The role of MicroScale Thermophoresis (MST) in Drug Discovery of protein kinase inhibitors

UNIVERSITA’ DEGLI STUDI DI PARMA DOTTORATO DI RICERCA IN “SCIENZE DEL FARMACO” CICLO XXXV THE ROLE OF MICROSCALE THERMOPHORESIS (MST) IN DRUG DISCOVERY OF PROTEIN KINASE INHIBITORS Coordinatore: Chiar.mo Prof. Marco Mor Tutori: Chiar.mo Prof. Marco Mor Dott.ssa Barbara Pioselli Dottorando: Elena Picchi Anni Accademici 2019/2020 – 2021/2022 2 3 TABLE OF CONTENTS ABSTRACT .............................................................................. 5 1. INTRODUCTION ..................................................................... 7 1.1 PROTEIN INTERACTIONS ....................................................................... 7 1.1.1 PROTEIN INTERACTIONS KEY ROLE AND CHARACTERIZATION ................................................. 7 1.1.2 BINDING THEORY: THE LAW OF MASS ACTION ........................................................................ 8 1.1.3 CORRELATION BETWEEN KD, Ki and IC50: THE CHENG-PRUSOFF EQUATION .......................... 10 1.1.4 FITTING MODELS .................................................................................................................... 11 1.1.5 KEY CRITERIA TO MEASURE BINDING REACTIONS .................................................................. 14 1.1.6 MODES OF INHIBITOR INTERACTIONS .................................................................................... 15 TIGHT BINDING ....................................................................................................................... 16 SLOW BINDING ....................................................................................................................... 19 COVALENT BINDING ............................................................................................................... 20 1.2 THE MICROSCALE THERMOPHORESIS (MST) ....................................... 21 1.2.1 GENERAL DESCRIPTION .......................................................................................................... 21 1.2.2 MST APPROACHES .................................................................................................................. 25 1.3 PROTEIN KINASES AS DRUG TARGETS................................................. 27 1.3.1 EPIDERMAL GROWTH FACTOR RECEPTOR (EGFR).................................................................. 29 1.3.2 RHO-ASSOCIATED PROTEIN KINASE (ROCK) ........................................................................... 31 2. AIM OF THE WORK .............................................................. 33 3. MATERIALS AND METHODS ................................................. 35 3.1 EGFR .................................................................................................. 35 3.1.1 MATERIALS ............................................................................................................................. 35 3.1.2 MATERIAL HANDLING AND STORAGE .................................................................................... 36 3.1.3 GENERAL PROTOCOL FOR EGFR INTERACTION ASSAYS ......................................................... 36 3.1.4 STOICHIOMETRY ASSAYS ........................................................................................................ 38 3.1.5 REVERSIBLE LIGAND RETENTION ASSAY ................................................................................. 40 3.2 ROCK .................................................................................................. 42 3.2.1 MATERIALS ............................................................................................................................. 42 3.2.2 MATERIAL HANDLING AND STORAGE .................................................................................... 42 3.2.3 GENERAL PROTOCOL FOR ROCK COMPETITION ASSAYS ........................................................ 43 3.2.4 LABELING PROTOCOL ............................................................................................................. 45 3.2.5 GENERAL PROTOCOL FOR ROCK BINDING ASSAYS ................................................................. 45 3.3 MST-ON TIME CHOICE ........................................................................ 47 4 3.4 DATA ELABORATION AND FITTING MODELS ....................................... 57 4. RESULTS AND DISCUSSION .................................................. 59 4.1 EGFR (Epidermal Growth Factor Receptor) ......................................... 59 4.1.1 INTERACTION WITH KINASE TRACER 199 ............................................................................... 60 4.1.2 COMPETITION ASSAYS ............................................................................................................ 72 GEFITINIB - TIGHT BINDING CONDITIONS .............................................................................. 72 LAPATINIB - SLOW BINDING TYPE II INHIBITOR ...................................................................... 80 OSIMERITINIB - IRREVERSIBLE BINDING ................................................................................. 87 4.2 ROCK (Rho-associated protein kinase) ................................................ 95 4.2.1 COMPETITION ASSAYS ............................................................................................................ 96 INTERACTION WITH KINASE TRACER 236 ............................................................................... 96 OPTIMIZATION OF THE INTERACTION WITH COMPOUND A .................................................. 97 COMPETITION WITH ALLOSTERIC COMPOUNDS .................................................................... 99 4.2.2 DIRECT BINDING ASSAYS ...................................................................................................... 102 ROCK-1 COVALENT LABELING AND BINDING COMPETENCE TEST ....................................... 102 INTERACTION WITH ALLOSTERIC COMPOUNDS ................................................................... 107 5. CONCLUSIONS ................................................................... 111 6. BIBLIOGRAPHY .................................................................. 113 5 ABSTRACT MicroScale Thermophoresis (MST) is one of the biophysical techniques commonly used for the characterization of ligand/protein interactions. The phenomenon of "thermophoresis" is the directed movement of molecules along a temperature gradient generated by an IR laser and MST can detect changes in charge, size, and hydration shell or conformation of a biomolecular complex caused by the interaction between a target protein and the cognate ligand. This phenomenon can be quantified by titrating the ligand to obtain a binding curve from which the dissociation constant (KD) can be derived. Even though the main application of MST is to determine binding parameters, the technique can also be employed to gain insights into other aspects of protein interactions such as stoichiometry, conformational states, time dependency, selectivity over mutations, and thermodynamics. In this project different types of interactions were investigated by using two model protein kinases of pharmaceutical interest, EGFR (Epidermal Growth Factor Receptor) and ROCK (Rho-associated protein kinase) in the presence of their well-known inhibitors. The main purpose was to investigate MST ability to characterize various binding modes with a particular emphasis on potent inhibitors, slow binders, covalent binders, and allosteric binders. First, EGFR system enabled the characterization of potent inhibitors while facing their intrinsic limitation of resulting in tight binding experimental conditions. Second, a time dependence analysis revealed the MST potential for describing conformational changes in protein kinases, whereas a targeted stoichiometry experiments facilitated the identification of various protein states as well as a difference in the propensity of Type I and Type II inhibitors to bind them. Finally, an alternative method to Jump Dilution for differentiating reversible and irreversible inhibitors was developed using MST. ROCK system, on the other hand, has been used to differentiate between orthosteric and allosteric binders by combining orthogonal approaches of competition and direct binding assays. Furthermore, an analysis of the best labeling conditions revealed differences between the two examined allosteric compounds, indicating a possible distinct interaction mechanism due to their different sensitivity to the dye position on the protein. These results enabled the development of a wide range of knowledge in the field of protein kinase interactions, providing a suitable background for future and unknown interaction systems. 6 7 1. INTRODUCTION 1.1 PROTEIN INTERACTIONS 1.1.1 PROTEIN INTERACTIONS KEY ROLE AND CHARACTERIZATION Proteins are macromolecules performing essential roles in the cell, including biochemical (enzymes), structural (cytoskeleton), mechanical (muscle), and cell signaling (hormones) functions[1]. Protein interactions are the basis of every biological function, from the simplest biochemical signal to the more complex cellular rearrangement. They are usually very specific and can either involve small molecules and cofactor or other proteins and macromolecules[2]. The intricate system of pathways and specific interactions typically contribute to the maintenance of the correct homeostasis of a whole organism but in certain cases can be the cause of undesired altered states, leading to disease mechanisms[3]. Indeed, almost every drug works by binding proteins (specific target) and interfere with their biological function. Therefore, a deep understanding of protein interactions mechanism and its characterization play a crucial role in biochemistry studies and Drug Discovery pipeline. To do this, orthogonal biophysical measures could be applied, allowing for a comprehensive description of the interactions in terms of affinity, kinetics, and thermodynamics. A complete characterization also includes data on conformational changes, stoichiometry, and the identification of specific binding regions[4]. The most appropriate biophysical technique for a specific goal must take into account a variety of factors, including the availability of protein and ligand in terms of quantity, purity, and concentration, all prior knowledge about the interaction system, and, most importantly, which output would best answer a specific question[2]. Once these aspects are clarified, it is possible to optimize an adequate experimental design and proceed with the assays. A leading approach is often chosen as the primary investigative method, while a few other techniques are applied for cross-validation and to gain complementary information[5]. 8 1.1.2 BINDING THEORY: THE LAW OF MASS ACTION Most of the binding interactions occur between proteins and smaller molecules generically called ligands. These interactions are typically reversible and can be quantified using the law of mass action, a simple physical-mathematical model in which a molecule A binds a molecule B to form the complex AB[6]: A + B  AB The binding process is described by the binding kinetics and depends on the rate of association and dissociation of the two binding partners. When a protein molecule (P) and its specific ligand molecule (L) find each other in solution the interaction system can be described as follow: where PL is the protein-ligand complex, kon is the association rate constant and koff is the dissociation rate constant. The units of measurement of these two parameters are respectively M-1s-1 and s-1. When the reaction reaches the equilibrium state the association rate is equal to the dissociation rate: Here the square brackets indicate the concentration of the molecule in solution. The rate of association kon depends on the concentration of both free protein and free ligand in solution (or the “active masses” of the reactants), because the higher the number of molecules, the higher the probability of their collision. Moreover, because larger proteins have a larger surface area of interaction, the association rate will be higher. Although not every encounter result in a beneficial interaction, once the two binding partners are in proximity, electrostatic forces can draw them towards one other and direct the charged ligand onto the binding site. Furthermore, the organization of the surface chemical groups on the protein can force a ligand that collides anywhere on the protein's surface towards the binding site. On the contrary, if only a fraction of the protein and ligand are in a binding-competent state, the rate of association may be reduced[2]. The rate of dissociation is a stochastic event and describes the probability of a complex to dissociate the next second and it is only proportional to the concentration of the complex. 9 Therefore, at equilibrium it is possible to rearrange the equation: 𝐾𝐷=𝑘𝑜𝑓𝑓 𝑘𝑜𝑛 𝐾𝐷=[𝑃][𝐿] [𝑃𝐿] where KD is the dissociation constant, a parameter directly related to the affinity of the ligand for the protein. When a ligand concentration is equal to the KD, the amount of the ligand present in the solution would be statistically enough to bind half of the total number of the binding sites. Knowing that the total protein concentration is the sum of the free protein in solution and the protein in complex with the ligand as described below: [𝑃]=[𝑃]𝑡𝑜𝑡−[𝑃𝐿] the equation can rearrange as follow: [𝑃𝐿]=[𝐿][𝑃]𝑡𝑜𝑡 [𝐿]+𝐾𝐷 Considering [PL] as a function of [L], this equation is that of a hyperbola called "Binding Isotherm", becoming a sigmoidal curve using a semi-logarithmic scale, the mostly used model to describe biophysical, biochemical and pharmacological interaction data (Figure 1). 0 2×10-6 4×10-6 6×10-6 8×10-6 -50 0 50 100 150 5×10-5 1×10-41.5×10-4 [ligand] M binding observable -10 -8 -6 -4 -2 -50 0 50 100 150 Log[ligand] binding observable Figure 1 Binding isotherm represented as a rectangular hyperbola (on the left) and as a sigmoidal curve (on the right). It is important to emphasize that the law of mass action is based on some prerequisites: 1. The reaction must be reversible, so an equilibrium state can be reached. 2. Every protein molecule needs to be equally accessible to the ligand. 3. The protein and the ligand must exist only in the bound or in the unbound state. 4. The interaction must not alter the protein or the ligand. 16 TIGHT BINDING[14][16][17][18][19] The tight binding experimental conditions occur when the analyzed inhibitor has an inhibition constant, Ki, lower than the specific protein concentration selected for the assay. As a result, tight binding is not an absolute characteristic inherent to the potent compound, but rather a relative condition strictly dependent to the experimental chosen concentrations. When tight binding occurs, at inhibitor concentrations lower than the protein concentration every ligand molecule added to the system is sequestered by the protein itself. Therefore, the proteininhibitor interaction will be governed by the amount of protein in the system rather than the actual affinity of the small molecule. When an interaction experiment is performed, the derived phenomenological IC50 value becomes strongly dependent on the protein concentration and, since the IC50 is the inhibitor concentration required to achieve 50% binding, it will never be less than half the protein concentration. In classical binding conditions, when Ki > [P], it is possible to assume that IC50 ~ Ki. Instead in tight binding conditions the protein concentration becomes a relevant parameter. If Ki/[P] is between 0.01 and 10, the IC50 depends on both the protein concentration, [P], and the Ki, according to the following equation: 𝐼𝐶50=[𝑃] 2+ 𝐾𝑖 When Ki/[P] is lower than 0.01 the IC50 becomes independent from the Ki value and the equation is rearranged: 𝐼𝐶50=1 2 [𝑃] In these conditions the system has entered the titration regime, the “IC50 wall” has been hit and the IC50 will not fall below the half the protein concentration (if it does, the latter has been probably overestimated in terms of active fraction). 17 Figure 3 Different aspect of interaction curves: in the “A” section is reported the classical binding curve (when KD > [P]), while in the “B” section is indicated the titration regime when KD << [P]. The steepness of the curve becomes higher and the best fitting model becomes the quadratic. Adapted from: Jarmoskaite, I. et al. (2020) ‘How to measure and evaluate binding affinities’, eLife, 9, pp. 1–34. doi: 10.7554/ELIFE.57264. When the system is in tight binding conditions, the data can be elaborated using the quadratic Morrison equation, which includes the protein concentration as a parameter into the fitting model. 18 One issue of debate is whether to leave the [P] float and so consider it as a variable parameter, or to keep it constant when applying the Morrison model. Kuzmic (2000) and Murphy (2004) reported distinct simulation studies in which they concluded that as the ratio of Ki to [P] drops, the inaccuracy in calculating Ki increases by fixing the protein concentration. The proposed solution involves working in two steps: first, fixing the protein concentration to its nominal value and observing the resulting inhibition constant: if the Ki value is greater than the fixed [P] value, the final result can be accepted. Alternatively, the fitting analysis must be repeated while allowing the model to fit the protein concentration as a variable parameter. Otherwise, according to Copeland (2005 and 2013), allowing the protein concentration to fluctuate may result in physically incoherent estimations of [P] as well as errors in estimating Ki. However, both points of view agree on the importance of accurately estimating the protein active fraction. The protein population is assumed to exist in at least two states, one of which is binding competent and the other unfolded or denatured. The competent state may be in turn populated by different conformational states. The active fraction can be determined by exploiting the tight binding situation itself. One of the two binding partners has to be fixed at a concentration greater than Ki (the optimum would be a factor 200) to achieve tight binding conditions. The concentration of the other binding partner must be titrated over a narrow range, which includes the first one. As a result, there will be two distributions of points that can be fit by two different linear regressions. The distribution at high titrant concentrations indicates the totally bound condition, whereas the distribution at lower titrant concentrations indicates the titrating regime. Assuming a stoichiometry of 1:1 the amount of the active fraction is determined by the breaking point between the two distributions. By increasing the tight binding condition, the breaking point becomes sharper, and the active fraction is more precisely defined (Figure 4). 19 Figure 4 The figure is referred to a case study reported by Jarmoskaite (2020) in which RNA is used as biological target and tested at the fixed concentration of 100 nM (on the left) and 10 nM (on the right). The measure of the protein active fraction was conducted by titrating the protein Puf4 while in tight binding conditions. By using a higher RNA concentration (on the left), the two distributions of points are better described, and the breaking point is sharper. Jarmoskaite, I. et al. (2020) ‘How to measure and evaluate binding affinities’, eLife, 9, pp. 1–34. doi: 10.7554/ELIFE.57264. Once the amount of active protein has been identified, the binding data can be fitted by using the Morrison equation, by including the actual amount of protein capable of binding into the model. SLOW BINDING[17] Slow binders are compounds that either associate or dissociate slowly from the protein. Therefore, due to their time dependence, affinity and potence need to be established only after the reaching of equilibrium. Slow binding can occur with different mechanisms: - Simple reversible slow binding: mechanism in which the association, the dissociation or both are slow; - Induced-fit: two-step reaction with a first rapid interaction and a following slow interconversion of the protein to a form that better accommodates the ligand; - Conformation selection: two-step reaction with a first slow conformational interconversion of the protein and then a rapid interaction with the inhibitor. Usually a complete characterization of slow binders yields the inhibition mechanisms, the true affinity based on enzymatic assays and the indication of the residence time, the period of time that the ligand spends bound to its target. 20 COVALENT BINDING[17][20][21][22] Covalent inhibitors are molecule with a reactive warhead aimed to form covalent bond with specific residues of target proteins. They often involve a two-step reaction, the first of which is a reversible interaction between the inhibitor and its binding pocket while the covalent bond itself is formed as a result of the second following reaction. The possible difficulty in selectivity is balanced by a longer duration of action, improved ligand efficiency, and the capacity to avoid drug resistance. Therefore, the covalent interaction consists of two components: the binding affinity between the inhibitor and the protein (Ki) and the chemical reactivity described by the rate of enzyme inactivation (kinact), the required time to covalently modify half of the protein[23]. Usually covalent inhibition is described with the covalent efficiency constant (kinact/Ki) a preferred parameter describing the inhibitory potency without the time dependence interference. However, in some cases the inhibitory concentration at a specific time, IC50(t), can be carefully used as surrogate of the parameter kinact/Ki, to can more easily correlate data with other activity assays[24]. Finally, the verification that the irreversible reaction has occurred is a critical element to take into account. The methods usually employed for this purpose are mass spectrometry analyses to identify the covalent adduct and the enzymatic test of activity recovery after a massive dilution (Jump Dilution). 21 1.2 THE MICROSCALE THERMOPHORESIS (MST)[25][26][27][28] Among the different biophysical approaches aimed at the characterization of protein-ligand interactions (e.g. ITC, SPR, fluorescence spectroscopy, etc.), the MicroScale Thermophoresis emerges as a versatile technique for the detection of binding between a large variety of molecular species. The approach main advantages are its minimal sample consumption, quickness of analysis, ability to utilize almost any type of buffer, and ability to discover binding of any nature (orthosteric, allosteric, etc.) as long as a sufficient ligand-dependent change in thermophoretic mobility occurs. 1.2.1 GENERAL DESCRIPTION The MicroScale Thermophoresis (NanoTemper) is an immobilization-free biophysical technique for the characterization of bimolecular interactions. The method is based on a physical phenomenon called thermophoresis, according to which every molecule in a fluid and subjected to a temperature gradient will start to migrate along the gradient, with a rate directly related to its physical characteristics; in particular it has been demonstrated the three aspects that mainly impact on the thermophoretic migration are: - Size - Charge - Hydration shell These characteristics are strongly influenced by external perturbation of the system; indeed, by considering the single molecule of a protein, its size, charge and conformation may significantly change in the presence of a specific interactor like an inhibitor. Therefore, it is by titrating the protein with a ligand molecule that is possible to monitor the change in thermophoresis migration of the resulting complex and to use that information as observable to obtain a binding curve and specific interaction parameters. In the MicroScale Thermophoresis (MST), the molecule movement start with the activation of an infrared laser (wavelength 1480 nm) and its energy absorption by the water molecule in solution. This rapidly results in the generation of a temperature gradient along the capillary tube inside of which the sample is loaded; the molecules start to migrate from the hotter to the colder zone as a consequence of the so-called Soret’s effect. The movement continues until it would be contrasted by the back-diffusion of the molecules. 22 The Soret’s effect is described by the following equation: ST=A kBT(−∆shyd(T)+βσeff 2 4εε0T λDH) where ST is the Soret coefficient, A is the molecule surface, λDH is the Debye length, kB is the Boltzmann constant, ∆shyd is the particle-area-specific hydration entropy, σeff is the effective surface charge density, ε is the dielectric constant, β is the coefficient that describe the correlation between the dielectric constant and the Debye length, T is the temperature (expressed in Kelvin)[29]. Once the IR ray has activated and generated the temperature gradient, the movement of the molecules is detected by their intrinsic fluorescence or through the use of fluorophores covalently attached to the protein. Indeed, a fluorescence detector is placed in the same spot of the IR activator and collects the information of decrease in the observed fluorescence in the center of the temperature gradient. The IR laser intensity is determined by the MST power, which may be regulated in three modes: low, medium, and high. The irradiation sample volume is approximately 2 nL, with a heated zone diameter of 100 µm, and the temperature gradient amplitude increases with intensity. In Figure 5 is reported a representation of every step of an MST experiment. Figure 5 Representation of an MST experiment and its phases. Adapted from M. Jerabek-Willemsen et al., “MicroScale Thermophoresis: Interaction analysis and beyond,” J. Mol. Struct., vol. 1077, pp. 101–113, Dec. 2014, doi: 10.1016/j.molstruc.2014.03.009. 23 The time along which the IR laser remains active is named MST-ON time and usually the experimenter can arbitrarily choose among six different MST-ON time, (1.5, 2.5, 5, 10, 15 and 20 seconds). Immediately after the IR activation the system detects a consistent drop in relative fluorescence called T-Jump. The initial change in fluorescence is most likely caused by the greater fluorophore collisional quenching at elevated temperatures (TRIC, Temperature Dependent Intensity Change) rather than the actual thermophoresis of the molecules[28]. The later migration times, instead, are most likely to be governed by the global properties of the molecule or complex in terms of size, charge and hydration shell. The fluorescence decrease at every MST-ON time (Fhot) is compared to the initial fluorescence (Fcold) and plotted as normalized fluorescence (FNorm), on a fractional or per mille scale: 𝐹𝑁𝑜𝑟𝑚= 𝐹ℎ𝑜𝑡 𝐹𝑐𝑜𝑙𝑑 ‰ As already mentioned, the thermophoretic mobility of a protein should change in the presence of a ligand that modifies the migration characteristics of the protein alone. Therefore, by titrating the interactor from an appropriate starting concentration, around 50-100-fold the estimated KD, it is possible to obtain a symmetrical distribution around the ligand concentration corresponding to the KD value. Then, by choosing an arbitrary MST-ON time, the relative experimental points set can be fitted to obtain an interaction curve and the relative parameters (Figure 6). Figure 6 Representation of how a binding curve is derived from thermophoresis traces. 24 The difference between the MST signal of the unbound and bound states is called Response Amplitude and depends on both the MST-ON time (later times are always characterized by a higher Response Amplitude) and on the MST power. An indication of the quality of the binding curve is the signal-to-noise ratio (S/N) defined as follow: 𝑆 𝑁= 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝐴𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑁𝑜𝑖𝑠𝑒 where the Noise is the standard deviation of the measurement and it is defined as the MST signal variation do not caused by the interactor: 𝑁𝑜𝑖𝑠𝑒= √∑(𝑟𝑖− 𝑟)2 𝑖𝑛−1 with ri indicating the residual of the fit, 𝑟 the average of all residuals and n the number of data points[30]. Another method to indicate the Noise is the standard deviation of the residuals (Sy.x), reported by software like GraphPad Prism: 𝑆𝑦.𝑥= √∑(𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙2) 𝑖𝑛−𝐾 where the residual is the vertical distance (Y units) of the experimental points from the curve and n-K is the number of degrees of freedom of the regression. 25 1.2.2 MST APPROACHES The MicroScale Thermophoresis can be used for two different approaches: the competition assay and the direct binding assay. The competition assay relies on the use of a fluorescent molecule to bind the protein target; then, a second non-fluorescent molecule is titrated to displace the first one, and an indirect indication of the affinity of the titrating molecule is collected. The direct binding assay relies on the use of a fluorescent molecule as one of the binding partners to obtain affinity information on the interaction system with a non-fluorescent molecule which is titrated. To employ the direct binding strategy, one of the interactors must be fluorescent, either intrinsically or extrinsically, and fall within the emission-excitation range in order to be detected correctly throughout the experiments. In Figure 7 are reported different types of fluorescent dyes and their excitation and emission ranges. Figure 7 Different types of dyes and their excitation and emission ranges. The instrument used for the analysis of this work and the potentially applicable fluorophores for measurements on the device are indicated in the red box. Adapted from NanoPedia - Monolith NT.115, NanoTemper Technologies GmbH. The binding approach frequently requires covalent labeling of the protein, which can be a difficult step due to the likelihood of dye interference with the structure and functionality of the protein, and thus with its binding competence[31]. Furthermore, the degree of labeling must be optimized because a low amount of dye covalently attached to the protein may result in a low fluorescence 32 33 2. AIM OF THE WORK This work aims to validate the MicroScale Thermophoresis (MST) as a biophysical technique to obtain valuable information on different binding modes in the field of protein kinase inhibitors. The relevance and the therapeutic potential that kinase inhibitors assumed in pharmaceutical research are due to the essential role these proteins play in life processes and in several pathologic conditions. Therefore, in the last decades medicinal chemistry directed its efforts to the characterization of kinase-inhibitor interaction in order to optimize drug design. Most of the information addressing SARs (Structure-Activity Relationship) choices result from in vitro studies[3]. Protein kinase inhibitors have a wide range of inhibition mechanisms, thus having a biophysical platform from which obtaining reliable information on these mode of action varieties is important for both a deeper understanding of already known mechanisms and a more trustable investigation of new potential drugs. MicroScale Thermophoresis is in most cases only applied to gain the affinity information of KD or IC50, but this perspective might be limiting considering the potential of the technique and its low sample consumption and rapidity of analysis. In this study, the MST has been employed to characterize, from different perspectives, the interaction modalities of two model protein kinases of pharmaceutical interest (EGFR and ROCK) and their well-known inhibitors to create a wide range knowledge to be used in future and unknown interaction systems. 34 35 3. MATERIALS AND METHODS The following methods are referred to the interaction experiments performed with the instrument Monolith NT.115 (NanoTemper). 3.1 EGFR 3.1.1 MATERIALS The EGFR (Epidermal Growth Factor Receptor) interaction system has been investigated through the competition assay and the isoforms considered were the catalytic domains of EGFR wild type and the mutate form EGFR L858R. The tested inhibitors were the three orthosteric compounds Gefitinib, Lapatinib and Osimertinib. For the competition assay the selected fluorescent species was the Kinase Tracer 199. In the table below are reported the characteristic of the starting material purchased for the experiments. Item Supplier Product Number Lot Tag Sequence (start-end) MW Da Concentration Purity Storage buffer EGFR (ErbB1) Wild Type Thermo Fisher PR7295B 2281133H GST 668-1210 90500 4.86 µM (0.44 g/L) 70 % 50 mM Tris (pH 7.5), 150 mM NaCl, 0.5 mM EDTA, 0.02% Triton® X-100, 2 mM DTT and 50% Glycerol EGFR (ErbB1) L858R Thermo Fisher PR7447A 2468986C GST 668-1210 90500 3.09 µM (0.28 g/L) 70 % 50 mM Tris (pH 7.5), 150 mM NaCl, 0.5 mM EDTA, 0.02% Triton® X-100, 2 mM DTT and 50% Glycerol Kinase Tracer 199 Thermo Fisher PV5830 - - - - 25 µM - DMSO Gefitinib (Iressa®) UniPR - - - - 446.90 - - Powder Lapatinib (GW572016) Med Chem Express HY-50898 - - - 581.06 10 mM - DMSO Osimertinib (AZD9291) Med Chem Express HY-15772 - - - 499.61 10 mM - DMSO The employed proteins are recombinant and fused to glutathione S-transferase (GST-tag), a commonly used method to improve purified protein solubility[51]. The constructs are the same employed in two cited works[52][53] and this allowed a direct comparison between the reported results and the different interaction systems to further validate MST as a valuable technique orthogonal to other biophysical approaches. 36 3.1.2 MATERIAL HANDLING AND STORAGE Both proteins and small molecules in DMSO are stored at -80°C upon their arrival. The Kinase Tracer 199 is stored at -20°C. Proteins For the first use (and for every thawing cycles) the proteins were thawed in ice, gently pipetted, and then centrifuged at 5000 rpm, 0°C, 5 minutes. Then, the proteins were aliquoted in PCR tubes 10 µL each and used only with a single further freezethaw cycle, besides the first one. Small molecules A 10 mM stock solution in DMSO was prepared also for the Gefitinib, provided in powder. The stock solutions of the three inhibitors were stored at -80°C in 100 µL aliquots and thawed and vortexed prior the use for every experimental session. The stock solution 10 mM in DMSO is the starting solution for the assays. Kinase Tracer 199 An intermediate solution of Tracer 199 is prepared at 250 nM by diluting 2 µL of stock solution (25 µM) in 198 µL of buffer 50 mM HEPES (pH 7.5), 150 mM NaCl. The solution is then stored in vial at - 20°C and thawed and vortexed prior the use for every experimental session. This intermediate solution is the starting solution for the assays. 3.1.3 GENERAL PROTOCOL FOR EGFR INTERACTION ASSAYS The interaction assays for the EGFR system are performed as follow: 1) Thawing of the protein, the Kinase Tracer 199 and the inhibitor of interest; centrifugation of the protein at 5000 rpm, 0°C, 5 minutes and vortex of the Tracer and the small molecules. 2) Preparation of the interaction buffer: 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20, by adding fresh TCEP and Tween every experimental session to the base buffer 50 mM HEPES (pH 7.5), 150 mM NaCl (stored at +4°C). 3) Preparation of three solutions for the titration: (T) or (PT) → Fluorescent molecule: Tracer alone or in complex with the protein. (P) or (L) → Titrating molecule: the protein alone or the inhibitor of interest. 37 (C) → Control solution at the same composition of the titrating molecule solution. NOTE: Every solution has to be prepared at twice the final desired concentration. • BINDING ASSAY BETWEEN EGFR AND THE KINASE TRACER 199 (25 nM or 5 nM) SOLUTION FUNCTION COMPOSITION INITIAL CONCENTRATION (2X) FINAL CONCENTRATION (IN THE ASSAY) (T) FLUORESCENT TRACER 199 50 nM or 10 nM 25 nM or 5 nM (P) TITRANT EGFR 3.76 µM 1.88 µM – 0.45 nM (C) CONTROL GLYCEROL 50 % glycerol 25 % glycerol • COMPETITION ASSAY WITH THE ORTHOSTERIC INHIBITORS SOLUTION FUNCTION COMPOSITION INITIAL CONCENTRATION (2X) FINAL CONCENTRATION (IN THE ASSAY) (PT) FLUORESCENT EGFR + TRACER 199 40 nM + 10 nM 20 nM + 5 nM (L) TITRANT GEFITINIB or LAPATINIB or OSIMERTINIB 200 nM 100 nM – 0.02 nM (C) CONTROL DMSO 0.002 % 0.001 % 4) Preparation of the titration on 8 points, with a dilution factor of 4; the last point is lacking in titrating molecule and is taken as control of the fluorescent molecule/complex thermophoretic mobility. - Number 8 PCR tubes from 1 to 8. - Add an appropriate volume (at least 10 or 15 µL) of (C) to the 2 - 8 tubes and the same volume of (L) or (P) to the tube number 1. - Titrate (L) or (P) from the tube 2 to the tube 7, by using a volume 3-times lower than the (C) volume. Then discard the last volume without adding it to 8. - Add (T) or (PT) to every tube, from 1 to 8, in the same volume used for (C). - Centrifuge the tubes at 15000 g, 22°C, 5 minutes. 5) Loading of the solutions into the capillaries (two capillaries for every concentration point) and set 10 minutes incubation inside the instrument at the desired temperature. 6) The assay is performed applying the following instrument setup. 38 Capillaries type Premium coated Excitation Power 100 %* MST-Power Medium Temperature 25°C *as mentioned in the Introduction paragraph 1.2.2, the choice of using 100 % of the Excitation Power was pondered for every experiment of this work, in order to maximize the yield at low fluorophore concentration, once the maintenance of a good signal-to-noise ratio had been verified. 3.1.4 STOICHIOMETRY ASSAYS The stoichiometry assays are performed in a narrow range of titrant concentration and at high nontitrant concentration, in order to exploit ligand depletion and to obtain stoichiometry information. This type of assay is performed as follow: 1) Thawing of the protein, the Kinase Tracer 199 and the inhibitor of interest; centrifugation of the protein at 5000 rpm, 0°C, 5 minutes and vortex of the Tracer and the small molecules. 2) Preparation of the interaction buffer: 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20, by adding fresh TCEP and Tween every experimental session to the base buffer 50 mM HEPES (pH 7.5), 150 mM NaCl (stored at +4°C). 3) Preparation of three solutions for the titration: (T) or (PT) → Fluorescent molecule: Tracer alone or in complex with the protein. (P) or (L) → Titrating molecule: the protein alone or the inhibitor of interest. (C) → Control solution at the same composition of the titrating molecule solution. NOTE: Every solution has to be prepared at twice the final desired concentration. • STOICHIOMETRY TITRATING THE PROTEIN OVER THE TRACER 199 SOLUTION FUNCTION COMPOSITION INITIAL CONCENTRATION (2X) FINAL CONCENTRATION (IN THE ASSAY) (T) FLUORESCENT TRACER 199 100 nM 50 nM (P) TITRANT EGFR 2 µM 1 µM – 35.18 nM (C) CONTROL GLYCEROL 50 % glycerol 25 % glycerol • STOICHIOMETRY TITRATING THE INHIBITOR OVER THE TRACER 199/EGFR COMPLEX SOLUTION FUNCTION COMPOSITION INITIAL CONCENTRATION (2X) FINAL CONCENTRATION (IN THE ASSAY) (PT) FLUORESCENT EGFR + TRACER 199 280 nM + 10 nM 140 nM + 5 nM (L) TITRANT GEFITINIB or LAPATINIB 200 nM 100 nM – 0.02 nM (C) CONTROL DMSO 0.002 % 0.001 % 39 4) Preparation of the titration on 16 points, with a dilution factor of 1.25: - Number 16 PCR tubes from 1 to 16. - Add an appropriate volume (at least 5 µL) of (C) to the 2 - 16 tubes and the same volume of (L) or (P) to the tube number 1. - Titrate (L) or (P) from the tube 2 to the tube 16, by using a volume 4-times higher than the (C) volume. Then discard the last volume. - Add (T) or (PT) to every tube, from 1 to 16, in the same volume used for (C). - Centrifuge the tubes at 15000 g, 22°C, 5 minutes. 5) Loading of the solutions into the capillaries (one capillary for every concentration point) and set 5 minutes incubation inside the instrument at the desired temperature. 6) The assay is performed applying the following instrument setup. Capillaries type Premium coated Excitation Power 100 % MST-Power Medium Temperature 25°C NOTE: In these experiments the time dependence (10 minutes and 3 hours) has been explored in two ways: 7) Preparation of the solutions (bullet points from 1 to 4) and setting of two timers, 10 minutes and 3 hours. Once past the 10 minutes loading of the first set of capillaries and incubation of further 5 minutes inside the instrument before the analysis. After 3 hours loading of the second set of capillaries from the same solutions, incubation of further 5 minutes inside the instrument before the analysis. 8) Preparation of the (PT) starting solution and setting of two timers, 10 minutes and 3 hours. Proceed with the bullet points from 4 to 6 two times, after 10 minutes and then after 3 hours. 40 3.1.5 REVERSIBLE LIGAND RETENTION ASSAY This assay has been developed as an alternative to the Jump Dilution for the differentiation of reversible and irreversible ligands. The approach is based on the size exclusion retention of small reversible molecules, while the large molecules and every small molecule covalently attached to it will not be retained and will be mechanically separated from the others without strong change in their concentration. The experiments have been performed as follow: 1) Thawing of the protein, the Kinase Tracer 199 and the inhibitor of interest; centrifugation of the protein at 5000 rpm, 0°C, 5 minutes and vortex of the Tracer and the small molecules. 2) Preparation of the interaction buffer: 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20, by adding fresh TCEP and Tween every experimental session to the base buffer 50 mM HEPES (pH 7.5), 150 mM NaCl (stored at +4°C). 3) Preparation of three solutions (at least 320 µL each): (P) → Protein alone with the same concentration of DMSO than the other two solutions. (PG) → Protein in complex with the Gefitinib. (PO) → Protein in complex with the Osimertinib. SOLUTION FUNCTION COMPOSITION CONCENTRATION (P) CONTROL EGFR 20 nM (PG) REVERSIBLE LIGAND EGFR + GEFITINIB 20 nM + 10 nM (PO) IRREVERSIBLE LIGAND EGFR + OSIMERTINIB 20 nM + 10 nM 4) Incubation of the solutions for 60 minutes and 150 minutes. 5) Preliminary Binding Check after every time point and before the Spin Column passages: addition of 0.6 µL Tracer 199 (stock solution 250 nM) to 29.4 µL of every solution in order to obtain 30 µL at Tracer concentration of 5 nM. Centrifugation at 15000 g, 22°C, 5 minutes. Every solution is used to load three capillaries and the capillaries are then incubated for 5 minutes inside the instrument before the analysis. 6) Size exclusion: for each time point 9 size exclusion Zeba™ Spin Desalting Columns, 7K MWCO, 0.5 mL, Thermo are used. Every Zeba™ Spin Desalting Column needs three steps of conditioning with the dilution buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20). After the conditioning, 130 µL of each solution (P), (PG), and (PO), 41 endures a passage in a Zeba Column each and then the eluted solutions undergo two further sequential passages in as many Zeba Columns (three Zeba column are used for every solution). 7) Final Binding Check in the same conditions of point number 5 after every incubation time and passage steps through the Zeba Spin Columns. 8) The assay is performed applying the following instrument setup. Capillaries type Premium coated Excitation Power 100 % MST-Power Medium Temperature 25°C 48 Prism software for the fitted parameters pIC50, Hill Slope, Top and Bottom, other than the S/N (obtained by dividing the Response Amplitude for the Sy.x) and the R2. The MST-ON time were then ranked on the basis of the narrowest interval confidence and goodness of fit to choose the most reproducible time point for that system. The same experimental set has been repeated for every different system of detection such as the covalent labeling of the protein. EGFR COMPETITION For the EGFR system (both wild type and L858R) three protein concentrations (70 nM, 20 nM, 5 nM) have been tested in order to consolidate the robustness of the method. The chosen reference is the Gefitinib, titrated starting from 100 nM. [EGFR] 70 nM, 20 nM, 5 nM [INHIBITOR] 100 nM – 0.02 nM + CONTROL POINTS [TRACER 199] 5 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005 % Tween 20 TEMPERATURE 25°C The binding curves were first analyzed as separated replicates in order to evaluate the macroscopic differences in fitting and parameters. EGFR WILD TYPE - EGFR WT, 70 nM + GEFITINIB 49 From the data collected at 70 nM it can be noticed that, except for the 5 s (coinciding with the crossover of the MST traces), all the MST-ON time yield a comparable pIC50 and Hill Slope information. The late MST-ON times however give a more robust fitting with a higher R2 value. At this high protein concentration, the fraction of the tracer that is bound to the protein is nearly equal to the total amount of the tracer leading to a high signal/noise. In these situations, every MST-on time gives a solid fit. The three replicates of each MST-ON time data set are then individually normalized from 0 to 100 and the obtained values are treated in GraphPad Prism as replicates values in side-by-side sub columns to get reliable confidence intervals (CI, 95%). Every MST-ON time is now ranked from 1 to 6, giving 1 to the time with the narrowest CI 95% for every fitting parameter (pIC50, Hill Slope, Top, Bottom) and the highest S/N and R2 value, and so on up to the worst MST-ON time in terms of confidence intervals and fitting at which is assigned the sixth value. The lowest sum of the ranked positions for every considered parameter gives the MST-ON time that yields the most reproducible information on three identical replicates. 1.5s 2.5s 5s 10s 15s 20s 8.0 8.5 9.0 pIC50 pIC50 1.5s 5s 10s 15s 20s 2.5s 1.5s 2.5s 5s 10s 15s 20s 0 1 2 3 Hill Slope Hill Slope 1.5s 2.5s 5s 10s 15s 20s 1.5s 2.5s 5s 10s 15s 20s 0.7 0.8 0.9 0.97 0.98 0.99 1.00 R2 R2 1.5s 2.5s 5s 10s 15s 20s 50 In this case the best MST-ON time in terms of reproducibility is the 20 s. - EGFR WT, 20 nM + GEFITINIB 51 Also by using 20 nM of protein the best MST-ON time results being the 20 s. The loss of fitting at 5 s is due to the cross-over of the thermophoretic traces. - EGFR WT, 5 nM + GEFITINIB 1.5s 2.5s 5s 10s 15s 20s 9.0 9.5 10.0 pIC50 pIC50 1.5s 5s 10s 15s 20s 2.5s 1.5s 2.5s 5s 10s 15s 20s 0.0 0.5 1.0 1.5 2.0 10 20 30 Hill Slope Hill Slope 1.5s 5s 10s 15s 2.5s 20s 1.5s 2.5s 5s 10s 15s 20s 0.5 0.6 0.7 0.8 0.96 0.98 1.00 R2 R2 1.5s 2.5s 5s 10s 15s 20s 52 By using a low protein concentration it is more evident that choosing the T-Jump region instead of the Thermophoresis region may lead to different conclusions in terms of potency and Hill Slope, especially when the number of replicates for a single conditions are not high. In this case a significant amount of Tracer is free in solution, therefore the noise is high and identifying the more robust MSTON time become relevant to rely the interaction interpretation on trustworthy fitting parameters. By applying the ranking method, the 20 s resulted once again as the best MST-ON time in terms of reproducibility. Therefore, it seems that the protein concentration had a little impact on which MST-ON time yield the most robust fitting parameters, even if at high concentrations, and consequently high signal-tonoise ratio, every time point seems to be able of yielding comparable information. The choice has 1.5s 2.5s 5s 10s 15s 20s 8 9 10 11 pIC50 pIC50 1.5s 5s 10s 15s 20s 2.5s 1.5s 2.5s 5s 10s 15s 20s 0.0 0.5 1.0 1.5 2.0 20 40 60 80 100 Hill Slope Hill Slope 1.5s 5s 10s 15s 20s 2.5s 1.5s 2.5s 5s 10s 15s 20s 0.0 0.1 0.2 0.3 0.75 0.80 0.85 0.90 0.95 1.00 R2 R2 1.5s 5s 10s 15s 20s 2.5s 53 more impact at low protein concentrations, condition that is the one more relevant for the investigation of potent inhibitors. - EGFR WT, 20 nM + LAPATINIB To confirm that the MST-ON time reproducibility was not dependent on the inhibitor, a triplicate with Lapatinib and EGFR WT 20 nM has been performed as well. Only the ranking table is reported below. The 20 s is confirmed to be also in this case the most reproducible MST-ON time for the EGFR wild type interaction system in competition with the Tracer 199. 54 EGFR L858R The mutant form of EGFR was also tested in triplicates with Gefitinib to identify the best MST-ON time in terms of reproducibility. Only the ranking tables are reported below. - EGFR L8583 70 nM + GEFITINIB - EGFR L8583 20 nM + GEFITINIB - EGFR L8583 5 nM + GEFITINIB From the collected data it is possible to conclude that also for the mutated form of EGFR the best MST-ON time for the competition assays with Tracer 199 is the 20 s. 55 ROCK-1 COMPETITION For the ROCK-1 competition system the triplicate has been performed with the following experimental conditions and the inhibitor used for this purpose is the orthosteric Compound A. [ROCK-1] 50 nM [INHIBITOR] 1 µM – 0.24 nM + CONTROL POINTS [TRACER 236] 15 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01 % Tween 20 TEMPERATURE 22°C The ranking table is reported below. On the basis of the reported results, the 10 s was the selected MST-ON time for the ROCK-1 competition experiments. ROCK-1 LABELED The ROCK-1 covalently labeled system has been investigated with the ranking method for both labeling reaction in the presence of the active site protector: #1 and #3. - ROCK-1 LB #1 (w/ active site protection) 56 - ROCK-1 LB #3 (w/ active site protection) It is interesting to note that two different labeling reactions, as far as resulting in two equally competent labeled proteins that yield comparable potency information, still could rely their most robust fitting parameters on distinct MST-ON times. Indeed, in the case of ROCK-1 LB #1 the interaction with the Compound A is better descripted by the 5 s timing, while for the ROCK-1 LB #3 the 2.5 s resulted the best MST-ON time to choose. Therefore, to graphically compare two binding curves originating in the two different labeled systems, it has been employed a normalization from 0 to 100 and the y-axis became indicated as “Fraction Bound”. The ranking test was not performed for the protein labeled in the absence of an active site protector (ROCK-1 LB #2) since the information acquired from the experiment with that labeled protein was used as qualitative indication of the influence of the protector during the labeling phase. As a result, the MST-ON time of 5 s was arbitrarily chosen for the analysis of the binding curves in those conditions. 57 3.4 DATA ELABORATION AND FITTING MODELS The main data elaboration has been performed by using GraphPad Prism 8.1.0 and the binding or competition data set have been fitted by using the Nonlinear regression (curve fit), log(inhibitor) vs response - Variable slope (four parameters) model. Quadratic fitting was performed through Excel Solver Add-In or with the MST analysis software (MO.Affinity Analysis). The Akaike's Information Criterion (AIC)[15], a method developed for comparing models and assisting in determining which model is more likely to be correct, has been used to help decide whether to treat the protein concentration in the quadratic equation as a constant or as an adjustable parameter. The method relies on three parameters: - N: the number of data points - K: the number of parameters fit by nonlinear regression plus 1 (constrained parameters are not counted) - SS: sum-of-square of the nonlinear regression The three values are then combined into the AIC equation: AIC=Nln(SS N)+2K+ 2K(K+1) N−K−1 Once the value has been calculated for all the model to compare, the AIC deriving from the different models is observed: the model having the lower AIC score is more likely to be correct. By considering the difference between two AIC scores (ΔAIC), it is possible to calculate the Evidence Ratio, a parameter indicating how many times one model is likely to be correct in comparison with the other: Evidence Ratio= 1 e−0.5 ΔAIC It is important to note that this method is not a statistical approach, it is an indication of which model is more likely to be correct and how much more likely. Therefore, in this work it has been considered as side support to better investigate the dependence of potency from the fixed parameter of protein concentration. 64 The experimental conditions for this assay are reported in the following table. [EGFR] 1 µM – 35.18 nM [TRACER 199] 50 nM TITRATION 4:1, 16 POINTS DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20 TEMPERATURE 25°C Figure 18 Stoichiometry assay in which EGFR WT is titrated from 1 µM with a dilution factor of 1.25 over a fixed concentration of Tracer 199 (50 nM). The breaking point indicated by the arrow is the concentration of EGFR WT sites capable of binding 50 nM of Tracer. The system clearly shows two different trends of points and the intersection of the two linear regression is the breaking point (Figure 18). The parameters of the linear regressions are indicated in the table below. LINEAR REGR. #1 LINEAR REGR. #2 Breaking Point (nM) Slope -0.087 0.0087 390.77 Y-intercept 762 724.4 R2 0.96 0.64 The nominal concentration of tracer is fixed at 50 nM, therefore if the breaking point appears at 390.77 nM the percentage of the protein that is able to bind the Tracer is the 12.8% of the total amount of the protein. 0 500 1000 720 730 740 750 760 770 STOICHIOMETRY - EGFR WT EGFR WT conc (nM) FNorm 20s 390.77 nM 65 % active sites (EGFR WT)= 50 nM 390.77 nM100=12.8 % The same experimental set has been applied to the EGFR L858R: Figure 19 Stoichiometry assay in which EGFR L858R is titrated from 1 µM with a dilution factor of 1.25 over a fixed concentration of Tracer 199 (50 nM). The breaking point indicated by the arrow is the concentration of EGFR WT sites capable of binding 50 nM of Tracer. In this case two distinct breaking points are registered: one at 135.18 nM and the second at 642.44 nM (Figure 19). It is interesting to notice that the non-changing Tracer mobility values are in between the two breaking points; this suggests that the change in mobility at higher protein concentrations may be due to a massive variation in some of the protein physicochemical features. A possibility is that in these conditions MST registered protein oligomerization. Indeed, the Tracer mobility at protein concentrations above 642.44 nM could be a combination of the Tracer completely bound to the monomer and the Tracer bound to the dimer, a form that increases with protein concentration. As a result, the second breaking point might be the concentration at which the system begins to be populated by that new type of interaction form (Figure 20). 0 500 1000 740 760 780 800 820 840 STOICHIOMETRY - EGFR L858R EGFR L858R conc nM FNorm 20s LINEAR REGR. #1 LINEAR REGR. #2 LINEAR REGR. #3 Breaking Point #1 (nM) Breaking Point #2 (nM) Slope -0.21 0.018 0.19 135.18 642.44 Y-intercept 777.8 746.5 633.2 R2 0.95 0.67 0.99 135.18 nM 642.44 nM 66 This hypothesis is consistent with the consolidated knowledge about the higher tendency of L858R to dimerization[41] and with the observation made during the Tracer affinity determination experiment for EGFR L858R, in which the highest concentration points exhibited anomalous behavior in comparison to the sigmoidal trend of the lower concentration points, indicating that above 500 nM the thermophoretic mobility begins to be governed by events beyond the simple Tracer-protein interaction. Figure 20 Interpretation of the stoichiometry assay with EGFR L858R: the first linear regression (orange)is related to the combination of Tracer mobility alone and bound to the monomer; the second linear regression (blue) is related to the mobility of the Tracer completely bound to the monomer; the third linear regression (pink) could be related to the coexistence of Tracer mobility both bound to EGFR monomer and dimer. Regarding the active fraction of the protein, also in this case the nominal concentration of the Tracer is fixed at 50 nM, therefore the breaking point that appears at 135.18 nM should be the percentage of the protein in monomeric form that is able to bind the Tracer. The value is the 37 % of the total amount of the protein. % active sites (EGFR L858R)= 50 nM 135.18 nM100=37 % These results indicate that the two isoforms of EGFR, wild type and L858R, exhibit two different percentage of active sites and two different tendency to dimerization. 0 500 1000 740 760 780 800 820 840 STOICHIOMETRY - EGFR L858R EGFR L858R conc nM FNorm 20s Tracer bound to the monomer + Tracer bound to the dimer Tracer bound to the monomer Tracer bound to the monomer + Tracer free 67 % active site (able to bind the Tracer 199) EGFR WT 12.8 % EGFR L858R 37.0 % As a result, the analysis of the direct binding data between the two isoforms of the EGFR and the Tracer 199 needs an adjustment in terms of protein concentration (x-axis). EGFR WT (nM nominal conc) 12.8 % active sites (nM) Log10 (M) 1880 240.64 -6.6 470 60.16 -7.2 117.5 15.04 -7.8 29.4 3.8 -8.4 7.3 0.9 -9.0 1.8 0.2 -9.6 0.5 0.06 -10.2 0.001* 0.0001* -12 *fictional value assigned to the zero concentration to obtain a real Log10 value EGFR WT ACTIVE FRACTION ADJUSTMENT (12.8%) T199 25 nM T199 5 nM LOGISTIC FITTING (four parameters) QUADRATIC FITTING LOGISTIC FITTING (four parameters) QUADRATIC FITTING pKD 7.97 8.01 7.99 8.12 KD (nM) 10.7 9.9 10.2 7.6 R2 0.99 0.99 0.96 0.96 HILL COEFF 0.85 / 1.34 / RESPONSE AMPLITUDE 58.04 54.89 53.45 55.75 MST-on TIME 20 20 20 20 fitting performed by letting “float” the Tracer concentration as adjustable parameter. By considering the real fraction of the protein competent for the Tracer binding the pKD of EGFR WT increases by an order of magnitude. Even in this case both the logistic and the quadratic fitting yields quite the same information: at 25 nM the system is probably in tight binding condition but the fact that the [Tracer]/2 value is 12.5 nM, thus close to the real affinity of the Tracer, might mask the affinity dependence for the Tracer concentration. EGFR L858R (nM nominal conc) 37 % active sites (nM) Log10 (M) 2160 799.2 -6.1 540 199.8 -6.7 135 50.0 -7.3 33.8 12.5 -7.9 8.4 3.1 -8.5 2.1 0.8 -9.1 0.5 0.2 -9.7 0.001* 0.0004* -12 *fictional value assigned to the zero concentration to obtain a real Log10 value 68 EGFR L858R ACTIVE FRACTION ADJUSTMENT (37%) T199 25 nM T199 5 nM LOGISTIC FITTING (four parameters) QUADRATIC FITTING LOGISTIC FITTING (four parameters) QUADRATIC FITTING pKD 7.72 8.20 8.55 8.65 KD (nM) 18.9 6.3 2.8 2.2 R2 0.87 0.85 0.94 0.94 HILL COEFF 1.37 / 0.76 / RESPONSE AMPLITUDE 33.15 32.67 21.51 19.09 MST-on TIME 20 20 20 20 fitting performed by letting “float” the Tracer concentration as adjustable parameter. In the case of EGFR L858R the affinity gain is less evident because the active fraction of the protein is larger and thus already closer to its nominal concentration. Observing the results at the lower Tracer concentration, 5 nM, both the logistic and quadratic fitting give similar information, most likely for the same reason as the wild-type experiment at 25 nM: the real affinity would be most probably close the [Tracer]/2 value, so, despite the tight binding conditions, a reliable fitting with the logistic fitting is still possible. Regarding the [Tracer] = 25 nM experiments, the quadratic fitting highlights the tight binding conditions that the logistic fitting partially masked (with the exception of the high Hill coefficient). In Figure 21 and 22 it is possible to observe the pKD values and their changes by considering the nominal protein concentration or its active fraction. NOMINAL PROTEIN CONCENTRATION [TRACER] EGFR WT (nominal conc.) EGFR L858R (nominal conc.) 25 nM 5 nM Figure 21 Comparison between the pKD (obtained considering the nominal concentration of the protein) at 25 nM and 5 nM of the Tracer. In light yellow the value obtained by the logistic fitting and in light purple the quadratic. LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR WT pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR L858R pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR WT pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR L858R pKd 69 ACTIVE FRACTION ADJUSTMENT [TRACER] EGFR WT (12.8%) EGFR L858R (37%) 25 nM 5 nM Figure 22 Comparison between the pKD obtained (considering the adjusted protein concentration on the basis of the stoichiometry assay) at 25 nM and 5 nM of the Tracer. In light yellow the value obtained by the logistic fitting and in light purple the quadratic. In summary, from this MST characterization emerged that EGFR wild type has less active sites that can bind Tracer 199 than EGFR L858R; even when this difference is taken into consideration in the data interpretation of the direct binding between the Tracer and the proteins, the EGFR L858R seems to have a higher affinity for the tracer than the EGFR WT, a condition that is more evident at lower Tracer concentrations. However, it was observed from these data that, in some cases, quadratic fitting yielded meaningless and indeterminate parameters, so the data resulting from those fitting (and reported in the tables as blue italic font) was obtained by treating the Tracer concentration as a floating parameter. The reason of this indetermination could be a non-sufficiently precise determination of the active fraction or the fact that the system has reached a level of complexity where a quadratic fitting is insufficient to reliably describe the interaction. A more complex fitting model, such as a cubic equation, may be required in these cases. The following competition assays with potent inhibitors will introduce a further element of complexity and a possible double tight binding condition will occur leading to an even more difficult parameter determination. Therefore, for the following paragraph (4.1.2 COMPETITION ASSAYS), the Akaike's Information Criterion (AIC) has been applied to help decide whether to treat the protein concentration in the quadratic equation as a constant or as an adjustable parameter and the values LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR WT pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR L858R pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR WT pKd LOGISTIC QUADRATIC 7.0 7.5 8.0 8.5 9.0 EGFR L858R pKd 70 reported in the results tables will take into account this information. Further considerations on the interaction system interpretation will follow at the end of the aforementioned paragraph. EGFR AND TRACER 199 TIME DEPENDENCE Another aspect evaluated was the change in Tracer thermophoretic mobility over time. The assay was performed by pre-incubating for three hours an appropriate volume of three solutions: - Tracer 199 alone 5 nM - Tracer 199 5 nM + EGFR WT 20 nM - Tracer 199 5 nM + EGFR L858R 20 nM During this time four capillaries for every condition were loaded and analyzed at every selected time point (10, 40, 60, 90, 120, 150 and 180 minutes). As usual, the thermophoretic mobility on the yaxis is referred to the Tracer mobility. Figure 23 Time dependence of the complex mobility protein-Tracer. In orange is reported the thermophoretic mobility of Tracer 199 alone during time as control. In green is reported the mobility over time of the complex EGFR WT + Tracer 199 and in pink the mobility over time of the complex EGFR L858R + Tracer 199. In Figure 23 it is evident that, also in this case, EGFR WT and EGFR L858R show different behaviors. Since the Tracer mobility in the bound state gets closer to the Tracer alone mobility, it is possible to assume the wild type undergoes some conformational change that either could cause: - A loss in Tracer binding. - A change in the complex mobility that casually approaches the Tracer alone. 0 760 780 800 820 PRE-INCUBATION: PROTEIN + T199 min FNorm 20s T199 T199 + WT T199 + L858R 10 40 60 90 120 150 180 71 In both cases the thermophoretic mobility reaches a steady state after two hours of pre-incubation. EGFR L858R also undergoes a conformational change resulting in thermophoretic mobility that is similar to Tracer alone, but the process is much slower. To determine if the conformational change was caused solely by the presence of the Tracer, the same experiment was repeated with the pre-incubation of the protein alone and the Tracer addition right before loading the capillaries at each time point. Figure 24 Time dependence of the complex mobility protein-Tracer prior incubation of the protein alone. In orange is reported the thermophoretic mobility of Tracer 199 alone during time as control. In green is reported the mobility over time of the complex EGFR WT + Tracer 199 and in pink the mobility over time of the complex EGFR L858R + Tracer 199. In Figure 24 it is evident that, even in the absence of Tracer 199 during the pre-incubation phase, EGFR WT undergoes anyway a conformational modification that brings Tracer-protein complex mobility even closer to mobility of Tracer alone. The EGFR L858R seems to undergo the same process, albeit at a slower and less pronounced rate. Therefore, Tracer 199 presence in the system is not the direct reason of the conformational change. However, in both cases, time plays an important role in the conformational changes of the two EGFR isoforms and must be considered during competition assays with orthosteric inhibitors or other experiments aimed at describing the time dependence of a specific inhibitor. Furthermore, it is important to emphasize that the Response Amplitude of a competition assay recorded at longer time points (> 120 minutes) will tend to decrease; indeed, the unbound points correspond to Tracer-protein mobility, whereas the high inhibitor concentration points correspond to Tracer alone mobility. This could be problematic in any assay that suffers from low Response Amplitude from the beginning. 0 760 780 800 820 PRE-INCUBATION: PROTEIN ALONE min FNorm 20s T199 T199 + WT 10 40 60 90 120 150 180 T199 + L858R The absence of Tracer 199 during the incubation led to a more evident change in the complex mobility leading it closer to the Tracer alone mobility 72 4.1.2 COMPETITION ASSAYS The Tracer 199 – EGFR interaction system is now described and it is possible to benefit of this information to approach the characterization of three EGFR inhibitors having different binding modes: a. Gefitinib: type I, reversible potent binder b. Lapatinib: type II, reversible slow binder c. Osimertinib, irreversible binder As previously mentioned in paragraph 4.1, the common feature of these inhibitors is to be orthosteric and to compete with the ATP, hence with the Tracer 199, for the binding to the ATP binding pocket. In a competition assay the protein and Tracer are fixed at a convenient concentration that enables to minimize fluorescence noise caused by free Tracer in solution. The orthosteric inhibitor is then titrated from a saturating concentration and competes with the Tracer for protein binding. The registered thermophoretic mobility will vary from the Tracer signal when fully associated to the protein and the signal related to the free Tracer in solution when the inhibitor concentration allows its complete displacement. GEFITINIB - TIGHT BINDING CONDITIONS Gefitinib is a reversible Type I inhibitor that is EGFR L858R selective[40][42][43]; therefore, the aim of this section was to investigate the Gefitinib-EGFR system with both the wild type and the mutated form to better understand how MST can describe highly potent interactions and if a good experimental design could highlight a difference in Gefitinib L858R selectivity over the wild type. A critical aspect to take into consideration during the design of a competition assay with a potent inhibitor is the corresponding affinities of Tracer and the inhibitor itself. To reduce the noise produced by free Tracer in solution, the protein and Tracer concentrations may be fixed excessively high, resulting in a tight binding condition as soon the potent inhibitor is added to the system. On the other side, it is possible that by decreasing too much the protein/Tracer concentrations the fluorescent difference between the bound and the unbound state will be hardly noticeable, leading to an uncertain situation. 73 A valuable strategy could be to perform a preliminary competition experiment by using high protein/Tracer concentration which certainly will results in a good signal-to-noise ratio. Once the binding of the inhibitor is confirmed is possible to decrease the protein/Tracer concentration and to observe the relative pIC50: if the system is in tight binding conditions the pIC50 will vary with the protein concentration. It is then possible to decrease the protein concentration until the stabilization of the pIC50 is reached or either when the noise becomes too high to obtain reliable information. In both cases the quadratic fitting could help to better interpret the competition experiments and to highlight an affinity value near to the real one. The Figure 25 reports a decision scheme that is possible to apply when studying a potent compound. Figure 25 Decision scheme to help in approaching a potent compound to avoid tight binding conditions. 80 LAPATINIB - SLOW BINDING TYPE II INHIBITOR Lapatinib has been chosen for the investigation, through MST, of time dependent Type II inhibition in EGFR wild type system. Since Lapatinib is known to have a time-dependence mechanism for the interaction with EGFR[42][44], its interaction during time has been monitored and the selected time points for the analysis were 10, 40, 75, 120 and 180 minutes. Moreover, a further 2.5-hour pre-incubation of the complex protein-Tracer alone was done before adding the inhibitor, based on previously registered protein-Tracer changes in thermophoretic mobility over time. The interaction was then examined for two hours after the inhibitor was added. In summary, the two experiments were performed as follow: 1. No preincubation: preparation of the titration (EGFR WT, Tracer199, inhibitor) and capillary loading of the same solution at the defined time points of 10, 40, 75, 120 and 180 minutes. 2. Preincubation (2.5h): preparation of the solution EGFR WT-Tracer 199 and preincubation for 2.5h. Then addition of the titrated inhibitor and capillary loading of the same solution at the defined time points of 10, 40, 75, and 120 minutes. The goal of the two analyses was to observe the interaction change over time while the proteinTracer equilibrium was changing, and then to compare the interaction time dependency after the protein-Tracer equilibrium was reached. Gefitinib was also tested in the same conditions as reference. [EGFR WT] 20 nM [INHIBITOR] 100 nM – 0.02 nM + CONTROL POINTS [TRACER 199] 5 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20 TEMPERATURE 25°C INTERACTION TIME (no preincubation) 10’, 40’, 75’, 120’, 180’ INTERACTION TIME (after 2.5h preincubation) 10’, 40’, 75’, 120’ 81 NO PREINCUBATION GEFITINIB LAPATINIB pIC50 IC50 (nM) Hill slope R2 pIC50 IC50 (nM) Hill slope R2 10’ 9.47 0.3 1.18 0.99 8.62 2.4 2.15 0.99 40’ 9.46 0.4 1.36 0.98 8.57 2.7 2.15 0.99 75’ 9.48 0.3 1.25 0.98 8.45 3.5 2.64 0.98 120’ 9.60 0.3 1.52 0.97 8.54 2.9 1.80 0.96 180’ 9.61 0.2 1.03 0.96 8.48 3.3 2.33 0.96 2.5H PREINCUBATION (PROTEIN-TRACER) GEFITINIB LAPATINIB pIC50 IC50 (nM) Hill slope R2 pIC50 IC50 (nM) Hill slope R2 10’ 9.46 0.3 1.59 0.99 8.99 1.0 1.29 0.99 40’ 9.47 0.3 1.13 0.97 8.98 1.0 1.90 0.98 75’ 9.43 0.4 1.10 0.96 9.05 0.9 2.10 0.98 120’ 9.40 0.4 1.08 0.95 8.95 1.1 2.13 0.97 -14 -12 -10 -8 -6 780 790 800 810 820 830 10' Log[inhibitor] FNorm 20s Gefitinib Lapatinib -14 -12 -10 -8 -6 780 790 800 810 820 830 40' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 780 790 800 810 820 830 75' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 790 795 800 805 810 815 820 120' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 795 800 805 810 815 820 180' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 800 810 820 830 10' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 800 810 820 830 840 40' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 800 810 820 830 75' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib -14 -12 -10 -8 -6 805 810 815 820 825 830 120' Log[inhibitor] FNorm 20s Gefitinib Lapatinib Osimertinib 10’ 40’ 75’ 120’ 180’ 10’ 40’ 75’ 120’ 82 By considering the "no preincubation" analysis, the pIC50 value for Gefitinib (light blue) and Lapatinib (yellow) does not change significantly over time. Interestingly, allowing the system protein-Tracer 199 to equilibrate for 2.5 hours prior to the addition of the inhibitors revealed a significant difference between the two compounds: whereas Gefitinib kept its pIC50 value constant, Lapatinib showed a 0.5 logarithm increase in potency. This new pIC50, however, is maintained in the time dependency analysis that follows. By observing the intermediate concentration of 1.5 nM over time (red circle in Figure 29 and Figure 30), this difference becomes more evident: Figure 29 Comparison between competition curves obtained with Gefitinib (blue) and Lapatinib (yellow). Up the condition without preincubation while down the condition with 2.5h of preincubation between protein and Tracer. In the red circles is indicated the 1.5 nM concentration. Figure 30 Focus on the 1.5 nM concentration and its contribution to the change in thermophoretic mobility of the Tracer over time. 83 As a result, the increased potency over time appears to be attributed to the slow conformational change of the protein alone over time, rather than the addition of the inhibitor itself, which appears to stabilize the system in whatever state it would be in when added. This observation is consistent with the literature evidence indicating the slow mechanism of Lapatinib as a result of the time required for EGFR conformational change to the inactive state[34][36][41]. The second explored aspect was the difference in binding mode of Gefitinib and Lapatinib. Gefitinib is known to bind EGFR in its active state (Type I inhibitor), while Lapatinib is a Type II inhibitor, binding the inactive form of the protein[42][57]. Thus, the goal of the following experimental section was to understand if MicroScale Thermophoresis could differentiate the binding of Type I and Type II inhibitors. To do so, stoichiometry experiments were performed to highlight differences of Gefitinib and Lapatinib in Tracer 199 displacement. Differently from the stoichiometry experiment used to identify the active fraction of the protein, in which EGFR was titrated over a fixed Tracer concentration, the titrating agent in this case is the inhibitor (Gefitinib or Lapatinib), and the proteinTracer complex is fixed at a concentration that guarantees tight binding conditions. With this experimental setup the breaking points will highlight the inhibitor concentration capable of displace Tracer 199. [EGFR WT] 140 nM [TRACER 199] 5 nM [INHIBITOR] 100 nM – 3.5 nM TITRATION 4:1, 16 POINTS DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20 TEMPERATURE 25°C 84 Figure 31 Stoichiometry assay in which Gefitinib (blue) or Lapatinib (yellow) are titrated over a fixed concentration EGFR WT and Tracer 199. GEFITINIB LINEAR REGR. #1 LINEAR REGR. #2 Breaking Point (nM) Slope 9.12 0.095 12.75 Y-intercept 693.4 808.4 R2 0.96 0.71 LAPATINIB LINEAR REGR. #1 LINEAR REGR. #2 LINEAR REGR. #3 Breaking Point #1 (nM) Breaking Point #2 (nM) Slope 0.57 4.47 0.29 14.18 35.07 Y-intercept 700.4 645.1 791.7 R2 0.47 0.96 0.87 Interestingly the data show a different behavior of Gefitinib and Lapatinib (Figure 31). By considering the high inhibitor concentrations, both drugs demonstrate the ability to displace Tracer 199, indeed the Tracer mobility does not change, settling its value around 800 units of normalized fluorescence. Then, by decreasing the inhibitor concentration, the two compounds start to differentiate: Gefitinib allows the binding of the first small amount of Tracer 199 at 12.75 nM, while Lapatinib at 35.07 nM. Lapatinib then reaches the concentration of 14.18 nM at which Tracer 199 is 0 10 20 30 40 50 60 70 80 90 100 650 700 750 800 850 STOICHIOMETRY GEFITINIB vs LAPATINIB Inhibitor concentration (nM) Tracer Mobility FNorm 20s Gefitinib Lapatinib 12.75 nM 35.07 nM 14.18 nM 85 not displaced anymore, leading to a second plateau. This can be explained by considering the presence of different protein states: - By titrating Gefitinib, the breaking point at 12.75 nM corresponds to the inhibitor concentration necessary to bind all the sites that Gefitinib shares with Tracer 199. Under that concentration a sufficient amount of sites are available for the Tracer binding. This fraction amounted to 9 %. % Gefitinib(like) sites= 12.75 nM 140 nM 100=9 % - In the Lapatinib experiment, the lowest concentrations of inhibitor up to 14.18 nM resulted non capable of Tracer displacing, then Lapatinib starts to bind the same sites of Tracer 199 displacing it. Thus, since the system is in tight binding conditions and Lapatinib must be sequestered and depleted even at low concentrations, under 14.18 nM the Lapatinib molecules must be bonded at a different site population than Tracer 199. As a result, at concentrations below 14.18 nM Tracer 199 is completely bound to its sites (plateau at 700 FNorm), whereas Lapatinib is bound to a not visible protein fraction that Tracer 199 does not have the ability to bind: the 10 % of the total protein amount. % Lapatinib(like)sites= 14.18 nM 140 nM 100=10 % The first breaking point corresponds to the concentration of Lapatinib that saturates the first population (Lapatinib-like sites); above that, Lapatinib begins to bind the Tracer-shared sites (probably the active conformation), for which the inhibitor has a lesser tendency to bind, preferring to saturate the others first. Then, the second breaking point at 35 nM is the Lapatinib concentration that does not allow Tracer binding anymore: thus, it corresponds to the total amount of Lapatinib sites (the sum of Lapatinib-like and Tracer-like). This fraction is around 25 % of the total amount. % Global Lapatinib sites= 35.07 nM 140 nM 100=25 % Finally, the residual amount, i.e. 75 %, is referred to the protein fraction not capable of binding Tracer 199 and not highlighted by Lapatinib binding. In this fraction can be listed the protein amount denatured and all the protein molecule in a conformational state not able to bind Tracer 199 and thus not visible. 86 Therefore, some assumption can be made: - Lapatinib binds two different protein states (it is possible to speculate the active and the inactive state) one of which is shared with Tracer 199 (otherwise displacement could not be possible); however Lapatinib prefers to bind the inactive state at low concentrations and starts to occupy active sites only after the saturation of the inactive form. - Tracer 199 binds only one of the two states highlighted by Lapatinib (presumably the active state), otherwise the plateau state at low Lapatinib concentration would not be visible. - Gefitinib binds first all the Tracer sites but it is not possible to know by these experiments if it occupies also other sites with a lower affinity once the Tracer sites are completely saturated. - The higher protein fraction is however constituted by protein non-active in terms of Tracer 199 binding. These considerations are summarized in Figure 32. Figure 32 Representation of the different EGFR population highlighted from Gefitinib and Lapatinib stoichiometry assays. For Lapatinib graph (on the right) the shadowed section is referred to the fraction obtained for difference from the Global and the Lapatinib-like sites. It is reasonable to assume that this fraction overlaps with the fraction identified by Gefitinib stoichiometry assay (on the left). In conclusions, MST allowed the differentiation of Type I and Type II inhibitors, by highlighting the presence of multiple conformational states and the inhibitors different propensity to bind them. % GEFITINIB-LIKE SITES % LAPATINIB-LIKE SITES 9 % GEFITINIB-LIKE 91 % OTHER 75 % OTHER 25 % GLOBAL 10 % LAPATINIB-LIKE 87 OSIMERITINIB - IRREVERSIBLE BINDING As discussed in the paragraph 1.1.6 - MODES OF INHIBITOR INTERACTIONS - COVALENT BINDING, the irreversible mechanism is usually described from a kinetic point of view, with the kinact/Ki constant, a potency indication that takes into consideration the inhibitor affinity (Ki) and the irreversible inactivation rate (kinact). By performing a classical binding or competition assay, is still possible to obtain a sigmoidal curve and an apparent IC50 value that, in this case, is dependent on the incubation time. Indeed, a proper potency measurement relies on specific kinetic assays, but that does not differ much from a reversible inhibitor binding curve in terms of shape and fitting. As it can be observed in Figure 33 (in which the same sample preparations have been loaded and tested in MST at 10’ and after three hours of incubation at room temperature), the difference in binding mechanisms of Gefitinib (reversible) and Osimertinib (irreversible) emerges only by observing their potency over time, not in any particular changes in their sigmoidal shape. [EGFR] 20 nM [INHIBITOR] 100 nM – 0.02 nM + CONTROL POINTS [TRACER 199] 5 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM TCEP, 0.005% Tween 20 TEMPERATURE 25°C Figure 33 Competition curves with Gefitinib (left) and Osimertinib (right) monitored over time at 10’ and 180’. GEFITINIB OSIMERTINIB pIC50 IC50 (nM) Hill slope R2 pIC50 IC50 (nM) Hill slope R2 10’ 9.5 0.3 1.33 1 8.7 2.2 0.80 1 180’ 9.7 0.2 1.33 0.98 9.2 0.7 0.78 0.97 Therefore, an important aspect to evaluate when dealing with unknown potentially covalent inhibitors, is to find a way to test the reversibility of the binding. This feature is usually investigated -14 -12 -10 -8 -6 0 50 100 GEFITINIB Log[inhibitor] M Fraction Bound 10' 180' -14 -12 -10 -8 -6 0 50 100 OSIMERTINIB Log[inhibitor] M Fraction Bound 10' 180' 88 with enzymatic assays which indicate the activity recovery of a target after the incubation with the inhibitor of interest and the subsequently rapid and massive dilution of the protein-inhibitor complex (Jump-Dilution)[17]. Nevertheless, this type of approach is challenging to perform in a competition setting like the one that has been applied so far. The main issues involve the protein and inhibitor concentrations and the high potency of the inhibitor. Indeed, the typical protocol would require a pre-incubation of the protein at a concentration 100-fold higher than the concentration used in the following competition assay and the inhibitor at a concentration 10 times higher than its IC50. By considering the EGFR WT system and the Gefitinib as the reference compound, the initial conditions should be: Initial conc. Final conc. [EGFR WT] 2 µM 100x the final conc. 20 nM [Gefitinib] 3 nM 10x its IC50 (0.3 nM) 0.03 nM [Tracer 199] / 5 nM Good final concentrations but not ideal initial concentrations due to the massive protein/inhibitor ratio. It is clear that in these experimental conditions (protein:inhibitor > 600-fold) the bound fraction of the protein can be neglected also in the initial state and with every type of inhibitor making the measurement incapable of distinguishing between reversible and irreversible interaction: Tracer 199 would be able to bind its small amount of protein even in the presence of Gefitinib. This problem would be overcome with a less potent inhibitor for which the 10-fold IC50 initial concentration might be near to the protein concentration. On the other hand, by decreasing the protein concentration at nanomolar values to allow Gefitinib binding at a significant number of sites (e.g. 40 nM, at which the active sites will be 5 nM, based on the 12.8% active fraction), the final concentration after a 100-fold jump dilution would be too small (0.4 nM), and the addition of Tracer 199 as fluorescent marker at its consolidated concentration of 5 nM would completely cover the complex signal. Moreover, decreasing Tracer 199 concentration below 5 nM would result in a poor fluorescence signal. Initial conc. Final conc. [EGFR WT] 40 nM 100x the final conc. 0.4 nM [Gefitinib] 3 nM 10x its IC50 (0.3 nM) 0.03 nM [Tracer 199] / 5 nM Good initial concentrations but not ideal final concentrations due to the low [EGFR WT] related to the [Tracer 199]. 89 For these reasons in the context of the present work an alternative method has been developed for the differentiation of reversible and irreversible inhibitors in the EGFR wild type system. This method, summarized in Figure 34, involves three phases: 1. Pre-incubation at two different times, 60 and 150 minutes, of the protein and the inhibitor (reversible or irreversible) at a low concentration that still allows the saturation of all the protein active sites. The system is then preliminary tested in a Binding Check experiment by adding 5 nM of the Tracer 199 to confirm the complete saturation of the protein. 2. Purification of the complex through a size exclusion spin column (Zeba™ Spin Desalting Columns, 7K MWCO, 0.5 mL, Thermo), which should retain every molecule smaller than 7 kDa not covalently bound to the protein. 3. A second Binding Check experiment with the addition of the Tracer 199 5 nM after the purification, in order to identify which inhibitors have been reversibly removed from the system and now allow the binding of the Tracer to the protein. Figure 34 Scheme of the developed method for the differentiation of reversible and irreversible ligands. For the setup of the experiment two inhibitor concentrations have been tested, 100 nM and 10 nM to confirm in a single concentration experiment that both were saturating concentrations. After the protein and inhibitor pre-incubation of 10 minutes, 5 nM of Tracer 199 were added. In the graph every dot corresponds to a single capillary scanned into the instrument and the y-axis is always referred to as the Tracer mobility that in every condition is added at a concentration of 5 nM. As it can be seen in Figure 35, when the Gefitinib is added to the system at both 100 nM (pink dots) and 10 nM (light blue dots) the response is the same, the Tracer cannot bind the protein; so [EGFR WT] 20 nM [Gefitinib] 100 nM or 10 nM 10’ pre-incubation of protein-inhibitor [Tracer 199] 5 nM 96 ALLOSTERIC INTERACTION Figure 41 With allosteric compounds a discrepancy between Competition and Direct Binding assays could occur, due to the missing ATP fluorescent displacement during the Competition Assay. 4.2.1 COMPETITION ASSAYS INTERACTION WITH KINASE TRACER 236 The interaction between ROCK-1 and its fluorescent ATP analog (Kinase Tracer 236, Thermo) has also been investigated in this case. In analogy with the EGFR system, two Tracer concentrations have been tested, 25 nM and 5 nM, and the results are reported in Figure 42. [ROCK-1] 15 µM – 3.6 nM + CONTROL POINTS [TRACER 236] 25 nM / 5 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01% Tween 20 TEMPERATURE 22°C Figure 42 Binding assay between ROCK-1 and T236 at 25 nM (on the left) and 5 nM (on the right). -14 -12 -10 -8 -6 -4 800 810 820 830 Tracer 236 25 nM Log[ROCK-1] (M) FNorm 10s T236 25nM T236 5nM -14 -12 -10 -8 -6 -4 815 820 825 830 835 Tracer 236 5 nM Log[ROCK-1] (M) FNorm 10s T236 25nM T236 5nM 97 ROCK-1 T236 25 nM T236 5 nM LOGISTIC FITTING (four parameters) QUADRATIC FITTING LOGISTIC FITTING (four parameters) QUADRATIC FITTING pKD 6.44 6.45 6.06 6.22 KD (nM) 359.4 352.0 878.11 598.27 R2 0.99 0.99 0.97 0.97 HILL COEFF 1.05 / 0.73 / RESPONSE AMPLITUDE 22.31 22.52 13.68 11.85 MST-on TIME 10 s 10 s 10 s 10 s In this case, the quadratic and the logistic fitting at 25 nM yield the same affinity values, well above the nominal Tracer concentration, suggesting that the system should not be in tight binding conditions. By decreasing the Tracer concentration to 5 nM, the affinity value does not increase, confirming that the system is not in tight binding conditions and the logistic fitting is sufficient to describe the interaction. OPTIMIZATION OF THE INTERACTION WITH COMPOUND A Three protein doses were tested to better characterize the interaction between ROCK-1 and the orthosteric Compound A: 200 nM, 50 nM, and 20 nM (Figure 43). The experimental conditions are detailed in the table below. The Tracer concentration of 15 nM was chosen since 5 nM tried in the preliminary testing produced an insufficient fluorescence signal. [ROCK-1] 200 nM, 50 nM, 20 nM [INHIBITOR] 1 µM – 0.24 nM + CONTROL POINTS [TRACER 236] 15 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01% Tween 20 TEMPERATURE 22°C 98 Figure 43 Competition assays with ROCK-1 tested at three concentrations (200 nM, 50 nM and 20 nM), Tracer 236 and Compound A. ROCK-1 (nominal conc.) LOGISTIC FITTING (four parameters) QUADRATIC FITTING pIC50 IC50 (nM) Hill slope R2 pIC50 IC50 (nM) R2 200 nM 7.11 77.3 1.81 0.97 7.02 94.5 0.96 50 nM 8.22 6.0 0.75 0.95 8.64 2.3 0.93 20 nM 7.92 12.0 0.37 0.77 7.79 16.3 0.69 When the protein concentration is reduced from 200 nM to 20 nM, the noise increases due to the increased concentration of free Tracer 236 in solution. Even with the best R2 value, the 200 nM condition has a Hill slope greater than one, indicating that the experimental conditions are not ideal for describing the system: the high protein concentration likely causes inhibitor depletion, which is reflected in the visible amount competing with the Tracer, however small it may be due to the tracer low affinity for the protein. As evidence, decreasing the protein concentration changes the pIC50 value of Compound A. In this case, 50 nM is probably the best compromise between a good fit and reliable information on potency, but due to the low affinity of the Tracer 236 for the protein, it is important to consider that every potent inhibitor might be underestimated in terms of pIC50. Therefore, as previously mentioned, in these conditions, the competition assay could be solely applied to obtain qualitative information about interaction, relying on the direct binding assay to obtain quantitative information. -14 -12 -10 -8 -6 -4 805 810 815 820 825 ROCK1 200 nM / T236 15 nM Log[Compound A] (M) FNorm 10s -14 -12 -10 -8 -6 -4 810 812 814 816 818 820 ROCK1 50 nM / T236 15nM Log[Compound A] (M) FNorm 10s -14 -12 -10 -8 -6 -4 805 806 807 808 809 810 ROCK1 20 nM / T236 15nM Log[Compound A] (M) FNorm 10s 99 COMPETITION WITH ALLOSTERIC COMPOUNDS Once the experimental conditions for the competition assay are chosen, the allosteric inhibitors (Compound B and Compound C) have been tested (Figure 44). [ROCK-1] 50 nM [INHIBITOR] 1 µM – 0.24 nM + CONTROL POINTS [TRACER 236] 15 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01% Tween 20 TEMPERATURE 22°C Figure 44 Competition assay with ROCK-1, Tracer 236 and Compounds A, B and C. pIC50 IC50 (nM) Hill slope Response Amplitude R2 COMPOUND A 7.11 77.3 1.81 5.8 0.97 COMPOUND B ~ 9.6 ~ 0.2 1.50 1.5 0.34 COMPOUND C - - - - - As can be seen in Figure 44, the only compound giving a robust competition curve is the Compound A. By testing the compounds in a single concentration test (Binding Check, Figure 45) at 1 µM, no significant difference is registered between the Tracer 236 mobility only bound to the protein (dark blue) and the Tracer 236 registered in the presence of the Compounds B and C (purple and red dots). -14 -12 -10 -8 -6 -4 810 815 820 ROCK-1 COMPETITION Log[inhibitor] (M) Tracer mobility Compound A Compund B Compound C Compound B Compound C 100 Figure 45 Binding Check with ROCK-1 in complex with Tracer 236 (dark blue), the same complex with the addition of orthosteric Compound A (light blue) and the allosteric Compounds B and C (purple and red). Binding Check P Value Significant? ROCK-1 vs. Compound A <0.0001 Yes (****) ROCK-1 vs. Compound B 0.28 No ROCK-1 vs. Compound C 0.95 No Compound A vs. Compound B <0.0001 Yes (****) Compound A vs. Compound C <0.0001 Yes (****) Compound B vs. Compound C 0.51 No However, in the competition assay, the Compound B showed a slight sigmoidal trend, albeit with a very low R2 and Response Amplitude, while the Compound C showed a tendency to increase Tracer mobility at higher concentrations. To verify if the trends of the two allosteric compounds would be confirmed in a replicate and at higher concentrations, both compound B and C were additionally tested with an initial titrating concentration of 10 µM. [ROCK-1] 50 nM [INHIBITOR] 10 µM – 2.4 nM + CONTROL POINTS [TRACER 236] 15 nM TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01% Tween 20 TEMPERATURE 22°C 101 Figure 46 Comparison between two competition assays with Compound B, one starting from 1 µM (purple) and the second starting from 10 µM (yellow). pIC50 IC50 (nM) Hill slope Response Amplitude R2 COMPOUND B (1 µM) ~ 9.6 ~ 0.2 1.50 1.5 0.34 COMPOUND B (10 µM) ~ 6.7 ~ 218.0 0.53 2.1 0.53 The point trends for Compound B in both tested conditions are not robust enough to conclude that Tracer 236 displacement is occurring (Figure 46). However, on the basis of the slight trend of the points toward lower Tracer mobility values, it is not possible to completely exclude the possibility that increased Compound B concentrations could impact the Tracer 236-ROCK-1 interaction or have some non-quantifiable distal effect on the ATP binding site. Figure 47 Comparison between two competition assays with Compound C, one starting from 1 µM (red) and the second starting from 10 µM (light pink). -14 -12 -10 -8 -6 -4 815 816 817 818 819 820 COMPOUND B Log[inhibitor] (M) Tracer mobility Compund B (10 M) Compound B (1 M) -14 -12 -10 -8 -6 -4 816 818 820 822 824 COMPOUND C Log[inhibitor] (M) Tracer mobility Compound C (10 M) Compound C (1 M) 102 pIC50 IC50 (nM) Hill slope Response Amplitude R2 COMPOUND C (1 µM) - - - - - COMPOUND C (10 µM) ~ 4.9 ~ 1263 1.08 6.10 0.65 Also for Compound C no robust fitting is occurring in the competition assay. At 10 µM an increased Tracer 236 mobility is registered (Figure 47). Since the Tracer is moving in the opposite direction of the free-Tracer mobility (registered at high concentrations of Compound A), it is not possible to assume that a displacement is taking place, therefore two other hypotheses can be made on this situation: - The highest concentrations points of Compound C induce an aberrant movement of the protein-Tracer complex due to its closeness to the solubility limit and to the potential formation of aggregates. - Compound C at higher concentrations is able to bind the protein in another site with respect to the Tracer 236 but also to induce a conformational change responsible for the different mobility of the ternary complex ROCK-1-Tracer 236-Compound C. In summary, from the competition experimental set it is possible to conclude that Compound B and Compound C have a different binding mode than the orthosteric Compound A. 4.2.2 DIRECT BINDING ASSAYS ROCK-1 COVALENT LABELING AND BINDING COMPETENCE TEST To perform a direct binding assay the protein ROCK-1 needed to be fluorescently labeled. The labeling protocol chosen for this purpose is the Monolith Protein Labeling Kit RED-NHS 2nd Generation (NanoTemper) and, after the labeling reaction, a binding test with Compound A was performed to verify the binding competence of the protein. Two labeling conditions were tested: 1. Labeling in the presence of a low potent orthosteric inhibitor (Compound X, pIC50 < 7) to protect the ATP binding pocket. 2. Labeling in the absence of any ligand (with the only addition of DMSO at the same percentage of the condition of the active site protection). 103 The reason of this comparison is to verify the possible interference of the covalent label with the binding competence of the protein. LABELING #1 (w/ active site protection) LABELING #2 [ROCK-1] 10 µM Dilution and labeling buffer Labeling Buffer NHS (NanoTemper) 130 mM NaHCO3, 50 mM NaCl, pH 8.2-8.3 Pre-incubation (active site protection) Compound X (pIC50 < 7) 100 µM DMSO 1 % Time of pre-incubation 20 min Dye RED-NHS 2nd Generation (NanoTemper) [Dye] 30 µM Incubation 30’ in the dark at 25°C Purification system PD SpinTrap G-25 (Cytiva) Purification buffer 50 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM DTT After the labeling reaction the proteins concentration and the Degree of Labeling (DOL) was measured through the acquisition of a UV spectra (Figure 48). Figure 48 UV spectra of two labeled protein: on the left ROCK-1 labeled in the presence of Compound X to protect the active site, while on the right ROCK-1 labeled in the absence of active site protector. LABELING #1 (w/ active site protection) LABELING #2 [ROCK-1 LB] 1.36 µM 1.19 µM DOL 0.28 0.34 The two different labeling reactions led to similar results in terms of protein concentration and DOL. The main difference is the presence of an absorption peak at 375 nm, probably due to an incomplete 300 400 500 600 700 0.00 0.02 0.04 0.06 0.08 0.10 LABELING #1 (w/ active site protection) Wavelenght (nm) Absorbance 300 400 500 600 700 0.00 0.02 0.04 0.06 0.08 0.10 LABELING #2 Wavelenght (nm) Absorbance 104 removal of Compound X from the system. The low affinity of Compound X is a necessary feature to facilitate its removal during the purification phase, but the use of a spin column for this phase probably reduced the efficiency of the process compared to a slower Gravity column (e.g. PD MiniTrap G-25, Cytiva). If necessary, it has been demonstrated that an additional purification step in a size exclusion spin column can be performed to better remove Compound X. However, the residual presence of the compound seems not to interfere in the subsequent interaction assay with the reference Compound A, probably due to Compound X low potency and the further dilution of the binding assay. Indeed, the ROCK-1 LB #1 demonstrated a robust interaction profile with Compound A after the direct binding assay was performed to confirm the binding competence of the labeled proteins. On the contrary, the ROCK-1 LB #2 seemed not to be able to yield reliable binding information. This can be due to interference from the fluorescent label in the interaction between the protein and the reference Compound A (Figure 49). [ROCK-1 LB] 20 nM [INHIBITOR] 1 µM – 0.24 nM + CONTROL POINTS TITRATION 1:3, 8 POINTS IN DUPLICATE DILUTION BUFFER 50 mM HEPES (pH 7.5), 150 mM NaCl, 2 mM DTT, 0.01% Tween 20 TEMPERATURE 22°C Figure 49 Direct binding assay with Compound A and ROCK-1 labeled in the presence of an active site protector (on the left) and in the absence of an active site protector (on the right). -14 -12 -10 -8 -6 -4 834 836 838 840 842 ROCK-1 LB #1 (w/ active site protection) + COMPOUND A Log[inhibitor] (M) FNorm 5s -14 -12 -10 -8 -6 -4 806 808 810 ROCK-1 LB #2 + COMPOUND A Log[inhibitor] (M) FNorm 5s 105 LABELING #1 (w/ active site protection) LABELING #2 pIC50 8.75 ~9.65 IC50 (nM) 1.8 0.2 R2 0.96 0.54 HILL COEFF 1.28 9.61 RESPONSE AMPLITUDE 5.03 2.29 MST-on TIME 5 s 5 s These data suggest that protection of the binding site with an orthosteric compound might be a valid choice in order to observe and quantitatively describe the interaction with orthosteric inhibitors and confirm the protein binding competence after labeling. The presence of a ligand during the labeling phase probably induces a reduction in protein mobility and a closure of its structure, avoiding the labeling of normally buried residues that are possibly involved in the formation of binding pockets. To verify the robustness and reproducibility of the labeling procedure, an additional labeling under the same conditions as Labeling #1 has been performed, and the results are reported in Figure 50. Figure 50 UV spectra of two labeled protein: on the left the already reported ROCK-1 labeled in the presence of Compound X to protect the active site, and on the right a replicate in the same experimental conditions. LABELING #1 (w/ active site protection) LABELING #3 (w/ active site protection) [ROCK-1 LB] 1.36 µM 1.20 µM DOL 0.28 0.28 300 400 500 600 700 0.00 0.02 0.04 0.06 0.08 0.10 LABELING #1 (w/ active site protection) Wavelenght (nm) Absorbance 300 400 500 600 700 0.00 0.02 0.04 0.06 0.08 0.10 LABELING #3 (w/ active site protection) Wavelenght (nm) Absorbance 112 from the perspectives of both the inhibitor-protein complex and the protein alone conformational changes over time. Furthermore, stoichiometry analysis with Gefitinib (Type I) and Lapatinib (Type II) revealed different protein states as well as a different propensity of inhibitors to bind them. The study of irreversible binding with Osimertinib enabled then the use of MST to develop an alternative approach to Jump Dilution for the differentiation of reversible and irreversible inhibitors, a potentially useful tool for the discovery of unknown covalent drugs. Finally, ROCK system permitted to apply MST for the differentiation of orthosteric and allosteric binders, leading to valuable complementary information through the combination of competition and direct binding assays. Furthermore, an investigation of the best labeling conditions found discrepancies between the two tested allosteric compounds, revealing a possible distinct interaction mechanism due to their different sensitivity to the dye position on the protein. In conclusion, MicroScale Thermophoresis emerged as a valuable tool for the characterization of several aspects of protein interactions, in addition to the simple determination of potency; thus, this project helped to better understand MST potentiality and applications through the establishment of benchmarks for several binding modes, and the acquired knowledge could potentially be applied to the future investigation of new chemical entity mechanisms, from the early to the late stages of Drug Discovery pipeline. 113 6. BIBLIOGRAPHY [1] X. Du et al., “Insights into protein–ligand interactions: Mechanisms, models, and methods” Int. J. Mol. Sci., vol. 17, no. 2, pp. 1–34, 2016, doi: 10.3390/ijms17020144. [2] M. A. Williams, “Protein-Ligand Interactions - Methods and Applications”, Methods in Molecular Biology, vol. 1008, 2013, doi: 10.1007/978-1-62703-398-5. [3] A. Copeland, “Evaluation of enzyme inhibitors in Drug Discovery - A Guide for Medicinal Chemists and Pharmacologists” Angew. Chemie, vol. 117, no. 40, 2005, doi: 10.1002/ange.200585328. [4] C. Velours et al., “Macromolecular interactions in vitro, comparing classical and novel approaches” Eur. Biophys. J., vol. 50, no. 3–4, pp. 313–330, 2021, doi: 10.1007/s00249-021-01517-5. [5] S. Markossian et al., “Assay Guidance Manual,” Eli Lilly & Company and the National Center for Advancing Translational Sciences, Bethesda (MD), 2020. [6] T. D. Pollard, “A guide to simple and informative binding assays”, Mol Biol Cell., vol. 21, no. 23, pp. 4061-4067, 2010, doi:10.1091/mbc.E10-08-0683. [7] T. Kenakin, “The mass action equation in pharmacology,” Br. J. Clin. Pharmacol., vol. 81, no. 1, pp. 41–51, 2016, doi: 10.1111/bcp.12810. [8] R. A. Copeland, “Enzymes: A Practical Introduction to Structure, Mechanism, and Data Analysis, 2nd Edition”, Wiley-VCH Verlag, 2000. [9] R. R. Neubig et al., “International Union of Pharmacology Committee on Receptor Nomenclature and Drug Classification. XXXVIII. Update on terms and symbols in quantitative pharmacology.” Pharmacological reviews vol. 55,4: 597-606, 2003, doi:10.1124/pr.55.4.4. [10] T. Kalliokoski, C. Kramer, A. Vulpetti, and P. Gedeck, “Comparability of Mixed IC50 Data - A Statistical Analysis,” PLoS One, vol. 8, no. 4, p. 61007, 2013, doi: 10.1371/journal.pone.0061007. [11] P. Newton, P. Harrison, and S. Clulow, “A novel method for determination of the affinity of protein: Protein interactions in homogeneous assays,” J. Biomol. Screen., vol. 13, no. 7, pp. 674–682, 2008, doi: 10.1177/1087057108321086. [12] E. C. Hulme and M. A. Trevethick, “Ligand binding assays at equilibrium: validation and interpretation.” British journal of pharmacology vol. 161,6, pp. 1219-37, 2010, doi: 10.1111/j.14765381.2009.00604.x. [13] E. Marchal, S. Roy, and L. Lafanechre, Chemogenomics and Chemical Genetics – “A User’s Introduction for Biologists, Chemists and Informaticians”. Springer Heidelberg Dordrecht London New York, 2011, doi: 10.1007/978-3-642-19615-7 [14] P. J. Tonge, “Quantifying the Interactions between Biomolecules: Guidelines for Assay Design and Data Analysis,” ACS Infect. Dis., vol. 5, no. 6, pp. 796–808, 2019, doi: 10.1021/acsinfecdis.9b00012. [15] H. Motulsky and A. Christopoulos, “Fitting Models to Biological Data using Linear and Nonlinear Regression - A practical guide to curve fitting.” GraphPad Software Inc., 2003. [16] I. Jarmoskaite, I. Alsadhan, P. P. Vaidyanathan, and D. Herschlag, “How to measure and evaluate binding affinities,” Elife, vol. 9, pp. 1–34, 2020, doi: 10.7554/ELIFE.57264. [17] R. A. Copeland, “Evaluation of enzyme inhibitors in Drug Discovery - A Guide for Medicinal Chemists and Pharmacologists” Wiley, 2013, doi: 10.1002/9781118540398. [18] P. Kuzmic, et al. “High-throughput screening of enzyme inhibitors: simultaneous determination of tight-binding inhibition constants and enzyme concentration.” Analytical biochemistry vol. 286,1, pp. 114 45-50, 2000, doi:10.1006/abio.2000.4685P. [19] D. J. Murphy, “Determination of accurate KI values for tight-binding enzyme inhibitors: An in silico study of experimental error and assay design,” Anal. Biochem., vol. 327, no. 1, pp. 61–67, 2004, doi: 10.1016/j.ab.2003.12.018. [20] A. Tuley and W. Fast, “The Taxonomy of Covalent Inhibitors,” Biochemistry, vol. 57, no. 24. American Chemical Society, pp. 3326–3337, Jun. 19, 2018, doi: 10.1021/acs.biochem.8b00315. [21] L. Boike, N. J. Henning, and D. K. Nomura, “Advances in covalent drug discovery,” Nat. Rev. Drug Discov., vol. 21, no. December, 2022, doi: 10.1038/s41573-022-00542-z. [22] P. Kuzmiˇ, “A two-point IC 50 method for evaluating the biochemical potency of irreversible enzyme inhibitors,” BioKin, 2020, doi: 10.1101/2020.06.25.171207. [23] J. M. Strelow, “A Perspective on the Kinetics of Covalent and Irreversible Inhibition,” Journal of Biomolecular Screening, vol. 22, no. 1, pp. 3–20, Jan. 01, 2017, doi: 10.1177/1087057116671509. [24] A. Thorarensen et al., “The advantages of describing covalent inhibitor in vitro potencies by IC50 at a fixed time point. IC50 determination of covalent inhibitors provides meaningful data to medicinal chemistry for SAR optimization,” Bioorganic Med. Chem., vol. 29, p. 115865, 2021, doi: 10.1016/j.bmc.2020.115865. [25] M. Jerabek-Willemsen, C. J. Wienken, D. Braun, P. Baaske, and S. Duhr, “Molecular interaction studies using microscale thermophoresis,” Assay and Drug Development Technologies, vol. 9, no. 4. pp. 342– 353, Aug. 01, 2011, doi: 10.1089/adt.2011.0380. [26] Y. Mao, L. Yu, R. Yang, L. B. Qu, and P. D. B. Harrington, “A novel method for the study of molecular interaction by using microscale thermophoresis,” Talanta, vol. 132, pp. 894–901, Jan. 2015, doi: 10.1016/j.talanta.2014.09.038. [27] SA. Seidel, et al. “Microscale thermophoresis quantifies biomolecular interactions under previously challenging conditions.” Methods (San Diego, Calif.) vol. 59,3, pp. 301-15, 2013, doi: 10.1016/j.ymeth.2012.12.005. [28] T. H. Scheuermann, S. B. Padrick, K. H. Gardner, and C. A. Brautigam, “On the acquisition and analysis of microscale thermophoresis data,” Anal. Biochem., vol. 496, pp. 79–93, Mar. 2016, doi: 10.1016/j.ab.2015.12.013. [29] S. Duhr and D. Braun, “Why molecules move along a temperature gradient”, PNAS, vol. 103, 19678– 19682, 2006. [30] NanoTemper Technologies GmbH., “NanoPedia - Monolith NT.115.” [31] B. López-Mndez, S. Uebel, L. P. Lundgren, and A. Sedivy, “Microscale Thermophoresis and additional effects measured in NanoTemper Monolith instruments,” Eur. Biophys. J., vol. 50, no. 3–4, pp. 653– 660, 2021, doi: 10.1007/s00249-021-01529-1. [32] J. M. Rainard, G. C. Pandarakalam, and S. P. McElroy, “Using Microscale Thermophoresis to Characterize Hits from High-Throughput Screening: A European Lead Factory Perspective,” SLAS Discovery, vol. 23, no. 3. pp. 225–241, Mar. 01, 2018, doi: 10.1177/2472555217744728. [33] G. Manning, et al. “The protein kinase complement of the human genome.” Science (New York, N.Y.) vol. 298,5600, pp. 1912-34, 2002, doi:10.1126/science.1075762G. [34] M. Congreve, C. W. Murray, and T. L. Blundell, “Structural biology and drug discovery,” Drug Discov. Today, vol. 10, no. 13, pp. 895–907, 2005, doi: 10.1016/S1359-6446(05)03484-7. [35] Z. Zhao and P. E. Bourne, “Progress with covalent small-molecule kinase inhibitors,” Drug Discov. Today, vol. 23, no. 3, pp. 727–735, 2018, doi: 10.1016/j.drudis.2018.01.035. 115 [36] I. Galdadas, et. al, “Structural basis of the effect of activating mutations on the EGF receptor,” Elife, vol. 10, pp. 1–24, 2021, doi: 10.7554/eLife.65824. [37] M. A. Morando et al., “Conformational Selection and Induced Fit Mechanisms in the Binding of an Anticancer Drug to the c-Src Kinase,” Sci. Rep., vol. 6, no. March, pp. 1–9, 2016, doi: 10.1038/srep24439. [38] S. Sigismund, D. Avanzato, and L. Lanzetti, “Emerging functions of the EGFR in cancer,” Mol. Oncol., vol. 12, no. 1, pp. 3–20, 2018, doi: 10.1002/1878-0261.12155. [39] Z. Zhao, et al. “Structural Insights into Characterizing Binding Sites in Epidermal Growth Factor Receptor Kinase Mutants.” Journal of chemical information and modeling vol. 59,1, pp. 453-462, 2019, doi: 10.1021/acs.jcim.8b00458. [40] L. Sutto and F. L. Gervasio, “Effects of oncogenic mutations on the conformational free-energy landscape of EGFR kinase,” PNAS, vol. 110, pp. 10616–10621, 2013, doi: 10.1073/pnas.1221953110. [41] Y. Shan et al., “Oncogenic mutations counteract intrinsic disorder in the EGFR kinase and promote receptor dimerization,” Cell, vol. 149, no. 4, pp. 860–870, May 2012, doi: 10.1016/j.cell.2012.02.063. [42] C. Becker et al., “Monitoring Conformational Changes in the Receptor Tyrosine Kinase EGFR,” ChemBioChem, vol. 17, no. 11, pp. 990–994, 2016, doi: 10.1002/cbic.201600115. [43] K. S. Gajiwala et al., “Insights into the aberrant activity of mutant EGFR kinase domain and drug recognition,” Structure, vol. 21, no. 2, pp. 209–219, Feb. 2013, doi: 10.1016/j.str.2012.11.014. [44] E. R. Wood et al., “A Unique Structure for Epidermal Growth Factor Receptor Bound to GW572016 (Lapatinib): Relationships among Protein Conformation, Inhibitor Off-Rate, and Receptor Activity in Tumor Cells,” Cancer research vol. 64,18, pp. 6652-9, 2004, doi: 10.1158/0008-5472.CAN-04-1168. [45] T. S. Beyett et al., “Molecular basis for cooperative binding and synergy of ATP-site and allosteric EGFR inhibitors,” Nat. Commun., vol. 13, no. 1, 2022, doi: 10.1038/s41467-022-30258-y. [46] Y. Yosaatmadja et al., “Binding mode of the breakthrough inhibitor AZD9291 to epidermal growth factor receptor revealed,” J. Struct. Biol., vol. 192, no. 3, pp. 539–544, 2015, doi: 10.1016/j.jsb.2015.10.018. [47] Y. Feng, P. V. Lograsso, O. Defert, and R. Li, “Rho Kinase (ROCK) Inhibitors and Their Therapeutic Potential,” J. Med. Chem., vol. 59, no. 6, pp. 2269–2300, 2016, doi: 10.1021/acs.jmedchem.5b00683. [48] C. Hahmann, T. Schroeter, T. “Rho-kinase inhibitors as therapeutics: from pan inhibition to isoform selectivity”. Cell. Mol. Life Sci. 67, 171–177, 2010, doi: 10.1007/s00018-009-0189-x. [49] Wang, Qing et al. “Advantages of Rho-associated kinases and their inhibitor fasudil for the treatment of neurodegenerative diseases.” Neural regeneration research vol. 17,12, 2623-2631, 2022, doi: 10.4103/1673-5374.335827. [50] L. Julian and M. F. Olson, “Rho-associated coiled-coil containing kinases (ROCK), structure, regulation, and functions,” Small GTPases, vol. 5, no. 2, 2014, doi: 10.4161/sgtp.29846. [51] J. K. Frank Schäfer et al. “Purification of GST-Tagged Proteins.” Methods in enzymology vol. 559, 12739, 2015, doi: 10.1016/bs.mie.2014.11.005. [52] F. Ferlenghi et al., “A sulfonyl fluoride derivative inhibits EGFRL858R/T790M/C797S by covalent modification of the catalytic lysine,” Eur. J. Med. Chem., vol. 225, 2021, doi: 10.1016/j.ejmech.2021.113786. [53] R. Castelli et al., “Balancing reactivity and antitumor activity: heteroarylthioacetamide derivatives as potent and time-dependent inhibitors of EGFR,” Eur. J. Med. Chem., vol. 162, pp. 507–524, 2019, doi: 10.1016/j.ejmech.2018.11.029. 116 [54] M. Jerabek-Willemsen et al., “MicroScale Thermophoresis: Interaction analysis and beyond,” J. Mol. Struct., vol. 1077, pp. 101–113, Dec. 2014, doi: 10.1016/j.molstruc.2014.03.009. [55] R. Magnez and C. Bailly, “Microscale Thermophoresis as a Tool to Study Protein Interactions and Their Implication in Human Diseases,” International journal of molecular sciences vol. 23,14 7672. 12 Jul. 2022, doi: 10.3390/ijms23147672. [56] X. Zhai et al. “Insight into the Therapeutic Selectivity of the Irreversible EGFR Tyrosine Kinase Inhibitor Osimertinib through Enzyme Kinetic Studies.” Biochemistry vol. 59,14, pp. 1428-1441, 2020, doi: 10.1021/acs.biochem.0c00104. [57] J. L. Macdonald-Obermann and L. J. Pike, “Allosteric regulation of epidermal growth factor (EGF) receptor ligand binding by tyrosine kinase inhibitors.” The Journal of biological chemistry vol. 293,35, pp. 13401-13414, 2018, doi: 10.1074/jbc.RA118.004139.