC Heteronuclear Single Quantum Correlation 2015) and this will allow for wider academic participation. HMQC Heteronuclear Multiple Quantum Correlation TROSY Transverse Relaxation Optimized SpectroscopY It is also important to mention the increasing role of docking CPMG Carr-Purcell-Meiboom-Gill Pulse Sequence simulations in FBDD screening. Docking will probably become SPR Surface Plasmon Resonance standard practise in initial fragment screening as it is fast and is not resource intensive. This enables different types of fragment libraries to be screened and an appropriate one selected before References biophysical screening, potentially saving a lot of time. Adams, J. M. and Cory, S. (2007) The Bcl-2 apoptotic switch in cancer development and therapy, Oncogene, 26 (9), 1324–1337. Final remarks Baell, J. B. (2010) Observations on screening-based research and some FBDD has seen many changes since its inception and has been concerning trends in the literature, Future Medicinal Chemistry,2 successful in developing inhibitors against a variety of protein (10), 1529–1546. targets. The ability to screen large chemical space whilst pro- Barker, A., Kettle, J. G., Nowak, T. et al. (2013) Expanding medicinal viding a flexible starting point for the medicinal chemistry is chemistry space, Drug Discovery Today, 18 (5), 298–304. what makes FBDD so attractive. The detection of weakly binding fragments and subsequent development into lead Bulfer, S. L., Jean‐Francois, F. L. and Arkin, M. R. (2016) Making FBDD molecules will continue to be the biggest challenges for work in Academia, in Jahnke W. and Erlanson D. A. (eds), Fragment- FBDD. Additionally, there are still some limitations with frag- based Drug Discovery: Lessons and Outlook, Wiley, Wiley-VCH, pp. ment screening when it comes to large macromolecular com- 223–240. plexes and membrane proteins. Due to the therapeutic Cala, O. and Krimm, I. (2015) Ligand-orientation based fragment selec- importance of these targets, there will likely be developments tion in STD NMR screening, Journal of Medicinal Chemistry, 58 (21), in biophysical techniques to aid screening of these targets. 8739–8742. We are already starting to see FBDD ideas being incorpo- rated into pharmaceutical companies and being used together Carolan, C. G. and Lamzin, V. S. (2014) Automated identification of with HTS (Rüdisser, Vangrevelinghe, Maibaum, 2016). crystallographic ligands using sparse-density representations, Academia are also participating in FBDD as access to bio- Acta Crystallographica Section D: Biological Crystallography,70, physical techniques needed for screening are available at 1844–1853. many institutions. In summary, FBDD is a rapidly evolving Carr, R. A. E., Congreve, M., Murray, C. W. et al. (2005) Fragment-based field and it will undoubtedly play a larger role in future drug lead discovery: leads by design, Drug Discovery Today, 10 (14), development. 987–992. Chavanieu, A. and Pugnière, M. (2016) Developments in SPR fragment Author’s biography screening, Expert Opinion on Drug Discovery, 11 (5), 489–499. Jacob Robson-Tull is a final year student studying BSc Chen, R., Wang, M., Wang, S. et al. (2015) A low cost surface plasmon Biochemistry (Hons) at Imperial College London. Through resonance biosensor using a laser line generator, Optics his studies and lab placements, he has developed an interest in Communications, 349, 83–88. structural biology and its role in identifying therapeutic tar- Constantine, K. L., Davis, M. E., Metzler, W. J. et al. (2006) Protein-ligand gets. Jacob plans to pursue a PhD in structural biology with NOE matching: a high-throughput method for binding pose evalu- the aim of carrying out research that could impact society. ation that does not require protein NMR resonance assignments, Journal of the American Chemical Society, 128 (22), 7252–7263. Funding Contreras‐Gómez, A., Sánchez‐Mirón, A., García‐Camacho, F. et al. (2013) Protein production using the baculovirus‐insect cell expres- The author received no specific funding for this work. sion system, Biotechnology Progress, 30 (1), 1–18. Craven, G. B., Affron, D. P., Allen, C. E. et al. (2018) High-throughput List of Abbreviations kinetic analysis for target-directed covalent ligand discovery, Angewandte Chemie International Edition, 57 (19), 5257–5261. FBDD Fragment-Based Drug Discovery ............................................................................................... .................................................................. 9 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Dalvit, C., Flocco, M. and Marina Veronesi, B. J. (2002) Fluorine-NMR Hann, M. M., Leach, A. R. and Harper, G. (2001) Molecular complexity competition binding experiments for high-throughput screening of and its impact on the probability of finding leads for drug discov- large compound mixtures, Combinatorial Chemistry & High ery, Journal of Chemical Information and Computer Sciences, 41 (3), Throughput Screening, 5 (8), 605–611. 856–864. Dalvit, C., Pevarello, P., Tatò, M. et al. (2000) Identification of com- Harner, M. J., Frank, A. O. and Fesik, S. W. (2013) Fragment-based drug pounds with binding affinity to proteins via magnetization transfer discovery using NMR spectroscopy, Journal of Biomolecular NMR,56 from bulk water*, Journal of Biomolecular NMR, 18 (1), 65–68. (2), 65–75. Davis, B. J. and Erlanson, D. A. (2013) Learning from our mistakes: the Hassell, A. M., An, G., Bledsoe, R. K. et al. (2006) Crystallization of ‘unknown knowns’ in fragment screening, Bioorganic & Medicinal protein-ligand complexes, Acta Crystallographica Section D: Chemistry Letters, 23 (10), 2844–2852. Biological Crystallography, 63, 72–79. de Sanctis, D., Beteva, A., Caserotto, H. et al. (2012) ID29: a high‐inten- Hopkins, A. L., Groom, C. R. and Alex, A. (2004) Ligand efficiency: a sity highly automated ESRF beamline for macromolecular crystal- useful metric for lead selection, Drug Discovery Today,9(10), lography experiments exploiting anomalous scattering, Journal of 430–431. Synchrotron Radiation, 19 (3), 455–461. Hopkins, A. L., Keserü, G. M., Leeson, P. D. et al. (2014) The role of ligand Derewenda, Z. S. (2010) Application of protein engineering to enhance efficiency metrics in drug discovery, Nature Reviews. Drug Discovery, crystallizability and improve crystal properties, Acta Crystallographica 13, 105. Section D: Biological Crystallography, 66, 604–615. Huang, R., Bonnichon, A., Claridge, T. D. W. et al. (2017) Protein–ligand Diamond Light Source. (2018) Fragment Screening - XChem. Diamond. binding affinity determination by the water LOGSY method: an Available from: https://www.diamond.ac.uk/Instruments/Mx/Fragment- optimised approach considering ligand rebinding, Scientific Screening.html [Accessed 12/12/2018]. Reports, 7, 43727. Erlanson, D. A. (2016) Fragments in the clinic: 2016 edition. Practical Hubbard,R.E.(2016) Theroleoffragment‐based discovery in lead Fragments. Available from: http://practicalfragments.blogspot.co.uk/ finding, in Jahnke W. and Erlanson D. A. (eds), Fragment‐based 2016/07/fragments-in-clinic-2016-edition.html [Accessed 09/09/2018]. Drug Discovery Lessons and Outlook,Wiley,Wiley-VCH,pp. 3–29. Erlanson, D. A., Fesik, S. W., Hubbard, R. E. et al. (2016) Twenty years on: the impact of fragments on drug discovery, Nature Reviews. Drug Jahnke, W., Floersheim, P., Ostermeier, C. et al. (2002) NMR reporter Discovery, 15, 605. screening for the detection of high‐affinity ligands, Angewandte Chemie International Edition, 41 (18), 3420–3423. Erlanson, D. A., Wells, J. A. and Braisted, A. C. (2004) Tethering: fragment-based drug discovery, Annual Review of Biophysics and Jhoti, H., Cleasby, A., Verdonk, M. et al. (2007) Fragment-based screen- Biomolecular Structure, 33 (1), 199–223. ing using X-ray crystallography and NMR spectroscopy, Current Opinion in Chemical Biology, 11 (5), 485–493. Evans, G., Axford, D., Waterman, D. et al. (2011) Macromolecular micro- crystallography, Crystallography Reviews, 17 (2), 105–142. Jhoti, H., Williams, G., Rees, D. C. et al. (2013) The ‘rule of three’ for fragment-based drug discovery: where are we now? Nature Friberg, A., Vigil, D., Zhao, B. et al. (2012) Discovery of potent myeloid cell Reviews. Drug Discovery, 12, 644. leukemia 1 (Mcl 1) inhibitors using fragment based methods and structure based design, Journal of Medicinal Chemistry,56 (1), 15–30. Khan, K. H. (2013) Gene expression in mammalian cells and its applica- tions, Advanced Pharmaceutical Bulletin, 3 (2), 257–263. Gee, C. T., Arntson, K. E., Urick, A. K. et al. (2016) Protein-observed 19F- NMR for fragment screening, affinity quantification and druggabil- Kim, M., Lee, S., Choi, H. et al. (2006) On-chip Escherichia coli culture, ity assessment, Nature Protocols, 11, 1414. purification, and detection of expressed proteins, European Biophysics Journal, 35 (8), 655–662. Giannetti, A. M. (2011) Chapter eight - from experimental design to val- idated hits: a comprehensive walk-through of fragment lead identi- Kiraly, P., Adams, R. W., Paudel, L. et al. (2015) Real-time pure shift (15)N fication using surface plasmon resonance, in Kuo L. C. (ed), HSQC of proteins: a real improvement in resolution and sensitivity, Fragment-Based Drug Design: Tools, Practical Approaches, and Journal of Biomolecular NMR, 62 (1), 43–52. Examples, Academic Press Elsevier, pp. 169–218. Lamoree, B. and Hubbard, R. (2017) Current perspectives in Hajduk, P. J., Augeri, D. J., Mack, J. et al. (2000) NMR-based screening of fragment-based lead discovery (FBLD), Essays in Biochemistry, proteins containing 13C-labeled methyl groups, Journal of the 61 (5), 453. American Chemical Society, 122 (33), 7898–7904. Lipinski, C. A., Lombardo, F., Dominy, B. W. et al. (2001) Experimental Hajduk, P. J., Huth, J. R. and Fesik, S. W. (2005) Druggability indices for and computational approaches to estimate solubility and perme- protein targets derived from NMR-based screening data, Journal of ability in drug discovery and development settings, Advanced Drug Medicinal Chemistry, 48 (7), 2518–2525. Delivery Reviews, 46 (1), 3–26. ............................................................................................... .................................................................. 10 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Bioscience Horizons � Volume 11 2018 Review article ............................................................................................... .................................................................. Locatelli-Hoops, S., Yeliseev, A. A., Gawrisch, K. et al. (2013) Surface plas- Patching, S. G. (2014) Surface plasmon resonance spectroscopy for mon resonance applied to G protein-coupled receptors, Biomedical characterisation of membrane protein–ligand interactions and its Spectroscopy and Imaging, 2 (3), 155–181. potential for drug discovery, Biochimica Et Biophysica Acta (BBA) - Biomembranes, 1838 (1, Part A), 43–55. Loll, P. J. (2014) Membrane proteins, detergents and crystals: what is the state of the art? Acta Crystallographica Section F, Structural Patel, D., Bauman, J. D. and Arnold, E. (2014) Advantages of crystallo- Biology Communications, 70, 1576–1583. graphic fragment screening: functional and mechanistic insights from a powerful platform for efficient drug discovery, Progress in Luchinat, E. and Banci, L. (2016) In-cell NMR: a topical review, IUCRJ,4, Biophysics and Molecular Biology,116 (0), 92–100. 108–118. Pelz, N. F., Bian, Z., Zhao, B. et al. (2016) Discovery of 2-Indole- Ma,R., Wang,P., Wu, J.etal. (2016) Processof fragment-basedleaddiscov- acylsulfonamide Myeloid Cell Leukemia 1 (Mcl-1) inhibitors using ery - a perspective from NMR, Molecules (Basel, Switzerland), 21 (7), E854. fragment-based methods, Journal of Medicinal Chemistry, 59 (5), 2054–2066. Macarron, R., Banks, M. N., Bojanic, D. et al. (2011) Impact of high- throughput screening in biomedical research, Nature Reviews. Drug Perkel, J. M. (2011) Surface Plasmon Resonance (SPR) and Other Discovery, 10, 188. Options for Label-free Detection. Biocompare. Available from: https://www.biocompare.com/Editorial-Articles/41803-Surface-Plasmon- Mahalingam, D., Swords, R., Carew, J. S. et al. (2009) Targeting Resonance-SPR-and-Other-Options-for-Label-free-Detection/ [Accessed HSP90 for cancer therapy, British Journal of Cancer, 100 (10), 09/09/2018]. 1523–1529. Pfaff, S. J., Chimenti, M. S., Kelly, M. J. S. et al. (2015) Biophysical meth- Mayer, M. and Meyer, B. (1999) Characterization of ligand binding by ods for identifying fragment-based inhibitors of protein–protein saturation transfer difference NMR spectroscopy, Angewandte interactions, in Meyerkord C. L. and Fu H. (eds), Protein–Protein Chemie International Edition, 38 (12), 1784–1788. Interactions: Methods and Applications, Springer New York, New McPherson, A. and Gavira, J. A. (2013) Introduction to protein crystal- York, NY, pp. 587–613. lization, Acta Crystallographica Section F, Structural Biology Philipp, H. T., Tate, M. W., Purohit, P. et al. (2016) High-speed X-ray Communications, 70, 2–20. imaging pixel array detector for synchrotron bunch isolation, Meng, X., Zhang, H., Mezei, M. et al. (2011) Molecular docking: a power- Journal of Synchrotron Radiation, 23 (2), 395–403. ful approach for structure-based drug discovery, Current Computer- Rees, D. C., Congreve, M., Murray, C. W. et al. (2004) Fragment-based Aided Drug Design, 7 (2), 146–157. lead discovery, Nature Reviews. Drug Discovery, 3, 660. Molecular Dimensions. (2018) Crystallisation Screens. Molecular Renaud, J., Chung, C., Danielson, U. H. et al. (2016) Biophysics in drug Dimensions. Available from: https://www.moleculardimensions. discovery: impact, challenges and opportunities, Nature Reviews. com/products/c255-Crystallization-Screens/ [Accessed 09/09/2018]. Drug Discovery, 15, 679. Murray, C. W., Carr, M. G., Callaghan, O. et al. (2010) Fragment-based drug Rüdisser, S., Vangrevelinghe, E. and Maibaum, J. (2016) An integrated discovery applied to Hsp90. Discovery of two lead series with high lig- approach for fragment‐based lead discovery: virtual, NMR, and and efficiency, Journal of Medicinal Chemistry, 53 (16), 5942–5955. high‐throughput screening combined with structure‐guided Murray, C. W. and Rees, D. C. (2009) The rise of fragment-based drug design. Application to the aspartyl protease renin, in Jahnke W. and discovery, Nature Chemistry, 1, 187. Erlanson D. A. (eds), Fragment‐based Drug Discovery: Lessons and Outlook, Wiley, Wiley-VCH, pp. 447–480. Murray, C. W., Verdonk, M. L. and Rees, D. C. (2012) Experiences in fragment-based drug discovery, Trends in Pharmacological Sciences, Scott, D. E., Christina, S. and Chris, A. (2016) Differential scanning fluori- 33 (5), 224–232. metry as part of a biophysical screening cascade, in Jahnke W. and Erlanson D. A. (eds), Fragment-based Drug Discovery: Lessons and Navratilova, I. and Hopkins, A. L. (2010) Fragment screening by surface Outlook, Wiley, Wiley-VCH, pp. 139–170. plasmon resonance, ACS Medicinal Chemistry Letters, 1 (1), 44–48. Segall, M. D. and Barber, C. (2014) Addressing toxicity risk when design- Neumann, T., Junker, H.-D., Schmidt, K. et al. (2007) SPR-based frag- ing and selecting compounds in early drug discovery, Drug ment screening: advantages and applications, Current Topics in Discovery Today, 19 (5), 688–693. Medicinal Chemistry, 7 (16), 1630–1642. Shah, U. V., Jahn, N. H., Huang, S. et al. (2017) Crystallisation via novel Olivier, B., Fatima, A., Christel, L. et al. (2012) Overexpression of mem- 3D nanotemplates as a tool for protein purification and bio- brane proteins using Pichia pastoris, Current Protocols in Protein separation, Journal of Crystal Growth, 469, 42–47. Science, 29 (1), 29.2.1–29.2.24. Shah, U. V., Parambil, J. V., Williams, D. R. et al. (2015) Preparation and Ostrem, J. M., Peters, U., Sos, M. L. et al. (2013) K-Ras(G12C) inhibitors characterisation of 3D nanotemplates for protein crystallisation, allosterically control GTP affinity and effector interactions, Nature, Powder Technology, 282, 10–18. 503 (7477), 548–551. ............................................................................................... .................................................................. 11 Downloaded from https://academic.oup.com/biohorizons/article/doi/10.1093/biohorizons/hzy015/5256437 by DeepDyve user on 14 August 2022 Review article Bioscience Horizons � Volume 11 2018 ............................................................................................... .................................................................. Shuker, S. B., Hajduk, P. J., Meadows, R. P. et al. (1996) Discovering high- Smith, J. L., Fischetti, R. F. and Yamamoto, M. (2012) Micro- affinity ligands for proteins: SAR by NMR, Science (New York, N.Y.), crystallography comes of age, Current Opinion in Structural Biology, 274 (5292), 1531. 22 (5), 602–612. Singh, V., Singh, K., Nand, A. et al. (2015) Small molecule microarray screen- Wu, B., Zhang, Z., Noberini, R. et al. (2013) HTS by NMR of combinatorial ing methodology based on surface plasmon resonance imaging, libraries: a fragment-based approach to ligand discovery, Chemistry Arabian Journal of Chemistry. https://doi.org/10.1016/j.arabjc.2014.12.020 & Biology, 20 (1), 19–33. ............................................................................................... ..................................................................

Journal

BioScience HorizonsOxford University Press

Published: Jan 1, 2018

References