Kinetic Optimization of Lysine-Targeting Covalent Inhibitors of HSP72

The covalent inhibition mechanism of action, which overcomes competition with high-affinity, high-abundance substrates of challenging protein targets, can deliver effective chemical probes and drugs. The success of this strategy has centered on exposed cysteine residues as nucleophiles but the low abundance of cysteine in the proteome has limited its application. We have recently reported our discovery that lysine-56 in the difficult-to-drug target HSP72 could form a covalent bond with a small-molecule inhibitor. We now disclose the optimization of these targeted covalent inhibitors using rational design. Essential to our optimization was the development of a new covalent fluorescence polarization assay, which allows for the direct measurement of the key kinetic parameter in covalent inhibitor design, kinact/KI, extrapolation of the underlying parameters, kinact and Ki, and direct comparison to reversible analogues. Using our approach, we demonstrate a >100-fold enhancement in covalent efficiency and key learnings in lysine-selective electrophile optimization.


Derivation of percentage occupancy equations for Figure 1
For TCIs that follow a two-step mechanism of inhibition, the observed rate constant of inactivation at a given inhibitor concentrations (k obs ) can be expressed in terms of the maximum rate constant of inactivation (k inact ), the concentration of inhibitor that gives a half-maximal rate of inactivation (K I ) and the inhibitor concentration (  were plotted and analysed using GraphPad Prism 6, graphical data represent the arithmetic mean ± curve fitting standard error of the mean for a single representative experiment.

K D determination
To each well, 5 μL of fluorescent probe molecule (N 6 -(6-aminohexyl)-ATP ATTO-488, Jena Bioscience; 20 nM in assay buffer, 10 nM final concentration) and increasing concentrations of HSP72-NBD protein (5 μL, two-fold dilution series) were added. Fluorescence polarisation values for tracer control wells (10 nM probe in assay buffer only) were subtracted from each data point prior to data analysis. K D determination was performed using non-linear regression analysis (GraphPad Prism 6, one site-specific binding model). pK D values are quoted as geometric mean ± standard error of the mean from 3 independent experiments.

Competitive binding experiments
Compounds (0.2 μL at 50 x screening concentration in DMSO) were dispensed using an ECHO 550 Liquid Handler (Labcyte Inc.). To the corresponding wells was added, 5 μL of probe molecule (20 nM in assay buffer, 10 nM final concentration) and 5 μL of protein (twice their final concentration in assay buffer) to give a 50% bound fraction. 8 Tracer controls (10 nM probe molecule only) and bound tracer controls (10 nM probe in presence of appropriate protein concentration) were included on each assay plate. IC 50 determination was performed using non-linear least squares curve fitting (GraphPad Prism 6, log(inhibitor) vs. responsevariable slope (four parameters)). K i values were calculated using the method described by Huang. 8 pK i values are quoted as geometric mean ± standard error of the mean from 3 independent experiments. Qualitative Analysis B.06.00 and graphs made using GraphPad Prism 6.

Worked example of kinetic FP assay
The output of the FP assay is in millipolarisation (mP) units, a measurement of the degree of polarization that the light retains after interacting with the rapidly tumbling fluorescent probe.
When the probe is bound to HSP72-NBD, it tumbles at a slower rate compared to when the probe is not bound. The greater the tumbling rate, the more the degree of polarisation is reduced. When the probe is bound to the protein, it tumbles slowly, leading to a high mP value.
When the probe is free, it tumble more quickly, leading to a lower mP value. Therefore, the degree of polarization of fluorescent light directly describes bound fraction of the protein.
Because the bound fraction of the fluorescent probe is proportional to the concentration of free Huang's Equation describes the relationship between IC 50 and K i : 8 Huang's equation was used to calculate K i values from the measured IC 50 s. See equation below: The equation states that the IC 50 for a ligand that is competitive for binding with the assay probe is related to the binding affinity of the ligand (K i ), the bound fraction of the probe (f 0 ), the binding affinity of the probe (K d ) and the concentration of the probe (L 0 ). For competition experiments, it is recommended that a protein concentration giving a bound fraction between 0.5 and 0.8 be selected. A bound fraction below 0.5 will often result in an assay that is not statistically robust due to the decreased size of the binding window, however as the bound fraction approaches 1 the relationship between K i and IC 50 deviates from linear and the resolvable range of the assay decreases. For these reasons, a bound fraction of 0.5 was used for all assays.

S8
Based on the analysis by Huang, 8 the free unbound protein concentration displays a non-linear relationship with the bound fraction. To carry out the kinetic FP-assay, an initial protein concentration that would lead to a high bound fraction (F b = 0.8) was selected as the reversible occupancy of the protein would rapidly displace the probe and reduce F b . At low bound fractions of probe (F b < 0.4), the assay is insensitive (Figure 1c), as large changes in F b are required for a small change in protein concentration. At high bound fraction of the probe (F b > 0.8), the relationship between F b and protein concentration is essentially non-linear and small changes in F b result in large changes in protein concentration (Figure 1b). Therefore, selection of an initial bound fraction of 0.8 ensures that the working range of the assay remains between these limits (0.4 < F b < 0.8, Figure 1d  The initial K i value was then used to focus a subsequent FP titration at TCI concentrations below the initial K i . A 1.5-fold dilution series from initial K i concentrations of TCI 14 was used to measure the time-dependence in mP values (Figure 4a). For TCI 14, the overall second-order rate constant for covalent bond formation with HSP72-NBD was k inact /K I = 35 M -1 s -1 . The k inact value was estimated using the initial K i value calculated from the t=0 extrapolation, which assumed that K i = K I . By multiplying k inact /K I by the initial K i value, k inact was calculated as 3.6x10 -4 s -1 . This is equal to a half-life at a theoretical 100% occupancy of 32 minutes (t 1/2  = ln2/k inact ).