High-Content Surface and Total Expression siRNA Kinase Library Screen with VX-809 Treatment Reveals Kinase Targets that Enhance F508del-CFTR Rescue

The most promising F508del-CFTR corrector, VX-809, has been unsuccessful as an effective, stand-alone treatment for CF patients, but the rescue effect in combination with other drugs may confer an acceptable level of therapeutic benefit. Targeting cellular factors that modify trafficking may act to enhance the cell surface density of F508-CFTR with VX-809 correction. Our goal is to identify druggable kinases that enhance F508del-CFTR rescue and stabilization at the cell surface beyond that achievable with the VX-809 corrector alone. To achieve this goal, we implemented a new high-throughput screening paradigm that quickly and quantitatively measures surface density and total protein in the same cells. This allowed for rapid screening for increased surface targeting and proteostatic regulation. The assay utilizes fluorogen-activating-protein (FAP) technology with cell excluded and cell permeant fluorogenic dyes in a quick, wash-free fluorescent plate reader format on live cells to first measure F508del-CFTR expressed on the surface and then the total amount of F508del-CFTR protein present. To screen for kinase targets, we used Dharmacon’s ON-TARGETplus SMARTpool siRNA Kinase library (715 target kinases) with and without 10 μM VX-809 treatment in triplicate at 37 °C. We identified several targets that had a significant interaction with VX-809 treatment in enhancing surface density with siRNA knockdown. Select small-molecule inhibitors of the kinase targets demonstrated augmented surface expression with VX-809 treatment.


MGnBu Photophysical properties:
*The final dye solutions for spectroscopy and biological experiments were diluted from a 1000x stock solution in 1% acidic ethanol and prepared the same day of experiments.
Quantum yield: Quantum yield of the dyes were determined using MG2p-dL5** as a standard with a quantum yield of 20% in PBS 7.4. The fluorescence emission of dye-dL5** solutions containing the same O.D at 630 nm (λex) in PBS 7.4 were obtained. The fluorescence spectra of the solutions were measured and the ratio of quantum yield of standard and unknown sample is: ‫ݔܨ‬ ‫ݏܨ‬ = ‫ݔܳ‬ ‫ݏܳ‬ Where F -fluorescence intensity, Q -quantum yield; x -sample, s -standard Binding affinity (Kd measurement): Binding affinity of MgnBu was determined by titrating it against a known concentration of dL5**. Triplicate fluorescence response was determined using a 96-well plate on a TECAN Infinite M1000 96-well plate reader fluorimeter using ex/em of 636 nm/664 nm. Analysis of fluorescence response was determined using a non-linear regression using One site -Total, accounting for ligand depletion. The model, originally used for radioactivity measurements was modified to fit fluorescence data by fixing the volume at 0.2 mL and SpecAct was set to 1.00 on GraphPad Prism 7.0 software, and the maximum fluorescent intensity was set to 5.0, representing the maximum concentration of complex that can be formed (based on 5 nM protein concentration). The ligand depletion model assumes that changes in complex formation are associated with complementary changes in free ligand and free receptor, and are a typical model for ligand-receptor interactions when one has to work at protein concentrations that are near the Kd value. The original formula and fitting can be found at: http://www.graphpad.com/guides/prism/6/curve-itting/index.htm?reg_one_site_total_depletion.htm

Analysis and Statistics
Pre-processing data: All well measurements had the mean background fluorescence subtracted. The surface CFTR fluorescence signal and total fluorescence signal were normalized to the relative quantity of cells (Hoechst 33342 signal) per scanning point, 16 total, inside a well. All values were then transformed to log scale to achieve a normal distribution required for the subsequent statistical tests. For each well, the 16 intra-well values were subjected to outlier analysis using robust Z*Score testing to identify and exclude significant anomalies within a well. These irregular measurements could arise from technical error or small debris. The final well sample measurement was calculated by the interquartile mean of the intra-well values. SiRNA screen quality assessment: The quality control (QC) of the siRNA screen was assessed through strictly standardize mean difference (SSMD). 3 This scoring method is favorable due to its ability to report consistent quality control results with positive controls that aren't necessarily very strong. In our system, our positive control is siRNA that targets CFTR, and the total amount of CFTR is reduced due to the F508d mutation. Even with VX-809 treatment, knockdown effect can be moderate in terms of total protein. Our QC criterion was based off of a moderate control effect assessing the VX-809 treated plate of a transfection pair, and using the CFTR siRNA wells as a positive control for total protein knockdown and scrambled NC siRNA wells as a negative control. A plate passes QC if it has good or excellent quality. Hit selection: Paired SSMD and mean fold change in CFTR surface expression were used to evaluate hits. 4 SSMD is the mean of differences (đ i ) divided by the standard deviation (ܵ ) of the differences between an siRNA and a negative reference. 5 The SSMD score generated is the average fold change in log scale penalized by the variability of fold change on the log scale.
VX-809 treated plate: The median of the negative siRNA VX-809 treated NC of a plate was subtracted from each sample of the same plate. The mean difference and standard deviation of an siRNA was calculated from the plate replicates. To select for strong siRNA effects with VX-809, we used a hit threshold of 2 for the kinase library screen, which used 3 replicates per siRNA. The siRNAs that were above the threshold were ranked based on their average fold change from the median VX-809 NC.
Vehicle (DMSO) treated plate: Without any chemical or temperature correction, there is near zero surface fluorescence in the NC wells alone. These vehicle treated plates are more sensitive to small increases in surface signal, which can potentially arise from media evaporation at the edges. In the kinase library screen, the negative controls are positioned in the first column, making it important to be able to identify and counter this potential effect. The % difference between the median of the NC and the interquartile mean of the whole plate is determined. The interquartile mean is used only when there is a larger than 15% difference between the surface fluorescence of the negative controls and the interquartile mean of the plate. This large difference infers that the characteristic NC near zero signal was likely aggravated by possible edge effects, making it important to utilize the plate based normalization. Normalizing using most of the samples on the plate is powerful because there are only very few instances where siRNA treatments promotes F508del-CFTR surface expression with DMSO alone. In addition, the large amount of samples in a plate promotes a more accurate measurement of the baseline surface signal. The median of the negative control siRNA or interquartile mean of a DMSO treated plate was subtracted from each sample of the same plate. The mean difference and standard deviation of an siRNA was calculated from the plate replicates. To identify strong siRNA effects with DMSO, we used a hit threshold of 2 for the kinase library screen, which used 3 replicates per siRNA. The siRNAs that were above the threshold were ranked based on their average fold change from the median DMSO NC or plate interquartile mean. The primary purpose of the vehicle treated plate was to establish the effect of an siRNA without VX-809 to determine the interaction significance.
Interaction testing: Interaction was assessed using a linear model that included a siRNA-drug interaction effect. 6,7 Surface Signal = siRNA(yes/no) + VX-809(yes/no) + siRNA*VX-809 (yes,yes) siRNA+VX-809 (yes,no) Vehicle siRNA (no,yes) VX-809 NC (no,no) Vehicle NC A two-way, balanced (3 replicates per condition) ANOVA was performed in Graphpad prism 6 to test for significant interaction using the log fold change from DMSO NC (NC well median or interquartile mean of plate). Equal variance was confirmed using Levene's test in Minitab. However, due to the variable nature of siRNA knockdown efficiency between replicates and a bare minimum of replicates, there were instances where replicate variance was too broad to calculate significant interaction. To help mitigate variance between replicates in the VX-809 condition, where there is often some variability in the VX-809 treatment itself, the overall average of VX-809 NC log fold increase from the DMSO baseline surface expression of each transfection pair (VX-809 and DMSO treated plate of a single replicate) was used to rescale each replicate to the overall average.