*2.4. Spectral Characterization of Substrate Binding and Reduction Potentials of CYP2C8 Polymorphisms*

We determined the reduction potential of the CYP2C8 variants using safranin T as a redox indicator as previously described [37]. Substrates binding to CYPs, such as CYP3A4, often perturbs water binding at the 6th coordinate position to produce a pentacoordinated high-spin heme (Figure 1). This results in an increase in the reduction potential of the heme and helps facilitate electron transfer from CPR [38]. The high-spin content can be observed as a shift in the heme absorbance from ~417 to ~390 nm. We did not observe a significant high-spin shift upon PAC binding to CYP2C8 in any of the variants, as was similarly observed for PUFAs binding to CYP2J2 [39]. We also determined that there is not a significant change in the reduction potential of CYP2C8\*1 upon PAC binding (Figure 4A, Table 3), which correlates to the poor high-spin content of the PAC-bound protein. The poor high-spin content and minimal change to the reduction potential together support the slow metabolism of PAC by CYP2C8 compared to other CYP-mediated drug metabolisms. Compared to the polymorphisms, CYP2C8\*2 had a slightly albeit not significantly lower reduction potential, and the reduction potential of CYP2C8\*3 was similar to WT (Figure 4, Table 3). Altogether, there is not a significant change to the intrinsic redox properties of the heme in these polymorphisms. We next proceeded to measure the first-electron transfer (FET) kinetics between CPR and CYP2C8.

**Figure 4.** Reduction potential of CYP2C8 variants. Redox titration of CYP2C8 variants was conducted in 0.1 M phosphate buffer (pH 7.4 at 25 ◦C) with or without 70 μM PAC. The potential was measured spectroscopically by using a safranin T as the redox indicator. Reduction was achieved using light and dithionite and oxidation was achieved using K3[Fe(CN)6)] as stated in the Methods section. Representative Nernst plots for (**A**) CYP2C8\*1 without paclitaxel (open black) and with paclitaxel (solid black), (**B**) CYP2C8\*2, and (**C**) CYP2C8\*3 from two experiments are shown. Data obtained from reduction is given as squares and data from oxidation is given as triangles. The zero intercept gives *E*◦-, the redox potential of the protein.

**Table 3.** Reduction potentials. Reduction potentials of the CYP2C8 variants was determined as described in the text. Error represents the SEM of two experiments.


#### *2.5. Polymorphisms Show Altered First Electron Transfer (FET) Kinetics as Determined by CO Stopped-Flow*

To further probe the effect of these polymorphisms on the CYP2C8 catalytic mechanism, we determined the kinetics of the FET from CPR to the CYP2C8 variants during the metabolism of 70 μM PAC. We firstly determined that there is not a significant change in the NADPH oxidation rates among the variants (Figure S3), which supports previous findings [19]. As NADPH is the initial step in the reaction, we then proceeded to conduct stopped-flow measurements in order to determine how CPR transfers electrons to the CYP2C8 variants.

CYPs display a characteristic shift in the heme absorbance to 450 nm upon CO binding to the heme. In order for CO to bind, the heme must be reduced to the ferrous state by CPR, which we determine by the appearance of the 450 nm absorbance band over time (Figure 5A). Therefore, the rate at which CO binds to the heme is directly related to the FET rate. The rate of CO binding was monophasic across all variants (Figure 5B–F). Compared to WT, CYP2C8\*2 and CYP2C8\*3 showed higher rates of CO binding, which may also explain the greater HOOH uncoupling and/or the higher levels of HOOH production by CYP2C8\*3. Contrariwise, the CYP2C8-R139K and CYP2C8-K399R variants showed lower rates compared to WT (Table 4).

**Figure 5.** Stopped-flow electron transfer rate to CYP2C8 variants. (**A**) Three-dimensional plot of representative absorbance spectra from 0 to 500 s showing the increase in absorbance near 450 nm following the mixing of NADPH with a pre-incubated complex of paclitaxel-bound CYP2C8 and CPR. (**B**–**F**) Representative plots of the change in peak absorbance at 450 nm from 0 s to 500 s. Experiments were conducted at either a 1:1 CPR:CYP or a 3:1 CPR:CYP ratio. Data points (grey squares) are overlaid with the line of best fit (solid black line) derived from a fit of the data to either monophasic (single exponential) or biphasic (two exponential) kinetic equations in OriginLab as indicated.

**Table 4.** Stopped-flow CO binding kinetics. Fast (k1) and slow (k2) rates are in units of ms<sup>−</sup>1. Experiments were conducted using a 1:1 and a 1:3 CYP:CPR ratios. Error represents the SEM of 3–6 experiments. Statistical significance compares to WT. \*\* *p* < 0.01, \*\*\* *p* = 0.001, \*\*\*\* *p* < 0.0001.


CYP metabolism studies are typically performed using a saturating 3:1 CPR:CYP ratio in order to achieve a maximum pseudo-zero-order electron transport kinetics between CPR and the CYP. Hitherto, we have used a 3:1 CPR:CYP ratio in our experiments. To determine if the FET is dependent on the

interactions of CYP2C8 with CPR, we repeated the experiments by lowering the CPR:CYP ratio from 3:1 to 1:1. CYP2C8\*1, CYP2C8-R139K, and CYP2C8-K399R all showed biphasic CO binding at the 1:1 CPR:CYP ratio and had similar fast (k1) and slow (k2) kinetics compared (Figure 5B,E,F, Table 4). Altering the CPR:CYP ratio has previously been shown to change the number of phases of the FET in certain situations likely by forcing CPR to associate with CYPs in unproductive confirmations at higher CPR:CYP ratios [40]. Likewise, the 3:1 ratio with CYP2C8\*1, CYP2C8-R139K, and CYP2C8-K399R may be promoting unproductive confirmations with CPR to produce these observations.

Interestingly, CYP2C8\*2 and CYP2C8\*3 only show the slow phase of the CO binding at the 1:1 ratio (Figure 5C,D, Table 4). The values of k2 for these polymorphisms are about half that compared to WT. However, the FET rate of CYP2C8\*2 and CYP2C8\*3 rates in the 3:1 CPR:CYP experiments resemble those of the WT slow phase in the 1:1 ratio. Therefore, it appears that more CPR is required for the CYP2C8\*2 and CYP2C8\*3 polymorphisms to rescue the WT activity of this slow phase. Together, these data suggest that these polymorphisms reduce either the binding of CPR to CYP2C8 or the transfer of electrons from CPR to CYP2C8. We will refer to these two events as the CYP–CPR redox interaction. Kaspera, et al. had previously determined that the apparent affinity of CPR for CYP2C8\*3 is greater than WT; however, the study also used a different recombinant system with cytochrome b5 as an auxiliary redox partner [19]. Homology modeling (Figure S4) reveals that R139, K399, and I269 all lie on the putative CYP-CPR binding interface, which means these mutations may be directly interfering with the CYP–CPR interaction. Further, K399 is located near the N-terminus at the protein-membrane interface (Figure 1A), and we showed previously that the N-terminus is essential for the CPR-mediated reduction of CYP2J2, which has a close homology to CYP2C8 [41].

#### **3. Conclusions**

CYP2C8\*2 and CYP2C8\*3 contain mutations that are not in the active site of CYP2C8. These mutations are instead located on the periphery of the protein. Therefore, these residues do not directly contribute to substrate binding and instead must affect CYP2C8 activity through their modulation of other steps in the CYP catalytic cycle. These mutations lie on the putative CYP–CPR interface and thus may be affecting the redox interactions between these proteins. We determined that CYP2C8\*2 is 47% as active towards PAC turnover compared to WT and that CYP2C8\*3 shows WT activity. The metabolism of PAC is 200% more uncoupled in the CYP2C8\*2 and CYP2C8\*3 polymorphisms, and CYP2C8\*3 produces significantly more HOOH compared to WT. Stopped-flow kinetics of the FET suggest that the polymorphisms reduce the transfer of electrons by CPR to CYP2C8. In conclusion, these in vitro studies demonstrate that these polymorphisms of CYP2C8 do not directly affect PAC binding to CYP2C8 but may, in fact, be affecting the redox interaction between CYP2C8 and CPR. The CYP2C8\*2 and CYP2C8\*3 polymorphisms reduce the CYP–CPR redox interaction and promote greater uncoupling of PAC metabolism. Therefore, not only is the FET disrupted in these polymorphisms, the electrons are being utilized towards ROS formation in lieu of PAC turnover. However, we were unable to definitively determine how the redox interaction is being altered, i.e., whether the polymorphisms affect the ability of CPR to bind and dock to CYP2C8, how the electrons are shuttled through CYP2C8, or both.

Another important finding is that the effects of the CYP2C8\*3 polymorphism on CYP2C8 activity cannot be explained by the additive contribution of the individual mutations themselves. In all experiments, CYP2C8\*3 had distinct activities compared to the linear combination of the R139K and K399R individual data. In fact, the single mutations either showed WT activity (especially concerning FET kinetics) or were significantly different than either WT or CYP2C8\*3 (e.g., the 265% increase in PAC metabolism by CYP2C8-R139K). Therefore, the mutation of these two residues do not additively contribute to the CYP2C8\*3 phenotype.

There are many mechanisms by which these polymorphisms effect the CYP2C8 redox interaction. They may be directly affecting the binding of CPR to CYP2C8. The mutations may also be altering the architecture of CYP2C8 such that it disfavors redox interactions with CPR while also destabilizing

the peroxy-heme intermediate to produce HOOH. CYP2C8\*2 contains an I269F mutation, which is a significant change to the physical and chemical properties of the residue. It is possible that this mutation has a profound effect on the folding of CYP2C8 and the interaction with CPR. CYP2C8\*3 contains a R139K/K399R double mutation, which interestingly swaps the Lys and Arg residues. While Lys and Arg are both positively charged residues, they differ significantly in their physical properties. For instance, Arg can form a greater number of electrostatic interactions and better maintains a positive charge compared to Lys. It has been shown that Arg substitution can increase the stability of GFP, which was shown in silico to be facilitated by a greater number of salt bridge interactions [42]. Arg also interacts differently to phospholipids and increases interfacial binding and perturbations to membranes [43]. Therefore, this polymorphism may be affecting how CYP2C8 associates with lipids as well as how it interacts with CPR.
