Results and Discussion

Because it was still not known whether atorvastatin and rosuvastatin are binding to human CAR, we measured the interaction of rosuvastatin, atorvastatin, and the positive control CITCO with hCAR using the SPR technique. The structures of the molecules are shown in Supplemental Fig. 1. Recombinant proteins GST-hCAR and GST-hRXRα-LBD were isolated and their purity assessed by SDS-PAGE and Western blot analysis (Supplemental Fig. 2). The SDS-PAGE analyses of recombinant proteins revealed high purity, and the Western blot analyses confirmed the presence of GST proteins.

The GST-hCAR/GST-hRXRα-LBD heterodimer was covalently attached to the surface of the SPR sensor chip and assayed with at least four different concentrations of compounds. Each concentration was injected 3 times to test the reproducibility. All binding responses and K_D values were reproducible, which indicates that immobilized complex did not lose activity over time.

As expected, CITCO, which is known to be a direct ligand for hCAR, showed specific binding to the complex. We also observed specific concentration-dependent binding of both statins of this study (Fig. 1). The measured K_D indicated that rosuvastatin binds to the CAR/RXR heterodimer with approximately 2-fold higher affinity than atorvastatin, but approximately 3-fold lower affinity than CITCO (Table 1). A comparison of measured maximal response with the theoretical maximal response indicated that several molecules of atorvastatin and CITCO bind to the hCAR/hRXRα-LBD complex (Table 1). By applying SPR we show for the first time that atorvastatin and rosuvastatin directly bind to the hCAR/RXR-LBD heterodimer.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

SPR analysis of atorvastatin, rosuvastatin, and CITCO binding to hCAR/hRXRα-LBD heterodimer. Different concentrations (1.56, 3.125, 12.5, 25, and 50 µM for CITCO and 6, 12.5, 25, 50, 75, 100, and 150 µM for atorvastatin and rosuvastatin) of analytes were tested for the binding (left panels). The binding curves (right panels) were generated by fitting the steady-state response levels at the end of the association phase versus the concentration of the injected analyte. The K_D values were obtained by fitting the data to the steady-state affinity model. For each analyte six or seven independent titration experiments were performed.

Binding of statins to hCAR is consistent with our previous transcriptome study, in which we showed that many genes were differentially expressed in human primary hepatocytes treated with atorvastatin and rosuvastatin and are under direct regulation of CAR (Hafner et al., 2011). Another study used surface plasmon resonance to test binding of atorvastatin and hCAR with coactivator (Hoffart et al., 2012). They preincubated atorvastatin or CITCO with CAR-LBD protein and injected it onto immobilized steroid receptor coactivator 1–receptor interaction domain (SRC-1-RID) protein. Under these conditions atorvastatin did not induce binding of hCAR-LBD with coactivator SRC-1. Further, they did not observe the assembly of human CAR-LBD with coactivators in COS1 cells.

The difference in results obtained by SPR between our study and that of Hoffart et al. (2012) was probably due to the different CAR dimerization partner and steps used. We used RXR as a dimerization partner, and studies show that RXRα may also have a role in stabilizing CAR-LBD in its active conformation via direct interactions across the heterodimer interface (Xu et al., 2004). RXR is also important for recruitment of coactivators such as SRC-1 (steroid receptor coactivator 1) (Dussault et al., 2002; Pavlin et al., 2014). Also, we tested the binding of atorvastatin to the CAR/RXR dimer whereas Hoffart et al. (2012) tested binding of the atorvastatin–CAR pair to the SRC-1 coactivator. It is important to note that none of the described experiments published on statins used physiologic concentrations of the drugs. All used micromolar concentrations, but in humans in serum only nanomolar concentrations were measured; certainly concentrations in the liver have not been determined, so it is unclear how the in vitro observations relate to the in vivo situation.

Our computational predictions using a structure-based or ligand-based approach indicated that statins are good potential ligands for CAR and PXR with little difference in potential between them. Furthermore, statins docked in CAR produced scores that correlate with the K_D values determined by SPR. Our data suggested variability in binding to these different models (Table 1, Supplemental Table 1, Supplemental Fig. 3).

Based on the docking results, the statins appeared to be good ligands for both CAR and PXR. Atorvastatin had the highest score against CAR (73.18), and rosuvastatin was lower (67.09) (Table 1). Similarly, the pharmacophores and Bayesian models suggest that all other statins will likely bind to CAR and PXR (Supplemental Table 1). These results broadly agree with the published data, confirming that the majority of statins are CAR/PXR activators, although the manner of this activation awaits experimental confirmation in the future. When comparing these scores with other studies using the same in silico approaches, we observed that the calculated scores for statins are within the same range as for known hCAR activators (Kortagere et al., 2009; Lynch et al., 2013). Interestingly, we and others have predicted that pravastatin is also a good potential ligand for CAR and PXR, although this statin has never been shown to have any ability to induce these two nuclear receptors or drug metabolism (Howe et al., 2011).

In conclusion, these findings are important for understanding the full biologic consequences of statin exposure due to molecular interaction of statins with nuclear receptors. We showed in silico that statins are potentially ligands of CAR and PXR and confirmed in vitro that atorvastatin and rosuvastatin directly bind to the CAR/RXRα dimer. As CAR and PXR regulate not only drug metabolism but also glucose and energy metabolism, this had led us to the assumption that they are potentially involved in the statins’ unwanted side effects such as the increase in developing type 2 diabetes in humans. Additionally, we conclude that computational approaches may be useful for assessing statin binding to CAR and PXR and that SPR can provide valuable insights on binding to validate these predictions.