Interactions of anti-COVID-19 drug candidates with hepatic transporters may cause liver toxicity and affect pharmacokinetics

Transporters in the human liver play a major role in the clearance of endo- and xenobiotics. Apical (canalicular) transporters extrude compounds to the bile, while basolateral hepatocyte transporters promote the uptake of, or expel, various compounds from/into the venous blood stream. In the present work we have examined the in vitro interactions of some key repurposed drugs advocated to treat COVID-19 (lopinavir, ritonavir, ivermectin, remdesivir and favipiravir), with the key drug transporters of hepatocytes. These transporters included ABCB11/BSEP, ABCC2/MRP2, and SLC47A1/MATE1 in the canalicular membrane, as well as ABCC3/MRP3, ABCC4/MRP4, SLC22A1/OCT1, SLCO1B1/OATP1B1, SLCO1B3/OATP1B3, and SLC10A1/NTCP, residing in the basolateral membrane. Lopinavir and ritonavir in low micromolar concentrations inhibited BSEP and MATE1 exporters, as well as OATP1B1/1B3 uptake transporters. Ritonavir had a similar inhibitory pattern, also inhibiting OCT1. Remdesivir strongly inhibited MRP4, OATP1B1/1B3, MATE1 and OCT1. Favipiravir had no significant effect on any of these transporters. Since both general drug metabolism and drug-induced liver toxicity are strongly dependent on the functioning of these transporters, the various interactions reported here may have important clinical relevance in the drug treatment of this viral disease and the existing co-morbidities.


Results
Interaction of potential anti-COVID-19 drugs with hepatic ABC transporters. MRP2 drug interactions were characterized by following the inhibitory effects of drugs on ATP-and benzbromarone-sensitive CDCF uptake into inverted membrane vesicles generated from MRP2-overexpressing Sf9 cells ( Fig. 2A and Table 1). We found that IVE strongly decreased the accumulation of the probe substrate at higher than 10 µM concentration, while LOP and FAV had no such effect. Interestingly, RIT and REM did not inhibit but rather increased the probe substrate transport at high concentrations, possibly due to an allosteric effect 32,33 . Since reduced glutathione (GSH) levels may affect MRP2-dependent substrate transport or drug interactions, we have www.nature.com/scientificreports/ also examined these transport processes in the presence of 2 mM GSH, but neither the vesicular uptake of CDCF, nor its inhibition by the drugs examined were affected (data not shown). MRP3 drug interactions were characterized by measuring ATP-and benzbromarone sensitive CDCF uptake into inverted HEK-293 membrane vesicles generated from MRP3-overexpressing HEK-293 cells ( Fig. 2B and Table 1). Only RIT inhibited the MRP3 transport function at a lower than 10 µM concentration, whereas LOP and IVE were effective only at higher concentrations (> 20 µM), while FAV and REM did not influence MRP3 transport at all.
MRP4 drug interactions were characterized by measuring ATP-sensitive DHEAS (dehydroepiandrosteronesulfate) uptake into inverted membrane vesicles generated from MRP4 overexpressing HEK-293 cells ( Fig. 2C and Table 1). We found that most drugs examined had no significant inhibitory effect on MRP4-dependent substrate transport (a maximum of 10-30% inhibition was observed by up to 50 µM of RIT, LOP, IVE or FAV), while REM had a strong inhibitory action on this transporter, with an IC 50 of about 2.3 µM (Fig. 2C).
BSEP drug interactions were characterized here by measuring the effects of drugs on ATP-and glyburide sensitive taurocholate (TC) uptake into inverted cell membrane vesicles generated from BSEP overexpressing HEK-293 cells ( Fig. 2D and Table 1). We found that LOP and RIT strongly inhibited TC transport at low concentrations (see Table 1 for IC 50 values) whereas IVE and REM were only effective above 20 µM. FAV did not modify BSEP-mediated TC transport. Cellular assays for SLC type transporters. OATP1B1 (Fig. 3A) and OATP1B3 (Fig. 3B) drug inhibition was studied by measuring the uptake of the fluorescent model substrate, pyranine, into A431 cells overexpressing the respective transporters and into mock-transfected control cells. As shown on Fig. 3, the drug inhibition panel for the two OATPs examined was somewhat different: IVE was less inhibitory for OATP1B1 than for OATP1B3, and RIT, LOP and REM had strong inhibitory effect at low micromolar range but with slightly different IC 50 values for both transporters (see Table 1). FAV did not modify OATP functionality in our assay. MATE1 (SLC47A1) interaction was characterized by metformin uptake into MDCKII cells overexpressing MATE1 and in mock-transfected controls. As shown in Fig. 4A, LOP, RIT and REM had a strong inhibitory action on this transporter at the low micromolar range, while IVE and FAV showed practically no inhibition of MATE1 transport activity (see also Table 1). Table 1. Summary of the transporter modulation properties of the repurposed anti-COVID-19 drugs examined in transport assays. Approximate IC 50 (μM) values were determined by nonlinear regression analysis of the data shown in the results section, using GraphPad prism software (version 5.01, GraphPad, La Jolla, CA, USA). Favipiravir (not listed) had no effect on any of the transporters examined. Estimation of the DDI effects was performed as described in the Methods section, and references for in vivo parameters were provided in the Discussion. For importers the R values (cut off value 1.1), while for the exporters C max /IC 50 values (cut off value 0.1) were calculated. For lopinavir and ritonavir R values were calculated for a combined treatment. www.nature.com/scientificreports/ OCT1 (SLC22A1) interaction was measured by metformin uptake into HEK-293 cells overexpressing OCT1 and mock-transfected controls. As shown in Fig. 4B, RIT and REM had a strong inhibitory effect on this transporter (below 10 µM), while LOP was less inhibitory, and IVE and FAV had no significant inhibitory action on OCT1 transport (see also Table 1).
NTCP/SLC10A1 interaction was investigated in HEK-293 cells overexpressing NTCP, where taurocholate (TC) uptake was measured in NTCP-overexpressing and in mock-transfected control cells. As shown in Fig. 4C, the repurposed anti-COVID agents had relatively weak inhibitory action on this uptake transporter: RIT, LOP and REM had IC 50 values at or above 50 µM, and IVE and FAV showed only minor inhibition, and observed only at higher concentrations.

Discussion
In the present work we have studied transporter interactions of several repurposed potential anti-COVID-19 agents. We focused on key membrane transporters involved in the liver uptake and excretion of endo-and xenobiotics and applied well documented in vitro assay techniques to characterize these interactions. We selected lopinavir, ritonavir, ivermectin, remdesivir and favipiravir as promising repurposed anti-COVID-19 agents for these studies. Some of the data presented here reinforces previously described transporter-drug interactions. In addition, our study on the key hepatocyte transporters working in a network at the uptake and export sites, may help decipher both potential risks of drug-induced liver injury (DILI) and alterations in general hepatocellular functions. Importantly, in the case of the clinically most promising agents, remdesivir and favipiravir, our findings may serve as a valuable addition to the previously scarce information about transporter interaction of these drugs.
In the case of ABC transporters, we employed a vesicular transport assay system based on inside-out vesicles generated of membrane isolates in which modulatory effects of the drugs on the accumulation of model substrates could be quantitatively examined. For SLC-type transporters we used whole-cell assays, in which drug modulation of a transporter specific substrate uptake was measured. In order to provide an estimation of potential in vivo drug effects from in vitro data, we followed the recommendations of the International Transporter Consortium (ITC) 34 and used the available pharmacokinetic data for the given compounds, to indicate potential drug-drug interaction risks (DDI) 34 . We used the recommended in vivo-in vitro extrapolation approach, based on the R values for OATP transporters and the C max /IC 50 values for ABC transporters (see the Methods section). In this extrapolation, our IC 50 values and the FDA-published pharmacokinetic data for these drugs were used.
In harmony with earlier investigations, our results show that IVE has a strong in vitro inhibitory effect on ABCB1, ABCG2 and OATP1B3 transporters ( 13,35-37 , Table 1). According to the FDA guideline for stromectol (IVE), we calculated the possible drug influence using a 165 µg/kg IVE dose and a peak plasma concentration value of 46 ng/ml (https:// www. acces sdata. fda. gov/ drugs atfda_ docs/ label/ 2008/ 05074 2s022 lbl. pdf). The predicted R value for IVE in the case of OATP1B3 (1.07) is near to the 1.1 cut-off value, and the C max /IC 50 value is above the 0.1 cut-off value for ABCB1 (0.17) but not for ABCG2 (0.047). These data suggest that higher dose IVE treatment may increase the risk of DDI related to the interactions with these transporters. As widely accepted, mainly the ABCB1/P-gp and the ABCG2/BCRP efflux transporters are responsible for reduced human IVE toxicity, by restricting the absorption of this drug in the intestine and protecting the brain as part of the blood-brain barrier 35,36,38,39 . Based on our results, MRP2 and BSEP inhibition appears only at higher IVE concentrations (above 20 µM), which probably does not occur in vivo, but compromised ABCB1 and ABCG2 transporter function (mutations, inhibition) may increase available drug plasma levels.
LOP and IVE were suggested to be used in combination (Kaletra) for antiviral treatment 5 . It has been reported that these drugs strongly inhibited several drug transporters in a wide array of in vitro assays 17,37,[40][41][42][43][44][45] , and here we have also found that LOP and RIT inhibited several hepatic transporters with high affinity (IC 50 -s below 10 µM, see 31 and Table 1). According to an FDA dataset for Kaletra, the potential treatment dose is 400 mg lopinavir and 100 mg ritonavir, added in combination as an antiviral treatment (https:// www. acces sdata. fda. gov/ drugs atfda_ docs/ label/ 2013/ 02125 1s046_ 02190 6s039 lbl. pdf). This dose results in about 11-20 µM of lopinavir and 1-1.73 µM of ritonavir peak plasma concentrations in vivo 46 . The estimated R values for OATP1B1 (2.39, combined) and OATP1B3 (1.74, combined), and the C max /IC 50 values for ABCB1 (18.6 for LOP, 2.66 for RIT) and ABCG2 (3.33 for LOP, 0.13 for RIT) indicate that LOP and RIT may exceed the cut-off values of these transporters and thus, inhibition of these transporters can be a relevant factor in DDI risk. Though the RIT dosage is relatively low in Kaletra, the measured low IC 50 values indicate that RIT may exacerbate lopinavir-mediated transporter inhibition in the combination treatment.
Estimation of clinical relevance from the in vitro inhibition data regarding the other investigated transporters is more difficult. However, the low in vitro IC 50 values (Table 1) for the OCT1 and the MATE1 transporters raise the possibility of potentially compromised in vivo function of these transporters. Previously, RIT was classified as a highly potential DILI causing compound, based on information in the NIH LiverTox Database 47 (livertox. nih.gov) and the Liver Toxicity Knowledge Base (fda.gov), although the cause of the observed liver enzyme elevations during ritonavir therapy has not yet been fully understood. The risk of DILI by RIT and LOP may occur via several mechanisms. It is known that these drugs are extensively metabolized via CYP3A4, and the production of toxic intermediates as well as their interactions with multidrug transporters may be involved in liver toxicity. Another significant source of DILI can be a compromised bile acid handling. If hepatocellular concentrations of these drugs reach the micromolar range, the bile acid exporter activity of BSEP may be decreased, because the IC 50 values of LOP and RIT regarding BSEP were 7.8 and 3.2 µM, respectively. Indeed, even low potency BSEP inhibition may be biologically significant if the local drug exposure is high. It has been reported that basolateral MRP3 and MRP4 provide a compensatory mechanism for bile acid efflux in the case of BSEP inhibition 48 and, although these transporters also interact with RIT and LOP, the high IC 50 values indicate that RIT and LOP may not inhibit the function of these transporters in vivo. However, the combined inhibition of BSEP and these transporters may lead to a clinically significant DILI, especially in the case of a wide-spread inflammation caused by a Sars-CoV-2 infection. In addition, pharmacokinetics of other drugs, optionally co-dosed in COVID patients, may be influenced by LOP/RIT. For example the transport of OATP1B substrate drugs, including statins, repaglinide, olmesartan and valsartan 28,49,50 , or the OCT1-and MATE1-mediated metformin transport 51 could also potentially be affected by LOP and/or RIT. Consistently with in vitro findings of transporter inhibition studies, Macias et al. found in a cross-sectional study a high overall frequency of DDI between ritonavir-boosted lopinavir for treating COVID-19 and other relevant medications such as statins, antibiotics, corticosteroids, or antiarrhythmic drugs 52 .
Since REM is a most promising repurposed drug in treating COVID-19, in this work we carefully assessed potential transporter inhibitions by this compound in various assay systems. REM did not show a significant inhibition of the ABC transporters, while we observed strong inhibition of OATP transporters (IC 50  www.nature.com/scientificreports/ and 1B3 were 2.9 and 4.3 µM, respectively) ( Table 1). Moreover, OCT1-and MATE1-mediated in vitro metformin transport was inhibited by REM in a concentration-dependent manner with an IC 50 of 6.1 and 2.3 µM. These in vitro IC 50 concentrations are close to the peak plasma concentration observed 30 min after intravenous administration of 200 mg REM for 30 min 53 , or close to the 3.7 µM C max after intravenous administration of a clinically applied dose of 100 mg remdesivir in healthy adults (https:// www. acces sdata. fda. gov/ drugs atfda_ docs/ label/ 2020/ 21478 7Orig 1s000 lbl. pdf). Based on our data, a combined inhibition of these transporters may affect the uptake and the biliary extrusion of both anionic and cationic drugs as well as endogenous substates, such as bile acids. REM also inhibited the NTCP-mediated uptake of taurocholate in a concentration-dependent manner. As NTCP and BSEP inhibition in vitro occurred only at high REM concentrations (40-50% inhibition at 50 µM) probably not achieved in vivo, this effect may not be considered clinically relevant. REM was shown to inhibit MRP4 in vitro (IC 50 was 2.9 µM), whereas it did not inhibit either MRP2 or MRP3. Since MRP4 has the ability to transport nucleoside analogue drugs 21,54 , this relatively selective effect was not unexpected. Regarding MRP4, there are no specific citations of clinically relevant DDIs ascribed to this transporter, and there is only limited information on the role of MRP4 in the clinical ADME of drugs. However, MRP4 is a multispecific pump extruding endogenous metabolites, including bile acids, urate, and conjugated steroid hormones into the bile canaliculi, therefore its potential involvement in adverse effects and DDI cannot be excluded. When treating co-morbidities in COVID-19, clinically important MRP4 drug substrates include cephalosporin, several antibiotics, diuretics like furosemide and hydrochlorothiazide, as well as olmesartan 18,[21][22][23][54][55][56][57][58] . In addition, MRP4 inhibitors include non-steroidal anti-inflammatory and cardiovascular drugs 54 . It should be noted that while in this work we focused on liver resident transporters, MATE1 and MRP4 are also highly expressed in the basolateral renal tubular membranes and these proteins have an important role in renal drug excretion. The current FDA guideline recommends evaluation of MATE1mediated interactions for drugs that undergo significant renal secretion (≥ 25% of total systemic clearance). Thus, REM inhibition of MRP4 may have multiple effects on general uric acid and drug clearance. In addition, MRP4 in the intestinal epithelia is involved in the oral absorption of various drugs, while in the blood-brain barrier MRP4 has a protective role for the central nervous system from drugs 20,57 .
In our REM studies, the data support earlier observations (https:// www. ema. europa. eu/ en/ docum ents/ other/ summa ry-compa ssion ate-use-remde sivir-gilead_ en. pdf) on the inhibition of OATP1B1, OATP1B3, BSEP, MRP4, and NTCP, while MATE1 inhibition is a new information. Since REM is rapidly cleared (T1/2 = 69 min) 53 from the human body, the potential of REM to be a perpetrator of clinically significant drug-drug interactions through these hepatic transporters is limited.
As a summary, we found that the examined repurposed anti-COVID-19 agents, with the exception of FAV, have multiple inhibitory effects on hepatic membrane transporters. These results may significantly help to decipher the effects of these transporter-drug interactions on general drug clearance and drug-induced liver toxicity. In addition, the various interactions quantitatively assessed here should be considered in clinical drug treatment strategies for the COVID-19 disease and for understanding existing co-morbidities.

Methods
The source of materials is provided in the detailed descriptions of the methods applied. Lopinavir, ritonavir and ivermectin were obtained from Sigma Aldrich Inc.; favipiravir (T-705) from MedChem Express; remdesivir was kindly provided by Lajos Szente, Cyclolab Ltd, Budapest, Hungary.
For the ABC-type transporter assays we used isolated membrane vesicle-based functional assays. ABCC2/ MRP2 vesicles were prepared from Sf9 cells infected by the human MRP2 coding baculovirus using a previously established method 33 . ABCB11/BSEP, ABCC3/MRP3 and ABCC4/MRP4 membranes were prepared from transporter expressing HEK-293 cells by Solvo Biotechnology (Budapest, Hungary). The vesicular uptake of transporter specific substrates was measured, and transporter related uptake was defined as the difference between the uptake with and without ATP (in the presence of AMP). Transporter specific inhibitors were also applied to control the specificity of transporters. For BSEP, radiolabeled taurocholate (1.5 µCi/sample of radiolabeled compound plus 2 µM unlabeled TC) (PerkinElmer Co); for MRP2 and MRP3, CDCF (5(6)-Carboxy-2′,7′dichlorofluorescein, Sigma) (5 µM); for MRP4, radiolabeled dehydroepiandrosterone sulfate (DHEAS) (0.2 µCi radiolabeled compound plus 0.5 µM unlabeled DHEAS) (PerkinElmer Co) substrates were used. Applied specific inhibitors were 100 µM glyburide for BSEP, 40 µM benzbromarone for MRP2 and MRP3, and 50 µM MK571 for MRP4. A 30 µg (MRP2, MRP3) or 50 µg (BSEP, MRP4) protein/sample was incubated for 10 or for 1.5 min (MRP4) with the specific substrate at 37 °C or at 32 °C (MRP4) in the presence of test drugs (added in 1 µl of DMSO) and incubation was started by pipetting 4 mM Mg-ATP or Mg-AMP to the samples. DMSO was added to the controls as well. The sample volume was 75 µl. After the incubation, samples were filtrated and washed rapidly on the Millipore manifold vacuum filter in MSFBN6B10 (Millipore, Burlington, MA, US) filter plate. After the plates were dried, Optiphase HiSafe (PerkinElmer) scintillation cocktail was added to each well for detection of radiolabeled compounds. CDCF was dissolved from the filter by 10% SDS and as a fluorescence stabilizer 0.1 N NaOH was added. The amount of radiolabeled substrates inside the filtered vesicles was determined by liquid scintillation counting by a MicroBeta Scintillation Counter (PerkinElmer, Waltham, MA, US). CDCF fluorescence was measured by VictorX (PerkinElmer Perkin-Elmer, Waltham, MA, US) plate reader at 492/635 nm.
For the SLC type transporter assays we used whole cell assays, by employing cell lines specifically expressing the investigated hepatic transporters and the relevant control, mock-transfected cell lines 59 .
For the OATP1B1 and OATP1B3 transporter assays, the interaction between potential anti-COVID-19 agents and OATPs was investigated in a microplate-based indirect assay employing pyranine as described 60  www.nature.com/scientificreports/ removed and the cells were preincubated in an uptake buffer in the presence or absence of anti-COVID-19 agents for 5 min at 37 °C. The reaction was started by the addition of an uptake buffer containing a final concentration of 10 μM (OATP1B1) or 20 μM (OATP1B3) pyranine and cells were incubated at 37 °C for 15 min (OATP1B1) or 30 min (OATP1B3). The reaction was stopped by removing the supernatant and the cells were washed with ice-cold PBS. The fluorescence of cells was determined using an Enspire fluorescent plate reader (PerkinElmer Co), Ex/Em: 460/510 nm. OATP-dependent transport was determined by extracting fluorescence measured in mock transfected cells.

MATE1 (SLC47A1), OCT1 (SLC22A1) and NTCP (SLC10A1) transporter inhibition assays
were performed as follows: overexpressing cell lines created by lentiviral method were used for these assays described in 61 . MATE1 was overexpressed in Madin-Darby canine kidney strain II (MDCKII) cells while OCT1 and NTCP were overexpressed in HEK-293 cells. Cells were seeded into 96-well tissue culture plates at a cell density of 1 × 10 5 cells per well. Experiments were performed 16-24 h after seeding. Before the experiments, cells were preincubated at 37 °C for 5 min in a Krebs Henseleit buffer (KHB, pH 8.0) or Hank's balanced salt solution (HBSS, pH 7.4) containing the tested drug at increasing concentration. DMSO was used for solvent control and the reference inhibitor (RI) for reference inhibitor wells. After the pre-incubation step, the solutions were removed. Experiments were performed at 37 °C in KHB or in HBSS containing the appropriate probe substrate and the drug at increasing concentrations. The organic solvent concentrations were equal in all wells. Cells overexpressing MATE1 and OCT1 were incubated with metformin (10 µM, including 0.2 or 0.1 μCi/mL 14 C-metformin) for 10 or 5 min, respectively, while cells overexpressing NTCP were incubated with TC (2 µM, including 2 µCi/mL 3 H-TC) for 2 min at 37 °C. MDCKII or HEK293-Mock (empty vector transduced) cells were used as a control. After the experiment, cells were washed twice with 100 µL of cold HBSS and lysed with 50 µL of 0.1 M NaOH and then analyzed by liquid scintillation.
Data and statistical analysis. All experiments were performed in duplicates and repeated in three biological replicates, thus each data point reflects six independent measurements. The group sizes were equal in all experiments. Normalization for the per cent (%) of control values was performed after subtracting the background levels as specified in the assay descriptions above, in order to avoid the inherent differences of the transporter activities and to properly reflect the inhibitory potential of each drug examined. IC 50 values were calculated by nonlinear regression analysis using the GraphPad prism software (version 5.01, GraphPad, La Jolla, CA, US https:// www. graph pad. com/ scien tific-softw are/ prism/).
For the estimation of potential in vivo drug effects, we have used the R values and the C max /IC 50 values, calculated as described in the International Transporterter Consortium guideline 34 , and the available pharmacokinetic data for the compounds. The R value is defined as 1 + (fu × Iin, max/IC 50 ), in which Iin, max is the estimated maximum inhibitor concentration at the inlet to the liver and is equal to: Iin, max = Cmax + (ka × Dose × FaFg)/ Qh/RB, where C max is the maximum systemic plasma concentration of the inhibitor; ka is the absorption rate constant. ka = 0.1/min was used as the worst-case estimate; Fa is the fraction absorbed, and Fa = 1 was used as the worst-case estimate. Fg is the intestinal availability, and Fg = 1 was used as the worst-case estimate; Qh is the hepatic blood flow rate. Qh = 1,500 ml/min was used; RB is the blood-to-plasma concentration ratio, assumed to be 1. IC 50 is the drug concentration which caused in vitro measured half maximum inhibition of the given transporter.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.