*2.2. The Globular Domain Converges to the Same Orientations and Positions with Respect to the Membrane in CG Simulations of Full Length and of N-Terminally Truncated Protein*

In order to ensure that the initial modelled structures did not bias the results, two separate simulations of CYP 2C9 and CYP 2C19 were performed using the N-terminally truncated form with only the globular domain (S3). In these simulations, the globular domain could explore more configurations before reaching a stable orientation. For both isoforms, the final orientations of the globular domains in the phospholipid bilayer were stabilized by insertion of the F'–G' helices in the bilayer. Snapshots from these CG simulations are shown in Figure 3A,B. The region of the F'–G' helices is one of the hydrophobic regions in CYPs that keep the globular domain anchored in the membrane, even after truncation of the TM-helix. In the simulations of truncated CYP 2C9, the orientation of the globular domain converged in 3.5 μs with a sharp decrease in the CoM distance of the F'–G' helices, shown by the arrow in Figure 3D. The orientation of the CYP 2C19 globular domain converged quickly in 200 ns and remained stable throughout the simulation (Figure 3B,F). In both CYPs, after the F'–G' helices developed contacts with the membrane, no further change was observed in the orientation of the globular domain. For both isoforms, the converged orientation of the globular domain was the same in the simulations of the globular domain only (S3) and of the full-length protein (S1, S2) (Table 1 and Figure 3C,E). The difference in the orientations of the two CYP isoforms in the membrane was maintained in the simulations of both the full length and truncated forms of the proteins. However, in the simulations of the truncated proteins (S3), the CoM distance was the same for both proteins

(46 ± 2 Å). Overall, the CYP 2C9 CG simulations showed final orientations differing from CYP 2C19 despite using the same simulation parameters, water models, and protein components, whether full length or globular domain only.

**Figure 3.** (**A**,**B**) Snapshots from CG simulations of globular domains (S3) of (**A**) CYP 2C9 and (**B**) CYP2 C19 showing exploration of different orientations followed by convergence to the same orientation as observed for CG simulations of the full-length wild-type proteins. The globular domain is shown with a silver surface representation, with the F'–G' helices shown as green VDW spheres and the I-helix (residues 286–316) shown as a blue cylinder with an arrow and a red sphere at the C-terminal end. The 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer is shown in cyan, with the phosphate atoms in the head groups shown as red spheres. (**C**–**F**) Convergence of the orientation and position of the globular domain during CG simulations of CYP 2C9 (**C**,**D**) and CYP 2C19 (**E**,**F**). The angles (◦) and distance (Å) values vs time (μs) are shown for selected trajectories from CG systems: (**C**,**E**) S1 (full-length proteins); (**D**,**F**) S3 (globular domain only). The angles α (black), β (red), and the TM-helix tilt angle (cyan) (defined in Figure 2 and Materials and Methods) are shown along with the axial distances of the bilayer CoM to the CoM of the globular domain (blue) and the F'–G' helices (green). The thick black arrows point to the decrease in the distance of the F'–G' helices from the membrane center, which is coincident with convergence to stable orientations.

### *2.3. Structural Di*ff*erences in the Interfacial Residues A*ff*ect the Protein-Membrane Interactions and Globular Domain Orientations in CG Simulations*

To test the effect of the initial structure of the globular domain on its positioning in the membrane, we next performed CG simulations starting with four different models of CYP 2C9 (systems: M1–M4 (see Table S1). These full-length models of CYP 2C9 differed slightly in the side chain conformations over the whole globular domain, and more significantly in the membrane-interacting regions due to the different templates used for modeling the protein (see Appendix A). We focused on CYP 2C9 for these tests because the F'–G' helices were missing in the crystal structure used (PDB 1R9O), and this region shows high structural variability in the crystal structures of CYP 2C9. Models M1 and M2 were built by employing two different strategies to use the template information from the structures of both CYP 2C9 and CYP 2C19, whereas M3 was built using the CYP 2C19 structure as a template, and M4 was built using a previous model [22] for the CYP 2C9 F'–G' helices. Important differences were observed in the conformations of β-hairpin 1, the B–C loop, and the F'–G' helices, which are critical for developing CYP-membrane interactions and thereby influence the final orientation of the globular domain in the membrane. The differences in the orientations in the CG simulations of the different models of CYP 2C9 are shown in Table 1, Table S2, and Figure S2.

In all these CG simulations (M1–M4), the distance of the CoM of the globular domain to the CoM of the lipid bilayer increased from 43 ± 2 to 46–48 ± 2 Å. The new CoM distance value for CYP 2C9 was the same as observed for CYP 2C19. The angles α and β increased in the CG simulations using modeled structures compared to those starting with the crystal structure of CYP 2C9 (S1, S2). In the CG simulations of M3, for which the CYP 2C19 crystal structure was the template, 50% of the simulations (three out of six trajectories) showed higher angles (class B), resembling the orientation of CYP 2C19 in the membrane (Table S2, Figure S2). The increased angle and distance values were attributed to the initial conformations of the globular domain, due to the selection of the modeling template. The CG simulations of the four CYP 2C9 models (M1–M4) indicated that it is not only the primary sequence but also the initial conformational differences in the linker, B–C loop, and F'–G' helix regions that influence the final positioning of the CYP globular domain in the membrane. However, it should be borne in mind that the different conformations come from templates with different sequences, and that conformational preferences are dependent on sequence.
