CYP-catalysed cycling of clozapine and clozapine-N-oxide promotes the generation of reactive oxygen species in vitro

Abstract Clozapine is an effective atypical antipsychotic indicated for treatment-resistant schizophrenia, but is under-prescribed due to the risk of severe adverse drug reactions such as myocarditis. A mechanistic understanding of clozapine cardiotoxicity remains elusive. This study aimed to investigate the contribution of selected CYP isoforms to cycling between clozapine and its major circulating metabolites, N-desmethylclozapine and clozapine-N-oxide, with the potential for reactive species production. CYP supersome™-based in vitro techniques were utilised to quantify specific enzyme activity associated with clozapine, clozapine-N-oxide and N-desmethylclozapine metabolism. The formation of reactive species within each incubation were quantified, and known intermediates detected. CYP3A4 predominately catalysed clozapine-N-oxide formation from clozapine and was associated with concentration-dependent reactive species production, whereas isoforms favouring the N-desmethylclozapine pathway (CYP2C19 and CYP1A2) did not produce reactive species. Extrahepatic isoforms CYP2J2 and CYP1B1 were also associated with the formation of clozapine-N-oxide and N-desmethylclozapine but did not favour one metabolic pathway over another. Unique to this investigation is that various CYP isoforms catalyse clozapine-N-oxide reduction to clozapine. This process was associated with the concentration-dependent formation of reactive species with CYP3A4, CYP1B1 and CYP1A1 that did not correlate with known reactive intermediates, implicating metabolite cycling and reactive oxygen species in the mechanism of clozapine-induced toxicity.


Introduction
Clozapine is the only effective medication for treatmentresistant schizophrenia (Galletly et al. 2016;Sultan et al. 2017), with a reported response in 60-70% of these patients (Lally et al. 2016).Despite its therapeutic effectiveness, clozapine is not recommended as a first-line therapy due to the high risk of adverse events including clozapine-induced cardiotoxicity (myocarditis and cardiomyopathy).
The introduction of a new monitoring protocol for clozapine-induced myocarditis (Ronaldson et al. 2011) revealed that it is more prevalent in Australia (>1%) than reported in other countries (<0.1%) (Ronaldson et al. 2015), whilst in New Zealand the reported incidences of clozapine-induced myocarditis and cardiomyopathy 3.8% and 1.3%, respectively, placing them into the category of common adverse reactions (Bellissima et al. 2021;Council for International Organizations of Medical Sciences 2001).No current evidence indicates a role for co-morbidities in the risk of developing clozapine-induced cardiotoxicity (Alawami et al. 2014;Bellissima et al. 2021).However, the development of myocarditis correlates with the rate of clozapine dose escalation (Chopra and de Leon 2016;de Leon et al. 2020;Ronaldson et al. 2012).
Drug withdrawal leads to recovery and clozapine rechallenge does not commonly induce a recurrence, provided the dose is carefully titrated during re-exposure (Manu et al. 2012), indicating the mechanism of toxicity is unlikely to result from IgE-mediated hypersensitivity (Kilian et al. 1999).The infiltrating eosinophils observed following histologic examination of clozapine-induced myocarditis (Kilian et al. 1999) may instead be a consequence of a directly toxic species causing cardiac-selective damage.Proposed mechanisms behind clozapine-induced cardiomyopathy include free-radical initiated myocardial injury (Merrill et al. 2005), and that cardiomyopathy may arise from long-term undetected myocarditis (Merrill et al. 2005).This is supported by the presence of both myocarditis and cardiomyopathy in some clozapine users (Kilian et al. 1999;Merrill et al. 2005).
Clozapine is metabolised extensively, with only 2.5% of the drug excreted unchanged (Sheehan et al. 2010).The major circulating metabolites include N-desmethylclozapine and clozapine-N-oxide (Dain et al. 1997;Pirmohamed et al. 1995).CYP3A4 and CYP1A2 are the primary hepatic enzymes attributed to the generation of both N-desmethylclozapine and clozapine-N-oxide, whilst CYP2C19 and CYP2D6 may also contribute to the N-desmethylation of clozapine (Pirmohamed et al. 1995;Sheehan et al. 2010;Tugnait et al. 1999).In vitro studies have demonstrated that human liver microsomes can bioactivate supra-therapeutic concentrations of clozapine to protein-reactive nitrenium and iminium ions, trapped using glutathione and cyanide, respectively (Dragovic et al. 2010(Dragovic et al. , 2013)).A number of extrahepatic CYPs can also metabolise supra-therapeutic concentrations of clozapine including CYP1A1, CYP1B1, CYP2C8, CYP2C9 and CYP2J2 (Dragovic et al. 2013).Notably, these CYPs are expressed in the myocardium or vasculature (DeLozier et al. 2007;Hausner et al. 2019).However, knowledge surrounding their contribution to clozapine toxicity is limited, particularly at therapeutic concentrations.Furthermore, the ability of the selected the isozymes to metabolise clozapine-N-oxide and N-desmethylclozapine have not previously been investigated.
A study of plasma concentrations in patients initiating clozapine therapy identified significantly higher clozapine-Noxide/N-desmethylclozapine ratios in patients diagnosed with myocarditis (cases) in comparison to those without a cardiotoxicity diagnosis (non-cases), (Bellissima 2018).This suggests that equilibria between metabolic pathways are altered to favour clozapine-N-oxide formation in patients developing myocarditis.
Additionally, clozapine-N-oxide can be reduced back to clozapine via a reversible metabolic process (Chang et al. 1998;Jann et al. 1993), although the mechanism by which this occurs is unknown.While oxidation-reduction (REDOX) cycling of clozapine and clozapine-N-oxide may prolong its therapeutic action (Jendryka et al. 2019), REDOX cycling can put cells under oxidative stress, generating reactive oxygen species (ROS) and free radicals that result in direct tissue damage and potentially an increased the likelihood of idiosyncratic adverse drug reactions (Uetrecht 2002).This has previously been proposed as a mechanism behind clozapine-induced cardiotoxicity (Abdel-Wahab and Metwally 2015; Arzuk et al. 2021;Williams et al. 2003).
We have undertaken in vitro assessment of the metabolism of clozapine, clozapine-N-oxide, and N-desmethylclozapine by a series of hepatic (CYP1A2, CYP2C19 and CYP3A4) and extrahepatic (CYP2J2, CYP1A1 and CYP1B1) CYP isoforms at both therapeutic and supra-therapeutic concentrations.The extrahepatic isoforms were selected as they have moderate to high expression in the myocardium (Chaudhary et al. 2009;DeLozier et al. 2007).
This study also measured the formation of non-selective reactive species to evaluate how each individual CYP isoform may contribute to the development of cardiotoxicity via this putative mechanism.Furthermore, the ability of each isozyme to generate reactive intermediate nitrenium and iminium ions was compared to overall reactive species formation.

Enzyme kinetics associated with the metabolism of clozapine, clozapine-N-oxide and N-desmethylclozapine
As the clinical concentrations of clozapine and its metabolites in cardiac tissue are unknown, the investigation into the CYP-catalysed metabolism of these compounds was conducted at a range of established therapeutic and supra-therapeutic plasma clozapine concentrations (Stark and Scott 2012).Turnover studies were conducted at 0, 0.9, 3.0, 9.0 and 15 mM, (300, 1000, 3000 and 5000 ng/mL, respectively).Each analysis was conducted in triplicate and two independent experiments undertaken on non-consecutive days.The combined data from the two experiments are reported (n ¼ 6).
Clozapine was pre-incubated with CYP1A2, CYP2C19, CYP3A4, CYP2J2, CYP1A1 or CYP1B1 supersomes TM (20 pmol/ mL) in 100 mM phosphate buffer (pH 7.4) for 5 min, followed by initiation of the reaction with NADPH (1 mM final concentration).The reaction mixture (100 lL final volume) was then incubated at 37 � C for 60 min.Following this the reaction was terminated, compounds extracted and then quantified with high performance liquid chromatography (HPLC) or liquid-chromatography mass spectrometry (LCMS) as described below.These procedures were also conducted with either clozapine-N-oxide or N-desmethylclozapine under identical conditions.

Reactive species generation associated with metabolism
To provide an indirect marker of potential oxidative damage associated with the REDOX cycling during metabolism of clozapine, the presence of ROS produced by the CYP-catalysed REDOX cycling of clozapine and clozapine-N-oxide were determined by co-incubation of clozapine or its plasma metabolites with 7.5 mg/mL 2 0 -7 0 -dichlorofluorescein diacetate (DCF-DA) in the presence or absence of CYP supersomes TM .Negative controls consisted of CYP supersomes TM and DCF-DA without clozapine, clozapine-N-oxide or N-desmethylclozapine.The oxidation of DCF-DA by free radicals results in the formation of a fluorescent product, 2 0 ,7 0 -dichlorofluorescein (DCF), detected with fluorescence spectrometry (485 nm excitation; 520 nm emission).The concentrations of DCF formed were ascertained from a validated standard curve (r 2 > 0.98).The linear range of the curve was 0.05-0.5 mg/mL and the lower limit of quantification was 0.05 mg/mL.As DCF-DA is a non-selective probe, it is also sensitive to reactive nitrogen species, such as reactive nitrenium or iminium ions.The cumulative generation of DCF at 60 min was assessed to ascertain the overall formation of CYP-catalysed reactive species generation from REDOX cycling of clozapine, clozapine-N-oxide or N-desmethylclozapine or the production of protein reactive intermediate metabolites.

Identification of reactive intermediate metabolites
To trap any potentially reactive intermediate metabolites, varying concentrations of clozapine or clozapine-N-oxide with either 15 mM glutathione (GSH) or 15 mM potassium cyanide were incubated with all CYP-isoforms in the presence of DCF-DA and NADPH.GSH has previously been reported to conjugate with reactive nitrenium ions, whilst cyanide will form adducts with reactive iminium ions formed through clozapine metabolism (Dragovic et al. 2013).All incubations proceeded for 60 min at 37 � C, and metabolites were extracted from supersomes TM , then detected with LCMS.

Extraction from supersomes TM
Clozapine and metabolites were extracted from the experimental samples (100 mL) using a protein precipitation procedure.Samples were spiked with 25 lL of internal standard (1000 ng/mL amoxapine) and immediately quenched via the addition of ice-cold acetonitrile (200 lL), followed by brief vortex mixing.Samples were then centrifuged at 15,000 g for 20 min.The supernatant was removed and evaporated under a gentle stream of nitrogen gas at 40 � C. The supernatant was reconstituted in 150 lL of the mobile phase, 70% 50 mM ammonium formate, pH 3.5: 30% acetonitrile-methanol mixture (90:10 V / V ).The reconstituted residue was then centrifuged at 15,000 g for 3 min, and the supernatant was transferred to a vial for analysis.

HPLC and LCMS assays
Clozapine, N-desmethylclozapine and clozapine-N-oxide concentrations were initially quantified using HPLC with UV-vis detection, but progression to LCMS was required for the detection of lower concentrations.Chromatographic separation was on a Grace Alltima HP C 18 (150 � 4.6 mm, 5 lm) reverse-phase highperformance LC column coupled with a Phenomenex security C 18 guard column.The mobile phase for both systems was 50 mM ammonium formate buffer, pH 3.5 [mobile phase A] and 90% ( V / V ) acetonitrile in methanol [mobile phase B] maintained at 70% [A] and 30% [B] throughout each run.An Agilent 1200 series HPLC system coupled with UV-vis set to a detection wavelength of 230 nm and a flow rate of 1.0 mL/min was utilised to generate chromatograms.The LCMS system comprised an Agilent 1200 HPLC coupled to an Agilent MSD model D quadrupole mass spectrum detector and Jet Stream electrospray ionisation source.The flow rate was 0.6 mL/min.Detection of compounds was performed in the positive electrospray ionisation mode with nitrogen drying and vaporising gas.Fragmentor voltage was set at 100 V, nebuliser pressure at 50 psig, gas temperature at 300 � C. The drying gas flow rate was 12.0 L/min.Extracted ion chromatograms were generated in selected ion monitoring (SIM) mode using the following (M þ H) ions: clozapine (327 m/z), clozapine-N-oxide (343 m/z), N-desmethylclozapine (313 m/z), nitrenium þ GSH (632 m/z), iminium þ CN (352 m/z) and amoxapine (315 m/z).A scan channel was also utilised to monitor for the formation of any unexpected metabolites (range 100-1000 m/z, fragmentor voltage 5 V).Agilent Chemstation software was used to access and process chromatograms.Validated standard curves (r 2 > 0.99) were developed to quantify clozapine, clozapine-N-oxide and N-desmethylclozapine.The linear range of the HPLC derived curves were 0.6-16 mM, and the lower limit of quantification was 0.6 mM (200 ng/mL) for all three substrates.The linear range of the LCMS derived curves were 0.08-1.6 mM and the lower limit of quantification was 0.08 mM (25 ng/mL) for all three compounds.Accuracy and precision were within ± 15% for all calibration standards, with the exception of the lower limit of quantification, (± 20%).

Data analysis
Following the quantification of metabolites, the product formation was normalised to a non-enzymatic control and the specific enzyme activity (pmol product/pmol CYP/min) calculated.Non-linear regression models were fitted to the specific activity data, allowing the calculation of the Michaelis-Menten parameters (V MAX and K M ) in GraphPad PRISM (v9.0.2).Where a rate could not be determined due to the lack of an authentic standard, comparisons of the peak-area ratio (PAR), between the analyte of interest and clozapine-N-oxide for the oxidation pathway or the clozapine for the reduction pathway, were conducted for all isoforms.
Time-dependent DCF formation was utilised as a marker of reactive species generation.All measured concentrations of DCF were normalised to a non-enzymatic control.Negative controls (no substrate) were included to ensure that the reported reactive species generation was attributable to the CYP-catalysed metabolism of clozapine, clozapine-N-oxide, or N-desmethylclozapine.
Statistical significance was assessed by two-way ANOVA, followed by Dunnett's multiple comparisons tests.Adjusted p-values � 0.05 were considered statistically significant.

Clozapine metabolism
All of the hepatic (Figure 1(A,B)) and extrahepatic (Figure 1(C,D)) CYP isoforms investigated catalysed the formation of clozapine-N-oxide and N-desmethylclozapine in a concentration-dependent manner; the Michaelis-Menten model non-linear regression fits for these data are shown.CYP3A4 had the highest activity for the formation of clozapine-N-oxide (V MAX ¼ 1.0 pmol/pmol CYP/min, K M ¼ 12.57 mM, Supplementary Table 1), while the other isoforms investigated were associated with substantially lower (>4-fold) catalytic activity (Figure 1(A,C)).Maximal velocity was not reached for all isoforms at the substrate concentrations assayed, resulting in the estimation of inaccurate V MAX, K M and intrinsic clearance parameters with wide 95% confidence intervals for some isoforms (Supplementary Table 1).Therefore, comparisons between CYP were undertaken using the specific activity at the highest clozapine concentration tested (V 15mM ).The overlap of their 95% confidence intervals indicates CYP1A2, CYP1B1, CYP2C19, and CYP2J2 all have a similar V 15mM for the production of clozapine-N-oxide (Table 1).In contrast, CYP2C19 had the greatest rate of N-desmethylclozapine formation in this system (V MAX ¼ 1.44 pmol/pmol CYP/min, K M ¼ 31.38 mM, Supplementary Table 1).All three hepatic CYP (Figure 1(B)) produced more N-desmethylclozapine than the extrahepatic isoforms (Figure 1(D)), with a relative order of catalytic activity of CYP2C19 > CYP1A2 >CYP3A4 > CYP2J2 > CYP1B1 > CYP1A1 (Table 1).CYP3A4 was the only isoform to favour clozapine-N-oxide formation over clozapine N-demethylation (Table 1).
rate of N-desmethylclozapine formation was close to the limits of detection for the extrahepatic CYP isoforms (Figure 2(D)) but was significantly above zero for all isoforms at 15 mM substrate (Table 2).This minimal turnover, in comparison to the hepatic isoforms, may reflect the lower production of a) clozapine from clozapine-N-oxide (Figure 2(B,C)) Ndesmethylclozapine from clozapine (Figure 1(D)) by these extrahepatic isozymes.

N-desmethylclozapine metabolism
There was no evidence of formation of clozapine (327 m/z) or clozapine-N-oxide (343 m/z) when N-desmethylclozapine was incubated with any of the CYP assayed.A full range mass scan did not result in the detection of additional metabolites derived from N-desmethylclozapine (data not shown).

Co-incubation of equimolar clozapine and clozapine-N-oxide
To confirm which pathway direction (oxidation or reduction) each isoform favoured, equimolar concentrations (0.9 mM or 3 mM) of clozapine and clozapine-N-oxide were co-incubated with individual CYP for 60 min.CYP3A4 strongly favoured the oxidation reaction at both substrate concentrations (Figure 3).Clozapine concentrations decreased by 65.34 ± 4.21% (0.9 mM co-incubation) and 41.77 ± 7.81% (3 mM co-incubation) whilst clozapine-N-oxide concentrations increased by 83.70 ± 22.79% (0.9 mM co-incubation) and 78.51 ± 12.69% (3 mM co-incubation).CYP3A4, CYP1A2, CYP2C19 and CYP1B1 produced statistically significant concentrations of N-desmethylclozapine (Figure 3).Whilst CYP2J2 and CYP1B1 catalysed changes in clozapine-N-oxide concentration in the 0.9 mM (Figure 3(A)) and 3 mM (Figure 3(B)) co-incubations respectively.No corresponding changes were observed in clozapine concentration for these isoforms.No isoforms assayed were associated with a significant change in the sum of measured concentrations relative to non-enzymatic control, suggesting no additional metabolites of note were produced in these incubations.

Formation of reactive species
Throughout each 60-minute incubation period the cumulative formation of DCF was measured as a marker of overall reactive species generation during clozapine and clozapine-N-oxide metabolism.CYP1A2 and CYP2C19 did not produce any detectable reactive species whilst metabolising either substrate (Figure 4(A,B)).However, concentration-dependent generation of reactive species was associated with the CYP3A4-catalysed metabolism of both clozapine (Figure 4(A)) and clozapine-N-oxide (Figure 4(B)).The maximal concentrations of DCF observed were higher in the clozapine-N-oxide incubations (0.52 ± 0.18 mg/mL) compared to those observed with clozapine as the substrate (0.32 ± 0.04 mg/mL) for this enzyme.The CYP2J2-catalysed metabolism of both clozapine (Figure 4(C)) and clozapine-N-oxide (Figure 4(D)) generated significant DCF concentrations, although notably this DCF formation was not dependent on substrate concentration in either direction.CYP1A1 and CYP1B1 did not produce any DCF with clozapine as the substrate (Figure 4(C)) but did so in a concentration-dependent manner in the presence of clozapine-N-oxide (Figure 4(D)).
The contribution of the previously reported (Dragovic et al. 2010(Dragovic et al. , 2013) ) reactive nitrenium and iminium intermediate species in this experimental system was then assessed by glutathione and cyanide trapping, respectively.Five chromatographic peaks were detected at the expected mass to charge ratio of the glutathione nitrenium adducts (632 m/z) in at least three replicate incubations and were labelled in order of their abundance relative to internal standard (PAR); CG-1 (retention time 3.7 min), CG-2 (retention time 5.9 min), CG-3 (retention time 3.1 min), CG-4 (retention time 6.5 min), CG-5 (retention time 4.1 min) (Data not shown).These species could not be quantified as authentic reference standards were not available.CG-1 was the only one of these adducts for which formation was significantly greater than control (95% confidence intervals > zero).CYP3A4 was associated with the highest formation of CG-1 from clozapine.It was the only CYP for which the formation of CG-1 increased in a concentration-dependent manner, and the kinetics of this process were non-linear as would be expected with the saturable formation of a metabolite.CYP3A4 produced significantly more CG-1 with clozapine as a substrate than clozapine-N-oxide, however the reverse was true for other isozymes assayed (Figure 5(A)).Formation of CG-1 from clozapine-N-oxide was highest in the CYP1A1 and CYP1B1 incubations (Figure 5

(B)).
A single chromatographic peak at the mass to charge ratio of the expected cyanide (CN) adduct of the iminium ion (352 m/z), referred to as CN-1, was detected following incubation with CYP3A4.CN-1 formation was minimal and not significantly different from control (data not shown).CN-1 was not detected following incubation of clozapine with other isoforms, nor was it detected in the presence of clozapine-N-oxide as substrate.
The inclusion of glutathione (trapping of nitrenium ion) in the clozapine incubations led to a significant increase in the formation of clozapine-N-oxide by CYP1A2, CYP3A4, and CYP1B1 (Figure 6(A)), and N-desmethylclozapine by CYP1A2, CYP3A4, and CYP2C19 (Figure 6(B)).No change in metabolite formation was observed for CYP2J2 and CYP1A1.Similarly, the addition of cyanide (trapping of iminium ion) to the clozapine incubations resulted in a significant increase in the CYP3A4 catalysed formation of clozapine-N-oxide (Figure 6(A)).This was selective for the N-oxidation pathway as no isoforms had a corresponding increase in the formation of the N-desmethylclozapine metabolite (Figure 6(B)).When clozapine-N-oxide was the substrate, the use of either trapping agent significantly decreased the reductive formation of clozapine catalysed by CYP1A2, CYP2C19, CYP3A4 and CYP1A1 (Figure 6(C)), and increased the formation of N-desmethylclozapine by CYP2C19 and CYP1B1 (Figure 6D).
There was a strong linear correlation between CYP3A4 catalysed rate of formation of clozapine-N-oxide and the concentration of reactive species detected as DCF (q ¼ 1.00, r 2 ¼ 1.00, p ¼ 0.0013).A linear correlation was also observed for CYP1A2 (q ¼ 0.97, r 2 ¼ 0.95, p ¼ 0.0274).For these hepatic isoforms there was also a strong linear relationship between the relative amount of CG-1 detected and the concentration of reactive species detected by DCF (CYP3A4: q ¼ 0.99, r 2 ¼ 0.98, p ¼ 0.0112; CYP1A2: q ¼ 0.99, r 2 ¼ 0.98, p ¼ 0.0093).In contrast, for the extrahepatic CYP, there was no correlation between any metabolic pathways assessed and the overall formation of reactive species [DCF].
Whilst CYP3A4-catalysed reduction of clozapine-N-oxide to clozapine correlated with concentration of reactive species formed [DCF], (q ¼ 0.96, r 2 ¼ 0.93, p ¼ 0.0383), there was no clear association with the relative abundance of CG-1.Furthermore, reduction of clozapine-N-oxide back to clozapine by the extrahepatic isoforms CYP1A1 and CYP1B1 also correlated with reactive species formation [DCF], (CYP1A1: q ¼ 0.99, r 2 ¼ 0.98, p ¼ 0.0104; CYP1B1: q ¼ 0.99, r 2 ¼ 0.99, p ¼ 0.0055).However, there was no correlation observed between the formation of CG-1 and DCF concentration for either of these isoforms.The remaining isoforms did not display any correlation between metabolic products and [DCF].

Discussion
The metabolism of clozapine has been implicated in the development of cardiotoxicity (Williams et al. 2003), as well as variation in plasma concentrations and altered clearance of the drug (Bondolfi et al. 2005;Reeves et al. 2023).The major plasma are expected to be predominately formed by hepatic CYP enzymes.Indeed, as has been reported previously (Eiermann et al. 1997;Fang et al. 1998) we observed that CYP1A2 and CYP2C19 predominately formed N-desmethylclozapine, whilst CYP3A4 favoured the formation of clozapine-N-oxide.The clozapine concentrations used in these in vitro experiments were selected based on clinical data.A concentration of 0.9 mM (300 ng/mL) is a commonly observed plasma concentration of clozapine, just below the therapeutic reference range of 350-600 ng/mL (1.07-1.84mM) (Hiemke et al. 2018) but often reported in patients during their first four weeks of therapy (Perry et al. 1991), a timeframe during which patients are most at risk of developing myocarditis (Bellissima et al. 2018).Moreover, 3.1 mM (1000 ng/mL), is the "laboratory alert level" for clozapine toxicity as suggested by the Arbeitsgemeinschaft f€ ur Neuropsychopharmakologie und Pharmakopsychiatrie (AGNP) consensus guidelines for therapeutic drug monitoring (Hiemke et al. 2018).Concentrations close to or above this value are also frequently observed in patients (Couchman et al. 2013).Additional supra-therapeutic concentrations were utilised in these in vitro experiments because during long-term clozapine use, it is expected that tissue concentrations of the drug will be higher than those observed in plasma.The physicochemical properties of this lipophilic drug suggest it is likely to undergo phospholipid binding and lysosomal distribution, leading to increased concentrations in the heart and in hepatocytes (Daniel 2003).Consistent with this, a recent study in mice has demonstrated that after 21 days of administration cardiac clozapine concentrations were 22-fold higher than plasma concentrations, whilst liver concentrations were �8-fold higher (Li et al. 2022).
As observed in these experiments, the CYP3A4-catalysed metabolism of both clozapine and clozapine-N-oxide was associated with the substantial formation of reactive species.Previous studies have linked CYP3A4 expression with variation in clozapine-dose required to achieve therapeutic concentrations (T� oth et al. 2017).Furthermore, clozapine is reported to induce CYP3A4 activity at concentrations within the range used in this experiment but does not have the same effect on CYP1A2 and CYP2C19 (Danek et al. 2020).If CYP3A4 is induced over the clozapine titration period it could lead to increased plasma concentrations of clozapine-N-oxide and increased reactive species production, thereby increasing the risk of inflammatory reactions.This may be amplified by the concomitant use of sodium valproate, an established risk factor for clozapine-induced myocarditis (Ronaldson et al. 2012), since therapeutic concentrations of valproic acid have also been reported to induce the expression and catalytic activity of CYP3A4 in primary hepatocytes (Cerveny et al. 2007).
As observed in the adduct-trapping experiments, CYP3A4 also produced the greatest amount of nitrenium ion and was the only isoform with detectable formation of the iminium ion, suggesting that CYP3A4 is the major enzyme responsible for the diverse metabolic profile of clozapine.
Among the extrahepatic isoforms, CYP1B1 also appeared to favour the formation of clozapine-N-oxide, similar to CYP3A4.However, unlike the hepatic isoform, this enzyme was not associated with substantial reactive species production unless incubated with high concentrations of clozapine-N-oxide.Furthermore, CYP1A1 catalysed the reduction of clozapine-N-oxide to clozapine, which was associated with reactive species production despite this isoform not catalysing the oxidative reaction.Neither of these extrahepatic isoforms displayed any correlation between nitrenium ion formation and the generation of reactive species (detected by DCF).The lack of correlation between known reactive intermediates and the overall generation of reactive species suggests that there are additional reactive species produced through the reduction of clozapine-N-oxide to clozapine.Furthermore, it is unlikely that the nitrenium ion is the causative factor for clozapine-induced myocarditis.Olanzapine, an alternative antipsychotic with a similar chemical structure, is also oxidised to a reactive nitrenium ion in the same manner as clozapine (Gardner et al. 1998), yet despite producing structurally similar reactive metabolites, olanzapine exhibits neither agranulocyte toxicity (Gardner et al. 1998) nor does it appear to be associated with myocarditis (De Las Cuevas et al. 2022).It is possible that the conversion of clozapine-Noxide to clozapine results in the formation of hydroxyl radicals and ROS that initiate tissue damage.The results of this investigation suggest that the oxidation of DCF-DA is strongly associated with the reduction of clozapine-N-oxide.Since nitrenium ions were not as readily produced following clozapine-N-oxide incubations, it is hypothesised that alternate reactive intermediates and free radicals associated with the loss of the oxide group (i.e.ROS) are more likely to contribute to reactive species mediated damage.
Oxidative stress initiated by ROS is central to the progression of many inflammatory conditions including myocarditis (Tada and Suzuki 2016), and prolonged oxidative stress produced by sub-clinical myocarditis concurrent with long-term clozapine use may be a pathological mechanism behind the cardiac remodelling associated with clozapine-associated cardiomyopathy (Merrill et al. 2005;Tada & Suzuki 2016).Previous studies have also linked clozapine exposure to ROS generation and lipid peroxidation in models of cardiotoxicity (Ahangari et al. 2022;Zhang et al. 2021), suggesting that further investigation into this as a mechanism of toxicity should be considered.Correlation analysis of CYP-catalysed metabolism with [DCF] as an marker of ROS indicated that ROS production was associated with the oxidative reaction (clozapine to clozapine-N-oxide) by hepatic CYP rather than the reduction reaction (clozapine-N-oxide to clozapine).This may suggest that elevated plasma concentration of clozapine-N-oxide may be a better predictor of tissue-associated damage and therefore the risk of patients developing cardiotoxicity.
Despite considerable CYP3A4 activity in hepatocytes and the potential for large clozapine-N-oxide concentrations to be formed in this tissue, the incidence of clozapine-induced severe hepatotoxicity is rare, although reports of elevated transaminases are common (Hummer et al. 1997;Wu Chou et al. 2014).This is likely due to the liver's substantial detoxication capacity, with high expression of conjugation enzymes and glutathione content.GSTA1 and GSTP1 are predominant antioxidant enzymes of the liver, but their expression in cardiomyocytes is low (The Human Protein Atlas 2019b, 2019a), suggesting the heart has limited conjugation capacity and is reliant on antioxidants, and other modulators of inflammation for protection against reactive compounds.Interestingly, the use of glutathione and cyanide to trap reactive intermediates resulted in altered enzyme activity.Whilst there was increased formation of both N-desmethylclozapine and clozapine-N-oxide from clozapine in the presence of glutathione, trapping of the reactive nitrenium ions also decreased the reduction of clozapine-N-oxide back to clozapine.This suggests that glutathione may stabilise clozapine-N-oxide and prevent REDOX cycling.Therefore, it may be that in patients with depleted glutathione there will be increased REDOX cycling and elevated ROS production, contributing to tissue damage.
Thioredoxin is an endogenous antioxidant and anti-inflammatory protein that can protect cardiac tissue from free-radical mediated adverse drug reactions (Das et al. 2021;Murata et al. 2022) through scavenging ROS and modulating chemotaxis (Whayne et al. 2015).However, it has a small molecular weight (�12kDa) and is readily eliminated from circulation, limiting its systemic antioxidant capacity (Murata et al. 2022).Similar cardioprotective effects are offered by epoxyeicosatrienoic acids, which are produced in cardiac tissue by CYP2J2-catalysed metabolism of arachidonic acid (Node et al. 1999;Wang et al. 2021).
CYP2J2 is the predominant isoform of the heart, with a 50-fold greater expression than any other isoform (Evangelista et al. 2013) and is therefore likely to contribute substantially to the metabolic profile of this tissue.Previous studies have identified that clozapine can inhibit this enzyme (Ren et al. 2013), but whether CYP2J2-catalysed clozapine metabolism has any impact on arachidonic acid turnover by this isoform is currently unknown.Whilst our results suggest clozapine is not a good substrate for CYP2J2, since it does not readily favour clozapine-N-oxide or N-desmethylclozapine, we cannot discount the possibility that CYP2J2 catalyses the reduction and oxidation at similar rates, resulting in no apparent change in substrate concentration, despite ROS generation detected as [DCF].
This study highlights the importance of investigating a drug and its metabolites in depth.Reactive clozapine intermediates have been detected in this study through the formation of glutathione conjugates and various cyanide adducts, which corroborates their previous identification in vitro and in vivo (Dragovic et al. 2010(Dragovic et al. , 2013;;Maggs et al. 1995;Pirmohamed et al. 1995;Williams et al. 2003).The ability of these reactive nitrenium and iminium ions to undergo protein adduction has often been hypothesised as a contributing factor to immunogenic reactions such as myocarditis (Williams et al. 2003).However, it is clear from this study that the formation of nitrenium and iminium conjugates do not correlate with the overall formation of reactive species associated with the metabolism of clozapine nor the CYP-catalysed reduction of clozapine-N-oxide back to clozapine.This novel finding suggests that these reactive ions alone are unlikely to be the sole contributor underpinning clozapineinduced cardiotoxicity, and that ROS produced through the REDOX cycling of clozapine and clozapine-N-oxide may instead induce oxidative stress and tissue damage leading to pro-inflammatory eosinophilic infiltrates.
Whilst there are differences worldwide around the regulations of clozapine use (Nielsen et al. 2016), it is not routine to monitor clozapine concentrations in early treatment.In the local region (Auckland, New Zealand), patients are monitored early in treatment for the development of agranulocytosis (Medsafe 2015).However, plasma concentrations of the drug or its N-desmethylclozapine metabolite are not monitored unless overt toxicity is suspected.The lack of specific symptoms or precise biomarkers makes clozapineinduced myocarditis challenging to diagnose causing a reluctance to use the drug, and its underuse in patients likely to benefit from its antipsychotic activity (Patel et al. 2019).Early detection of myocarditis is essential, since some reports have myocarditis case fatalities of more than 50%, although this is thought to be less than 10% when appropriate monitoring is in place (Ronaldson 2017).The results of this in vitro work, combined with the observation of elevated plasma clozapine-N-oxide/N-desmethylclozapine ratios in patients with myocarditis (Bellissima 2018), suggest that the inclusion of clozapine-N-oxide plasma concentrations alongside measuring clozapine in routine monitoring may improve the determination of cardiotoxicity risk.However, this work is limited by its use of single CYP isoforms to predict the true risk of reactive species generation in vivo.More holistic cell-and tissue-based models, as well as a clearer picture of free concentrations within human cardiac tissue, are required to better understand the likely risk of cardiac damage as a result of CYP-catalysed REDOX cycling of clozapine and clozapine-Noxide.
In conclusion, this study has identified a strong correlation between the CYP-catalysed metabolism of clozapine and clozapine-N-oxide with the formation of reactive species in vitro.This work implicates the contribution of REDOX cycling of the N-oxide metabolite as a mechanism of toxicity and highlights the importance of monitoring plasma concentrations of clozapine and clozapine-N-oxide to identify the potential for elevated ROS and risk of cardiotoxicity.

Figure 1 .
Figure 1.Formation of clozapine-N-oxide and N-desmethylclozapine following incubation of clozapine with isolated CYP isoforms.Michaelis-Menten models for the formation of (A) clozapine-N-oxide by hepatic CYP isoforms, (B) N-desmethylclozapine by hepatic CYP isoforms, (C) clozapine-N-oxide by extrahepatic CYP isoforms, and (D) N-desmethylclozapine by extrahepatic CYP isoforms are shown.DCF-DA was included in all incubations for matched detection of reactive species.Rate of metabolite formation (pmol/pmol CYP/min) is reported normalised to a non-enzymatic control.Data are presented as mean ± SEM (n ¼ 6).
(B)), this conversion was substantially slower (�6-fold) than for the formation of clozapine.The chemical structure of N-desmethylclozapine likely precludes its formation directly from clozapine-N-oxide, hence detection of Ndesmethylclozapine (Figure2(B)) may instead be a result of the initial reductive biotransformation of clozapine-N-oxide to clozapine (Figure2(A)), which would then be a substrate for efficient N-demethylation by the same CYP enzyme (Figure1(B)).This putative mechanism is biochemically feasible(Bickel 1969;Manvich et al. 2018;Raper et al. 2017).The

Figure 2 .
Figure 2. Formation clozapine and N-desmethylclozapine following incubation of clozapine-N-oxide with isolated CYP isoforms.Michaelis-Menten models for the formation of (A) clozapine by hepatic CYP isoforms, (B) N-desmethylclozapine by hepatic CYP isoforms, (C) clozapine by extrahepatic CYP isoforms, and (D) N-desmethylclozapine by extrahepatic CYP isoforms are shown.DCF-DA was included in all incubations for the matched detection of reactive species.Rate of metabolite formation (pmol/pmol CYP/min) was normalised to a non-enzymatic control.Data are presented as mean ± SEM (n ¼ 6).

Figure 3 .
Figure 3. Change in measured concentrations of substrate and N-desmethylclozapine following the co-incubation of equimolar amounts of clozapine and clozapine-N-oxide with CYP isoforms.(A) Co-incubation of 0.9 mM clozapine þ 0.9 mM clozapine-N-oxide with individual CYP-isoforms and non-enzymatic control (NEC).(B) Co-incubation of 3 mM clozapine þ 3 mM clozapine-N-oxide with individual CYP-isoforms and non-enzymatic control (NEC).Measured concentrations were derived from a validated standard curve (R 2 > 0.98).Data are presented as mean ± SEM (n ¼ 6).Statistically significant differences relative to the non-enzymatic controls were assessed with a two-way ANOVA and Dunnett's multiple comparisons post-hoc test � p < 0.05, �� p < 0.01, ��� p < 0.001 and ���� p < 0.0001.

Figure 4 .
Figure 4. Cumulative generation of reactive species associated with CYP-catalysed metabolism with either clozapine or clozapine-N-oxide as substrate, as detected by formation of DCF.(A) Formation of reactive species associated with metabolism of clozapine by hepatic CYP-isoforms.(B) Formation of reactive species associated with metabolism of clozapine-N-oxide by hepatic CYP isoforms.(C) Formation of reactive species associated with metabolism of clozapine by extrahepatic CYP isoforms.(D) Formation of reactive species associated with metabolism of clozapine-N-oxide by extrahepatic CYP isoforms.CYP enzymes were incubated with varying concentrations of clozapine or clozapine-N-oxide and 7.5 mg/mL DCF-DA.Subsequent DCF production was detected with fluorescence spectroscopy (485 nm EX; 520 nm EM).Concentrations of DCF were calculated from a valid standard curve (R 2 > 0.99), then normalised to background CYP-catalysed conversion of DCF-DA to DCF and non-enzymatic controls.Data are presented as mean ± SEM (n ¼ 6).Statistically significant differences relative to the negative controls were assessed with a two-way ANOVA and Dunnett's multiple comparisons post-hoc test � p < 0.05, �� p < 0.01, ��� p < 0.001, ���� p < 0.0001.

Figure 5 .
Figure 5. Formation of nitrenium-glutathione adduct 1 (CG-1) following incubation of clozapine or clozapine-N-oxide with isolated CYP isoforms.Michaelis-Menten models for the formation of (A) CG-1 by CYP isoforms in the presence of clozapine, (B) CG-1 by CYP isoforms in the presence of clozapine-N-oxide.Concentrations of CG-1 could not be calculated due to a lack of authentic standard.Data are presented as mean peak area ratio ± SEM (n ¼ 6).

Figure 6 .
Figure 6.Comparison of the effect of trapping nitrenium (GSH adduct) or iminium (CN adduct) ions on the rate of oxidation, reduction or demethylation of clozapine and clozapine-N-oxide.(A) Effect on the oxidation pathway (clozapine-N-oxide formation from clozapine).(B) Effect on the demethylation pathway (N-desmethylclozapine formation from clozapine substrate).(C) Effect on the reduction pathway (clozapine formation when clozapine-N-oxide is a substrate).(D) Effect on the demethylation of clozapine formed from the reduction of clozapine-N-oxide.The rate of velocity was compared from 15 mM substrate concentrations.Measured metabolite concentrations were derived from a validated standard curve (R 2 > 0.98) and baseline-corrected to the non-enzymatic control to account for spontaneous turnover of substrate.Data are presented as mean ± SEM (n ¼ 6).Statistically significant differences between test conditions were assessed with a two-way ANOVA and Dunnett's multiple comparisons post-hoc test � p < 0.05, ��� p < 0.001 and ���� p < 0.0001.