Cultured rat astrocytes were used to investigate the mechanisms involved in the modulation of the Mrp1-mediated export of GSH, GSSG and GS-B by the antiviral compound ritonavir. The release of GSH from viable astrocytes was strongly stimulated in the presence of 10 µM ritonavir as indicated by a 4-fold increased specific release rate, which is consistent with literature data [48]. Also formaldehyde has previously been reported to stimulate Mrp1-mediated GSH export from cultured astrocytes and this stimulation was characterised by a strong increase in the Vmax value for the export of GSH, while the Km values remained almost unchanged [84]. For this formaldehyde-induced stimulation, an increase in the number of active Mrp1 molecules in the cell membrane was postulated [84]. An increased number of Mrp1 proteins in the cell membrane could in principle also explain the ritonavir-induced stimulation of GSH export observed in this study. At least for ritonavir-treated human intestinal carcinoma cells, an upregulation of Mrp1 expression and activity after chronic exposure to ritonavir has been reported [90]. However, this appears unlikely to be the case for ritonavir-treated astrocytes, due to the short incubation times in the amino acid-free incubation medium that has been used. A recruitment of intracellular Mrp1-containing vesicles to the plasma membrane upon incubation with ritonavir, as reported for bilirubin-treated astrocytes [91], cannot contribute to the stimulated GSH export either, because of three reasons: 1) The Vmax values calculated for the GSH export remained very similar for control and ritonavir-treated astrocytes; 2) the stimulation was only observed for GSH and not for the other Mrp1 substrates GSSG and GS-B; and 3) no obvious difference was observed for the cellular localization of Mrp1 by immunocytochemical staining of control and ritonavir-treated astrocytes (Fig. S8).
Kinetic analysis of the export of GSH from viable astrocytes revealed a Km value of around 200 nmol/mg, which corresponds to around 50 mM GSH considering the reported cytosolic volume of 4.1 µL/mg protein for cultured astrocytes [92]. This is similar to values previously reported for the release of GSH from cultured astrocytes [84, 85]. However, in the presence of ritonavir the Km value for GSH export was drastically lowered by 85% to around 28 nmol/mg (corresponding to 6.7 mM GSH). Thus, half-maximal transport velocity is established in ritonavir-treated astrocytes at much lower GSH concentrations. The cytosolic GSH concentration in cultured astrocytes has been reported to be 8 mM [92]. Using the calculated kinetic parameters for GSH export from control and ritonavir-treated astrocytes to calculate the GSH export velocity for the cytosolic concentration of 8 mM GSH, we obtain initial export rates of 3.3 nmol/(mg h) (control) and 10.9 nmol/(mg h) (ritonavir). These values account for 14% and 54% of the calculated Vmax, respectively, and match with the observed almost 4-fold stimulation of GSH export in ritonavir-treated astrocytes.
In contrast to the astrocytic GSH export, the Mrp1-mediated export of GSSG and GS-B, [16, 18, 33, 34, 52, 93] was not stimulated in the presence of 10 µM ritonavir. A high concentration of 100 µM ritonavir even inhibited these Mrp1-mediated export processes. Similarly, also the known Mrp1 inhibitor MK571 has been observed to lower the astrocytic GSSG and GS-B export [16, 33, 34] and competition of MK571 with these substrates has been proposed as the reason for the observed inhibition [17]. A potentially stronger inhibitory effect of ritonavir in concentrations above 100 µM on the GSSG and GS-B export could unfortunately not be investigated due to the limited solubility of ritonavir. For peripheral cells, it was shown that ritonavir is itself a substrate of Mrp1 and thereby inhibits the transport of other Mrp1 substrates such as the chemotherapeutic drugs doxorubicin and etoposide [94–97]. The ritonavir-dependent inhibition of Mrp1-mediated export of GSSG and GS-B from astrocytes is consistent with these literature data. Thus, ritonavir differently modulates the transport properties of different Mrp1 substrates; it namely stimulates GSH export but inhibits GSSG and GS-B export from rat astrocytes.
To shed light onto this surprising finding, atomistic details of the above-mentioned diverse transport processes were investigated by MD simulations using the homology model of rat Mrp1 in complex with its docked ligands GSH, GSSG and/or ritonavir. The predicted binding position of GSH in the P-pocket of rat Mrp1 is similar to those found by Johnson and Chen [12] for the GS-moiety of LTC4 in bovine Mrp1 resolved by cryo-EM experiments, including hydrogen bond formation to residues R1198 and R1250. Therefore, the generated rat homology model was considered a useful tool for further rat Mrp1-ligand interaction studies. In the absence of other ligands, GSH was previously suggested to bind preferentially to the P-Pocket of Mrp1 [12]. In our simulations we have observed also a high affinity of GSH for the H-pocket (Figs. 5 and 7), which is likely caused by hydrogen-bond interactions to the amino acids Y1244, R1250, W554 and R594 (Fig. S7). Although R1250 and R594 actually belong to the P-pocket due to their positive charges, their spatial proximity to residues of the H-pocket enables the stabilization of GSH within the cavity of the H-pocket. In addition, residues W554 and R594 belong to the transmembrane domain (TMD) 1 bundle, whereas Y1244 and R1250 are classified within TMD2. Simultaneous binding of GSH to R1198, K333, H336, Y441 and R1250 (GSH binding to the P-pocket) or W554, R594, Y1244 and R1250 (GSH binding to H-pocket) would bridge TMD1 and TMD2 in both cases. This mechanism was suggested to initiate an overall conformational change in Mrp1, which is crucial for the transport process and also applies to the predicted GSH binding configurations in this study [12]. Due to the discovery of two translocation pathways, one from each pocket, which merge into the extracellular facing site of the transmembrane domain of Mrp1, transport of GSH from either pockets is theoretically possible [98].
The observed binding mode of GSSG to the P-pocket of rat Mrp1 resembled that of GSH within the P-pocket. However, the binding was much less stable, and GSSG was observed to migrate out of the pocket to diffuse away into the cytosol after a few hundred ns of simulation time (Fig. S6). Moreover, no binding of GSSG to the H-pocket could be identified.
Moreover, ritonavir preferentially binds to the P-pocket (Fig. S6 and 7). Binding to the H-pocket might be hindered due to ritonavir’s relatively large size (98 atoms). Indeed, the H-pocket only consists of five amino acids, whereas the P-pocket consists of up to ten amino acids, including R594 (Fig. S4c). The same explanation likely applies to GSSG (70 atoms), whereas the much smaller GSH molecule (36 atoms) readily binds to both the P-pocket and the H-pocket, as highlighted above.
As a consequence, the simultaneous binding of GSH and ritonavir to rat Mrp1 is sterically possible (Figs. 5 and 7). Additionally, the residence time and binding stability of GSH to rat Mrp1 was increased in presence of ritonavir (Fig. 6). The binding of ritonavir to Mrp1 may therefore increase the affinity of GSH for the H-Pocket of Mrp1, thereby enhancing GSH binding and lowering the Km value for GSH transport. Seemingly, ritonavir acts like a plug when bound to the P-pocket, locking a GSH molecule in its position within the H-pocket until the transport process is initiated. This concept is in agreement with the recently reported hypothesis of sequential binding of GSH and modulators of Mrp1-mediated GSH export [14]. Amino acids W554, M1094, W1247 and R1250 were observed to spatially restrict the movement of GSH into the direction of the cytosol, while ritonavir blocks the movement towards the P-pocket, thereby preventing the diffusion of GSH out of the binding site. Application of ritonavir in higher concentrations (100 µM) to astrocyte cultures resulted in a decreased acceleration of Mrp1-mediated GSH export compared to applications of lower concentrations (30 µM) (Fig. 3a,b). This experimental finding could be explained by the predicted binding model discussed above. If ritonavir binds to the P-pocket before a GSH molecule is able to bind to the H-pocket, access of GSH to the entire binding site might be blocked by ritonavir and the transport process be brought to a halt for the duration of ritonavir binding.
In contrast, ritonavir is likely to compete with GSSG for binding at the same P-pocket of the Mrp1 binding site, thereby inhibiting the binding and the subsequent transport of this Mrp1 substrate (Fig. 7 and S6). An inhibitory mechanism on the GSSG export was already postulated for MK571. However, it was suggested that MK571 would bind to the H-pocket [17]. The compound GS-B (61 atoms) is structurally similar to GSSG and its export was indeed reduced upon addition of ritonavir. Although GS-B was not simulated explicitly, it is plausible that it also preferentially occupies the P-pocket, thus competes for binding with ritonavir.
In conclusion, the experimentally observed interference of the antiretroviral drug ritonavir with the export of the different Mrp1 substrates GSH, GSSG and GS-B can most likely be explained by direct binding of ritonavir into the polar part of the biphasic binding site of Mrp1. While binding of ritonavir into the hydrophilic P-pocket of Mrp1 might stabilise and thereby stimulate the export of GSH that is bound to the hydrophobic H-Pocket, the export of larger substrates including GSSG or GS-B is inhibited due to the occupancy of the P-Pocket by ritonavir, since these larger substrates require binding to the P-Pocket in order to be transported (Fig. 7).
Understanding the principles behind the transport mechanism of the ubiquitously expressed transport protein Mrp1, which is responsible for the export and detoxification of various types of (chemotherapeutic) drugs and potentially toxic xenobiotic substances with a broad variety of chemical characters, is crucial for predicting and understanding the occurrence of tolerance and/or side-effects of various types of drugs, especially in combinational prescription as it is currently the case for paxlovid. For instance, binding of the anticancer drugs doxorubicin and etoposide to the P-pocket of Mrp1, similar to ritonavir, would explain the reported transport inhibition of Mrp1 by these compounds and the simultaneous occurrence of cytotoxicity [96, 97]. And indeed, doxorubicin in addition to other anticancer drugs have very recently been shown in in silico docking experiments to bind to Mrp1 [99].
We would like to note that the observations and conclusions made in our study are most likely also applicable to transport processes mediated by human Mrp1, as both the amino acid sequence and the domain topology of rat and human Mrp1 show very high similarity. In the future, a more in-depth insight into the transport mechanism of various Mrp1 substrates and evaluation of multiple ligand binding poses with accurately determined binding free energies would be desirable. This requires computationally demanding enhanced-sampling MD techniques such as alchemical free energy calculations or funnel-shaped restrained metadynamics with Hamiltonian replica-exchange (fun-SWISH) [100], which are left out for further, more in-depth studies.
For the brain, a treatment with ritonavir might have severe consequences for the GSH metabolism, since it accelerates the export of GSH from astrocytes and neurons [48, 101]. But at least under conditions of optimal amino acid supply, GSH synthesis can fully compensate the ritonavir-induced acceleration of GSH export from astrocytes [48]. However, accelerated export of GSH results in increased extracellular GSH concentrations, which could have multiple consequences to neurons. On the one hand, GSH is hydrolysed by the astrocytic ectoenzyme γ-glutamyltransferase generating increased levels of the dipeptide CysGly and the neurotransmitter glutamate [35, 51, 102, 103] of which the latter is believed to induce excitotoxicity by overstimulation of glutamate receptors [104, 105]. On the other hand, an increased extracellular concentration of GSH might serve as precursor donor for neuronal GSH synthesis [102, 106]. Furthermore, ritonavir has been reported to induce oxidative stress [107, 108] as well as HIV infection itself [109–112]. Since astrocytes generate substantial amounts of GSSG during oxidative stress, which is subsequently released via Mrp1 to sustain the cellular thiol reduction potential [16, 35], a blockage of GSSG export through ritonavir in the brain of treated patients might therefore additionally reduce oxidative stress resistance of these cells. Therefore, chronic oxidative stress and GSH deficiency caused by HIV infection might be even more detrimental for brain cells under prescription of ritonavir due to its interference with Mrp1-mediated transport processes.