Structural analogues of roscovitine rescue the intracellular traffic and the function of ER-retained ABCB4 variants in cell models

Adenosine triphosphate binding cassette transporter, subfamily B member 4 (ABCB4) is the transporter of phosphatidylcholine at the canalicular membrane of hepatocytes. ABCB4 deficiency, due to genetic variations, is responsible for progressive familial intrahepatic cholestasis type 3 (PFIC3) and other rare biliary diseases. Roscovitine is a molecule in clinical trial that was shown to correct the F508del variant of cystic fibrosis transmembrane conductance regulator (CFTR), another ABC transporter. In the present study, we hypothesized that roscovitine could act as a corrector of ABCB4 traffic-defective variants. Using HEK and HepG2 cells, we showed that roscovitine corrected the traffic and localisation at the plasma membrane of ABCB4-I541F, a prototypical intracellularly retained variant. However, roscovitine caused cytotoxicity, which urged us to synthesize non-toxic structural analogues. Roscovitine analogues were able to correct the intracellular traffic of ABCB4-I541F in HepG2 cells. Importantly, the phospholipid secretion activity of this variant was substantially rescued by three analogues (MRT2-235, MRT2-237 and MRT2-243) in HEK cells. We showed that these analogues also triggered the rescue of intracellular traffic and function of two other intracellularly retained ABCB4 variants, i.e. I490T and L556R. Our results indicate that structural analogues of roscovitine can rescue genetic variations altering the intracellular traffic of ABCB4 and should be considered as therapeutic means for severe biliary diseases caused by this class of variations.

ABCB4, also called MDR3, is a member of the ATP Binding Cassette (ABC) superfamily. Its expression is mainly restricted to the canalicular membrane of hepatocytes where its function is to flop phosphatidylcholine (PC) from the inner leaflet to the outer leaflet of the plasma membrane, thus allowing its secretion into bile (for reviews, see 1,2 ). At the molecular level, secreted PC has an essential role since it forms mixed micelles with the other co-secreted hydrophobic bile components, such as bile salts and cholesterol, as first proposed by Admirand and Small 3 . Indeed, impairment of PC secretion into bile leads to the formation of cholesterol crystals and gallstones, as well as loss of protection from the detergent effects of free bile salts on biological membranes of the biliary tree 1,4,5 . These clinical characteristics have been largely reported for patients with rare biliary diseases such as progressive familial intrahepatic cholestasis type 3 (PFIC3), low-phospholipid associated cholelithiasis syndrome or intrahepatic cholestasis of pregnancy [6][7][8] . These diseases are correlated with genetic variations of the ABCB4 gene, most of them being missense variations (see 2,9,10 and references therein, as well as http://abcmutations.hegelab. org/proteinDetails?pid=63). For patients with ABCB4-related biliary diseases, the only pharmacological treatment is the administration of ursodeoxycholic acid (UDCA), a bile acid with low hydrophobicity 11 . Although this therapy is efficient in the milder forms of the diseases 12,13 , UDCA is not -or poorly -efficient in the majority of identified in a homozygous PFIC3 patient 7 , and further characterized in our laboratory 17,26 . Indeed, ABCB4-I541F has been shown to be retained in the ER as an immature and high-mannose glycosylated protein 26 , characterized by the absence or low abundance of a mature protein band on immunoblot, compared to the WT protein (Fig. 1A). We have previously demonstrated that the maturation and the localisation at the canalicular membrane of ABCB4-I541F could be partially rescued upon temperature shift of cell culture at 27 °C or treatment with the ABCB1/MDR1 substrate CsA 17,26 . Since roscovitine at 100 µM was described to correct the function of the F508del ER-retained CFTR/ABCC7 variant 25 , we investigated the potential correction of ABCB4-I541F by this molecule. Such treatment led to the correction of the maturation of ABCB4-I541F in HEK cells (Fig. 1A), as shown by the quantification of immunoblots (Fig. 1B). These results were confirmed by indirect immunofluorescence analyses in HepG2 cells, a hepatocellular carcinoma derived cell line forming pseudo-bile canaliculi in cell culture 27 . Indeed, the increased maturation of ABCB4-I541F was associated with its relocalisation at the canalicular membrane after treatment with 100 µM roscovitine in HepG2 cells (Fig. 1C). However, roscovitine displayed important dose-dependent cytotoxicity (Fig. 1D) and inhibition of ABCB4-WT-mediated PC secretion activity in HEK cells (Fig. 1E), whichcompromises its relevance for further investigations.
Structural analogues of roscovitine rescue the maturation and the canalicular localisation of ABCB4-I541F. The cytotoxic effect induced by roscovitine might be explained by its CDK inhibition activity 22 . We thus synthesized structural analogues ( Fig. 2A) potentially devoid of this feature. Aftin-4 is a methyl-roscovitine with reduced kinase inhibition activity 28,29 , and M3 is the main hepatic metabolite of roscovitine. All other products are carboxylated analogues displaying much reduced kinase inhibition compared to the parent compound, roscovitine (Table 1). This indicates that carboxylation at the hydroxyl level and/or methylation at position N6 ( Fig. 2A and Supplementary Fig. S1) strongly reduce CDK inhibition activity of these molecules. As expected, all roscovitine analogues with decreased CDK inhibition activity were less cytotoxic (Table 1 and Supplementary Fig. S1). Interestingly, in HEK cells, some of these analogues (MRT2-235, -237, -239, -241, -243, -245 and -249) were able to significantly increase the expression of a mature form of ABCB4-I541F while other analogues (Aftin-4, M3, MRT2-163 and MRT2-164) were not (Fig. 2B). The quantification of these results indicated a ~2.5 fold increase of expression of the mature form of ABCB4-I541F after treatment with 100 µM of the most potent roscovitine analogues (Fig. 2C). We then analysed the efficacy of the most potent analogues to correct the canalicular targeting of ABCB4-I541F in HepG2 cells. After 16 hours of treatment with 100 µM of these compounds, we observed partial relocalisation of ABCB4-I541F at bile canaliculi, as shown by its partial co-localisation with ABCC2 (Fig. 3). Altogether, these results demonstrate that the maturation and the canalicular localisation of ABCB4-I541F can be rescued by new structural analogues of roscovitine.
Structural analogues of roscovitine are less cytotoxic and less inhibitory of ABCB4 activity than roscovitine. To pursue this study, we decided to focus on three roscovitine analogues (MRT2-235, MRT2-237 and MRT2-243) which displayed significant rescue of the maturation and partial relocalisation of ABCB4-I541F at bile canaliculi (Figs 2B,C and 3). At concentrations lower than 100 µM, these analogues were much less cytotoxic than roscovitine in HEK cells (Fig. 4A). Indeed, the cell viability was still higher than 80% at 25 µM of the analogues while it was reduced to less than 60% at the same concentration for roscovitine (Fig. 4A). Furthermore, measurement of ABCB4-WT-mediated PC secretion in HEK cells indicated that the three roscovitine analogues were less inhibitory than the original molecule when used at 10 µM and 25 µM (Fig. 4B). The IC 50 on ABCB4-WT activity was 7.5 µM for roscovitine while it was 2.5 to 4.0 fold higher for its analogues ( Fig. 4C and Table 2). It is also important to notice that treatment with roscovitine and these three analogues at 10, 25 and 100 µM do not www.nature.com/scientificreports www.nature.com/scientificreports/ alter the plasma membrane localisation of ABCB4-WT in HEK cells ( Supplementary Fig. S2). Thus, inhibition of ABCB4-mediated PC secretion by roscovitine and its analogues may be due to a direct effect of these compounds on ABCB4 function. Altogether, these results indicate that roscovitine analogues might be interesting ABCB4 correctors with low cytotoxicity and low inhibition activity of ABCB4. www.nature.com/scientificreports www.nature.com/scientificreports/ Roscovitine analogues rescue the maturation, the localisation and the activity of ABCB4-I541F.
The inhibition of ABCB4 activity induced by MRT2-235, MRT2-237 and MRT2-243 when used at 100 µM led us to investigate these analogues at lower concentrations. After treatment with 5, 10 and 25 µM of the three analogues, we observed a dose-dependent correction of ABCB4-I541F maturation in HEK cells ( Fig. 5A; quantification in Fig. 5B). Treatment with 25 µM of these analogues triggered partial relocalisation of ABCB4-I541F at bile canaliculi in HepG2 cells (Fig. 5C). Finally, we measured the PC secretion activity of ABCB4-I541F in HEK cells after treatment with or without roscovitine analogues at 25 µM. Importantly, we noticed a marked correction www.nature.com/scientificreports www.nature.com/scientificreports/ of ABCB4-I541F activity from 7.1 ± 0.8% of residual activity with control treatment (vehicle) to 19.0 ± 3.4%, 13.0 ± 3.5% and 22.5 ± 6.2% after treatment with 25 µM of MRT2-235, MRT2-237 and MRT2-243, respectively ( Fig. 5D). Altogether, our results demonstrate that the three selected roscovitine analogues are able to retrieve the maturation and the canalicular localisation of the ER-retained ABCB4-I541F variant and to significantly correct its PC secretion activity.

Roscovitine analogues rescue the function of two other ER-retained ABCB4 variants.
To extend this study, we analysed the effect of roscovitine analogues on two other ER-retained missense ABCB4 variants, I490T and L556R, identified in patients with liver cancer 30 or PFIC3 7 . Without treatment, these variants were mainly detected as immature proteins by immunoblot in HEK cells (Fig. 6A,B). However, after treatment with 5 to 25 µM of the three roscovitine analogues, we observed a dose-dependent rescue of the maturation of these two variants (Fig. 6A,B), as confirmed by the quantification of these experiments (Fig. 6C,D). These results were further validated by indirect immunofluorescence assays in HepG2 cells: both I490T and L556R variants were partially relocalised at bile canaliculi upon treatment with 25 µM of MRT2-235, MRT2-237 or MRT2-243 (Fig. 6E,F). As expected, the PC secretion activity of these two ER-retained variants in HEK cells was strongly impaired in vehicle-treated cells (Fig. 6G,H). However, after treatment with 25 µM of the three roscovitine analogues, we observed a significant increase of the residual activity of these variants (Fig. 6G,H). These results provide evidence that roscovitine analogues are able to rescue the localisation, the maturation and the PC secretion activity of several ER-retained ABCB4 variants.

Discussion
PFICs and hereditary cholestatic diseases in general represent a considerable fraction of indications for liver transplantation 31,32 . In the absence of surgery and efficient pharmacological treatment able to cure patients, or at least to delay liver transplantation, the life expectancy is low 33 . Roscovitine has been shown to partially rescue intracellular traffic and chloride channel activity of CFTR-F508del 25 . Thus, in this work, we investigated the possibility to repurpose this molecule in the frame of ABCB4-related diseases. While we observed that roscovitine was able to correct maturation and localisation of the ER-retained ABCB4-I541F variant in cell models, its important cytotoxicity and inhibition of ABCB4-WT function ruled out its use for further analyses. We then synthesized structural analogues of roscovitine, devoid of intrinsic CDK inhibition capacity (Table 1). Interestingly, three of these analogues (MRT2-235, MRT2-237 and MRT2-243) were less cytotoxic and less inhibitory of ABCB4-WT activity than the original molecule, and allowed partial rescue of the maturation and the localisation of distinct ER-retained ABCB4 variants. More importantly, these corrections were associated with a significant rescue of PC secretion for these variants. Thus, roscovitine analogues may be considered as good candidates in preclinical studies.
We and others have previously shown that cyclosporins, essentially CsA, are able to rescue the maturation and the cell surface localisation of ABCB4 defective variants 9,17-19 . However, until now, attempts to rescue functional ABCB4 variants using substrates and/or inhibitors of ABCB1 such as CsA, FK506/Tacrolimus or Tariquidar were not successful ( 18 ; our unpublished data), strengthening the requirement for developing new targeted pharmacotherapies. Here, we provide evidence that roscovitine analogues are efficient correctors of the intracellular traffic as well as the function of ER-retained ABCB4 variants. The chemical chaperone 4-phenylbutyrate, already used in clinics 34,35 , has been shown to rescue other ER-retained ABCB4 variants 36 , as well as defective variants of different ABC transporters ( 10 and references therein). However, for patients who do not respond or comply to this therapy, or develop major side effects 37,38 , roscovitine analogues may constitute good alternatives in the frame of personalized medicine.
Roscovitine has been shown by direct co-cristallisation approaches to bind to CDK2 and CDK5 in their catalytic site, and to directly compete with ATP 39,40 . We can speculate that ABCB1 substrates, such as CsA, correct ABCB4 defective variants but inhibit its function by similar mechanisms. In the present study, while roscovitine www.nature.com/scientificreports www.nature.com/scientificreports/ analogues allowed a better folding of ER-retained ABCB4 variants and thus their cell surface relocalisation, these molecules could impair the PC secretion activity by additional molecular mechanisms. Thus, roscovitine analogues seem to be dual molecules with two distinct functions: one allowing a better folding and traffic of ER-retained ABCB4 variants and the other interfering with the PC secretion activity of the transporter. www.nature.com/scientificreports www.nature.com/scientificreports/ Roscovitine analogues may correct ABCB4 function through different molecular mechanisms, compared to the original molecule. Indeed, we observed that the immature band of ABCB4 is less pronounced after treatment with roscovitine, which was not the case with treatment at 27 °C, CsA or roscovitine analogues (Figs 1A and 2B). www.nature.com/scientificreports www.nature.com/scientificreports/ This would reflect the involvement of distinct molecular mechanisms between roscovitine and its structural analogues. We also observed different rescuing efficiencies of roscovitine analogues on the three distinct ABCB4 variants (compare Figs 5A,B and 6A-D) which may reflect different structural defects induced by the missense variations. It is interesting to note that the presence of a carboxyl which prevents kinase inhibition ( Fig. 2A Table 2. IC 50 of roscovitine and its analogues on the PC secretion activity of ABCB4-WT. IC 50 were calculated from dose-response curves shown in Fig. 4C. www.nature.com/scientificreports www.nature.com/scientificreports/ does not prevent rescue of ER-retained ABCB4 variants. Moreover, the presence of a cyclopentyl at position 9 and substitutions (F, Cl, CF3) seem to contribute to the ABCB4 rescuing efficacy of the analogues. Although roscovitine has been reported to partially correct CFTR-F508del function 25 , we did not observe such effect on ABCB4-I541F. These results may illustrate the difference of molecular mechanisms between the floppase activity of ABCB4 and the chloride channel function of CFTR. However, here, we provide evidence that roscovitine analogues are able to partially correct the function of ER-retained ABCB4 variants even if the mechanisms of their cell entry, passive diffusion or active uptake are not yet characterized. This opens the perspective to the use of these and related compounds as targeted pharmacotherapies for patients with ABCB4-related diseases.
The research of pharmacological therapies for ABCB4-related diseases as alternatives to liver transplantation is mainly focused on PFIC3, the most severe form of these diseases, for which the majority of homozygous or compound heterozygous patients have defects in both ABCB4 alleles (for a review, see 41 ). PC concentrations in bile of PFIC3 patients are very low due to a strongly impaired ABCB4 activity 7 . Thus, targeting patients harbouring this category of genetic variations using roscovitine analogues could be a promising therapeutic strategy in order to improve PC secretion into bile. Indeed, it has been proposed that patients with enough residual ABCB4 activity (≥7%), are better responders to UDCA therapy 42 , as this is the case for patients with milder forms of ABCB4-related diseases such as low-phospholipid associated cholelithiasis syndrome and intrahepatic cholestasis of pregnancy [41][42][43][44] . It is thus tempting to speculate that the partial rescue of ABCB4 function using roscovitine analogues would be sufficient for patients to positively respond to UDCA therapy. It is also interesting to note that UDCA treatment may contribute to the relocalisation of ABCB4 at the canalicular membrane of hepatocytes in in vivo systems, as proposed for ABCB11 and ABCC2 in cholestatic rat models 45,46 and ABCB11 in PFIC2 patients 47 .
The effect of roscovitine are currently evaluated in preclinical and clinical trials in the frame of several pathologies 48 , including a recent phase 2 clinical trial in patients with cystic fibrosis (ROSCO-CF; https://clinicaltrials. gov/ct2/show/NCT02649751). The results obtained from these studies would allow to consider the repositioning of roscovitine, and more especially its more potent structural analogues, to treat ABCB4-related rare biliary diseases.

Materials and Methods
DNA constructs and mutagenesis. The subcloning of wild type (WT) ABCB4, isoform A (NM_000443.3), into pcDNA3 vector has been described 26 . The I541F, I490T and L556R missense ABCB4 variants were previously reported and described 9,26,49 . Site directed mutagenesis were performed using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Les Ulis, France), following manufacturer's instructions, and using the primers (Eurogentec, Angers, France) described in Supplementary Table S1. The sequences of all constructs were systematically verified by automated sequencing.
For transient transfection of HEK cells, they were seeded at subconfluent levels in the adequate tissue culture wells at least six hours before transfection. Turbofect (Thermo Fisher scientific) was used at a ratio of reagent:DNA of 2:1 according to manufacturer's instructions. For transient transfection of HepG2 cells, subconfluent cultures were seeded in the adequate culture wells 24 hours before transfection. Lipofectamine 3000 (Thermo Fisher Scientific) was used at a ratio of reagent:DNA of 1.5:1 according to manufacturer's instructions. Cell treatments and processing for further analyses were performed at least 16 hours post-transfection.  (Roscoff, France). Each compound was solubilized in dimethylsulfoxide (DMSO) as 1000X concentrated stock solutions in order to treat cells with 5 µM to 100 µM final concentrations, using DMSO as control vehicle at the same dilution (0.1% DMSO for all conditions). The cells were treated 24 hours post-transfection with these drugs during 16 hours, except for the viability assays for which cells were treated during three days. After drug treatment, cells were used for immunoanalyses, cell viability assays or PC secretion assays.

Chemicals and cell treatments. Cyclosporin
Protein kinases and their activators or regulators (see Table 1) were expressed and purified, and their catalytic activity was assayed in the presence of a range of concentrations of each roscovitine analogue, as previously described 28,29,40 . IC 50 values were calculated from the dose-response curves.
Immunoanalyses. Western blots and indirect immunofluorescence were performed as previously described 9,50 . For immunoblot analyses, total cell lysates were prepared in denaturing and reducing sample buffer 51 , separated on 7.5% SDS-PAGE and transferred on nitrocellulose membranes using Trans-Blot system (Bio-Rad Laboratories, Hercules, CA). Saturated membranes were incubated with mouse monoclonal anti-ABCB4 (clone P 3 II-26; Enzo Life Sciences, Villeurbanne, France) or anti-α-tubulin (clone 1E4C11; ProteinTech, Manchester, United Kingdom) antibodies and then with peroxidase conjugated anti-mouse secondary antibodies (Sigma). Signals were detected with ECL prime western blotting detection reagent (GE Healthcare, www.nature.com/scientificreports www.nature.com/scientificreports/ Velizy-Villacoublay, France) and quantified in the linear range of detection using ImageJ 1.50i software (U.S. National Institutes of Health, Bethesda, MD).
For indirect immunofluorescence, HepG2 cells were grown on glass coverslips and after transfection and treatments, they were fixed and permeabilized during 1 min at −20 °C in ice-cold methanol. Then ABCB4 and ABCC2 were immunolabelled using the mouse monoclonal P 3 II-26 (IgG2b) and M 2 I-4 (IgG1) antibodies (Enzo Life Sciences), respectively. AlexaFluor ® 555 and AlexaFluor ® 488 conjugated isotype-specific secondary antibodies (Thermo Fisher Scientific) and Hoechst 33342 (Thermo Fisher Scientific) were used to label ABCB4, ABCC2 and nuclei, respectively. Images were acquired using an IX83 inverted fluorescence microscope from Olympus (Rungis, France), equipped with a UPLSAPO 60XS2 silicone immersion objective and a Hamamatsu ORCA Flash4.0 digital CMOS camera, and analysed using Olympus CellSens Dimension Desktop version 1. 16 and Adobe Photoshop version 8.0.1. For each experiment, all images were acquired with constant settings (acquisition time and correction of signal intensities).
Cell viability assays. Cell viability was assessed by the conversion of MTT (3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide) into formazan crystals by living cells, as described 52 . In brief, HEK cells were seeded in 96-well plates in triplicate for each tested condition, including controls (no cells, no treatment, treatment with vehicle). After transient expression of ABCB4-WT as described above (to mimic the experimental conditions of the other experiments) and drug treatment during 72 hours, 125 µg/ml MTT (final concentration) was added in each well and cells were re-incubated at 37 °C during 2 hours. These conditions were optimized in order to maintain the absorbance at 540 nm of the blanks below 0.1 OD unit (negative controls) and the absorbance at 540 nm for untreated cells between 1.0 and 1.3 OD units (positive controls). Then culture media were gently washed out, cells were lysed in 100 µL of pure DMSO and absorbance at 540 nm was measured using a multiplate cytofluorimeter SpectraFluor from Tecan (Männedorf, Switzerland). Cell viability was calculated for each triplicate after background subtraction and expressed as percentage of the mean for cells treated with vehicle only.
Measurement of ABCB4-mediated phosphatidylcholine secretion. HEK cells were seeded in 0.01% poly-L-lysine (Sigma) pre-coated 6-well plates of 10 cm² area per well (1.3 × 10 6 cells per well). After transient transfection of the empty vector (control), plasmids encoding ABCB4-WT or its variants, cells were incubated during 24 hours at 37 °C in serum-free medium supplemented with 0.5 mM sodium taurocholate (Sigma) and 0.02% fatty acid-free bovine serum albumin (Sigma). Then, the secreted PC was quantified from the collected media using a fluoro-enzymatic assay previously described 50 . Each condition was analysed in triplicate, after background subtraction, and results were normalized to the expression levels of ABCB4 determined by immunoblots of the corresponding cell lysates.
Statistics. Graphics and non-parametric analyses of variance tests (Kruskall-Wallis) were performed using Prism version 7.00 (GraphPad Software, La Jolla, CA). A P value of less than 0.05 was considered significant. If not specified in figure legends, symbols indicate the comparison between the control (or vehicle-treated) and the other tested conditions.

Data Availability
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.