Ligand discrimination between active and inactive activation loop conformations of Aurora-A kinase is unmodified by phosphorylation† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c8sc03669a

Activation loop phosphorylation changes the position of equilibrium between DFG-in-like and DFG-out-like conformations but not the conformational preference of inhibitors.


Calculating the threshold between optimizing for active and inactive enzyme conformations
On a surface characterized by equation (4) (here rewritten as equation (S1) in terms of x, y and z for ease of notation), a point (x,y) is on the surface a distance of ∆x from the contour line z = C (C = constant) in a direction parallel to the x-axis, and a distance of ∆y from the contour line z = C in a direction parallel to the y-axis.
Thus coordinates of point A are (x -∆x, y) and point B are (x,y -∆y).
(1 ) eq eq x y K z x K y ⋅ + = ⋅ + (S1) Substituting coordinates into (S1), at A where (S7) is the equation of the contour line z = C, and (S8) is a straight line with slope = 1 which intersects z = C at eq y x K = .
Estimating the inclusion of two co-localized single-labelled molecules in our analysis (the false positive rate) All single molecule traces included in our analysis exhibited two-step (and only two-step) photobleaching to background-subtracted zero, with each photobleaching step reducing the original intensity by around half. This selection step ensures that all analyzed traces originate from molecules possessing two fluorophores (ie double-labelled protein samples), and excludes most single-labelled proteins (originating from incomplete labelling). Nevertheless, two single-labelled protein molecules co-localized within the same diffraction-limited spot also fulfil this criteria. We refer to the mistaken inclusion of two co-localized single-labelled molecules in our analysis as our false positive rate.
It is important to note that false positive traces will never transition to a quenched intensity (will never show dynamic behavior) since TMR quenching requires the two dyes to be very close to (within ~16 Å of) each other 18 . Thus false positive traces will only be included in our fluorescence intensity histograms, not our dwell time histograms (our kinetic measurements). For the same reason, false positive traces will only contribute to the population of the unquenched peak in the intensity histogram.
In order to estimate our false positive rate we reanalyzed the data from our single-labelled control experiments 11 . The experimental conditions for these data are identical to those reported in this paper, except that the samples used have only one labelling site (K224C or S283C depending on the data set). Since all molecules in these experiments are single-(or un-) labelled, any traces showing two photobleaching steps must be due to two co-localized molecules As expected, there were several traces for our single-labelled protein samples which had two photobleaching steps, and which otherwise met all of our standard criteria for inclusion as an experimental trace. For K224C this was 3 ± 1 % of total traces, for S283C this was 2 ± 1 % of total traces (values reported are mean and standard deviation of five experiments). By eye, these two experiments had different immobilization densities (K224C an unusually dense field of view, S283C a less dense 'normal' field of view), giving us confidence that the co-localization rate we measured is a good estimate of the co-localization rate in our other experiments. For these two experiments, around 10 % of all traces met the inclusion criteria (including the presence of a single photobleaching step) for the histograms which we previously published 11 meaning that the ratio of co-localized:dispersed molecules for K224C (dense field of view) is 3:10 and for S283C (normal density) is 1:5.
In our experiments with double-labelled protein, only co-localization events involving two singlelabelled protein molecules contribute to our false positive rate. Co-localization of two doublelabelled proteins (4 photobleaching steps) or one single and one double-labelled protein (3 photobleaching steps) would already be excluded from analysis. We therefore used our labelling efficiencies to calculate the number of single-labelled molecules as a proportion of the total number of labelled proteins and scaled the co-localized part of the co-localized:dispersed ratio accordingly. We assumed that the ratio of co-localized:dispersed molecules for slides of similar immobilization density remained constant throughout our experiments, and therefore used this rescaled ratio to estimate our false positive rate (Supplementary table SIV).
However, there is no simple relationship between our false positive rate and the error introduced by this on the reported values from our intensity histograms. In other words, a 5% change in the number of traces included in a peak will not result in a 5% change in the area of that peak. Additionally, each experimental trace is of a different length (since photobleaching is stochastic), so a 5% change in the number of traces does not translate into a 5% change in the number of frames included.
In order to estimate the effect of our false positive rate on our experimental results, we compared our results for K224C/S283C and M373C/S283 constructs. Both phosphorylated and unphosphorylated experiments report the same conformational distributions, yet these samples have enormously different labelling efficiencies (Supplementary a Assuming random labelling of sites b Assuming that the ratio of co-localized : dispersed molecules among traces meeting the criteria for analysis remains similar to that measured in S283C control (ie 1:5). False positive rate = probability of trace being due to co-localization (1/6) * probability of co-localization event resulting from two single-labelled molecules Supplementary a Statistics for intensity histograms (Figure 1b-d, f-h). b A molecule was scored as interconverting if a transition from high intensity to quenched intensity (or vice versa) was observed before the first photobleaching event occurred. The reported numbers should be considered a lower bound, since many traces photobleached relatively quickly. Supplementary figure S6: When designing a potent ligand, the ligand should be optimized to bind the conformation of the kinase to which it has greatest affinity (panels a&g). In the hypothetical situation of designing a less potent ligand, it may be more productive to optimize binding to the conformation of the kinase to which it has initially least affinity. This depends on the position of the starting ligand on the surface relative to dashed white line (shown in figure). For ligands which are located on the surface between the dashed white line and the diagonal Kd,inactive = Kd,active, the shortest distance to the target affinity (target contour line) is not by optimizing binding to the conformation to which the initial ligand binds tightest. Eg in panel (e), the initial compound binds more tightly to the inactive conformation (Kd,active = 950 nM, Kd,inactive = 800 nM) but the distance to the contour line Kd,overall = 500 nM is shorter along the x-axis (optimizing Kd,active) than along the y-axis (optimizing Kd,inactive). Interestingly, the position of the dashed white line (the line along which optimization against active or inactive conformations is equidistant from the stated contour line) changes with the overall potency being targeted. a-f) Contour plots drawn for phosphorylated kinase (Keq = 0.3). a) Kd,overall = 1 nM. b) Kd,overall = 100 nM. c) Kd,overall = 200 nM. d) Kd,overall = 300 nM. e) Kd,overall = 500 nM. f) Kd,overall = 700 nM. g-l) Contour plots drawn for a hypothetical kinase preferentially adopting the inactive activation loop conformation (Keq = 2). g) Kd,overall = 1 nM. h) Kd,overall = 100 nM. i) Kd,overall = 200 nM. j) Kd,overall = 300 nM. k) Kd,overall = 500 nM. l) Kd,overall = 700 nM. The equidistant line (given by equation (S8)