Identification of new correctors for traffic-defective ABCB4 variants by a high-content screening approach

ABCB4 is located at the canalicular membrane of hepatocytes and is responsible for the secretion of phosphatidylcholine into bile. Genetic variations of this transporter are correlated with rare cholestatic liver diseases, the most severe being progressive familial intrahepatic cholestasis type 3 (PFIC3). PFIC3 patients most often require liver transplantation. In this context of unmet medical need, we developed a high-content screening approach to identify small molecules able to correct ABCB4 molecular defects. Intracellularly-retained variants of ABCB4 were expressed in cell models and their maturation, cellular localization and function were analyzed after treatment with the molecules identified by high-content screening. In total, six hits were identified by high-content screening. Three of them were able to correct the maturation and canalicular localization of two distinct intracellularly-retained ABCB4 variants; one molecule was able to significantly restore the function of two ABCB4 variants. In addition, in silico molecular docking calculations suggest that the identified hits may interact with wild type ABCB4 residues involved in ATP binding/hydrolysis. Our results pave the way for their optimization in order to provide new drug candidates as potential alternative to liver transplantation for patients with severe forms of ABCB4-related diseases, including PFIC3.

The measurement of ABCB4-mediated phosphatidylcholine (PC) secretion was already described 2,3 .Each tested condition was analyzed in triplicate and the amount of secreted PC was obtained after background subtraction.Then, final results were normalized to the expression levels of ABCB4 that were determined in parallel by immunoblot analyses of the corresponding cell lysates for each experiment.

Immunoanalyses in 384 well plates
Wells of 384-well plates were aspirated using a Bravo automated liquid handling platform (Agilent, Santa Clara, United States).Twenty microliters of rabbit anti-FLAG antibodies (Merck; 1:300 dilution) were added and plates were incubated at room temperature.After 1.5 hours, cells were washed four times with 65 µL phosphate-buffered saline (PBS) using a plate washer (Biotek EL406, Agilent, Santa Clara, CA, USA).Cells were fixed by adding 1% formaldehyde for 15 minutes and washed four times with PBS.Forty microliters of PBS containing 0.1% bovine serum albumin (BSA) were added.After 15 minutes, wells were emptied and 40 µL of a mixture of Hoescht 33342 (Thermo Fisher Scientific; 1 µg/mL) and anti-rabbit AlexaFluor TM 488-coupled secondary antibodies (Thermo Fisher Scientific; 1:400 dilution) in 0.05% BSA-containing PBS were added.After 1 hour, cells were washed eight times with a plate washer and plates were directly processed for image acquisition.) were stirred at room temperature in dichloromethane in presence of triethylamine (96 µL, 2.2 equiv.).After 24 hours and 72 hours, 40 µL (1 equiv.) of triethylamine were added.

Chemical synthesis
The reaction was complete after 7 days.The reaction mixture in dichloromethane was washed 3 times with H2O.The organic phase was dried on MgSO4 and concentrated under vacuum.

Cell viability assays
Toxicity assessment of drug candidates was based on a colorimetric assay using the tetrazolium salt 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), which is reduced to an insoluble purple precipitate (formazan) by metabolically active cells 4 , as published 2 .Briefly, HEK cells were seeded in a 96-well plate and treated with the candidate drugs at different concentrations (or DMSO as a control), including positive (1:5 DMSO) and negative (1:1,000 DMSO and no treatment) controls.After 16 hours of treatment, 100 μl of 0.25 mg/ml MTT (Merck, Saint-Quentin-Fallavier, France) was added to each well, and the cells were further incubated for 2 hours at 37°C.After incubation, culture media was washed out, and the cells were lysed in 100 μl of pure DMSO.Then, the absorbance at 550 nm, proportional to the number of alive cells in each well, was measured using a Wallac Victor 3 multilabel plate reader (PerkinElmer, Massy, France).

( 2 : 2 Supplementary
1:1) lipid bilayer and solvated in water with NaCl concentration at 0.154 M. ABCB4 models were minimized using a 4-step protocol: (i) water O-atom minimization, (ii) H-atom minimization, (iii) water minimization and (iv) whole system minimization.Ligands were initially optimized at the M06-2X/6-31+G(d,p) density functional theory to ensure proper bond distances and angles.This was achieved using the Gaussian16 software5 .Frequency calculations were then carried out to ensure local minima by the absence of imaginary frequency.Supplementary Fig. 2 Quantifications for high-content screening set-up.a-c The percentage of mCherry-ABCB4-positive cells (a) and signal intensities of mCherry (b) and Alexa 488 (anti-FLAG antibodies) (c) were determined from automated analyses using Columbus TM software.For each condition, means (± SD) of at least 48 independent wells (from 384-well plates) are represented.Cyclosporine A (CsA) was tested at 10 µM.Supplementary Fig. 3 Compound screening in mCherry-ABCB4-FLAG-I541Fexpressing HEK cells.Each 384-well plate was validated using strictly standardized mean difference (SSMD) values, calculated between DMSO-treated (controls) and 10 µM cyclosporin A (CsA)-treated conditions.The plates numbered A1 to A4 correspond to the Prestwick library, those numbered B1 to B10 to the second library.SSMD values were calculated using the mean µ and the variance σ of the distribution of Alexa 488 (anti-FLAG) signal intensities with the following equation: Fig. 13 Small compound rotatable dihedral angles for molecular docking calculations.Dihedral angle depicted in red were allowed to free rotate during molecular docking pose search.

Table 4 Number of calculated atomic contact fractions per residue between compounds #1, #2 and #3 and ABCB4 if from molecular docking calculations.
Contact analyses were performed over selected poses.

Supplementary Table 6 Predicted binding pockets of ABCB4 if obtained from PrankWEB online webserver, including color code for Supplementary Figures 9 & 11.
Predicted binding pocket with a probability lower than 0.1 were not considered for comparison with molecular docking calculations.

Supplementary Table 7 Predicted binding pockets of ABCB4 cc obtained from PrankWEB online webserver, including color code for Supplementary Figures 9 & 11.
Predicted binding pocket with a probability lower than 0.1 were not considered for comparison with molecular docking calculations.