Results
Inhibition of TKIs on
rivaroxaban metabolism
Three TKI concentrations (1, 10, and 100 μM) were used to investigate
the inhibitory intensity. Imatinib and gefitinib showed strong
inhibition of the metabolism of rivaroxaban, but sunitinib only exerted
inhibitory activity in the incubation with HLM and
CYP3A4. Imatinib exerted the
strongest inhibitory effect on CYP3A4-mediated rivaroxaban metabolism,
with undetectable formation of M1 in the incubation with 100 μM imatinib
(Figure 1B). Gradient concentrations were used to determine the
IC50 values. In
CYP3A4-mediated rivaroxaban metabolism, imatinib showed the strongest
inhibitory activity with an IC50 value of 4.35 μM, and
gefitinib showed the strongest inhibitory effect on metabolism by CYP2J2
with an IC50 value of 3.72 μM. In the metabolism
associated with HLM incubation, imatinib was also the strongest
inhibitor with an IC50 value of 1.70 μM. Broadly, both
imatinib and gefitinib showed a strong inhibitory effect on the
rivaroxaban metabolism mediated by CYP3A4 and CYP2J2. In contrast,
sunitinib only exerted an inhibitory effect against CYP3A4-mediated
metabolism, while the effect on CYP2J2-mediated metabolism was not
obvious with an IC50 value of 397.70 μM (Figure 2E, F).
The detailed IC50 values are shown in Table 1.
Reversible inhibition
behaviour of TKIs on CYP3A4 and CYP2J2
The Ki value of the TKIs was fitted from the
kinetic curve and the R2 values and inhibition modes
are shown in Tables 2. As sunitinib did not exert more than 50%
inhibition even at 250 μM, the Ki value was not
measured. All three TKIs showed non-competitive inhibition on CYP3A4-
and CYP2J2-mediated rivaroxaban metabolism with good correlation (Figure
3 and 4). This was corroborated by the
respective
Dixon and Lineweaver–Burk plots. As with the IC50results, imatinib showed the strongest inhibition on CYP3A4-mediated
rivaroxaban metabolism with a Ki value of 1.92
μM. The Ki values of sunitinib and gefitinib with
CYP3A4 were 13.24 and 4.91 μM, respectively, which were similar to their
IC50 values.
IC50-shift
assays on CYP2J2 and CYP3A4
IC50 shift assays of CYP2J2 and CYP3A4 were performed to
explore the mechanism-dependent inhibition. Compared with direct
inhibition, the 30-min pre-incubation with NADPH in the CYP2J2
incubation did not significantly change the inhibitory effect of
imatinib and gefitinib (Table 3). While the inhibitory effect of
sunitinib was slightly increased, which resulted in 4.00-fold decrease
in the IC50 value (Figure 5B and Table 3). In brief, the
30-min pre-incubation did not significantly affect the inhibition of
CYP2J2 by the three TKIs, with all IC50 shifts being
less than 1.5-fold. In contrast, all IC50 values for the
inhibition of CYP3A4 by the TKIs decreased by more than 1.5-fold (Table
3). Notably, sunitinib showed the largest change in IC50shift (Figure 5E), with an IC50 value decreasing to 2.73
μM following the 30 min pre-incubation.
Time-dependent inactivation
of CYP3A4 by sunitinib
Given the 4.00-fold IC50 shift of sunitinib on CYP3A4
following 30-min pre-incubation with NADPH, time-dependent inactivation
constants of sunitinib were further determined (Figure 6A). The maximum
inactivation rate (Kinact) and the inhibitor
concentration needed to cause half of Kinact(KI) were fitted using the non-linear regression method.
As shown in Figure 6B, the Kinact and KIvalues of sunitinib were 0.0339 min–1 and 2.901 μM,
respectively. The Kinact value indicated that
approximately 3.4% of CYP3A4 was inactivated per minute when it was
incubated with the saturating concentration of sunitinib.
Molecular docking
simulations
Molecular
docking simulations were used to elucidate the binding conformations for
the interactions between the TKIs and CYP2J2 or CYP3A4. In the docking
simulation between CYP2J2 and the TKIs, gefitinib had the lowest
ChemScore value, followed by imatinib and then sunitinib. This ChemScore
ranking was consistent with the inhibitory intensity of the TKIs on
CYP2J2. Additionally, imatinib had the lowest ChemScore value in the
docking simulation with CYP3A4, followed by gefitinib and then
sunitinib, which was also consistent with the inhibitory intensity of
these TKIs on CYP3A4 (Figure 7).
Estimation
of the DDI risk between rivaroxaban and the three TKIs
According to the inhibition constants of the TKIs for CYP2J2- and
CYP3A4-mediated rivaroxaban metabolism, the AUC fold-changes when the
TKIs were combined with rivaroxaban were
predicted.
Imatinib was predicted to at most result in 244% increase in
rivaroxaban exposure (Table 4). Sunitinib and gefitinib were predicted
to at most result in rivaroxaban exposure increases of 2% and 11%,
respectively.
Discussion
All three TKIs were found to have a remarkable inhibitory effect on
CYP3A4-mediated rivaroxaban metabolism. Imatinib showed the strongest
reversible inhibitory effect towards CYP3A4 with a
Ki value of 1.92 μM. The inhibition of CYP3A4 by
sunitinib or gefitinib was also potent, with Kivalues of 13.24 and 4.91 μM, respectively. In addition to reversible
inhibition, the three TKIs also exerted time-dependent inactivation of
CYP3A4, especially sunitinib. Compared with reversible inhibition,
mechanism-dependent inactivation is always more frequently related to
unfavourable DDIs in clinical practice (Kalgutkar, Obach & Maurer,
2007). Importantly, the expression of CYP3A4 which is abundantly
expressed in the liver showed more than 100-fold population variability
(Zanger & Schwab, 2013). More importantly, it has been reported that
sunitinib has a 10-fold higher concentration in the liver than in blood
(Lau et al., 2015). Thus, being cautious about the CYP3A4 enzyme
expression in patients who take both rivaroxaban and TKIs is necessary.
The present study found that CYP2J2, which dominates the metabolism of
rivaroxaban, was inhibited by the TKIs. Imatinib and gefitinib were
potent inhibitors of CYP2J2 with Ki values of
3.53 and 2.99
μM,
respectively. Additionally, there was no irreversible inhibition of
CYP2J2 by the three TKIs, with the results showing IC50shifts of less than 1.5-fold. Although the three TKIs did not
irreversibly inactivate CYP2J2, the DDI risk produced by inhibiting
CYP2J2 cannot be ignored due to the distribution of CYP2J2 in vivo.
CYP2J2 was identified as an enzyme that mainly distributed in the heart
(Das, Weigle, Arnold, Kim, Carnevale & Huff, 2020): the mRNA levels of
CYP2J2 in the cardiovascular system exceed those of other detected
isozymes by 3 million to 62 times (Michaud, Frappier, Dumas & Turgeon,
2010; Wu, Moomaw, Tomer, Falck & Zeldin, 1996). Although CYP2J2 is not
usually considered as a DDI target due to its lower content in the
liver, if the heart is set as the organ for potential DDIs, there may be
an extremely high risk of DDI between rivaroxaban and TKIs. In addition
to its physiological distribution in the cardiovascular system, CYP2J2
has also recently been found to be highly expressed in various tumour
tissues (Allison et al., 2017; Karkhanis, Hong & Chan, 2017). Moreover,
TKIs would accumulate in the tumour tissue, thus the DDI risks of
rivaroxaban in combination with TKIs for cancer patients may be higher
than in our predictions. Therefore, it is necessary to include CYP2J2 in
DDI research of rivaroxaban.
Imatinib was predicted to have moderate DDI risk when combined with
rivaroxaban. Based on the inhibitory constants of imatinib on CYP2J2 and
CYP3A4, and the metabolic contributions of two isoforms, imatinib was
predicted to yield at most a 2.44-fold rivaroxaban AUC increase.
Therefore, according to the FDA’s guidelines for the relative risk of
DDIs, a moderate DDI risk may exist for the combination of rivaroxaban
and imatinib (Table 4). This result was predicted from the
pharmacokinetic data of patients with solid tumours who were received
TKIs. Notably, cancer patients have poorer metabolic function than
normal patients, which may directly lead to higher plasma concentrations
of TKIs. More importantly, patients who are refractory to standard
therapy have the much higher plasma concentrations (Scheffler, Di Gion,
Doroshyenko, Wolf & Fuhr, 2011). Therefore, using pharmacokinetic data
from cancer patients may make our prediction more accurate. However,
metabolic enzyme activity is subject to individual variability, so
factors such as hepatic blood flow and genetic polymorphism cannot be
neglected (Zanger & Schwab, 2013). Factors that influence the enzyme
activity of cancer patients may be more difficult to determine due to
the more complicated in vivo environment. All these factors would
produce great diversity in the pharmacokinetic data of TKIs in cancer
patients. Therefore, an individual physiologically-based pharmacokinetic
model for cancer patients would be encouraged in the prediction of the
pharmacokinetic behaviour of rivaroxaban in combination with TKIs.