CYP2J2 Molecular Recognition: A New Axis for Therapeutic Design

Abstract

Cytochrome P450 (CYP) epoxygenases are a special subset of heme-containing CYP enzymes capable of performing the epoxidation of polyunsaturated fatty acids (PUFA) and the metabolism of xenobiotics. This dual functionality positions epoxygenases along a metabolic crossroad. Therefore, structure-function studies are critical for understanding their role in bioactive oxy-lipid synthesis, drug-PUFA interactions, and for designing therapeutics that directly target the epoxygenases. To better exploit CYP epoxygenases as therapeutic targets, there is a need for improved understanding of epoxygenase structure-function. Of the characterized epoxygenases, human CYP2J2 stands out as a potential target because of its role in cardiovascular physiology. In this review, the early research on the discovery and activity of epoxygenases is contextualized to more recent advances in CYP epoxygenase enzymology with respect to PUFA and drug metabolism. Additionally, this review employs CYP2J2 epoxygenase as a model system to highlight both the seminal works and recent advances in epoxygenase enzymology. Herein we cover CYP2J2’s interactions with PUFAs and xenobiotics, its tissue-specific physiological roles in diseased states, and its structural features that enable epoxygenase function. Additionally, the enumeration of research on CYP2J2 identifies the future needs for the molecular characterization of CYP2J2 to enable a new axis of therapeutic design.

1. Introduction

Epidemiological evidence suggests that a diet rich in polyunsaturated fatty acids (PUFAs) is beneficial for various health conditions (Bowen et al., 2016; Kris-Etherton et al., 2003; Maki et al., 2018; Simopoulos, 2008). The biochemical mechanisms facilitating these effects are being gradually unveiled (Fischer et al., 2014; Schunck et al., 2018). Previous studies demonstrated that PUFAs are converted to bioactive lipid metabolites through both oxidative and non-oxidative routes of metabolism. The oxidative metabolism of PUFAs are broadly initiated by three classes of enzymes: cyclooxygenase (COX), lipoxygenase (LOX), and cytochrome P450 (CYP) epoxygenase (EPOX) (Figure 1) (Brash, 2001; Spector, 2009b; Spector and Kim, 2015). While the COX and LOX pathways are best known for converting arachidonic acid (AA) into pro-inflammatory mediators, termed “prostanoids” and “leukotrienes” (Brash, 2001), the EPOX pathway converts AA to anti-inflammatory epoxide mediators “epoxyeicosatrienoic acid (EETs)” (Spector et al., 2004; Spector and Kim, 2015; Xu et al., 2011a). Mechanistic studies on the COX and LOX enzymes have led to the development of several drugs such as non-steroidal anti-inflammatory drugs (NSAIDs) (Luong et al., 1996) and anti-allergic drugs (Bruno et al., 2018; Charlier and Michaux, 2003). Additionally, they improved the understanding of pro-resolving mediators, such as the resolvins (Serhan et al., 2008; Serhan and Levy, 2018).

Figure 1. Overview of fatty acid and endocannabinoid metabolism by cytochrome P450 classes.

A lipidomic view of prostaglandin H2(PGH2), anandamide (AEA), arachidonic acid (AA), docosahexaenoic acid (DHA) and leukotriene B₄ metabolism by functionally distinct, membrane-bound CYPs. The CYPs are organized into classes based on their primary metabolism pathway and are shown as colored ovals. The substrates of each reaction are shown on the left-hand side of the enzyme, while the products are shown to the right of the enzyme. (A) The prostanoid- and thromboxane-producing CYPs include CYP5A1 and CYP8A1. While CYP5A1 catalyzes the conversion of PGH2 to Thromboxane A2 (TXA₂), a potent vasoconstrictor, CYP8A1 produces PGI2-the opposing vasodilatory metabolite. (B) CYP2C8, CYP2C9, andCYP2J2 display epoxygenase activity and can metabolize AA, DHA, and the endocannabinoid AEA to bioactive metabolites in a regioselective and enantioselective manner. (C) The CYP4F class consisting of CYP4F12, CYP4F2 and CYP4F3. In particular, they catalyze hydroxylation reactions to produce the hydroxyeicosatetraenoic acids (HETEs) and hydroxylated leukotriene metabolites, such as leukotriene B₄. Although hydroxylation is the primary metabolic pathway for the CYP4F class, they have been shown to metabolize polyunsaturated fatty acids such as AA.

In contrast, there are limited therapeutics targeting the EPOX pathway. This is likely due to an incomplete understanding of the total biosynthetic machinery used to generate bioactive epoxides, structure-function of the enzymes involved, as well as the overall physiology and pathology of the bioactive epoxides. Past reports indicate that the ω−6 PUFA epoxides such as EETs have anti-inflammatory and cardioprotective properties (Zeldin, 2001). Likewise, the ω−3 PUFA epoxides protect against cardiac arrhythmia, reduce inflammatory and neuropathic pain, and prevent tumor metastasis (Fischer et al., 2014; Spector and Kim, 2015). An excellent review summarizing the recent advances on omega-3 PUFA epoxides, including their generation by EPOX enzymes, their metabolic fate, known biological activities, and roads to clinical applications, has been published elsewhere (Schunck et al., 2018).

The PUFA-epoxides are rapidly degraded via soluble epoxide hydrolase (sEH) to diols. Advances in studying the physiological role of the PUFA-epoxides have come from the use of potent selective sEH inhibitors (sEHi) that indirectly increase the levels of the PUFA-epoxides (Harris et al., 2008; Imig and Hammock, 2009). Bioisosteric analogues of PUFA-epoxides analogs with urea or oxamide groups, in lieu of the epoxide, were applied to target cardiovascular diseases and pain (Falck et al., 2009). However, no current drugs exist that directly regulate the EPOX enzymes. There will be benefits for directly inhibiting EPOX enzymes in diseases states where increases in the EETs leads to poor disease outcomes. Given that past research efforts have been directed towards inhibiting PUFA-epoxide degradative enzymes, epoxide lifetime and activity could be enhanced by drugs which could modulate EPOX enzymes for elevated or more targeted biosynthesis. Several reports suggest that particular CYP isoforms contribute to cancer (Jiang et al., 2007), neurodegenerative disease (Yan et al., 2015), and cardiovascular disease pathology (Aliwarga et al., 2018). Targeting these CYPs could offer a new approach for addressing these disorders. Co-administration with existing pharmaceuticals may enhance therapeutic efficacy and decrease off-target effects (Sevrioukova and Poulos, 2013c).

Establishing structure-function relationships is imperative to design direct therapeutics and inhibitors that target EPOX enzymes. These therapeutics could increase select epoxide metabolites, while indirect methods increase all regioisomers. Specific isomers display distinct functions and isoform-selective inhibitors can tune the pool of metabolites to access desirable biological effects. Herein, select EPOX enzyme structure-function relationships will be reviewed with respect to PUFA-lipid metabolism to aid in designing therapeutics targeting this pathway.

Human CYP isoforms exhibit polymorphisms. CYPs also exhibit high degrees of substrate permissiveness, enabling them with a potential to perform lipid hydroxylation and/or epoxidation. Several of the cytochrome P450 (CYP) epoxygenase enzymes characterized as lipid metabolizers belong to the CYP2 and CYP4 classes. Although “lipid” is a broad term, for this review, we refer to lipids as either polyunsaturated fatty acids (PUFAs) or endocannabinoid (eCB) signaling molecules. While lipid-metabolizing CYPs perform a great array of catalytic function on lipid substrates, we will specifically focus on the monooxygenase reactions of these enzymes to generate PUFA- or eCB-epoxide products (Table 1 and Table 2). The focus of this review to highlight epoxygenase function does not preclude any physiological roles of epoxygenase-derived hydroxylation products. Many of these CYP isoforms including CYP2J2 have the potential to perform subterminal mono-hydroxylation reaction on the PUFA substrate. For instance, it was shown previously that CYP2J2-mediated PUFA metabolism led to the formation of 19-HETE that is cardioprotective in nature. In particular, it was shown that 19(R)-HETE acts as a 20-HETE antagonist which is involved in the development of stroke (Cheng et al., 2008); 19(S)-HETE has been shown to protect against Ang II-induced cardiac hypertrophy (Shoieb and El-Kadi, 2018). Separately, it was shown that mouse CYP2J9 is abundant in brain is involved in the biosynthesis of 19-HETE, an eicosanoid that inhibits activity of P/Q-type Ca²⁺ channels (Qu et al., 2001). Therefore, the PUFA-hydroxylated product synthesized by CYP2J2 and other CYPs play important physiological roles in the body.

Table 1.

Rates and product distributions of EET-like epoxide generation by human CYP epoxygenases.

Human Cytochrome P450s		Total metabolism		Vmax of individual epoxide regioisomer formation [% product distribution]
	Enzyme Preparation	Substrate	Vmax	ω−3 epoxide	ω−6 epoxide	ω−9 epoxide	ω−12 epoxide	ω−15 epoxide	Ref
2J2	Recombinantly expressed in Sf9 insect cells	AA	65 A	n/a	53.3 ± 2.9 A	34.8 ± 1.5 A	8.9 ± 1.1 A	3.1 ± 0.3 A	(Wu et al., 1996)
	Recombinantly expressed in E. coli DH5α cells	AA	100 A	n/a	[37]	[18]	[24]	[21]	(McDougle et al., 2014)
	Recombinantly expressed in E. coli DH5α cells	AA	189 ± 9 A	n/a	[40]*	[20]*	[10]*	[5]*	(Arnold et al., 2016)
	Supersome coexpressing CPR	AA	56 ± 4 A	n/a	[30]	[25]	[19]	ND	(Arnold et al., 2010)
	Recombinantly expressed in E. coli DH5α cells	AEA	135 ± 14 A	n/a	[51.7]	[36.6]	[11.3]	[0.4]	(Arnold et al., 2010)
	Recombinantly expressed in E. coli DH5α cells	DHA	254 ± 11 A	ND	ND	ND	ND	ND	(Arnold et al., 2016)
	Supersome coexpressing CPR	DHA	228 ± 8 A	[27]	[7]	[3]	[5]	[2]	(Arnold et al., 2010)
	Recombinantly expressed in E. coli DH5α cells	EPA	259 ± 77 A	ND	ND	ND	ND	ND	(Arnold et al., 2016)
	Supersome coexpressing CPR	EPA	943 ± 17 A	[62]	[15]	[5]	[4]	ND	(Arnold et al., 2010)
	Recombinantly expressed in E. coli DH5α cells	LA	146 ± 17 A	n/a	ND	ND	n/a	n/a	(Arnold et al., 2016)
	Recombinantly expressed in E. coli DH5α cells	2-AG	390 A	n/a	265.4 ± 12.3 A	125.5 ± 5.2 A	0	0	(McDougle et al., 2014)
2B1	Sprague-Dawley rat microsomes	AA	1616 ± 96 A	n/a	[54.5 ± 1.5]	[18.9 ± 1.9]	[16.7 ± 1.3]	[9.8 ± 0.9]	(Laethem et al., 1994)
	Sprague-Dawley rat microsomes	AA	0.30 B	n/a	110 B [30]	110 B [37]	80 B [27]	[<6]	(Capdevila et al., 1990b)
2B2	Sprague-Dawley rat microsomes	AA	404 ± 65 A	n/a	[32.9 ± 0.7]	[22.5 ± 1.1]	[20.8 ± 1.5]	[23.7 ± 2.0]	(Laethem et al., 1994)
	Sprague-Dawley rat microsomes	AA	0.10 B	n/a	30 B [42]	30 B [26]	30 B [27]	[<5]	(Capdevila et al., 1990b)
2B6	Recombinantly expressed in E. coli C41 DE3 cells	AA	17.316 C [12]	n/a	2.040 C [11.7]	7.560 C [43.7]	3.12 C [18.0]	4.596 C [26.6]	(Sridar et al., 2011)
2B12	Recombinantly expressed in E. coli (cell type not specified)	AA	ND	n/a	ND	[80]	[20]	ND	(Keeney et al., 1998)
2C8	Human liver microsomes (HLM)	AA	96 ± 15 D [66]	n/a	27 ± 4 D	ND	ND	ND	(Rifkind et al., 1995)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	AA	ND	n/a	[48]	[52]	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	EPA	ND	[45]	[30]	[22]	[3]	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DPA	ND	[35]	[26]	[14]	[25]	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DHA	ND	[39]	[28]	[20]	[13]	ND	(Arnold et al., 2010)
2C9	Human liver microsomes (HLM)	AA	56 ± 4 D [52]	n/a	24 ± 4 D	ND	ND	ND	(Rifkind et al., 1995)
	Human liver microsomes (HLM)	AA	1028 ± 92 A	n/a	[60.5]	[26.3]	[13.2]	ND	(Daikh et al., 1994b)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	AA	ND	n/a	[52]	[30]	[18]	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	EPA	ND	[7]	[45]	[28]	[20]	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DPA	ND	0	[16]	0	[73]	[10]	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DHA	ND	[0]	[16]	[9]	[70]	[6]	(Arnold et al., 2010)
2C10	Recombinantly expressed with a MAXBAC Baculovirus Expression System	AA	0.75 B [90]	n/a	[52]	[30]	[17]	[≤1]	(Zeldin et al., 1995)
2C19	Supersome coexpressing CPR	AA	20.5 B*	n/a	[25]	[4]	[17]	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	EPA	23.0 B*	[11]	[8]	[8]	[8]	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DPA	20.0 B*	[13]	[20]	[20]	[20]	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DHA	19.0 B*	[11]	[8]	[16]	[17]	ND	(Arnold et al., 2010)
2D6	Recombinantly expressed in E. coli C41 DE3 cells	AEA	ND	n/a	1.3 C	1.1 C	1.6 C	0	(Snider et al., 2008)
2E1	Human liver microsomes; Supersome coexpressing CPR	AA	0	n/a	ND	ND	ND	ND	(Rifkind et al., 1995)
	Supersome coexpressing CPR	EPA	ND	[37]	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DPA	ND	[54]	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DHA	ND	[47]	ND	ND	ND	ND	(Arnold et al., 2010)
1A1	Sprague-Dawley rat microsomes	AA	40 A	n/a	10 A [61]	10 A [22]	10 A [17]	[<1]	(Capdevila et al., 1990b)
1A2	Recombinantly expressed in Human HepG2 hepatoma cells	AA	39 ± 7 D [49]	n/a	12 ± 2 D	ND	ND	ND	(Rifkind et al., 1995)
	Sprague-Dawley rat microsomes	AA	70 A	n/a	20 A [23]	40 A [58]	10 A [15]	[<4]	(Capdevila et al., 1990b)
3A4	Transfection into MCF-7 cell line using siRNA	AA	2.279 C	n/a	0.11 ± 0.798 C [4.8]	0.849 ± 0.13 C [37.3]	0.632 ± 0.09 C [27.7]	0	(Mitra et al., 2011)
	Recombinantly expressed in E. coli (cell type not specified)	AEA	6.23 C	n/a	2.13 C	2.56 C	0.85 C	0.69 C	(Pratt-Hyatt et al., 2010)
4A11	Supersome coexpressing CPR	AA	ND	n/a	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	EPA	ND	[10]	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DPA	ND	ND	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing CPR	DHA	ND	[8]	ND	ND	ND	ND	(Arnold et al., 2010)
4A12A	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	AA	ND	n/a	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	EPA	ND	[58]	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DPA	ND	[37]	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DHA	ND	[31]	ND	ND	ND	ND	(Arnold et al., 2010)
4A12B	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	AA	ND	n/a	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	EPA	ND	[72]	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DPA	ND	[48]	ND	ND	ND	ND	(Arnold et al., 2010)
	Recombinantly coexpressed with CPR in baculovirus/Sf9 insect cells	DHA	ND	[44]	ND	ND	ND	ND	(Arnold et al., 2010)
4F2	Supersome coexpressing cytochrome b5	AA	ND	n/a	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing cytochrome b5	EPA	ND	[4]	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing cytochrome b5	DPA	ND	[3]	ND	ND	ND	ND	(Arnold et al., 2010)
	Supersome coexpressing cytochrome b5	DHA	ND	[1]	ND	ND	ND	ND	(Arnold et al., 2010)
4X1	Recombinantly expressed in E. coli DH5α	AEA	200 A	n/a	n/a	200 A [100]	0	0	(Stark et al., 2008)
	Recombinantly expressed in E. coli DH5α	AA	27 A	n/a	18 A	0	9 A	0	(Stark et al., 2008)
	Recombinantly expressed in E. coli DH5α	2-AG	< 5 A	n/a	0	0	0	0	(Stark et al., 2008)

^A- pmol product / min / nmol CYP;

^B- nmol product / min / nmol CYP;

^C- pmol product / min / pmol CYP;

^D- pmol / mg CYP (in microsome);

^*- estimated from graph; exact numbers not provided;

^‡- estimated from chromatogram;

n/a – not applicable (given structure);

ND – no data provided

Note: Most of the CYP enzymes listed in this table simultaneously generate subterminal hydroxy-metabolites, which were not included for calculating the relative formation of the regioisomeric epoxy-metabolites.

TABLE 2.

Enantioselectivity of arachidonic acid metabolism by CYP epoxygenases.

	ω−6 epoxide enantioselectivity		ω−9 epoxide enantioselectivity		ω−12 epoxide enantioselectivity
CYP	(R,S)	(S.R)	(R,S)	(S.R)	(R,S)	(S.R)	Ref
2J2	76	24	49	51	47	53	(Wu et al., 1996) (Zeldin et al., 1996)
2J3	43	57	62	38	60	40	(Wu et al., 1997)
2B1	64	36	14	86	14	86	(Capdevila et al., 1990b)
2B2	65	35	16	84	10	90	(Capdevila et al., 1990b)
2B12	ND	ND	30	70	61	39	(Keeney et al., 1998)
2C8	82	18	82	18	ND	ND	(Zeldin et al., 1995; Zeldin et al., 1996)
	86.2	13.8	81.1	18.9	ND	ND	(Daikh et al., 1994b)
2C9	63	37	31	69	ND	ND	(Zeldin et al., 1996)
	62.5	37.5	30.5	69.4	ND	ND	(Daikh et al., 1994b)
2C10	63	37	31	69	34	66	(Zeldin et al., 1995; Zeldin et al., 1996)
2C11	49	51	54	46	42	58	(Capdevila et al., 1990b)
	38	62	79	21	54	46	(Holla et al., 1999)
2C23	25	75	89	11	94	6	(Holla et al., 1999)
2C24	67	33	75	25	39	61	(Holla et al., 1999)
2C44	ND	ND	94	6	95	5	(DeLozier et al., 2004)
2C50	63.8	36.2	58	42	57.1	42.9	(Wang et al., 2004)
2C54	48.7	51.3	46.5	53.5	46.6	53.4	(Wang et al., 2004)
2C55	40.4	59.5	57	43	59	41	(Wang et al., 2004)
2D18	71	29	76	24	48	52	(Thompson et al., 2000)
2N1	49	51	92	8	91	9	(Oleksiak et al., 2000)
2N2	32	68	70	30	90	10	(Oleksiak et al., 2000)
1A1	85	15	13	87	13	87	(Capdevila et al., 1990b)
1A2	63	37	95	5	7	93	(Capdevila et al., 1990b)

While the epoxyeicosatrienoic acids (EETs) generated from CYPs are mostly anti-inflammatory and vasodilatory (Xu et al., 2011a), the hydroxylated products (19- and 20-HETE) are vasodilatory (Tunaru et al., 2016) and vasoconstrictive, respectively (Fan et al., 2016) (Escalante et al., 1989; Ishizuka et al., 2008; Kim et al., 2013; Schwartzman et al., 1989). Meanwhile, CYP2 and CYP4 class enzymes mediate lipid metabolism of PUFA alkenes into PUFA-epoxides that are primarily anti-inflammatory in nature (Node et al., 1999). In addition to performing these epoxidation reactions, several CYP4 class enzymes can hydroxylate PUFAs yielding pro-inflammatory hydroxy-metabolites (Kalsotra and Strobel, 2006). Certain CYPs have been linked with PUFA epoxidation, primarily through the association of the enzyme’s physiological role and their characterized epoxygenase function (Spector, 2009a; Spector and Kim, 2015).

To develop therapeutics that directly target the EPOX pathway, a thorough understanding of the tissue-specific expression, structure-function relationships, biochemical mechanisms, and physiological roles of both EPOX bioactive metabolites and their biosynthetic machinery is required. In this review, we use human CYP Family 2 Subfamily J Member 2 (CYP2J2) as a model epoxygenase to better understand the biochemical and physiological functions, the kinetics of PUFA metabolism, and the structural details, of CYP epoxygenases. CYP2J2 is chosen due to its tissue-specific expression patterns; the regioisomeric production of its metabolites; expansive structure-function and biophysical studies covering lipid and xenobiotic metabolism; and its history in computational modeling, where the takeaways from each can be applied to the study of CYP epoxygenases for therapeutic development at large.

2. Functional classification of cytochrome P450s and their interactions with redox partners

CYPs comprise a superfamily of membrane-bound hemoproteins that are located in either the endoplasmic reticulum (ER) or the mitochondria (Gonzalez and Gelboin, 1992). In humans, 57 CYPs have been identified (Guengerich et al., 2005) although there are discrepancies in the number of enzymes considered “functional” due to differences in genetic processing and stability. Thus, not all CYPs are studied proportionately. CYPs are named for their characteristic Soret absorbance at 450 nm when carbon monoxide is bound to the reduced CYP (Omura and Sato, 1964). CYPs have universal ability to insert an oxygen atom at C-H, C=C, or N- bonds, and sometimes perform degradative functions such as N- or O-dealkylation (Rittle and Green, 2010). CYPs can catalyze a wide variety of chemical reactions, ranging from carbon hydroxylation, heteroatom release and oxygenation, functional group migration, aromatization, and epoxidation (Denisov et al., 2005; Guengerich, 2001; Guengerich and Munro, 2013; Makris et al., 2005; Meunier et al., 2004; Munro et al., 2013; Poulos, 2014). In addition, many other types of “atypical” CYP reactions are also possible. For example, CYP5A1 and CYP8A1 are atypical CYPs since they are isomerases as opposed to oxidases (Yokoyama et al., 1991) (Das et al., 2014; Meling et al., 2015b; Yokoyama et al., 1996).

Although the reactions are chemically diverse, many display a common feature – monooxygenase activity. CYPs can insert an oxygen atom into unactivated C-H bonds at physiological temperatures and pressures to convert olefinic substrates into oxygen-containing products, such as epoxides, via oxidative reactions.

CYPs are broadly classified into two categories based on their function: (1) biosynthetic CYPs are involved in synthesizing endogenous oxy-lipids or steroids and (2) xenobiotic-metabolizing CYPs that are involved in drug metabolism. While these two classes of CYPs utilize the same overall catalytic mechanism, their ability to uniquely position substrates for metabolism enables different physiological functions.

There are several CYPs involved in steroidogenesis. These include CYP11A1 (conversion of cholesterol to pregnenolone) (Hanukoglu, 1992), CYP17A1 (hydroxylation or C-C bond scission of pregnenolone and progesterone) (Hall, 1991), CYP19A1 (conversion of testosterone to estrogen) (Glubb et al., 2017), and CYP11B1 (conversion of 11-deoxycortisol and 11-deoxycorticosterone to cortisol and corticosterone, respectively) (Hanukoglu et al., 1981). Additionally, CYP2R1 hydroxylates vitamin D to calcifediol (Cheng et al., 2004) and CYP27B1 hydroxylates calcifediol to calcitriol (Takeyama et al., 1997). Importantly, these biosynthetic steroidogenic CYPs are very selective for their substrate recognition and generally do not metabolize other substrates (de Ronde and de Jong, 2011; Glubb et al., 2017; Howell et al., 2005) (DeVore, 2012).

The xenobiotic-metabolizing CYPs are involved in phase 1 drug metabolism (Galetin and Houston, 2006; McDonnell and Dang, 2013). Contrary to steroid-synthesizing CYPs, the xenobiotic CYPs, such as CYP3A4, can effectively metabolize any foreign substance that enters the body. CYPs in the 2C family, as well as 3A4, can leverage their promiscuity through cooperative binding mechanisms, or use their larger active sites, to accommodate more, as well as bulky, molecules. Hence, the “one enzyme one substrate” hypothesis breaks down for several xenobiotic CYPs such as CYP3A4, as they exhibit complex substrate-substrate interactions due to the concurrent, or cooperative, binding of multiple drugs at allosteric sites or even at the active site (Atkins, 2005; Denisov et al., 2009; Korzekwa et al., 1998). For example, the modulation of midazolam binding by carbamazepine in CYP3A4 leads to a shift in the site of metabolism from 1’-hydroxylation of midazolam to 4-hydroxylation (Denisov and Sligar, 2016; Denisov and Sligari, 2017; Roberts et al., 2011).

CYPs metabolizing polyunsaturated fatty acids (PUFAs) are at the crossroads of bioactive oxy-lipid synthesis and drug metabolism. Hence, it becomes imperative to study these CYPs, as they can show complex drug-PUFA interactions. PUFA-metabolizing CYPs can function as either omega-hydroxylases or epoxygenases, where differences in product distribution bear physiological significance. Several CYPs function as both omega-hydroxylases and epoxygenases. Epoxygenases convert alkenes, located along the fatty acid chain, to epoxides (Fischer et al., 2014; Ozawa et al., 1986; Spector, 2009b; Spector et al., 2004; Spector and Kim, 2015). Most of the PUFA-metabolizing CYPs are also involved in drug metabolism. Hence, it is apparent that drugs binding to PUFA-metabolizing CYPs will affect PUFA metabolism and vice versa through some form of cooperative interactions. For instance, 6-hydroxyflavone was shown to noncompetitively inhibit CYP2C9 and it was predicted to bind to a site just outside the heme pocket and preventing substrate ingression (Si et al., 2009). The impact of lipid-drug interactions on the metabolism of PUFAs by these CYPs is an area of active interest and highlights the importance of studying CYP epoxygenases if these enzymes are to be targeted or modulated by therapeutics.

Most human CYPs, with the exception of isomerases, require redox partners to deliver two electrons to the heme iron at distinct points in the catalytic cycle (Figure 1A). In particular, cytochrome P450 reductase (CPR) is the primary redox partner of various microsomal CYPs (Lu et al., 1969). Other reductases, such as adrenodoxin (ADX) and adrenodoxin reductase (ADXR), perform redox transfer for mitochondrial CYPs (Hanukoglu, 1992). Cytochrome b5 can also act as an electron mediator between CPR and CYPs and subsequently modulate CYP substrate metabolism, as seen with CYP17A1 (Duggal et al., 2018; Duggal et al., 2016; Estrada et al., 2013; Porter, 2002). In addition to the interplay among CYPs and their redox proteins, the membrane itself is an important component in controlling the putative activity of CYPs (Scott et al., 2016). Together, the intricate protein-protein and protein-lipid interactions help shuttle high energy electrons to the CYP catalytic center.

3. Discovery and functional elucidation of CYP epoxygenase activity

In 1981, non-prostanoid and leukotriene AA metabolites were identified in microsomal CYP systems (Capdevila et al., 1981; Morrison and Pascoe, 1981). It was confirmed that CYP activity was responsible for PUFA monooxygenation based on the ratio of the rate of NAPDH oxidation and oxygen consumption during PUFA monooxygenation. These hydroxylated metabolites, called the HETEs, were later joined by a novel class of epoxidized metabolites—the epoxyeicosatrienoic acids (EETs). The discovery of EETs was thus the first indication of CYP epoxygenase function in AA oxidative metabolism (Chacos et al., 1982; Oliw et al., 1982).

Product identification improved following developments in mass spectrometry, where authentic oxy-lipid standards were synthesized to elucidate the structure of the products (Oliw, 1994). Although it was recognized that liver microsomal activity was due to the presence of a suite of multiple CYP enzymes, it was unknown which specific CYP isoforms were responsible for epoxygenase function. As the emerging field of metabolomics elucidated novel products of CYP epoxygenase-mediated reactions, identifying the regioselectivity of endogenous epoxidation served as a metric to differentiate between epoxygenase activity of CYP isoforms (Capdevila et al., 1990a). The comparison of CYP epoxide regio- and enantio-selectivity between microsomal epoxygenases and their specific isoforms, as well as the use of antibody-based inhibition experiments, confirmed the role of CYP epoxygenases in PUFA metabolism (Figure 1B) (Capdevila and Falck, 2002; Zeldin et al., 1995).

The most well-studied epoxidation reaction is the metabolism of AA to all four cis-EET (5,6-, 8,9-, 11,12-, and 14,15-EET) regioisomers by CYPs. However, there are differences in the ratio of these regioisomers depending on the CYP isoform (Capdevila et al., 2000; Node et al., 1999; Roman, 2002; Spector, 2009b). For instance, most CYP epoxygenases generate a higher ratio of the terminal epoxide, 14,15-EET, compared to the others. This is likely due to the facilitated entry of the fatty acid tail towards the heme active site as opposed to the carboxylic acid end of the molecule, as the CYP channels must be able to direct lipophilic molecules into the deeply buried active site cavity (Johnson and Stout, 2005; Otyepka et al., 2012; Otyepka et al., 2007; Poulos, 1988; Urban et al., 2018). In mammals, the epoxidation of PUFAs to non-allylic, cis-epoxides is unique to the P450 system, as other epoxides that are formed from hydroperoxide-dependent CYP oxidation, or from lipoxygenase oxidation of AA, can form both cis- and trans- EETs (Vaz et al., 1998; Vaz et al., 1997).

(R,S) or (S,R) epoxide stereoisomers are produced depending on the CYP isoform (Blum et al., 2019; Cinelli et al., 2018). Determining the biological effects of individual EET enantiomers and regioisomers is lacking because most biological studies are done with racemic EET mixtures (Chen et al., 2014; Ding et al., 2014; Gross et al., 2008; Kiss et al., 2010; Liu et al., 2014; Sisignano et al., 2012; Spector et al., 2004; Sudhahar et al., 2010). Enantioselectivities of individual CYP epoxygenases with respect to AA metabolism are summarized in Table 2.

Interestingly, the ratio of the regioisomers produced in vivo depends on which tissues the EETs were extracted from, indicating differential CYP expression in different tissues. For example, the ratio of 14,15-EET and 11,12-EET in the heart was revealed to be approximately 34% and 37% of all total EETs, respectively. On the other hand, it has been shown that 5,6-EET is the predominant EET found within the lung (45.4%) (El-sherbeni et al. 2013). A major factor underscoring these differences in EET regioisomer bioavailability is the presence of specific CYP epoxygenases. When present in a given tissue, these epoxygenases express their signature regioselective product distributions. Therefore, knowledge of the tissue-distribution of EET regioisomers may aid in the identification of CYPs, which can then be targeted or modulated for treating pathological conditions where AA metabolism is altered.

Although the majority of characterized CYP epoxygenases belong to the CYP2J or CYP2C families, other families of CYPs have been shown to generate PUFA-epoxides at detectable levels, albeit to a lesser extent. These include the CYP2D and CYP4F classes (Guengerich et al. 2005). The CYP4 family, including CYP4A and CYP4F, mainly utilizes AA as a substrate, but some isoforms can accommodate larger PUFA substrates, such as docosahexaenoic acid (DHA) and docosapentaenoic acid (DPA). The CYP4A and CYP4F classes were originally characterized for their roles as prostaglandin and leukotriene hydroxylases. For example, CYP4F was shown to be the major LTB₄ degradative pathway (Figure 1C) and CYP4A as a key player in prostaglandin hydroxylation (Kalsotra and Strobel 2006). However, it was shown that CYP4F12 and CYP4F8 metabolized DHA in a NADPH-dependent manner to the four epoxides 19,20-epoxydocosapentaenoic acid (19,20-EDP), 16,17-EDP, 13,14-EDP, and 10,11-EDP. CYP4F8 also hydroxylates AA to 18(R)-HETE, and CYP4F12 metabolizes DPA to four epoxides, 19,20-epoxydocosatetraenoic acid (19,20-EDPA), 16,17-EDPA, 13,14-EDPA, and to a lesser extent 10,11-EDPA (Stark et al. 2005). Additionally, CYP4F2 can convert AA, EPA, DHA, and docosapentaenoic acid (DPA) to epoxy and hydroxyl metabolites (Arnold et al. 2010). CYP4F2 functions almost exclusively as a hydroxylase, indicating that the formation of PUFA epoxides is a minor product in comparison. For instance, CYP4F2 converts DHA into 22-hydroxydocosahexaenoic acid (22-HDoHE), 21-HDoHE, and 19,20-EDP.

The identification of epoxygenase function in the CYP4 family raises an important point in that the earlier declarations of “primary epoxygenases” was biased from the extent of screening certain CYPs in certain tissues from certain organisms. However, one should note that the epoxygenase activity of CYP4Fs is a minor reaction and is also selective for certain PUFAs.

With the advent of recombinant expression of membrane-bound human CYPs, researchers became interested in in vitro AA metabolism by human enzymes. Earlier research had established the CYP 2C family as the primary epoxygenases in rat and rabbit livers, which prompted the purification of CYPs 2C8 and 2C9 (same as CYP2C10) from human livers for further studies (Daikh et al., 1994a). Both CYP2C8 and CYP2C9 favor the production of 14,15-EET compared to the non-terminal epoxides (Chang and Waxman 1995). Human liver microsomal CYP2C8 was found to generate only 14,15- and 11,12-EETs in a ratio of 5:4, while CYP2C9 made 14,15-, 11,12-, and 8,9-EETs in ratios of 2.3:1.0:0.5 (Daikh et al., 1994a). Zeldin and coworkers heterologously expressed and purified human CYP 2C8 to determine the absolute configuration of its resulting EET metabolites(Zeldin et al., 1995). The metabolic trends previously found for CYP2C8 liver microsomes were in agreement with Zeldin group’s data, though it was found that the (R,S) enantiomer for both 11,12- and 14,15-EETs were preferred. CYP2C8 also selectively generates only the (R,S) enantiomers of 14,15-EET and 11,12-EET. Similarly, CYP2C9 exhibits a catalytic preference towards the formation of the 14(R),15(S)-EET enantiomer and it produces a higher ratio of 11(S),12(R)-EET compared to the 11(R),12(S)-EET regioisomer (Kalsotra and Strobel 2006).

CYP2J2 is another major EET-producing CYP2 enzyme. It generates all four EET regioisomers. It displays unique enantiomeric selectivity with the terminal 14,15-EET and 11,12-EET. CYP2J2 can only produce the 14(R),15(S)-EET enantiomer, but it can produce an approximately equal ratio of the 11,12-EET enantiomers (Wu et al., 1996; Zeldin et al., 1996). Table 1 lists the differences in epoxygenase activity and substrate preference of different CYP epoxygenases.

As the lipid-metabolizing CYPs from CYP2 and CYP4 classes are also involved in drug metabolism, it is difficult to identify structural features for strictly defining CYP epoxygenases. While CYP2 enzymes are mainly classified as primary epoxygenases, they are still mainly characterized by their ability to perform monooxygenation on a large variety of xenobiotics (Lewis et al., 1999). Importantly, the definition of a CYP epoxygenase is not limited to any specific substrate class, per se, but by the tissue-specific expression of CYP epoxygenases, along with the disease-specific physiological roles associated with PUFA epoxide metabolites. This refined definition of a CYP epoxygenase will aid in studying the pharmacology of these epoxides throughout different regions of the body (i.e. the cardiovascular system, brain, intestines, etc.). Moving forward, this review will contextualize PUFA-epoxides with respect to the current body of work on the expression, physiological role, biochemical function, and structures associated with modeling CYP2J2.

4. Expression of human CYP2J2 epoxygenase, known polymorphisms, and physiological function

4.1. Expression of CYP2J2 in extrahepatic tissues

The 2C and 2J families contain enzymes which have been characterized as primary CYP epoxygenases that produce EETs, PUFA epoxides, and epoxy-eCBs (McDougle et al., 2017; Xu et al., 2011a; Zelasko et al., 2015). In humans, the most well-studied of these enzymes include CYP2J2 (the only 2J enzyme in humans), CYP2C8, CYP2C9, and to a lesser extent CYP2C19 (Xu et al., 2011a). The 2C enzymes, much like their other xenobiotic counterparts, have the highest levels of expression in the liver (Galetin and Houston, 2006; Graves et al., 2017).

CYP2J2 was first discovered in the liver. However, it is considered an extrahepatic P450 with predominant expression in the heart (Delozier et al., 2007a; Evangelista et al., 2013; Node et al., 1999; Wu et al., 1996). Specifically, CYP2J2 is highly expressed in the aortic endothelium and cardiomyocytes, and more broadly in the right ventricle (Michaud et al., 2010). CYP2J2 is also expressed in the brain (Dutheil et al., 2009b), and the gastrointestinal (GI) tract (Matsumoto et al., 2002; Paine et al., 2006; Walker et al., 2016; Zeldin et al., 1997).

Genomic analysis by the GTEx database corroborates the tissue-specific expression of CYP2J2 (Table S1). In order of abundance, primary expression of CYP2J2 was 23.9 RPKM in heart muscle (n=412), followed by the liver (19.19 RPKM, n=119), small intestine (18.5 RPKM, n=88), and brain (Consortium, 2013) (Table S1). Importantly, the GTEx database sampled from a large number of patient samples. Other databases, such as the FANTOM5 project database, suggested primary CYP2J2 expression in the small intestine (275.5 tags/million, n-1), followed by liver (223.6 tags/million, n=1), brain, and then the heart (48.4 tags/million, n=1) (Lizio et al., 2015) (Table S2). Collectively, human genome analysis coupled with expression analysis by various research groups worldwide indicate CYP2J2 is primarily expressed in the heart, suggesting cardiovascular implications.

CYP2J2 mRNA levels were found to be almost 20,000-fold higher than CYP2C8 or CYP2C9 in the myocardium. Furthermore, CYP2J2 is expressed in the vascular endothelium (Chaudhary et al., 2009; DeLozier et al., 2007b; Michaud et al., 2010). Importantly, it was found that the EET profile observed in the heart matches the EET profile produced by CYP2J2, thus demonstrating CYP2J2 is likely the primary epoxygenase in the heart (Wu et al., 1996). This type of analysis – coupling specific CYP expression and EET regioisomer profiles – can be extended to the study of any CYP epoxygenase to identify the primary metabolizers within any given tissue. It is important to consider that regardless of the extent of expression in different tissues, a CYP epoxygenase will contribute to the localized production of EETs, and therefore the resulting physiology.

CYP2J2 is found in several regions of the brain including the hippocampus, cerebellum, and astrocytes (Dutheil et al., 2009a; Ferguson and Tyndale, 2011; Nishimura et al., 2003). Human CYP2J2 activity is orthologously compared to the activity of a variety of CYP2J enzymes in rodents, as rats and mice have many different CYP2Js whereas humans only have CYP2J2. In rodents, these CYP2J enzymes have been shown to also be expressed in various rat brain-specific regions and cell types (Graves et al., 2013; Liu et al., 2017a; Qu et al., 2001). Similar to the heart, CYP2J in the brain has been shown to provide a protective role against various brain diseases, such as Parkinson’s (Li et al., 2018a). Despite its dominance in the heart, the biochemical functions of human CYP2J2 persist in other tissues in which it is expressed, warranting further studies on the physiological relevance of CYP2J2 function in other tissues.

4.2. CYP2J2 homologs in other vertebrates and transgenic models

To assess the physiological function of CYP2J2, it is important to also understand the cross-species differences between phylogenetically-related CYP2J enzymes. Vertebrates with the closest homology sequence to human CYP2J2 include rhesus macaques or chimpanzees (Bult et al., 2015; Finger et al., 2017; Smith et al., 2018; Solanki et al., 2018a). CYP2J2 isoforms between humans and monkeys are reportedly 95% similar (Uno et al., 2007), an observation supported by genome sequencing databases (Bult et al., 2015; Finger et al., 2017; Smith et al., 2018).

Several CYP2J2 homologs in mice and rat rodent models facilitate studies exploring the physiological roles of PUFA-epoxides. For instance, there are seven genes within the mouse CYP2J locus, including CYP2J5, CYP2J6, CYP2J8, CYP2J9, CYP2J11, CYP2J12, and CYP2J13 (Graves et al., 2013). CYP2J6 is the closest homolog to CYP2J2 with 84% homology at the primary amino acid sequence level (Solanki et al., 2018a). The Rat Genome Database indicates that rat homologs to human CYP2J2 include CYP2J3, CYP2J4, CYP2J10, CYP2J13, and CYP2J16 (Shimoyama et al., 2015), where CYP2J4 is closest in homology sequence at 76% similarity (Solanki et al., 2018a). To better understand the rodent genome and cell-specific expression of genes, a recent study conducted single cell transcriptomics on 20 tissues from eight C57BL/6 mice (Consortium et al., 2018). Exploring the cell-specific expression dataset may provide insight into cell-specific expression of human CYPs. In C57BL/6 mice, CYP2J6 was highly expressed in glial cells, such as oligodendrocyte progenitor cells, oligodendrocytes, and astrocytes. As the cross-talk of glia plays an important role in neuroinflammatory and neurodegenerative diseases (Domingues et al., 2016), this data suggests that PUFA epoxides in astrocytes and oligodendrocytes may act to encourage microglial polarization towards an anti-neuroinflammatory microglial phenotype.

4.3. Physiological role of CYP2J2

CYP2J2 transgenic models provide key insights into CYP2J2 activity in diseased states (Table S3). For example, there exist a few studies demonstrating that mice expressing CYP2J2 are more resistant to experimentally-induced diseases (Azevedo et al., 2016; Chen et al., 2015; Chen et al., 2014; He et al., 2015; Ma et al., 2013; Westphal et al., 2013; Zhao et al., 2012; Zheng et al., 2010). The Zeldin lab was the first to generate a transgenic CYP2J2 (CYP2J2-Tr) animal model (Seubert et al., 2004; Xiao et al., 2004). To do this, CYP2J2 cDNA was cloned into pBS-αMHC-hGH, a vector containing the αMHC promotor for cardiomyocyte-specific expression of the transgene. In addition, the human growth hormone (hGH)/polyA was incorporated into the vector to enhance the transgene mRNA stability. This transgene was then injected into the pronuclei of a C57BL/6 mouse embryo which was implanted into pseudo-pregnant mice (Seubert et al., 2004; Xiao et al., 2004). The resulting transgenic mice expressing CYP2J2 became a powerful tool that has been utilized since its inception.

Studies using these transgenic mice were the first to verify the role of epoxygenases in vivo. For example, CYP2J2-Tr mice displayed a greater postischemic recovery of left ventricular function as compared to wild type mice (Seubert et al., 2004). This confirmed the previous finding that CYP2J2 is essential in generating cardioprotective EETs. Another study examined the effects of EETs on cardiomyocyte L-type Ca²⁺ channels, which are a type of channel involved in maintaining proper excitation-contraction cycles of the heart (Bodi et al., 2005).

For instance, one study reported that EETs increase I_Ca in rat cardiomyocytes (Xiao et al., 1998), while another reported that EETs have inhibitory properties on porcine cardiac L-type Ca²⁺ channels (Chen et al., 1999). Utilizing CYP2J2-Tr mice, it was demonstrated that CYP2J2-mediated generation of EET metabolites had a stimulatory effect on I_Ca in cardiomyocytes, and thus in modulating basal cardiac L-type Ca²⁺ channel activity (Xiao et al., 2004). Collectively, CYP2J2-Tr mice proved to be a powerful resource in understanding the roles of P450 epoxygenases in vivo.

Given its expression in the heart, CYP2J2 plays an important role in cardiovascular diseases. For instance, cardiac remodeling is a physiological process implicated in cardiovascular diseases, resulting in numerous complications including cardiac dysfunction and arrythmia (Azevedo et al., 2016). In transgenic mice overexpressing human cardiomyocyte-specific CYP2J2, arrhythmia inducibility was reduced compared to WT mice (Westphal et al., 2013). An additional study demonstrated that CYP2J2 was able to inhibit cardiac remodeling altogether, a process that inevitably leads to arrhythmia (He et al., 2015). Other studies demonstrated that overexpression of CYP2J2 alleviated cardiac injury induced by TNFα (Zhao et al., 2012), a common cytokine produced and involved in systemic inflammation (Kalliolias and Ivashkiv, 2016). Overexpression of CYP2J2 has also been shown to reduce cardiotoxicity induced by chemotherapeutics, such as doxorubicin (DOX). As EETs prevent the myocardium from ischemia-reperfusion injury, transgenic mice overexpressing CYP2J2 specific to cardiomyocytes were generated to test whether CYP2J2 was able to reduce DOX-induced cardiotoxicity (Zhang et al., 2009). Convincingly, transgenic mice overexpressing CYP2J2 exhibited less cardiomyocyte apoptosis and improved echocardiography readings compared to that of wild type mice hearts (Zhang et al., 2009). CYP2J2 overexpression also improves cardiopulmonary function and protects against ischemia (Chen et al., 2014; Zheng et al., 2010). Further involvement of CYP2J2 in cardiovascular diseases has been reviewed extensively (Aliwarga et al., 2018) (Solanki et al., 2018a).

Within the brain, CYP2J2 is reported to be the second highest expressed CYP, implicating potential roles in mediating neurodegenerative and cerebrovascular diseases. Neurodegenerative diseases such as Alzheimer’s diseases are associated with the CYP2J2 polymorphisms found in the Chinese Han population (Yan et al., 2015). In vivo, CYP2J enzymes in rodents protected against Parkinson’s disease models (Li et al., 2018b). However, studies of brain diseases utilizing CYP2J2 transgenic mice with overexpression specific to the brain are currently unavailable. Nevertheless, endothelial overexpression of CYP2J2 protected against experimental cerebral ischemia in mice (Li et al., 2012). Coupling these works with the importance of tissue-specific expression will hopefully usher in new work elucidating the role of CYP2J2 epoxygenase function in combatting brain diseases.

In addition to cardiovascular and cerebrovascular properties (Jia et al., 2016; Zhao et al., 2018), CYP2J2 is implicated in chronic diseases, such as diabetes. While PUFA epoxides are effective against diabetic diseases, there are only a few studies directly examining the effects of CYP2J2 in diabetes (Ma et al., 2013), or in high-fat diet-induced non-alcoholic fatty liver disease (NAFLD) (Chen et al., 2015). Given the role of CYP epoxygenases and their EET products in maintaining homeostasis, it is likely that epoxygenases like CYP2J2 participate in the physiological responses to disease in other tissues where the basal expression is low.

Despite tissue-specific expression differences, CYP2J2 and other epoxygenases are upregulated in cancer, promoting angiogenesis and tumorigenesis and can therefore be considered as cancer-promoting. In particular, CYP2J2 expression was elevated as measured by mRNA and protein levels of human-derived cancer cell lines and human cancer tissues as compared to healthy human cell lines or tissues (Jiang et al., 2005). Follow-up studies demonstrated that CYP2J2 overexpression promoted tumor metastasis in four different human cancer cell lines via production of EETs (Jiang et al., 2007). Interestingly, miRNAs such as let-7, a common dysregulated in human cancers (Balzeau et al., 2017; Boyerinas et al., 2010), increased the expression of CYP2J2 in human cancers (Chen et al., 2012). Although used to suppress tumorigenesis (Mizuno et al., 2018), promotion of let-7b miRNA therapy similarly reduced cancer phenotypes as well as CYP2J2 activity post-translationally regulating CYP2J2 expression (Chen et al., 2012), suggesting there exists greater complexity in the regulatory networks involving CYP2J2 and its activity in cancer. As CYP epoxygenases produce multiple EET regioisomers, it is likely that the regioisomers display different physiological behaviors based upon their subsequent anabolic and catabolic pathways. Rand et al. found that 8,9-, 5,6-, and 11,12-EET were substrates for both COX-1 and COX-2, in decreasing order, resulting in products that can be further metabolized by sEH, ultimately forming an angiogenic lipid species known as ct-8,9-E-11-HET (Rand et al., 2017a). While a single downstream omega-6 metabolite was found to be angiogenic following CYP bioactivation, the formation of EDP omega-3 epoxides has shown to display anti-cancer properties (Kim et al., 2016; Zhang et al., 2013) suggesting that fatty acid structure may lead to distinct properties. It has been shown that the tumor-promoting properties of EETs are in part mediated by their further metabolism by COX enzymes (Rand et al., 2017b).

In summary, while CYP2J2 and PUFA epoxide-generating epoxygenases may be detrimental in cancer due to their angiogenic potential (Wang and Dubois, 2012), overwhelming evidence suggests that PUFA epoxides are important in modulating cardiovascular and inflammatory diseases (Aliwarga et al., 2018; Thomson et al., 2012; Yang et al., 2015; Zeldin, 2001). However, it is important to note here that the functional interactions between a single epoxygenase and a substrate class do not necessarily follow clear-cut rules leading to a specified physiological implication. This is perhaps one of the greatest challenges faced within CYP enzymology, which is confounded even further when considering how CYP epoxygenase structure-function results in the accommodation of many different substrates, ultimately to derive similar looking molecules with sometimes contrasting bioactivity. Because of their lipophilic nature, endogenous CYP-epoxygenase products either immediately interact with receptors given their CYP-derived form or are further metabolized before interacting with their target receptor as downstream metabolites (i.e. interactions within the endocannabinoid system). Accordingly, the presence and activity of a specific CYP epoxygenase can provide insight into their physiological functions. But to measure how epoxygenase function contributes to or detriments from a given physiological state, it is necessary to contextualize epoxygenase presence and activity with that of other enzymatic and receptor machineries associated with the assumed bioactivity. This rationale also applies to functions specified by different epoxygenase polymorphisms. More information on CYP2J2 expression and influence on physiological states has been reviewed previously (Xu et al., 2013).

4.4. Known polymorphisms of CYP2J2

Although CYP2J2 is implicated in diabetes (Hashizume et al., 2002; Li et al., 2015b; Pucci et al., 2003) and inflammatory gastrointestinal diseases (Hashizume et al., 2002; Walker et al., 2016), most studies have focused on its cardioprotective role, given its expression in heart tissues and early publication history of studies targeting CYP2J2 cardiovascular function. Accordingly, most of the known CYP2J2 polymorphisms are associated with cardiovascular diseases (Borgel et al., 2008; Dreisbach et al., 2005; Fava et al., 2010; King et al., 2005). An example is the CYP2J2*7 polymorphism found in Chinese populations. While this polymorphism associated with cerebrovascular physiology through increased risk of ischemic stroke, it is inherently cardiovascular in nature (Wang et al., 2017). The CYP2J2*7 polymorphism has a “T” substituted with “G” at the 50^th position in the promoter region (King et al., 2002). Additionally, distinct ethnic populations display differences in allele frequencies (Berlin et al., 2011). This single nucleotide polymorphism (SNP) is associated with a wide array of other cardiovascular diseases including coronary artery disease (King et al., 2002; Lee et al., 2007; Marciante et al., 2008; Xu et al., 2011b), hypertension (Wu et al., 2007), and myocardial infarction (Li et al., 2015a; Wu et al., 2007; Xu et al., 2011b). In addition, stroke patients were observed to have a higher frequency of the CYP2J2*7 allele compared to a healthy human control group (Li et al., 2015a). As it pertains to CYP2J2*7, evidence points to an imbalance of PUFA epoxides generation for its physiological malfunction. While only the CYP2J2*7 allele is associated with cardiovascular diseases, other variations impair metabolic function. For example, CYP2J2 *2, *3, *4 and *6 variations all exhibit reduced AA metabolism (Berlin et al., 2011; King et al., 2002). Additionally, *8 and *10 had complete loss (Lee et al., 2005), or reduced protein function (Gaedigk et al., 2006). Reduced protein function or ability to metabolize arachidonic acid suggests that the body would have lower circulating PUFA epoxides, and therefore a decrease in the amount of anti-inflammatory and cardioprotective mediators. Given the importance of substrate turnover in generating these metabolites, much of the availability and bioactivity of PUFA epoxides are defined by epoxygenase kinetics. And while we discuss a few instances in which biochemical function is implicated in the physiological roles of specific CYP2J2 polymorphisms, we refer the reader to another review in which the functional consequences of CYP2J2 polymorphisms are discussed in greater detail (Murray, 2016).

5. Kinetics of metabolism of PUFAs and their endocannabinoid derivatives by CYP2J2 epoxygenase

CYP2J2 was discovered in 1996 (Wu et al., 1996) and much of its initial characterization was based on exploring its role in human physiology. A scalable method to purify membrane-bound CYP2J2 was therefore required to examine the kinetics of CYP2J2-mediated PUFA and eCB metabolism. CYP2J2 was first recombinantly expressed using the baculovirus system (Wu et al., 1996). The expression of recombinant protein under these conditions was 100–150 nmol/liter of infected Sf9 cells. Following protein purification however, the final yields were approximately 10% of that from cell harvesting (Wu et al., 1996). In 2002, five variants of CYP2J2, including R158C, I192N, D342N, N404Y, and T143A were identified that were recombinantly expressed with the Sf9/baculovirus system. The yields were low, with P450 expression levels ranging from 3 to 10 nmol/liter of Sf9 cells (King et al., 2002). Due to the need for large amounts of protein for structure-function studies, CYP2J2 was expressed in bacterial expression systems. Several constructs of N terminus-deleted mutants of CYP2J2 with silent mutations were expressed in bacterial expression systems (McDougle et al., 2013). All constructs were co-transformed into DH5α E. coli cells with the pTGro7 plasmid encoding for the GroEL-GroES chaperonin system. The protein yields of the N-terminally modified constructs ranged up to 200 nmol/liter protein, a yield that would allow for detailed structure-function studies. Using the CYP2J2-CPR-Nanodisc system, a model lipid bilayer that mimics the endogenous membrane environment of the protein, rates of ebastine (EBS) metabolism were confirmed that were similar to that of previously reported rates. Collectively, this was the first report of high yielding expression of functional full-length CYP2J2 protein (McDougle et al., 2013).

The large-scale expression of CYP2J2 subsequently facilitated several kinetic studies on direct metabolism of PUFAs and drugs by CYP2J2, which are outlined below.

5.1. CYP2J2-mediated metabolism of PUFA and omega-6 endocannabinoids -asymmetric binding and metabolism

CYP2J2 metabolizes arachidonic acid (AA) to all four EET regioisomers, with the preferred product being the terminal epoxide 14,15-EET. On the contrary, 2-arachidonyl glycerol (2-AG) (an AA-derived eCB) is epoxidized only at the 14,15- and 11,12- positions, presumably because it has a larger head group at the carboxy terminus of the AA. CYP2J2 metabolizes anandamide (AEA) with a V_max of 53.3 ± 2.9 pmol product∙min⁻¹∙nmol⁻¹ P450 (McDougle et al., 2014), AA with a V_max of 189 ± 9 pmol product∙min⁻¹∙nmol⁻¹ P450, linoleic acid (LA) with a V_max of 146 ± 17 pmol product∙min⁻¹∙nmol⁻¹ P450, EPA with a V_max of 259 ± 77 pmol product∙min⁻¹∙nmol⁻¹ P450, and DHA with a V_max of 254 ± 11 pmol product∙min⁻¹∙nmol⁻¹ P450 (Arnold et al., 2016). These values are consistent with other reported rates of metabolism of PUFAs by CYPs, which typically range between 100 to 1000 pmol product∙min⁻¹∙nmol⁻¹ P450 (Table 1). Previously, a similar rate of 200 pmol product∙min⁻¹∙nmol⁻¹ P450 for eicosapentaenoic acid (EPA) metabolism was reported using rat CYP2J3, which is 70% similar in amino acid sequence to CYP2J2 (Wu et al., 1997). Although these fatty acid substrates can assume multiple conformations in the active site due to flexibility in the alkyl chain, changes in the structure of the lipid head group can influence overall substrate orientation and atomic distance from the heme, thereby influencing the different product distributions seen for AA, DHA, EPA, 2-AG and AEA. Molecular Operating Environment (MOE) modelling done between CYP2J2 and 2-AG had revealed that greater polar interactions between the glycerol moiety and residues within CYP2J2 restricted orientations to favor epoxidation at selected positions (McDougle et al., 2014). Meanwhile, AA and AEA interactions with a single polar residue, Q228, enabled more freedom for other sites of epoxidation. With an increase in the number of studies that explored the metabolism of different fatty acids and their eCB derivatives, it was revealed that changing head groups of the carboxylic acid changes the rate of metabolism of the lipids (Arnold et al., 2016; 2017). Previous data suggested that a three-carbon spacer length between the carboxylic head group and the neighboring alkene moiety caused a relatively fixed orientation to favor quicker terminal epoxidation due to polar contacts, ultimately influencing regioselectivity of the epoxidation. Specifically, a triad of the amino acids T318, R321, and S493 directly contacted and trapped the PUFA carboxylic end to a certain extent of motions during the simulations involving CYP2J2. Regioselectivity is also determined by the steric constraints imposed by the presence of multiple methylene bridges along a PUFA alkyl chain. From these molecular dynamics (MD) studies, some conformational freedom provided to the lipid substrate enables the generation of multiple epoxides, while enough rigidity within the protein structure biases regioselectivity. From the modeling and experimental studies, the terminal epoxide regioisomer of both DHA and EPA were predicted to be formed preferentially by CYP2J2. However, future studies are required to confirm the regiospecificy of the different CYP epoxygenase with respect to DHA and EPA epoxidation.

We additionally showed that CYP2J2 was able to differentiate reactive moieties within the active site (i.e. different olefinic bonds and their methylene bridges) so that individual oxygenations can be selected. The rates of metabolism of PUFAs and regioisomers formed by human CYP epoxygenases are reported in Table 1. Similar data for non-human CYP epoxygenases can be found in Supplemental Information (Table S4).

5.2. CYP2J2-mediated metabolism of omega-6 and omega-3 endocannabinoids

The functions of PUFAs are partially mediated by their non-oxidative conversion to eCBs, which act on the same cannabinoid receptors as the principle components of Cannabis sativa. Functionalization of the PUFA headgroups leads to dramatic changes in biological activities, rate of metabolism, and receptor selectivity (Devane et al., 1992; Leishman and Bradshaw, 2015; Mechoulam et al., 1995; Sugiura et al., 1995). Anandamide (AEA) consists of an AA scaffold conjugated to an N-linked ethanolamine, whereas 2-AG is AA conjugated to glycerol at the secondary alcohol position. Ethanolamide and glycerol derivatives of other PUFAs such as EPA and DHA have been discovered (Devane et al., 1992; Mechoulam et al., 1995; Sugiura et al., 1995) (Balvers et al., 2010; Berger et al., 2001). Further eCB-like modifications include amide-linked derivations with amino acids and taurine (Leishman and Bradshaw, 2015; Saghatelian et al., 2006). Together, eCBs mediate several divergent physiological responses primarily through cannabinoid receptors 1 and 2 (CB1 and CB2). The non-selective transient receptor potential (TRP) channels, a class of non-selective cation channels, have also been shown to be activated by eCBs, notably TRP vanilloid 1 (TRPV1) (De Petrocellis and Di Marzo, 2015; Maione et al., 2007; Rimmerman et al., 2009; Ryskamp et al., 2014; Toth et al., 2009).

CYP2J2, CYP2D6, and CYP4X1, are capable of metabolizing eCBs to form eCB epoxides, which exhibit anti-inflammatory and vasodilatory properties partly through interactions with cannabinoid receptors, as well as with an unknown PUFA epoxide receptor (Maccarrone, 2017) (McDougle et al., 2017; Roy et al., 2018; Snider et al., 2010). Furthermore, 5,6-EET-EA (AEA-epoxide) and many of the ω−3 eCB epoxides show a greater selectivity for CB2 binding, suggesting that the epoxidation of eCBs switches CB1/CB2 preference (McDougle et al., 2017; Roy et al., 2018; Snider et al., 2009). Thus, the metabolism of eCBs and eCB-like molecules by several CYPs may be important in maintaining homeostasis during inflammation-based pathologies.

It is important to note that the bioactivities of the epoxide metabolites can be distinct from those of their parent molecule substrates metabolized by CYP epoxygenases. Combining these relative pools of unreacted and bioactivated parent compound scaffolds in specific tissues can impose variable consequences on the handling of different inflammatory states. This systems-level metabolic equilibrium becomes further complicated when inhibitors are introduced that modulate epoxygenase function through singular or cooperative interactions.

6. Inhibition of CYP2J2 by drugs and endogenous lipids

Given that CYPs are generally capable of binding multiple substrates due to large active sites, and that lipid conformational flexibility enables active CYP conformations, it is important to understand the molecular mechanisms of lipid-drug interactions through inhibition studies. CYPs can be inhibited through the formation of reversible, pseudo-irreversible, or irreversible complexes with an inhibitor. Reversible inhibitors interfere with CYP function prior to electron transfer through non-covalent interactions with certain residues and regions of the protein, usually by binding into the active site. Pseudo-irreversible inhibitors form a coordinate bond to the heme, the extent of inhibition being dependent upon the affinity of the lone pair of electrons on a heteroatom in the inhibitor scaffold. Alternatively, an inhibitor could covalently modify active site residues within the CYP active site to induce a catalytically inactive protein (irreversible inhibition). In some cases, the inhibitors can prevent the oxy-heme intermediates from proceeding productively to product generation; they essentially “trap” the enzyme in an unproductive state upon reaction (Brueggemeier, 2002; Correia and Ortiz de Montellano, 2005; Fontana et al., 2005; Grime et al., 2009; Hollenberg, 2002; 2009; Hollenberg et al., 2008; Kalgutkar et al., 2007; Kent et al., 2001; Obach et al., 2007; Vanden Bossche, 1992). Such mechanism-based inhibitors are difficult to design but can be very specific and therefore can facilitate physiological studies related to the enzyme. One such mechanism-based inhibitor is reported for CYP19A1 (Ghosh et al., 2009; Hong et al., 2007; Maurelli et al., 2011). Other forms of inhibition result from binding to lipophilic regions of the protein, which impairs the CYP’s ability to sample conformations necessary for catalysis. Regardless of the overall mode of inhibition, inhibitors compete with CYP substrates for access to the heme. Further information regarding CYP inhibition is reviewed in the book chapter by Correia and Hollenberg (Correia and Hollenberg, 2015).

6.1. Specific and mechanism-based inhibitors of CYP2J2

CYP epoxygenases metabolize both lipids and drugs and are able to recognize a wide variety of ligands, including those that are either polycyclic, aliphatic, or both (Correia and Hollenberg, 2015; Lewis et al., 1999; Smith et al., 1997a; b). To this point, it has been postulated that some drug substrates of CYP2J2 interfere with its epoxygenase activity and contribute to cardiotoxicity (Solanki et al., 2018b) (Arnold and Das, 2018). Therefore, it is of pharmacological interest to determine which drugs’ off-target effects involve CYP2J2 inhibition, which may contribute to cardiotoxicity. Additionally, CYP2J2 expression is upregulated in many cancers and EETs are implicated in tumor progression, and so the inhibition of CYP2J2 may be important in combating cancer (Karkhanis et al., 2017). This subsection provides examples of various drugs that target CYP2J2 specifically or with unique implications. Figure 2 shows the structures of some of these drugs and how they inhibit CYP2J2. The rest of Section 6 will outline specific interactions of drugs regarding CYP2J2’s endogenous metabolisms.

Figure 2. Xenobiotic ligands of CYP2J2. Structures of key ligands mentioned in the text are given.

Each is categorized based on their mode of CYP2J2 inhibition (mostly of astemizole metabolism). Xenobiotics that are known to have cardiovascular effects are indicated with a yellow square and those that are cardiotoxic are indicated with a red square. Abbreviations for the type of drug are given in parentheses beside each name and are as follows: aAR, antiarrhythmic; aCAN, anticancer; aHIS, antihistamine; aHYP, antihypertensive; aINF, anti-inflammatory; aRV, antiretroviral; CCB, calcium channel blocker; HOR, hormonal; PSY, psychoactive. Compounds indicated with asterisks are mechanism-based inhibitors.

Commercially available xenobiotic substrates were determined as CYP2J2 substrates using either human liver microsomes or screens against CYP2J2 activity. Preliminary work with CYP2J2 identified several inhibitors/substrates, many of which are antihypertensive, antiarrhythmic, or cardiotoxic drugs. Many of these cardiotoxic drugs are Class II antihistamines. Some of these inhibitors/substrates include albendazole, amiodarone (Lee et al., 2012; Lee et al., 2010), apixaban (Wang et al., 2010), astemizole (Lee et al., 2010; Matsumoto et al., 2002), cyclosporine (Lee et al., 2010), danazol (Lee et al., 2010; Lee et al., 2015), ebastine (EBS) (Evangelista et al., 2013; Hashizume et al., 2002; Liu et al., 2006), fenbendazole (Wu et al., 2013), mesoridazine (Lee et al., 2010), nabumetone (Lee et al., 2010), ritonavir (Kaspera et al., 2014), terfenadine (Chen et al., 2009a; Evangelista et al., 2013; Lafite et al., 2006; Lafite et al., 2007c; Lee et al., 2010; Matsumoto and Yamazoe, 2001), tamoxifen (Lee et al., 2010), thioridazine (Lee et al., 2010), and vorapaxar (Ghosal et al., 2011).

In 2010, Totah’s group led a large 139 compound screen to identify novel substrates of CYP2J2, where specific inhibitors for predominant hepatic CYPs – 1A2, 2C8, 2C9, 2C19, 2D6, and 3A4, all of which also exhibit epoxygenase function – were implemented to identify 2J2-specific activity within supersomes (Lee et al., 2010). The ten compounds identified from the 139 compounds screened were not specific to CYP2J2. The catalytic turnover of these ten molecules was less than that for previously identified CYP2J2 substrates terfenadine and astemizole. Furthermore, besides astemizole and danazol forming five and six metabolites, respectively, the remaining eight selected substrates were metabolized to form primarily one or two products when metabolized by CYP2J2. This initial screen provided the postulate that CYP2J2 possesses a large active site to accommodate larger, bulkier substrates.

In another study, terfenadine hydroxylation and astemizole O-demethylation, both of which are primarily performed by CYP2J2, were used as probe reactions to screen for CYP2J2 inhibition (Lee et al., 2012). In the screen, 42 compounds inhibited terfenadine hydroxylation by 50% or more, whereas eight compounds reduced CYP2J2 activity by 90% or more. Of the compounds that inhibited terfenadine hydroxylation by 50% or more, 24 of them overlapped with those that inhibited astemizole O-demethylation by 50% or more. The candidate compounds inhibiting astemizole metabolism by greater than or equal to 90% were danazol, ketoconazole, loratadine, miconazole, and nicardipine. To further elucidate the specificity of the observed inhibition, the top 20 inhibitors from the previous screen were further screened using human liver microsomes metabolizing terfenadine and astemizole. Danazol demonstrated the greatest selectivity against CYP2J2-mediated metabolism. Later, it was shown that danazol could be a potential index inhibitor of CYP2J2, as it inhibited several CYP2J2 based reactions with sub-micromolar IC₅₀ values, making it a substrate-independent inhibitor (Lee et al., 2015). While this work was critical for developing tools to study CYP2J2, for clinical work, it underscores the difficulty in extending CYP epoxygenase research beyond specific case studies as danazol also inhibited CYP2C9 and CYP2D6, which also have epoxygenase function (Lee et al., 2012) (Ren et al., 2013).

A more recent study suggested some structure-activity relationships among antihypertensive drugs to further characterize the requirements for inhibition of CYP2J2 (Ikemura et al., 2019). In a luciferin-2J2/4F12 O-dealkylase activity assay, Ikemura et al. evaluated the inhibitory capacities of drugs amlodipine, azelnidipine, barnidipine, benidipine, cilnidipine, efonidipine, felodipine, manidipine, nicardipine, nifedipine, nilvadipine, nisoldipine, nitrendipine, telmisartan, and delapril.

While the rationale for screening these drugs against CYP2J2 was to suggest that CYP2J2-mediated interactions with antihypertensive drugs may explain some of the side effects and bioavailability of these therapeutics, the study’s choice of substrates recapitulated the major findings of the CYP2J2 substrate/inhibitor screenings from 2010. Each of the molecules were polycyclic, containing two to six rings. Those molecules which acted as competitive inhibitors all possessed bulky and largely asymmetric scaffolds, allowing them to take up more space within CYP2J2’s larger active site. With largely asymmetric structures containing a long linker to separate a bulky substitution from their tetrasubstituted dihydropyridine core, these competitive inhibitors share a trend of having dihydropyridine moieties with complex substitutions. Meanwhile, the drugs considered to be mixed inhibitors shared a trend of having mostly dimethylated planar cores, para substituted benzenes containing electron-withdrawing groups, and more or less symmetric non-bulky meta substitutions. With less bulk and more freely accessible lone pairs, these mixed inhibitors have more opportunities to bind to various parts of CYP2J2 to induce mixed inhibition.

While structure-activity trends from previous studies suggest an importance of bulk and relative symmetry on how a ligand may interact with CYP2J2, they do not necessarily provide strict guidelines beyond examples of strongly CYP2J2-specific drugs, nor do they suggest exact physiological consequences which may translate following CYP2J2 inhibition. Terfenadine/tefenadone and derivatives thereof demonstrate remarkable specificity in inhibiting CYP2J2 (Lafite et al., 2007a; Lee et al., 2018) (Chen et al., 2009b; Lafite et al., 2007c). Importantly, terfenadine is a Class II antihistamine that is cardiotoxic, but another Class II congener, EBS, is not. Many of these terfenadone derivatives show greater inhibition of CYP2J2 even when compared to CYP2J2’s closest epoxygenase relatives, CYP2C8 and CYP2C9. These derivatives also show many modes of reversible and irreversible inhibition. The ketone moiety of terfenadone is key to binding to CYP2J2 and positioning the drug near the heme (Lafite et al., 2007a) (Li et al., 2008) while the tail is important for determining reversible or mechanism-based inhibition (Lafite et al., 2007b). Another CYP2J2-specfic inhibitor, LKY-047, was developed as a competitive and an uncompetitive inhibitor of CYP2J2. It inhibited CYP2J2-mediated metabolisms with IC50 values less than 3 μM, whereas the IC50 of all other tested CYPs was greater than 50 μM (Phuc et al., 2017). More CYP2J2-preferring inhibitors appear to demonstrate multiple modes of inhibition when maintaining specificity.

Some mechanism-based inhibitors of CYP2J2 have also been found. This class of inhibitors inactivate CTP2J2 irreversibly and would thus have long-lasting implications. These contain a highly reactive functional group, such as an alkynyl or quinone moiety. 17α-Ethynylestradiol, a component of oral contraceptives, was found to inactivate CYP2J2 by forming two reactive intermediates. During the oxidation of the ethynyl group, mass spectrometry analysis suggested that the oxidation of the terminal ethynyl carbon forms adducts with the CYP2J2 protein and oxidation of the penultimate carbon forms adducts with the heme (Lin et al., 2018). Dronedarone was developed as an antiarrhythmic but was discontinued due to an increased risk of heart failure. Like its congener, amiodarone, dronedarone is metabolized by CYP2J2. Amiodarone showed noncompetitive inhibition while dronedarone showed mixed inhibition. Both amiodarone and dronedarone were also mechanism-based inhibitors, with dronedarone showing the greatest potency. This was thought to have arisen from the formation of a reactive quinone-oxime (Figure 2) (Chan et al., 2016; Karkhanis et al., 2016). Lastly, the calcium-channel blocker azelnidipine was identified as a CYP2J2 mechanism-based inactivator. Several factors required for inactivation suggested that inhibition of CYP2J2 occurs during metabolism of this drug. More specifically, after oxidation of its dihydropyridine ring, the resulting pyridine ring is postulated to form an unstable moiety with an amino group (Ikemura et al., 2019).

Typically, a probe substrate is used for screening inhibitors. The use of probe substrates to determine inhibition of CYP epoxygenases can confound efforts towards inhibitor discovery because an epoxygenase may assume different conformations in response to one class of molecules (e.g. polycyclic xenobiotics) that may not account for a different substrate type (e.g. PUFAs or lipids). For example, both AA and ebastine (EBS) are structurally dissimilar substrates of CYP2J2. Doxorubicin (DOX), a cardiotoxic chemotherapeutic, is converted to 7-deoxydoxorubicin aglycone (7-de-aDOX) in the body. While DOX inhibits EBS and AA metabolism and demonstrates competitive inhibition, 7-de-aDOX only inhibits AA, but not EBS, metabolism (Arnold et al., 2017). From this example, it can be noted that multiple molecules can concurrently bind to CYP2J2, as supported by kinetic and fluorescence polarization (FP) measurements, but the conformation that these molecules occupy the active site is highly dependent on the structures of the molecules in the study (Figure 3). Taken together, interactions between a specific CYP and a specific ligand are unique with respect to that pairing, and this observation needs to be taken into consideration when extrapolating from studies using probe substrates. Inhibition studies must then highlight the value of determining the effects of inhibitors or drugs directly on the metabolism of endogenous substrates on CYP epoxygenases, such as AA for CYP2J2, instead of relying solely on probe substrate reactions. This might also indicate why screening of inhibitors using probe substrate reactions does not always translate into in vivo studies. Overall, it is important to screen for inhibitors that are the most specific for certain CYPs using both exogenous and endogenous substrates as probe substrates.

Figure 3. Modulation of arachidonic acid by CPR-mediated generation of 7-deoxydoxorubicin aglycone.

Occupation of the PUFA-binding pocket by 7-de-aDOX results in altered regioselectivity in metabolism of AA by CYP2J2 (left box). AA metabolism occurring in the absence of 7-de-aDOX assumes native regioselectivity preferences (right box). Image reproduced after slight modifications with permission from the original publisher (Arnold et al., 2017).

The physiological response to a drug-epoxygenase interaction extends beyond an instance or type of inhibition, where a broad relationship between drug shape and function suggests an altered production of CYP-derived bioactive lipids. To date, there is a limited amount of structure-function data to accurately predict how a drug will inhibit CYP2J2. Indeed, Figure 2 shows that CYP2J2 binds ligands with diverse chemical structures. In general, drugs that are more linear in shape tend to be competitive inhibitors. However, a drug’s geometry cannot always predict the ensuing enzymatic interactions and subsequent therapeutic activity. Some of these drugs have similar structures, like EBS and terfenadine, and inhibit CYP2J2 in similar ways. Others, like amiodarone and dronedarone, inhibit CYP2J2 in unique ways. Mechanism-based inhibition cannot always be predicted as well. Both danazol and 17α-ethynylestradiol contain an ethynyl moiety, but only the latter has been shown to inactivate CYP2J2. Meanwhile, a drug like azelnidipine, which has a biphenyl terminus similar to EBS and terfenadine, acts as a mechanism-based inhibitor. These nontransitive actions further exemplify the need to investigate ligand-ligand interaction on a per-need basis. Furthermore, Figure 2 also demonstrates that inhibiting CYP2J2 does not always lead to an observed cardiotoxic event, as not all cardioactive drugs that modulate CYP2J2 are cardiotoxic. For example, the antihypertensive drug telmisartan is a mixed inhibitor of CYP2J2 and has shown no adverse cardiotoxicity (Ren et al., 2013). Instead, we conjecture that cardiotoxic drugs may precipitate a toxic event and the inhibition of CYP2J2 exacerbates the toxicity by preventing the recovery (Arnold and Das, 2018). Despite the unpredictive nature regarding CYP2J2 inhibition and cardiotoxicity, the preponderance of cardiotoxic drugs modulating CYP2J2 exemplifies the need to include CYP2J2 in toxicity studies.

6.2. CYP2J2-ligand interactions

6.2.1. Molecular Basis of lipid-lipid interactions in CYP2J2

The inhibitor screen of CYP2J2 provided evidence that the CYP2J2 active site is large and can therefore bind structurally divergent xenobiotics. This implies that the CYP2J2 active site will be large enough to accommodate lipids and drug simultaneously as seen for CYP3A4 (Denisov et al., 2007a; Frank et al., 2011). Therefore, these drug-lipid interactions would impose noticeable effects on PUFA metabolism. Early instances of drug-lipid interactions within CYP2J2 were shown when AA was found to inhibit EBS hydroxylation (Hashizume et al., 2002). The rate of EBS hydroxylation by CYP2J2 was significantly greater than other microsomal CYPs, allowing for EBS to act as a probe substrate of CYP2J2. Use of 50 μM AA reduced EBS hydroxylation by ~90% in human intestine microsomes, whereas 100 μM AA nearly shut down the CYP2J2-mediated EBS hydroxylation (Hashizume et al., 2002). In a separate study, astemizole O-demethylation was inhibited by 33% relative to control in the presence of 100 μM AA (Lee et al., 2010). Taken together, these studies were the first examples of lipid-based inhibition of CYP2J2 activity, but without any mechanistic detail. These studies highlighted that endogenous lipids are inhibitors of drug metabolism by CYP2J2, and that endogenous ligands can prove to be strong inhibitors due to their physiological and evolutionary relevance to epoxygenase function.

We recently reported the metabolism of endogenous PUFAs and their eCB derivatives by CYP2J2 and how exogenous drugs modulate these lipids’ metabolism. We provided molecular details which contextualize epoxygenase action when presented with both endogenous lipids and exogenous drugs. The first study investigated mixtures of different PUFAs and how they inhibit each other’s metabolism. Overall among the pairs of PUFAs, the inhibition best followed a competitive inhibition model (Arnold et al., 2016). Experiments were conducted such that the PUFAs docosahexanoic acid (DHA), eicosapentanoic acid (EPA) and linoleic acid (LA), were individually co-incubated with AA. In these experiments, AA acted as either an inhibitor or a substrate. It was found that DHA inhibited AA metabolism the strongest with a K_i^DHA->AA of 16.5 ± 2.2 μM. However, in experiments with DHA as the substrate, AA inhibited DHA metabolism with four-fold less potency with a K_i^AA->DHA of 65.2 ± 15.8 μM. EPA and AA did not noticeably inhibit each other’s metabolism. We speculated this to be caused by their similarities in lipid chain length and variations in the degrees of unsaturation (Arnold et al., 2016). Competitive titrations against probe substrate EBS were then performed to evaluate this asymmetric trend in inhibition amongst PUFAs and AA. Here, DHA bound twice as tightly as AA to CYP2J2. Following this mode of analysis, LA and EPA bound more loosely to CYP2J2 than AA and DHA, which accurately reflects the trends seen from AA inhibition studies. Using MD simulations, few critical residues in CYP2J2 were identified as interacting with DHA, EPA and AA, namely T318, S493, and R321. From the modeling studies the terminal epoxide regioisomers of both DHA and EPA were expected to form preferentially over the other regioisomers. Specifically, a CYP2J2-T318V single mutant and a CYP2J2-T318V/S493A double mutant were made and the inhibitory binding constants for DHA and AA were found to be weaker for both the single and double mutants during the EBS competitive binding assay. However, the K_i for DHA in the single mutant was ~50% that of AA, but approaches ~80% that of AA when tested on the double mutant. Trends in weaker binding also translated to weaker inhibition of AA metabolism. As mutation of T318 and S493 resulted in DHA binding CYP2J2 with affinities similar to AA, it can be implied that these residues are involved in maintaining CYP2J2 selectivity between different fatty acid substrates. By focusing on these evolved interactions between a class of preferred lipid substrates and CYP2J2, we were able to uncover a “PUFA binding pocket”, the existence of which was able to justify an emerging pathway of DOX-induced cardiotoxicity mediated through CYP2J2 (Arnold and Das, 2018).

6.2.2. Molecular basis of doxorubicin-lipid interactions in CYP2J2 and potential alternate mechanism of doxorubicin-induced cardiotoxicity

Given CYP2J2 expression in the heart, we postulated that cardiotoxic drugs such as DOX modulated AA metabolism to incite cardiotoxicity (Arnold et al., 2017). This hypothesis was primarily supported by the observation that dose-dependent DOX toxicity is reduced either through CYP2J2 overexpression or EET exposure (Xu et al., 2011a; Zhang et al., 2009). This study suggested that CYP2J2 produced EETs or metabolized DOX directly. How CYP2J2 interacts with DOX is further complicated by the fact that DOX gets metabolized by CYP reductase (CPR) to form 7-de-aDOX(Bartoszek, 2002). We determined that DOX and 7-de-aDOX inhibit AA metabolism with an IC₅₀ and K_i of 10.2 ± 0.1 and 5.78 ± 0.03 μM, respectively. Using competitive fluorescence polarization experiments, 7-de-aDOX was found to inhibit AA binding to CYP2J2 with half the efficacy of DOX. However, in both experiments 7-de-aDOX does not fully inhibit AA metabolism, even at saturation. Given that the 7-de-aDOX did not completely inhibit AA metabolism raised the possibility that that there existed an alternative CYP2J2 binding site for 7-de-aDOX that allows concurrent binding with AA. Taken together, the inhibitory data suggested DOX to be a competitive inhibitor of AA. Meanwhile the concurrent generation of 7-de-aDOX during the reaction resulted in an incomplete inhibition, suggesting that 7-de-aDOX binding allows for the partial binding of AA. The partial effect of AA binding was further supported by a change of regioselectivity that increased 5,6-EET production while decreasing 14,15-EET. In comparison, DOX was a competitive inhibitor of EBS metabolism by CYP2J2 whereas 7-de-aDOX exhibited no inhibitory potential on the turnover, suggesting that 7-de-aDOX does not bind to the same site as EBS. MD simulations were used to further elucidate potential binding modes for DOX and 7-de-aDOX to CYP2J2, and to contextualize the effects of DOX and 7-de-aDOX on AA metabolism regioselectivity. The simulations were used to examine how these anthracyclines interacted with the previously identified PUFA-binding pocket. 7-de-aDOX was shown to bind into the PUFA binding pocket (T318, R321, and S493) near the heme and alter the positioning of AA to skew regioselectivity during metabolism to favor 5,6- or 8,9-EET generation. Interestingly, these regioisomers are less capable in serving cardioprotective roles compared to 14,15- or 11,12-EETs (Figure 3) (Arnold et al., 2017). The simulations also depicted EBS as sampling space opposite the PUFA-binding pocket and peripheral to the heme, therefore not interacting with the site of 7-de-aDOX binding. All simulations performed were in agreement with experimental data, as increasing DOX exposure did result in a significant decrease in terminal 14,15-EET formation (Figure 3). Lastly, the generation of 7-de-aDOX illustrated potential complexity of CYP-based inhibitions, as this inhibitor was formed through interactions with CPR to then inhibit CYP2J2 epoxygenation of AA. Taken together, the inhibitory capacities of DOX and 7-de-aDOX were shown to be specific to AA metabolism because it was found that 7-de-aDOX occupied a binding site specific for fatty acid and lipid substrates. Thus, multiple events occurring within the CYP2J2 active site during DOX metabolism ultimately led to skewed regioselectivity to favor EETs with altered bioactive properties from those otherwise produced during basal conditions.

6.2.3. Molecular basis of endogenous lipid virodhamine interactions with CYP2J2

PUFAs are examples of endogenous substrate-inhibitors of CYP2J2; that is, they function as substrates of CYP2J2 that also inhibit each other’s metabolism. Molecules that function primarily as inhibitors (without being metabolized) are instrumental in shutting down and regulating physiological functions. DOX is an example of such an inhibitor as we could not identify a CYP2J2-mediated metabolism. While it is of importance to find exogenous inhibitors, the body has its own repertoire of endogenous inhibitors that are not well characterized. Recently it was shown that virodhamine (O-AEA), an endogenous AA derivative, is an endogenous inhibitor of CYP2J2 (Carnevale et al., 2018). The overall catalytic turnover of O-AEA into epoxy-O-AEAs was very low (V_max = ~ 5 pmol product∙min⁻¹∙nmol⁻¹ P450). When co-incubated with AEA for inhibition studies, O-AEA was found to be a potent inhibitor and inhibited AEA metabolism (K_i of 8 ± 2 μM), while the V_max for AEA turnover remained the same. The kinetic data suggested competitive inhibition of AEA by O-AEA. Competitive titrations against EBS revealed that O-AEA produced a type II bathochromic (“red”) shift in the CYP2J2 UV-Vis spectra. Type II shifts are typically indicative of direct binding to the heme prosthetic group. Direct binding to the heme also explains the high inhibitory potential of O-AEA on EBS binding to CYP2J2, as its K_i was 13 ± 4.7 μM. MD studies showed that while protonated O-AEA interacted with E314, F310, and T219 aid in directing O-AEA to assume a catalytically competent conformation, some snapshots of neutral O-AEA positioned the amine group as being able to directly bind heme. Thus, O-AEA competitively inhibits AEA metabolism by occupying the space near the heme, which is facilitated through hydrophobic contacts with the I-helix and/or through direct heme ligation (Carnevale et al., 2018). Furthermore, a dose-dependent wound healing experiment revealed that O-AEA dose-dependently inhibits wound healing in hCYP2J2-transfected HUVEC cells with an IC₅₀ value of 0.61 μM. Taken together, the molecular characterizations of the interactions between O-AEA and CYP2J2 are just one instance of an endogenous interaction between a PUFA and a CYP epoxygenase bearing physiological relevance. It is likely that there are other endogenous molecules which can regulate epoxygenase function through selective inhibition, which might have variable physiological consequences depending upon the specific CYP and tissue localization involved in the inhibition event.

6.2.4. Molecular basis of phytocannabinoid metabolism by CYP2J2

While the previously discussed studies focus on metabolism of endogenous substrates, it is important to recognize that CYP epoxygenases also metabolize xenobiotic substrates. Phytocannabinoids (pCBs) are the chemical constituents of Cannabis sativa and represent an additional class of xenobiotic substrates for CYPs. The extent of cyclization of the pCB rings (Figure 2) can become a useful handle in differentiating the subclasses of pCBs. Specifically, it is well-known that maintaining a tricyclic pCB scaffold maintains cannabinoid receptor partial agonism, whereas ring opening is one factor which can lead to abolished receptor activation (Prandi et al., 2018). Furthermore, the opening of these rings provide aliphatic alkyl chains aside from the side chain off the resorcinol moiety, introducing additional lipophilic character (Hanus et al., 2016). Such inherent lipophilic character can provide the basis of molecular interactions with CYPs.

Although pCBs are known to inhibit CYP-mediated drug metabolism in microsomes (Zendulka et al., 2016), there previously was no biochemical study of CYP2J2-mediated metabolism of pCBs. pCBs have known cardiovascular effects (Adams et al., 1976; Dewey, 1986; Fredericks et al., 1981; Jones, 2002; Mendizabal and Adler-Graschinsky, 2007; Randall et al., 2004), which led us to believe that they interacted with CYP2J2. Given CYP2J2’s role in mediating homeostasis within the endocannabinoid system, we studied the mechanism of CYP2J2-pCB interactions. Additionally, study of CYP2J2 would serve as a model enzyme for observing how interactions with pCBs affect eCB metabolism. In light of increased medical and recreational Cannabis usage worldwide, these studies provide cardiovascular insights into how pCBs interact with the endocannabinoid system.

We showed that CYP2J2 can metabolize six of the most naturally predominant pCBs: Δ9-tetrahydrocannabinol (Δ9-THC), Δ8-tetrahydrocannabinol (Δ8-THC), cannabinol (CBN), cannabidiol (CBD), cannabichromene (CBC), and cannabigerol (CBG) (Arnold et al., 2018). End-point metabolism resulted in the discovery that each pCB was monohydroxylated to form a primary metabolite, although there were other dioxygenated side products formed as well. Mass spectrometry results suggested that despite the structural differences between each of the pCBs tested, the primary metabolites derived from each species were likely hydroxylated along the pentyl side chain to form 1’/1”-OH products when compared with well-established fragmentation patterns published for phytocannabinoids (Harvey, 1987).

In the presence of AEA, all pCBs, with the exception of CBG, demonstrated noncompetitive inhibition against the generation of EET-EAs from anandamide. CBG competitively inhibited AEA metabolism with a K_i of 10.8 ± 1.4 μM. The only pCB with a tighter inhibition constant was Δ9-THC at 9.86 ± 1.4 μM. The remaining pCBs inhibited AEA metabolism less efficiently than Δ9-THC. In decreasing potency, AEA metabolism was inhibited by CBD, followed by CBN, Δ8-THC, and CBC. Trends in catalytic efficiency mirrored those seen for inhibition constants across the pCBs, except the order of greatest catalytic efficiency between Δ8-THC and CBD were opposite that seen within the inhibition data. Across all the binding and metabolism studies, the pCBs binding affinities and turnovers were equal to or better than AEA. Taken together, pCBs effectively shut down AEA metabolism by CYP2J2 through a noncompetitive model and are metabolized by CYP2J2 (Arnold et al., 2018).

The pCBs provide a noncanonical example of noncanonical inhibitors, as the pCBs are also substrates of CYP2J2. Noncompetitive inhibitors typically act on an allosteric site and therefore are not substrates themselves; thus, the pCBs may either be binding two sites or their binding to the active site does not preclude AEA binding. CBG, however, is a competitive inhibitor; perhaps its more linearized structure gives it more flexibility and a more lipid-like structure. Thus, it binds similarly as AEA and functions as a competitive inhibitor. The generation of a single, general metabolite suggests that despite their structural differences, all pCBs tested were able to position their alkyl chains for C-H hydroxylation by CYP2J2 in a similar, if not identical, manner.

7. Delineating the role of the membrane environment and macromolecular crowding in CYP-CPR interactions.

All human CYPs are membrane-associated proteins, where the surrounding membrane and macromolecular environments have been shown to affect membrane and soluble protein functions (Brignac-Huber et al., 2016; Scott et al., 2016). Mammalian CYPs interact with their redox partners – including CPR, ADX-ADR – in the membranes. Measuring interaction between the CYP and the membrane bilayer is an active area of both experimental and computational research. Notably, P450 interactions with membranes have been shown to regulate many important aspects of molecular recognition of CYPs including, but not limited to, ligand binding, membrane depth of insertion, redox potentials, enzyme stability, structure, and overall enzyme orientation (Srejber et al., 2018). Each of these variables has a direct effect on the function of CYPs, and, in the case of CYP epoxygenases, the generation of bioactive epoxides. Additionally, advanced computational approaches have revealed the dynamic nature of CYPs towards interactions with the membrane bilayer (Baylon et al., 2013; Berka et al., 2011; Berka et al., 2013; Cojocaru et al., 2011; Denisov et al., 2012; Lonsdale et al., 2014). Previous studies have revealed that CYPs are inserted within the membrane, with a partially imbedded active site and deeply immersed N-terminus and F-G loop (Bayburt and Sligar, 2002; Berka et al., 2011; Ozalp et al., 2006). Therefore, including the membrane as a factor in influencing the CYP activity is imperative. Although there are several studies indicating that anionic lipids in the membrane influence CYP-CPR association and activity (Balvers et al., 1993; Blanck et al., 1984; Das and Sligar, 2009; Ingelman-Sundberg et al., 1981; Kim et al., 2003; Whited and Johs, 2015), the role of the membrane in these interactions is not totally resolved.

CYP2J2 is an integral membrane protein which is anchored into the membrane bilayer via its N-terminus hydrophobic residues and F-G loop. Recently, the membrane binding and insertion of CYP2J2 was captured using a highly mobile membrane-mimetic (HMMM) model to run molecular dynamics (MD) simulations (McDougle et al., 2015). The MD simulations revealed three key residues—W235, I236, and F239—in the F-G loop, as well as W48 in the N-terminus that are responsible for association with the membranes (Figure 4A). Noting I236 and F239 are hydrophobic, both were mutated to hydrophilic residues – aspartic acid (D) and histidine (H) which can potentially form a stable salt-bridge between them. MD simulations revealed that I236D causes W235 to begin detaching from the membrane while F239H has less of an effect. The double mutant containing both I236D and F239H has a similar effect to I236D (McDougle et al., 2015). These CYP2J2 mutants were then expressed in E. coli and characterized spectroscopically. The mutations did not interrupt the P450-fold or the heme-thiolate binding motif to a large extent and were found to be stable. AA metabolism was used to assess the functionality of each mutant where it was discovered that I236D mutant had greater catalytic efficiency compared to the base CYP2J2 construct while F239H dramatically exhibited decreased activity despite the membrane insertion predictions from the MD simulations. The double mutant had a similar catalytic efficiency to the CYP2J2 base construct and none of the mutants changed the ratio of EET regioisomers, indicating little change at the heme active site. Using fluorescence spectroscopy and MD simulations, high resolution interactions between the membrane and CYP2J2 F-G loop were provided. Overall, the results suggested that the introduction of hydrophilic residues in the F-G loop may help in engineering a soluble CYP2J2 without major alteration of the site for substrate recognition or product egress (McDougle et al., 2015).

Figure 4. Effect of membrane association on CYP2J2 function.

(A) Membrane binding of CYP2J2 captured with the HMMM membrane in 40 ns. The membrane-bound model of CYP2J2was extended to a POPC membrane and simulated for another 100 ns. Residues Trp-48, Trp-235, Ile-236 and Phe-239, identified to interact with the membrane upon binding, are shown in stick representation (McDougle et al., 2015). (B) Three-dimensional plot of representative absorbance spectra from 0 to 1000 s following mixing of NADPH with a CYP2J2–CPR pre-incubated complex (Meling et al., 2015a). (C) Association-dependent electron transfer kinetics of the formation of Fe(II)–CO peak at 449 nm over time (100 s) (Meling et al., 2015a). (D) Measurement of rate of NADPH oxidation and rate of metabolite formation from ebastine metabolism by CYP2J2-CPR-nanodiscs with varied lipid compositions (Huff et al., 2019). (E) Michaelis-Menten curves for 80% POPC / 20% POPS nanodiscs crowded with 100 mg/ml of macromolecular crowding agent (Huff et al., 2019).

Likewise, the N-terminus hydrophobic amino acid residues of CYP2J2 are important for CYP-CPR interactions and catalytic efficiency. A previous investigation using stopped-flow spectroscopy studied the role of the N-terminal domain in electron transfer from CPR to CYP, as well as the effects of substrate type and oligomeric state on both full-length and truncated CYP2J2 (denoted by fl and tr, respectively) (Meling et al., 2015a). CO reduction kinetics from 0.2–1000 seconds revealed that the association-dependent rate k₁ is ~200-fold faster for CYP2J2fl than for the truncated protein, indicating the essential role of the N-terminus in CYP-CPR association (Figure 4B and 4C). This remains true when AA is used as a substrate. Kinetics were also similar between the detergent reconstituted system and CYP2J2 in nanodiscs, showing that oligomerization state is not a key factor (Meling et al., 2015a). The authors hypothesized that the truncated N-terminus caused the CYP-CPR association step to be rate-determining, rather than fast, as it was for CYP2J2fl, implying that CYP2J2 and CPR may still associate rapidly in vivo. They also noted that CYP2J2tr may have adopted a less than optimal conformation with CPR, which would signify the importance of the N-terminus in docking. This is supported by both CYP2J2fl and CYP2J2tr having a K_s along the same order of magnitude. Overall, these studies highlight the functional importance of CYP N-terminus integrity, as the existence of hydrophobic residues along this region of the protein allow for efficient association with the CYP obligate redox partner and the membrane bilayer.

Aside from the structural elements which allow for CYP2J2 and CPR association, it is essential to take into consideration the role of the membrane bilayer composition in overall CYP function. In a recent study, it was shown that lipid membrane composition alters the rate of metabolism of EBS by CYP2J2, as well as CYP2J2s coupling efficiency with CPR (Huff et al., 2019) (Figure 4D). Nanodiscs containing 20% and 40% anionic lipids (POPS) had the highest rates of NADPH oxidation and EBS hydroxylation, respectively. Above 40% anionic lipids, there was a noticeable drop in enzymatic activity, resulting in speculation that electrostatic repulsion between anionic phospholipids can destabilize the nanodisc assembly (Roy et al., 2015).

Nanodiscs containing cholesterol produced slightly less product, and sphingomyelin (SM)-containing nanodiscs produced less than half the primary EBS metabolite, hydroxyebastine, compared to 20% POPS discs. Notably, the secondary metabolite of EBS, carebastine, was only quantifiably produced by nanodiscs containing SM (Huff et al., 2019). It was postulated that a lipid environment including SM promotes processive metabolism or that SM serves as a potential allosteric regulator. Extending these thoughts, it is likely that the ability for individual CYPs to sample catalytically competent conformations is heavily reliant on membrane composition, as this environment commands how a CYP folds around substrates and associates with CPR. And as CYPs are expected to be in lipid rafts associated with high concentration of SM, these results are relevant to in vivo CYP functionality. Membrane insertion then becomes an important factor as well to overall CYP function, holding equally valuable potential for translating to in vivo systems.

Taking into consideration the influence of the crowded cellular milieu beyond just membrane content, the same study also examined the effects of macromolecular crowding on CYP2J2 and CPR through basic characterization experiments in the presence of microviscogens. Microviscogens are viscous reagents which alter the relative solution volume available to proteins with which they share space (Blacklow et al., 1988), essentially simulating the effects of molecular crowding by changing effective concentrations (a phenomenon sometimes referred to as the excluded volume effect) in an in vitro setting (Christiansen and Wittung-Stafshede, 2013; Ellis, 2001; Schneider et al., 2015). Experiments in a lipid-solubilized-system (LSS) revealed an increased K_m for all crowding agents tested (Huff et al., 2019). However, studies were also performed in nanodiscs to further mimic a membrane-like environment. Ficoll 400 increased the K_m of EBS binding to 13.3 μM compared to 9.44 μM for uncrowded nanodiscs. Both Dextran 75 and Dextran 500 slowed the V_max, and Dextran 75 significantly decreased the K_m (1.40 nmol product∙min⁻¹∙nmol⁻¹ P450 compared to 4.05 nmol product∙min⁻¹∙nmol⁻¹ P450 for uncrowded). This difference in effect was attributed to two factors—volume and compressibility. Though similar in molecular weight, Ficoll is known to be more compressible than Dextran and thus takes up less volume overall (Venturoli and Rippe, 2005). An expansion of this study examined whether macromolecular crowders would affect the stability of CYP2J2 and CPR through ΔG unfolding experiments. Data revealed that both Ficoll 70 and sucrose had a destabilizing effect on CYP2J2, whereas CPR was stabilized by sucrose (Huff et al., 2019).

Ultimately, the study concluded that the surrounding environment of integral membrane proteins such as CYP2J2 is critical to protein function and should be taken into account when examining other aspects of the protein. True for all CYPs, the composition of the membrane environment is a critical determinant for how the protein can sample different conformations, associate with CPR, and therefore how it perceives different substrates. Subtle differences between CYP epoxygenase function, product preference, and regio- and enantio-selectivity must then all be influenced, and perhaps partially explained, by native lipid environments and extent of membrane insertion. It is therefore likely that the association with human membranes was a critical determinant for the evolved structural architectures of human CYP epoxygenases.

8. Structural analysis of CYP epoxygenases and homology models of CYP2J2

8.1. Comparison of crystal structures of CYP2 epoxygenases and CYP3A4

Given how malleable CYPs are in their ability to perform a variety of reactions, it becomes critical to obtain accurate crystal structures so that specific functions can be related to structural features, namely domain orientations and folds. A crystal structure from the CYP2J family is currently unavailable. However, many other CYPs homologous to CYP2J2 have been crystallized, including the other human epoxygenases, CYP2C8 and CYP2C9. Thus, parts of epoxygenase structures have provided the basis for CYP2J2 in silico homology models. In this section, the key aspects of the known crystal structures of CYPs used to develop CYP2J2 homology models will be discussed. These crystal structures are of the following CYPs: 2C5, 2C8, 2C9, 2C19, 2A6, 2B4, 2D6, 2R1. Crystal structures of CYP3A4 are also discussed in order to illustrate the importance of experimentally determined active site volumes for explaining observed substrate permissiveness. The factors governing ligand binding to CYPs cannot be generalized, per se, as specific CYPs will have a different interaction with each ligand. However, a special emphasis will be given to structures that show common interaction features, multiple ligand binding, and unique ligand recognition. These features are especially important in evaluating homology models of CYP2J2 obtained from these structures.

All CYPs contain a general conserved CYP-fold (Werck-Reichhart and Feyereisen, 2000). Although the sequence identity among CYPs can be as low as 20%, they possess a high degree of structural similarity, particularly around the heme region with regards to the precise positioning of key helices. CYPs generally contain 4 β-sheets and 13 α helices. The conserved core is comprised of a 4-helix bundle (designated as the D, E, I, and L-Helices), the J and K helices, two sets of β-sheets, and a coil. The heme binding loop is comprised of a Phe-X-X-Gly-X-Arg-X-Cys-X-Gly sequence, in which the Cys is crucial for heme ligation. The Glu-X-X-Arg patch on the K-Helix is also conserved that is involved in stabilization of the protein core. The variable regions consist of the 6 substrate recognition sequences (SRS) and the membrane anchoring (N-terminal) region. Further membrane anchoring is facilitated by the F-G loop region. More importantly, the helices that directly contact heme are conserved – especially the I-and L-helices. The B’ helix is the least conserved across different CYPs, as this helix controls substrate specificity.

For most of the mammalian CYPs that have been crystallized, the hydrophobic N-terminus is truncated and is replaced by a soluble hydrophilic sequence to increase expression in E. coli and in vitro solubility following purification (Zelasko et al., 2013). These modifications generally do not alter CYP activity and allow them to still maintain association with membranes through other membrane binding regions (McDougle et al., 2013; Pernecky et al., 1993).

Rabbit CYP2C5 was the first mammalian CYP to have its crystal structure resolved (Williams et al., 2000). The modified CYP2C5 construct, CYP2C5/3LVdH showed remarkable spatial conservation at the heme binding site to bacterial CYP102A1, given that bacterial CYPs are water soluble (Cosme and Johnson, 2000; vonWachenfeldt et al., 1997). This is remarkable as there have been millions of years of divergence between bacteria and rabbits; furthermore, bacterial CYPs are soluble while mammalian CYPs are membrane-bound. As CPR has a negatively charged surface, a positively charged surface near membrane-bound CYP heme was designated as the CYP-CPR binding interface, which was based also on previous mutational studies demonstrating reduced CPR binding (Williams et al., 2000). Thereafter, structures of CYP2C5/3LVdH were solved with different drugs such as DMZ (Wester et al., 2003a) and diclofenac (Wester et al., 2003b). The carboxylate moiety of diclofenac formed an extensive H-bond network with water among residues N204, K241, S289, and D290. Such H-bond and charge interactions are key to carboxylate recognition with many CYPs (Wester et al., 2003b).

CYP2C8 and CYP2C9 are two of the predominant epoxygenases in humans. The closest in homology to CYP2J2 is CYP2C8 followed by CYP2C9. Eric Johnson and coworkers elucidated the structure of substrate-free CYP2C8 in 2004 (Schoch et al., 2004). CYP2C8dH crystallized as a dimer with extensive interactions between helices F to G of each monomer. Importantly, two palmitic acid molecules were bound at the interface of these two dimers which may or may not represent a functional fatty acid binding site. Later, they determined the structures of the CYP2C8dH construct bound to different drugs such as montelukast, troglitazone, felodipine, and 9-cis-retinoic acid (Schoch et al., 2008). However, two molecules of 9-cis-retinoic acid were bound concurrently to CYP2C8 in the active site as this ligand is smaller than the other ligands that were tested (Figure 5A). Experimental data corroborated two binding sites for 9-cis-retinoic acid in CYP2C8 (Schoch et al., 2008). These data are key for establishing the ability for epoxygenases to bind multiple ligands.

Figure 5. Structures of the most characterized CYP2 human epoxygenases.

(A-C) Structures of (A) CYP2C8 (crystal, PDB: 1PQ2), (B) CYP2C9 (crystal, PDB: 1OG2), and (C) CYP2J2 (homology model, (Lafite et al., 2007a)). The hemes are shown as red sticks. The I-Helices are highlighted in orange. The palmitic acid (PA) in the CYP2C8 structure is shown as black sticks. (D-E) Key active site residues for substrate binding as mentioned in the text. Some residues have been omitted for clarity. (D) Residues (blue) involved in 9-cis-retinoic acid (RA, orange) binding to CYP2C8 (PDB: 2NNH). Two molecules of RA are bound in the active site and support homotropic cooperativity. (E) Residues (green) involved in flurbiprofen (FBF, orange) binding to CYP2C9 (PDB: 1R9O). These residues are believed to contribute to the selectivity of CYP2C9 for negatively charged substrates. Notably, R108 swings inward by almost 180° to interact with FBF, as compared to the apo structure (PDB: 1OG2). (F) Residues (green) comprising the “PUFA binding pocket” in CYP2J2 as revealed by MD simulations (Arnold et al., 2016). Arachidonic acid (AA, orange) is bound. For each crystal structure, one monomer is shown for clarity.

The structure of CYP2C9 was originally resolved with and without warfarin bound (Williams et al., 2003). An important feature of this structure was that S-warfarin was bound in a pocket that was too far away from the heme to be properly metabolized but could allosterically modulate the binding of other substrates. In silico modelling confirmed that another warfarin molecule or fluconazole could bind near the heme with warfarin bound in this pocket (Williams et al., 2003). A separate study later found that 6-hydroxyflavone noncompetitively inhibited CYP2C9 by binding near this warfarin allosteric site and preventing diclofenac access to the heme (Si et al., 2009). These data helped to establish that one ligand’s binding near the heme can allosterically modulate the binding of another. Later, another CYP2C9 structure was obtained with an anionic substrate, flurbiprofen, bound to the active site (Wester et al., 2004). The original structure obtained by Williams et al. could not explain the anionic selectivity of CYP2C9 due to the lack of basic residues in the active site (Williams et al., 2003). However, the new structure by Wester, et al. showed that flurbiprofen induces conformational changes that allow residues such as R108, D293, and N289 to swivel into position (Figure 5B) (Wester et al., 2004). Another structure obtained by Liu et al. shows that R108 also interacts with the thiophene group of TCA007, facilitating substrate binding (Liu et al., 2017b). Therefore, these structures demonstrated that substrates can induce different conformations upon binding.

CYP2C19 is 91% homologous to CYP2C9. The structure of CYP2C19 bound to 0XV was determined in 2012 (Reynald et al., 2012). CYP2C19 contains an antechamber that resembles the warfarin site in CYP2C9. However, in CYP2C8 smaller residues merge these two chambers into one large active site. CYP2C8, therefore, has the largest active site, followed by CYP2C19 and then CYP2C9.

Like CYP2C5, rabbit CYP2B4 represents another pinnacle CYP that has become a model CYP for biophysical and structural studies. Particularly, the structures of CYP2B4 have been used to model open and closed confirmations of mammalian CYPs. CYP2B4 was crystallized as a dimer in the open (substrate-free) conformation in 2003, due to the presence of a cleft in the F-G region (Scott et al., 2003) and its structure was further resolved in the closed conformation in 2004 by binding to 4-CPI (Scott et al., 2004). The putative closed conformation was crystallized with 4-CPI bound to the protein and achieved by mutating a histidine residue in CYP2B4dH (CYP2B4dH-H266Y) (Scott et al., 2004). The cleft seen in the open confirmation moved in towards 4-CPI to isolate it and the active site from the bulk solution. This “closed” structure more closely resembles the structures of CYP2C5, CYP2C8, and CYP2C9 obtained earlier. Later, a substrate-free structure in the closed conformation was obtained (Wilderman et al., 2010). Interestingly, CYMAL-5, a common detergent used to solubilize CYP2B4, was found not only to bind a peripheral site near the F-G region of WT and mutant (F202W or F224W) CYP2B4 but also at the active site (Gay et al., 2010; Liu et al., 2016; Shah et al., 2016; Shah et al., 2012; Wilderman et al., 2010). This demonstrates that detergents often act as ligands and bind to CYP active sites. For instance, it has been reported that nonionic detergent Emugen 913 can be replaced for phospholipid as an activator of N-demthylase activity of CYP2B4 at low concentrations (Myasoedova et al., 2006). Finally, several mutational studies provided further insight into substrate recognition by CYP2B4.

CYP2D6 is not considered a primary epoxygenase; however, it metabolizes PUFA-derived eCBs. CYP2D6 is known to produce all four regioisomer epoxides of AEA (EET-EAs), as well as hydroxylated products of AEA (Snider et al., 2008; Sridar et al., 2011). CYP2D6 is a primary CYP of the brain and is known to selectively metabolize basic and nitrogen-containing substrates that includes psychoactive and psychotherapeutic drugs such as 3,4-methylenedioxymethamphetamine (MDMA) (de la Torre et al., 2012), tricyclic antidepressants (Dean, 2012), and cannabidiol (Jiang et al., 2011). The structure of CYP2D6 (allelic variant V347M) was solved in 2006 (Rowland et al., 2006). The CYP2D6 architecture was closest to that of CYP2C9, but with key differences. The most prominent difference is in the F-G region. Importantly, negatively charged residues were identified in the active site and were proposed to attract basic ligands to the active site (Rowland et al., 2006). Thereafter, several structures of ligand-bound CYP2D6 were reported (Wang et al., 2012; Wang et al., 2015). The first study demonstrated prinomastat bound to CYP2D6 in a closed conformation (Wang et al., 2012). Later, structures of thioridazine demonstrated open and closed structures, with two molecules of thioridazine bound in the open structure (Wang et al., 2015).

CYP2J2 has a high amount of sequence homology to CYP2R1, despite these enzymes having distinct functions. CYP2R1 is responsible for hydroxylating the vitamin D series, notably producing calcifediol (Cheng et al., 2004). CYP2J2 likewise metabolizes Vitamin D₂, D₃, and 1α-OH-D₃, with a higher preference for the exogenous D₂ over the endogenous D₃ (Aiba et al., 2006). There are 3 PDB entries for the structure of CYP2R1 (Strushkevich et al., 2008). The substrate access channel in crystals with vitamin D₃ bound formed an “extended” active site. The conformation of Vitamin D₃ was in a non-steroidal conformation, consistent with the inability of CYP2R1 to metabolize cholesterol or 7-dehydrocholesterol (Strushkevich et al., 2008). A unique structural aspect of the CYP2R1 structure is that the I-Helix is not distorted, whereas I-helix distortion in other mammalian and bacterial CYPs is proposed to help facilitate proton delivery to O₂ during the catalytic cycle (Meling et al., 2015b).

CYP3A4 is a highly promiscuous enzyme and the concurrent binding of several substrates has been proposed to be the basis for some drug-drug interactions (Denisov et al., 2007b; Denisov et al., 2009; Guengerich, 1997). Although CYP3A4 is not highly homologous to CYP2J2 and has not been used to generate a CYP2J2 homology model, it is one of the most studied human CYPs with respect to drug metabolism. Furthermore, it has a large active site similar to that proposed for CYP2J2. Despite having less than 40% homology to CYP2Cs, CYP3A4 has a structure that is rather similar to CYP2C8 (Yano et al., 2004). Both have large active sites, i.e. 1438 ų for CYP2C8 and 1386 ų for CYP3A4. However, the key difference is that the active site of CYP3A4 is more open at the vicinity of the heme due in part to shorter F- and G-Helices that do not pass over the active site. This larger volume allows for the accommodation of a large array of substrates and substrate orientations that have been observed experimentally (Yano et al., 2004). Since then, many structures of CYP3A4 with other ligands have been solved with various drugs and endogenous ligands bound, such as progesterone and metyrapone (Williams et al., 2004), ketoconazole and erythromycin (Ekroos and Sjogren, 2006), ritonavir and analogs (Sevrioukova and Poulos, 2010; 2012a; 2013a; b), bromoergocryptine (Sevrioukova and Poulos, 2012b), progesterone and citrate (Sevrioukova and Poulos, 2015), midazolam (Sevrioukova and Poulos, 2017), and metformin (Guo et al., 2017). Progesterone bound at a peripheral site near the access channel and membrane interface (Sevrioukova and Poulos, 2015; Williams et al., 2004). Later, progesterone binding at this site was shown to increase the activity of CYP3A4 and contribute to cooperativity (Denisov et al., 2015).

Capturing the different crystal structures of CYP2 enzymes and others has enabled a deeper understanding of important characteristics regarding CYP function. One of the most striking and defining features of CYPs is their promiscuous substrate selectivity and conformational plasticity. This is perhaps best exemplified by the open, closed, and quasi-closed states of CYP2B4, demonstrating that substrates can induce key structural changes (but not always). Additionally, CYP2C9 demonstrates how a substrate (flurbiprofen) can induce a key interactive residue to swivel into the active site. Complex allosteric binding has also been revealed through these crystal structures. CYP2B4, CYP2C8, and CYP3A4 demonstrate peripheral binding sites for CYMAL-5, palmitic acid, and progesterone, respectively, and warfarin shows allosteric binding to CYP2C9 near the heme. Multiple-ligand binding to CYPs is demonstrated by CYP2C9 (retinoic acid), CYP2D6 (thioridazine), and CYP2B4 (CYMAL-5) providing insight into the complex cooperativity observed with CYPs. Overall, it is this structural malleability between CYPs and their ligands that make generalizations regarding ligand binding difficult. Ligand binding, therefore, must be studied on a per-basis need to gain insight into the unique molecular interactions governing respective binding. This may especially be true regarding PUFAs, which are highly malleable substrates whose conformations may be highly influenced by interactions with another ligand. More crystal structures of CYPs with PUFAs or PUFA-like ligands may help to provide insight into this conformational pliability. However, the crystal structure of one CYP cannot suffice in explaining or predicting the ligand interactions of another CYP due to these complex, unique interactions. Therefore, no current crystal structure alone can provide insight into how ligands bind to CYP2J2. Other techniques may further our understanding of the dynamic interactions between PUFAs and epoxygenases. Homology modeling and molecular dynamic simulations can provide valuable insight into interactions between ligands and proteins with respect to time, especially if the protein has no known crystal structure. In the next sections, the homology models and molecular dynamics simulations of CYP2J2 are discussed, which can help bridge the gap in this knowledge.

8.2. In silico modeling of the CYP2J2 structure and use of molecular dynamics to perform structure-function studies

Computational approaches can provide insight into the dynamics of molecular interactions, which are hard to ascertain by the static frames of crystallographic data. Several homology models have been developed and have been used in performing molecular docking and molecular dynamic simulations to better describe CYP2J2 structure-function, as this epoxygenase has no available crystal structure. These models rely on the crystal structures of various CYPs as outlined in the previous section. In this section, the development of CYP2J2 homology models and insights provided from these models are described.

One of the earliest models was developed by Lafite et al. (Lafite et al., 2007a). The purpose of the model was to help explain the regioselective hydroxylation of terfenadone derivatives by CYP2J2. Many CYP structures were used to develop the model, which are as follows: CYP2A6 [PDB 1Z11, (Yano et al., 2005)], CYP2B4 [PDB 2BDM, (Zhao et al., 2006)], CYP2C5 [PDB 1NR6, (Wester et al., 2003b)], CYP2C8 [PDB 1PQ2, (Schoch et al., 2004)], and CYP2D6 [PDB 2F9Q, (Rowland et al., 2006)]. The authors used the SWISS-MODEL v3.5 server to build the CYP2J2 model. To test the robustness of the model builder, the authors built a model of CYP2B4 using this software and compared it to its crystal structure. They were able to obtain a model of CYP2B4 that matched within 1.05 Å of the known closed structure of CYP2B4 [PDB 1SUO, (Scott et al., 2004)] despite the reference structures having only 51%−54% sequence identity to CYP2B4. Molecular docking was then performed on select terfenadone derivatives. One of the most important findings of the study was that R117 is important for forming H-bonds to the ketones of these derivatives. This helped to position the molecule near the heme for metabolism and would also aid in the understanding of terfandone-CYP2J2 structure-function for future inhibitor development (Figure 5C).

Later, a model was developed by Li, et al. based on the structure of CYP2C9 bound to warfarin [PDB 1OG5, (Williams et al., 2003)] using Modeller 8v2 software (Li et al., 2008). They truncated CYP2J2 at the first 43 residues of the N-terminus to reflect the missing residues of CYP2C9 in the crystal structure. MD simulations were then used on the resulting model for 1 ns to test the stability of the structure. Furthermore, molecular docking with terfenadone and related substrates of CYP2J2 was done and, again, the importance of R117 in directing ligand binding was observed. However, it was also observed that R117 formed a salt bridge with E222, which is analogous to other observations for similar CYPs. While the identification of this salt bridge did not occur in the work of Lafite et al., similar results between both efforts ascribe the same amino acid with functional importance in maintaining structural interactions between CYP2J2 and its substrates. Again, homology models may not necessarily replace the accuracy of a crystal structure in a specific conformation. But, as shown from these studies, the ability of models to accurately describe biophysical phenomenon can be validated by cross-referencing results obtained from multiple models, where each model was constructed by different software packages and computational approaches.

It can become difficult for experimentalists to know which homology models to trust when it comes to the best interpretation of a protein-ligand interaction. For example, another CYP2J2 model (Lee et al., 2010) based on the structures of CYP2A6 [PDB 1Z10, (Yano et al., 2005)], CYP2B4 [PDB 1PO5, (Scott et al., 2003)], and CYP2C8 [PDB 1PQ2, (Schoch et al., 2004)] was prepared. The purpose for generating a model in this study was to compare the active site volumes between CYP2J2 to CYP3A4. Referencing active site volumes between these two CYPs served as the authors’ basis for comparison of in vitro ligand binding, and so no docking or MD studies were performed. Thus the stability of the model was not robustly tested. To accurately represent experimental observations or predict outcomes of CYP structure-function studies it is important to understand a model’s design and purpose.

As reiterated throughout this review, the binding of lipid substrates to CYP epoxygenases is distinct from that of more rigid substrates due to differences in ligand conformational flexibility. Accordingly, homology models should be developed to account for intricacies inherent to the desired application. Homology models were later developed to understand AA binding to CYP2J2. The first study investigated the polymorphisms T143A and N404Y (Cong et al., 2013). The sequence of CYP2J2 was submitted to PDB to look for sequence homologies. CYP2R1 bound to vitamin D₃ was used for the template based on sequence identity [PDB 3C6G, (Strushkevich et al., 2008)]. The model was then tested for stability with a 1 ns MD simulation. Molecular docking was performed for AA in the protonated form to each polymorphism, and then 10-ns MD simulations were performed. The group found that 5 of the 6 CYP substrate recognition sequences (SRS) were involved in AA binding, which mostly involved hydrophobic interactions. It was suggested that the T143A polymorphism could affect water access and N404Y polymorphism could affect the positioning of residues interacting with the headgroup of AA. The group later developed another model to investigate more CYP2J2 polymorphisms and obtained similar results (Xia et al., 2014).

Another model was developed later by Proietti et al. to further investigate AA binding to CYP2J2 (Proietti et al., 2016). This model was built based on the structure of CYP2B4 bound to 4-(4-chlorophenyl)imidazole [PDB 1SUO, (Scott et al., 2004)]. This structure was chosen based on sequence identity to CYP2J2 (39%), completeness of the sequence, and high resolution of the structure. This active site, however, was too small to accommodate a large molecule like arachidonic acid. Instead, the authors used the active site of CYP2R1 bound to vitamin D₃ [PDB 3C6G, (Strushkevich et al., 2008)] and likewise used vitamin D₃ as the model ligand to develop the CYP2J2 active site. The study then investigated AA binding to CYP2J2 in the unprotonated form using 30–50-ns MD simulations. It was found that R117 was important for AA binding to CYP2J2, which orients the 11,12-alkene position closest to the heme for metabolism.

A model of CYP2J2 was made by See-Hyoung, et al. based on the structure of CYP2C8 [PDB: 2NNI, (Park et al., 2017; Schoch et al., 2008)]. The authors, however, did not provide details into their model validation. The study demonstrated a noncompetitive inhibition of astemizole demethylation in human liver microsomes by a compound found in the gromwell plant, acetylshikonin. As astemizole demethylation is predominantly performed by CYP2J2, the authors then modelled astemizole and acetylshikonin binding to CYP2J2. Acetylshikonin, interestingly, bound in the active site near the heme and prevented astemizole from adopting a productive conformation, similar to the noncompetitive inhibition of CYP2C9 by 6-hydroxyflavone (Si et al., 2009). As phytocannabinoids (pCBs) are noncompetitive inhibitors of AEA metabolism and are also substrates, we suspect that the pCBs similarly inhibit AEA metabolism by binding in a pocket in the active site and preventing AEA metabolism (Arnold et al., 2018).

These several homology models vary in their approaches regarding the use of CYP templates and in the software used for modeling. They also vary in the degree of testing, validation, and employment of the models, the latter of which can give valuable insight into the integrity of the model and how they can be used. All the models employed molecular docking or molecular dynamics simulations in their studies to validate them, except for the model proposed by Lee et al. The models developed by Lafite et al., Cong et al., Xia et al., and Proietti et al. have all been used in MD simulations in addition to molecular docking. One of the most striking differences among the models is the range of active site volumes. From largest to smallest, the reported active site volumes are as follows: 1420 ų (Lee et al., 2010), 945 ų (Lafite et al., 2007a), 330 ų (Xia et al., 2014), 320 ų (Cong et al., 2013), and 234 ų (Proietti et al., 2016). It was suggested that the vast differences in active site volume estimation could be arising from how each program defines the active site volume and implied that including solvent or substrate channels and the heme pocket as defining the active site volume can give larger estimates (Proietti et al., 2016). It is important for both experimental and theoretical researchers across the spectrum of basic and translational research to be cognizant of the variability between different homology models when attempting to describe CYP structure-function. Confidence in a model can always be improved through validation measures, which can also include computational sampling techniques, like molecular docking and molecular dynamics simulations for characterizing biophysical interactions.

8.3. Molecular docking and molecular dynamics simulations

Despite the large variations in the models, especially with respect to active site volume estimations, some common features are apparent. Notably, R117 was shown to be an important residue for ligand recognition. This residue was crucial for forming H-bond interactions with the ketone of terfenadone and positioning it to the heme, which was later confirmed by a follow-up study using 200-ps MD simulations (Lafite et al., 2018; Lafite et al., 2007a). Similarly, R117 was identified in the binding of terfenadone and several similar ligands (Li et al., 2008). R117 was also shown to form an important salt bridge with E222, and it was demonstrated that in vitro mutations of R117 undermined the stability of CYP2J2 (Lafite et al., 2018; Li et al., 2008). R117 was implicated in directing AA binding to CYP2J2. However, the conformation of AA that interacted with R117 presented the 11,12- position closest to the heme for oxidation at the end of their simulations. The major site of metabolism is at the 14,15- position, which the authors mentioned was not replicated in their simulations (Proietti et al., 2016). One possibility is that their model has the smallest active site volume, which may be constraining AA into this position. Without further structural data, the true active site volume of CYP2J2 remains otherwise unknown. The other simulations performed in the protonated state of AA may support the predominance of hydrophobic interactions in their results (Cong et al., 2013; Xia et al., 2014). Still, these models have limited abilities in explaining the exact interactions between CYP2J2 and AA without being accompanied with rigorous thermodynamic and structural experiments.

A difficulty in accurately recapitulating CYP activity in vitro is mimicking the membrane environment. The previous simulations were performed in aqueous solutions. However, CYP2J2 exists in a membrane environment, which can influence the architecture of CYP2J2 and its activity. Therefore, Lafite’s CYP2J2 model was used to perform MD simulations of CYP2J2 in a highly mobile membrane-mimetic model (HMMM). From the simulations, we identified F-G loop residues which are important for the association of CYP2J2 to the membrane, namely I236 and F239. Building upon the truncated CYP2J2 construct (Δ34-CYP2J2), a double mutant, Δ34-CYP2J2-I236D/F239H, was produced that maintained enzymatic activity while also being soluble in aqueous media (McDougle et al., 2015).

We further carried out MD simulations to investigate the binding of lipid ligands to CYP2J2, namely AA, EPA, DHA, and O-AEA in their charged states (Arnold et al., 2016; Carnevale et al., 2018). These studies led to the determination of a “PUFA binding pocket” comprised of residues T318, R321, and S493, which direct PUFA binding to CYP2J2. The importance of these residues was confirmed by mutational analysis using a T318V and a T318V/S493A double mutant. Meanwhile, other binding clusters were observed where PUFAs bound to sites previously reported terfenadone and EBS binding sites (Arnold et al., 2016), and also resembled previously reported orientations assumed by AA in other homology models (Proietti et al., 2016). However, no significant interactions with R117 were observed in the simulations of CYP2J2 and the lipid ligands mentioned here. This discrepancy in binding interactions is likely the result of several differences between the previous simulations versus ours. Firstly, our lab’s work used the Lafite homology model, which has a much larger reported active site volume than the other models. Secondly, initial docking of each PUFA was done with the predominant position (the ultimate alkene) closest to the heme before running the simulations for 100–200 ps. Lastly, the simulations were carried out using a membrane-mimetic model. Controlling for substrate geometry and the native membrane environment prior to simulation were quality checks which we believe amounted to a more representative model (Jo et al., 2008; Morris et al., 2009).

Using these precautions in designing in silico experiments allow for more accurate computation that can accompany experimental data. When hypothesizing that DOX-induced cardiotoxicity was mediated by CYP2J2, substrate-bound CYP2J2 MD simulations were performed using different coincubations of PUFAs, DOX and 7-de-aDOX (Arnold et al., 2017). It was shown that 7-de-aDOX modulated the site of AA metabolism by CYP2J2 towards greater 5,6-epoxidation. MD simulations demonstrated that 7-de-aDOX can bind into the PUFA binding pocket of CYP2J2, thus shifting AA to place the headgroup closer to the heme and bias regioselective metabolism (Arnold et al., 2017). Experimentally observed shifts in metabolism validated the simulation results, providing a more solid rationale to explain the known cardiotoxicity of DOX than simulations or modelling would on their own (Arnold and Das, 2018).

Overall, computational analyses give information on the dynamics of ligand binding that are usually corroborated through crystal structure or NMR studies. MD simulation studies increasingly reported no major conformational changes in CYP2J2, which may reflect the rigidity of the CYP2J2 active site. Although they may not fully represent the system perfectly, in silico modeling can aid experimental data to develop a more comprehensive understanding of physically observed molecular interactions. This is especially true regarding multiple-ligand interactions, such as AA and 7-de-aDOX, (Arnold et al., 2017) as well as for single-ligand interactions, as was done for determining the primary mode of CYP2J2 inhibition by O-AEA (Carnevale et al., 2018). Simulations can help to bridge gaps in comprehending how a ligand engages with conformationally lithe CYPs. But, given their very short timescales and sampling approaches, these computational simulations are most effective when they equip a validated crystal structure or homology model, account for nuanced information which best mirror the context of the experiment (e.g. membranous versus aqueous environments), and accompany wet bench experimental data that the simulations are attempting to describe.

9. Conclusion and future outlook

Mammalian CYPs are the “organic chemists of the body” that are equipped with an impressive arsenal of synthetic capabilities, which allows them to metabolize a wide diversity of substrates. The biochemical roles assumed by CYPs are a consequence of their complex structure-function relationships, their large repertoire of substrates, and the membrane environment which encapsulates them. CYP epoxygenases are a specific subclass of CYPs that are worthy of singular attention, as their large active sites and ability to metabolize endogenous lipids, results in the formation of PUFA-epoxides with diverse pharmacological properties.

This review highlights the recent advances in the structure-function studies of CYP epoxygenases, focusing on CYP2J2 due to its prominent role in PUFA epoxygenation. Here, we have shown that only a subset of the 57 human CYPs have the ability to metabolize PUFAs into PUFA-epoxides (Table 1). The CYP2 class has been designated as the primary epoxygenases based on their widespread expression across different tissue types in murine models(Graves et al., 2013), which then translated into further research showing that CYP2 enzymes demonstrate an important physiological function related to the epoxidation of PUFAs into bioactive lipid mediators.

Despite functional differences, all CYPs utilize the same generalized CYP catalytic cycle to metabolize their substrates. It was noted that CYPs can be broadly classified into two categories based on their function. They can be classified either as biosynthetic CYPs that have evolved to synthesize endogenous steroids, or as xenobiotic-metabolizing CYPs that are involved in exogenous drug metabolism. Interestingly, CYP epoxygenases are well-endowed with both functionalities, as they are at the crossroads of bioactive oxy-lipid synthesis and drug metabolism. Most of the lipid-metabolizing CYPs, such as CYP2C8 and CYP2C9, are well known for their important role in drug metabolism. This fluid function of lipid-metabolizing CYPs, on one hand points to our need for understanding the determinants for molecular recognition and metabolism of lipids; on the other hand, this dual functionality demonstrates a physiological need for more information on the tissue distribution and activity of these CYP epoxygenases during different disease states. Overall, more studies in nearly every research area – gene expression, in vivo animal studies, in vitro biochemical assay, biophysical characterizations, crystal and spectroscopic structure determination, improved computational modeling and simulation capacities – are critically needed to understand lipid metabolism by CYP epoxygenases and its physiological relevance in health and disease.

Initial studies in the 1980s focused on identifying the regioselectivity of endogenous epoxidation by the CYPs in lieu of identifying specific CYP isoforms in microsomes that were responsible for the observed epoxidation. This classification of epoxide regioselectivity and enantioselectivity was later used as a metric to differentiate between key CYP epoxygenase activities in different tissues as well as diverse disease states. Overall, while most CYP research has been “liver-centric”, the role of extrahepatic CYPs in localized endogenous lipid metabolism is emerging. Currently there are very few studies that link the expression of CYP epoxygenases in tissues or in disease states with the type of epoxide regioisomers or enantiomers respectively formed to the outcome of disease progression. For instance, while CYP2J2 has been most widely studied with respect to its cardiovascular role, it is noted here that CYP2J2 expression is present in other extrahepatic tissues, such as the brain and small intestine. It is intriguing that single cell brain expression studies show that CYP2J2 (CYP2J6 in mice) is highly expressed in oligodendrocyte progenitor cells (OPC) as well as endothelial cells, suggesting that CYP2J2, as well as other CYPs and CYP epoxygenases, carry out specific functions within specific cell types (Toselli et al., 2016). That CYPs are expressed and active in extrahepatic tissues, like the brain, is a stance which researchers need to emphasize further; otherwise, the importance of CYP function across different disease states in different tissues will be stalled. The function of CYP2J2 in the brain and its dynamic expression in brain tissues during neuroinflammation is unknown. As PUFA-epoxides have been detected in brain, it is highly plausible that CYP2J2 produces PUFA-epoxides that act on microglial cells to reduce inflammation and promote pathological resolution. However, the mechanism of CYP2J2 and other epoxygenases to generate PUFA-epoxides in response to an insult or inflammatory event in the brain needs to be studied further. Given differences in regio- and enantioselectivity in epoxidizing endogenous substrates, it is very likely that different epoxygenases are differentially involved in the regulation of specific inflammatory responses across all tissues. Hence, there is a need to further elucidate how PUFA-epoxides are locally produced in different tissues to help modulate these responses.

CYP2J2 and other epoxygenases are membrane-bound proteins. Lipids and proteins behave differently in a membrane environment when compared to free fatty acids. Membrane bilayers introduce spatial restrictions to the proteins and thus change the relative diffusion of hydrophobic molecules such as PUFAs. Precisely how lipids or other substrates enter CYPs has not been experimentally proven, although computational studies have been referenced in this review when discussing the influence of the membrane environment on epoxygenase structure-function. Additionally, enzymes that release free fatty acids, those that metabolize these lipids into bioactive molecules, and the receptors which interact with these lipids, are all co-localized to the membrane. Given their close proximity, the effective concentration of PUFAs at the site of action becomes an important variable to consider for in vitro and in vivo research. A common criticism to in vitro structure-function studies on the interactions between bioactive lipid mediators and their protein targets is with respect to experimental design. These experiments are conducted on the micromolar scale, whereas the endogenous production of these molecules is several-fold lower in blood and extracted tissues. However, it is important to take into consideration macromolecular crowding within the membrane environment. Yes, many of the bioactive lipid mediators covered in this review are synthesized on-demand, which results in arguably reduced half-lives and chemical stability. But, the membrane provides a platform to facilitate interactions between these rapidly degrading bioactive lipids and their cognate receptors through dynamic and crowding effects (Cabanos et al., 2017; Comoglio et al., 2014; Hilgemann, 2007). Note that a single lipid synthesized within 10 nm of its target in the membrane has an effective concentration of ~3.5 mM (Cabanos et al., 2017).

Regardless of its localization, CYP2J2 expression and activity leads to the formation of non-allylic, cis-epoxides of PUFAs. PUFAs are different from other small or biological molecules. There are many unknowns of PUFA biology with respect to their release from and location within membranes throughout the body. PUFAs are also highly flexible, often adopting various conformations which lead to rattling and dynamically variable motions when associating with membranes and CYP active sites. These motions, which are due to high degrees of conformational flexibility, cooperate with the inherent structural malleability of CYP epoxygenases to influence the formation of various epoxides and hydroxides. Therefore, the PUFAs have less discriminatory molecular handles or reactive moieties aside from their head groups, as any modifications at the carboxylic ends of lipid substrates, such as eCBs, changes the binding interactions and biological properties of PUFAs. This makes their study more challenging because most of the molecular interactions between PUFAs and CYPs become difficult to experimentally modulate. More importantly, PUFAs are hydrophobic substrates with poor solubility in buffer conditions normally used to study biochemistry of protein-lipid interactions, which fundamentally limits the extent of possible biochemical characterizations that can be done on these molecules. Hydrophobicity and flexibility due to unbranched structures with varying degrees of unsaturation makes these molecules highly versatile (they bind several different receptors) and malleable (they can bind into a variety of cavities and can allosterically modulate substrate metabolism). These same features also make it difficult to predict how lipids and PUFAs will interact with a protein. Therefore, there needs to be a greater dataset of lipid-CYP interactions which will help to create better tools by which we may predict how these lipids bind to the epoxygenases, as well as other protein receptors which directly mediate downstream physiological responses and regulatory schemes.

The initial characterization of CYP2J2 was based on exploring its role in human cardiovascular physiology. CYP2J2 was recombinantly expressed using the baculovirus system which was not sufficient for biophysical studies. Later, it was recombinantly expressed in bacteria to yield enough protein for detailed structure-function studies involved in drug and lipid metabolism. It was shown that CYP2J2 was able to metabolize PUFAs and eCBs. There has been a rigorous search for CYP2J2 inhibitors by several and separate research groups using probe substrate reactions. The inhibitor screens lead to identification of danazol and terfenadine derivatives as the semi-specific inhibitors of CYP2J2. The diversity of identified inhibitors also demonstrated that CYP2J2 has a large active site which can provide greater conformational freedom to small molecules at the active site when compared to other epoxygenases. However, the identification of a specific mechanism-based inhibitor and a crystal structure of CYP2J2 are still ongoing and will be invaluable in delineating the role of this CYP in the presence of other epoxygenases such as CYP2C8 and CYP2C9. The Ikemura et al. study (2019) referenced during the introduction of CYP2J2-specific substrates and inhibitors characterized azelnidipine as a CYP2J2 mechanism-based inhibitor. Within that same paper other drugs such as ritonavir, dronedarone, amiodarone, 17-ethynylestradiol, 17-octedcynoic acid, N-(methylsulfonyl)-6-(2-proparglyoxyphenyl)-hexanamide), as well as some unnamed compounds discovered by Lafite et al. (Lafite et al., 2007b), were cited as mechanism-based inhibitors, although with respect to inhibiting specific CYP2J2-ligand interactions. Recently, a near-infrared fluorescent probe for monitoring CYP2J2 activity was developed, showing promise that other molecular design studies can be carried out to develop more specific probes and inhibitors of CYP2J2 (Ning et al., 2018). It is these types of biochemical advances that will facilitate several physiological studies in other disease states, such as cancer, where overall expression of most CYPs are elevated and specific contribution of the different CYP isoforms remains poorly understood.

One of the primary problems throughout the pharmaceutical industry is drug-drug interactions where concurrent use of two drugs or one drug and one endogenous molecule may mimic, magnify, or oppose the effect of the other drug. Whether intended or unintended targets of small molecules, drug-drug interactions accommodated by CYP active sites could result in impaired homeostatic maintenance due to altered pathways of endogenous metabolism (Branden et al., 2014; Guengerich, 2002). CYP2J2 is the primary cardiovascular CYP but is not included regularly in drug toxicity screening as drug metabolism studies are commonly “liver-centric”. There is increasing awareness that many current drugs are substrates for CYP2J2 and this enzyme plays a significant part in the metabolism of drugs known to cause cardiotoxicity. Only doxorubicin-CYP2J2 interactions have been studied using both in vivo and in vitro models. Therefore, our understanding of the role of CYP2J2 and EET formation is based largely on the data from this single drug. Fully understanding the role and possible protection by CYP2J2 in the heart against drug toxicity requires extended molecular characterization studies of how CYP2J2 discriminates between xenobiotic and endogenous substrates when presented with either ligand type. The biochemical studies on the doxorubicin-lipid, phytocannabinoid-lipid, and lipid-lipid interactions with CYP2J2 demonstrate that CYP2J2 cooperative recognition of small molecules should be taken into account, especially with respect to evaluation of the cardiotoxicity of chemotherapeutics drugs. These studies showed that multiple molecules can concurrently bind to CYP2J2, but the selected conformation of these molecules occupying the active site is highly dependent on the structures of each molecule studied. Overall the structure-function studies of CYP2J2 have led to a better understanding on how ligands and proteins, specifically CYPs, interact, which can give valuable insight into their functions. More specifically, these experimental findings may lead to better drug design or therapeutics that can exploit those features involved in modes of molecular recognition and conformational sampling by the ligand and protein when in complex.

Considering the variety of health benefits that the epoxides provide to inflammation and the cardio/cerebrovascular systems (Davis et al., 2017), there is a lot of interest in exploiting this pathway for therapies. Importantly, more research will need to be done regarding the potential effects of drugs. A more provocative approach would be to design a drug that can bind and potentiate epoxygenase activity. Positive heterotopic interactions have been observed for other CYPs, mostly the drug-metabolizing CYP3A4. Therefore, eventually developing drugs that can increase endogenous CYP activity may be an effective therapeutic route to increase the levels of PUFA-epoxides, especially during disease conditions such as stroke.

Overall, CYP2J2 is one of the most highly expressed CYPs of the heart and is also highly expressed in the brain compared to other epoxygenases. Therefore, CYP2J2 must be a source of producing in these tissues PUFA epoxides with vasodilatory, anti-inflammatory and other beneficial pharmacologies. An important feature is that CYP2J2 produces all four regioisomers of EETs and can metabolize a variety of different lipids due to its large active site varying from fatty acids to large eCBs, such as 2-AG. Epoxygenases CYP2C8 and CYP2C9 are more limited by their substrates and regioselectivities. Additionally, CYP2J2 has specific enantioselectivity and produces a 74:26 ratio of (R,S)-:(S,R)-14,15-EET and an almost racemic mixture of the other regioisomers. Taken together, the targeting of selective or generalist epoxygenases by modulating their activities can prove to be a potential strategy for therapeutic design due to the relationship between CYP-derived products and tissue-specific physiology. Selective product distributions leading to different epoxide regioisomers and enantiomers can influence bioactive lipid binding to downstream receptors, where the result of metabolism often ends with biotransformed molecules exhibiting altered pharmacologies from their parent structures. Although the identity of the EET receptor remains unknown (Liu et al., 2017c), it is known to show enantioselective preferences not only in binding but also in potential downstream signaling. Therefore, the unique ability of CYP2J2 of producing all the different regioisomers and 19-hydroxy product may be important for cardiovascular and anti-inflammatory function. Alternatively, selective regioisomer formation by select CYP epoxygenases in select tissues could be critical in elucidating endogenous mechanisms for responding to the development of different disease pathologies.

Biochemically, CYP2J2 is also a very slow enzyme with respect to lipid metabolism as compared to drug metabolism, which is physiologically relevant as high PUFA-epoxide production can lead to adverse toxicity. There are several reasons that can be hypothesized that might explain why CYP2J2 is a slow enzyme. Firstly, the solubility of the PUFA substrates is low in aqueous solution. For instance, rate of metabolism of anandamide by CYP2J2 is higher than the rate of metabolism for arachidonic acid by CYP2J2 as anandamide is more soluble in the reaction buffer compared to arachidonic acid (McDougle et al., 2014). Secondly, the kinetic data agrees between PUFA metabolizing cytochrome P450 enzyme systems (Dhar et al., 2008). Therefore, we observe that PUFA kinetics with CYP2J2 and similar membrane-bound CYP systems are slow compared to bacterial soluble CYPs. Upon considering the rates and reactive preferences of different CYP epoxygenases, it is imperative to ponder about the extent of crosstalk between all the epoxygenases across the cognate tissues in which they are expressed, as well as their potential lipid substrates and the resulting downstream interactions with receptors following CYP metabolism.

There is an overall consensus in the scientific community that acute inflammation needs to be followed by a timely resolution otherwise it leads to chronic inflammation. This inflammation resolution plays a fundamental part in the body’s response to inflammation caused by tissue injury, infection, and aging. Hence, it can be postulated that there is complex cross-talk and competition between the CYP2J2 and other epoxygenases, along with cyclooxygenase and lipoxygenase enzymes, for PUFA substrates during an inflammatory process. Therefore, as can be seen from this review, there are many unanswered questions with regards to CYP epoxygenases: (a) interactions with PUFA and drugs, (b) tissue specific roles, (c) interactions with membranes, (d) specific mechanism-based inhibitors and inducers and (e) interactions amongst other epoxygenases and with different lipid-metabolizing enzymes especially during diseases conditions. The enumeration of the research on CYP2J2 presented here describes the current knowledge on CYP epoxygenases and identifies the future needs for the molecular characterization of CYP2J2 to enable a new axis of therapeutic design.

Abbreviations

2-AG

2-arachidonoyl glycerol

7-de-aDOX

7-deoxydoxorubicin aglycone

ų

cubic ångström

AA

arachidonic acid

ADX.

adrenodoxin

ADXR

adrenodoxin reductase

AEA

anandamide

CB1

cannabinoid receptor 1

CB2

cannabinoid receptor 2

CBC

cannabichromene

CBD

cannabidiol

CBG

cannabigerol

CBN

cannabinol

COX

cyclooxygenase

CPR

cytochrome P450 reductase

CYP

cytochrome P450

CYP2J2

CYP Family 2 Subfamily J Member 2

CYP2J2fl

Full length/native N-terminus CYP2J2

CYP2J2tr

N-terminus truncated CYP2J2

DHA

docosahexaenoic acid

DOX

doxorubicin

DPA

docosapetaenoic acid

EBS

ebastine

eCB

endocannabinoid

EDP

epoxydocosapentaenoic acid

EET

epoxyeicosatrienoic acid

EET-EA

epoxyeicosatrienoic ethanolamide (synonymous with AEA-epoxide)

EPA

eicosapentaenoic acid

EPOX

epoxygenase

ER

endoplasmic reticulum

FP

fluorescence polarization

HDoHE

hydroxydocosahexaenoic acids

HMMM

highly mobile membrane-mimetic

LA

linoleic acid

LOX

lipoxygenase

MD

molecular dynamics

MDMA

3,4-methylenedioxymethamphetamine

MOE

Molecular Operating Environment

NAFLD

non-alcoholic fatty liver disease

NSAID

non-steroidal anti-inflammatory drug

O-AEA

virodhamine

pCB

phytocannabinoid

PDB

protein data bank

PGH₂

prostaglandin H2

PGI₂

prostaglandin I2

POPC

1-palmitoyl-2-oleoyl-glycero-3-phosphocholine

POPS

- 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine

PUFA

polyunsaturated fatty acid

sEH

soluble epoxide hydrolase

sEHi

soluble epoxide hydrolase inhibitors

SM

sphingomyelin

SNP

single nucleotide polymorphism

TRP

non-selective transient receptor potential cation channel

TRPV1

TRP vanilloid 1

Δ8-THC

Δ8-tetrahydrocannabinol

Δ9-THC

Δ9-tetrahydrocannabinol